The Ellipsis Approach to Coordination (Wilder 1994). 292. 3.2.2 ...... Smith (1991), Benincà (1994), Ménard (1994, 177-179), Cocchi (1994), and Squartini &.
ÉTUDES ROMANES DE LUND 53
Verner Egerland
THE SYNTAX OF PAST PARTICIPLES A Generative Study of Nonfinite Constructions in Ancient and Modern Italian
Egerland, Verner, The Syntax of Past Participles. A Generative Study of Nonfinite Constructions in Ancient and Modern Italian. Études Romanes de Lund 53, Lund 1996. 352 pages. Written in English. Monograph. The purpose of this thesis is double. First, it will be shown that a number of problems of the syntax of ancient, literary Italian lend themselves readily to an analysis in terms of the Theory of Principles and Parameters. Second, the Italian data, supported by comparative remarks on modern Romance and Germanic, confirm the essential correctness of one particular version of the above theory namely the Antisymmetry framework of Kayne (1994). The issue is predicative past participles in two contexts: A. The past participle in periphrastic constructions with the verb avere ‘have’, for example, ho chiuso la finestra ‘I have closed the window’ as opposed to ho la finestra chiusa ‘I have the window closed’. B. The past participle in the absolute construction chiuse le finestre, me ne sono andato ‘closed the windows, I went out’. The theoretical and empirical problems that will be discussed can be summarized in four general points: functional structure of the participle, word order, agreement patterns, and interpretation especially with regard to Aktionsart or verbal aspect; the analysis will focus on the Affectedness Constraint. It will be argued that the participle clause hosts an agreement projection (AgrOP) and an aspect projection (AspP), but there will also be reason to consider the relevance of Tense and Comp for participial syntax. The discussion of word order will concentrate on the question of how the arguments are placed in relation to the predicate, that is, the placement of S, V, and O. The analysis takes a synchronic perspective, and compares three stages of Italian: the period that runs from the late 13th century to 1400, and the first half of the 16th century. These periods of time will be referred to as Medieval Italian and Renaissance Italian, as opposed to Contemporary Italian. It is not the purpose of this thesis to analyze the mechanisms of language change. Lund University Press Box 141, S-221 00 Lund Sweden Art. nr. 20422 ISSN 0347-0822 Lund University Press ISBN 91-7966-368-0 Chartwell-Bratt Ltd ISBN 086238-436-2 © 1996 Verner Egerland Printed in Sweden Lunds Grafiska AB Lund 1996
To my Parents
TABLE OF CONTENTS Acknowledgments
11
INTRODUCTION 1 General Presentation 2 Aims and Limits 2.1 Defining the Research Agenda 2.2 Limits 2.3 Historical Syntax and Generative Grammar 2.3.1 How this Thesis Relates to Previous Generative Approaches to Historical Syntax 2.3.2 Synchrony and Diachrony 2.4 The Organization of this Thesis 3 General Discussion on the Framework and the Theoretical Issues 3.1 Case Theory and The Minimalist Program 3.2 X’-Theory 3.2.1 Antisymmetry and Bare Phrase Structure 3.2.2 Word Order Alternations: Subject-Object-Verb / Subject-Verb-Object 3.3 Functional Structure 3.3.1 Split-Infl and the Structure of Participles 3.3.2 The AgrO node: Case and Agreement 3.3.3 The Aspect node
13 13 14 14 15 17
27 29 29 31 32
PART A: AVERE + PAST PARTICIPLE
37
CHAPTER 1 Word Order and Agreement in the Periphrastic Tense: 14th and 16th Centuries 0. Introduction 1 The Corpus Survey of Lucchesi (1962/63) 2 Organization of the Present Corpus 2.1 The Constructions 2.2 The texts 2.2.A The A-Corpus 2.2.B The B-Corpus 2.3 How the Survey is Carried Out 3 The Result of the Survey 3.1 Type 1: Participle+Object
37 37 39 41 41 42 43 44 45 47 47
17 18 20 22 23 25 25
3.2 3.2.1 3.3 3.4 3.5 3.6 3.7 3.8 4
Type 2: Object+Participle Conclusions on the Basis of Construction Types 1 and 2 Type 3: Objects in Operator Contexts Type 4: Pronominal Objects Type 5: Preposed Participle Type 6: Preposed Object Type 7: Participle+Infinitive+Object Type 8: Object gap Concluding Remarks
53 59 60 64 65 66 67 67 68
CHAPTER 2 The AgrO Node: Word Order and Spec-Head Agreement in Auxiliary + Participle Constructions 0 Introduction 1 Preliminary discussion 1.1 Spec Head Agreement, Inversion, and Rightward Agreement 1.2 Some Further Notes on ‘Free’ Inversion 2 The Data 2.1 Avere + Past Participle in Medieval Italian 2.2 Tener + Past Participle in Iberoromance 2.3 Få + Past Participle in Swedish 3 Analysis I: Extraposition or Free Branching? 4 Analysis II: Object Pro 5 Word Order Variations in Old Italian, and Remarks on the Comparison with Modern Germanic 5.1 Construction types 1, 2, 5, and 6 5.2 XP-Preposing, Long Head Movement, or Stylistic Fronting? 5.3 An Additional Note on Construction Type 5 5.4 Construction Type 7 6 Conclusion
89 89 97 98 99 100
Appendix 1 to Chapter 2: Inflectional Structure in the Swedish Participle Phrase Appendix 2 to Chapter 2: SOV and Equi-Distance
101 104
CHAPTER 3 The Aspect Node: Accusative Checking, AspP and the Affectedness Constraint 0 Introduction: The Affectedness Constraint and Case Theory 1 Theoretic Background to the Discussion on Affectedness and its Correlation with Aspect, Case, and Argument Structure 2 One Affectedness Construction in Italian and Scandinavian: Possessive have + participle
69 69 69 69 73 75 75 79 79 81 85
106
107 113
2.1 2.2 3 3.1 3.2 3.3 3.4 3.5 4 4.1 4.2 4.3 5
The Lexical Restrictions on PAP A Note on Få + Participle Analysis Passive Participles as Case Checkers Affectedness, DP-Movement, and Case Affectedness and the Lexical Suppression of External Arguments Aspect, the Extended Projection Principle and Syntactic Externalization of Arguments Conclusion Some Additional Notes on Delimited Asp AspP and Theta Theory Delimited AspP and the Lexical-Functional Distinction The Status of EPP in MPLT The Romance Causative: Fare + Infinitive
CHAPTER 4 The Interaction Between Aspect and Agreement: Asp, AgrO, Greed, and Participial Agreement 0 Introduction 1 A Theoretical Issue: Antisymmetry and Bare Phrase Structure 2 The Empirical Issue: Agreement in Operator Contexts 2.1 A Note on the Analysis of Kayne (1989a) 3 Analysis 3.1 Aims 3.2 Structural Case Checking and Split Infl 3.2.1 The Nominative Case Checking Parameter 3.2.2 The Accusative Case Checking Parameter 3.3 Summary 4 Extensions to other contexts 4.1 Agreement Patterns in ECM Contexts 4.2 Participial Agreement with Clitic Pronouns 4.3 Agreement in the Passive 5 Medieval and Renaissance Italian 5.1 Agreement in Operator contexts 5.2 Pronominal Agreement 6 Extension to Germanic 7 PAP 8 Summary
115 120 121 121 123 129 132 139 139 140 141 142 142
148 148 149 150 152 153 153 154 154 156 161 162 162 164 166 167 168 170 172 173 174
Appendix to Chapter 4: Further Observations on Agreement Phenomena in Medieval and Renaissance Italian 1 Agreement by Attraction 2 Agreement in Arbitrary Contexts
176 176 177
PART B: THE ABSOLUTE PAST PARTICIPLE
183
CHAPTER 5 Participial Absolute Small Clauses in the History of Italian: Data 0 Introduction to Part B and Chapter Five 1 Organization of the Data 2.1 Number of Lexical Arguments 2.1.1 Summary 2.2 Word order 2.2.1 Summary 2.3 Participial Agreement 2.3.1 Summary 2.4 Lexical Restrictions 2.4.1 Summary 2.5 Negation 2.5.1 Summary 2.6 The Agentive by-Phrase 2.6.1 Summary 2.7 Comp field elements 2.7.1 Complementizers 2.7.2 Wh-expressions 2.7.3 Summary 2.8 Null Argument Subjects 2.8.1 Summary 2.9 Null Argument Objects Table 5.I: Major Syntactic Patterns of PASC: MI, RI, CI
183 183 185 186 188 188 193 193 198 198 203 203 205 205 209 209 209 210 215 215 221 221 223
Appendix to Chapter 5: The Problem of Punctuation
224
CHAPTER 6 Participial Absolute Small Clauses in the History of Italian: Analysis 0 Introduction 1 Preliminary Discussion on the Theoretical Issues 1.1 Belletti’s (1990) Account for Absolute Participles 1.1.1 The Absence of Tense, V-to-Comp, and Rightward Agreement
227 227 228 228
1.1.2 1.1.2.1 1.1.2.2 1.1.2.3 1.1.2.4 1.2 1.2.1 1.2.2 1.2.3 2 2.1 2.2 2.2.1 2.2.2 2.2.3 2.2.4 2.2.5 2.2.6 2.3 2.3.1 2.3.2 2.3.3 2.4 2.4.1 2.4.2 2.5 2.5.1 2.5.2 2.6 3 4
The Effects of the Absence of Tense Voice in PASC Unergative Predicates in PASC Two Lexical Arguments in PASC Negation in PASC Modification of Belletti’s (1990) Analysis V-to-Comp and Spec Head Agreement: A Unified Account of Transitive and Ergative PASC V-to-Comp and the Absence of Tense The Case Conflict in Comp Structure and Derivation of Participial Absolute Clauses in Three Stages of Italian General Proposals: Tense, Aspect, Voice, and Object pro The Internal Structure of Participial Absolute Clauses; the Tense Node Tense and the Wh-Criterion Tense and Negation Tense and the Copular V A Layered Case Theory The Temporal Reference of Nonfinite Verb Forms Conclusion The derivation of PASC Verb-Raising in CI and MI DP-Movement in MI Verb-Raising and DP-Movement in RI Case, Voice, and Aspect ‘Open Voice’ in MI and RI Contemporary PASC as a Passive Construction Some Problems and Theoretical Implications of the Passive Analysis Belletti’s Arguments for an Active Analysis of CI PASC Distinguishing PAP from PASC, and Remarks on the Notion of Affectedness Tense and the Suppression of External Arguments Old Italian Gerundival Clauses Argument pro and the Problem of Recovery
Appendix to Chapter 6: A Further Note on the Problem of Voice in Ancient Italian
230 230 231 232 232 233 233 235 236 239 239 240 241 242 243 243 245 248 249 249 250 254 256 256 259 261 261 265 269 270 273
278
PART C: NULL PRONOUNS AND NOTES ON THE SYNTAX OF ADJECTIVES CHAPTER 7 Object pro 0 Introduction 1 Preliminary Discussion 1.1 Null DPs and Word Order 1.2 Recovery of Null Pronouns 2 Null Objects in Coordination Structures: the Data 2.1 Medieval and Modern Italian 2.2 NOC in Old and Modern Scandinavian 2.3 Similarities and differences between NOC in Medieval Italian and in Old Scandinavian 3 Analysis 3.1 Aims 3.2 Some Hypotheses that could be relevant for NOC 3.2.1 The Ellipsis Approach to Coordination (Wilder 1994) 3.2.2 The Topic-Drop Analysis 3.2.3 Phrasal Pronouns in Old Spanish (Rivero 1992) 3.3 Analysis: Recovering Dependencies, Object pro, and Overt Object Shift 4 A’-Bound Null Objects Further Data from Old Italian and Scandinavian 5 Conclusion of Chapter 7
281
281 281 281 281 283 284 284 289 291 292 292 292 292 293 295 297 302 306
CHAPTER 8 Remaining issues: On the Object pro of Rizzi (1986a) and Resultative Adjectival Constructions 0 Introduction 1 Object pro of Rizzi (1986a) 2 Resultative Adjectival Constructions 2.1 Adjectival PAP 2.2 Resultative Secondary Predicates
308 308 308 316 316 323
CONCLUDING REMARKS
327
Ancient Texts cited References
331 332
Acknowledgements.
First of all, I am deeply indebted to my supervisors Kjell-Åke Gunnarson, who was the first to introduce me to generative grammar and has followed my work attentively since then, and Maria Rita Manzini, whose teaching and valuable suggestions inspired and influenced a great deal of this thesis. They have both been very generous with their time and interest. I am also indebted to all the others who throughout the years have discussed the content of these pages with me, providing useful criticism and helpful comments. First and foremost, I wish to thank Christer Platzack for this and also for his part (which is considerable) in making the linguistic environment of Lund interesting. Many thanks also to Anna Cardinaletti, Guglielmo Cinque, and Paola Benincà. This work would never have seen the light of day, had I not been able to spend some years in Florence, and for this I owe gratitude to those who in one way or another made this possible, and especially to Leonardo Maria Savoia and Suzanne Schlyter. Special thanks goes to Luciana Brandi, who made me realize the interest in studying Old Italian from a generative view point. My sincere gratitude to Vetenskapssocieteten i Lund and the Fondazione Famiglia Rausing who funded my stay. Having been able to benefit from two interesting and yet very different academic environments has been a great privilege. First, I must thank my fellow Ph.D. students (some of whom are no longer students) for their interesting discussions, moral support, pleasant company, good food and native intuitions; above all my gratitude goes to Gloria Cocchi, Cecilia Falk, and Eva Wiberg, who have also been kind enough to read some of my pages. Thanks to Lars Olof Delsing, Gunlög Josefsson, Vesta Sandberg, Anders Bengtsson, and Carla Cariboni Killander for pleasant evenings past and to come. For the same reasons, I wish to thank Giuseppina Turano, Caterina Donati, Grazia Gianelli, Andrea Manuelli, and Luisa Melis. Many thanks to Peter Taylor and Liliete Martins for help with English and Portuguese data, to Sven Ekblad and Roberto Cerqui for interesting discussions on Medieval Italian, and to Claire Gronemeyer and Thane Sinclaire for having corrected my English. My sincere gratitude to the kind and helpful personnel of the library at the Accademia della Crusca, Castello. It would be impossible to mention all the friends who have supported me during these years and helped me to organize my life. The list is long and includes a lot of people already mentioned. Expressing gratitude for friendship is needless, but for their hospitality I owe a lot to Roberto Cerqui, Helena Börnfors, Margherita Ferraris and Raffaele Riccardi, Johan Nilsson and Maria Rosvall, Karin Ohlson, Lena Halldenius, and Kjell Eriksson. I am indebted to all of those who have helped me in my constant battles against the computer, above all, Johan Nilsson, Elisabetta Carpitelli, and Kjell Eriksson. I have saved the most important point for last, namely my parents. Without them, this work would never have been started and would never have come to an end. I am also very grateful to my aunt Ulla-Britt, for her devoted interest and financial support throughout the years. Thank you all.
INTRODUCTION
1
General Presentation
Whenever one tries to embrace two fields of research such as historical grammar and generative linguistics in our case, applying the theoretical apparatus of the latter to analyze the former, one does so with the hope that the encounter is mutually beneficial. In this thesis, I will discuss various problems of participial syntax in three different stages of Italian, and I will do so with a double aim: 1. I hope to show in the pages that follow that a number of crucial problems of the syntax of ancient, literary Italian lend themselves readily to an analysis in terms of the Theory of Principles and Parameters. Whereas a long tradition of grammarians have emphasized the artificial and latinizing aspects of this language, some of its most striking syntactic properties do not escape a principled and coherent explanation. 2. It is my intention to use the Italian data, supported by comparative remarks on modern Romance and Germanic, to argue for the essential correctness of one particular version of the above theory. In a few words, the study of ancient Italian may contribute in an interesting way to develop the theory, and the theory may provide some deeper understanding of the properties of ancient Italian. Such a double aim is, I believe, natural for a generative approach to the field of historical syntax. The terminology requires further explanation. As for the notion of ancient, literary Italian, by the term literary I mean written texts of a middle or high register. The question of register will be touched in chapter 1. Of the Italian varieties, I will consider Tuscan, and, among Tuscan varieties, mainly Florentine. The expression literary ancient Italian is used to indicate literary ancient Florentine, and this is indeed what one usually implies. When exceptions are justified, and sometimes they are, non-Florentine and nonTuscan texts will be discussed. Ancient is a quite general label that here comprehends two major periods, roughly, the Medieval ages and the Renaissance.1 The analysis will consider two periods in particular: 1. the period that runs from the late 13th century to 1400, and 2. the first half of the 16th century. I will use three abbreviations that may be useful for the reader to memorize from the beginning: MI (Medieval Italian), RI (Renaissance Italian) and CI (Contemporary Italian). It will become obvious, though, already at an early stage, that the labels MI and RI do not exactly correspond to the two periods of time. Instead, MI 1
I use the terms ancient and old in a quite general fashion. In the philological tradition the term italiano antico does not refer to a well defined period of time (as corresponding terminology in the Swedish tradition is often used), but there is rather the habitude of dividing the stages of the Italian language in centuries: the 13th, the 14th and so on. This is the organization of the Storia della lingua italiana of Migliorini (1988).
13
comprehends the first period until the mid 14th century and RI to the second half of the first period together with all of the second period. All of this will be illustrated and explained from chapter 1 onwards. The theoretic dimension of this work also requires some definition. In the last few years, the Principles and Parameters Theory has undergone a period of rapid change. A number of influential proposals have rendered the premises for generative analysis radically different from what they were only some years ago, at the beginning of this decade. Generally, this work can be defined as an attempt to analyze the syntax of participles in a framework inspired by the Minimalist Program of Chomsky (1993) and the Antisymmetry theory of Kayne (1994). As these approaches are compatible in some but not all respects, some choices will have to be made. I will argue that my data lend themselves to a successful and coherent description in terms of Antisymmetry. The analysis that follows will to a large extent also be compatible with minimalism in its first formulation, Chomsky (1993), but will diverge from the Bare Phrase Structure approach of Chomsky (1995) on some crucial points.2 The above cited works are to some extent programmatic, and our task is to apply them and check predictions on a limited set of data. Numerous directly or indirectly related issues will have to be left open in my demonstration.
2
Aims and Limits
2.1
Defining the Research Agenda
The field of participial syntax is extensive, and it is therefore necessary to limit the discussion to some well-defined topics of research. I have tried to draw the boundaries in a fashion as natural and intuitively correct as possible. The issue I deal with is predicative past participles in two major contexts: A.
The past participle in periphrastic constructions with the verb have, i.e., avere+past participle. e.g. Ho mangiato le mele. ‘I have eaten the apples’
Under A, I will not only treat the periphrastic tense but also the case of have of possession followed by a resultative participial phrase, as in I have the apples sold. The fact that these two constructions cannot be clearly distinguished in the earliest Italian texts (as well as some other Romance and Germanic varieties) is of some importance for the discussion that follows.
2
I will make separate references to Chomsky (1991), (1993) and (1995), which however appeared together in the volume The Minimalist Program (=Chomsky 1995) during the last stages of my work.
14
B.
The past participle in the absolute construction. e.g. Mangiate le mele, me ne sono andato. ‘eaten the apples, I went away’
While the periphrastic construction is familiar from many languages, the absolute participle construction is more particular to Italian; in those Romance languages where it exists it exhibits quite different properties, and it is less common in Germanic languages. The theoretical and empirical problems that concern us are numerous, but can be summarized in four general points: 1. Functional structure of the participle 2. Word order 3. Agreement patterns 4. Interpretation, especially with regard to Aktionsart or verbal aspect; the analysis will focus on the Affectedness Constraint. As for point 1, functional structure, I will dedicate most of my demonstration to the functional structure relevant for points 3 and 4, namely the agreement projection and the Aspect projection, but there will also be reason to discuss the relevance of Tense and Comp in the participial clause. Furthermore, the question of how point 2, word order, correlates with 3 and 4, agreement and interpretation, will be of primary interest. It should be pointed out from the beginning that the issue of word order will be limited to the question of how the arguments are placed in relation to the predicate, that is, the placement of S, V, and O. The position of non-arguments will by and large be disregarded. Ancient and Modern Italian data will on some points be compared with other Romance and Germanic languages. Even if an in depth analysis of Germanic word order is impossible here, I will use Germanic evidence as a point of reference at several occasions during my demonstration. Above all, I believe the comparison between modern Scandinavian and Medieval Italian with regard to participial agreement patterns is fruitful, and it will be discussed in chapter 4. The discussion on word order in the periphrasis requires some attention as to the similarities and differences between Medieval Italian and Modern German and Dutch; this will be dealt with in chapter 2.
2.2
Limits
There are many aspects of Old Italian syntax that deserve further investigation. The majority does not enter under participial syntax, but some will prove directly relevant at one point or another of my demonstration. Some remarks and tentative assumptions will therefore have to be made, even on issues that go beyond my field of research. The analysis will respect the following limitations:
15
1. As I am treating predicative participles, the syntax of attributive participles is not on my agenda. This is motivated, apart from considerations of time and space, by the simple fact that the two issues chosen, the periphrasis and the absolute construction, appear to be the most extensive and complex. In particular, I will not be concerned with the so called reduced relative, or pseudo relative. 2. The discussion on the periphrasis, point A above, will be limited to avere ‘have’ + past participle. This excludes from the discussion other kinds of periphrastic participial constructions, such as those involving modals, and more importantly, those involving the copula essere ‘be’. This last point is particularly difficult. An accurate discussion on agreement patterns in the avere+past participle periphrasis must take into account some aspects of essere+past participle; the issue of agreement in the passive cannot be totally neglected and will be brought up in chapter 4. I will disregard, however, the question of word order in copular constructions as well as all problems linked to the choice of auxiliary in the compound tense avere-essere, which is quite an interesting problem but far too extensive to be dealt with here. As essere + participle is excluded from my research, agreement patterns in the expletive, impersonal and existential constructions are also excluded.3 3. It is important to make a distinction between participial syntax and adjectival syntax. Although the theory does not provide straightforward and unambiguous criteria for making this distinction, it must be made (even at the price of an arbitrary choice) as the discussion otherwise would be much too extensive. As will become obvious to the reader, the comparison between participial and adjectival syntax cannot be altogether neglected, and I will therefore address some issues of particular importance in the final chapter. 4. Trivially, I do not propose to analyze other nonfinite verb forms, infinitives, gerunds and present participles. However, some of the conclusions about past participles will have consequences relevant to the syntax of the nonfinite verb in general. Some comparative remarks will be made, primarily on the ancient Italian gerund, during the discussion of the absolute constructions (chapter 6). 5. It is quite clear that Old Italian had verb second properties (studied above all by Benincà 1994, 1995), a circumstance which of course has relevance for a variety of
3
Some long-discussed problems of Old Italian syntax belong to these groups. Among these is the well known fu fatto beffe di loro ’(there)was made scorn of them’; see Parodi (1957), Ageno (1964), and Rohlfs (1969), among others. It is true that avere in the ancient prose is often used as an existential verb, but this, on the other hand, is less relevant to participial syntax. The study of existential avere is more relevant to the analysis of finite verb agreement.
16
syntactic issues in Old Italian; above all, it must be remembered that verb second is relevant to topicalization, scrambling, dislocation, or more generally, preposing processes. I will avoid the discussion of verb second, and my analysis will therefore be silent on the exact nature of the various attested cases of preposing. Finally, as our program of research has two sides, developing the theory of syntax and explaining the features of Old Italian grammar, there will inevitably be parts of the analysis that are more important for one than for the other. The discussion of chapter 3 is dedicated to a problem of syntactic theory and uses data from modern languages; its relevance for Old Italian will not become obvious until chapter 6. Vice versa, some of the discussion on word order and agreement in chapters two and four are mainly intended to show that Old Italian data can be systematically dealt with under our hypothesis; it may be that they contribute less to the development of theory.
2.3
Historical Syntax and Generative Grammar
2.3.1
How this Thesis Relates to Previous Generative Approaches to Historical Syntax
The list of generative studies dedicated to the problems of historical syntax has grown considerably in the last ten years. Because of the way I have defined the research program of this thesis, many of these have turned out little relevant for my demonstration. This is so for mainly two reasons: a lot of generative research on historical syntax has concentrated on issues such as verb second phenomena, word order in the main clause and null subjects (cf. van Kemenade 1987, Adams 1987, Vance 1989, 1995, Santorini 1989, Weerman 1989, Rivero 1991, Falk 1993, and Platzack 1995, among others), all of which are relevant for the syntax of the finite verb and nominative Case. Many of these are or minor importance for my research program and analysis in this thesis. There are above all two reasons for this. Firstly, my own work concentrates on the field of nonfinite syntax and accusative Case. A second point of divergence is the fact that a lot of generative research on historical syntax has been primarily interested in the diachronic aspect (cf. Lightfoot 1979, 1991, Kroch 1989, 1994, Santorini 1989, Falk 1993, Roberts 1993, and Clark & Roberts 1993, among others), whereas I focus on the synchronic systems at different stages of the language. A third difference between the present work and previous approaches is of course the theoretical apparatus which inevitably changes over time.
17
2.3.2
Synchrony and Diachrony
As we approach the problem of historical syntax and wish to study different stages in the history of a given language L, there are three major questions to address: 1. What are the differences between the various stages of L? 2. When did L change from one stage to the other? 3. Why did L change from one stage to the other? The first of these questions has priority over the following two. We cannot explain why or when differences between historic stages of L emerge, without knowing wherein the differences lie. The approach to ancient language can be either synchronic or diachronic. In other words, one may study the grammar of L considering the different centuries as stable systems, leaving aside, at least in part, the question of how the language community has developed or proceeded from a previous system to a following one. Alternatively, one chooses to analyze the change itself, attempting to elucidate the mechanisms that bring a language from one stage to another. In the former case, the scholar studies the grammar of a certain period of time which may include some decades or even some centuries as a synchronically working system, and limits himself to the observation that some diachronically working process has led the language to this stage. In the latter case, he or she is primarily interested in how the change is brought about, and how the diachronically working mechanisms are to be understood. I have chosen the first of these approaches; in brief, my aim is to answer the question of what differences exist between different periods of L. This choice is motivated by limited time and space. I will indicate a series of parametric resettings that distinguish three stages of Italian: the 14th century, the 16th century, and the 20th century. To some extent, I will also comment on the second question: when the grammar of Italian changed. It will become obvious that some properties of the Italian grammar changed around the middle of the 14th century, whereas others remained stable until the first half of the 16th century, and consequently, must have changed at some point between 1550 and present times. The time issue, however, will not be addressed in very exact terms. The third question listed above, how the change was brought about, states the diachronic problem and will not be discussed in this thesis. Although I leave the matter for future research, I believe a brief reflection on some current theories of language change may be useful. For a general survey of these approaches, I refer the reader to Falk (1993, ch. 2 esp. 34-41, 52-61).
18
One school of thought within generative theory has viewed syntactic changes as catastrophic events, in the words of Lightfoot (1979, 1991). Behind the term lies the intuition that changes in syntax tend to be relatively fast and that several changes occur more or less contemporaneously. This observation lends itself very well to an analysis in terms of principles and parameters; if a number of ‘surface’ patterns in a natural language depend on one underlying parametric setting, and if syntactic changes are due to parametric resettings, it is quite natural that several (apparently unconnected) changes occur contemporaneously and in a restricted period of time. A somewhat different view has been advanced in more recent years and is usually referred to as the Variationist Hypothesis of Kroch (1989, 1994) and Santorini (1989). According to this view, variation in the linguistic environment in which the language learner acquires his grammar may cause him to learn more than one grammatical system. The competition between grammatical systems thus takes place not only at the level of the language community (a fact which is straightforward) but also at the individual level.4 Crucial for the Variationist Hypothesis is the idea that syntactic changes «ordinarily proceed gradually, with innovative linguistic forms and word order patterns ousting older ones only slowly over centuries.»5 Furthermore, syntactic changes tend to proceed at a constant rate, in the sense that frequencies of forms increase and decrease with the same speed. For this latter view, the reader is also referred to Clark & Roberts (1993). The choice between these approaches is to some extent an empirical matter; it remains to be empirically clarified how fast or how gradual syntactic changes normally are. The corpus I will present in chapters 1 and 5 is intended to illustrate some clear differences between the three stages of Italian. It is my aim to show that a number of minor syntactic changes in Italian can be explained on the assumption of a few parametric resettings, and at least in this sense my demonstration may appear close in spirit to the ‘catastrophic’ view on syntactic change. However, the corpus survey is not sufficient to shed light on the question of whether the grammatical change proceeds at a constant rate or not. I believe that a future investigation on Italian participial syntax from the diachronic perspective could be promising. Finally, one might ask why a study is dedicated to historical syntax if the aim is synchronic analysis. The simple answer to this is that the quite peculiar patterns of some participial constructions in the varieties of Old Italian are not attested in any modern
4
«… language learners are not constrained to abduce a single grammar from the primary data when the positive evidence that they hear contains evidence for more than one grammatical system. This assumption is clearly independently motivated by the fact that bilingual and multilingual children successfully acquire more than one grammar.» (Santorini 1989, 13) This way of reasoning has the consequence that the single individual’s alternation between different forms of his language is compared with the bilingual individual’s alternation between different languages. 5 Kroch (1994, 2).
19
language, to the best of my knowledge, and can therefore be studied only from the ancient texts.
2.4
The Organization of this Thesis
This thesis is divided in three parts, of which the first two correspond to our two main issues: participles in the periphrastic tense, avere+ past participle (Part A), and in the absolute construction (Part B). Each of these will be introduced by a chapter where the crucial data are presented, chapter 1 and chapter 5. The chapters that follow in each part will mainly be concerned with the analysis of these data. The claim that Medieval Italian had object pro will be made in chapter 2 and is relevant for the demonstration throughout the whole of the thesis. For expository reasons, I have chosen to save a more detailed discussion of this claim and some of the empirical justification for it until the third part, C, immediately followed by the discussion on adjectival constructions. The discussion of Part A aims to show that word order systematically correlates with agreement and with Aktionsart. These observations justify the claim that the participial clause hosts an agreement element and an aspectual element. The Old Italian data collected in chapter 1 support above all the first claim, concerning the Agreement-Object node, as I will show in chapter 2. The second claim, concerning the Aspect node, is the topic for chapter 3. The demonstration in chapter 3 builds exclusively on contemporary evidence. Once we have accepted a ‘split’ of the participial Infl into AgrO and Asp, we will turn to the interaction between these two heads in chapter 4.
PART A
AVERE + PAST PARTICIPLE
Chapter 1 Word order and Agreement in the Periphrastic Tense, 14th and 16th centuries In this chapter, I will present the results of a corpus survey of the avere+past participle construction in two periods of time, the Medieval era and the Renaissance. The survey comprises all of the different contexts in which avere + past participle + direct object may appear, and is the basis for the analysis in the following chapters. Chapter 2 The AgrO Node. Word Order and Spec-Head Agreement in Auxiliary + Participle Constructions In chapter 2, I will analyze the data of chapter 1 concerning word order and participial agreement in the periphrasis. The Italian data will be compared to constructions in Modern Scandinavian and Modern Iberoromance, and I will argue that they are best explained in an Antisymmetric framework and on the assumption that certain grammars have the property of
20
licensing an expletive object pro. The variation in word order patterns relates primarily to the alternation between VO and OV in the participle phrase, and some more complex patterns that, I will argue, are cases of preposing of the participle phrase. Chapter 3 The Aspect Node. Accusative Checking, AspP, and the Affectedness Constraint The discussion in chapter 3 is dedicated to the relevance of aspect in syntax. The assumption of an aspect node makes it possible to link the aspectual interpretation of a construction to word order patterns. In particular, the explanation of OV word order in participle phrases in modern languages such as Italian and Swedish (already mentioned in chapter 2) derives from this idea. The ancient Italian varieties are left aside throughout the discussion of AspP, but the Asp node is crucial for the analysis of absolute participles in chapter 6 and has relevance for the discussion in chapter 8. Chapter 4 The Interaction Between Aspect and Agreement: Asp, AgrO, Greed, and Participial Agreement The last chapter of Part A deals with agreement in relative clauses, agreement with clitic pronouns and, to some extent, in the passive. The Old Italian data will be compared with Modern Romance and Germanic. The analysis will be carried out in terms of Antisymmetry.
PART B
THE ABSOLUTE PAST PARTICIPLE
Chapter 5 Participial Absolute Small Clauses in the History of Italian: Data In chapter 5, I will present data from the ancient varieties compared with contemporary Italian. The ancient texts referred to are to a large extent the same as were used in chapter 1, with some additions. Chapter 6 Participial Absolute Small Clauses in the History of Italian: Analysis The analysis of the absolute construction in three different stages of Italian is presented in chapter 6. The demonstration builds on the arguments for two separate functional heads in the participle phrase, Asp and AgrO, and introduces the proposal for participial Tense and Comp. The central claim of chapter 6 is that most of the crucial differences between the ancient use of the absolute participle and the modern use can be derived from the presence of Tense in the former and its absence in the latter.
21
PART C
NULL PRONOUNS AND NOTES ON THE SYNTAX OF ADJECTIVES
Chapter 7 Object pro The issue of null pronouns will be introduced in the early stages of the thesis. The analysis in chapter 2, and that in chapter 6 to some extent, relies on the idea of an expletive object pro; argument pro is briefly discussed in the last section of chapter 6. Chapter 7 introduces the explicit claim that Medieval Italian had null argument objects. We have to distinguish between argument object pro, which is found in coordinated structures, and other instances of null objects which, I will assume, are operator-bound and hence presumably variables. Chapter 8 Remaining issues: On the Object pro of Rizzi (1986a) and Resultative Adjectival Constructions I save two issues until the end. The first concerns the object pro of Rizzi (1986a): my analysis builds on the assumption that some differences between the 14th century grammar and the modern one are related to the property of licensing object pro in the former but not in the latter. This claim must be made compatible with the pro hypothesis of Rizzi (1986a) who convincingly argues that object pro is also licensed in the modern grammar. The second issue regards the syntax of adjectives. I will briefly suggest that my analysis of word order patterns and the Affectedness Constraint carries over to the field of adjectival syntax as well. In addition, there are five appendices. The two appendices to chapter 2 treat theoretical issues concerning functional structure in the participle clause and the issue of minimality. The three remaining appendices are more descriptive in nature. In the appendix to chapter 4, I will point out two peculiar agreement patterns in Old Italian; one is the phenomenon of so called agreement ‘by attraction’ and the other concerns arbitrary expressions. In the appendix to chapter 5, I will address the problem of punctuation, and in the appendix to chapter 6, I will make a brief comparative remark on a phenomenon I have called ‘open voice’. The content of these appendices is speculative in nature, dealing with intuitions rather than formalized proposals.
3
General Discussion on the Framework and the Theoretical Issues
At present, some basic ideas and concepts of the Government and Binding framework, (Chomsky 1981, 1982, 1986a and 1986b), are rapidly changing: Kayne (1994) advances a restrictive approach to X’-theory, and Case theory is drastically reformulated by Chomsky (1993).
22
Inevitably, the way my research is conducted depends on how these ideas are interpreted and to what extent they are assumed. In the following section, I will briefly review some of the most influential proposals and explain how my work relates to these. In this introductory section I will concentrate on MPLT and Antisymmetry. A thorough presentation of these frameworks is however impossible, and the reader is therefore referred to Chomsky (1993, 1995) and Kayne (1994) for those concepts and assumptions that are not explained here. Finally, in 3.3.3, I will briefly discuss the premises for my use of the aspect node.
3.1
Case Theory and The Minimalist Program
Some influential work in the field of Case theory during the last years has brought about a major change in the view on structural Case, agreement and word order phenomena. Following mainly the proposal of Chomsky (1993), I will assume a few basic concepts of MPLT: 1. Case assignment is a presyntactic process. Nouns are inserted into the syntactic component carrying their structural Case feature. They move in the syntax for reasons of Case checking which occurs in functional projections against inflectional features carried by the verb. Likewise, verbs enter the syntax already inflected and move in order to check their features against φ-features of the noun; see di Sciullo & Williams (1987), for the lexicalist hypothesis. Movement operations that occur prior to Spell-Out are called overt and those that take place between Spell-out and LF covert. 2. Checking takes place in the configurations (1) and (2), that is, in the Spec-head relation and in the head-head relation. (1)
[XP YP X°]
(2)
[X° Y° X°]
3. The checking hypothesis expressed in (1) and (2) replaces the earlier version of Case theory based on government. The concept of government is eliminated from the theory of grammar. This innovation, one of the most far reaching in Chomsky (1993), naturally has important consequences for the following account of word order and agreement in Old Italian, Romance and Germanic, as it will have for any other account.
23
4. I will assume two fundamental Principles of Economy, introduced by Chomsky (1993): a. Procrastinate ensures that only movement operations that are forced in overt syntax are allowed in overt syntax. The system chooses covert operations whenever possible. b. Greed states that movement operations are egoistic, in the sense that an element only moves to satisfy its own needs for checking. Movement of any element into any position is allowed only if it is required. In Chomsky (1995), Greed is incorporated into the movement hypothesis through the introduction of Move F. I assume 1-4 without further discussion. Some other ideas of the Minimalist Program will be restated or assumed only partially: 5. The Minimal Link Condition (MLC) is put forward in Chomsky (1993), and in effect forces an element to move by making only the shortest step. There is a tension between the MLC and Greed, as MLC would force movement into positions that are not required by Greed. In subsequent work, Chomsky (1995) resolves the tension through the introduction of Move F, which in essence means that Greed is privileged over MLC, and this is indeed the most natural interpretation for my purposes. The problem will prove relevant for my account of participial agreement in operator contexts (chapter 4). 6. In the different versions of the Minimalist Program, the role of agreement has changed considerably. In the first formulation, agreement is the driving force behind verb movement; the V raises to Infl in order to check its Agr features. A distinction is drawn between weak and strong agreement, where strong triggers overt raising and checking of the V and weak covert checking. Quite different is the last version of minimalism, where agreement, so it is said, can actually be dispensed with in the theory of grammar. I will not follow the last suggestion for reasons to which I return in 3.3. below and which will become obvious during the demonstration. Nor will I assume the first in its strict formulation; we must accept, in this work and independently, that V-raising is triggered in many contexts and many languages where no visibly strong agreement is present to justify the movement. However, the issue of finite V-raising is not directly relevant to our discussion; whether Chomsky’s original proposal is assumed or somehow reformulated is therefore not of crucial importance. Crucial are instead the agreement configurations stated in (1)/(2) above. Feature checking can take place in two local relations only: Spec head and head-head. It thereby follows that these are
24
the only configurations where agreement can be triggered.6 I will assume this throughout: a predicate under a functional head X can show agreement with an argument YP if (a) YP moves to the Spec of X or (b) Y adjoins to X, that is, in (3) and (4) but not in (5). (3) (4) (5)
[XP YP X°] [X° Y° X°] *[XP X° [YP Y°]]
Agreement Spec-to-head Agreement head-to-head *Agreement Compl-to-head
Among recent contributions to the debate on agreement, the third alternative, Head Complement agreement, is assumed by Belletti (1990) for Italian and Sigurðsson (1992b) for Scandinavian. I will instead follow the idea that (5) is indeed excluded; I will argue for the correctness of this in chapters 2, 4, and 6.
3.2
X’-Theory
3.2.1
Antisymmetry and Bare Phrase Structure
Kayne (1994) formulates a highly restrictive version of the X’-module, on the basic assumption that there is rigid correspondence between hierarchical structure and linear order. This correspondence is stated as the Linear Correspondence Axiom (Kayne 1994, 6), which has the consequence that the following configurations are universally excluded for principled reasons: (6)
*
More than one adjunction to XP:
XP
YP
XP
ZP
X’
6
Triggered, used in this way, would actually be a pre-minimalist terminology, as we now assume that agreement features are already present on the elements when they are inserted into the syntax. But the result is the same if we assume that whenever a V agrees overtly with a DP, the φ-features corresponding to agreement (on the DP and on the V) count as strong and require overt checking. For our purposes, the choice between saying that overt movement of DP is triggered by the overt realization of agreement, and saying that overt realization of agreement is triggered by overt movement of DP, is equal.
25
(7)
*
Rightward projection of the specifier:
XP
X’
Spec
(8)
*
Leftward projection of the complement:
X’
Compl
X
(9)
*
Rightward adjunction:
XÞ
XÞ
YÞ
(6) amounts to saying that no material can be adjoined to an XP that already has a specifier. Put differently, the distinction between specifiers and adjuncts actually disappears. (7) states that natural languages universally have the specifier to the left, and (8) that they are headinitial. The prohibition on rightward adjunction in (9) is of course valid for adjunction at any level, not only to a head as indicated in (9). Consider that a case like (10), which represents the current analysis of right-dislocation or extraposition, is therefore doubly excluded.
(10) *
XP
XP
Spec
YP
X’
Assume in (10), for instance, that X=V or Infl, and YP a right-dislocated element; Spec X is occupied by the subject under the subject-internal-to-VP hypothesis; see Manzini (1988) and
26
Koopman & Sportiche (1991, among many others). (10) is excluded under (9) as a case of rightward adjunction and obviously also by (6) because both YP and ZP have adjoined to XP. In the Bare Phrase Structure approach of Chomsky (1995) conclusions (7) and (8) are upheld, though Kayne’s way to derive them is not available. On the contrary, (6) is allowed in the last version of minimalism through the introduction of the Multiple-Spec hypothesis. Chomsky (1995) and Kayne (1994) converge in excluding (7) and (8) but their approaches diverge on (6). For our purposes, the most adequate move is to accept Kayne’s conclusions about the structures (6)-(9). On the basis of my data, I will argue for a solution in terms of Antisymmetry and against Chomsky (1995), where these two prove to be incompatible. The effects of the LCA will have direct consequences for the analysis of chapters 2, 4, and 8. The case of right dislocation (9), which is perhaps the most difficult point for Antisymmetry, is not directly relevant to my data. On the contrary, (7) and (8) have drastic consequences for the analysis of the word order facts I am interested in. I will review some of them in 3.2.2.
3.2.2
Word Order Alternations: Subject-Object-Verb / Subject-Verb-Object
As we will see, Old Italian word order patterns (above all in nonfinite contexts) are somewhat unstable, alternating between SVO and SOV. In Government and Binding Theory, such variations would have been captured by the postulation of mainly two kinds of parameters: first, a Case assigning parameter, determining in which structural relation Case assignment can take place: (11)
Case X is assigned a. under Spec-head agreement, b. under government. (cf. Hoekstra & Mulder 1990, Cardinaletti & Roberts 1991, Roberts 1993)
Most research has been concerned with nominative Case. In order to account for our data we would need to assume X=accusative.7 The second parametric choice, which was been assumed for the last fifteen years or so, concerns branching. Following a widespread view, German and Dutch would differ from English, Scandinavian and Romance in being head final (cf. Koster 1975, Zwart 1993, 57-63):
7
This was the view taken in the first version of this work, presented at the Grammatik i Fokus colloquium, Lund, February 1992.
27
(12)
(13)
V’
Comp l
V
V’
Comp l
V
Thus, German and Dutch word order patterns in compound tenses and subordinated clauses are explained on the assumption that the nonfinite V remains in situ (the auxiliary in (14) raises overtly to a higher position that may be either AgrS or C). (14)
Ich habe das Buch gelesen. I have the book read
(15)
… das Ich das Buch lese. that I the book read
(16)
V’
V lese/gelesen
DP(O) das Buch
This approach is abandoned by Zwart (1993) who builds on the Antisymmetry model of Kayne (1994). Zwart argues, primarily on the basis of Dutch data, that word order patterns such as those in (14) and (15) are the result of overt object shift to Spec AgrO, and that a head initial approach is preferable (whether the V raises to AgrO or remains under V is irrelevant). (17)
AgrOP
AgrO’
Spec DP(O) AgrO
VP
(11) is excluded in minimalism, and (13) and (16) are excluded in minimalism and Antisymmetry. Let us begin with a comment on (13) and (16): As we have said, the head final approach is universally excluded on principled grounds in Kayne (1994, 33-37). Chomsky (1993, 1995) hypothesizes that different word
28
order patterns may be derived from a deep head-initial structure merely by assuming differences in overt or covert raising of the relevant elements: subject, object and verb. SOV surface order is derived from underlying SVO in the way indicated in (17); VSO can be derived by overt raising of the verb (subject and object in situ) or by overt raising of verb and subject (object alone in situ); OSV, by overt raising of the object alone, and so on and so forth. Limiting our discussion to Romance, the idea is descriptively very successful, as we will show in the following. The Case parameter (11) obviously can not be maintained if government is eliminated from the theory and Case-checking in the Spec-head relation remains as the only option. If overt object shift generally applies in German and Dutch, the accusative feature carried by the object in (14) and (15) is such that it requires overt checking, that is, strong in minimalist terms, and the difference between SVO and SOV languages thus reduces to the weak - strong parameter. The question then arises how to account for SVO/SOV-alternations observed in one and the same grammar, and this problem will occupy us for a large part of chapters 2 and 6.
3.3
Functional Structure
3.3.1
Split-Infl and the Structure of Participles
At least since Chomsky (1986b) it has commonly been assumed that the finite sentence S corresponds to the projection of finite inflection IP. The functional head Infl thus hosts not only Tense T but also various inflectional features as person, number or others generally indicated as Agr. Pollock (1989) proposed to resolve the problem of double-headedness of Infl by splitting it into two functional projections: T and Agr, where AgrP is the complement of T. The Split-Infl hypothesis was assumed by a variety of scholars though with a different ordering of the elements; for Belletti (1990), Chomsky (1991) and Rizzi (1990), TP is the complement of Agr. With the introduction of abstract functional structure, the view of nonfinite verb forms has changed, and is still changing rapidly. In the classic analysis of Stowell (1981), a small clause contained by definition a predicate and a subject and did not host an inflectional node. Later on, some influential and independent proposals have converged to introduce inflectional structure into the lower part of the tree. Baker, Johnson & Roberts (1989) argued for an inflectional element in the participle phrase, and subsequently, evidence has emerged for the existence of an additional projection in the tree, higher than VP but lower than IP (e.g., Déprez 1989 and Johnson 1991). (18)
[IP Infl. [XP X [VP V]]]
29
Kayne (1989a) suggested that the participial V in Romance hosts an Agr element. Abstract Tense, aspect and agreement heads have henceforth been suggested for the syntax of participles and infinitives; see among many others Pollock (1989), Belletti (1990), Kayne (1991, 1993), and Junker & Martineau (1992). The sentential structure assumed by Chomsky (1991, 1993), and those following his framework, is the one indicated in (19): (19)
[CP C [AgrSP AgrS [TP T [AgrOP AgrO [VP V]]]]]
The lower portion of this structure, [AgrOP AgrO [VP V]], corresponds to the participial phrase. In this system, Aux is the same as T. Belletti (1990) proposes that an Aspect Phrase is also included in the functional structure of participles. Furthermore, the absolute participle is truly clausal, implying the presence of CP as in (20): (20)
([CP C )[AgrP Agr [AspP Asp [VP V]]]]
Kayne (1993) observes that the choice of auxiliary verb in some Romance varieties is sensitive both to Tense and to the person specification of the subject. He suggests (21) as the structure of the participial phrase alone. The C head of (21) has properties of both D and P, and the exact label is not of primary interest: (21)
Aux [CP C/D/P [AgrSP AgrS [TP T [AgrOP AgrO [VP V]]]]]
Inspired by the Antisymmetry program, Cinque (1994) gives evidence that the structure of participles is even richer than what has previously been assumed. On the basis of adverbial distribution in Italian, Cinque shows that the participial phrase hosts multiple functional structure:
(22)
Mod Neg Quant Quant Adv I bambini non hanno detto mica piú tutti tutto bene the children have not said at all no longer all everything well
alla maestra. to the teacher
The participle in (22) (underlined) has moved through a number of functional projections, and the adverbs and quantifiers fill the specifiers of these projections. As a consequence, there are more functional projections than indicated in (21). The proposal of Chomsky (1995) goes in the opposite direction breaking with the Split-Infl hypothesis. It is argued that the Agr based theory can be abandoned for a Multiple-Spec Theory and that Agr might be eliminated from UG altogether. The minimalist
30
framework comes into conflict with Antisymmetry on this point, and a considerable part of this thesis is concerned with the discussion of this contrast. The view I will take in the present work is the following: a lot of work on agreement phenomena carried out in recent years has emphasized the link that seems to hold between the overt realization of agreement and surface word order. This link is indeed strong, and in the following (chapters 2, 4 and 6) I will add empirical evidence in support of the conclusion that UG principles regulate the realization of agreement and that these correlate with DP movement.
3.3.2
The AgrO Node: Case and Agreement
As mentioned, it was first suggested by Kayne (1989a) that the participial structure contains an Agr node. Chomsky (1991, 1993) assumes this to be AgrO. There is an important difference between Kayne (1989a) and Chomsky (1991, 1993) with respect to the use of AgrO. For Kayne (1989a), AgrO is the functional node that corresponds to participial agreement, whereas for Chomsky (1991, 1993) AgrOP is the projection where accusative checking takes place. This difference may give rise to confusion: it is obvious that the checking of participle agreement cannot always be ‘equal’ to the checking of accusative Case. In many languages, participle agreement is triggered in the passive by a DP carrying nominative, for instance. For this reason, agreement and Case must be dissociated (cf. Chomsky 1995, 284).8 I will quite simply assume that the checking of Agreement may or may not coincide with the checking of Case and that Spec AgrO may or may not be a Case position. In the passive, AgrO is the locus for agreement features, but not for the Case feature. In clitic constructions in Romance (where participles agree with accusative clitics) AgrO may host both agreement features and the Case feature.9 If Greed is to be taken seriously, it is dubious whether the realization of agreement can be seen as a reflex of the derivation, that is, as the effect of a DP moving through a Spec of the relevant type. If the DP makes this move, it is because the movement is required for convergence. If agreement is realized and therefore checked in a construction independently of Case checking (as in the passive), this must be because checking of Agr forced for reasons independent of Case. Intuitively, this is not very strange. It is natural that some syntactic mechanism ensures that an argument is properly combined with its predicate. 8
There are different ways to capture this fact: Friedmann & Siloni (1993) assume an Agr phrase corresponding to participial agreement, AgrPartP, distinct from AgrOP. They are followed by Belletti (1993). This proposal, interesting as it may be, is however quite different from the one I will follow here. 9 Likewise, Borer (1993; 1995) in her work on AspP, reaches the conclusion that Spec Asp may or may not be a Case position which seems to be a reasonable assumption. The checking of Asp will therefore equal the checking of Case in certain contexts and not in others. See chapter 3 for further discussion.
31
Consider that in a passive construction, the DP checks Nominative against the T (or AgrS) headed by the copula and not by the predicate. Presumably, the φ-features on the argument DP (which are visible in LF), need to be matched against the predicate. In the spirit of minimalism I assume this the reason why agreement is spelled out, and checked in the passive. Finally, I will not follow Chomsky (1993, 1995) in assuming that Agr is one category realized in two different positions.10 I will assume that AgrS is involved in nominative checking together with Tense, and in a parallel fashion, that AgrO is involved in accusative checking together with Asp. AgrO and AgrS, rather than being two manifestations of the same category, are categorially distinct and linked to Asp and T, respectively.
3.3.3
The Aspect node
The minimalist approach as outlined in Chomsky (1993, 1995) does not take into consideration Aspect as a possible functional category. The claim that Aspect is syntactically realized has been advanced by a variety of scholars, beginning with Emonds (1985) and Tenny (1987); for discussion of AspP in Romance, see Belletti (1990), Cinque (1994), and Uriagereka (1995); see also Diesing & Jelinek (1993), Borer (1993; 1995), Arad (1995), and Travis (1995) among others. As aspectual relations are indeed relevant for the interpretation of a linguistic expression, an Aspect projection can be assumed with the same right as Tense, intuitively speaking. In other words, it is motivated to assume that the aspectual content of a representation must meet conditions of the conceptual-intentional interface, or Logical Form. I therefore follow the above studies in introducing an Asp head.11 The sentential structure is then (23) (specifiers omitted) where the participial phrase corresponds to [AgrOP AgrO [AspP Asp [VP V]]]:12
10
In the Minimalist Program, AgrS and AgrO are the same category appearing in two different positions or, in different terms, «mnemonics with no theoretical status» (Chomsky 1995, 341); see among others Rouveret (1991a) and Friedmann & Siloni (1993) for some argumentation against Chomsky’s view. 11 «… postulation of a functional category has to be justified, either by output conditions (phonetic and semantic interpretation) or by theory-internal arguments» (Chomsky 1995, 240). I will defend the view that AgrO and Asp are justified by output conditions of both kinds: the former is justified by the extensive correlation between agreement and word order, the latter by the fact that aspectual interpretation puts limits on various syntactic processes. Antisymmetry provides a theory-internal argument for both (Kayne 1994, 29-30) 12 (23) illustrates Belletti’s (1990) and Uriagereka’s (1995) ordering of the participial functional projections, where AgrOP dominates AspP. I will assume this throughout.
32
(23)
CP
Comp
AgrSP
AgrS
TP
T
AgrOP
AgrO
AspP
Asp
VP
However, building on Kayne (1993) and Cinque (1994), I maintain that the participial portion of the tree is richer than what is indicated here. Firstly, I assume with Kayne that the participle may have both T and AgrS; secondly, judging from Cinque’s adverbial data it is quite possible that the participial structure is even richer. There is also a possibility that some node is freely recursive (e.g., Belletti 1990, Cardinaletti & Roberts 1991, for functional structure in the finite clause). I will concentrate on showing that the participial phrase hosts AgrO, Asp, and T, since these are clearly present in my data.13 A major aim of this work is thus to explore some possible consequences of the assumption of an Aspect Phrase in syntax. In this sense, I wish to continue a discussion that has been present in the literature for a decade or so. The consequences of AspP are numerous and bear on a variety of issues, as we shall see. Needless to say, a thorough investigation of all of its implications goes far beyond the limits of time and space available for this study. This thesis should be regarded as an attempt to capture some fundamental generalizations concerning aspectual interpretation, above all what has been referred to in the literature as the Affectedness Constraint, and its correlation with word order and the projection of argument structure. Hopefully, the results are interesting, though the approach is sometimes slightly programmatic. I will point out possible extensions and alternative formulations of the proposal. To begin with, I believe some clarification of terminology is in order. First and foremost, most of the literature on the issues of Tense and Aspect emphasizes the necessity of distinguishing between the two, recognizing however that they correlate extensively. The distinction is often made in terms of time notions internal to the
13
In chapter 6, I will suggest that the presence of Tense is subject to parametrization.
33
situation or the Event (Aspect) and external to the situation or the Event (Tense) (cf. Comrie 1976, 5 and Hornstein 1990, 9). Secondly, in the pioneering work of Tenny (1987), the fundamental aspectual notion that regulates the mapping of arguments from lexicon to syntax is called delimitedness. Borer (1993, 1995) assumes the Aspect node in syntax to host the feature +/-delimited. Delimitedness comes very close to what was previously referred to as boundedness or telicity, a fact which may create terminological confusion. Consulting some references I take to be standard - Bertinetto (1986) for Italian, and Platzack (1979) for Swedish - we find a distinction between Aspetto and Azione in the former corresponding to Aspect and Aktionsart in the latter. The simplified and incomplete scheme of (24) illustrates how both scholars categorize some of the most used dichotomies: (24)
Bertinetto (1986):
Aspetto
imperfettivo
Platzack (1979):
non-perfective
Azione
perfettivo
telico
Aspect
atelico
Aktionsart
perfective
bounded
unbounded
Evidently, Aspect covers the perfective-imperfective distinction, whereas the telic/atelic or bounded/unbounded distinctions enter under Aktionsart or Azione. The topic for our study, delimitedness is not included in what these scholars understand as Aspect. However, Comrie (1976, 6 footnote 4, and chapter 2), who does not use the term Aktionsart, treats the notion of telicity as a problem of inherent meaning or inherent aspect, which is regarded as a subpart of the general issue. Following Borer (1993), I will assume that AspP is the locus for the delimitedness feature. Some other approaches have pursued the idea further, introducing more than one AspP. Cinque (1994) assumes two aspectual nodes, and in the approach of Travis (1995), there is an Aspect category as well as an Event category. Moreover, Borer (1993, 1995) argues that AspP has the function of ordering the verb’s arguments, which are projected into the syntax unordered (see section 1.4, chapter 3 for a brief presentation of this idea), and that AspP is involved in the checking of accusative Case. I follow these ideas in claiming that there are two aspectual nodes in syntax; there is one Asp node internal to the VP-shell which
34
carries the delimitedness feature, and there is one Asp node external to the VP-shell. The purpose of this assumption is to avoid a conceptual problem linked to the use of AspP: If, on the one hand, the Asp node is responsible for the ordering of argument structure, one is inclined to assume that Aspect makes up part of the predicate phrase; on the other hand, if AspP can be the locus for Accusative checking, it should rather have the status of a functional projection in a minimalist view.14 In order to avoid ambiguity with regard to the lexical/functional distinction of AspP, I thus assume two Asp nodes, one inside the VP-shell that is involved in the ‘ordering’ of argument structure (though in a way slightly different from what is assumed by Borer 1993, 1995), and the other outside and above VP, where accusative Case is checked. Clearly, such a move opens the possibility of saying that the lower Asp is the locus of Aktionsart (the delimited/non-delimited distinction), whereas the higher hosts aspectual specification, according to the distinction made by Bertinetto (1986) and Platzack (1979), among others. This last idea will not be crucial for my demonstration, however, but is a possible interpretation of the AspP nodes assumed. Note that, if this line of reasoning is pursued, sentence structure would contain at least the meaningful elements listed in (25), disregarding Agr nodes: (25)
[CP C [TP T [AspP2 Aspect [AspP1 Aktionsart [VP V]]]]]
The AspP corresponding to Aktionsart is assumed to be part of the VP-shell, and it is this element that will be of primary interest for the following discussion. Having said that my analysis lies in the field of Aktionsart/Azione, it must immediately be made clear that, although the matter of Aktionsart/Azione appears to be distinct from Aspect/Aspetto, the two interact in ways that are obvious but nevertheless intricate. It is useful to recall that telicity to some extent depends on perfectivity, in the sense that the telic reading is most clear when a predicate is used perfectively.15 Even more important for us is the fact that certain predicates can be understood as telic only when they appear with a perfective reading. The relevance of this will become clear in chapter 6. In brief, I assume Aktionsart/Azione and Aspect/Aspetto are not independent of one another, and accept it as intuitively clear they interact, though I will not make any attempt to formalize the intuition. Nor will I try to formalize the relation between Aktionsart, Aspect, and Tense.
14
This issue, that I refer to as ‘ambiguity’, is not always considered a problem, however. Travis (1995) takes the view that the categories Aspect and Event make up part of the VP-shell, that they participate in processes of lexical syntax (in the sense of Hale & Keyser 1993) and furthermore, that Aspect and Event are actually distinct from lexical categories and functional categories. I do not follow her on this point. 15 «I verbi telici hanno un’importante proprietà: quando sono coniugati secondo il paradigma di un Tempo imperfettivo, essi non consentono di derivare alcuna implicazione circa la positiva conclusione del processo; e dunque non permettono neppure di dedurre che il protagonista ha veramente compiuto il processo in questione.» (Bertinetto 1986, 91)
35
Finally, further investigation on these matters should shed light on the interrelation between Aspect and Voice. It is clear that passive formation brings about changes in the interpretation of Aspect or Aktionsart. The matter has been addressed already by Comrie (1976, 1981), and Borer (1995) makes use of the delimitedness notion in her analysis of passive. On this point, my model will be worked out so as to capture some correlations between passive and delimitedness that I believe to be crucial, making some predictions on how these correlate when mapped onto syntax and what consequences they have for word order phenomena. My work will contribute relatively little to the understanding on how this correlation is to be interpreted in terms of aspectual or lexical semantics. We are here dealing with the syntactic relevance of aspectual notions rather than with the understanding of the notions themselves. This distinction, I believe, is not only natural but also necessary.
36
PART A AVERE + PAST PARTICIPLE
CHAPTER 1 Word Order and Agreement in the Periphrastic Tense: 14th and 16th Centuries 0
Introduction
In the Romance languages, a transitive past participle tends to agree with its DP object when the object appears to the left of the participle, but not when it appears to the right. It is a wellknown fact that participles in Modern (standard) Italian cannot agree in (1) but cannot fail to agree in (2): (1a) (1b)
(2a) (2b)
Ho aperto la finestra I have opened[-Agr] the window *Ho aperta la finestra I have opened[+Agr] the window Ho la finestra aperta I have the window open(ed)[+Agr] *Ho la finestra aperto I have the window open(ed)[-Agr]
In Modern Italian, (1a) and (2a) correspond to two different readings similar to the different readings associated with the same options in word order in many other languages. (1a) is a compound tense, and (2a) which instantiates the possessive have is followed by a resultative small clause. This issue will be discussed extensively in chapter 3. There are Romance varieties that defy this generalization. In ancient stages of Romance (1b) is attested, and it is also grammatical in some contemporary dialects; see Rohlfs (1969, § 725), Parkinson (1988, 162), Wheeler (1988, 194, 270), Kayne (1989a), Smith (1991), Benincà (1994), Ménard (1994, 177-179), Cocchi (1994), and Squartini & Bertinetto (forthcoming). I will henceforth refer to (1b) as a case of rightward agreement as opposed to leftward agreement (2a).1 One of the Italian dialects in question is Friulano. Benincà (1994, 83-84) reports that in the grammar of Friulano, the participle agrees with the following object in (3a).
1
For the problem of auxiliarization and the development of the compound tense, the reader is referred to Salvi (1982), Harris (1982), Vincent (1982), and Bertinetto (1986).
37
Crucially, the object can also appear in a position to the left of the participle as in (3b) (examples from Benincà 1994, 83 footnote 10): (3a) (3b)
O ai comprade une biele giachete. I have bought[+Agr] a nice jacket O ai une biele giachete comprade. I have a nice jacket bought[+Agr]
The curiosity of these data is that rightward agreement on the participle appears to correlate with ‘free’ word order in the sense that (3a) and (3b) do not correspond to two distinct readings; both of them express the meaning of a compound tense. It is now of some interest to see if this is an isolated phenomenon or if there is some generalization to be captured here. I will argue in the following chapters that the pattern illustrated in (3a) and (3b) is not an isolated phenomenon, and that there are several grammars that permit (1b), and also allow (2a) with a reading that makes (1b) and (2a) synonymous. In chapter 2, I will give examples of the same phenomenon from modern languages like Swedish and Iberoromance. In the rest of this chapter, I will present Old Italian data that point at the same kind of correlation as the one illustrated in Friulano. We know that rightward agreement was common in Florentine texts from the 14th century. This is illustrated in (4a):
(4a)
Aux (…) come che tu abbi even if you have
part. O perduti i tuoi denari (…) lost[+Agr] your money (Dec; II:5)
In the relevant period, the object could also appear in leftward position:
(4b)
Aux (…) co’ denari avresti with your money you would have
O part. la persona perduta. your life lost[+Agr] (Dec; II:5)
I will now argue on the basis of 14th and 16th century data that the following generalizations hold: 1. Rightward agreement, as in (4a), is largely predominant in the earliest stage of Italian in the sense that it is very rare to find non-agreeing participles in cases like (4a). 2. As long as participles always agree to their right, as in (4a), (4a) and (4b) do not correspond to two distinct readings. This point will be further explained and clarified in chapter 2.
38
3. Even though examples of rightward agreement are found in written Italian until this century, there is a clear break point between the two varieties I will refer to as Medieval Italian and Renaissance Italian. The change appears to have taken place by the middle of the 14th century, and there are thus systematic differences between the written language of the first half of the 14 century, and the one used from the late 14th century onwards. In order to illustrate my point, I will present a corpus survey of literary texts. We are now addressing an issue that has already attracted the interest of traditional grammarians, but the traditional studies on the matter are somewhat unclear with regard to the correlation between agreement and word order, and the differences over time.2 Among the descriptive studies we may rely on, the most exhaustive appears to be the one of Lucchesi (1962/63). Before proceeding with my own data, I will therefore summarize some of the crucial observations of Lucchesi.
1
The Corpus Survey of Lucchesi (1962/63)
Lucchesi (1962/63) considers a large corpus consisting mainly of 13th century texts. On the basis of these, he reaches the following conclusions: 1. Participial agreement is highly regular in Tuscany in the 13th century, in the sense that transitive participles very frequently agree with their objects (in some texts without exceptions, as it were) independently of where the object appears in relation to the participle (Lucchesi 1962/63, 205). 2. In the beginning of the period, there does not appear to be any clear difference in meaning linked to the two crucial word order alternatives: I have opened the window, cf. (1a) and (4a) above, and I have the window opened, cf. (2a) and (4b). This means that the two constructions were, or could be synonymous (Lucchesi, 205-206; we will return to the matter in chapter 2, section 2.1).
2
Reading Rohlfs’ (1969, § 725) treatment of participial agreement, one has the impression that there are no clear tendencies of the kind mentioned in the text. Thus, Rohlfs claims that it is unimportant whether the object precedes or follows the participle: «Non ha dunque importanza se l’oggetto segue o precede il verbo.» (Rohlfs, § 725) «Le regole stabilite dai grammatici, secondo cui la variazione del participio è legata alla posizione dell’oggetto, sono artificiose.» (§ 725, footnote 2). Furthermore, it emerges from Rohlfs’ discussion that agreement is predominant in the ancient times, but that there is no clear break point in which the participial agreement pattern changes: «Anche scrittori posteriori e moderni si mantengono in parte fedeli a questa regola.» (§ 725). Judging from these lines, neither does the tendency to participial agreement correlate systematically with word order, nor are there systematic differences between the historic stages of Italian. The survey of this chapter will challenge this view.
39
3. When non-agreement participle-object is attested in 13th century Tuscan texts, the contexts in which this happens are of essentially two kinds: a. When the participle does not show agreement, the object is always in rightward position in relation to the participle. Hence, non-agreement on the participle is attested in cases such as I have opened the window, (1a)/(4a), but not in I have the window opened, (2a)/(4b) (Lucchesi, 212-214) b. Agreement is more easily lost when the object is a quantified noun, and very often when the noun is cosa ‘thing’ (which is feminine in Italian), in compounds as ogni cosa ‘each thing’, molte cose ‘many things’, and similar. This is to say that non-agreement is more easily found in a structure as I have seen each thing than in I have seen one/the thing. Lucchesi argues that each thing is semantically equivalent to everything and uses the term ‘mental equivalence’ to define the phenomenon (Lucchesi, 200, 205). The lack of agreement in this context is hardly surprising, given that cosa in Modern Italian has masculine gender when it is used as a quantifier, as in qualcosa ‘something’; see Cinque (1990a, 76). c. Agreement is lost earlier in Medieval French and in the northern Italian varieties than in Tuscan. During the 13th century, there is also variation within Tuscany, in the sense that participles agree more regularly in southern Tuscan varieties than in northern ones.3 Lucchesi’s study concentrated on 13th century use and was less extensive on the 14th century. We am now interested in continuing this survey, comparing the 14th century with the 16th century. Furthermore, the way of organizing the material and classifying the different construction types will be slightly different from Lucchesi’s study.
3
«I testi settentrionali sembrerebbero indicare che, complessivamente, a Nord dell’Appennino l’abolizione dell’accordo ha nel ‘200 una diffusione paragonabile a quella segnalata, nello stesso secolo, per il francese. In Toscana il fenomeno ha una estensione più limitata.» (Lucchesi, 1962/63, 205) «Molto più rigoroso, o addirittura privo di eccezioni, si presenta invece l’accordo in tre testi del ‘200 che ci sono pervenuti in copie antiche, e di cui abbiamo edizioni diplomatiche o molto autorevoli: i Conti d’antichi Cavalieri (codice Martelli), il Tristano Riccardiano e le Storie de Troja et de Roma. La redazione originale di tutte queste opere è avvenuta a Sud della Toscana o ai suoi margini meridionali: aretina la prima, umbra settentrionale la seconda, romanesca la terza; sicché verrebbe fatto di pensare che la tendenza a omettere l’accordo cresce salendo dal Sud al Nord. È un’ipotesi attraente che permetterebbe di inserire la Toscana a mezza strada fra le tendenze conservatrici del Sud e quelle innovatrici del Nord.» (Lucchesi, 251-252) Observe that when Lucchesi arrives at the conclusion that the loss of agreement goes from south to north («cresce salendo dal Sud al Nord»), he is certainly making a lapse. His demonstration is built up to show precisely the opposite, namely that the loss of agreement starts in the north and spreads south.
40
2
Organization of the Present Corpus
2.1
The Constructions
We are interested in knowing when a participle agrees with its direct object, and will therefore consider any construction that involves an auxiliary verb (avere), a transitive participle and a direct object. Such constructions obviously present themselves in a variety of ways and the material will be organized in the following way: 1. Participle+Object, which is the ‘straight’ word order, from a contemporary view point. Constructions that belong to this group can in principle be of two kinds: those where the auxiliary V precedes the VO complex (auxiliary+participle+object: I have seen the woman) and those where it follows the VO complex (participle+object+auxiliary: seen the woman I have). 2. Object+Participle, which represent at the same time a more archaic pattern, but is also superficially similar to the construction illustrated in (2a), perfectly grammatical also in modern use. As under the previous point, there are two options: the auxiliary verb may precede (auxiliary+object+participle: I have the woman seen) or follow (object+participle+auxiliary: the woman seen I have). 3. Objects in Operator Contexts. By operator contexts I mean relative clauses (The woman that I have seen…) and interrogatives (which woman have you seen? or how many women …?). The second type of clause is quite infrequent in my material. There is of course a third category that should qualify namely Focus; trivially, in a written material Focus cannot be distinguished from other preposing operations such as Topicalization and this possibility must therefore be disregarded. 4. Pronominal objects. All cases where the direct object is represented by a clitic pronoun. 5. Preposed participles. Under this group, I will consider those structures where the participle has been preposed, and the object appears on its own to the right of the auxiliary, that is, participle+auxiliary+object: seen (I) have the woman. In addition, I have counted to this group some cases of a certain type of preposing where the participle is followed by the complementizer. I will have very little to say on this case in the analysis:
41
(5)
Udito che ebbono tutti questa sentenzia… heard that had everyone this sentence (Trec: VII)
6. Preposed objects. By this, I mean the structure object+auxiliary+participle: the woman (I) have seen where the object appears alone in sentence initial position being separated from the participle by the auxiliary. 7. Participle+infinitive+object. In the relevant stages of Italian, a participle may agree with a direct object embedded under an infinitive, as in (6): (6)
Voi non avrete compiuta ciascuno di dire una sua novelletta … You have not finished[fem.] each one to tell a novel[fem.] (Dec; I: Intro)
8. Object gap. The last group is only present in a few of the texts examined. It covers cases where no phonetic object is present. The adequate analysis of these will be discussed in chapter 7. In all of these categories, we will find examples of agreeing and non-agreeing participles. We will see below how the agreeing and non-agreeing cases are distributed over the eight construction types. Before we proceed, recall the properties of Modern Italian with respect to these categories. Modern Italian does not have agreement in types 1 and 3. Agreement is obligatory in 2 and 4, but the use of construction type 2 in Contemporary Italian is distinctly different from the ancient texts, as we shall see later. As for 5, 6, and 8, these constructions are not as productive in the contemporary system as some of them appear to have been in the ancient ones. Also, it is difficult to tell how they relate to the dislocation or Focus structures of modern language.
2.2
The texts
I have chosen eleven literary texts that cover a period of 120 years, from the late 13th century until around 1400, and another twelve from a period of around 60 years, from 1500 until 1560. The former will be referred to as the A-corpus, and the latter as the B-corpus. The texts consulted are the following (cf. also the list of Ancient texts cited in the references on p. 330). In the following paragraph, I give the name of the text and the abbreviation I use for it within parenthesis. Then follow the approximate year(s) in which the text was written or published, the name of the author, and his lifetime.
42
2.2.A
The A-Corpus
A.1 Il Novellino (Nov.no.), Late 13th century, anonymous author. A.2 Il Libro de’ Vizî e delle Virtudi (Vizi), Late 13th century, Bono Giamboni 1235 ca. - 1295 ca. A.3a Vita Nuova (VN), 1292-1293, Dante Alighieri 1265-1321. A.3b Il Convivio (Conv.), ca. 1304, Dante Alighieri 1265-1321. A.4 Cronica (Cron.), 1320 ca. - 1348, Giovanni Villani 1280 ca. - 1348. A.5 Decamerone (Dec.), 1350 ca., Giovanni Boccaccio 1313 - 1375. A.6 I fioretti di San Francesco (Fior.), 1370 -1390, anonymous author. A.7 Il Pecorone (Pec.), Late 14th century, Ser Giovanni. A.8 Il Trecentonovelle (Trec.), 1392-1397, Franco Sacchetti 1332 ca - 1400. A.9 Il Novelliere (Nov.re), 1400, Giovanni Sercambi 1347 ca. - 1424. A.10 I Ricordi (RicM.), 1393-1411, Giovanni di Pagolo Morelli 1371 - 1444. This corpus has been sorted in accordance with the following criteria: 1. Hoping to obtain a material which is somewhat uniform, I have chosen to limit the field geographically. Whereas Lucchesi’s corpus contained non-Tuscan texts, both northern and southern, this one is limited to Tuscany. As a matter of fact, the geographic extension of the A-corpus turns out to be even more narrow, as the texts are mostly of a central Tuscan origin. Out of ten authors, seven are, as far as we know, Florentine. The remaining three are: the Novellino, the Fioretti, and the Novelliere. The origin of the Novellino has been a controversial matter for a long time; Ricciardi (1988, 77) finds a Florentine origin plausible, and so also for the Fioretti (Ricciardi, 78).4 Only the Novelliere is written in an occidental variety of Tuscan (the dialect of Lucca). 2. Poetry has been excluded, as considerations of rhythm and rhyme have influence on the syntactic form of a text.5 Those parts of the works presented above and below that are written in verse, are excluded from the survey.
4
Ceserani & de Federicis (1979a, 415) mention the possibility of a Senese origin of the Fioretti. In each event we are dealing with a central Tuscan text; see also the comment of the editor Don Giuseppe de Luca, Fior; 533-534. 5 Lucchesi’s (1962/62) material also includes poetry. In the analysis of syntactic patterns (and especially if the topic is word order) there is always some uncertainty with respect to the reliability of poetic texts; the doubt that the demands of verse might have influence on syntactic form is constantly repeated in, for instance, Ménard’s (1994) descriptive work on Old French.
43
3. In order to offer interesting material, that is, a certain number of examples from all of the construction types listed in 2.1., the texts must be sufficiently long. In particular, they must offer a total of 50 cases (see 2.3. below). 4. Furthermore, if a certain number of examples from all of the construction types are to be found, the language of the texts must be sufficiently varied. Certain types of preposing, for instance, are not likely to be found at all if the style of the text is too colloquial.6 The last point means, in essence, that the material must be of a certain degree of syntactic complexity in order to be interesting. Therefore, the texts chosen are predominantly of a middle or high register.7 There is an important observation to be made in this connection; because of the nature of the constructions we are interested in, we are forced to search in distinctly literary texts. In such contexts, there are problems of style, genre and register that in a different kind of material would have been less noticeable. It must be emphasized that we want to capture the correct generalizations, and that curiosities are of less interest. Still, when confronted with stylistic variation, we cannot pretend it is not there, and we are forced to comment upon it. In particular, it cannot be excluded that stylistic variation due to literary genre or other factors conceal the structures of the grammar, and it is then of some importance to discover the patterns of the grammatical system underlying the written style. These observations have some relevance, above all, when we address the 16th century prose. 2.2.B
The B-Corpus:
B.1 Il Principe (Princ.), 1513, Niccolò Machiavelli 1469-1527. B.2 Dialogo del reggimento di Firenze (Dial.), 1521-1526, Francesco Guicciardini 1483-1540. B.3 I Ragionamenti (RagF.), 1525, Agnolo Firenzuola 1493-1543. B.4 Lettere (LettA.), 1420 - 1550, Pietro Aretino 1492-1556. B.5 Vita (Vita), 1558 ca, Benvenuto Cellini 1500-1571. B.6 Le Cene (Cene), 1540 ca., Antonfrancesco Grazzini (il Lasca) 1503-1584. B.7 La Zucca (Zucca), 1551-1552, Anton Francesco Doni 1513-1574. B.8 Il Libro del Cortegiano (Cort.), 1513-1528, Baldassar Castiglione 1478-1529. B.9 Prose della volgar lingua (Prose), 1525, Pietro Bembo 1470-1547. 6
It is the same corpus, with some additions, that is at the basis for the discussion of Part B of this thesis. When we consider the absolute participle, the need for somewhat advanced texts becomes even more evident, as a colloquial register might not offer any examples at all of what we search for. 7 D’Achille (1990a, 1990b), who studies the development of dislocation in Italian, builds on a corpus which, not unexpectedly, contains several of the works included in ours. D’Achille (1990b, 24) uses a system of stylistic classification of the texts that recognizes three levels: A. writings closest to spoken language; B. writings of a middle register; C. writings of an elevated, high register. According to d’Achille’s classification, the corpus we are working with falls under the categories B and C (D’Achille 1990b, 40-56).
44
B.10 I Lucidi (Luc.), 1549, Agnolo Firenzuola 1493 - 1543. B.11a Dieci lettere private (LettM), 1510-1520, Niccolò Machiavelli. B.11b La Mandragola (Mand), 1518, Niccolò Machiavelli. The criteria for the B-corpus are equal to those of the A-corpus for the first seven texts. The authors are all from the central Tuscan area, and the texts have the sufficient length and variety to offer the kind of material that we are interested in. As we shall see, there is some unexpected variation among the 16th century authors. This circumstance, for reasons that we will discuss as we proceed, will call for a comparison between the Tuscan works and two nonTuscan ones, the Cortegiano (B.8) and the Prose (B.9). Finally, I will add to the B-corpus two texts for the theatre and a short collection of letters, the Lucidi (B.10), the Mandragola (B. 11b), and Lettere Private of Machiavelli (B.11a) that differ from the rest of the corpus with regard to register. The relevance of these additions will be better explained below.
2.3
How the Survey is Carried Out
1. We will study these texts until we have found the first 50 cases of constructions that fall into the categories defined above. The mass of text one has to read in order to reach 50 is highly variable. 50 cases are soon attested in some of the texts; in the Decameron (A.6), we reach that point around the fourth novel of the second day. 50 cases are not as easily attested in others. As a matter of fact, in the Vita Nuova (A.3a), only 24 cases are found. We will continue then, adding other 26 cases from the Convivio (A.3b). Dante is represented in the corpus by 50 cases, taken from two sources, as indicated in the list in 2.2.A. In the B-corpus, the same problem will arise with the texts added as point B.10 and B.11a and b. From these texts, that are introduced in order to offer a point of reference, we will only consider 30 cases out of the two first construction types, object-participle and participle-object (cf. 2.1. above). In order to attest 30 cases of the first two types, we need to put the Mandragola together with the Lettere Private, as indicated in 2.2.B.. Since the A-corpus contains ten texts and we attest fifty cases in each, the Acorpus is 500 examples large. The B-corpus is actually of 510 examples as we count 50 cases of nine texts, and 30 each of the remaining two. The total corpus presented in this chapter thus consists of 1010 examples. What we wish to find out is if there are enough clear tendencies to make a distinction between two grammatical systems justified. As we shall see, there are. We have thereby limited the scope of interest to the first of the three questions formulated in
45
2.3.2. in the Introduction.8 We are interested in showing that there are different stages of Italian, and in defining the differences between them. The corpus is sufficiently large to allow such conclusions. As it happens, the corpus will also give a rough indication about when the change took place. If however, we wanted a deeper understanding of the following two problems; when Italian changed from the first stage to the second, and why it did so, we would need a corpus which (a) is larger, (b) comprehends texts of a broader register, and (c) covers up the period between 1400 and 1500. 2. As for the way to count occurrences, the following circumstances must be remembered: a. We will of course only consider objects that are not singular masculine, since the uninflected participle and the participle agreeing in singular masculine both end in -o. b. All verbal forms, finite and nonfinite, in the ancient texts normally agree with the closest DP in case the argument, subject or object, consists of coordinated nouns. Therefore, only the closest DP counts for our purposes. (7a) must be considered a regular case of agreement, and (7b) is not to be counted at all given that masculine singular objects are excluded. (7a) … egli non abbia tanto grido e tanta fama raccolta dalle genti … he has not much acclamation[m.s.] and fame[f.s.] received[f.s.] from people (Prose: II; III) (7b) dovessero gli antichi uomini considerazione e risguardo avere avuto … should the ancient men consideration[f.s.] and respect[m.s.] have had[m.s.] (Prose: I; V) c. Crucially, I count occurrences of participles. If two participles are coordinated, as in (8), they are counted as two cases: (8)
E non solamente hai uccise e divorate le bestie … and not only have you killed[f.pl.] and devoured[f.pl.] the animals[f.pl.] (Fior; XXI)
8
The questions were: 1. What are the differences between the various stages of a language L?; 2. When did L change from one stage to the other?; 3. Why did L change from one stage to the other?
46
3
The Result of the Survey
3.1
Type 1: Participle+Object
3.1.a
The A-Corpus
In the A-corpus, construction type 1 is attested 225 times in total. These are distributed among the texts in the following way: Total number of occurrences of participle+object in corpus A: A.1 A.2 A.3a/b A.4 A.5 A.6 A.7 Nov.no Vizi VN/Conv Cron Dec Fior Pec 19 18 19 31 12 27 19
A.8 Trec 30
A.9 Nov.re 24
A.10 RicM 26
In table I, the column representing the agreeing form is white, and that representing the nonagreeing form is dark: Table I Part-obj
30
Agreement
26
Non-agreement 22
21 20
18 16
16
15
14
11
16
11 9
10 3
4
4
8
5 3
2
1 0
A.1 A.2 A.3 A.4 A.5 A.6 A.7 A.8 A.9 A.10
It is quite clear that something happens by the middle of the 14th century. In this material, a change is visible between the Decameron (A.5) and the Fioretti (A.6). The agreeing forms are highly predominant in the first five texts, but heavily reduced in the following five. If we
47
divide the corpus in two groups, A.1-A.5 vs. A.6-A.10, the total number of agreeing and nonagreeing forms are the following: Total number of agreeing and non-agreeing forms, participle+object, corpus A: A.1-A.5: A.6-A.10: agreeing participle 82 agreeing participle 33 non-agreeing part. 17 non-agreeing part. 93 Table II 93
100 90
Agreement Non-agreement
82
80 70 60 50 33
40 30 20
17
10 0
A.1-A.5
A.6-A.10
The data indicate that there is a new grammatical system taking over by the second half of the century.9 The differences noted by Lucchesi between quantified and non-quantified nouns are not very clear in my material. In the first period, nouns of both categories tend to agree, but in the last period they generally do not. It should be mentioned that in the Trecentonovelle (A.8), all of the nine agreeing cases are definite nouns, like (9a), whereas the participle never agrees with a bare noun: (9a)
… avea per assedio costretta la città di Forlí … he had by siege forced[+Agr] the town of Forlí (Trec; VII)
9
Also Lucchesi indicates a break point in the literary language after the Decameron: «Va innanzitutto notato che nelle cinque opere scritte prima del 1360 l’accordo, come in quasi tutti i testi toscani del ‘200, prevale anche quando il participio precede l’oggetto.» (Lucchesi 1962/63, 269)
48
(9b)
… come se della porta avesse aúto paura… as if of the door he had had[-Agr] fear (Trec; XII)
There is however a tendency toward non-agreement with cosa in many construction types as we shall see: (9c)
Maestro, i’ ho veduto cosa che molto mi dispiace… sir, I have seen thing[-Agr] that does not please me (Nov.no; LXVIII)
3.1.b
The B-Corpus
We will begin with the seven Tuscan texts. The total number of examples of construction type 1 are the following: Total number of occurrences of type 1, participle+object, in corpus B.1-B.7: B.1 B.2 B.3 B.4 B.5 B.6 Princ Dial RagF LettA Vita Cene 26 28 32 29 21 19
B.7 Zucca 32
In table III, the column representing the agreeing form is white, and the one representing the non-agreeing form is dark: Table III Part-obj 29
30
26
Agreement Non-agreement
20
17
17
16
15 13
13 13
13 8
10
2
2
3
0 B.1
B.2
B.3
B.4
B.5
B.6
49
B.7
We find different patterns in the B-corpus as well. The relation between agreeing and nonagreeing participles in B.1, B.3, and B.4 is around fifty-fifty, whereas the non-agreeing form is clearly predominant in B.2, B.6, and B.7. Generally, the texts of B.2, B.6, and B.7 are quite similar to those of the late 14th century, and therefore also to modern use, since participles in the relevant context do not agree in contemporary standard language. There is a crucial difference between this variation in the B-corpus and the variation of the A-corpus we just saw. In the B-corpus, these differences are not chronologically related since the non-agreeing pattern is predominant in texts B.2, B.6, and B.7 which are the second and the last two texts, chronologically speaking. Therefore, the variation found in the 16th century texts B.1-B.7 can hardly be ascribed to a change in the grammatical system that occurred at that time, but has presumably some different source. I suggest that the grammatical system of the relevant period is best reflected by the texts B.2., B.6., and B.7, and that the relative frequency of agreement in the remaining texts reflects the archaic use, however this is to be understood more precisely; see the discussion in 2.3.2. of the Introduction. There was a controversy in the Renaissance between essentially three different views on how the literary language was supposed to be formed, and what its models should be. Of these schools of thought, two can be called ‘modernist’ in the sense that they proposed that literary language should be founded on the contemporary spoken language, whereas the third view held that the literary language should build upon the style of Petrarch and Boccaccio; hence, that 16th century authors should write in a 14th century language.10 Consider now two non-Tuscan texts: B.8, the Cortegiano, represents one of the modernist views, that of Castiglione, and B.9 is the Prose della Volgar Lingua of Bembo, which is the text that proposes and exemplifies the archaic style. Total number of occurrences of type 1, participle+object, in corpus B.8-B.9: B.8 B.9 Cort. Prose 38 4
10
On the questione della lingua, see Ceserani & de Federicis (1979b, 383-388), Migliorini (1988, 309328), and Ricciardi (1988, 225-227).
50
Table IV Part-obj 40
Agreement Non-agreement
34
30 20 10 0
4
3 B.8
1 B.9
Note first of all that construction type 1 is extremely rare in B.9. This is significant for reasons that we will see in the following pages. In the text of the Cortegiano (B.8), which explicitly tries to reproduce spoken style, the agreeing participle is infrequent when word order is participle+object. This is the same pattern we have already found in the late 14th century texts and, in B.2, B.6, and B.7 Are there reasons to think, then, that the agreeing forms of B.1-B.7 are due at least partially to imitation of archaic style? There may indeed be some reasons to believe this in spite of the fact that some of the authors, as Machiavelli, declared themselves opposed to the 14th century imitation of Bembo. An interesting point of reference would be other texts written by the same authors of B.1-B.7 at a different stylistic level. I have chosen two such points of reference namely from the theatrical production of Firenzuola and Machiavelli.11 In the comedian prose, the presence of archaic imitation is with all certainty less important than it might be in the literary genres where the 14th century had defined the models, such as the short stories. I have taken the Lucidi, a play of Firenzuola, and the Mandragola of Machiavelli, and I have read the first 30 cases of construction type 1. Since
11
Both Firenzuola and Machiavelli are of course inspired by the linguistic example of the 14th century as is most of the 16th century literature. Whereas Firenzuola is closer to the ‘Bembist’ line in the debate, Machiavelli is more inclined to the use of contemporary Florentine as a literary language. The interesting point here is that while the 14th century models were valid for the writing of short stories (novelle), poetry or political texts, there were no such models for the theatrical text since authors of the 14th century did not write for the theatre. This is the simple reason why we can expect there to be differences between the theatrical production and other works of the same authors. Migliorini points out this circumstance with reference to a text of Machiavelli: «Quando poi i non Toscani affrontino generi letterari in cui manchino modelli antichi, com’è specialmente la commedia, debbono ricorrere al toscano …» (Migliorini 1988, 321) Clearly, this holds not only for the nonTuscans, but also for the Tuscans themselves.
51
the latter text is not sufficiently long to offer thirty examples, I have added a collection of private letters from the same period as the Mandragola (as indicated above, 2.2B.): Total number of occurrences of type 1, participle+object, in corpus B.10-B.11: B.10 B.11a/b Luc LettM/Mand 30 30 Table V Part-obj 24
24 24
Agreement Non-agreement
20 16 12 8
6
6
4 0
B.10
B.11
In the theatrical prose, as well as in the private letters, Machiavelli and Firenzuola also conform to the non-agreeing pattern. Compare B.10 and B.11 with B.1 and B.3 in table III. (10a)
…e di Pisa aveva presa la protezione. and of Pisa (he) had taken[+Agr] the protection (Princ; VII)
(10b)
… tu hai scacciata la tempesta… you have chased[+Agr] the tempest (RagF; I:Intro)
(10c)
… e aver trovato la casa mia… and having found[-Agr] my house (Luc; I:2)
52
(10d)
Ho lasciato dunque i pensieri delle cose grandi … I have left[-Agr] thus the thoughts of great things (LettM; VII)
Presumably then, the change we have observed by the second half in the 14th century texts reflects a change in the grammatical system, that is, the change from the ancient grammar where participial agreement in construction type 1 is obligatory, to the new grammar where it is not required and successively dies out. It is possible that the variation attested among 16th century authors is a variation in literary use rather than a case of optionality in the grammar. The agreeing forms of construction type 1 may be the fruit of stylistic imitation having the status of a superficial phenomenon external to the grammatical system. There are however alternative accounts to explore. The authors that make use of agreeing participles in type 1 may do so because they make reference to two grammatical systems, both of which internal. This idea is in line with the theory of diachronic change outlined by Kroch (1989, 1994) and Santorini (1989). It is not my intention to discuss these possibilities.
3.2
Type 2: Object+Participle
3.2.a
The A-Corpus
Construction type 2 is less frequent than type 1 and certainly represents a more ancient system. It is, namely, more frequent in the first half of the 14th century than in subsequent literature. The total number of cases is 37, divided among the authors in the following way: Total number of occurrences of type 2, object+participle, in corpus A: A.1 A.2 A.3a/b A.4 A.5 A.6 A.7 A.8 Nov.no Vizi VN/Conv Cron Dec Fior Pec Trec 2 10 3 2 13 2 0 0
53
A.9 Nov.re 2
A.10 RicM 3
Table VI
Obj-part 20
Agreement Non-agreement 13 10
10
3
2 0
0
2 0
1 1
0
3
2 0
0
0
0
0
0 A.1 A.2 A.3 A.4 A.5 A.6 A.7 A.8 A.9 A.10 The total number of agreeing cases is 36. Thus, the participle always agrees with the preceding object, except in (11), which is from Villani (A.4): (11)
… avendo alcuna cosa spirato… having some thing aspired[-Agr] (Cron; 40)
(11) falls under the phenomenon Lucchesi called mental equivalence (see 1, point 3.b. above). In other terms, it is clear that cosa, at least when preceded by quantifying expressions, in the ancient grammar can have the status of QP rather than DP, which is similar to the state of affairs in Modern Italian. I conclude that the participle agrees obligatorily in construction type 2. What is interesting here is that the construction itself seems to have been reduced in a way which is more or less parallel to the reduction of participle agreement in construction type 1. If we compare the frequency of construction type 2 in the two halves of the century, we find that the relation is 30:7. Total number of occurrences of type 2, object+participle, in corpus A: A.1-5 A.6-10 30 7
But the difference between the first and the second half of the century is more profound than the statistics suggest. It is important to point out that construction type 2 covers structures of
54
essentially two kinds: firstly, there are cases that must be understood as compound tenses. These are ungrammatical in Modern Italian with the word order of construction type 2:
(12)
Aux O part. (…) co’ denari avresti la persona perduta. with your money you would have your life lost[+Agr] (Dec; II:5)
Secondly, there are cases where construction type 2 is ambiguous between a compound tense and a possessive, resultative reading. This goes for (13):
(13)
Aux O … chi è quel Vizio che ha già le sue genti who is that vice that has already his people (Vizi; XXX)
part. schierate … gathered
(13) can be understood as who has already gathered his people or who has his people gathered.12 If we look more closely at the 30 occurrences attested in A.1-A.5 and the 7 occurrences of A.6-A.10, it is obvious that the clear cut cases of compound tenses generally belong to the first group. In A.1-A.5, I have counted 20 unambiguous examples of compound tenses in construction type 2. Consider some of them (others will be given in chapter 2, section 2.1.): (14a)
Messere, perch’elli ha bene morte servita. Sire, because he has well death deserved (Nov.no; LXIII)
(14b)
… che avessero sì leggiere le non fittizie parole apprese… that they had so easily the non-false words learned (Conv; II:XII)
(14d)
… e perciò aveano la detta moneta recata. and therefore they had the mentioned money given (Cron; 101)
12
(13) is only apparently a clear case of possessive have followed by a resultative small clause. Given that the relevant word order could correspond to a compound tense reading, witness (12), we cannot be sure of the status of (13); see below chapter 2, section 2.1.
55
(14e)
hai tu mai testimonianza niuna falsa detta contra alcuno… have you ever a false testimony said against someone (Dec; I:1)
Construction type 2 in the second part of the A-corpus conforms to modern use in the sense that the value is no longer that of a compound tense. Consider four of the seven examples from A.6-A.10: (15a)
…essendo uomo di grande contemplazione, avea la mente sospesa e being a man of profound contemplation, he had his mind raised and levata a Dio… elevated toward God (Fior; III)
(15b)
… ch’ell’abbia le spese assegnate … because she has her expenses assigned (RicM; 218)
(15c)
… se non hanno anni quinidici compiuti… if they do not have fifteen years accomplished (RicM; 223)
The only certain cases of compound tenses are the following two: (16a)
Lo sommo e potente Dio, dal quale tutti i beni derivano, the highest and almighty God, from whom all good derives, ha lla natura umana creata … has the human nature created (Nov.re; Intro)
(16b)
A colui il quale sen cagione ha di molte ingiurie sostenute … to he who without reason has many offences supported (Nov.re; 10)
Furthermore, there is another circumstance that deserves comment. Some structures belonging to construction type 2 have not been included in the survey, namely those where the object is a quantifier without following DP, unless of course the quantifier is inflected as in (14b) above. Such cases are quite frequent in the first half of the A-corpus, but less so in the second half:
56
(17a)
Quelli rispuose ch’avea tutto donato. he answered that he had all given (Nov.no; XIX)
(17b)
… nella sua corte avea molto dispeso e donato. in his court he had much spent and given (Nov.no; XXX)
(17c)
Perché m’hai tanto tolto? why have from-me[cl.]-you so much taken (Nov.no; LXXII)
(17d)
… io era quelli che li avea ciò mandato. I was the one who to-her[cl.] had that sent (VN; III)
(17e)
… e dico che io hoe ciò perduto. and I say that I have that lost (VN; VII)
(17f)
E quando avemmo assai veduto, … and when we had enough seen (Vizi; XV)
(17g)
E quando ebbe così detto … and when she had so said (Vizi; XV)
These observations confirm our general impression, namely that there are systematic differences between the first and the second half of the 14th century. The difference consists in the fact that construction type 2 after the middle of the century becomes more infrequent, and above all, it takes on the meaning which is associated with it in modern language.
3.2.b
The B-Corpus
Construction type 2 in the B-corpus is very rare. Once again, the prose of the 16th century authors is similar to that of the second half of the 14th.
57
Total number of occurrences of object+participle in corpus B: B.1 B.2 B.3 B.4 B.5 B.6 B.7 Princ Dial RagF LettA Vita Cene Zucca 0 1 0 0 0 3 0
B.8 Cort 0
B.9 Prose 14
In Luc and Mand/LettM, there are no examples of construction type 2. The three cases of object+participle attested in the Cene and the 14 cases of the Prose are all agreeing. It is now clear why construction type 1 turned out to be so infrequent in the Prose (B.9.); the periphrastic tense in the Prose is normally expressed as construction type 2. (18) is what we have called an unambiguous case of a compound tense: (18)
… egli ha senza sua fatica quella lingua nella culla he has without effort that language in his cradle e nelle fascie apparata… and swaddling-clothes learned (Prose; I:XVI)
This means that the word order pattern typical of Boccaccio (A.5), is uncommon among 16th century writers with the exception of Bembo (B.9). Interestingly, it is distinctly more frequent in Bembo than it was in the first half of the 14th century. The single example from Guicciardini is somewhat curious. The preposing of object+participle to the auxiliary is very rare among 16th century authors. The noun is evidently cosa, however without the quantifier, and agreement is not triggered on the participle: (19)
… che non ci pare, …, cosa più strana veduto avere. that it does not seem to us thing more strange seen have (Dial; I)
As for the cases attested in the Cene (B.6), they are all of the ‘modern’ kind, as the adequate reading is presumably a possessive-resultative one: (20a)
Ma poi ch’egli ebbe le sue cose acconce e divisate … but after he had his things settled and divided (Cene; I:2)
(20b)
… avendo la tenera gola aperta … having the tender throat open (Cene; I:5)
58
3.2.1
Conclusions on the Basis of Construction Types 1 and 2
We are now in a position to draw the following conclusions: 1. Agreement in construction type 1, participle+object, is highly regular in the beginning of the 14th century, but gets heavily reduced by the second half of the same century. The 16th century system is similar to the system of the latter half of the 14th century. 2. Agreement is obligatory in construction type 2, object+participle, in both periods, but the construction type itself is reduced by the second half of the 14th century, and is not attested at all in most of the texts from the Pecorone (A.7) onwards. Hence, ‘rightward’ agreement participle-object was regular in Old Italian as long as the object could also appear in a position to the left of the participle. The pattern of Friulano illustrated at the beginning of this chapter, examples (3a)-(3b), was typical for the Old Italian grammar, and a correlation between word order and agreement does indeed seem to exist. Furthermore, construction types 1 and 2 could correspond to meanings that are synonymous. In other words, they could both have the meaning of a compound tense, until the middle of the 14th century. The possessive-resultative value of construction type 2 seems to have been consolidated after the Decameron (A.5). 3. The Prose (B.9) makes the only clear exception from the rule as agreement is regular and construction type 2 is frequent. The grammar of the Prose is similar to the early 14th century system. The intention was to descriptively define two grammatical systems, and in order to do so we have studied a corpus of the 14th century and one of the 16th century. What we have attested is indeed two grammatical systems, but not quite as we had expected. It so happens that the second system takes shape by the middle of the 14th century. The labels Medieval Italian (henceforth: MI) and Renaissance Italian (henceforth: RI) cannot be used in the way we first intended. Rather, with regard to the phenomena under discussion, the line between the two grammatical systems is to be drawn between A.5 and A.6, as indicated by table VII:
59
Table VII
MI
RI
Corpus A 1250
1300
1350
Corpus B 1400
1450
1500
1550
1600
The result is not totally surprising, however. Migliorini (1988, 182) treats the 14th century until the year 1375, which thus becomes the line between the medieval period and the new era.13
3.3
Type 3: Objects in Operator Contexts
3.3.a
The A-Corpus
The cases to be considered under construction type 3 are almost exclusively relative clauses. The cases of the MI corpus are predominantly agreeing. Consider table VIII: Total number of construction type 3, objects in operator contexts, in the A-corpus: A.1 A.2 A.3a/b A.4 A.5 A.6 A.7 A.8 A.9 Nov.no Vizi VN/Conv Cron Dec Fior Pec Trec Nov.re 6 9 8 9 7 10 9 8 5
13
A.10 RicM 6
«… perché l’ultimo quarto del secolo meglio si ricongiunge, per l’umanesimo ormai dominante, con le tendenze del Quattrocento.» (Migliorini 1988, 182)
60
Table VIII
Operator Obj. 9 9
8
8 7
Agreement
8
Non-agreement
7 6
6
6
6
5
5
4 4
4
4
3
3 2 1 0
1
1
1
1
1
1
1
0 A.1 A.2 A.3 A.4 A.5 A.6 A.7 A.8 A.9 A.10
As we can see from table VIII, agreement is regular in the relative clause for the MI authors through the major part of the century. Note that the clear break observed in construction type 1 between A.5 and A.6 is not reflected in the relative clause. However, among the last five texts of corpus A there are differences. In A.6, A.7, and A.10, agreement is regular as in the first five texts. In A.8, agreement appears to be optional and in A.9, the non-agreeing cases prevail. (21a)
le pietre ch’avea donate allo ’mperadore the stones that he had given[+Agr] to the emperor (Nov.no; II)
(21b)
… quella medesima risposta la quale avea avuta frate Silvestro. the very same answer that had had[+Agr] brother Sylvester (Fior; XVI)
(21c)
tutti i luoghi santi che oltre mare avea vicitato all the sacred places that he had beyond the seas visited[-Agr] (Trec; X)
The material is too small to allow for any definite conclusions, but the general impression of variation in use is strengthened when we consider the B-corpus.
61
3.3.b
The B-Corpus
The total number of examples from construction type 3 in the B-corpus is 79. Total number of occurrences of type 3, objects in operator contexts, in the B-corpus: B.1 B.2 B.3 B.4 B.5 B.6 B.7 B.8 B.9 Princ Dial RagF LettA Vita Cene Zucca Cort Prose
16 9 2 14 14 1 8 7 8 When we look closer at the relation between agreeing and non-agreeing participles in construction type 3, there is an interesting variation among the texts: Table IX Operator Obj. 14 14
Agreement 11
12
Non-agreement 9
10 7
8
7 6 5
6
4 3
4 2 2
3 2
2 1
2 0
1 0
0 B.1 B.2 B.3 B.4
B.5 B.6 B.7 B.8 B.9
Recall the conclusion from the previous section: the period we refer to as RI begins by the middle of the 14th century. In RI, participles tend not to agree in construction type 1, and construction type 2 becomes infrequent.14 In construction type 3, the RI texts, which by now means all of the texts from A.6 to B.9, appear to be divided into two categories: those in which the participle agrees, and those in which it does not. The former of these two reminds strongly of the use of Modern French where participles do agree in construction type 3, whereas the latter are closer to Modern Italian where agreement is not realized in the relevant context. Let us call the former RI1 and the latter RI2. 14
In particular, construction type 2 with the meaning of a compound tense by and large disappears after the middle of the 14th century.
62
(22) RI1 Agreement in operator contexts: Fioretti (Fior, A.6) Ser Giovanni (Pec, A.7) Morelli (RicM, A.10) Machiavelli (Princ, B.1) Aretino (LettA, B.4) Doni (Zucca, B.7)
RI2 Non-agreement in operator contexts: Sercambi (Nov.re, A.9) Guicciardini (Dial, B.2) Cellini (Vita, B.5) ?
For Sacchetti, Castiglione, Firenzuola, and Grazzini it is impossible to tell what the general patterns are. It appears as if Cellini’s (Vita, B.5) use is rather of the RI2 kind. It must be emphasized that the material is far too small to allow any definite generalizations, even for the authors listed in (22). In spite of the fact that the corpus is restricted, there is one point to be made though, and it is quite sufficient for present purposes: the loss of agreement in construction type 1 is not immediately followed by a loss of agreement in construction type 3. Agreement in the periphrastic tense, construction type 1, is very infrequent in A.7, A.10, and B.7, texts in which agreement is predominant in construction type 3. Consider some of the examples: (23a)
la fede che il re avea data al papa the faith the king had given[+Agr] to the pope (Princ; III)
(23b)
gli altri vini che mi avete mandati the other wines that you have sent[+Agr] me (LettA; IX)
(23c)
quelle paure che tu hai detto those fears that you have said[-Agr] (Dial; I)
(23d)
quelle tale opere, che lui aveva visto those works that he had seen[-Agr] (Vita; XIV)
Two conclusions can be drawn from this: 1. Agreement in construction type 1 and construction type 3 are distinct phenomena, in the sense that the latter obviously does not depend on the former. 2. Whereas agreement in construction type 1 is reduced more or less together with the objectparticiple word order, agreement survives in construction type 3.
63
The analysis, that follows in chapter 4, must capture both of these facts. Note that both of these observations would have to be made independently on the basis of modern data, since Modern French has participle agreement in type 3, but not in type 1. However, our survey has brought about a novelty into the study of participle agreement; it appears as if Italian on its way from an original stage (MI), where participles agree with the DP object in virtually all contexts, to the modern stage, where agreement is limited mainly to construction types 2 and 4, passes through a stage similar to Modern French (RI1), where agreement is regular in construction type 3, but not in type 1.
3.4
Type 4: Pronominal Objects
There are no surprises in the agreement patterns of construction type 4: there are very few exceptions to the rule that participles must agree with clitic objects. In the A-corpus, there is a total of 107 cases out of which 102 are agreeing forms and 5 non-agreeing. In the B-corpus, the corresponding figures are: 77 for the total number of examples, 70 agreeing and 7 nonagreeing. Agreement appears to be optional, in some but not all texts, at least when the clitic pronoun is ci and ne/‘us’ and ‘thereof’. This is not unexpected as we know that the same pronouns allow for optional realization of agreement even in Modern Italian (Salvi 1991, 239). It is interesting to notice however, that there is only one case found of non-agreeing participles before A.6. The example is from the Novellino, and the pronoun is a clitic si with the value of a reciprocal: (24)
Questi due cavalieri s’aveano lungamente amato. these two knights si[cl.]-had for a long time loved[-Agr] (Nov.no; XXXIV)
It is not excluded that a larger descriptive survey of ancient Italian use would discover differences even on this point between the first and the second half of the 14th century. I have no interesting suggestions to make about these cases. (25a)
ci ha qui guidati it us[cl.]-has here guided[+Agr] (Dec; 28)
(25b)
per molti dì ci hanno ingannato for many days they have us[cl.]-deceived[-Agr] (Cort; XVI)
64
3.5
Type 5: Preposed Participle
3.5.a
The A-Corpus
The preposed participle is not frequently found in the earliest texts. It is attested only 10 times in the A-corpus. Total number of occurrences of type 5, preposed participle, in corpus A: A.1 A.2 A.3a/b A.4 A.5 A.6 A.7 A.8 Nov.no Vizi VN/Conv Cron Dec Fior Pec Trec 0 0 2 0 1 3 0 2
A.9 Nov.re 2
A.10 RicM 0
Interestingly, the three cases attested in A.3 and A.5 are all of the kind illustrated in (26): (26)
Perduta ho la fatica… lost (I-)have the fatigue (Dec; I: 2)
From the Fioretti (A.6) onwards, the other type of participle preposing (27) is occasionally found: (27)
E, letta che l’ebbe … and read as (he) it[cl.]-had (Fior; V)
Agreement in these few cases is somewhat irregular. When the construction involves a clitic pronoun, as in (27), the participle agrees in three cases out of three. When the object is DP, the participle agrees in three cases out of seven.
3.5.b
The B-Corpus
The construction is somewhat more current in the B-corpus, mainly due to its relative frequency in the Cene (B.6). The total number of occurrences is 18. The total number of occurrences of type 5, preposed participle, in corpus B: B.1 B.2 B.3 B.4 B.5 B.6 B.7 B.8 Princ Dial RagF LettA Vita Cene Zucca Cort 2 0 0 0 2 9 0 2
B.9 Prose 3
Out of 18 cases, 6 involve a pronominal clitic, and the participle agrees in all of these. In 12
65
cases, the object is a DP, but only in four the participles show agreement. Since the preposed participle is a somewhat rare phenomenon, it is not very interesting to present the result in a diagram. The Cene is the only text in which construction type 5 appears with some frequency, and the tendency is toward non-agreement. The nine examples from the Cene do all instantiate a DP object, and agreement is attested only in two of these nine cases.
3.6
Type 6: Preposed Object
3.6.a
The A-Corpus
Construction type 6, the preposed object, is attested 22 times in the A-corpus. Total number of occurrences of type 6, preposed object, in the A-corpus: A.1 A.2 A.3a/b A.4 A.5 A.6 A.7 A.8 Nov.no Vizi VN/Conv Cron Dec Fior Pec Trec 8 1 4 0 3 2 0 0
A.9 Nov.re 3
A.10 RicM 1
Generally, the participle agrees. I find only three non-agreeing cases out of 22; there are two from the Novellino (A.1), and one from the Convivio (A.3b). It is clear then, that the participle agrees when the object is preposed just as it does in clitic left dislocation structures in Modern Italian. Once again, one of the exceptions instantiates cosa. (28a)
La signoria di Roma t’ho data. the seigniory of Rome have I given[+Agr] to you (Nov.no; LXXII)
(28b)
… c’hai vinto colui che tutte l’altre cose ha vinto. that you have won him who all the other things have won[-Agr] (Nov.no; LXX)
3.6.b
The B-Corpus
There are 17 cases of construction type 6 in the B-corpus. Total number of occurrences of type 6, preposed object, in the B-corpus: B.1 B.2 B.3 B.4 B.5 B.6 B.7 B.8 Princ Dial RagF LettA Vita Cene Zucca Cort 1 0 0 1 5 1 1 0
66
B.9 Prose 8
In 15 out of these cases, the participle agrees with the preposed object. The two non-agreeing examples are the ones from the Cene (B.6) and the Zucca (B.7). (29a)
O mio padre, quella fanciulla ho desiderata e amata… o my father, that girl have (I) desired[+Agr] and loved[+Agr] (Vita; III)
(29b)
… alquanti anni della mia fanciullezza ho fatti nella Provenza … some years of my childhood have (I) made[+Agr] in Provence (Prose; I:VII)
(29c)
… una favola ci abbia raccontato… a fairy tale to-us[cl.]-has (he) told[-Agr] (Cene; I:2)
3.7
Type 7: Participle+Infinitive+Object
3.7.a/b
A+B-Corpus
In the A-corpus, I have attested 4 cases where a participle agrees with a DP object which is embedded under an infinitive, versus 17 cases where it does not. In the B-corpus, the relation is 13 cases of non-agreement versus 1 case of agreement. It is clear that agreement in this context was not predominant at any time, but does occasionally occur in my material. Construction type 7 differs from the others in having optional, and very rare, agreement throughout. 3.8
Type 8: Object gap
3.8.a/b
A+B-Corpus
In the entire corpus, there are three cases of participles agreeing with non phonetic objects. These are attested in the Novellino (A.1), in the Decameron (A.5), and in the Fioretti (A.6). (30a)
E noi avemo rifatta la cittade e raforzata _ … and we have remade[f.s.] the town[f.s.] and reinforced[f.s.] _ (Nov.no; LXXXI)
(30b)
… avendo già ragunati molti compagni e ricevuti _ all’Ordine… having already gathered[pl.] many companions[pl.] and received[pl.] _ to the order (Fior. XV)
67
(30c)
… ancora che vecchio fosse sentì subitamente non meno cocenti even though he was old, (he) felt immediately not less boiling gli stimoli della carne che sentiti avesse _ il suo giovane monaco. the ‘stimuli’ of the flesh than felt[pl.] had _ his young monk (Dec; I:4)
4
Concluding Remarks
I conclude from these data that it is justified to distinguish between two major grammatical systems in the material under discussion, and that the first is replaced by the second by the middle of the 14th century. I have chosen to refer to the first grammar as MI and to the latter as RI (cf. table VII under 3.2.1 above). The difference is visible, above all, in construction types 1 and 2. With some probability, there is more than one grammatical system in circulation among the RI authors, and this is possible to detect in construction type 3 where I have distinguished between two varieties, RI1 and RI2, according to (22) in section (3.3b) above. Construction type 4 has quite the same properties as in Modern Italian, and for the remaining types the material is too small to indicate any clear tendencies. As for the analysis of the data, I will discuss construction types 1, 2, 5, 6, and 7 in chapter 2. Construction types 3 and 4 will be analyzed in chapter 4, and type 8 will be saved for chapter 7. Chapter 3 is also dedicated to the discussion of construction type 2, but on the basis of modern data.
68
CHAPTER 2 The AgrO Node: Word Order and Spec-Head Agreement in Auxiliary + Participle Constructions
0
Introduction
In this chapter, I will begin to analyze the data presented in chapter 1. Although the ancient Italian texts under examination instantiate numerous cases of ‘rightward’ agreement, it will be argued that a Spec-head agreement analysis is nevertheless adequate. Both word order variations and agreement patterns in the medieval period are readily explained in the Antisymmetry framework of Kayne (1994), on the assumption of a null object expletive in Spec AgrOP. The analysis draws some inspiration from Chomsky (1993), but will however diverge from Chomsky (1995), as I continue to assume the Agr category. The chapter is organized as follows: in section 1, I will discuss some basic assumptions of the theory of null subjects as elaborated in the Government and Binding format and discuss how to interpret them in MPLT. In 2.1., I turn to some medieval Italian data, which will be briefly compared with Modern Iberoromance 2.2. and Modern Swedish 2.3. The analysis will be presented in sections 3 and 4, and in 5 we will discuss how to capture the similarities and differences between the word order patterns in Medieval Italian and in Modern Germanic.
1
Preliminary Discussion
1.1
Spec-Head Agreement, Inversion, and Rightward Agreement
At least since Chomsky (1986b), it is commonly assumed that agreement between a DP subject and the finite verb is the syntactic reflex of the local relation between a functional head (AgrS) and its specifier. In Kayne (1989a), the conclusion is generalized to participial agreement through the introduction of an Agr node corresponding to the object, assumed also by Chomsky (1993). Agreement between X and YP would then be spelled out in configuration (1) and not (2): (1)
[XP YP [X’ X]]
(2)
[XP [X’ X YP]]
69
The idea is descriptively successful, as it has been repeatedly observed that agreement between a verbal head and a DP argument, in Germanic and Romance for instance, is typically manifested when the argument appears to the left of the V and rarely when it appears to the right. Following Kayne’s (1994) version of X’-theory, ‘to-the-left’ would mean in a specifier and ‘to-the-right’, presumably, in a complement (but see below for other possibilities). Under such assumptions, the theory must offer some additional device to explain the phenomenon of rightward agreement as in cleft sentences or inversion structures: (3)
Sono io che … it is[+Agr] I that …
(4)
Hanno telefonato i tuoi amici. have[+Agr] telephoned your friends
(5)
Sono arrivati alcuni dei miei amici. have[+Agr] arrived some of my friends
In work by Chomsky (1981, 1982), Rizzi (1982), and Burzio (1986), the solution was linked to the theory of null subjects.1 This was resolved so as to derive the phenomena of ‘rightward agreement’ and ‘free subject inversion’ from the property of pro-drop. It was thus crucially assumed that the inversion of the subject in (4)-(5)2 is possible because Italian can license a null copy of the subject that occupies the Spec I position in which nominative is assigned. The fact that the verb agrees with the subject to its right is readily explained on the same assumption: V raises to a functional head X and shows agreement with a null element in Spec X that copies the overt DP left in situ. The Minimalist Program (Chomsky 1993, 1995) changes the premises of our discussion in predominantly two ways. 1
For discussion on pro, Case chains and inversion phenomena, the reader is also referred to Chomsky (1986a), Brandi & Cordin (1981, 1989), Shlonsky (1990), Cardinaletti (1994a, 47-50; 1994b), as well as Rizzi (1986a) for inversion phenomena in adjectival small clauses (to which we return in chapter 8). See also Belletti (1988), Lasnik (1992), and Moro (1993) for alternative views. Notice that the cleft construction in (3) is slightly different from the others (cf. Moro 1993, 29), as the subject cannot appear in the leftward position. 2 The question whether the DP subject is in its basic position in (5), right-adjoined to VP in (4), or occupies some other position is not crucial. There must in all circumstances be a null element in Spec I. Even if nominative can be assigned under government to the subject in the adjoined position in (4), and even if the subject of (5) carries some Case different from nominative (e.g. Belletti 1988 and Lasnik 1992), the Extended Projection Principle requires the presence of an expletive in Spec I. For some discussion on the Extended Projection Principle in MPLT, I refer to Chomsky (1993, 1995), Bures (1992), Platzack (1994), and my own discussion in chapter 3, section 4.3. below.
70
1. The notion of Greed makes it difficult to consider the realization of agreement as a mere reflex of the derivation. If agreement is spelled out, it is because its realization is required for the derivation to converge, otherwise it would be blocked. Through the introduction of Greed, the basic assumptions of the pro theory are strengthened, or this is at least one possible interpretation. When reformulated in terms of the MPLT, the conclusions of the pro theory, as originally assumed in the above studies, will actually be forced on theoretical grounds: if a DP can raise to subject position, Spec AgrS, in overt syntax, it is because it has to. There must be a feature present on the DP itself that requires overt checking. Otherwise its raising would be blocked by both Procrastinate and Greed. If then, the DP can also appear somewhere else, generally in final position, we must assume that the relevant feature is checked anyway, and checking can only take place in a Spec-head relation with AgrS, if the DP is a subject. If Procrastinate, Greed or feature checking in the Spec-head configuration, all of which are crucial to MPLT, are to be maintained, the assumption of a null element in Spec AgrS that copies the overt DP will be forced, thus distinguishing languages that allow for null subjects, rightward agreement, and free inversion from those that do not. Furthermore, as head government is eliminated from the theory, a possible alternative is no longer available, namely the direct Case assignment from Infl to a right adjoined subject, as in (4). 2. In MPLT, fully inflected words are inserted from the lexicon and move in syntax for reasons of checking. An obvious interpretation is to say that whenever a predicate carries overtly realized agreement morphology, the corresponding abstract φ-features are strong (on Agr and on the DP) and must be checked in overt syntax. The minimalist approach to agreement would then be a mere restatement of Kayne’s (1989a) agreement hypothesis. There is an alternative, however, namely to assume that overtly realized agreement may be either strong or weak.3 The consequence of this would be that a pair of items, V and DP, could be inserted in syntax with overt Agr features, which however count as weak and are checked in covert syntax. If V is a participle and DP a direct object, we would have VO word order, and as result, rightward agreement. If V is a finite verb, DP the external argument, and the features in question count as strong only on the V, V will be raised in overt syntax and DP will be left in situ. Again, we would have a case of rightward agreement. In such a view, the explanatory force of the Spec-head agreement hypothesis is
3
Or rather, correspond to features that are either strong or weak «For ease of exposition, I sometimes speak of a functional category as strong when I mean, more explicitly, that one of its features is strong.» (Chomsky 1995, 232-233) Henceforth, I will not remind the reader of this distinction.
71
lost.4 The fact that predicates with striking regularity agree with DPs to their left and not to their right, would become coincidental. To be more precise, there are languages where verbal forms (finite or not) generally agree with DPs or nominal expressions appearing to their left and only in this case; French and certain Italian varieties can be characterized in this way, for instance. What we do not know of is a grammar where verbs always agree to their right and never to their left, a situation that one would expect to find, if the realization of agreement morphology and word order are independent in principle. It would be perfectly possible for the spell out of agreement to be required for some non-syntactic reason, while the features corresponding to Agr count as weak. It is important to recall that in Romance and Germanic, where we otherwise find a wide variation in agreement patterns, such a grammar has not yet been attested. The empirical basis for a correlation between word order and agreement, in the sense that V generally agrees ‘leftward’, appears to be solid, and I will insist on it throughout the demonstration. This observation makes sense in a framework that only admits Spec-head agreement and where the linear ordering specifier-head is not subject to cross-linguistic variation.5 I will thus assume that overtly realized agreement morphology always counts as strong and requires overt checking. Generally speaking, if we follow the Spec-head agreement hypothesis, there are two ways to address the problem posed by rightward agreement, both of which originate in Rizzi (1982): 1. As above, we can posit a null pronominal that moves in overt syntax allowing the DP to stay in situ. 2. We can assume some version of the V-to-Comp strategy; that is, that the DP is overtly raised to the specifier of an Agr head, after which the verb is raised to this Agr head and beyond it, in order to appear higher up in the structure than the DP (Kayne 1989, Belletti 1990). The first of these solutions, 1 but obviously not 2, predicts a correlation between rightward agreement and the inversion property. There is little doubt that both of these approaches can be defended empirically. We are now primarily interested in the first and the aims of this chapter are the following: I will argue that the predicted correlation between ‘inversion’ and rightward agreement is fundamentally correct and that it is valid for 4
Chomsky’s view on the matter is not entirely clear: «There is at least a tendency for φ-features to be overtly manifested when raising to the checking domain is overt rather than covert, as in verbal agreement with subject versus object in nominative-accusative languages with the EPP, or visible participial agreement in French as a reflex of overt raising. In the Move F theory, the difference reduces to [Spec, H] versus [H F H] constructions, φ-features tending to be overt on H in the former but not the latter. Let us tentatively assume this to be the case, though a principled explanation is lacking, and the empirical facts plainly require much closer scrutiny over a far broader range.» (Chomsky 1995, 277) 5 Kayne (1994, 49-52) adds empirical arguments to the one mentioned in the text. One case of correlation between word order and (number) agreement is stated as Greenberg’s (1966, 94) Universal 33. There are languages where a DP subject preceding the verb triggers agreement on the latter while a DP subject following the verb does not. The opposite situation is not attested.
72
accusative marked DPs in participial constructions. Finite contexts will not be discussed. Moreover, I will show that the phenomena observed in Medieval Italian pattern with similar constructions in Modern Iberoromance and Modern Germanic. The discussion will support an analysis in terms of Spec-head agreement and of Antisymmetry.
1.2
Some Further Notes on ‘Free’ Inversion
The null subject theory thus derives a cluster of properties from essentially one assumption. Against this, it is argued by Safir (1986) that free inversion and null subjects are distinct properties and should be treated separately. Arguing on the basis of Romance data, Safir claims that there are languages that have null subjects but do not admit free inversion, and vice versa, that some languages have free inversion but do not have null subjects. As Safir, I will limit my discussion to Romance and begin with the last objection. Safir proposes as an example of a grammar which has free inversion but not null subjects certain Italian varieties described and discussed by Brandi & Cordin (1981, 1989), Vanelli, Renzi & Benincà (1985), Rizzi (1986b), Poletto (1993), and Benincà (1994; see above all chapters 2 and 5). The relevant varieties have subject clitics that Safir takes to have the same grammatical status as French pronominal subjects, but still display the inversion property with DPs as opposed to French. His conclusion goes against that of Brandi & Cordin, who argue that subject clitics in the Italian dialects are heads, or more precisely, the overt realization of the Agr element in Infl. The conclusion that northern Italian dialects differ from French in this regard is also defended by Vanelli, Renzi & Benincà (1985), Poletto (1993), and Benincà (1994).6 It seems highly unlikely that the grammatical status of subject clitics in dialectal Italian is identical to that of pronominal subjects in French. As Safir himself notes (p. 338 f.n. 7), subject clitics of Trentino cannot be coordinated, whereas French pronominal subjects can. Furthermore, it is significant that some subject clitics in some of the Italian dialects can appear between the negation and the finite verb, which is not the case for French pronominal subjects. I take it that the conclusion of the above cited authors holds (contra Safir), namely that the pronominal elements in dialectal Italian are indeed clitics and that the relevant dialects are null subject languages. The ‘free inversion’ issue is more intricate. Safir points out that Portuguese, which is a null subject language, does not permit free inversion, but can only invert indefinite subjects ((6a) and (b) from Safir 1986, 341):
6
Rizzi (1986b) takes a third view, assuming that French subject pronouns cliticize to the verb in PF. This modification is not of crucial importance here; see also Rizzi & Roberts (1989).
73
(6a) (6b)
*Existe Deus. (there) exists God Não existe tal ilha. (there) exists no such island
My own informants accept inverted names in list-readings, (7) being the answer to the question ‘who is coming tonight?’: (7)
Vem a Maria, o João … comes Mary, John …
It is important to point out here that the term free inversion has probably been used inappropriately, as it is dubious whether inversion is totally free in any language. Differences in word order may reflect differences in, for instance, information structure or even aspect. This was shown to be the case in Italian by Calabrese (1992), and Belletti (1988) made it clear that Italian inversion structures are also sensitive to the Definiteness Effect; see also Delfitto & Pinto (1992). Thus, even if (8a) and (b) are roughly synonymous, we should therefore not draw the conclusion that they are identical in all respects, and it might be best to avoid the term ‘free’: (8a)
Gianni viene. John comes
(8b)
Viene Gianni. comes John
The issue here is slightly different though: the null subject languages under discussion have in common the fact that structures like (6)-(8) are possible, that is, a DP may appear in sentence final position while the typical subject position remains phonetically empty. It suffices to say that subject pro is a necessary condition for all of (6)-(8), though not a sufficient one. Some independent differences between Italian and Portuguese (relating, perhaps to Focus or information structure generally), have the effect that (6a) and (8b) are possible in Italian but not in Portuguese. The fundamental difference between a null subject language such as Portuguese and Swedish, for instance, is still obvious; examples like (6a), (6b), (7) and (8b), where the preverbal subject position is not filled, are impossible in Swedish. The issue of free inversion has also been relevant to the discussion on Old Romance. It was argued by Adams (1987) that Old French did not allow the South Romance type of subject inversion (‘free’ inversion) but only the Germanic type of subject inversion (‘verb second’ inversion), and this in spite of the fact that Old French had null subjects. However, it is convincingly shown by Vance (1989, 1995) that inversion structures of the
74
South Romance type are also attested in Old French, and that Old French is therefore not an exception to the generalization captured by the null subject hypothesis.
2
The Data.
2.1
Avere + Past Participle in Medieval Italian
Let me briefly restate the introductory discussion from chapter 1. The past participle in Modern (standard) Italian cannot agree in (9) but must agree in (10): (9a) (9b)
(10a) (10b)
Ho aperto la finestra I have opened[-Agr] the window *Ho aperta la finestra I have opened[+Agr] the window Ho la finestra aperta I have the window open(ed)[+Agr] *Ho la finestra aperto I have the window open(ed)[-Agr]
In modern usage, two different interpretations are associated with (9a) and (10a). For present discussion one crucial observation is needed: the participial predicate in (10a) has an arbitrary agent, whereas the agent of (9a) is the surface subject, that is, (10a) does not crucially presuppose that I myself opened the window. I will from now on call (10a) PAP (Possessive Auxiliary+Participial clause), in order to distinguish it from the compound tense. As we have already mentioned, there are several varieties of Romance that permit agreement in (9b) (cf. the preceding chapter), and among those, we find historic varieties of Italian. I believe that the most interesting and relevant evidence should come from a variety where the participle always agrees, that is, in a variety where only (9b)/(10a) are found, whereas (9a)/(10b) are attested only exceptionally or not at all. As we have seen, this is the case in Medieval Italian before 1350 but not afterwards. From the middle of the 14th century, agreement in (9b) is highly irregular and varies with stylistic level. This state of affairs legitimates the suspicion that the agreeing form of (9b) in Renaissance Italian no longer belongs to the grammatical system and that the observed variation is a matter of literary style (see chapter 1 for some discussion on the possible implications of this). As for 14th century texts, the first important observation is that the direct object of a participial verb could freely appear both to the right and to the left of this participle.
75
(11a)
(11b)
Aux part. O (...) non veggiono, per ciò che hanno chiusi li occhi (...) they cannot see because they have closed[+Agr] there eyes (Conv; I:IV) Aux (…) come che tu abbi even if you have
part. O perduti i tuoi denari (…) lost[+Agr] your money (Dec; II:5)
Along with the rightward agreement cases (11a) and (b), there are constructions of the type in (12a) and (b), where the object occurs in a position to the left of the participle:
(12a)
(12b)
Aux O part sì come afferma chi ha li occhi chiusi ... as he says who has the eyes closed[+Agr] (Conv; II:IV) Aux O part. … co’ denari avresti la persona perduta. with your money you would have your life lost[+Agr] (Dec; II:5)
Judging from word order, we might be led to believe that the sentences (11a) and (b) correspond to a modern compound tense, whereas (12a) and (b) are rather equivalent to the PAP construction in (10a) above. There is reason to believe, however, that this is not the case; rather, the constructions in (11)-(12) are all essentially equivalent from a syntactic point of view. Consider (13a)-(c):
(13a)
(13b)
Aux ... senza avere without having
O alcuna dottrina di medicina any education in medicine (Dec; I:Intro.)
part. avuta giammai ... had[+Agr] ever
Aux O part. …se io ho bene la ‘ntenzione di tutte compresa… if I have well the wish of everyone understood[+Agr] (Dec; I: 4)
76
(13c)
Aux lo ‘nquisitor domandò se egli avesse the inquisitor asked if he had (Dec; I:6)
O la messa the mass
part. udita quella mattina. heard[+Agr] that morning
Despite word order Aux-O-part as in (10a), it is quite clear that (13a)-(c) can hardly be interpreted as anything but compound tenses: 1. The external argument of the participial Vs is obviously not arbitrary for any of them. 2. The participial predicates, avere and udire, cannot be used in a PAP-like construction, but only in compound tenses; compare the grammaticality of I have had sth or I have heard sth with *I have sth had or *I have sth heard (the aspectual properties of the construction will be discussed in more detail in chapter 3). 3. It is furthermore quite clear that the event expressed by the compound Aux-part. is anterior to the tense of narration (at least in (13a) and (13c)). This observation is of course less compelling, as the Tense value of the periphrasis is variable. There are differences within Romance and Germanic with regard to the ‘present relevance’ of the past action; see Squartini & Bertinetto (forthcoming) and references cited there. I conclude that a PAP reading of (13a)-(13c) is excluded. There is now reason to suspect that word order might have been irrelevant for the interpretation of such constructions. Clearly, both Aux-O-part. and Aux-part.-O could have the meaning of a compound tense. The next question is then whether the inverse situation is also attested, that is, if not only the string Aux-O-part. but also Aux-part.-O could correspond to PAP as in (10a). Lucchesi (1962/63) has convincingly shown that this was the case (see Lucchesi 1962/63, 206 and 194 for a list of the texts cited). Consider (14a)-(c):
(14a)
(14b)
Aux part. O menelao abe traforata la coxa. Menelaos had pierced[+Agr] the leg (Lucchesi, Storie de Troja et de Roma; p. 207) Aux part. O como è morto lo corpo de colui cha ha tajà la testa. as is dead the body of him who has cut-off[+Agr] the head (Lucchesi, La Regola dei Servi della Vergine; p. 207)
77
(14c)
Aux part. O era assai lasso e aveva ingrossate assa’ le sue braccia. he was quite tired and had swollen[+Agr] much his arms (Lucchesi, Tavola Ritonda; p. 207)
Examples (14a)-(c) have the word order of a compound tense, but cannot be interpreted as such. None of the surface subjects in (14) are the agent of the participial predicate; they are rather the ‘possessors’ of the DPs. Menelaos in (a) did not pierce his leg, but had (got) his leg pierced; (b) does not speak of someone who cut someone else’s head off, but of he who has (got) his own head cut off (by someone else); likewise in (c), the subject has (got) his arms swollen, and no other reading is available. These observations are important, as they show that the two different word orders, Aux-part.-O and Aux-O-part., did not correspond to two different readings, as opposed to the situation in Modern Italian; compare (9a) and (10a) above, where the former can only correspond to a compound tense and the latter only to PAP. Rather, what we are dealing with is a sort of ‘object inversion’ property; the deep object of the participial predicate can appear both to the right and to the left of that predicate. Both Aux-part.-O and Aux-O-part. could have the meaning of either PAP or a compound tense. Given the discussion of paragraph 1.2., we cannot exclude that some subtle difference in interpretation actually did distinguish between the constructions of (12) and (13), at the level of information structure, for instance. The crucial point is that they were not linked to different interpretations comparable to the difference in contemporary language. The nature of this inversion phenomenon still needs to be properly defined; we will turn to the analysis in sections 3 and 4 below. The conclusion that we derive from this is interesting for the Spec-head agreement hypothesis: rightward participial agreement with a direct object in Italian was regular in a period where this direct object could appear freely either to the right or to the left of its participial predicate. That is, agreement in a phrase like (15a) was regular in Medieval Italian as long as (15b) was also grammatical and as long as the two constructions did not correspond to two distinct readings, at least in the modern sense. (15a) (15b)
Ho perduti i soldi. I have lost[+Agr] the money Ho i soldi perduti. I have the money lost[+Agr]
Having established this, I will now go on to argue that this state of affairs is not a peculiarity limited to an ancient grammar during the course of some parametric resetting, but is indeed attested in contemporary grammars, which appear to be stable, synchronically working
78
systems. I will exemplify with Modern Spanish and Modern Swedish. We will go through the relevant data briefly in 2.2 and 2.3, before turning to the analysis in 3 and 4.
2.2
Tener + Past Participle in Iberoromance
Whereas in Italian, French, English and Scandinavian among other languages, the verb have is used to form both the compound tense and PAP, in Spanish it occurs only in the compound tense, having been replaced by tener in its possessive reading. (16a) (16b)
He escrito las cartas. I have written[-Agr] the letters Tengo las cartas escritas. I have the letters written[+Agr]
In Portuguese, ter has taken the place of have in (16a) as well (cf. Parkinson 1988). Examples from Spanish are sufficient for present discussion. The syntactic patterns of these constructions are nevertheless the same: agreement is impossible in the compound tense (16a) and obligatory in PAP (16b). An important difference is however that the object in (16b) can also appear in final position, where agreement is still obligatory: (17)
Tengo escritas las cartas (=(16b))
As there is no significant difference in reading between (17) and (16b), the Spanish pattern in tener+past participle constructions is strikingly similar to avere+p.p. attested in Old Italian.7
2.3
Få + Past Participle in Swedish
In Swedish, as in English, the verb hava ‘have’ has an aspectual variant in få. The use of få is quite complex and resembles English get in its transitive uses, in spite of some differences. 7
The similarity is strengthened by the fact that tener in (certain varieties of) Spanish seems to be in an auxiliarization process and sometimes takes on the value of past tense. This becomes clear in the study of Harre (1991), who reports judgments of informants that accept both (i) and (ii), for instance: (i) %Ya te tengo dicho que no hagas eso. I have already told you that you shouldn’t do this (ii) %Tiene conocidas muchas chicas durante sus viajes. He has known[+Agr] many girls during his travels. I choose to mark these examples %. For my own informants (European and South American) the judgment is clearly * for (ii), but most tend to accept (i).
79
On this topic, see Ekbo (1943) for a traditional treatment, and in recent frameworks, Platzack (1989), Vikner & Sprouse (1988), Christensen & Taraldsen (1989) for Norwegian data, and Hedlund (1992).8 Through a process of auxiliarization, clearly similar to that of have in Germanic and Romance, få has been brought from its original meaning of receiving in (18) (still present in modern language) to the use illustrated in (19), which is a causative construction. (18) (18a) (18b)
(19)
Full, semantic use: Jag fick brevet. I got the letter = ‘I received the letter’ Jag fick brevet skrivet.9 I got the letter written[+Agr] = ‘I received the letter that was written’ Auxiliarized use I: causative Jag fick brevet skrivet. I got the letter written[+Agr] = ‘I made someone write the letter’
Importantly, the DP argument in (19) can freely appear both to the right and to the left of the participle. Word order is irrelevant for the interpretation of the construction; (19) and (20) are thus synonymous. Agreement is obligatory in both cases: (19) (20)
Jag fick brevet skrivet = I got the letter written[+Agr] Jag fick skrivet brevet = I got written[+Agr] the letter (=19)
The process of auxiliarization has finally brought få to (21): (21)
Auxiliarized use II: Aspect mark Jag fick skrivit brevet I got written[-Agr] the letter
8
My own intuitions for Swedish do not always correspond to those reported by Vikner & Sprouse (1988) for Danish, or those of Christensen & Taraldsen (1989) for Norwegian. Differences in the participial inflectional paradigm between Scandinavian varieties may be the cause of this. There even seems to be some variation within Swedish (cf. Hedlund 1992). 9 The inflected participle in Swedish ends in -et (singular neuter), -en (singular uter), or -na (plural either gender). Skrivit is the uninflected form, which the grammatical tradition calls the supinum. I will refer to the uninflected form as the active participle. See Källström (1993) for a descriptive study on agreement in Swedish. We will return to this issue in chapters 3 and 4. Notice that the participle is clearly attributive in (18b), where få has its full, semantic reading, but predicative in (19), where få is auxiliarized. Postposed attributive participle phrases normally need to be heavier than in (18b), which is therefore stylistically somewhat awkward.
80
The meaning of (21) is approximately I wrote the letter;10 the agent is no longer arbitrary, and skrivit is an active participle. (21) has the same syntactic patterns as the compound tense formed with have: the object is obligatorily to the right of the participle which cannot agree. In summary, we have compared auxiliary+participle constructions in Old Italian, Modern Spanish and Modern Swedish. There is a parallel pattern among these, which can be summarized as follows: When supposed rightward agreement is found, as when a participle agrees with a DP object to the right as in (15a), (17) and (20), this object is also attested in a position to the left of the participle, witness (15b), (16b) and (19). In all of these examples, variations in word order do not correspond to a difference in interpretation; (16b) is synonymous to (17), (19) to (20), and as far as we can tell, no significant difference in interpretation can have been linked to (15a) and (15b). There is indeed a correlation between word order, agreement, and interpretation that so far has been stated in terms of linear ordering. ‘Rightward’ and ‘leftward’ positions now require a proper, structural definition.
3
Analysis I: Extraposition or Free Branching?
This section considers some possible analyses that are closely connected to what has been independently assumed in the literature. These analyses will be rejected in this section and I assume a different solution in section 4. 1. One possibility that comes to mind is some version of a free-branching analysis. This could be carried out in either of two ways: 1.a Building on Old Italian, we may assume that the parametric change from Latin word order SOV to Modern Romance SVO is the origin of the observed variation. We then entertain that the complement of V in Old Italian, at an intermediate stage of diachronic development, branched freely either to the left or to the right of V.11 (22)
(a) (b)
[V’ V Compl] [V’ Compl V]
1.b Instead of the complement, what branches freely might be a specifier. It has been suggested by Contreras (1987) that the specifier of Spanish small clauses has this property. 10
I prefer to define få as an aspectual marker in this case. See Hedlund (1992) for a discussion on the semantic implications of the construction. 11 In a different framework, Antinucci (1977) hypothesized that the auxiliarization process was linked to, or even driven by, the shift in word order.
81
(23)
(a) [SC Spec pred] (b) [SC pred Spec] (cf. Contreras 1987, Koopman & Sportiche 1991, among others)
2 Apart from the free-branching alternatives, there is a third solution: the righthand object could be extraposed or right-dislocated in a position adjoined to the XP, where XP=VP/IP or perhaps AP:
XP
(24)
DP(O)
XP
Spec
X’
The structures (22), (23) and (24) are excluded under the more restrictive approaches to X’theory that I wish to follow.12 Furthermore, all three structures can be rejected on empirical grounds. We will see this in 3.1.-3.4., and then turn to a third proposal in 4. 3.1 The first objection is of course that parallel data are preferably to be captured in a unified hypothesis. The SOV/SVO analysis may be viable for Old Italian, but has less to say for Modern Spanish. The free-branching-Spec hypothesis may be valid for Spanish, but fails to account for Scandinavian data. Uniform data from the three languages calls for a different kind of approach. As far as I know, there is no independent evidence for a free-branching analysis in VP in Swedish. 3.2 The extraposition or dislocation hypothesis proves untenable for both Swedish and Spanish. In Spanish, such a construction would require a resumptive pronoun as in Romance generally. (25)
Las tengo escritas, las cartas.
Right dislocation, or extraposition, is quite ungrammatical in Swedish as well if the DP is not repeated in the form of a weak pronoun. Moreover, the extraposed DP generally needs to be heavier than in (26)/(27): 12
A further possibility that will be awkward in the Antisymmetry format is the adjoined-predicate approach advanced for Romance by Merlo (1989). I will disregard this suggestion in this and the following chapter and return to it, briefly, in chapter 8.
82
(26) (27)
*Jag fick skrivet, breven. Jag fick dem skrivna, breven.
3.3 Both of the free-branching alternatives predict that the ordering of the elements is irrelevant for agreement. According to (23), the participle would show agreement in both the (a) and the (b) cases, if we assume Spec-head agreement (henceforth SHA). (22) predicts the same if we were to assume head-complement agreement (henceforth HCA). The combination of word order in the participial complement and agreement or non-agreement yields four logical possibilities: A. Object to the right of a non-agreeing participle. ho aperto la finestra I have opened[-Agr] the window. B. Object to the left of an agreeing participle: ho la finestra aperta I have the window opened[+Agr] C. Object to the right of an agreeing participle: ho aperta la finestra I have opened[+Agr] the window D. Object to the left of a non-agreeing participle: ho la finestra aperto I have the window opened[-Agr] If we instead assume only SHA, and exclude HCA, and if the specifier is always in the lefthand position, alternative (D) is ruled out on principled grounds. If the DP object fills the specifier position, agreement should be obligatory on the participle.13 As we have seen above, modern standard Italian lacks both (C) and (D). More interestingly, the Swedish få + past participle construction can appear in the shape of (A), (B) and (C) as we have seen in section 2, but does not admit (D): (28) (29) (30) (31)
Jag fick skrivet brevet (=C) Jag fick brevet skrivet (=B) Jag fick skrivit brevet (=A) *Jag fick brevet skrivit (=D)
13
Only those grammars are relevant where the predicative participle has the property of showing agreement. English and German data cannot be considered here.
83
Moreover, recall that tener + past participle shows tendencies toward auxiliarization (see footnote 7 above). Harre (1991, 67) reports that some of her informants were inclined to accept both (32) and (33): (32) (33)
%Tengo vistos (a) los hombres. (=C) I have seen[+Agr] the men %Tengo (a) los hombres vistos. (=B) I have the men seen[+Agr]
Some did accept a non-agreeing participle as well, but only when the object appeared in final position. Thus, the same informants that accepted (34) refused (35) (but accepted (33) above): (34) (35)
%Tengo visto a los hombres. I have seen[-Agr] the men *Tengo a los hombres visto. I have the men seen[-Agr]
(=A) (=D)
This is totally unexpected under a free-branching hypothesis of either kind.14 Note also that ter + participle in Portuguese show the same pattern: (36a) (36b) (36c) (36d)
Tenho a janela aberta I have the window opened[+Agr] Tenho aberta a janela. = (36a) I have opened[+Agr] the window Tenho aberto a janela. ≠ (36a)/(36b) I have opened[-Agr] the window *Tenho a janela aberto. I have the window opened[-Agr]
(= B) (= C) (= A) (= D)
In Old Italian texts from the 13th and the 14th centuries, both (C) and (B) are frequently attested. From the late 14th century onwards, (A) becomes widespread. Apart from two exceptions that were discussed in chapter 1 (see chapter 1, sections 3.2a and 3.2b), alternative (D) appears to be absent. Even though the fact that a form is not attested cannot in principle be taken as evidence that it was ungrammatical, its absence, in texts that otherwise are rich of inversions of many kinds, is clearly significant. The data in these languages are explained on the assumption that the complement is always to the right of the head and that UG only permits SHA. 14
Note, coincidentally, that the small clauses discussed by Contreras (1987) do not have participial predicates. His conclusions are based on small clauses with non-agreeing categories as predicates.
84
3.4 This is furthermore supported by the following data: recall that Spanish tener can take a clausal complement (cf. footnote 7, this chapter): (37)
%Ya te tengo dicho [CP que no hagas eso]. I have already said that you should not do this
This is also possible with the få-construction in Swedish: (38)
Han fick skrivet att beslutet var olagligt (i tidningen) He got written that the decision was illegal (in the newspaper)
The word order indicated in the above examples is the only possible. The CP can never occur in pre-participial position: (39)
*Ya te tengo [CP que no hagas eso] dicho. I already have [that you shouldn’t do this] said
(40)
*Han fick [CP att beslutet var olagligt] skrivet. He got [that the decision was illegal] written
Constructions of the kind in (39)-(40) are not attested in Old Italian to my knowledge; see also Zwart (1993, 297) for the same kind of argumentation with reference to Dutch. I therefore conclude that any kind of free-branching analysis, (22) or (23), is inadequate. We will outline and adopt a different proposal in 4.
4
Analysis II: Object Pro
The first and crucial assumption I will make, is that the participial predicate in the constructions under examination takes its DP argument as a complement, quite a natural assumption since we are dealing with transitive predicates. I suggest therefore that the rightward position is always a complement, and that the DP(O) here is in its base position. The leftward position is a specifier corresponding to an Agr head, to which the participle raises. We thereby account for the grammaticality of (A) - the participle cannot agree with its complement, the grammaticality of (B) - the participle in Agr must agree with the specifier of Agr, and the ungrammaticality of (D) - see 3.3 above. It follows that Case theoretic motivations are responsible for the differences in word order: the object can either be shifted in overt syntax yielding (B), or in covert syntax yielding (A). We are left with two problems: 1. Why is (C) at all possible? 2. Why is it that, when (C) is attested, we also find (B) and they are synonymous?
85
Having dispensed with the free branching alternatives, we need to find the solution elsewhere. Following a suggestion from Rizzi (1986a), I assume the structures in question involve an expletive pronominal that moves in overt syntax allowing the object to stay in situ.15 This amounts to saying that the classic analysis of nominative expletive chains in the finite clause carries over to the participial clause and in this I diverge from Chomsky (1995) in the sense that I assume that Spec AgrO can host an expletive element.16 However, Chomsky does not exclude the idea of an expletive accusative chain on principled grounds. Rather, on minimalist assumptions, an accusative expletive is a natural conclusion: if the DP(O) can raise overtly, it is because it must do so. If, then, DP(O) has the possibility of remaining in situ, whatever feature requires overt checking must still be checked in some other way. The conclusion that a pro takes the place of DP(O) in overt syntax, and that the relevant feature is checked on the chain follows naturally. What we have here is then another instantiation of an expletive dependency, involving, however, the object. The overt expletive in Swedish is det ‘it’, which is the same as the definite article in neuter singular. Consider again (19)/(20), repeated here as (41)/(42). (41)
Jag fick brevet skrivet I got the letter[neut. sing.] written[neut. sing.]
(42)
Jag fick skrivet brevet I got written[neut. sing.] the letter[neut. sing.]
The noun letter is singular neuter. If we replace it with a plural noun, things change. When the noun appears to the left of the participle, the participle agrees in number; when the noun appears to the right, the participle continues to show neuter singular agreement: (43)
Jag fick breven skrivna. I got the letters[plural] written[plural]
(44)
Jag fick skrivet breven. I got written[neut. sing.] the letters[plural]
15
Rizzi (1986a) claimed that expletive pro was responsible for DP inversion in adjectival small clauses, and my proposal is above all inspired by this idea. What I suggest here is however different from Rizzi’s (1986a) original analysis of expletive object pro. The question arises how to distinguish the Old Italian system from the modern one. We will turn to that problem in chapter 8. 16 Chomsky (1995, 341) claims that the specifier of AgrS can host both an expletive and a raised noun, whereas the specifier of AgrO only hosts raised nouns. My account makes AgrO and AgrS perfectly similar in this respect.
86
It is important to remember that neuter singular is an agreeing, inflected form of the participle; the uninflected form would be skrivit, which is ungrammatical in (44) in the relevant reading (compare (21) above)17 and in (43) under any reading. We therefore conclude that the participle in (44) is neither a non-agreeing form, nor does it agree with the DP on its right; these are evident conclusions. I suggest that Swedish has a null pronominal, carrying the φ-feature specification of det, which raises to the Spec of the participial phrase (=Spec AgrO) in (44) triggering agreement on the participle in singular neuter. The hypothesis I adopt for Swedish carries over to Old Italian and Spanish and is the following: the agreement features on the participles in the relevant constructions, and the accusative Case features on the corresponding nouns, count as strong in the terminology of Chomsky (1993) and, therefore, require overt checking. When the DP appears to the left of the participle, it has been overtly raised to what I take to be the specifier of AgrO. In addition, these grammars have the option of licensing a null object that may raise in overt syntax allowing the noun to remain in situ. A chain is formed in either of two ways, DP(O)i … ti or proi … DP(O)i. The structures I propose are those indicated in (45)/(46):
(45)
AuxP
Aux
AgrOP
AgrO’
Spec DP(O)j
VP
AgrO part. i V ti
tj
17
Sigurðsson’s (1992b) conclusion that the Swedish participle has two non-agreeing forms therefore needs some modification. Clearly, both skrivit and skrivet are non-agreeing in the sense that they do not show agreement with the DP in an example like (43) but only skrivit is an uninflected form. Skrivet in (44) is indeed an agreeing form although it does not agree with the DP to the right. Note that the pattern conforms to Greenberg’s Universal 33; cf. footnote 5 above.
87
(46)
AuxP
Aux
AgrOP
AgrO’
Spec pro j AgrO part. i
VP
DP(O) j
Vi
The (a)-examples in (47)-(49) correspond to the structure given in (45) and the (b)-examples to (46). (47a)
Tengo las cartasi escritas ti I have the lettersi written[pl.f.]ti
(47b)
Tengo proi escritas las cartasi I have proi written[pl.f.]the lettersi ‘
(48a)
Jag fick breveti skrivet ti. I got the letteri written[sing. neuter] ti
(48b)
Jag fick proi skrivet breveti I got proi written[sing. neuter]the letteri
(49a)
Aveva gli occhii chiusi ti he had the eyesi closed[pl.] ti
(49b)
Aveva proi chiusi gli occhii he had proi closed[pl.] the eyesi
Assuming that free inversion is actually object shift triggered by the necessity to check Case, we explain the ungrammaticality of a CP in leftward position at the same time. Recall the contrast between (37)/(38) and (39)/(40) above, repeated here as (50)/(51) and (52)/(53):
88
(50) (51)
%Ya te tengo dicho [CP que no hagas eso]. Han fick skrivet [CP att beslutet var olagligt] (i tidningen)
(52) (53)
*Ya te tengo [CP que no hagas eso] dicho. *Han fick [CP att beslutet var olagligt] skrivet.
If we assume that CP does not carry Case and therefore does not move to check any, the contrast follows nicely in this account. I conclude that the expletive-associate solution is descriptively superior to its alternatives. In the following section, I will show that it can explain more complex word order patterns and, at the same time, some of the differences and similarities between Old Italian and Germanic.
5
Word Order Variations in Old Italian, and Remarks on the Comparison with Modern Germanic
5.1
Construction types 1, 2, 5, and 6
As has become obvious by now, word order patterns of Medieval and Renaissance Italian (MI and RI) are to some extent similar to those of certain modern Germanic languages such as German and Dutch. It is of course not my aim to explain German or Dutch data, but it must be shown that my account of Old Italian in principle can handle the comparison with modern Germanic. The crucial difference between my 14th century Italian material and Modern German and Dutch is that MI shows a greater freedom in word order. I will suggest here that the MI system shares with Modern German or Dutch the fundamental property of having overt object movement to Spec AgrO (following Zwart 1993) but differs in that MI Spec AgrO can host an expletive element at Spell-Out. On the first assumption, we will derive the similarities between the grammars, on the latter the differences. We will go through the various possibilities in turn, in 5.1.1.-5.1.5. 5.1.1 We have seen that the object in Medieval Italian Aux-participle constructions could occupy a position to the left of the participle (Aux-object-participle), and this property is regular in German:
(54)
Aux O part. … co’ denari avresti la persona perduta. with your money you would have your life lost[+Agr] (Dec; II:5)
89
(55)
Ich I
Aux O habe das Buch have the book
part. gelesen. read (= ‘I have read the book’)
As in German and Dutch, according to Zwart (1993) and Den Dikken (1995), I take this to be an instance of A-movement to Spec AgrO. 5.1.2 Old Italian is still different from the German system in that objects also could appear to the right of the participle (Aux-participle-object):
(56)
Aux (…) come che tu abbi even if you have
(57)
Aux part. *Ich habe gelesen I have read
part. O perduti i tuoi denari (…) lost[+Agr] your money (Dec; II:5)
O das Buch. the book
Another striking circumstance is that there does not seem to be any difference in the 14th century between main and subordinate clauses with respect to (54)/(56), or at least, no such difference emerges from my material. Other word order combinations have been found but we will now have a problem in defining them properly: in the literature on Germanic and Romance syntax, distinctions are made between Topicalization, Scrambling, Dislocation, and Focus (see among many others Cinque 1990a and Den Dikken 1995 for some recent references). In Old Italian texts, we observe a variety of preposing structures, and it is quite difficult to determine whether these are scrambling, topicalization, or dislocation. The problem has been extensively discussed in recent years; cf. among others, Marcantonio (1976), Vanelli (1986), D’Achille (1990a, 1990b), and for recent discussion in a generative framework, Benincà (1994). The exact analysis depends to some extent on issues that are not part of my research agenda, and would bring the discussion much too far away from the main problem which is participial syntax. In particular, a proper understanding of the preposing mechanisms will not be reached without further analysis of the verb second property of Old Italian; see above all Benincà (1994) for this aspect. I take (54) and (56) to be object shift, thus A-movement to Spec AgrO (in (54) the DP moves, in (56) pro moves and the DP associate stays in situ). All of the following constructions will be referred to as ‘preposing’, which appears to be the most neutral term. Consequently, I will not make a precise claim about their landing site, which will be labeled
90
XP throughout, as the exact nature of the various types of preposing remains an open issue. In some cases, XP will be a functional projection higher than IP and lower than Comp, and sometimes it appears in sentence initial position. Spec X may thus correspond to any of Spec Topic, Spec Focus or Spec C. As before, I will disregard the base position of the subject, and I will indicate the base position of the auxiliary as an Aux, head of its own projection, assuming raising of Aux to Infl. If, instead, AuxP = IP, nothing crucial changes. 5.1.3 We have observed the word order object-participle-Aux in MI, which is the rule in German and Dutch embedded clauses:
(58)
(59)
(60)
O part. …quel dí stesso che la buona confessione fatta that same day that the good confession made (Dec; I:1)
Aux avea… (he-)had
O part Aux …come la corte di Roma veduta avesse… as the court of Rome seen (he-)had (Dec; I:2)
... daß du ... that you
O part. Aux das Buch gelesen hast. the book read have
However, as we have already said, auxiliary-participle-object is also present in MI. Again, we are comparing an optional phenomenon in MI with an obligatory one in German and Dutch. Consider (61)/(62):
(61)
(62)
Aux part. O …a colei piacerà che quel giorno avrà avuta la signoria… to her will it please who that day will have had the command (Dec; Intro.)
*... daß du ... that you
Aux hast have
part. O gelesen das Buch. read the book
Both similarities and differences follow on this account. First of all, I assume that Zwart’s (1993) account is correct for German and Dutch. This amounts to saying that Old Italian,
91
German and Dutch are similar in having basic VO word order (under Antisymmetry, UG would not admit parametrization on this point) and strong accusative Case, that is, overt checking of accusative. (54) and (55) are therefore equally grammatical in the languages under discussion. If MI has an expletive object pronominal that raises in overt syntax thus permitting the phonetic object to stay in situ, and if a similar strategy is not available in German or Dutch, the contrast between (56) and (57) follows in accordance with my claim. As for the remaining Old Italian data, they can be shown to follow from the assumptions already made for (54) and (56). I will now assume the most simple structure for the participial phrase which is (63). A more thorough account will have to take other projections in consideration. We will discuss some implications in the following sections and chapters: (63)
Aux [AgrOP AgrO [VP V]]
First, we need to decide whether the participial V stays in situ or moves overtly to AgrO. Consider that it was possible to insert adverbial material between the participle and the object, and post-participial stranded quantifiers are attested as in (64):
(64a)
(64b)
Aux part. Adv. O … avendo veduta a una festa una bellissima donna … having seen at a feast a very beautiful woman (Dec; I: 10) Aux part. Quant Voi avete ditto tutti you have said all (Nov.re; I)
O la verità. the truth
I conclude that the participle moved to AgrO in Medieval Italian as in Modern Italian (cf. Belletti 1990, Cinque 1994). Consider example (59), la corte di Roma veduta avesse ‘the court or Rome seen had (he)’. Movement of the object to Spec AgrO in overt syntax and the subsequent preposing of the entire participial phrase to XP above the finite Aux yields the word order objectparticiple-Aux.
92
(65)
XP
Spec AgrOPj
Spec DP(O)i la corte di Roma
X’
AgrO vedutal
InflP
X
AgrO’ VP V tl
Infl avessek
ti
AuxP
Aux tk
tj
(65) is the structure of (59), with the subject omitted. The process is clearly optional. If the object stays in situ and no preposing takes place, word order will be as in Modern Italian. Take (61), avrà avuta la signoria ‘(will) have had the seigniory’, and see (46) above. (66)
[AuxP avrà [AgrOP proj [AgrO’ avutai [VP ti la signoriaj]]]]
5.1.4 (67) instantiates object-Aux-participle and would be a case of preposing of the object alone, while other constituents stay in situ. This corresponds to construction type 6, see section 3.6 in chapter 1:18
(67)
O Tre battaglie di campo three battles
Aux part. ho poi fatte… have (I) then done… (Nov.no; XXI)
18
This particular case is discussed in detail by Benincà (1994) who argues that it targets the specifier of Comp.
93
(68)
XP
X’
DP(O) i tre battaglie X
InflP
Infl hoj
Aux P
Aux tj
AgrOP
Spec ti
AgrO’
AgrO fatte
VP
All of the examples we have seen so far can thus be analyzed as preposing of either the object or the entire participial phrase, where we assume the participial phrase = [[AgrOP [VP]]. There is of course reason to suspect a richer structure for the ancient participle as for the modern one. Such a move is necessary for (69), where an adverbial PP has been inserted between the object and the participle:
(69)
Aux. O PP (Adv) Poscia che Dio ebbe Adamo ed Eva, per lo peccato ch’aveano After God had Adam and Eve, for the sin they had part. fatto, tratti di paradiso … committed, taken[+Agr] from Paradise… (Vizî; VI)
94
(70)
AgrOP AgrO’
Spec DP(O)i
YP
AgrO
Y’
Spec PP
VP
Y Vj tj
ti
One way of resolving this is to assume that the Agr node is freely recursive, hence YP = AgrOP. It would then be possible to insert material into the lower recursion of AgrO while the object moves higher up in the structure but remains within the participial phrase, that is, within some recursion of AgrOP. 5.1.5 Consider now that there is a fifth possible combination, participle-auxiliaryobject, corresponding to construction type 5 in the corpus survey:
(71)
part. Perduta lost
Aux. O ho la fatica… (I-)have the fatigue (Dec; I: 2)
This kind of preposing is quite uncommon in the texts of both centuries; cf. section 3.5, chapter 1. In (71) the participle occupies the sentence initial position on its own, apparently stranding the object to the right of the auxiliary. For (71) there are two solutions that immediately come to mind; it might be a case of Long Head Movement or an instantiation of what is usually referred to as Stylistic Fronting. However, I will argue in 5.2. that (71) can neither be Long Head Movement nor Stylistic Fronting. I will first show that (71) follows from assumptions already made. (71) can be derived on the same basis as (69): if there can be more than one AgrOP in the structure, it would be possible for the object to reach the specifier of the higher AgrO, while the lower AgrOP is preposed just as before:
95
(72)
[AuxP Aux [AgrOP AgrO2 [AgrOP AgrO1 [VP V]]]]
Suppose the object reaches Spec AgrO2. Preposing of the lowest recursion of AgrOP, indicated as 1, yields the construction in (71), giving the impression that the object has been stranded to the right of the auxiliary. XP
(73)
X’
AgrOP1i AgrO1 perduta
VP
X
InflP Infl hoj
AuxP Aux tj
AgrOP2
Spec DP(O) la fatica
AgrO’2 AgrO2
ti
All word order variations of MI are in this way derived on three assumptions: (a) overt object shift, which is obligatory; (b) the licensing of an expletive object pronominal, which is optional; (c) preposing of either of DP(O) and AgrOP, which is optional. Note finally, that the derivation of (73) does not depend crucially on the licensing of object pro. Furthermore, (73), or some structure similar to (73), may also be adequate if the object DP moves out from VP for reasons independent of Case (thus, presumably not to AgrO but higher up). This is the desired result: we have seen that overt object shift and participle agreement are regular in Old Italian until the middle of the 14th century. The preposing of participles is attested throughout the 16th century, and it is therefore clear that the derivation yielding (71) is in principle also available after the loss of object pro and overt object shift.
96
5.2
XP-Preposing, Long Head Movement, or Stylistic Fronting?
At the same time, construction type 5 (71) is reminiscent of Long Head Movement (cf. Lema & Rivero 1990, Rivero 1991, and Roberts 1993) on the one hand, and of Stylistic Fronting (cf. Maling 1980 and Holmberg & Platzack 1995) on the other. It can be shown, however, that the Medieval Italian construction of (71) does not correspond to either of these phenomena. Lema & Rivero (1990) maintain that LHM is attested in Balkan languages as well as in historical and archaic Romance varieties including early Italian. They insist however that the phenomenon in question is limited to root contexts. (71) is a root context, but consider (74) and (75) (the latter from Rohlfs 1969, § 985) where the participles have been preposed after a complementizer ((74) is a passive construction and (75) is a compound tense):
(74)
(75)
Comp Poi che since
… ma but
part. cop. purgato è cleaned is (Conv; I: V) Comp poi che after that (Dec; III:8)
DP(S) questo pane da le macule … this bread from the spots…
part. mangiato eaten
Aux ebbe … had
The stranded object is often a clitic as in (76)/(77).
(76)
(77)
…ma che un qualche dio con le proprie mani but that some God with his own hands (Cort; XIV)
part. formati formed
cl. Aux. gli abbia… them[cl.]-has
part. cl. Aux. …che la natura per dispetto o per ludibrio produtti gli abbia … that nature by mistake or scorn produced them[cl.]-has … (Cort; XIV)
Among RI authors, preposing of the participle with stranding of the DP object is found above all in Lasca (B.6) and Bembo (B.9):
97
(78)
Comp … io ho udito dire più volte che gl’italiani uomini I have heard saying many times that the Italians part. Aux. O apparata hanno questa arte, … learned have this art (Prose; I: VII)
It is evident that the phenomenon applies freely to embedded contexts; therefore, there must be some difference between the Italian varieties under discussion and the grammars treated by Lema & Rivero (1990). The data presented in (74)-(78) are sufficient to exclude the second possibility, Stylistic Fronting, which is known to exist in Sardinian among Romance languages. Fronting in Sardinian, however, obeys the same restriction that holds in Icelandic, namely that the canonical subject position must be left free (cf. Jones 1988, 339, for Sardinian and Holmberg & Platzack 1995, for Icelandic; see also references cited there). The subject is inverted in (74) and omitted in (75), but it is realized in (76)-(78) as well as in (79a) and (b): (79a)
… né io sofferto avria pena ingiustamente. nor I suffered would-have punishment unjustfully (Conv; I:III)
(79b)
…era più grande che nessuna che veduto io ne avessi… it was bigger than anyone that seen I thereof[cl.]-had (Vita; XXVII)
The lexical subject is thus attested both in pre- and in post-participial position. In all of (76)(79) however, it precedes the finite auxiliary and therefore cannot be in an inverted position. I conclude that an analysis in terms of LHM or Stylistic Fronting is not adequate. The data are best analyzed as preposing of XP. It remains somewhat unclear to me what the landing site of the preposed element is in (74)-(79). Of course, whether or not the mechanisms that I propose for Italian are available in modern languages, depends not only on whether they have overt object shift or object pro, but also on their topicalization or scrambling properties in general.
5.3
An Additional Note on Construction Type 5
Among the cases of participle preposing, I have also counted the following ones, where the single participle precedes a complementizer (cf. Rohlfs 1969, § 990):
98
(80a)
(80b)
Udito che ebbono tutti questa sentenzia… heard that had everyone this sentence (Trec: VII) Preso che ebbe el duca la Romagna… taken that had the duke Romagna (Princ: VII)
For this particular case, I have little to suggest. Note first that (80a) and (b) could follow on the same assumptions as (71) and (74)-(79), that is, as a case of XP preposing. There are doubts on the adequacy of this analysis, however. There is a clear difference between this kind of preposing and the preposing of a participle without the following complementizer, namely the fact that, when there is a complementizer, the participle is never found together with a DP object. When there is no following complementizer, DP object+participle are preposed quite frequently in the earliest part of my material (the first part of the A-corpus). Note that in Modern Swedish, where a similar construction is possible, the opposite situation is attested, since participle+object cannot be separated by preposing: (81a) (81b)
(82a) (82b)
Köpa bil som han skulle … buy a car as he was going to … *Köpa som han skulle bil … buy as he was going to a car … Trodde på spöken som han gjorde … believed in ghosts as he did … *Trodde som han gjorde på spöken … believed as he did in ghosts …
As opposed to the former case, this construction might perhaps qualify as a sort of Long Head Movement in Old Italian, whereas it is definitely XP-preposing in Swedish. There are some unclear points to sort out in order to pursue such an analysis, and I have no interesting suggestions to make.
5.4
Construction Type 7
Construction type 7 refers to the case in which participles agree with a DP object which is the direct argument of an infinitive V embedded under the participle.
99
(83a)
(83b)
… avendo già fatti i familiari … davanti chiamarsi having already made[pl.] the servants in front of her call = ‘she made (somebody) call the servants’ (Dec; I:Intro) Voi non avrete compiuta ciascuno di dire una sua novelletta … You have not finished[fem.] each one to tell a novel (Dec; I: Intro)
Kayne (1989a) proposes to account for this case as well as construction type 1, Aux+participle+object, as instantiations of «right dislocation with a phonetically unrealized object clitic» Kayne (1989a, 96).19 Considering the Old Italian material, there is one obvious difference between Aux+participle+object, construction type 1, and construction type 7, namely that agreement was never as regular in the latter as it was in the former. Kayne’s analysis suggests that the agreement phenomenon was optional, and this appears to be the correct generalization for (83a) and (b), though not for construction type 1. I will follow Kayne (1989a) on this point, assuming that agreement on the participle in (83a)-(b) is triggered by a null clitic pronominal. 6
Conclusion
Summing up, our discussion leads us to four conclusions: 1. The correlation between agreement phenomena and word order (hence, DP movement) is solid and can be taken as an argument that the Agr category should not be excluded from the theory of Grammar. 2. Among the cases of supposed rightward agreement attested, we have yet not seen any compelling argument in favor of head complement agreement. The data can readily be explained in terms of SHA, as we suspected. 3. Object inversion phenomena in three different grammars have given us reason to believe that there is no substantial difference between nominative and accusative dependencies; both can be built up by an expletive and an associate. I conclude that Spec AgrO may host expletive elements in the same way as AgrS. This idea will be further developed and justified in following chapters. 4. Furthermore, all data follow under a strict version of Antisymmetry, as only two positions, complement and specifier, need to be assumed in each projection and there is no parametrization of branching. 19
According to Kayne (1989a), the grammaticality of (83a)-(83b) is actually due to two factors: (a) a null pronominal object, and (b) the possibility of reconstruction (cf. Rizzi 1982, Kayne 1989b). Reconstruction must have been a general property of Old Italian, and what distinguishes Old from Modern Italian on this point is therefore the null pronominal clitic. This is in line with my demonstration.
100
Appendix 1 to Chapter 2: Inflectional Structure in the Swedish Participle Phrase It is highly plausible that the Swedish participle (whether active - the so called supinum - or passive) hosts inflectional material. We assumed this in section 4 of this chapter, and it has been argued by Hedlund (1992) that the nonfinite verb regularly moves into a functional projection. This can be shown by the following data: in Swedish, impersonal constructions are grammatical with both unergative and ergative predicates (cf. Falk 1989): (1)
Ergative: Det kom tre män på vägen. there came three men on the road
(2)
Unergative: Det dansade tre män på vägen. there danced three men on the road
Under standard assumptions about argument structure, the difference between ergative and unergative predicates is that the former project their single argument as a complement whereas the latter take an external argument (Burzio 1986; I disregard the question of whether unergatives subcategorize for a null object as claimed by Hale & Keyser 1993, Kayne 1993, and Chomsky 1995):
(3)
(4)
VP
Erga tive:
VP
Spec DP
V’
V
Une rgative :
Comp l DP
V’
V
This difference would never appear visibly in the Swedish main clause with a finite verb. Although the subject DP in the impersonal remains in its basic position, the verb moves higher so as to appear to the left of the subject. In the compound tense, however, it has been widely held that the nonfinite V stays in situ (e.g. Holmberg 1986, Falk 1993, and Holmberg & Platzack 1995; see also Vikner 1990, Jonas & Bobaljik 1993 and Chomsky 1995 for related discussion). If we put (1) and (2) in the compound tense, we obtain the following contrast:
101
(5) (6) (7) (8)
Det har kommit tre män på vägen. there has come three men on the road *Det har tre män kommit på vägen. there has three men come on the road Det har dansat tre män på vägen. there has danced three men on the road *Det har tre män dansat på vägen. there has three men danced on the road
The contrast between (5) and (6) is expected: if both the verb and the DP stay in situ, the DP argument of the ergative predicate must appear to the right, in the complement. As for (7) and (8), we actually expect the opposite result, given (4) above. The conclusion is that the nonfinite verb actually moves overtly to some position higher than VP while the subject stays in situ in Spec V. The structure of (7) must then be (9): (9)
[InflP har [YP dansati [VP [Spec tre män] ti ]]]
There is more to this however: (9) builds on the assumption that auxiliaries are inserted directly under T or Infl, which is a difficult assumption for Swedish for the following reasons: first, in the subordinate clause, finite verbs as well stay in situ (cf. Platzack & Holmberg 1989, Holmberg & Platzack 1995). Second, the negation in Swedish, as generally in Scandinavian, is normally assumed to be adjoined to VP, or to head its own projection between VP and AgrOP. Recall that object shift of DP in Icelandic, and of weak pronouns in Swedish, targets a position higher than negation (cf. Holmberg 1986, Jonas & Bobaljik 1993; see also Watanabe 1993). Since the auxiliary regularly appears to the right of negation in subordinate clauses, we must assume that the auxiliary is base generated lower than NegP (hence; lower than AgrO): (10)
… att han inte har kommit. that he not has come
The structure of (10) at Spell-out would then be (11), omitting the subject:20 (11)
… Comp … [TP T ([AgrOP AgrO) [NegP Neg [AuxP har [VP kommit]]]]]
Now, if (5)-(8) are reformulated in an embedded and negated context, it is still clear that the participle V must precede the argument:
20
AuxP could be equal to the light VP of Chomsky (1995).
102
(12)
… att det inte har kommit tre män på vägen. …that there not has come three men on the road *… att det inte har tre män kommit på vägen. …that there not has three men come on the road … att det inte har dansat tre män på vägen. …that there not has danced three men on the road *… att det inte har tre män dansat på vägen. …that there not has three men danced on the road
(13) (14) (15)
If the auxiliary is now in situ, the relevant position into which the nonfinite verb moves in (14) must actually be between VP and AuxP, hence lower than both NegP and TP: (16)
TP
T
Neg P
Neg inte
Aux P
Aux har
YP
Y dan sat
VP
Spec tre män
V’
V t Given that AgrOP is presumably above NegP (when it is projected) as indicated in (11), the most natural candidate for YP in (16) is AspP for which I will argue in the following chapter. The data suffice for a first conclusion, namely that the Swedish participial phrase indeed hosts abstract inflectional material, and that the structure is richer than what suggested in Chomsky (1993).
103
Appendix 2 to Chapter 2: SOV and Equi-Distance
In the second chapter, I have formulated a hypothesis that is in conflict with some approaches to Locality. Crucially, I suggest that the direct object in (1) moves in overt syntax to the specifier of a functional projection that I assume to be AgrOP. If we follow the subjectinternal-to-VP hypothesis of Manzini (1988), Koopman & Sportiche (1991), and many others, the object would have to move over the subject, base generated in Spec V as illustrated in (1). As the derivation continues, the subject moves higher in the structure but this is less relevant here. (1)
AgrOP
Spec DP(O)i
AgrO’
AgrO
VP
V’
Spec DP(S) V
ti
The derivation of (1) is excluded under Relativized Minimality (Rizzi 1990), if both dependencies, object-trace and subject-trace, are A-dependencies, as indeed we have suggested they are. The subject (or its trace) in Spec V will count as a barrier for the trace of the object, which thus fails to be antecedent governed by DP(O) in Spec AgrO. Still, there are constructions where movement operations like (1) seem to be allowed, as for instance object shift in Icelandic. Chomsky (1993) proposes to overcome minimality violations as the one illustrated in (1) through the introduction of Equi-distance. In essence, the proposal is that DP(O), like any other element, obeys the Minimal Link Condition and must move by making the shortest step. Thus, it can only move to Spec V. However, DP(O) can arrive at Spec AgrO skipping Spec V, if and only if Spec V and Spec AgrO can be considered equally distant from DP(O). Equi-distance is defined in terms of domains where the domain of V is VP, the domain of AgrO is AgrOP and so on. If the verb raises to AgrO, the domain of [AgrO V AgrO] comprehends all of AgrOP and VP, in which case the specifiers of V and AgrO count as equally distant for DP(O). In this way, movement of the object to Spec AgrO across the
104
subject in Spec V is possible if and only if the verb raises to AgrO. Subsequent raising of the verb higher than AgrO will create new domains, making it possible for the subject to move over the object in Spec AgrO. Chomsky’s analysis captures the generalization advanced by Holmberg (1986) that object shift in Scandinavian is parasitic on verb-raising; see also Déprez (1989), Vikner (1990), Bures (1992), and Jonas & Bobaljik (1993). The premises for this discussion change radically when Antisymmetry is assumed, hence not only for Kayne (1994) but also for Chomsky (1995). If left-headedness is universal - if all human languages are SVO - derivations such as (1) must be generally allowed also in languages where the verb does not raise overtly. This must be true for any language and any construction where both subject and object appear to the left of the verb. For present purposes, it is sufficient to show that my account can be pursued in spite of the minimality problem. This is indeed so. As far as I know there are at least two ways to circumvent the problem, both of which are compatible with my account in principle. The first alternative is to follow the argumentation of Ferguson & Groat (1994) and Chomsky (1995) by saying that although Spec V intervenes between the basic position of DP(O) and Spec AgrO, DP(O) can move to Spec AgrO over Spec V as Spec AgrO is the closest position where DP(O) can check its features. Crucially, DP(O) can only move to a position to which it is attracted. Since DP(O) cannot check any features in Spec V, this position does not count for it and can be skipped. The second alternative is to maintain Chomsky’s (1993) notion of Equi-distance assuming a representational view in the following sense: Spec V and Spec AgrO are made equally distant for DP(O) as a dependency is formed out of AgrO and V, but not necessarily through overt verb-raising. Suppose UG allows for some other mechanism that combines V and AgrO, creating a domain out of AgrOP and VP. In the spirit of Guéron & Hoekstra (1988) or Manzini (1992), we might assume that AgrO and V, and other functional heads, form a dependency through the percolation of some feature. Alternatively, we can follow the more recent reformulation of the same hypothesis suggested by Manzini (1994a, 1994b). For Manzini, a long dependency involving [C, I, V] is formed out of the subdependencies [I, V] and [C, I]. The first of these, [I, V], is independently formed as a case of saturation and the second, [C, I], through the anchorage of tense. The long dependency, like its subdependencies, is thus formed without overt movement of V to I and C. I conclude that the theory, at present, must permit derivations as (1) quite independently of the approach outlined here, and that a solution to this particular problem of minimality can be built on assumptions already made in the literature.
105
CHAPTER 3 The Aspect Node: Accusative Checking, AspP, and the Affectedness Constraint
0
Introduction: The Affectedness Constraint and Case Theory
As I stated in the Introduction, I will assume a syntactic Asp head, following many scholars, e.g. Emonds (1985), Tenny (1987), Belletti (1990), Diesing & Jelinek (1993), Borer (1993; 1995), and Travis (1995). In the following section, I will investigate some of the empirical support for such a move and possible theoretic implications. The issue is of great importance regardless of whether the Asp projection is assumed instead of AgrO, as in Borer (1995), or along with AgrO, as in Belletti (1990). As mentioned, I will follow the second option assuming both AgrO and Asp. Moreover, along the lines of Cinque (1994) among others, I will assume that clause structure hosts more than one Asp projection. The discussion in this chapter will however only deal with an Asp head linked to Aktionsart or more precisely, the locus of ‘delimitedness’. Leaning on previous work of Tenny (1987) and Borer (1993; 1995), I will use the notion delimitedness to analyze the long-discussed matter of affectedness contexts, i.e. those where an argument (usually internal, direct) is in some sense affected by the event expressed by the predicate. References to the extensive literature on the subject will be given as we proceed. Inevitably, the discussion will bring us to other issues that may be directly or indirectly related to affectedness, and some of these cannot be thoroughly sorted out for reasons of limited time and space. I will dedicate the major part of my demonstration to the construction mentioned in chapters one and two, which I have called PAP (Possessive Auxiliary + Participle phrase); PAP is shown to be subject to the Affectedness Constraint but has not yet been discussed in the literature. The affectedness constructions that will be considered here are mainly three: 1. PAP; 2. The English Middle (‘this car parks easily’); 3. The nominal passive (‘the city’s destruction by the Mongols …’). To these, I will add a fourth by the end of this chapter, the Romance causative, formed out of fare + infinitive (far riparare la macchina da qualcuno, lit. ‘to make repair the car by someone’). The conclusions of this discussion are essential for the following analysis of the ancient Italian data (Part B, chapter six) where I will argue that Modern Italian absolute small clauses formed out of past participles enter under the general analysis and can therefore be considered as affectedness constructions along with the rest. Resultative, secondary predicates (‘hammer the metal flat’) constitute a fifth context related to the affectedness issue, but I will save my remarks on these for the last chapter of this thesis (Part C, chapter eight).
106
This chapter is organized as follows: after some background discussion on affectedness and AspP in section 1, I will turn, in section 2, to PAP (‘Possessive Auxiliary + Participial clause’). It will be shown that PAP is subject to the Affectedness Constraint and this observation has some non-trivial consequences for the general analysis of affectedness. Under 3, I will formulate a hypothesis that permits us to derive the Affectedness Constraint in PAP, the English Middle, and the nominal passive, and, at the same time, explains why the constraint is not operative in the verbal passive, the adjectival passive and the compound tense (the comparison between the compound tense and PAP will be of crucial importance, for reasons that will become obvious as we proceed). I propose to capture DP-movement and its correlation with the suppression of external arguments in a unitary account, presented in 3.4.. As this has yet not been achieved in the generative model (or any other), the program of research is, I believe, theoretically warranted. In section 4, I will briefly discuss how the hypothesis can be accommodated to theoretic frameworks and general assumptions that are different from those I will follow in the analysis. Finally, in section 5, I will discuss how the account extends to the Romance Causative, the fare + infinitive construction.
1
Theoretic Background to the Discussion on Affectedness and its Correlation with Aspect, Case, and Argument Structure
In this section, I will give some background to the problem of affectedness and a brief review of previous approaches. My own account, which follows in section 3, draws inspiration from some works in particular, and I will therefore present only those suggestions that I have chosen to follow (directly or indirectly). Other alternatives that have been explored throughout the years, that have less in common with the approach I wish to develop will be left aside.1
1.1
The Affectedness Constraint (Anderson 1979)
The notion of affectedness was brought up in generative grammar through Anderson’s (1979) treatment of the noun phrase. She observed that certain predicates that are compatible with the verbal passive are ungrammatical in the nominal passive. ‘Perform’, ‘destroy’, ‘avoid’, and ‘pursuit’ can all be used in the verbal passive but only ‘performance’ and ‘destruction’ are compatible with the nominal passive (examples from Anderson 1979, 43 and Tenny 1987, 60).
For some, affectedness is incorporated into θ-theory. Rizzi (1986a), followed by Roberge (1991), assumes a particular, affected θ-role, a subclass of Theme, assigned only to affected arguments.
1
107
(1a) (1b) (2a) (2b)
The company’s performance of the play … The play’s performance by the company … The Mongols’ destruction of the city … The city’s destruction by the Mongols …
(3a) (3b) (4a) (4b)
John’s avoidance of Bill … *Bill’s avoidance by John … Sally’s pursuit of the cat … *The cat’s pursuit by Sally …
Anderson noticed that the origins of this contrast lie in a difference in event type. She formulated this difference in terms of how an event affects its participants: «‘Bill’ is the object of ‘avoidance’ but the person referred to by ‘Bill’ is not directly affected by the action described. ‘Destruction’, the noun usually used as an example of NP-preposing, is an extreme example of an object being affected by the action of the nominal.» In order to be affected, an argument «… must be changed or moved by the action of the head nominal».2 Nominal predicates as ‘performance’, ‘definition’, ‘publication’ and ‘translation’ enter under the same class, as they «… have this feature in common: that although they do not imply a change in the abstract object, they do imply the creation of an object. Thus in one sense the play is not changed by any performance of it, but each performance creates an entity that is also ‘the play’.»3 Anderson recognizes that the term ‘affected’ sometimes must be understood in an extended sense, and writes: «Although I am not prepared to state the principle more specifically I think that the basic idea that the complement NP must be affected in some way is correct.»4 As a matter of fact, Anderson’s statement appears to be the most specific description ever given to the notion of affectedness. Numerous scholars have noticed the effects of affectedness elsewhere in syntax (we will go through the most well known of them in this chapter), but have had little to contribute to the definition of the concept. In many approaches, the effect has been assumed as an independent principle, the Affectedness Constraint, but attempts have also been made to capture the fundamental generalizations behind it, reducing the constraint to more general mechanisms. Scholars have repeatedly reflected upon the fact (see for instance Tenny 1987, 9) that the Affectedness Constraint systematically correlates with: 1. DP-movement 2. Case, and 3. the syntactic organization of argument structure, more precisely with the suppression of the external argument. 2
Anderson (1979, 44). Anderson (1979, 44). 4 Anderson (1979, 45). 3
108
The program of research I will pursue here goes back primarily to Tenny (1987), but draws inspiration from other models as well. Under 1.2- 1.4, I will give a brief review of these, setting aside other approaches - there are numerous - that have less in common with the ideas I wish to develop.
1.2
Delimitedness (Tenny 1987)
Tenny (1987) tries to get rid of a large and somewhat confusing terminology by imposing the distinction delimited/non-delimited to replace earlier dichotomies: «Affectedness verbs have the property that they describe an event in which the direct argument of the verb is capable of delimiting the event.»5 Tenny concentrates her demonstration on showing that delimitedness indeed subsumes affectedness. It is obvious that the delimited/non-delimited distinction comes close to some other familiar dichotomies as telic/atelic or the bounded/unbounded (the former used by Comrie 1976 and Bertinetto 1986, the latter by Platzack 1979; cf. also Dahl 1981). For Comrie (1976), a telic event describes «a process that leads up to a well-defined terminal point, beyond which the process cannot continue».6 Tenny comments on the similarity, but her discussion is not extensive on this point. The question of exactly how delimited/nondelimited relates to telic/atelic, and indirectly, how the notion of telicity relates to affectedness, are open issues after Tenny’s demonstration. This uncertainty will be of some importance as we proceed with the issue in chapter six (cf. p. 261). Tenny proposes to capture the correlation between affectedness, argument structure and Case through the Aspectual Interface Hypothesis, which determines how aspectual interpretation is mapped onto syntax. In other words, aspectual interpretation (the delimited/non-delimited distinction) regulates (a) how an argument is projected in syntax (internal or external) and (b) what Case it receives: «Each event participant must be marked as to (…) whether it undergoes change or not, and if so, if it undergoes change in such a way that it can measure out the event; i.e. whether the event participant can be an internal argument or not.»7 «(…) map an argument that is marked as a possible direct argument into the direct argument position. It receives structural accusative case.»8 Tenny’s proposal has the effect of depriving the Affectedness Constraint of its theoretic status. There is no independent Constraint operating in some syntactic module, rather the effects observed are the result of the interaction of more general principles concerning aspectual interpretation and argument projection. Moreover, Tenny’s rules of 5
Tenny (1987, 112). Comrie (1976, 45). 7 Tenny (1987, 307). 8 Tenny (1987, 309). 6
109
aspectual mapping onto syntax are intended to replace thematic roles, which she criticizes as loose; beyond the aspectual relations, syntax does not need to make reference to any further semantic properties of the arguments.
1.3
Resultative Case (De Hoop 1992)
Tenny examines some languages where aspectual information has overt morphological realization. She notes, for instance, that Finnish partitive is morphologically distinct from accusative (she discusses this on pp. 47-50). Thus, the difference between aspects in Finnish does not only emerge from the way arguments are projected but also from the morphological Case they carry. Tenny does not pursue a possible consequence of this observation: that the aspectual information is encoded not only in the argument structure of a predicate but also in Case itself.9 This step is taken by de Hoop (1992). First of all, de Hoop adopts the MPLT dichotomy between weakness and strength of features, but gives semantic content to the terms (the reader is referred to her discussion in chapters 1 and 3). Moreover, de Hoop argues against the Partitive Hypothesis of Belletti (1988), noticing, for instance, that the alternation between the two objective Cases in Finnish under discussion, accusative and partitive, correlates not only with the definiteness/indefiniteness distinction but also with regard to resultativity versus irresultativity of the predicate. Thus, in sentences such as (5)/(6) resultativity, and not definiteness, is crucial for the choice of Case (de Hoop 1992, 64). (5)
Anne rakensi taloa Anne built house[part]
Irresultative ‘Anne was building a/the house’
(6)
Anne rakensi talon Anne built house[acc]
Resultative ‘Anne built a/the house’
To de Hoop, the objective Case in (5) (the partitive), rather than being inherent, is still a structural Case (that in her model is defined ‘default’). There is one idea in her demonstration that above all has relevance for the analysis that follows below, namely that the objective Case in (5), the Finnish partitive, is to be defined a weak structural Case (p. 74), and the objective Case of (6), the accusative, is strong. In other words: the irresultative or default objective is weak and the resultative objective is strong. De Hoop’s claim that weak Case is checked at D-structure and strong Case at S-structure (p. 77) is best translated into MPLT in
9
In some approaches, morphological Case is assumed to be inherent. See Belletti (1988) on the Finnish partitive, but also Lasnik (1992) who suggests that partitive may be structural.
110
terms of overt or covert checking; if a Case is strong it requires checking before Spell-Out, if it is weak it stays in situ until after Spell-Out.10 Even in the case just discussed, it is clear that parametrization must be assumed in two senses: First, the semantic specification of Cases varies cross-linguistically. The Finnish partitive is specified irresultative, whereas Turkish partitive is not, for instance (De Hoop, 68). Second, a Case which is defined weak in some languages may actually be defined strong in others.
1.4
AspP and the Syntactic Ordering of Arguments (Borer 1993)
A somewhat different view is taken by Borer (1993, 1995) and Arad (1995). Borer creates a link between Aspect and Case theory and she postulates an Asp node (which replaces the AgrO of Chomsky 1993). Borrowing Tenny’s (1987) terminology, Borer assumes a delimitedness feature that may be generated in this Asp head. Its specifier is the checking site for delimited arguments. Importantly though, the delimitedness feature may be distinct from Case itself, as Spec Asp, when projected, may or may not be a Case position. Typically, delimited arguments that do not carry accusative move through a Spec Asp which is specified +delimited but -accusative. The assumption of AspP is crucial for Borer’s approach to argument structure. The basic idea of her hypothesis is that arguments are syntactically projected unordered with respect to the internal/external distinction of Williams (1981). Against the standard assumption (cf. Burzio 1986, and Koopman & Sportiche 1991 for the subject-internal-to-VP-hypothesis) that ergatives, unergatives and transitives are projected into syntax as (7), (8) and (9) respectively, she assumes an unordered projection such as (10) and suggests that the arguments are ordered by syntactic movement through Asp in accordance with the aspectual relation they have to the predicate:11
10
De Hoop reaches this conclusion, and I take it she is right. Obviously, differences in word order do not emerge from (5)/(6). 11 Borer’s proposal is the continuation of a series of attempts to derive the projection of argument structure while dismissing with θ-theory; apart from Tenny (1987), cf. also Hale & Keyser (1993) and Hoekstra & Mulder (1990). Among these attempts to explain argument structure, Borer’s Aspect approach appears to be the most drastic. In the model of Hale & Keyser (1993), argument structure is defined lexically, but in a part of the Lexicon, the Lexical Relational Structure (LRS), which, in their words, functions itself as a syntax, in the sense that it conforms to X’-theory, and processes taking place in LRS obey the principles of Unambiguous Projection (binary branching) and Full Interpretation. I wish to point out that the line of research that I will pursue is in principle independent of these ideas; the account I will present below can, as far as I can see, be stated with or without reference to thematic roles.
111
(7)
(8)
Ergat ive:
VP
Tran sitive:
VP
V’
Spec DP(S)
V
(10)
(9)
Unergati ve:
Comp l DP(S)
VP V’
V
Spec DP(S)
V’
V
Comp l DP(O)
Unordered projection of arguments (Borer 1993, 1995):
AspP
Spec
Asp’
Asp
VP V, DP 1, DP 2, …
Apart from these proposals, a variety of different ideas have been defended concerning AspP. Firstly, it has been suggested that Asp makes up part of the VP shell, rather than being external to VP; this is the claim of Travis (1995). Secondly, it is assumed by Cinque (1994), among others, that there are more than one AspP in clause structure. Both of these claims have some relevance for the discussion of this and the following chapter.
1.5
The Central Claims of this Chapter
I will make use of AspP and the notion of delimitedness in a way that is inspired by the accounts presented above, and this work is in many ways a continuation of these. My analysis builds on the following assumptions: 1. Disregarding AgrO, I assume that (11) corresponds to the lower part of sentence structure. (11)
T … [AspP Asp2 [AspP Asp1 [VP V]]]
112
I assume that the lower AspP1 in (11) is the locus for the delimitedness feature. 2. I will not follow the idea that arguments are projected into syntax unordered. Thus, for reasons that will become obvious below, I maintain the external/internal distinction as stated in (7), (8), and (9). 3. I assume AspP1 makes up part of the VP-shell and is thus internal to VP, whereas the higher AspP2 is a purely functional projection, external to VP. As a consequence of this, I assume that the specifier of AspP1 may count as the subject of VP. I will discuss below how these projections relate to Case checking.12 The points in 1-3 will turn out to have interesting consequences for the analysis of the Affectedness Constraint, and they will allow us to reduce the constraint to more general mechanisms at work in syntax.
2
One Affectedness Construction in Italian and Scandinavian: Possessive have + participle
Let us now examine the participial construction mentioned before, which is present in Romance, Scandinavian and English.13 Consider the examples in (12): (12a) (12b)
It: Sw:
Ho le valigie fatte. Jag har väskorna packade. I have the suitcases packed[+Agr]
(12) instantiates a Possessive Auxiliary verb that takes a participle clause as its complement; I will henceforth refer to this as PAP. I have chosen the generic definition possessive auxiliary; there are obvious parallels to similar constructions not involving have and not being
12
Ultimately, I build on a suggestion from Maria Rita Manzini (p.c.) that the recursions of VP in the Larsonian shell (Larson 1988, Chomsky 1995) may be considered Asp Phrases, where the various Asp heads host information about the aspectual relation between the arguments and the predicate. Travis (1995) make a similar claim, as she assumes that AspP belongs to l-syntax, or the Lexical Relational Structure of Hale & Keyser (1993). How far such ideas can be pursued remains an open question. 13 Among English speakers, there is some variation concerning the interpretation of (12). To some of my informants, have in the English example corresponding to (12) - see the gloss - can clearly be both possessive and causative. To others, a possessive reading of (12) is awkward; it would be better expressed I have got my suitcases packed. Some informants have a causative reading in both. The difference may have to do with variety. For those English speakers who accept the gloss of (12) with a possessive reading, the English PAP is parallel to the same construction in other European languages under examination, and the following analysis proposes to account for their judgments. I have no suggestions to make with regard to the causative reading of (12).
113
exclusively possessive; tener/ter + past participle in Iberoromance (Lois 1990, Harre 1991), to some extent få+participle in Scandinavian (Vikner & Sprouse 1988, Christensen & Taraldsen 1989, Platzack 1989, and Hedlund 1992) and get + participle in English (e.g. Haegeman 1985). PAP occurs in several, in fact most Romance and Germanic languages (the exceptions are German and Dutch), which also make frequent use of (13), the compound tense. (13a) (13b)
It: Sw:
Ho fatto le valigie. Jag har packat väskorna. I have packed[-Agr] the suitcases
PAP is generally distinguished from the compound tense, formed with have, by a number of characteristics: 1. As for participial agreement, the reader is referred to the discussion in chapter 2 (we will return to the issue in chapter 4). In PAP the participle agrees with the DP appearing to its left in those languages where participles have this property (agreement is visible in Romance and Scandinavian). In the compound tense, the participle never agrees with an object to its right. (14a) (14b)
It: Sw:
*Ho le valigie[pl.fem] fatto[-Agr]. *Jag har väskorna[pl.] packat[-Agr].
2. The difference in word order will be crucial for the following analysis: In PAP the DP object (henceforth DP(O)), the suitcases of (12), appears obligatorily to the left of the participle, whereas in the compound tense obligatorily to the right. 3. In the compound tense (henceforth CT), the DP subject (henceforth DP(S)) carries the θrole assigned to it by the main verb, agent in (13), whereas in (12) it is the possessor of the object. The PAP construction denotes the subject’s possession of the result of a past action. The agent of this past action remains implicit and arbitrary; it appears therefore that the participial predicate in PAP has undergone passivization. 4. There are lexical restrictions on PAP, quite different from those on the compound tense; in languages like English or Swedish, where have is generally used in CT, practically any verb can combine with have to form a compound tense, with few exceptions (such as English modals, which will not concern us here), but not any verb can be used in PAP. I believe these differences can be resolved on minimal assumptions concerning mainly Case theory. I will begin the discussion backwards, starting with the last issue, the lexical restrictions, in section 2.1. From now on, I will concentrate my demonstration on Italian and
114
Swedish, leaving English aside. The analysis will carry over to some but not all English speakers, see footnote 13. I am not aware of whether other Romance and Scandinavian languages differ substantially from the facts presented on Italian and Swedish.
2.1
The Lexical Restrictions on PAP
First of all, the participle in the PAP-construction has to be of a transitive verb; ergatives in PAP are generally awkward: (15a) (15b) (16a) (16b) (17a) (17b)
*?Abbiamo gli ospiti arrivati. *?Vi har gästerna anlända. We have the guests arrived *?Gli alberi hanno le foglie cadute. *?Träden har sina löv fallna. The trees have their leaves fallen *?Ho le scarpe scivolate (sul ghiaccio). *?Jag har skorna glidna (på isen). I have the shoes glided (on the ice)
The same is true for unergative Vs:14 (18a) (18b) (19a) (19b)
*?Abbiamo un tango ballato. *?Vi har en tango dansad. we have a tango danced *?Abbiamo tre chilometri corsi. *?Vi har tre kilometer sprungna. we have three miles run
PAP is thus only possible with transitives. However, we also need to distinguish between different event types. The DP of the participial clause in PAP must be an internal argument that undergoes a change of state. Hence, the construction is subject to a restriction of affectedness. This can be shown by confronting PAP with two affectedness constructions usually advanced as a point of reference, the English Middle and the nominal passive, cf. Anderson (1979), Keyser & Roeper (1984), and Tenny (1987):
14
The ungrammaticality of Swedish unergative verbs in PAP may also be due to morphology in some cases, as there is no fledged participial form corresponding to unergative verbs, but only the active participle, referred to as the supinum in traditional grammar.
115
1. Verbs which can be used in PAP are the same that can appear in the English Middle: (20) (21) (21a) (21b)
(22) (23) (23a) (23b)
The Middle: This car parks easily (on the street) PAP: Ho la macchina parcheggiata (per strada). Jag har bilen parkerad (på gatan). I have the car parked (on the street) The Middle: Certain letters write easily. PAP: Ho le lettere scritte. Jag har breven skrivna. I have the letters written
Perception verbs, psychological activity verbs and Exceptional Case Marking verbs are all excluded: (24) (25a) (25b) (26) (27a) (27b) (28) (29a) (29b)
*The mountains see beautifully after rain. *Abbiamo le montagne viste. *Vi har bergen sedda. we have the mountains seen *This theorem learns fast. *Abbiamo questo teorema imparato. *Vi har detta teorem lärt. we have this theorem learnt *John believes to be intelligent easily. *Ho Gianni considerato intelligente. *Jag har Johan ansedd intelligent. I have John considered intelligent
2. PAP participial predicates are equivalent to those nominal predicates that admit the nominal passive: (30) (31) (31a) (31b)
The Nominal Passive: The city’s destruction by the Mongols … PAP: Avevano la loro città distrutta. De hade staden ödelagd. they had the city destroyed
116
(32) (33) (33a) (33b)
(34) (35) (35a) (35b)
The Nominal Passive: The natives’ conversion by the missionaries … PAP: I missionari hanno gli indigeni convertiti. Missionärerna har infödingarna omvända. the missionaries have the natives converted The Nominal Passive: *Bill’s avoidance by John … PAP: *Gianni ha Carlo evitato. *Johan har Karl undviken. John has Bill avoided
Verbs of consumption, creation, and similar types also show the analogy between the middle, the nominal passive, and PAP: (36) (37) (38a) (38b)
This play performs easily. The play’s performance by the company … L’autore avrà la commedia rappresentata … Författaren kommer att ha sin kommedi framförd … the author will have his comedy performed
(39) (40) (41a) (41b)
This poem translates easily. The poem’s translation by John … Abbiamo il poema tradotto. Vi har dikten översatt. we have the poem translated
(42) (43) (44a) (44b)
Hot bread cuts easily. We’d better leave the cake’s cutting to Mary. Abbiamo il pane tagliato. Vi har brödet skuret. we have the bread cut
Verbs with unaffected direct arguments cannot be used in PAP. Statives are excluded, and so are predicates where the external argument, but not the internal, undergoes change: (45)
*Vi har matematik gillad. we have mathematics liked
117
(46a) (46b) (47) (48a) (48b)
*Ho il tuo fratello conosciuto. *Jag har din bror känd. I have your brother known *Ha il passato ricordato. he has the past remembered *?Abbiamo gli uccelli osservati. *?Vi har fåglarna observerade. we have the birds observed
Recall that ergatives are generally awkward in PAP, even though their direct arguments are affected or delimit the event (cf. Borer 1993, 1995). In sharp contrast with ergative verbs of motion, transitive verbs of motion are grammatical in PAP: (49a) (49b) (50a) (50b)
*Ho la macchina andata (dal meccanico). *Jag har bilen åkt/åken (till reparatören). I have the car gone (to the mechanic) Ho la macchina depositata (dal meccanico). Jag har bilen avlämnad (till reparatören). I have the car delivered (to the mechanic)
The same contrast is equally clear with verbs of the spray/load type, referred to by Tenny (1987, 67) as verbs of ‘total affectedness’. When transitive, these permit PAP with either of their two internal arguments: (51a) (51b) (52a) (52b)
Abbiamo il camion caricato di frutta. we have the truck loaded with fruit Abbiamo la frutta caricata sul camion. we have the fruit loaded on the truck Vi har färgen sprejad på väggen. we have the paint sprayed on the wall Vi har väggen sprejad med färg. we have the wall sprayed with paint
When ergative, such verbs do not permit PAP in any of the possible combinations. The ungrammaticality of (53)/(54) should be compared with the grammaticality of gli uccelli abbondano nel bosco/il bosco abbonda di uccelli for Italian, and bina svärmar i trädgården/trädgården svärmar med bin for Swedish:15 15
All of the logically possible combinations are ungrammatical: *Gli uccelli hanno il bosco abbondato/*Il bosco ha gli uccelli abbondati, *Bina har trädgården svärmad/*Trädgården har bina svärmade.
118
(53a) (53b) (54a) (54b)
*Abbiamo il bosco abbondato di uccelli. We have the wood abounded with birds *Abbiamo gli uccelli abbondati nel bosco. We have the birds abounded in the wood *Vi har trädgården svärmad med bin. we have the garden swarmed with bees *Vi har bina svärmade i trädgården. we have the bees swarmed in the garden
It is therefore motivated to assume that the derivation of well-formed PAP-constructions with verbs like burn and melt (AVB/BV-alternation; Burzio 1986) are ‘deep’ transitive constructions and not ergatives. (55a) (55b)
Avevamo il ghiaccio sciolto. Vi hade isen smält. we had the ice melted
The evidence for this is that in Swedish, where transitive and ergative verbs of the relevant class are often morphologically distinct, only the transitive is admissible in PAP: (56) (57) (58a) (58b) (59a) (59b)
Ho la casa bruciata. I have the house burnt Abbiamo la nave affondata. We have the ship sunk Vi har ljusen brända. We have the lights burnt[transitive] *?Vi har ljusen brunna. We have the lights burnt[ergative] Vi har skeppet sänkt. We have the ship sunk[transitive] *?Vi har skeppet sjunket. We have the ship sunk[ergative]
(cf. bränna ‘burn’, transitive) (cf. brinna ‘burn’, ergative) (cf. sänka ‘sink’, transitive) (cf. sjunka ‘sink’, ergative)
It is clear that PAP is subject to the Affectedness Constraint. The parallel between PAP, the middle, and the nominal passive is obvious. In 2.2., I will make a comparison with another Swedish construction, få (‘get’) + past participle (presented in chapter two) in order to show that the lexical restrictions on få + p.p. and have + p.p. are the same.
119
2.2
A Note on Få + Past Participle
Få + participle has causative meaning and does not have direct equivalents in Romance.16 Nevertheless, I will use the term PAP to indicate it. Få + participle is possible with transitive verbs of affectedness: (60) (61) (62) (63)
Jag fick väskorna packade. I got the suitcases packed Vi fick frukten lastad på lastbilen. We got the fruit loaded on the truck Dom fick staden förstörd. They got the city destroyed Jag fick bilen skjuten (till reparatören). I got the car pushed (to the mechanic)
It is impossible with unergatives and ergatives: (64) (65) (66)
*?Vi fick tre kilometer sprungna. We got three kilometers run *Vi fick gästerna anlända. We got the guests arrived *Vi fick bilen åkt/åken (till reparatören). We got the car gone (to the mechanic)
It is furthermore impossible with unaffecting predicates of the types already illustrated: (67) (68) (69)
*Vi fick bergen sedda. We got the mountains seen *Vi fick Johan undviken. We got John avoided *Vi fick läxan lärd. We got the lesson learned
Not unexpectedly, the causative PAP-få patterns with the possessive PAP-hava in Swedish. I conclude that both of these can be defined as affectedness constructions.
16
Få+participle is quite similar to both get and have + participle in English. Some English speakers tend to use have in a fashion quite similar to Swedish få, see f.n. 12 above. As opposed to get, få does not admit the passive use of His leg got broken, though the equivalent to He got his leg broken is perfect (cf. Haegeman 1985 and Hedlund 1992).
120
3
Analysis
The aims of my analysis are those stated in the Introduction, which I repeat: the analysis of the Affectedness Constraint should account for how the constraint correlates with17 1. DP-movement, 2. Case, and 3. the syntactic organization of argument structure, more precisely with the suppression of the external argument. In addition to this, we need to explain why ergative predicates are disallowed in PAP, although ergatives may be telic or delimited (cf. Borer 1993, 1995). I will begin my demonstration in 3.1., by explaining why ergatives are excluded. In 3.2., I will discuss the correlation between affectedness, DP-movement and Case, and under 3.3.-3.4., I will link the analysis to the suppression of external arguments. The result of this discussion is that the Affectedness Constraint is deprived of theoretic status; its effects are the results of some underlying, more general principles of grammar.
3.1
Passive Participles as Case Checkers
One possibility that naturally offers itself is to analyze PAP as an ECM context. In a previous framework, it has been independently suggested that the governing V can assign accusative down into the participial complement; see Hedlund (1992) on få+participle in Swedish and Lois (1990) for tener+participle in Spanish. However, I believe that this is not correct and that the apparent similarity between PAP and ECM contexts is misleading. The reasons for this are the following: First of all, it is unclear if the V have has the property of assigning accusative at all. Kayne (1989a, 88) argues that the Tense auxiliary have is not a Case assigner. If a unified account of Tense auxiliary have and possessive have is to be pursued (Guéron 1986, Kayne 1993, Mahajan 1994, and Cocchi 1994), we should avoid positing a difference in Case assigning/checking properties. These proposals essentially mean that expressions of possession are deep existentials of sorts: possessive have is derived from be through the incorporation of a prepositional element. It is therefore possible that, in a normal expression of possession such as John has a sister, the nominal complement of have is not an argument but a predicate, and as such does not need Case (following the lines of Guéron 1986 and Guéron & Hoekstra 1988). Second, when we assume have to be Exceptional Case Marker, we must explain why PAP resists passivization differently from normal ECM contexts. Compare (70) to (71):
17
Cf. Tenny (1987, 9).
121
(70a) (70b)
*Le lettere sono state avute scritte. *Breven har varit havda skrivna. The letters have been had written
(71)
John was believed to be intelligent.
The impossibility extends to PAP constructions of the Iberoromance type:18 (72)
*Las cartas han sido tenidas escritas.
Third, recall that ergative predicates are awkward in PAP, despite their argument being affected. (15) and (16) are here repeated as (73) and (74): (73a) (73b) (74a) (74b)
*?Abbiamo gli ospiti arrivati. *?Vi har gästerna anlända. we have the guests arrived *?Gli alberi hanno le foglie cadute. *?Träden har sina löv fallna. the trees have their leaves fallen
If Case checking depends on the participial V, (73) and (74) would follow on the current assumption that ergatives are intrinsically incapable of assigning accusative. If accusative comes from the higher V, be it have, få or tener, the prohibition on ergatives will not fall out. Hedlund (1992) therefore cannot adequately explain the ungrammaticality of (75) and (76):19 (75) (76)
*?Vi fick gästerna anlända. we got the guests arrived *?Träden fick sina löv fallna. the trees got their leaves fallen
Kjell-Åke Gunnarson (p.c.) points out that an ergative predicate is possible in a prepositional small clause as (77) in Swedish: (77)
Med gästerna anlända satte vi oss till bords. with the guests arrived sat we down at the table
18
This circumstance is of particular interest as tener otherwise can be passivized, at least for some speakers. In Egerland (1994) I have speculated about the origins of this difference. 19 Also Lois (1990) must face the same problem for tener+participle in Spanish. It is not straightforwardly clear what rules out (i) if not a failure of Case assignment/checking. (i) *Tenemos el coche ido (a el mecanico). we have the car gone (to the mechanic)
122
This supports my thesis: in (77) the preposition is responsible for Case assignment, whereas in (75)-(76) there is no Case assigner or (checker) available. As we have excluded that possessive have is responsible for Case checking, we are left with the assumption that the DP’s accusative Case in PAP is checked inside the participial clause, even though the participle is passive.20
3.2
Affectedness, DP-Movement, and Case
The direct object of the PAP construction has turned into the subject of the participial phrase, and the minimal assumption is therefore that it is projected as such, in the complement position of the participial V. Theoretical simplicity thus favors an analysis according to which the object of PAP is promoted to subject of the participial clause through syntactic movement.21 Suppose this movement is forced by the need to check the delimitedness feature in an Asp head. The delimitedness feature is compatible only with predicates of affectedness
20
The idea that UG allows for a ‘second’ accusative or objective Case (for instance, in the sense that V may retain its property to assign Case even after passivization) has been repeatedly suggested in the literature, cf. Jaeggli (1986) and De Hoop (1992); the Partitive Hypothesis of Belletti (1988), Mahajan (1990) and Lasnik (1992) is a variation on the same theme. On the contrary, Mahajan’s (1994) assumption that participial Vs are not Case assigners cannot be followed. The possibility of having accusative marked DPs in Italian absolute small clauses (Belletti 1990) corroborates the view I am taking here. 21 In previous approaches the conclusion is necessary: according to Tenny’s (1987) Aspectual Interface Hypothesis, see 1.2. above, affected or delimited arguments are necessarily projected into syntax as internal direct arguments; theta-based approaches as Rizzi (1986a) and Roberge (1991) will have the same result. The conclusion may also have direct empirical support. Guglielmo Cinque (p.c.) points out that extraction of ne in (i), PAP, is awkward as opposed to (ii), CT: (i) *Ne ho il cofano _ rotto. (ii) Ne ho rotto il cofano _. ne I have the bonnet broken ne I have broken the bonnet However, if we consider a truly adjectival small clause, complement to a lexical verb, we find that the clitic ne can be extracted out of this position (from Chomsky & Lasnik 1991). (iii) Ne ritengo molti _ intelligenti. ne I consider many _ intelligent The contrast might follow if ne has been extracted from its basic position in (iii) and (ii) but from a derived position in (i). Consider, however, that this kind of evidence is questioned by Borer (1993; 1995). Moreover, the nature of the complex predicate certainly plays a role; we will return to the issue briefly in section 5. As for Swedish, the participial predicate can be preposed as already discussed in chapter 2. (iv) Han har sina skjortor strukna. he has his shirts ironed (v) Strukna tror jag inte att han har sina skjortor (möjligen tvättade). Ironed think I not that he has his shirts (possibly washed) If a Long Head Movement analysis is to be excluded, (v) is preposing of the participial XP, in which case the object must have left this XP at surface structure. Thus, even on empirical grounds can we claim that the object of PAP with all certainty is in a derived position.
123
like destroy and not with verbs of unaffectedness like see. If the verbal predicate is delimited, the derivation, at the point of Spell-Out, is (78), and (78) thus illustrates the first step of the derivation of PAP.
(78)
AspP
Spec DP(O)i
Asp’
VP
Asp [+del.]
ti
V
(78) is the crucial portion of structure. I assume (with Kayne 1993) that the surface subject DP (the possessor) is not generated as the external argument of V but higher up in the structure. Two questions arise immediately. Let us state them as (79): (79a) Is movement of DP(O) to Spec Asp forced for reasons of Case; in other words, is Spec Asp a position where accusative Case is checked on DP(O)? (79b) If affectedness boils down to the delimitedness feature, why is the Affectedness Constraint valid for PAP and not for the compound tense or any other construction? In order to sort out these issues, we need to consider the entire range of constructions that obey the Affectedness Constraint and, of course, those that do not obey it. I will begin with the first of these questions: is Spec Asp a Case position? As a first attempt, we could assume that Spec Asp is indeed a position where accusative is checked. This would come close to the approach of De Hoop (1992), as we are practically assuming that the delimitedness distinction is incorporated in Case Theory: we would assume a ‘delimited accusative Case’ quite reminiscent of De Hoop’s ‘resultative objective Case’. However, I will not follow this idea. It is clear from the data we have seen already that delimitedness effects are not confined to contexts where an argument carries and checks accusative. In order to make the account more general, we should follow another possibility. Consider that the delimited Asp head can be assumed for all contexts that contain a delimited predicate; above all, it must be assumed for the Affectedness Constructions. We can also suppose that an internal argument is externalized, in the sense that it is triggered to move through Spec Asp in all of the Affectedness Constructions.
124
Consider first the passive nominal: in (80), the DP the city is generated as the internal argument of the nominal predicate destruction. The DP moves in syntax to a position where it checks genitive. As above, we assume that [+ delimited] is generated under an Asp node, and that the DP the city moves through the specifier of this Asp node. As for the nature of the genitive checking head, it remains an open issue. For concreteness, I call it GenP, following Cornilescu (1993) and Longobardi (1995) among others.22 (80)
The Passive Nominal:
DP
D
GenP
Gen ’
Spec DP the cityi Gen ’s
AspP
Spec ti
Asp’
Asp [+del.]
NP
destruction t i by the Mongol s I will now make an additional assumption: suppose that the relevant Asp node is projected if and only if the predicate is delimited. In other words, if the predicate is non-delimited, there is no Asp feature to check. Consider the structure of an active nominal with a non-delimited predicate.
22
Recall that Kayne (1993) takes English genitive ‘-s’ to correspond to the AgrO node. The GenP could be equal to the PossP of some authors, cf. Valois (1991) and Delsing (1993). An AspP inside the DP is also assumed by Valois (1991, 128) in another context and with a role different from the one I am suggesting here.
125
(81)
The Active Nominal:
DP
D
GenP
Spec DP Johni
Gen’
Gen ’s
NP
ti avoidance of Bill Evidently, an unaffecting predicate such as avoidance can only appear in (81) and not in (80): avoidance is incompatible with the delimitedness feature. This idea will be further developed and explained below; for the moment we are only discussing the question raised in (79a). We have established that the delimitedness feature is not limited to accusative contexts but can also be involved in a derivation where DP moves to check genitive. If DP moves to check genitive, as it does in the nominal passive, the involvement of accusative would give rise to a Case conflict.23 In answer to the first question stated in (79a), I take the following view: the delimitedness feature, rather than being a Case feature, is ‘added’ during the derivation and can therefore combine, in principle, with any Case.24 This permits us to propose essentially the same analysis for the English Middle, where the affected DP carries nominative and moves to Spec AgrS to check it. I assume it also checks a [+del.] feature and I take it that
23
In other terms, the derivation would be ruled out because of a mismatch of features (Chomsky 1995, 308-309). It has been suggested in the literature that the same nominal expression, under certain circumstances, may carry two distinct Cases, cf. Belletti’s (1988) approach to Icelandic Quirky subjects and Sigurðsson (1992a) for discussion. Our issue is different, since we take the involved Cases to be structural, not inherent as Belletti’s partitive. 24 By saying this, we avoid a possible conflict with Chomsky’s (1995) version of minimalism: Chomsky assumes that Case is not interpreted in Logical Form and that Case features, which rule over overt syntactic movement, are therefore invisible after Spell-Out (Chomsky 1995, 277-279). If we had assumed that the aspectual interpretation was indeed encoded in Case, Case features ought to be visible at LF as information of event types presumably is. Instead, I assume that the [+del.] feature, which has content relevant for LF, is crucially distinct from Case itself. So far, we can therefore follow Chomsky by saying that Case features are indeed invisible at LF, whereas the Asp feature is visible.
126
[+del.] is checked during the derivation, in an Asp projection, which at this point must also be assumed in the finite clause. (82)
The English Middle:
IP
I’
Spec DPi I
AspP
Spec ti
Asp’
Asp [+del.]
VP
V
ti
If (82) is assumed, [+del.] is checked by the subject during the derivation. I choose to follow Stroik (1992, 1995) and Hoekstra & Roberts (1993) in assuming that the argument of the English Middle is generated as an object and syntactically promoted to subject position; I have assumed the same to be true for PAP, and for the passive nominal the issue looks fairly uncontroversial. Let us turn to the second question formulated in (79b): as mentioned, the proposal must also account for contexts where the Affectedness Constraint is not operative. The adjectival passive is one such context. The adjectival passive does not obey the Affectedness Constraint (cf. Hoekstra & Roberts 1993, 202); the argument of an adjectival passive may or may not be affected:
127
(83)
The problemi is [AP ti unresolved]. cf. These problemsi resolve ti easily … We had the problemi resolved ti …
(84)
The answeri is [AP ti unknown]. cf. *The answeri knows ti easily … *We have the answeri known ti …
This state of affairs has interesting implications for the present proposal, and I believe the adjectival passive gives an argument in favor of the idea that affectedness is related to externalization to a Spec Asp position. What we have observed in the affectedness constructions, is that the internal argument DP is syntactically promoted to the subject position of a small clause (Spec Asp): in PAP for instance, DP becomes the subject of the participial clause. I will tentatively define this ‘externalization internal to the small clause’, which will do as an informal label for the process. For the adjectival passive, it is commonly held (e. g. Levin & Rappaport 1986) that the passivized argument of AP is pre-syntactically promoted to the subject position of AP. As indicated in (83)/(84), the trace of the DP is not in the complement of the adjectival predicate. I take this to indicate that the first step of the derivation, the one that promotes the DP from object to subject position inside the small clause (movement to Spec Asp[+ del.], in other words), is the crucial one for affectedness. This conclusion is in line with our above line of reasoning. Let us make a first attempt to capture this generalization. It is schematically expressed in (85): (85)
Adjectival passive English Middle Nominal passive PAP have PAP få
Sensitivity to the Affectedness Constraint + + + +
Syntactic externalization of DP + + + +
We have thereby taken the first step toward the analysis we are searching for: there may indeed be a correlation between syntactic externalization (DP-movement of an internal argument) and affectedness. The correlation is captured on the hypothesis that externalization is driven by the delimitedness feature. However, the question of (79b) has not yet received a satisfactory answer and a lot remains to be said about (85). As we know, there is one construction that clearly defies the generalization expressed in (85), namely the verbal passive. In the verbal passive, an internal argument is promoted to subject position, checking nominative in Spec AgrS; the
128
Affectedness Constraint is evidently not valid, witness (86) and (87), which are equally grammatical. The argument of (87) is affected by the event expressed by the predicate, that of (86) is not: (86) (87)
Weird noises have been heard. The city has been destroyed.
(86) poses a problem for the analysis as developed so far, and we therefore need to sort out the facts so as to account for (86) also. In order to arrive at this, we must first enter the third problem relevant for the Affectedness Constraint: the organization of argument structure and the suppression of external arguments.
3.3
Affectedness and the Lexical Suppression of External Arguments
It has been observed that the requirement of affectedness is typical for contexts where an external argument has been suppressed (cf. Jaeggli 1986 and Guasti 1992). I will now assume that the external arguments of affectedness constructions are suppressed in the sense of being lexically suppressed, hence not syntactically represented, following the suggestion that the crucial difference between the verbal passive and the affectedness constructions lies in the projection of argument structure.25 I will go through the relevant evidence assuming the view of Rizzi (1986a), Roberts (1987), Giorgi & Longobardi (1991), Stroik (1992), Guasti (1992), and Hoekstra & Roberts (1993) among others: a syntactically active category can bind anaphors26 and control PRO into a purpose clause. These two tests for syntactic presence, or activity, of the external argument distinguish between the affectedness constructions, where the result is negative, and the verbal passive (together with the active nominal), where the external argument seems indeed to be present. Let us go through the relevant data: Consider first the verbal passive. Manzini (1983) and Chomsky (1981), note that the implicit argument of the verbal passive can control PROarb in the purpose clause:
25
The empirical issue is indeed intricate as the tests usually advanced as criteria for the syntactic absence or presence of arguments, based on binding and control, have been challenged. See Williams (1985, 1987) and Grimshaw (1990, 85) who both assume ‘event control’, and also Ackema & Schoorlemmer (1995), among others. Nevertheless, I will follow the authors cited in the text in assuming that these criteria are valid. 26 This argument is used also by Belletti (1990) in her treatment of Italian absolute small clauses. We will return to the matter in chapter 6.
129
(88)
The boat was sunk to PROarb collect the insurance.
In active nominals, control of PRO and binding of anaphors are possible without a genitive DP as (89), or with as in (90) ((89)-(91) from Giorgi & Longobardi 1991, 135; see also Williams 1985, Jaeggli 1986, 618, and Grimshaw 1990, 84): (89)
I disapprove of the destruction of the ship to PROarb collect the insurance.
(90)
I disapprove of their destruction of the ship to PROarb collect the insurance.
In the nominal passive the external argument is not syntactically present, witness (91): (91)
*the ship’s destruction to PROarb collect the insurance …
The agent DP that surfaces as genitive can of course bind an anaphor: (92)
John’s purchase of these drugs for himself …
Giorgi & Longobardi (1991, 138) report the difference between the active nominal in (93), where the anaphor is bound by an arbitrary agent, and the passive nominal in (94) where control does not obtain: (93) (94)
The purchase of these drugs for oneself is not prohibited in this state. *Their purchase for oneself is not prohibited in this state.
Hoekstra & Roberts (1993) reach the conclusion that the external argument in the Middle is syntactically inactive on the basis of the ungrammaticality of (95)/(96):27 (95) (96)
*Houses don’t paint easily for oneself. *In this country, bureaucrats bribe easily [PROarb to avoid paying taxes].
27
But see Stroik (1992, 1995) for a different view, and also Zribi-Hertz (1993) and Ackema & Schoorlemmer (1995) for discussion. Among the constructions under examination, the English middle is perhaps the one on which opinions diverge most. We will return to the matter further down.
130
Likewise, the external argument is syntactically inactive in PAP.28 Anaphors cannot be bound by the implicit agent of the participial predicate: (97a) (97b)
*Jag har brevet skrivet till en självarb. *Ho la lettera scritta a se stessiarb. I have the letter written to oneself
(98a) (98b)
*Vi har frukten lastad på ens egenarb lastbil. *Abbiamo la frutta caricata sul proprioarb camion. We have the fruit loaded on one’s own truck
It is impossible to obtain control of an adjunct PRO from the presumed arbitrary element inside the participial clause. In all of (99)-(102), PRO is obligatorily controlled by the surface subject, the possessor, and not by the implicit, arbitrary agent: (99)
Jagi måste ha/få breven skrivna ikväll Ii must have/get the letters written tonight [för att PRO*arb/i skicka dem imorgon]. [to PRO*arb/i send them to morrow]
(100)
Vii måste få nyheten bekräftad genast wei must get the news confirmed at once [för att PRO*arb/i kunna publicera den i morgon]. to PRO*arb/i be able to publish it tomorrow
(101)
proi Avremo la macchina riparata stasera Wei will have the car repaired tonight [per PRO*arb/i partire domani]. [to PRO*arb/i leave tomorrow]
(102)
proi Aveva la casa bruciata hei had his house burnt [per PRO*arb/i riscuotere l’assicurazione] [to PRO*arb/i collect the insurance]
28
The data rule out a suggestion, hypothesized for PAP by Christensen & Taraldsen (1989), namely that the external argument is realized as arbitrary PRO. In the light of (97)-(102) above, the idea is hardly viable.
131
Thus, the external argument is syntactically present in the verbal passive and in the active nominal. It is absent, hence suppressed, in all of the affectedness constructions, PAP included.
3.4
Aspect, the Extended Projection Principle and Syntactic Externalization of Arguments
I believe that the correlation between the Affectedness Constraint and the projection of argument structure can be derived on assumptions that have been fairly common in generative literature for the last fifteen years or so. I will suggest in the following pages that the Affectedness Constraint can be derived on mainly three assumptions. The first one remains from the discussion of the previous section: 1. The delimitedness feature triggers externalization of the internal argument to Spec Asp. 2. AspP carrying the delimitedness feature makes up part of the predicative XP-shell, where X can be V, N, or A. This means that an element in the specifier of Asp counts as external argument, or subject of VP, NP, and AP. 3. Predicate phrases, or shells, must have a subject. Let us begin with the last point. In Government and Binding Theory, the Extended Projection Principle of Chomsky (1982, 10) claims that the clause must have a subject; see also Stowell (1983). Suppose this requirement is valid also for small clauses, or predicates. Consider a condition such as (103):29 29
(103) does not have the status of a principle, but is derivative in a variety of frameworks. For instance, (103) is reminiscent of the Predication Condition of Rothstein (1995): «Every syntactic predicate must be syntactically saturated.» This requirement is based on the Principle of Full Interpretation: «A syntactic predicate is an open maximal projection that needs to be saturated by being linked to a syntactic argument, its subject.» (Rothstein 1995, 503). For Rothstein, all of the categories that concern us here - V, N, and A - can be syntactic predicates. For discussion on Chomsky’s EPP, Rothstein’s Predication Condition and Full Interpretation, see also Giorgi & Longobardi (1991, above all 15-16, 187-191). Not only for Rothstein, but for many recent approaches, the claim that predicates must have subjects is derivative. It follows from the Conditions on Aspectual Realization of Borer (1993) and Arad (1995, 222), that there must be one Spec Asp in the derivation. The view taken by Hale & Keyser (1993, 74-77) has the same effect: principles ruling over Lexical Relational Structure ensure that the Spec position of a VP must be filled. At the origins of this requirement are the Principles of Full Interpretation and Unambiguous Projection (binary branching); in a model that does not make use of the notion ‘thematic role’, the agent of a predication is the argument projected as Spec, asymmetrically c-commanding the inner argument (cf. also Hoekstra & Mulder 1990, 75, cited by Borer 1993, 25-26). If there is no higher, external argument c-commanding the inner argument, the latter could not be interpreted. Note finally, that the same conclusion can also be reached in Antisymmetry. Kayne’s assumption that «the highest element of a chain of heads must have a specifier, in the sense of having a phrase that asymmetrically c-commands it within its maximal projection» (Kayne 1994, 31), can be interpreted so as to yield the same result: a chain of lexical heads (including Asp) is created inside the VP-shell and requires a specifier in VP. Subsequently, a chain of functional heads is created external to the VP shell and requires a specifier in IP.
132
(103)
Every predicate must have a subject.
Now, there are two ways of giving VP (or any predicate XP) a subject, or an external argument: either by base generating an argument as external or by moving it to subject position in syntax. In the latter case, externalization of an internal argument to subject position inside the VP shell can be driven by the delimitedness feature. The hypothesis I am advancing can then be formulated as follows: when the external argument is suppressed, only those constructions where the internal argument is promoted to subject position through DP movement to Spec Asp[+del.] survive. Derivations involving non-delimited predicates crash. In order to arrive at this conclusion, we consider AspP a part of the VP shell, assuming that an element in Spec Asp counts as the external argument of VP. If this line of reasoning is pursued coherently, we reach the conclusion that a structure like (104) is ungrammatical; if DP moves directly to Spec I, the predicate does not have an external argument and the condition formulated as (103) is violated:
(104)
* IP
I’
Spec DPi I
VP
V
ti
The derivation is ‘saved’ through externalization of the DP inside the predicative XP, of which AspP is part:
133
(105)
IP
Spec
I’
I
AspP
Spec DPi
Asp’
Asp [+del.]
VP
V
ti
The conclusion is valid not only for VP but for any predicative phrase. We have derived that only a delimited argument can be syntactically promoted to subject position in VP, that is, only a delimited argument can move to Spec Asp[+del.]. Recall, moreover, that the VP shell assumed by Chomsky (1995) has as its highest projection a light verb, in the specifier of which the external argument is normally base generated. If such an argument is base generated in the Spec of V, the VP has a subject and respects (103) straightforwardly. In our line of reasoning, the light VP is not projected if the external argument is suppressed, and (103) is violated unless something else happens. If, then, a delimited Asp is projected, triggering object shift of the internal argument to Spec Asp, the VP receives a subject and (103) is respected. In brief, in order to satisfy the requirement that predicates have a subject, there must be an external argument which is either base generated (normally agent) or syntactically externalized (which means delimited, on our assumptions). In the following structures, I will simplify by omitting the light V of Chomsky (1995). Consider now the structure of a transitive, active V (106), as compared to PAP (107): (106)
Transitive V:
[VP [Spec DP(S) ] V DP(O) ]
(107)
PAP:
[VP [V’ V DP(O)]]
134
The transitive predicate of (106) has a subject and will satisfy the condition in (103), whereas (107) does not respect the condition. However, externalization is triggered by the delimitedness feature on Asp. Movement of the object to Spec Asp provides the predicate with a subject.30 (108)
PAP:
AspP
Spec DPi
Asp’
VP
Asp [+del.] V
ti
When the external argument is suppressed, the requirement (103) can be met through externalization of the object in AspP and only in this way. If the predicate of (107) is nondelimited or unaffecting, the derivation will crash. I suggest this is true also for the English Middle where the internal argument moves to check nominative in Spec Infl. By doing so, it gives the clause a subject and would satisfy Chomsky’s (1982) EPP, but does not satisfy (103) unless it first moves to Spec Asp, as indicated in (109):31
30
I will from now on simplify all trees, omitting X’ when there is no Spec, as in (108). Here we part ways with other approaches to the Middle, assuming that the internal argument is indeed projected as such (with Hoekstra & Roberts 1993 and contra Fagan 1988) but that the external argument is lexically suppressed (with Fagan 1988 and contra Hoekstra & Roberts 1993). Likewise, I agree with Ackema & Schoorlemmer (1995) that the external argument of the Middle is lexically suppressed, but disagree on their non-movement analysis. The arguments given by Ackema & Schoorlemmer in favour of an unergative analysis of middle verbs are based on Dutch. It is important to remember, I believe, that Dutch differs from English in allowing middles to be formed out of adjuncts (Hoekstra & Roberts 1993). This circumstance clearly shows that there is some fundamental difference between the languages with respect to affectedness.
31
135
(109)
the English Middle:
AspP
Asp’
Spec DPi
VP
Asp [+del.]
ti
V
The predicate referred to in (103) can evidently also be nominal (cf. Rothstein 1995, 503). We now assume, in a coherent fashion, that an external argument is projected in the active event nominal, following Giorgi & Longobardi (1991), Valois (1991), and Delsing (1993), among others.32 In DP there are two possibilities: (a) an argument of the nominal predicate is base generated as external, or (b) it is lexically suppressed, as before. The active nominal predicate is base generated as illustrated in (110): (110)
the Active Event Nominal:
NP
Spec DP
N’
N
PP
In (110) it is irrelevant if the predicate is delimited or not, that is, if AspP is projected or not.
32
Different views have been expressed with respect to the syntactic presence of the external argument in DP (cf. Grimshaw 1990). I set aside these as well as the discussion of possible differences between Germanic and Romance (cf. Giorgi & Longobardi 1991 and Bottari 1992).
136
The corresponding structure of the passive nominal will then be (111), with the external argument suppressed: (111)
the Nominal Passive:
Gen P
Gen ’
Spec
AspP
Gen
Spec
Asp’
NP
Asp [+del.]
N destruction
DP
the city
In this structure, the city moves to Spec Asp, and by doing so it provides the nominal predicate with a subject, by virtue of (103). This movement is triggered by the delimited Asp head, and the nominal passive thus enters under our general description. If the nominal predicate of (111) was non-delimited, the derivation would not converge. Consider now the adjectival passive. Following Levin & Rappaport (1986), the internal argument is lexically promoted to subject position, and the predicate AP is base generated as in (112):
137
(112)
the Adjectival Passive:
AP
Spec DP
A’
A (112) respects the requirement that the predicate has an external argument, and it is irrelevant if the predicate is delimited or not. The verbal passive contrasts with the others in that the external argument shows signs of being syntactically active, as we have seen. Suppose, then, that the external argument of the verbal passive is indeed syntactically projected, as frequently suggested (cf. Jaeggli 1986, Baker, Johnson & Roberts 1989, and Hoekstra & Roberts 1993). (113)
the Verbal Passive:
VP
Spec (DP)
V’
V
DP(O)
I leave open the question of what exactly happens with the external argument in syntax (it may be absorbed by some morpheme or perhaps, realized as pro, cf. Guasti 1992, Hoekstra & Roberts 1993). As the predicate already has a subject, (113) is well formed with regard to the condition (103); the externalization of DP(O) in the VP-shell becomes irrelevant. Consider also the difference between unergative and ergative predicates:
138
(114)
Unergative:
(115) Ergative:
VP Spec DP(S)
VP V’
V
V’
V
Compl DP(S)
I refer to Borer’s (1993, 1995) distinction between these two classes: ergatives are intrinsically delimited, and non-agentive, whereas unergatives are non-delimited, and agentive. As a result, unergatives satisfy (103) as their argument is base generated in the specifier of V. Ergatives, however, do not. In my model, this implies that the argument of an ergative predicate is externalized as above, through movement to Spec Asp. This conclusion is consistent with the fact that ergatives are predicates of delimitedness, while unergatives are not.
3.5
Conclusion
We have captured (a) the correlation between affectedness and DP-movement, and (b) the correlation between affectedness and the lexical suppression of external arguments. The Affectedness Constraint is derivative and deprived of theoretic status, as desired. It follows on principled grounds when an internal argument can and must be promoted to subject position, through movement to Spec Asp[+del.]. We have reached a unified account under the assumption that not only clauses but also predicates require the projection of an external argument, a subject, in the spirit of a lot of recent proposals in the literature. Predicates that are base generated with an external argument satisfy this condition as they stand; it is then irrelevant if the predicate is delimited or not. If the external argument is lexically suppressed (or not available at all, as in ergatives), only predicates expressing an event that affects the internal argument are admissible, because only in such structures is the internal argument externalized through movement to Spec Asp.
4
Some Additional Notes on Delimited Asp
In this section, I will briefly discuss how my hypothesis relates to some other ideas on argument structure, functional structure, and Case checking. Without giving a ready analysis
139
or any definite answers, I will simply point out that the present hypothesis is compatible or, with some additional assumptions, can be made compatible with some other views, even though this may not be clear at first glance. I will address three general questions that I find too important to be ignored: (a) how the hypothesis relates to Theta theory; (b) what the status of AspP is with respect to the functional-lexical distinction; (c) how the present hypothesis can be understood in an MPLT framework.
4.1
AspP and Theta Theory
I have already mentioned (footnote 11, introduction to this chapter) that some of the approaches that have inspired this work belong to a current of generative grammar that wishes to eliminate the theory of thematic roles, replacing it with aspectual rules of argument mapping (as Tenny 1987), or reducing it to general principles of branching and Full Interpretation (as Hale & Keyser 1993), or Asp triggered DP-movement (as Borer 1993, 1995) - for the exact proposals, the reader is referred to these works. Although my account of affectedness does not share all crucial assumptions of these studies, it is clearly in line with them and respects the general program of research as defined by these scholars. The question is whether my hypothesis is in principle incompatible with Theta Theory. The answer to this question is negative; what has been stated above is quite viable even in a framework that makes reference to thematic relations. Compare my analysis with the Affectedness Constraint as stated by Jaeggli (1986, 607): (116)
If a complement of X is unaffected, it is impossible to eliminate the external theta role of X.
This formulation of the Affectedness Constraint captures the fact that the effects of the Constraint are seen in contexts where an external argument has been ‘eliminated’, in the words of Jaeggli. Note that in my account, (116) is derivative; in a model where externalization is understood as DP-movement to Spec Asp[+del.], it follows that only the external argument of an affecting predicate can be suppressed, since only an affected argument can take its place, so to speak. Clearly, the idea that a predicate requires an external argument can be pursued regardless of whether we assume that this external argument carries a thematic label or not, and from this we can quite simply conclude that my hypothesis is in
140
principle independent of what view we assume on Theta roles. The above account does not crucially rely on the existence of thematic roles, nor does it exclude them.33
4.2
Delimited AspP and the Lexical-Functional Distinction
I have assumed that Spec Asp[+del.] is not a position where Case is checked (or assigned, if one wishes to diverge from the minimalist program on this point). The reason for this is, quite simply, that the Affectedness Constraint, and hence movement to Spec Asp[+del.], applies in constructions where the DP at the final point of derivation checks nominative (the middle), genitive (the nominal passive), or accusative (PAP). I assume that the DP in each of these cases is triggered to move higher than Spec Asp[+del.] for reasons of Case checking, to Spec AgrS (the middle) or Spec Gen (the nominal passive). I assume the same holds for PAP, namely that the checking site of the accusative-marked DP is the Spec of a functional projection higher than the delimited AspP and external to the VP shell (I will be explicit about the nature of this projection in chapter 4). Note that this is not a necessary conclusion. If we were to follow e.g. Borer (1993; 1995), we could say that Spec Asp is a position where accusative Case is optionally assigned (in Borer’s terms). There are independent circumstances determining whether the derivation converges or not: suppose, with Borer, that an independent principle requires that nominative is always assigned. If the delimited DP of the English middle receives accusative in Spec Asp, the derivation crashes since the same DP must also receive nominative. If the delimited DP of PAP does not receive accusative in Spec Asp, the derivation crashes since no other Case can be assigned to this DP, and so on and so forth. In order to assimilate this line of reasoning to MPLT, we would have to make reference to Case checking rather than Case assignment, but this is essentially a problem of reformulation. My hypothesis, or the central idea behind it, could be also restated in this way. Nevertheless, I have adopted a view closer to Travis (1995), by saying that the delimited Asp, which is the locus of Aktionsart or inherent, verbal aspect, makes up part of the VP-shell. The delimited AspP is ambiguous with regard to the lexical-functional distinction, if we say both that Spec Asp counts as the external argument of the predicate V, and that accusative Case is assigned or checked in the same position. Therefore, I avoid the second assumption. I take it that the delimited AspP is lexical, at least in the crucial sense ‘internal to VP’, and I stick to the assumption that accusative Case is checked in a functional portion of the structure, higher 33
My proposal is intuitively closely related to Jaeggli, who however develops his hypothesis in a theta theoretic format: «… if the external θ-role is missing in a particular structure, it would then become impossible to compute the internal θ-role. Thus, such a structure would be illegitimate because the internal thematic relation would not be well defined. It is only well defined when the external θ-role is also present. If this line of reasoning is correct, the Affectedness Constraint would simply be an empirical generalization following from a more articulated theory of thematic roles. It would not have the status of an independent principle of grammar.» (Jaeggli 1986, 608)
141
than and external to VP. Again, after a slight reformulation my proposal would be compatible with more than one view on this point as well.
4.3
The Status of EPP in MPLT
We have hypothesized that the claim that a predicate must have a subject may be derivative. In the Minimalist Program, EPP is indeed derivative but in a different sense. The full clause, IP, must have a subject because the features of T must be checked (Chomsky 1993, Bures 1993, Platzack 1994a). Now, there may also be an alternative way of deriving the idea that predicates too must have a Spec - (103), or something similar to it - which is closer in spirit to Chomsky (1993, 1995). Disregarding possible Agr nodes and the Comp field, the sentence structure we assume would now host two meaningful categories, T and Asp. If Chomsky’s EPP is derived on the assumption that T needs checking, the principle of (103) might follow from the need to check Asp. The above account would then boil down to a requirement that the heads hosting temporal and aspectual information, T and Asp, must be checked.34 We are therefore in a position to claim that the difference between this account and Chomsky’s (1995) version of minimalism with regard to the treatment of EPP can be overcome if some additional assumptions are made. I have shown in this section that the general ideas behind the affectedness hypothesis presented above are also viable under assumptions different from those made in the text and can be restated in different frameworks. I leave the theoretical discussion at this, and turn to the Romance causative which is the last empirical issue that I will address in this chapter.
5
The Romance Causative: Fare + Infinitive
Guasti (1992) argues that the Romance causative construction formed out of fare ‘make’ + infinitive is subject to the Affectedness Constraint. In the relevant construction, the external argument of the infinitival V is expressed as an agentive ‘by’-phrase.35
34
If such a view is taken, the light verb of Chomsky might correspond to a different Asp projection; cf. the Asp Originator of Borer (1993, 1995) and Arad (1995). A more thorough account must pursue such ideas much further than is possible here. 35 There exists another causative construction with different properties, where the agent is expressed as an oblique argument (Guasti 1992): (i) Ho fatto riparare la macchina a Gianni. I have made repair the car to John
142
(117)
Ho fatto riparare la macchina da Gianni. I have made repair the car by John
(117) is subject to the Affectedness Constraint; if, instead of repair, we choose a predicate of unaffectedness, the result is deviant ((117)-(118) from Guasti 1992). (118)
*Ho fatto vedere le foto da Gianni. I have made see the photos by John
In the causative fare + infinitive we face a more intricate puzzle, which I will be able to account for only partially. As for the suppression of the external argument, Guasti (1992) shows that the fare + infinitive indeed contrasts with the verbal passive in the predicted way: the external argument is syntactically present in the verbal passive and absent in the relevant causative construction. In the Italian verbal passive, anaphors can be bound and PRO can be controlled inside a purpose clause. Consider her Italian examples (Guasti 1992, 110-112): (119)
Questo genere di privilegi fu sempre riservato solo a se stessi. this kind of privilege was always kept only to oneself
(120)
Questo monumento è stato costruito dall’architetto Nervii this monument has been built by the architect Nervi per PROi ottenere appoggi politici. to obtain political support
The external argument is however not active in the + infinitive construction, witness (121) and (122) (Guasti, 110-112): (121)
*In questo spettacolo il regista fece flagellare se stessi (dagli attori). in this show the stage director made flagellate oneself (by the actors)
(122)
Il sindacoj ha fatto costruire il monumento dall’architetto Nervii the mayor has made build the monument by the architect Nervi, per PRO*i/j ottenere appoggi politici. to obtain political support
This observation is also crucial for Guasti’s analysis. In the spirit of Jaeggli (1986), Guasti emphasizes the correlation between the suppression of external arguments and the Affectedness Constraint. She proposes to explain why the external argument is suppressed in affectedness constructions with the assumption that the external argument in the passive is projected as pro, licensed by the participial ‘en’-morpheme (cf. also Hoekstra & Roberts
143
1993). In her view, affectedness constructions are passives where there is no ‘en’-morpheme that can license pro, that is, passive constructions formed out of predicates other than past participles; a finite V in the middle, a noun in the nominal passive, and an infinitive in the Romance causative. As a consequence, pro cannot be licensed in affectedness constructions, and it follows that the external argument cannot be syntactically projected. The generalization suggested by Guasti (1992) is the following: (123)
Verbal passive (English) Middle Nominal passive Causative fare + inf
Sensitivity to the Affectedness Constraint + + +
‘-en’ morpheme + -
The problem for this view when confronted with PAP is obvious. PAP obeys the effects of affectedness; the external argument is indeed suppressed, but in contrast with the other constructions, the predicate is participial. Following Guasti, the external argument of PAP could be projected as pro, licensed by the ‘en’-morpheme. This, however, does not correspond to the data attested in paragraph 3.3, where we saw that the external argument of PAP fails to control PRO or bind anaphors. My account has the advantage that it does not predict any particular link between affectedness and the morphologic make-up of the predicates. If we consider the Romance causative in my model, we are now forced to the assumption that the internal argument of the infinitive must be externalized in accordance with the condition assumed in (103). We should however pay attention to one circumstance: as opposed to previous cases, we are now dealing with a complex predicate.36 One difference in the derivation between the Romance causative and the other affectedness constructions that may be imputable to this fact is that the lower predicate itself, the infinitive, has raised to the higher one, fare, so as to appear to the left of, or higher than the affected object. We can assume that the infinitival V raises to the causative V through the intervening Asp head:
36
The extensive literature on complex predicates has been concerned with the question of a lexical or a syntactic derivation; see among others Rouveret & Vergnaud (1980), Manzini (1983), Rizzi (1986a), DiSciullo & Williams (1987), Gunnarson (1992), and Aarts (1992). A further problem, which cannot be addressed here, is of course how these accounts relate to minimalism and Antisymmetry.
144
(124)
V’ AspP
VÞ VÞ Infi
VÞ tj
Spec
Asp’
Asp [+del.] ti
VP
V ti
DP(O)
(124) is in essence the surface structure advanced by Guasti (1992) and Watanabe (1993),37 to which we add the Aspect phrase. [V tj] in (124) is the trace of fare that raises overtly to the matrix Infl. In the other affectedness constructions, DP(O) moves in overt syntax, and at this point, it would be desirable to draw the conclusion that it does so also in the causative since the causative, then, would conform to the general analysis. Judging from superficial word order, it is indeed possible that the object moves overtly in the causative. I will limit my discussion to two possible objections to the claim that the object raises overtly: adverbial distribution facts and extraction of ne apparently raise problems for an analysis in terms of overt object shift to Spec Asp. I will however show that the data do not offer counterevidence solid enough to motivate the general proposal being abandoned. 1. Belletti (1990) defines a category of adverbials in Italian that she calls ‘lower adverbs’. To this group belong spesso ‘often’ and completamente ‘completely’; see also Pollock (1989) and Guasti (1992). These are supposedly adjoined to the category that they modify. In both cases, they obviously modify the event expressed by the verb, and Belletti (1990), therefore, takes them to be adjoined to VP. For Belletti (1990), these adverbs are ‘on top of’ VP, but she does not consider the possibility of a richer structure inside VP. When we introduce Asp, assuming that it hosts information of event type and hypothesizing that it makes up part of the VP shell, lower adverbs would naturally be adjoined to AspP. If adjunction is not admitted, they would fill the specifier of a different AspP or some other functional projection:
37
(124) obviously does not follow Watanabe in assuming covert movement of the object to a matrix Spec AgrO; this question can be left entirely open for our purposes.
145
(125)
[VP fare [AspP spesso/completamente[AspP Asp [VP Infinitive DP]]]]
Clearly, the object can move to the specifier of Asp and still appear to the right of these adverbs, as indicated in (128) ((126)/(127) from Guasti): (126) (127)
(128)
Ho fatto riparare spesso la macchina … I have made repair often the car Hanno fatto restaurare completamente il teatro … They have made restore completely the theatre [AspP spesso/completamente[AspP DPi … [VP ti ]]]
Thus, (126)/(127) do not give clear indications of where the object is situated at Spell-Out. This line of reasoning conforms to Cinque’s (1994) account of adverbial distribution in the Italian participial clause. 2.
Extraction of ne is possible from the infinitival phrase:38
(129)
Di macchine ne ho fatte riparare molte _ (da Gianni). of cars ne I have made repair many _ (by John) Di compiti ne ho fatti fare molti _ (dagli alunni). of homework ne I have made make many _ (by the pupils)
(130)
Again, this does not force any particular conclusion with regard to the surface position of the object; recall from footnote 21 above that ne-extraction is possible from specifiers in a complex predicate like consider … intelligent (131), while impossible for PAP (132), for instance: (131)
Ne ritengo molti _ intelligenti. ne I consider many _ intelligent
(132)
*Ne ho il cofano _ rotto. ne I have the bonnet _ broken
38
These structures are sometimes given a question mark by my informants. I take this to be imputable to the fact that the fare + da … construction is stylistically marked when compared to fare + a … or at least less colloquial. By this I mean that those informants who judge (130) marginal, for instance, tend to do so also before ne has been extracted.
146
Even though the analysis of these contrasts remains somewhat obscure, I take the crucial point to be clear; ne-extraction facts do not give decisive counterevidence to the proposal that the object in the fare + infinitive has been overtly shifted. It appears that the delimitedness feature, when present, forces overt checking; the Romance causative does not defy this generalization. However, it may be necessary to assume parametrization with respect to the strength of the delimitedness feature. To be more precise, I now assume that the delimitedness feature triggers overt movement in the languages we are discussing here, not universally. There may of course be languages where the delimitedness feature is weak. In conclusion, although some aspects of the fare + infinitive construction must be left for future research, I hold that it enters under my general description of the Affectedness Constraint. As the entire range of constructions involved has not yet received a uniform account, the analysis I propose is theoretically warranted and should therefore be pursued.
147
CHAPTER 4 The Interaction Between Aspect and Agreement: Asp, AgrO, Greed, and Participial Agreement
0
Introduction
We have argued for the presence of an Agr head inside the participial clause on the basis of word order facts in some Romance and Germanic languages, and have seen that word order actually correlates also with the event type expressed by certain predicates. Furthermore, the Affectedness Constraint can be reduced to general mechanisms of feature checking through the introduction of a functional head that carries an aspectual feature to be checked off by affected nouns. The result is independent arguments for two distinct functional elements in the participial structure: one of them carrying agreement, and the other aspectual specification. I assume that the claim that participial clauses host multiple inflectional structure has sufficient empirical support, and I follow therefore a variety of scholars who assume a split of the participial Infl; see section 3.3. of the Introduction. The main issue of chapter three was an AspP locus for information of Aktionsart or verbal aspect. It is now my purpose to bring the analysis further, introducing a broader set of data. The fourth chapter is dedicated, first and foremost, to the checking of accusative in active constructions. I will continue to assume an AspP which is crucially involved in the checking of accusative along with AgrOP. The AspP under discussion at this point, has properties clearly different from the one proposed in chapter three, and I will therefore assume two Aspectual Phrases: the first one is linked to Aktionsart and makes up part of the VP shell (as I assumed in chapter 3); the second one, which I introduce now, is a purely functional projection, external to VP. If the first AspP corresponds to Aktionsart, the second might correspond to Aspect, according to a current distinction; see section 3.3.3. of the Introduction. However, this last idea will not be crucial for the discussion that follows. If we accept the participial Split-Infl hypothesis it will help us understand some well known Romance data concerning participial agreement in construction type 4 (see chapter 1), that is, in relative clauses. After some theoretical discussion that bears mainly on the Antisymmetry framework of Kayne (1994) and the most recent developments of the Minimalist Program put forward by Chomsky (1995), I will present the Romance data under section 2, and propose in section 3 what I consider to be a simplified account of these data. Again, the solution is closer in spirit to the Antisymmetry framework of Kayne (1994) than to Chomsky (1995). Crucially, I will continue to assume that there is no distinction between adjuncts and specifiers.
148
1
A Theoretical Issue: Antisymmetry and Bare Phrase Structure
Our discussion relates to Chomsky (1993, 1995) and Kayne (1989a, 1994) and regards mainly two theoretical issues: 1. In the Antisymmetry framework, the distinction between specifier and adjunct disappears, as only one segment of XP is admitted. In the Bare Phrase Structure approach advanced by Chomsky (1995), XP can have more than one specifier. The analysis outlined below does not make reference to the distinction between specifier and adjunct and presupposes that there can be only one specifier in XP. 2. In the minimalist model, there is a tension between the Minimal Link Condition (MLC) and Greed, which can be described as follows: if the DP is to move from Spec Z to Spec X in (1) below, MLC forces it to proceed by the shortest step, that is, through Spec Y: (1)
XP
Spec DPi
X’
X
YP
Spec ti
Y’
Y
ZP
Spec ti
Z’
Greed, on the other hand, requires that the DP raises in order to check its own features. If the DP has no feature to check in Spec Y, this movement would be unnecessary and would violate Greed, if this principle is to be understood in a strict fashion. The derivation would thus be (2), where DP takes one step from Spec Z to Spec X and the irrelevant Spec Y is not projected:
149
(2)
XP
Spec DPi
X’
X
YP
ZP
Y
Spec ti
Z’
Chomsky (1995) incorporates the notion of Greed in the definition of movement; ‘Move α’ is substituted by ‘Move F’. If movement of DP in (2) is triggered by ‘Move F’, and if F is checked in Spec X, DP cannot move to Spec Y.1 The second kind of derivation should then be pursued in a systematic way, assuming that Greed is privileged over MLC. If Greed/Move F do not require the projection of Spec Y, the correct derivation is (2) and not (1). Indeed, some works in the minimalist framework previous to Chomsky (1995), to which we will make reference below in 3.2.1., already point in this direction. The following demonstration will show that a set of Romance data are successfully explained under (2) rather than (1), confirming the correctness of this conclusion.
2
The Empirical Issue: Agreement in Operator Contexts
At least since Kayne (1989a), some efforts have been made to explain participial agreement phenomena in operator contexts, sometimes referred to as A’-agreement.2 Agreement in operator contexts refers to the French data in (3a)-(b) as opposed to Italian (4a)-(b) and Spanish (5a)-(b). In A’-dependencies, relative clauses (a-examples) and wh-questions (bexamples), the active participle of a compound tense shows agreement with the object in
1
This follows from the formulation of Last Resort: «Move F raises F to a target K only if F enters into a checking relation with a sublabel of K.» (Chomsky 1995, 280) 2 See also Branigan 1992, Watanabe 1993, Cocchi 1994, and Pettiward 1995. The problem of participial agreement is addressed by La Fauci (1989) within the framework of Relational Grammar.
150
French, but not in Italian and Spanish:3 (3a)
Fr:
(b)
Fr:
(4a)
It:
(b)
It:
(5a)
Sp:
(b)
Sp:
Les lettres que j’ai écrites … the letters[+fem. pl.] that I have written[+fem. pl.]… Combien de lettres as-tu écrites … how many letters[+fem. pl.] have you written[+fem. pl.] … Le lettere che ho scritto … the letters[+fem. pl.] that I have written[-Agr]… Quante lettere hai scritto … how many letters[+fem. pl.] have you written[-Agr] Las cartas que he escrito …[-Agr] the letters[+fem. pl.] that I have written[-Agr]… Quantas cartas has escrito …[-Agr] how many letters[+fem. pl.] have you written[-Agr]
The picture turns out to be more complex than this however, when other French data are considered. Take the following examples from Kayne (1989a, 90-91): (6)
Je me demande combien de couleurs il a déteint (*es) … I wonder how many colours it/there has run[-Agr] …
(7)
Il sera repeint(*es) beaucoup de chaises cette année. There will be repainted[-Agr] many chairs this year
(8)
Je me demande combien de chaises il sera repeint (*es) I wonder how many chairs there will be repainted[-Agr]
(6)-(8) are impersonal expletive constructions, (6) with have as auxiliary and (7)/(8) with ‘be’. Participial agreement is not possible in any of these examples. Furthermore, the participle of faire cannot agree in weather-expressions, nor can the participle of falloir in the expression il faut que ‘it is necessary’ (examples from Kayne 1989a, 90-91):
3
Branigan (1992) and Pettiward (1995) emphasize that agreement in French (3a) and (3b) is optional, and their analyses build on this assumption. It so happens that participles in (4a) and (4b) agree also in some Italian varieties as well as in Ancient Italian (see chapter 1 and section 5 below, this chapter), but participle agreement is not generally optional in Romance. Generally speaking, it is of course difficult to handle optionality in the present model. What we are treating here, rather than three ‘national languages’ are three abstract grammatical systems, and it might have been justified to call them G1, G2 and G3 instead of ‘French’, ‘Italian’ and ‘Spanish’, recognizing the obvious fact that these systems compete within one and the same geographic area. Still, for ease of exposition I find ‘French’, ‘Italian’ and ‘Spanish’ the most suitable labels.
151
(9a) (9b)
(10a) (10b)
Il a fait une chaleur atroce. it has made an atrocious heat Quelle chaleur atroce qu’il a fait (*e)! what an atrocious heat it has made Il me faut ces chaises. it is necessary-to me these chairs les chaises qu’il m’aurait fallu (*es) … the chairs that it would have been necessary-[-Agr]-to me
Consider also that the higher participle cannot agree in (11), reported by Branigan (1992) where the Wh-expression has moved from an embedded clause. (11)
*la lettre qu’il a dite que Claire lui a envoyée … the letter[fem.sing.] that he has said[fem.sing.] that Claire has sent[fem.sing.] to him
The ungrammaticality is due to agreement on dire. My informants accept agreement on the lower participle. There are other data to consider, but I will begin with those listed in (3)-(11) when formulating a hypothesis in the following section. We will return to empirical observations in 4.
2.1
A Note on the Analysis of Kayne (1989a)
Kayne (1989a) builds on the idea that participle agreement and finite verb agreement can be dealt with in a maximally unified way and reduced to the same fundamental mechanism. Crucially, the participle in (3a) and (b) does not agree ‘directly’ with the operator, but with an empty category adjoined to a participial AgrP (in quite the same way as the finite V under Infl agrees with a DP in Spec I). In Kayne’s terms, agreement is mediated through this empty category and the derivation of a French Wh-construction as (12), is shown in (13) (Kayne 1989a, 89 - Kayne uses IP instead of AgrP): (12)
Fr:
Combien de lettres as-tu écrites … ? how many letters[+fem. pl.] have you written[+fem. pl.] …
(13)
Whi Aux [AgrP [e]i [AgrP AGRi participle [e]i ]]
152
Kayne assumes that the empty category is adjoined to AgrP as it has to be in an A’-position. This in turn, follows from a principle that ensures that «if a Case-marked chain is headed by an A-position, then that A-position must be assigned Case» (Kayne 1989a, 89). In Kayne’s model, the intermediate position occupied by the highest [e] in (13) is not a position where Case is assigned, and it thereby follows that it must be A’-position. A trace in the relevant position is present in both (3a) and (b) in French. As for (6)-(8), (9b), and (10b) where agreement on the participle is not possible, although movement has applied, Kayne gives an explanation based on the assumption that expletives must be replaced by their associates in LF.4 The given structure (valid for all of (6)-(8), (9b), and (10b)) is shown in (14) (Kayne 1989a, 90): (14)
Whi il Aux [AgrP [e]i [AgrP AGRi participle [e]i ]]
In this structure, Kayne argues that the expletive should be replaced by the upper [e], adjoined to AgrP. This gives rise to improper movement as [e] would move from an A’-position to an A position. Therefore, the derivation has to take the form of (15): (15)
Whi il Aux [AgrP participle [e]i ]
In (15), [e] replaces the expletive in LF and there is no intermediate [e] adjoined to AgrP that triggers agreement on the participle.
3
Analysis
3.1
Aims
The solution of Kayne (1989a) is not straightforwardly viable under Antisymmetry (Kayne 1994), and in fact, is not needed. I will outline a different approach under the following sections. With Kayne, I assume that finite verb agreement and past participle agreement are indeed to be captured in a unified account, and that nominative Case checking and accusative Case checking generally should be analyzed in a maximally unified way. I intend to show that the analysis does not need to make reference to the distinction between adjuncts and specifiers. Under 3.2., I will begin by presenting the basic assumptions of my approach which draws on some developments of the minimalism program. First, I will briefly review what I call the nominative Case checking parameter in 3.2.1. Then, in 3.2.2., I will show how the discussion carries over to the checking of accusative Case, and thereby give a solution to 4
See Chomsky (1986b). In Chomsky (1993), the associate adjoins to the expletive in LF, and in Chomsky (1995, 272-276), the feature F on the associate raises covertly to Infl.
153
the general problem defined in section 2. Lastly, I will turn to exceptional cases and deviances and show how they fit in with my account.
3.2
Structural Case Checking and Split Infl
3.2.1
The Nominative Case Checking Parameter
In the GB model, nominative on the DP subject is assigned by Infl, present as one functional head. When, after Pollock (1989), Belletti (1990), Chomsky (1993) among others, Infl is divided in two or more heads, the question arises whether the head crucial for nominative assignment or checking is either one of AgrS or T or both. In the Minimalist Program as presented by Chomsky (1993), there is some uncertainty on this point. Nominative is checked against AgrS into which T has been incorporated, but it is unclear whether the DP subject passes through Spec T or not, and consequently, if Spec T is at all projected; for Branigan (1992) and Watanabe (1993), it is not. Even though Chomsky leaves the question open, we realize that the checking site can be regarded as a parameter. This is indeed a natural assumption since the checking process would now involve two features on the V, AgrS and T, but only one on the corresponding DP, nominative. In subsequent work in the minimalist frame this claim becomes explicit. For Jonas (1992), Bures (1993), and Jonas & Bobaljik (1993), the choice of checking site for the DP subject is parametrized. The effect of these proposals is quite simply, that when two or more features on the Verb correspond to the nominative feature on the subject, only one of these verbal features is crucial for the checking of nominative, and that the other is secondary for purposes of nominative checking. We can state this as the Nominative Case Checking Parameter: (16)
Nominative Case on the DP is checked against X.
X in (16) would normally have the value T or AgrS. Given our introductory discussion on Greed and Minimal Link Condition, where we held (with Chomsky 1995) that the former is privileged over the latter, the parameter in (16) now means that only one of the functional heads in question will have a specifier position taken by the subject, namely Spec X where nominative is checked. In a language where the crucial feature is AgrS and not T, nominative checking is illustrated in (17):
154
(17)
Nominative Case on the DP is checked against AgrS.
AgrSP
Spec DP(S)
AgrS’
AgrS
TP
T
(…)
In a language where T is crucial, we have the inverse situation: (18)
Nominative Case on the DP is checked against T.
AgrSP
AgrS
TP Spec DP(S)
T’
T
(…)
(18) illustrates the parametric choice Jonas (1992), Bures (1993), and Jonas & Bobaljik (1993) suggest for Icelandic. The Nominative Case Checking Parameter would then be stated as (19): (19) Language: English/Italian/French … Icelandic
Nom checked against: AgrS T
For Branigan (1992), nominative in English is checked in Spec AgrS, but the subject moves to the specifier of a higher position, ΠP. Watanabe (1993) maintains the view that the finite V generally raises to C in verb second languages. The idea implies that the nominative checking
155
head, X in (16), is C in V2 languages. Zwart (1993) takes a different stand point, assuming AgrS to be the nominative checking head generally, also for verb second grammars.5 All of these suggestions are variations on the idea that the locus of nominative checking is parametrically decided and that the relevant parameter is responsible for some cross-linguistic variation in word order patterns in the finite clause. Thus, the parameter of (16) expresses some assumptions that have been quite common in the literature for some years.
3.2.2
The Accusative Case Checking Parameter
Word order and functional structure have been the focus of attention in the generative literature for a long time. Whereas most of that attention has concentrated on issues concerning nominative checking in the finite clause (mainly relevant for word order patterns involving finite V and the subject), we are now mainly interested in the syntax of participles, and the checking of accusative (relevant for word order patterns involving the nonfinite V and the object). I will claim that the general conclusions about sentence structure of the finite clause and its correlation with nominative checking carry over to the syntax of nonfinite clauses and the checking of accusative. This is what we ought to do, if structural Case checking together with the agreement issue is indeed to be analyzed in a maximally unified way. More precisely, let us assume a split of the participial Infl in two functional heads, Asp and AgrO. Suppose the checking of accusative is parametrized in the same fashion as the checking of nominative, that is, that the head against which accusative is checked is either of AgrO or Asp, and that the choice between them varies cross-linguistically. This would mean in brief, that (20) is valid along with (16) above: (20)
Accusative on DP is checked against Y.
Suppose some of the crucial differences in participial agreement patterns between French on the one hand, and Italian and Spanish on the other, derive from this parametric setting. Let us assume Y is AgrO in French, and Asp in Italian and Spanish. Accusative checking in these Romance languages can then be summarized as in (21):
5
I will not enter into details of nominative checking in Germanic which are not crucial for my demonstration. The general idea that the locus of nominative checking is parametrically decided should be viable regardless of whether the principal feature against which nominative is checked is C, into which AgrS has been incorporated (in which case we have V-to-Comp) or some of T and AgrS (in which case the V stays in T or AgrS). For recent discussion on V2 and further references, see Vikner (1990), Zwart (1993, 1994), Gärtner & Steinbach (1994), and Holmberg & Platzack (1995).
156
(21) Language: French Italian Spanish
Acc checked against: AgrO Asp Asp
The consequences of this for our data should now be obvious. Consider again (3)-(4) restated as (22)-(23). As Spanish and Italian are equal on this point, Italian examples will represent them both: (22a)
Fr:
(22b)
Fr:
(23a)
It:
(23b)
It:
Les lettres que j’ai écrites … the letters[+fem. pl.] that I have written[+fem. pl.]… Combien de lettres as-tu écrites … how many letters[+fem. pl.] have you written[+fem. pl.] … Le lettere che ho scritto … the letters[+fem. pl.] that I have written[-Agr]… Quante lettere hai scritto … how many letters[+fem. pl.] have you written[-Agr]
The a-examples and the b-examples of (22)-(23) are both operator-variable dependencies. Variables are Case marked and the accusative Case feature must be checked. In (22)-(23), checking of accusative is overt and it is so for independent reasons. This is my first assumption: accusative is checked overtly in (22)-(23), not because the Case feature is strong and requires overt checking, but rather vice versa: accusative must be checked overtly because operator-variable dependencies are created overtly. It follows from (21), that the checking of accusative Case in French requires projection of the Spec AgrO, and in Spec Asp in Spanish and Italian. The Spec Asp in French does not need to be projected, and more importantly, the Spec AgrO in Spanish and Italian needs not be present since there is no checking process taking place in the projection of AgrO in these languages. Greed requires the projection of Spec AgrO and not of Spec Asp in French, and of Spec Asp and not of Spec AgrO in Italian and Spanish. The Minimal Link Condition would require the projection of both positions in all languages, and the difference would not follow. The crucial portion of structure is given in (24)/(25). For concreteness, I now assume Belletti’s (1990) and Uriagereka’s (1995) ordering of Asp and AgrO. The conclusion would of course hold even if the ordering were the opposite.
157
French: (24a) (24b)
Les lettresi Opi que j’ai [AgrOP ti [AgrO’ écrites ti ]] (=22a) Combien de lettresi as-tu [AgrOP ti [AgrO’ écrites ti]]? (=22b)
(24c )
AgrOP
AgrO’
Spec ti
AspP
AgrO
VP
Asp
Italian and Spanish: (25a) Le letterei Opi che ho [AspP ti [Asp’ scritto ti … ]] (25b) Quante letterei hai [AspP ti [Asp’ scritto ti ]]
(25c )
(=23a) (=23b)
AgrOP
AgrO
AspP
Spec ti
Asp’
Asp
VP
(25c) corresponds to Italian (23a)/(23b), and to the Spanish examples (5a) and (5b) above. This is a straightforward way to derive participial agreement in A’-contexts in French, and its absence in Italian and Spanish with respect to Antisymmetry. It follows from general assumptions on structural Case checking adopted ultimately from the Minimalist approach to nominative checking. Note that the crucial part of the derivation, the one that ‘triggers’ agreement on the participle, is not A’-movement to an adjoined position but A-movement to the specifier where Case is checked. On this point my account diverges from Kayne (1989a),
158
as we have already seen, but is equal to Branigan (1992).6 There is another remark to be made: in the spirit of Chomsky (1995, 276-279), although we do not follow his general approach, we might say that the content of some but not all functional heads is interpretable at LF. In order to be coherent, we ought to assume that the content of Asp and T is always checked, since Asp and T are meaningful elements. Hence, they are visible in LF, whereas AgrO and AgrS are not. In a language like French where Acc is checked in Spec AgrO, the content of Asp must be checked independently and we may assume it is so as Asp raises to AgrO. Similarly, it could be argued that in those languages where nominative is checked against AgrS, T must be checked for independent reasons because T raises to AgrS. Thus, there is a perfect parallel between accusative checking and nominative checking on this point as well. Returning to the data of section 2, on this account we can also give a natural explanation to those contexts where participial agreement in French does not obtain. Consider again the examples (6)-(8), (9b) and (10b), here restated as (26)-(30): (26)
Je me demande combien de couleurs il a déteint (*es) … I wonder how many colours it/there has run[-Agr] …
(27)
Il sera repeint(*es) beaucoup de chaises cette année. There will be repainted[-Agr] many chairs this year
(28)
Je me demande combien de chaises il sera repeint (*es) I wonder how many chairs there will be repainted[-Agr]
(29)
Quelle chaleur atroce qu’il a fait (*e)! what an atrocious heat it has made
(30)
les chaises qu’il m’aurait fallu (*es) … the chairs that it would have been necessary-[-Agr]-to me
Respecting Antisymmetry, (26)-(28) can be excluded on either one of two grounds, and (29)/(30) on the latter of these two: 1. In the view taken by Chomsky (1986b) and shared by Kayne (1989a) among others, expletive constructions as those in (26)-(28) crucially involve assignment of nominative (hence, checking of nominative in the present model). If there is no accusative to be checked on the associate, the projection of Spec AgrO is not required, hence blocked. 6
Moreover, the structures given in (24c) and (25c) are not conform to Kayne’s (1994, chapter 8) analysis of relative clauses. There is no problem here, though, since my account can be restated in terms of the Antisymmetric approach. I prefer illustrating my idea with the structures in (24c) and (25c) as they presumably are more familiar to the reader.
159
2. Chomsky (1993), and subsequent work, takes traces of a DP to be copies of the DP. As expletive constructions host two phonetic elements, the expletive and the associate, there is no clear reason as to why eventual traces should copy the φ-features of the associate. As a matter of fact, there is empirical evidence suggesting that traces in expletive constructions may be copies of the expletive. Consider a Swedish example: (31)
Sw: Många brev blev skrivna. many letters became written[+plural]
The verb bli is close in meaning and syntax to English become. The DP many letters has moved to a position where it checks nominative, and has triggered agreement on the participle. Recall from chapter two above that Swedish makes a distinction between active and passive participles in the following sense: passive participles agree in gender and number and active participles do not show agreement. The non-agreeing form of the participle is morphologically distinct from singular neuter.7 In (32)/(33), the former is realized as it and the latter as et. (32)
Jag har skrivit brevet. I have written[-Agr] the letter
(33)
Brevet är skrivet. The letter[neuter, singular] is written[neuter, singular]
Romance participles make no distinction between the non-agreeing form and the masculine singular. Swedish participles have a distinction between the non-agreeing form and the neuter singular, and this makes it possible to check our prediction on Swedish data. If (31) is expressed as an impersonal, the result is (34): (34)
Det blev skrivet många brev. There was written[neuter, singular] many letters
As can be seen above, the participle in (34) agrees but not with the noun left in situ. Neuter singular is instead the specification of the Swedish expletive pronoun det, and we assume that the participle in (34) agrees with the expletive pronoun: (35)
7
Deti blev [ti skrivet många brevi].
I continue to call active participle what is often referred to as the supinum.
160
The expletive chain consists of three elements, [Det … t … DP], and the intermediate trace carries φ-features of the expletive det. The same holds for (36), where the associate has been wh-moved: (36)
Hur många brev blev det skrivet? How many letters became there written[neuter, singular]
(36) contrasts with (37): (37)
*?Hur många brev blev det skrivna? How many letters became there written[plural]
This shows that the trace also carries φ-features of the expletive when the wh-expressions has moved overtly. We will return to the discussion of the passive below. Even (38) presented above as (11), follows straightforwardly: (38)
*la lettre qu’il a dite que Claire lui a envoyée … the letter[fem.sing.] that he has said[fem.sing.] that Claire has sent[fem.sing.] to him
This is to be expected if la lettre checks Case against the lower participle. As there is nothing to check in the specifier of Agr corresponding to dire, this step of the derivation is blocked by Greed.8 3.3
Summary
To summarize so far, the general hypothesis holds until now. The agreement data follow on the assumption of an accusative Case checking parameter that has different settings for French on the one hand and Italian and Spanish on the other. There has been no need so far, to resort to the distinction between adjuncts and specifiers, or to the notion of improper movement. I take this to be a step forward. The agreement problem is of course wider; there are differences between these and other types of structures and some interesting cross linguistic variation. The account must therefore be extended essentially in three ways: 1. We need to show how the present assumptions can handle the agreement problem in other contexts. First, we will consider some data from French ECM constructions, 4.1, and then we will go through the cases of agreement with clitics 4.2, and agreement in the passive, 4.3. 2. In section 5, we will turn to the comparison between Modern Romance languages and Medieval and Renaissance Italian that we touched in chapters 1 and 2. 8
For discussion on this case, see also Chomsky (1995, 325).
161
3. Lastly, we will see that generally the proposed analysis can also handle agreement patterns of the Germanic languages. This will be shown in section 6.
4
Extensions to Other Contexts
4.1
Agreement Patterns in ECM Contexts
There is a different set of data remaining to be explained. These can be summarized in three points: 1. Agreement on participial faire ‘make’ is impossible in causative constructions (Guéron & Hoekstra 1988, 69, and Branigan 1992, 34): (39)
*la femme qu’il a faite partir … the woman that he has made[fem. sing.] leave
(40)
la femme qu’il a fait partir … the woman that he has made[-Agr] leave
2. Agreement on the participle is impossible in (41), compare this with (42) where ‘dire ‘say’ takes an infinitival complement (Ruwet 1982, Kayne 1989a, 92). (41)
*Une femme qu’on a dite ne pas être belle … a woman that one has said[+fem. sing.] not to be beautiful
(42)
Une femme qu’on a dit ne pas être belle … a woman that one has said[-Agr] not to be beautiful
3. Agreement is optional in (43) and (44) where dire takes an adjectival clause as its complement (Kayne 1989a, 92). (43)
Une femme qu’on a dite belle … a woman that one has said[+fem. sing.] beautiful
(44)
Une femme qu’on a dit belle … a woman that one has said[-Agr] beautiful
The explanation of these contrasts depends on what analysis we assume for Exceptional Case Marking and complex predicates. Since the participles fait and dit do not agree in (40), (42), and (44), the relative operators in (40), (42) and (44) have never reached the Spec AgrO
162
corresponding to fait and dit, but have been extracted from some lower position. This observation is perfectly coherent with the claim of Junker & Martineau (1992), based on accusative-with-infinitve constructions in historical French syntax. Junker & Martineau argue that, in accusative-with-infinitive constructions, Case assignment takes place inside the infinitival clause. I suppose they are correct in this claim and that the general conclusion holds for both faire … partir in (39)/(40) and dire … être… in (41)/(42). Now, as accusative must be overtly checked on the dependency, the relative operator moves into the Spec of an abstract head inside the infinitival clause and from there up to the matrix Spec C. The projection of the specifiers of fait and dit is not required, hence blocked.9 The somewhat puzzling state of affairs illustrated in (43)/(44) is still unexplained. Why is it the participle in (43)/(44) agrees optionally? On this point, I will limit my analysis to a conjecture. The difference between the agreeing (43) and the non-agreeing (44) variant may be the following: (44) is a case of ellipsis, that is, equal to (42) with the omission of être. In such a case, the participial V cannot agree as the wh-expression must stop in the infinitival clause (headed by a phonetically null être) and is extracted directly from there to Spec C as above. (43) on the contrary is interpreted as AP without any infinitival head in which accusative can be checked. Then, accusative is checked in Spec AgrO corresponding to dire and triggers participial agreement. My suggestion is that the case of optionality in (43)/(44) is only apparent: (44) is an ellipsis structurally equivalent to (42), and accusative is checked in the infinitival clause without involving the matrix AgrOP. (43) is not an ellipsis, the object DP cannot be Case checked against the Adjective but moves into the matrix Spec AgrO. My general analysis of participial agreement in French carries over. The choice between (43) and (44) would then be apparently optional in the sense that the participle of dire must agree in the first case (43), whereas in the second (44) it cannot.10 Optionality is itself a problem, as I mentioned in footnote 3. Branigan (1992) and Pettiward (1995) build their accounts of participle agreement in the relative clause on the observation that participle agreement is generally optional in Modern French. However, the choice between different approaches is ultimately a choice of explanatory adequacy and theoretical simplicity. Participial agreement in the relative is not generally optional in Romance, and it is therefore unclear how the explanation of Branigan (1992) and Pettiward (1995) for French carries over to Italian or other Romance varieties. From the empirical view point, the accusative Case checking parameter has the advantage of capturing systematic 9
This hypothesis is obviously in conflict with the minimalist account for Exceptional Case Marking. In the minimalist frame, ECM is analyzed as covert raising of the accusative marked DP to a Spec AgrO dominating the Exceptional Case Marker (cf. Chomsky 1993 and Branigan 1992). Instead, I am of the opinion that the idea of Junker & Martineau (1992) is intuitively appealing and I believe that further investigation along these lines is warranted. See also chapter 6, section 2.4.1 footnote 14. 10 On these assumptions, the data follow and I have no other suggestions for this case. Some doubts remain, but I will not dedicate any discussion to them here. Above all, the Case assigning/checking properties of adjectives must be clarified, so as to allow us to define in what contexts A may and may not assign accusative. In chapter 8, I will suggest that A has the property of assigning/checking accusative though in a context quite different form the present one.
163
differences between the grammatical systems that otherwise fail to be captured.11 At the same time, the Case checking parameter makes accusative checking perfectly parallel to nominative checking.
4.2
Participial Agreement with Clitic Pronouns
Consider first pronominal agreement. It is a well known fact that participles agree with clitic pronouns in French and Italian: (45) (46)
Je les ai écrites. I them[cl.] have written[fem. pl.] Le ho scritte. I them[cl.] have written[fem. pl.]
However, they do not in Spanish: (47)
Las he escrito. I them[cl.] have written[-Agr]
Previous discussion brings us now to saying that agreement on the participle is spelled out because it is required in French and Italian, but not in Spanish. We must assume, for coherency, that checking processes require its presence in the two former grammars, but not in the latter one. The clitic pronoun, being a D° element, is commonly assumed to adjoin to the functional X° hosting the finite Aux. In the model I am developing, this will have to mean that the clitic needs to check some feature against the relevant functional head. Suppose that, when the clitic pronoun corresponds to a direct object, it also needs to check accusative, and that this is distinct from the checking process that takes place in the functional head to which the pronominal cliticizes. If pronouns as well as DPs are subject to a Case parameter as the one proposed in (20), the difference may follow, if the parameter can have different settings for pronouns and DPs: (48)
Accusative on D° (clitic pronouns) is checked against Z.
We then posit that the locus for accusative checking on pronouns is different from that of DPs as described in (49): 11
Furthermore, the accusative Case checking parameter obviates the need for assuming that traces move; see Pettiward (1995) and also Kayne (1989a), cited in the text. I follow Chomsky (1995, 304) in assuming that the trace is immobile.
164
(49) Language: French: Italian: Spanish:
On DP: AgrO Asp Asp
Acc checked against: On D° (clitics): AgrO AgrO Asp
The suggested parameter is quite in line with the intuitions behind some previous approaches to pronominal syntax. That clitic pronouns in Romance adjoin to different functional heads in the infinitival clause was argued for by Kayne (1991); in present terms, this means that they are triggered to check off features by different functional heads. Consider also that there may be independent reasons to believe that the checking sites for pronouns and DPs are also different in languages that do not have clitic pronouns. In Scandinavian, it is possible to raise (shift or scramble) both weak pronouns (generally), and DP objects (in Icelandic). But, the landing sites for these elements are presumably different. DP objects in Icelandic are shifted to a position that Déprez (1989) and Jonas & Bobaljik (1993) take to be Spec AgrO, whereas weak pronouns (I suppose in all Scandinavian) can go higher up in the structure; see above all Holmberg (1986), Déprez (1989), Josefsson (1992, 1993), and Holmberg & Platzack (1995). I take it then, that clitic pronouns check accusative against AgrO in Italian differently from DP objects, but that the checking of accusative in French is always against AgrO for clitics and DPs alike, and that the head Asp has the same two functions in Spanish. What happens more precisely in the clitic construction will now depend on how the clitic moves to its Spell-Out position. There are two options: (a) the clitic moves Spec-to-Spec until the final point of derivation when it adjoins to the Infl head; (b) the clitic moves head-to-head. Among the numerous scholars who have dealt with, or commented on the problem, I refer to Kayne (1989a, 1989b, 1991, 1994), Roberts (1991), Branigan (1992), Cardinaletti (1994a, 1995), Chomsky (1995).12 If we assume Spec-to-Spec-movement, the specifier of AgrO will be projected in French and Italian but not in Spanish. If we assume head-to-head-movement, the head AgrO will be projected in French and Italian but not in Spanish.13
12
Recall that we assume Spec-head agreement and head-head agreement. Trivially, if a head raises to another head there must be an agreement relation between them in the sense that they must be coindexed. Therefore, it is perfectly natural to assume that the clitic could trigger the spell out of agreement on the participle even if it were to raise through the Agr head and not the Spec Agr.
13
Cocchi (1994) has argued that participle agreement patterns in Spanish are due to the absence of AgrO in this language.
165
4.3
Agreement in the Passive
Let us now turn to the case of passive. Participles agree in all of the languages under discussion: (50)
Les lettres sont écrites. the letters[fem.pl.] are written[fem.pl.]
(51)
Le lettere sono scritte. the letters[fem.pl.] are written[fem.pl.]
(52)
Las cartas estan escritas. the letters[fem.pl.] are written[fem.pl.]
A passivized DP moves to check nominative against the finite Infl. The overt realization of agreement on the participle is obviously not linked to the checking of accusative. In the model I propose, it does not have to be, since the category of AgrO may or may not be involved in accusative checking, hence, checking of φ-features is not necessarily equal to the checking of Case. Suppose that when the DP direct argument moves to check nominative as a result of passive formation, there is a second requirement that has to be met, namely that the predicate must check that it is properly matched or paired with the argument. Then, a passive construction involves two checking processes that are equally forced, namely Case checking in the finite IP and agreement checking in AgrOP. (53)
[InflP DPj Aux … [AgrOP tj [AgrO’ Participlei [AspP ti [VP ti tj ]]]]]
This is the derivation I suggest for all three of the Romance languages taken into consideration. Something more needs to be said about this derivation however. There are alternatives that may be taken into account. One of them is to say that the Agreement head in the participial passive clause is not of the AgrO kind but rather AgrS. This may appear to be a question of labels - especially if Chomsky (1993, 1995) is right in claiming that ‘O’ and ‘S’ do not have theoretical status but are only mnemonics. It has been argued, on the other hand, that these functional categories, AgrS and AgrO, do indeed have distinct properties (cf. Rouveret 1991a and Friedmann & Siloni 1993), and the two alternatives would actually give rise to different predictions in our case. Consider that (54) and (55) are both ungrammatical: (54)
*Nous avons achetés le livre. we have bought[masc.pl.] the book
166
(55)
*Nous avons dansés toute la nuit. we have danced[masc.pl.] all night
It is now in line with our general way of reasoning to assume that the surface subjects of (54) and (55) must check that they are properly matched or paired with the participial predicates. If we suppose that participial agreement morphology corresponds to AgrO as we have above, the ungrammaticality of (54) would follow on the assumption that AgrO cannot be taken by the subject as it must be kept free to be taken by the object. Kayne (1993) and Cocchi (1994) explain (55) on the same ground, by saying that unergative predicates subcategorize for a null object; see also Hale & Keyser (1993) and Chomsky (1995), among others. Spec AgrO must be left free for this null object in (55). If, on the contrary, we were to assume that participial agreement morphology corresponds to AgrS, this explanation is obviously not viable. I therefore maintain that AgrO is the correct label. A further remark that has to be made on proper pairing is that if we pursue the idea coherently, we will have to assume that pairing is checked against functional heads that do not have overt morphological realization.14 I will return to this idea in section 6, when the comparison with Germanic is brought up. Summing up so far, we have seen that the agreement data of all of the relevant contexts lend themselves to a description in terms of the suggested parameters (20) and (48). We will now show that what we have said also covers agreement patterns observed in the ancient Italian texts.
5
Medieval and Renaissance Italian
When we turn to Italian texts from previous centuries, it is quite obvious that what we are facing is not unitary grammars, but the intersection and mixture of many grammatical systems. Dialectal and stylistic factors conspire to make the picture more complex and make it difficult to arrive at a correct generalization. As before, I limit my discussion to central Tuscan in two crucial centuries: the 14th (that I call Medieval, MI) and the 16th (Renaissance, RI). Recall that a parametric difference between Medieval and Renaissance Italian regarding the strength of the accusative Case feature was already assumed in chapter two. Accusative is strong in Medieval Italian, as we assume it is in Modern German and Dutch (cf. Zwart 1993), whereas it is weak in Renaissance Italian as in Modern Romance generally. The first parameter makes the following division between languages: 14
The latter proposal is identical to Kayne (1993). Kayne argues that the effects of abstract AgrS and T elements inside the participial clause are visible in some Romance varieties as it inflects upon the choice of Auxiliary.
167
(56) The Accusative strength parameter: Acc strong: Medieval Italian Acc weak: Renaissance Italian, Contemporary Italian (and Romance) This was assumed in order to explain the now familiar pattern. In the 14th century, word order object-participle (construction type 2; see chapter 1) is attested with some frequency: 14th century, MI: (57) Io ho, vivendo, tante ingiurie fatte a Domenedio… I have, living, many insults[fem. pl.] made[fem. pl.] to God (Dec; I:1) Already by the middle of the 14th century, this word order becomes more infrequent and is by and large absent in the 16th century (with exception for Bembo, see section 3.2.1, chapter 1). The assumption illustrated in (56), that MI had strong accusative and RI/CI weak accusative, is justified by this observation. Our task at present is to establish how the Case checking parameter proposed in (20) and (48) is set for the relevant varieties of Italian. We will do so under the following two paragraphs, beginning with Agreement in operator contexts.
5.1
Agreement in Operator Contexts
The view on agreement in operator contexts has been modified. It has nothing to do with the A-A’-distinction as such, but is imputable only to the fact that a dependency is overtly created. Along the lines already suggested by Branigan (1992), I assume that movement of a DP to Spec AgrO is A-movement. I continue to talk about agreement in operator contexts for ease of exposition. The term does not imply, as I have already explained, that agreement is triggered on the participle by a DP (or its trace) in an A’ position. In Medieval Italian, participles agree with great regularity in all contexts including relative clauses: 14th century, MI: (58) … che tutte quelle ch’io ho trovate … that all those[fem. pl.] that I have found[fem. pl.] (Nov.no; LXX)
168
(59)
… le rimate parole ch’io avea dette innanzi… the words[fem. pl.] on rhyme that I had said before[fem. pl.] (VN; XXXIX)
The second parametric choice that distinguishes between the systems under discussion regards the locus of accusative Case checking: in AgrOP or in AspP. As I have already mentioned, agreement is regular in MI and by this I mean that transitive participles always agree with their objects.15 If this is stated in terms of the Case checking parameter suggested in section 3.2.2., it will have to mean that the functional head against which accusative is checked in MI is AgrO. We arrive at the following subdivision: (60) Acc strong: Acc weak:
Acc checked in AgrOP Medieval Italian French,
Acc checked in AspP
Contemporary Italian, Spanish
In Renaissance Italian (or, to be more precise, from the mid 14th onwards) we find a lot of variation on this point. In certain authors, it seems clear though, that when participles generally tend not to agree in the compound tense, they continue to do so in relative clauses. RI, 16th: (61)
… che io non avessi udito le pazzie che egli ha dette… that I had not heard[-Agr] the nonsense[fem. pl.] that he has said[fem. pl.] (Mand; III:7)
The usage varies considerably between authors and stylistic levels. Generally, the Renaissance grammar does not have rightward agreement or the word order OV in participials. Some authors maintain however agreement in operator contexts and the Renaissance Italian emerging from those texts is clearly similar to Modern French. In other texts we find a use which is very close to Contemporary Italian where participles do not agree in relative clauses. The texts from the mid 14th century onwards, all those we have arbitrarily called RI, can therefore be divided in two groups, say RI1 and RI2 where the former maintains operator agreement and the latter does not.
15
We have already drawn the conclusion that rightward agreement in the most ancient stage of Italian crucially depended on the possibility of word order OV. The pro-strategy proposed in chapter two will not be discussed here, but we will return to it in chapters six, seven, and eight.
169
(62) RI1 Agreement in operator contexts: Ser Giovanni (Pecorone) Morelli (Ricordi) Machiavelli (Principe) Aretino (Lettere) Doni (Zucca)
RI2 Non-agreement in operator contexts: Sercambi (Novelliere) Guicciardini (Dialogo) Cellini (Vita) ?
For others, the material is either too small or the variation too large to admit for any interesting generalizations. This goes for Sacchetti, Castiglione, Firenzuola, and Grazzini. We capture the RI system in the present model by saying that RI1 and RI2 differ on the checking parameter: (63)
Acc strong: Acc weak:
Acc checked in AgrOP: Medieval Italian RI1, French
Acc checked in AspP:
RI2, Contemporary Italian, Spanish
Finally, the syntax of Bembo (see chapter 1, sections 3.1b, 3.2b, and 3.3b) is clearly different from both RI1 and RI2. It would be pointless, however, to introduce a third category, RI3, as Bembo’s prose on each of the relevant points is equal to that of the early 14th century: participles agree in all contexts and word order Aux-Obj.-part. is frequent. In this sense, leaving aside other aspects of his style, Bembo differs from other RI authors by simply using the MI grammar. For Bembo, accusative is checked in Spec AgrO and the accusative feature is strong.
5.2
Pronominal agreement
As far as I can tell, there is no substantial difference between the three stages of Italian on this point. Participles agree regularly with clitic pronouns in all three stages: MI, 14th century: (64) …ci ha fatte fuggire… they us[cl.]-have made[fem.pl.] flee (Dec; I: Intro)
170
(65)
…tu gli hai qui potuti vedere… you them[cl.]-have here could[masc. pl.] see (Dec; I: 2)
RI, 16th century: (66) … che l’abbia meglio governata … that it[cl.]-he had better governed[+Agr] (Dial; I) (67)
… ci arebbe di necessità tenuti uniti … us[cl.]-he would have for necessity held[+Agr] united[+Agr] (Dial; I)
(68)
… se la cosa sta come voi me l’avete scritta … if the things are as you me[cl.]-it[cl.]-have written[+Agr] (LettM; V)
(69)
… senza avermi adornata … without having-me[cl.] adorned[+Agr] (RagF; I: Intro)
Some cases of non-agreement are attested, although mainly in RI. These tend to involve clitics of first or second person, especially plural, or clitic ne. This is to say that RI is not to different from Modern Italian, where the participle has the option of not agreeing when the clitic is mi, ci, ti, and vi ‘me’, ‘we’, and ‘you’ (sing. and pl.), cf. Salvi (1991, 239). In MI on the contrary, agreement appears to be regular also with clitics that are first and second person. The evidence is however not clear enough to permit general conclusions. A survey of a larger material will have to show if there are indeed systematic differences in this sense: (70)
MI, 14th: …ci ha qui guidati… (they) us[cl.]-have here brought[masc. pl.] (Dec; I:Intro)
(71)
RI, 16th: Egli ci ha oramai preso … he us[cl.]-have now taken[-Agr] (RagF; I:1)
171
6
Extension to Germanic
The variation within the Romance languages under discussion falls out from the interaction between two parameters. The same is presumably true for most Germanic varieties. Building on Zwart (1993), but obviously also on Chomsky (1993, 1995) and Kayne (1994), the Germanic group would be divided first and foremost with regard to the strength of the accusative feature: (72) Acc strong: Acc weak:
German, Dutch English, Scandinavian
Germanic languages do not show variation on the second point. The phenomenon of operator agreement observed in French and ancient Italian varieties, is not encountered among the Modern Germanic varieties usually discussed in the literature. The Scandinavian, German, and English systems function as the Italian and Spanish ones: accusative is checked by Asp and agreement on the participle is not spelled out in the relative clause. Consider (73) and (74). (73) (74)
Sw. Breven som jag har skrivit… Eng. The letters that I have written…
[-Agr] [-Agr]
In my model, that will have to mean that accusative in Germanic is generally checked against Asp. (75) Acc checked in AgrOP: Acc strong: Medieval Italian Acc weak: RI1, French
Acc checked in AspP: German, Dutch RI2, Scandinavian, English, Contemporary Italian, Spanish
This is to say that Modern Romance standard varieties generally do not know of strong accusative, and that Modern Germanic generally does not recur to AgrO for accusative checking. Medieval Italian, and certainly also Modern Italian dialects, clearly show that a strong accusative feature is found in the Romance group as well (cf. Cocchi 1994). The question arises whether accusative checking in AgrOP is attested in Germanic varieties, but that variation goes beyond the aims of this study. I conclude from this that the languages from the Germanic group presented do not defy a classification in present terms. The crucial patterns observed can be reduced to two
172
fundamental parameters: a checking parameter ruling over the locus of the checking process, and a strength parameter deciding when the checking takes place, either covertly or overtly. As for the Germanic passive, Swedish follows the Romance pattern. The passive participle agrees with the moved DP: (76)
Sw. Breven är skrivna. the letters[pl.] are written[pl.]
It may be that AgrO in German and English is syntactically represented as a phonetically null morpheme. Finally, pronominal structures are clearly not comparable, as Germanic languages do not have clitic pronouns but weak pronouns in the terminology of Cardinaletti (1994a, 1995), and these do not raise overtly in the compound tense where the V itself stays in situ. 7
PAP
We must now return to the PAP construction discussed earlier in chapter three. It was stated that the DP in (77a) and (b) moves to Spec Asp in order to check the delimitedness feature, for reasons that are familiar by now. (77a) (77b)
It: Sw:
Ho le valigie fatte. Jag har väskorna packade. I have the suitcases packed
Since the participles agree in Italian and Swedish, AgrO is obviously also involved in the derivation, and we are now in a position to state this in terms of our account. In Italian and Swedish, accusative Case on the DP is checked in the AspP which is external to VP according to the scheme in (75). The participle V is passive in PAP, and in Italian and Swedish there is then an independent requirement that the passivized DP checks its φ-features in Spec AgrO. As a result, these three requirements, delimitedness, Case and pairing or matching, force the 2 1 projection and checking of AgrO, Asp and Asp , the delimited Asp. When we take all checking processes into consideration, the derivation of PAP in Scandinavian and Romance can be summarized as (78):
173
(78)
AgrOP
Spec DPi
AgrO’
AgrO
AspP
Asp’
Spec ti Asp
AspP
Spec
Asp’
ti VP
Asp [+del.] V
8
ti
Summary
Throughout the discussion, I have followed and defended two ideas concerning the nature of agreement: 1. Agreement is not only a reflex of derivation, nor is it spelled out only because some element has passed through the specifier of a relevant functional head. The content of Agr is something that must be checked in certain grammars and contexts, and cannot be checked in others. In claiming this view, we follow the spirit of minimalism, though diverging from some previous accounts. 2. Checking of Agr cannot be considered as universally equal to the checking of Case. Hence, taking AgrOP to be the locus of accusative checking may for some languages be appropriate but is misguiding for others. The verbal passive construction gives a clear indication for the correctness of point two. In the passive of certain languages, an Agr element is visibly involved in the
174
participial clause although the DP moves to check nominative against the finite Infl. The realization of agreement is thus required but not for reasons of Case. In section 3, we saw that some differences in participial agreement between Romance languages can be derived on the assumption that the locus of accusative checking can be either AspP or AgrOP where the choice is parametrically decided and divides French from Italian and Spanish. The proposal is quite similar to what has been suggested for Nominative checking and the relation between T and AgrS. The account was extended to agreement in other contexts, as those involving clitic pronouns, and the verbal passive construction. In section 5, it was shown that the analysis also takes care of differences between the modern Romance grammars and the historical varieties of Italian under discussion. Lastly, in section 6, I claimed that cross-linguistic variation among Germanic languages does not pose any crucial problem for our view, and that Scandinavian, Dutch, German and English probably fall under the general categories defined by two parametrical choices: First, it is parametrically decided whether accusative checking takes place in AspP or in AgrOP, and second, whether the accusative feature is weak or strong.
175
APPENDIX TO CHAPTER 4: Further Observations on Agreement Phenomena in Medieval and Renaissance Italian
1
Agreement by Attraction
The phenomenon sometimes referred to as agreement per attrazione, ‘by attraction’, deserves a comment here, though I have few precise claims to make. It was observed by Lucchesi (1962/63, 208 f.n. 1) that participles in the eldest texts, above all from the 13th century, sometimes are seen to agree with surface subjects in unergative and transitive contexts. In my own material, examples are very rare. (1) instantiates a transitive participle agreeing with a postponed subject (cf. the edition of Cesare Segre, Nov.no p. 61 f.n. 2). In (2) and (3), the unergative participles of segnoreggiare ‘rule’ and correre ‘run’ agree with their subjects: (1)
Questo libro tratta d’alquanti fiori di parlare, di belle cortesie e di be’ this book tells about ‘flowers’ of speech, nice courtesies and beautiful risposi e di belle valentie e doni, secondo che per lo tempo passato answers and virtues and gifts, according to what in past times have hanno fatti molti valenti uomini. made[m.pl.] many valent men[m.pl.] (Nov.no; I)
(2)
Temo la infamia di tanta passione avere seguita, quanto concepe I fear the infamy[f.sing.] of so much passion having followed, as sees chi legge le sopra nominate canzoni in me avere segnoreggiata; … he who reads the above songs in me having ruled[f.sing.] (Conv; I: II)
(3)
…e cavatemi di queste pene ove vissuta sono and liberate-me from these pains in which I have lived e corsa ho quel corso che la fortuna m’ha dato. and run[f.sing.] have (I) the course that fortune has given to me [subject: feminine singular] (Fatti; XIX)
If we accept that Spec AgrO is a position reserved for the object, both in the transitive case and in the unergative (assuming that unergatives subcategorize for null objects), data such as (1)-(3) seem to indicate that the realization of agreement on the participle does not necessarily depend on AgrO alone. If Spec AgrO is blocked in (1)-(3), Spell Out of agreement must be triggered by DP-movement through Spec AgrS. Thus, there may be grammars where
176
participial agreement corresponds to AgrS (see our above discussion, 4.3., this chapter). Such data are however sporadic in the texts I have consulted. Although they can definitely not be considered productive or general in any of the authors under discussion, it is not excluded that they indeed were productive at some previous stage, and that their occurrence is in some sense a trace of some stable system, chronologically prior to the period treated here.16 In Modern Italian, there is one context where participial agreement can be assumed to correspond to participial AgrS rather than AgrO, namely the absolute ergative participle to which we will turn in Part B.
2
Agreement in Arbitrary Contexts
The realization of agreement in arbitrary constructions is distinctly different in the ancient texts from what we are used to in modern language. It is possible that a minor parametric change is responsible for this. Among languages there seems to be a variety of ways to identify arbitrary meaning morphologically. In Modern Italian, the φ-feature specification identified with arbitrary meaning is plural masculine (Rizzi 1986a): (4) (5)
È difficile essere giovani. It is difficult to be young[pl. masc.] Questa musica rende allegri. this music makes happy[pl. masc.]
In a modern Italian impersonal construction with the clitic pronoun si, the finite V shows singular agreement whereas the nonfinite predicate, adjective or participle, shows agreement in plural masculine (cf. Burzio 1986, 59; Manzini 1986, 242; Cinque 1988a; Salvi 1991, 236238): (6)
(7)
Quando si è giovani, … when si is[sing.3 pers.] young[pl. masc.] (= ‘when one is young’) Quando si è arrivati, … when si is[sing.3 pers.] arrived[pl. masc.] (= ‘when one has arrived’)
16
Lucchesi (1962/63, 207-208) mentions that cases of agreement by attraction are very frequent in some texts included in his own survey. As mentioned above, his material is mainly 13th century and therefore, more remote than the one treated here. Furthermore, the texts he refers to are, as far as I can understand, not Florentine. If these patterns could be considered general for some ancient dialectal variety remains to be seen.
177
According to the analysis of Cinque (1988a, 529), impersonal si is morphologically specified for plural and masculine but unspecified for person. Further, si under Infl identifies an arbitrary pro in Spec Infl (cf. Cinque 1988a, 550):
(8)
IP
I’
Spec pro
VP
I°
si
I° è
The arbitrary pronominal element has the specification of first person plural, we (Cinque, 552), but has no effect on the morphological features of si. Hence, the finite V, that has an inflectional paradigm expressing person will show default agreement which is third person singular. Since participles and adjectives carry morphologic features of number and gender, these will be triggered by pro and will be realized as masculine plural. Cinque (1988a, 534 f.n. 15) suggests however that the number value of si is subject to parametric variation, and it seems plausible that the φ-feature specification of arb is parametrically decided. In the Italian texts of both MI and RI periods, do we find arbitrary contexts where participles and adjectives show singular masculine agreement: (9)
Ora, concludendo, queste sopra dette cose sono utile a divenire isperto Now, concluding, these above mentioned things are useful to become expert[sing.] e intendente al mondo, a farsi bene volere e essere onorato e riguardato; … and wise[sing.] of the world, to make oneself loved and to be honoured[sing.] and respected[sing.] (RicM; III, 265)
(10) … avea una colonna in messo murata, in nella quale si potea venire e udire e veder there was a pilaster in the middle, in which one could come and hear and see tutto ciò che faceano chi in quella camera era, sensa esser veduto. everything that made who in that room was, without being seen[sing.] (Nov.re. I)
178
It is of course very tempting to link this observation to the fact that arbitrary meaning in the ancient varieties was frequently expressed by the pronominalized noun uomo ‘man’ as in Modern French or several Germanic languages (cf. Rohlfs 1969, § 737): (11)
Così sono alquanti, e non pochi, so there are some, and not few, che vogliono che l’uomo li tegna dicitori … who want that man holds them for authors (Conv; I:XI)
One possibility would be that the arbitrary pro of the ancient Italian texts carries the φfeatures of uomo, third person singular masculine, meaning quite simply that the φ-feature value of arb corresponds to man and not to we. There are also some other data pointing in this direction. In (12), we have an example of the arbitrary object pro discussed by Rizzi (1986a). This is another context where the participle in Modern Italian must agree in plural: (12)
Da poppa stava il celestial nocchiero in the stern stood the celestial helms-man tal che faria beato pur descripto such that the description would make blessed[sing.] (Purg; II: 43-44)
The meaning of (12) is that the vision of him would beatify people. The participle carries however singular agreement. Consider also that Modern Italian recurs to a special kind of clitic cluster in order to express an impersonal reading of a reflexive predicate. The clitic normally used in such contexts is si both in the reflexive and in the impersonal. When these appear together, the phonological outcome is not si si but ci si. Take fidarsi ‘trust’ which is reflexive in Italian: (13)
Non ci si può fidare in loro. ci si can’t trust them = ‘one can’t trust them’
In my 14th century material, when an impersonal construction is formed out of either a reflexive predicate (as vestirsi ‘dress oneself’ in (14)) or a reciprocal (as tagliarsi ‘cut each other’, in (15)), only one clitic is realized: (14)
nel vostro mondo giù si veste … in your world (people) dress themselves[cl.] (Par; III: 99)
179
(15)
In inferno si taglia,… In hell (people) cut each other[cl.] (Trec; IV)
I find it highly plausible that the clitic si spelled out is the reflexive one, and not the impersonal. We find cases where the clitic si appears together with the impersonal uomo. (16)
E così nelli amichevoli modi de’ nemici non si dee uomo fidare. and therefore in the friendly ways of the enemies si should not man have faith (Nov.no; LXXVI)
Moreover, impersonal si in 14th century Italian can be omitted. It often is in the prose of Villani in expressions like (17)-(18): (17)
… furono molte cose come innanzi _ farà menzione. there were many things as ahead[=below] _ will mention (Cron; V: 23)
(18)
… come innanzi _ farà menzione … as ahead _ will mention (Cron; V: 38)
It may be therefore, that the arbitrary pro of these varieties is quite simply a null realization of uomo. It must be underlined, however, that there seem to have been several other ways to express arbitrary or impersonal constructions. Sometimes we find altri ‘other(s)’ used as an impersonal pronoun triggering singular agreement (for (19), see the edition of Cesare Segre, Nov.no p. 123, f.n. 3).17 (19)
onde tu piagni te medesima, e assai è laida cosa because you weep for yourself, and it is not nice piangere altri se stesso. to weep other for oneself[sing.] (Nov.no; LXXI)
The use of impersonal constructions varies considerably throughout the period, and it is highly plausible that what we observe in 14th century texts is the confusion due to an ongoing 17
An additional curiosity about (19) is that altri (having quite simply the meaning of ‘people’) appears as the subject of the infinitive V.
180
parametric resetting, and/or the competition between different dialectal and synchronically working grammars. It is sometimes striking how authors alternate different arbitrary expressions in the same context or even in the same sentence. It takes a longer text to illustrate this. Consider (20): (20)
Questo ho voluto recare a memoria per esempro di chi legge: This have I wanted to recall as an example for him who reads cioè che niuno, o maschio o femmina, né per paura né per lusinghe né namely that no one, man or woman, neither for fear nor for flattering per veruno modo mai si spogli di suo avere o di sue ragioni, con ciò nor for any other reason, should ever abandon his belongings or his sia cosa che, de’ cento, e novantanove ne rimangono disfatti, e pure rights, since out of hundred, ninety-nine are ruined thereby, and even da’ più istretti parenti o amici; però che sono quelli in cui altri si by the closest relatives or friends; because they are those in which fida, e però se ne rimane ingannato e tradito, e in ultimo perduto other si[refl.cl.] trusts, and therefore si[impers. cl.] remains il suo avere, rimanendo di tutto nimico di chi te l’ha tolto. fooled[sing.] and betrayed[sing.], and finally lost ones belongings, remaining enemy[sing.] of whom took it away from you. (RicM. III 188)
The first impersonal pronoun used meaning ‘people’ is altri that appears together with the reflexive si, both in bold face. The construction is similar to (19) above. In the following sentence, the impersonal is expressed with si, as in Modern Italian, triggering however singular agreement on both participles ingannato e tradito. In the final portion, the author Morelli ends up with an arbitrary tu ‘you’. Not surprisingly, nimico ‘enemy’ shows the morphological specification of masculine singular. This continuous gliding between different person specifications is attested elsewhere. Whereas in some of the examples above the impersonal si appears together with pronouns like uomo or altri, there are cases in the prose of Boccaccio where impersonal si alternates with noi. In the following lines, a discourse is initiated with a noi ‘we’ which is transformed into a probably impersonal si:18
18
The example is noted by the editor Giorgio Padoan, Espos; I: II footnote 131; See also Espos; VI: II footnote 72.
181
(21)
Le quali cose, sommamente considerate, assai aperte dimostrano noi these things, together considered, clearly show (that) we dover potere per lo leone, al nostro autore aparito, could be able to, for the lion, that appeared in front of our author intendersi il vizio della superbia. understand-si[cl.] the vice of pride (Espos; I: II)
(21) may very well be an error, but is nevertheless revealing. If we assume that arb has the φfeature specification of uomo ‘man’ in ancient grammar, and receives the specification of noi ‘we’ in the modern one, cases as (22) may reflect a change in arb interpretation from man to we. This is of course nothing but a guess, though it is suggestive. I conclude by stating simply that a further study of the historic varieties might prove fruitful for our understanding of the syntax of arbitrary expressions.
182
PART B THE ABSOLUTE PAST PARTICIPLE
CHAPTER 5 Participial Absolute Small Clauses in the History of Italian: Data
0
Introduction to Part B and Chapter Five
We will now turn to the syntax of the participle in its absolute use. The first part of this thesis was introduced by a descriptive section chapter 1, and this second part will also begin with a presentation of the data in chapter five. The construction we are interested in is the one studied by Belletti (1990) among others, and is illustrated in (i) (example (i) from Belletti 1990, 89):1 (i)
Arrivata Maria, Gianni tirò un sospiro di sollievo. arrived Maria, Gianni was relieved
The term absolute is not uncontroversial. It is intended to mean independent from the matrix in some sense, but as has been often observed the notion of independency can be understood in a variety of ways. Absolute constructions may of course depend to some degree on the finite clause to which they are attached semantically, since their temporal value is settled in relation to the finite verb, or grammatically, as arguments of the absolute and the matrix may be co-referential. In addition, there is a high degree of ambiguity in the syntax of ancient literary Italian. A brief look at the 14th century authors is sufficient to make us understand that the syntax of coordination, subordination or relativization is distinctly different from modern norms. The participial clauses we will go through in the following chapter are often strikingly ambiguous in their grammatical function, a circumstance that casts some doubt on the adequacy of the term ‘absolute’. Still, I believe the traditional terminology is preferable for ease of reference and will henceforth use the abbreviation PASC for a Participial Absolute Small Clause. I will consider Italian absolute constructions of three centuries: the 14th, the 16th and the 20th. The general patterns of PASC during these periods are evident enough to allow us to recognize three distinct and roughly stable grammatical systems. However, the reader must pay attention to a fundamental circumstance that we have already touched on in chapter one. There is no perfect correspondence between the three systems and the three 1
See Ruwet (1982) for absolute constructions in French and Gunnarson (1995) for the same issue in Spanish. The semantic implications of absolute clauses are discussed in Raposo & Uriagereka (1993).
183
centuries. As we have already seen in the periphrastic past tense, there is some major break by the middle of the 14th century. Authors of the late 14th century use patterns strikingly similar to those of the 16th century. On the other hand, there are 16th century texts that remind us to some extent of the first half of the 14th century. As previously mentioned, we have the choice of describing and analyzing different grammars, calling them simply G1, G2 etc., but I believe that such a choice would complicate the reading. Therefore, the labels from Part A will be maintained: I define Medieval Italian (MI) all texts until the mid 14th century, the period from the late 14th until the 16th century will be called Renaissance (RI), and 20th century will be labeled Contemporary (CI). The definitions are arbitrary but handy. The reader is again asked to consider that what follows in the next two chapters is a descriptive overview and analysis of three synchronically working grammars and the parameters that presumably distinguish them, not of the passage from one grammar to the other. It is highly plausible that the break between MI and RI occurs around 1330-50, but I do not know when the 16th century system, RI, is replaced by the 20th, CI. The emerging picture as we go through the data will be quite complex. As we turn to the analysis in the following section, we must try to abstract away from infrequent irregularities that to some extent may be due to stylistic factors. Indeed it is not a trivial task to distinguish the general rule from the exceptions. The issue is particularly intricate in the 16th century prose, and I will signal the uncertainties as we proceed. The corpus from part A has been extended. As for the texts added, the reader is referred to the list of ancient texts cited by the end of this thesis. The demonstration of part B is not based on a statistic survey; I have not counted occurrences of the constructions that will be discussed, but I rely on the general impressions that in many cases are quite clear. The reasons for this have to do with problems of interpreting ancient punctuation, and will be explained in section 2.2. below and in the appendix to this chapter. Moreover, some of the patterns we are interested in, as for instance the negation in PASC, are so infrequent that a statistic calculus would be meaningless. 20th century data are taken from Belletti (1990), Cinque (1990b), Bertuccelli Papi (1991, in Renzi & Salvi eds.), from discussions at Class Lectures of Maria Rita Manzini (1993-1994), and to some limited extent, from my informants.2
2
As some of the constructions reported in the texts are quite complex, the crucial portion of the sentence is put in bold face in order to facilitate the reading. When the Italian text extends over more than one line, the English gloss will be given in italics. Moreover, the relevant participle verb is underlined in the English gloss.
184
1
Organization of the Data
The syntactic patterns of PASC that we are interested in can be listed in nine points:3 1. How many lexical arguments are can appear in PASC? The issue is to establish whether the construction admits for both a lexical subject and a lexical object or only one of them. I disregard oblique arguments as well as adverbial material. 2. Where are the arguments, subject and object, placed in relation to the participial verb? Is the verb obligatorily in first position, or can it be preceded by any of S or O? 3. Does the participle agree with its arguments? There are two cases two consider: (a) agreement or non-agreement transitive participle-object, and (b) agreement or non-agreement ergative participle-subject. 4. What are the lexical restrictions on PASC? As before, we are primarily interested in the transitive-ergative-unergative distinction. We will also see if the participial form of the copular V, stato ‘stayed’, ‘been’ (< stare, essere; ‘stay’, ‘be’) is possible in PASC. 5. Is negation possible? 6. Is the agentive by-phrase possible? We distinguish, of course, between PASC (a)with a lexical argument, and (b)without a lexical argument. 7. Can PASC host Comp-field elements? I intend by ‘Comp-field elements’ two categories of words: (a) complementizers and (b) relative and interrogative expressions. In each of these categories we wish to know whether lexical argument(s) are possible. Under point seven we will therefore have four subgroups of constructions: 7.1.a. PASC with complementizer and lexical argument; 7.1.b. PASC with complementizer but without lexical argument; 7.2.a. PASC with wh-Comp and lexical argument; 7.2.b. PASC with wh-Comp but without lexical argument. 8. Can PASC have an argument null subject? In other words, can the null subject of PASC be disjoint in reference from the matrix subject? Again, there are two cases to consider: (a) null subject with ergative V, and (b) null subject with transitive (and unergative) V. 3
I henceforth use the terms subject and object with reference to Case theory and not to the internal/external distinction. A subject is a category, lexical or empty, that carries nominative, and an object is a category carrying accusative.
185
9. Can PASC have an argument null object? We wish to know whether an empty category marked accusative can be free in reference. On all of these points, we are interested to know about any eventual differences between MI, RI and CI.
2.1
Number of Lexical Arguments
In MI and RI, the transitive PASC can host two arguments, subject and object. This is not so in the 20th century CI according to Belletti (1990) and Bertuccelli Papi (1991): 20th century, CI: (1)
*Conosciuta Elena Gianni, tutti cominciarono a fare pettegolezzi. known Helen John, everybody started to gossip (Bertuccelli Papi 1991, 596)
It has been observed (Manzini, Class 1994), that transitive PASCs with two realized arguments are not totally excluded in Modern Italian: (2)
Restituiti io i libri, … returned I the books, …
For Manzini, the acceptability of (2) may be a matter of register. Furthermore, (2) is better with a pronoun than with an R-expression (Anna Cardinaletti and Gloria Cocchi, both p.c.). Constructions of this kind are however abundant in early 14th century texts. From the late 14th onwards, they decrease in frequency but are nevertheless attested. Both subject and object can obviously be R-expressions. 14th-16th centuries, MI-RI: (3a)
Fatte le comandamenta la Fede Giudea, e la fe’ dell’idoli morta Made[f.sing] the commandments the Jewish faith, and the pagan faith e spenta, cominciò la Fede Cristiana a segnoreggiare tutto ’l mondo… dead began the Christian faith to rule all the world (Vizi; XLII)
186
(3b)
Vinta la Fede Pagana tutta la terra d’oltremare …, Won[f.sing] the Pagan faith all the land beyond the sea colse baldanza sopra la Fede Cristiana … (she) became bold towards the Christian faith (Vizi; XLVII)
(3c)
Racquistata e rivinta la Fede Cristiana Conquered[f.sing] and re-won[f.sing] the Christian faith tutta la terra di qua da mare … all the land on this side of the sea (Vizi; LIII)
(3d)
ed Enea presa la lancia, … and Enea taken[fem.sing.] the lance, … (Fatti; LVII)
(3e)
Ordinate le schiere de’ due re nel piano … e ciascuno de’ detti ordered the lines of the two kings on the field and each of the named signori ammonita la sua gente di ben fare, … lords admonished his men to do well (Cron; VII: 9)
(3f)
e per comandamento di lei, Dioneo preso un liuto e la Fiammetta and on her commandment, Dioneo taken[m.sing.] a lute and Fiammetta una viuola, cominciarono soavemente una danza a sonare… a viola, (they) began sweetly a dance to play (Dec; Intro)
(3g)
ricevuta ser Ciappelletto la procura received[f.sing.] ser Ciappelletto the proxy e le lettere favorevoli del re … and the favours of the king (Dec; I:1)
(3h)
e il santo data la benedizione a uno de’ fedeli … and the saint given the benediction to one of the faithful furono menati dinanzi a’ farisei … they were brought before the Pharisees’ (MMin.; XII)
187
(3i)
Veduto adunqua i soprascritti omini e donne, frati e preti, seen thus the before mentioned men and women, friars and priests la pestilensia multiplicare, … pensonno … the plague multiply they thought… (Nov.re; Intr.)
(3j)
Consumato il ditto Manasse i suoi ducati, spent the mentioned Manasse his money, … e di quine ne trasse uno … … and therefrom he took one … (Nov.re; I)
(3k)
E veduto il Calí la bella apariensia de’ giovani, … and seen the Calif the nice look of the youngsters (Nov.re; I)
(3l)
Acquistata adunque el re la Lombardia, Conquered thus the king Lombardy si riguadagnò subito quella reputazione … he immediately regained the reputation (Princ. III)
(3m)
saputo il Soderini il meraviglioso ingegnio di mio padre … known Soderini the marvelous talent of my father (Vita; VI)
2.1.1
Summary
The correct generalization seems to be that the transitive PASC can host two lexical arguments, subject and object, in MI and RI. In CI, this is generally not possible, except under special circumstances, that is, if the subject of the transitive PASC is pronominal.
2.2
Word order
Belletti (1990) shows that the participial V of PASC must precede its argument, regardless of whether it is external or internal:
188
20th century, CI: (4a) (4b) (4c) (4d)
Arrivata Maria, … *Maria arrivata, … Conosciuta Maria, … *Maria conosciuta, …
(Arrived Mary …) (M. arrived …) (Known M. …) (M. known …)
VS *SV VO *OV
Bertuccelli Papi (1991) gives an example of SV word order from contemporary literary language: (5)
Di solito, lei uscita, io finivo la colazione usually, she gone out, I finished breakfast e Millo mi rivedeva i compiti. and Millo looked over my homework (Sciascia, Candido; cit. Bertuccelli Papi 1991, 600)
Notice that the example she cites instantiates a preposed pronominal subject. There may be differences in movement between pronouns and R-expressions to which we will return below in this section. Quite different is the MI system, where both object and subject occur to the left and to the right of the participial V: 14th century, MI; Ergatives: (6a)
Per ciò che levatesi tutte… for this reason stood up all of them (Dec; I:10)
VS
(6b)
Appresso ciò non molti dì passati, … after this not many days passed (VN; XXII)
SV
(6c)
… molti de’ vicini avanti destisi e levatisi … many of the neighbours woken and stood up (Dec; II:5)
SV
14th century, MI; Transitives: (7a)
… e con festa da lei preso commiato, si partì. and with joy from her taken farewell, he went away (Dec; I:10)
189
VO
(7b)
… con senno e con ordine l’uficio commesso, … with sense and with order the office accomplished (Dec; II:8)
OV
The Old Italian structures are strikingly different from the example reported by Bertuccelli Papi (5) as preposed objects and subjects in the elder texts are normally ‘heavier’, and can appear in preparticipial position together with other material such as adverbs and prepositional phrases. When, in the ancient texts, we have two lexical arguments, subject and object, there are six logical combinations: VSO, VOS, SVO, SOV, OSV, and OVS. Out of these, I have attested VOS, VSO and SVO ((8a), (8b), and (8c) equal (3a), (3b) and (3f) above):
(8a)
(8b)
(8c)
V Fatte Made
O S le comandamenta la Fede Giudea, …, the commandments the Jewish faith, … (Vizi; XLII) V S O Vinta la Fede Pagana tutta la terra d’oltremare …, Won the Pagan faith all the land beyond the sea, … (Vizi; XLVII) S V O Dioneo preso un liuto, … a lute Dioneo taken (Dec; Intro)
We do not find SOV, OSV or OVS unless either of S or O is a relative pronoun, see examples below in 2.7.2. Interestingly, RI respects largely the same rule as Modern Italian; the argument is found in rightward position regardless of the transitive/ergative distinction: 16th century, RI: (9)
… morti tutti quelli che, …, lo potevano offendere … VS dead all those who could do him harm (Princ; VIII)
(10)
… preso el ducato di Urbino … taken the duchy of Urbino (Princ; VII)
VO
190
This is with little doubt the predominant pattern. It must be remembered that RI data vary considerably with stylistic level as already stated, but the variation on this point is not striking; see below this section. When in RI we have two lexical arguments, word order seems less free than in MI. Consider again some cases from the late 14th and the 16th centuries; (11a)-(11d) equal (3j)-(3m) above):
(11a)
(11b)
(11c)
(11d)
V S Consumato il ditto Manasse the mentioned Manasse spent (Nov.re; I) V S E veduto il Calí and seen the Calif (Nov.re; I)
O i suoi ducati, … his money, …
O la bella apariensia de’ giovani, … the nice look of the youngsters
V S Acquistata adunque el re Conquered thus the king (Princ. III)
O la Lombardia, Lombardy
V S O saputo il Soderini il meraviglioso ingegnio di mio padre … known Soderini the marvellous talent of my father (Vita; VI)
Word order in the relevant examples is obviously VSO. A major problem is of course that we cannot rely upon the reading suggested by punctuation. In cases like (12a)-(b), it is impossible to tell whether the subject belongs to the matrix or to the participial clause. For discussion on this problem, see for instance Herczeg (1972, 193), Ageno (1964), and the appendix to this chapter: (12a)
Saputo questo mio padre, armato, lo andò a trovare … known this my father, armed, went to see him (Vita; IX)
(12b)
Quelli che lo portavano udite le parole, those who carried him heard the words spaventati, lasciarono andare la bara … horrified, let the the coffin go (Cene; II: 2)
191
Any analysis that tries to explain ancient Italian syntax will inevitably suffer from this uncertainty. This is the reason why we cannot calculate reliable statistics on word order in these constructions, we do not know how to count cases like (12a) and (12b) which constitute a good part of the material. I will return to this problem in the appendix to this chapter. Notice that if the DP subjects of (12a) and (12b) make part of the absolute clauses, we have two other examples of PASC with two lexical arguments in RI which has been independently attested (see section 2.1.). On the other hand, if they are included in the PASC, we have examples of VOS and SVO, the independent motivation for which is less solid. 16th century syntax might have had the option to prepose a pronominal subject to the participle: (13a)
… e lui per ispavento caduto per terra, dissi: … and he for the fright fallen to the ground, I said … (Vita; XVII)
(13b)
… ed ella postasi lo’ in mezzo, and she put-herself[cl.] between them buona pezza ragionarono di più varie e diverse materie for a long while they spoke about several things (Cinq. Fortini; p. 331)
(13c)
anco ella assicuratasi seco, gli rendeva parte degli amorosi baci even she taken-herself[cl.] confidence with him, returned his kisses (Cinq. Fortini; p. 331)
(13d)
… egli accortosi per mille segni, che il padrone era egli, he recognized by a thousand signs that the master was he, per colorir com’io mi credo un suo disegno, to hide as I believe a plan of his se mai la occasione gli venisse, pensò … in case the occasion presented itself, thought … (RagF; I:1)
However, it cannot be excluded that the pronominal subjects of the last two constructions, (13c) and (13d), belong to the matrix clause and not to the absolute clause. Only the first two are unambiguous in this sense: the pronominal subjects of (13a) and (13b) must belong to the absolute clause, since they are disjoint in reference from the matrix subject. Still, verb-subject word order is also predominant when the subject is pronominal:
192
(14)
Così, partito lui, si mosse a ragionare meco … so, went out he, he came to speak with me… (LettA; III)
2.2.1
Summary
It is quite clear that CI PASC is subject to a verb-first-restriction, at least if the discussion is limited to word order of the predicate and its arguments. It is equally clear that no such restriction was valid for MI PASC. In the MI period, subjects, objects as well as adverbial material could precede the V. In PASCs with a single argument, we attest VO/OV, and VS/SV. When it comes to PASCs with two lexical arguments out of the six logically possible word order combinations, namely: VSO, VOS, SVO, SOV, OSV, and OVS, I have attested the first three, VSO, VOS, and SVO, but not SOV, OSV, and OVS. The important generalization then is that if there is a subject present in the PASC, the object cannot appear to the left of the verb. As for RI, it appears that the restriction on V1 is already valid by and large from the mid 14th century onwards. In PASC constructions with a single argument, word order is VO and VS, and when there are two arguments, VSO. However, the picture is less obvious in RI than in the other periods for reasons discussed above. There are clear counterexamples, SV-word order, at least when the preceding subject is pronominal.
2.3
Participial Agreement
In a Modern PASC, the participial V obligatorily shows agreement with its DP argument be it nominative or accusative: 20th century, CI: (15) (16)
Arrivata (*-o) Maria, … arrived[+Agr] M. Conosciuta (*-o) Maria, … known[+Agr] M.
This property is also attested in medieval texts, where PASCs do not fail to show agreement regardless of where they appear in relation to the verb (cf. examples (6) and (7) above section 2.2.):
193
14th century, MI; Ergatives: (17a)
E occorsagli una nuova malizia, … and happened[fem.sing.]-to-him[cl.] another misfortune (Dec; I: 4)
(17b)
…, venuta l’ora del desinare, … arrived[fem.sing.] the hour of supper (Dec; I: 5)
14th century, MI; Transitives: (18a)
e in esilio morí in Francia (ch’aveva a fare di là ed era compagno de’ and in exile he died in France (because he had to do there and was a Pazzi) e tutti i suoi beni disfatti, e certi partner of the Pazzi) and all of his goods destroyed[+Agr], and certain altri popolani accusati con lui … other men of the people accused[+Agr] with him’ (Cron: VIII; 8)
(18b)
… serrata la cella con la chiave … locked[+Agr] the cell with the key (Dec; I:4)
(18c)
… il re di Francia, molte triegue fatte con gli alamanni, morì … the king of France, many truces made[+Agr] with the Germans, died (Dec; II: 8)
In the earliest texts, until the Decameron, the exceptions to the rule are extremely few: (19a)
E dettoli le donzelle essere dimonî, … and said[-Agr]-to him[cl.] the girls be demons (Nov.no. XIV)
(19b)
conosciuto tantosto costei esser femina … known[-Agr] soon her be woman (Dec; II: 3)
(19a) and (b) are accusative with infinitive constructions.
194
At some point however, object agreement is lost in the transitive PASC. The typical RI pattern is the one formalized in (20) where the postposed object does not trigger agreement on the participle V: (20)
RI: participle[-Agr]-DP(O)
The most thorough investigation of the issue is due to Škerlj (1932), see also Lucchesi (1962/63) and Rohlfs (1969, § 725). Considering Škerlj’s study, we notice that the examples of the relevant kind are strikingly few during the first half of the 14th century in Florentine prose.4 But, by the second half of the century, non-agreement appears to be the rule for some authors (e.g. Morelli and Sercambi). Non-agreement in the transitive PASC remains solid throughout the RI period also being predominant in the 16th century. Generalizing on the data, the ergative participle tends to agree with a postposed subject, whereas the transitive participle tends not to agree with a postposed object. (21a)-(21g) is strikingly more frequent than (22), and (23a)-(23c) more frequent than (24): 14th-16th centuries, RI; Transitives: (21a)
Mostrato la via al garzone, vi andò malvolentieri showed[-Agr] the way to the boy, he went there reluctantly (Trec; XVII)
4
Škerlj (1932) reports only two examples from a text of Bono Giamboni. In Il Libro dei Vizî e delle Virtudi used in my survey, non-agreeing participles in PASC are hard to find. The infrequency if not absence of this pattern in Bono Giamboni is significant as the PASC construction itself is highly frequent. It should be pointed out that my result is slightly different from the one reached by Lucchesi (1962/62) on this point. Lucchesi claims that non-agreement in the relevant context is rare in the 13th and early 14th century, but nevertheless, reports some exceptions to the rule. For Il Libro dei Vizî e delle Virtudi , Lucchesi relies on the edition of Tassi from 1836, whereas I have used the edition of Cesare Segre 1968. (i), for instance, is given as non-agreeing by Lucchesi, but shows agreement in Segre’s edition. (i) is reported as (66) below in this chapter: (i) … e giurato le sue commendamenta … … and sworn[-Agr] her commandments … (Lucchesi, Il Libro dei Vizî e delle Virtudi, 276) Likewise, Lucchesi gives examples of non-agreeing PASC constructions from the Decameron that in my edition are agreeing. Among others, Lucchesi gives the following example where the participle does not show agreement with a clitic object. (ii) is given as (1) in the appendix to this chapter: (ii) … e io messogli in una mia cassa… … and I put[-Agr]-them[cl] in a box … (Lucchesi, Decameron, 261) Lucchesi refers to the Decameron edited by Vittore Branca in 1951/52. My own copy is the more recent edition of Vittore Branca from 1989, where (ii) has been changed to an agreeing form.
195
(21b)
E tolto moglie, uscito di manovaldi e mancando la roba … and taken[-Agr] wife, gone out of servants and missing stuff … (RicM: III, 161)
(21c)
… gittato sopra una panchettina la tovaglietta … thrown[-Agr] on a bench a towel (RagA: I)
(21d)
Però, lasciato questi ragionamenti … however, left[-Agr] these arguments (Dial: I)
(21e)
Di poi la mattina fatto colazione‚ … then in the morning made[-Agr] breakfast (Cinq. Fortini; p. 324)
(21f)
… serrato le finestre se ne andarono… closed[-Agr] the windows they went away (Cene: Introd.)
(21g)
Alla fine dato licenza al compagno, … at the end given[-Agr] farewell to his friend (Cene; I:1)
(22)
e stangata la porta perchè il volpone non scappasse … and barred[+Agr] the door so that the fox could not escape (RagA: I)
14th-16th centuries, RI; Ergatives: (23a)
… quelli cardinali… che, diventati papi, those cardinals that, become[+Agr] popes, …avessino ad avere paura di lui. had reasons to fear him (Princ: VII)
(23b)
E fra pochi giorni venute le robe, … and after a few days arrived[+Agr] the dresses (Cinq. Fortini; p. 335)
196
(23c)
…, venuta la fine delle nozze, Lavinia se ne ritornò a casa, … arrived[+Agr] the end of the matrimony, Lavinia went home (RagF: I:2)
(24)
… a lui passato la stizza e a me la paura… for him passed[-Agr] the harm and for me the fear (Vita: XXIV)
Notice the contrast between the two participles in (25) taken from the same page as the example (21g) above: (25)
I gioveni, veggendo sparite le donne e le finestre the youngsters, seeing disappeared[+Agr] the women and the windows serrate, subito, lasciato la impresa, se ne tornarono in camera; closed[+Agr], immediately, left[-Agr] the enterprise, went back to their room (Cene: Intro.)
The second line put in bold face, is presumably an absolute participle, and the first may be a complement of veggendo ‘seeing’, coordinated with the preceding participial phrase). When the argument is in rightward position, the participle fails to agree. But, when it appears to the left of the participle, we find agreement. In the first conjunct after veggendo, there is an ergative participle agreeing with its subject to the right. Consider also the following alternation in Sercambi, typical not only for his prose, but generally from the mid 14th century onwards:5 (26a)
Venuta l’ora della cena, … arrived[+Agr] the hour of dinner (Nov.re: I)
(26b)
E veduto l’ora da doversi partire … and seen[-Agr] the hour to leave (Nov.re: IIII)
5
This circumstance of course has not gone unnoticed in the philological literature. The wide spread lack of rightward agreement is not limited to participial contexts, but extends to adjectival and copular constructions in general. The editor of the Cene, Riccardo Bruscagli, notes at several occasions that predicates (including the absolute participle) fail to agree with a postposed noun (e.g. Bruscagli; Cene, p. 7 footnote 1, and p. 17 footnote 4). In his comment to the Ricordi of Morelli, Vittore Branca notes: «participio e predicato aggettivale non concordano quasi mai, quando sono preposti, col soggetto; e del resto la desinenza -o ha un valore neutro … » (Branca; RicM, 163 footnote 1).
197
2.3.1
Summary
The participle in CI PASC must show agreement with the postposed subject, if ergative, or with the postposed object, if transitive. Likewise, the participle in MI always agrees with the object in transitive constructions and with the subject in ergatives regardless of where the argument appears in relation to the verb. In RI, the absolute transitive participle generally does not agree with its object, whereas the ergative one agrees with the subject. So, in spite of some exceptions, we can claim that MI and CI are clearly distinct from RI.
2.4
Lexical Restrictions
It is commonly held that Modern Italian PASCs do not admit unergative predicates. In the literature, (27a)-(27e) are given as ungrammatical and to these I add (27f)-(27i) with the judgments of my informants: 20th century, CI; Unergatives: (27a)
*Telefonato Gianni, Maria andò all’appuntamento. telephoned John, Mary went to the appointment (Belletti 1990, 89)
(27b)
*?Parlato con Mario, andò a casa. spoken with Mario, he went home (Cinque 1990b, 24)
(27c)
*Lavorato tutto il giorno, Gianni si sentiva stanco. worked all day, John felt tired (Bertuccelli Papi 1991, 594)
(27d)
*Camminato nel parco, Gianni tornò a casa. walked in the park, John went home (Bertuccelli Papi 1991, 594)
(27e)
*Piovuto tutto il giorno, non potemmo uscire. rained all day, we couldn’t go out (Bertuccelli Papi 1991, 594)
(27f)
*?Cantato 20 anni alla Scala, il tenore si ritirò. sung 20 years at the Scala, the tenor retired
198
(27g)
*?Pensato alle consequenze delle sue azioni, se ne pentí. thought of the consequences of his acts, he had regrets
(27h)
*?Risposto a tutto davanti al giudice, fu assolto. answered to everything in front of the judge, he was absolved
(27i)
*?Regnato sul paese per dieci anni, il re abdicò. reigned over the country for ten years, the king abdicated
It appears as if judgments among native speakers are somewhat elastic being sensitive to the aspectual value of the construction. Some of these verbs are notoriously ambiguous between an unergative and a transitive reading. The ambiguity holds for a number of predicates, as pensare ‘think’, rispondere ‘answer’, and cantare ‘sing’, which all tend towards an atelic reading in (27). Bertuccelli Papi uses the terms ‘durative’ and ‘atelic’ instead of ‘unergative’ and these may indeed be more adequate definitions. Telicity can be imposed on these predicates when they are followed by a nominal expression, an object for instance, that delimits or measures out the event in the terms of Tenny (1987) or Borer (1993, 1995). Verbs like pensare ‘think’ or cantare ‘sing’ acquire transitivity, pensare una soluzione ‘think out a solution’, cantare una aria ‘sing an air’, and can appear in PASC when used in the passive voice. (28)
Cantata molte volte dalla Tebaldi, l’aria fu incisa su disco. sung many times by Tebaldi, the air was recorded
Bertuccelli Papi (1991, 596) notes the contrast between pensare used as a passivized transitive in which case it can appear in PASC, and when it is active where it cannot be used in PASC. As pointed out by Cinque (1990b, 26 f.n. 25), unergatives improve when introduced by an adverbial element as una volta ‘once’. Both of (29) and (30) are taken from Cinque (1990b): (29)
?Una volta parlato a Piero, Maria … once spoken to Piero, Mary …
(30) also becomes possible, where the unergative predicate is followed by a lexical argument: (30)
?Una volta cenato anche i bambini, once dined the children too, potremo parlare con maggior tranquillità. we will be able to speak with more serenity
199
What happens in (29) and (30) is that the adverbial expression puts an endpoint to the event introducing a meaning of telicity. All of (27a)-(27i) improve somewhat when introduced by una volta according to my informants. In the ancient texts, constructions such as the following are abundant: 14th-16th centuries, MI/RI: (31a)
Parlato la Filosofia cosí profondamente sopra la materia del spoken Philosophy so deeply about the reason mio rammaricamento, …sí mi sforzai di difendere il mio errore … for my despair, I tried to defend my error (Vizi; IX)
(31b)
Cenato ogni gente, e rassettate a sedere, disse la Fede … dined everybody, and had a seat, the Faith said … (Vizi; XVI)
(31c)
Risposto alla Prudenzia a tutte le sue adomandagioni … disse answered to Prudence to all her questions … she said (Vizi; LXIX)
(31d)
… e lui regnato nello imperio ott’anni, morí … and he governed in the empire for eight years, died… (Cron; V: 18)
(31e)
E lui riposato e soggiornato in Firenze alquanti dí, and he rested and stayed in Florence some days sí richiese il comune di volere la signoria e guardia della cittade … asked again the commune for the government and the custody of the town (Cron; VIII: 49)
(31f)
… e quivi, cautamente domandato della donna alla quale and here, cautiously asked about the woman to whom the daughter la figliuola lasciata avea e del suo stato, trovò la Giannetta moglie left he had and about her state, he found G. married del figliuolo … to his son (Dec; II:8)
(31g)
E cosí pensato, una mattina si misse in cammino … and thought in this way, one morning he went away (Trec: III)
200
(31h)
E intrati in camino e caminato alquanto, and entered on the road and gone for a while l’uno delli imbasciadori parlò dicendo: … one of the ambassadors spoke and said’ (Nov.re: IIII)
(31i)
Mangiato, prima che da taula si partisseno, Calidonia … parlò … Eaten, before they went away from the table, Calidonia spoke (Nov.re.; IIII)
(31j)
Il quale, appena camminato il mezzo, who, walked half way, pervenne al piè della montata … came to the foot of the hill (Cinq. Cademosto, p. 153)
(31k)
«Sta bene» disse il maestro; e sopra ciò pensato alquanto, «That’s well» said the master; and on this thought somewhat, si dispose … he decided (Cene; I:1)
(31l)
La sera intanto ne venne, e Salvestro, tornato a casa e cenato, … the evening came, and Salvestro, returned home and dined (Cene; I:1)
(31m)
I becchini, messo che t’aranno nella bara, the gravediggers, laid as they have you in the coffin e alla fossa condotto, e cantato, … and brought to the grave, and sung (Cene; II:2)
(31n)
Pur saltato inanzi e lanciandosi su per le scale, then jumped forward and throwing himself up the stairs giudicò quasi quasi che la stanza non facesse per lui. he found almost that the room was not fit for him (Zucca: II; ‘Grillo Primo’)
Two observations may be of some relevance: Apart from (31j), these PASCs do not have adverbial modifiers comparable to una volta (alquanto = ‘somewhat’/‘for a while’ in (31h)/(31k) may have a similar function, however). Furthermore, some are constructed not
201
with direct arguments, hence, not as transitives, but with oblique ones: regnato nell’impero, domandato della donna, pensato sopra ciò.6 Notice also that PASCs in the historic varieties can be formed with the participle of essere/stare ‘be’ with the meaning of ‘be’, ‘stay’ or ‘remain’; and stare + adjective where the modern grammar would need the support of a gerund, essendo stato assente ‘having been absent’ to create the meaning of duration: 14th-16th centuries, MI/RI: (32a)
Martellino, …, stato alquanto, cominciò … M., stood/remained for a while, began a far sembiante di distendere l’uno de’ diti … to pretend to twist one of his fingers (Dec; II:1)
(32b)
… e poi che ebbero udito la messa, e istati in orazione and after they had heard the mass, and stayed praying infino a terza, il prete … prese il messale… until ‘terza’ the priest took the missal … (Fior. II)
(32c)
Ma pure stato un pezzo, ed e’ disse: … but remained/waited for a while, and he said (Pec; II: 2)
(32d)
Dapoi, stato quello che gli piacque, prese commiato … Then, been/remained as long as he wished, he took farewell (Trec; III)
(32e)
… la santa e cattolica Parte guelfa, la quale Idio mantenga the holy and catholic ‘Parte guelfa’, which God saves come sua divota insegna, in quanto al mondo, as his devout, in secular affairs, sempre in favore della Santa Chiesa istata. always in favor of the holy church been (RicM: III, p. 197)
6
Domandare ’ask’ can take a direct complement in ancient Italian and can thus be a transitive verb (Ageno 1964, 48). This is not relevant for our case: domandare in (31f) is used unergatively.
202
(32f)
E stato alquanto tempo, insieme la ditta nuova cameriera and stood for a while, together with the mentioned new maid dormendosi colla reina … sleeping with the queen… (Nov.re.; IIII)
(32g)
E stati a Lucca alcuni dí, passonno per la via di Pistoia. and stayed in Lucca for some days, they continued to Pistoia (Nov.re.; IIII)
(32h)
… e stato el piú tempo della età mia assente dalla patria… and been for most of my life time absent from the fatherland (Dial; Proemio)
(32i)
… se ne andò co i vetturali insieme a Lione; dove stato he went off with the coachmen together a Lyon, where stayed alquanti giorni, mise mano a i suoi sacchetti, … for some days, he laid hands on his sacks … (Cene; I:5)
2.4.1
Summary
PASC in CI is subject to some lexical restrictions that were not valid in ancient times. Both MI and RI PASCs could be formed out of an unergative predicate. This is marginally possible in modern times if, and only if, the construction is introduced by a delimiting adverbial expression such as una volta ‘once’. Moreover, the participial form stato ‘stayed’ or ‘been’, followed by a prepositional phrase, an adjective or a temporal adverb, can be used in PASC in CI together with a gerund essendo stato ‘having been’, but only in this instance. In MI and RI PASCs, stato appears frequently alone, without the support of a gerund.
2.5
Negation
Negation is ungrammatical in 20th century PASCs; see (33a) and (33b) from Belletti (1990, 95): 20th century, CI: (33a)
*Non arrivata Maria, Gianni tirò un sospiro di sollievo. not arrived M., John was relieved
203
(33b)
*Non mai conosciuta Maria, … not ever known M., …
Negation is attested in both 14th and 16th century PASCs. In some cases it is followed by a lexical argument. 14th-16th centuries, MI/RI: (34a)
E trattato e messo in assetto col papa e col re Carlo il passaggio and treated and agreed between the pope and king Charles the passage di Cicilia alla primavera vegnente, … , il papa non dimenticato to Sicily the following spring,… , the pope not forgotten lo sdegno preso contro alla parte bianca di Firenze, the resentment against the white part of Florence, non volle che soggiornasse e vernasse invano. did not want him to stay and spend the winter in vain. (Cron; VIII: 49)
(34b)
… mai non riconosciutala, (…) not ever recognized-her[clitic] lei nelle braccia ricevuta lagrimando teneramente basciò. her in his arms received crying he kissed tenderly (Dec; II:6)
(34c)
… non spaventato dal ricente peccato da lui commesso, not scared by the recent sin by him committed con le mani ancor sanguinose allato le si coricò… with the hands still bloody beside her he laid down (Dec; II:7)
(34d)
… non dimenticata l’impromessa fatta alla moglie, not forgotten the promise made to his wife, disse di tutto … he told anything (Cinq. Alamanni; p. 281)
(34e)
… per la malignità, non uscito quasi sangue, ebbe per certo che for his serious condition, not come-out almost blood, he took for sure egli fusse, come egli era veramente, morto. that he was, as he actually was, dead (Cene; I:5)
204
One should pay attention to the fact that negation is attested in PASC not only in the shape of a simple non, but also together with the negative adverb mai non, as in (34b) (cf. also non…quasi in (34e)). Whereas the former case (non alone) in principle could be a constituent negation, the latter is undoubtedly sentence negation, under the distinction made by Cardinaletti & Guasti (1992b, 2). 2.5.1
Summary
Negation is attested in both MI and CI PASCs, and could be of the sentential kind. In CI, negation of any kind is excluded.
2.6
The Agentive By-Phrase
Belletti reports a contrast between (35) and (36), and argues that 20th century PASCs do not admit an agentive by-phrase together with a lexical DP. 20th century, CI: (35)
Salutata da tutti, Maria lasciò la sala. greeted by everybody, Maria left the room
(36)
*Salutata Maria da Gianni, tutti uscirono dalla sala. greeted Maria by Gianni, everybody left the room
But, there seems to be some disagreement on the status of these data. It is my impression that judgments tend to diverge among speakers along two dimensions, register and aspect. Consider first that PASC generally improve when introduced by the adverbs una volta and appena. In addition to this, when introduced by the by-phrase, PASCs improve the more affected the internal argument is by the event. This was first noticed by Dini (1994). The examples given in (37a)-(37d) summarize the judgments of my informants. (37a)
?Una volta fatti i compiti dagli alunni, once made the homework by the pupils, l’insegnante li correggerà. the teacher will correct them
(37b)
?Appena tradotto l’articolo dal lettore, verrà subito pubblicato. once translated the article by the lecturer, it will immediately be published
205
(37c)
?Appena riparata la macchina dal meccanico, partirò. once repaired the car by the mechanic, I will leave
(37d)
?Una volta arrestato dalla polizia il colpevole dell’omicidio, once arrested by the police the guilty of the homicide, il caso fu riaperto. the case was reopened
If the predicate is unaffecting, the result is also awkward also when PASC is introduced by the appropriate adverbial:7 (38a)
*Una volta vista da Gianni l’amica di Maria, once seen by John Mary’s friend, se n’ è innamorato perdutamente. he fell in love
(38b)
*?Appena conosciuta Maria dai miei genitori, once known Mary by my parents, sono diventati amici. they became friends
Obviously, there is a problem of register involved as well, however this is to be understood more precisely. Bertuccelli Papi (1991, 595) gives the following case as well formed: (39)
Terminati i ringraziamenti e pronunciato finished the thanks and pronounced dal presidente dell’associazione il discorso ufficiale, by the president of the association the official speech, ebbe inzio la cerimonia di consegna dei premi. the ceremony of consignment of the prizes began
(39) is of a formal register, which normally means at a stylistic level closer to ancient use. If the discourse is brought up to a formal, or generally ‘higher’ level, predicates like conoscere or vedere, ‘know’ and ‘see’, also improve:
7
In chapter 6, section 2.5.2., I will assume that predicates such as conoscere and vedere may become delimiting predicates under the right circumstances. It may be, then, that the crucial difference between the predicates of (38) and those of (37) is a difference in agentivity rather than delimitedness.
206
(40a)
Viste dal giudice conciliatore le documentazioni prodotte seen by the judge of peace the documentation produced dalle due parti, la sentenza fu pronunciata. by the two parts, the sentence was pronounced
(40b)
Una volta ascoltati dal giudice i testimoni della difesa, Once heard by the judge the testimonies of the defence, l’imputato fu assolto.8 the accused was freed
For reasons I cannot account for, there seems to be some preference for scrambling of the agentive by-phrase, placing it between the verb and the object; see also the example of Bertuccelli Papi above. The agentive by-phrase is attested in ancient texts, not unexpectedly at this point. Among the 14th century writers it appears to be frequent only in Boccaccio, and in the 16th century, only in Bembo. It is occasionally reported elsewhere. Note the position of the by-phrase, which is generally scrambled, and the predicates used in (41c) and (41f):9 14th-16th centuries, MI/RI: (41a)
…, commendata da tutti la novella di Neifile, ella si tacque … commended by everybody the novel of Neifile, she fell silent (Dec; I:3)
(41b)
La novella da Neifile detta … The novel by N. told[fem.] (Dec; I:3)
(41c)
Ma questo da’ ciciliani conosciuto, subitamente egli But this (=he) by the Sicilians recognized, immediately he e molti altri amici … furono per prigioni dati … and many other friends were given as prisoners (Dec; II:6)
8
It should be remembered however, that there are differences in lexical semantics between ascoltare musica and ascoltare un testimone, roughly, ‘hear music’ and ‘hear a witness’. The latter expression lends itself to a delimited reading more easily than the former. 9 Herczeg (1972, 193-194) gives examples from the prose of Masuccio Salernitano (15th century), and observes that the agentive by-phrase is actually most common with the predicates ascoltare and vedere ‘listen’ and ‘see’ as opposed to what we have seen in Modern Italian.
207
(41d)
Fatta adunque la concession dal soldano a Sicurano, … Made[fem.] thus the concession by the sultan to Sicurano (Dec; II:9)
(41e)
… e recatovi da’ famigliari le sedie … and brought-there by the servants the chairs (Prose; I: I)
(41f)
A che ripostogli da Giuliano che così era; … to what answered-him[cl.] by G. that it was so (Prose; I: I)
(41g)
essi da sedere si levarono, and they stood up, e preso da tutti il passo verso le scale, … and taken by everyone the step towards the stairs (Prose; I: XX)
(41h)
… cedeva a le tenebre perpetue; onde da se stesso chiesta he gave in to the eternal darkness; why by himself asked la estrema unzione, ricevuto cotal sacramento disse: … the extreme unction, received this sacrament he said (LettA; 3)
We must also consider those cases where the agent is expressed with per ‘for’ and not with da ‘by’ (cf. Rohlfs 1969, § 810): (42a)
E poi si levò dritto il detto Bavaro, e letto per lo detto vescovo and then the Bavarian stood up and read by the mentioned bishop in una carta il decreto che a confermazione del papa si costuma, on a piece of paper the decree used to confirm the election of the pope l’appellò il detto Bavaro Nicola papa quinto … he proclaimed the mentioned Bavarian Pope Nicola the fifth (Cron; X: 72)
(42b)
… fu sopra ciò provveduto, it was on this taken measures e fatti per certi ufficiali certi ordini molto forti … and made by certain officials certain orders very harsh (Cron; X: 150)
208
The by-phrase is of course also attested in PASCs with no lexical argument ((43a) and (b) are similar to the modern type (36)): (43a)
(= (48h) below) dalla quale aiutati e sollevati, sono pervenuti ad altissimi gradi … by which helped and promoted, they have reached high positions (Gal: I)
(43b)
Giove, udito simil querele e ragionamenti fastidiosi, Jupiter, heard such quarrels and tiresome reasoning tratto da la collera e aventato dalla stizza, andò … moved by the anger and pushed by the resentment, went … (Zucca; I: ‘Baia Ultima’)
2.6.1
Summary
The agentive by-phrase in PASC, which in the ancient texts corresponds to both da- and perphrases, is attested in all of the varieties, MI, RI, and CI. The use in CI is restricted when there is a lexical argument in PASC. The grammaticality of the by-phrase is then sensitive to semantics in the sense that it improves with affectedness, or perhaps, with agentivity. No such restriction appears to have been operative in RI and MI where we attest the by-phrase in PASCs with lexical arguments regardless of the event type expressed by the predicate.
2.7
Comp Field Elements
2.7.1
Complementizers
PASCs of all the relevant periods can be introduced by a filled Comp. Both transitive and ergative PASCs can host complementizers such as anche se ‘even if’ and benché ‘although’, but not a complementizer and a lexical argument together; see the (a)-examples from Cinque (1990b, 25) and the (b)-examples from Belletti (1990, 99): (44a)
Anche se arrivata in ritardo, Maria non si scusò. even if arrived late, M. did not apologize
(44b)
Benché partita da sola, Maria si divertì moltissimo. although left alone, M. amused herself a lot
(45a)
*Anche se arrivata Maria, tutti si rifiutarono di partire. even if arrived M., everybody refused to leave
209
(45b)
*Benché partita Maria, ci divertimmo ugualmente. although left M., we amused ourselves anyway
Complementizers introducing PASCs are attested also in MI, with the important difference that they can be followed by lexical arguments. (46a) should be compared to (45b): 14th century, MI: (46a)
… dicono li soprascritti savi che, bene che quelli cotali uomini the above mentioned learned that, in spite those men diventati animali, … la mente dentro rimaneva loro umana, … become animals, … their mind inside remained human (Fatti: XXVII)
(46b)
Per ciò che levatesi tutte e lui invitato, For which reason stood-up everyone and him invited, in una fresca corte il menarono … in a fresh court they brought him (Dec; I:10)
2.7.2
Wh-expressions
Modern Italian PASCs do not generally admit a wh-Comp; that is to say, the Comp field of a Modern Italian PASC cannot host relative or interrogative expressions. Neither can an absolute participial clause function as an indirect interrogative (47a), nor can it be introduced by relative elements as in (47b), (47c):10 20th century, CI: (47a)
*Si è chiesto dove andati (gli altri) he wonder where gone (the others)
(47b)
*Ho cercato Gianni, al quale dati i soldi, non avrò più debiti. I have looked for John, to whom given the money, I will have no more debts
10
To some informants (47b) improves, but is still marginal when the relative expression appears in postparticipial position: (i) ??Ho cercato Gianni, una volta dati i soldi al quale, non avrò più debiti. I have looked for John, once given the money to whom, I will have no more debts
210
(47c)
*Siamo andati in spiaggia, dove incontrati degli amici, we went to the beach, where met some friends, ci siamo messi a parlare. we sat down to speak
In (47a), the participial clause would no longer be absolute in a strict sense; here, it would have the grammatical function of an argument. As we shall see, there is some evidence that this was actually possible in the ancient varieties. There is a striking difference between 20th century Italian and MI/RI on this point. MI and RI seem to have allowed for a larger variety of Comp field elements apart from complementizers. Wh-expressions and relative pronouns are frequently attested. In (48c), (d), (e) and (j), the relative expression has the meaning for which reason or something similar. In (48h), we have a relative expression in the function of agentive by-phrase. In (48a), (g) and (i)-(m), dove ‘where’, onde ‘from where’/‘whereupon’ and perché ‘why’ have the function of relative expressions meaning respectively to/in which, whereupon and for which reason: 14th-16th centuries, MI/RI: (48a)
Onde partiti costoro, ritornaimi a la mia opera, … whereupon left they, I went back to my work (VN; XXXIV)
(48b)
per la qual cosa in sul palagio della podestà letta la detta for which reason in the palace of the podestà read the mentioned prosciogligione, e condannato messer Simone … il popolo minuto verdict and condemned master Simone … the people gridò: «muoia la podestà» shouted: «death to the podestà» (Cron; VIII: 8)
(48c)
La quale il giudice menata con grandissima festa a casa sua, … whom the judge taken with great joy to his house (Dec; II:10)
(48d)
La qual cosa molti de’ vicini avanti destisi e levatisi … For which reason many of the neighbors before woken and stood-up (Dec; II:5)
211
(48e)
Per la qual cosa Eva mangiato del frutto proibito, For which reason Eve eaten of the prohibited fruit, e datone ad Adamo, incontanente s’apersero gli occhi loro, and given of it to Adam, immediately their eyes opened e cognobbero che essi erano ignudi… and they recognized that they were naked (Espos; IV: I, 45)
(48f)
I quali raunati a concestoro fu mandato per frate Michele … who gathered to the assembly, brother M. was sent for ‚ (MMin; VIII)
(48g)
E come disse, cosí fu la verità, onde passati tre mesi e tre dì and as he said, so it happened, whereupon passed three months and e l’abate fe caricare la nave … three days and the abbot let load the ship (Navig; 16)
(48h)
[la loro piacevole e graziosa maniera…] [their pleasant and graceful ways…] dalla quale aiutati e sollevati, sono pervenuti ad altissimi gradi … by which helped and promoted, they have reached high positions (Gal; I)
(48i)
presero il camino verso casa, dove arrivati, … si posero a tavola … they took the road home, where arrived, they sat down at the table (RagF; I:Intro)
(48j)
e’ si pensò che fusse qualche gran bacalare, perché trattolo he thought that it was some great scholar, wherefore taken-him[cl.] in disparte, e’ lo prese segretamente a domandar … away, he began secretly to ask him… (RagF; I:2)
(48k)
se n’andò nella Nunziata, dove ritrovato uno amico suo, he went to the Annunziata, where found a friend of his, che tutto il dì usava con lo Abbate … that all of the day stayed with the Abbot (RagF; I:3)
212
(48l)
se ne tornarono in camera; dove trovato acceso un buon fuoco, they returned to the room, where found lighted a nice fire chi attese a rasciugarse, … somebody thought to dry himself (Cene; Introd.)
(48m)
Trovasene alcuni che … fanno nei primi anni troppo il savio, there are those who are too learned in their early years onde venuti in età mandono a effetto il corso della fanciullezza. whereupon come to age they start to put in effect their youth (Zucca; ‘Cicalamento Primo’)
The exact nature of relative expressions as il quale ‘which’, with fledged forms, onde, cui etc. is somewhat unclear in spite of extensive discussion in the literature. It has been suggested that the frequent use of il quale in sentence initial position falls under the general problem of distinguishing between coordination and subordination; see Ghinassi (1971) for the notion of ‘relative parahypotax’. Segre (1963, 211) defends the view that clauses introduced by il quale are often to be considered main clauses;11 see also among many others, Alisova (1967) who by and large adopts Segre’s view. As for Modern Italian, Cinque (1988b, 478 and 1991, 201202) points out that il quale in a restrictive relative can appear only as a relativized oblique argument. Otherwise, the relative pronoun is che (examples from Cinque 1988b, 478): 20th century, CI: (49a) (49b)
*La penna la quale è sul tavolo è mia … the pen that is on the table is mine … La penna alla quale pensavi costa troppo … the pen that you were thinking on costs too much …
Therefore, it may be of some relevance that, in the MI/RI examples reported above, the relativized noun is often oblique. Apart from this, the contrast between (49a) and (49b) is evidently not valid in the ancient texts. Il quale in (50) undoubtedly introduces a restrictive relative clause:
11
«E infine il pronome relativo in principio di proposizione, col valore latino di ET + dimostrativo, diviene un facile mezzo di coordinazione, di cui la prosa del tempo fa un vero abuso.» (Segre 1963, 211)
213
14th century, MI: (50)
Sotto la quale rubrica io trovo scritte le parole under which title I find written the words le quali è mio intendimento d’assemplare in questo libello. which it is my intention to write down in this book (VN; I)
An observation of some relevance is that 14th century Italian has strong verb second tendencies, which roughly means that a main clause with the subject in initial position does not admit for other preverbal material. Consider (51): (51)
i quali per la loro richezza tutti furono cavalieri which due to their wealth all became chevaliers (Cron; 80)
The pronoun i quali is immediately followed by an adverbial clause and the quantifier tutti. The subject position of (51) is the one where the quantifier has been stranded, and the relative expression must then be higher in the structure. This is coherent with the assumption that (51) is indeed a relative and not a main clause.12 A more accurate analysis of the status of these constructions must address the Old Italian Comp field, an issue that we have excluded from our agenda. Still, without entering such a discussion in depth, I assume that pronouns such as il quale, dove, perché can be regarded as Spec C elements introducing relative clauses (even though we cannot claim that this is always true). The crucial point here is that the ancient Italian PASC could host a wh-Comp. This conclusion is supported not only by (48a)-(m) above, but also by examples as (52a) and (52b) where the participial clause has the value of an embedded interrogative: 14th century, MI/RI: (52a)
… senza sapere chi la moglie tolta gli avesse o dove portatala. without knowing who the wife taken from him had or where taken-her[cl.] (Dec; II:10)
12
This is assuming i quali to be in Spec C, the quantifier in Spec AgrS and the adverbial PP in a Topic position.
214
(52b)
… fece dire a Salamone come ella era e perché quivi venuta. he made Salamone say how she was and why here come (Trec; II)
I have chosen to introduce (52a) and (b) among the absolute constructions, although it can be questioned whether they actually belong here. The idea that will become transparent in the analysis is quite simply that (52a) and (b) were grammatical in the ancient grammars for the same underlying reasons as the rest of the constructions discussed in this section. Note also, that judging from the data, the 14th and 16th century grammars do not seem to have any restriction on a phonetically realized argument in a PASC introduced by a Comp element, regardless of whether this is a complementizer or a relative expression. 2.7.3
Summary
To summarize, the CI PASC can be introduced by a complementizer if and only if there is no lexical argument. It can never be introduced by a wh-expression or a relative pronoun. RI and MI PASCs can host complementizers, wh- and relative expressions, and this regardless of whether there are lexical arguments in the clause.
2.8
Null Argument Subjects
Contemporary PASCs are analyzed as control contexts by Belletti (1990).13 Thus, in (53) the subject of the ergative in (53a) and of the transitive in (53b), must be co-referential with the matrix subject. (53)-(55) are all from Belletti (1990, 120-121): 20th century, CI: (53a)
Scesa dal treno, Maria prese subito un taxi. gotten down the train, M. immediately took a cab
(53b)
(Maria venne in treno e …) abbandonato il proprio M. came by train and … quit scompartimento, corse a prendere un taxi. her compartment, rushed to get a cab
Belletti mentions two exceptions to this rule: first, quasi argument pros can be licensed, for example, in weather expressions: 13
For null subjects in PASC, see also Salvi (1986, 39-40).
215
20th century, CI: (54)
Finito di nevicare, partiremo. stopped snowing, we will go
Second, a null argument is also well formed in (55) with an ergative participle, and where the subject is present in context: 20th century, CI: (55)
(Maria è venuta in treno e …) M. came by train and … scesa dal treno, è cominciato uno sciopero di due giorni. gotten down the train, started a two-day long strike
In addition, Manzini (Class, 1994) points out that a null argument pro is possible in a construction such as (56) which, in a sense, is the inverse of (54). The matrix subject is quasiargumental: 20th century, CI: (56)
Una volta partiti, ha smesso di piovere. once left, it stopped raining
The null arguments in question seem to be limited to ergative contexts. In a transitive PASC, the non-lexical subject cannot be disjoint in reference from the matrix. Belletti (1990, 120) gives (57) in contrast to (53b): 20th century, CI: (57)
*(Maria è venuta in treno e …) abbandonato il proprio M. came by train and …quit her scompartimento, è cominciato uno sciopero di due giorni. compartment started a two-day long strike
The comparison between MI and RI on this point is difficult. It is clear that the null subject of PASC in at least MI can be free in reference from the matrix clause, and hence is not a control context. But, it is less clear to what precise extent the ancient grammars are different from CI.
216
First, consider (58) which is comparable to Belletti’s (55) above. The null subject of the ergative appears in the immediately preceding discourse, and the matrix that follows is a passive construction: (58)
E partiti, fu serrata la prigione. and gone away, the prison was closed = ‘As they had gone away, the prison was closed’ (MMin.; X)
Quite often, the subject of the PASC is plural and includes the singular subject of the main clause (cf. Herczeg 1972, 192). The subject in (59a) and (59b) is one of the persons referred to in the participial clause: 14th century, MI: (59a)
E ismontati e assettati a sedere, disse:… and dismounted[pl.] and seated[pl.], (she) said (Vizi; XIV)
(59b)
E a lui rivolti, disse l’uno: … and to him turned[pl.], one [of them] said (Dec: II; 5)
However, this is not the case in (60) where the matrix subject is disjoint in reference from the plural subject of the absolute: (60)
E giunti nel cospetto del giudice, e fattoli la riverenza debita, and arrived before the judge, and made-him[cl.] reverence il giudice conobbe subito messer Gianetto, the judge recognized immediately Gianetto, ma messer Gianetto non conobbe lui… but Gianetto did not recognize him = ‘As they had arrived in front of the judge, the judge recognized Giannetto.’ (Pec; IV: 1)
Somewhat more awkward to the modern speaker is (61), where two absolute constructions precede the main clause; the null subject of piaciuta is co-referential with the object of the gerundival small clause, but absent from the main clause.
217
(61)
Avendo veduta a una festa una bellissima donna …e piaciutagli having seen at a feast a beautiful woman … and pleased[sing.f.] him sommamente, non altrimenti che un giovinetto quelle very much, non differently from a youngster those [le amorose fiamme] nel maturo petto ricevette… [’the flames of love’] in his mature breast (he) received = ‘As he had seen a beautiful woman, and as she pleased him, he received…’ (Dec; I:10)
A null subject, disjoint in reference from the matrix subject and hence not controlled, is obviously allowed in MI with active transitive participles as well as with unergatives when this subject has an arbitrary flavor, someone or people: (62a)
Il conte Ugolino … ordinata la cacciata di giudice, se n’andò fuori di Pisa … the count U., ordered the expulsion of the judge, left Pisa (Cron; VII: 121)
(62b)
Martellino … sopra santo Arrigo fa vista di guerire e, M. on saint A. pretends to be cured and conosciuto il suo inganno, è battuto e poi preso. known his bluff, (he) is beaten and then arrested = ‘As they had understood his bluff (as his bluff had been known), he was beaten’ (Dec; II:1)
(62c)
Egli, fatto dì chiaro, mostrando di venire di più lontano, He, made clear day, showing to come from far away, aperte le porte, entrò nel castello … opened the gates, entered the castle = ‘As they had opened the gates (as the gates were opened), he entered the castle’ (Dec; II:2)
(62d)
E mandato pel santo alla prigione, e tratto fuori de’ ceppi, and sent for the saint in prison, and taken out from the stocks fue menato dinanzi da loro … (he) was brought before them = ‘As they had sent for the saint (as the saint had been sent for) and as he had been taken out of the stocks he was brought …’ (MMin.; VIII)
218
(62e)
Mangiato, prima che da taula si partisseno, Calidonia … parlò … Eaten, before they went away from the table, Calidonia spoke = ‘As they had eaten, …, Caledonia spoke…’ (Nov.re.; IIII)
(62e) was reported above as (31i), and instantiates an unergative predicate with an arbitrary subject disjoint in reference from the matrix. The transitive PASC with arbitrary implicit subject is not disallowed in CI either. Witness examples such as (63a) (63b): 20th century, CI: (63a)
Visto che non viene, è meglio che te ne vada. seen that he does not come, you had better leave = ‘As we have seen/as it is obvious that…’
(63b)
Considerato anche questo aspetto del problema, considered also this aspect of the problem, la soluzione sembra meno facile di quanto pensassimo. the solution seems less easy than we thought
However, expressions such as visto che and other similar expressions could be lexicalized and have the status of a complementizer. This might also go for cases like detto questo, fatto questo ‘said this’, ‘done this’. The implicit subject of these expressions, in the ancient prose, can refer to a specific individual: 14th century, MI: (64a)
Fatto questo, Sibilla lo menò allo ‘nferno done this, S. sent him to the Inferno = ‘As he had done this, she sent him…’ (Fatti; XXV)
(64b)
E dette queste parole eglino udivano boci che gridavano … and said these words they heard voices that shouted = ‘As he had said these words they heard …’ (Navig; 24)
Apart from these cases, there is solid evidence that the transitive and unergative PASC of MI can host a null subject which is neither arbitrary nor controlled from the matrix; hence, a
219
subject pro. There is a clear contrast between the following examples and those judged ungrammatical by Belletti (1990, 121). (65a)
Risposto alla Prudenzia a tutte le sue adomandagioni…disse: -Figliuolo mio … answered to Prudence to all her questions she said: My son … = ‘As I had answered, she said…’ (Vizi; LXIX)
(65b)
Poi il notaio cominciò a leggere i processi overo condannagioni; then the clerk began to read the accusations e letto il prolago, frate Michele stava cheto. and read the introduction, brother Michele remained silent = ‘As the clerk had read, brother Michele remained silent’ (MMin.; XVII)
(65c)
Tornati l’imbasciadori al medico e tutto narrato, fu contento. returned[pl.] the ambassadors to the doctor and everything told(he) was satisfied = ‘As they had told him everything, he was satisfied’ (Nov.re; IIII)
Consider also a slightly more complex case as (66): (66)
Ricevuto per fedele da la Fede Cristiana, received[sing.m.] as a believer by the Christian Faith, e giurate le sue comendamenta, n’andammo a letto and sworn[pl.f.] her commandments, we went to bed = ‘As I had been received … and as I had sworn …, we went to bed.’ (Vizi: XIX)
Here, two transitive participial clauses, the first passive and the second active, precede the finite clause. The null subject of the first participle is co-referential with the null subject of the second, but they are both disjoint in reference from the subject of the main clause. Whereas the change in word order and agreement patterns seems to have taken place by around the middle of the 14th century, the null arguments are attested throughout the 14th century and are reduced only in the 16th. Generalizing on the data, MI on this point means the entire 14th century and RI means the 16th century. Furthermore, this is perhaps the point where 16th century use shows the greatest variation between authors. It is my distinct impression, however, that Renaissance texts largely respect the modern rule. Hence, null
220
subjects of Renaissance PASC are generally not disjoint in reference from the matrix clause. Obvious exceptions are found in Cellini. 2.8.1
Summary
Subject pro is clearly possible in MI PASCs, both with ergatives and transitives. As for RI and CI, it appears as if subject pro is generally more difficult, above all with transitive predicates. Even when a null subject of a non-ergative PASC, disjoint in reference from the matrix subject, is found grammatical, it seems to be so in limited cases as with weather verbs and with certain expressions where it takes on arbitrary meaning. I suggest the correct generalization is that the pro option is fully productive only in MI, but the conclusion must be taken with some caution. This issue will not play an important role in the general analysis.14
2.9
Null Argument Objects
We have thus detected, in MI, a null argument that may be (a) the external argument of a transitive V, and (b) the internal argument of an ergative V, in each case a category marked nominative. It remains to be established whether the internal argument of a transitive predicate can also be null, or more precisely, if there is evidence for an argument pro carrying accusative. The hypothetic case would be of the type in (67), which has not been attested in any of the varieties:15 (67)
*Conosciuta _ (Gianni), ha cambiato il suo stile di vita. known _ (John), he has changed his life style pro=Mary.
Consider (68), where the null element is indeed the internal argument of the transitive verb prendere ‘take’: (68)
14 15
E preso, così fatto, da Marchese e da Stecchi, And taken[sing.m.], in that way, by M. and by S., verso la chiesa si dirizzarono… towards the church (they) went (Dec; II:1)
For subject pro in absolute small clauses in Spanish, see Gunnarson (1995, 154-158). Thanks to Maria Rita Manzini for pointing out the relevance of the absence of this pattern.
221
But, the presence of the agentive by-phrase, da M. e da S., is a clear sign that the participle is passive (compare 2.6. above) and that the pro in question is still a subject pro, carrying nominative. Consider again (69) (= (62d) above): (69)
E mandato pel santo alla prigione, e tratto fuori de’ ceppi, and sent for the saint in prison, and taken out from the stocks fue menato dinanzi da loro … (he) was brought before them’ = ‘As they had sent for the saint (as the saint had been sent for) and as he had been taken out of the stocks he was brought …’ (MMin.; VIII)
The subject of the first PASC is disjoint in reference from the matrix subject. The second PASC tratto fuori de ceppi ‘taken out from the stocks’ has a null argument which is internal, but, as in the previous case, the participle could be as passive as the main clause that follows. In that case, the null argument is still nominative, and we do not have any evidence in favor of a null argument carrying accusative. In conclusion, a null argument in the ancient PASC can most presumably only be nominative, and the internal argument of a transitive V can be null only if passivized.
222
Table 5.1
General Syntactic Patterns of PASC: MI, RI, CI
1. Two lexical arguments 2. V1 word order obligatory 3. Agreement a. (ergative) part.-subj b. (transitive) part.-obj 4. Lexical restrictions a. Unergative (or atelic) V b. The copular V, stato ‘been’ 5. Negation 6. Agentive by-phrase a. with lexical argument b. without lexical argument 7. Comp field elements 7. 1. Complementizers: a. with (lexical) argument b. without (lexical) argument 7. 2. Wh Comp a. with (lexical) argument b. without (lexical) argument 8. Argument null subject a. with ergative V b. with transitive V 9. Argument null object
1300 (MI) + -
1500 (RI) + +17
1900 (CI) -16 +
+ +
+ -18
+ +
+ + +
+ + +
-
+ +
+ +
+19 +
+ +
+ +
+
+ +
+ +
-
+ + -
-
-
16
With exception, perhaps, for pronominal subjects; see 2.1., (2). With exception, perhaps, for pronominal subjects; see 2.2., (13). 18 Non agreement participle-object in RI is largely predominant but not exclusive; see 2.3., (22). 19 The agentive by-phrase in CI is not always possible but sensitive to semantics; see 2.6. (35)-(40). 17
223
Appendix to Chapter 5: The Problem of Punctuation
Before turning to the analysis of the material in terms of generative theory, I believe it is useful to comment upon a more general problem that regards the classification of the data. Modern norms of punctuation, which differ already between modern written languages, were of course not generally valid in the historic varieties. For Old Italian, the problem has been addressed by above all Ageno (1964, ch. 14), and Herczeg (1972, 193) discusses punctuation in PASC. In some of the examples above, it is possible to imagine a different analysis, more coherent with Modern Italian, by quite simply replacing commas. Consider a case like (1): (1)
… e io messigli in una mia cassa senza annoverare, ivi bene and I put[+Agr]-them[cl] in a box without counting, a month later a un mese trovai ch’egli erano quatro piccioli più che non doveano. (I) found that there were four ‘piccioli’ more than there should be (Dec; I:1)
The construction is ungrammatical in CI with the reading suggested by the bold face, that is, if the subject belongs to the absolute clause (io messigli in una mia cassa senza annoverare, trovai … ‘I put them in a box without counting, (I) found …’). It is more natural for a modern speaker, if a comma is added after the subject io (e io, messigli in una mia cassa senza annoverare, trovai … ‘and I, put them in a box without counting, found …’). Considerations of this kind are relevant for many of the cases of the preceding data section of chapter 5. As I mentioned earlier, this insecurity with respect to the proper analysis of absolute clauses in the history of Italian is the reason why a statistic survey of word order patterns is indeed difficult; see section 2.2. of chapter 5. A good part of the material will be of precisely this kind: a subject preceding a participial predicate happens to be subject both of the participial predicate and of the following matrix clause. However, we should not draw the conclusion that deviances in punctuation offer a general explanation for the differences attested between the ancient grammars and the present one. Only a part of the data can be ‘explained’ in this manner and it is quite clear that constructions of the type in (1) were also perfectly grammatical with a reading that makes io belong to the participial clause because we also have seen constructions like (2):
224
(2)
e per comandamento di lei, Dioneo preso un liuto e la Fiammetta and on her commandment, Dioneo taken[m.sing.] a lute and Fiammetta una viuola, cominciarono soavemente una danza a sonare… a viola, (they) began sweetly a dance to play (Dec; Intro)
The same line of reasoning holds for (3): (3)
… e menaronnelo a palagio; dove molti seguitolo che da lui si and they brought him to the palace; where many followed-him[cl.] tenevano scherniti, avendo udito che per tagliaborse era stato who by him considered themselves bluffed, having heard that he had preso, non parendo loro avere alcun altro più giusto titolo a fargli been taken as a thief, not seeming them to have any other more just dare la mala ventura, similmente cominciarono a dir ciascuno da lui reason to make him have a bitter end, all in the same way (they) began essergli stata tagliata la borsa. to say each one by him having been stolen the purse (Dec; II:1)
I have put in bold face what I believe to be the absolute participial clause. There is one uncertainty here. The wh-word dove ‘where’ could actually belong to the main clause dove, …, similimente cominciarono a dir ‘where‚ …, everybody began to say’. If dove belongs to the main clause it would be separated from the rest of the main clause by three absolute clauses; one participial and two gerundival. On the contrary, it can be excluded that the subject molti belongs to the main clause. That is to say, we cannot understand the construction as in (4). (4)
dove molti, … , similmente cominciarono a dire … where many, … , together began to say
This analysis is excluded by the fact that a relative clause belonging to molti is attached to the right of the participial V: molti seguitolo che da lui si tenevano scherniti ‘many followed[participle] him who by him considered themselves bluffed’. Hence, molti in (3) is inside the participial clause.20 A certain part of those complex structures that the traditional analysis refers to as anacoluthon, obviously follow patterns different from modern grammar, but they do so in a 20
And a fair guess would be that also dove is inside it, with the meaning to which rather than where. This is not a necessary conclusion, though.
225
quite systematic fashion and do not escape a principled explanation. Consider (5), noticed by the editor of the Esposizioni, Giorgio Padoan: (5)
Erano i due angeli, quando alla casa di Lot pervennero, The two angels were, when they came to the house of Lot in forma di due speziosissimi giovanetti, li quali in the shape of two beautiful youngsters, which by the Sodomites seen, da’ Sogdomiti veduti, incontanente corsono alla casa di Lot, (they) immediately ran to the house of L., adomandando d’aver questi giovani. asking to have these youngsters (Espos.; XI (36))
The construction is curious for the modern speaker who tends to interpret the relative pronoun li quali as the subject of the finite V corsono ‘they ran’, which of course, it is not. The modern grammar favours an analysis of the kind i quali, visti dai sodomiti, corsono alla casa di Lot ‘who, seen by the sodomites, ran to the house of Lot …’. As the commas of the texts indicate, what we have here is two coordinated main clauses, erano i due angeli ‘the angels were …’ and incontanente corsono ‘they ran immediately …’, and the string li quali da’ Sogdomiti veduti ‘whom by the sodomites seen’ is an absolute clause embedded under the second main clause. The subject of corsono ‘they ran’ is a null pronominal element and as such free in reference. This reading of (5) is coherent with the observation that relative pronouns could generally introduce absolute participial clauses.
226
CHAPTER 6 Participial Absolute Small Clauses in the History of Italian: Analysis
0
Introduction
In the first chapter of part B, we have seen how the general patterns of the absolute participle vary over three centuries. I believe that many surface differences between the historic grammars and the modern one follow from a few simple parametric resettings, in spite of the complexity of the data. Throughout my demonstration in this chapter, I refer the reader to the list of properties given as table 5.1, page 223, which summarizes what I here set out to discuss.1 The first section is dedicated to the analysis of absolute participles given in Belletti (1990). This will be presented in 1.1 and modified in 1.2. I will present my own analysis under 2. Section 2 is divided into the following parts: under 2.2, I will argue that the ancient Italian PASC, in both Medieval Italian (MI) and Renaissance Italian (RI), contained a Tense node, contrary to the PASC in Contemporary Italian (CI). In 2.3, I will claim that the observed differences in word order between the three periods boil down to two underlying differences between MI on the one hand, and RI/CI on the other; namely that (a) V-raising in MI targets AgrS whereas the verb raises higher in the other varieties (2.3.1), and (b) in the MI period, but not in RI and CI, an expletive object pro is licensed allowing for ‘free’ inversion in MI but not afterwards (2.3.2). In (2.4) it will be argued that the MI and RI PASC can be both active and passive, whereas the contemporary PASC is always passive. The consequences of the passive analysis for PASC in CI will be discussed in 2.5. In 3, I will make a brief comparison between the ancient Italian participial constructions and the absolute gerund, and in 4 finally, I will comment on the issue of null arguments.
1
Preliminary Discussion on the Theoretical Issues
Before turning to participial constructions from Medieval and Renaissance Italian, it is necessary to briefly review Belletti’s (1990) analysis of the corresponding constructions in 1
Of course, this does not exclude the possibility that other data are introduced and explained during the demonstration, when they can be considered directly relevant for our discussion.
227
Contemporary Italian. The central claims of her demonstration will be listed in 1.1.1-1.1.2 below. When we have done this, we will comment on some theoretic issues and propose a reformulation and a modification of Belletti’s approach, under 1.2.1.-1.2.2., in order to make it compatible with our framework and more suited to deal with the historic data.
1.1
Belletti’s (1990) Account for Absolute Participles
1.1.1
The Absence of Tense, V-to-Comp, and Rightward Agreement
Belletti (1990) assumes that participial absolute small clauses (PASCs) are CPs. However, they do not include all functional projections of a finite CP. The structure she proposes is (1). (1)
[CP C [AgrP Agr [AspP Asp [VP V]]]]
(1) differs from the finite clause structure advocated by Chomsky (1993) in two ways: First, it has no Tense projection; secondly, it contains only one Agr node.2 Belletti (1990) argues that participial agreement with nominative subjects in modern PASC is the result of this nominal subject being raised to the specifier of Agr, while the participle moves up to the C head of the small clause, which is assumed to contain a Tense feature. verb-raising to C is necessary for Case reasons: no Tense Phrase is present in PASC, and a Tense feature is generated in C. Verb-raising to C, marked [+T], enables nominative Case to be assigned to the DP in Spec Agr under government from C; see Belletti (1990, 89) for the example (2), and (Belletti, 96) for the structure. (2)
Arrivata Maria, Gianni tirò un sospiro di sollievo. arrived Maria, Gianni was relieved
2
Belletti does not use the labels ‘S’ and ‘O’. Given that PASC can be both transitive and ergative, the label, in minimalist terms, could be either ‘S’ or ‘O’.
228
(3)
CP
Comp arrivata i
AgrP
Agr’
Spec DP(S) Mariaj Agr ti
VP
V ti
tj
The participle may also agree with a structural object, as in (4). The question is then whether agreement is realized in the same way, which would be the desirable conclusion, or through some different device. (4)
Conosciuta Maria/me, Gianni ha subito cambiato il suo stile di vita. known Maria/me, Gianni has immediately changed his lifestyle (Belletti 1990, 89)
According to Belletti (p. 108), the S-structure of (4) corresponds to (5), where the object is left in situ: (5)
Comp
CP AgrP
Agr cono sciutai V ti
VP
DP(O) Maria/me
229
Here Belletti introduces the hypothesis of agreement between the head and the complement, a marked Case-marking strategy, which she argues makes accusative available to the object. Belletti signals a problem for this view, namely that (4) as well as (2) instantiates verb-raising to C; witness the equal ungrammaticality of (6) and (7), where Comp is occupied by a complementizer: (6)
*Anche se arrivata Maria, tutti si rifiutarono di partire. (Belletti, 99) even if arrived M., everybody refused to leave
(7)
*Anche se conosciuta Maria, Gianni non è cambiato. (p. 110) even if known M., G. did not change
It seems that the C position must be kept free for the verb in both cases. Belletti resolves this by assuming that verb-raising to C applies in overt syntax for reasons of Case-marking in (6), and in LF for reasons of Case checking in (7) (Belletti, 111). The general conclusion is that the V always must incorporate [+T] in order to check structural Case, accusative as well as nominative, and that checking and assignment of Case are distinct syntactic processes; they coincide in (6) but not in (7). The crucial word order restriction on contemporary PASC, namely the verb first pattern, is thus ultimately derived from the absence of the T head in PASCs; in the absence of T, the verb must move to C in order to assign nominative in overt syntax and to check accusative in covert syntax.
1.1.2
The Effects of the Absence of Tense
On the assumption that Tense is absent from PASC and that the verb - as a consequence must raise to Comp, Belletti derives four other properties of PASC. I will list them as 1.1.2.1 1.1.2.4. 1.1.2.1
Voice in PASC
PASC constructions can clearly be passive, witness (8). (8)
Salutata da tutti, Maria lasciò la sala. (Belletti,116) greeted by everybody, Maria left the room
However, PASC can host an agentive by-phrase if and only if the argument is not phonetically realized:
230
(9)
*Salutata Maria da Gianni, tutti uscirono dalla sala. (p. 105) greeted Maria by Gianni, everybody left the room
The by-phrase is also grammatical when an overt complementizer is added, as in (10): (10)
Anche se salutata da tutti, Maria lasciò la sala. (p. 116) even if greeted by everybody, Maria left the room
Belletti explains the ungrammaticality of (9) by saying that the passive participle absorbs accusative Case and hence, cannot move into C in overt syntax; the cooccurrence in C of accusative and the nominative assigning element [+T] would give rise to a Case conflict (Belletti, 114-116). Therefore, the passive participle must stay in a lower head, Agr. It follows that PASC with a passive participle can be introduced by an overt complementizer as in (10). Since the participle does not reach C, however, it can neither assign nor check structural Case on any argument. The contrast between (9) and (8)/(10) is correctly predicted. 1.1.2.2
Unergative Predicates in PASC
In a similar way, Belletti (pp. 112-113) excludes unergative predicates from PASC: (11)
*Telefonato Gianni, Maria andò all’appuntamento. (Belletti, 89) telephoned Gianni, Maria went to the appointment
As unergative verbs are taken to be assigners of accusative, the corresponding past participle absorbs accusative just as the passive participle of a transitive verb. For reasons already mentioned, a Case conflict then arises when the unergative participle reaches C. As a matter of fact, in Belletti’s model all past participles of transitives and unergatives absorb accusative. This is consistent with the view that they do not reach C in overt syntax; only ergative participles do. The analysis gives rise to interesting questions, to which Belletti pays some attention. How is it that an object can be Case-marked by a transitive participle in (12) and a ‘cognate’ object by an unergative in (13)? (12)
Ho conosciuto Maria. I have known Maria
(13)
Hanno appena telefonato una brutta notizia. (Belletti, 145 f.n. 32) they have just telephoned a bad news
Belletti assumes that in (12), the auxiliary V assigns accusative to the participial ending and that the accusative of the verb therefore remains free to be assigned to the object:
231
Acc. (14) I hav e know n Maria
Acc. The same analysis carries over to (13). 1.1.2.3
Two Lexical Arguments in PASC
Assuming that the participial ending of a transitive V always carries accusative also explains why the modern Italian PASC cannot host two lexical arguments. (15)
*Conosciuto/-a Elena Gianni, … known Helen John, …
In order to assign nominative to the subject, the participle V in (15) must reach the T-feature in Comp. However, since the participle ending carries accusative a Case conflict will arise if the participle reaches C in overt syntax. Thus, the ungrammaticality of (15), as well as the above cases (9) and (11), derives from V-to-Comp. 1.1.2.4
Negation in PASC
Moreover, Modern Italian PASCs cannot host negation: (16) (17)
*Non arrivata Maria, … (Belletti, 94-95) not arrived Maria, … *Non arrivata, Maria … not arrived, Maria …
Belletti accounts for this restriction (p. 140 f.n. 8) simply by stating that PASCs do not host a NegP. Following Zanuttini (1991), who claims that NegP requires the presence of a T head (see also Cardinaletti & Guasti 1992b), Belletti is able to relate the ungrammaticality of (16) and (17) to the absence of Tense in PASC. The general idea behind Belletti’s approach appears to be, quite simply, that all of the central properties of PASC can be derived, directly or indirectly, from the absence of TP. The prohibition on negation derives directly from the absence of Tense, and so also the necessity of V-to-Comp-raising. From V-to-Comp and thus indirectly from the absence of Tense, it follows that unergative predicates, two lexical arguments, and the agentive by-phrase are disallowed in PASC.
232
1.2
Modification of Belletti’s (1990) Analysis
Although I believe that the central claim of Belletti’s analysis - the absence of Tense - is fundamentally correct, I choose not to follow the above outlined account in all regards, mainly for theoretical reasons. Belletti’s (1990) model will here be modified in the following ways: 1.2.1
V-to-Comp and Spec Head Agreement: A Unified Account of Transitive and Ergative PASC
We accept the following two contexts as ‘agreement’ contexts: (18)
[XP [Spec YP] [X’ X] …]
(19)
[X° Y° X°]
X agrees with YP in (18) and X° with Y° in (19); this is to say that we accept Spec head agreement (SHA) and head-head agreement. We exclude agreement between X° and YP and between X° and Y° in (20), head complement agreement (HCA): (20)
[X’ X° [YP Y°]]
Belletti assumes (20) to hold for accusative Case assignment in PASC. Theoretical simplicity favors a unified treatment of the accusative and the nominative PASCs, in the sense that the V can be assumed to raise to Comp in both cases, as argued by Kayne (1989a) and Cinque (1990b).3 Trying to reformulate Belletti’s proposal in terms of the MPLT, we need to reject the HCA hypothesis for two theoretical reasons:
3
The issue is of course also an empirical one. Belletti adds ne-cliticization facts that would suggest that the object in (i) has remained in situ, on the assumption that ne is extractable only from complement positions. (i) Conosciutene [molte_], … (Belletti; p. 104) known of them+cl. many, … (ii) *?Arrivatine [parecchi _], …(Belletti; p. 101) arrived of them+cl. many, … There seems to be some variation among speakers on this point, as Belletti notices in f. n. 31. These intuitions are shared by Anna Cardinaletti but not by Rita Manzini (Class; 1994), Gloria Cocchi and Giuseppina Turano (all p.c.). I have no explanation for this but will follow the conclusion I take to be theoretically warranted. Of course, it cannot be excluded that we are dealing with two distinct grammatical systems.
233
1. Following the minimalist framework of Chomsky (1993), we can no longer make a distinction between overt movement for Case assignment and covert movement for Case checking; movement applies for reasons of checking only. 2. The HCA is theoretically motivated in the following way: «…we can speculate that these [SHA and HCA] are the two agreement configurations made available by UG. More specifically, we can suppose that the fundamental relation regulating agreement processes is just government by the head or by a head projection.»4 In a theory that tries to eliminate head government, it is necessary to reduce these alternatives to one, since the government relation, intended to capture the two mechanisms in one, is no longer assumed. Moreover, whereas the SHA analysis is empirically supported by a variety of facts from Germanic and Romance, the status of the HCA alternative is more obscure; if UG admits it, then we would expect it to be generally available. In order to support the HCA thesis, one might attempt to put forward Old Italian constructions such as those discussed in chapters 1 and 2. Belletti suggests that cases like (21) indicate that HCA was indeed productive during the ancient stages of Italian: (21)
aveva rubati danari (from Machiavelli, cit. Belletti; p. 144, f.n. 28) he had stolen[pl.] money
But, as we have seen, these do not offer compelling evidence in favor of HCA. The phenomenon is highly irregular in Machiavelli, and in fact almost absent in some of his contemporaries. Belletti cites Rohlfs (1969, §725) who, however, does not give information about this clear difference in use between the 14th century and the 16th century (cf. our comment in chapter 1 footnote 2). The most compelling evidence should come from a variety where participial agreement is regular, that is, where participles always agree. For Italian, this means going back to the 14th century. In the relevant period there are strong reasons to suspect that the grammaticality of rightward agreement, as in (21), was in some sense dependent on the possibility of putting the object in pre-participial position. See our discussion in chapters 1 and 2. I maintain that a unified analysis of transitive and ergative PASCs, based on overt V-to-Comp in both cases, is warranted. I assume therefore (5) is not correct, but that the derivation of both is essentially the one indicated in (3).
4
Belletti (1990, 109).
234
1.2.2
V-to-Comp and the Absence of Tense
Any account that assumes Case assignment from C to Spec I will have a problem when head government is taken away from the theory. The problem will of course go far beyond the discussion of Italian absolute clauses, but I will concentrate on these. It might be useful to point out that V-to-Comp (hence, the verb first restriction) could, in principle, be explained without reference to head government. Consider that the verb during the derivation must check its features in two functional heads - namely, φ-features in AgrS and the Tense feature in T - whereas the DP checks nominative in Spec AgrS. If TP is absent from the structure, and T is generated under Comp, it follows that the V must reach Comp in order to make the derivation converge. The DP is triggered to move to Spec AgrS in order to check nominative and stops there, under the assumption that nominative in Italian is checked in Spec AgrS, cf. the Nominative Checking Parameter assumed in chapter 4, 3.2.1. As further movement of DP is not needed, it is blocked by Greed. The crucial portion of structure is given in (22): (22)
CP
Comp Vi
AgrSP AgrS’
Spec DP(S) AgrS ti
V-raising to C follows on general assumptions in the MPLT framework, and there is no need to assume head government from C of Spec AgrS. Unfortunately, though, it is difficult to insist on the correlation between the presence/absence of TP and the necessity of verb-raising to C, and this issue must to some extent be left open in the present work. Furthermore, there is a problem with other absolute constructions, for instance the Italian absolute gerund. Absolute gerunds instantiate V-toComp as they are subject to a V1 condition, similar to PASC. However, they correspond positively to the negation test: (23)
Sono riuscita a farlo tacere non rispondendo alle sue provocazioni. I managed to make him silent not answering to his provocations (from Lonzi 1991, 578)
235
Given the correctness of Zanuttini’s (1991) claim that Negation depends on Tense, the conclusion is that the absolute gerund contains TP. Furthermore, the absolute gerund may include the copular verb, which is inconceivable in the modern PASC: (24) (25)
Essendo stanco, … being tired *Stato stanco, … been tired
This makes sense if the copular V depends on the Tense node (either because the copular is generated under T or because it has to move there). Again, the conclusion is that the absolute gerund has Tense whereas the absolute participle does not. Note also that, unlike PASC, absolute gerunds can host two lexical arguments: (26)
*Conosciuta Elena Gianni, tutti cominciarono a fare pettegolezzi. known Helen John, everybody started to gossip (Bertuccelli Papi 1991, 596)
(27)
Conoscendo Elena Gianni, … knowing Helen John, …
The contrast between (26) and (27) is presumably also related to the presence/absence of Tense, and the conclusion turns out to be the same: absolute gerunds have TP, absolute participles have not. I will return to this issue in 2.2.4. Therefore, if T is absent from participial absolutes and present in gerundival ones, we can explain, for instance, why negation is impossible in PASC and possible with absolute gerunds. However, from this we cannot derive V-to-Comp which is equally forced in both.
1.2.3
The Case Conflict in Comp
I argued above against the idea that have is an accusative Case-assigner. The discussion was of course limited to PAP contexts and to have of possession, but, coherently, I take the same to be true for the tense auxiliary. This assumption, together with the claim that modern Italian PASC instantiates V-to-Comp generally, has immediate consequences for the explanation of (28)-(31) ((28)-(30) from Belletti 1990, 89, (31) from Bertuccelli Papi 1991, 596)
236
(28)
Salutata da tutti, Maria lasciò la sala. greeted by everybody, Maria left the room
(29)
*Salutata Maria da Gianni, tutti uscirono dalla sala. greeted Maria by Gianni, everybody left the room
(30)
*Telefonato Gianni, Maria andò all’appuntamento. telephoned Gianni, Maria went to the appointment
(31)
*Conosciuta Elena Gianni, tutti cominciarono a fare pettegolezzi. known Helen John, everybody started to gossip
The view that (29), (30), and (31) are excluded because a Case conflict arises when the participle, having absorbed accusative, moves to Comp cannot be maintained here. There are two reasons for this: first, we have now assumed that the participle V raises to Comp overtly even in the transitive case, and we therefore cannot admit that such a move would give rise to a Case conflict. Second, the idea that accusative is generally absorbed by the participle ending cannot be maintained under minimalism without some serious difficulties. Belletti’s account of accusative assignment in (12) and (13), repeated here as (32) and (33), is not viable in the minimalist framework: (32)
Ho conosciuto Maria. I have known Maria
(33)
Hanno appena telefonato una brutta notizia. they have just telephoned a bad news
For Belletti, the auxiliary assigns accusative to the participle ending, and the accusative of the V remains free to be assigned to the object, as we have said:
Acc. (34) I hav e know n Maria
Acc. Taking away the notion of head government from the theory, we also take away the possibility of solutions of this kind. If ‘have’ cannot assign accusative to the participial ending, the objects of both (32) and (33) would remain without Case. We could try to say that the participial ending carries accusative only when it is argumental, i.e. only when it also
237
carries an external θ-role; no accusative has been assigned to the ‘-en’ morpheme of an active participle. In that case, no problem arises with (32) and (33), but (30) and (31) do not follow. In addition, there are empirical doubts concerning (29): we have seen in section 2.6 in chapter 5 that the agentive by-phrase is not generally excluded but improves if, for instance, we approach a delimited reading or formulate the phrase at a higher stylistic level (cf. Dini 1994).5 My informants generally distinguish between (36) and (37). (38) is given as grammatical by Bertuccelli Papi (1991, 595) ((35)=(28) and (36)=(29)): (35)
Salutata da tutti, Maria … greeted by everyone, Mary …
(36)
*Salutata Maria da Gianni, … greeted Mary by John, …
(37)
?Appena tradotto l’articolo dal lettore, verrà subito pubblicato. once translated the article by the lecturer, it will immediately be published
(38)
Terminati i ringraziamenti e pronunciato finished the thanks and pronounced dal presidente dell’associazione il discorso ufficiale, by the president of the association the official speech, ebbe inzio la cerimonia di consegna dei premi. the ceremony of consignment of the prizes began
If the contrast between (35) and (36) is to be explained on the assumption of a Case conflict in Comp, the contrast between (36) and (37) does not follow. An explanation in terms of a Case conflict in Comp, due to the concurrent presence of accusative and nominative features, is awkward and I choose to abandon the idea. Thus, the present analysis must offer some device to exclude (30) and (31) on other grounds. Anticipating the following discussion, I will hypothesize that (31) is indeed ungrammatical because of the absence of Tense, and for (30) I will pursue an analysis based on verbal aspect. As for (29)/(36) and (37), I believe the correct generalization is that the agentive by-phrase is not generally excluded from PASC but only in some contexts; my demonstration will be less thorough on this point.
5
Gunnarson (1995, 144) gives a Spanish example of a well formed transitive PASC with agentive byphrase. The example is of a formal register.
238
2
Structure and Derivation of Participial Absolute Clauses in Three Stages of Italian
2.1
General Proposals: Tense, Aspect, Voice, and Object pro
The analysis which will occupy us for the rest of this chapter is based on five general claims, some of which have already been made above (namely three and four): 1. PASC in the historic varieties has a Tense node. Some of the above differences follow from the presence of TP in MI and RI and its absence in CI. 2. There are differences in verb-raising between the three grammars in the sense that the participle raises to AgrS in MI, whereas in RI and CI it raises higher in the structure. 3. The accusative Case feature in the earliest texts is strong which entails that DP(O) moves overtly. This means quite simply that MI shows some signs of a SOV system. Some patterns are explained if accusative checking is overt in MI and covert in RI. 4. As was suggested and argued for in chapter 2, the medieval participial clause hosts some functional head that licenses object pro. The claim distinguishes MI, on the one hand, from RI and CI, on the other. 5. Accusative checking in CI, however, is presumably overt. Given the model outlined in chapter 3, this could follow from differences in aspect and voice between CI and MI/RI. Specifically, if there is always a delimitedness feature in the derivation of contemporary PASC, the delimited DP will always be forced to move overtly. Thus I claim that the contemporary PASC is subject to the Affectedness Constraint, as analyzed in chapter 3, differently from PASC in MI and RI. In addition to this, I will assume that nominative Case checking is always overt.6 The sections below will be organized in the following way: in 2.2. I will focus on the relevance of the first of these claims - the presence of TP in the historic PASC. The structure I assume will be motivated in 2.2.1.-2.2.6. The word order issue will be dealt with in 2.3., where I first concentrate on the MI system with respect to verb-raising (2.3.1.) and DPmovement (2.3.2.). In this latter section, the conclusions from chapter 2 concerning expletive pro (object and subject) will be repeated and further supported. The RI data will be brought up in 2.3.3. In 2.4., we will turn to the issues of Aspect and Voice, discussing, in turn, MI/RI in 2.4.1. and CI under 2.4.2., where I will argue for a passive analysis of the contemporary PASC.
6
Borer (1993, 1995) assumes that nominative is always assigned (which is a way to ensure that clauses must have subjects and hence, to derive the EPP). Note the difference between Borer’s point and mine: in my model, nominative is checked overtly, if there is nominative to check. Trivially, this is not always the case in an absolute participial clause. No element receives nominative in conosciuta me ….
239
2.2
The Internal Structure of Participial Absolute Clauses; the Tense Node
In order to deal with the data presented above, I will work with a structure which is richer than the one assumed by Belletti (1990), relying on independent motivation provided by Kayne (1993) and Cinque (1994), among others. Consider again the structure proposed by Belletti: (39)
[CP C [AgrP Agr [AspP Asp [VP V]]]]
The contemporary participle agrees with its sole lexical argument and hosts one Agr node which is either ‘S’ or ‘O’. MI and RI PASCs could host two lexical arguments. Recall that we continue to follow Chomsky (1993) in assuming two Agr nodes, linked to nominative and accusative respectively. Disregarding Asp for the moment, we conclude from the evidence in chapter 5 that two separate Agr nodes were present in Old Italian PASCs in order to check nominative and accusative. The structure of these in Old Italian must then at least be (40): (40)
[AgrSP AgrS [AgrOP AgrO [VP V]]]
We will return to the question of agreement in sections 2.3. and 2.4. Belletti (1990) argues that PASCs are CPs because they can be introduced by a complementizer. We have seen that ancient varieties of Italian PASCs were no exception and that the presence of Comp field elements was indeed highly frequent. (41)
… dicono li soprascritti savi che, bene che quelli cotali uomini the above mentioned learned say that, in spite of those men diventati animali, … la mente dentro rimaneva loro umana, … become animals, … their mind inside remained human (Fatti: XXVII)
It must therefore be assumed that PASC in the historic varieties is CP, as is the modern equivalent: (42)
[CP C [AgrSP AgrS [AgrOP AgrO [VP V]]]]
We will now see that the above data call for yet an assumption, namely that MI and RI PASCs had Tense, differently from the contemporary PASC. This assumption accounts for some of the differences between Old and Modern Italian PASCs, as will be demonstrated below.
240
2.2.1
Tense and the Wh-Criterion
As mentioned, older Italian varieties permitted a variety of Comp elements, among which whwords and relative expressions. (43)
… fece dire a Salamone come ella era e perché quivi venuta. he made Salamone say how she was and why here come[fem.] (Trec. II)
(44)
La quale il giudice menata con grandissima festa a casa sua, … Whom the judge taken[fem.] with great joy to his house (Dec; II:10)
Whereas in CI only complementizers are allowed, MI/RI PASC can be introduced by a whword. Evidently, in (43) the participial clause is no longer ‘absolute’; it has become argumental and taken on the value of an embedded interrogative. These facts may have to do with the presence of T if, as claimed by Rizzi (1991), Infl (in our case T) is the locus of whfeatures which are raised to Comp, as illustrated in (45) (adapted from Rizzi 1991; we disregard agreement nodes). (45)
CP
C’
Spec Op
TP
Comp
T’
Spec
T [+wh] Raising of T[+wh] to C is forced by the Wh-Criterion which requires that a Spec-head relation holds between the two. In a structure where T is not available, the Wh-Criterion cannot be satisfied. It follows from our hypothesis that the Medieval and Renaissance Italian PASCs obey these principles as does a contemporary finite main clause. The modern Italian PASC, on the contrary, is correctly predicted to be more restrictive on this point, if the T
241
parameter has a negative value; thus, complementizers are allowed but not wh-expressions or relative pronouns. I suggest, moreover, that it is precisely this Tense specification that enables the participial construct (once it is +wh) to function as an argument of an interrogative verb, as in (43).
2.2.2
Tense and Negation
The structure proposed in (45), if valid for PASC in the ancient varieties, entails that negation should have been possible in previous stages of the grammar, following Zanuttini’s (1991) suggestion that Neg depends on the presence of T. As we have already seen, this corresponds to the attested data. (46)
E trattato e messo in assetto col papa e col re Carlo il passaggio and treated and agreed between the pope and king Charles the passage di Cicilia alla primavera vegnente, … , il papa non dimenticato to Sicily the following spring,… , the pope not forgotten lo sdegno preso contro alla parte bianca di Firenze, the resentment against the white part of Florens, non volle che soggiornasse e vernasse invano. did not want him to stay and spend the winter in vain. (Cron; VIII: 49)
I take this to indicate that the structure of MI/RI did indeed contain NegP and, as a consequence, a Tense head. The latter is no longer present in Modern Italian, or alternatively, is still present but in some sense ‘inert’ and incapable of licensing negation. (47)
MI/RI:
[CP C [AgrSP AgrS [TP T [NegP Neg [AgrOP AgrO [VP V]]]]]]
Note also that the negative element in the ancient grammar is presumably of a sentential kind and not an adverbial, according to the distinction made by Cardinaletti & Guasti (1992b). We can tell this because the Neg head non is attested together with the negative adverbial mai (presumably a Spec element) in (48): (48)
… mai non riconosciutala, (…) not ever recognized her[clitic] lei nelle braccia ricevuta lagrimando teneramente basciò. her in his arms received crying he kissed tenderly (Dec; II:6)
242
In the adjectival small clause, where negation is adverbial, non and mai are in complementary distribution (Cardinaletti & Guasti 1992b, 6).
2.2.3
Tense and the Copular V
Recall that the contemporary PASC does not allow the participial form of essere/stare. An absolute construction like (49) is ungrammatical without the support of a gerund as in (50): (49)
*Stato in Germania per dieci anni, … been in Germany for ten years
(50)
Essendo stato in Germania per dieci anni, … having been in Germany for ten years
Structures like (49) are frequently attested in the old prose: 14th century, MI: (51) Martellino, …, stato alquanto, cominciò … M., stood/remained for a while, began a far sembiante di distendere l’uno de’ diti … to pretend to twist one of his fingers (Dec; II:1) 16th century, RI: (52) … e stato el piú tempo della età mia assente dalla patria… and been for most of my life time absent from the fatherland (Dial; Proemio) Even this follows from the absence of a Tense node in modern Italian PASC and the presence of Tense in the older varieties, if the copula requires the presence of T.
2.2.4
A Layered Case Theory
Notice that the comparison between Ancient and Modern Italian data might give support to Watanabe’s (1993) layered Case-theory, or some version of that idea (the reader is referred to his discussion, above all chapter 4, and to Chomsky (1995, 385 footnote 49). In Watanabe’s system, the checking of structural Case in a functional projection must be followed by an additional checking process in a higher one, a process sometimes referred to as validation of
243
Case checking. This is a way to capture the hypothesized relevance of Tense for accusative and the relevance of Comp for nominative; accusative checking in AgrOP requires the presence of T, nominative checking in AgrSP requires the presence of C. Belletti (1990) assumes that the T head is absent in the contemporary PASC, and that the system as a result of this resorts to C for Case checking. If we follow her on this point, the difference between MI and CI absolute participles is then illustrated in (53)/(54): (53)
MI/RI (14/16th c.):
[CP C [AgrSP AgrS [TP T [AgrOP AgrO [VP V]]]]]
(54)
CI (20th c.):
[CP C [AgrS/OP AgrS/O [VP V]]]
Medieval Italian PASC has both T and C (each dominating an Agr node) whereas the modern PASC only C. If the checking of accusative and nominative depend on two distinct layers, T and C respectively, the expected difference between the grammars is that in the medieval PASC two structural Cases could be checked, but in the modern one only one structural Case; either nominative or accusative but not both in the same construction. The prediction is born out, as we have seen in 2.1. of chapter 5.7 14th century, MI: (55) ed Enea presa la lancia, … and Enea taken[fem.sing.] the lance, … (Fatti; LVII) 16th century, RI: (56) Acquistata adunque el re la Lombardia, Conquered thus the king Lombardy si riguadagnò subito quella reputazione … he immediately regained the reputation (Princ. III) 20th century, CI: (57) *Conosciuta Elena Gianni, tutti cominciarono a fare pettegolezzi. known Helen John, everybody started to gossip (Bertuccelli Papi 1991, 596)
7
Rita Manzini’s example (i), given as (2) in chapter 5, does not follow, however: (i) Restituiti io i libri, … returned I the books, … (i) is possible if the subject is pronominal and carries Focus. I have no explanation for this.
244
This account gives rise to a second prediction: as we have mentioned, overt complementizers cannot occur with overt nominative or accusative arguments in the Modern Italian PASC: (58)
Benché partita da sola, Maria si divertì moltissimo. although left alone, Maria amused herself a lot
(59)
*Benché partita Maria, ci divertimmo ugualmente. although left Maria, we amused ourselves anyway (Belletti 1990, 99)
(59) should be compared to (41), repeated here as (60): (60)
… dicono li soprascritti savi che, bene che quelli cotali uomini the above mentioned learned say that, in spite of those men diventati animali, … la mente dentro rimaneva loro umana, … become animals, … their mind inside remained human (Fatti: XXVII)
Recall that the contrast between (58) and (59) is derived from the necessity of V-raising to C in Belletti’s model. If we assume that nominative and accusative in contemporary Italian are checked in Spec Agr after which the participle must raise to C, (59) is also correctly excluded here. Suppose checking of nominative in Medieval Italian takes place in Spec AgrS to which T has raised. Further movement to C is not required and (60) is predicted to be possible. The same goes for accusative checking in MI; accusative checking in Spec AgrO is followed by raising of AgrO to T and the derivation converges. The analysis is straightforward for MI. For RI something else has to be said, and we will return to the issue in 2.3.3. In my demonstration, I have avoided the formalism of Watanabe’s proposal, but I do not exclude its being pursued. In all circumstances, the properties of Italian participial clauses of the relevant periods may be taken as support for the idea of a layered Case Theory in some version.
2.2.5
The Temporal Reference of Nonfinite Verb Forms
In traditional literature on the subject, it is often observed that the temporal value of the ancient absolute constructions had a higher degree of independence in relation to the matrix, finite clause than is the case in modern grammar. This has been emphasized for gerundival absolutes (Herczeg 1972‚ 144-153) and may to some extent be true also for other nonfinite verb forms. The infinitival verbs in (61a) and (61b) have a temporal value that in modern
245
Italian is expressed with a compound infinitive (examples noted by the editors Enzo Esposito (61a) and Antonio Lanza (61b)): (61a)
… e no lli parve mai vedere piú begli occhi che quegli. and it didn’t appear to him see more beautiful eyes than those (vedere = di aver visto= to have seen) (Pec: III; 1)
(61b)
E con questo vedere, preso gran sospetto, tanto fece che … and with this see, had a suspicion, he did so much that … (vedere = avendo visto = having seen) (Trec; XVI)
As for participles, the matter is more delicate. There are of course many cases of ambiguity. The three PASCs in (62) can naturally be understood as referring to three subsequent events: (62)
Venuta l’ora della cena, i preditti fratelli posti come the hour of dinner, the mentioned brothers sat a mensa in nella ditta camera, il Calí entrato in to eat in the mentioned room, the Calif entered nella colonna, Manasse … stimò subito … into the pilaster, Manasse immediately thought (Nov.re; I)
(62) could thus be understood as it was/became time to eat, the brothers sat down at the table, Calí entered into the pilaster and Manasse thought …. Sometimes, the participle can be replaced by a gerund without changing substantially the meaning of the construction: (63)
fu - … - affrontato e ferito d’un pugnale sopra la poppa manca: he was aggressed and wounded with a knife above his left chest onde il poverello, sentitosi ferito, si misse a fuggire… wherefore the poor man, felt-himself[cl.] injured, fled (Cene; I: 5)
Felt himself injured is roughly equivalent to feeling himself injured. But cases like (62) and (63) do not give any decisive indication; they are compatible with a reading consistent with the modern system. Absolute participles may, under certain circumstances, express a succession of events in modern language also (Bertinetto 1986, 266-267). The temporal ambiguity of participles is however more evident in periphrastic constructions, a matter thoroughly discussed in the traditional literature. The temporal
246
ambiguity of (classic and vulgar) Latin, where amatus sum has the double option of being understood I am loved and I have been loved, is maintained in the earliest centuries of Italian (among others, Parodi 1957, Ambrosini 1960/61, Ageno 1964, 1971, Tekavčić 1980). An often cited example is (64): (64)
E vo’ che sappi che, dinanzi ad essi, and I want you to know that before them spiriti umani non eran salvati. other human spirits were not saved (Inf; IV: 62-63)
where non eran salvati ‘were not saved’ means non erano stati salvati ‘had not been saved’. The phenomenon can be taken as a sign that the participle has a Tense specification of its own, which enables it to denote an event anterior to the point of reference (given by the copular verb in the past tense) (Bertinetto 1991, 17-20, 56-57). If contemporary participial phrases do not host a temporal element, they are unspecified for Tense. In this case, the temporal content of the clause hangs on the Tense of the finite verb alone. It follows from my analysis that there should be a difference between the ancient grammars and the contemporary one with respect to the temporal specification of participles. In fact, what we observe is what we expected, in the sense that the participle of MI and RI shows signs of having a temporal feature on its own which makes it temporally independent of the finite V. The claim is furthermore corroborated by the fact that ancient Italian grammar seems to have had a participial NegP in periphrastic constructions also: (65)
… avrebbono non fatta festa, ma il contrario. they would have not made feast, but the contrary (Cron; IX: 181)
In modern Italian, the negation can appear either in sentence initial position (the normal position) or immediately preceding the DP, but not in the position attested in (65)/(66c):8 20th century, CI: (66a) Non avrebbero fatto festa, ma il contrario. not they would have made feast, but the contrary (66b)
Avrebbero fatto … non festa, ma il contrario. they would have made not feast, but the contrary
8
(66b) is acceptable only under the right intonational circumstances, which means that the DP must be focussed. To some of my informants, (66b) is still marginal.
247
(66c)
*Avrebbero non fatto festa, ma il contrario they would have not made feast but the contrary
One must recognize, of course, that a complete understanding of the data depends not only on the Tense of such constructions but also on the Aspect as well as the interaction between the two; this is the program of research pursued by Ambrosini (1960/61). Our major concern here, however, is to elucidate how Tense and Aspect are codified in syntax and what consequences they have for movement and the licensing of various elements, such as negation and wh-expressions. I take it that the evidence is sufficiently solid to support our main point, namely that MI and RI participle clauses host a Tense node. Note finally, that cliticization to the participle is attested (both in MI and RI) in the periphrastic tense as well: 16th century, RI: (67) … avendo propostosi di stare immobile … having decided-si[cl.] to remain immobile (Cene; I:1) In Modern Italian, the clitic must raise and cliticize to avendo. This is yet another piece of evidence in favour of our hypothesis: there is some crucial difference in functional structure between the ancient and the modern participle, which under current approaches to the syntax of clitics could boil down to the presence/absence of a Tense node (cf. Rouveret 1991b, Kayne 1991).9
2.2.6
Conclusion
On the assumption that there is a T head in the MI and RI PASC, but not in CI, we have derived a number of differences between the grammars. The reader is referred to table 5.1. in chapter 5, where the central data from the three stages of Italian are summarized:10 9
Moreover, in the modern grammar the auxiliary be would be selected: essendosi proposto …. I will not discuss the possible implications of this for Kayne’s (1993) hypothesis for auxiliary selection and functional structure in participle clauses. 10 One additional remark is in place: the differences between ancient and modern grammars have been derived on the assumption that Tense is present in the participial clause in the former and absent in the latter, an assumption which is quite sufficient in order to analyze the data I have proposed to explain. Building on some already mentioned studies (cf. Kayne 1993 and Cinque 1994), it may be feasible to say that TP is always present in the participial clause. If this is correct, the parametric difference we have discussed does not concern the presence of T, but rather its properties. I assume the above analysis could then be reformulated in other terms, although our general conclusion, namely that T is indeed relevant for what we have found, holds anyway. See also Gunnarson (1995) for discussion on Tense and absolute constructions in Spanish.
248
1. MI and RI but not CI PASC may have a wh-Comp. This difference is stated as point 7.2. in table 5.1., and it follows from the Tense hypothesis, if wh-features are base generated in T or otherwise depend on the presence of T. 2. Negation is possible in the ancient varieties but not in the modern one. If the presence of negation requires the presence of Tense, the difference is captured on the assumption that MI and RI PASC have Tense whereas CI does not. We have thereby derived point 5 in table 5.1. 3. Two lexical arguments are allowed in MI and RI PASC, but only one in CI. This corresponds to point one in table 5.1. In a ‘layered’ Case Theory, two lexical arguments require the presence of both T and C. In a grammar where only C is projected, only one lexical argument can be Case checked. The absence of T in CI and its presence in MI/RI explains the difference between MI/RI and CI stated as point 1 in table 5.1. 4. If we accept that the co-occurrence of a lexical argument and a filled Comp depends on the presence or absence of Tense, we have shed some light on point 7.2.a in table 5.1, that is, why PASC in the ancient stages could host lexical arguments together with a complementizer, whereas this is excluded in CI.
2.3
The Derivation of PASC
2.3.1
Verb-Raising in CI and MI
I accept without further argumentation Belletti’s (1990) conclusion that the modern Italian ergative PASC instantiates V-to-Comp-raising in overt syntax. Generalizing the conclusion, I assume (contra Belletti, but with Kayne 1989a and Cinque 1990b) that V-to-Comp holds also for the transitive PASC - see above 1.2.1. Let us begin with 14th century grammar. In the case of an overt complementizer or other CP elements, MI word order may still be (Comp-)SVO. The natural conclusion is that the participial verb in Old Italian absolute constructions did not move as far as to C, but stayed in a lower position, which, for concreteness, I take to be Agr (‘S’ in the presence of a nominative subject, ‘O’ if an accusative object alone is present).11 The structure of an MI PASC hosting an ergative V and a nominative subject would thus be (68): (68)
[CP C [AgrSP DP(S)j Vi [TP tj ti [AgrOP ti [VP ti tj ]]]]]
A transitive PASC with an accusative object would have the structure (69): (69)
[CP C [TP T [AgrOP DP(O)j Vi [VP ti tj ]]]]
11
The discussion is limited to the absolute participle; it so happens that the conclusion carries over to absolute constructions in general. In 3, it will be shown that MI did not have V-to-Comp-raising with absolute gerunds either.
249
In the relevant period, the V did not obligatorily appear in first position, but could be preceded, apart from the complementizer, by both DP arguments and prepositional material. (70)
(…) e con festa da lei preso commiato, si partì. and [with joy] [from her] taken farewell, he went away (Dec; I:10)
(71)
… con senno e con ordine l’uficio commesso, … [with sense and with order] the office accomplished (Dec; II:8)
(72)=(70)
[CP PP, PP participle DP(O)… ]
(73)=(71)
[CP PP, PP DP(O) participle… ]
This is of course expected under present assumptions; if the verb did not move as far as to C, but to AgrS, prepositional or adverbial material could appear to the left of the participial V adjoining to AgrS, if adjunction is admitted, or as the specifiers of some recursion of AgrS (cf. Belletti 1990, Cinque 1994, and Kayne 1994 for discussion); in a coherent fashion, I assume the second alternative.
2.3.2
DP-Movement in MI
Since both subjects and objects could appear to the left of V in MI, I suppose that checking of both nominative and accusative was overt in the MI period; this conclusion was also reached in chapters 2 and 4. Notice that agreement with the DP (‘O’ if the verb is transitive, ‘S’ if ergative) was obligatory in MI as in Modern Italian. If I am correct in claiming that AgrO could license object pro (cf. section 4, chapter 2), the inversion property attested in participial complements to ‘have’ should also be present in PASCs. This is indeed confirmed, since both (74) and (75) are frequently attested. (74)
Part.[+Agr]-DP(O), … ‘Written the letters, …’
(75)
DP(O)-Part.[+Agr], … ‘The letters written, …’
(76)
… serrata la cella con la chiave … locked[fem.] the cell with the key (Dec; I:4)
250
VO
(77)
… e queste arme trovate, …, and these arms found[pl.f.] ancora più s’incominciò a meravigliare even more he began to wonder (Dec; II:5)
OV
The structures of (76)/(77) correspond to the trees given in (78)/(79). (78)
AgrOP
AgrO’
Spec proj
AgrO serrata i
VP
V ti (79)
DP(O)j la cella
AgrOP
Spec que ste armej
AgrO’
AgrO trova tei
VP
V ti
tj
The same must be true for nominative subjects in Old Italian ergative PASCs, since Old Italian AgrS licensed null subjects. Consider (80) and (81): (80)
Onde partiti costoro … whereupon left they (VN; XXXIV)
VS
251
(81)
Appresso ciò non molti dì passati, … SV after this not many days passed (VN; XXII)
In these cases, the same analysis will apply to AgrS (other possible projections are left out for reasons of space): (82)
AgrSP
AgrS’
Spec pro j
AgrS partiti i
VP
DP(S) costoro j
V ti (83)
AgrSP
Spec non molti dìj
AgrS’
AgrS passatii
VP
V ti
tj
Other possibilities are also predicted. As can be seen from (84) and (85), reported in chapter 5 as (8a) and (8b), the subject of a transitive PASC can appear to the right of the object (VOS) and between the object and the participle (VSO).
(84)
V O Fatte le comandamenta Made[f.pl.] the commandments (Vizi; XLII)
S la Fede Giudea, …, the Jewish faith, …
252
(85)
V S Vinta la Fede Pagana Won[f.sing] the Pagan faith (Vizi; XLVII)
O tutta la terra d’oltremare …, all the land beyond the sea, …
Both (84) and (85) are expected under our analysis. As indicated, the V raises to AgrS in all cases. In (84) we assume that DP(O) raises overtly, and DP(S) stays in situ: (86)
= (84) by Spell-out: [AgrSP proj Vi [TP tj ti [AgrOP DP(O)k ti [VP DP(S)j ti tk ]]]]]
In (85), on the other hand, the V has raised overtly to AgrS, whereas both DP(S) and DP(O) remain in situ: (87)
= (85) by Spell-out: [AgrSP proj Vi [TP tj ti [AgrOP prok ti [VP DP(S)j ti DP(O)k ]]]]]
As indicated in (86) and (87), arguments that stay in situ are always doubled by an expletive pro that moves in overt syntax. The third pattern attested, SVO, is trivially predicted if all elements move overtly: S in Spec AgrS, V in AgrS, O in AgrO. Alternatively, of course, O could stay in situ being doubled by an expletive pro. Interestingly, we have not found examples of SOV, OSV, or OVS unless one of S or O is a relative pronoun. This is consistent with my demonstration; if there is a nominative subject, the V needs to reach AgrS and the object must then appear to its right, or lower than the V. Summing up so far, the proposed analysis diverges from the most recent version of MPLT (Chomsky 1995) on some crucial points. I have chosen not to eliminate the Agr category from UG. Turning back to our discussion in chapter 2, the correlation between rightward agreement and ‘free’ word order seems to be solid. ‘Free’ word order is the informal indication of the option for a DP to move overtly or stay in situ being doubled by an expletive element. As a consequence, I assume both expletive nominative chains and expletive accusative chains; the minimalist program assumes expletive elements in nominative dependencies only (Chomsky 1995, 341). Checking of both accusative and nominative Case is overt in both MI and CI PASCs, but the MI PASC generally admits for two phonetically realized arguments. We have now derived the difference between CI and MI in point 2 of table 5.1, page 223. The RI pattern remains to be explained.
253
2.3.3
Verb-Raising and DP-Movement in RI
Verb-first word order is predominant in RI. From the mid 14th century onwards, word order in PASC is roughly similar to modern use. Assuming the argument carrying nominative is in Spec AgrS, the verb in RI raises higher than AgrS. On the other hand, we do have examples of pre-posed pronominal subjects: (88)
… ed ella postasi lo’ in mezzo, and she put-herself[cl.] between them buona pezza ragionarono di più varie e diverse materie for a long while they spoke about several things (Cinq. Fortini; p. 331)
Moreover, lexical arguments and Comp elements can co-occur in RI PASC: (89)
se ne tornarono in camera; dove trovato acceso un buon fuoco, they returned to the room, where found lit a good fire chi attese a rasciugarse, … somebody thought to dry himself (Cene; Introd.)
This induces us to believe that verb-raising did not target C after all. The implications of this are rather interesting. Suppose the landing site for pronominal subjects and R-expressions is not the same, but that pronouns target a position higher than AgrS but lower than C:
254
(90)
Comp
CP
XP
X’
Spec Prono minal Subject X
AgrSP
Spec R-Subject
AgrS’
AgrS If pronominal subjects in RI move to Spec X but R-expressions to Spec AgrS, and if the participial verb moves to X, some patterns of RI become more comprehensible. As for the exact nature of X, there are several possibilities coming to mind: X could be part of the Comp field, and in such a case, one might invoke a CAgr node as for instance Platzack (1994b) and Uriagereka (1995). Alternatively, if we assume a split of AgrS into its distinct φ-features (Person, Number), XP could be projected from one such. Spec X would be the landing site for pronominal subjects, and given that pronouns are distinct from R-expressions in the relevant languages by showing person specification, XP could be a Person Phrase. This latter suggestion would be in the spirit of Poletto (1993, 90-97) who cites Ur Shlonsky for the idea of a split of AgrS. The observation that the grammar makes distinction between referential subjects and pronominal subjects is not new. Platzack (1995), and Hulk & van Kemenade (1995) discuss the relevance of pronominal subjects for the scope of verb-raising and the loss of verb second.12 I will leave this issue open, although there will be reason to comment on it again in section 3, below. The hypothesis in (90) also accounts for another of our observations: when RI PASCs have two lexical arguments, word order is VSO (see the examples in (11), section 2.2 in chapter 5):
12
It is possible that the occurrence of pre-verbal pronominal subjects was a trigger for the loss of verb second in Old French. However, note that the historical change in Italian nonfinite constructions goes in the opposite direction with regard to Old French finite clauses: the verb raises to AgrS in MI and to Comp in CI.
255
(91)
V S O saputo il Soderini il meraviglioso ingegnio di mio padre … Soderini the marvellous talent of my father known (Vita; VI)
This pattern is correctly derived if (a) the V raises to X as indicated in (90), (b) the nominative DP raises to Spec AgrS, and (c) the object stays in situ. The last point requires further discussion. In order to explain differences in agreement between CI and RI, we must again address the question of voice and aspect.
2.4
Case, Voice, and Aspect
I will consider the issues of Case, voice, and aspect in the same section. In the model I am developing (I continue to build on the analysis of chapter 3), a treatment of these points in separate paragraphs would turn out to be difficult. The relevant data are those listed as points 2, 3, 4, and 6 in table 5.1 on page 223, and they are derived from the interaction between the «Accusative strength parameter» and the suppression of external arguments.
2.4.1
‘Open Voice’ in MI and RI and the ‘Accusative Strength Parameter’
Beginning with the question of voice, different ideas have been defended on absolute participles in the ancient Italian varieties. In one of the traditional approaches to the problem, Škerlj (1932) argued that the PASC construction was originally passive and that it took on an active meaning at some point of the medieval era. The simple fact that the participle originally agrees, in MI, and then looses agreement in RI, could be seen as an argument in favor of this hypothesis, considering that passive participles normally show agreement: 14th century, MI: (92) … serrata la cella con la chiave … locked[+Agr] the cell with the key (Dec; I:4) 16th century, RI: (93) … gittato sopra una panchettina la tovaglietta … thrown[-Agr] on a bench a towel (RagA: I)
256
However, agreement cannot be used as a criterion for establishing the voice of a construction. In French relatives, the active participle agrees with the relative pronoun, for instance. Moreover, the transitive participle in the modern PASC obligatorily agrees with the direct argument, and Belletti (1990) argues that the Modern PASC is an active construction. The issue of voice is of great relevance in my model as well. My analysis will diverge on empirical and theoretical grounds, from Škerlj’s (1932) account of Medieval and Renaissance Italian on the one hand, and from Belletti’s (1990) account of Modern Italian on the other. On the basis of fairly uncontroversial criteria, we can establish that PASCs of MI and RI can be both active and passive. (94) and (95) are examples of passive voice, since they take an agentive by-phrase.13 (96) and (97) are active, as they have two lexically realized arguments: (94)
…, commendata da tutti la novella di Neifile, ella si tacque … commended by everybody the novel of Neifile, she fell silent (Dec; I:3)
(95)
essi da sedere si levarono, and they stood up, e preso da tutti il passo verso le scale, … and taken by everyone the step towards the stairs (Prose; I: XX)
(96)
Ordinate le schiere de’ due re nel piano … e ciascuno de’ detti ordered the lines of the two kings on the field and each of the named signori ammonita la sua gente di ben fare, … lords admonished his men to do well (Cron; VII: 9)
(97)
saputo il Soderini il meraviglioso ingegnio di mio padre … known Soderini the marvellous talent of my father (Vita; VI)
Škerlj’s (1932) claim that a diachronic development brings the construction from passive to active is problematic in the light of (94)-(97), which show that PASC can be both active and passive in both MI and RI. This circumstance, namely that voice in the ancient varieties is not
13
I thus assume that a construction is passive if the agentive by-phrase is possible. It does not follow, of course, that a construction is active if the by-phrase is not possible. We will return to this briefly in 2.5.1.
257
‘fixed’ but can be either active or passive, is of crucial importance. I will informally refer to this as ‘open voice’. Consider first the active constructions: I have assumed that Medieval, Renaissance and Contemporary Italian are divided by some parametric settings, one of which is the strength of the accusative Case feature (cf. p. 168): The Accusative strength parameter: (98) Acc strong: Medieval Italian Acc weak: Renaissance Italian, Contemporary Italian (and Romance)
Repeating our conclusions from chapters 2 and 4, the parametric shift from strong to weak probably takes place around the first half of the 14th century; word order and agreement patterns change more or less simultaneously in the compound tense and in PASC, as we have already seen. Recall also that the accusative checking head in MI is AgrO (chapter 4, 5.1.), and that the participial V moves only to the Agr node (if there is only a DP(O) to check, the verb raises to AgrO). The accusative strength parameter (98) immediately accounts for three facts: 1. In MI, the accusative-marked DP moves overtly to Spec AgrO. In RI, it stays in situ. Hence, objects can appear to the left of the participle in MI but not in RI. 2. In MI, Spec AgrO is overtly filled either by the DP(O) or by an expletive object pro carrying the φ-features of the DP(O) associate. This is not the case in RI. Therefore, transitive participles agree regularly with their objects in MI but not in RI.14 3. In the RI transitive PASC, we have actually observed some optionality with regard to agreement: (99)
Alla fine dato licenza al compagno, … at the end given[-Agr] farewell to his friend (Cene; I:1)
14
This explains the general pattern. Some exceptions remain. Recall that non-agreeing MI participles are attested in accusative-with-infinitive constructions; see chapter 5, section 2.3: (i) conosciuto tantosto costei esser femina … known[-Agr] soon her be woman (Dec; II: 3) Data such as (i) confirm the hypothesis of Junker & Martineau (1992), according to whom the DP of accusative-with-infinitive constructions is Case checked inside the infinitival clause, that is, not in the AgrOP that corresponds to the participle V conosciuto; see chapter 4, section 4.1. Non-agreement in (i) is predicted under this hypothesis.
258
(100)
e stangata la porta perchè il volpone non scappasse … and barred[+Agr] the door so that the fox could not escape (RagA; I)
(99) and (100) show that RI transitive PASCs may passive. If they are passive, the lexical argument carries (or may carry) nominative. Checking of nominative is by hypothesis always overt; the DP moves overtly and agreement is spelled out. If the transitive PASC of RI is active, the direct argument DP carries accusative which is checked covertly. Agreement is not spelled out in such a case, similarly to the modern compound tense. In this manner, the optionality of agreement in RI receives an explanation. In brief, I assume the DP carries accusative in (99) and nominative in (100).
2.4.2
Contemporary PASC as a Passive Construct
At this point, we run into a problem which might appear serious at first glance: if Contemporary Italian shares with RI the property of having weak accusative, and if CI PASC is an active construct as Belletti claims, contemporary PASC would be identical to the 16th century PASC in not showing agreement. This is however a false prediction as we have seen, since the participle in the contemporary PASC obligatorily agrees with the DP argument. We arrive at the correct predictions if we assume that the derivation of PASC in Contemporary Italian obligatorily involves the delimitedness feature. This in turn would follow if the external argument of PASC is lexically suppressed. Recall the hypothesis of chapter 3: if the external argument (the subject of V) is lexically suppressed, the derivation is safe only if the internal argument becomes subject of V through DP-movement to the specifier of the delimited Asp. Furthermore, we have already assumed that there is a requirement that the argument of a passive V must check that its φ-features correspond to those of the predicate (cf. chapter 4, section 4.3, p. 166). By hypothesis, the checking of matching or pairing, which is independent of Case, takes place in the specifier of an Agr head. Thus, the difference between MI/RI and CI follows if PASC in the modern grammar, as opposed to the ancient ones, is not optionally passive but obligatorily. If the external subject of modern PASC is suppressed, only those structures where the internal argument can be externalized through movement to Spec Asp are acceptable. Therefore, in PASC the delimited aspect node will always be brought into action as in PAP and other constructions discussed above. On this assumption, word order and agreement in CI PASC can be derived from our general hypothesis. This amounts to saying that (101) and (102) are both passive constructions:
259
(101)
Salutata da tutti, Maria … greeted by everybody, Maria…
(102)
Salutata Maria, Gianni … greeted Mary, John …
In Belletti’s (1990) account, only (101) is assumed to be passive. If we assume that the lexical suppression of the external argument has applied in both (101) and (102), some other facts follow straightforwardly: 1. The DP object in CI moves overtly, since it is triggered to do so by the delimitedness feature in Asp. The fact that the transitive participle must agree with its object is also derived if we accept a passive analysis and the proposal that passive arguments must check that their φ-features correspond to those of the predicate. The participle in the modern PASC shows the same agreement pattern as any other passive participle in modern Italian; agreement is furthermore a subcase of the Spec head mechanism. Consider point (3b) in table 5.1. Agreement on the transitive participle with the DP object is obligatory in MI and CI, but not in RI. This state of affairs is now explained: In MI, DP(O) moves overtly to Spec AgrO because accusative is strong and is checked in Spec AgrO - agreement is spelled out. In RI, DP(O) stays in situ because accusative is weak agreement is not spelled out. In CI, if the subject is lexically suppressed, DP(O) is triggered to move to Spec Asp[+del.] as well as to check its φ-features against the participial predicate in Spec AgrO - agreement is spelled out. 2. Consider point 4 in table 5.1. Unergative predicates are excluded in CI, but not in the other time periods. My account has brought us to the conclusion that PASC is a sort of affectedness construction; it enters under our general analysis of delimitedness and DPmovement to Spec Asp outlined in chapter 3. This is an important achievement, since the analysis now explains the lexical restrictions on PASC in contemporary grammar and why these were not valid in the ancient grammars. If the participle in CI PASC carries the delimitedness feature, the construction is incompatible with durative or atelic predicates; PASC in 20th century grammar can be formed from ergatives and telic transitives, and only these. This result has been obtained without reference to a Case conflict in Comp which I consider a step forward. Obviously, unergative predicates such as mangiare or cantare ‘eat’ and ‘sing’ can be made telic through transitivization, i.e. by the adding of a direct argument, in which case they are allowed in the contemporary PASC as well. In addition, there seems to be another strategy at hand, namely the adding of a delimiting adverbial expression, like una volta ‘once’ (cf. Cinque 1990b). I will not make any attempt to formalize this intuition. If MI and RI transitive PASCs had the option of being active constructions, we correctly predict that the lexical restrictions were not operative in those varieties. Since active
260
transitive and unergative predicates are projected with an external argument, the restriction stated as (103) on page 132 was satisfied. Hence, there was no ban on atelic predicates.
2.5
Some Problems and Theoretical Implications of the Passive Analysis
Two further issues remain to be sorted out: 1. Belletti (1990) argues in favor of an active analysis of PASC, and this deserves some comment. 2. The lexical restrictions on PASC are not identical to those on PAP, and the differences must be shown to follow in some way. The problem, I believe, relates to the way we have used the term ‘delimited’, that is, in a way that makes it synonymous to ‘telic’ and ‘affected’ at the same time. The problem was mentioned in the introduction to chapter 3, and we have now reached a point where some distinction must be made. 2.5.1
Belletti’s (1990) Arguments for an Active Analysis of CI PASC
There are five facts about the modern PASC that may support an active analysis (all the following examples (103)-(109) are from Belletti 1990, 104-106): A. Object pronominals can cliticize to the absolute participle, which they cannot in the passive. (103)
Conosciutami, … known-me[cl.]
B. Idiom chunks are normally excluded from the passive but are possible in PASC: (104)
Gianni non riusciva piú a sbarcare il lunario. John was unable to pass the calender = ‘John was unable to financially survive’
(105)
*Il lunario non era piú sbarcato da Gianni. the calender was not passed by John
(106)
Sbarcato il lunario anche in quella difficile occasione, Gianni … passed the calendar also in that difficult situation, John … = ‘financially survived also in that difficult situation, John…’
261
C. Anaphors can be bound in PASC: (107)
Elogiata solo se stessa, Maria restò del tutto isolata. praised only herself, M. remained completely isolated
For Belletti, the external argument of the participial predicate is projected as PRO which binds the anaphor in PASC. D. The agentive by-phrase is ungrammatical in (9), repeated here as (108): (108)
*Salutata Maria da Gianni, tutti uscirono dalla sala. greeted Maria by Gianni, everybody left the room
E. The participle assigns accusative: (109)
Conosciuta me, … known me[Acc], …
Let us begin with the last property, (E). I assumed already in chapter 3, section 3.1., that passive participles maintain the property of assigning (or checking) accusative Case, unlike ergative participles, which do not show this property at all. A close examination of the restrictions on PAP led us to this conclusion. The contrast between (110) and (111) follows straightforwardly, iff DP(O) must check accusative (or receives it) against the participial predicate and not against possessive have: (110)
Ho le valigie fatte. I have the suitcases packed
(111)
*?Abbiamo gli ospiti arrivati. We have the guests arrived
This entails, of course, that the passive participle but not the ergative can check/assign accusative. As for the enclicis of pronouns (A), it is true that this is not possible in the passive, but neither is it possible with the active participle. (112) and (113) are equally excluded: (112) (113)
*Ho fattolo. I have made-it[cl.] *Sono statolo. I have been-it[cl.]
262
Thus, the cliticization of the pronoun does not offer a criterion for establishing passive-hood. Presumably, the contrast between (103) and (112)/(113) is due to the lack of relevant functional structure in periphrastic tenses. Consider also the fact that enclisis of pronouns is possible in reduced relatives as in (114), where a dative clitic has cliticized to a transitive passive participle, cf. Burzio (1986) and Cocchi (1994). (114) is from Cinque (1988b, 499500): (114)
La persona presentatami prima … the person presented-to me[cl.] before…
A solid criterion for establishing whether or not a construction is passivized is presumably not offered by idiom chunks either, (B). Idiom chunks seem to resist, to various degrees, many kinds of transformations, or at least those that separate V from O (For discussion on idiom chunks, see also Ruwet 1983;15 Benincà, Salvi & Frison 1988, 151; and Cinque 1990a, 89): (115a)
??Il lunario, che prima sbarcavo abbastanza bene, the calender, that I managed to pass quite well once non lo sbarco piú da qualche tempo. I do not pass it very well nowadays
(115b)
??Le cuoia, che Gianni aveva già tirato, the leather, that John had already drawn, poco dopo le tirò anche Maria. after a little while, also Mary drew it (tirare le cuoia = ‘draw the leather’ = ‘die’)
Although (115a) and (115b) are both active contexts, they are deviant, obviously for some reason independent of voice. The same must be true for the agentive by-phrase, (D) - the grammaticality or ungrammaticality of a PASC with the by-phrase must to some extent be independent of the question of voice. If we part from the assumption that the by-phrase is only allowed in a passive construction, we must admit that among passive constructions there are some that allow the by-phrase and others that do not; the by-phrase is grammatical in the verbal passive and the nominal passive, but not in the English middle, for instance.16 We have already observed that the by-phrase improves under certain circumstances (see examples (37) and (38), this chapter, and Dini 1994), and this fact indicates of course that the construction tends 15
When it comes to establishing whether this or that ‘transformation’ is possible, Ruwet (1983, 47-48) in particular emphasizes the difficulties linked to the use of idiomatic expressions as criteria. 16 This is true for adverbial phrases in general. For the claim that the by-phrase is adverbial, see Guasti (1992); cf. also Grimshaw & Vikner (1993).
263
toward a passive reading. If the by-phrase shows sensitivity to the aspectual or semantic properties of predicates (as, for instance, delimitedness or agentivity), this will have to be explained independently of the question of voice. The only substantial difficulty for my proposal is (C) - binding of anaphors. One might attempt here to make reference to those who claim that anaphoric binding is not a reliable test for syntactic presence of an argument (cf. Williams 1987); but in doing so, we would seriously undermine the analysis of chapter 3 above. Fortunately, such a move is probably not necessary. Consider that an anaphor (or an anaphoric genitive) can also be bound in a prepositional small clause in Swedish: (116)
Med sig själv/sitt eget namn på listan, gick han till direktören. with himself/his[+a-p.] own name on the list, he went to the director.
Unless we wish to assume PRO inside the prepositional small clause, we are forced to assume that the matrix subject binds the anaphor inside the PP. Suppose the PP, functioning as an adverbial, is generated lower in the structure and is subsequently pre-posed. The matrix subject can then c-command and bind the anaphor at D-structure, or, in the MPLT framework, at LF. If we accept this for Swedish (116), we can also do so for Italian (107): I suggest that the explanation as to why the anaphor can be bound in (107) lies in reconstruction of some kind, as suggested by Kayne (1989a, 100 f.n. 27). The absolute participle, being a sort of sentence adverbial, is base-generated in a lower position and moved to sentence initial position. The matrix subject of (107) could therefore bind into PASC, and there is no need to assume PRO inside the PASC. Consider finally that Belletti’s account and this one may or may not make the same predictions, depending on how we define the notion of governing category or binding domain. The anaphor of (107) must be bound in its governing category or binding domain. For the definition of these notions, see, among many others, Chomsky (1981), Manzini & Wexler (1987), and Cinque (1990a). Suppose that the governing category or binding domain must contain a subject accessible to the anaphor. In Belletti’s active analysis of PASC, the external argument of (107) is syntactically projected as PRO. Thus, PASC has a subject and counts as the binding domain of the anaphor. If, instead, we adopt the passive analysis outlined here, the external argument of (107), the agent, is pre-syntactically suppressed and not projected in syntax. Under such assumptions, PASC in (107) does not have a subject, and is therefore not the binding domain of the anaphor. Under these premises, Belletti’s account and this one equally derive (107). If, however, the definition of binding domain makes reference to the notion of Tense, or root Tense (see Manzini & Wexler 1987 for discussion of this possibility), the binding domain of (107) (and of course also (116)) must be the matrix clause, as we have argued extensively that the modern Italian PASC does not host a Tense node, let alone root Tense. I assume therefore that the binding domain of (107) and (116) alike is the matrix clause.
264
I conclude that the arguments against a passive analysis of the contemporary PASC are not compelling. If the passive analysis is accepted, PASC fits in with our general description of lexical restrictions on participle constructs, and the differences between PASC of the three stages of Italian have received a principled explanation. I suggest this is so and proceed to discuss other implications of the passive analysis and the aspectual interpretation of participles.
2.5.2
Distinguishing PAP from PASC, and Remarks on the Notion of Affectedness
Recall again our discussion in chapter 3: if the external argument of CI PASC has been presyntactically eliminated, an internal argument has to be externalized in syntax by movement into the specifier of a delimited Asp node. A problem now arises in the way we use the term ‘delimitedness’. I have so far avoided a critical discussion of aspectual terminology, but some general observations are now in order. In Tenny (1987), the term ‘delimited’ is intended to replace above all ‘affected’. The notion of delimitedness, so it is argued, comes close to other aspectual concepts like ‘telicity’ or ‘resultativity’, but Tenny does not make any explicit claim that they are identical. Borer (1993, 1995) uses the term delimited as synonymous to telic, or telicity is at least derived from the delimitedness feature.17 As we crucially rely on both Tenny’s and Borer’s analyses (assuming both to some extent) we have implicitly accepted that delimitedness underlies both affectedness and telicity, if not that it actually equals both of these. This has not created any problem for my demonstration until this point, but there are some inconsistencies to straighten out. Following among others Bertinetto (1986, 302), there are numerous predicates of hybrid verbal aspect, that is, predicates that are ambiguous between various aspectual classes depending on the environment in which they appear. Some of these take direct arguments optionally, like sing (cf. above 2.4.2.), whereas the verbal aspect of others is sensitive to sentential aspect (which by the way seems to be a quite normal state of affairs). Some verbs of hybrid aspect are admitted in PASC but excluded from PAP. I will illustrate the point with vedere and conoscere ‘see’ and ‘know’ which we have used to exemplify both constructs:
17
Of course, Borer discusses the projection of argument structure and does not enter into the issue of the Affectedness Constraint.
265
PAP: (117) (118)
PASC: (119) (120)
*Ho il tuo fratello conosciuto. I have your brother known *Abbiamo gli amici visti. we have our friends seen
Conosciuto il tuo fratello, … known your brother, … Visti tutti i nostri amici, … seen all of our friends, …
Although see and know would not normally be telic, I would like to assume that they may acquire telicity when used in a perfective context (this has been suggested by Mourelatos 1978): cf. I saw it/them in the sense I visited it/I met them and I knew her in the sense I made her acquaintance. (121)
Il museo era grande; ci ho messo quasi un giorno a vederlo. The museum was big; it took me almost a day to see it (= to see everything that was in it, to visit it)
(122)
Ha tanti amici a Roma He has many friends in Rome ci ha messo più di una settimana a vederli tutti. it took him more than a weak to see them all (= to meet them)
(123)
Maria è una persona molto socievole Mary is a very sociable person ci ho messo meno di due ore a conoscerla alla cena ieri. it took me less than two hours to know her at the dinner yesterday (= to make her acquaintance)
Understood in this way, see and know correspond to criteria for telicity (cf. Bertinetto 1986, ch. 4.2., esp. 27618): they can appear in a context such as metterci X tempo a _ ‘take X time to _’. When vedere and conoscere are used in PASC, it is typically with the telic reading. 18
Bertinetto expresses a different opinion on the predicate vedere (Bertinetto 1986, 307 f.n. 30). I recognize that my way of reasoning here is less straightforward for a predicate like see than for know. Still, I follow the proposal of Mourelatos (1978, 423-424) that see as well is potentially telic. I have nothing to say about possible differences between Italian and English on this point.
266
(119)/(120) can of course only be understood as I had made his acquaintance and as we had met all of our friends and not as I was knowing him or as we were watching/meeting all of our friends. I assume that the delimitedness feature is operative in both PAP and PASC and that the result is a requirement on telicity. Now, if we take it that the delimitedness feature is responsible for this, we must add something in order to explain why PAP with the corresponding predicates is still excluded, witness the ungrammaticality of (117)/(118). The notion of affectedness, as it has been commonly used in the literature, implies something in addition to telicity; in many cases, it appears that affectedness presupposes a certain degree of concreteness in the noun, and hence also in the event. Going back to the definitions of Anderson, an affected DP «must be changed or moved by the action of the head nominal».19 Notions such as telic or bounded are usually taken to mean that the event has duration and an endpoint and little else. Even though predicates like see or know may acquire telicity or boundedness in the sense that they may have duration and an endpoint, it is difficult to say that their objects are affected by the event expressed. In brief, my cutting of the bread affects the bread, but my seeing the museum does not affect the museum, in any comparable sense. Consider again some of the affectedness constructions, with which we are now familiar: (124)
PAP have (possessive: Swedish and Italian) Ho la macchina riparata. I have the car repaired
(125)
PAP få (causative: only in Swedish) Jag fick bilen reparerad. I got the car repaired
(126)
Fare+infinitive (causative: only in Italian) Ho fatto riparare la macchina. I have made repair the car
Consider that, in these cases, the subject owns or has in its possession the result of a telic event or, alternatively, the subject causes the coming about of this event. Although we are dealing with an abstract notion of possession and causation, it is still feasible that the objects of such predicates must have a certain degree of concreteness, and that the constructions themselves improve with concreteness. This means that we are insisting on a claim that has followed us from the beginning of our discussion in chapter 3, namely that the Affectedness Constraint does not have independent theoretical status, but is the name given to various 19
Anderson (1979, 44).
267
effects of the interaction between aspectual and lexical semantic factors. At the basis of Affectedness lies the concept of delimitedness (which by now means ‘telicity’ in the sense of Comrie 1976 and Bertinetto 1986 or ‘boundedness’ in the sense of Platzack 1979; see 3.3.3. of the Introduction). We must now return to the remaining affectedness constructions, the English middle and the nominal passive. In the derivation of these, there is always a delimited Asp node for reasons discussed above; only predicates that project the delimited AspP are admitted. Predicates like see and know are also excluded when they tend to a telic reading, and therefore some other factors must be involved; in brief, if a telic reading is available for the predicate know, the ungrammaticality of *Mary knows easily is unaccounted for. The reason such predicates are still excluded from the middle and from the nominal passive could be due to the interaction between verbal aspect and sentential aspect, and thereby also to the interaction between verbal aspect and tense. The relevant predicates, know and see, acquire a telic reading when used perfectively but not imperfectively. In other words, I knew Mary can definitely be telic, but a telic reading of I know Mary is more difficult or perhaps, excluded. In Italian, the telic reading ought to be available in conobbi Maria or ho conosciuto Maria but hardly in conosco Maria or conoscevo Maria (see Bertinetto’s 1986 diagram, 298-299). The telic reading in these cases crucially depends on the perfective reading. If the construction in question does not allow for a perfective reading, the relevant predicates cannot acquire telicity and will be excluded. It will be possible for know and see to acquire telicity in PASC, where the perfective reading is obligatory, but not in the nominal passive or the English middle where the imperfective-perfective distinction is neutralized. Recall that the English middle obligatorily has generic tense reading; it cannot be used in the compound tense, for instance: (127)
*These chickens have killed easily …
Summing up, PASC falls under our general description as it is subject to a requirement imposed on it by the delimitedness feature. Delimitedness can now be understood quite simply as telicity, and it is the origin of all constructions where the Affectedness Constraint has been observed. The notion of affectedness itself is derivative as we suspected; the intuition that the object of affectedness constructions must be affected in a more concrete sense - moved, touched or altered - has nothing to do with the syntactic derivation of such constructions, but emerges from lexical semantics, namely from the independent circumstance that some affectedness constructions imply possession or causation. Those affectedness constructions that do not imply causation or possession, the middle and the nominal passive, are peculiar in a different sense, namely that their sentential aspect cannot be perfective. A group of predicates, including see and know, will thereby be correctly excluded from all affectedness contexts except PASC, PASC being the only one that does not imply causation or possession and has a perfective reading.
268
I thus conclude the discussion on AspP and affectedness with the claim that, although many aspects of the problem remain unresolved, the general hypothesis seems promising and future research along these lines is warranted.
2.6
Tense and the Suppression of External Arguments
On the assumption that the modern PASC is passive and furthermore that the external argument is lexically suppressed, we have predicted that the delimitedness feature is operative in PASC, and from this a number of the crucial differences between the modern PASC and the ancient PASCs have been derived. In this paragraph I will comment further on two problems linked to our analysis: 1. My analysis of PASC in CI builds on the assumption that the external argument is suppressed and must be so. If the external argument could optionally be projected, unergative predicates would be grammatical according to my hypothesis. However, the lexical suppression of the external argument, crucial for my hypothesis, has been assumed ad hoc. We would like to know why the external argument must be suppressed in PASC. 2. The data discussed in section 2.2. were derived on an essentially different ground, namely from the absence of Tense in the modern PASC. These two explanations, the suppression of the external argument and the absence of Tense, are in principle independent from one another. However, there may be ways of linking them, making the former derive from the latter. Beginning with the issue in 1 above, the easiest assumption appears to be that lexical suppression of the external argument is always an option. Suppose that, in principle, an external argument can always be syntactically projected or lexically suppressed, and that the outcome depends on independent factors.20 If the external argument is suppressed in the periphrastic participial constructions we discussed in chapter 3, the result is PAP; if the external argument is projected, the result is the compound tense. Suppression of the external argument in an event nominal yields the passive nominal; projection of the external argument yields the active nominal. Suppression of the external argument would in principle be optional in PASC also, but for some reason, the case in which the external argument is projected is ruled out. I have argued at length that the contemporary Italian PASC does not have Tense and that the fact that the CI PASC cannot host two arguments is due to the lack of Tense. In 2.2.4. above, I suggested (following other scholars, e.g. Belletti 1990 and Watanabe 1993) that the checking of structural Case requires the presence of Tense and/or Comp, and that the ban on two lexical arguments in CI PASC derives from the fact that only one of these ‘layers’ is present, namely Comp. Thus, it is natural that either nominative or accusative can be 20
This claim is reminiscent of the rule of arb-assignment of Rizzi (1986a) and Fagan (1988).
269
checked in the modern PASC but not both. Furthermore, I assumed in chapter 3, section 3.4., that the external argument is indeed projected in the verbal passive (following Jaeggli 1986, Baker, Johnson & Roberts 1989, and Hoekstra & Roberts 1993, among others). (128) was given in chapter 3 as (113): (128)
the Verbal Passive:
VP
V’
Spec (DP) V
DP(O)
Whatever happens with the external argument in syntax, it is clear that (128) involves the checking of two structural Cases. Under standard assumptions, accusative is ‘absorbed’ by the passive morphology or assigned to it; in either event, accusative must be checked and we can assume this happens as the passive V raises to the functional head that corresponds to passive morphology. The internal argument in (128) moves to check nominative. If the checking of two structural Cases requires the presence of both T and Comp, it is now clear why (128) cannot be the basic structure of PASC in Modern Italian; structural Case checking is limited to one Case, because of the absence of Tense. Thus, even if the suppression of the external argument is optional in PASC as elsewhere, the derivation of a PASC construction where the external argument is projected will crash for Case reasons. If this is correct, the analysis presented in this chapter becomes considerably more stringent since the hypothesis discussed in 2.4.2. concerning the suppression of the external argument and its consequences, is actually derived from the absence of Tense. The external argument must be suppressed because of the absence of Tense, and when the external argument is suppressed, only a derivation where the delimitedness feature is activated can converge. The last two points of this chapter are dedicated to the comparison between PASC and the absolute gerund. In 3, we will see that there are interesting parallel patterns between the absolute participle and the absolute gerund in the grammars of medieval and renaissance Italian. In 4, I will discuss the issue of argument pro in PASC.
3
Old Italian Gerundival Clauses
I have followed Belletti (1990), Kayne (1989a), and Cinque (1990b), in assuming that the Vto-Comp hypothesis of Rizzi (1982) applies to Modern Italian PASCs. I have argued that
270
Medieval Italian PASCs do not instantiate V-to-Comp raising and that word order patterns are to be analyzed in terms of a pro strategy. Which alternative we choose may of course be relevant not only to participial small clauses, but to non finite absolute constructions in general. Interestingly, gerundival constructions follow the same patterns as the participial ones. Whereas in Modern Italian the V in gerundival absolute constructions must be in initial position, Old Italian admits the word orders SV (129a)/(b) and OV (130a)/(b). (129a)
… e l’abate non vegnendo, … and the abbot not coming (Dec; I:7)
SV
(129b)
… Andreuccio putendo forte, disse l’uno … Andreuccio smelling badly, one of them said … (Dec; II:5)
SV
(130a)
…, quella menando la Lauretta, … that leading Lauretta (Dec; I:concl.)
OV
(130b)
… sé giovane e fresca sentendo … herself young and fresh feeling (Dec; II:8)
OV
When the object is some kind of relative pronoun, we find both OVS (131) and OSV (132). Even if it is possible to assume V-to-Comp in (131), it is more difficult in (132):
(131)
(132)
O(rel) V S La qual cosa veggendo Stecchi e Marchese … Which thing seeing Stecchi and Marchese (Dec; II:1) O(rel) S La qual cosa costor Which thing they (Dec; II:5)
V vedendo, … seeing, …
Trivially, we of course also find constructions showing the modern word order. Crucial differences in word order patterns between Medieval Italian and Contemporary Italian PASCs are thus attested both with participial and gerundival absolutes.
271
The conclusion that Medieval Italian did not have V-to-Comp in absolute constructions seems to be essentially correct. As for the Renaissance pattern, it appears that our general conclusions for PASC hold. This is to say that the gerundival V in RI raises higher in the structure than AgrS (thus, higher than in MI) but not as high as to Comp (it stays lower than in modern grammar). I proposed the structure (90) for RI PASC above, repeated here as (133): (133)
Comp
CP
XP
X’
Spec Prono minal Subject X
AgrSP
Spec R-Subject
AgrS’
AgrS The conclusion that the absolute V in RI raises to X, pronominal subjects to Spec X and referential subjects to Spec AgrS, may be tenable not only for participles but for gerunds as well. Whereas referential subjects are usually found in post-verbal position, pronominal subjects sometimes appear to the left of the gerund: (134)
… ed elli con gran timore tornandovi, and he with great fear turning back, gli diede lo inquisitore ad intendere … to-him gave the inquisitor to understand… (Trec; XI)
(135)
Io gli ricordai e’ mia danari: lui sbeffandomi; I reminded him of my money: he mocking-me[cl.]; a il quale io dissi che … to whom I said that… (Vita; XIV)
272
(136)
… questa ruina, la quale io prevedendo e giudicandola perniziosa this ruin, which I foreseeing and judging-it injurious non solo a lui e agli amici ma ancora alla città… not only to him and to his friends but to the town (Dial; I)
Recall our discussion from chapter 5 section 2.2., and the appendix to chapter 5. We often cannot tell from punctuation whether an item belongs to PASC or to the matrix. In (137), the referential subject Alberto is at the same time subject of the gerund and of the matrix: (137)
… e ‘l detto Alberto essendo a cavallo tra la brigata and the mentioned Alberto being on the horse among sanese e bene armato, scese da cavallo … the senese soldiers and well armed, went down from the horse… (Trec; XIII)
It is natural to analyze (137) as Alberto, essendo a cavallo…, scese ‘Alberto, being on the horse…, went down’. Notice that the examples (134)/(136) on the contrary are ‘safe’ in the sense that the pronominal subjects of (134)/(136) must belong to the gerund; they are disjoint in reference from the matrix subject. In (136), the subject pronoun is obviously lower than the relative expression.
4
Argument pro and the Problem of Recovery
Our discussion on word order in the ancient participle constructions has led us to conclude that both subject and object pro are licensed in the MI PASC. Rizzi (1986a) makes a distinction between the licensing and recovery of pro, in the sense that pro is formally licensed under Case-marking by a given head according to the parameter in (138): (138)
Pro is Case marked by Xy
Rizzi accounts for the recovery of pro in the form of (139): (139) (139a) (139b) (139c)
Recovery of pro according to Rizzi (1986a) expletive pro: no φ-features. quasi argument pro: [+ number] argument pro: [+ number, + person]
273
This means that Case marking (formal licensing) is a necessary and sufficient condition for expletive pro, but necessary and not sufficient for argument and quasi-argument pro, since the latter two also require recovery by the appropriate φ-features. Given this distinction, in saying that subject and object pro are licensed in the 14th century participle clause, I have only predicted that expletive subject and object pro are possible in that grammar. My analysis, so far, remains silent on the possibility of quasi-argument and argument pro. Still, it is obvious that subject pro could be argumental in the finite sentence in both MI and RI, to the same extent as in CI. As for the participial absolute construction under discussion here, we have solid evidence for an argument subject pro in the 14th century with all kinds of predicates: ergative, unergative and transitive. (140)
Subject pro, ergative PASC: E giunti nel cospetto del giudice, e fattoli la riverenza debita, and arrived before the judge, and made-him[cl.] reverence il giudice conobbe subito messer Gianetto, the judge recognized immediately Gianetto, ma messer Gianetto non conobbe lui… but Gianetto did not recognize him ( = ‘As they had arrived in front of the judge, the judge recognized Giannetto.’) (Pec; IV: 1)
(141)
Subject pro, unergative PASC: Risposto alla Prudenzia a tutte le sue adomandagioni … answered to Prudence to all her questions disse: -Figliuolo mio … (she) said: My son … ( =‘As I had answered, she said…’) (Vizi; LXIX)
(142)
Subject pro, transitive PASC: e giurate le sue comendamenta, n’andammo a letto and sworn[pl.f.] her commandments[pl.f.], we went to bed ( =‘As I had sworn …, we went to bed.’) (Vizi: XIX)
According to Rizzi’s recovery mechanism in (139) above, this null subject of 14th century Italian cannot have been recovered by the φ-features contained in the participial agreement. The participle does not carry specification for person, which should be necessary for the
274
recovery of argument pro. As a matter of fact, the hypothesis that subject pro in (140)-(142) is recovered by agreement is clearly excluded for three reasons: First, the participle does not agree with the subject either in the transitive PASC, where it agrees instead with the object, or in the unergative PASC, where it does not agree at all. Second, if participial agreement was indeed sufficient to recover an argument pro, and if AgrO of the MI period actually licensed pro as we have reasons to suspect, argument object pro should have been possible in the transitive PASC of MI. In section 2.9 of chapter 5, we concluded that it probably is not possible. We do not find evidence of the kind exemplified in (143): (143)
Conosciuta _ (Gianni), ha cambiato il suo stile di vita. known _ (John), he has changed his life style pro=Mary.
Accusative marked argument null objects are not attested in PASC,21 and nominative marked argument null subjects are attested regardless of whether the participial V agrees or not. Third, in the 14th century argument subject pro is also recovered with absolute gerunds that do not have inflectional morphology of any kind: (144)
… aspettando la donna che Buondelmonte venisse, waiting the woman that B. would come e non venendo, cominciò aver paura … and not coming, (she) began to fear ( = ‘As he did not come, she began to fear’) (Pec; II; 2)
To summarize, we are left with the following conclusions: 1. The discussion on word order in Medieval Italian participle clauses has led to the conclusion that both AgrS and AgrO license pro; I am assuming both object and subject expletive pro. 2. Argument object pro is not attested, and I take this to indicate that it could not be recovered. 3. Argument subject pro is attested and hence recovered, but not by agreement; argument subject pro is allowed regardless of whether the participle agrees or not. It is also found with other nonfinite verb forms, such as the gerund, that do not carry agreement. 21
Anticipating the following discussion, argument object pro is however attested in MI in a precise and quite different context, namely coordination constructions. This particular case, which will be discussed in chapter seven, does not seem to be directly relevant to the issue of argument subject pro in the participial clause. In chapter seven, I will attempt an explanation in terms of a recovering dependency, rather than recovering agreement.
275
Some alternative recovery strategy must be available. I have no precise claims to make as to the nature of this mechanism or why it is available only to recover a null subject but not a null object. Notice that various approaches to recovery will have the same problem. Cole (1987), developing ideas of Huang (1984, 1989), assumes the existence of two groups of languages: in the first, a Generalized Control Rule (GCR) applies to both PRO and pro, in the second, GCR applies only to PRO. In this second group, where we find Imbabura Quechua, Korean and Thai (Cole 1987), null object pronominals are permitted without identifying agreement on the verb. Cole’s analysis makes explicit the claim that there exists a category of languages where null pronominals need only be Case marked and where no additional requirement on identification is to be met. Under all circumstances, it is unclear why nominative argument pro but not accusative argument pro was allowed in Old Italian. Sigurđsson (1992c), who deals with Old Icelandic, suggests the identification, or recovery condition on pro in (145) for Old Icelandic:22 (145)
pro is identified by free co-indexing with any NP in preceding discourse.
(145) states what seems to be a property of overt pronominals in general (as long as principle B of Binding Theory is respected) and boils down to the claim that there is no particular requirement on identification. In any case, (145) has little to say about our Old Italian data, as the asymmetry between null subjects and null objects remains unexplained. Having said this, we may conclude that the recovery problem, seen in the light of some current approaches, remains unresolved, and we can leave the issue for future research. There is however a circumstance that may be of some interest and deserves comment before closing this chapter. It has been observed that the recovery of argument subjects correlates with the scope of verb movement in another Romance variety, namely Old French. Crucially, null argument subjects were restricted to the inverted, post-verbal position in the earliest stage of French, during the period when this language obeyed the verb second constraint. When verb second is lost, null argument subjects increase; there is a variety of suggestions for how this is to be captured, see for example Adams (1987), Vance (1989, 1995), Rivero (1991), Cardinaletti (1994a, 105-107), Platzack (1995, 216-219), and Hulk & van Kemenade (1995); see also Poletto (1993, 95-96) for the correlation between null subjects and V-raising. All of these accounts build on the assumption that the verb in V2 clauses raises to Comp. When V2 is lost in Old French, and null argument subjects increase, the verb raises to AgrS. If this is correct, and if our conclusions about the ancient Italian PASC are equally correct, we have discovered the same kind of change, proceeding however in the opposite direction with respect to Old French: in MI, namely, null argument subjects are frequently attested in PASC, and in the same period the participle V does not raise as far as to 22
He is followed by Sprouse & Vance (1993) on Old French.
276
Comp, but stays in a lower functional projection that I have taken to be the participial AgrS. At some point between Medieval and Modern Italian, PASC becomes a V-to-Comp construction, and argument subjects are by and large lost. It is of some difficulty to establish when these two changes take place. As we have already stated in section 2.3.3., RI has an intermediate status between the two. In RI, argument null subjects are more restricted than in MI, and it seems that the participle V raises higher than AgrS though not to Comp but to some intermediate functional head, cf. 2.3.3. above. One might be tempted to suggest that the changes in the Italian participle clause have the same source as the changes in the Old French finite clause. In brief, the subject pro-drop option is restricted in grammars where the V raises to Comp. One way to deal with this is to suggest that there are inherent, abstract φ-features present on the AgrS head, sufficient for the recovery of pro though not phonetically realized in spite of what we said above in this section). Similar claims have been made in the literature. Farrell (1990), discussing Brazilian Portuguese (see also Bianchi & Figueiredo Silva 1993 on the same topic), attempts a solution in these terms. In the absence of identifying Object Agreement in Brazilian Portuguese, the claim is that intrinsic, phonetically unrealized features on an Agr head may suffice for identification.23 Once again, the proposal contributes little to our understanding of why null argument subjects are attested, even in the absence of overt identifying agreement features, while null argument objects are not, even in the presence of overt identifying agreement features. Furthermore, Farrells’s claim (p. 344) that pro, in the absence of φ-features on its licensing head, is intrinsically specified for third person does not apply to our case; the null subjects of (141) and (142) are first person, for instance. I leave the discussion at this, convinced that the current approaches to the null argument problem, or at least the cited ones, will not bring us any further.
23
See also Platzack (1992 and 1995, 218), for the idea that C is marked with an abstract feature of person in Icelandic but not in German.
277
Appendix to Chapter 6: A Further Note on the Problem of Voice in Ancient Italian
I have established with an informal terminology that voice is ‘fixed’ in modern language but ‘open’ in the ancient grammars, see section 2.4.1. By this I mean that PASC is obligatorily passive in CI, but could be both active and passive in MI and RI. This is of course a purely descriptive statement; to explain this in terms of the syntactic theory and identify the mechanisms underlying this difference remains a challenge for future research. However, I will give some further justification for this generalization. The difference between the ancient grammars and the contemporary one is striking, and it is not limited to the absolute use of the participle. 1. Recall, first of all, that we detected a similar ambiguity in the avere + participle periphrasis in chapters 1 and 2. It is notoriously difficult in the most ancient texts to distinguish between PAP and the compound tense. The participial predicates in the following two examples from the Convivio (given in chapter (2) as (11a) and (12a)), can both be understood as active or passive: (1)
(…) non veggiono, per ciò che hanno chiusi li occhi (...) they cannot see because they have closed[+Agr] the eyes (Conv; I:IV)
(2)
sì come afferma chi ha li occhi chiusi (...) as he says who has the eyes closed[+Agr] (Conv; II:IV)
Given the evidence presented in chapters 1 and 2 that word order is ‘free’, (1) and (2) may be synonymous. 2. The fact that voice in the ancient participle was not clearly defined, but could be interpreted as either active or passive in a given construction, is also obvious from the attributive contexts, i.e. where the participle appears as the attribute of a noun. We have avoided the issue of attributive participles, but a brief comparison is not out of place here. A participle used as an attributive adjective as in (3)-(5), has passive meaning in modern language:
278
(3) (4) (5)
Una cosa risaputa … a thing known … = ‘a thing which is known (by …)’ Un uomo concosciuto … a man known … = ‘a man who is known’ Un vino bevuto … a wine drunk … = ‘a wine which is/has been drunk (by …)’
The past participle in similar constructions was not clearly defined with regard to the activepassive distinction in vulgar Latin and in ancient varieties of Romance. Rohlfs (1969, § 724) gives the following classic examples: (6)
L:
(7)
L:
Vinum potum … a wine drunk … = ‘a wine which is/has been drunk (by …)’ Homo potus … a man drunk … = ‘ a man who has drunk’
The voice of the attributive past participle was subsequently fixed as passive, but there are numerous exceptions from the ancient prose as well as from (archaic) dialects. Rohlfs (1969, § 724) reports, among other examples, un uomo saputo ‘a known man’ (= ‘a man who knows a lot’), un soldato ardito ‘a dared soldier’ (= ‘ a soldier who dears’), and un frate mal vissuto ‘a badly lived monk’ (= ‘a monk who has lived badly’); Rohlfs ascribes the last example to Machiavelli). (8), from Sacchetti, is reported by Ageno (1964, 292): (8)
… ma egli fu antiveduto e circunspetto. but he was foreseen and cautious (Trec: 148)
In modern standard language, and in the dialects, the active attributive participle belongs mostly to idiomatic expressions, as om navigao ‘a sailed man’ (= ‘a man who has sailed’, Venetian example cited in Rohlfs § 724). Cases of ‘open voice’ (attributive past participles being used both in active and passive sense) are attested both from the 14th and the 16th centuries, i.e. in the same periods for which I have concluded that the participle in PASC is freely active or passive. How these data are to be captured in a generative model remains to be seen. If we view the facts in a broader perspective it seems more than likely that what we have observed are the consequences of the break down of the Latin passive system. Considering that the voice changes in Vulgar Latin and Old Romance coincide with a major upheaval in the temporal and aspectual systems as well, a more general understanding of these phenomena
279
depends on our understanding of how aspect, voice and tense relate to each other, a matter that, at present, appears somewhat obscure.24 A final remark is that this curiosity of voice in the ancient texts was not limited to participles. Ageno (1964, 289-294) gives examples of adjectives having the same ambiguity; she discusses those ending in -abile/-evole ‘-able’ and -oso ‘-ous’.
24
The phenomena under discussion are well-known also in some other languages; see Wheeler (1988, 194) and Ménard (1994, 175-176) for some Romance examples. It is attested in Modern Swedish in expressions that are close to the idiosyncratic register as in en drucken man ‘a drunk man’(= ‘a man who has been drinking’). The fact is also clear in the English construction. For the change in the temporal and aspectual systems mentioned in the text, see Tekavčić (1980) among many others.
280
PART C NULL PRONOUNS AND NOTES ON THE SYNTAX OF ADJECTIVES
CHAPTER 7 Object pro
0
Introduction
In order to explain some syntactic differences between Modern and Old Italian, I have assumed, in previous chapters, that the licensing of a null object pronominal distinguishes the latter from the former. I have argued that this move is theoretically motivated, and perhaps necessary, when the MPLT frame is adopted. In this chapter, the move will be empirically defended on the basis of other data. It will be shown that Medieval Italian had argument object pro in one specific context, namely in coordinated structures. This was a property not only of ancient Italian varieties, but has also been reported from Scandinavian, mostly in ancient or literary contexts. I will attempt a unified hypothesis for the Italian and the Scandinavian data, based on the hypothesis of object shift and object pro that I have already developed in chapters 2 and chapter 6.
1
Preliminary Discussion
1.1
Null DPs and Word Order
Let us repeat briefly the discussion from chapter 2: the inversion property of the subject attested in the Modern Italian finite clause, and the same property of the object in participial clauses in Medieval Italian, can be accounted for in a maximally unified way, in terms of a pro-drop strategy. In the finite clause, the DP subject has an option: either to move to Spec I (or Spec AgrS) in overt syntax in order to check nominative, or to stay in situ in which case the subject position, Spec I, is filled by an expletive pro in overt syntax. In each case, a nominative dependency is created involving Spec I and the complement of V, assuming for concreteness an ergative predicate:
281
(1a)
IP
IP
(1b)
Spec DPi
I’
Spec proi
I’
VP
I
I
V
Compl DPi
V
VP
Compl ti
The analysis of subject inversion illustrated in (1a) and (1b) builds on Chomsky (1981, 1982), Rizzi (1982), Burzio (1986); see also, among many others, Cardinaletti (1994a, 1994b), and section 1.1 in chapter 2 of this thesis. Rizzi’s (1986a) approach to object pro and Kayne’s (1989a) account of participle agreement open the possibility of extending the hypothesis to nonfinite small clauses. In those languages where we find optional word order in participial clauses similar to the so called “free inversion” in the Italian finite clause, we can hypothesize an expletive object pro and, hence, an accusative expletive dependency quite similar to the nominative one: (2a)
(2b)
AgrOP
Spec DPi
AgrO’
Spec proi AgrO
AgrOP
AgrO
VP
V
AgrO’
Comp l DPi
VP
V
Comp l ti
In this thesis, the idea that null pronouns and free word order actually correlate has been followed and pursued in a systematic fashion. We have seen that the analysis summarized in (2a)/(b) is empirically corroborated by the fact that there are languages where participles show rightward agreement, and where the object they agree with can also appear in a position to the left of the participle. My account covers word order and agreement patterns in both periphrastic and absolute participle constructions (see section 4, chapter 2 for the former, and section 2.3.2, chapter 6 for the latter). This account thus allows us to capture the correlation between word order and agreement in a unified hypothesis, quite in line with the proposals of
282
Rizzi (1986a) and Kayne (1989a). Recall however that the subject pro hypothesis for Modern Italian includes null expletive subjects and null argument subjects. So far, I have assumed a null expletive object only; the claim has not been supported by evidence for a null argument object. On the contrary, I concluded in chapter 6, section 4 that participial agreement is unable to recover null pronouns. In this chapter, however, I will discuss some evidence for null object arguments in Medieval Italian. Before turning to the data, I will again consider some general ideas on licensing and recovery of null pronouns (under 1.2); the reader is also referred to the discussion in section 4 of chapter 6. Under 2.1, I will present the Medieval Italian data and in 2.2 and 2.3 these will be compared with similar Scandinavian data. I will turn to the main analysis in 3. Section 4 is dedicated to null objects in some other contexts, arguably operator-bound and hence different from the issue of sections 2 and 3.
1.2
Recovery of Null Pronouns
In section 4 of the preceding chapter, we considered some of the approaches to the recovery of pro. In Rizzi (1986a), a null pronoun is recovered by phonetically realized φ-features on the licensing head; however, we have reached the conclusion that null subjects in the absolute participle construction are not recovered in this way. As for null objects, the languages under discussion here (Germanic and Romance, mostly ancient varieties) do not present evidence of the kind illustrated in (3): (3)
*Gianni ha detestata _. John has detested[fem. sing.] _
Had (3) been attested, it would have given a straightforward argument for a null object pro, recovered by participial φ-features. Unfortunately, the available evidence is much less straightforward than (3). Other scholars consider a null argument to be licit without any requirement of recovery that is locally defined; it may be freely co-indexed with any DP in preceding discourse or pragmatically recovered in some way (e.g. Sigurðsson 1992c). A third, influential idea, defended for instance by Huang (1984, 1989), Campos (1986), and Cardinaletti (1994a), is that null objects which are not recovered through agreement are not pronominal but variables. The proposal, sometimes referred to as the Topic drop or empty Topic approach, builds on the idea that the DP object is moved to an operator position, Spec C, and is dropped from there. The trace of the DP in Spec C A’-binds its trace, thus licensing it as a variable. In the following section, I will present evidence for null objects in medieval Italian that cannot be accounted for on any of these grounds.
283
A number of scholars have argued for the existence of null pronominal objects. Apart from the approach of Rizzi (1986a), see Farrell (1990) for Brazilian Portuguese; in Scandinavian, the claim has been made by Åfarli & Creider (1987) for Norwegian, Rögnvaldsson (1990) and Sigurðsson (1992c) for Icelandic (see also Authier 1992, f.n 2). Since there are often doubts as to the status of null objects, be they variables or pronominal, it is usually of major interest to present evidence of null objects in contexts that do not involve operators, i.e. contexts that are not analyzable as operator-variable dependencies. In this chapter, I will make a distinction between (a) null objects that in my opinion can be considered operator-bound (I will discuss these in section 4) and thus do not give compelling evidence for object pro,1 and (b) null objects that are not operator-bound and hence, cannot be analyzed as variables (this will be the main topic of sections 2 and 3). A solid piece of evidence for null objects not being variables, which has been put forward in the above literature, consists of null objects in coordination structures. I will show in the following pages that Medieval Italian permitted argument null objects in coordination structures, in a way that is quite similar to what has been reported from Old Scandinavian. I will argue that we can deal with them in a satisfactory manner within our general approach to participial syntax. More precisely, a principled account of null objects in coordination constructions can be built on the structures (2a) and (2b) above.
2
Null Objects in Coordination Structures: the Data
2.1
Medieval and Modern Italian
In Modern Italian, an object clitic needs to be repeated when it corresponds to two or more coordinated verbs. Thus, in a construction like (4), the object clitic cannot be left out in the second phrase, witness (5): (4)
Carlo la detesta e la considera una stupida. C. her[cl]-hates and her[cl.]-considers stupid
(5)
*Carlo la detesta e _ considera una stupida. (Calabrese 1988, 553)
For Italian, clitic pronouns in coordinated structures have been discussed by Benincà & Cinque (1993) and Poletto (1993). Poletto (1993, 6) holds that coordination at the Agr° level is possible with the verbs leggere and rileggere, ‘read’ and ‘read again’, distinguished only by the prefix -ri. 1
It is important to point out, on the other hand, that the fact that a null object is operator-bound does not exclude its being pro, see footnote 15.
284
(6)
(Questo libro) lo leggo e _ rileggo. (This book) it[cl.]-(I) read and _ (I) read again
I will henceforth refer to (5) and (6) as null objects in coordination structures (NOC), as opposed to phonetic objects in coordination structures (POC) in (4). It has been pointed out to me that coordinations without repetition of the object clitic may be acceptable in Modern Italian not only in (6) but also under other circumstances. The following example is judged grammatical by Maria Rita Manzini and Giuseppina Turano (both p.c.): (7)
Lo baciai e _ abbracciai. (compare (8d) below) Him[cl.]-(I) kissed and _ (I) embraced
What we are dealing with here is, as in many other cases, a marginal Modern Italian phenomenon that resembles what in Old Italian was a general option, if not actually the rule. In Italian texts from the 13th-14th centuries, coordination of verbs without repetition of the object clitic is the normal case. (8a)
E io vi dico e _ prometto che … and I you[cl.]-say and _ promise that … (Vizi; VI)
(8b)
E ti guarderanno e _ salveranno da’ detti nemici. and you[cl.]-(they) will watch over and _ (they) will save … (Vizi; XI)
(8c)
all’uscio della casa … se n’andò, … e molto il dimenò e _ percosse. to the door of the house he went, and a lot it[cl.]-shook and _ struck. (Dec; II:5)
(8d)
… l’abbracciò e _ baciò … he embraced her[cl.] and kissed _ (Dec; II: 6)
(8e)
… molto la vide e _ ricevette. much he saw her[cl.] and received _ (Dec; II: 6)
285
(8f)
… la videro e _ sentirono … they saw her[cl.] and heard _ (Dec; II:7)
(8g)
… tu mi riprenda e _ corregga aspramente de’ miei difetti. you blame me[cl.] and correct _ harshly for my faults (Fior. III)
(8h)
… d’ogni offesa ch’io ho fatta, for any offence that I have done m’accuso e _ rendo in colpa al mio salvatore… I accuse myself[cl.] and declare _ guilty in front of my savior’ (Fior. VI)
(8i)
… che tu lo vincesti e _ disfacesti that you him[cl.] defeated and _ destroyed. (Trec; V)
This property was shared by the clitic si: (8j)
si possa o _ debba ragionevolmente stimar piú è cosa vana … (it) can itself[cl.] or _ should reasonably consider more a vain thing (Trec; VII)
(8k)
nel vostro mondo giù si veste e _ vela in your world (people) dress themselves[cl.] and _ cover perché fino al morir si vegghi e _ dorma so that until death they[cl.] are awake and _ sleep (Par; III: 99-100)
It must be pointed out, first of all, that the syntactic analysis should distinguish between the general property and the exceptional cases. There is little doubt that POC (4) instantiates the general property of modern Italian, and that NOC, coordination without the object clitic as in (6) or (7), is exceptionally grammatical. It is equally clear that the opposite situation pertains in the texts from the 13th and 14th centuries; in coordination contexts the object clitic is generally not repeated. Some authors alternate between POC and NOC (e.g., Bono Giamboni), whereas in others POC is by and large absent (e.g., Boccaccio). There are also other reasons to suspect that the superficial similarity between the modern Italian examples in (6) and (7), and the Old Italian ones, illustrated in (8a)-(k), is misleading. Whereas the NOC property of (8a)-(k) seems to have been fully productive in the
286
ancient variety, in modern language it is quite sensitive to context. Notice first of all that in Modern Italian, the well formed cases of NOC have at most two coordinated terms. More importantly, the coordinated verbs in (6)/(7), baciare and abbracciare ‘kiss’ and ‘embrace’, leggere and rileggere ‘read’ and ‘read again’, are semantically closely related. If we were to add a third item without repeating the clitic, the acceptability decreases, and especially if the third predicate does not share this semantic relation with the others:2 (9a)
??lo baciaci e _ abbracciai e _ incontrai di nuovo il giorno dopo. Him[cl.]-(I) kissed and _ embraced and _ met again the day after
(9b)
??lo leggo e _ rileggo e _ apprezzo molto (This book) it[cl.]-(I) read and _ read again and _ appreciate a lot
Consider now the following NOC examples from Old Italian: (10a)
… tutte per ordine le m’ insegnò, e _ disse e _ ridisse molte volte… all these [things], them[cl.]-me[cl.]-(she) taught and _ (she) told and _ (she) told again many times3 (Vizi; XIV)
(10b)
… tutte le cose vedute e imaginate All the things seen and imagined si conoscono e _ sentenziano e _ giudicano. ’si’-know and _ consider and _ judge (= ‘they are known and considered and judged’) (Vizi; XI)
(10c)
In inferno si taglia, _ squarta, _ arraffia e _ impicca … In hell ‘si’ [cl.] cut, _ rip, _ tear, and _ hang4 (= ‘in hell people cut, rip, tear, and hang each other’) (Trec.; IV)
In Modern Italian, there may be some lexical reanalysis at work, underlying (6) and (7), that is, a lexical operation that turns the pair of verbs into a sort of compound, projected into
2
A fair guess would be that these coordination structures are more sensitive to the semantic affinity between the predicates than to the number of coordinates. (i) is acceptable for Giuseppina Turano (p.c.), compare (9a) in the text with the following: (i) Lo baciai e _ abbracciai e _ strinsi forte a me. Him[cl.]-(I) kissed and _ (I) embraced and _ (I) clasped to my breast 3 This is presumably a case of coordination of both the accusative and the dative clitic. 4 The construction is impersonal and reflexive and would require two clitics, ci si, in modern Italian. The same example is given in the appendix to chapter 4 as (15).
287
syntax as such. For discussion on this possibility, see Benincà & Cinque (1993, 2319) and Kayne (1994, 62-63):5 (11)
[V leggo e rileggo]
The fact that the coordination possibilities are sensitive to the semantics of the verbs involved suggests that the process is not purely syntactic. It is in fact difficult to insert other material inside the coordinated clauses. Acceptability again decreases if we put in a prepositional phrase: (12)
?? Lo leggo e con molto interesse _ rileggo. it[cl.]-(I) read and with great interest (I) read _ again
This is however attested in Old Italian: (13)
tanto lo stropicciò e con acqua calda _ lavò … so much him[cl.]-(she) brushed and with hot water _ (she) washed (Dec; II:4)
For a description of the restrictions on clitics in coordinated structures, the reader is referred to Benincà & Cinque (1993).6
5
Coordination of two heads in syntax is excluded under the Linear Correspondence Axiom; cf. Kayne (1994, 63). 6 Furthermore, it seems to be the general case in Modern Italian that coordinations without a repeated clitic are symmetric in the sense that the two predicates must have the same tense and φ-feature specification (Benincà & Cinque 1993, 2316-2318). If we make the two coordinated verbs differ with respect to tense, the construction will again call for a clitic in both clauses: (i) ?? Lo baciai e _ abbraccio. Him[cl.]-(I) kissed[pret.] and _ (I) embrace[pres.] (ii) ?? Lo lessi e _ rileggo. it[cl.]-(I) read[pret.] and _ (I) read[pres.] again Again, Old Italian does not observe this restriction, but allows coordination of verbs in different tenses without a second resumptive clitic: (iii) e chiamaronlo e chiamano _ san Ciappelletto. (Dec; I:1) and (they) called[pret.]-him[cl.] and (they) call[pres.] _ san Ciappelletto. (iv) … io nol so né _ seppi già mai. (Dec; II:7) I know it not, and I never knew _ (iv) is not well formed in modern language for Giuseppina Turano (personal communication), who however gives (v) as acceptable: (v) Non lo sa né _ saprà mai. he doesn’t know it, nor will he ever know _ Some other cases of well formed coordination structures where the two conjuncts differ with regard to tense are given by Benincà & Cinque (p. 2317, footnote 4). These share with (v) the fact that the second conjunct is future and the first present. My analysis will not contribute to the understanding of the tense problem.
288
I take the evidence to be sufficiently clear on the crucial point, namely that the phenomena attested in medieval and Modern Italian are essentially different in nature. If productivity itself can be used as a criterion for distinguishing between syntactic and lexical operations (cf. Rizzi 1986a7), the conclusion seems to be that coordination of verbs without repetition of the resumptive pronoun in Medieval Italian, but not in Contemporary Italian, was the result of some syntactic device. This conclusion is strengthened by the sensitivity to the semantics of the coordinated predicates we have demonstrated in contemporary language and the apparent insensitivity to it that emerges from the ancient texts. I will assume this is right and continue.
2.2
NOC in Old and Modern Scandinavian
Consider the following examples (from Åfarli & Creider 1987, 339-341): (14a)
nogen arbeider … som kjørte snej fra gaterne some workers who hauled snow from the streets og hvælvet [e]j i sjøen. and tipped _ into the sea
(14b)
Siðan flutte þeir þorgils líkitj upp með ánni then moved Torgeil and his men the corpse up along the river ok grófu [e]j þar niðr. and buried _ there down
(14c)
Magnús konungr lét gera skrínj og búa [e]j gulli Magnus king had made a chest and decorated _ with gold ok silfi ok setja [e]j steinum. and silver and set _ with stones
It seems as if the phenomenon is mostly attested in Old Scandinavian and varieties that can be defined conservative if not archaic. (14a) is taken from 19th century literary language, and (14b)/(14c) are Old Norse. The phenomenon is quite restricted in modern Scandinavian, but it is possible to find examples in Modern Swedish. In sentences like (15a)/(b):
7
This is the difference between English and Italian null arbitrary objects according to Rizzi (1986a). Arbitrary null objects were however fully productive in previous stages of English, an observation that fits in with my description: «… the observed development follows a rather familiar pattern. What was a fully productive syntactic option in previous stages of the language became a limited lexical option.» (Rizzi 1986a, 533)
289
(15a) (15b)
Han tog boken och han läste den. he took the book and he read it Hon tog äpplet och hon åt det. she took the apple and she ate it
The subject of the second clause can of course be omitted, as in English: (16a) (16b)
Han tog boken och _ läste den. He took the book and _ read it Hon tog äpplet och _ åt det. she took the apple and _ ate it
The object cannot be omitted if the subject is present: (17a) (17b)
?? Han tog boken och han läste _. He took the book and he read _ ??Hon tog äpplet och hon åt _. she took the apple and she ate _
But the coordination is perfect if both subject and object are missing: (18a) (18b)
Han tog boken och _ läste _. He took the book and _ read _ Hon tog äpplet och _ åt _. she took the apple and _ ate _
All of the examples from Norwegian varieties, Old Norse and Icelandic presented by Åfarli & Creider (1987) and Rögnvaldsson (1990) are similar to the Swedish data, as the subject in the second clause is generally missing. Further, in Modern Scandinavian, it seems that the construction is sensitive to the semantics of the second predicate: generally, the second verb of the conjunct is a transitive that admits an absolute use, such as eat or read (cf. I am eating/reading). (19) contrasts with (18a)/(b):8 (19)
*Hon träffade Johan och avskydde _. she met John and hated _
No such restriction seems to have been valid in the ancient literary varieties, and most of the examples reported by Åfarli & Creider and Rögnvaldsson are not grammatical in Modern Swedish, in my opinion.
8
It also seems that the modern coordination is more natural when the first verb of the coordination is taga ’take’.
290
2.3
Similarities and Differences Between NOC in Medieval Italian and in Old Scandinavian
The typical pattern, quite clear in the above examples, is that the second or third clitics are omitted, whereas the first clitic is always present. Further, a lexical subject only occurs in the first conjunct. Notice also that we are dealing exclusively with coordination of verbs in the simple tense. We cannot have a compound tense in the second conjunct in Modern Swedish, nor in Italian, and this is not attested in the ancient texts: (20) (21)
*Han tog boken och _ har läst _. he took the book and _ has read _ *Lo vidi e _ ho rivisto. I saw him and have _ seen again
There is a difference of crucial importance concerning the nature of the object. In Scandinavian, the object is a DP presumably left in situ. In Italian, the object is a clitic. The (near) absence of clitic objects in Scandinavian makes it impossible to verify whether the Italian type of construction was possible there, but the Scandinavian type of NOC ought to have been possible in Old Italian. (22) confirms this:9 (22)
Amai tua figliuola e amo _ e amerò _ sempre … I loved your daughter and love and will always love (Dec; II: 6)
(23a) and (b) instantiate coordination of participles, where the lexical object is the complement of the first conjunct: (23a)
E noi avemo rifatta la cittade e raforzata _ … and we have remade the town and reinforced (Nov.no; 131)
(23b)
… avendo già ragunati molti compagni e ricevuti _ all’Ordine… having already gathered many companions and received _ to the order (Fior. XV)
9
This example is reported in Benincà (1993, 255 footnote 7). Contrary to what Benincà suggests in the note, the analysis I propose for these cases does not presuppose that verbal inflection had anything to do with the possibility of null objects as those discussed in the text. It is however quite possible that the phenomenon we are studying has some relevance to the omission of resumptive pronouns in Clitic Left Dislocation structures, typical for Old Italian; see Marcantonio (1976), Benincà (1994, 1995).
291
3
Analysis
3.1
Aims
The phenomenon of NOC appears to have been generally wide spread in Old Italian and Old Scandinavian varieties and to have been largely lost in the modern ones. It is tempting to look for a unified solution, saying that the changes in the Scandinavian and the Italian grammatical systems are essentially the same. If this is possible to pursue, we would have made a step towards a better understanding of the general syntactic change from ancient to contemporary Romance and Germanic. The points we should account for are stated in (24): (24)
3.2
1. The account must capture the fact that the phenomenon is attested both with clitic pronouns (in Italian) that are raised to some functional projection, and DPs (in Scandinavian) that presumably are left in situ. 2. The account must explain why the subject can be present only in the first clause and not in the following ones. 3. We must explain why the object, clitic or DP, is always present in the first clause and not in the following ones; why do we not find structures where the object is missing in the first clause and phonetically realized in the second, for instance? 4. The account must capture why there is a general ban on compound tenses, as illustrated above, cf. (20)/(21).
Some Hypotheses that could be relevant for NOC
I will begin by excluding some possible approaches that come to mind; first, it will be shown in 3.2.1. that an ellipsis approach such as that of Wilder (1994) is not adequate; second, we will reject the Topic-drop solution in 3.2.2.; third, I will argue in 3.2.3. that the explanation of NOC in Old Italian cannot build on the assumption that the relevant pronouns were phrasal and not clitic - this means that the analysis advanced by Rivero (1992) for Old Spanish pronouns is not viable for Old Italian. My own analysis will be outlined under 3.3. 3.2.1
The Ellipsis Approach to Coordination (Wilder 1994)
I will here build on Kayne (1994) and Wilder (1994). The latter approach is viable in the Antisymmetry format. Wilder argues for an ellipsis-based account for coordination structures. A distinction is made between forward and backward ellipsis, exemplified in (25)/(26) (from Wilder 1994, 23):
292
(25)
Forward deletion: Mary came in and _ sat down.
(26)
Backward deletion: John bought _ and Sue read the day’s newspaper.
For obvious reasons, the forward deletion case will typically concern subjects and backward deletion objects, though not exclusively. Forward deletion can target objects according to Wilder, but an example is given where the object is deleted together with the whole of VP. (27)
John bought a book for Mary, and Pete _ for Sue. (Wilder, 33)
In Wilder’s terminology, NOC of Old Scandinavian and Italian would be characterized as forward deletion with an object antecedent. This is however a case that Wilder explicitly excludes by imposing the condition of (28) (Wilder, 33): (28)
Head Condition on Forward Deletion. An ellipsis site may not be c-commanded by an overt (non-deleted) head in its domain (= conjunct).
The condition of (28) will exclude (29) and our cases of NOC on the same ground, namely because there is a non-deleted head in the second conjunct, the V, that c-commands the ellipsis site: (29)
*John bought the book and Mary read _. (Wilder, 33)
(30)
E noi avemo rifatta la cittade e raforzata _ … and we have remade the town and reinforced _ (Nov.no; 131)
Wilder’s account rules out (29) and (30) correctly, as the result is indeed awkward in the contemporary grammars under discussion. I take it therefore that his account is correct, and I conclude that NOC in ancient varieties of Scandinavian and Italian is not a case of ellipsis in Wilder’s sense.
3.2.2
The Topic-Drop Analysis
One possibility would be to link the observations made here to verb-second, building on the following intuition: the phenomenon of NOC may be an instantiation of Topic drop or, in any case, require that Spec C remains free. If V2 grammars have V-to-Comp-raising in finite clauses, the V would not stop in Infl but move on to C. In order for an A’-chain to be created,
293
the specifier of the same C must then be left free, that is, it cannot be occupied by the subject. There are problems though: if we are dealing with coordination at the CP level, and if it is assumed that the subject is projected, we must explain why the subject cannot be realized in the inverted position as is usual in V2 grammars. A construction such as (31) is decidedly ungrammatical in Swedish and is not attested in any of the varieties under discussion: (31)
**Han tog bokeni och [e]1i läste han [e]2i he took the book and read he
[e]2 is the basic position of the object, [e]1 is the topic position, Spec C. Given the ungrammaticality of (31), it is plausible that the subject in NOC is projected only once. Topicalization of a null object could be assumed in (32), where also the lexical object of the first conjunct is also topicalized (cf. Rögnvaldsson 1990, for Icelandic examples): (32)
?Bokeni tog han och [e]1i läste han [e]2i the book took he and _ read he
(32) is not perfect in my opinion, but requires subject inversion in both conjuncts, in order to be acceptable. What we are dealing with in (32) is certainly CP coordination. It is quite possible that the second Spec C (the lower one) is occupied by a null topic [e]1, which licenses [e]2 as a variable. The difference between (32) and the NOC data we have seen in 2.1./2.2. is, I think, quite clear.10 It also needs to be pointed out that the parallel between verb second and NOC is not perfect, since Italian has lost V2 whereas Modern Scandinavian languages maintain this property. (33) Old Scandinavian Old Italian Modern Swedish Modern Italian
NOC + + -
V2 + + + -
This means, of course, that the disappearance of null objects in these contexts cannot be attributed to the loss of V2.
10
Sigurðsson (1992c) also argues against a Topic-drop analysis of the phenomenon. Further, Cecilia Falk (p.c.) points out that Topic-drop of subjects in Swedish is ungrammatical in coordinated main clauses.
294
3.2.3
Phrasal Pronouns in Old Spanish (Rivero 1992)
One possibility to consider is that the pronouns of Medieval Italian might not have been clitic in the modern sense. Given that coordination is often used as a test for clitic-hood (cf., among many others, Kayne 1975, Brandi & Cordin 1981, Poletto 1993), the Medieval Italian data might be interpreted in a different sense; if the pronouns of MI are XPs, the coordination possibilities attested are comprehensible. Rivero (1992) has shown that “clitic” pronouns of Old Spanish were phrasal and occupied a Spec position in syntax. The evidence Rivero presents in favor of an XP analysis for the relevant pronouns in Old Spanish is not available for Medieval Italian. In Old Spanish it was possible to insert other material between the pronoun and the finite V, for instance, a pronominal subject (34a), or an R-subject together with a prepositional phrase (34b) (the pronoun in bold face and the interpolated element in italics): (34a)
Dixe que lo yo avía muerto. (Rivero 1992, 243) he said that him I had killed
(34b)
Et si lo el rey por bien toviere, mándeme quemar. (p. 244) and if it the king for good had, let him order that I burn
I have not found cases of interpolation of this kind in the Medieval Italian texts I have consulted, and the phenomenon can hardly be relevant for the stages of Italian we are discussing here.11 Some pronouns of Old Italian probably had an ambiguous status with regard to the clitic-XP distinction. This appears to be true for loro ‘they’/‘to them’, lui and lei ‘he’/‘to him’ and ‘she’/‘to her’; see Cardinaletti (1991) for an analysis of loro in modern grammar. The clitics we have seen in NOC are: mi, ti, lo, la, si, ci, vi. Of these, none appears to have phrasal properties, as far as we can tell judging from the texts. The interpolation evidence brought forward by Rivero (1992) relevant for Old Spanish is not attested in Old Italian. Clitics in the MI and RI texts that I have studied can be separated from the finite verb in one specific case, namely with the adverbial element pure, which roughly corresponds to English ‘only’, ‘however’ or ‘thus’. There are numerous examples of this. Consider a couple of them:12
11
It cannot be excluded, as in other cases, that interpolation of XP between a clitic and its host was indeed possible at some previous stage of Italian, or in some historical varieties. 12 Benincà & Cinque (1993, 2324-2325) have examples of interpolation of pure and the negation non. They also report that the adverbial elements gnanca ‘not even’ and sai ‘quite’ can separate clitics from the verbal host in a northern Italian dialect (Modern Triestino).
295
(35a)
Ma dacché vi pur piace, ubbidirò … but since (it) you[cl.]-pur-pleases, I will obey (Nov.no. LXV)
(35b)
E, se egli si pur confessa, i peccati suoi son tanti … and, if he himself[cl.]-pur-confesses, his sins are plenty (Dec; I:1)
From RI prose we have evidence that pure can separate the finite V from negation. (36)
Voi udite come non pur piove, anzi diluvia il cielo. you hear how not pur rains, but inundates the heaven (Cene; Intr.)
If we follow Belletti (1990), the negation non cliticizes to the finite V in Infl, and nothing in my material suggests that there is any significant difference between MI/RI and CI on this point. Since pure is the only element that can separate clitic pronouns and the negation from V in MI, there is no ground for the conclusion that the pronouns in question have phrasal status. The exact structure of (35)-(36) is a different question. Under the Antisymmetry hypothesis, we must assume that cliticization in these cases involves more than one functional head. According to the structure in (37), the clitic pronouns and the negation are attached to a higher functional head, and pure occupies the Spec of the lower functional head:
(37)
AgrP 2
Agr’2
Spec
AgrÞ2 AgrP 1 cl.
AgrÞ2
Agr’1
Spec pure AgrÞ1 V
296
This is in line with Kayne’s (1991, 1994) claim that clitics can indeed adjoin to empty functional heads. There is little reason to assume that the pronominal items under discussion were anything but clitic in the medieval grammar. I assume this and proceed with the analysis.
3.3
Analysis: Recovering Dependencies, Object pro, and Overt Object Shift
I will begin my analysis with the observation that the object in NOC, clitic or DP, is always present in the first conjunct; this is stated as point 3 in (24). Intuitively, the first object is present in order to recover the following ones, and this makes sense if the recovering mechanism builds on c-command, and if c-command of the null objects from the phonetic object is generally obtained in the ancient grammars but not in the contemporary grammars. Shlonsky (1990) proposes that pro, in order to be recovered, must be in a chain relation with an element with phonetically discrete grammatical features. The information necessary for the recovery of pro is then present in the chain. Consider Shlonsky’s (1990, 270) Recoverability Condition: (38)
Feature Assignment/Recoverability Condition: Coindex pro with an element in pro’s CHAIN bearing phonologically discrete grammatical features.
In the typical case, pro is recovered under Spec head agreement. If the Spec head relation between subject and Infl can be considered a link in an extended chain (Chomsky 1986a), Shlonsky’s Recoverability Condition includes recovery of an argument subject pro in Spec I.13 Consider also the consequence of the Recoverability Condition for expletive pro: for Rizzi (1986a), expletive pro does not need φ-features to be recovered. If we interpret this in terms of (38), we can take a different view, with Shlonsky (1990), saying that expletive pro is automatically recovered since expletive pro (in inversion structures) is in a chain relation with a phonetic DP. Note also that, for Shlonsky (1990), the recovering relation is a Case-chain. I will follow the intuition expressed in (38), with the modification that it should not make reference solely to Case-chains, but to dependencies in a broader sense. The analysis of NOC can now be built on two assumptions: 1. I assume that null object pronouns are licensed in Old Italian, and this has also been suggested for Old Scandinavian. I continue to consider the occurrences of null objects in the modern languages as a limited phenomenon (with exception, perhaps, for Icelandic). 2. If our analysis of overt object shift is on the right track, this option must have been available generally in Old Scandinavian (recall that it still exists in Icelandic) as well as in 13
As for the nature of the grammatical features, we can very well assume Rizzi’s (1986a) recovery conditions in addition to (38).
297
Old Italian, but only to some limited extent in modern Swedish and Italian. Recall that these two properties, object pro and overt object shift, are in principle independent of one another; I follow Zwart (1993) in assuming that German and Dutch have overt object shift, but do not assume they have object pro. Suppose the ancient varieties that have Null Objects in Coordination structures also had overt object shift and object pro. This generalization is stated in (39): (39)
Old Scandinavian Medieval Italian Modern Swedish Modern Italian
Object pro + + -
NOC + + -
Overt object shift + + -
I follow Kayne (1994) in assuming that coordination structures are subject to binary branching and the LCA. Further, I assume, with Kayne (1994, 57-59) and Wilder (1994), that the conjunct is the head of its own phrase, as illustrated in (40): (40)
XP Spec
X’
X and
Comp l
In a structure with two conjuncts, one would be the Spec of X and the other the complement. Under Antisymmetry, coordination of participial clauses would have the structure of (41), where the two first participle phrases are specifiers of the conjunctions and the last participle phrase is a complement:
298
(41)
Aux P
Aux
XP
Spec PartP.
X’
YP
X and
Y’
Spec PartP
Compl. PartP.
Y and
I assume, moreover, that PartP in our model corresponds to AgrOP. The minimal assumption, it appears, is that the NOC constructions attested in Old Scandinavian and Old Italian are of the same kind: at the AgrOP level. Is this possible to pursue? The structure would be (42):
(42)
IP
Infl
XP Spec AgrOP X’
AgrO
VP X and
AgrOP
AgrO
VP
(42) takes the shape of a small conjunct analysis of the kind explicitly rejected by Wilder (1994); in his ellipsis approach, coordination is generally at CP level. We have good reasons to believe, following Wilder’s demonstration, that NOC structures do not instantiate ellipsis,
299
and it seems plausible that they are not coordination at the CP level. If we assume (42), the two first predictions are: 1. The coordination is at a level below the point where the auxiliary is inserted (regardless of whether Aux is inserted under T or heads a projection of its own); an auxiliary V could be present only in the structure above XP in (42). 2. The same goes for the subject, which in (42) can be projected only once as nominative must be checked in the higher portion of structure which is not included in the coordination. Two of the questions raised in (24) have thereby received an answer on the assumption that what we have observed in Old Italian and Scandinavian texts is coordination at the participial level, or AgrOP level. The remaining two depend on what type of recovery mechanism we are dealing with. Alternatives such as Topic-drop or the like are certainly to be excluded, since we assume that the conjuncts are not CP. If we bring in ideas on non-configurational recovery (free coindexing with DP in preceding discourse, pragmatic identification or some such), we have no straightforward explanation for why the first element (clitic or DP) is always present and the following ones missing; if these null arguments are recovered through free coindexing or pragmatically in some manner, it should have been equally possible to omit the first clitic in a coordinated structure, spelling out the second, or even omitting all of them. The most reasonable guess is instead that the first object is present in order to recover the content of the others. Whatever the recovery mechanism is, I conclude that it is built on c-command: the first object must c-command the following ones and the difference between ancient and modern grammars might be that c-command does not obtain in the latter. This follows on assumptions already made; I have argued that the historic grammars generally instantiate object shift and that an expletive object pro can be licensed which raises to Spec AgrO in overt syntax. If this mechanism was general in Medieval Italian and Old Scandinavian, the coordinated structure is (43) for Medieval Italian:
300
(43)
XP X’
Spec AgrOP Spec proi
X &
AgrOP
AgrO’
AgrO
Spec proi
AgrO’
VP
AgrO
V
ti
VP
ti
V
I propose (44) for Old Scandinavian: (44)
XP X’ Spec AgrOP
Spec proi
X & AgrO’
AgrO
AgrOP
Spec proi VP
V
AgrO’
AgrO DP(O)i
VP
V
ti
Under Kayne’s (1994, 23) definition of c-command, the highest pro in Spec AgrO actually ccommands everything contained in XP. The idea behind this account is, quite simply, that NOC structures can be described as extended expletive chains of sorts. Consider again the Recoverability Condition of Shlonsky (1990), repeated here as (45):
301
(45)
Feature Assignment/Recoverability Condition: Coindex pro with an element in pro’s CHAIN bearing phonologically discrete grammatical features.
The condition applies to our case if we understand it in a broader sense, that is, if it does not make reference to Case-chains only, but to dependencies generally. Suppose, in particular, that pro is recovered if it is coindexed with a c-commanding element bearing phonologically discrete grammatical features.14 Consider (43) and (44): the null objects of the second and third conjuncts are in an A-chain, headed by the expletive object pro. In the Italian structure, the first pro is recovered by the c-commanding clitic pronominal higher up in the tree; in the Scandinavian one, the first pro is already in an A-chain with the associate DP left in situ. The Medieval Italian examples (22) and (23a)/(b) above are of the ‘Scandinavian’ kind and fall under the structure of (44). The structures in (43)/(44) are obviously possible in a language where expletive object pronouns are overtly shifted to the position indicated as Spec AgrO. If the difference between ancient and modern grammars is precisely this, NOC follows from our general account of Old Italian, and so the two remaining points of (24) (namely questions one and three) have also received their answer. The coordination data discussed in this section are thus derived within the same hypothesis from which we derived word order and agreement patterns in the periphrastic construction in chapter 2 and in the absolute construction in chapter 6. The range of data derived on essentially two assumptions is now considerable, and I conclude that the thesis is empirically corroborated.
4
A’-Bound Null Objects. Further Data from Old Italian and Scandinavian
As previously mentioned, it has been argued that Old Scandinavian varieties had null objects that did not obey any requirement on local recovery (Sigurðsson (1992c)). It is now important to exclude this possibility; apart from any theoretic doubt there may be on such a claim, it
14
This means that A-binding can be a recovering dependency. What I argue for in the text comes close to the notion of long dependencies of Manzini (1994a, 1994b), cf. also appendix 2 to chapter 2. NOC structures consist of a number of subdependencies (pro … t, in (43), pro … DP in (44)), and a long dependency is created as the highest pro of each subdependency is A-bound by the preceding pro. Pro, being subject to binding condition B, can be A-bound but not in its binding domain. My idea thus holds if conjuncts count as binding domains. If the Antisymmetry approach to coordination is to be pursued, this must be assumed anyway: in Swedish, where there are pronominal and anaphoric possessives, only the pronominal one is grammatical in a coordinated structure: (i) Johan och hans bror … (ii) *Johan och sin bror … John and his[+p-a] brother … John and sin[-p+a] brother … Thanks to Lars-Olof Delsing for pointing out the relevance of this.
302
must be excluded that a mechanism such as “free coindexing” operates in NOC. Otherwise, an important generalization would be lost. My account of NOC takes the dependencies to be A-dependencies. NOC in Old Scandinavian and Old Italian thus falls under (45). There are instantiations of null argument objects in both grammars that do not follow from (45). There is however no solid evidence for a “non-configurational” recovery strategy. If it is possible to assume that the null objects in question are operator-bound, which I will assume, we can take them off our research agenda, since they would not, in that case, imply anything for our general pro-hypothesis.15 Consider first a context where null argument objects are attested in Old Italian, namely in absolute gerundival clauses. Trivially, there is no overt morphology available to recover pro. (46a)
… ebbero pietre preziose, le quali, portando _ in mano they had precious stones, which, carrying _ in hand a carne ignuda, fanno l’uomo invisibile nude make man invisible (= ‘when you carry them in your hand they make you invisible if you are nude’) (Fatti; IX)
(46b)
Amore, veramente pigliando _ e sottilmente considerando _ , Love, really (seriously) taking _ and subtly considering _ non è altro che unimento spirituale de l’anima e de la cosa amata. is nothing but a spiritual union of the soul and the thing loved (Conv; III: 2)
Notice, however, that in the first case (46a), the gerundive clause is embedded under a relative; we can assume that the missing object is licensed from the relative operator. In (46b), it is feasible to assume that the subject amore has been Topicalized (or perhaps Focused; it is anyhow separated from the finite V by the gerundive clause). In this case as well can we think of an A’-dependency involving the topicalized DP and the null objects inside the gerundivals. Clearly, it is possible that the null objects of (46a) and (b) are operator bound.
15
Some recent proposals have the effect that operator-binding of an empty category does not necessarily exclude that the empty category be defined [+ pronominal, - anaphoric], i.e. pro. Cinque (1990a, 113-115) suggests that A’-binding is a recovery mechanism for null pronominal elements offered by UG, and this analysis is proposed for parasitic gaps. The same view is defended by Authier (1992) who claims that arbitrary null arguments are A’-bound pros. If this is assumed, the premises for our discussion change. If the claim that operator-binding is a recovery mechanism for pro is correct, Shlonsky’s Recoverability Condition (45) would have to be reformulated so as to make reference to both A-dependencies and A’-dependencies. I will not investigate on the consequences of such a move for my hypothesis.
303
Consider now two somewhat more complex constructions: (47a)
… ancora che vecchio fosse sentì subitamente non meno cocenti even though he was old, (he) felt immediately not less boiling gli stimoli della carne che sentiti avesse _ il suo giovane monaco. the ‘stimuli of the flesh than felt[pl.] had _ his young monk (Dec; I:4)
(47b)
La donna gli fece apprestare panni … li quali, The woman borrowed him clothes … that, come vestiti s’ebbe _, a suo dosso fatti parevano.16 as he worn[pl.] _ had seemed as made for him (Dec; II:2)
Again, the above cases clearly have the flavor of parasitic gaps. There is one curiosity about them; whereas parasitic gaps are typically licensed in infinitival complements, the deleted objects in (47a) and (b) are inside a tensed CP, where parasitic gaps are usually deviant (for the Tense Effect, see Cinque 1990a, 136-143, Manzini 1992, 118-119, and 1994a). (48a) (48b)
I bought the clothes that I liked [without wearing _ ]. ?*I bought the clothes that I liked [when I had worn _ ].
If an A’-dependency is blocked in (48a) and (b) due to the presence of a finite Tense, the null objects in (47a) and (b) may be recovered by the φ-features on the participle. But if we draw this conclusion, we predict that participial agreement could generally recover object pro in Medieval Italian. This cannot be correct; argument null objects have not been attested, in absolute participles or in the compound tense. As Tense islands do not always give rise to clear cut judgments, and as we have only two examples of the relevant kind, the alternative conclusion is preferable, namely that (47a) and (b) are indeed parasitic gaps.17 I conclude that none of (46a)/(b)-(47a)/(b) offers decisive evidence that Old Italian had argument object pro that escapes the Recoverability Condition stated in (45); a parasitic gap analysis can be pursued for all of them. Whether parasitic gaps are pronominal 16
(47a)/(47b) instantiate two other peculiar properties of this Italian variety: the use of avere with a reflexive verb in (47a) and preposing of the participle (in both). The former is not directly relevant for our discussion, see chapter 2 for the latter. 17 A construction like (47a) is also possible in Modern Swedish, but sensitive to lexical semantics, very much as NOC in Modern Swedish. (i), which is the direct translation of (47a), is acceptable according to my intuitions, whereas (ii) is more awkward: (i) ?Han kände köttets lustar mer än vad hon hade känt _. He felt the lust of the flesh more than she had felt _ (ii) *?Han avskydde Kalle mer än vad hon hade avskytt _. he hated Charles more than she had hated _
304
or not is not directly relevant here (cf. footnote 15), the crucial issue being only whether a wh dependency can be assumed or not. It can be so for Old Italian (46a)/(b)-(47a)/(b) and perhaps also for some of the Old Icelandic evidence advanced in the literature. Consider (49a)-(c), from Sigurðsson (1992c, 22): (49a)
hann spyrr huart hon leti _ yt bera. he asked whether she let _ out bring (= ‘whether she let (someone) bring it out’)
(49b)
dvergrinn mælti at sá baugr skyldi vera hverjum höfudsbani, the dwarf said that that ring should to anybody bring death er atti _ who possessed _
(49c)
ætla ek, at þú nytir eigi boga minn þottú spyrnir fótum í _. I believe that you cannot use my bow even if you push with-feet in _
My above account for NOC maintains that NOC structures involve A-dependencies; the three cases of null argument objects in (49) do not seem to follow. Sigurðsson argues that they cannot be analyzed as wh dependencies either and uses them as an argument in favor of the idea that object pro in Old Icelandic was non-locally recovered. However, the evidence is not compelling. Sigurðsson excludes the possibility that the clauses inside which the null objects are found could host a recovering operator. It is not clear whether this option can be excluded. The fact that these null objects are typically attested with a +wh Comp ((49a) is an embedded interrogative, (49b) is a relative clause; compare with Old Italian (46a) and (47b)) is probably not coincidental. On the one hand, Sigurðsson notes that the relative operator in (49b) ought to be coindexed with the subject anybody and presumably cannot recover the null object the ring. On the other, (49b) (which has some resemblance to the Old Italian examples (46a) and (47b)) is reminiscent of cases of apparent DP-extraction from islands (Cinque 1990a, ch. 3). (50a) (50b)
Il libro che conosco l’autore che scrisse _. Boken som jag känner författaren som skrev _. the book that I know the author who wrote _
Furthermore, the context of (49a) is actually a dialogue (though indirect) of which the null object is the discourse topic; the typical context where null objects are actually allowed in modern Scandinavian and several other languages. A Topic-drop analysis cannot be a priori excluded for (49a). As for (49c), it strongly resembles particle verb constructions, put in sth., which often allow for null prepositional objects also in contemporary Scandinavian. (51a) För att känna om plattan var varm lade hon handen på _.
305
to feel if the plate was hot she laid her hand on _ (51b)
Vi kan inte öppna dörren när du ställer stolen för _. We can’t open the door when you put the chair for (against) _
(51c)
Jag tror inte du kan spänna bågen ens om du sätter foten i _. I believe you cannot use my bow even if you put your foot in _
The exact analysis of (51a)-(51c) can remain open for our purposes. It is not straightforwardly obvious that the missing objects of (51a)-(51c) are structurally represented at all. Summing up, the null objects of Old Italian and Old Scandinavian considered in section 4 are not very different from structures that are familiar from modern grammars, such as parasitic gaps, apparent DP-extraction from islands and null prepositional objects. At least in the two former cases, it is natural to assume a recovering A’-dependency. If there were some general difference between the ancient literary languages and the modern ones with respect to locality conditions, this is not possible to tell on the basis of what we have seen. The material is too small to allow any general conclusions; structures of the relevant complexity are quite simply too infrequent.
5
Conclusion of Chapter 7
The hypothesis is descriptively successful as it allows us to capture a number of changes from Medieval to Contemporary Italian on essentially two assumptions: (a) in MI, accusative is strong and forces overt Object shift; (b) AgrO in MI licenses pro. At least the first of these properties is lost at some point during the 14th century. In order to insist on this generalization, I have made a distinction between null objects in coordinated structures, which are pro recovered by an A-dependency, and null objects in other contexts, which I have argued can be analyzed as operator bound (this, in turn, can mean either that they are variables or that they are A’-bound pro; see footnote 15). I conclude that there is empirical support for my claim. Moreover, without the two preceding assumptions, the syntactic changes under discussion defy a unified account. Possible alternatives, such as linking the grammaticality of certain instances of null objects to the V2 phenomenon or to ellipsis, fail to capture the entire range of data in a satisfactory way. In my view, the V2 properties of MI are irrelevant. What we need to assume is a change in the Formal Licensing parameter, saying that AgrO licensed pro in the Old Italian grammar but does not in the modern one, and a change in the property of the accusative Case feature, from strong to weak. Having said this, it must now be shown how cases of null objects in Contemporary Italian can be dealt with in this model, and more precisely, how my account can be made compatible with Rizzi (1986a).
306
CHAPTER 8 Remaining issues: On the Object pro of Rizzi (1986a) and Resultative Adjectival Constructions
0
Introduction
I will finally turn to two questions, and also potential difficulties for my analysis that the reader might have noticed at an early stage of the demonstration. The first is how the proposed object pro parameter is compatible with Rizzi’s (1986a) claim that object pro is indeed possible in Modern Italian. The second problem to resolve regards the adjectival clause, since, as we know, adjectival constructions may also be resultative. First, PAP, or PAP-like constructions, can be formed with adjectival small clauses. Second, the concept of delimitedness is clearly also relevant for secondary predicates discussed by Tenny (1987) and, for Romance languages, by Merlo (1989). I have saved these issues until last, and have decided to treat them in the same chapter as I believe that the answers to them are related and depend on previous demonstration.
1
Object pro of Rizzi (1986a)
I have crucially assumed that word order differences between Old and Modern Italian are imputable to the fact that AgrO licenses pro in the former but not in the latter variety. Obviously, there is a possible conflict between this claim and the object pro hypothesis of Rizzi (1986a). Rizzi argues that V in Italian licenses a null pronominal object under government. This property distinguishes Italian from English, for example, where null objects are less productive and do not show signs of syntactic activity. In English, theta grids corresponding to null objects are lexically saturated, and the object, therefore, is not projected in syntax. Take some of the crucial data of Rizzi (1986a). The cases to explain are the following: 1. Arbitrary Object pro is possible in adjunct small clauses under the condition that the finite verb is an affecting predicate: (1)
Un dottore serio visita _ nudi. (Rizzi 1986a, 505) a serious doctor visits _ nude
(2)
*Gianni conosce _ giovani. (p. 537) John meets _ young
308
2. Arbitrary Object pro is possible in argument small clauses under the condition that these are causative and it is not possible in epistemic small clauses: (3a) (3b)
(4a) (4b)
Questa musica rende _ allegri (p. 507) this music makes _ happy Talvolta la stampa lascia _ perplessi. (p. 533) sometimes the press causes-to-remain _ puzzled *Spesso il medico considera _ sani. (p. 533) often the doctor considers _ healthy *Talvolta la stampa ritiene _ perplessi. (p. 533) sometimes the press believes _ puzzled
The difference between (3a)-(b) and (4a)-(b) is presumably due to the same origin as the difference between (1) and (2) above, since epistemic predicates are unaffecting and causatives affecting. 3. If pro is expletive, however, things are different. Expletive pro is possible in the small clause complement of both epistemic and causative predicates: (5a) (5b)
Gianni ritiene [ pro probabile [che Mario venga]]. (Rizzi 1986a, 527)) John believes likely that Mario comes Il tuo comportamento ha reso [pro improbabile [che Mario venga]]. Your behavior rendered unlikely that Mario comes
Rizzi attributes the inversion property of (6) to an expletive null element in the Spec of the adjectival small clause: (6a) (6b)
Ritengo [suo fratello più intelligente]. (p. 529) I believe his brother more intelligent Ritengo [ _ più intelligente suo fratello]. I believe _ more intelligent his brother
4. Arbitrary object pro is possible not only in complex contexts, but also with single predicates as in (7): (7a) (7b)
Talvolta Mario spaventa/preoccupa/impressiona/meraviglia _.(p. 536) sometimes M. frightens/worries/impresses/amazes _. Un successo può dotare _ di maggior entusiasmo. (p. 537) a success can endow _ with more enthusiasm
309
5. On the other hand, it should be emphasized that the predicates of (1) and (3) are generally incapable of licensing pro on their own, that is, when they are not followed by a small clause. Witness that (8a)-(8b) are ungrammatical with the relevant reading: (8a) (8b) (8c)
*Un dottor serio visita _. a serious doctor visits _ *Questa musica rende _. this music makes _ *Talvolta la stampa lascia _. sometimes the press causes-to-remain _
Rizzi’s account for (1)-(7) (excluding (8a)-(c) that he does not discuss) is built on the assumption that V counts as a licensing head in Italian, and that pro is licensed in all of (1)-(7) by Case marking under government from V. Rizzi’s formal licensing parameter is stated as (9) (Rizzi 1986a, 524): (9)
pro is Case-marked by X
y
X has the value V in Italian but not in English. On this assumption, Rizzi explains (1), (3), (5), (6), and (7), but not (2) and (4). For Rizzi, (2) and (4) are independently excluded by the Affectedness Constraint that requires that the null argument is affected by the event. He claims further that the Affectedness requirement lies on the recovery of pro and not on the formal licensing. The structure Rizzi proposes for the complex predicate, take (3a) for example, is (10), where V licenses pro inside the small clause under government and Exceptional Case marking: (10)
[VP V [SC pro A]]
In the minimalist framework, V cannot license pro through government in (10) but this problem is only apparent. What we can suppose is that the small clause contains an Agr head which licenses pro in its specifier. Recall furthermore, that Chomsky (1993) suggests a distinction between participial agreement AgrO, and adjectival agreement AgrA. A first attempt to translate Rizzi’s model into MPLT terms would be to say that in Modern Italian object pro is licensed by AgrA. Instead of (10), I assume the structure (11) which has been independently argued for by Cardinaletti & Guasti (1992a), cf. also Chomsky (1993):
310
(11)
VP
V rende
AgrAP
AgrA’
Spec pro AgrA
AP
A allegri Suppose the licensing head in Modern Italian, X of (9), is AgrA. In this way we have taken care of (1), (3), (5), (6), and (8). There may be a point in making a distinction in this sense. Consider again PAP. To my informants, inversion of the direct argument (which for us means: ‘leaving the object in situ’) is unnatural: (12a) (12b)
(13a) (13b)
(14a) (14b)
Ho le camicie stirate. I have my shirts ironed *? Ho stirate le camicie. I have ironed my shirts Abbiamo la macchina portata dal meccanico. we have the car driven/brought to the mechanics *?Abbiamo portata la macchina dal meccanico. we have driven/brought to the mechanics Abbiamo la frutta caricata sul camion. we have the fruit loaded on the truck *?Abbiamo caricata la frutta sul camion. we have loaded the fruit on the truck
The b-sentences of (12)-(14) are supposed to be interpreted as synonymous to the a-sentences. One must keep in mind, however, that certain Italian varieties still admit for rightward agreement on the participle, and that some speakers therefore might recognize the b-examples of (12)-(14) as compound tenses. For my informants, the result is still deviant.
311
Inversion appears to be quite more natural, and even perfectly grammatical when the predicate is an adjective: (15a) (15b)
(16a) (16b)
(17a) (17b)
(18a) (18b)
Gianni non ha le idee chiare. John does not have his ideas clear Gianni non ha chiare le idee. John does not have clear his ideas Ho la cena pronta. I have the dinner ready Ho pronta la cena. I have ready the dinner Aveva la casa piena di ospiti. he had his house full of guests Aveva piena la casa di ospiti. he had full his house of guests In questo modo, avevano i viaggiatori salvi. in this way, they have the passengers safe In questo modo, avevano salvi i viaggiatori. in this way, they had safe the passengers
In PAP contexts, there seems to be a general contrast between inversion possibilities in the adjectival small clause and in the participial small clause, in the sense that inversion is well formed only in the former. This gives a further confirmation to what we have suspected and argued for at length, namely that the Case properties of these constructs do not depend on the verb ‘have’ but on the non finite predicate. Furthermore, the data suggests that the licensing head cannot be AgrO. The inversion facts of (15)-(18) are due to some property that adjectives have in Italian, but that participles do not share, given the ungrammaticality of (12)-(14). I suggest that the explanation to this contrast lies in the structure (11) above, and more precisely, in Cardinaletti & Guasti’s claim that object pro is licensed by a functional head inside the small clause under Spec-head agreement and not from the finite V under government. In chapter 2, we saw that the word order exemplified in (12)-(14) most probably was grammatical in the earliest variety of Italian, and this brings us to conclude that pro was licensed by AgrO and AgrA in Medieval Italian, but only by AgrA in Contemporary Italian. In Iberoromance languages, both Spanish and Portuguese, the inversion property in the phrasal complement of tener/ter, ‘hold’ or ‘have’, is attested both with adjectives and participles:
312
(19a) (19b)
(20a) (20b)
Tengo las camisas planchadas. I have my shirts ironed Tengo planchadas las camisas. (= (19a)) I have ironed my shirts Tengo la casa limpia. I have the house clean Tengo limpia la casa. (= (20a)) I have clean the house
We conclude that Modern Iberoromance languages are similar to Medieval Italian, in the sense that both AgrO and AgrA are licensers of pro. Still, the account is far from exhaustive. We have not explained why pro is possible in (7) and why it is impossible in (2) and (4). I believe the two PAP constructions in Swedish may give some useful indication at this point. Inversion of the above kind is quite more natural in the Swedish få-construction than it is in the ha-construction. (21a) (21b)
(22a) (22b)
Jag fick brevet skrivet. I got the letter written Jag fick skrivet brevet. (= (21a)) I got written the letter Jag har brevet skrivet. I have the letter written *?Jag har skrivet brevet. (= (22a)) I have written the letter
Obviously, the participial predicates in all of (21a)/(b)-(22a)/(b) are identical with regard to agreement and aspectual properties. In (21), where the finite V is få, the object may appear in rightward position that I take to be the basic position (see the above discussion in chapters 1 and 3). In (22), this is not possible, but the object has to move overtly. This will mean that expletive pro is licensed in få-participle but not in ha-participle. This is unexpected from one point of view as we have had good reasons to believe that Case marking properties of such constructions, and hence licensing of pro, do not depend on the finite V but on the nonfinite predicate. I suggest the solution lies in the notion of complex predicates, and more precisely that we should admit for complex licensers. The idea is by itself quite in line with Rizzi’s way of reasoning (see his section 5.1.). Recall that ha is a verb of possession, whereas få is causative. The conclusion is then, that pro may be licensed by a functional head in the participle phrase under condition that there is a CAUSE feature in the structure. How to
313
capture this formally is however somewhat unclear to me. Consider again the Formal Licensing according to Rizzi (1986a, 524): (23)
y
pro is Case-marked by X
It is not sufficient, in Swedish, to say that the licensing head X is AgrO. If it were, (21b) and (22b) would be equally possible. The data indicates that the licensing head may be complex, which is to say that X may have the value of two heads that are syntactically or lexically combined somehow. The licensing context for pro in Swedish would then be [CAUSEi … AgrOi], where the two features are unified through the formation of a complex predicate. Thereby, we rule out expletive pro in (22b) and with that the inversion property. The question is then how the complex predicate is formed. The finite verb and the participial phrase do not form a syntactic constituent, and we conclude (as Rizzi 1986a) that the formation of a complex predicate may take place in syntax. The idea is crucially dependent on Rizzi’s reasoning on the formation of a complex predicate, which I assume to be viable also in a minimalist or an Antisymmetric framework in principle. The coindexation mechanism of Rizzi might of course be understood in a variety of ways. Following Guasti (1992), we could assume abstract raising and incorporation of the nonfinite predicate to the higher predicate. One may also think of some percolation device (cf. Guéron & Hoekstra 1988), but the choice will remain open here. By now it should be clear that it is highly doubtful whether the correct generalization can only be captured by saying that pro is licensed by a functional head (AgrO or AgrA) under Spec head agreement or by a lexical head (V) under government. Above all, Rizzi’s observation that null objects in Modern Italian obey the effects of affectedness is crucial and shows that more intricate mechanisms are involved. I will limit my discussion to some general remarks, concluding that the entire range of the problem must be left for future research. For Rizzi, the affectedness requirement lies on identification and has to do with arbitrary interpretation. The model we are developing here where information of Aktionsart or verbal aspect is syntactically realized as a feature opens a different possibility, namely that the licensing head for pro could be precisely this feature, that is, Asp when endowed with delimited reading. In addition, it seems probable that Asp is not responsible for the licensing of pro alone, but rather together with the non finite predicate. The licensing head for pro in the relevant cases of Italian is then the complex head [Asp[+del.]i … AgrAi]. An interesting difference between this proposal and Rizzi’s model now arises as Rizzi assumed affectedness to be relevant for the recovery of pro. Actually, there is no conflict between the proposals. If pro is recovered by features on the licensing head, it is natural to say that Asp[+del.] both licenses and recovers the null argument, or is involved in both licensing and recovering. If expletive pro is not subject to a requirement that recovering features are present on the
314
licensing head, AgrA alone would be sufficient for the licensing of expletive pro and this is what we attest in (5) and (6). I am now following the analysis of Rizzi (1986a). What remains to be explained is why pro is possible with the single predicates in (7), but not with those in (8). It may be that what Rizzi has analyzed as a unitary phenomenon in Italian is less unitary than what it appears at first sight. The example (7a), where an arbitrary null object is instantiated with experiencer verbs, is one of the typical cases where the object can also be null in Swedish where null objects otherwise are not as productive as in Italian (Rizzi, p. 536): (24a)
Denna nyhet skrämmer _. this news frightens _
(24b)
Hans påståenden förvånar _. his claims surprise _
(24c)
Deras förmåga imponerar _. their capability impresses _
(24d)
Dessa händelser överraskar _. these events amaze _
Nevertheless, we cannot exclude that the null arbitrary arguments of Swedish (24a)-(d) and Italian (7a) are of the same kind, namely lexically suppressed according to Rizzi’s analysis. This also appears to be true for some benefactive predicates (cf. Rizzi, 536): (25)
Det händer att direktören befordrar/tackar/hjälper _ utan anledning. It happens that the director promotes/thanks/helps _ without reason
Furthermore, according to Guéron (1986), it is a general property of benefactive arguments that they are optional. In any event, there are differences showing that the phenomenon is sensitive to argument structure. The data suggest that the property of object drop was more general in ancient varieties of Italian, and it could therefore be claimed that the licensing head of those varieties was AgrO. In modern language, null objects appear to be more sensitive to structure and lexical semantics. Generally speaking, there are two restrictions interacting: First, if null objects are arguments, they are well formed only with affectedness predicates; second, the typical context of null objects, argument and expletive, is a complex context, that is, where the finite V forms a complex with a non-finite predicate, or where there is a secondary predicate. This second predicate is typically adjectival. It appears difficult to derive the entire range of data merely on the assumption that V license pro under government or that
315
any functional head does so under Spec head agreement. Intuitively, what we see in Modern Italian is a null object restricted by more specific requirements. The typical context for an argument arbitrary pro is one where a delimited predicate takes an adjectival small clause, and the licensing and recovering head could in this case be the complex [Asp[+del.]i … AgrAi]. When pro is not argument, the head AgrA could be sufficient. I leave the discussion at this and proceed with the problem of adjectival predicates.
2
Resultative Adjectival Constructions
2.1
Adjectival PAP
We now arrive at the second question of this section, which naturally follows our discussion of the first: what are the properties of the possessive auxiliary + adjective constructions as opposed to the participial ones? Cinque (1990b) claims that the ergative/unergative distinction is valid not only for verbal predicates as generally assumed, but also for adjectives. He proposes, on the basis of Italian evidence, that adjectives such as probabile ‘probable’, oscuro ‘obscure’, and ovvio ‘obvious’ are ergative and ingiusto ‘unjust’, pericoloso ‘dangerous’, and fedele ‘faithful’ unergative. Recall, now, that PAP is generally awkward with participial predicates of both kinds: The ergative PAP: (26a) *?Abbiamo gli ospiti arrivati. (26b) *?Vi har gästerna anlända. We have the guests arrived The unergative PAP: (27a) *?Abbiamo tre chilometri corsi. (27b) *?Vi har tre kilometer sprungna. we have three miles run
The ergative and unergative adjectives I have mentioned are all equally excluded which appears to be consistent with my demonstration. The Italian examples are as usual ambiguous between an attributive and a predicative reading of the adjective; the judgments presuppose a predicative one:
316
The ergative adjectival PAP: (28a) *Abbiamo i risultati probabili. (28b) *Vi har resultaten sannolika. we have the results probable
(29a) (29b)
*Ha le sue intenzioni oscure. *Han har sina avsikter dunkla. he has his intentions obscure
(30a) (30b)
*Aveva il suo fallimento ovvio. *Han hade sitt misslyckande uppenbart. he had his failure obvious
The unergative adjectival PAP: (31a) *Avevano la condanna ingiusta. (31b) *De hade domen orättvis. they had the condemnation unjust
(32a) (32b)
*Aveva le idee pericolose. *Han hade tankarna farliga. he had the ideas dangerous
(33a) (33b)
*Ha la moglie fedele. *Han har hustrun trogen. he has his wife faithful
Well formed adjectival PAP constructions instantiate a different group of adjectives including, among others, pronto ‘ready’, chiaro ‘clear’, sicuro ‘secure’, pieno ‘full’, and salvo ‘safe’. Of these, at least two are what Cinque defines as ergative, namely chiaro and sicuro. The relevant adjectives also appear to be possible in PAP cross-linguistically. (34a) (34b)
Ho la cena pronta. Jag har middagen färdig. I have the dinner ready
317
(35a) (35b)
Gianni non ha le idee chiare. Johan har inte sin uppfattning klar1. John does not have his ideas clear
(36a) (36b)
Abbiamo gli investimenti sicuri. Vi har investeringarna säkra2. we have the investments secure
(37a) (37b)
Aveva la casa piena di ospiti. Han hade huset fullt av gäster. he had his house full of guests
(38a) (38b)
In questo modo, avevano i viaggiatori salvi. På så vis hade de passagerarna säkra. in this way, they had the passengers safe
The adjectival PAP is therefore possible with some of the adjectives Cinque defines as ergative, but not all. A fair guess would be that it is generally awkward with unergative adjectives, but I am less certain on this point. Future research will have to resolve this issue together with cross-linguistic differences. It appears to me that the class of adjectives that are allowed in PAP corresponds to those studied by Bosque (1990) on Spanish. Bosque intends to show that a category of Spanish adjectives are lexically derived from corresponding verbal forms, and that they share some fundamental aspectual property with the verbal counterpart. The category in question includes lleno ‘full’, limpio ‘clean’, contento ‘satisfied’, and some adjectives of dialectal or literary use as calmo ‘calm’ and salvo ‘safe’ (Bosque 1990, 183). In Spanish, as well as in other languages, these adjectives are morphologically and semantically close to verbal predicates, as ‘full’ is related to ‘fill’ or ‘clean[A]’ to ‘clean[V]’, for example. For Bosque (1990, 186), the adjectival form is derived from the participial form of the verb rather than vice versa. The lexical derivation goes llenar>llenado>lleno instead of lleno>llenar>llenado, through a process of conversion, or truncamiento, in Bosque’s terms, that turns the participle into A; hence, ‘fill’>‘filled’>‘full’, and not ‘full’>‘fill’>‘filled’. Although the idea may appear counterintuitive to some, I assume Bosque is right in this. The crucial point here is only that A
1
In Swedish, PAP is well formed with klar when the meaning is close to ‘ready’, perhaps less so when it means ‘clear’: (i) Jag har middagen klar. (=(34b)) I have the dinner ready 2 Note that Swedish säker corresponds both to ‘secure’ and ‘safe’.
318
and V share the same aspectual property which is perfectivity.3 However, for our purposes, Bosque’s hypothesis must be extended to Germanic (Bosque 1990 comments this on p. 182183) given that the adjectival forms that may appear in PAP appear to be roughly the same in Swedish as in Italian4. By hypothesis, the adjectival predicates in question inherit perfectivity from the verbal stem from which they are derived. By doing so, these adjectives have the prerequisite for being telic or delimited (cf. Bertinetto 1986, 298-299 and our discussion in chapter 6), and when they are used in PAP they undoubtedly acquire such a reading. If this idea, or some version of it is valid, our above discussion in chapter 3 would now have some non-trivial consequences for the analysis of the adjectival PAP. The ideal situation would be that adjectival PAP and participial PAP are not distinct syntactically; and as a result, that the single argument of the adjectival predicate is base generated as the complement.5 As for the structure of I have the dinner ready in the languages under discussion, a first approximation is then (39): (39)
Aux P
AP
Aux
A’
Spec
A
DP
The structure must however be sufficiently rich to allow for the possessor to be base generated higher than AP; I continue to follow Guéron (1986), Kayne (1993) and Mahajan (1994) in their claim that the possessor is linked to an abstract prepositional element, although I continue to leave the question open of where exactly it originates. Crucially, I assume the possessor does not count as the external argument of A. The predicative AP is thus base generated without an external argument. What will happen in (39) according to my 3
A further possibility to investigate is that A and V are both derived from a common source originally unspecified for the lexical features. Intuitively speaking, this seems preferable for certain Swedish adjectives such as färdig ‘ready’ that do indeed correspond to a verb förfärdiga ‘make-ready’. The adjectival form, if we were to follow Bosque coherently, would then be derived from the V through a process that deletes the prefix för- and the suffix -a. 4 There is one obvious exception to this, about which I have nothing to say, namely those possessive expressions that in Swedish can only be formed with an attributive A, but in Romance have at least the appearance of being predicative. As is well known, this concerns mainly colour adjectives as in aveva gli occhi verdi ‘she had green eyes’. 5 Recall also that this is a necessary conclusion according to Tenny’s (1987) Aspectual Mapping Principle.
319
hypothesis, is that the internal argument will be externalized. As with participles, overt externalization is driven by the delimited Asp feature we introduced in chapter 3. DP therefore, moves to Spec Asp which by hypothesis counts as subject of the predicative phrase, that is, the AP shell: (40)
AspP
Spec DPi
Asp’
Asp [+del.]
AP
A
ti
By saying this, we have proposed a syntactic derivation of adjectival PAP which is identical to that of participial PAP. It follows that adjectival PAP can be formed out of the class of adjectival predicates individuated by Bosque given that the A as well as the V must be delimited in order to appear in this construction. The grammaticality of Ho la cena pronta ‘I have the dinner ready’ and the ill-formedness of *aveva le idee pericolose ‘he had the ideas dangerous’ is thereby predicted. A consequence of this proposal is that word order variations as those exemplified in (15)-(18) above are due to the licensing of an expletive pro that may take the place of the DP in overt syntax, quite in line with Rizzi (1986a). As opposed to the constructions discussed by Rizzi (1986a), we have to assume for this case that the rightward position is the basic position of the DP in the complement of the adjectival predicate. This amounts to saying that adjectival PAP is a transitive or perhaps, transitivized construction. Further, the DP argument is Case checked against the adjective and not against possessive have. I assume therefore that adjectives of at least this class, and in this context, can assign structural accusative. The adjectival predicates of (34)-(38) thus have the syntactic status of participial predicates. I believe that this account has some empirical support, and I continue to build my argumentation on Italian and Swedish: 1. Apart from the fact that an expression such as having the dinner ready naturally lends itself to a transitive reading in the languages that make use of it, the transitivization process is straightforwardly clear in some Romance varieties. Rohlfs (1968, § 629) mentions southern Italian dialects where the avere + adjective periphrasis takes on the value of a compound tense. The phenomenon is, according to Rohlfs, most normal with pieno ‘full’. (41a) I have
320
full the sack thus means I have filled the sack, and (41b) means he has filled it (examples from Rohlfs 1969, § 629):
(41a)
calabrese:
agghiu chijnu u saccu. I have full the sack = ‘I have filled the sack’
(41b)
laziale:
j’a pino. (he) it[cl.]-has full = ‘he has filled it
In the ancient literary texts, there is evidence of other adjectives having this property. Rohlfs reports (42a) and (b) from the Mandragola: (42a)
… mi ha fracido … (he) me[cl.]-has wet = ‘He has made me wet’ (Mand; II:5)
(42b)
…tu mi hai chiaro… you me[cl.]-have clear = ‘you have made me clear’ ‘you have enlightened me’ (Mand; IV:2)
The construction avere + chiaro ‘clear’ and fracido ‘wet’, has taken on a transitive, causative reading in (42a) and (b). 2. In central Tuscan Italian, the adjective may loose agreement with the postponed DP (Luciana Brandi, p.c.): (43)
Ho pronto la cena. I have ready[-Agr] the dinner
This may happen however, only when the DP appears in final position: (44a) (44b)
*Ho la cena pronto. I have the dinner ready[-Agr] Ho la cena pronta. I have the dinner ready[+Agr]
321
Interestingly, the phenomenon also occurs in Swedish, and appears to be quite wide spread (cf. Hedlund 1992, section 2.5). If we translate the very same sentence in Swedish, we also discover the same contrast. The result is better if få ‘get’ is used instead of hava ‘have’: (45a) (45b)
Jag fick middagen färdig. I got the dinner ready[+Agr] Jag fick färdigt middagen. I got ready[-Agr] the dinner
The agreeing form of the adjective is not excluded when the DP appears to the right. In my opinion, (46b) is still less natural than (45b), but the non-agreeing adjective is ungrammatical if the DP is in leftward position, (46a): (46a) (46b)
*Jag fick middagen färdigt. I got the dinner ready[-Agr] ?Jag fick färdig middagen. I got ready[+Agr] the dinner
The final ‘-t’ could actually be understood in two ways: ‘-t’ is the normal ending of the active participle, the supinum. However, ‘-t’ is also the mark of neuter gender. The noun middagen is common gender, and the agreeing adjective should therefore have the shape färdig. Agreement in common gender corresponds in this case to null morphology. I have glossed färdigt as non-agreeing, although it may be discussed whether this form corresponds to an adjective marked for neuter or an adjective assuming the morphology of an active participle. The choice is not crucial though, because either way the conclusion will be that the adjectival construction is transitivized. If the adjective is to be understood as agreeing in neuter, the analysis of the få-construction from chapter 2 carries over. We would then have an expletive pro defined singular, neuter in Spec AgrA. If the adjective has been reinterpreted and has the status of an active participle or supinum, things are if possible even more obvious. In any event, the conclusion will be that rightward and leftward positions are structurally different. ‘Leftward’ means the specifier of an Agr element, and ‘rightward’ means the complement of A, that is, the basic position of the DP.6 These patterns are productive in the sense that it is often possible in the fåconstruction to let an adjective take the relevant morphology when the DP is postponed (cf. Hedlund 1992, section 2.5):
6
Note finally that the ‘-t’ in Swedish is the ending of adverbs as well. However, I choose to disregard the possibility mentioned by Hedlund (1992, 75) that the adjective in constructions such as (45b) (or (53b) below) can be reinterpreted as an adverb. If this possibility actually exists, I can see no principled reason to why the adjective could not be reinterpreted as an adverb also in (46a) (or in (54a) below).
322
(47a) (47b)
(48a) (48b)
Fick du mattan ren? did you get the carpet clean[+Agr]? Fick du rent mattan? did you get clean[-Agr] the carpet? Jag fick artikeln klar i går. I got the article ready[+Agr] yesterday Jag fick klart artikeln i går. I got ready[-Agr] the article yesterday
Adjectives as färdig, klar and ren take on a perfective meaning when used in the relevant context, and describe telic events in all of (45)-(48).7 Summing up, in the spirit of Bosque (1990) I have assumed that the derivation of adjectival resultative clauses as in having the dinner ready is essentially equal to that of participial resultative clauses such as having the shirts ironed, in Italian and Swedish. I assume along with Bosque that lexical processes identify a class of adjectives that have the property necessary for this, which is perfectivity. Adjectives of the perfective class, in the relevant construction, show properties similar to transitive, passive participles, which in our model means: (a) that their single argument is a direct object, and (b) that this single argument checks accusative in the adjectival small clause. This claim is corroborated by data from Scandinavian and Italian varieties where the transitive nature of these adjectival clauses is reflected both by word order and morphology, apart from semantics. Still, given what was said in the first section of this chapter, we must assume the A-V distinction is not neutralized, but is still visible in syntax as there seems to be a functional head in the adjectival clause that licenses pro, whereas there is no such head in the participial clause.
2.2
Resultative Secondary Predicates
Resultative secondary predicates are discussed by Tenny (1987), Merlo (1989) for Italian, and Hedlund (1992) for Swedish, among many others; see references cited in these works. The
7
Again, there are factors of lexical semantics that make the picture more obscure. An adjective as lugn ‘calm’, which describes a psychological state, can be used in the causative construction, but cannot appear in final position, regardless of whether the adjective agrees or not. (i) Fick du gästerna lugna? did you get the guests calm[+Agr]? (ii) *Fick du lugnt gästerna? did you get calm[-Agr] the guests? (iii) *Fick du lugna gästerna? did you get calm[+Agr] the guests? I have no interesting proposals to make concerning this difference.
323
construction under discussion is exemplified in (49):8 (49)
Resultative SP: John hammered the metal flat.
(49) is subject to the Affectedness Constraint since the predicates that may be used are, by and large, the same that can appear in the English Middle. Merlo’s (1989) account in terms of an adjoined predicate is not viable under either of Antisymmetry or Bare-Phrase-Structure. She proposes the following structure (pp. 34-35): (50)
Resultative SP:
IP
NP John
VP
V hammer
NP the metal
Pred flat
If we accept without discussion that the Affectedness Constraint is operative in (49), the conclusion in our framework must be that there is an Asp head with the feature [+ delimited] involved in the derivation. The event expressed in (49) is obviously of the delimited kind, but I would like to suggest in line with our way of reasoning above that the object in hammer the metal flat is affected not only by hammer, but by the complex hammer … flat.9 This permits us to say that the delimited Asp node is brought into act inside the adjectival clause. In chapter 3, it was argued that externalization of the internal argument is movement to Spec Asp triggered by the delimitedness feature. I suggest that the following happens: the DP the metal is the internal argument of the complex hammer … flat and is base generated as the complement of the lowest member of the complex, the adjective:
8
I will disregard here the other two constructions discussed by Merlo (1989), among others, namely the ‘descriptive SP of the object’, (i), and the ‘descriptive SP of the subject’, (ii): (i) Descriptive SP (of the object): John served the meat overcooked. (ii) Descriptive SP (of the subject): John served the meat angry. How Merlo’s account is going to be reinterpreted in terms of Antisymmetry or Bare-Phrase-Structure, is unclear to me. Furthermore, there are cross-linguistic differences with regard to (i) and (ii) that go far beyond our research agenda. Since they are not subject to Affectedness, they are also less relevant for our discussion. 9 The idea is reminiscent of Chomsky’s (1986a) claim that the argument of adjectival small clauses is the object of a complex V … A.
324
(51)
VP
Spec DP John
V’
V hammer
AP
A flat
DP the metal
The lowest member of the complex evidently does not have an external argument (John is external argument of the highest predicate, V). The affected DP is externalized to subject position in the predicate AP shell, triggered by the need to check the Asp[+del.] feature: (52)
VP
Spec DP(S)
V’
V
AspP
Spec DP(O) i
Asp’
Asp [+del. ]
AP
A
ti
Following the analysis outlined in chapter 3, this is possible only in constructions that involve a delimitedness feature. If the event expressed by the complex is non-delimited, externalization of DP(O) to Spec Asp[+del.] will not take place and the derivation will crash under the assumption that the predicate requires the presence of a ‘subject’; see our discussion in chapter 3, section 3.4. In essence, this account means that the lowest member of the complex inherits the delimitedness feature from the highest member.
325
The hypothesis expressed in (52) furthermore, is supported by Swedish data. In Swedish, the direct object of such constructions can appear on both sides of the resultative secondary predicate, witness (53a) and (b): (53a) (53b)
Han tvättade golven rena. he washed the floors clean[+Agr] Han tvättade rent golven. he washed clean[-Agr] the floors
The same difference with regard to agreement appears in (53a) and (b) as the one we have already seen in (43)-(48) and in chapter 2; the adjective obligatorily agrees in number when the DP is to the left, but not when DP appears to the right of A: (54a) (54b)
??Han tvättade golven rent. he washed the floors clean[-Agr] ?Han tvättade rena golven. he washed clean[+Agr] the floors
This contrast, which by now is familiar, shows as in previous cases that the two positions are structurally different; on this point, the same conclusion is reached by Hedlund (1992, section 2.5). In (53a), the DP has moved through the Spec of an Agr node, and in (53b), the DP stays in situ. As in previous cases, I assume the adjectival clause hosts more functional structure than indicated in (52), and more precisely, that there is an AgrAP above the AspP. I conclude that the analysis can be successfully adapted to Antisymmetry in this way as word order and agreement facts are dealt with better under this approach than under (50) above.
326
CONCLUDING REMARKS
My concluding remarks will be brief. The discussion of this thesis has been concerned with four major issues of participial syntax: 1. Functional structure of the participle 2. Word order 3. Agreement patterns 4. Interpretation, especially with regard to Aktionsart or verbal aspect. I have argued that there is solid empirical evidence that word order correlates with agreement (see chapters 2, 4, and 6). It has been shown that a wide range of data receive a natural explanation in a theory that makes use of agreement nodes and builds on the Spec-head agreement hypothesis. My account has followed the Antisymmetry Theory of Kayne (1994). In chapters 3 and 6, I argued that the fact that the Affectedness Constraint correlates with DP-movement and the projection of argument structure can be captured in a unified hypothesis on the assumption of an Aspect node carrying the delimitedness feature of Tenny (1987) and Borer (1993, 1995). It has also been shown that the Affectedness Constraint itself does not have to be assumed as an independent principle of grammar, but that its effects derive from general and principled mechanisms. It is awkward to derive these results in a theory that does not admit any functional structure in the participle clause beyond the VP itself, as in the Multiple-Spec hypothesis of Chomsky (1995). The presence of AgrO and Asp is empirically justified by output conditions on phonetic and semantic interpretation. Furthermore, the assumption of multiple functional structure in the participial clause is justified on theoretical grounds in the Antisymmetry framework. The correlation between accusative Case checking and participial agreement and that between nominative Case checking and finite verb agreement are captured in a maximally unified hypothesis. Crucially, I argued in chapters 2 and 6 that Case dependencies of both the accusative and the nominative kind can consist of an expletive and an associate DP. This means essentially that the central claims of the theory of subject pro drop in finite clauses (cf. Chomsky 1981, 1982, Rizzi 1982, Burzio 1986) may be assumed for object pro in participial clauses. In chapter 4, I assumed that both accusative and nominative checking are subject to a Case checking parameter, ensuring in what projection the Case feature is checked. The idea that nominative Case checking may take place in either of ArgSP or TP, and that the choice is parametrically decided (cf. Jonas 1992, Bures 1993, and Jonas & Bobaljik 1993) carries over to accusative checking.
327
In chapter 6, I showed that the complex patterns of the absolute participle in three stages of Italian can to some extent be derived on the assumptions justified in the preceding chapters. In addition, I argued that the participial clause may host a Tense head and that the effects of this element are clearly visible in the ancient varieties of Italian. My demonstration has given further empirical support to the hypothesis that Tense is crucially relevant for the checking of both accusative and nominative Case (cf. Roberts 1987, Belletti 1990, and Watanabe 1993, among many others). It was shown in chapter 7 that the object pro analysis proposed in chapters 2 and 6 accounts for null objects in coordinated structures. Again, the result was derived in the Antisymmetry format, and would not have lent itself as easily to an analysis in terms of the Minimalist Program of Chomsky (1995). Finally, in chapter 8 I discussed how the present hypothesis can be made compatible with the pro theory of Rizzi (1986a). I concluded that object pro in Modern Italian is mainly attested in adjectival environments and preferably as the complement of an affecting predicate. In terms of my analysis this means that object pro in Modern Italian is crucially dependent on AgrA or Asp, or both. The relevant restrictions can hardly have been valid in the ancient grammar. I believe that this analysis has contributed to our knowledge of the history of Italian in an interesting way. A number of very complex and apparently unrelated syntactic phenomena in ancient, literary Italian are presumably imputable to a few fundamental properties of the relevant grammatical system. Above all, four of these properties have been identified and discussed at some length: 1. Medieval Italian had a null pronominal element with the grammatical function of an object. This property was responsible for the complex inversion structures attested in literary prose, and to the omission of the lexical object in coordinated structures, at the same time. 2. A number of properties that distinguish the absolute participle in the ancient Italian grammars from the contemporary grammar are derived on the assumption that the absolute construction of ancient Italian included an abstract temporal element. 3. The OV patterns of participial constructions in Medieval Italian are explained on the assumption that accusative Case at the original stage of language was strong in the sense of Chomsky (1993). 4. The idea that abstract Case is checked in either of AspP or AgrOP and that the choice is parametrically determined, accounts for cross-linguistic variation in participial agreement patterns, including the differences between the three stages of Italian. My analysis has been primarily concerned with the problem of defining the properties of the three stages of Italian. On some points, the data has indicated when the change from the first to second system occurred, namely by the middle of the 14th century. My analysis does not
328
address the reasons why this change occurred, an aspect of the problem that belongs to the diachronic analysis of language. Given the amount of data derived on these assumptions, and given the complexity of the data, I conclude that the discussion has led to some interesting results, and I hope that the encounter between two disciplines such as historical syntax and theoretical linguistics can be considered mutually beneficial.
329
Ancient texts cited: Le Cene, Antonfrancesco Grazzini (Il Lasca). A cura di Riccardo Bruscagli, Salerno Editrice, Roma 1976. (I Novellieri Italiani, volume 27) Cinq Novelle italiane. Il Cinquecento. A cura di Mercello Ciccuto, Garzanti, Milano 1982. References to this anthology are given with the name of the novellist and page: ‘Cinq. Cademosto, p. ….’ Conv Il Convivio, Dante Alighieri. Presentazione, note e commenti di Piero Cudini. Garzanti 1980; III ed. 1990. Cort Il Libro del Cortegiano, Baldassar Castiglione. Introduzione di Amedeo Quondam. Note di Nicola Longo. Garzanti 1981; IV ed. 1991. Cron Cronica di Giovanni Villani. Scelta, introduzione e note di Giovanni Aquilecchia. Einaudi, Torino 1979. Dec Decamerone, Giovanni Boccaccio. A cura di Vittore Branca. Arnoldo Mondadori, Milano 1985; I ed. Oscar classici 1989. Dial Dialogo del reggimento di Firenze, Francesco Guicciardini. A cura di Gian Mario Anselmi & Carlo Varotti. Bollati Boringhieri, Torino 1994. Espos Esposizioni sopra la Comedia di Dante, Giovanni Boccaccio. A cura di Giorgio Padoan. Arnoldo Mondadori, Milano 1965. Fatti I fatti di Enea, Guido da Pisa. A cura di Francesco Fòffano, Sansoni, Firenze 1900; nuova presentazione di Franca Ageno, 1968. Fior I fioretti di San Francesco. In Scrittori di religione del trecento: Volgarizzamenti, tomo quarto. Classici Ricciardi. A cura di Don Giuseppe di Luca, Einaudi 1977. Gal Il Galateo, Giovanni Della Casa. A cura di Stefano Prandi. Einaudi, Torino 1994. Inf Inferno, Dante Alighieri. In La Commedia secondo l’antica vulgata. A cura di Giorgio Petrocchi. Arnoldo Mondadori, Milano I ed. 1966. LettA Lettere, Pietro Aretino. Introduzione, scelta e commento di Paolo Procaccioli. Rizzoli, Milano, 1 ed. 1990. LettM Dieci lettere private, Niccolò Machiavelli. A cura di Giovanni Bardazzi. Salerno Editrice, Roma, 1992. Lucchesi References to texts cited in Lucchesi’s work are given in the form: ‘Lucchesi, Tavola Ritonda …’. The page number refers to Lucchesi (1962/63). Luc I Lucidi, Agnolo Firenzuola. A cura di Adriano Seroni. Sansoni Editore, Firenze 1958; I ed. Le Betulle 1991. Mand La Mandragola, Machiavelli. Introduzione e note di Gennaro Sasso. Biblioteca universale Rizzoli, Milano 1980; II ed. 1983. MMin Storia di fra’ Michele Minorita. A cura di Emanuele Trevi. Salerno Editrice, Roma 1991. Cene
330
Navig Nov.no Nov.re Par Pec Princ Prose Purg RagA RagF RicM Trec Vita Vizi VN Zucca
La navigazione di San Brandano. A cura di Maria Antonietta Grignani. Bompiani, II ed. Milano 1992. Il Novellino. A cura di Cesare Segre. Ricciardi editore, Milano-Napoli 1959. In Prosatori del duecento. Einaudi, Torino 1976. Il Novelliere, Giovanni Sercambi. A cura di Luciano Rossi. Salerno Editrice, Roma 1974. (I Novellieri Italiani, volume 9; tomo I) Paradiso, Dante Alighieri. In La Commedia secondo l’antica vulgata. A cura di Giorgio Petrocchi. Arnoldo Mondadori, I ed. Milano 1967. Il Pecorone, Ser Giovanni. A cura di Enzo Esposito. Longo Editore, Ravenna 1974. Il Principe, Niccolò Machiavelli. A cura di Luigi Firpo. Einaudi, Torino 1961; XII ed. 1981. Prose della volgar lingua, Pietro Bembo. A cura di Carlo Dionisotti, UTET, Torino 1966; I classici Italiani TEA, II ed. 1993. Purgatorio, Dante Alighieri. In La Commedia secondo l’antica vulgata. A cura di Giorgio Petrocchi. Arnoldo Mondadori, I ed. Milano 1967. Ragionamenti, Pietro Aretino. Introduzione di Giorgio Bàrberi Squarotti. Commento di Carla Forno. Biblioteca universale Rizzoli, Milano 1988. I Ragionamenti, Agnolo Firenzuola. A cura di Adriano Seroni. Sansoni Editore, Firenze 1958; I ed. Le Betulle 1991. I Ricordi di Giovanni di Pagolo Morelli. A cura di Vittore Branca. Le Monnier, Firenze 1956. Il Trecentonovelle, Franco Sacchetti. A cura di Antonio Lanza. Sansoni Editore, Firenze 1984; I ed. Le Betulle 1990. Vita, Benvenuto Cellini. A cura di Ettore Camesasca. Rizzoli, Milano 1985. Il Libro de’ Vizî e delle Virtudi, Bono Giamboni. A cura di Cesare Segre, Giulio Einaudi Editore, Torino 1968. Vita Nuova, Dante Alighieri. A cura di Fredi Chiappelli. Mursia Editore, Milano 1965. La Zucca, Anton Francesco Doni. A cura di Carlo Cordié, Riccardo Ricciardi Editore, Milano-Napoli 1976. (In: Folengo, Aretino, Doni, tomo II)
331
References: Aarts, B.
(1992) Small Clauses in English. Mouton de Gruyter, Berlin.
Ackema, P. & M. Schoorlemmer (1995) Middles and Nonmovement. Linguistic Inquiry 26, 173-197. Adams, M.
(1987) Old French, Null Subjects, and Verb Second Phenomena. Doctoral dissertation UCLA.
Adams, M.
(1989) Verb second effects in Medieval French. In Kirschner & De Cesaris (eds.)
Ageno, F. B.
(1964) Il Verbo nell’italiano antico. Riccardo Ricciardi Editore, MilanoNapoli.
Ageno, F. B.
(1971) Osservazioni sull’aspetto e il tempo del verbo nella ‘Commedia’. Studi di Grammatica Italiana vol I; pp. 61-100.
Alisova, T.
(1967) Studi di sintassi italiana I; Forme di subordinazione relativa nell’italiano antico (sec. XIII-XV). Studi di Filologia Italiana XXV, 223313.
Ambrosini, E.
(1960/61) L’uso dei tempi storici nell’italiano antico. L’Italia dialettale XXIV, 13-124.
Anderson, M.
(1979) Noun Phrase Structure. Ph.D. dissertation, University of Connecticut.
Antinucci, F.
(1977) Fondamenti di una teoria tipologica del linguaggio. Il Mulino, Bologna.
Arad, M.
(1995) On the projection of ditransitive verbs. University College of London Working Papers in Linguistics 7, 215-233.
Authier, J.-M.
(1991) V-governed expletives, Case Theory, and the Projection Principle. Linguistic Inquiry 22, 721-740.
332
Authier, J.-M.
(1992) A Parametric Account of V-Governed Arbitrary Null Arguments. Natural Language and Linguistic Theory 10, 345-374.
Baker, M., K. Johnson & I. Roberts (1989) Passive arguments raised. Linguistic Inquiry 20, 219-251. Banfi, E. & P. Cordin (a c. di) (1990) Storia dell’italiano e forme dell’italianizzazione. Atti del XXIII congresso internazionale di studi. Bulzoni, Roma. Battye, A. & I. Roberts (eds.) (1995) Clause Structure and Language Change. Oxford University Press, New York - Oxford. Belletti, A.
(1988) The Case of Unaccusatives. Linguistic Inquiry 19.
Belletti, A.
(1990) Generalized Verb Movement. Aspects of Verb Syntax. Rosenberg & Sellier, Torino.
Belletti, A.
(1993) Case Checking and Clitic Placement. Geneva Generative Papers 1:2, 101- 118.
Benincà, P.
(1993) Sintassi. In A. Sobrero (a c. di) Introduzione all’italiano contemporaneo, Le strutture. Laterza, Roma-Bari 1993, 247-290.
Benincà, P.
(1994) La variazione sintattica. Il Mulino, Bologna.
Benincà, P.
(1995) Complement Clitics in Medieval Romance: the Tobler-Mussafia Law. In Battye & Roberts (eds.), 325-344.
Benincà, P. (ed.) (1989) Dialect Variation and the Theory of Grammar. Foris, Dordrecht. Benincà, P., G. Salvi & L. Frison. (1988) L’ordine degli elementi della frase e le costruzioni marcate. In Renzi (a c. di) 1988, 115-226. Benincà, P. & G. Cinque (1993) Su alcune differenze fra enclisi e proclisi. In Omaggio a Gianfranco Folena. Editoriale Programma, 2313-2326. Bertinetto, P. M. (1986) Tempo, Aspetto e Azione nel verbo italiano. Il sistema dell’indicativo. Accademia della Crusca, Firenze. 333
Bertinetto, P. M. (1991) Il verbo. In Renzi, L. & G. Salvi (a c. di) 1991, 13-162. Bertuccelli Papi, M. (1991) Il participio passato. In Renzi, L. & G. Salvi (a c. di) 1991, 593604. Bianchi, V. & M. C. Figueiredo Silva (1993) On Some Properties of Agreement-Object in Italian and in Brazilian Portuguese. Ms. Scuola Normale di Pisa and University of Genève. Borer, H.
(1991) The Causative-Inchoative Alternation: A Case Study in Parallel Morphology. The Linguistic Review 8, 119-158.
Borer, H.
(1993) The Projection of Arguments. In Benedicto, E. & J. Runner (eds.) Functional Projections, University of Massachusetts 1993.
Borer, H.
(1995) Passive without Theta Grids. Ms. University of Massachusetts, Amherst and Hebrew University, Jerusalem.
Borer, H. (ed.)
(1986) The Syntax of Pronominal Clitics. Syntax and Semantics 19. Academic Press, Orlando etc.
Bosque, I.
(1990) Sobre el aspecto en los adjetivos y en los participios. In Bosque (ed.) Tiempo y aspecto en español, Cátedera, Madrid 1990.
Bottari, P.
(1992) Romance ‘passive nominals’. Geneva Generative Papers 0:0. 66-76.
Brandi, L. & P. Cordin (1981) Dialetti e italiano: un confronto sul Parametro del Soggetto Nullo. Rivista di Grammatica Generativa 6, 33-87. Brandi, L. & P. Cordin (1989) Two Italian Dialects and the Null Subject Parameter. In Jaeggli, O. & K. Safir (eds.), 111-142. Branigan, P.
(1992) Subjects and Complementizers. Ph.D. dissertation, MIT.
Bures, A.
(1993) There is an Argument for an LF cycle here. In CLS 28 vol. 2, The Cycle in Linguistic Theory, 14-35.
334
Burzio, L.
(1986) Italian Syntax. A Government-Binding Approach. Reidel, Dordrecht.
Calabrese, A.
(1988) I pronomi clitici. In Renzi (a c. di) 1988, 549-592.
Calabrese, A.
(1992) Some Remarks on Focus and Logical Structures in Italian. Harward Working Papers in Linguistics 1, 91-127.
Campos, H.
(1986) Indefinite Object Drop. Linguistic Inquiry 17, 354-359.
Cardinaletti, A. (1991) On Pronoun Dative Movement. The Italian Dative Loro. Probus 3, 127-153. Cardinaletti, A. (1994a) La sintassi dei pronomi. Uno studio comparativo delle lingue germaniche e romanze. Il Mulino, Bologna. Cardinaletti, A. (1994b) Agreement and Control in Expletive Constructions. Ms. MIT. Cardinaletti, A. (1995) Pronouns in Germanic and Romance Languages: an Overview. Ms. University of Venice. Cardinaletti, A. & I. Roberts (1991) Clause Structure and X-Second. Ms. Universities of Venezia and Genève. Published in Chao, W. & G. Horrocks (eds.) Levels, Principles, and Processes: The Structure of Grammatical Representations. Foris, Dordrecht. Cardinaletti, A. & M. T. Guasti (1992a) Epistemic Small Clauses and Null Subjects. ESCOL Proceedings. Cardinaletti, A. & M. T. Guasti (1992b) Negation in Small Clauses. University of Venice Working Papers in Linguistics I.5, 1-34. Cardinaletti, A., G. Cinque & G. Giusti (eds.) (1988) Constituent Structure. Foris, Dordrecht. Ceserani, R. & L. De Federicis (1979a) Il materiale e l’immaginario. La società urbana. Loescher Editore, Torino. Ceserani, R. & L. De Federicis (1979b) Il materiale e l’immaginario. La società signorile. Loescher Editore, Torino. 335
Chomsky, N.
(1981) Lectures on Government and Binding. Foris, Dordrecht.
Chomsky, N.
(1982) Some Concepts and Consequences of the Theory of Government and Binding. Linguistic Inquiry Monograph 6, MIT Press, Cambridge, Massachusetts.
Chomsky, N.
(1986a) Knowledge of Language. Praeger, New York.
Chomsky, N.
(1986b) Barriers. Linguistic Inquiry Monograph 13, MIT Press, Cambridge, Massachusetts.
Chomsky, N.
(1991) Some Notes on Economy of Derivation and Representation. In R. Freidin (ed.) Principles and Parameters in Comparative Grammar, MIT Press, Cambridge, Massachusetts, 417-454. Reprinted in Chomsky 1995.
Chomsky, N.
(1993) A Minimalist Program for Linguistic Theory. In Hale, K. & S. J. Keyser (eds.) 1993. Reprinted in Chomsky 1995.
Chomsky, N.
(1995) The Minimalist Program. MIT Press, Cambridge, Massachusetts.
Chomsky, N. & H. Lasnik (1993) The Theory of Principles and Parameters. In Jacobs, J., A. von Stechow, W. Sternefeld & T. Vennemann (eds.) Syntax: An International Handbook of Contemporary Research. Walter de Gruyter, Berlin & New York 1993. Reprinted in Chomsky 1995. Christensen K. K. & K. T. Taraldsen (1989) Expletive Chain Formation. In Benincà (ed.) 1989, 53-83. Cinque, G.
(1988a) On Si Constructions and the Theory of Arb. Linguistic Inquiry 19, 521-581.
Cinque, G.
(1988b) La frase relativa. In Renzi (a c. di) 1988, 443-506.
Cinque, G.
(1990a) Types of A’-dependencies. Linguistic Inquiry Monograph 17. MIT Press, Cambridge, Massachusetts.
336
Cinque, G.
(1990b) Ergative Adjectives and the Lexicalist Hypothesis. Natural Language and Linguistic Theory 8, 1-40.
Cinque, G.
(1991) Teoria linguistica e sintassi italiana. Il Mulino, Bologna.
Cinque, G.
(1992) On leftward movement of tutto in Italian. University of Venice Working Papers in Linguistics 2:6.
Cinque, G.
(1994) Movimento del participio e struttura funzionale della frase. Talk presented at the XX Incontro di Grammatica Generativa, Padova 1994.
Clark, R. & I. Roberts (1993) A Computational Model of Language Learnability and Language Change. Linguistic Inquiry 24, 299-345. Cocchi, G.
(1994) Sintassi della selezione dell’ausiliare. Tesi di dottorato di ricerca di linguistica italiana, Firenze.
Cole, P.
(1987) Null Objects in Universal Grammar. Linguistic Inquiry 18, 597-612.
Comrie, B.
(1976) Aspect. Cambridge University Press.
Comrie, B.
(1981) Aspect and Voice: Some Reflections on Perfect and Passive. In Tedeschi, P. & A. Zaenen (eds.), Syntax and Semantics 14, Academic Press, New York 1981.
Contreras, H.
(1987) Small Clauses in Spanish and English. Natural Language and Linguistic Theory 5 (2), 225-243.
Cornilescu, A.
(1993) Notes on the Structure of the Romanian DP and the Assignment of the Genitive Case. University of Venice Working Papers in Linguistics 3:2, 107-133.
D’Achille, P.
(1990a) I fenomeni di ‘tematizzazione’ di fronte alla codificazione cinquecentesca. In Banfi & Cordin (a c. di), 283-294.
D’Achille, P.
(1990b) Sintassi del parlato e tradizione scritta della lingua italiana. Analisi di testi dalle origini al secolo XVIII. Bonacci, Roma.
337
Dahl, Ö.
(1981) On the Definition of the Telic-Atelic (Bounded-Nonbounded) Distinction. In Tedeschi, P. & A. Zaenen (eds.), Syntax and Semantics 14, Academic Press, New York etc. 1981, 79-90.
Dardano, M.
(1963) L’arte del periodo nel Cortegiano. La Rassegna della letteratura italiana 67, 456-458.
Dardano, M.
(1967) Note sul Bestiario toscano. Italia dialettale XXX, 29-117.
De Hoop, H.
(1992) Case Configuration and Noun Phrase Interpretation. Ph.D. Dissertation, University of Groningen.
Delfitto, D. & M. Pinto (1992) How free is free inversion? Recherches de Linguistique Française e Romance d’Utrecht 11. Delsing, L.-O.
(1993) The Internal Structure of Noun Phrases in the Scandinavian Languages: a Comparative Study. Doctoral Dissertation, Department of Scandinavian Languages, Lund.
Déprez, V.
(1989) On the Typology of Syntactic Positions and the Nature of Chains. MIT Ph.D dissertation.
Diesing M. & Jelinek (1993) The Syntax and Semantics of Object Shift. Working Papers in Scandinavian Syntax 51, 1-54. Dini, L.
(1994) Aspectual Constraints on Italian Absolute Phrases. Quaderni del Laboratorio di Linguistica 8, 52-88.
DiSciullo, A. M. & E. Williams (1987) On the definition of Word. Linguistic Inquiry Monograph 14, MIT Press, Cambridge, Massachusetts. Egerland, V.
(1994) Sul verbo ‘avere’ ed il participio passato nell’italiano antico: un approccio generativo. Quaderni del Dipartimento di Linguistica, Università di Firenze 5, 95-113.
Egerland, V.
(1995) Spec-Head Agreement, Case Theory and the Syntax of Participles: Absolute Participial Clauses in Old Italian. Rivista di Grammatica Generativa 20, 33-68. 338
Ekbo, S.
(1943) Studier över uppkomsten av supinum i de germanska språken med utgångspunkt i fornvästnordiskan. A. - B. Lundequistska Bokhandeln. Uppsala Universitets Årsskrift 7.
Emonds, J.
(1985) A Unified Theory of Syntactic Categories. Foris, Dordrecht.
Fagan, S.
(1988) The English middle. Linguistic Inquiry 19, 181-203.
Falk, C.
(1989) On the Existential Construction in the Germanic Languages. Working Papers in Scandinavian Syntax 32.
Falk, C.
(1993) Non-Referential Subjects in the History of Swedish. Ph.D. dissertation, Department of Scandinavian Languages, Lund.
Farrell, P.
(1990) Null Objects in Brazilian Portuguese. Natural Language and Linguistic Theory 8, 325-346.
Ferguson, K. S. & E. Groat (1994) Defining ‘Shortest Move’. talk presented at the GLOW colloquium, Wien 1994. Glow Newsletter 32. Friedmann, M. & T. Siloni (1993) AgrO is not AgrP: Complex Tenses in Hebrew and French. Ms. Université de Genève. Ghinassi, G.
(1971) Casi di paraipotassi relativa in italiano antico. Studi di Grammatica Italiana vol I, 45-60.
Giorgi, A. & G. Longobardi (1991) The Syntax of Noun Phrases. Cambridge University Press, Cambridge etc. Greenberg, J. H. (1966) Some Universals of Grammar with Particular Reference to the Order of Meaningful Elements. In Greenberg J. H. (ed.) Universals of Language, 73-113, MIT Press Cambridge, Massachusetts. Grimshaw, J.
(1990) Argument Structure. Linguistic Inquiry Monograph 18, MIT Press, Cambridge, Massachusetts.
339
Grimshaw, J. & S. Vikner (1993) Obligatory Adjuncts and the Structure of Events. In Reuland & Abraham (eds.), 143-155. Guasti, M. T.
(1992) Causative and Perception Verbs. Ph.D. dissertation, University of Geneva.
Guéron, J.
(1986) Le verbe avoir. In P. Coopmans, I. Bordellois & B. Dotson Smith (eds.) Formal Parameters of Generative Grammar II, Dordrecht, ICG Printing, 83-105.
Guéron, J. & T. Hoekstra (1988) T-Chains and the Constituent Structure of Auxiliaries. In Cardinaletti, A., G. Cinque & G. Giusti (eds.), 35-99. Guéron, J. & J.-Y. Pollock (eds.) (1991) Grammaire Générative et Syntaxe Comparée. Éditions de CNRS, Paris. Gunnarson, K.-Å. (1992) Adjectival Participle Phrases and Small Clauses. Studia Linguistica 46, 63-71. Gunnarson, K.-Å. (1995) Small Clauses and Absolute Constructions in Spanish. Probus 7, 125-171. Gärtner, H.-M. & M. Steinbach (1994) Economy, Verb Second, and SVO-SOV Distinction. Working Papers in Scandinavian Syntax 53, 1-59. Haegeman, L.
(1985) The get-passive and Burzio’s generalization. Lingua 66, 53-77.
Hale, K. & S. J. Keyser (1993) On Argument Structure and the Lexical Expression of Syntactic Relations. In Hale & Keyser (eds.) 1993. Hale, K. & S. J. Keyser (eds.) (1993) The View from Building 20: Essays in Linguistics in Honor of Sylvain Bromberger. MIT Press, Cambridge, Massachusetts. Harre, C. E.
(1991) Tener + Past Participle. A case study in linguistic description. Routledge, London & New York.
Harris, M.
(1982) The ’Past Simple’ and the ’Present Perfect’ in Romance. In Vincent & Harris (eds.) 340
Harris, M. & N. Vincent (1988) (eds.) The Romance Languages. Croom Helm London & Sydney. Hedlund, C.
(1992) On Participles. Ph.D. dissertation, Department of Linguistics, University of Stockholm.
Herczeg, G.
(1972) Saggi Linguistici e stilistici. Accademia toscana di scienze e lettere ‘La Colombaria’, XXIV, Leo Olschki Editore, Firenze.
Hoekstra, T. & R. Mulder (1990) Unergatives as Copular Verbs. Locational and Existential Predication. The Linguistic Review 7, 1-79. Hoekstra, T. & I. Roberts (1993) Middle Constructions in Dutch and English. In Reuland & Abraham (eds.) Holmberg, A.
(1986) Word Order and Syntactic Features in the Scandinavian Languages and English. Ph.D. dissertation, University of Stockholm.
Holmberg, A. & C. Platzack (1995) The Role of Inflection in Scandinavian Syntax. Oxford University Press, New York, Oxford. Hornstein, N.
(1990) As Time Goes By. Tense and Universal Grammar. MIT Press, Cambridge, Massachusetts.
Huang, J.
(1984) On the Distribution and Reference of Empty Pronouns. Linguistic Inquiry 15, 531-575.
Huang, J.
(1989) Pro-Drop in Chinese: a Generalized Control Theory. In Jaeggli, O. & K. Safir (eds.) 1989.
Hulk, A. & A. van Kemenade (1995) Verb Second, Pro-Drop, Functional Projections, and Language Change. In Battye, A. & I. Roberts (eds.) 1995. Jaeggli. O.
(1986) Passive. Linguistic Inquiry 17, 587-622.
Jaeggli, O. & Safir, K. (eds.) (1989) The Null Subject Parameter. Kluwer, Dordrecht.
341
Johnson, K.
(1991) Object Positions. Natural Language and Linguistic Theory 9, 577636.
Jonas, D.
(1992) Checking Theory and Nominative Case in Icelandic. Harward Working Papers in Linguistics 1, 175-196.
Jonas, D. & J. Bobaljik (1993) Specs for Subjects. The Role of TP in Icelandic. In Bobaljik, J. & C. Philips (eds.) MIT Working Papers in Linguistics 18. Papers on Case and Agreement I. Department of Linguistics and Philosophy, MIT. Jones, M.
(1988) Sardinian. In Harris, M. & N. Vincent (eds.), 314-350.
Josefsson, G.
(1992) Object Shift and Weak Pronominals in Swedish. Working Papers in Scandinavian Syntax 49, 59-94
Josefsson, G.
(1993) Scandinavian Pronouns and Object Shift. Working Papers in Scandinavian Syntax 52, 1-28.
Junker, M.-O. & F. Martineau (1992) The Structure of Infinitives. Probus 4. Kayne, R. S.
(1975) French Syntax. The Transformational Cycle. The MIT Press, Cambridge, Massachusetts.
Kayne, R.S.
(1989a) Facets of Romance Past Participle Agreement. In Benincà (ed.) 1989, 85-103.
Kayne, R. S.
(1989b) Null Subjects and Clitic Climbing. In Jaeggli, O. & K. J. Safir (eds.), 239-261.
Kayne, R. S.
(1991) Romance Clitics, Verb Movement, and PRO. Linguistic Inquiry 22, 647-686.
Kayne, R. S.
(1993) Toward a Modular Theory of Auxiliary Selection. Studia Linguistica 47, 3-31.
Kayne, R. S.
(1994) The Antisymmetry of Syntax. Linguistic Inquiry Monograph 25, MIT Press, Cambridge, Massachusetts.
342
Kemenade, A. van (1987) Syntactic Case and Morphological Case in the History of English. Foris, Dordrecht. Keyser, S. & T. Roeper (1984) On the Middle and Ergative Constructions in English. Linguistic Inquiry 15, 381-416. Kirschner, C. & J. De Cesaris. (eds.) (1989) Studies in Romance Linguistics. Current Issues in Linguistic Theory 60, John Benjamins Publishing Company, Amsterdam & Philadelphia. Koopman, H. & D. Sportiche (1991) The Position of Subjects. Lingua 85, 211-258. Koster, J.
(1975) Dutch as an SOV-Language. Linguistic Analysis 1, 111-136.
Kroch, A.
(1989) Reflexes of Grammar in Patterns of Language Change. Language Variation and Change 1, 199-244.
Kroch, A.
(1994) Morphosyntactic Variation. to appear in K. Beals et al., (eds.) Papers from the 30th Regional Meeting of the Chicago Linguistics Society: Parasession on Variation and Linguistic Theory.
Källström, R.
(1993) Kongruens i svenskan. Acta Universitatis Gothoburgensis, Göteborg.
La Fauci, N.
(1989) Ausiliari perfettivi e accordo del participio passato in italiano e in francese. In Foresti, F., E. Rizzi & P. Benedini ((a c. di)) L’italiano tra le lingue romanze. Atti del XX Congresso SLI 27, 213-242.
Larson, R.
(1988) On the Double Object Construction. Linguistic Inquiry 19, 335-391.
Lasnik, H.
(1992) Case and Expletives: Notes Toward a Parametric Account. Linguistic Inquiry 23, 381-406.
Lema, J. & M.-L. Rivero (1989) Long Head Movement: ECP vs. HMC. North Eastern Linguistic Society 20, 333-347. Levin, B. & M. Rappaport (1986) The Formation of Adjectival Passives. Linguistic Inquiry 17, 623-662.
343
Lightfoot, D.
(1979) Principles of Diachronic Syntax. Cambridge University Press, Cambridge, London, New York, Melbourne.
Lightfoot, D.
(1991) How to Set Parameters. Arguments from Language Change. MIT Press, Cambridge, Massachusetts.
Lois, X.
(1990) Auxiliary Selection and Past Participle Agreement in Romance. Probus 2, 233-255.
Longobardi, G. (1995) The Syntax of Genitive. Part one: Construct State across Languages. a Minimalist Interpretation. Draft, University of Venice. Lonzi, L.
(1988) Tipi di gerundio. Rivista di Grammatica Generativa 13, 59-80.
Lonzi, L.
(1991) Frasi subordinate al gerundio. In Renzi, L. & G. Salvi (a c. di) 1991, 571-592.
Lucchesi, V.
(1962/63) Participio e accordo nel volgare antico. Atti dell’accademia di scienze e letteratura ‘La Colombaria’ 27, 193-278.
Mahajan, A.
(1990) The A/A’-Distinction and Movement Theory. MIT Doctoral dissertation.
Mahajan, A.
(1994) Split Ergativity, Auxiliary Selection and Word Order Directionality. talken given at the GLOW colloquium Wien. GLOW Newsletter 32, 38-39.
Maiden, M.
(1995) A Linguistic History of Italian. Longman, London and New York.
Maling, J.
(1980) Inversion in Embedded Clauses in Modern Icelandic. Íslenskt mál 2, 175-193.
Manzini, M. R. (1983) Restructuring and Reanalysis. MIT Doctoral dissertation. Manzini, M. R. (1986) On Italian Si. In Borer, H. (ed.), Syntax and Semantics 19, 241-262. Manzini, M.R.
(1988) Constituent Structure and Locality. In Cardinaletti, A., G. Cinque & G. Giusti (eds.) 1988.
344
Manzini, M. R. (1992) Locality: A theory and some of its empirical consequences. Linguistic Inquiry Monograph 19, Cambridge Mass. MIT Press. Manzini, M. R. (1994a) Locality, Minimalism, and Parasitic Gaps. Linguistic Inquiry 25. Manzini, M. R. (1994b) Syntactic Dependencies and their properties: weak islands. Quaderni del Dipartimento di Linguistica di Firenze 5, 75-94. Manzini, M. R. & K. Wexler (1987) Parameters, Binding Theory, and Learnability. Linguistic Inquiry 18, 413-444. Marcantonio, A. (1976) Un aspetto dell’ordine delle parole nell’italiano del due-trecento. Rivista di Grammatica Generativa 1, 55-77. Ménard, P.
(1994) Syntaxe de l’ancien français. Éditions Bière, Bordeaux.
Merlo, P.
(1989) Risultative in italiano e in inglese. Rivista di Grammatica Generativa 14, 29-53.
Migliorini, B.
(1988) Storia della lingua italiana: I. Introduzione. Di Ghino Ghinassi, Sansoni, Firenze.
Moro, A.
(1993) I predicati nominali e la struttura della frase. Unipress, Padova.
Mourelatos, A. P. D. (1978) Events, Processes and States. Linguistics and Philosophy 2, 415434. Parkinson, S.
(1988) Portuguese. In Harris & Vincent (eds.), 131-169.
Parodi, E. G.
(1957) Lingua e letteratura. Studi di teoria linguistica e di storia dell’italiano antico. (a c. di) Gianfranco Folena, Biblioteca di cultura 15-16, Venezia 1957.
Pettiward, A.
(1995) Anti-Procrastinate effects and optional agreement in French. talk presented at The University of Amsterdam is going Romance Amsterdam 1995.
345
Platzack, C.
(1979) The Semantic Interpretation of Aspect and Aktionsarten. Foris, Dordrecht.
Platzack, C.
(1989) The Swedish Supine: An Active Verb Form or the Non-agreeing Form of the Past Participle?, in Jaspers, D., W. Klosters, Y. Putseys & P Seuren (eds.), Sentential Complementation and the Lexicon, Foris, Dordrecht 1989, 305-319.
Platzack, C.
(1992) Complementizer Agreement and Argument Clitics. Working Papers in Scandinavian Syntax 50, 25-54.
Platzack, C.
(1994a) Null Subjects, Weak Agr and Syntactic Differences in Scandinavian. Working Papers in Scandinavian Syntax, 53, 85 -106.
Platzack, C.
(1994b) Syntactic differences in the Comp-domain: Mainland Scandinavian evidence for Split Comp. talk given at the GLOW Colloquium, Wien 1994. GLOW Newsletter 32.
Platzack, C.
(1995) The Loss of Verb Second in English and French. In Battye & Roberts (eds.) 1995, 200-226.
Platzack, C. & A. Holmberg (1989) The Role of Agr and Finiteness. Working Papers in Scandinavian Syntax 43, 51-76. Poletto, C.
(1993) La sintassi del soggetto nei dialetti italiani settentrionali. Unipress, Padova.
Pollock, J.-Y.
(1989) Verb Movement, UG and the Structure of IP. Linguistic Inquiry 20, 365-424.
Raposo, E.
(1987).Case Theory and INFL-to-COMP. Linguistic Inquiry 18.
Raposo, E. & J. Uriagereka (1993) Two types of small clauses. Talk presented at the GLOW Colloquium, Lund 1993. Renzi, L. (a c. di) (1988).Grande grammatica italiana di consultazione. I. Il Mulino, Bologna.
346
Renzi, L. & G. Salvi (a c. di) (1991).Grande grammatica italiana di consultazione. II. Il Mulino, Bologna. Reuland, E. & W. Abraham (eds.) (1993) Knowledge and Language. II: Lexical and Conceptual Structure. Kluwer Academic Publishers 1993 Ricciardi, M.
(1988) La Letteratura in Italia. Profilo Storico. Bompiani, Milano.
Rivero, M.-L.
(1991) Long Head Movement vs. V2 and Null Subjects in Old Romance. paper presented at the Workshop on Null Subjects, Xth International Conference on Historical Linguistics. Vrije Universiteit, Amsterdam, Holland 1991.
Rivero, M.-L.
(1992) Clitic and NP Climbing in Old Spanish. In Campos, H. & F. Martinez-Gil (eds.) Current Studies in Spanish Linguistics. Georgetown University Press, Washington D.C.
Rizzi, L.
(1982) Issues in Italian Syntax. Foris, Dordrecht.
Rizzi, L.
(1986a) Null Objects in Italian and the Theory of pro. Linguistic Inquiry 17, 501-557.
Rizzi, L.
(1986b) On the Status of Subject Clitics in Romance. In Jaeggli, O. & C. Silva-Carvalán (eds.) Studies in Romance Linguistics, Foris, Dordrecht, 391-419.
Rizzi, L.
(1990) Relativized Minimality. Linguistic Inquiry Monograph 16, MIT Press, Cambridge, Massachusetts.
Rizzi, L.
(1991) Residual Verb Second and the Wh-Criterion. Technical Reports in Formal and Computational Linguistics, 2. Université de Genève.
Rizzi, L. & I. Roberts (1989) Complex Inversion in French. Probus 1, 1-30. Roberge, J.-Y.
(1991) On the Recoverability of Null Objects. In Wanner & Kibbee (eds.), 299-312.
Roberts, I.
(1987) Implicit and Dethematized Subjects. Foris, Dordrecht. 347
Roberts, I.
(1991) Excorporation and Minimality. Linguistic Inquiry 22, 209-218.
Roberts, I.
(1993) Verbs and Diachronic Syntax. Kluwer, Dordrecht.
Rohlfs, G.
(1968) Grammatica storica della lingua italiana e dei suoi dialetti; Morfologia. Einaudi, Torino.
Rohlfs, G.
(1969) Grammatica storica della lingua italiana e dei suoi dialetti; Sintassi. Einaudi, Torino.
Rothstein, S. D. (1995) Pleonastics and the Interpretation of Pronouns. Linguistic Inquiry 26, 499-529. Rouveret, A.
(1991a) Functional Categories and Agreement. The Linguistic Review 8, 353-387.
Rouveret, A.
(1991b) Clitisation et temps en portugais européen. In Guéron & Pollock (eds.), 147-170.
Rouveret, A. & J.-R. Vergnaud (1980) Specifying Reference to the Subject: French Causatives and Conditions on Representations. Linguistic Inquiry 11, 97202. Ruwet, N.
(1982) Grammaire des insultes et autres études. Seuil, Paris.
Ruwet, N.
(1983) Du bon usage Linguistiques 11, 5-84.
des
expressions
idiomatiques.
Recherches
Rögnvaldsson, E. (1990) Null Objects in Icelandic. In Syntax and Semantics 24, Maling, J. & A. Zaenen eds, Academic Press, San Diego etc. Safir, K.
(1986) Subject Clitics and the Nom-Drop Parameter. In Borer, H. (ed.) Syntax and Semantics 19, 333-356.
Salvi, G.
(1982) Sulla storia sintattica della costruzione romanza habeo + participio, Révue Romane 17, 118-133.
348
Salvi, G.
(1986) Asimmetrie soggetto/tema in italiano. In H. Stammerjohann (ed.) Tema-Rema in italiano, Gunter Narr Verlag, Tübingen 1986, 37-53.
Salvi, G.
(1991) L’accordo. In Renzi & Salvi (a c. di), 227-244.
Santorini, B.
(1989) The Generalization of the Verb-Second Constraint in the History of Yiddish. Doctoral dissertation, University of Pennsylvania.
Segre, C.
(1963) Lingua, Stile and Società. Milano, Feltrinelli.
Shlonsky, Ur
(1990) Pro in Hebrew Subject Inversion. Linguistic Inquiry 21, 263-275.
Sigurðsson, H.
(1992a) The Case of Quirky Subjects. Working Papers in Scandinavian Syntax 49, 1-26
Sigurðsson, H.
(1992b) Agreement as Visible F-government. Working Papers in Scandinavian Syntax 50, 55-74.
Sigurðsson, H.
(1992c) Argument-Drop in Old Icelandic. Ms. Institute of Linguistics, University of Iceland.
Škerlj, S.
(1932) Costrutti participiali del tipo ’veduto la bellezza’. L’Italia dialettale VIII; 117-178.
Smith, J. C.
(1991) Thematicity and ‘Object’-Participle Agreement in Romance. In Wanner & Kibbee (eds.), 335-352.
Speas, M.
(1991) Functional Heads and Inflectional Morphemes. The Linguistic Review 8, 389-417.
Sprouse, R. A. & B. Vance (1993) An Explanation for the Loss of Null Subjects in Certain Romance and Germanic Languages. Talk presented at the GLOW Colloquium, Lund 1993. Squartini, M. & P. M. Bertinetto (forthcoming) The Simple and Compound Past in Romance Languages. to appear in Dahl, Ö. (ed.) Tense and Aspect in the Languages of Europe. Mouton de Gruyter, Berlin.
349
Stowell, T.
(1981) Origins of Phrase Structure. Doctoral Dissertation, MIT.
Stowell, T.
(1983) Subjects across Categories. The Linguistic Review 2: 285-312.
Stroik, T.
(1992) Middles and movement. Linguistic Inquiry 23; 127-137.
Stroik, T.
(1995) On Middle Formation: A Reply to Zribi-Hertz. Linguistic Inquiry 26, 165-171.
Tekavčić, P.
(1980) Grammatica storica dell’italiano II: Morfosintassi. Il Mulino, Bologna.
Tenny, C.
(1987) Grammaticalizing Aspect and Affectedness. Doctoral dissertation, MIT.
Travis, L.
(1995) Binding Categories, Event Structure, and L-syntax vs. S-syntax. Talk presented at Clausal Architecture: Temporal, Aspectual and Verbal Projections, Bergamo 1995.
Uriagereka, J.
(1995) Aspects of the Syntax of Clitic Placement in Western Romance. Linguistic Inquiry 26, 79-123.
Valois, D.
(1991) The Internal Syntax of DP. PhD dissertation, University of California, Los Angeles.
Vance, B.
(1989) Null Subjects and Syntactic Change in Medieval French. Doctoral dissertation, Cornell University.
Vance, B.
(1995) On the Decline of Verb Movement to Comp in Old and Middle French. In Battye & Roberts (eds.), 173-199.
Vanelli, L., L. Renzi & P. Benincà (1985) Typologie des pronoms sujets dans les langues romanes. Actes du XVIIe Congrès International de Linguistique et Philologies Romanes 3, 164-176. Vanelli, L.
(1986) Strutture tematiche in italiano antico. In H. Stammerjohann (ed.) Tema-Rema in italiano, Gunter Narr Verlag, Tübingen 1986.
350
Vikner, S.
(1990) Verb Movement and the Licensing of NP-Positions in the Germanic Languages. Doctoral Dissertation, Université de Genève.
Vikner, S. & R. Sprouse (1988) Have/Be-Selection as an A-Chain Membership Requirement. Working Papers in Scandinavian Syntax 38. Vincent, N.
(1982) The development of the Auxiliaries HABERE and ESSE in Romance, in Vincent & Harris (eds.) 1982, 71-96.
Vincent, N. & M. Harris (eds.) (1982) Studies in the Romance Verb. Croom Helm, London & Canberra. Wanner, D. & D. A. Kibbee (eds.) (1991) New Analysis in Romance Linguistics. Current Issues in Linguistic Theory 69, John Benjamins Publishing Company, Amsterdam-Philadelphia. Watanabe, A.
(1993) Agr-Based Case Theory and Its Interaction with the A-bar System. MIT Doctoral dissertation.
Weerman, F.
(1989) The V2 Conspiracy. Foris, Dordrecht.
Wheeler, M. W. (1988) Catalan., 170-208, and Occitan., 246-278, in Harris & Vincent (eds.) 1988. Wilder, C.
(1994) Some Properties of Ellipsis in Coordination. Geneva Generative Papers 2:2, 23-61.
Williams, E.
(1981) Argument Structure and Morphology. The Linguistic Review 1, 81114.
Williams, E.
(1985) PRO and Subject of NP. Natural Language and Linguistic Theory 3, 297-315.
Williams, E.
(1987) Implicit Arguments, the Binding Theory, and Control. Natural Language and Linguistic Theory 5, 151-180.
Zanuttini, R. (1991) Syntactic Properties of Sentential Negation. A Comparative Study of Romance Languages. Doctoral Dissertation, University of Pennsylvania. 351
Zribi-Hertz, A.
(1993) On Stroik’s analysis of English middle constructions. Linguistic Inquiry 24, 583-589.
Zwart, J.-W.
(1993) Dutch Syntax. A Minimalist Approach. Ph.D. dissertation, University of Groeningen.
Zwart, J.-W.
(1994) The Minimalist Program and Germanic Syntax. A Reply to Gärtner and Steinbach.
Åfarli, T & C. Creider (1987) Nonsubject Pro-Drop in Norwegian. Linguistic Inquiry 18, 339345.
352
ÉTUDES ROMANES DE LUND SÉRIE FONDÉE PAR ALF LOMBARD ED. ALF LOMBARD 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12.
13. 14. 15. 16.
MALMBERG, BERTIL, Le roman du Comte de Poitiers, poème français du XIIIe siècle, publié avec introduction, notes et glossaire. 1940. THORDSTEIN, ARVID, Le bestiaire d’amour rimé, poème inédit du XIIIe siècle, publié avec introduction, notes et glossaire. 1940. NILSSON-EHLE, HANS, Les adverbes en -ment compléments d’un verbe en français moderne. Étude de classement syntaxique et sémantique. 1941. SCHLYTER, BÖRJE, La vie de Thomas Becket par Beneit. Poème anglo-normand du XIIe siècle, publié d’après tous les manuscrits. 1941. RONSJÖ, EINAR, La vie de saint Nicolas par Wace. Poème religieux du XIIe siècle, publié d’après tous les manuscrits. 1942. THORNÉ HAMMAR, EVA, Le développement de sens du suffixe latin -bilis en français. 1942. MALMBERG, BERTIL, Le système consonantique du français moderne. Études de phonétique et de phonologie. 1944. BRANDT, GUSTAF, La concurrence entre soi et lui, eux, elle(s). Étude de syntaxe historique française. 1944. NILSSON-EHLE, HANS, Les propositions complétives juxtaposées en italien moderne. 1947. MALMBERG, BERTIL, Études sur la phonétique de l’espagnol parlé en Argentine. 1950. ANDERSSON, SVEN, Études sur la syntaxe et la sémantique du mot français tout. 1954. BOSTRÖM, INGEMAR, Les noms abstraits accompagnés d’un infinitif et combinés avec avoir. Étude historique sur la syntaxe des articles et des prépositions dans ce genre de constructions françaises. 1957. NEUMANN, SVEN-GÖSTA, Recherches sur le français des XVe et XVIe siècles et sur sa codification par les théoriciens de l’époque. 1959. ANDERSSON, SVEN, Nouvelles études sur la syntaxe et la sémantique du mot français tout. 1961. BORNÄS, GÖRAN, Trois contes français du XIIIe siècle, tirés du recueil des Vies des Pères. 1968. JACOBSSON, HARRY, L’expression imagée dans Les Thibault de Roger Martin du Gard. 1968.
17. 18. 19.
NILSSON, ELSA, Les termes relatifs et les propositions relatives en roumain moderne. Étude de syntaxe descriptive. 1969. Mélanges de philologie offerts à Alf Lombard. 1969. BRODIN, GRETA, Termini dimostrativi toscani. Studio storico di morfologia, sintassi e semantica. 1970.
ED. ÖSTEN SÖDERGÅRD 20.
24. 25.
GUNNARSON, KJELL-ÅKE, Le complément de lieu dans le syntagme nominal. 1972. WESTRIN, MAIBRIT, Étude sur la concurrence de davantage avec plus dans la période allant de 1200 à la Révolution. Comparaison avec l’usage actuel. 1973. SCHLYTER, KERSTIN, Les énumérations des personnages dans la Chanson de Roland. Étude comparative. 1974. ROBACH, INGER-BRITT, Étude socio-linguistique de la segmentation syntaxique du français parlé. 1974. BRODIN, BRITA, Criaturas ficticias y su mundo, en « Rayuela » de Cortázar. 1975. UNDHAGEN, LYDIA, Morale et les autres lexèmes formés sur le radical moral-
26.
étudiés dans des dictionnaires et dans des textes littéraires français de la seconde moitié du XVIIIe siècle. Étude de sémantique structurale. 1975. SANDQVIST, SVEN, Études syntaxiques sur la Chronique des Ducs de Normandie
21. 22. 23.
27. 28. 29. 30. 31. 32. 33. 34.
par Benoit. 1976. SWAHN, SIGBRIT, Proust dans la recherche littéraire. Problèmes, méthodes, approches nouvelles. 1979. LARSSON, EVA, La dislocation en français. Étude de syntaxe générative. 1979. SWEDENBORG, EKY, Jean Barois de Roger Martin du Gard. Étude des manuscrits et des techniques narratives. 1979. GRAUMANN, GUNNAR, « La guerre de Troie » aura lieu. La préparation de la pièce de Giraudoux. 1979. KELLNER, SVEN, « Le Docteur Pascal » de Zola: Rétrospective des RougonMacquart, Livre de Documents, Roman à Thèse. 1980. LLAVADOR, YVONNE, La poésie algérienne de langue française et la guerre d’Algérie. 1980. BIRGANDER, PIA, Boris Vian romancier. Étude des techniques narratives. 1981. GRELSSON, SIGVARD, Les adverbes en -ment. Étude psycho-mécanique et psycho-systématique. 1981.
35. 36. 37. 38. 39.
JOSEFSON, EVA-KARIN, La vision citadine et sociale dans l’œuvre d’Emile Verhaeren. 1982. WIJK, MARGARETH, Guillaume Apollinaire et l’esprit nouveau. 1982. HEED, SVEN-ÅKE, Le coco du dada. Victor ou les Enfants au pouvoir de Roger Vitrac : texte et représentation. 1983. ORFALI, INGRID, Fiction érogène à partir de Klossowski. 1983. SANDQVIST, SVEN, Notes textuelles sur le Roman de Tristan de Béroul. 1984.
ED. LARS LINDVALL 40. 41. 42.
43. 44. 45.
BORNÄS, GÖRAN, Ordre alphabétique et classement méthodique du lexique. Étude de quelques dictionnaires d’apprentissage français. 1986. LARSSON, BJÖRN, La réception des Mandarins. Le roman de Simone de Beauvoir face à la critique littéraire en France. 1988. SANDQVIST, SVEN, Le Dyalogue saint Gregore. Les Dialogues de saint Grégoire le Grand traduits en vers français à rimes léonines par un Normand anonyme du XIVe siècle. Édition avec introduction, notes et glossaire. 2 vol. 1989. SANDQVIST, OLLE, La Vie saint Gregore. Poème normand du XIVe siècle, publié avec introduction, notes et glossaire. 1989. ANGELFORS, CHRISTINA, La Double Conscience. La prise de conscience féminine chez Colette, Simone de Beauvoir et Marie Cardinal. 1989. Actes du Xe Congrès des Romanistes Scandinaves, Lund, 10-14 août 1987, édités par LARS LINDVALL. 1990.
ED. SUZANNE SCHLYTER 46. 47. 48. 49. 50.
SWAHN, SIGBRIT, Balzac et le merveilleux. Étude du roman balzacien 1822-1832. 1991. ELGENIUS, BERNT, Studio sull’uso delle congiunzioni concessive nell’italiano del Novecento. 1991. SANDQVIST, SVEN, La Vie de saint Évroul. Poème normand du XIVe siècle, publié avec introduction, notes et glossaire. 1992. HERMERÉN, INGRID, El uso de la forma en RA con valor no-subjunctivo en el español moderno. 1992. LARSSON, BJÖRN, La place et le sens des adjectifs épithètes de valorisation positive. 1994.
51. 52. 53. 54. 55. 56.
EKBLAD, SVEN, Studi sui sottofondi strutturali nel Nome della rosa di Umberto Eco. Parte I. La Divina Commedia di Dante. 1994. ZETTERBERG, ANDERS, Les propriétés des choses selon le Rosarius (B.N. f. fr. 12483). Édition revue et complétée par SVEN SANDQVIST. 1994. EGERLAND, VERNER, The Syntax of Past Participles. A Generative Study on Nonfinite Constructions in Ancient and Modern Italian. 1996. BENGTSSON, ANDERS, La Vie de sainte Bathilde. Quatre versions en prose des XIIIe et XVe siècles, publiées avec introduction, notes et glossaire. 1996. SANDQVIST, SVEN, Le Bestiaire et le Lapidaire du Rosarius (B.N. f. fr. 12483). 1996. JÖNSSON, NILS-OLOF, La Vie de saint Germer et la Vie de saint Josse de Pierre de Beauvais. Deux poèmes du XIIIe siècle, publiés avec introduction, notes et
59.
glossaire. 1997. LARSSON, BJÖRN, Le bon sens commun. Remarques sur le rôle de la (re)cognition intersubjective dans l'épistémologie et l'ontologie du sens. 1997. WIBERG, EVA, Il riferimento temporale nel dialogo. Un confronto tra giovani bilingui italo-svedesi e giovani monolingui romani. 1997. SANDBERG, VESTA, Temps et Traduction. Étude contrastive des temps de
60.
l'indicatif du français et du suédois. 1997. DITVALL, CORALIA, Études sur la syntaxe et la sémantique de "tot" en roumain
61.
ancien et moderne. 1997. BARDEL, CAMILLA, La negazione nell’italiano degli svedesi. Sequenze
57. 58.
62. 63. 64.
65. 66. 67.
68.
acquisizionali e influssi translinguistici. 2000. CARIBONI KILLANDER, CARLA, De la théorie de la description à la description chez Julien Gracq. 2000. FORNÉ, ANNA, La piratería textual. Un estudio hipertextual de Son vacas, somos puercos y El médico de los piratas de Carmen Boullosa. 2001. LENNARTSSON, VIVI-ANNE, L’Effet-sincérité. L’Autobiographie littéraire vue à travers la critique journalistique. L’Exemple de La Force des choses de Simone de Beauvoir. 2001. MÖRTE ALLING, ANNIKA, Le désir selon l’Autre. Étude du Rouge et le Noir et de la Chartreuse de Parme à la lumière du « désir triangulaire » de René Girard. 2003. JARLSBO, JEANA, Ecriture et altérité dans trois romans de J. M. G. Le Clézio : Désert, Onitsha et La quarantaine. 2003. GRANFELDT, JONAS, L’Acquisition des catégories fonctionnelles. Étude comparative du développement du DP français chez des enfants et des apprenants adultes. 2003. WESTIN, EVA, Le récit conversationnel en situation exolingue de français Formes, types et fonctions. 2003.
69. 70.
BÖRJESSON, ANNE, La syntaxe de seul et seulement. 2004. WILHELMI, JUAN - ENKVIST, INGER, Literatura y Compromiso. Serie de estudios hispánicos. 2004.
ED. INGER ENKVIST, BJÖRN LARSSON, SUZANNE SCHLYTER 71. 72. 73. 74. 75.
BERNARDINI, PETRA, L’italiano come prima e seconda (madre)lingua. Indagine longitudinale sullo sviluppo del DP. 2004. ÁLVAREZ SALAMANCA, MARÍA DEL PILAR, De Sobremesa, 1887-1896. José Asunción Silva: El poeta novelista. 2004. CONWAY, ÅSA, Le paragraphe oral en français L1, en suédois L1 et en français L2. Étude syntaxique, prosodique et discursive. 2005. JABET, MARITA, L’omission de l’article et du pronom sujet dans le français abidjanais. 2005. BOZIER, CHRISTINE, La sollicitation dans l’interaction exolingue en français. 2005.
