Amsterdam: John Benjamins. Kegl, Judy, Ann Senghas, and Marie Coppola. ..... Donald Weisz, Gregory A. Elder, James Schmeidler, Rita. De Gasperi, Miguel A.
DISTRIBUTED MORPHOLOGICAL MECHANISMS OF
LABOVIAN VARIATION
IN
MORPHOSYNTAX
A Dissertation submitted to the Faculty of the Graduate School of Arts and Sciences of Georgetown University in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Linguistics
By
Jeffrey K. Parrott, B.A.
Washington, DC January 9, 2007
Copyright 2007 by Jeffrey K. Parrott All Rights Reserved
ii
DISTRIBUTED MORPHOLOGICAL MECHANISMS OF LABOVIAN VARIATION IN MORPHOSYNTAX Jeffrey K. Parrott, B.A. Thesis Advisor: Donna Lardiere, Ph.D.
ABSTRACT This dissertation takes a biolinguistic perspective on Labovian variation in morphosyntax, an important phenomenon discovered by sociolinguists (Chambers, Trudgill and Schilling-Estes 2002). The dissertation asks how an account of the mechanisms underlying such variation can be incorporated into a Minimalist theoretical model of the human language faculty (Chomsky 1995, et seq.). To address this question, three cases of morphosyntactic variation in English are investigated using an adapted combination of variationist and theoretical methods. A primary empirical focus is on the moribund English variety spoken in the community of Smith Island, Maryland, where both phonological and morphosyntactic variants are currently undergoing a process of rapid change as the insular dialect approaches death due to ongoing population attrition (Schilling-Estes and Wolfram 1999). This dissertation considers two morphosyntactic variants on Smith Island: weak expletive it (Parrott 2002) and leveled weren’t (Schilling-Estes 2000, Mittelstaedt 2006). The dissertation also examines the puzzle of pronominal caseform mismatches and pronoun-specific ordering asymmetries that occur in English coordinate noun phrases (Emonds 1986). This dissertation adopts a particular Minimalist theory of syntax (Chomsky 2000) augmented with the independently motivated and well-articulated theory of Distributed Morphology (DM, Halle and Marantz 1993, Embick and Noyer to appear). In this theoretical model, mechanisms of Labovian variation can be located in the features of syntactic terminals and their combination in the narrow syntax, as proposed by Adger and Smith (Adger and Smith 2005, Adger 2006). Evidence from the three case studies supports the additional hypothesis that mechanisms of variation are located in the morphological component of the language faculty, at the interface between the syntactic computation and the phonological component. Specifically, it is claimed that variation can arise from the inventory iii
and feature structure of non-competing Vocabulary Items and their interactions with ordered operations during the morphological computation to the Phonetic-Form interface. By showing how actual cases of morphosyntactic variation might receive a plausible analysis within the Minimalist-theoretical framework of DM, an overarching goal of the dissertation is to advocate further cooperative research efforts toward bridging the gap between biolinguistic theory and the empirical study of Labovian variation and change in progress.
iv
ACKNOWLEDGEMENTS Many thanks to my advisor Donna Lardiere and the rest of my committee, David Adger, Kleanthes Grohmann, and Natalie Schilling-Estes. Thanks to everyone at Georgetown, including but not limited to Jen Mittelstaedt (and Andy Robinson, and the Mittelstaedt family), Matt Bauer, AnnaMarie Trester, Dominik Rus, Takae Tsujioka, Manuela Diez, Sam Candon, Raffaella Zanuttini, Hector Campos, Lisa Zsiga, Paul Portner, Elena Herburger, Ralph Fasold, Cathy Ball, Mits Ota, Bruce Moren, Alkistis Fleischer, Phillip Handmaker, Phillip Levine, Carlota Bernal, Shiraz Felling, Elliot Lash, Dan Beckett, Miok Pak, the Graduate School of Arts and Sciences, David Lightfoot, Chris Quigley, and all the faculty, staff, and students in the Department of Linguistics. Thanks to everyone at Portland State University, including but not limited to Tucker Childs, Tom Dieterich, Lynn Santelmann, and the rest of the faculty, staff, and students in the Department of Applied Linguistics, Maji Rhee, the faculty and students in Japanese language and literature, the Department of Philosophy, and Byron Haines. Thanks to my teachers in Corvallis, including but not limited to Ken Gouveia, Blue Carmen, and Rachel Kirby. Thanks to everyone at Yonsei University, Han Ji Sook, her sister’s family (hi Youngyo and Songyee!), Ha Hyun Ju, and all of my Korean friends, students, and teachers. Thanks to all the linguists, including but not limited to Andrew Nevins, everyone at the Harvard Linguistics Department, Cilene Rodrigues, Tom McFadden, David Embick, Rolf Noyer, everyone in the Penn Linguistics Department, Philip Angermeyer, John Singler, everyone in the NYU Linguistics Department, Laura Rupp, Tommy Grano, Jennifer Smith, Emily Bender, everyone in the Stanford Linguistics Department, Masayuki Oishi, Dennis Ott, Asaf Bachrach, Omer Preminger, everyone in MIT linguistics, Scott Fults and Danielle and Frankie, Ian Roberts, Christina Tortora, everyone in CUNY linguistics, Oliver Dudas, Walt Wolfram, Morris Halle, Victoria Fromkin (RIP), Norbert Hornstein, everyone at the UMD Linguistics Department, Artemis Alexiadou, Manuela Schoenenberger, everyone in Uni Stuttgart linguistics, everyone in Uni Cologne linguistics, Andrea Moro, Masayuki Oishi, everyone at Penn Working Papers, the Linguist List, all the participants and organizers at the ‘97 LSA Summer Institute, WECOL ‘99, NWAV ‘99, ‘01, ‘02, ‘05, ICKL ‘00, Asymmetry ‘01, GURT ‘04, PLC ‘06, and InterPhases ‘06. Thanks to everyone at Facts on File and Today’s Science on v
File, including but not limited to Jonathan Taylor, Amy Perry, and Margie Bank. Thanks to all my dear friends, including but not limited to Jason Park, Jinny Kim, Erik Henrikson, Peter Weber and Jen Furl, Alex Carlson and Memorie and family, Bobbie-Jo Webb-Robertson and family, Arik Goff, Alex Hamilton, Alex Hall and Mary Beth, Allison Dell and Rob Swainston, Alyssa Wright and Tom Kotik, Stephanie Pisias and Jeremy Norris and family, Josh Paulson, Andy O’Neil and Alex, Rene Morales and Wava, Joe and Viki Taylor and family, Melanie Mauro, Joy Grohmann, Justin Paulson, Robert-Sensei and Brooklyn Aikikai, Jen Dunham, Mark Dion, Morgan Puett, Cristy Gast, Dana Sherwood, Bob Braine and Leslie Reed, Stefan Sonnberger, Christian Schön, Isa Reiss, Catriona Shaw, Viola Thiele, Thomas Priermeyer and Karin Knopp and family, Patrick Gruban and Rosie, Karen Brandl, Stefanie Trojan, Notburger Karl, Tom Früchtl and Simone Lanzenstiel, Andreas Baernthaler and Kim, Naomi Grundke, Bela Lee (RIP), Suzka Jung, Michi Mattes, and Gunther Massuri. Thanks to my family, including but not limited to Keith and Wanda Parrott, Shannon Parrott, Kim Parrott, Elmer and Grace Parrott (RIP), Tom and Karen Soderling and family, Harold Otta, Sonnica Otta (RIP), Jean (RIP), Arlene, Rick and Anne Otta and family, Eldon and Debbie Otta and family, Peg and Larry Phelps and family, LaRayne DeJulis, all of our family friends, Jiří and Milada Hobza, Babi Kristina, Babi Dáša and Deda Tonda, Tony Hobza and family, and Tonda and Milena Valeš and family. Thanks to my feline companions Fuzzy (RIP), Ruby (RIP), Damon (RIP), Domino, and Fritz. Most of all, thanks to Klara Hobza, my beloved wife, my closest friend, my partner in crime, and my greatest inspiration. This dissertation is dedicated to you! ♥
vi
Table of Contents Chapter 1: The Variation Gap .............................. 1 1.
Introduction ........................................ 1
1.1
Into the Gap ....................................... 3
1.2
Goals and Questions for Research .................. 11
1.3
Organization and Summary of Chapters .............. 13
Chapter 2: Labovian Variation and Change in Progress on Smith Island ............................................. 18 2.1
What Labovian Variation Is ........................ 19
2.1.1
Labovian vs. parametric variation.............. 20
2.1.2
Labovian vs. allomorphic variation............. 21
2.1.3
Social vs. semantic meaning.................... 24
2.2
Variationist Methods and Apparent Time ............ 26
2.2.1
Vs. Theoretical methods........................ 26
2.2.2
Sociolinguistic interviews..................... 27
2.2.3
Quantitative analysis.......................... 29
2.2.4
The Apparent-time method....................... 30
2.3
Dialect Death and Concentration on Smith Island ... 32
2.3.1
Smith Island: Geography, history, community.... 34
2.3.2
Dialect death and concentration................ 39
2.3.3
Two phonological changes in progress........... 40
2.4
Conclusion ........................................ 44
Chapter 3: Biolinguistics and the Distributed Morphology Model .................................................... 46 vii
3.
Introduction ....................................... 46
3.1
The Biolinguistic Perspective ..................... 46
3.1.1
Five questions for biolinguistic inquiry....... 48
3.1.2
The Scope of biolinguistic theory.............. 49
3.1.3
The Scope of sociolinguistics.................. 51
3.1.4
Competence/performance vs. methodology......... 52
3.1.5
The Scope of present inquiry................... 55
3.2
A Minimalist, Distributed Morphology Model ........ 59
3.2.1
A Minimalist program for biolinguistic theory.. 59
3.2.2
The Theory of Distributed Morphology........... 62
3.2.3
The DM architecture of the language faculty.... 64
3.2.4
Listed primitives 1: Syntactic terminals....... 66
3.2.5
The Lexical Array.............................. 70
3.2.6
Narrow syntax 1: Merge and phrase structure.... 71
3.2.7
Narrow syntax 2: Agreement and feature checking 73
3.2.8
Spell Out and the LF-interface computation..... 78
3.2.9 Ordered operations of the morphological component ...................................................... 80 3.2.10 Listed primitives 2: Vocabulary Items and Insertion............................................. 92 3.2.11 3.3
Listed primitives 3: The Encyclopedia......... 98
Minimalism, DM, and Labovian Variation ............ 99
Chapter 4: Weak Expletive It on Smith Island ............ 103 4.
Introduction ...................................... 103
viii
4.1
Expletives and Agreement in English .............. 105
4.1.1
Two expletive forms and agreement............. 105
4.1.2
The Definiteness Effect and agreement......... 108
4.1.3 Associate-agreement variation with expletive there ................................................ 111 4.1.4
Summary....................................... 116
4.2 WEIT: Variation and Change in Progress on Smith Island ................................................ 116 4.2.1
WEIT in English............................... 117
4.2.2
WEIT on Smith Island.......................... 117
4.2.3
WEIT and categorical 3s agreement............. 119
4.2.4
A Methodological digression................... 122
4.2.5
Concentration of WEIT in apparent time........ 127
4.2.6
WEIT in real time............................. 137
4.2.7 Summary and implications for variation analysis ..................................................... 139 4.3
WEIT as Morphosyntactic Mismatch ................. 141
4.3.1
A Minimalist theory of agreement.............. 142
4.3.2
Chomsky’s analysis of expletives.............. 146
4.3.3
WEIT as morphosyntactic mismatch.............. 150
4.4
A DM Approach to Expletives in English ........... 152
4.4.1 Morphosyntactic features of expletives it and there ................................................ 152 4.4.2
A DM analysis of WEIT and the Case paradox.... 157
4.4.3
A DM analysis of there and associate agreement 158
ix
4.4.4 Speculation on variable 3s agreement with there ..................................................... 164 4.4.5 Summary and questions for morphosyntactic theory ..................................................... 166 4.5
Conclusion ....................................... 173
Chapter 5: Distributed Morphological Mechanisms of Smith Island Weren’t Leveling ................................. 174 5.
Introduction ...................................... 174
5.1
Agreement Leveling in English Varieties .......... 176
5.1.1
3s Agreement Leveling......................... 177
5.1.2
Allomorphs of past-tense be ................... 180
5.1.3
Was leveling.................................. 182
5.1.4
Were leveling................................. 184
5.2
Adger and Smith’s Minimalist, Lexical Approach ... 187
5.2.1
A Minimalist syntax........................... 188
5.2.2
Was leveling in Buckie, Scotland.............. 192
5.2.3
A Lexical analysis of Buckie was leveling..... 193
5.2.4
Summary....................................... 197
5.3
Weren’t leveling on Smith Island ................. 198
5.3.1
Past-tense be leveling on Smith Island........ 199
5.3.2
Weren’t leveling on Smith Island.............. 201
5.3.3
Two apparent-time studies..................... 201
5.3.4
Summary....................................... 206
5.4
Distributed Morphological Mechanisms ............. 207
5.4.1
Lexical analyses of weren't leveling.......... 207
x
5.4.2
Past-tense be suppletion and allomorphy....... 211
5.4.3
Contracted -n’t ............................... 215
5.4.4
Weren’t leveling as suppletion of negation.... 218
5.4.5
Fusion and the ordering of operations......... 220
5.4.6 Extension to ain’t and other leveled auxiliaries ..................................................... 227 5.4.7 5.5
Was leveling and non-competition of Vocabulary 229
Conclusion ....................................... 235
Chapter 6: Distributed Morphological Mechanisms of PronounCase Variation .......................................... 239 6.
Introduction ...................................... 239
6.1
Salient and Stigmatized Sociolinguistic Variation 240
6.1.1
Pronominal allomorphy in English.............. 241
6.1.2 Mismatched pronominal case forms in coordinates ..................................................... 242 6.1.3 Angermeyer and Singler’s observational methodology.......................................... 244 6.1.4
Mismatch Specimens............................ 247
6.1.5
Summary....................................... 252
6.2
Theoretical perspectives on case mismatch ........ 254
6.2.1
Case in Minimalist Syntax..................... 254
6.2.2
Pronoun-case variation as mismatch............ 259
6.2.3
Previous Theoretical Approaches............... 260
6.2.3.1
Deviant prestige constructions ............ 261
6.2.3.2
Extraordinarily unbalanced coordination ... 264
xi
6.3
6.2.3.3
Grammatical viruses ....................... 270
6.2.3.4
Viral infection with DM default case ...... 275
Distributed Morphological Mechanisms ............. 276
6.3.1
A DM theory of case........................... 276
6.3.2 A Case-less DM analysis of English pronominal forms................................................ 278 6.3.3
Supplemental pronoun Vocabulary............... 284
6.3.4
Cross-linguistic predictions.................. 291
6.3.5
Predictions for acquisition research.......... 294
6.4
Conclusion ....................................... 296
Chapter 7: Bridging the Gap ............................. 298 7.
Introduction ...................................... 298
7.1
Summary of Conclusions ........................... 298
7.2 Future research on mechanisms of Labovian variation ...................................................... 303 References .............................................. 309
xii
Chapter 1: The Variation Gap
1.
Introduction
This dissertation takes a biolinguistic perspective on Labovian variation in morphosyntax, an important phenomenon discovered by sociolinguists (Chambers, Trudgill and Schilling-Estes 2002). The dissertation asks how an account of the mechanisms underlying such variation can be incorporated into a Minimalist theoretical model of the human language faculty (Chomsky 1995, et seq.). To address this question, three cases of morphosyntactic variation in English are investigated using an adapted combination of variationist and theoretical methods. A primary empirical focus is on the moribund English variety spoken in the community of Smith Island, Maryland, where both phonological and morphosyntactic variants are currently undergoing a process of rapid change as the insular dialect approaches death due to ongoing population attrition (Schilling-Estes and Wolfram 1999). This dissertation considers two morphosyntactic variables on Smith Island: weak expletive it (Parrott 2002) and leveled weren’t (Schilling-Estes 2000, Mittelstaedt 2006). The dissertation also examines the puzzle of pronominal 1
case-form mismatches and pronoun-specific ordering asymmetries that occur in English coordinate noun phrases (Emonds 1986). This dissertation adopts a particular Minimalist theory of syntax (Chomsky 2000a) augmented with the independently motivated and well-articulated theory of Distributed Morphology (DM, Halle and Marantz 1993, Embick and Noyer to appear). In this theoretical model, mechanisms of Labovian variation can be located in the features of syntactic terminals and their combination in the narrow syntax, as proposed by Adger and Smith (Adger and Smith 2005, Adger 2006). Evidence from the three case studies supports the additional hypothesis that mechanisms of variation are located in the morphological component of the language faculty, at the interface between the syntactic computation and the phonological component. Specifically, it is claimed that variation can arise from the inventory and feature structure of non-competing Vocabulary Items and their interactions with ordered operations during the morphological computation to the Phonetic Form interface. By showing how actual cases of morphosyntactic variation might receive a plausible analysis within the Minimalist-theoretical framework of DM, an overarching goal of the dissertation is to advocate further cooperative
2
research efforts toward bridging the gap between biolinguistic theory and the empirical study of Labovian variation and change in progress.
1.1
Into the Gap
"Unfortunately," as Schilling-Estes notes in The Handbook of Language Variation and Change (2002a: 203-204), "it has proven difficult to incorporate variation into linguistic theory or theoretical linguistic models into variation study." Although there has been some substantial progress in "work to bridge the gap between phonological theory and variation study" (see Anttila 2002 for an overview, and see also Bauer 2005 for an especially innovative research program for phonetic variation ), Schilling-Estes observes that "variation analysis has had little impact on syntactic theory over the years." The "longstanding gap between variation study and theoretical syntax" remains a serious impediment to the incorporation of sociolinguistic approaches under the biolinguistic perspective. (For further discussion of this gap from various empirical and theoretical perspectives, see Meechan and Foley 1994, Labov 1995, Hudson 1995a, 1997, Winford 1996, Wilson and Henry 1998, Bender 2001, Henry 2002, Cornips and Corrigan 2005,
3
Rupp 2005, Adger 2006, Adger and Trousdale to appear,
and
Parrott to appear-a.) Over the last four decades, efforts to bridge the unfortunate variation gap have been somewhat few, but are far from new. Indeed, the early variationist literature attempted to provide a theoretical explanation for the newly scrutinized phenomena of socially significant linguistic variation (Labov 1963, 1966). Labov (1969, 1972, see also Cedergren and Sankoff 1974) proposed that variant forms are produced by the variable application of rules similar to those familiar from generative phonology (Chomsky and Halle 1968) and early transformational syntax (Chomsky 1965). Such rules state that a linguistic process or operation (e.g., deletion of phonological features) applies obligatorily in a specified linguistic environment. In contrast, variable rules apply probabilistically: a variable rule includes a weighting for the probability of application in a range of specified environments. Moreover, the environments specified in a variable rule may consist of social and not only linguistic information. Variable rules were viewed with skepticism outside of variationist sociolinguistics (see e.g. Bickerton 1971 for early criticism), however, and were never adopted in syntactic theory, which had already begun to move away from
4
transformational rule-based theories (Chomsky 1971) and toward the modular Government and Binding theories of the 1980s (Chomsky 1981, 1982, 1986). Variation study focused on social and empirical matters as the field became increasingly estranged from syntactic theory, and largely abandoned the task of developing a linguistic-theoretical explanation of variation. Thus, variable-rule theories of variation succumbed to a “quiet demise” (Fasold 1991) in sociolinguistics as well. Meanwhile, Kroch (1989, 1994) proposed that variation does not arise in a single grammar as suggested by the variable rule approach, but instead that individuals may have access to multiple grammars. On this theory, simultaneous multiple grammars are considered to be diachronically unstable, and inevitable competition between the grammars in a population yields morphosyntactic change over time (Kroch 2001). The theory has been quite influential in historical syntax and some other fields outside of variationist sociolinguistics (for example, inspiring attempts to model first-language acquisition in terms of competing grammars as in Yang 2004 and other work). Notably, the competing-grammars model attempts to explain morphosyntactic changes that took many centuries to complete, and is based entirely on data from historic texts
5
and not from variationist studies of change in progress (e.g. as conceded by Labov 1994: 3). However, it remains unclear how a theory that posits obligatory competition between multiple grammars can account for the many welldocumented cases of diachronically stable variation observed by sociolinguists (Schilling-Estes 2002a, Henry 2002). A related theoretical approach to variation came forward with the advent of the Principles and Parameters framework in the late 1980s and early 1990s (Chomsky and Lasnik 1993). In this framework, the human genetic endowment for language (Universal Grammar) includes both invariant principles and parametric options that must be set by exposure to environmental linguistic input during the development process. Thus, parametric theory is meant to account for observed cross-linguistic variation. Parameters were originally conceived as binary ‘switches’ whose settings affected a large cluster of linguistic properties (e.g. Baker 1996, 2001). Syntactic theorists engaged in the comparative study of closely related dialects proposed the existence of so-called ‘microparameters’ whose settings affected a much smaller set of linguistic properties, perhaps even just one per microparameter (e.g. Kayne 1996). Following this idea,
6
Henry (1995, 1996, Wilson and Henry 1998) furthermore proposed that variation arises because an individual’s microparameters can have multiple settings. In other words, multiple grammars are multiple parameter settings; contra Kroch, however, Henry does not claim that these multiple grammars must be in competition. Another theoretical approach to cross-linguistic variation, based on the work of Borer (1984), rejects the concept of parameters as switches. Instead, differences between languages are to be explained by differences in the properties of lexical items. On the strongest version of this theory, parametric variation should be entirely limited to the morphosyntactic features of functional lexical items. Chomsky adopted such a lexical theory of parametric variation at the dawn of his Minimalist program for biolinguistic theory, but noted that it will be “no trivial matter to make it explicit” (1993: 44, fn. 4). Such a theory would seem quite compatible with the ideas of Kroch and Henry: Labovian variation arises when an individual has access to multiple lexical items with different microparametrically significant morphological features. But as Chomsky correctly predicted, the theory has indeed proven somewhat difficult to make explicit. Thus, this potential approach to bridging the gap has not
7
been widely adopted or applied to actual cases of Labovian variation and change in progress until very recently. The last five years have witnessed a renewed interest in bridging the variation gap. In her analysis of copula deletion in African-American English, Bender (2001) proposes an articulated new version of variable-rule theory that includes probabilistic social weightings in lexical items (see Adger 2006 for criticism) within the theoretical framework of Head-driven Phrase Structure Grammar (Pollard and Sag 1994). The papers collected by Cornips and Corrigan (2005) make a significant contribution toward “reconciling the biological and the social” by utilizing a diverse array of empirical methodologies and analytical techniques. Among them, for example, Henry (2005) considers agreement variation in English weak-expletive constructions and suggests the resurrection of variable-rule theory in the form of variable agreement operations (though she does not specify exactly how such theoretical mechanisms would work, and see Parrott to appear-a for criticism on methodological grounds). Rupp (2005) and Tortora (2006) have also examined dialectal variation in agreement with weak expletives from a syntactic-theoretical perspective. Finally, Bresnan, Deo and Sharma (to appear) have proposed that morphosyntactic variation in the form of English copula/auxiliary be arises
8
from probabilistic re-ranking of output constraints within the theoretical framework of (stochastic) Optimality Theory (Prince and Smolensky 1993, Archangeli and Langendoen 1997). Despite these previous and recent efforts to bridge the gap, and with the turn of the 21st century already more than half a decade behind us at this writing, the chasm still appears to yawn widest between the empirical study of variation and Minimalist theoretical syntax (Chomsky 1995, et seq.). Variationist sociolinguistics has had virtually no detectable influence on Minimalist theorizing, nor vice versa. Minimalists have typically dismissed variationist sociolinguistics as a field solely concerned with matters of usage (and phonology) that are irrelevant to syntactic theory. Variationists, on the other hand, have typically regarded their well-documented phenomena of variation and change in progress as fundamentally incompatible with Minimalist theories, where optional syntactic operations are ruled out in principle. At first glance, then, Minimalist theories of syntax and the facts of morphosyntactic variation appear irreconcilable. If so, it would seem that the Minimalist program is substantially disconfirmed. Labovian variation in morphosyntax is a ubiquitous empirical phenomenon that is moreover observed
9
to be involved with language change over time. Any adequate biolinguistic theory must be able to incorporate some account of its mechanisms. However, a new and explicitly Minimalist approach to Labovian variation has recently emerged from the ongoing collaboration between syntactic theorist David Adger and sociolinguistic variationist Jennifer Smith. In a paper entitled “Variation and the Minimalist Program” and published in the Cornips and Corrigan volume, Adger and Smith argue that Minimalist syntactic theory “melds extremely well with the kinds of data that variationists study” (2005: 173). Working in cooperation, these two linguists have advanced an analysis wherein mechanisms of Labovian variation are located in the semantically uninterpretable features of lexical items and their interactions with feature checking in the syntax (see also Adger 2006). Hearkening back to the lexical parametric theories of Borer, Kroch, Henry, and Chomsky, Adger and Smith’s Minimalist approach allows “the orderly patterns of variation seen across (groups of) individuals” to be reduced to “lexical choice by an individual speaker of functional elements with particular feature specifications” (2005: 173).
10
This dissertation embarks upon a somewhat different approach to Labovian variation, also following the Minimalist program but utilizing the specific theoretical framework of Distributed Morphology (DM, Halle and Marantz 1993, Embick and Noyer to appear). DM offers an articulated theory of the primitive items computed by the narrow syntax and the post-syntactic PF-interface component. Crucially, the theory of DM holds that both the morphological properties of lexical primitives and the morphological operations that construct and modify them are distributed throughout the model of grammar. In such a framework, morphosyntactic variation can also arise in the features of lexical items and their combinations in the syntax as proposed by Adger and Smith. However, the central hypothesis pursued in this dissertation is that additional mechanisms of Labovian variation can be located in the inventory and feature structure of DM’s Vocabulary Items and their interactions with the operations of the morphological component.
1.2
Goals and Questions for Research
Bridging the variation gap is an immense and necessarily collaborative endeavor. This dissertation aims to make a modest contribution to the task by attempting to identify
11
and locate theoretical mechanisms of Labovian variation within a Minimalist, DM framework. Thus, the following general research questions will guide inquiry throughout the dissertation. First, what kind of morphosyntactic objects are Labovian variant forms? Second, what are the morphosyntactic structural environments where Labovian variation can occur? Third, how can we explain apparent morphosyntactic mismatches in the variant forms? Finally, how can we explain the possibility of sociolinguistic choice of Labovian variants, as opposed to the deterministic, complementary distribution of allomorphic variants? This dissertation adopts a biolinguistic perspective and is primarily addressed to a theoretical audience. It is my hope that the dissertation will demonstrate the usefulness of variationist data and methods in addressing theoretical issues, especially those questions relevant to the Minimalist program and the theory of DM. I expect that sociolinguistic variationists will find much to disagree with in my approach, and the dissertation sets aside many of the issues that concern them most. However, the overarching goal of this dissertation is to advocate further cooperation between researchers on opposite sides of the gap by showing how actual case studies of Labovian
12
variation might be plausibly analyzed in a particular biolinguistic-theoretical framework.
1.3
Organization and Summary of Chapters
This dissertation is organized around three central case studies of Labovian variation in morphosyntax, presented in Chapters 4, 5, and 6. Chapters 2 and 3 begin by surveying some territory on either side of the variation gap, and Chapter 7 concludes with suggestions for further research. Succinct summaries of each chapter are given below. Chapter 2: Labovian Variation and Change in Progress on Smith Island begins by defining Labovian variation and contrasting this phenomenon with parametric and allomorphic variation. This is followed by a brief review of the empirical methods used in the study of variation and change in progress. Finally, the chapter introduces the moribund community of Smith Island, Maryland, providing examples of phonological variation and change currently in progress there. Chapter 3: Biolinguistics and the Distributed Morphology Model turns to the other side of the gap, introducing the biolinguistic perspective and then giving a brief review of the Minimalist, DM theoretical model of the
13
language faculty that is adopted throughout the dissertation. Chapter 4: Weak Expletive It on Smith Island presents the dissertation’s first case study of morphosyntactic variation and change, the variant weak-expletive form it on Smith Island. Although weak expletive it (WEIT) is a commonly documented feature of several modern and historical English varieties, there have been no variationist or theoretical studies of WEIT and its agreement properties. This chapter provides both perspectives on WEIT. In addition to quantitative observational data, elicited judgments confirm that verbal agreement is categorically third singular with WEIT. Apparent-time analysis, supported by preliminary real-time data, shows that the WEIT variant is part of the socially motivated concentration process on Smith Island (Parrott 2002). After reviewing a current Minimalist analysis of expletives and agreement (Chomsky 2000a), the chapter points out a mismatch paradox. WEIT should either leave the associate DP’s uninterpretable Case feature unchecked, causing a crash at the interface, or allow plural associate agreement, contrary to fact. The chapter proposes a DMtheoretical analysis of WEIT that resolves the apparently paradoxical Case mismatch by following a radical approach
14
that denies the existence of uninterpretable syntactic Case features (McFadden 2004). The chapter concludes that the mechanism of WEIT variation is located in the choice of terminals selected into the pre-syntactic Lexical Array. Chapter 5: Distributed Morphological Mechanisms of Smith Island Weren’t Leveling examines weren’t leveling, another case of morphosyntactic variation and change in progress on Smith Island. Although weren’t leveling is not unique to Smith Island, this pattern of apparent variation in agreement is apparently anomalous: leveling occurs only with clitic negation (e.g. She weren't scared) and in the total absence of concurrent were leveling (e.g., *She were scared, *She were not scared). This chapter outlines some facts about agreement leveling variation in varieties of English, with a focus on the especially common phenomenon of leveling in the paradigm of past-tense be. The chapter provides a review of Adger and Smith’s (2005) Minimalist, lexical analysis of was leveling in the community of Buckie, Scotland, which locates mechanisms of variation in the uninterpretable features of syntactic terminals and their checking combinations in the narrow syntax. Weren’t leveling is examined, and two variationist studies of its usage in apparent time on Smith Island are briefly reviewed. Adopting the Distributed Morphology theoretical
15
framework outlined in Chapters 3 and 4, the chapter then presents a DM analysis where leveled weren’t is claimed to be a suppletive form, resulting from a non-competing Vocabulary Item that requires morphological Fusion of the be, Tense, and Negation terminals during the morphological component. In this case, the mechanism of weren’t-leveling variation is located in the choice of non-competing Vocabulary Items and their interaction with ordered operations in the morphological component. Chapter 6: Distributed Morphological Mechanisms of Pronoun-Case Variation leaves Smith Island to examine the (in)famous case of mismatched pronominal case-form variation in English coordinate phrases (e.g. Me and him were doing calls together; He talked John and I into going; Him and I were working at the time; This is starting to make him and I both feel really bad). Despite its sociolinguistic salience, pronoun-case variation has attracted little attention from variationists. Methodological problems have hampered theoretical approaches, relegating pronoun-case variation to the margins of inquiry despite the Minimalist theory of Case checking. This chapter introduces the known facts of pronoun-case variation, reviews the observational method used in Angermeyer and Singler’s (2003) variationist study
16
of 1st person singular pronouns used in object coordinates, and presents some specimens of pronoun-case mismatch collected using their methodology. After a brief review of previous attempts to capture pronoun case mismatch in syntactic theories, the chapter proposes a DM-mechanistic analysis of this variation, again concluding that in this case the mechanisms of Labovian variation are located in the feature structure and individual inventories of noncompeting Vocabulary Items. Chapter 7: Bridging the Gap concludes the dissertation with a summary of conclusions a few suggested directions for future research aimed at bridging the variation gap.
17
Chapter 2: Labovian Variation and Change in Progress on Smith Island
1.
Introduction
The purpose of this chapter is to introduce and exemplify the empirical phenomenon under theoretical investigation in this dissertation, which I have chosen to refer to as Labovian variation. With his seminal work of the 1960s and 70s (Labov 1963, 1966, Weinreich, Labov and Herzog 1968, Labov 1969, 1972), William Labov pioneered the study of language change in progress and established the field of variationist sociolinguistics (for surveys of this robust field see Chambers 2002a, Chambers, Trudgill and SchillingEstes 2002). Labov continues to produce important and influential research on phonetic and phonological variation and change (Labov 1994, 2001, Labov, Ash and Boberg 2005). The chapter is structured as follows. In Section 2.1, Labovian variation is defined and distinguished from other kinds of variable phenomena studied in linguistics. Section 2.2. sketches the empirical methods used to study variation and the apparent-time method for observing language change in progress. Finally, Section 2.3 introduces the community of Smith Island, Maryland, and briefly provides an example 18
of phonological variation and change currently in progress. Chapters 4 and 5 return to Smith Island for a more detailed examination of two morphosyntactic changes underway in this moribund English variety.
2.1
What Labovian Variation Is1
Labovian variation has also been referred to as ‘variability,’ ‘inherent variation,’ and ‘sociolinguistic variation.’ This ubiquitous and well-documented linguistic phenomenon can be essentially described as “alternate ways of saying ‘the same’ thing” (Labov 1972: 118), or when “speakers use different forms to express the same meaning” (Labov 1995: 115). Below is a more formal working definition of Labovian variation that will be adopted throughout this dissertation. (1)
1
Labovian variation a.
(Populations of) individuals use variant linguistic forms;
b.
The variant forms appear in the same linguistic environment (variants are not allomorphs in complementary distribution);
c.
The variant forms do not express different lexical or truth-conditional semantics, nor different morphosyntactic functions.
I am borrowing this phrase, with much respect, from Mayr (2001).
19
The sections that follow consider each of the three different parts (1a-c) given in this definition.
2.1.1
Labovian vs. parametric variation
(1a) states that variant forms are used by individuals (and populations of individuals). Crucially, Labovian variation must be distinguished from the use of variant linguistic forms between different individuals (and populations of individuals). In other words, Labovian variation is not the same phenomenon as cross-linguistic or cross-dialectal variation. Whether and to what extent the theoretical mechanisms underlying Labovian variation are identical to, or overlap with, the mechanisms underlying parametric variation remains to be determined. It is unlikely that the mechanisms of these two phenomena are totally distinct. As noted in the first chapter, it has been claimed that Labovian variation in morphosyntax arises from multiple settings for microparameters (Henry 1995, 1996, Wilson and Henry 1998), and that competition between multiple grammars, perhaps as defined by parameter settings, is the cause of morphosyntactic change over time (Kroch 1989, 1994). Under the Minimalist approach proposed by Adger and Smith (2005), Labovian variation arises from the properties
20
of multiple lexical items, consistent with parametric theories following Borer (1984) and Chomsky (1993). The DM approach pursued here is consistent with the idea that parametric variation is restricted to the distributed properties of the objects and operations of the morphological component. Whatever the case, this matter cannot be decided a priori and must be addressed with empirical research on Labovian variation as a distinct phenomena.
2.1.2
Labovian vs. allomorphic variation
(1b) states that variant forms are not in complementary distribution, but appear in the same linguistic environment. In other words, Labovian variation must be distinguished from the familiar phenomena of allophony and allomorphy, which also involve variant linguistic forms that express the same meaning or morphosyntactic function. The occurrence of variant allomorphic forms is completely determined by aspects of their phonological and morphosyntactic environments. For a very simple example, consider the two allomorphs of the English indefinite article, a and an. (2)
a.
a phantom
b.
an apparition
21
In many varieties of English, these two forms occur in complementary distribution according to their phonological environment: an cannot occur in the pre-consonant environment of a, while a cannot occur in the pre-vocalic environment of an. (3)
a. * an phantom2 b. * a apparition In contrast, Labovian variation is “the non-
deterministic choice of form” (Adger 2006): variants occur in the same linguistic environment and not in complementary distribution like allomorphs. For comparison, consider a simple example of stable (i.e., not changing over time) morphophonological variation in English, “the alternation of an apical and velar nasal consonant as -in [In] and -ing [I]” (Labov 1995: 117). Both of these variant forms appear in the same phonological environment of unstressed syllables. (4)
a.
some horrible thing
b. * some horrible thin’
2
Here and throughout, I use '*' to mean that a form is both unattested and judged ill-formed according to the intuitions of some informant (possibly the author or another researcher). It may sometimes (but not always) be inferred that an unattested form will be judged 'ill-formed' in the relevant population. However, it may not be inferred that a form judged unacceptable by some individual or population will also be so judged by another individual or population. More on methodology follows below.
22
Just as with allomorphy, there is no change in semantics or morphosyntactic function expressed by this variation in form; unlike allomorphy, the variants do not appear in complementary distribution. (5)
a.
a shrieking banshee
b.
a shriekin’ banshee
Adger (2006) points out that linguistic theory has almost universally adopted "the assumption that either variants are determined by their context or they are in free variation." However, the empirical study of Labovian variation shows clearly that this is not the case. Although Labovian variation is non-deterministic, the appearance of Labovian variants can be probabilistically influenced by aspects of their linguistic environment (as well their social meaning, on which more below). For example, Labov (1995) reports that the -in’ variant is most likely to appear in progressive constructions (e.g. is working), less likely to appear in adjectives (e.g. interesting) and gerunds (e.g. swimming), with the lowest probability of appearance in nominals (e.g. ceiling). However, and crucially if we are to distinguish Labovian variation from allomorphy, a variant’s linguistic environment never determines whether it can appear. Thus, as Labov (1995: 118) notes, “one is far more likely to say ‘I was workin’
23
on it’ than to say ‘Good mornin’!’ although ‘Good mornin’!’ is a perfectly possible utterance.” Much variationist research tries to discover how the interaction of such linguistic contextual factors correlates with usage of variants. However, perhaps wrongly, these issues will be completely set aside in this dissertation. Maintaining a strict conceptual distinction between the mechanisms of the language faculty and their use, the dissertation is solely concerned with the theoretical question of distinguishing between the mechanisms of Labovian and allomorphic variation. It is this aspect of the present approach to which variationists will most likely object. The following chapter offers some additional justification for the omission of linguistic context from the theoretical analysis of Labovian variation pursued in this dissertation.
2.1.3
Social vs. semantic meaning
(1c) states that variant forms do not express distinctions in semantic meaning or morphosyntactic function. However, even though variants “convey exactly the same grammatical meaning” they are often used to “convey very different social meanings” (Chambers 2002b: 3-4). In other words,
24
Labovian variation involves the sociolinguistic choice of form. A large majority of variationist research attempts to discover the interacting social factors that correlate with usage of variants. This sociolinguistic approach has identified many kinds of social factors that influence variation, such as age, sex, ethnic group, economic class, social-network relationships, etc. seemingly ad infinitum (for overviews, references, and illustrative cases, see Chambers 2002a, Chambers, Trudgill and Schilling-Estes 2002). Even extremely localized social factors can probabilistically influence the appearance of Labovian variants. For example, in the insular community of Ocracoke, North Carolina, Wolfram and Schilling-Estes (1995) showed that the statistical probability of certain phonological variants occurring was significantly increased if the speaker was a member of a small group of regular poker players. Furthermore, Schilling-Estes’s research (2001, 2002b, 2004) indicates that an individual’s perception of the conversational topic or discourse event has a significant effect on their use of Labovian variants. And in fact, individuals have varying degrees of conscious, volitional control over their use of Labovian variants in various social context (see Wolfram and Schilling-Estes
25
1995, Schilling-Estes and Wolfram 1999 for examples of 'demonstration' of variants). The social significance of Labovian variation is a matter of usage outside the scope of linguistic theory on the biolinguistic perspective adopted here (as discussed in the following chapter), and thus the issue is not directly addressed in this dissertation. Nevertheless, it should be borne in mind that we cannot fully understand language change without at least some notion of how the mechanisms of variation are deployed for social purposes in human populations. The next section turns to consider this issue.
2.2
Variationist Methods and Apparent Time
During the last four decades, variationists developed and refined methods for the empirical study of sociolinguistic variation. These methods have yielded the landmark apparent-time method (Bailey et al. 1992) for observing language change in progress, “one of the cornerstones of research in language variation and change” (Bailey 2002: 312).
2.2.1
Vs. Theoretical methods
Variationist methods are quite different from those used by theoretical linguists. Intuitions about the meaning and
26
well- (or ill-) formedness of linguistic expressions have always been the main form of empirical evidence in linguistic theory. Known as ‘grammaticality’ or ‘acceptability’ judgments, linguistic intuitions can be self-observed by the researcher or elicited from others. The advantages and disadvantages of this kind of evidence have sometimes been discussed (Schütze 1996, Cowart 1997, Cornips and Corrigan 2005, Gervain and Zemplén 2005, Parrott to appear-a), but this has not led to the widespread adoption of new methodologies or sources of evidence in theoretical linguistics (despite efforts e.g. Cowart 1997). Variationist methods do not typically involve linguistic intuitions, but rather the observation and quantitative analysis of what is assumed to approximate ordinary linguistic usage.
2.2.2
Sociolinguistic interviews
Most variationist data is gathered using sociolinguistic interviews. These consist of recorded, informal conversations between a researcher and one or more members of a community under investigation. Variationist researchers conducting a sociolinguistic interview may sometimes elicit particular forms from speakers (for example, a word’s pronunciation). However, variationists
27
rarely attempt to elicit acceptability judgments of the sort employed in theoretical linguistics.3 More typically, sociolinguistic interviews are designed so as not to draw interviewees’ attention to the language they are using. Interviewers often introduce conversation topics (such as community controversies, near-death experiences, the supernatural, etc.) that are intended to produce excited and non-self-conscious speech. The goal of a sociolinguistic interview is to elicit language usage that is as naturalistic as possible. The question of how best to observe "everyday speech" has generated much discussion among variationists. There is an unavoidable observer’s paradox--how can we (ethically) record the way people talk when they are not being recorded? This paradox is very important to sociolinguists, who do not want the ordinary sociolinguistic behaviors of their subjects to be obscured by different sociolinguistic behaviors in the presence of interviewers. However, the observer’s paradox is of less importance to theoretical inquiry, which is less concerned with the social aspects of linguistic behavior, and more with what this behavior reveals about the mechanisms of the language faculty.
3
Attempts to do so can result in frustration for all parties; see Chapter 4 and Schilling-Estes (2004).
28
2.2.3
Quantitative analysis
Using recordings and/or transcripts of sociolinguistic interviews, researchers identify instances of variation for quantitative analysis. Researchers count the number of times a variable’s environment occurs, and the number of times each variant appears in the environment. They then calculate the frequency with which a particular variant appears, dividing the number of a particular variant’s appearances by the number of environments where it could have appeared (= #variant/#environs). The resulting value is expressed as a usage percentage. This percentage can be calculated across all speakers in a community, or used to compare the usage of individuals or sets of speakers. Quantitative variation data can be furthermore subjected to computer-aided statistical analysis, typically using a specially designed program like VARBRUL (Fasold 1993). The purpose of this analysis is to determine whether any specific factors make a particular variant statistically more or less likely to appear. As discussed above, factors that are known to probabilistically influence variation include aspects of the variable’s linguistic context, as well as the variants’ (local) social meanings. A program like VARBRUL detects whether any hypothesized factors have a statistically-significant
29
influence (favoring, disfavoring, or zero) on the probability of a variant’s occurrence. Computer-aided statistical analysis is an important tool in variationist sociolinguistics. Since the focus of this dissertation is theoretical rather than sociolinguistic, VARBRUL results are generally not reported in the chapters that follow.
2.2.4
The Apparent-time method
The empirical study of sociolinguistic variation, using the methods above, resulted in a major discovery: language change can be observed as it happens, and not merely reconstructed from historical texts. This is possible using the apparent-time method, a substantive advance in the study of language change. According to Bailey, the apparent-time method “has had an enormous impact both on our knowledge of the mechanisms of change and on our understanding of its motivations” (2002: 312). The apparent-time method requires collecting sociolinguistic interviews from a cross section of age groups in the population under investigation. The variation data are then analyzed quantitatively. If the usage frequency of a variant increases or decreases across age groups (with statistical significance as verified by VARBRUL), it can be inferred that a change is taking place.
30
The speed of a change can be determined by looking at the slope of the increase/decrease in variant usage between age groups. When one variant is used 100% of the time in a population, the change is complete. The apparent-time method rests on the reasonable and empirically verifiable assumption that older individuals use variants at approximately the same frequencies as when they were younger. Of course, this means that change in apparent time must be distinguished from ‘age grading,’ a sociolinguistic phenomenon where individuals change their variant usage frequencies as they grow older. The distinction requires comparison with real-time data from the (longitudinal or sample) study of individuals, which can be a challenging task. In the variation literature, the validity of the apparent-time method has been consistently supported when apparent-time data have been checked against available real-time data (e.g. Bailey et al. 1992, Bailey 2002, Parrott 2002, Schilling-Estes 2005). Because such results have been replicated, the apparent-time method is by now a standard tool in variationist research (e.g. Labov 1994, 2001). Most variationist research using the apparent-time method has focused on phonology and phonetics rather than morphosyntax. For example, Labov’s epic two-volume work
31
Principles of Linguistic Change (1994, 2001) presents only studies of sound changes in progress, apparently conceding the study of morphosyntactic variation and change to “diachronic studies of texts” (1994: 3).4 And indeed, most theoretical research on morphosyntactic change has focused on historical changes reconstructed from textual evidence (e.g. Lightfoot 1999, Kroch 2001). Thus far, there has been little theoretical interest (especially within the Minimalist program) in utilizing the apparent-time method to observe morphosyntactic change in progress.
2.3
Dialect Death and Concentration on Smith Island
In order to provide an example of the phenomenon of Labovian variation and the methods used to study it, this section introduces the moribund community of Smith Island, Maryland. Smith Island has several characteristics that make it an ideal laboratory for investigating Labovian variation and language change in progress. Smith Island’s 4
In a footnote, Labov further states that the "best vantage point for studying current syntactic change is found in developing creoles" (1994: 3, fn. 1). Although the study of creoles has certainly informed linguistic theory (e.g. Bickerton 1990, Pinker 1994, DeGraff 1999), Labov does not elaborate on why creoles are necessarily better than apparent-time studies of morphosyntactic variation and change. Insofar as the abominable social conditions that gave rise to traditionallystudied creoles (Holm 1988, 1989) no longer obtain, research on creoles has also relied heavily upon the study of historical texts rather than observing anything like creolization in progress. However, researchers have recently been able to observe the spontaneous emergence of a creole sign language in deaf Nicaraguan children (Kegl, Senghas and Coppola 1999, Senghas, Kita and Ozyurek 2004).
32
homogenous and relatively isolated population minimizes issues of dialect contact and mixture. The small (and decreasing) population size means that it might be possible in the future to interview many or most of the residents. Longitudinal real-time studies are also possible. In the meantime, we can be somewhat confident that extant samples are representative of the community. Indeed, there are a fairly large amount of data from Smith Island already available for analysis. Georgetown University graduatestudent Rebecca Setliff recorded sociolinguistic interviews on Smith Island during 1983 (with the assistance of island residents recruited as interviewers). Professor Natalie Schilling-Estes of Georgetown University conducted more interviews on Smith Island from 1999 to 2001 with the assistance of this author and other graduate students. In total, there are over 100 hours of recorded sociolinguistic interviews (though not all of these have been transcribed). Because these data straddle a 20 year period, they allow for both apparent- and real-time analysis. Moreover, the process of concentration was first documented on Smith Island and is so far unique to this community (Schilling-Estes 1997, Schilling-Estes and Wolfram 1999, Schilling-Estes 2000, Parrott 2002, Wolfram and Schilling-Estes 2003, Schilling-Estes and Wolfram 2003,
33
Trester 2003, Mittelstaedt to appear, in progress). Concentration is a process of variation and change in progress, whereby usage of particular variants (assumed to be socially characteristic) increases in apparent time as a dialect dies. As a result, “linguistic distinctiveness is heightened among a reduced number of speakers” (SchillingEstes and Wolfram 1999: 488) for social reasons. On Smith Island, linguistic concentration seems to express solidarity in a moribund community. This process allows us to observe accelerated language change. Both phonological and morphosyntactic variables involved in concentration have been documented
on Smith Island. Two morphosyntactic
variables are the topics of Chapters 4 and 5, and include the weak expletive it (Parrott 2002) and weren’t leveling (Schilling-Estes and Wolfram 1994, Schilling-Estes 2000, Wolfram and Schilling-Estes 2003, Mittelstaedt to appear, in progress). In the last section below, we will examine two phonological changes: /ay/ with a raised and/or backed nucleus e.g. [sId]/[s>Id] ‘side’, and /aw/ with a glidefronted and/or raised nucleus e.g. [saInd]/[sEInd] ‘sound’ (Schilling-Estes 1997, Schilling-Estes and Wolfram 1999).
2.3.1
Smith Island: Geography, history, community
34
One in name and community identity, Smith Island is actually a small cluster of islands and wetland marshes located in the Chesapeake Bay, just on the Maryland side of the Virginia - Maryland border. Although many of the Chesapeake Bay islands were once inhabited, only Smith Island and nearby Tangier Island, Virginia are currently populated.
35
Map 1.
Smith Island (Ì), Chesapeake Bay, and surroundings
Map 2.
Smith Island (Ì): Ewell, Rhodes Point, Tylerton
36
English, Cornish, and Welsh immigrants first established settlements in 1657, and Smith Island has been continuously populated since then. The original population consisted of a few groups of farmers, grew to 19 families by 1808, and then to 300 families by the end of U.S. Civil War. In more recent decades, Smith Island’s population had remained steady at about 650 residents, but began to fall in 1980, reaching 459 in 1990. Smith Island had just 364 residents in 2000, and this population is quite homogenous. Over 98% of Smith Island’s residents identified themselves as ‘white’ and native to the state (presumably they are island natives) (U.S. Census Bureau, 2000). Smith Island is separated from Crisfield, MD, the closest town on the mainland, by a 45-minute boat ride. The ride is sometimes impossible during the winter, when the waters of the bay can freeze up for days at a time. There is no airport, but automobiles are shipped onto Smith Island by boat for use by residents. There are three small towns on Smith Island: Ewell, Rhodes Point, and Tylerton. Ewell, the largest town and de facto capital, is connected with Rhodes Point by a short road. Tylerton, the most isolated of the three communities, can be reached only by boat.
37
Because of Smith Island’s geographical isolation, social contact with the mainland has been limited in nature. Children attend elementary and middle school in Ewell, but travel daily to Crisfield for high school. Mainlanders are regarded as “foreigners,” although tourism, a modern development, draws small numbers of them. Unlike Tangier Island, VA or North Carolina’s Outer Banks Islands, Smith Island has never catered to tourism. Ferry service is infrequent and scheduled for islanders rather than for visitors, making days trips inconvenient. There are two small “bed and breakfast” style hotels on the island, both in Ewell, and no tourist facilities other than a small visitors center. Very few mainlanders live on Smith Island year round, or settle for long periods of time. The Smith Island community is not likely to remain intact. Political and environmental factors beyond the control of the islanders are having a devastating effect on the independent small-scale oyster and crab harvesting that is the community’s economic backbone. For this reason, residents are leaving Smith Island in growing numbers to find employment; many young people leave Smith Island permanently after graduating from high school. Moreover, the islands are being eaten away by the constant erosion of the Chesapeake Bay. Flooding is increasingly common, and
38
Smith Island will almost certainly be uninhabitable in less than a century. (For non-linguistic sources on Smith Island's history and culture, see Horton 1987, Dize 1990, Wennersten 1992, Horton 1996, or Sheenan 1994. See also http://www.smithisland.org.)
2.3.2
Dialect death and concentration
Smith Island’s geographic and social isolation has allowed the independent development of a dialect that is relatively distinct from English varieties spoken on the mainland. This insular variety is in the process of dying as the speech community becomes so dispersed that there are no longer any native speakers. The death of a language or dialect is a “sociolinguistic phenomenon” (Wolfram 2002) that can teach us about how languages change over time. Moribund languages and dialects are known to undergo a process of dissipation, whereby usage of locally characteristic variants gradually decreases while usage of variants from an encroaching dominant or majority dialect simultaneously increases. In other words, as a dialect dies by dissipation, it becomes more like the dialect that will eventually replace it. Dialect death by dissipation is currently taking place in the community of Okracoke, one of North Carolina’s Outer
39
Banks Islands (Wolfram and Schilling-Estes 1995, SchillingEstes 1997, Schilling-Estes and Wolfram 1999). Dialect death is proceeding via a very different process called concentration (Schilling-Estes 1997, Schilling-Estes and Wolfram 1999) in the Smith Island population, where the apparent time method shows that usage of certain locally characteristic variants is increasing rapidly over a few generations while usage of mainlandcharacteristic variants decreases. In other words, as the Smith Island dialect approaches death because of ongoing population attrition, it becomes less, not more, like the encroaching mainland dialects that will eventually replace it. Dialect death by concentration was first documented on Smith Island but has not been documented elsewhere. This is probably because the set of social circumstances that lead to death by concentration are more rare than the social circumstances that lead to death by dissipation.5
2.3.3
Two phonological changes in progress
Schilling-Estes and Wolfram (1999) discuss two phonological changes in detail, including spectrographic and VARBRUL analyses not reported here. Although these particular 5
For more on the sociolinguistics of concentration on Smith Island, see Schilling-Estes (1997, 2000, 2005) Schilling-Estes and Wolfram (1999, 2003), Wolfram and Schilling-Estes (2003) and Mittelstaedt (to appear, in progress).
40
variables are found on both Smith Island and Ocracoke, their patterning in apparent time is different in each community. On Ocracoke, usage of the distinctive variants is declining over time due to dialect death by dissipation; on Smith Island, usage of the distinctive variants is increasing over time due to dialect death by concentration. Simplifying for purposes of discussion, the two variables are as follows. (6)
/ay/ with a raised or backed nucleus, e.g. [sId] or [s>Id] ‘side’ The raised/backed variants ([eI]/[>I]) are counted
together in the quantitative analysis below. Unlike Ocracokers, Smith Islanders are not generally aware of variation in /ay/ and do not demonstrate or comment upon it. (7)
/aw/ with a glide-fronted and/or raised nucleus, e.g. [saInd] or [sæInd] ‘sound’ Schilling-Estes and Wolfram (1999: 498) observe that
variation does not occur in “word-final...(e.g. [naU] ‘now’, [haU] ‘how’)” nor in “word-medial syllable-final...(e.g. [thaUr] ‘tower’)” environments. They suggest that “/aw/ has undergone an allophonic split, in which fronting is restricted to closed syllables.” Thus, open syllables are a linguistic environment that determines the appearance of
41
[aU]; closed syllables are an environment in which variants ([aU]/[aI]/[æI]) occur non-deterministically. The glide-fronted/backed variants ([aI]/[æI]) are counted together in the quantitative analysis below. The /aw/ variable is especially salient on Smith Island compared to /ay/. Variation in /aw/ is frequently commented upon, and the different variants can even be demonstrated by some speakers (see Schilling-Estes and Wolfram 1999 for examples and discussion). Schilling-Estes and Wolfram divided the speakers in their sample into three generation groups by year of birth; these generation groups are maintained throughout the dissertation for purposes of comparison. (8)
Generation Groups Generation I: b. 1899 - 1916 (age 84 - 67 at time of interviews, 1983) Generation II: b. 1944 - 1961 (age 39 - 22 at time of interviews, 1983) Generation III: b. 1966 - 1971 (age 17 - 12 at time of interviews, 1983) Looking at these phonological variables in apparent
time, we can observe the changes in progress. Only usage percentages are given (#variant/#environs), not VARBRUL statistical probabilities. Usage percentages are totals within generation groups. Usage percentages are similar
42
between individuals in these generation groups, with outlier individuals typically explainable in social terms (Mittelstaedt in progress, Schilling-Estes 2005). Rebecca Setliff organized the 1983 sociolinguistic interviews from which these data were extracted. For more detailed information on the data and quantitative analyses, see Schilling-Estes and Wolfram (1999).
100
Usage Percentage
90 80 70
raised /ay/
60 50
glide-fronted /aw/
40 30 20 10 0 Gen. I
Gen. II
Gen. III
Generation Group
Figure 1. Raised /ay/ and Glide-fronted /aw/ on Smith Island, in apparent time, 1983 interviews (Schilling-Estes and Wolfram 1999, graph adapted from Schilling-Estes 2000) For both of these phonological variables, usage of the socially distinctive variants increases in apparent time. As evidence that dialect death by concentration is really an accelerated form of language change, Schilling-Estes and Wolfram
point out that the patterning of these variables
(especially /aw/) in apparent time “approximates the Scurve that characterizes the diffusion of new language 43
forms in healthy language varieties (e.g. Bailey 1973)” (1999: 513). On Smith Island there is a more dramatic increase in the use of glide-fronted /aw/ variants ([aI]/[æI]), whose usage jumps by 50% between Generation I and II, than raised /ay/ variants ([eI]/[>I]), whose usage rises by 16% between Generation I and II. This difference in the rate of change can be attributed to social factors--as noted above, /aw/ variation is highly salient on Smith Island and contributes to the islanders’ sense of unique linguistic identity. However, despite their relative unawareness of
variation
in /ay/, islanders’ use of the distinctive variants ([eI]/[>I]) also increases in apparent time for this variable. This is characteristic of death by concentration: multiple variables are affected, not only socially salient demonstration variables.
2.4
Conclusion
This chapter has defined the phenomenon of Labovian variation, reviewed the methods used by variationists to study change in progress, and given an example from Smith Island and its process of dialect death via a process of concentration. The next chapter introduces the
44
biolinguistic perspective, discusses the place of sociolinguistics on this perspective, and gives an overview of the theoretical model that will be used to analyze Labovian variation in morphosyntax, on Smith Island and elsewhere.
45
Chapter 3: Biolinguistics and the Distributed Morphology Model
3.
Introduction
This chapter orients toward the other side of the variation gap, and is structured as follows. Section 3.1 briefly introduces the biolinguistic perspective and discusses the complementary scopes of linguistic theory and sociolinguistics under this framework of inquiry. Section 3.2 gives an outline of the Minimalist, Distributed Morphology (DM) theoretical model adopted in the dissertation. Section 3.3 concludes the chapter by considering some reasons for using this model to analyze Labovian variation in morphosyntax.
3.1
The Biolinguistic Perspective
An explicitly biological perspective on language and its scientific study was first articulated by Noam Chomsky (e.g. 1959, 1965, 1986, and much other seminal work), who continues to have a leading influence on research and theory in linguistics (e.g. Chomsky 2000b, 2002, Hauser, Chomsky and Fitch 2002, Chomsky 2004b, 2005a, Fitch, Hauser and Chomsky 2005, and much other work). On this 46
perspective, the capacity or faculty of language is understood to be a genetically endowed biological system that is both unique to Homo sapiens1 and universal among individuals of the species (with rare exceptions involving neurological damage, psychological pathology, or genetic developmental disorder). The human language faculty is conceived as a specialized cognitive ‘organ’ (see especially Anderson and Lightfoot 1999, 2002, who argue that linguistic theory is therefore a kind of 'cognitive physiology'). Thus, the scientific study of language, like the scientific study of any biological system, should utilize the methods and insights of the biological sciences (on which, see e.g. Mayr 1982, 1997). This so-called biolinguistic perspective (Jenkins 2000) is by now the consensus view among a wide community of linguists and other scientists concerned with understanding the human trait of acquiring, using, and understanding language (Jenkins 2004). Biolinguistics is a multidisciplinary field of inquiry situated within the study of human biology, including cognitive neuroscience (e.g. Grodzinsky and Finkel 1998, Ullman 2004, Phillips, Kazanina and Abada 2005), genetics and molecular biology (e.g. Gopnik and 1
Notwithstanding controversies about the alleged linguistic abilities of non-human primates and other animals (e.g. Terrace 1979, Pinker 1994: 334-342, Hauser 1997, Ramus et al. 2000, Hauser, Chomsky and Fitch 2002, Fitch and Hauser 2004, Gentner et al. 2006).
47
Crago 1991, Vargha-Khadem et al. 1995, Vargha-Khadem et al. 1998, Lai et al. 2001, Enard et al. 2002, Shu et al. 2005), and ultimately within evolutionary theory (e.g. Bickerton 1990, Pinker and Bloom 1990, Hauser 1997, Hauser, Chomsky and Fitch 2002, Pinker 2003, Fitch and Hauser 2004, Pinker and Jackendoff 2005, Fitch, Hauser and Chomsky 2005, Jackendoff and Pinker 2005, Gentner et al. 2006).
3.1.1
Five questions for biolinguistic inquiry
Three (and later four) guiding research questions have been repeatedly formulated throughout the biolinguistic literature (e.g. Chomsky 1986, Chomsky and Lasnik 1993, Jenkins 2000, etc.). As Hauser (1997: 2) points out in his comprehensive study of animal communication systems, these same questions can be traced to the work of biological ethologist Tinbergen (1952), and they constitute “the only fully encompassing and explanatory approach to communication in the animal kingdom, including human language.” The four questions are reformulated below. I have added a fifth question about populational phenomena, an issue that has been seldom addressed in biolinguistics despite the importance of ‘population thinking’ in the biological sciences (see Mayr 1982, 1997, 2001). As discussed in the following sections, these five questions
48
provide a framework of inquiry that encompasses both linguistic theory and sociolinguistics within the biolinguistic perspective. (1)
i.
Mechanistic questions: What are the internal, cognitive mechanisms of the human language faculty? How do these correspond to neurological mechanisms?
ii.
Developmental questions: What is the genetically endowed initial state of the language faculty (UG)? How does the language faculty develop to a mature state based on environmental linguistic input?
iii. Functional questions: How are the mechanisms of language faculty used, and for what purposes? What are the social and other functions of the language faculty? iv.
Evolutionary questions: How did the language faculty evolve in our species? Which linguistic mechanisms have precursors in other species, and which are unique to humans?
v.
Population questions: What is the role of populations in language phenomena? How do linguistic changes disperse through populations?
3.1.2
The Scope of biolinguistic theory
On this biolinguistic perspective, linguistic theory is primarily concerned with the kinds of mechanistic and developmental questions given in (1i-ii) above.
49
Biolinguistic theoretical research tries to develop explicit2 formal models of the cognitive systems that constitute the language faculty in both its initial and mature states, before and after environmental linguistic experiences. The scope of linguistic theory encompasses all the modules of the language faculty: its interpretive systems of sound/sign, its syntactic combinatorial systems, and its systems of compositional semantics. As we will see in what follows, morphosyntactic theory is specifically concerned with mechanisms involved in the interface between the syntactic computation and the levels of sound/sign. One goal of biolinguistic research is to provide a unified account of the cognitive and neurological mechanisms of the language faculty. Because linguistic theory addresses such mechanistic questions (1i), there has been some, albeit limited, interaction with the study of cognitive neuroscience; at present, these fields of inquiry proceed independently pending eventual unification (for discussion of this relationship, see e.g. Marantz 2005, Poeppel and Embick 2005). Furthermore, because it explicitly addresses developmental questions (1ii), linguistic theory has had a somewhat more robust and still 2
“I have always understood a generative grammar to be nothing more than an explicit grammar” (Chomsky 1995: 162, fn 1); see also (Chomsky 1965: 4).
50
ongoing interaction with the study of (both native and adult) language acquisition (for some examples, see Lust 1977, Lust and Mervis 1980, MacWhinney 1992, Poeppel and Wexler 1993, Wexler 1994, 1998, Rispoli 1994, 1995, Schütze 1997, Markson and Bloom 1997, DeGraff 1999, Lardiere 2000, 2005, White 2003, Yang 2004, Senghas, Kita and Ozyurek 2004, Sorace 2005, among a great deal of other work).
3.1.3
The Scope of sociolinguistics
Biolinguistic inquiry ultimately seeks answers to all five questions about the human language faculty, including an understanding of its use for communication of information, representation of thought, maintenance of social identities, transmission of culture, and numerous other functions. The social meaning expressed by Labovian variation (Chambers 2002a) surely contributes to the explanation of language usage patterns observed in human populations, for example the motivation and spread of language changes. And the social functions of language may very well enter into an eventual account of language evolution. Thus, sociolinguistics can be seen as fitting very naturally into the biolinguistic perspective by addressing questions of language use and social function (1iii), as well as addressing questions about phenomena
51
such as the motivations for and dispersal of language changes in populations (1v). These questions fall outside the ambit of biolinguistic theory, which is limited to mechanistic (1i) and developmental (1ii) questions; these mechanisms are internal to the minds/brains of individuals, and not a property of populations.
3.1.4
Competence/performance vs. methodology
A very contentious aspect of the biolinguistic perspective is what Chomsky (in)famously called the “fundamental distinction between competence (the speaker-hearer’s knowledge of his language) and performance (the actual use of language in concrete situations)” (1965: 4, italics in original). Controversies about competence/performance are a major reason for the variation gap’s persistence over 4 decades.3 Bender (2001) argues that “the distinction is not based on empirical evidence and should be subject to reevaluation.” But this fundamental distinction is not an empirical claim. As reflected in the guiding questions of (1) above, the biolinguistic perspective entails a conceptual distinction between the mechanisms of the language faculty (1i)--‘competence’--and the use and function of these mechanisms (1iii)--‘performance.’ 3
For more perspectives on this debate, see Newmeyer (2003) and subsequent discussion.
52
Certainly such a distinction is justified by its usefulness for scientific inquiry, and has thus been widely adopted in the study of biological systems, including the study of language. Only such a “shift of focus...from behavior or the products of behavior to states of the mind/brain that enter into behavior” (Chomsky 1986: 3) allows us to construct theoretical models of the language faculty’s mature and initial mechanisms. Theoretical explanations addressing the mechanistic and developmental questions of (1i-ii) can be, and are, pursued independently of social and other explanations addressing the functional questions of (1iii,v). Unfortunately, and on both sides of the gap, the competence/performance distinction has often been misinterpreted as a kind of restriction on legitimate empirical methods used in linguistic inquiry. And indeed, sharp methodological differences are also largely responsible for the continued estrangement of syntactic theory and sociolinguistics. For example, Labov (1995: 114) states that syntactic theory is hindered by “a methodological commitment to the elicitation of introspections as a primary data base” and the attendant rejection of “actual behavior as relevant data.” It is true that the empirical basis of linguistic theory is almost
53
entirely built from the experimental elicitation of intuitions about the meaning and well- (or ill-) formedness of linguistic structures from informants (very often from the researchers themselves). Evidence from linguistic intuitions is frequently needed in order to evaluate specific theoretical hypotheses, especially when the relevant structures are not readily observable in everyday usage. However, the experimental collection and analysis of linguistic intuition data for syntactic theory is all too typically conducted in a haphazard and uncontrolled fashion (Gervain and Zemplén 2005, Parrott to appear-a). Occasional calls for reform (Schütze 1996, Cowart 1997) have not led to the widespread adoption of new methods in theoretical linguistics. Biolinguistic theorists should take such methodological criticisms seriously. It should be understood that the necessary conceptual distinction between mechanisms and function in biolinguistics does not entail that data from usage cannot be evidence for theoretical models, only that usage is not itself the subject of linguistic theory. Theorists are sometimes skeptical of evidence from language use, but objections are misplaced. Because we cannot directly observe the faculty of language, we must infer its properties by interpreting
54
whatever empirical data are available to us. There is no direct or privileged view of the language faculty provided by reflection, introspection, intuitions, or any other source of data. Indeed, it must be recognized that acceptability judgments themselves are merely the observation of a kind of meta-linguistic performance. As in all scientific inquiry, it is up to analysts to determine the evidential value of any empirical data. Apparent conflicts between sources of evidence must be aired and evaluated.
3.1.5
The Scope of present inquiry
The primary purpose of this dissertation is to identify mechanisms of morphosyntactic variation within a Minimalist, DM-theoretical framework as outlined below. The three following chapters attempt to provide a theoretical account of the mechanisms of Labovian variation, as distinguished from the mechanisms of allomorphic and parametric variation. Several different kinds of empirical methods are utilized to gather evidence, including sociolinguistic interviews, variationist quantitative and apparent-time analyses, linguistic intuitions elicited from informants, and observed specimens of language in use. The dissertation will not “reject observations when they
55
conflict with introspections” (Labov 1995: 114), nor vice versa. However, this dissertation takes the biolinguistic perspective as discussed above, assuming a sharp conceptual distinction between the mechanisms of the language faculty internal to individuals and their use and function in human populations. Therefore, the scope of the dissertation is strictly limited to the mechanistic and developmental questions of (1i-ii). This dissertation does not attempt to provide any theoretical explanation for the observed probabilistic effects of linguistic context on the frequency of Labovian variation. Nor does the dissertation attempt to provide any theoretical explanation for the social functions of Labovian variation. Both of these issues are taken to be outside the scope of present inquiry, which is restricted to the mechanisms underlying Labovian variation. Especially from a variationist perspective, this total exclusion of social and linguistic context from the theoretical analysis can be seen as a very serious drawback. On this view, these influencing factors are central to the phenomenon of variation itself. Variationist inquiry critically must include quantitative study of how multiple factors affect the frequency of variant usage (Labov 2005). The empirical fact that frequency of variant
56
usage is probabilistically influenced by factors of linguistic and social context has convinced many variationists that the competence/performance distinction of biolinguistic theory is untenable. Thus, as noted above, Bender (2001) argues on such grounds for a theoretical model of variation that includes probability weightings sensitive to both social and linguistic context. But on the biolinguistic perspective adopted here, such devices cannot be included in the theoretical mechanisms of Labovian variation. Empirical facts do not themselves compel us to abandon the useful conceptual distinction between mechanisms and use of mechanisms. Therefore, the theories of variation offered here account for which morphosyntactic objects can be variants, and how this variation differs from allomorphy. These mechanistic theories do not give any account of why individuals choose one variant over another during performance, nor of the social meanings or linguistic contexts that influence this usage choice. The narrow aims of this dissertation should not be taken to entail that variationist issues are outside the scope of biolinguistics, only that they are outside the scope of questions (1i-ii) that are the concern of biolinguistic theory. Sociolinguistics can and should address the functional and populational questions of
57
(1iii,v) within the biolinguistic perspective. For example, I believe that a unified account of the theoretical mechanisms of Labovian variation and their social functions in human populations will eventually become important for questions of language evolution (1iv). In a similar way, a unified account of how variable and categorical mechanisms internal to the language faculty are processed by external performance systems will be necessary for an eventual explanation of the probabilistic effect of linguistic context on the usage frequency of variants. That biolinguistic inquiry into this empirical characteristic of Labovian variation should focus on processing systems follows from the competence/performance distinction: mechanisms of variation are part of linguistic competence and the usage of these mechanisms are part of performance. Indeed, this should be seen as a useful hypothesis that can generate testable predictions to be investigated using psycholinguistic and even neurolinguistic methods (such as those used by Phillips 2004, Phillips, Kazanina and Abada 2005 to study syntactic processing). For example, we might ask whether Labovian variation is processed at different speeds or by different brain regions depending on the linguistic context of the variants. We might also ask how patients suffering from brain damage, aphasias, or lesions
58
process and perform Labovian variation. Could we observe a kind of dissociation effect, whereby such a patient has the same Labovian variables available but does not use them at the frequencies predicted by the variants’ linguistic contexts? These and similar avenues of investigation fall far outside the restricted scope of this dissertation, but should be addressed by future collaborative research from a biolinguistic perspective.
3.2
A Minimalist, Distributed Morphology Model
This section outlines the Minimalist, DM-theoretical model adopted in this dissertation. This presentation is intended to prepare the ground for the DM analyses of Labovian variation proposed in the next three chapters, and so it highlights aspects of the model that will be relevant for subsequent chapters. For this reason, many important issues that are currently being investigated in syntactic and morphological theory are simply set aside. Additionally, discussion of some specifics will be postponed until later chapters.
3.2.1
A Minimalist program for biolinguistic theory
At the outset, I want to stress the endlessly reiterated point that Minimalism is not a theory. It is a research
59
program, a strategy to guide theoretical inquiry into the human faculty of language within the biolinguistic perspective. As Hornstein, Nunes and Grohmann (2005) remark in their recent textbook Understanding Minimalism, the program is intended to “promote a research environment in which various alternative, equally ‘minimalist’ yet substantively different, theories” can be compared and evaluated. This dissertation follows such a research program: I adopt a specific Minimalist, DM-theoretical model based on Embick and Noyer (to appear), contrast it with a substantively different Minimalist theory based on Chomsky (2000a), and evaluate these models by using them to analyze empirical cases of Labovian variation in morphosyntax. However, no non-Minimalist theories of the language faculty--such as Head-driven Phrase Structure Grammar (HPSG) (Pollard and Sag 1994, Bender 2001), Optimality Theory (OT) (Prince and Smolensky 1993, Archangeli and Langendoen 1997), theories based on Kayne’s Linear Correspondence Axiom (LCA) (Kayne 1994, Moro 2000), and Jackendoff’s Parallel Architecture theory (Jackendoff 1997, 2002)--are considered or compared in this dissertation. Such an endeavor is left to future work.
60
The Minimalist program4 (Chomsky 1993, 1995, 2000a, 2001, 2004a, 2005a, b, Chomsky, Hauser and Fitch 2005) follows a general consensus developed within the Principles and Parameters framework (e.g. Chomsky and Lasnik 1993) and earlier Government and Binding models (e.g. Chomsky 1981, 1982, 1986). The human faculty of language is understood to be a specialized computational system that links sound/sign and semantic meaning by generating hierarchically organized phrase structures built from bundles of lexical features. This computational system, hereafter referred to as the syntax, must interface with at least two external systems: the Sensory-Motor (S-M) systems of sound/sign and the Conceptual-Intensional (C-I) systems of meaning. While these systems are external to the language faculty, they are internal to the minds/brains of individuals. The Minimalist program proceeds from a conjecture that the mechanisms of syntax are perfectly optimal, constituting a maximally economical mapping of sound/sign and meaning representations to be interpreted by the language-external S-M and C-I systems. The objective of Minimalist research is to determine the extent to which this conjecture is 4
Note that different versions of this research program's name appear in the literature, including with both terms in capitals (Minimalist Program), with abbreviation (MP), and with no capitals (minimalist program). Throughout, I capitalize ‘Minimalist’ since I regard this term as having a unique referent; I do not capitalize ‘program’ since it is a generic term.
61
true. This research program places severe constraints on theoretical models of the language faculty. Postulated mechanisms should be reduced to a bare minimum, and these mechanisms should be restricted either by independent principles of computational efficiency or motivated by requirements imposed at the interfaces. Therefore, a Minimalist theoretical architecture of the language faculty will include only a set of lexical primitives, a syntactic computational system that combines these primitives into more complex structures, and two levels of representation that are required to function as interfaces with language-external systems: the S-M interface level of Phonetic Form5 (PF) and the C-I interface level of Logical Form (LF).
3.2.2
The Theory of Distributed Morphology
The term ‘morphology’ refers to theories about lexical primitives and the principles that govern their combination and structure. The last two decades have seen the development of explicit morphological theories about the mapping of hierarchical syntactic structures onto linear phonological forms (Baker 1988, Anderson 1992, Aronoff 5
I continue to use the term PF; perhaps a more accurate name for this level would be Perceptual Form, following Stroik (forthcoming, see Stroik and Putnam 2006).
62
1994, Beard 1995, Spencer and Zwicky 1998). Any adequate theory of the language faculty must include some account of this mapping procedure. Thus, the task of morphological inquiry is to discover how the syntax-phonology mapping is carried out, and to characterize the place of morphology in the architecture of language. Assuming the Minimalist architecture introduce above, the theory of Distributed Morphology (DM) provides a substantively different and much more articulated model of lexical primitives, their syntactic combination into complex structures, and the operations that apply during the post-syntactic computation to the PF interface. This theory was first presented comprehensively by Halle and Marantz (1993) and has been subsequently refined by a community of researchers over the last 15 years (e.g. Bobaljik 1995, Halle 1997, Marantz 1997, Harley and Noyer 1999, Wiklund 2001, Embick and Noyer 2001, McFadden 2004, 2006, Kandybowicz 2006, Embick and Marantz 2006, Nevins to appear, Embick to appear-a, b, Embick and Noyer to appear, and many others). DM’s treatment of lexical primitives is a major departure from standard Minimalist theories. Most importantly, in DM there is no single generative lexicon that feeds the syntax. Lexical information is distributed among three distinct lists, and the combination of lexical
63
primitives into complex word structures is accomplished by syntactic and post-syntactic mechanisms. Furthermore, DM follows other morphological theories (see especially Beard 1995) in assuming that all or some phonological features are not part of the lexical primitives prior to or during the syntactic computation. Instead, DM holds that the phonological features of abstract functional elements are late inserted during the post-syntactic morphological computation to PF. This dissertation adopts, with some modifications, the Minimalist, DM model of the language faculty that is presented by Embick and Noyer (to appear, see also Embick and Noyer 2001). The remainder of this section gives an overview of this DM-theoretical model. I make no attempt to be comprehensive, focusing instead on aspects of the theory that are relevant for the analyses of Labovian variation to follow, and postponing consideration of many specifics until subsequent chapters. Moreover, as Embick and Noyer (to appear) review, much of the current research in DM investigates “operations on argument structure and related areas” (e.g. Marantz 1997); this dissertation says almost nothing about such issues.
3.2.3
The DM architecture of the language faculty
64
The diagram given below depicts the Minimalist, DM architecture of the language faculty adopted in this dissertation. The sub-sections that follow review its components and computational mechanisms.
65
Syntactic terminals Roots Abstract morphemes
Lexical Array
Í
Multiple Spell Outs
Merge Move (= Internal Merge) Agree?
Î
Narrow Syntax
Phonetic Form (PF) Phonology, prosody Merger Fusion Vocabulary Insertion
Vocabulary
Morphological Component Sensory-Motor (S-M) systems
Encyclopedia
ConceptualIntensional (C-I) systems
Logical Form (LF) Semantics
Figure 1: The DM architecture of the language faculty
3.2.4
Listed primitives 1: Syntactic terminals
The first list of lexical primitives in this model is called the syntactic terminals. These are the input to the 66
combinatorial syntactic computation, and correspond to syntactic heads (X0) in Minimalist phrase structure (Chomsky 1995, more on this below); they may also be referred to as morphemes or terminal nodes. Syntactic terminals consist of unstructured bundles of semantic, syntactic, morphological, and/or phonological features. The full possible set of lexical features is endowed by UG, and developmental exposure to linguistic environmental input allows individuals to acquire the particular subset of these universal features that will comprise their inventory of syntactic terminals. In early versions of DM theory (Halle and Marantz 1993), all syntactic terminals lacked phonological features until post-syntactic late insertion of phonological features. However, this dissertation follows Embick and Noyer (to appear) in assuming that syntactic terminals come in two types.6 The first are called abstract or functional terminals; these contain semantic and morphosyntactic features but lack phonological features. Thus, the phonological features of abstract morphemes must be supplied by Vocabulary Insertion in the post-syntactic morphological component. The second type are called Roots; 6
See especially Beard (1995) for persuasive arguments that lexemes (i.e. Roots) must be distinguished from functional morphemes in any adequate morphological theory.
67
these contain both semantic and phonological features. On this theory, Roots do not contain any syntactic categorical information, but instead are combined with abstract category heads in the syntax. The DM theory of Roots and categorizing heads is intended to account for a range of facts involving argument structure and category-changing derivational morphology (see especially Marantz 1997). I will not review these issues here, since the analyses of Labovian variation pursued in this dissertation focuses on the properties of abstract terminals and their behavior in the post-syntactic morphological component. To give a simple example of how Roots, category heads, and abstract morphemes combine, consider the nominal phrase a ghost. The semantic features of its Root terminal are represented below with small capitals inside of a squareroot symbol, and phonological features are indicated with subscripted slash brackets. (2)
Root terminal √GHOST/ost/ The abstract terminals in this phrase consist of a
nominalizing category head, indicated with a lower-case, italicized n, and a Determiner (D) morpheme with a semantic definiteness feature indicated by subscripted [Def:-]. Both of these abstract terminals lack phonological features. 68
(3)
Abstract terminals n
=
category head
D[Def:-]
=
functional head
The Root terminal combines with the category head in the syntax (see below), yielding an nP that corresponds to the noun ghost. (4)
Noun = ghost nP 3 n √GHOST/ost/ Next, the nP is combined with the semantically
indefinite D terminal in the syntax, yielding a Determiner Phrase (DP). Note that this is the standard analysis of nominal phrase structure after Abney (1987), and I continue to assume a DP analysis throughout the dissertation. D will be provided with its phonological exponent a in the postsyntactic morphological component. (5)
Determiner Phrase (DP) = a ghost DP 3 nP D[Def:-] 3 n √GHOST/ost/ Because of the narrow scope of inquiry pursued here,
for reasons of space, and for simplicity of exposition, I will abstract away from Root and category-head structures
69
in the remainder of this dissertation. Syntactic category will be represented hereafter with a (subscripted) capital letter, and the phonological features of Roots are indicated with italics. (6)
Determiner Phrase (DP) = a ghost DP 3 D[Def:-] Nghost
3.2.5
The Lexical Array
The syntactic computation does not access the entire listed inventory of terminals. Instead, each session of syntactic computation may access only the subset of terminals selected into an unordered set called the Lexical Array (following Chomsky 2000a, cf. the Numeration of Chomsky 1995). For example, consider the Lexical Array for the sentence A ghost haunts this house. As noted above, all terminals are given with their phonological exponents in italics and their category label as a subscript. Analysis of verbal tense and agreement features is postponed until the following sections, and represented here by the terminal T for Tense. (7)
Lexical Array = A ghost haunts this house {Nhouse,
Dthis, Vhaunt,
T,
70
Nghost, Da}
3.2.6
Narrow syntax 1: Merge and phrase structure
Hierarchical phrase structure is a universal characteristic of human language. Therefore, by ‘virtual conceptual necessity,’ (Chomsky 1995) a minimal structure-building operation must be among the mechanisms of syntax. Following the bare phrase structure theory presented in Chomsky (1995), hierarchal phrase structures are derived by recursive application of the operation Merge. Syntactic terminals are input to the syntax from the Lexical Array. Merge combines two syntactic terminals, or phrases already constructed from them, to yield a new, more complex syntactic object. For a simple example of how Merge builds hierarchical syntactic structure, consider the verb phrase haunt this house. Application of the operation Merge, indicated below with an arrow, combines the D this with the N house. On this theory of phrase structure, one of the Merged constituents ‘projects’ its category to label the new phrase. Here, the result is a DP this house. Phrase structures are given in both bracket and tree notation below for clarity. (8)
{...
Nhouse, Dthis
...}
MERGE Î [DP
Dthis
[Nhouse]] 71
=
DP 3 Dthis Nhouse Next, the DP this house is Merged to the Verb (V)
haunt, yielding the VP haunt this house (still postponing analysis of verbal tense and agreement). Notice that the hierarchical structure of VP comes as an automatic consequence of the operation Merge: DP is the complement of V, and contained within VP. (9)
[DP
Dthis
[Nhouse]], {...
Vhaunt
...}
MERGE Î [VP
Vhaunt
[DP
Dthis
[Nhouse]]]
=
VP 3 DP Vhaunt 3 Dthis Nhouse
Very recent Minimalist theory (Chomsky 2005a, Chomsky, Hauser and Fitch 2005) has advanced a conjecture that the narrow syntactic computation is extremely sparse, consisting only of Merge and no other mechanisms.7 Movement is just the result of the internal application of Merge, targeting a phrase inside the structure already constructed rather than a new lexical item. On such an austere theory, 7
Hauser, Chomsky, and Fitch (2002) hypothesized that recursive Merge is the only linguistic mechanism that is an evolutionary innovation, unique to humans. This launched an ongoing comparative empirical research program to test the hypothesis in animals, so far consisting of Fitch and Hauser’s (2004) experiments with tamarin monkeys and the recent experiments with starlings reported by Gentner et al. (2006).
72
the observed variety of human languages must be explained by the requirement to interface with external systems. Of course, such a hypothesis cannot be addressed or evaluated here, given the narrow focus of the dissertation. But this strong Minimalist hypothesis seems at least consistent with the DM theoretical approach, which locates many linguistic mechanisms in the post-syntactic computation to the PF interface with S-M systems.
3.2.7
Narrow syntax 2: Agreement and feature checking
This sub-section presents a specific Minimalist theory of case and verbal agreement morphology following Chomsky (2000a, see also Adger 2003, Pesetsky and Torrego to appear). On this theory, by now the standard,8 the features of abstract syntactic terminals come in two flavors: those that are interpretable at the LF-interface, and those that are uninterpretable at the LF-interface. Features of the latter sort must be ‘checked’ during the narrow syntactic computation before they reach the LF interface, or else the derivation will ‘crash’ (and the structure is predicted to be unattested and/or judged ill-formed). Following Pesetsky and Torrego (to appear), uninterpretable features do not have a value until they are checked in the syntax; however, 8
But compare, for example, the theory of Grohmann (2003), which maintains the earlier theory of specifier-head agreement.
73
as we will see below, the value of a checked feature is available in the post-syntactic morphological component. Notation of features and checking is as follows: uninterpretable features are indicated with a prefixed italicized u, uninterpretable features checked in the syntax are indicated with a strikethrough font, and a feature’s value is indicated following a colon. A schematic of this feature checking notation is given below. (10) Syntactic feature checking [u(ninterpretable)Feature:value] On this theory, the abstract syntactic terminal corresponding to finite Tense (represented here by T[Past:±]) enters the narrow syntax with semantically uninterpretable phi (φ) features of Person and Number. These same φ features are semantically interpretable on D: a Person feature can be valued first (1), second (2), or third (3), and a Number feature can be valued singular (s) or plural (p). The uninterpretable φ features of T must be checked by Agreement with the valued, interpretable φ features of D. On this theory, Case is an uninterpretable feature of D that is valued Nominative when checked by finite T, Accusative when checked by V, and Oblique when checked by P. (11) Finite Tense, with uninterpretable φ features T[Past:±,
uPers: , uNum: ]
74
(12) D with valued φ features and uninterpretable Case D[Pers:(1,2,3),
Num:(s,p), uCase: ]
Following Chomsky (2000a, see also Adger 2003), the operation Agree is the mechanism for checking uninterpretable features in the narrow syntax. When this operation applies, the uninterpretable feature set of a lexical item ‘probes’ its c-command domain for an identical set of features on another lexical item, the ‘goal.’ Finding such a matching set, the uninterpretable features of the probe are checked and valued. Only a full set of features on the goal can induce checking of features on the probe; Agree is an ‘all-or-nothing’ or ‘one-fell-swoop’ operation on this theory. The Agree operation is schematized below, with finite T as the probe and DP as the goal. The arrow indicates application of the Agree operation in the narrow syntax, α represents the value of φ features for Person and Number, and uninterpretable Case is valued Nominative by the probe of finite T. (13) Agree in narrow syntax [PROBE = T[Past:±,
uφ: ]
... [GOAL = DP[φ:α,
uCase: ]]]
uφ: ]
... [GOAL = DP[φ:α,
uCase:Nom]]]
AGREE Î [PROBE = T[Past:±,
Consider the example sentence from above, A ghost haunts this house. The subject DP a ghost has interpretable 75
φ features valued 3s, and an uninterpretable, unvalued Case feature. Adopting the now-standard VP-internal subject analysis, the DP a ghost is Merged to the VP haunt this house. This DP subject will later move to Merge with TP (further discussion of this movement is postponed until Chapter 4). For reasons of space, lexical category labels and the internal structure of DPs are hereafter omitted, and hypothetical expanded internal structures of VP and TP are omitted throughout the dissertation. For purposes of exposition, X-bar (X’) notation is used to indicate intermediate projections, which have no status in this theory of phrase structure. (14) [VP haunt this house], [DP a ghost[φ:3s,
uCase: ]]
MERGE Î [VP [DP a ghost[φ:3s, =
uCase: ]]
[V’ haunt this house]]
VP 3
DP 6 a ghost[φ:3s, uCase:
]
V’ 5 haunt this house
Next, finite T is Merged to the VP a ghost haunt this house. T has a semantically interpretable feature valued [Past:-], and a set of uninterpretable, unvalued φ features that must be checked. (15) [VP [DP a ghost[φ:3s, uCase: ]] [V’ haunt this house]], {... T[Past:-, uφ: ] ...}
76
MERGE Î [TP T[Past:-, uφ: ] [VP [DP a ghost[φ:3s, [V’ haunt this house]]]
uCase: ]]
=
TP 3 T[Past:-, uφ: ] VP 3 DP V’ 6 5 a ghost[φ:3s, uCase: ] haunt this house Now, the operation Agree applies: the uninterpretable φ features of T probe and find the matching φ features of the goal DP a ghost. T’s uninterpretable φ features are checked and valued 3s by Agreement with DP, and the DP’s uninterpretable Case feature is checked and valued Nominative by Agreement with finite T. (16) [TP T[Past:-, uφ: ] [VP [DP a ghost[φ:3s, [V’ haunt this house]]]
uCase: ]]
AGREE Î [TP T[Past:-, uφ:3s] [VP [DP a ghost[φ:3s, [V’ haunt [DP this house]]]]
uCase:Nom]]
=
TP 3 T[Past:-, uφ:3s] VP 3 DP V’ 6 5 a ghost[φ:3s, uCase:Nom] haunt this house As discussed in Chapter 5, Adger and Smith (2005, see also Adger 2006) analyze a case of Labovian variation using
77
this syntactic theory. However, an Agreement featurechecking analysis of case and subject-verb agreement morphology leads to mismatch in the three cases of Labovian variation examined in the following chapters. Therefore, this dissertation will adopt a current version of DM theory that denies the existence of uninterpretable features in the narrow syntax.9 On such a DM theory, semantically meaningless ‘ornamental’ case and verbal agreement morphology is determined by the post-syntactic insertion of so-called Dissociated features by language-specific rules in the morphological component (following Embick and Noyer to appear, McFadden 2004). Details and discussion are postponed until Chapter 4, which presents a DM analysis of Dissociated subject-verb agreement in English, and Chapter 6, which argues that English lacks even Dissociated case morphology and presents an alternate analysis of pronominal allomorphy.
3.2.8
Spell Out and the LF-interface computation
At some stage during in the narrow syntactic computation, the phrasal structures constructed from terminal nodes must be split onto two separate computational branches as they 9
It is not only DM theorists who want to dispense with semantically uninterpretable syntactic features, for example see Sigurðsson (2006); conversely, not all DM theorists want to eliminate syntactic feature valuing, for example see Nevins (2006, to appear).
78
are mapped to the interface levels of LF and PF. Thus, the operation Spell Out applies, transferring the hierarchical syntactic structures so-far constructed to the morphological component where they undergo further computation required by the PF interface. Further computations are also required to map syntactic objects onto semantic representations at LF. Phonologically ‘covert’ LF-branch operations, for example quantifier raising, are not represented at the PF interface because they occur after Spell Out. No more will be said here about the LF-branch computation or semantics generally. Uriagereka (1999) suggested that Spell Out can apply multiple times in the course of a syntactic computation, transferring separate chunks of syntactic structure to the PF interface. This suggestion launched a Minimalist research program to determine exactly which chunks of structure are Spelled Out, still an open question.10 Chomsky (2001, 2004a, 2005a, 2005b) has developed a proposal that Spell Out applies to two syntactic ‘phases,’ which consist of all the material contained in the complements of the complementizer head C and the transitive verbal head (socalled ‘little v’). Grohmann (2003) proposes an alternative theory where Spell Out applies to three ‘prolific domains’ 10
For example, the first InterPhases conference in Nicosia, Cyprus (2006) was entirely devoted to this question.
79
in clausal structure; these consist of the verbal theta (θ) layer where argument structure is established, the agreement phi (φ) layer where inflectional features are checked, and the complementizer omega (Ω) layer where discourse and clause-typing information occurs. Evaluating such contrasting theories of multiple Spell Out will certainly be relevant for the DM theory of post-syntactic computation in the morphological component, as noted for example by Embick (to appear-a). However, multiple Spell Out is not relevant for the analyses of Labovian variation that are pursued in the following chapters. Thus, nothing more will be said about the issue here, leaving it for future research.
3.2.9
Ordered operations of the morphological component
Syntactic terminals combined into hierarchical phrase structures by Merge in the narrow syntax are transferred to the morphological component by Spell Out. During the postSpell-Out morphological computation to the PF interface, terminals are subject to further operations that “modify and elaborate syntactic structure in limited ways” (Embick and Noyer to appear). These morphological operations differ in their effects on terminal structure, the structural conditions under which they apply, and their consequences 80
for Vocabulary Insertion (Vocabulary Items and their insertion are discussion in the next subsection). Following Embick and Noyer (2001), morphological operations are ordered with respect to the linearization of terminals, which is triggered by the first Vocabulary Insertion of phonological features; a completely linearized string of phonological features is required by the PF interface. Operations that are sensitive to hierarchical syntactic phrase structures must apply before linearization, while operations that are sensitive to linear ordering of terminals must apply after linearization. The former, hereafter referred to as ‘structural’ operations, include Dissociated feature/morpheme-insertion rules (discussed in Chapter 4) and lowering Merger (exemplified below). The latter, hereafter referred to as ‘repair’ operations, include local-dislocation Merger11 (see Embick and Noyer 2001, Embick to appear-a), and Fusion (following Kandybowicz to appear, exemplified below, and discussed in Chapter 5). Determining the ordering of morphological operations and the precise mechanics of terminal linearization are major topics of current research in DM; for detailed proposals and discussion see the DM literature 11
This operation will not be discussed in the dissertation; nor will the operation of Fission, which splits the features of terminals (Halle and Marantz 1993, Halle 1997).
81
already cited and also Embick (to appear-b) and Embick and Marantz (2006). The diagram below illustrates the proposed ordering of morphological operations to be discussed in the following chapters. The computation from Spell Out to the PF interface is indicated with a dashed vertical arrow; application of the different types of morphological operations is indicated with solid horizontal arrows. In Chapter 5, it is claimed that repair operations and postlinearization Vocabulary Insertion are iterated until all the abstract terminals in a structure have been supplied with phonological features. Spell Out Í
Structural operations Insertion of Dissociated features/morphemes, Lowering Merger
Í
Vocabulary search, Insertion
Í
Linearization of terminals
Í
Repair operations (iterated) Fusion, Local-dislocation Merger
Í
Vocabulary search, Insertion (iterated)
PF Figure 2: Ordering of morphological operations
82
Consider the operation of lowering Merger. Following Embick and Noyer (2001, see also Embick and Noyer to appear, Embick to appear-a), this operation lowers a terminal head X to adjoin with the head Y of its complement YP. Notice that the operation crucially refers to hierarchical syntactic phrase structure. The features of X and Y remain in distinct terminal nodes, so that two separate Vocabulary Items (discussed in Section 3.2.10 below) must insert phonological features into X and Y. However, lowering Merger creates a complex terminal head: the lowered terminal X is contained in a maximal projection of the terminal Y. Thus, the operation of lowering Merger is among the mechanisms of affixation in DM theory. Following Embick and Noyer (2001, see also Embick to appear-a), morphological operations and Vocabulary Items (see below) can be sensitive to the constituency of complex heads that result from morphological operations such as lowering Merger (and also from syntactic head movement, as below and in Chapter 5). This dissertation adopts Embick and Noyer’s definition of Maximal-Word (M-Word) as the highest terminal projection not dominated by any other terminal projection. As Embick and Noyer acknowledge, this is the same definition given for H0max in Chomsky (1995). Lowering Merger in the morphological component is
83
schematized below, where X and Y are terminal nodes, XP, YP, and ZP are syntactic phrases, and YM indicates the maximal projection of the complex terminal Y+X created by Merger. (17) Lowering Merger of X and Y XP Î 3 X YP 3 Y ZP 5 ...
XP 3 YP 3 ZP YM 2 5 Y X ...
This schematic is also illustrated in bracket notation below, where the M-Word boundary is indicated with the notation [M ...]. (18) Lowering Merger of X and Y [XP X [YP Y [ZP ...]]]
Î
[XP
[YP [M Y [X]] [ZP ...]]]
For a concrete example of lowering Merger, consider the example sentence A ghost haunts this house. A major empirical motivation for Embick and Noyer’s (2001) analysis is the morphological lowering of the tense and agreement affix to main verbs in English, where syntactic head movement cannot raise main verbs to affixal finite T. As is well known, certain adverbial phrases adjoined to VP evidently do not interfere with the lowering of finite T to full verbs in English. (19) a.
A ghost often haunts this house. 84
b. * A ghost haunts often this house. However, non-adjoined phrases such as negation prevent this operation, requiring the insertion of do to affix with finite T (see Embick and Noyer 2001, and also Halle and Marantz 1993, for more on do support and negation). (20) a. * A ghost not often haunts this house. b.
A ghost does not often haunt this house.
b.
A ghost doesn’t often haunt this house.
The diagrams below depict the syntactic structure for this example sentence when it is input to the morphological component after Spell Out. Both trees and bracket notation are given for clarity. At this stage in the morphological computation, finite T has φ features valued 3s. As discussed in Section 3.2.7 above, these φ features are provisionally assumed to have been semantically uninterpretable on T and therefore checked and valued in the narrow syntax; however, in Chapter 4 it will be proposed that φ features are copied into finite T after Spell Out by English-specific Dissociated agreement-feature insertion rules in the morphological component. The difference between the two analyses is not relevant for the present discussion of lowering Merger. For purposes of exposition, the syntactic movement site of the subject DP is indicated in trace notation; moreover, as noted above, the expanded internal 85
structures of DP, TP, and VP are omitted throughout, and some terminals are given with phonological exponents in italics and category labels in subscript. (21) Spell Out to the morphological component [TP [DP a ghost] [T’ T[Past:-, φ:3s] [VP tDP [V’ Vhaunt [DP this house]]]]] =
TP 3
DP T’ 5 3 a ghost T[Past:-, φ:3s] VP 3 tDP V’ 3 DP Vhaunt 6 this house Next, the operation Merger lowers the finite T terminal to adjoin with the terminal head of its complement VP, creating a complex verbal terminal V+T. (22) Lowering Merger of finite T and V T’ Î 3 T[Past:-, φ:3s] VP 3 tDP V’ 3 haunt DP V 5 ...
T’ 3 VP 3 tDP
V’ 3 VM DP 2 5 Vhaunt T[Past:-, φ:3s] ...
This example is again illustrated in bracket notation below, with the M-Word boundary indicated. As discussed further below, this operation allows a Vocabulary Item to 86
insert the phonological exponent -s into the still-distinct terminal for finite T, now affixed to the main verb haunt. (23) Lowering Merger of finite T and V [TP [DP ...] T[Past:-,
φ:3s]
[TP [DP ...] [VP tDP [M
[VP tDP
Vhaunt
Vhaunt
[T[Past:-,
[DP ...]]] Î φ:3s]]][DP
...]]]
Because the Merger operation is sensitive to hierarchical syntactic structure, it can lower T to V over an intervening VP adjunct. The Adverbial Phrase (AdvP) often is not the complement of finite T, so its terminal head is ‘invisible’ to the operation. (24) Spell Out, VP adjunct [TP [DP a ghost] [T’ T[Past:-, φ:3s] [VP [AdvP often] [VP tDP [V’ Vhaunt [DP this house]]]]] =
TP 3
DP T’ 5 3 a ghost T[Past:-, φ:3s] VP 3 AdvP VP 5 3 V’ often tDP 3 DP Vhaunt 6 this house However, Merger cannot lower T to V over an intervening Negation Phrase (NegP). Unlike an adjunct, NegP is the complement of T, so its terminal head Negation (Neg) is ‘visible’ to the Merger operation. In this case, Neg is 87
not the right target for affixal finite T, so do support is triggered (but not analyzed here, see Embick and Noyer 2001). (25) Spell Out, NegP complement [TP [DP a ghost] [T’ T[Past:-, φ:3s] [NegP Neg [VP tDP [V’ Vhaunt [DP this house]]]]] =
TP 3
DP T’ 5 3 a ghost T[Past:-, φ:3s] NegP 3 Neg VP 3 V’ tDP 3 DP Vhaunt 6 this house Chapter 5 contains some further discussion of negation and lowering Merger, and Chapter 6 proposes that English pronominal Vocabulary Items can refer to hierarchical syntactic structure in the same way as the Merger operation. Now, we turn to consider the morphological operation of Fusion. Following Kandybowicz (to appear), it is argued here and in Chapter 5 that the Fusion operation is triggered by the need to ‘repair’ a morphological terminal structure where the phonological features of Vocabulary Items could not otherwise be inserted. On this analysis,
88
the featural structures of Vocabulary Items are ‘visible’ to the Fusion operation. Therefore, Fusion must apply late in the morphological computation, after the first application of Vocabulary Insertion when terminals are linearized. This is contrary to some earlier DM analyses (e.g. Halle and Marantz 1993, Halle 1997) where Fusion applies before Vocabulary Insertion and linearization. In other words, as Kandybowicz (to appear) points out, the operations of the morphological component are even “more highly distributed than previously believed.” Fusion combines the features of two sub-terminals X+Y contained in a M-word into a single terminal [X, Y]. The new terminal contains all the features of both X and Y, so a single Vocabulary Item may insert phonological features at [X, Y]. A structural condition on Fusion is that terminals X and Y must be sisters under a terminal projection, so Fusion requires the prior application of a structural operation that creates complex terminals--either syntactic head movement or lowering Merger as discussed above. Thus, Fusion is among the mechanisms of suppletion in DM theory. The Fusion operation in the morphological component is schematized below. Because this operation is claimed to apply after linearization, only bracket notation is given. The precise mechanics of linearization are an
89
open research question in DM theory (see especially Embick to appear-a for detailed proposals); here and throughout, morphosyntactic terminal structures are simply assumed to have been correctly linearized, and linear adjacency between terminal nodes is notated with ‘*.’ Note that while late Fusion is not sensitive to the hierarchical structure of syntactic phrases, the constituency of complex terminal heads remains ‘visible’ to the Fusion operation after linearization. Below, X and Y are sub-terminals contained in the M-word of X, and X is right adjacent to Y. (26) Morphological Fusion of X and Y ...[M [Y] * X]...
Æ
...[M Y, X]...
Kandybowicz (to appear) gives a late Fusion analysis of verbal focus in Nupe, a Nigerian language. In this dissertation, the evidently suppletive forms of the English finite auxiliary and copular verb be are an empirical motivation for the late Fusion analysis. In the present tense, these forms are 1s am, 2s are, 3s is, and 1,2,3p are. English be is discussed at some length in Chapter 5, so some details of the analysis are postponed. For present purposes, consider the example sentence A ghost is haunting this house. Here, be is an auxiliary, the head of an Auxiliary Phrase (AuxP). Analysis of the morphosyntactic featural content of auxiliary and copular be is omitted 90
here and throughout; the terminal node for be is given in an italic font. Following Adger and Smith (2005) for purposes of comparison in Chapter 5, it is assumed that be raises by syntactic head movement to adjoin with finite T. Note that like the morphological Merger operation discussed above, head movement in the narrow syntax can create complex terminal heads. Below, both tree and bracket notation are provided for clarity. Both finite T and be are sub-terminals contained within the M-word of T as result of syntactic head movement. As above, for exposition purposes, movement sites are given in trace notation, the internal structures of DP and VP are omitted, and some terminals are given with their phonological exponents in italics. (27) Spell Out to the morphological component, auxiliary be [TP [DP a ghost] [T’ [M [be] T[Past:-, φ:3s]] [AuxP tbe [VP tDP haunting this house]]]] =
TP 3 DP T’ 5 3 AuxP a ghost T M 2 2 be T[Past:-, φ:3s] tbe VP 6 tDP haunting this house
Later during the morphological computation to PF, after the first Vocabulary Insertion and linearization of terminals, the operation Fusion applies to head-adjoined be
91
and finite T. As discussed in the subsection immediately below, this operation applies in order to allow a single Vocabulary Item for present-tense be to insert the suppletive phonological exponent is at the single terminal [be, T] that results from Fusion. (28) Late morphological Fusion of be and T[Past:-, [M [be] * T[Past:-, 3.2.10
φ:3s]]
Æ
[M be, T[Past:-,
φ:3s]
φ:3s]]
Listed primitives 2: Vocabulary Items and Insertion
The second list of lexical primitives in this model is called the Vocabulary. Vocabulary are the list of phonological exponents for abstract terminals. Each Vocabulary Item contains a set of phonological features, along with a set of morphosyntactic features that identify a terminal where the exponent will be inserted. Vocabulary Items may also contain contextual information specifying the hierarchical or linear configuration of its target terminal. A schematic Vocabulary Item is given below: the phonological features /fon/ on the right of the arrow will be inserted into an abstract terminal X with φ features valued α as identified by the morphosyntactic features on the left of the arrow, when X is right adjacent to a terminal Y as specified by the contextual features to the right of the slash.
92
(29) Schematic Vocabulary Item [X, φ:α] Morphosyntactic features
Ù
/fon/
/
Phonological features
Y * __ Contextual features
The operation of Vocabulary Insertion supplies phonological features to all of the abstract terminals in a morphological structure. As noted above and discussed further in Chapter 5, Vocabulary Insertion may apply at multiple stages in the morphological computation to the PF interface. After the first application of Vocabulary Insertion, morphosyntactic structures are linearized. Therefore, subsequent insertion of phonological features may involve Vocabulary Items that contain linear information in their contextual features, as proposed in Chapter 6. Vocabulary Items must compete for insertion into an abstract terminal node. However, the morphosyntactic features of a Vocabulary Items can be underspecified, so that “a single phonological exponent may appear in more than one syntactico-semantic context” (Embick and Noyer to appear). In other words, competition of underspecified Vocabulary Items is among the mechanisms of allomorphy in DM theory. The phonological features of Vocabulary Item can be inserted into an abstract terminal node so long as the
93
morphosyntactic features of the Vocabulary Item contain a subset of the morphosyntactic features on the target terminal and no contradictory features. Underspecified Vocabulary Items compete for insertion at a terminal node according to a version of the familiar elsewhere principle. The most specified Vocabulary Item, with the most morphosyntactic features matching the target terminal, always wins the competition and inserts its exponent first. This prevents the insertion of exponents from lessspecified Vocabulary Items. ‘Elsewhere’ exponents are inserted by default whenever there are no more-specified Vocabulary Items that match the features of the target terminal. It is important to recognize that the syntactic terminals and the Vocabulary are separate and distinct lists in this model, and are accessed at different points in the derivation. Syntactic terminals are input to the narrow syntactic computation, and are later transferred to the morphological component by the operation Spell Out. The morphosyntactic features of terminals are fully specified, although these features may be modified by morphological operations as discussed above. Vocabulary Items contain phonological features and the instructions for their insertion into abstract terminals during the morphological
94
computation to the PF interface. These instructions include morphosyntactic features that may be underspecified, but the morphosyntactic features listed in the Vocabulary are completely distinct from, and do not modify or interact with, those listed in the syntactic terminals. To see how underspecified Vocabulary Items compete for insertion of phonological features into fully specified syntactic terminals, consider the examples from above. First, following Halle and Marantz (1993), the following Vocabulary Items compete to insert the exponent for the finite tense and agreement affix on main verbs in English. The most specified Vocabulary Item inserts the phonological features /z/ when the distinct abstract terminal for present-tense T has φ features valued 3s. The elsewhere Vocabulary Item is totally unspecified for φ features, inserting no phonological features (a zero exponent) into present-tense T with any other φ-feature values. Notice that even though the morphosyntactic features of this elsewhere Vocabulary Item contain no φ features whatsoever, the abstract syntactic terminal for present-tense T will have a full set of valued φ features.12
12
Again, it is irrelevant for present purposes whether the φ features of T are valued by Agreement in the syntax, as on the standard Minimalist theory discussed above, or inserted by a Dissociated
95
(30) Postulated Vocabulary for present-tense T [Past:-, Num:s, Pers:3]
⇔
/z/
[Past:-] (elsewhere)
⇔
Ø
For example, consider the example sentence A ghost haunts this house. As discussed above, the morphological operation Merger lowers T to affix with the main verb. Present-tense T has φ features valued 3s, so the most specified Vocabulary Item wins the competition, inserting its phonological exponent /z/ into the abstract terminal.13 This is illustrated below with a left arrow indicating the application of Vocabulary Insertion. (31) ...[M
Vhaunt
[T[Past:-,
φ:3s]]]...
Í
/z/
Consider a slightly different sentence, Ghosts haunt this house, where present-tense T has a full set of φ features valued 3p. Now the underspecified elsewhere Vocabulary Item inserts its zero exponent by default. (32) ...[M
Vhaunt
[T[Past:-,
φ:3p]]]...
Í
/Ø/
Next, consider the following postulated Vocabulary for present-tense be in English. Notice that these Vocabulary Items are suppletive, inserting a single exponent into an single abstract terminal that contains the features of both
feature-copying rule in the morphological component, as on the DM theory and as proposed in Chapter 4. 13 On this analysis, an English-specific rule will later devoice -z in the phonological component.
96
be and present-tense T. The most specified Vocabulary Item for present-tense be inserts the phonological features /æm/ when T’s φ features are valued 1s; an equally specified Vocabulary Item inserts the phonological features /Iz/ when T’s φ features are valued 3s; and an elsewhere Vocabulary Item inserts the phonological features /ar/ when T’s φ features have any other values. (33) Postulated Vocabulary for present-tense be [be, Past:-, Num:s, Pers:1]
⇔
/æm/
[be, Past:-, Num:s, Pers:3]
⇔
/Iz/
[be, Past:-] (elsewhere)
⇔
/ar/
As discussed above and in Chapter 5, this dissertation follows the late Fusion analysis of Kandybowicz (to appear). For example, consider the example sentence A ghost is haunting this house. As above, the auxiliary terminal be is raised to finite T by head movement in the narrow syntax, so that be and T are distinct terminals contained in the complex M-word terminal T after Spell Out. At some stage during the morphological computation, the inventory of Vocabulary are searched for the first application of Vocabulary Insertion, with subsequent linearization of the morphosyntactic terminals. This search retrieves the
97
Vocabulary Items for present-tense be, but none of these can be inserted into the distinct terminals for be and T. (34) ...[M [be] * T[Past:-,
φ:3s]]...
Therefore, late application of Fusion combines the features of be and T into a single terminal, ‘repairing’ the morphosyntactic structure so that these suppletive Vocabulary Items can insert their phonological exponents into the newly created abstract terminal. Since T’s φ features are valued 3s, the Vocabulary Item specified with 3s features inserts the exponent is. (35) ... [M be, T[Past:-,
φ:3s]]...
Í
/Iz/
Consider a slightly different sentence, Ghosts are haunting this house, where present-tense T has a full set of φ features valued 3p. Now the underspecified elsewhere Vocabulary Item inserts the exponent are by default. (36) ... [M be, T[Past:-,
3.2.11
φ:3p]]...
Í
/ar/
Listed primitives 3: The Encyclopedia
Finally, the third list of lexical primitives is called the Encyclopedia. According to Embick and Noyer (to appear), the Encyclopedia is accessed “subsequent to the output of LF/PF,” and contains information about the interpretation of Roots and idiom structures. For example, an Encyclopedia entry for the Root √GHOST might contain detailed information 98
about ghosts: that they are the disembodied spirits of dead people, that they haunt houses and other locations, that they leak ectoplasm, etc. Because of its limited scope, this dissertation will not discuss the Encyclopedia any further, focusing instead on the morphosyntactic properties of syntactic terminals and Vocabulary Items. However, I would like to at least speculate here that the Encyclopedia might be a repository for social meaning. Thus, a future DM theory might explore the conjecture that the social interpretation of Labovian variants is listed in the Encyclopedia.
3.3
Minimalism, DM, and Labovian Variation
This chapter has outlined the Minimalist, DM-theoretical model that will used to analyze cases of morphosyntactic variation in the three chapters that follow. The DM theory reviewed above provides deterministic mechanisms for the familiar phenomena of allomorphy, affixation, and suppletion. However, the present DM theory lacks mechanisms that can account for Labovian variation, where variant forms appear probabilistically in the same morphosyntactic environment as discussed in Chapter 2. Moreover, in Minimalist theories including DM, the application of operations in the narrow syntactic computation (Merge,
99
Move, and possibly Agree) is strictly determined by the morphosyntactic features of terminal items selected into the Lexical Array. The Minimalist program prohibits nondeterministic syntactic operations in principle, so observed Labovian variation in the narrow syntax could be considered evidence against Minimalist theories.14 In closing, then, we might ask why this particular theoretical model is a good one with which to address the problem of Labovian variation in morphosyntax. The answer, I suggest, has two parts. First, a Minimalist theory of syntax makes clear predictions about where we should expect to find mechanisms of morphosyntactic variation. As noted immediately above, there can be no variation in the operations of narrow syntax on a Minimalist theory. And as we have seen in Chapter 2, Labovian variation does not affect semantic meaning, so mechanisms of variation cannot be located in the post-Spell-Out computation to the LF interface. Therefore, a Minimalist theoretical model predicts that mechanisms of variation should be limited to two locations: variation can arise in the features of syntactic terminals, 14
However, apparent instances of Labovian variation in the narrow syntax might instead be considered evidence that mechanisms of the observed variation are in fact located elsewhere, in the syntactic terminals or the morphological component. In other words, Labovian variation in morphosyntax might be a kind of empirical diagnostic for the kinds of theoretical mechanisms proposed in this dissertation.
100
as proposed by Adger and Smith (2005, and also Adger 2006) and discussed in Chapters 4 and 5, or in the objects and operations of the morphological computation to the PF interface, as proposed in Chapters 5 and 6. Indeed, these are exactly the predicted locations for mechanisms of parametric variation in a Minimalist theory. If the reader will indulge me, I would like to reproduce a lengthy quote from Chomsky (1993: 3), who argues that parametric variation must be determined by what is "visible" to the child acquiring language [...]. It is not surprising, then, to find a degree of variation in the PF component, and in aspects of the lexicon: Saussurean arbitrariness (association of concepts with phonological matrices), properties of grammatical formatives (inflection, etc.) and readily detectable properties that hold of lexical items generally (e.g., the head parameter). The same is clearly true of Labovian variation, which often differs between (populations of) individuals. For example, my neighbors in Brooklyn do not use leveled weren’t. Evidently, then, Labovian variation must be acquired on the basis of exposure to linguistic environmental input. This entails that the mechanisms of Labovian and parametric variation overlap considerably, as suggested in Chapter 1 and 2. Chomsky continues (1993: 3), Variation would be
in the overt syntax or LF component more problematic, since evidence [for
101
acquisition] could only be quite indirect. A narrow conjecture is that there is no such variation...variation is limited to nonsubstantive parts of the lexicon and general properties of lexical items. Let us tentatively adopt that assumption...as another element of the minimalist program. I suggest we adopt the parallel Minimalist assumption that mechanisms of Labovian variation are also limited to “properties of lexical items” and “properties of grammatical formatives” like inflection “in the PF component.” This is the second part of the answer to the question posed above: DM allows these Minimalist predictions to be formulated in an articulated theory of lexical properties and the PF-interface component. The DM theoretical framework holds the promise of implementing a unified analysis of the mechanisms of both parametric and Labovian variation in the features of syntactic terminals, Vocabulary Items, and the operations of the morphological component.
102
Chapter 4: Weak Expletive It on Smith Island
4.
Introduction
This chapter examines the morphosyntactic variant weak expletive it (WEIT) and its usage on Smith Island. Although WEIT is a commonly documented feature of several modern and historical English varieties, there have been no variationist or theoretical studies of WEIT and its agreement properties. This chapter provides both perspectives on WEIT, and concludes that in this case the mechanisms of Labovian variation are located in the features of syntactic terminals. The chapter is structured as follows. Section 4.1 outlines some known facts about expletive constructions and agreement in English. Section 4.2 presents a study of WEIT on Smith Island using variationist and other methods.1 In addition to the quantitative observational data, elicited judgments confirm that verbal agreement is categorically third singular (3s) with WEIT. Apparent-time analysis,
1
The research reported here and in Parrott (2002) was financed in part with State Funds from the Maryland Historical Trust, an agency of the Department of Housing and Community Development of the State of Maryland. However, the contents of this chapter and the dissertation as whole do not necessarily reflect the views or policies of the Maryland Historical Trust or the Department of Housing and Community Development.
103
supported by preliminary real-time data, shows that the WEIT variant is part of the socially motivated concentration process on Smith Island (Parrott 2002). Section 4.3 considers theoretical issues. After a review of a current Minimalist analysis of expletives and agreement (Chomsky 2000a), I point out a mismatch paradox: WEIT should either leave the associate DP’s uninterpretable Case feature unchecked, causing a crash at the interface, or allow plural associate agreement, contrary to fact. Section 4.4 proposes an analysis of WEIT that resolves the apparently paradoxical Case mismatch. Adopting the theoretical framework of Distributed Morphology (DM, Halle and Marantz 1993, Embick and Noyer to appear), I follow a radical approach that denies the existence of uninterpretable syntactic Case features; language-specific case morphology is the result of feature-insertion rules that apply after the narrow syntax (McFadden 2004). This DM analysis is next extended to agreement. I propose morphological agreement-feature insertion rules for English that can account for both categorical 3s agreement with it and rightward associate agreement with there. Finally, Section 4.5 returns to the dissertation’s overall goal of identifying the mechanisms of Labovian variation in a Minimalist-DM framework. In this case, the mechanism of
104
WEIT variation are located in the choice of terminals selected into the pre-syntactic Lexical Array.
4.1
Expletives and Agreement in English
To begin, this section outlines some facts about expletives and their agreement properties in varieties of English.
4.1.1
Two expletive forms and agreement
An expletive is an apparently nominal form that functions as a structural subject, but has no thematic role (such as agent or patient, etc.) in the argument-structure semantics of the sentence. Expletives are believed to satisfy a parameterized syntactic requirement, called the Extended Projection Principle (EPP) (Chomsky 1981), that (English) sentences must have a phrase in the subject position. In English, there are two morphologically distinct expletive subjects. Expletive it is homophonous with the third-person singular (3s) gender-neutral pronoun it, while the ‘weak’ expletive there is homophonous with the locative form there. The two expletives have different distributions and agreement properties. As shown in the examples below, expletive it can appear as the subject of weather predicates (e.g. rain), and as the subject of a raising predicate with a finite
105
complement clause. Raising predicates are verbs (e.g. appear) and adjectives (e.g. unlikely) that do not assign a thematic (theta = θ) role to their DP subject. The EPP requires that the subject position of a raising predicate must be filled, either by an expletive or by movement of a DP from the complement clause; but this movement is impossible when the complement clause is finite. Verbal agreement morphology always takes the 3s form in an expletive-it sentence, presumably reflecting ordinary subject-verb agreement with the 3s subject it. (1)
Weather predicates It is (*are) stormy tonight.
(2)
Raising predicates w/finite complements a.
It seems (*seem) that these isles are haunted. Cf. * These isles seem(s) that are haunted.
b.
It is (*are) likely that wraiths are among us. Cf. * Wraiths are (is) likely that are among us.
At first glance, the so-called ‘weak’ expletive there seems to have a wider distribution than it. The examples below give a non-exhaustive sampling of the syntactic environments where expletive there can appear, including the subject of a copular existential sentence, the subject of a raising predicate with a non-finite complement clause, the subject of a passive sentence, and the subject of an unaccusative verb. Like passive verbs and raising
106
predicates, unaccusative verbs (e.g. arrive) do not assign any θ role to their subject positions, which the EPP compels to be filled. This can be accomplished either with an expletive or by movement of a post-verbal thematic argument, the so-called ‘associate’ DP; this movement is possible only when the lower clause is not finite. The weakness of expletive there is due to its characteristic property of inducing “apparently anomalous rightward agreement” (Chomsky 1995, Chapter 1 with Lasnik: 66): verbs in there-expletive constructions can agree in number with the associate DP. (3)
Copular existentials There are poltergeists in this house. Cf. Poltergeists are in this house.
(4)
(5)
Raising predicates w/non-finite complements a.
There seem to be will-o’-the-wisps everywhere. Cf. Will-o’-the-wisps seem to be everywhere.
b.
There are likely to be shades in the cemetery. Cf. Shades are likely to be in the cemetery.
Passives There were phantoms seen on this beach. Cf. Phantoms were seen on this beach.
(6)
Unaccusatives Every year, there appear on this bridge malevolent apparitions. Cf. Every year, malevolent apparitions appear on this bridge.
107
For those English varieties that do not use weak expletive it,2 the two expletive forms seem to occur in complementary distribution: one form does not appear in the other’s syntactic environment. (7)
Weather predicates * There is stormy tonight.
(8)
Raising predicates w/finite complements a. * There seems that these isles are haunted. b. * There is likely that wraiths are among us.
(9)
Copular existentials * It are poltergeists in this house.
(10) Raising predicates w/non-finite complements a. * It seem to be will-o’-the-wisps everywhere. b. * It are likely to be shades in the cemetery. (11) Passives * It were phantoms seen on this beach. (12) Unaccusatives * Every year, it appear on this bridge malevolent apparitions.
4.1.2
The Definiteness Effect and agreement
2
I avoid the term “standard English” here and throughout. The ideology of 'standard' languages is worthy of sociolinguistic investigation; see Lippi-Green (1997) for a noteworthy example. But this social notion should not be the object of biolinguistic inquiry.
108
Weak expletive there also induces a restriction such that the associate DP may not be semantically definite; this fact is described as the ‘definiteness effect.’3 (13) Copular existentials * There are the poltergeists in this house. (14) Raising predicates w/non-finite complements a. * There seem to be the will-o’-the-wisps everywhere. b. * There are likely to be the shades in the cemetery. (15) Passives * There were the phantoms seen on this beach. (16) Unaccusatives * Every year, there appear on this bridge the malevolent apparitions. Notice that the definiteness effect means we cannot observe whether associate agreement involves person or just number. English pronouns are morphologically distinguished by person and number (and apparently case, the topic of Chapter 6). (17) English pronouns by person and number
1st
Singular
Plural
I/me
we/us
3
As Kleanthes Grohmann points out (p.c.), it may be proposed that the definiteness effect is due to the associate being a NP rather than a DP. Theories of the definiteness effect will not be considered here; instead, I merely use it as a diagnostic.
109
2nd
you
you4
3rd
she/her
they/them
All other DPs are third person, and are morphologically distinguished only by number: (18) English DPs by number
3rd
Singular
Plural
ghost
ghosts
But pronouns are inherently definite, and thus pronouns are independently ruled out in weak-expletive constructions due to the definiteness effect. (19) Copular existentials * There are we in this house. (1p) Cf. We are in this house. (20) Raising predicates w/non-finite complements a. * There seem to be y’all everywhere. (2p) Cf. Y’all seem to be everywhere. b. * There are likely to be they in the cemetery. (3p) Cf. They are likely to be they in the cemetery. (21) Passives * There was I seen on this beach. (1s) Cf. I was seen on this beach. (22) Unaccusatives * Every year, there appear on this bridge you. (2s) Cf. Every year, you appear on this bridge. 4
Some English varieties have a distinct 2p pronoun, for example y'all, yous, yins, and possibly you guys. Because these various pronoun forms are not evidently relevant to the matters under investigation here, only the homophonous 2p form you appears henceforth.
110
4.1.3
Associate-agreement variation with expletive there
Variable 3s agreement with plural associates in weak expletive there constructions has been documented in many varieties of English, as exemplified below (attested, from Rupp 2005). (23) a.
There’s so many useless degrees you can do now.
b.
Well, there is children, but not babies.
c.
There was pits everywhere.
Research in theoretical syntax has mostly ignored agreement variation in weak expletive there constructions.5 Chomsky’s (1995: 384) footnote on the matter dismisses such variation as superficial and therefore irrelevant for syntactic theory. Chomsky claims that the variation does not occur with interrogatives (*Is there three books on the table?) or negation (*There isn’t any books on the table), and implies that variation cannot occur with other forms of be (*There was three books on the table). However, these claims are not supported by experimental or observational evidence. Chomsky concludes that there’s is a “frozen option,” but he does not specify any mechanisms by which this form might be derived. 5
An exception is Sobin’s (1997) ‘virus theory’ of associate agreement with weak expletives, refuted by Schütze (1999). I criticize virus theory in Chapter 6.
111
Most of what is known about associate agreement variation with there expletives has come from variationist research. On this approach, however, there expletive constructions have typically been analyzed as one of several linguistic factors potentially conditioning variable 3s verbal agreement in general. As Meechan and Foley observe in their classic study (1994: 64), Virtually all previous quantitative studies of variable concord in existentials in English are subsumed within larger studies of variable concord in all verbal environments, a feature of non-standard dialects. Joined recently by a few others (e.g., Henry 2002, Cornips and Corrigan 2005, and Rupp 2005), Meechan and Foley argue that agreement variation in weak expletive there constructions should instead be analyzed separately from agreement variation with non-expletive subjects. All of these researchers point to the broad theoretical consensus that there-expletive constructions have a different syntactic structure than other kinds of sentences, and that this structural difference causes the rightward associate agreement unique to weak there expletives. But even from a variationist perspective, it is clear that “existential constructions exhibit different properties than other types of concord” (Meechan and Foley 1994: 64).
112
First of all, variable usage of 3s agreement in weak expletive there constructions can be completely dissociated in apparent time from usage of variable 3s agreement with non-expletive subjects. For example, consider Hay and Schreier’s (2004) study of variable agreement forms of be in New Zealand English over the past 150 years. Hay and Schreier found that usage of 3s agreement forms with nonexpletive plural subjects declined to near zero in the 20th century. There was a concomitant decline in usage of 3s agreement forms with plural associates in there-expletive constructions, to about 50% usage at the turn of the century. But while usage of 3s forms with plural nonexpletive subjects remains at near zero, usage of 3s forms with plural associates in there-expletive constructions has independently rebounded to about 80% in modern New Zealand English. Furthermore, linguistic factors conditioning 3s agreement variation with weak there expletives are different from those that condition agreement variation with non-expletive subjects. For example, Rupp (2005) provides a review of this topic and a new empirical study in a Midlands variety of British English using variationist observational methods augmented with various experimental techniques for eliciting acceptability judgments. Rupp
113
found that 3s forms with plural associates were less likely to be used in there-expletive constructions containing contracted and full forms of clausal negation (e.g., There isn’t/is many boats). 3s agreement was also less likely in interrogative there-expletive constructions with plural associates (e.g., Is there many boats?). These linguistic factors were not found to have a significant effect on agreement variation with non-expletive subjects. Finally, variationist conflation of agreement variation in there-expletive constructions with agreement variation in general has obscured an important question: does variable 3s agreement in weak expletive there constructions ever occur in English varieties lacking concurrent 3s agreement variation with non-expletive subjects? Currently, the answer to this question remains unknown. The Midlands English dialects examined by Rupp (2005) exhibit 3s agreement variation with non-expletive subjects. For their study, Meechan and Foley (1994) combined data from African-American enclave communities in Nova Scotia (Poplack and Tagliamonte 1991) with data from Canadian English speakers of mostly British descent in Ottawa, Ontario. The African-American Nova Scotians have agreement variation with non-expletive subjects (Tagliamonte and Smith 2000). Citing DeWolf (1992), Meechan
114
and Foley claim that “quintessential standard” (Chambers 1991) Canadian English “is not a dialect that generally exhibits a lack of concord between subject and verb” (1994: 70). Their Ottawa speakers do nonetheless exhibit 3s variation with plural associates in there-expletive constructions, suggesting an affirmative answer to the question posed above. Unfortunately, however, Meechan and Foley’s quantitative analyses do not separate the two varieties, so their study does not conclusively settle the matter. The question of whether agreement variation in thereexpletive constructions can occur independently of general agreement variation must be addressed by future variationist studies. A definitive answer will be particularly important for deciding upon a theoretical analysis, and so variable associate agreement in weakexpletive there constructions clearly warrants much more cooperative variationist and theoretical research of the kind advocated by Meechan and Foley, Rupp, Henry, Cornips and Corrigan, and others. It is not my intention to carry out such investigations here; this chapter focuses more narrowly on the variant weak-expletive form it and its agreement properties. However, in the last section I offer
115
some tentative speculation about the mechanisms of variable 3s agreement with weak there expletives.
4.1.4
Summary
For present purposes, it suffices to emphasize the difference between the expletive forms it and there. Rightward number agreement with a plural associate is at least possible with the weak-expletive form there, if variable for some individuals and populations. However, agreement forms are categorically 3s with the expletive form it. This is summarized below. (24) Two English expletive forms and agreement properties
4.2
a.
there = variable number agreement with associate DP
b.
it = categorical 3s agreement
WEIT: Variation and Change in Progress on Smith Island
This section examines the morphosyntactic variant weak expletive it (hereafter referred to with the acronym WEIT) and its usage on Smith Island. I claim that agreement is categorically 3s when the form it is used as a weak expletive. Furthermore, as reported in Parrott (2002), WEIT is evidently part of the socially motivated process of concentration on Smith Island.
116
4.2.1
WEIT in English
Variable usage of WEIT has been documented in several varieties of English (e.g. Wolfram and Schilling-Estes 1998). For a well-known example, WEIT is a pervasive feature of African-American English (AAE) varieties (e.g. Martin and Wolfram 1998, Green 2002).6 (25) It was about three times that Chauncey and Rip had to tell me where to go on certain plays.7
4.2.2
WEIT on Smith Island
On Smith Island, the variant form it appears in all the syntactic environments of weak expletive there. All examples are attested in sociolinguistic interviews (from Parrott 2002). (26) Copular existentials In winter, it’s nothing to do. (27) Raising predicates w/non-finite complements It just happened to be a EMT on this part of the Island....
6
Use of they as a variant form of the weak expletive has also been documented in AAE and in other varieties such as Appalachian English (Tortora 2006). 7 Attested, Rasheed Wallace speaking to the Oregonian, collected by the author 3/3/2004 (oregonlive.com). Wallace is talking about learning a new system after being traded to the Detroit Pistons from the Portland Trail Blazers. In this quote, the forward remarks that there were approximately three times during the game when Detroit guards Chauncey Billups and Richard “Rip” Hamilton had to tell Wallace where to go on the court during particular set plays.
117
(28) Passives And it was sharks seen down there that day.8 (29) Unaccusatives Then you go straight on down, and it comes this white house here....9 WEIT is unattested in the environment of the homophonous locative form there, showing that the WEIT variant is limited to the syntactic weak-expletive environment. (30) a.
It’s only a handful of ‘em down there.
b. * It’s only a handful of ‘em down it. (31) a.
...it was [a] cat in there...
b. * ...it was [a] cat in it... WEIT is also unattested with a definite associate DP, further evidence that WEIT is really a variant form of the weak expletive. (32) Copular existentials * In summer, it’s the big barbecue. (33) Raising predicates w/non-finite complements * It just happened to be the doctor on the island. 8
Note that in this example, the speaker is using a passivized expletive construction to assert that unspecified people saw sharks in the location and time in question. In other words, the example sentence could be paraphrased as People saw sharks down there that day, or as Sharks were seen down there that day. Nothing in the discourse suggests that this sentence is actually a cleft; the speaker is clearly not, for example, contrasting the sharks with some other sea creature, like It was sharks, not whales, (that were) seen down there that day. 9 Note that in this example, this is being used as an indefinite--as in, for example, There's this book that I want to consult.
118
(34) Passives * And it was that shark seen down there. (35) Unaccusatives * Then you go straight on down, and it comes John’s house. To confirm that the non-attestation of definite DPs with WEIT is due to the definiteness effect, I elicited an acceptability judgment from one Smith Islander during a sociolinguistic interview conducted in 2000. This informant, a 39-year-old female, rejected WEIT sentences with a definite associate DP. In the transcript extracts below, this author is fieldworker JP and the informant is F39. (36)
Judgment on definiteness effect with WEIT (F39) JP:
What about, what about these. Say it’s a big crab in the pot? F39: Yeah okay. JP: What’s about it’s the big crab in the pot? F39: It’s the big crab? It’s the big crab, well JP: I want it as the, there is the big crab. F39: Uh huh. They wouldn’t say it’s the big crab in the pot. It’s a big crab maybe.
4.2.3
WEIT and categorical 3s agreement
119
Consistent with above-mentioned findings for other English varieties, number agreement with plural associates in weakexpletive there constructions is variable on Smith Island.10 (37) a. b.
There are two older than me and one younger. I believe there’s spirits though.
However, unlike weak expletive there, WEIT occurs with categorical (i.e. non-variable) 3s agreement. Rightward agreement with a plural associate DP is completely unattested with WEIT on Smith Island (see below for quantitative figures). (38) a.
It’s no...separate burial plots on Tylerton.
b. * It’re no separate burial plots on Tylerton. Categorical 3s agreement with WEIT is not limited to contracted agreement forms. (39) a.
Is it any funny things you remember...?
b. * Are it any funny things you remember? Parrott (2000a) analyzed 46 weak-expletive constructions with plural associate DPs,11 extracted from 1983 interviews with 12 Smith Islanders. 42 out of 46 verbal elements had the 3s form, while only four had the plural form. Of the 3s forms, half occurred with there and
10
This kind of variation was not quantitatively analyzed in Parrott (2002), nor is it here. For a review of previous research along with original quantitative data and syntactic analysis, see Rupp (2005). 11 Without contracted negation, see Chapter 5.
120
half with WEIT; all four of the plural forms occurred with there. Weak-expletive form WEIT there
Plural verbal form 0 4
3s verbal form 21 21
Table 1. Verbal agreement in weak-expletive sentences with plural associate DPs (adapted from Parrott 2000a) Parrott (2000a) also analyzed past-tense be (was/were) data from Schilling-Estes (2000), which were extracted from interviews with 24 Smith Islanders. As above, only instances of weak-expletive sentences with plural associate DPs and past-tense be were counted.12 50 of 60 total pasttense be forms were singular, while 10 were plural. All 10 plural forms occurred with there. Weak-expletive form WEIT there
Plural were 0 10
Singular was 28 22
Table 2. Past-tense be agreement in weak-expletive sentences with plural associate DPs, (adapted from Parrott 2000a) Again, to confirm the non-attestation of rightward plural associate agreement with WEIT, I elicited acceptability judgments from two Smith Islanders during sociolinguistic interviews conducted in 2000. The first informant was the same quoted above. She immediately rejected the full plural form are, but wondered about the 12
Again, without contracted negation, see Chapter 5.
121
acceptability of contracted -‘re for older generations. The informant makes an analogy with expletive it in a weather predicate (quite intriguing considering that she has no training in linguistics), before finally concluding that contracted -‘re is unacceptable with WEIT. Below, FW is another fieldworker, and ‘xxx’ indicates sequences that were unintelligible to the transcriber. (40)
Judgments on plural agreement with WEIT (F39) JP: F39: JP: F39:
What about it are a lot of crabs in the pot? I don’t know. No. I don’t think it would go. What about it’re a lot of crabs? I don’t know, it might be all right. I am trying to think, some of the older generation may say that. JP: It’re a lot of crabs in the pots? F39: I don’t want to, see now, I am trying to think too much. I think they’d lean more to it’s a lot, it’s a lot of crabs in the pot, more so than it’re, it are. I don’t know. It might sound like it’re storm. Like it’re stormy out here. No. xxx FW: Could that be like it, it were... F39: No no. I think it’s, an it’s, I think it would be it it, like it’s stormy out here. I don’t think so. No I don’t think so.
4.2.4
A Methodological digression
Here I would like to digress briefly in order to call attention to a methodological issue that arises when investigating variation in morphosyntax. The second informant was a 43-year-old male, and his interview vividly
122
illustrates the difficulties of eliciting acceptability judgments during sociolinguistic fieldwork. In response to test sentences, this informant immediately volunteered alternatives using Smith Island’s ironic “backwards talk” (see Schilling-Estes 2004 for more discussion of this interview and the sociolinguistics of backwards talk on Smith Island). The alternatives express roughly the same semantic meanings as the test sentences (i.e., that many crabs exist), but do not contain the relevant expletive and agreement forms. Below, the informant is M43 and ‘[...]’ indicates an elided section of the transcript. (41)
1st attempted judgment elicitation for WEIT (M43) JP:
And um, I am interested in one particular thing. Which is this. I know on Smith Island, where some other dialects of English say like, there’s a lot of crabs in the bay? M43: Yeah. JP: And Smith Islanders can say, it’s a lot of crabs in the bay, right? M43: Yeah. It ain’t neither crab. JP: Huh? M43: Alright, they’d say, it ain’t neither crab. JP: Mm-hm. Oh, that’s right. With the backwards statement. M43: Yep. [...] M43: Or they would say all you want. Or they would say all you want. Somebody got on the radio and said yup, all you want xxx At this point, the fieldworker attempted to redirect the informant toward the test sentences, but the informant
123
volunteered another alternative and asserted that it is possible to say anything in the local dialect. (42)
2nd attempted judgment elicitation for WEIT (M43) JP:
M43: JP: M43: JP: M43: JP: M43:
So it’s a lot of crab in the bay, um, if you, you were gonna stress it, that is suppose someone said, you know, it’s not a lot of crabs in the bay this year, xxx oh no it is a lot of crabs in the bay this year... Yeah, I would say no, it aint narry one. Well, but, could you, is it possible to say? Oh yeah. Yeah? It’s possible to say all those things. Smith Island dialect, be hard to find something that you couldn’t say. Well, we’ll see about that. Okay.
Finally, the fieldworker used contrastive phonological stress (indicated with CAPITALS) to highlight the relevant plural and 3s agreement forms in a WEIT construction. When presented in this way, the informant rejected the plural agreement form are with WEIT. (43)
1st judgments on plural agreement with WEIT (M43) JP:
So, so it’s possible for a Smith Islander to say it IS a lot of crabs in the bay? M43: Oh yeah, sure. JP: What about it ARE a lot of crabs in the bay this year? M43: It ARE a lot of crabs? No, they wouldn’t say it like that. JP: What about it are a lot of crabs in the bay this year? M43: It are a lot of crabs in the bay. No, they don’t say it like that.
124
At this point, elicitation attempts ceased and other fieldworkers spoke to the informant about various topics. Toward the end of the interview, fieldworker JP requested additional judgments from the informant. But the informant resisted, stating his view that backwards talk is the most important feature of the Smith Island dialect. (44)
3rd attempted elicitation attempt (M43) M43: But I think you’re way off track with what you’re doing right now. I don’t know, I mean, I don’t know, are you writing a paper or working on a thesis or what? Is this speech pattern you recognize from somewhere else I mean? JP: I’m trying to figure out what the speech pattern is. M43: Yeah. Cause you’re way off base I think. JP: You think? M43: Yeah. I mean I don’t think comparing these two sentences are gonna get you any closer to, uh, understanding the Smith Island dialect at all. They don’t talk like that. I mean not even close, uh, I would, reconsider whatever. I would say well, if you’re really gonna figure it out you’d really need to spend more time, listening to, what they say and how they say it. Um, because it’s very. I don’t want to say it’s unique cause I haven’t been to that many places, I don’t know maybe somewhere they talk, do exactly the same thing but. It’s unique for the places I’ve been, I don’t know about you. I’ve never been anywhere else where people talk backwards. JP: Well see I’m, she’s the one interested in the backwards actually. M43: Yeah. JP: I’m not so interested in the backwards myself. M43: But that’s the way they talk. I mean, you know, its not a question of being interested in the backwards or not interested in the backwards. That’s how they talk. I mean, you know, you have to be, if you’re gonna do Smith Island dialect you have to be interested in the backwards, uh,
125
or otherwise you’re never gonna, you’re never gonna get a handle on it. But you know, you’re the linguistics. I don’t want to be telling you what to do. Fieldworker JP persisted (with assistance from another fieldworker), and presented the test sentences as pairs. The informant finally rejected the plural agreement form are and also the past-tense plural form were, this time with WEIT in interrogative sentences. (45)
2nd judgments on plural agreement with WEIT (M43) JP: FW: M43: JP:
M43: JP: M43: JP: M43:
JP: M43: JP: M43: JP: M43: JP: M43: JP: CP:
Do you want to hear some more anyway? Do more anyway? Why not, you wrote em up? He wrote ‘em up, so go ahead. These are questions, well I mean this is anything, you have two pairs of sentences, of the questions, so just tell me which one sounds the more natural. So the first one is, so, are it a lot of crabs in the bay this year? Say that one again. Are it a lot of crabs in the bay this year? No, they wouldn’t say that. Is it a lot of crabs in the bay this year? Yeah, it would be the second one. Well, they actually talk frontwards for lack of a better xxx, they’re not that horribly off in their grammar, uh, sometimes they are, but you know. How about, um, it were a lot of crabs in the bay this year? No, they wouldn’t say that. It was a lot of crabs in the bay this year? It was a lot of crabs in, How about were it a lot of crabs in the bay last year? Were it? Were it a lot of crabs in the bay last year? No they wouldn’t say that no Was it a lot of crabs in the bay last year? They would use the second one.
126
To conclude this digression, I would like to reiterate that a variety of empirical methods should be utilized for inquiry into Labovian variation in morphosyntax. From these excerpts, it is clear that asking untrained informants for acceptability judgments can be problematic in a fieldwork setting. Informants may not understand the requested tasks, or they may have opinions about their local dialect that are not relevant to the morphosyntactic structures under investigation. In this case, acceptability judgments were eventually obtained from the informant after some prodding. Although these judgments might otherwise be regarded as less reliable because of the way they were elicited, here they accord with the non-elicited observational findings as well as the judgments of another informant.
4.2.5
Concentration of WEIT in apparent time
This section reports on an apparent-time analysis of variable WEIT usage on Smith Island that was originally presented in Parrott (2002). The data were extracted from transcripts of sociolinguistic interviews with 17 Smith Islanders. These interviews were conducted in 1983 by Rebecca Setliff and another interviewer who was an island resident. All and only weak expletives were extracted and analyzed. Potential
127
weak expletive tokens were collected by using the ‘find’ function of word processor to search for all instances of there and it in a transcript. Non-weak-expletive instances of both there and it were discarded. These included locative there (e.g., I was there), pronominal it (e.g., I love it), and non-weak expletive it (e.g., It’s raining). Instances of it that were ambiguous between a pronominal and a weak-expletive interpretation were also discarded (e.g., “...it was a lot to learn....”). Repeated expletives, expletives isolated in sentence fragments, and self corrected expletives were counted only when both the expletive and an agreeing verb occurred. For example, both expletives in the following sentence would have been counted: It’s, it’s a lot of crabs in the pot. However, only the second expletive would have been counted in this sentence: It, it’s a lot of crabs in the pot.13 The data were analyzed according to an individual’s age, utilizing the apparent time method as discussed in Chapter 2 (Bailey et al. 1992, Bailey 2002). In order to facilitate comparison, the 17 speakers were divided into 13
In hindsight, this was not the correct decision. Immediate repetitions should have been discarded from the count so that they would not artificially inflate the usage frequency of WEIT. Parrott (2002) did not code or count the number of immediate repetitions, so it is not possible to determine whether the few instances of repetition in those data significantly affected the apparent-time results. Future quantitative research on WEIT will certainly discard immediate repetitions.
128
the same generation groups used by Schilling-Estes and Wolfram (1999) and Schilling-Estes (2000). The generation groups are repeated here from Chapter 2: (46) Generation Groups Generation I: b. 1899 - 1916 (age 84 - 67 at time of interviews, 1983) Generation II: b. 1944 - 1961 (age 39 - 22 at time of interviews, 1983) Generation III: b. 1966 - 1971 (age 17 - 12 at time of interviews, 1983) The table below displays, for each generation group, the total number of weak expletives, the number of WEIT tokens, and the percentage of WEIT usage out of all weak expletives. These figures are totals for all speakers in the generation group. WEIT accounts for over 70% of the 446 weak expletives in this sample. The results show that individuals in Generation I used the WEIT variant for 54% of their weak expletives. The percentage of WEIT usage jumps to 77.3% for Generation II, and remains only slightly higher for Generation III at 78.4%.
129
Generation Group Generation I b. 1899-1916 (4 persons) Generation II b. 1944-1961 (6 persons) Generation III b. 1966-1971 (7 persons) Totals (17 persons)
#WEIT/#Weak-expletive environs
%WEIT
73 / 135
54%
133 / 172
77.3%
109 / 139
78.4%
315 / 446
70.6%
Table 3. Numbers and WEIT percentages (adapted from Parrott 2002) Parrott also performed a statistical variable-rule (VARBRUL) analysis (Fasold 1993), which is included here for the sake of completeness. The VARBRUL analysis of these data shows that the correlation between an individual speaker’s age and their use of WEIT is statistically significant, and that there is a .72 overall probability of WEIT usage over all the individuals in the sample. The WEIT variant is disfavored for individuals from Generation I, with a .32 probability of usage. Mirroring the usage percentages above, the probability of WEIT jumps to .57 for individuals from Generation II and remains slightly higher at .59 for Generation III.
130
Application = WEIT Non-application = there Factor group: Generation Input probability
= .72
Generation Generation Generation Generation
= .32 = .57 = .59
group I II III
Chi-square per cell Total Chi-square
= .000 = .000
Table 4. VARBRUL results (adapted from Parrott 2002) On Smith Island, then, WEIT is used more commonly than there as a weak expletive. Generations II and III account for the majority of WEIT usage, while Generation I only uses WEIT about half the time. The VARBRUL analysis confirms this pattern and its statistical significance. Following the apparent time method, these results indicate that usage of WEIT is increasing. We may conclude that there is a morphosyntactic change in progress on Smith Island, such that WEIT replaces there as a weak expletive. This change is nearly complete for three of the Generation II and III speakers in this sample, who use WEIT for over 90% of their weak expletives. These findings for WEIT are consistent with those of Schilling-Estes and Wolfram (1999) discussed in Chapter 2, adding to the mounting evidence that the Smith Island dialect is undergoing a process of concentration as it dies.
131
A characteristic of the concentration process, according to Schilling-Estes and Wolfram, is that although the pace may be accelerated, linguistic changes progress in an otherwise normal fashion. The pattern of rapid change under concentration “approximates the S-curve that characterizes the diffusion of new language forms in healthy language varieties” (Schilling-Estes and Wolfram 1999: 513). The increase in WEIT usage over apparent time is proceeding as predicted under the concentration process. The graph below shows an approximately S-shaped increase. The steepest slope of the S-shaped curve is the 23% jump in WEIT frequency between Generation I and II. Usage of WEIT continues to increase between Generation II and III, but only very slightly at 1%.
Usage Percentage
100 80 60 40 20 0
Gen. I
Gen. II
Gen. III
Generation Group
Figure 1. Percent WEIT usage on Smith Island, in apparent time (adapted from Parrott 2002)
132
WEIT’s rate of increase is similar to that of the two phonological changes studied by Schilling-Estes and Wolfram (1999), where usage of the distinctive Smith Island variant increases sharply between Generations I and II, followed by only a slight increase between Generations II and III. Below, I repeat their graph from Chapter 2 for ease of comparison. It shows that usage of the glide-fronted/raised variant of /aw/ jumps by 50% from Generation I to Generation II, but only increases by 1% from Generation II to III. Usage of the raised/backed variant of /ay/ increases by 16% and then 4%.
100
Usage Percentage
90 80 70
raised /ay/
60 50
glide-fronted /aw/
40 30 20 10 0 Gen. I
Gen. II
Gen. III
Generation Group
Figure 2. Percent raised /ay/ and glide-fronted /aw/ usage on Smith Island, in apparent time (Schilling-Estes and Wolfram 1999, graph adapted from Schilling-Estes 2000) Another morphosyntactic change in progress on Smith Island, and the topic of Chapter 5 to follow, is the leveling in all persons of past-tense be to the form
133
weren’t with contracted negation (Schilling-Estes 2000, Wolfram and Schilling-Estes 2003, Mittelstaedt in progress, to appear). WEIT’s pattern of increase is also similar to that observed for the weren’t leveling change. SchillingEstes (2000) demonstrates that weren’t leveling has proceeded at a lightning pace, faster than either the change to WEIT or the phonological changes discussed above. But the pattern is nonetheless similar to both: an S-shaped upward curve featuring a sharp increase of 25% between Generation I and II. The graph below also includes data from an additional generation group, Generation IV (born 1975-1987), who were interviewed in 1999 and 2000 by Natalie Schilling-Estes and colleagues. Much more will be said about weren’t leveling in Chapter 5.
Usage Percentage
100 80 60 40 20 0 Gen. I
Gen. II
Gen. III
Gen. IV
Generation Group
Figure 3. Percent leveled weren’t usage on Smith Island, in apparent time (adapted from Schilling-Estes 2000)
134
A sharp increase in usage appears between Generation I and II for every concentrating phonological and morphosyntactic variant that has been investigated to date on Smith Island. It seems very likely that this increase signals the beginning of the concentration process on Smith Island. According to Schilling-Estes and Wolfram, concentration may occur because speakers “seek, (consciously or unconsciously) to heighten their already increasing dialectal distinctiveness as a sort of linguistic ‘self-defense’ against the encroachment of the outside world” (1999: 510). Generation II was the first generation to experience real population attrition on Smith Island, growing up during the decades when the island’s population first began to decline significantly. Generation II was also the first generation to come into near daily contact with speakers of mainland varieties. Smith Islanders from Generation II were the first to attend high school on the mainland; prior generations attended school on the island, and visited the mainland less often. Direct and frequent exposure to mainland dialects would have highlighted the distinctive features of the Smith Island variety by contrast, providing a target for divergence. Generation II’s usage of distinctive dialect features would increase in response to their new awareness that the Smith
135
Island community, and hence its unique dialect, is threatened by population decline. Thus, the constellation of circumstances first experienced by Generation II might explain why these speakers appear to initiate the concentration process on Smith Island. There is, however, a major difference between weren’t leveling and the other changes. The changes involving raised /ay/, glide-fronted /aw/, and WEIT are not complete, leveling off at approximately 40%, 56%, and 77% respectively between Generations II and III. Presumably these changes will be completed in Generation IV or even later, if the concentration process continues until the Smith Island dialect dies. In contrast, weren’t leveling races to its conclusion between Generations II and III, leveling off at nearly 100% between Generations III and IV. It is not clear why weren’t leveling is so much more rapid than the other changes, and no explanations will be attempted here. According to Schilling-Estes and Wolfram (1999), another characteristic of the concentration process is that accelerated changes are found throughout the socially distinctive features of the dialect. That is, concentration must be distinguished from ‘focusing,’ where accelerated usage may be confined to just a single (and extremely
136
socially salient) variant. The WEIT results reported here replicate previous findings for an additional morphosyntactic variable, bringing to four the number of phonological and morphosyntactic variants known to be undergoing concentration on Smith Island.14
4.2.6
WEIT in real time
For a real-time comparison to support and confirm the apparent-time results, Parrott (2002) analyzed a small set of additional data. Some of these data come from reinterviews of one Generation I and one Generation III individual from the 1983 sample, carried out in 1999 and 2000 by Natalie Schilling-Estes and Laurie Zimmermann. Data were also extracted from a group interview with four Generation IV individuals (born 1982 - 1987), conducted by this author in 2000. Only interviews with females were used, in order to abstract away from sex as a potential social factor in such a small data set. The data were analyzed using the methodology outlined above. This produced too few tokens for VARBRUL analysis; the numbers and percentages are given below.
14
If the ain’t and don’t leveling cases reported in Mittelstaedt’s dissertation and discussed in Chapter 5 are to be considered separate variables, then the number is six.
137
Generation Group Generation I b. 1911 1 female Generation III b. 1971 1 female Generation IV b. 1982-1987 4 females
#WEIT/#Weak-expletive environs
%WEIT 52.6%
20 / 38
1983 = 48.3% 97.3%
36 / 37
1983 = 100% 61.4%
43 / 70
1983 Gen. III females = 84.3%
Table 5. Numbers and WEIT percentages (adapted from Parrott 2002) The individual Generation I and Generation III females use WEIT at a level very close--within 5%--to their 1983 usage. This suggests that the change observed in apparent time is real. It is evidently not an instance of age grading. However, the rate of WEIT usage is unexpectedly low for the Generation IV females. While Generation III females in 1983 used WEIT for an average of 84.3% of their weak expletives, for these Generation IV females WEIT usage declines to 61.4%. The concentration process predicts that WEIT usage should increase over time, and that prediction is born out for Generations I-III, where the trend is clearly upward. For this reason the apparent Generation IV decline is puzzling, but it could be merely the result of small sample size. If so, the average WEIT usage rate is predicted to increase as more Generation IV data are added
138
to the sample. This task, and more real-time analysis, needs to be undertaken in future research.
4.2.7
Summary and implications for variation analysis
The case of WEIT on Smith Island is an example of morphosyntactic variation and change in apparent time. Using the methods of variationists and the structural analyses of syntactic theory, we can observe that two variant morphosyntactic forms, it and there, appear in the syntactic environments of weak expletive there. The variant forms are not in complementary distribution, nor do they express different semantics or morphosyntactic functions. Plural associate agreement with there is variable on Smith Island, but there is not a single instance of plural associate agreement with WEIT in all of the data that have been analyzed to date. Moreover, the informants we have consulted so far reject plural associate agreement with WEIT. Although more data need to be examined, these preliminary results support the conclusion that that verbal agreement is categorically 3s with WEIT on Smith Island; that is, plural associate agreement with WEIT is impossible. 3s associate agreement is categorical with WEIT even independently of general agreement leveling variation on
139
Smith Island. As discussed further in Chapter 5, variable leveling to the 3s agreement form is observed in verbs, past-tense be, present-tense be, and present-tense have. For all these, 3s agreement leveling is at low usage levels, and is stable or declining in apparent time (Schilling-Estes 2000, Parrott 2001c, Trester 2003, Mittelstaedt in progress, to appear). For variationist quantitative analysis, the implication of these facts about agreement and WEIT is that expletive sentences should not be counted together with non-expletive sentences when analyzing subject-verb agreement. On Smith Island, including WEIT sentences in an analysis of agreement variation would inflate counts of 3s leveling, which are otherwise low and declining. WEIT should also be counted separately from there expletives in studies of agreement variation in weak-expletive constructions. There is another implication. When asked about their dialect, Smith Islanders invariably discuss and can demonstrate the use of glide-fronted /aw/, but they have never mentioned or demonstrated WEIT. As the sociolinguistic-interview transcript extracts above illustrate, Smith Islanders do not seem to be consciously aware of this variant (with its categorical 3s agreement)
140
even when asked for acceptability judgments about it. However, the variant form evidently expresses a locally relevant social meaning. This is evidenced by WEIT’s participation in the rapid, socially motivated process of concentration. The same is true not only of WEIT, but also of the non-salient but still-concentrating phonological variant raised /ay/. It is also true of the morphosyntactic variant leveled weren’t, the topic of the next chapter. Therefore, socially motivated changes in progress can be a diagnostic for the socially significance of variation even when individuals are not conscious of the variants.
4.3
WEIT as Morphosyntactic Mismatch
In the remaining two sections, I consider some theoretical issues raised by WEIT variation and agreement. English expletive constructions are a central and perennial topic in Minimalist syntactic theory (Chomsky 1995, 2000a, 2001, 2004a, and others). There have been relatively few attempts to reconcile theories with variable 3s associate in weakexpletive there constructions (e.g., Sobin 1997, Schütze 1999, Wilson and Henry 1998, Henry 2002, 2005, Rupp 2005), but no theoretical attempts to account for WEIT and its
141
agreement properties.15 As established above, verbal agreement is categorically 3s with WEIT. This fact needs theoretical explanation, since the impossibility of plural associate agreement with WEIT constitutes a morphosyntactic mismatch on a current Minimalist analysis of expletives and agreement (Chomsky 2000a).
4.3.1
A Minimalist theory of agreement
This section examines more closely the Minimalist Agree mechanism for feature checking that was very briefly introduced in Chapter 3 (Chomsky 2000a, see also Adger 2003). The operation Agree applies in the narrow syntax, checking semantically uninterpretable features before Spell Out to the interface components. Agree applies to elements in a C-command configuration. The uninterpretable features of the higher element ‘probe’ the interpretable features of a lower element, the ‘goal.’ If the all of the features on the probe match up with (a subset of) features on the goal, the probe’s features are checked, and valued the same as the goal’s. In this system, Case is an uninterpretable feature on the DP goal, checked and valued upon being probed. The role of Case is to make the interpretable features of the goal “visible” for a probe. By stipulation, 15
See Parrott (2000a) for one exception.
142
the Case feature is valued Nominative when probed by finite Tense. More will be said about Case in Chapter 6. Repeating from Chapter 3, syntactic feature checking by Agreement is schematized below. (47) Syntactic feature checking [u(ninterpretable)feature:value] (48) Agree in narrow syntax [PROBE = T[Past:±,
uφ: ]
... [GOAL = DP[φ:α,
uCase: ]]]
uφ: ]
... [GOAL = DP[φ:α,
uCase:Nom]]]
AGREE Î [PROBE = T[Past:±,
In this system, Case does not drive movement. Movement is the result of an uninterpretable feature EPP on T, checked by Merge of a DP. Also a stipulation, the EPP feature follows from nothing else in the theory and is merely a technical mechanism needed to express the empirical fact that English sentences need structural subjects. According to the now widely accepted VP-internal subject hypothesis, subject DPs are Merged in the specifier of VP and later moved to their overt position in the specifier of T. To give an example of how this system works, consider the examples below, with relevant structure shown in bracket notation for reasons of space. I will not discuss any internal structure of VP or TP here, focusing on 143
expletives and subject-verb agreement. At this derivational stage, the subject DP monster crabs is in its Merged position at the specifier of VP. Tense (T) consists of a semantically interpretable Tense feature and a set of uninterpretable agreement (phi = φ) features: Person (Pers:) and Number (Num:). T’s φ features are not valued. The DP monster crabs has an identical set of interpretable φ features, all of which are valued: its Person feature has the value Third (3) and its Number feature has the value Plural (p). The DP’s Case feature is uninterpretable and unvalued. (49) Monster crabs eat mainlanders. [TP T[uφ: , EPP]... [VP monster crabs[φ:3p,
uCase: ]
eat mainlanders]]
Applying the operation Agree, T’s uninterpretable φ features probe and find the interpretable φ-feature set of the goal DP monster crabs. T’s φ features, matching the set in DP, are valued 3p and then checked. DP’s Case feature, probed by finite T, is valued Nominative (Nom) and checked. Applying the operation Move, DP raises to the specifier of T (its pronounced position) in order to check T’s EPP feature. (50) Monster crabs eat mainlanders. AGREE Î
144
[TP T[uφ:3p, EPP]... [VP monster crabs[φ:3p,
uCase:Nom]
eat mainlanders]]
MOVE Î [TP monster crabs[φ:3p, uCase:Nom] T[uφ:3p, [VP t eat mainlanders]]
EPP]...
For present purposes, what is relevant about this feature-checking system is the way it explains verbal agreement. Recall from Chapter 3 that after Spell Out, T will be lowered to adjoin with the main verb; the phonological exponent for affixal T will depend on the value of its agreement φ features in the morphological component. On this Minimalist feature-checking theory, as reviewed in Chapter 3, the uninterpretable φ features of T are valued in the narrow syntax by Agreement with the interpretable φ features of the subject DP. Chomsky takes this to reflect the intuition that verbs agree with their subjects, and not vice versa. Thus here, because T is valued 3p by the subject, the finite verbal morphology appears in the 3p form -Ø (eat). If the subject DP had 3s φ features (the monster crab), T’s φ features would be valued 3s by Agree, and the finite verb would take the 3s agreement form -s (eats). (51) The monster crab eats mainlanders. [TP T[uφ:
, EPP]...
145
[VP the monster crab[φ:3s,
uCase: ]
eat mainlanders]]
AGREE Î [TP T[uφ:3s, EPP]... [VP the monster crab[φ:3s,
uCase:Nom]
eat mainlanders]]
MOVE Î [TP the monster crab[φ:3s, uCase:Nom] T[uφ:3s, [VP t eat mainlanders]] 4.3.2
EPP]...
Chomsky’s analysis of expletives
On Chomsky’s (2000a) analysis, expletives are Determiners (D) which can be Merged from the Lexical Array to check the EPP feature of T. Expletive it is identical to pronominal it, with interpretable 3s φ features and an uninterpretable Case feature. Weak expletive there has a single uninterpretable Person feature. (52) a. it b. there
=
D[φ:3s,
uCase: ]
=
D[uPers:
]
Chomsky’s analysis connects the complementary distribution of English expletive forms with the weak expletive’s characteristic property of rightward associate agreement. Consider the following examples. Expletive there is Merged at the specifier of VP. But T’s uninterpretable φ features cannot be checked, since the weak expletive does not have a full set of matching φ features. (53) There are monster crabs in the pot.
146
[TP T[uφ: , EPP] ... [VP there[uPers:
]
... monster crabs[φ:3p,
uCase: ]
... ]]
Now, there Moves to T in order to check T’s EPP feature. The uninterpretable Person feature of there probes T, finds a set of features to match with (i.e., a single feature person), and is checked. T’s uninterpretable φ features probe the lower associate DP monster crabs. This checks T’s uninterpretable φ features--and, crucially, the DP’s uninterpretable Case feature. T’s features are valued 3p by the lower DP monster crabs. This system explains why agreement in there-expletive constructions is with the associate DP: there has too few φ features to value T, which must always be valued by a lower DP with a full set of φ features. (54) There are monster crabs in the pot. AGREE, MOVE Î [TP there[uPers:3] T[uφ:3p, EPP] ... [VP t ... monster crabs[φ:3p,
uCase:Nom]]]]]]
Now, consider expletive it, which is Merged in the specifier of the highest VP. The uninterpretable features in the embedded clause have all been checked independently of the higher clause. Expletive it’s complete φ-feature set is probed by the highest T, valuing T’s φ features 3s and checking them (along with it’s Case feature). The system
147
explains why agreement in it-expletive constructions is categorically 3s: T’s features are always valued by it, which has a full set of φ features valued 3s. (55) It seems that monster crabs eat mainlanders. [TP T[uφ: , EPP] ... [VP it[φ:3s, uCase: ] ... seem [CP that [TP monster crabs[φ:3p, uCase:Nom] T[uφ:3p, [VP t ... eat mainlanders]]]]]
EPP]
...
EPP]
...
AGREE, MOVE Î [TP it[φ:3s, uCase:Nom] T[uφ:3s, EPP] ... [VP t ... seem [CP that [TP monster crabs[φ:3p, uCase:Nom] T[uφ:3p, [VP t ... eat mainlanders]]]]] This analysis gives a clear account of the complementary distribution of English expletives it and there. If the wrong expletive form is Merged, uninterpretable features will not be checked, causing a crash at the interfaces. What happens if expletive there is Merged to an itexpletive structure with a finite complement clause? Uninterpretable φ and Case features in the embedded clause have been independently checked. In the highest clause, T probes there. But the weak-expletive’s single Person feature cannot check the complete φ set of T. T still lacks a goal with a full set of matching φ features, so T’s uninterpretable φ-feature set is never checked or valued. There raises, but it cannot check its uninterpretable
148
Person feature since T is not valued. The structure crashes with uninterpretable features at the interface (unchecked features are double underlined in the example structures). (56) * There seem that monster crabs eat mainlanders. [TP T[uφ: , EPP] ... [VP there[uPers: ,] ... seem [CP that [TP monster crabs[φ:3p, uCase:Nom] T[uφ:3p, [VP t ... eat mainlanders]]]]]
EPP]
...
EPP]
...
AGREE, MOVE Î * [TP there[uPers: ,] T[uφ: , EPP] ... [VP t ... seem [CP that [TP monster crabs[φ:3p, uCase:Nom] T[uφ:3p, [VP t ... eat mainlanders]]]]]
Now consider what happens if expletive it is Merged into a weak-expletive there structure. T probes it, valuing and checking T’s uninterpretable φ set. But now there is nothing to probe the DP monster crabs, and so its uninterpretable Case feature remains unchecked. The structure crashes with an uninterpretable feature at the interface. The same fate should befall all weak-expletive structures with it: T’s φ-feature set is checked by probing it, and so an uninterpretable Case feature of the associate DP remains unchecked at the interface. (57) * It is monster crabs in the pot. (non-WEIT varieties) [TP T[uφ: , EPP] ... [VP it[φ:3s , uCase: ] ... monster crabs[φ:3p, uCase: ] ... ]] AGREE, MOVE Î * [TP it[φ:3s,
uCase:Nom]
T[uφ:3s,
EPP]
149
... [VP t ...
monster crabs[φ:3p,
uCase: ]
... ]]
To summarize, Chomsky’s analysis accounts for the apparent complementary distribution of expletive forms in English, as well as for their different agreement properties.16 Weak expletive there has a single uninterpretable Person feature which cannot check the φ features of T. T’s φ features are valued and checked when T probes the associate DP, resulting in rightward associate agreement. There can occur with copular and other nonfinite complement clauses, since DP’s Case feature is checked by T. In contrast, expletive it has its own complete set of interpretable φ features, valued 3s, that can check and value the uninterpretable φ features of T. The result is categorical 3s agreement with expletive it. It can occur with finite complement clauses because there are no lower DPs whose Case features are left unchecked without the probe of T.
4.3.3
WEIT as morphosyntactic mismatch
Given this Minimalist analysis of English expletives, WEIT variation (on Smith Island and elsewhere) constitutes a paradoxical case of morphosyntactic mismatch. The two 16
Chomsky (2000) does not address variable 3s agreement in weakexpletive constructions. As mentioned above, Chomsky (1995) dismisses such variation in a footnote.
150
expletive forms are not in complementary distribution as predicted by the theory: the form it occurs variably in the environment of weak expletive there (although the form there does not appear in the environment of expletive it). Moreover, 3s agreement is categorical with WEIT (although associate agreement is variable with expletive there). But these properties are a paradox on the analysis outlined above. Suppose WEIT has a full set of 3s features, as suggested by the fact of categorical 3s agreement with WEIT. Then WEIT’s occurrence in weak-expletive environments is not predicted: just as we saw above, WEIT should be probed by T, leaving an unchecked Case feature on the associate DP. (58) It is monster crabs in the pot. (WEIT varieties) [TP T[uφ: , EPP] ... [VP WEit[φ:3s , uCase: monster crabs[φ:3p, uCase: ] ... ]]
]
...
AGREE, MOVE Î * [TP WEit[φ:3s, uCase:Nom] T[uφ:3s, EPP] ... [VP t ... monster crabs[φ:3p, uCase: ] ... ]] On the other hand, suppose WEIT has only a single person feature like weak expletive there. Then T should probe the associate DP and check its Case feature. Since T’s φ features will be valued by the associate, this predicts at least the possibility of plural associate 151
agreement with WEIT (even if it is variable like with there). The prediction is contrary to fact: as established above, 3s agreement is categorical with WEIT. (59) * It are monster crabs in the pot. (WEIT varieties) [TP T[uφ: , EPP] ... [VP WEit[uPers:
]
... monster crabs[φ:3p,
uCase: ]
... ]]
AGREE, MOVE Î [TP WEit[uPers:3] T[uφ:3p, EPP] ... [VP t ... monster crabs[φ:3p,
uCase:Nom]]]]]]
Therefore, WEIT on Smith Island (also a common feature of many other English varieties as noted) is prima facie empirical evidence against Chomsky’s Minimalist featurechecking analysis of expletives and associate agreement in English.
4.4
A DM Approach to Expletives in English
This final section adopts the DM theoretical framework as outlined in Chapter 3 (Halle and Marantz 1993, Embick and Noyer to appear), and presents analyses of English expletives and their agreement properties.
4.4.1
Morphosyntactic features of expletives it and there
152
First of all, it seems safe to assume that WEIT is expletive it. The Oxford English Dictionary (OED)17 reports that the form it was “formerly” used as a variant of the weak expletive form there. The earliest citation given is from 1300, and the last is from 1617. All of the citations given by the OED have a singular associate and 3s agreement. (The one exception is an entry from 1435: “It were two dragons stiff and strong....”) It seems reasonable to suppose that WEIT was retained on Smith Island from the English dialects spoken by British settlers who first settled there in the 1650s. Because virtually no African Americans have ever lived in this isolated insular community, I conclude that the use of WEIT on Smith Island is not due to contact with AAE varieties. Indeed, WEIT is independently retained in many English varieties of the United States unrelated to Smith Island. WEIT may in fact be another piece of evidence for the hypothesis that modern AAE developed from exposure to earlier British dialects (Poplack 2000) rather than from a creole (Rickford 1998). Furthermore, it seems both plausible and parsimonious to assume that expletive it is identical to pronominal it. Both are abstract terminals with the syntactic category D, and both evidently have agreement features valued 3s during 17
I consulted the online edition.
153
the morphological component. What about expletive there? The syntactic category of there is unclear, and I want to leave this question unresolved here. As discussed above, Chomsky (2000) analyzes expletive there as a φ-feature deficient D because it can satisfy the EPP as a structural subject. However, Kleanthes Grohmann points out (p.c.) that phrases of categories other than DP can also satisfy the EPP. For example, consider that a Preposition (P) phrase can fill the subject position of a copular construction (a) as well as an expletive there (b) or a DP (c).18 (60) a.
[PP Under the bed] is a good place to hide.
b.
There is a good place to hide [PP under the bed].
c.
[DP A good place to hide] is under the bed.
Moreover, as Donna Lardiere points out (p.c.), the form there never agrees with anything, and nothing ever agrees with there. So let us conclude, at least, that expletive there is not a D and that it has no agreement features whatsoever. Arguably, whatever its category, expletive there has the same features as its locative homophone there in the morphological component. For concreteness, I will claim that the features in question are Proximate (near the speaker, Prox:±) and Distal (away
18
For similar examples see Moro (2000), who proposes an analysis of expletives based on Kayne’s Linear Correspondence Axiom theory (1994).
154
from the speaker, Dist:±). Locative/expletive there has the features [Prox:-, Dist:-], and locative here has the features [Prox:+, Dist:-].19 To summarize, on the DM approach taken here, both expletive subjects it and there are abstract syntactic terminals (not Roots). Expletive/pronominal it is a D with 3s agreement features. Expletive/locative there has an undetermined category (X), the features [Prox:-, Dist:-], and no agreement features. Vocabulary Items for it and there are provided below. Because these Vocabulary Items share no substantive morphosyntactic features, they do not compete.20 (61)
Postulated Vocabulary Items for it and there a.
[D, Pers:3, Num:s]
Ù
/ɪt/
b.
[X, Prox:-, Dist:-]
Ù
/ðɛɹ/
19
In this analysis I am following Nevins (to appear), who observes that two binary-valued features can yield a three-way distinction, with one possible feature-value combination ruled out semantically. Here, the combination [Prox:+, Dist:+] is impossible because nothing can be simultaneously close to and far from the speaker. Unlike other languages such as Korean, English does not morphologically distinguish the combinations [Prox:-, Dist:-] and [Prox:-, Dist:+]. It could be that there is a homophone with both meanings available. Alternately, English could lack any syntactic terminal item with (or, exponent for) the features [Prox:-, Dist:+], requiring a periphrastic strategy for expressing this semantic meaning (e.g., over there). I will not attempt to resolve the matter here. 20 For consistency, and because nothing about the analysis of expletives being developed in this chapter hinges on it, I will continue to use Pers:(1,2,3) and Num:(s,p) to represent agreement features. A more standard DM analysis treats person and number as binary-valued feature combinations (e.g. Halle 1997, Nevins to appear).
155
An important question remains. On the above analysis, the expletive subjects it and there arrive in the morphological component with the same features as their homophones. But these features have a semantic interpretation on their non-expletive counterparts. That is, pronominal it denotes something like a singular sex/gender-neutral entity that is neither the speaker nor addressee, while locative there denotes something like a location neither close to nor far from the speaker. But when used as expletive subjects, it and there do not seem to have any semantic denotation whatsoever. So what happens to their features at LF? I cannot adequately resolve this issue here, but I will suggest one possible solution following Nevins (to appear), who suggests that certain “features may be deleted on the LF branch, while surviving intact on the PF branch.” Thus, let us suppose that it and there do indeed have semantically interpretable features when they fulfill their syntactic function of satisfying the EPP as a phrase in a non-thematic subject position. These features are then spelled out to the PF-interface branch of the computation. At the LF interface, however, these otherwise interpretable features must be deleted by an operation Nevins calls ‘Deprivation’ because they occur on a terminal without any
156
θ-role. If it can be properly fleshed out in future research, such an analysis would explain why non-thematic expletive subjects do not have any semantic interpretation while apparently having the same features as their interpretable homophones during the morphological component.
4.4.2
A DM analysis of WEIT and the Case paradox
Continuing, how do we escape the Case paradox discussed above? Why doesn’t the unchecked Case feature of the associate DP cause a crash at the interfaces when 3s it appears in a weak expletive construction? I want to conclude that there are simply no such objects as uninterpretable syntactic Case features. In other words, no uninterpretable features on the associate DP must be checked in the narrow syntax. Such a conclusion diverges quite radically from standard Minimalist theoretical treatments of Case. But it is exactly the conclusion predicted by the theory of DM: case morphology is merely ‘ornamental’ and Case plays no role in the narrow syntax or the semantics (Halle and Marantz 1993, Embick and Noyer to appear). On this approach, so-called ‘Dissociated’ case features or morphemes are inserted by language-specific rules in the
157
morphological component, after the narrow syntax (see McFadden 2004 for a comprehensive DM theory of case).21 Of course, the DM theory of post-syntactic case applies to the human language faculty, not just the English language. However, because case morphology is not endowed by UG, language-specific Dissociated case rules must be learned during development from adequately transparent environmental linguistic input. For this reason, I will argue in Chapter 6 that English lacks even Dissociated case morphology, despite its pronominal forms. Adopting the DM treatment of Dissociated case, WEIT with categorical 3s agreement is not a morphosyntactic mismatch and the apparent paradox disappears. Either of it or there will satisfy EPP in a weak-expletive construction. Categorical 3s agreement with WEIT is due to the ordinary mechanisms of subject-verb agreement in English, which are discussed below.
4.4.3
A DM analysis of there and associate agreement
Continuing with this DM approach, let us suppose that in addition to Case, there are furthermore no uninterpretable Agreement features in the narrow syntax. Only those Person and Number features necessary for semantic interpretation 21
See also Beard (1995) for a similar post-syntactic approach to case in a different theory of morphology.
158
of DP22 are retained in this theory. For languages with ornamental verbal agreement morphology, Dissociated agreement features or morphemes are inserted by languagespecific rules during the morphological component (Halle and Marantz 1993, Halle 1997, Embick and Noyer to appear). Again, though this DM theory denies that there are uninterpretable syntactic agreement features, specific Dissociated agreement insertion rules must be learned for any particular language. Now consider English, which does have ornamental verbal agreement morphology. Unlike in many other languages such as German, verbal agreement exponence is never distinct from tense exponence in English. For example, as illustrated below, the form -s is a productive exponent of 3s present tense on regular verbs. Note that the following examples represent English varieties that lack agreementleveling variation. Finiteness in T will be represented by the semantically interpretable feature Past; past tense is represented by the feature value ‘plus’ (Past:+) and present tense by the feature value ‘minus’ (Past:-). (62) Agreement and tense on regular verbs in English
1s
Past:-
Past:+
I haunt
I haunted
22
Or perhaps Noun (N) phrase (NP), depending on the loci of agreement features.
159
2s
you haunt
you haunted
3s
(s)he haunts
(s)he haunted
123p
we/you/they haunt
we/you/they haunted
The forms am, is, and are are the exponents of person and number in the present-tense of be, and the forms was and were are the exponents of agreement in the past-tense. (63) Agreement and tense on be in English Past:-
Past:+
1s
I am
I was
2s
you are
you were
3s
(s)he is
(s)he was
123p
we/you/they are
we/you/they were
Thus, there is no obvious reason to think that Dissociated agreement rules insert a distinct agreement morpheme for English. Moreover, although the -ed exponent of past-tense on verbs does not distinguish agreement at all, we still need agreement for the past-tense forms was/were. Thus, we should not restrict the insertion of Dissociated agreement features to the present tense in English. Since the exponent of non-finite tense (to) does not distinguish person or number, however, we must limit Dissociated agreement feature insertion to finite tense (i.e., T with a feature Past:±).
160
As a first approximation for English, let us suppose that Dissociated agreement features are copied into T[Past:±] from the features of the subject DP. This is formalized in the rule below, which states that (Dissociated) Person and Number features valued α are inserted into the terminal morpheme T[Past:±] when a DP with (interpretable) Person and Number features valued α is in the specifier of T[Past:±]. (64) Postulated rule for Dissociated agreement in English T[Past:±] Æ T[Past:±,
Pers:α, Num:α]
/ [TP DP[Pers:α, Num:α] [ T[_] [...]]]
This rule explains why 3s agreement is categorical with pronominal and (weak) expletive it: whenever it is a subject, even in a weak-expletive construction, the dissociate agreement rule for English will copy 3s features into T during the morphological computation. Now let us suppose that rightward associate agreement in English is the result of a Dissociated agreement rule that copies agreement features from an associate DP lower in the VP whenever the subject phrase has no agreement features. This is formalized in the rule below, which states that (Dissociated) Person and Number features valued α are inserted into the terminal morpheme T[Past:±] when a phrase in the specifier of T[Past:±] has zero agreement
161
features (φ=Ø) and a DP with (interpretable) Person and Number features valued α is in the VP complement of T[Past:±]. (65) Postulated rule for associate agreement in English T[Past:±] Æ T[Past:±,
Pers:α, Num:α]
/ [TP XP[φ=Ø] [ T[_] [VP ... DP[Pers:α,
Num:α]]]]
This rule does not mention expletive there, and so it should apply whenever any phrase without agreement features occupies the subject position. The examples below illustrate that this is indeed the case: plural associate agreement occurs with PP subjects (a), just as it does with weak-expletive subjects (b) and plural DP subjects (c). (66) a.
[PP Under the bed] are good places to hide.
b.
There are good places to hide [PP under the bed].
c.
[DP Good places to hide] are under the bed.
This rule also helps explain the semi-complementary distribution of the expletive forms it and there in English. As we have seen, in WEIT varieties the form it can appear in all of the syntactic environments of the weakexpletive form there, with categorical 3s agreement. However, I am aware of no English variety that allows the form there to appear in the syntactic environments of expletive it, regardless of verbal agreement (this is certainly unattested on Smith Island). (67) a. * There is (are) raining over the bay tonight.
162
b. * There seem(s) that the islands are cursed. The associated agreement rule given above copies features from a DP in the VP complement of finite T. But in a weather sentence, there is no DP in the VP complement of finite T, only the adjectival or verbal weather predicate itself and possibly a prepositional or adverbial phrase. In a raising predicate with a finite complement phrase, finite T has a CP complement rather than a VP complement. Therefore, the associate agreement insertion rule cannot apply in either type of it-expletive construction. Using an expletive it subject in these constructions allows the ordinary Dissociated agreement rule to apply, since the terminal it has 3s agreement features to be copied. But the expletive there terminal never has agreement features, only the features [Prox:-, Dist:-], and so it prevents either agreement rule from applying. Without the insertion of Dissociated agreement features, no Vocabulary Item can be inserted in the terminal morpheme T[Past:±] (more will be said about the Vocabulary for finite T in the following chapter). Thus, on this analysis the expletive form there is predicted not to occur, and to be judged ill-formed, in the syntactic environments of expletive it. Notice that the analysis refers to the abstract features of the expletive terminals prior to Vocabulary Insertion, and not to the
163
Vocabulary Items that will supply phonological exponents for the expletives later during the morphological computation. Finally, the Dissociated agreement rules proposed above crucially must refer to hierarchical syntactic phrase structures that are input to the morphological component after Spell Out from the narrow syntax (as do the rules for Dissociated case proposed by McFadden 2004, Embick and Noyer to appear). Therefore, Dissociated feature insertion rules must be among the ‘structural’ operations that apply prior to the first round of Vocabulary Insertion and subsequent linearization of the morphosyntactic terminals (as discussed in Chapter 3, another structural morphological operation is lowering Merger, following Embick and Noyer 2001). More will be said about the ordering of morphological operations in Chapter 5, where it will be argued that Fusion is a ‘repair’ operation that applies to linearized terminal structures in order to allow insertion of Vocabulary Items (also following Kandybowicz 2006).
4.4.4
Speculation on variable 3s agreement with there
What can be said about variable 3s associate agreement with weak-expletive there? Although the main focus of this
164
chapter is WEIT, 3s variation with there is also found on Smith Island and in many other varieties of English, as noted. This kind of variation has been documented in all English varieties that also have general 3s agreement variation with non-expletive subjects. Thus, the mechanisms of 3s agreement variation with weak expletives could simply be the same mechanisms responsible for general 3s agreement variation with non-expletive subjects. However, if it can be conclusively demonstrated that 3s agreement variation with weak expletives occurs in English varieties lacking general 3s agreement variation, another analysis would be needed. Perhaps such variation is the result of an additional dissociated-agreement rule that inserts 3s features, instead of copying them from an associate DP, when the subject phrase has no agreement features? This rule would have to vary with the associate agreement rule above, raising the issue of whether and how dissociated feature insertion rules might compete. (68) Hypothetical 3s rule for dissociated agreement T[Past:±] Æ T[Past:±,
Pers:3, Num:s]
/ [TP XP[φ=Ø] [ T[_] [...]]]
The correct analysis of 3s agreement variation with weak expletive there will depend on answers to the unresolved empirical questions raised above. And note that
165
these questions can likely be addressed only by using variationist and sociolinguistic observational methods.
4.4.5
Summary and questions for morphosyntactic theory
On the DM analysis of expletives presented above, WEIT is just pronominal/expletive it, a D with semantically interpretable 3s agreement features. Expletive there is not a D, and has the semantically interpretable features [Prox:-, Dist:-], but no agreement features at all. There are no uninterpretable Case or Agreement features in the narrow syntax, only semantically interpretable agreement features on DP. In languages that have it, ornamental case and agreement morphology is the result of Dissociated feature/morpheme insertion rules that apply in the morphological component, after the narrow syntactic computation. On such a DM approach, WEIT with categorical 3s agreement poses no Case paradox. English verbal agreement morphology is the result of feature insertion rules that copy agreement features into finite T from a DP. The ordinary rule copies agreement features from a DP in the spec of finite T. An additional associate agreement rule copies agreement features from a DP in the VP complement of
166
finite T whenever the subject has no agreement features. The two rules are repeated below. (69) Postulated rules for Dissociated agreement in English a.
Ordinary agreement
T[Past:±] Æ T[Past:±, b.
Pers:α, Num:α]
Associate agreement
T[Past:±] Æ T[Past:±,
Pers:α, Num:α]
/ [TP DP[Pers:α, Num:α] [ T[_] [...]]] / [TP XP[φ=Ø] [ T[_] [VP ... DP[Pers:α,
Num:α]]]]
Therefore, 3s verbal agreement is categorical with 3s pronominal/expletive it. Expletive there has no agreement features, triggering Dissociated agreement with an associate DP. The form it can occur in weak-expletive environments since it has 3s features to be copied. The form there cannot appear in it-expletive environments because the there-expletive terminal has no agreement features and no other DP is available for copying, so that neither Dissociated agreement rule can apply. One significant difficulty for this analysis regards the apparent ‘ungrammaticality’ of WEIT in many English dialects. The analysis given above seems to predict that use of it as a weak expletive should be possible in any variety of English, as long as verbal agreement is categorically 3s. I have claimed that the mechanism of Labovian variation in WEIT dialects is the choice between
167
syntactic terminals with φ features valued 3s (it) or with non-φ features valued [Prox:-, Dist:-] (there), and that either terminal item can function as a weak expletive in a non-θ subject position. In the post-syntactic morphological component, non-φ-featured there triggers the Dissociated associate-agreement rule, while 3s φ-featured it always triggers the ordinary Dissociated subject-agreement rule. However, as reviewed in the first section of the chapter, it and there appear in complementary distribution in many English varieties, where WEIT is both unattested and judged unacceptable. In other words, what is the mechanism of parametric variation that distinguishes varieties with WEIT from those without WEIT? I cannot adequately address this question here, but I would like to speculate that the operation of Deprivation might be involved. As briefly mentioned above, this postulated operation deletes interpretable syntactic features at the LF interface, parallel to feature deletion rules in phonology and morphology (Nevins to appear). Deprivation accounts for cases, like the expletives considered here, where otherwise interpretable morphosyntactic features appear at PF but not at LF. We might suppose that in historic and contemporary dialects of English with WEIT, the Deprivation operation deletes the 168
interpretable features of both there and it at LF. In dialects that lack WEIT, Deprivation only deletes the interpretable [Prox:-, Dist:-] features of there and not the interpretable 3s φ features of it. Thus, the presence of semantically interpretable features on a non-thematic DP at LF causes the ungrammaticality of it in such varieties. Again, why does Deprivation apply to WEIT in some dialects but not others? Presumably, Deprivation is triggered by properties of the terminals in question, which can differ between (populations of) individuals. Thus, a more complete and therefore convincing account of parametric variation with WEIT awaits future research on the postulated Deprivation operation and its application in the case of English expletives. For now, I must leave these problematic issues unresolved. Of course, the most outstanding question raised in this chapter concerns the status of case and agreement in morphosyntactic theory. The analysis presented above is based upon the independent DM hypothesis that there are no semantically uninterpretable Case or φ features in the narrow syntax. This implies an even-stronger hypothesis that there are simply no uninterpretable syntactic features whatsoever, contrary to current standard Minimalist theory (Chomsky 1995, et seq., Pesetsky and Torrego to appear). Is 169
this going too far? Minimalist syntax relies upon uninterpretable Case and Agreement features to perform many tasks, for example the identification of phrases and landing sites for movement; even movement itself is theorized as the requirement of an uninterpretable EPP feature (Chomsky 2000a). Eliminating uninterpretable features as theoretical objects might seem excessively radical from a Minimalist perspective, and such a move would certainly force a great deal of mechanical modifications to the theory. There are intuitively compelling Minimalist conceptual arguments against uninterpretable features, however. Why do such semantically superfluous objects exist in an ‘optimal’ computational system? Why don’t they violate Full Interpretation, or the Inclusiveness Condition? How can features with no meaning be acquired, most especially if they are not pronounced? From a morphology-theoretic perspective, moreover, the elimination of uninterpretable features may not appear so radical. Semantically meaningless morphophonological forms, such as person and number agreement on verbal elements and case on nominal elements, are pervasive in human language. Their empirical properties--including paradigmatic syncretisms and other kinds of mismatches--must be accounted for by an
170
articulated theory of morphosyntax like DM.23 As Embick and Noyer (to appear) point out, it is already well accepted that a great deal of non-semantic information relevant to phonology, such as syllabic and prosodic structure, is established after syntax (Chomsky 1995: 228, 381 fn. 10). And the absence of semantic meaning in case and verbal agreement morphology can be easily explained if such morphology is established by language-specific mechanisms after the narrow syntax, during the PF-interface computation. Therefore, it is my view that a Minimalist theory of the human language faculty should do without such objects. In narrow syntax, the minimally necessary structurebuilding operations Merge and Move apply to terminal elements whose features are interpretable at the LF interface. Phrasal movement is motivated by LF-interface requirements of semantics and information structure, as suggested in some of Chomsky’s most recent work (2004a, 2005a, 2005b).24 Alternative explanations should be sought for the EPP in English and other languages where it is 23
And not just DM: for other examples of other morphological theories, see Beard (1995) and Anderson (1992). 24 We may wonder, what about head movement? Many species of head movement do not seem semantically motivated; would it be possible to relegate head movement to the morphological component? I return briefly to this puzzling issue in the last chapter. Meanwhile, see Parrott (2001a) for speculation about a DM theory of V-to-I head raising, and Matushansky (2006) for more recent theoretical proposals that split mechanisms of head movement between syntax and morphology.
171
observed (perhaps these explanations might also be found in the morphological component). Relegating the theoretical mechanisms of morphological phenomena to the PF-interface component is a strongly Minimalist hypothesis on this view. The core syntactic computation of the human language faculty is extremely minimal and efficient, reduced to the absolute conceptual necessity of a recursive operation Merge. Mechanisms of all other linguistic phenomena (including, significantly, both parametric and Labovian variation) are motivated by the requirement to interface with language-external systems of meaning and form. It is beyond the scope of this dissertation to adequately evaluate or defend the strongest hypothesis that there are no uninterpretable features in the narrow syntax, and I do not claim to offer any conclusive evidence on the issue. However, the DM analyses being developed in this dissertation can extend Minimalist theoretical coverage to the empirical domain of Labovian variation and change in progress. This lends support, in my view, to the weaker hypothesis that there are no uninterpretable Case or Agreement features in narrow syntax. Thus, I will adopt the DM analysis of Dissociated case and agreement throughout the remainder of the dissertation.
172
4.5
Conclusion
Let us pull back for a final perspective on the overall purpose of the dissertation, which is to identify some mechanisms and locations of Labovian variation in a Minimalist, DM theoretical framework. This chapter has examined a case of morphosyntactic variation and change in progress, namely WEIT on Smith Island. According to the theoretical analysis of English WEIT variation given above, the mechanism of Labovian variation is located in the abstract syntactic terminals (as on the approach advocated by Adger and Smith 2005, Adger to appear). The sociolinguistic “choice” is between whether it or there will be selected into the pre-syntactic Lexical Array, where the terminal item will be subject to syntactic and morphological operations that apply deterministically. The choice of which words to use is trivially available in any theory of syntax. The next chapter will examine the case of weren’t leveling, where I claim that the mechanisms of Labovian variation are located in the objects and operations of the morphological component. As we will see, these kinds of choices are only available in the theory of DM.
173
Chapter 5: Distributed Morphological Mechanisms of Smith Island Weren’t Leveling
5.
Introduction
This chapter examines weren’t leveling, another case of morphosyntactic variation and change in progress on Smith Island. Although weren’t leveling is not unique to Smith Island, the pattern of variation itself seems rather anomalous. At first glance, weren’t leveling might appear to be a part of agreement leveling variation in the paradigm of past-tense be. However, this chapter shows that this is not the case. Weren’t leveling occurs only with the contracted form of negation -n't (e.g. She weren't scared) and in the total absence of concurrent were leveling (e.g., *She were scared, *She were not scared). Moreover, variationist studies show that leveling to weren’t patterns independently of was-leveling in apparent time on Smith Island. This chapter provides a DM-mechanistic analysis of weren’t leveling that locates mechanisms of Labovian variation in the inventory and features of non-competing Vocabulary Items and their interaction with ordered morphological operations in post-syntactic morphological component. 174
The chapter is structured as follows. Section 5.1 outlines some facts about agreement leveling variation in varieties of English, with a focus on the especially common phenomenon of leveling in the paradigm of past-tense be. Section 5.2 reviews Adger and Smith’s Minimalist, lexical analysis of was leveling in the community of Buckie, Scotland, which locates mechanisms of variation in the uninterpretable features of syntactic terminals and their checking combinations in the narrow syntax. Section 5.3 examines weren’t leveling and briefly reviews two variationist studies of its usage in apparent time on Smith Island. Section 5.4 adopts the Distributed Morphology theoretical framework outlined in Chapters 3 and 4. The section presents a DM analysis where leveled weren’t is claimed to be a suppletive form, resulting from a noncompeting Vocabulary Item that requires morphological Fusion of the be, Tense, and Negation terminals during the morphological component. A non-competing Vocabulary approach to was leveling is briefly considered, but no analysis is advocated in this chapter (the penultimate section contains additional speculation on was leveling in DM). Finally, Section 5.5 returns to the dissertation’s overall goal of identifying the mechanisms of Labovian variation in a Minimalist-DM framework. In this case, the
175
mechanism of weren’t-leveling variation is located in the choice of non-competing Vocabulary Items.
5.1
Agreement Leveling in English Varieties
To begin, this section outlines some facts about so-called ‘leveling’ variation in English subject-verb agreement morphology. Leveling (a.k.a. ‘analogical leveling’ or ‘regularization’ in the variation literature) is a descriptive term referring to a pattern of morphosyntactic variation whereby one form in a paradigm (the ‘leveled’ or ‘pivot’ form) appears variably in the morphosyntactic environments of the other forms in the paradigm.1 100% usage of the leveled form can result in categorical (i.e. nonvariable) syncretism across the paradigm. However, actual usage of the leveled variant form need not increase: leveling variation may remain stable over time, decline over time, or even reverse trajectory over time (e.g. as shown by Hay and Schreier 2004, see Chapter 4). Moreover, leveling variation often does not affect every cell in a 1
Note that ‘paradigm’ is also intended here solely as a descriptive term, referring to a matrix of possible values for morphosyntactic features; cells in a paradigm may or may not be realized by any distinct morphological forms. Under the DM framework, paradigms are not theoretical objects and no rules or operations refer to them (see e.g. Bobaljik 2001, 2002 for arguments against any theoretical status for paradigms). Although I cannot pursue the point here, patterns of leveling documented in the variationist literature might constitute a source of empirical evidence regarding the theoretical status of paradigms. This is another reason for theorists to bridge the gap with variation study.
176
paradigm. For example, as discussed below, variable leveling to the form was occurs with 1p (we) pronominal and plural DP subjects, but is categorically non-attested with 3p (they) pronominal subjects in Buckie, Scotland (Smith 2000).
5.1.1
3s Agreement Leveling
Variation in subject-verb agreement morphology (a.k.a. ‘concord’ in the variationist literature) has been documented in many varieties of English. For example, as noted in the preceding chapter, variable leveling to the 3s agreement form is observed in all verbal paradigms on Smith Island. This includes main verbs in present tense, presenttense and past-tense copula/auxiliary be, and present-tense auxiliary have (examples attested in sociolinguistic interviews, from Schilling-Estes p.c.). (1)
a.
When the crabs comes to the bank.
b.
The ones that did it was the people that I knew.
c.
Buckrams is on the order of soft crabs but thicker shells.
d.
A lot of places I used to know has washed away.
Usage of leveled 3s verbal agreement forms is minimal and appears to be stable or declining in apparent time on
177
Smith Island, as confirmed by several variationist studies.2 Leveling in past-tense copula/auxiliary be (Schilling-Estes 2000, Mittelstaedt 2006) will be discussed in Section 5.4 below. Trester (2003, 2006) provides a variationist study of 3s leveling in present-tense verbs and present-tense copular/auxiliary be; negated auxiliaries ain’t and don’t were excluded along with expletive sentences. Trester found no instances of 3s leveled forms with 1p (we) or 3p (they) pronominal subjects. The following examples were attested in sociolinguistic interviews (from Trester 2006). (2)
a.
Some people likes it better down here than they do to Ewell.
b.
Me and Dad usually buys cars together and shares the insurance.
c.
They’ll tell them how things is going.
For purposes of comparison in apparent time, Trester divided her sample using the same generation groups as Schilling-Estes and Wolfram (1999) and Schilling-Estes (2000) (see Chapters 2 and 4). The table below shows the number and usage percentage of leveled forms (i.e., 3s
2
Parrott (2001c) gives a very preliminary sketch of 3s agreement leveling on Smith Island. Although not enough tokens were collected for statistical analysis, findings were consistent with other studies of agreement leveling on Smith Island (Schilling-Estes 2000, Trester 2003, 2006, Mittelstaedt 2006). Parrott found no instances of 3s leveling with 1p (we) or 3p (they) pronominal subjects.
178
singular forms with a plural DP subject) for each generation group in her sample.3 Generation Group Generation I b. 1899-1916 Generation II b. 1944-1961 Generation III b. 1966-1971 Generation IV b. 1967-1983 Totals
#3s leveled/ #leveling environs
%3s leveled
33/63
52%
41/112
36%
34/160
21%
32/104 140/439
30% 31%
Table 1. Present-tense 3s agreement leveling on Smith Island, numbers and usage percentages in apparent time (adapted from Trester 2003) As graphically illustrated below, 3s leveling in present-tense verbs and be declines in apparent time, dropping 31% from Generation I (52%) to Generation III (21%). Usage of leveled 3s forms increases slightly, by 9%, for Generation IV (30%) in Trester’s sample.
3
VARBRUL analyses confirmed that generation group was a statistically significant factor influencing agreement-leveling variation, and the probability of leveled-form usage decreases in each generation as expected from the usage percentages. VARBRUL results are not reported here, see Trester 2006 for details.
179
Usage Percentage
100 90 80 70 60 50 40 30 20 10 0 Gen. I
Gen. II
Gen. III
Gen. IV
Generation Group
Figure 1. Percent present-tense 3s agreement leveling on Smith Island, in apparent time (adapted from Trester 2003)
5.1.2
Allomorphs of past-tense be
Below is a paradigm for copula/auxiliary be in the pasttense, representing varieties of English without any leveling variation. Pronominal number values (singular and plural) are given in columns, and person values (1st, 2nd, 3rd) are given in rows. As mentioned above and discussed further below, quantitative analysis of agreement leveling often reveals a categorical distinction between pronominal and DP subjects.4 Thus, an additional row with singular and plural DPs is given in the paradigm. (3)
Past-tense be Singular
Plural
4
Variationist analyses consistently show a probabilistic difference between agreement leveling with pronominal vs. DP subjects. Probabilistic frequency effects in morphosyntactic variation are set aside throughout this dissertation, perhaps wrongly.
180
1st
I was
we were
2nd
you were
you5 were
3rd
she was
they were
DP
a boat was
boats were
In varieties without agreement leveling, the two distinct morphological forms of past-tense be occur in complementary distribution: was appears only with a 1s or 3s subject, and were appears only with a 2s or plural subject. Thus, was and were can be considered allomorphs of past-tense be. (4)
a. * I were scared.
(1s)
b. * You was scared. c. * She were scared.
(2s) (3s)
d. * We/you/they was scared.
(1,2,3p)
e. * People was scared. (Plural DP) Notice that the same forms occur regardless of the presence or type of negation, whether the full form not or the contracted form -n't. In other words, in many English varieties there is no apparent interaction between negation and the agreement allomorphy of past-tense be. A schematic
5
As noted in the previous chapter, some English varieties have a distinct 2p pronoun, for example y'all, yous, yins, and possibly you guys. Because these various pronoun forms are not evidently relevant to the matters under investigation here, only the homophonous 2p form you appears henceforth.
181
paradigm for past-tense be allomorphy, including negation, is given below. (5)
Past-tense be, with negation
5.1.3
Singular
Plural
1st
I was not I wasn’t
we were not we weren’t
2nd
you were not you weren’t
you were not you weren’t
3rd
she was not she wasn’t
they were not they weren’t
DP
a boat was not a boat wasn’t
boats were not boats weren't
Was leveling
Agreement-leveling variation is especially common in the past-tense paradigm of copula/auxiliary be. Leveling to the form was with 2s and plural subjects has been documented in varieties as diverse as African-American English, Appalachian and Ozark English, Southern White English, New York City English, and various dialects of British English (for reviews and references see Chambers 2002a). For example, consider was leveling in the isolated AfricanAmerican enclave communities of Nova Scotia, Canada (Poplack and Tagliamonte 1991). In this variety, leveled was is attested with 2s (you), 1p (we), and 3p (they) pronominal subjects, as well as with plural DP subjects
182
(examples attested in sociolinguistic interviews, from Tagliamonte and Smith 2000: 143). (6)
a.
You was in the choir with Melanie and Nellie.
b.
And we was the only colour family.
c.
They was picking up wood and thing.
d.
The books was different from the slates that we use.
Below is a schematic paradigm representing English varieties that have was-leveling variation across the pasttense be paradigm (i.e., with both pronominal and DP subjects). Leveled forms are given in bold (was) and variation is indicted with a percentage sign (%were). (7)
Was leveling Singular
Plural
1st
I was
we was (%were)
2nd
you was (%were)
you was (%were)
3rd
(s)he was
they was (%were)
DP
a boat was
boats was (%were)
Given that negation does not interact with past-tense be allomorphy in many English varieties, we might expect the same in was-leveling varieties. This expectation holds in Nova Scotian African-American English (example attested in a sociolinguistic interview, from Tagliamonte and Smith 2000: 143).
183
(8)
They wasn't prejudiced up there then. Below is a schematic paradigm including negation,
representing English varieties that have was-leveling variation with both pronominal and DP subjects. (9)
Was leveling, with negation Singular
Plural
1st
I was not I wasn’t
we was (%were) not we wasn’t (%weren’t)
2nd
you was (%were) not you wasn’t (%weren’t)
you was (%were) not you wasn’t (%weren’t)
3rd
(s)he was (s)he wasn’t
they was (%were) not they wasn’t (%weren’t)
DP
a boat was not a boat wasn’t
boats was (%were) not boats wasn’t (%weren’t)
5.1.4
Were leveling
Leveling to the form were with 1s and 3s subjects is evidently much less common than leveling to was, but has been documented in some British and U.S. English varieties (see e.g. Britain 2002, Wolfram and Schilling-Estes 2003). For example, consider (the remnants of) were leveling in the insular community of Ocracoke, NC (examples attested in sociolinguistic interviews, from Schilling-Estes and Wolfram 1994: 280). (10) a. b.
I were afraid I was going to miss something. The neighborhood she was in were just like the old Germans.
184
Below is a schematic paradigm representing English varieties that have were-leveling variation with both pronominal and DP subjects. Were leveling can be thought of as the mirror image of was leveling. (11) Were leveling Singular
Plural
1st
I were (%was)
we were
2nd
you were
you were
3rd
(s)he were (%was)
they were
DP
a boat were (%was)
boats were
As with was leveling, we might expect the same forms to occur regardless of the presence or type of negation, as in the hypothetical paradigm given below. (12) Were leveling, with negation (hypothetical) Singular
Plural
1st
I were (%was) not I weren't (%wasn't)
we were not we weren't
2nd
you were not you weren't
y'all were not y'all weren't
3rd
(s)he were (%was) not (s)he weren't (%wasn't)
they were not they weren't
DP
a boat were (%was) not (s)he weren't (%wasn't)
boats were not they weren't
However, it is not possible to evaluate this prediction on Ocracoke. Schilling-Estes and Wolfram report
185
that they observed no instances of leveled were with the full form of negation (not), and that informants “explicitly rejected” such sentences “while accepting their contracted counterparts” (1994: 281). (13) a. * I were not there yesterday. Cf. I weren’t there yesterday. b. * She were not on the boat yesterday. Cf. She weren’t there yesterday. Anticipating the presentation of Smith Island weren't leveling to follow in Section 5.3, there are a combination of reasons for this pattern of past-tense be leveling on Ocracoke. First, leveled were is used at extremely low levels by only the older speakers in the community (1.3% usage in Generation I and .08% usage in Generation II); were leveling is categorically absent in the youngest generation (0% usage in Generation III). Second, was leveling continues to be attested on Ocracoke, although it has declined to 9.4% usage in Generation III. Finally, leveling to the agreement form were with the contracted form of negation (-n’t) and a 1s or 3s subject is holding steady at nearly 50% for all generations on Ocracoke. Despite their total lack of were leveling, Generation III has the highest level of leveled-weren't usage at 56%. The table below gives the numbers and usage percentages of leveled variants was and were on Ocracoke by generation, as
186
well as separate numbers and usage percentages for the leveled forms wasn't and weren't.6 Generation Group Generation I b. 1911-1933 (5 persons) Generation II b. 1934-1958 (6 persons) Generation III b. 1967-1983 (8 persons) Totals (19 persons)
#was/ #leveling environs %was 13/91
#wasn’t/ #leveling environs %wasn’t 0/10
#were/ #leveling environs %were 6/464
#weren’t/ #leveling environs %weren’t 18/37
14.3% 23/142
0.0% 2/11
1.3% 6/733
48.6% 27/58
16.2% 8/85
18.2% 0/4
0.8% 0/262
46.6% 14/25
9.4% 44/318 13.8%
0.0% 2/25 8%
0.0% 12/1459 0.08%
56.0% 59/120 49.1%
Table 2. Past-tense be leveling on Ocracoke, NC, in apparent time (table adapted from Wolfram and SchillingEstes 2003, data from Schilling-Estes and Wolfram 1994)
5.2
Adger and Smith’s Minimalist, Lexical Approach
This section reviews Adger and Smith’s (2005) Minimalist, lexical analysis of was leveling in the Scottish village of Buckie (Smith 2000). On this approach to Labovian variation in morphosyntax, variant morphological forms are different terminals prior to and during the syntactic computation. Variant terminals have the same semantically interpretable features, but may have different uninterpretable features;
6
Schilling-Estes and Wolfram’s VARBRUL results are not reported here, but they did confirm the statistical significance of these generational usage percentages; see Schilling-Estes and Wolfram (1994) for details.
187
thus, variant terminals can be ‘spelled out’ at PF with correspondingly different morphophonological forms.7
5.2.1
A Minimalist syntax
Adger and Smith adopt the Minimalist theoretical framework outlined in Chapters 3 and 4 (Chomsky 2000a, Adger 2003). Their approach to variation relies on post-syntactic insertion of phonological content, although it does not make use of any mechanisms specific to DM theory. I follow Adger and Smith’s standard assumptions about the syntax of be. Copular be is the abstract functional head of a verbal phrase (VP) that takes a small clause, adjectival, or other kinds of phrasal complements; auxiliary be is the abstract functional head of a auxiliary phrase (AuxP) that takes a VP complement. I also follow them in setting aside questions about expanded structure internal to TP and VP. Adger and Smith report that the “copula vs. auxiliary status of the verb” was not statistically “significant for the use of [leveled] was” (2005: 174, fn. 6); likewise, none of the other studies of past-tense be leveling mentioned in this chapter
7
This approach was further developed in a subsequent article by Adger entitled “Combinatorial Variability” (2006), which unfortunately cannot be reviewed in detail here. Adger and Smith are currently working on a book about their Minimalist, lexical approach to Labovian variation in morphosyntax.
188
(Schilling-Estes 2000, Parrott 2001c, Trester 2003, 2006, Mittelstaedt 2006) found that the copula vs. auxiliary status of be made any difference in the patterns of variation. Therefore, I hereafter disregard the copula/auxiliary distinction. As illustrated below with tree notation (adapted from Adger and Smith 2005: 165), both kinds of be must raise by syntactic head movement to adjoin with finite Tense (T); the subject DP is Merged as the specifier of VP, and must raise by the syntactic operation Move to the specifier of T. (14) Syntax of auxiliary be TP 3 DP T’ 3 T AuxP 2 2 VP be T tbe 6 tDP ... Adger and Smith adopt a standard feature-checking theory as reviewed in Chapters 3 and 4: semantically uninterpretable features must be checked and valued by the operation Agree during the narrow syntactic computation, before reaching the LF interface. In their system, the abstract syntactic terminal T consists of a semantically interpretable Tense feature valued Past ([tense:past]), an uninterpretable Case feature valued Nominative 189
([ucase:nom]), and uninterpretable, unvalued Number and Person features ([unum:, upers:]).8 (15) T, unchecked and unvalued T[tense:past, ucase:nom, unum:, upers:] In Adger and Smith’s system, a Person feature can have a positive (+) value of 1 or 2 (first and second person) or a negative (-) value (third person). (16) Person features [pers:+]
[pers:-]
[pers:1] = 1st
[pers:-] = 3rd
[pers:2] = 2nd During the narrow syntactic computation, the uninterpretable features of T will be checked and valued by the interpretable features of the subject via the operation Agree. After syntax, PF “morpheme spell-out rules” yield a different morphophonological form for each configuration of T’s feature values when adjoined to be (adapted from Adger and Smith 2005: 165-167). (17) Morpheme spell-out rules for [be T] [be T[tense:past, ucase:nom, unum:sing, upers:1]] Æ
spells out as was
[be T[tense:past, ucase:nom, unum:sing, upers:2]]
8
Because it is not relevant for the phenomena and analyses considered here, Case will not be discussed again until Chapter 6.
190
Æ
spells out as were
[be T[tense:past, ucase:nom, unum:sing, upers:-]] Æ
spells out as was
[be T[tense:past, ucase:nom, unum:pl, upers:1]] Æ
spells out as were
[be T[tense:past, ucase:nom, unum:pl, upers:2]] Æ
spells out as were
[be T[tense:past, ucase:nom, unum:pl, upers:-]] Æ
spells out as were
To see how Adger and Smith’s system works, let us consider the following simple example. (18) They were running. Here, the pronominal subject they has semantically interpretable Person and Number features valued 3p. (19) pron[ucase: , num:pl, pers:-] T’s uninterpretable Person and Number features are checked by Agree in the narrow syntax and assume the same values as those of the subject pronoun. This is illustrated below using bracket notation (omitting analysis of the verb running). (20) [TP T[tense:past, ucase:nom, unum:, upers:] [AuxP be [VP pron[ucase: , num:pl, pers:-] [V running]]]] AGREE, MOVE Î [TP pron[ucase:nom, num:pl, pers:-] [be T[tense:past, ucase:nom,
unum:pl, upers:-]]
191
[AuxP tbe [VP tpron [V running]]]] At PF, morpheme spell-out rules yield the forms they for the pronoun and were for T. (21) Morpheme spell-out rules for pronoun and [be T] pron[ucase:nom, num:pl, pers:-] Æ
spells out as they
[be T[tense:past, ucase:nom, unum:pl, upers:-]] Æ
5.2.2
spells out as were
Was leveling in Buckie, Scotland
Adger and Smith use this Minimalist theoretical system to analyze an unusual case of was leveling in the small village of Buckie in northeastern Scotland (Smith 2000). As in the previous section, Buckie was-leveling variation exhibits no interaction with the presence or type of negation. However, Buckie was leveling shows a “relatively rare...variable/categorical split” (Adger and Smith 2005: 167) in the morphosyntactic environment for leveling: leveled was is completely unattested with 3p pronominal subjects (*they was), but occurs variably with full NP plural subjects (%boats was). (22) Buckie was leveling Singular
Plural
192
1st
I was not
we was (%were)
2nd
you was (%were)
you was (%were)
3rd
(s)he was
they were
DP
a boat was
boats was (%were)
Adger and Smith report that Buckie was leveling is declining slightly in apparent time. The oldest generation in their sample shows high usage of leveled was at about 80%, decreasing to approximately 50% usage in the two younger generations.
5.2.3
A Lexical analysis of Buckie was leveling
Adger and Smith propose that was-leveling variation arises because speakers of the Buckie English variety (and any other was-leveling variety) have an additional lexical item for Tense. This syntactic terminal, labeled T2, has no uninterpretable Number feature but is otherwise identical to the terminal T discussed above. T2 can check and value its uninterpretable Person feature with any subject during the narrow syntax. (23) T2, unchecked and unvalued T2[tense:past, ucase:nom, upers: ] At this point in their analysis, Adger and Smith must invoke PF morpheme spell-out rules. On their approach, “the featural content of [be T2] differs from that of [be T],
193
and the morphology can be sensitive to this, spelling out the former as was” (2005: 166). In Buckie, the morpheme spell-out rule for [be T2] is specified with a positive value for uninterpretable Person. (24) Morpheme spell-out rule for [be T] [be T2[tense:past, ucase:nom, upers:+]] Æ
spells out as was
Thus, [be T2] can only spell out as was if the Person feature of T2 has been valued [upers:+] by syntactic Agreement with the 1p pronoun we or the (homophonous) 2s and 2p pronouns you. If the Person feature of T2 has been valued [upers:-] by Agreement with the 3p pronoun they, no spell out morpheme is available for [be T2] at PF. This explains the categorical non-attestation of leveled was with the subject pronoun they in Buckie. (25) Buckie morpheme spell-out rules for [be T2] [be T2[tense:past, ucase:nom, upers:1]] Æ
spells out as was
[be T[tense:past, ucase:nom, upers:2]] Æ
spells out as was
[be T[tense:past, ucase:nom, upers:-]] =
no spell out
Adger and Smith claim that was-leveling variation with DP subjects in Buckie arises not from “multiple lexical 194
entries for T,” but rather from “multiple lexical entries for D” (2005: 170). On their analysis, D can inherit an uninterpretable Number feature with the value of the interpretable Number feature on the N head of its NP complement. However, in Buckie and some other varieties, an additional D2 can be lexically specified with an uninterpretable Number feature valued singular (adapted from Adger and Smith 2005: 168). (26) D D2
= [DP D[unum:α] [NP N[num:α]]] = [DP D[unum:sing] [NP N[num:α]]]
As evidence for this analysis of D, Adger and Smith point out that demonstratives show variable singular number agreement in Buckie; the following examples were attested in sociolinguistic interviews (2005: 169). (27) a. b.
My mam had all this stories o’ ootside folk. Cf. My mam had all these stories of outside folk. But I’d piles of that photos of the dancing. Cf. But I’d had piles of those photos of the dancing.
Therefore, when Buckie’s D2 is selected into the Lexical Array, the uninterpretable Number feature of T will be checked and valued singular and [be T] will be spelled out as was at PF. Notice that T2 is not involved in this variation at all, since both D and D2 will always have the value [pers:-] and [be T2] has no spell out morpheme with
195
this value for Person. Was leveling with D2 is exemplified below (adapted from Adger and Smith 2005: 170). (28) The boats was... [TP T[tense:past, ucase:nom, unum:, upers:] [AuxP be [VP [DP D2[ucase: , unum:sing, pers:-] [NP boats[num:pl, [V ... ]]]]
pers:-]]]
AGREE, MOVE Î [TP [DP D2[ucase:nom, unum:sing, pers:-] [NP boats[num:pl, [be T[tense:past, ucase:nom, unum:sing, upers:-]] [AuxP tbe [VP tDP [V ... ]]]]
pers:-]]]
(29) Morpheme spell-out rules for [D2], [NP boats]], [be T] D2[ucase:nom, unum:sing, pers:-] Æ
spells out as the
NP[boats[num:pl, pers:-]] Æ
spells out as boats
[be T[tense:past, ucase:nom, unum:pl, upers:-]] Æ
spells out as was
To account for the more commonly attested pattern of was-leveling variation across the past-tense be paradigm (i.e. with both plural DP and pronominal subjects, as in the Nova Scotian African-American English varieties mentioned above from Tagliamonte and Smith 2000), Adger and Smith need to posit a different morpheme spell-out rule for [be T2] that yields the form was “irrespective of the value of [upers]” on T2 (2005: 167). They do not specify what
196
such a rule would look like. Presumably it could refer to a [upers] feature with the value ±, as below. (30) Across-the-paradigm spell-out rule for [be T2] [be T2[tense:past, ucase:nom, upers:±]] Æ
spells out as was
Alternately, it could be that in such was-leveling varieties, the lexical item T2’ and its morpheme spell out rule when adjoined to be are totally unspecified for Person and Number, as below. (31) T2’, unchecked and unvalued T2’[tense:past, ucase:nom] (32) Morpheme spell-out rule for T2’ [be T2’[tense:past, ucase:nom]] Æ 5.2.4
spells out as was
Summary
On Adger and Smith’s Minimalist lexical approach, the mechanisms of Labovian variation in morphosyntax are located in the uninterpretable features of syntactic terminals and their checking combinations in the narrow syntax. This captures the “multiple form/single meaning notion of a linguistic variable” (2005: 173), and casts the sociolinguistic ‘choice’ of variants as a choice between terminals selected into the pre-syntactic Lexical Array. As
197
observed in the preceding chapter, the choice of words to use must be available in any theory of syntax. Adger and Smith’s lexical approach provides a straightforward analysis of was-leveling variation in Buckie and other English varieties. As we will see in the next section, the analysis of weren't leveling is rather less straightforward.
5.3
Weren’t leveling on Smith Island
This section introduces the phenomenon of weren’t leveling found on Smith Island and in a few other English varieties. This unusual pattern of morphosyntactic variation involves leveling to the form weren't in the absence of concurrent were leveling, and with or without concurrent was leveling. Weren’t leveling appears to have a significantly more limited geographic distribution than was leveling, which is quite common in English as noted above. To date, weren't leveling has been documented in just a few varieties spoken in Mid-Atlantic coastal regions of the United States, including Smith Island, Maryland, as well as Ocracoke Island, Harkers Island, and Hyde County, North Carolina (Schilling-Estes and Wolfram 1994, 2003, Wolfram and Schilling-Estes 2003); weren't leveling has also been
198
documented in a variety spoken in the Fens region of England (Britain 2002). Weren't leveling is perhaps the most dramatic morphosyntactic change in progress on Smith Island, having been nearly completed in 3 to 4 generations because of the rapid process of concentration. Usage of the leveled variant weren't is at or close to 100% for the youngest speakers in two apparent-time studies reviewed below (Schilling-Estes 2000, Mittelstaedt 2006).
5.3.1
Past-tense be leveling on Smith Island
As in many English varieties, was leveling is attested with plural DP subjects on Smith Island. Schilling-Estes (2000) reports 5 instances of leveled was with the 3p pronoun they and none with the 1p pronoun we; Mittelstaedt (2006) found no instances of was leveling with plural pronominal subjects, consistent with the findings of Parrott (2001c) and Trester (2003, 2006). Below are examples of was leveling attested in sociolinguistic interviews on Smith Island (from Mittelstaedt, p.c.). (33) a. b.
The boats was a lot slower. Two of ‘em was born to Crisfield and two was born to Salisbury.
199
As expected, was leveling shows no apparent interaction with negation; leveling to the form wasn’t is also attested in sociolinguistic interviews on Smith Island (from Mittelstaedt 2006). (34) There just wasn’t enough oysters.9 In every study to date, leveling to the form were with 1s and 3s pronominal or DP subjects is completely unattested on Smith Island. (35) a. * I were scared. b. * She were scared. c. * The boat were slower. Both the full (not) and contracted (-n’t) forms of negation are attested with past-tense be on Smith Island. (36) No it wasn’t, I can assure you this John Dunne poem was not in any way shape or form meant as a threat. However, leveling to the form were with the full form of negation is completely unattested on Smith Island. (37) a. * I were not scared. b. * She were not scared. c. * The boat were not slower.
9
This expletive sentence is the only instance (1/4) of leveled wasn’t in Mittelstaedt’s sample. Expletive sentences are typically included in quantitative analyses of agreement leveling, as in all the variationist studies cited in this chapter. However, there are reasons to suspect that expletive sentences are involved with independent patterns of variation (see Chapter 4); if so, expletive sentences might be inflating was/wasn’t leveling counts. This does not affect the theoretical points made here.
200
5.3.2
Weren’t leveling on Smith Island
On Smith Island, leveling to the form weren’t is well attested indeed, occurring with 1s and 3s pronominal and 3s DP subjects (from Mittelstaedt 2006 and p.c.). (38) a.
I weren’t able to answer.
b.
I weren’t very old.
c.
She weren’t that close to you.
d.
He weren’t expecting a boat.
e.
The man weren’t there every day.
f.
Ma weren’t doing no laughing.
The following schematic paradigm represents cooccurring was and weren’t leveling on Smith Island. (39) Smith Island was and weren’t leveling
5.3.3
Singular
Plural
1st
I was not I weren’t (%wasn’t)
we was (%were) not we wasn’t (%weren’t)
2nd
you was (%were) not you wasn’t (%weren’t)
you was (%were) not you wasn’t (%weren’t)
3rd
she was not she weren’t (%wasn’t)
they was (%were) not they wasn’t (%weren’t)
DP
a boat was not boats was (%were) not a boat weren’t (%wasn’t) boats wasn’t (%weren’t)
Two apparent-time studies
Two variationist quantitative studies on Smith Island (Schilling-Estes 2000, Mittelstaedt 2006) show that was
201
leveling is declining very slightly in apparent time, to about 13% usage in Generation IV (averaging across both studies). However, both studies show that usage of leveled wasn’t has dropped much more drastically, plunging to categorical 0% percent usage in generation groups III and IV. The rapid loss of leveled wasn’t stands in stark contrast to the rapid increase of leveled weren’t on Smith Island. Both studies found that total usage of leveled weren’t averaged 61% across all generation groups. Schilling-Estes (2000) found categorical 100% usage of leveled weren’t in Generation III, and near-categorical 96.4% usage in Generation IV. Mittelstaedt (2006) found near-categorical 90.9% use of leveled weren’t in Generation IV.10 Tables and graphs for both studies are given below.
10
Mittelstaedt’s sample contained two outlier individuals in Generation III, both of whose leveled weren’t usage was substantially reduced for social and stylistic reasons. For details, see Mittelstaedt (2006, to appear); for more on stylistic variation and change in apparent time, see Schilling-Estes (2002b, 2005).
202
Generation Group Generation I b. 1899-1932 (7 persons) Generation II b. 1942-1961 (7 persons) Generation III b. 1965-1971 (9 persons) Generation IV b. 1975-1987 (6 persons) Totals (29 persons)
#was/ #leveling environs %was 34/99
#wasn’t/ #leveling environs %wasn’t 5/6
#were/ #leveling environs %were 0/418
#weren’t/ #leveling environs %weren’t 6/27
34.3% 17/116
83.3% 2/9
0.0% 0/462
22.2% 17/36
14.7% 11/49
22.2% 0/2
0.0% 0/214
47.2% 12/12
22.4% 6/51
0.0 0/2
0.0% 0/254
100% 27/28
11.8% 68/315 21.5%
0.0% 7/19 36.8%
0.0% 0/1348 0%
96.4% 62/103 60.1%
Table 3. Past-tense be leveling on Smith Island, MD, in apparent time (data from Schilling-Estes 2000, table adapted from Wolfram and Schilling-Estes 2003)11
100 90
Usage Percentage
80 70
leveling to was
60
leveling to wasn't
50
leveling to were
40
leveling to weren't
30 20 10 0 Gen. I
Gen. II Gen. III Gen. IV Generation Group
Figure 2. Past-tense be leveling on Smith Island, MD, in apparent time (data from Schilling-Estes 2000, graph adapted from Wolfram and Schilling-Estes 2003)
11
The number of speakers per generation group was not indicated in the cited publications; that information comes from Schilling-Estes (personal communication).
203
Generation Group Generation I b. 1923-1933 (4 persons) Generation II b. 1938-1958 (7 persons) Generation III b. 1960-1970 (5 persons) Generation IV b. 1973-1987 (13 persons) Totals (29 persons)
#was/ #leveling environs %was 9/61
#wasn’t/ #leveling environs %wasn’t 1/4
#were/ #leveling environs %were 0/119
#weren’t/ #leveling environs %weren’t 28/42
14.8% 6/40
25% 0/9
0% 0/107
66.7% 17/31
15% 9/55
0% 0/8
0% 0/107
54.8% 7/24
16.4% 6/41
0% 0/10
0% 0/134
29.1% 30/33
14.6% 30/197 15%
0% 1/34 2.9%
0% 0/467 0%
90.9% 82/130 63%
Table 4. Past-tense be leveling on Smith Island, MD, in apparent time (table adapted from Mittelstaedt 2006) 100 90 Usage Percentage
80 70
leveling to was
60
leveling to wasn't
50
leveling to were
40
leveling to weren't
30 20 10 0 Gen. I
Gen. II
Gen. III
Gen. IV
Generation Group
Figure 3. Past-tense be leveling on Smith Island, MD, in apparent time (data from Mittelstaedt 2006) In Schilling-Estes (2000), the Generation I-III samples are extracted from sociolinguistic interviews conducted in 1983, while the Generation IV sample is from interviews conducted during 1999-2001. In Mittelstaedt
204
(2006), all of the Generation I-IV samples are from interviews conducted during 1999-2001. Thus, comparing the two studies gives us a snapshot of change in real time for Generations I-III (and in apparent time, when comparing Generations I-III with Generation IV).12 The figure below shows that weren’t leveling seems to be increasing in real time for Generation I, but remains relatively steady for Generation II. What looks like a real-time decline in weren’t leveling for Generation III is more likely due to style-shifting outlier individuals in Mittelstaedt’s sample (see footnote 10 above). The rapid change in apparent time
Usage Percentage
is quite evident for Generation IV. 100 90 80 70 60 50 40 30 20 10 0
Gen. I Gen. II Gen. III Gen. IV (1999-2001)
Schilling-Estes (Gen. I-III 1983)
Mittelstaedt (Gen. IIV 1999-2001)
Year(s) of Interviews
Figure 4. Weren’t leveling on Smith Island, MD, in real and apparent time, two samples (Schilling-Estes 2000, Mittelstaedt 2006) 12
Thanks to Jennifer Mittelstaedt for pointing this out.
205
5.3.4
Summary
Despite appearances, several characteristics of Smith Island weren’t leveling indicate that it is not an ordinary instance of agreement-leveling variation. The first and most remarkable distinguishing characteristic of weren't leveling is that it occurs in the total absence of leveling to the form were. Moreover, weren’t leveling occurs variably with both singular DP and pronominal subjects. In contrast, 3s agreement leveling in all verbal paradigms occurs extremely rarely, if at all, with pronominal subjects on Smith Island. As noted above, Schilling-Estes (2000) found only 5 instances of leveled was with the 3p pronoun they, while Mittelstaedt (2006) found none; neither study observed was-leveling with the 1p pronoun we. Finally, weren’t leveling patterns independently of was leveling in apparent time on Smith Island. The change to categorical usage of leveled weren’t has been very nearly completed at the expense of leveled wasn’t (that is, usage of leveled was with contracted negation has simultaneously plummeted to 0%). At the same time, was leveling is relatively stable at low usage levels, perhaps declining very slightly in apparent time. These facts suggest that the mechanisms of weren't leveling are different from those
206
underlying was leveling and the general process of 3s agreement leveling on Smith Island.
5.4
Distributed Morphological Mechanisms
This section presents a DM-mechanistic analysis of Smith Island weren’t leveling, essentially following an original proposal by Mittelstaedt and Parrott (2002, see also Parrott to appear-b). I adopt the Minimalist, DM theoretical model that was outlined in Chapter 3 and implemented in Chapter 4. Adger and Smith’s (2005) analysis of was-leveling reviewed above does not utilize mechanisms of DM and it is unclear whether their lexical approach is compatible with this theory. In the last section below, I tentatively suggest some ways that Adger and Smith’s analysis of was-leveling might be implemented in a DM framework.
5.4.1
Lexical analyses of weren't leveling
The facts reviewed in the preceding section make it clear that a simple lexical analysis of were leveling will not suffice to capture weren’t leveling variation. For example, we could straightforwardly account for across-the-paradigm were leveling in Adger and Smith’s system by positing a variant syntactic terminal T3 that has no uninterpretable
207
Person feature (or, no agreement features whatsoever). T3 will check its uninterpretable features with those of any subject and spell out as were when adjoined to be. Of course, this analysis predicts the possibility of concurrent were leveling, contrary to fact on Smith Island. (40) Hypothetical T3, unchecked and unvalued T3[tense:past, ucase:nom, (unum: )] (41) Hypothetical morpheme spell-out rule for T3 [be T3[tense:past, ucase:nom, (unum:±)]] Æ
spells out as were
One way to account for weren’t leveling on Adger and Smith’s lexical approach would be to posit an additional lexical item T-Neg. This terminal has no uninterpretable Person and Number features, but contains an uninterpretable Negation feature that can be checked by Agreement with the interpretable [neg] terminal during the narrow syntax. At PF, the morpheme spell-out rule for [be T-Neg] yields the form weren’t (Adger p.c.). (42) Hypothetical T-Neg, unchecked and unvalued T-Neg[tense:past, ucase:nom, uneg] (43) Hypothetical morpheme spell-out rule for [be T-Neg] [be T-Neg[tense:past, ucase:nom, uneg]] Æ
208
spells out as were
However, this analysis is problematic for several reasons. First, the existence of an uninterpretable negation feature in Minimalist syntax is not independently motivated and seems somewhat ad hoc in this analysis.13 Moreover, T-Neg contains an uninterpretable Case feature in addition to its uninterpretable Negation feature. But the matching Case features of a subject DP do not appear on the same terminal as interpretable Negation. Therefore, T-Neg’s uninterpretable features would seem to be uncheckable if the uninterpretable features of a Probe must be checked all at once (as in Chomsky 2000); if not (e.g. as in Castillo, Drury and Grohmann 1999), the T-Neg analysis seems to predict the possibility of were leveling with the full form of negation not, contrary to fact on Smith Island. Furthermore, a strong version of DM denies the existence of uninterpretable syntactic features altogether (see Chapter 4). But the T-Neg analysis would directly contradict even the weaker version of DM theory adopted in this dissertation, by apparently requiring the assembly of primitive features (Tense, Case, and Negation) in a generative pre-syntactic lexicon. Thus, such a lexical analysis of weren’t leveling cannot be implemented in a DM 13
Although Parrott (2000b) also proposed an uninterpretable negation feature as an ad hoc mechanism for capturing Negative Inversion in AAE, I no longer agree with that analysis. For a better attempt to motivate something like an uninterpretable negation feature, see Brown (1999).
209
framework. Lastly, even though Adger and Smith’s lexical approach requires the late insertion of phonological features, they do not adopt the theory of DM and of course are not bound to this framework. Nonetheless, the T-Neg analysis goes against this author’s intuition that Tense and Negation are separate terminals and must be combined syntactically and/or post-syntactically to yield contracted negation (-n’t) on past-tense be.14 Finally, another way to account for weren’t leveling in Adger and Smith’s system might be to propose that a morpheme spell-out rule yields the form weren’t only when the variant terminal T3 from above is adjoined to be and the syntactic terminal [neg]. (44) Hypothetical morpheme spell-out rule for [[be T3] neg] [[be T3[tense:past, ucase:nom, (unum:±)]] neg] Æ
spells out as weren’t
But this analysis seems to undermine Adger and Smith’s lexical approach to Labovian variation. Why do we need a distinct syntactic terminal T3 if it will only spell out at PF after being adjoined with [neg]? Instead, that is, why not simply have a distinct morpheme spell-out rule for
14
Such an intuition may not be worth much. As David Adger points out (p.c.), the empirical facts observed by Zwicky and Pullum (1983) can be interpreted as supporting a lexical approach like the T-Neg analysis. Zwicky and Pullum do not offer such a specific theoretical analysis of -n’t, however.
210
ordinary T when it is adjoined to [neg], yielding the leveled form weren’t at PF? (45) Hypothetical morpheme spell-out rule for [[be T] neg] [[be T[tense:past, ucase:nom, unum:±, upers:±] neg] Æ
spells out as weren’t
This suggests a DM approach where mechanisms of Labovian variation are located in the post-syntactic morphological component. Below, I offer a very similar analysis of weren't leveling as a kind of suppletion.
5.4.2
Past-tense be suppletion and allomorphy
At this point, I adopt the theory of DM as outlined in Chapters 3 and 4. Finite Tense is represented with the feature T[Past:±], and Person and Number continue to be represented by the features [Pers:(1,2,3)] and [Num:(s,p)].15 There are no uninterpretable Agreement or Case features in the syntax. The Person and Number features of T are copied from the interpretable features of the subject by an English-specific dissociated agreement rule
15
As mentioned in a Chapter 4 footnote, a more standard DM analysis treats both person and number as a combination of binary-valued features (e.g. Halle 1997, Embick and Noyer to appear, Nevins to appear). I continue to set this matter aside, since it does not affect the analysis of weren’t leveling being developed in this chapter. However, as will be briefly discussed in the penultimate section, the featural analysis of person and number is certainly relevant for an account of agreement leveling variation in English be (e.g. Adger 2006, Nevins and Parrott in progress).
211
that applies in the first round of operations during the morphological component. As reviewed in Chapter 3, the morphological operation Fusion is a mechanism of suppletion in DM that can be used to account for the evidently suppletive forms of the English copula/auxiliary be in the present (am, is, are) and past tense (was, were). We can straightforwardly assume that adjoined be and T[Past:±] undergo Fusion in the morphological component after narrow syntax. Following Adger and Smith’s analysis of be, Fusion of be and T[Past:±] is fed by syntactic head movement. The Fusion operation can also be fed by morphological Merger, as we will see below. Fusion of the adjoined terminals [[be] T[Past:±]] yields a single terminal [be T[Past:±]] in the morphological component. This operation is necessary in order to insert the Vocabulary Items for English be, which do not insert exponents into distinct terminals for be and finite T[Past:±]. Thus Fusion acts as a kind of repair operation, modifying the structural configuration of morphosyntactic terminals in order to prepare them for Vocabulary Insertion. According to the analysis given in more detail below (and following Kandybowicz to appear), Fusion applies late in the morphological computation, after the first round of Vocabulary Insertion and subsequent linearization of the
212
hierarchical syntactic structure. As reviewed in Chapter 3, constituency of linearized complex terminals is given in bracket notation, with ‘[M ...]’ representing the maximal terminal boundary16 and ‘*’ representing linear adjacency following Embick (to appear-a). (46) Morphological Fusion of be and T[past:±] ...[M [be] * T[past:±]]...
Æ
...[M be T[past:±]]...
As reviewed in Chapter 3, competition of underspecified Vocabulary Items is a mechanism of allomorphy in DM. One way to capture the allomorphy of English past-tense be is by postulating that the Vocabulary Items given below compete for insertion into the single terminal created by Fusion of be and T[Past:+]. Notice that the substantive morphosyntactic features of these Vocabulary Items (on the left side of the double arrow) contain the features of both be and finite T[Past:+]. If these terminals were not Fused in the morphological component, the suppletive Vocabulary Items for English be could not be inserted at all. (47) Postulated Vocabulary for past-tense be [be, Past:+, Num:s, Pers:2]
⇔
/w/
[be, Past:+, Num:p]
⇔
/w/
16
On Embick’s approach, adopted here, a maximal terminal is the X0Max of the bare phrase structure given in Chomsky (1995).
213
⇔
[be, Past:+] (elsewhere)
/wz/
In this postulated Vocabulary block, the most specified Vocabulary Item inserts the phonological features of the form were when the Person and Number features of the target node are valued 2s. This is meant to account for the fact that form were is exceptional in 2s, involving horizontal syncretism in the paradigm. The next and lessspecified Item in this Vocabulary block inserts the form were when the Number feature of the target terminal [M be T[past:+]] is valued plural. Finally, the least-specified elsewhere Vocabulary Item inserts was by default. Andrew Nevins (p.c.) points out another way that was/were allomorphy could be captured. In DM, so-called Impoverishment rules can delete features of terminals during the morphological component (Halle 1997). Here, an Impoverishment rule could delete the singular Number feature of T when it occurs in the context of a Person feature valued 2nd. (48) Hypothetical Impoverishment rule for past-tense be [Num:s]
Æ
Ø
/
[ __ Pers:2]
Now we would need only two Vocabulary Items to insert the exponent forms of past-tense be. (49) Hypothetical Vocabulary for past-tense be (Impov.) [be, Past:+, Num:s]
Ù 214
/wz/
(elsewhere)
Ù
/w/
Since nothing about the analysis of weren’t leveling to follow apparently hinges on the matter, I will continue to assume the non-Impoverishment analysis. However, comparing the two analyses for a DM-theoretical account of agreement leveling variation in English be is part of the research currently pursued by Nevins and Parrott (in progress). 5.4.3
Contracted -n’t
A full account of the morphosyntax of negation in a DM framework is unfortunately too large a task to be undertaken in this dissertation.17 For the purpose of analyzing weren’t leveling, I will simply make the following assumptions about the morphosyntax of English negation. First, Negation is an abstract syntactic terminal Neg (roughly following Zanuttini 1997, 2001); second, Neg is Merged after Aux, so that the Negation Phrase (NegP) is above AuxP (following Adger 2003); third, the postsyntactic operation of morphological Merger lowers finite T[Past:±] over Neg to adjoin with main verbs in English (following Embick and Noyer 2001); and finally, syntactic
17
Providing such an account is part of my research currently in progress.
215
head movement raises English auxiliaries like be to finite T[Past:±] over Neg (Roberts 2001). (50) Syntax of auxiliary be and negation TP 3 DP T’ 3 T NegP 2 2 be T[Past:±] Neg AuxP 3 tbe VP 6 tDP ... I furthermore presume that there is a significant morphosyntactic difference between the phonologically reduced form of negation -n’t and the full form not in English. The full form not is a maximal projection of the terminal Neg, perhaps Merged or Moved into the specifier of another syntactic terminal (Zanuttini 2001). However, the reduced form -n’t results from the head adjunction of Neg to [M [be] T[past:±]] at some point in the narrow syntax or post-syntactic morphological computation. It could be that the adjunction of Neg to [M [be] T[past:±]] is accomplished via syntactic head movement. However, in keeping with the DM-framework adopted here, I would like to assume instead that post-syntactic morphological Merger lowers the adjoined [M [be] T[past:±]] terminal to Neg, as shown below. Semantic and other differences between negation with not 216
and -n’t (see Zwicky and Pullum 1983) suggest that this Merger operation is not optional, as I will assume without specifying what triggers or constrains it. (51) Morphological Merger of Neg and [[be] T] T’ Æ 3 NegP TM 2 3 be T Neg AuxP 6 tbe ...
T’ 3 NegP 3 AuxP Neg M 2 5 T Neg tbe ... 2 be T
I further assume that the morphophonology of English is sensitive to the adjunction of Neg. This idea can be implemented by positing a separate Vocabulary Item that inserts the exponent -n’t when Neg is right-adjacent to maximal finite T[Past:±] inside maximal Neg; on this analysis, not is treated as the elsewhere exponent of Neg.18 (52) Postulated Vocabulary for English negation [Neg]
⇔
/nt/ / [M [T[Past:±]] * __ ]
[Neg] (elsewhere)
⇔
/nat/
Again, nothing about the analysis of weren’t leveling should hinge on these matters, as long as [M [be] T[past:±]] 18
Alternately, it might be that a PF readjustment rule simply phonologically reduces the form not when it is right adjacent to finite T. However, Zwicky and Pullum (1983) give convincing empirical arguments against such a simple phonological contraction or cliticization analysis. Accordingly, I will continue to assume that the reduced form -n’t results from some kind of morphosyntactic affixation process, however implemented.
217
and Neg end up head adjoined in the morphological component.
5.4.4
Weren’t leveling as suppletion of negation
Schilling-Estes and Wolfram (1994: 294) suggest that weren’t leveling results from a ‘remorphologization’ of negation that yields “suppletive-like negators that function as unanalyzable units,” analogous to the form ain’t. (53)
Ain't analogy (Schilling-Estes & Wolfram 1994: 290) am is are
:
ain't
::
was
:
x (x = weren't)
were
Mittelstaedt and Parrott (2002, see also Parrott to appear-b) implement Schilling-Estes and Wolfram’s suggestion in a DM framework. On this analysis, weren’t leveling results from a Vocabulary Item for past-tense be that has no Person or Number agreement features whatsoever, but includes a Negation feature. (54) Postulated Vocabulary Item for leveled weren’t [be, Past:+, Neg]
⇔
/wnt/
In order for the leveled form weren’t to be inserted, this suppletive Vocabulary Item requires morphological Fusion of the Neg terminal with the [M be T[past:+]] terminal. In this case, the head adjunction of [M [be] T[past:+]] and 218
Neg (whether via syntactic Movement or morphological Merger) for the reduced form -n’t feeds Fusion. As shown above, [M [be] T[past:+]] must be Fused in any case because of the suppletive Vocabulary Items for past-tense be. (55) Morphological Fusion of [M be T[past:+]] and Neg [M [be T[past:+] ] * Neg ] Æ
[M be T[past:+] Neg ]
The suppletive Vocabulary Item for leveled weren’t should not compete for insertion. Its substantive morphosyntactic features are a subset of any target terminal only when Fusion has applied, in which case weren’t will be the only insertable Vocabulary Item.19 The non-competition of weren’t with Vocabulary for past-tense be and Neg is illustrated below. Non-competition of Vocabulary Items is notated hereafter with a dotted-dashed line. (56) Non-competition of weren’t and was/were (Non-Impov.) [be, Past:+, Neg]
⇔
/wnt/
[be, Past:+, Num:s, Pers:2]
⇔
/w/
[be, Past:+, Num:p]
⇔
/w/
[be, Past:+] (elsewhere)
⇔
/wz/
(57) Non-competition of weren’t, not/-n’t
19
Note that for the same reasons, the suppletive Vocabulary Item for leveled weren’t does not compete with the Vocabulary Items of the Impoverishment analysis mentioned above.
219
[[be, Past:+, Neg]
⇔
/wnt/
[Neg]
⇔
/nt/ / [M [T[Past:±]] * __ ]
[Neg] (elsewhere)
⇔
/nat/
On this analysis, Labovian variation arises from the non-competition of Vocabulary Items. In other words, the variant form leveled weren’t is a sociolinguistic choice for individuals who have this non-competing Vocabulary Item in their inventory.
5.4.5
Fusion and the ordering of operations
The suppletion analysis of weren’t leveling raises an important question about the morphological operation of Fusion. I have claimed that the non-competing Vocabulary Item for weren’t can be chosen by speakers because it does not compete with other Vocabulary for insertion. However, the operation Fusion must apply to the [M be T[past:+]] and Neg terminals in order for weren’t to be inserted. Does this mean that an individual can “choose to Fuse” (in the words of Donna Lardiere, p.c.)? That is, can mechanisms of Labovian variation be located in a variable Fusion operation? On the contrary, I would like to maintain that in this case the mechanisms of Labovian variation are located in the choice of non-competing Vocabulary and the
220
interactions between Vocabulary Items and the ordered, iterative operations of the morphological component. Following Embick and Noyer (2001), operations such as lowering Merger refer to the hierarchical structure of morphosyntactic terminals and therefore apply early in the morphological computation, before the first round of Vocabulary Insertion and subsequent linearization of the terminals. As in Chapter 4 (and following McFadden 2004, Embick and Noyer to appear), Dissociated feature insertion is another structural operation that applies early in the morphological component. As discussed above, the operation of Fusion applies in order to ‘repair’ terminals in preparation for insertion of suppletive Vocabulary.20 Therefore, Fusion must refer to the features of Vocabulary Items retrieved by searching the Vocabulary inventory. I conclude (also following Kandybowicz to appear) that Fusion applies later in the morphological computation, after the first round of Vocabulary Insertion and subsequent linearization of the terminals. Moreover, I propose that Fusion (and other repair operations) can apply iteratively after search and insertion of Vocabulary. Iteration of operations will 20
Embick and Noyer (2001) propose another kind of late repair operation called ‘Local-dislocation Merger.’ for details including the mechanics of linearization, see also Embick (to appear-a).
221
continue until phonological exponents have been inserted into all of the abstract terminals in the structure. Notice that on this kind of theory, the application of morphological operations is determined by the feature structure of the Vocabulary Items in the inventory of an individual. Below, the proposed ordering of morphological operations is repeated from Chapter 3. (58) Ordering of morphological operations Spell Out Í
Structural operations Insertion of Dissociated features/morphemes, Lowering Merger
Í
Vocabulary search, Insertion
Í
Linearization of terminals
Í
Repair operations (iterated) Fusion, Local-dislocation Merger
Í
Vocabulary search, Insertion (iterated)
PF To see how this analysis of Labovian variation as noncompeting Vocabulary works, let us consider the morphosyntactic computation of the varying forms leveled weren’t and non-leveled wasn’t in the following simple examples. (59) a.
She weren’t scared. 222
b.
She wasn’t scared.
The structure below is spelled out to the morphological component from narrow syntax. I omit analysis of the Adjectival Phrase (AdjP) scared. The pronominal subject she has semantically interpretable 3s agreement features, and T has an interpretable feature [Past:+]. (60) Stage 1: Spell out to the morphological component TP 3 she[Pers:3, Num:s] T’ 3 T NegP 2 2 be T[Past:+] Neg AuxP 3 tbe AdjP 6 tshe ... Now structural operations apply. I assume that the English rule for Dissociated agreement applies first,21 copying the 3s features of the pronominal subject she to T[Past:+] (see Chapter 4). (61) Stage 2: Dissociated agreement T[Past:+] Æ
T[Past:+, Pers:3, Num:s]
21
This assumption is not trivial. Presumably there is ordering within the structural operations, perhaps in the repair operations as well. I will not elaborate on the matter here, leaving it for future work.
223
The structural operation Merger applies next, lowering Neg to [M [be] T
[Past:+, Pers:3, Num:s]].
22
(62) Stage 3: Lowering Merger of Neg and [M [M be] T ]23 T’ Æ 3 TM NegP 2 3 be T Neg AuxP 6 tbe ...
T’ 3 NegP 3 Neg M AuxP 2 5 T Neg tbe ... 2 be T
Now the Vocabulary inventory is searched and the first round of Vocabulary Insertion applies. The Vocabulary Item for a 3s feminine pronoun inserts the exponent she at this stage of the computation. Postponing further analysis of pronouns until Chapter 6, the following Vocabulary Items are retrieved by the initial Vocabulary Search. (63) Stage 4: 1st Vocabulary search [be, Past:+, Neg]
⇔
/wnt/
[be, Past:+, Num:s, Pers:2]
⇔
/w/
[be, Past:+, Num:p]
⇔
/w/
[be, Past:+] (elsewhere)
⇔
/wz/
[Neg]
⇔
/nt/ / [M [T[Past:±]] * __ ]
[Neg] (elsewhere)
⇔
/nat/
22
As mentioned above, I presume without explanation that this Merger operation is obligatory and not optional. 23 The features [Past:+s, Pers:3, Num:s] are omitted from these trees for reasons of space.
224
None of these Vocabulary Items can be inserted into the configuration of morphosyntactic terminals at this stage of the computation. Next, the hierarchical morphosyntactic structure is linearized, yielding the following ordering of terminals (for the detailed mechanics of linearization in DM see Embick to appear-a). (64) Stage 5: Linearization of terminals ...[M [[be] * T
[Past:+, Pers:3, Num:s]]
* Neg ]...
As above, I assume that Fusion must now apply to [M [be] * T
[Past:+, Pers:3, Num:s]]
to allow any insertion of English
Vocabulary. (65) Stage 6: Fusion of be and T[past:±] [M [be] * T
Æ
[Past:+, Pers:3, Num:s]]
[M be T
[Past:+, Pers:3, Num:s]]
The choice between non-competing Vocabulary Items arises at this stage. If the suppletive weren’t Vocabulary Item is chosen, Fusion applies to [M [be T
[Past:+, Pers:3, Num:s]]
* Neg ]. (66) Stage 7a: Fusion of [M be T [M [be T
[Past:+, Pers:3, Num:s]]
[M be T
[Past:+, Pers:3, Num:s]]
* Neg ]
[Past:+, Pers:3, Num:s]
and Neg
Æ Neg ]
This allows Vocabulary Insertion of the leveled form weren’t. (67) Stage 8a: Vocabulary Insertion of weren’t 225
[M be T
[Past:+, Pers:3, Num:s]
Neg ]
Í
/wnt/
All abstract terminals have now been supplied with phonological exponents, so morphological operations stop and the linearized phonological features are sent to the phonological component (PF) for still further computation (i.e., readjustment rules and other phonological operations). (68) Stage 9a: PF of weren’t .../wnt/... If the suppletive weren’t Vocabulary Item is not chosen, the Vocabulary Items given above insert the elsewhere form was along with the form n’t at the stilldistinct terminals for [M be T
[Past:+, Pers:3, Num:s]]
and Neg
(shown on separate lines for clarity). (69) Stage 7b: Vocabulary Insertion of was and n’t * Neg ]
[M [be T
[Past:+, Pers:3, Num:s]]
Í
/nt/
Í
/wz/
All abstract terminals have now been supplied with phonological exponents, so morphological operations stop and the linearized phonological features are sent to PF. (70) Stage 8b: PF of wasn’t ... /wz/ * /nt/...
226
5.4.6
Extension to ain’t and other leveled auxiliaries
It seems reasonable to conjecture that the weren’t variant arose by analogy to ain’t, rather than resulting from an agreement leveling process. This is also supported by the fact that ain’t is historically derived from aren’t. In other words, the leveled variant is the plural form, as with weren’t (and unlike was leveling).24 The DM-theoretical analysis of weren’t leveling can be extended straightforwardly to account for ain’t and other leveled auxiliaries with contracted -n’t. For example, Mittelstaedt’s dissertation (2006) shows that the presenttense auxiliaries be, have, and do are leveling with -n’t, evidently taking part in the socially-motivated process of concentration on Smith Island.
24
As pointed out by Jennifer Mittelstaedt (p.c.).
227
Generation Group Generation I b. 1923-1933 (4 persons) Generation II b. 1938-1958 (8 persons) Generation III b. 1960-1970 (5 persons) Generation IV b. 1973-1987 (13 persons) Totals (30 persons)
#weren’t/ #leveling environs %weren’t 27/41
#ain’t/ #leveling environs (be) %ain’t 4/5
#ain’t/ #leveling environs (have) %ain’t 1/5
#don’t/ #leveling environs %don’t 16/20
65.9% 16/31
80% 23/23
20% 7/10
80% 12/15
51.6% 7/24
100% 25/28
70% 3/7
80% 20/30
29.2% 30/33
89.3% 107/107
42.9% 35/37
66.7% 40/41
90.9% 80/129 62%
100% 159/163 97.5%
94.6% 46/59 77.9%
97.6% 88/106 83%
Table 5. Clitic-negated auxiliary leveling on Smith Island, MD (table adapted from Mittelstaedt 2006) This fact suggests that there is a single mechanism underlying all of these cases. The following Vocabulary would account for the Smith Island patterns of auxiliary leveling. (71) Postulated Vocabulary for Smith Island ain’t [be, Past:-, Neg]
⇔
/eint/
[have, Past:-, Neg]
⇔
/eint/
(72) Postulated Vocabulary Item for Smith Island don’t ⇔
[do, Past:-, Neg]
/dont/
For another example, in African-American English (AAE) (Green 2002, Martin and Wolfram 1998), both present- and past-tense do variably level to ain't (in addition to present-tense be and have). The following specimen exemplifies ain't leveling of present-tense do.
228
(73) This is my first summer [with] no trouble. I ain't go to jail for speeding. Didn't go to jail for DUI. I didn't break my foot. I didn't break my other foot. I'm one step ahead of the game already.25 The following Vocabulary Items would account for AAE patterns of ain't leveling. (74) Postulated Vocabulary for AAE ain’t
5.4.7
[be, Past:-, Neg]
⇔
/eint/
[have, Past:-, Neg]
⇔
/eint/
[do, Neg]
⇔
/eint/
Was leveling and non-competition of Vocabulary
One question remains. How do we handle (across-theparadigm) was leveling on this non-competing Vocabulary approach to Labovian variation? We might posit a totally underspecified Vocabulary Item for leveled was. (75) Hypothetical Vocabulary for was leveling (1) [be, Past:+]
⇔
/wz/
[be, Past:+, Num:s, Pers:2]
⇔
/w/
[be, Past:+, Num:p]
⇔
/w/
[be, Past:+]
⇔
/wz/
(elsewhere)
If this Vocabulary Item had to compete for insertion, it would be the default. To maintain the present approach to Labovian variation, we would have to stipulate non25
Attested, Kwame Brown quoted in the Washington Post, collected by the author 10/1/2005.
229
competition, perhaps as a kind of additional structure in the Vocabulary. Instead, it might be possible to derive the non-competition of Vocabulary by making the morphosyntactic features of Labovian variant Vocabulary Items identical to those of competing, non-variable Vocabulary Items. (76) Hypothetical Vocabulary for was leveling (2) [be, Past:+, Num:s, Pers:2]
⇔
/wz/
[be, Past:+, Num:p]
⇔
/wz/
[be, Past:+, Num:s, Pers:2]
⇔
/w/
[be, Past:+, Num:p]
⇔
/w/
[be, Past:+]
⇔
/wz/
(elsewhere)
Here the price is two Vocabulary Items for leveled was, rather than just one. It might be objected that this treatment misses a generalization about leveling. Another way to capture was leveling might be to simply adapt Adger and Smith’s approach to DM. Essentially, we could substitute their underspecified lexical items and morpheme spell-out rules for underspecified Vocabulary Items. The following Vocabulary Item, for the Buckie pattern of was leveling discussed above, will insert was at any past-tense be terminal with a positive value for Person and any value for Number (indicated here with s^p). (77) Hypothetical Vocabulary for Buckie was leveling
230
[be, Past:+, Num:s^p, Pers:+]
⇔
/wz/
[be, Past:+, Num:s, Pers:2]
⇔
/w/
[be, Past:+, Num:p]
⇔
/w/
[be, Past:+]
⇔
/wz/
(elsewhere)
To capture the more common pattern of was leveling across the paradigm, we need a Vocabulary Item that inserts was at any past-tense be terminal. (78) Hypothetical Vocabulary for was leveling (3) [be, Past:+, Num:s^p, Pers:±]
⇔
/wz/
[be, Past:+, Num:s, Pers:2]
⇔
/w/
[be, Past:+, Num:p]
⇔
/w/
[be, Past:+]
⇔
/wz/
(elsewhere)
These Vocabulary Items do not compete for insertion because their substantive morphosyntactic features are as specified as the most-specified competing Vocabulary Item. Of course, the price here is the non-trivial assumptions that the substantive morphosyntactic features of Vocabulary Items valued Num:s^p, Pers:+, or Pers:± are equivalent to those valued Num:s or Pers:2 with respect to competition for Vocabulary Insertion. At this point, we should consider adopting a different featural analysis of Person and Number φ features. For 231
example, Adger (2006) proposes a Minimalist lexical analysis of (Buckie) was leveling that differs in this respect from the analysis given by Adger and Smith (2005). Adger adopts the following binary-valued features for Person and Number. (79) Person [Participant:±][Author:±] (80) Number [Singular:±] These features combine as follows in the pronominal paradigm for English (adapted from Adger 2006). (81) 1s I
= [Sing:+, Part:+, Auth:+]
2s you
= [Sing:+, Part:+, Auth:-]
3s (s)he/it
= [Sing:+, Part:-, Auth:-]
1p we
= [Sing:-, Part:+, Auth:+]
2p you
= [Sing:-, Part:+, Auth:-]
2s they
= [Sing:-, Part:-, Auth:-]
Adger proposes a specific learning algorithm that constrains the acquisition of lexical items by reducing “optionality, synonymy, and the size of the lexicon” (2006). In order to acquire the set of English lexical items for past-tense T with uninterpretable φ features, the child uses this algorithm to analyze the distribution of agreement forms was/were; this yields lexical items for 232
past-tense T that are underspecified for Person and Number. Adger notes that his approach can be implemented in a DM framework, “with fully specified syntactic feature bundles, but underspecified Vocabulary Items” (2006). Thus, following Adger’s suggestion, we might hypothesize the following set of Vocabulary Items for past-tense be in Buckie.26 The first Vocabulary Item is most specified with the morphosyntactic features [Part:-, Sing:+], so it always wins the competition when T’s φ features are valued 3s [Sing:+, Part:-, Auth:-]. But the remaining Vocabulary Items have the same number of morphosyntactic features and therefore do not compete with each other. Homophonous exponents are numbered with a subscript for further reference. (82) Hypothetical Vocabulary for Buckie past-tense be [be, Past:+, Part:-, Sing:+]
⇔
/wz/1
[be, Past:+, Part:+]
⇔
/wz/2
[be, Past:+, Auth:+]
⇔
/wz/3
[be, Past:+, Auth:-]
⇔
/w/1
[be, Past:+, Sing:-]
⇔
/w/2
These Vocabulary will insert was/were exponents into the morphosyntactic terminal [be T[Past:+]] when T has
26
Thanks to David Adger (p.c.) for pointing this out.
233
particular φ-feature values, as illustrated below (the left arrow indicates Vocabulary Insertion). Because they are maximally underspecified for φ features, more than one Vocabulary Item can insert phonological features into the same terminal. For some φ-feature values of T, the same exponent can be inserted by different Vocabulary: for example, either of the two Vocabulary Items [be, Past:+, Part:+] ⇔ /wz/2 or [be, Past:+, Auth:+] ⇔ /wz/3 can insert its phonological features when T has the φ-feature values [Sing:+, Part:+, Auth:+] (either copied from, or in Agreement with, the 1s pronoun subject I). For other φfeature values of T, different exponents can be inserted by different Vocabulary: for example, either of the two Vocabulary Items [be, Past:+, Part:+] ⇔ /wz/2 or [be, Past:+, Auth:-] ⇔ /w/1 can insert its phonological features when T has the φ-feature values [Sing:+, Part:+, Auth:-] (either copied from, or in Agreement with, the 2s pronoun subject you). Notice that with these Vocabulary Items, the exponent was cannot be inserted when T has the φ-feature values [Sing:-, Part:-, Auth:-] (either copied from, or in Agreement with, the 3p pronoun subject they). This is how Buckie was leveling can be captured in such a theory. 234
(83) I...[be T[Past:+,
Sing:+, Part:+, Auth:+]]
you...[be T[Past:+,
Sing:+, Part:+, Auth:-]]
(s)he/it...[be T[Past:+, we...[be T[Past:+,
Sing:+, Part:-, Auth:-]]
Sing:-, Part:+, Auth:+]]
you...[be T[Past:+,
Sing:-, Part:+, Auth:-]]
they...[be T[Past:+,
Sing:-, Part:-, Auth:-]
Å /wz/2 OR /wz/3 Å /w/1 OR /wz/2 Å
/wz/1
Å /w/2 OR /wz/2 OR /wz/3 Å /w/1 OR /w/2 OR /wz/2 Å /w/1 OR /w/2
The general question of how to handle agreement leveling variation on a DM approach requires further research,27 and I will not attempt to resolve the issue here.
5.5
Conclusion
Let us again retreat for a final overview on the purpose of the dissertation, which is to identify some mechanisms and locations of Labovian variation in a Minimalist, DM theoretical framework. This chapter has examined a case of morphosyntactic variation and change in progress, namely 27
As mentioned above, my current research on these issues in collaboration with Andrew Nevins attempts to capture patterns of agreement leveling in English be with variable feature-deleting Impoverishment rules.
235
weren’t leveling on Smith Island. According to the DMtheoretical analysis proposed above, the mechanisms of weren’t leveling variation are located in the inventory and features of non-competing Vocabulary Items and their interaction with the ordered operations of the morphological component. These Vocabulary-based mechanisms of Labovian variation are in addition to mechanisms of variation located in the syntactic terminals (as proposed in Chapter 4, and similar to the analysis of Adger and Smith 2005 as reviewed above). On this DM analysis, leveled weren’t is analyzed as a suppletive Vocabulary Item whose exponent can only be inserted after Fusion of the be, T[Past:+], and Neg terminals during the morphological computation. Weren’t leveling takes place without concurrent were leveling because the weren’t Vocabulary Item lacks Person and Number agreement features. In other words, weren’t leveling is a ‘remorphologization’ of negation exponence as suggested by Schilling-Estes and Wolfram (1994), and not really an instance of agreement-leveling variation. Weren’t leveling only occurs with the reduced form of negation -n’t because the weren’t Vocabulary Item contains a Negation feature in its substantive morphosyntactic features. Head adjunction of [M [M be] T[past:+]] and Neg (whether by syntactic head
236
movement or morphological Merger) feeds Fusion of [M be T[past:+]] and Neg for insertion of weren’t. This analysis provides additional support for the late application of Fusion (Kandybowicz to appear), a morphological operation which modifies terminals to allow the insertion of suppletive Vocabulary Items. Because its morphosyntactic features are a subset of a target terminal only if Fusion applies, the Vocabulary Item for weren’t does not compete for insertion, and therefore can be a sociolinguistic choice for individuals. This explains the empirical distinction between allomorphy, (competition between Vocabulary Items) and Labovian variation (non-competition between Vocabulary Items). On Adger and Smith’s analysis, mechanisms of Labovian variation are located “within properties of lexical items,” so that social “patterns of variation seen across (groups of) individuals reduce to...lexical choice” (2005: 173). On the DM analysis proposed here, mechanisms of Labovian variation are located within properties of Vocabulary Items, and social patterns of variation reduce to choice of non-competing Vocabulary Items. Smith Islanders choose the weren’t Vocabulary Item with great frequency. Thus weren’t leveling is evidently part of the socially motivated
237
process of concentration on Smith Island, along with other concentrating phonological and morphosyntactic variants. The next chapter examines mismatched pronoun-case variation in English coordinates, where I will again claim that the mechanisms of Labovian variation are located in the inventory and features of non-competing Vocabulary Items.
238
Chapter 6: Distributed Morphological Mechanisms of Pronoun-Case Variation
6.
Introduction
In this chapter, we turn away from Smith Island to examine a
more (in)famous case: variable pronominal case-form
mismatches in coordinate phrases and other syntactic environments in English. Given the goals of the dissertation, variable pronoun-case mismatches can be seen as a productive point of contact between syntactic theory and the empirical study of Labovian variation. Despite its sociolinguistic salience, pronoun-case variation has attracted little attention from variationists. Methodological problems have hampered theoretical approaches, relegating pronoun-case variation to the margins of inquiry despite standard Minimalist featurechecking theories of Case. This chapter presents a caseless Distributed Morphology (DM) analysis of English pronouns, and argues again that significant mechanisms of Labovian variation must be located in the inventories and feature structures of non-competing Vocabulary Items. The chapter is organized as follows. Section 6.1 introduces the phenomenon of English pronominal case-form 239
mismatch in coordination, and presents some specimens of pronoun-case mismatches collected using a variationist observational methodology from Angermeyer and Singler’s (2003) study of 1s pronouns in object coordinates. Section 6.2 gives a theoretical perspective on pronoun-case mismatch, showing why it poses a serious problem for Minimalist feature-checking theories of Case before briefly reviewing some previous theoretical approaches to the problem. Section 6.3 proposes a totally case-less DM analysis of English pronouns and variable mismatch in coordinates, and Section 6.4 concludes with a summary of the chapter.
6.1
Salient and Stigmatized Sociolinguistic Variation
This section introduces the phenomena of pronominal caseform variation in English coordinates and other syntactic environments, and points out the salience and social significance of this variation. Adapting an observational methodology from Angermeyer and Singler’s (2003) variationist study of 1s pronouns in object coordinates, the section presents selected collected specimens of pronominal case-form mismatch in coordinates. Pronounspecific ordering asymmetries are observed, and the summary
240
of attested patters will form the empirical basis for the theoretical analysis given in this chapter.
6.1.1
Pronominal allomorphy in English
In most varieties of English,1 pronouns have a Subject Form (SF) and an Object Form (OF) allomorph, as illustrated in the following familiar paradigm (for a similar paradigm see e.g. Quirk and Greenbaum 1973: 102):2 (1)
English pronominal case forms Subject Form (SF)
Object Form (OF)
1s
I
me
3s
she/he
her/him
1p
we
us
3p
they
them
The distribution of these case forms is determined by the pronoun’s syntactic position. English pronouns appear in the SF when they are the subject of a tensed clause, and
1
This discussion does not consider pidgin and creole varieties. It would be interesting to do a systematic study of pronoun case in English (or, Germanic) lexified creoles. 2 I exclude 2s, 2p pronouns because they are homophonous for case form; for the same reason, I exclude the 3s Neut. pronoun it. Also, I do not address here the interesting question of Genitive Case, which is apparently marked on both pronouns and full NPs in English (making it ‘Morphologically Transparent,’ more on this concept below). See Bernstein and Tortora (2005) for a recent theoretical treatment of Genitive case morphology in English. Finally, I do not discuss variation in the case form of who/whom (see e.g. Lasnik and Sobin 2000 for a 'virus' theory account).
241
otherwise they appear in the OF. In other words, SF and OF pronouns seem to occur in complementary distribution. (2)
a.
We are a curse upon them.3
b. * Us are a curse upon they.4 It has been almost universally assumed in linguistics that pronoun case-form allomorphs in English are isomorphic with abstract syntactic Case features. On this view, SFs correspond to Nominative Case and OFs correspond to Accusative and Dative/Oblique Cases.5 Hereafter, following the standard convention, syntactic features are referred to as ‘Case,’ including Nominative (Nom), Accusative (Acc), Dative (Dat), and Oblique (Obl). The morphological forms of English pronouns are referred to as ‘case,’ including Subject Forms (SFs) and Object Forms (OFs).
6.1.2
Mismatched pronominal case forms in coordinates
English pronouns variably occur with mismatched case-forms in certain syntactic environments, most (in)famously coordination. For example, 1s OFs variably occur in Nom subject coordinates and 1s SFs variably appear in Acc and 3
“The descriptive linguists are a curse upon their race, who of course think that what the people say is the law.” John Simon, on PBS’s Do You Speak American? (http://www.pbs.org/speak/transcripts/1.html) 4 Throughout, unless explicitly stated otherwise, ‘*’ means ‘unattested’ and/or ‘unacceptable’. 5 Dative/Oblique Case is not discussed in this chapter. As will become apparent, this has no relevance for the analysis of English pronouns, since I claim that English lacks case (even in the DM sense). Matters are of course different in German and other transparent case languages.
242
Dat/Obl object coordinates. This is illustrated below with attested examples collected by author using a methodology adapted from Angermeyer and Singler (2003), as reported below. (3)
a.
[J.] and me were talking about that yesterday. Cf. J. and I were talking about that yesterday.
b.
It didn’t surprise Mark and I.6 Cf. It didn’t surprise Mark and me.
Pronominal case-form variation in coordinates is especially salient to all native speakers of English. Usage of mismatched case forms in coordination is harshly stigmatized and a perennial target of prescriptivist ire. Scornful social attitudes toward pronoun-case mismatch are illustrated in almost any prescriptive usage guide. Here is just one example taken virtually at random. [...] when a pronoun isn’t alone, instinct goes down the drain, and grammar with it. So we run into abominations like The odds were against you and I, although no one would dream of saying “against I.” (O'Conner 1996a: 11, italics and bold in original) This book’s title (Woe is I: The Grammarphobe’s Guide to Better English in Plain English) refers to pronoun-case variation in a different syntactic environment, postcopular nominals (a.k.a. predicate nominals, see Emonds 1986). The focus of this chapter is on pronoun-case
6
From an unscripted television interview show.
243
mismatch in coordination, but we will return briefly to post-copular nominals below. Prescriptivists claim that post-copular nominals are in the Nom Case and therefore that post-copular pronouns must appear in the SF. But OF pronouns are well attested in this environment, as illustrated below. It really is just him [...].7 Cf. ?? It really is just he.
(4)
For even more vitriolic condemnation of pronoun-case mismatches, see Garner (1998: 527, 345). See Loving (1990), Honey (1995), and Redfern (1994) for prescriptivist attempts to grapple with the undeniable ubiquity of pronoun-case mismatches in coordination, especially “hypercorrection” of SFs in Acc/Obl coordinates by elite speakers.8 Angermeyer and Singler (2003) provide additional examples from this genre, as does Emonds (1986).
6.1.3
Angermeyer and Singler’s observational methodology
Despite its remarkable salience and social significance, English pronoun-case variation has attracted very little
7
From an unscripted television interview show. Even prescriptivists confess to occasional mismatches in this environment. For example, O’Conner challenges her friends to “catch me in the occasional ‘between you and I’ (OK, I admit it),” but reassures her readers that this is “only a grammatical error, not a drive-by shooting” (O'Conner 1996b).
8
244
attention from variationists.9 A recent and welcome exception is Angermeyer and Singler (2003). However, they limit their study to 1s pronouns in object coordinates, categorizing the variants into three patterns (2003: 178): the ‘Vernacular’ (me and X), the ‘Standard’ (X and me), and the ‘Polite’ (X and I).10 Angermeyer and Singler point out one of the major reasons for the variationist neglect of case-form variation in English. Standard variationist data-collection methodologies are difficult to employ because coordinated pronouns appear rarely in spoken discourse, are difficult to elicit, and mismatches are socially stigmatized. Thus, Angermeyer and Singler conclude that the sociolinguistic interview, the basic unit of quantitative research in a Labovian paradigm, simply yields too few instances of co-ordinate NPs in object position to be an adequate source of data. (2003: 182) Angermeyer and Singler took steps to get around this problem. Instead of sociolinguistic interviews, two different methodologies were used to gather data. First, they designed a novel ‘sociolinguistic experiment’ to
9
Quattlebaum (1994) gives an interesting quantitative (but not variationist) study in an instructional setting, following Sobin (1994b). See Hudson (1995b) for some discussion from a sociolinguistic point of view. Most recently, Grano (2006a, 2006b) presents quantitative analyses based on corpus studies. 10 Myself variants (myself and X; X and myself) were also counted when they did not have a subject antecedent.
245
covertly elicit coordinated 1s pronouns using a kind of language game, which produced about a quarter of their data. Their second data-collection methodology “consisted simply of listening for occurrences of case-marked pronouns in co-ordinate NPs in object position in everyday speech” (Angermeyer and Singler 2003: 183).11 This observational methodology was adapted in order to collect specimens of mismatched pronominal case forms in coordinates, as reported in the next section below. Angermeyer and Singler provide a sociolinguistic variationist analysis of 1s pronouns in object coordinates that they collected using the two methods above. I do not report their quantitative results here, but to summarize very briefly, Angermeyer and Singler found that sociolinguistic factors, including the education and age of the speaker, had a statistically significant effect on the usage of these variants. This is expected given the social salience of pronoun-case variation, as noted above. Moreover, Angermeyer and Singler found no evidence that there is any change in progress, nor that the variation is a recent innovation. Rather, they conclude, “as they have apparently done for more than 400 years, the Vernacular,
11
Rickford et al. (1995) is cited as another variationist study that employs a similar methodology.
246
the Polite, and the Standard seem to be continuing in a dynamic state of stable ternary variation” (2003: 201).
6.1.4
Mismatch Specimens
As we will see in the following section, variation in the case form of English pronouns poses a serious mismatch problem for standard Minimalist feature-checking theories of Case (Chomsky 1995, 2000a, et seq.). An adequate theoretical analysis of such variation must account for the full range of attested pronominal case-form mismatches in coordinates (and other syntactic environments). However, Angermeyer and Singler (2003) only collected data on 1s pronouns in object coordinates. Thus, in order to prepare the ground for the DM-theoretical analysis given below, I adapted their observational methodology to collect specimens of coordinated pronoun-case mismatches that were observed in everyday speech and some writing over several years. Only specimens exhibiting case-form mismatch were collected, but these specimens were not restricted to 1s SF pronouns in object coordinates. Rather, collected specimens include 1s, 3s, and 3p OF pronouns in Nom subject coordinates, and 1s and 3s SF pronouns in Acc/Dat/Obl object coordinates. No quantitative or sociolinguistic analysis of these specimens is performed or intended.
247
Rather, the purpose of these specimens is to determine the attested range of mismatches in pronominal case forms, to determine all of the attested syntactic positions for coordinates where mismatches occur, and to provide attested examples of pronoun-specific ordering effects. To repeat for emphasis, no conclusions should be drawn about variant frequency on the basis of the following specimens. To begin, OF pronouns are well attested in Nom subject coordinates. (5)
a.
Me and her party!
b.
That’s why me and him still talk.
Mismatched pronominal OFs are attested in either conjunct, as exemplified below for 1s OFs. (6)
a.
I dreamed last night that you and me went on a canoeing trip.
b.
Me and my fellow researchers have tried to defend our research.12
Mismatched 3s OFs are also attested in either conjunct, regardless of the pronoun’s gender. (7)
Him and her have this gallery, on Lorimer.
(8)
a.
My sister and her don’t have any mutual friends.
b.
Her and Britney are trying to grow up.
a.
Him and the zombie hunter are fighting.13
b.
[D.R.] and him couldn’t share the same facility.14
(9)
12 13
From a written but unedited blog comment online. From a British speaker.
248
Coordinated plural pronouns are extremely rare in spoken and written discourse (see also Grano 2006a). This is apparently for pragmatic reasons: if there is enough context for plural reference, than the whole coordinate can be replaced by a plural pronoun. Nonetheless, mismatched 3p OFs are indeed attested in Nom subject coordinates. (10) a.
[S.] and them I guess are getting in tonight....
b.
When Castro and them took over, [....]15
c.
Her brothers and them was [standing] over there.16
I collected just one specimen of a mismatched 1p OF pronoun in a Nom subject coordinate, which occurs in the fourth of five conjuncts. Thus it cannot be determined from these specimens whether mismatched plural OF pronouns are also attested in the first conjunct of a coordinate. (11) The Times, the News and the Post (who reported that quote, incidentally), the Voice, us, and dozens of others keep and maintain their own news racks on the city streets.17 Proceeding, mismatched SF pronouns are well attested in both Acc and Dat/Obl coordinates. However, pronominal SFs exhibit a pronoun-specific ordering asymmetry that has been noted in all previous and current work on this topic
14
A spoken quotation from an online newspaper article. From a talk-radio guest, unscripted. 16 From a litigant on an unscripted television show. 17 From a presumably edited newspaper article that appeared online and in print. 15
249
(e.g., Emonds 1986, Sobin 1997, Angermeyer and Singler 2003, Grano 2006a, b). The mismatched 1s SF pronoun (I) is only attested in the second conjunct, and totally unattested in the first conjunct in these data. (12) a.
[...] he proceeded to start whipping [name] and I in the legs with the stick.18
b. * He proceeded to start whipping I and [name] in the legs with the stick. (13) a.
And if our troops do lose, it’s Night of the Living Dead for you and I.19
b. * It’s Night of the Living Dead for I and you. In contrast, the mismatched 3s SF pronouns (he/she) are attested only in the first conjunct, and are totally unattested in the second conjunct in these data. (14) a.
Dr. Mohammed Hazim in Baquba, pleaded for his governor to protect he and his colleagues from “organized terrorism of the police and army.”20
b. * Dr. Mohammed Hazim in Baquba, pleaded for his governor to protect his colleagues and he from “organized terrorism of the police and army.” (15) a.
He thought I was coming between he and his wife.21
b. * He thought I was coming between his wife and he. Notably, mismatched SF plural pronouns were totally unattested in Acc and Dat/Obl object coordinates in these data. Thus, it is not possible to determine from these 18
From an unscripted television interview show. From a written but unedited blog comment online. 20 From a presumably unedited article by the independent journalist Dahr Jamail, published online at http://dahrjamailiraq.com/. 21 From an unscripted television show. 19
250
collected specimens whether any conjunct ordering effects are attested with plural pronouns. Finally, mismatched 3s OF and 1s SF pronouns are well attested together, in both Nom subject coordinates and Acc/Obl object coordinates. All attested specimens exhibit the pronoun-specific conjunct-ordering asymmetry noted above. That is, the 1s SF pronoun (I) always appears in the second conjunct while the 3s OF pronoun (him/her) always appears in the first conjunct; other possible combinations are totally unattested in these data. (16) a.
Him and I were working at the time.22
b. * I and him were working at the time. c. * He and me were working at the time. d. * Me and he were working at the time. (17) a.
This is starting to make him and I both feel really bad.23
b. * This is starting to make I and him both feel really bad. c. * This is starting to make he and me both feel really bad. d. * This is starting to make me and he both feel really bad. (18) a.
I don’t like to talk about him and I developing chemistry.24
22
From an unscripted television show. From an unscripted television show. 24 A spoken quotation from a newspaper article published online and in print. 23
251
b. * I don’t like to talk about I and him developing chemistry. c. * I don’t like to talk about he and me developing chemistry. d. * I don’t like to talk about me and he developing chemistry.
6.1.5
Summary
This section has shown that pronoun-case variation is salient, well attested, and socially stigmatized in English. Adapting an observational methodology from Angermeyer and Singler (2003), I collected specimens of mismatched pronominal case-forms in coordinates. Based on these mismatch specimens as presented above, we can infer the following patterns of pronoun-case mismatch in coordinates. The patterns marked ‘*’ below were completely unattested in the collected specimens. As noted, coordinated plural pronouns are rare overall. Mismatched plural OF pronouns were unattested in the first conjunct, and mismatched plural SF pronouns were totally unattested. Setting aside the exceptional plural pronouns, notice that while (1s and 3s) OFs are attested in both conjuncts, the conjunct ordering asymmetries involve only the (1s and 3s) SFs. (19) Mismatched 1s OFs in Nom subject coordinates
252
a.
X and me
b.
me and X
(20) Mismatched 3s OFs in Nom subject coordinates a.
X and him/her
b.
him/her and X
(21) Mismatched 1s SFs in Acc/Dat/Obl object coordinates a.
X and I
b. * I and X (22) Mismatched 3s SFs in Acc/Dat/Obl object coordinates a. * X and she/he b.
she/he and X
(23) Mismatched 1,3p OF pronouns in subject coordinates a.
X and them
b.
Xn, us, and X
c. * us/them and X (24) Mismatched 1,3p SF pronouns in object coordinates a. * X and we/they b. * we/they and X This summary of attested mismatches is not intended to entail that the starred SF patterns cannot occur in a Nom subject coordinate, and these patterns occasionally do appear in edited writing. Angermeyer and Singler state in a footnote that the pattern I and X “cannot occur in object position and it shows up only rarely in subject position in
253
speech, but it does show up in subject position in writing” (2003: 203, fn. 17). For example, they point out that the I and X pattern is famously exemplified in the writing of William Labov: “[...] I and my associates have interviewed many thousands of speakers [...]” (2001: 6). Further research will be required to determine the attestation and usage frequency of the I and X, X and he/she, and plural pronoun patterns in usage. So far as I am aware, these patterns are not commonly attested.
6.2
Theoretical perspectives on case mismatch
The purpose of this section is to review a standard Minimalist theory of Case, point out that English pronominal case-form variation poses a serious mismatch problem on this theory, and briefly review a few previous theoretical approaches to this problem.
6.2.1
Case in Minimalist Syntax
Case theory was a central module of Government and Binding (Chomsky 1981, 1982) and Principles and Parameters (Chomsky and Lasnik 1993) theories of syntax. The role of Case was diminished somewhat in Minimalist theories following Chomsky (Chomsky 1995, 2000a et seq.), where Case is seen as an instance of general uninterpretable feature checking.
254
Regardless, in all permutations of these theories, abstract syntactic Case features are always present on DPs25 and active during the narrow syntactic computation, regardless of whether Case is realized morphophonologically in a particular language. Moreover, as mentioned above, it has been widely and standardly assumed that the morphological case forms of English pronouns are isomorphic with their abstract syntactic Case features. For example, a recent textbook on Minimalist syntax so interprets “the empirical fact that DPs may have different phonetic shape depending on the type of Case they bear” (Hornstein, Nunes and Grohmann 2005: 105). Let us review a particular Minimalist theory of Case in English, following Chomsky (2000a), Adger (Adger 2003), and Pesetsky and Torrego (to appear). English pronouns are syntactic terminals of the category D. Pronouns enter the syntax with semantically interpretable φ features of Pers:(1,2,3) and Num:(s,p), and interpretable Gender features (indicated here with ♂ and ♀). Like all DPs, pronouns also have a semantically uninterpretable Case feature that must be checked and valued by Agreement with a corresponding functional terminal prior to the LF
25
Determiner (D) heads may have Noun Phrase (NP) complements, the standard analysis following Abney (1987).
255
interface. Uninterpretable features and Agreement checking are notated as in the following schematic diagrams, repeated from Chapters 3 and 4. (25) Syntactic feature checking [u(ninterpretable)feature:value] (26) Agree in narrow syntax [PROBE[uF:
, uG: ]...[GOAL[F:α, G:β, H:γ]]]
AGREE Î [PROBE[uF:α,
uG:β]...[GOAL[F:α, G:β, H:γ]]]
Uninterpretable syntactic Case features are valued by Agreement feature checking on this theory. Thus, a DP’s Case feature is valued Nom by the probe of finite T, Acc by the probe of V26, and Dat/Obl by the probe of various Ps. Following Adger and Smith’s lexical Minimalist approach from the previous chapter, pronouns are spelled out at PF based on their valued syntactic Case features. It is important to notice that the isomorphism of syntactic Case and English case-form allomorphs follows directly from the features of lexical items on this kind of theory. Pronominal features are illustrated schematically below, with their phonological spell outs. (27) Nom/Acc Case-checking heads, pronouns, and spell outs 26
Or more accurately, little v (or whatever verbal head takes a direct object and has the subject DP Merged in its specifier). As mentioned in Chapter 3 and elsewhere, I abstract away from the internal structure of VP and TP in this dissertation.
256
a.
Nominative Case (Nom) =
T[Past:±]
b.
D[φ:1s, D[φ:3s, D[φ:3s, D[φ:1p, D[φ:3p,
uCase:Nom] ♂, uCase:Nom] ♀, uCase:Nom] uCase:Nom] uCase:Nom]
Æ Æ Æ Æ Æ
/ai/ /hi/ /i/ /wi/ /e/
Æ Æ Æ Æ Æ
/mi/ /hIm/ /h/ /s/ /m/
Accusative Case (Acc) =
V
D[φ:1s, D[φ:3s, D[φ:3s, D[φ:1p, D[φ:3p,
uCase:Acc] ♂, uCase:Acc] ♀, uCase:Acc] uCase:Acc] uCase:Acc]
To see how this theory of Case checking works, take a simple example for Nom case, given in bracket notation below (analysis of VP omitted). Finite T has uninterpretable φ features, and the pronoun Merged in the specifier of VP has an uninterpretable and unvalued Case feature. (28) He haunts this house. [TP T[Past:-, uφ: , EPP]... [VP D[φ:3s, ♂, uCase:
]
haunts this house]]
Applying the operation Agree, T’s uninterpretable φ features probe and find the interpretable φ-feature set of the goal pronoun D he. T’s φ features are checked and valued 3s. The pronoun’s Case feature is valued Nom and checked after being probed by finite T. Applying the operation Move, the pronoun raises to the specifier of T (its 257
pronounced position) in order to check T’s EPP feature. The structure can be spelled out to the interfaces with all uninterpretable features checked. At PF, the 3s ♂ pronoun with a Case feature valued Nom is spelled out as he. (29) He haunts this house. AGREE Î [TP T[Past:-, uφ:3s, EPP]... [VP D[φ:3s, ♂, uCase:Nom] haunts this house]] MOVE Î [TP D[φ:3s, ♂, uCase:Nom] T[Past:-, uφ:3s, [VP t haunts this house]] PF
=
D[φ:3s,
♂, uCase:Nom]
Æ
EPP]...
/hi/
This theoretical system cannot generate a OF pronoun in the subject of a tensed clause. The Case feature of a pronoun probed by finite T has been valued Nom, and therefore cannot be spelled out as the OF at PF. If the pronoun were somehow probed by V instead, valuing its Case feature Acc and allowing spell out as the OF at PF, then the uninterpretable φ features of finite T would be left unchecked to crash at the LF interface. Either way, the ungrammatical result is unacceptable and unattested. (30) * Him haunts this house. The system also cannot generate a SF pronoun in a verbal (or any other) object position. The Case feature of a pronoun probed by V has been valued Acc, and therefore 258
cannot be spelled out as the SF at PF. If the pronoun were somehow probed by finite T instead, valuing its Case feature Nom and allowing spell out as the SF at PF, then the uninterpretable Case feature of the subject DP would be left unchecked to crash at the LF interface. Either way, the ungrammatical result is unacceptable and unattested. (31) * This house destroyed he.
6.2.2
Pronoun-case variation as mismatch
Given the Minimalist theory of Case just sketched, which is quite standard in all relevant aspects, the pronoun-case variation discussed in the preceding section constitutes a straightforward but serious instance of morphosyntactic mismatch. Keeping just to 1s pronouns, such unexpected mismatches are extremely well attested: as we have seen, OF pronouns appear variably in either conjunct of subject coordinates. But this morphological case form does not correspond to the Nom-valued syntactic Case feature that pronouns in the specifier of finite T must have according to the theory. (32) a.
b.
Me and my fellow researchers have tried to defend our research. Cf. * Me has tried to defend our research. [J.] and me were talking about that yesterday. Cf. * Me was talking about that yesterday.
259
Conversely, SF pronouns appear variably in verbal direct (and other) object coordinates. But again, this morphological case form does not correspond to the Accvalued syntactic Case feature that pronouns probed by V must have according to the theory. However, 1s SF pronouns appear only in the second conjunct of Acc coordinates, but are unattested in the first conjunct. The opposite is true of 3s SF pronouns. This pronoun-specific conjunct ordering asymmetry is even more mysterious given a standard Minimalist feature checking theory of Case. (33) a.
That’s what fucked [J.] and I, the fours.27
b. * That’s what fucked I and [J.], the fours. c. * That’s what fucked I, the fours.
6.2.3
Previous Theoretical Approaches
Efforts to analyze pronoun-case mismatches in coordinates have remained at the margins of theoretical inquiry, largely because of methodological problems. Strong social stigma makes acceptability judgments unreliable for this very salient variation, and syntactic theorists have generally declined to employ the observational methods of variationists. This subsection briefly reviews a few previous theoretical approaches to pronoun-case variation, 27
Spoken to the author during a game of dominos.
260
establishing some necessary background for the DM analysis given in the next section.28
6.2.3.1
Deviant prestige constructions
In a little-known paper entitled “Grammatically deviant prestige constructions,” Emonds (1986) gave what I believe to be the first syntactic theoretical account of pronouncase mismatch in coordination. In fact, Emonds identified five syntactic environments where he claimed that OF pronouns are “normal” for English, but that the usage of “grammatically deviant” SF pronouns is prescribed and thus socially prestigious. These five environments are listed below in Emonds’s terms and with his examples (1986: 96). (34)
Deviant prestige constructions a.
Conjoined Subjects Mary and him/*he are late.
b.
Subjects of understood predicates Students smarter than her/*she get no scholarship.
c.
Predicate nominals It is just us/*we who John says are late.
28
The most comprehensive recent investigation of pronoun case is Quinn’s (2005) book The Distribution of Pronoun Case Forms in English. However, Quinn used questionnaires to collect acceptability judgments, a methodology that I believe is unavoidably compromised in this instance by the strong social stigma attached to usage “errors” in pronoun case. She moreover gives a constraint-based theoretical account which I do not review here, since my focus is on Minimalist theories.
261
d.
First person demonstratives Us/*we commuters are often blamed for smog.
e.
Appositives to subjects Judy thinks that the best math student, namely her/*she, ought to get a scholarship.
Emonds points out that mismatched pronominal (or any) case forms do not appear in German, where unlike English, all DPs are morphologically case marked. On the basis of this observation, he claims that the morphological form of English pronouns is not a realization of their syntactic Case features. According to Emonds, syntactic features can be the source of a morphological distinction only if those syntactic features are “Morphologically Transparent,” as defined below. (35) Morphological transparency (Emonds 1986: 106-107) Definition. A syntactic category C [Case] is “Morphologically Transparent” on B [a lexical item] if and only if a productive number of pairs of simple B which contrast with respect to C also differ phonologically. An additional constraint on acquisition prevents English-acquiring children from realizing non-transparent Case features on their lexical items. (36) Transparency constraint (Emonds 1986: 106-107) An abstract (e.g. case) feature C of a category B is realized on the lexical head of B in a language if and only if the C is Morphologically Transparent on B.
262
On Emonds’ analysis, when English lost productive case marking as result of independent phonological changes in DP (Allen 1995), syntactic Case features became morphologically non-transparent on DP. However, case marking on the closed set of pronouns was suppletive, and thus was unaffected by the phonological change. Syntactic Case features could no longer be responsible for the distribution of morphologically distinct pronoun case-form allomorphs, so the distinction must be maintained by another mechanism. Using the theoretical machinery available at the time, Emonds proposes that a “local transformation” took over this function in English. English pronouns can appear in the SF only when they are “an immediate constituent of a sentence (S) which contains an inflected verbal element” (1986: 97). Therefore, the local transformation rule Emonds proposes for English will yield a SF only if the pronoun is directly governed by Inflection; otherwise, the pronoun will appear in the OF. Assuming that the phrase structure of coordination blocks government of a internal conjunct by Inflection, this rule correctly generates the normal distribution of pronominal OFs in all of Emonds’s prestige constructions above. Emonds’s crucial insight is that the distribution of SF
263
pronouns in English can be described as extremely local to finite tense, with the OF appearing in all other syntactic environments. For Emonds, socially prescribed usage of SF pronouns in any syntactic environment non-local to Inflection is deviant. In other words, these SFs are not generated by the grammar, and Emonds does not propose any mechanisms to account for such forms. Moreover, he does not propose any mechanisms to account for mismatched SF pronouns in Acc/Dat/Obl object coordinates, regarding this as “overcorrection” caused by the deviant use of SFs in Nom subject coordinates. As Angermeyer and Singler correctly observe, both Emonds and Johannessen (1998) (see below) “seemed to feel that labeling something ‘overcorrection’ freed one from the obligation of having to provide a formal account of it” (2003: 201, fn. 9). Naturally, this criticism applies to any proposed theoretical account of these phenomena: mechanisms of case-form mismatch and the observed pronoun-specific ordering asymmetries must be provided.
6.2.3.2
Extraordinarily unbalanced coordination
According to Angermeyer and Singler (2003: 175), Johannessen’s (1998) book Coordination provides “the most
264
thorough of the formal treatments” of pronoun-case mismatch in coordinates.29 Johannessen’s book deals with a wide range of phenomena involving coordination, for a number of languages. However, a significant amount of her data and analysis involves case-form mismatch in English coordinates. Regarding this phenomenon as parametric variation, Johannessen distinguishes two types of case-form mismatch in coordinate structures. The first she calls “Unbalanced Coordination” (UC), where just one conjunct seems to have “deviant” (i.e. mismatched) Case features. Note that Johannessen standardly assumes the isomorphism of pronoun case forms and their abstract syntactic Case features. However, she furthermore assumes that the DP in the first conjunct has non-mismatched syntactic Case features, even though English lacks morphological case on non-pronominal DPs. Johannessen (1998: 15) provides an attested example of UC, as given below. (37) Unbalanced Coordination (UC) = Can someone help my wife and I find housing in Texas...? The second type is called “Extraordinarily Balanced Coordination” (EBC), where both conjuncts have the same
29
See also Zoerner (1996) for a similar treatment involving the phrasal structure of the coordinate phrase.
265
mismatched Case features. Johannessen (1998: 62) provides a concocted but not attested example of EBC, as given below. (38) Extraordinarily Balanced Coordination (EBC) = Them and us are going to the game together. For Johannessen, coordination is a functional head Conjunction (Co0) that projects a standardly asymmetrical Xbar (X’) phrase structure, Conjunction Phrase (CoP). The first conjunct is an XP in the specifier of Co0, while the second conjunct is a YP complement of Co0.30 (39) Conjunction Phrase (CoP) CoP 3 XP Co’ 3 Co0 YP Johannessen claims that the facts of English pronouncase mismatch (her UC and EBC) can be explained by a rather confusing welter of mechanisms interacting with the structure of CoP. Her analysis of English requires mechanisms of Case assignment, Case checking, default Case licensing, and Case “overcorrection” (as well as additional mechanisms needed to account for phenomena in other languages). Besides being inconsistent with current Minimalist theoretical approaches, Johannessen’s analysis
30
See Munn (1994) for more on the X’ syntax of coordination, or cf. Goodall (1987) for a very different “parallel structures” approach.
266
is self contradictory and makes the wrong empirical predictions for English. Summarizing very roughly, Johannessen claims that that there is a “CoP Sensitivity” parameter. When this parameter is set to +CoP Sensitive, the CoP is a barrier to the assignment of outside Case features. Thus, default Case (Acc for English) is licensed on both conjuncts. This yields EBC, with OFs in both conjuncts of coordinate subjects and objects. Johannessen (1998: 120-123) claims that default case is only licensed by Co0 in its complement, where Case is not assigned or checked. But it is not clear why default Acc case can be licensed on the specifier of Co0 in these +CoP Sensitive EBC cases. When the parameter is set to -CoP Sensitive, CoP is a not a barrier to the assignment of Case features from outside. Co0 receives outside Case features, but can only check these features on the first conjunct in its specifier. Default Case is licensed on the second conjunct in its complement. This is supposed to yield UC in coordinate subjects and objects. However, Johannessen claims that overcorrection in English is possible “only where the grammar seems to lack the possibility of direct checking (as when there is no case or default case)” (1998: 123-126). In such instances, the Co0 assigns Nom Case to its
267
complement. But it is not clear why overcorrection allows the Co0 head to assign Nom Case to its complement if default case is already licensed on the complement of Co0. Based on these mechanisms, Johannessen makes several empirical predictions about possible coordinated pronoun case-form mismatches in English. For example, she claims that there are no occurrences of UC where both conjuncts are pronouns and only the first conjunct has mismatched case (1998: 63). But her theory seems to predict that this would be possible in a subject coordinate: with the parameter set to +Cop sensitive, the first conjunct gets default Acc Case and overcorrection assigns Nom case to the second conjunct. Indeed, as we saw above, such specimens are well attested when the second conjunct is the 1s SF pronoun I. (40) Him and I were working at the time. However, with the same configuration of mechanisms, Johannessen’s theory predicts that such a coordinate should be possible even if the second conjunct were a 3s SF pronoun, apparently contrary to fact. (41) a. * Me and he were working at the time. b. * Him and he were working at the time. In general, because her analysis refers only to abstract syntactic Case features, it cannot account for the
268
pronoun-specific ordering asymmetries discussed above. For example, Johannessen furthermore claims that it is impossible for SF pronouns to appear in both conjuncts of an object coordinate. In such a structure, the second conjunct could receive Nom Case from the Co0 head via overcorrection. If the parameter is set to +Cop Sensitive, the first conjunct receives outside Acc Case; if the parameter is set to -Cop Sensitive, default Acc Case is licensed on the first conjunct. Either way, the prediction is clear enough. Johannessen gives the following judgments, which she confirmed with native-speaker informants (1998: 69, fn. 8).31 (42) a. * Gramps will kiss he and she. b. * Gramps expected he and she to sing. Acceptability judgments for coordinated pronouns have a serious risk of distortion because of the social stigma attached to mismatches, as noted above. Nevertheless, Johannessen herself provides the following attested “exception” to this prediction (1998: 63, #112j).32 (43) Joe wouldn’t explain it to [she and I]. In general, the prediction is apparently disconfirmed whenever the second conjunct is the 1s SF
31 32
She cites Schwartz (1985: 167) as the source for these examples. Also noted by Angermeyer and Singler (2003: 176, 202 fn. 11).
269
pronoun I. Johannessen’s starred examples seem quite a bit more acceptable if the second pronoun is I: (44) a. % Gramps will kiss he and I. b. % Gramps expected he and I to sing. Johannessen furthermore claims that a subject coordinate cannot occur with an OF pronoun in the first conjunct, giving the following judgments from her informants.33 (45) a. * Him and she will drive to the movies. b. * It would be better if her and he drove to the movies. But again, as we saw directly above, such specimens are well attested when the second conjunct is the 1s SF pronoun I, and this certainly improves the acceptability of both examples. (46) a. % Him and I will drive to the movies. b. % It would be better if her and I drove to the movies.
6.2.3.3
Grammatical viruses
Sobin (1994a, 1994b) and associates (Quattlebaum 1994) were virtually the only linguists to conduct research on pronominal case-form mismatches in coordinates and other syntactic environments after Emonds (1986). This research 33
Again citing Schwartz (1985: 165) as the source for these examples.
270
led to Sobin’s (1997) proposal of “grammatical virus” theory to account for pronoun-case mismatch in coordinates. Sobin also used virus theory to analyze other variable phenomena such as rightward agreement in expletive constructions (but cf. Schütze 1999, who argues that "expletive constructions are not infected"), and in later collaborative work (Lasnik and Sobin 2000) the virus theory was further extended to account for variation between who and whom. Viruses are claimed to be extra-grammatical rules which can “infect” a host grammar and thus “subvert output in that they license sentences which are ungrammatical in the strict sense” (Lasnik and Sobin 2000: 352). Sobin’s (1997) analysis of English pronouns in coordinates essentially follows Emonds (1986), but provides additional virus mechanisms to account for prestigious and overcorrected SF pronouns in coordinates. Sobin stipulates that ordinary syntactic Case-checking mechanisms cannot apply inside coordinates in English. In a footnote, he speculates (following Emonds 1986) that the transparency of case morphology in German and other languages somehow allows Case to be checked inside of coordination. But Sobin does not explain exactly how such mechanisms would work and calls for further research on the problem. He proposes a default Acc case-checking rule that is specific to English,
271
but not itself a virus. This rule is intended to explain the appearance of OF pronouns in either conjunct of a subject or object coordinate (Sobin 1997: 336). (47) The Default Accusative Rule If: ...[NP ACC]... 1 then:
check ACC on 1.
In order to account for the mismatched 1s SF pronoun I, which appears only in the second conjunct as we have seen above, Sobin proposes that the following extragrammatical virus rule can check Nom Case on a 1s pronoun when it follows and (1997: 336). This virus is intended to yield and I in both subject and object coordinates. (48) The “...and I...” Rule ...and [Prn +1, +sg, NOM]... 1 2
If: then:
check NOM on 2.
In order to account for 3s SF pronouns in coordinates, Sobin proposes another virus rule that checks Nom Case on the 3s pronoun when it follows the finite clause complementizer that, which can be phonologically null and still trigger the rule (1997: 336). (49) The “that she” Rule If:
...that [Prn +3, +sg, NOM]... 1 2
272
then:
check NOM on 2.
Notice that this virus is intended to yield 3s SFs in the first conjunct of subject coordinates, and cannot explain mismatched 3s SFs in the first conjunct of object coordinates, where there is no that. Sobin does not offer any explanation for such well-attested specimens, other than to speculate that the ‘and I’ virus might somehow “generalize” to 3rd person. (50)
He thought I was coming between he and his wife. Sobin also proposes a virus rule for certain cases of
apparently mismatched but socially prescribed SF pronouns used as post-copular nominals. This virus is intended to yield the prestigious phrase It is I, but it does not check Nom Case in any other sentences. Thus, the default Acc Case checking rule will yield OFs in other post-copular nominal constructions, for example That was just him/*he (1997: 337). (51) The “it is I” Rule If:
...that [Prn +3, +sg, NOM]... 1 2
then:
check NOM on 2.
In my view, serious conceptual and theoretical problems doom the virus theory.34 First, virus rules are
34
See also Parrott (2001b) for earlier criticism of virus theory.
273
external to the ordinary mechanisms of the grammar. This is a rather extraordinary claim that would seem to require correspondingly empirical evidence. The properties attributed to viruses by Sobin (1997) and Lasnik and Sobin (2000) include lexical specificity and linear ordering effects, but these are just the observed empirical properties of the phenomena, such as pronominal case-form mismatches in coordinates, that virus theory is meant to explain. Thus there is no independent empirical motivation for these extra-grammatical virus mechanisms. Second, there are no obvious cross-linguistic predictions made by virus theory. There are no apparent connections between virus mechanisms and the syntactic environments they can infect, nor between viruses and the languages they can infect. Moreover, there are no Minimalist motivations for extragrammatical virus mechanisms. Viruses are certainly not motivated by interface conditions, conceptual necessity, or computational efficiency. Finally, the actual mechanisms of the virus theory are unclear and likely inconsistent with Minimalist theories. For example, Lasnik and Sobin claim that virus rules can check syntactic features “at a point in the derivation where the order of overt elements is fixed, most likely at ‘spellout’”
(2000: 355). But
following Chomsky (2000, et seq.), Spell Out is a syntactic
274
operation that delivers feature sets to the LF and PF interfaces. Uninterpretable features such as Case must be checked and valued prior to reaching the interfaces, or they cause the derivation to crash. On such a theory, it is very difficult to see how additional virus rules could check features ‘at’ Spell Out, which is an operation and not a level of representation. Viruses could not check uninterpretable features after Spell Out, since the derivation would already have crashed. And if viruses check features during the narrow syntax, their crucial sensitivity to linear ordering is unexplained.
6.2.3.4
Viral infection with DM default case
Schütze (2001) adopts Sobin’s virus analysis for SF pronouns in coordinates, including “overcorrection” of SF pronouns in Acc/Dat/Obl object coordinates. However, instead of a default Case checking rule, Schütze uses the theory of DM (Halle and Marantz 1993) to account for OFs as post-syntactic default case. At PF, the pronominal Vocabulary Items of English can insert elsewhere OF forms into pronouns whose Case features have not been checked and valued in the narrow syntax. By stipulation, Case is not checked in a range of syntactic environments including coordinates, post-copular nominals, and the other
275
environments first identified by Emonds (1986). Schütze does not explain why these environments prevent normal Case checking. The DM portion of his analysis is a step in the right direction, and below it will also be claimed
that
OFs are the default Vocabulary Items for English pronouns. However, Schütze’s analysis still must rely on the virus theory to account for mismatched and coordinated SF pronouns. Thus, his approach ultimately suffers from all the flaws of the virus theory it adopts.
6.3
Distributed Morphological Mechanisms
This section presents a DM-mechanistic analysis of pronominal case-form mismatches in coordination and other syntactic environments. From this point on, I adopt a Minimalist syntax mostly following Chomsky (2000), augmented by a DM theory of lexical primitives and the morphological component mostly following Embick and Noyer (to appear).
6.3.1
A DM theory of case
Following the DM approach as outlined in Chapters 4 and 5, I assume that there are no uninterpretable Person, Number, or Case features in the narrow syntactic computation. ‘Dissociated’ case features are assigned by language-
276
specific rules that operate during the morphological component, after Spell Out. McFadden (2004) provides a comprehensive DM theory of dissociated case in German, Icelandic, and other languages with productive case morphology on all DPs. However, McFadden (to appear) proposes that the morphological case rules of English differ from those of German and similar case languages. In English, dissociated Nom case features are only assigned to pronouns that are “maximally close to finite T,” and Acc/Obl features are assigned by default when this structural condition does not hold. This is similar but not identical to the default-case analysis proposed by Schütze (2001) and briefly discussed above. McFadden’s analysis of English explains the appearance of mismatched OF pronouns in subject coordinates and other environments with intervening material between the pronoun and finite T. However, it is unclear how such a defaultcase analysis would account for mismatched SF pronouns in object coordinates, or the pronoun-specific conjunctordering asymmetries. McFadden attributes SF mismatches and ordering effects to “prescriptive pressure” resulting in an “artificially archaic distribution of the forms,” but does not specify the responsible mechanisms. As above, Schütze’s (2001) default-case analysis must appeal to dubious and
277
extra-grammatical virus mechanisms (Sobin 1997, Lasnik and Sobin 2000) to account for these facts. Neither McFadden’s nor Schütze’s default-case analysis explains why the English and German case systems are so different, or makes cross-linguistic predictions about the kind of case-form mismatches and asymmetries observed in English.
6.3.2
A Case-less DM analysis of English pronominal forms
I propose an analysis that takes the DM approach to case one step further. Essentially following Emonds (1986), I deny that English even has morphological case assignment rules like those in German and other case languages. On the DM theory of case, dissociated morphological case assignment rules are language-specific and not endowed by UG. Therefore, case assignment rules and any Vocabulary Items that refer to case features must be acquired on the basis of linguistic environmental input. Again following an insight from Emonds (1986), I hypothesize that the difference between English and German is the relative ‘transparency’ of case morphology on D. In German, enough of all DPs are phonologically distinguished according to case so that the developing child has sufficient environmental evidence from which to acquire dissociated case assignment rules and corresponding Vocabulary Items.
278
In English, only the pronouns remain phonologically distinguished as the historical remnant of long-lost transparent morphological case. By hypothesis, this handful of distinctive pronoun forms in English is not sufficient evidence from which to acquire dissociated case assignment rules and Vocabulary such as in German. Thus, in order to acquire the complementary distribution of pronominal Subject and Object Forms in English, the developing child must instead learn a very different set of Vocabulary Items. On this analysis, the exponent of English pronouns is determined by the contextual information of Vocabulary Items, and not by the assignment of morphological (or syntactic) case features. As schematized below, SF exponents can be inserted if the pronoun D is itself the specifier of finite T during the morphological component; elsewhere, OF exponents are inserted. In other words, pronominal OFs really are the default in English, but these forms are not the result of default case assignment mechanisms in the senses of Johannessen (1998), Sobin (1997), or McFadden (2004, to appear). This is essentially the analysis of Emonds (1986) implemented using DM mechanisms. Because the analysis lacks syntactic Case or
279
grammatical viruses, it is similar to but not identical to Schütze’s (2001) DM default-case analysis. (52) Schematic Vocabulary for English pronouns [D, Pers:_, Num:_]
Ù
/SF/
/
[D, Pers:_, Num:_]
Ù
/OF/
elsewhere
[TP __ [ T[Past:±] ...]]
The English Vocabulary Items for 1s and 3s pronouns are given below (plural pronouns are not given here, but are as in the schematic Vocabulary above). (53) Vocabulary Items for English 1s pronouns [D, Pers:1, Num:s]
Ù
/ai/
/
[D, Pers:1, Num:s]
Ù
/mi/
elsewhere
[TP __ [ T[Past:±] ...]]
(54) Vocabulary Items for English 3s pronouns [D, Pers:3, Num:s, ♂] Ù /hi/ / [TP __ [ T[Past:±] ...]] [D, Pers:3, Num:s, ♂] Ù /hIm/ elsewhere [D, Pers:3, Num:s, ♀] Ù /i/ / [TP __ [ T[Past:±] ...]] [D, Pers:3, Num:s, ♀] Ù /h/ elsewhere This analysis explains why apparently mismatched OFs can occur in either conjunct of a subject coordinate. A subject Coordinate Phrase (CoP) itself is the specifier of finite T, as illustrated.
280
(55) CoP and finite T TP 3 CoP T’ 6 3 D and D T[Past:±] VP 5 ... Any pronoun inside of a CoP is either the specifier of the coordinate head Co or its complement, as illustrated below. Therefore, the Vocabulary for English pronouns cannot insert SF exponents inside of coordinates, regardless of the CoP’s syntactic position. Pronouns in any conjunct of a coordinate will take the elsewhere OF by default. (56) Structure of CoP CoP 3 D Co’ 3 D Co This analysis moreover applies straightforwardly to all the syntactic environments identified by Emonds (1986), whose only common property is that none are the specifier of finite T. As predicted, default OF pronouns appear in all of these environments, for example the post-copular nominal position mentioned above. (57) It was just him.
281
Furthermore, this analysis explains why SF exponents can be inserted when a TP adjunct intervenes between the pronoun and finite T. (58) He sometimes wails. As mentioned in Chapter 3, it has long been observed that VP adjuncts do not interfere with the morphological lowering of finite T to full verbs in English, but nonadjuncts like negation prevent this operation and thus require insertion of do. (59) a.
He often wails.
b. * He not often wails. c.
He does not often wail.
To account for this and other such cases where adjuncts apparently do not interfere with morphological operations, Bobaljik (1995) proposed a relation of morphosyntactic Adjacency that combines both linear and hierarchical information available in the morphological component. However, Embick and Noyer (2001) eliminated this relation from DM theory by allowing the lowering Merger operation to refer to hierarchical structures during the early stages of the morphological computation, before linearization. On their analysis, English finite T can simply be lowered to the head of its VP complement. Since it is not in the complement of finite T, a VP adjunct is
282
structurally ‘invisible’ to the lowering Merger operation. NegP is not a VP adjunct but is itself the complement of T, explaining why negation requires do support with full verbs in English. The present analysis of English pronominal Vocabulary does not have to rely on Bobaljik’s concept of morphosyntactic Adjacency.35 Instead, following Embick and Noyer (2001), the contextual features of postulated SF Vocabulary Items refer to hierarchical syntactic structures, which are the input to the morphological component. Although it evidently disrupts the ‘maximum closeness’ of the pronoun and finite T, a TP adjunct is structurally invisible to pronominal SF Vocabulary. Therefore, the pronoun is still in the specifier of T[Past:±] and the SF exponent can be inserted. (60) Pronoun, TP adjunct, and finite T TP 3 D TP 3 Adj T 5 3 sometimes T[Past:-] VP 6 wails 35
An earlier version of the analysis (see e.g. Parrott 2006) used the relation of morphosyntactic Adjacency in English pronominal Vocabulary. This was somewhat problematic because the supplementary Vocabulary Items for mismatched SF pronouns in coordinates, introduced immediately below, crucially refer to linear adjacency with the coordinate head and.
283
6.3.3
Supplemental pronoun Vocabulary
So far, the proposed analysis will account for the allomorphy of non-coordinated SF and OF pronouns, as well as for OF pronouns in either conjunct of a subject or object coordinate. The analysis also accounts for normal OF pronouns in all of Emonds’s “prestige constructions” (see above). We must now consider mismatched SF pronouns in object coordinates, as well as the well-attested pronounspecific conjunct-ordering asymmetries that are when mismatched SF pronouns appear in object coordinates.. As discussed above, Are SF pronouns in coordinates simply deviant, the result of viral mechanisms external to the ordinary mechanisms of syntax and morphology? I would like to propose a different analysis. English speakers may--but need not necessarily--learn supplementary Vocabulary Items for pronouns. These supplemental Vocabulary are not acquired during the period of language development along with the ordinary Vocabulary for pronouns as given above. Instead, supplementary Vocabulary are learned because of their social prestige, either through explicit prescriptive instruction or from frequent and socially salient environmental exposure (see Grano 2006a, b for discussion of the relationship between frequency and
284
prescription). Crucially, the contextual information in these supplementary Vocabulary Items refers to a linearly adjacent coordinate head and, and not to the hierarchical position of finite T. Therefore, supplemental Vocabulary can insert specific SF exponents for pronouns in both subject and object coordinates, accounting for SF mismatches and pronoun-specific ordering asymmetries. Supplementary Vocabulary Items are given below for 1s and 3s pronouns. As in previous chapters, linear adjacency is indicated with ‘*’ following Embick (to appear-a). (61) Supplemental Vocabulary for English 1s and 3s pronouns a.
[D, Pers:1, Num:s]
Ù
/ai/
/
[ænd] * __
b.
[D, Pers:3, Num:s, ♂]
Ù
/hi/
/
__ * [ænd]
c.
[D, Pers:3, Num:s, ♀]
Ù
/i/
/
__ * [ænd]
By hypothesis, these supplementary Vocabulary Items do not compete with ordinary pronominal Vocabulary. Their contextual features are no more specified than the contextual features of the Vocabulary for SF pronouns. However, an important difference is that the supplementary Vocabulary refer to liner adjacency, and therefore must be inserted after linearization of terminals during the morphological component. As in Chapter 5, then, noncompletion of Vocabulary accounts for the difference 285
between Labovian variation and allomorphy on this DM approach. Furthermore, English pronominal case-form variation between individuals can arise from differing inventories of Vocabulary Items on this analysis. First, an individual’s Vocabulary inventory can contain no supplementary SF pronoun Vocabulary Items whatsoever, as illustrated below (plural pronouns omitted, see schematic Vocabulary above). (62) English pronoun Vocabulary, Inventory I [D, Pers:1, Num:s]
Ù /ai/ / [TP __ [ T[Past:±] ...]]
[D, Pers:1, Num:s]
Ù /mi/ elsewhere
[D, Pers:3, Num:s, ♂]
Ù /hi/ / [TP __ [ T[Past:±] ...]]
[D, Pers:3, Num:s, ♂]
Ù /hIm/ elsewhere
[D, Pers:3, Num:s, ♀]
Ù /i/ / [TP __ [ T[Past:±] ...]]
[D, Pers:3, Num:s, ♀]
Ù /h/ elsewhere
An individual with pronoun Inventory I is predicted to use OF pronouns, but not SF pronouns, in subject and object coordinates. Next, an individual’s Vocabulary inventory could contain only the supplementary ‘and I’ Vocabulary Item but no other SF supplements. As in previous chapters, noncompetition of Vocabulary Items is indicated below with a dotted-dashed line.
286
(63) English pronoun Vocabulary, Inventory II [D, Pers:1, Num:s]
Ù /ai/ / [ænd] * __
[D, Pers:1, Num:s]
Ù /ai/ / [TP __ [ T[Past:±] ...]]
[D, Pers:1, Num:s]
Ù /mi/
[D, Pers:3, Num:s, ♂]
Ù /hi/ / [TP __ [ T[Past:±] ...]]
[D, Pers:3, Num:s, ♂]
Ù /hIm/ elsewhere
[D, Pers:3, Num:s, ♀]
Ù /i/ / [TP __ [ T[Past:±] ...]]
[D, Pers:3, Num:s, ♀]
Ù /h/ elsewhere
elsewhere
An individual with pronoun Inventory II is predicted to use OF pronouns in any conjunct of a subject or object coordinate. It is also predicted that such an individual can optionally use a 1s SF pronoun I in the second conjunct, but never the first conjunct, of a subject or object coordinate. Individuals with pronoun Vocabulary Inventory II can use her and I in a subject or object coordinate, but cannot use *I and her because her supplementary Vocabulary only inserts I in a coordinate if it is right adjacent to and. Such an individual also cannot use *she and me because she lacks supplemental Vocabulary for the 3s pronoun in a coordinate. Next, an individual’s Vocabulary inventory could contain both of the optional ‘and I’ and ‘(s)he and’ supplemental Vocabulary Items.
287
(64) English pronoun Vocabulary, Inventory III [D, Pers:1, Num:s]
Ù /ai/
[D, Pers:1, Num:s]
Ù /ai/ / [TP __ [ T[Past:±] ...]]
[D, Pers:1, Num:s]
Ù /mi/
elsewhere
[D, Pers:3, Num:s, ♂]
Ù /hi/
/
[D, Pers:3, Num:s, ♂]
Ù /hi/ / [TP __ [ T[Past:±] ...]]
[D, Pers:3, Num:s, ♂]
Ù /hIm/ elsewhere
[D, Pers:3, Num:s, ♀]
Ù /i/
[D, Pers:3, Num:s, ♀]
Ù /i/ / [TP __ [ T[Past:±] ...]]
[D, Pers:3, Num:s, ♀]
Ù /h/ elsewhere
/
/
[ænd] * __
__ * [ænd]
__ * [ænd]
An individual with pronoun Inventory III is predicted to use OF pronouns in any conjunct of a subject or object coordinate. She can optionally use a 1s SF pronoun in the second conjunct, and/or a 3s SF pronoun in the first conjunct, of a subject or object coordinate. It is predicted that such an individual can use she and I in a subject or object coordinate, but never *I and she because of her supplemental Vocabulary for SF pronouns in coordinates. On this theory, it is predicted that an individual with Inventory III could produce both her and I and she and me. This usage is apparently non-attested or extremely rare, but this can be explained by extra-
288
linguistic social factors. The supplementary Vocabulary Items that yield SF pronouns in coordinates are highly salient and socially prestigious, so it is unlikely that an individual with Inventory III would use one but not both of their SF supplements in a coordinate. Notice that on this analysis, there is predicted to be no difference in the pronominal case forms that appear in syntactic subject vs. object coordinates. Because of social stigma and prescriptive education, there would seem to be good reasons to doubt self-reporting of pronoun case-form usage in coordinates. However, some individuals claim to consistently use SF pronouns in subject coordinates and OF pronouns in object coordinates. It is not certain that such individuals actually exist, although this question could be certainly be addressed using various empirical methods. Assuming for the sake of argument that they do exist, I would claim that such individuals have a still-different Vocabulary for pronouns. Perhaps they have learned morphological rules and Vocabulary Items similar to a transparent case language like German (something like the system of McFadden 2004, to appear). Alternately, let us suppose that such individuals have simply learned a supplementary Vocabulary Item that inserts SF pronouns anywhere inside CoPs when they occur in the specifier of
289
finite T; OFs are still the elsewhere exponents. This Vocabulary is illustrated schematically below. (65)
English pronoun Vocabulary, Inventory IV [D, Pers:_, Num:_] Ù /SF/ / [TP [CoP __ ][ T[Past:±] ... ]] [D, Pers:_, Num:_] Ù /SF/ / [TP __ [ T[Past:±] ...]] [D, Pers:_, Num:_] Ù /OF/ elsewhere This theoretical analysis makes several predictions
about usage frequency in populations when taken in combination with social factors. The first prediction is that the pronoun Vocabulary Inventories II and III are in an implicational relationship: if an individual has Inventory III then they have Inventory II, but no individual has Inventory III and not Inventory II. In other words, no one who is prescriptively moved enough to learn ‘she and’ fails to have also learned ‘and I.’ However, many people are forced to learn ‘and I’ but do not bother learning additional supplements. A second prediction is that Inventory II is much more widespread among the English-speaking population than Inventory III. Finally, I would predict that Inventory IV is extremely rare in the English-speaking population, and is limited to the most linguistically educated and prescriptively attentive, for example English teachers, writers, editors, and some linguists. 290
We can now tie up a few loose ends. First, there is no theoretical reason why an individual could not learn a supplementary Vocabulary Item for I and. However, I maintain that such persons are rare in English-speaking populations and that usage of I and is exceptional and mostly limited to edited writing. Similarly, I claim that few individuals in English-speaking populations have learned supplementary Vocabulary Items for plural pronouns SF in coordinates, although again, nothing prevents them from doing so on this analysis. Finally, notice that this kind of supplementary Vocabulary analysis can easily be expanded to cover apparently mismatched SF pronouns in the other prestige environments discussed by Emonds (1986) and Sobin (1997). For example, a very dedicated prescriptivist might be able to learn a supplementary Vocabulary Item that inserts the 1s SF I when the pronoun is right adjacent to it is. In any other environment, however, the default OF will still appear. (66) a.
6.3.4
It is I.
b.
It’s me (*I).
c.
It is just me (*I).
c.
It was me (*I).
Cross-linguistic predictions
291
By following the insights of Emonds (1986), the present DM analysis of English pronoun-case mismatches makes crosslinguistic predictions, a major advantage over some previous analyses reviewed above. Mismatched pronominal case forms are predicted to be completely unattested in languages with transparent case morphology on open-class DPs, such as German, Icelandic, Czech, etc. Pronominal case-form mismatches in coordinates and possibly other syntactic environments are predicted to occur only in languages that retain remnant case-like distinctions on a closed subset of pronouns, such as English, Norwegian, Danish, and Dutch. Preliminary inquiry suggests that these predictions are indeed confirmed. Several native speakers of German have informed me that pronoun-case mismatch in coordinates is unacceptable and totally non-attested. (67) a. * Mich und Stefan haben Bier getrunken. Me and Stefan drank beer. b. * Stefan und mich haben Bier getrunken. Stefan and me drank beer. (68) a. * Die Polizei hat ich und Stefan verhaftet. The police arrested I and Stefan. b. * Die Polizei hat Stefan und ich verhaftet. The police arrested Stefan and I.
292
In contrast, a native Norwegian speaker informed me that pronoun-case mismatch in coordinates is acceptable and occurs regularly. (69) a.
Meg og John drakk øl i går kveld. Me and John drank beer last night.
b.
John og meg drakk øl i går kveld. John and me drank beer last night.
Furthermore, a grammar suggests that such variation also occurs in Danish (Allan, Holmes and Lundskear-Nielsen 1995: 145, bold in original). In colloquial language, the objective form mig is sometimes used as subject [...]. This happens mostly in coordination with a noun phrase, irrespective of the order of the two (or more) coordinated elements, though it is felt to be even more informal when the personal pronoun appears in first place [...]. (70) a.
Min bror og mig er gode venner. My brother and me are good friends.
b.
Mig og min bror er gode venner. Me and my brother are good friends.
Much more research is needed, and especially variationist and sociolinguistic research into pronoun-case variation in Norwegian, Danish, and Dutch. More generally, the DM analysis presented here makes testable predictions about remnant versus productive morphology that should be investigated cross-linguistically and historically.
293
6.3.5
Predictions for acquisition research
Finally, the present DM analysis of English pronoun-case mismatches makes predictions about acquisition that can be tested experimentally, another advantage over previous analyses reviewed above. Again following Emonds (1986), it was hypothesized above that dissociated-case insertion rules and corresponding Vocabulary Items cannot be learned unless case morphology is sufficiently transparent, as in languages like German. Taken in conjunction with this hypothesis, the DM analysis of English pronominal forms presented above makes the following prediction. Children who are acquiring English should not produce SF pronouns in coordinate DPs until they have been taught to do so, and thereby learned some number of supplemental pronoun Vocabulary. Thus we predict that young, pre-kindergarten children will not produce SF pronouns in coordinates. This prediction might be tested with an experiment designed to elicit coordinated subject DPs containing pronouns, in order to see whether these pronouns will be in SF or OF. Participants would be children of about 4-5 years old, who have acquired non-coordinated pronoun case forms, but have not attended kindergarten. The children should be given an
294
elicitation task, since coordinated pronouns occur rarely in ordinary discourse. Such an experiment could be designed as follows. A child sits in a room with a puppet stage and two researchers, one male and one female. The child is not told the researchers’ names. There is some candy in clear bowl on the stage. A familiar puppet character asks the child to please keep an eye on the candy, and then leaves the stage. A different familiar puppet character appears on stage. This puppet talks to the researchers, encouraging them to eat the candy. After some comical persuasion, the researcher agrees, and together with the puppet they eat up all the candy. The original puppet reenters the stage. Acting shocked to see the missing candy, the puppet asks the child “What happened? Where is the candy?” Because the researchers are never named, this should elicit a coordinate DP with a pronoun, for example as below. (71) Fozzie and [her/she] ate up the candy! The number and sex of the researchers eating candy can be manipulated to elicit different pronouns, for example as below. (72) a. b.
Fozzie and [him/he] ate up the candy! Fozzie and [them/they] ate up the candy!
295
Again, the prediction is that children will produce only OF pronouns in their coordinates. If so, this would constitute support for Emonds’s transparency hypothesis and the present DM analysis of English pronominal forms.
6.4
Conclusion
In this chapter, I have again argued that significant mechanisms of Labovian variation are located in the inventory and features of non-competing Vocabulary Items. On the DM analysis of English pronominal case-form mismatches developed above, the variant case forms are inserted by different Vocabulary Items during the morphological component. Apparent morphosyntactic mismatches result because English lacks dissociated morphological case assignment rules. Instead, the pronominal Vocabulary of English inserts SF exponents when a pronoun is the specifier of finite T, and default OFs are inserted in all other environments. OFs are inserted by default in either conjunct of subject and object coordinates because pronouns in a CoP are not in the specifier of finite T. Specific SFs can occur inside coordinates only if the speaker has learned supplementary Vocabulary Items. The supplementary Vocabulary for coordinated 1s pronouns insert the SF if the pronoun is
296
right adjacent to and, whereas the supplementary Vocabulary for coordinated 3s pronouns insert the SF if the pronoun is left adjacent to and. As in Chapter 5, Labovian variation is distinguished from allomorphy because variant supplementary Vocabulary Items do not compete for insertion. In this case, it would seem that the mechanisms of parametric and Labovian variation overlap in DM’s Vocabulary Items. Individual speakers may have no, one (‘and I’), two (‘and I’ and ‘(s)he and’), or more supplementary pronoun Vocabulary in their inventory. Following Emonds, I have claimed that independent phonological changes in English rendered case features nontransparent for the acquisition of dissociated case rules and corresponding Vocabulary Items. The proposed set of pronoun Vocabulary, with contextual features that refer to the hierarchical position of finite T, was then required in order to maintain the complementary distribution of suppletive pronominal case-form allomorphs in English.
297
Chapter 7: Bridging the Gap
7.
Introduction
This chapter gives a final overview on the dissertation. The primary motivation for this dissertation was to contribute toward bridging the unfortunate gap that still separates biolinguistic theory from the empirical study of Labovian variation and change in progress. Descending into this longstanding gap, the dissertation posed the question of how an account of the mechanisms underlying Labovian variation in morphosyntax can be incorporated into a Minimalist theoretical model of the human language faculty. The remainder of this chapter is structured as follows. Section 7.1 gives a summary of conclusions, and Section 7.2 closes the dissertation with a call for further research comparing proposed theoretical mechanisms of Labovian variation, emphasizing the dissertation’s overarching goal of advocating cooperative efforts toward bridging the variation gap.
7.1
Summary of Conclusions
This dissertation adopted a particular Minimalist theory of syntax (Chomsky 2000a) augmented with the independently 298
motivated and well-articulated theory of Distributed Morphology (DM, Halle and Marantz 1993, Embick and Noyer to appear), as outlined in Chapter 3. Minimalist and DM theories of morphosyntax provide mechanisms for the familiar phenomenon of allomorphy, where variant morphological forms appear deterministically in a certain morphosyntactic environment. However, current theories lack mechanisms for Labovian variation, where variant forms appear probabilistically in the same morphosyntactic environment as discussed in Chapter 2. Chapters 4, 5, and 6 adapted variationist observational and quantitative methods in order to investigate cases of morphosyntactic variation. The proposed DM-mechanistic analyses of these phenomena have identified some specific morphosyntactic objects that constitute Labovian variant forms, located some specific morphosyntactic structural environments where Labovian variation occurs, explained some specific morphosyntactic mismatches in variant forms, and partially explained the possibility of sociolinguistic choice of Labovian variants, as opposed to the deterministic, complementary distribution of allomorphic variants. Chapter 4 examined weak expletive it (WEIT) on Smith Island, a case of morphosyntactic variation and change in progress. On the analysis of WEIT proposed in this chapter,
299
the mechanism of Labovian variation is located in the features of abstract syntactic terminals (similar to the approach advocated by Adger and Smith 2005, Adger 2006). The sociolinguistic ‘choice’ is between whether the terminal it (with φ features valued 3s) or the terminal there (with non-φ features valued [Prox:-, Dist:-]) will be selected into the pre-syntactic Lexical Array. As Adger and Smith point out, such a choice must be available in any theory of syntax. Once it has been chosen, the expletive terminal will be subject to syntactic and morphological operations that apply deterministically. Following a radical version of DM theory, this chapter denied the existence of uninterpretable Case and φ features that are checked and valued by the operation Agree in the narrow syntax (Embick and Noyer to appear, McFadden 2004). Instead, adopting the DM theory of Dissociated agreementfeature insertion, it was proposed that an English-specific rule copies the (interpretable) φ features of a subject DP into finite T during the post-syntactic morphological computation. If the subject phrase has no φ-features, as with the expletive there, an alternate Dissociated agreement rule of English copies the φ features of an associate DP in the VP complement of finite T. Thus, verbal agreement is categorically 3s with WEIT: because the 300
expletive terminal it has φ features valued 3s, it always triggers the ordinary subject agreement rule in English. To repeat, in this case the mechanism of Labovian variation is located in the choice of abstract syntactic terminals. Chapter 5 argued that mechanisms of Labovian variation are additionally located in the objects and operations of the morphological component. The chapter examined weren’t leveling, another case of morphosyntactic variation and change in progress on Smith Island. On the analysis of weren’t leveling proposed in this chapter, the mechanism of Labovian variation is a non-competing Vocabulary Item and its interaction with the operation of Fusion in the morphological component. Following Kandybowicz (to appear), late Fusion modifies morphosyntactic terminal structures to allow the insertion of suppletive Vocabulary Items, as outlined in Chapter 3. Leveled weren’t is a suppletive Vocabulary Item whose exponent can only be Inserted after Fusion of the be, T[Past:+], and Neg terminals. The Vocabulary Item for leveled weren’t lacks φ features, but contains Neg among its morphosyntactic features. Head adjunction of [M [be] T[past:+]] and Neg (whether by syntactic head movement or morphological Merger) feeds Fusion of [M be T[past:+]] and Neg for Insertion of weren’t. Thus, Smith Island weren’t leveling occurs without concurrent were leveling, and only 301
with the reduced form of negation -n’t The Vocabulary Item for weren’t does not compete for Vocabulary Insertion because its morphosyntactic features are a subset of any target terminal only if Fusion applies. Thus, leveled weren’t can be a sociolinguistic choice for individuals. To repeat, in this case the mechanism of Labovian variation is located in the inventory and features of non-competing Vocabulary Items, and their interactions with the operations of the morphological component. Finally, Chapter 6 examined mismatched pronoun-case variation in English coordinates, and argued again that the mechanism of Labovian variation is located in the inventory and features of non-competing Vocabulary Items. On the analysis of English pronominal case-form mismatches proposed in this chapter, the variant pronoun-case forms are inserted by different Vocabulary Items in the morphological component. Apparent morphosyntactic mismatches result because English lacks Dissociated morphological case assignment rules (contra McFadden to appear). English pronominal Vocabulary Items insert Subject Form (SF) exponents when a pronoun is the specifier of finite T. Default Object Form (OF) exponents are inserted in all other environments. OFs are inserted by default in either conjunct of subject and object coordinates because
302
pronouns in a CoP are not in the specifier of finite T. Specific SFs can occur inside coordinates only if the speaker has learned supplementary Vocabulary Items. The supplementary Vocabulary for coordinated 1s pronouns insert the SF if the pronoun is right adjacent to and, whereas the supplementary Vocabulary for coordinated 3s pronouns insert the SF if the pronoun is left adjacent to and. These variant supplementary Vocabulary Items do not compete for Insertion, and individuals may have no or several supplementary pronoun Vocabulary in their inventory. To repeat, in this case the mechanism of Labovian variation is located in the inventory and features of non-competing Vocabulary Items.
7.2
Future research on mechanisms of Labovian variation
I want to close the dissertation by calling for long-term cooperative research that negotiates the divide between sociolinguistics and biolinguistic theory. As discussed in Chapter 3, sociolinguistics fits naturally into the biolinguistic perspective by addressing questions of linguistic use, as well as population phenomena, that fall outside the ambit of linguistic theory. Thus, biolinguistics should encourage collaboration between theorists and variationists. Future research should
303
establish a robust body of data on morphosyntactic variation and change in progress, and proceed to contrast and evaluate theoretical approaches to these phenomena on an empirical basis. As an example, we can contrast the substantively different theoretical approaches to Labovian variation that have been proposed within the Minimalist-DM framework. As discussed in Chapter 3, this theoretical model predicts that possible mechanisms of Labovian variation should be limited to the features of syntactic terminals and to the objects and operations of the morphological component. As we saw in Chapter 5, Adger and Smith (2005, Adger 2006) have proposed that mechanisms of Labovian variation can be located in the uninterpretable features of syntactic terminals and their checking combinations in the narrow syntax. On such an analysis, social “patterns of variation seen across (groups of) individuals reduce to...lexical choice” (Adger and Smith 2005: 173). Adger (2006) provides a schematic diagram illustrating this approach to Labovian variation, which he calls ‘combinatorial variability.’ As reproduced below, the diagram shows a lexical item (LI1) with a set of interpretable features F1, F2, F3 that can check the uninterpretable features uF1, uF2, uF3 on distinct lexical items LI2, LI3, and LI4 in the narrow syntax
304
(Agreement is indicated by ‘...’). LIs 2-4 are spelled out with different phonological exponents (indicated by an arrow): the Phonetic Form (PF) of LI2 is x, PF of LI3 is y, and the PF of LI3 is z. In other words, variant forms x, y, and z correspond to variant lexical items LI2-4; the variants appear in the same morphosyntactic environment of Agreement with LI1; and the variants do not express any difference in semantics because the features of the variants LI2-4 are uninterpretable at LF. (1)
Combinatorial variability (Adger 2006)
LI1{F1,F2,F3} ...
LI2{uF1}
Æ
PF(LI2) = x
LI3{uF2}
Æ
PF(LI3) = y
LI4{uF3}
Æ
PF(LI4) = z
In Chapter 4 of this dissertation, evidence from a case study of WEIT on Smith Island supported the hypothesis that mechanisms of Labovian variation can indeed be located in the features of syntactic terminals. This analysis is very similar to Adger and Smith’s combinatorial variability, except that it denies any role for uninterpretable Case and Agreement feature checking in the narrow syntax. In Chapters 5 and 6, however, evidence from case studies of weren’t leveling on Smith Island and English pronominal case-form mismatches supported the additional hypothesis that Labovian variation can arise in 305
the inventory and feature structure of non-competing Vocabulary Items and their interactions with ordered operations in the morphological component. On this DM approach, a mechanism of allomorphy is the competition between Vocabulary Items, and a mechanism of Labovian variation is non-competition between Vocabulary Items. Social patterns of variation thus reduce to choice of noncompeting Vocabulary Items. This approach is illustrated below with a schematic diagram. Here, an abstract syntactic terminal ST1 has a set of one or possibly more interpretable features F1 (and/or Fn). Vocabulary Insertion (VI, indicated with an arrow) will provide ST1 with its phonological exponent in the morphological component. A set of Vocabulary Items contain a subset of morphosyntactic features [F1(Fn)] that identify ST1 for insertion of phonological features /x/, /y/, or /z/. The first Vocabulary Item for ST1 does not compete with the others to insert its exponent /x/ (indicated with a dotted-dashed line). Thus, by hypothesis, this non-competing Labovian variant form can be a sociolinguistic choice. The remaining Vocabulary Items must compete to insert their allomorphic exponents /y/ and /z/.
306
(2)
Labovian Variation in Vocabulary
ST1{F1(Fn)}
Å
ST1 [F1(Fn)]
Ù
/x/
[F1(Fn)]
Ù
/y/
else
Ù
/z/
VI
The plausible hypothesis of non-competing Vocabulary Items explains the empirically necessary distinction between allomorphy and Labovian variation. But this DM approach raises the important question of whether we need mechanisms of Labovian variation in both the syntactic terminals and the morphological component. Is it possible, or desirable, to locate variation in only one of these components? Future research should address this issue. On the empirical basis of three case studies, this dissertation has argued for mechanisms of Labovian variation in both the syntactic terminals (similar to Adger and Smith’s lexical approach but without uninterpretable feature checking) and in the post-syntactic insertion of non-competing Vocabulary Items. Both kinds of mechanisms are independently motivated and consistent with the Minimalist program. Some notion of lexical choice is clearly necessary in any theory of syntax, as is some articulated procedure for mapping syntactic structures onto phonological ones. Presently, therefore, there is no 307
obvious reason to exclude either lexical or Vocabularybased mechanisms from the theory of Labovian variation.
308
References 2000. Table DP-1. Profile of General Demographic Characteristics: 2000, Geographic area: Smith Island CDP, Maryland: U.S. Census Bureau. Abney, Steven. 1987. The English Noun Phrase in its Sentential Aspect. Doctoral Dissertation, Massachusetts Institute of Technology. Adger, David. 2003. Core Syntax: A Minimalist Approach. Oxford: Oxford University Press. Adger, David. 2006. Combinatorial variability. Journal of Linguistics 42: 503-530. Adger, David. to appear. Combinatorial variability. Journal of Linguistics. Adger, David, and Jennifer Smith. 2005. Variation and the Minimalist Program. In Syntax and Variation: Reconciling the Biological and the Social, eds. Leonie Cornips and Karen P. Corrigan, 149-178. Amsterdam: John Benjamins. Adger, David, and Graeme Trousdale. to appear. Variation in English syntax: theoretical implications. English Language and Linguistics. Allan, Robert, Philip Holmes, and Tom Lundskear-Nielsen. 1995. Danish: A Comprehensive Grammar. New York: Routledge. Allen, Cynthia L. 1995. Case Marking and Reanalysis. Oxford: Oxford University Press. Anderson, Stephen R. 1992. A-Morphous Morphology. Cambridge: Cambridge University Press. Anderson, Stephen R., and David W. Lightfoot. 1999. The Human language faculty as an organ. Annual Review of Physiology 62: 697-722. Anderson, Stephen R., and David W. Lightfoot. 2002. The Language Organ. Cambridge: Cambridge University Press. Angermeyer, Philipp S., and John Victor Singler. 2003. The case for politeness: Pronoun variation in co-ordinate NPs in object position in English. Language Variation and Change 15: 171-209. Anttila, Arto. 2002. Variation and phonological theory. In The Handbook of Language Variation and Change, eds. Jack K. Chambers, Peter Trudgill and Natalie Schilling-Estes. Malden, MA: Blackwell. Archangeli, Diana, and D. Terence Langendoen eds. 1997. Optimality Theory: An Overview. Malden, MA: Blackwell. Aronoff, Mark. 1994. Morphology by itself: stems and inflectional classes. Cambridge: MIT Press.
309
Bailey, Charles-James N. 1973. Variation and linguistic theory. Ms. Arlington, VA: Center for Applied Linguistics. Bailey, Guy. 2002. Real and apparent time. In The Handbook of Language Variation and Change, eds. Jack K. Chambers, Peter Trudgill and Natalie Schilling-Estes. Malden, MA: Blackwell. Bailey, Guy, Tom Wikle, Jan Tillerly, and Lori Sand. 1992. The apparent time construct. Language Variation and Change 3: 241-264. Baker, Mark C. 1988. Incorporation: A Theory of Grammatical Function Changing. Chicago: The University of Chicago Press. Baker, Mark C. 1996. The Polysynthesis Parameter: Oxford Studies in Comparative Syntax. Oxford: Oxford University Press. Baker, Mark C. 2001. The Atoms of Language: The Mind's Hidden Rules of Grammar. New York, NY: Basic Books. Bauer, Matthew J. 2005. Prosodic Strengthening of Consonants in Iron Range English. Doctoral Dissertation, Georgetown University. Beard, Robert. 1995. Lexeme-Morpheme Base Morphology: A General Theory of Inflection and Word Formation. Oxford: Oxford University Press. Bender, Emily M. 2001. Syntactic Variation and Linguistic Competence: The Case of AAVE Copula Absence. Doctoral Dissertation, Stanford University. Bernstein, Judy B., and Christina Tortora. 2005. Two types of possessive forms in English. Lingua 115: 1221-1242. Bickerton, Derek. 1971. Inherent variability and variable rules. Foundations of Language 7: 457-492. Bickerton, Derek. 1990. Language and Species. Chicago: University of Chicago Press. Bobaljik, Jonathan. 1995. Morphosyntax: The Syntax of Verbal Inflection. Doctoral Dissertation, Massachusetts Institute of Technology. Bobaljik, Jonathan. 2001. Syncretism without paradigms: Remarks on Williams 1981, 1994. Ms. McGill University. Bobaljik, Jonathan. 2002. Realizing Germanic inflection: Why morphology does not drive syntax. The Journal of Comparative Germanic Linguistics 6: 129-167. Borer, Hagit. 1984. Parametric Syntax. Dordrecht: Foris. Bresnan, Joan, Ashwini Deo, and Devyani Sharma. to appear. Typology in variation: A Probalistic Approach to be and n't in the Survey of English Dialects. English Language and Linguistics.
310
Britain, David. 2002. Diffusion, leveling, simplification and reallocation in past tense BE in the English Fens. Journal of Sociolinguistics 6: 16-43. Brown, Sue. 1999. The Syntax of Negation in Russian: A Minimalist Approach. Stanford: CLSI Publications. Castillo, Juan Carlos, John Drury, and Kleanthes K. Grohmann. 1999. Merge Over Move and the Extended Projection Principle. In University of Maryland Working Papers in Linguistics, eds. Sachiko Aoshima, John Drury and Tuomo Neuvonen, 63-103. Cedergren, Henrietta J., and David Sankoff. 1974. Variable rules: Performance as a statistical reflection of competence. Language 50: 333-335. Chambers, Jack. 1991. Canada: Introduction. In Englsih Around the World, ed. Jenny Cheshire, 89-107. Cambridge: Cambridge University Press. Chambers, Jack K. 2002a. Sociolinguistic Theory: Linguistic Variation and its Social Significance (2nd Edition). Malden, MA: Blackwell. Chambers, Jack K. 2002b. Studying Language Variation: An Informal Epistemology. In The Handbook of Language Variation and Change, eds. Jack K. Chambers, Peter Trudgill and Natalie Schilling-Estes. Malden, MA: Blackwell. Chambers, Jack K., Peter Trudgill, and Natalie SchillingEstes eds. 2002. The Handbook of Language Variation and Change. Malden, MA: Blackwell. Chomsky, Noam. 1959. A review of Verbal Behavior, by B. F. Skinner. Language 35: 26-58. Chomsky, Noam. 1965. Aspects of the Theory of Syntax. Cambridge: MIT Press. Chomsky, Noam. 1971. Formal Discussion Of Miller and Ervin's The Development of Grammar in Child Language. In Child Language A Book Of Readings, eds. A. Bar-Adon and W. F. Leopold. New Jersey: Prentice-Hall, Inc. Chomsky, Noam. 1981. Lectures on Government and Binding. Dordrecht: Foris. Chomsky, Noam. 1982. Some Concepts and Consequences of the Theory of Government and Binding. Cambridge: MIT Press. Chomsky, Noam. 1986. Knowledge of Language: Its Nature, Origin, and Use. Westport: Praeger Publishers. Chomsky, Noam. 1993. A Minimalist program for linguistic theory. In The View from Building 20: Essays in linguistics in honor of Sylvain Bromberger, eds. Ken Hale and S. J. Keyser, 1-52. Cambridge: MIT Press.
311
Chomsky, Noam. 1995. The Minimalist Program. Cambridge, MA: MIT Press. Chomsky, Noam. 2000a. Minimalist inquiries: The Framework. In Step by Step: Essays on Minimalist Syntax in Honor of Howard Lasnik, eds. Roger Martin, David Michaels and Juan Uriagereka, 89-155. Cambridge, MA: MIT Press. Chomsky, Noam. 2000b. New Horizons in the Study of Language and Mind. Cambridge: Cambridge University Press. Chomsky, Noam. 2001. Derivation by phase. In Ken Hale: A Life in Language, ed. Michael Kenstowicz, 1-52. Cambridge: MIT Press. Chomsky, Noam. 2002. On Nature and Language. Cambridge: Cambridge University Press. Chomsky, Noam. 2004a. Beyond explanatory adequacy. In Structures and Beyond: The Cartography of Syntactic Structures, ed. Adriana Belletti, 104-131. Oxford: Oxford University Press. Chomsky, Noam. 2004b. Language and mind: Current thoughts on ancient problems. In Variation and Universals in Biolinguistics, ed. Lyle Jenkins. Amsterdam: Elsevier. Chomsky, Noam. 2005a. Three factors in language design. Linguistic Inquiry 36: 1-22. Chomsky, Noam. 2005b. On phases. Ms. MIT. Chomsky, Noam, and Morris Halle. 1968. The Sound Pattern of English. New York: Harper & Row. Chomsky, Noam, and Howard Lasnik. 1993. The theory of principles and parameters. In Syntax: An International Handbook of Contemporary Research, eds. Joachim Jacobs et al. Berlin: Walter de Gruyter. Chomsky, Noam, Marc D. Hauser, and W. Tecumseh Fitch. 2005. The Minimalist Program: Unpublished appendix to Fitch, Hauser and Chomsky (to appear). Cognition. Cornips, Leonie, and Karen Corrigan. 2005. Toward an integrated approach to syntactic variation: A Retrospective and prospective synopsis. In Syntax and Variation: Reconciling the Biological and the Social, eds. Leonie Cornips and Karen P. Corrigan. Amsterdam: John Benjamins. Cowart, Wayne. 1997. Experimental Syntax. Thousand Oaks, CA: Sage Publications. DeGraff, Michel ed. 1999. Language Creation and Language Change: Creolization, Diachrony, and Development. Cambridge, MA: MIT Press. DeWolf, Gaelan Dodds. 1992. Social and Regional Factors in Canadian English. Toronto: Canadian Scholar's Press. Dize, Frances W. 1990. Smith Island, Chesapeake Bay. Centerville, MD: Tidewater Publishers.
312
Embick, David. to appear-a. Linearization and Local Dislocation: Derivational mechanics and interactions. Linguistic Analysis. Embick, David. to appear-b. Blocking effects and analytic/synthetic alternations. Natural Language and Linguistic Theory. Embick, David, and Rolf Noyer. 2001. Movement operations after syntax. Linguistic Inquiry 32: 555-595. Embick, David, and Alec Marantz. 2006. Architecture and blocking. Ms. University of Pennsylvania and Massachusetts Institute of Technology. Embick, David, and Rolf Noyer. to appear. Distributed Morphology and the Syntax/Morphology Interface. In The Oxford Handbook of Linguistic Interfaces, eds. G. Ramchand and C. Reiss. Oxford: Oxford University Press. Emonds, Joseph. 1986. Grammatically deviant prestige constructions. In A Festschrift for Sol Saporta, eds. Michael Brame, Heles Contreras and Fredrick J. Newmeyer. Seattle: Noit Amrofer. Enard, Wolfgang, Molly Przeworski, Simon E Fisher, Cecilia S L Lai, Victor Wiebe, Takashi Kitano, Anthony P Monaco, and Svante Paabo. 2002. Molecular evolution of FOXP2, a gene involved in speech and language. Nature 418: 869-872. Fasold, Ralph. 1991. The quiet demise of variable rules. American Speech 66: 3-21. Fasold, Ralph. 1993. Manual for Microcomputer VARBRUL 2 System, MS-DOS version. Ms. Georgetown University. Fitch, W. Tecumseh, and Marc D. Hauser. 2004. Computational Constraints on Syntactic Processing in a Nonhuman Primate. Science 303: 377-380. Fitch, W. Tecumseh, Marc D. Hauser, and Noam Chomsky. 2005. The Evolution of the language faculty: Clarifications and implications. Cognition 97: 179-210. Garner, Brian A. 1998. A Dictionary of Modern American Usage. New York: Oxford University Press. Gentner, Timothy Q., Kimberly M. Fenn, Daniel Margoliash, and Howard C. Nusbaum. 2006. Recursive syntactic pattern learning by songbirds. Nature 440: 1204-1207. Gervain, Judit, and Gábor Zemplén. 2005. Focus raising: A Paradigmatic example of the treatment of syntactic variation. In Syntax and Variation: Reconciling the Biological and the Social, eds. Leonie Cornips and Karen P. Corrigan. Amsterdam: John Benjamins. Goodall, Grant. 1987. Parallel Structures in Syntax. Cambridge: Cambridge University Press.
313
Gopnik, Myra, and Martha B. Crago. 1991. Familial aggregation of a developmental language disorder. Cognition 39: 1-50. Grano, Thomas. 2006a. “Me and her” meets “he and I”: Case, person, and linear ordering in English coordinated pronouns. Honors thesis, Stanford University. Grano, Thomas. 2006b. Her and I, *she and me: Case, person, and coordination. Paper presented at New Ways of Analyzing Variation (NWAV) 35, Ohio State University. Green, Lisa J. 2002. African American English: A Linguistic Introduction. Cambridge, UK: Cambridge University Press. Grodzinsky, Yosef, and Lisa Finkel. 1998. The Neurology of empty categories: Aphasics' failure to detect grammaticality. Journal of Cognitive Neuroscience 10: 281-292. Grohmann, Kleanthes K. 2003. Prolific Domains: On the AntiLocality of Movement Dependencies. Amsterdam: John Benjamins. Halle, Morris. 1997. Distributed Morphology: Impoverishment and Fission. MIT Working Papers in Linguistics. Halle, Morris, and Alec Marantz. 1993. Distributed morphology and the pieces of inflection. In The View from Building 20: Essays in Linguistics in Honor of Sylvain Bromberger, eds. Ken Hale and Samuel Jay Keyser, 111-176. Cambridge, MA: MIT Press. Harley, Heidi, and Rolf Noyer. 1999. Distributed Morphology. Glot International 4: 3-9. Hauser, Marc D. 1997. The Evolution of Communication. Cambridge: MIT Press. Hauser, Marc D., Noam Chomsky, and W. Tecumseh Fitch. 2002. The faculty of language: What is it, who has it, and how did it evolve? Science 298: 1569-1579. Hay, Jennifer, and Daniel Schreier. 2004. Reversing the trajectory of language change: Subject-verb agreement with be in New Zealand English. Language Variation and Change 16: 209-235. Henry, Alison. 1995. Belfast English and Standard English: Dialect Variation and Parameter Setting. Oxford: Oxford University Press. Henry, Alison. 1996. Imperative Inversion in Belfast English. In Microparametric Syntax and Dialect Variation, eds. James R. Black and Virginia Motapanyane. Amsterdam: John Benjamins. Henry, Alison. 2002. Variation and syntactic theory. In The Handbook of Language Variation and Change, eds. Jack
314
K. Chambers, Peter Trudgill and Natalie SchillingEstes. Malden, MA: Blackwell. Henry, Alison. 2005. Idiolectal variation and syntactic theory. In Syntax and Variation: Reconciling the Biological and the Social, eds. Leonie Cornips and Karen P. Corrigan. Amsterdam: John Benjamins. Holm, John. 1988. Pidgins and Creoles: Volume I: Theory and Structure. Cambridge, UK: Cambridge University Press. Holm, John. 1989. Pidgins and Creoles: Volume II: Reference Survey. Cambridge, UK: Cambridge University Press. Honey, John. 1995. A New rule for the Queen and I? English Today 11: 3-8. Hornstein, Norbert, Jairo Nunes, and Kleanthes K. Grohmann. 2005. Understanding Minimalism. Cambridge, UK: Cambridge University Press. Horton, Tom. 1987. Bay Country. Baltimore: The Johns Hopkins University Press. Horton, Tom. 1996. An Island Out of Time: A Memoir of Smith Island in the Chesapeake. New York: W.W. Norton & Co. Hudson, Richard. 1995a. Syntax and Sociolinguistics. In Syntax: An International Handbook of Contemporary Research Vol. 2, eds. Joachim Jacobs et al., 15141528. Berlin: Walter de Gruyter. Hudson, Richard. 1995b. Does English really have case? Journal of Linguistics 31: 375-392. Hudson, Richard. 1997. Inherent variability and linguistic theory. Cognitive Linguistics 8: 73-108. Jackendoff, Ray. 1997. The Architecture of the Language Faculty. Cambridge: MIT Press. Jackendoff, Ray. 2002. Foundations of Language: Brain, Meaning, Grammar, Evolution: Oxford University Press. Jackendoff, Ray, and Steven Pinker. 2005. The Nature of the Language Faculty and its Implications for Evolution of Language (Reply to Fitch, Hauser, & Chomsky). Cognition 97: 211-225. Jenkins, Lyle. 2000. Biolinguistics: Exploring the Biology of Language. Cambridge, UK: Cambridge University Press. Jenkins, Lyle ed. 2004. Variation and Universals in Biolinguistics. Amsterdam: Elsevier. Johannessen, Janne Bondi. 1998. Coordination. New York: Oxford University Press. Kandybowicz, Jason. 2006. On Fusion and the architecture of PF. Paper presented at Penn Linguistics Colloquium (PLC) 30, University of Pennsylvania. Kandybowicz, Jason. to appear. Fusion and PF architecture. In Proceedings of the 30th Annual Penn Linguistics
315
Colloquium. Philadelphia: University of Pennsylvania Working Papers in Linguistics. Kayne, Richard S. 1994. The Antisymmetry of Syntax. Cambridge: MIT Press. Kayne, Richard S. 1996. Microparametric syntax: Some introductory remarks. In Microparametric Syntax and Dialect Variation, eds. James R. Black and Virginia Motapanyane, ix-xviii. Amsterdam: John Benjamins. Kegl, Judy, Ann Senghas, and Marie Coppola. 1999. Creation through contact: Sign language emergence and sign language change in Nicargua. In Language Creation and Language Change: Creolization, Diachrony, and Development, ed. Michel DeGraff. Cambridge, MA: MIT Press. Kroch, Anthony. 1989. Reflexes of grammar in patterns of language change. Language Variation and Change 1: 199244. Kroch, Anthony. 1994. Morphosyntactic variation. In Papers from the 30th Regional Meeting of the Chicago Linguistics Society, ed. Katharine Beals, 180–201. Chicago: Chicago Linguistic Society. Kroch, Anthony. 2001. Syntactic Change. In The Handbook of Contemporary Syntactic Theory, eds. Mark Baltin and Chris Collins. Malden, MA: Blackwell. Labov, William. 1963. The Social motivation of a sound change. Word 19: 273-309. Labov, William. 1966. The Social Stratification of English in New York City. Washington, DC: Center for Applied Linguistics. Labov, William. 1969. Contraction, deletion, and inherent variability of the English copula. Language 45: 715762. Labov, William. 1972. Sociolinguistic Patterns. Philadelphia: University of Pennsylvania Press. Labov, William. 1994. Principles of Linguistic Change: Internal Factors. Cambridge, MA: Blackwell. Labov, William. 1995. The two futures of linguistics. In Linguistics in the Morning Calm 3: Selected Papers from the Seoul International Conference on Linguistics (SECOL) 1992, ed. Ik-Hwan Lee, 113-147. Seoul: Hanshin Publishing Company. Labov, William. 2001. Principles of Linguistic Change: Social Factors. Cambridge, MA: Blackwell. Labov, William. 2005. Quantitative analysis of linguistic variation. In Sociolinguistics: An International Handbook of the Science of Language and Society, 2nd
316
Edition, eds. Ulrich Ammon et al., 6-21. Berlin: Walter de Gruyter. Labov, William, Sharon Ash, and Charles Boberg. 2005. The Atlas of North American English: Phonetics, Phonology and Sound Change. Berlin and New York: Walter de Gruyter. Lai, Cecilia S L, Simon E Fisher, Jane A Hurst, Faraneh Vargha-Khadem, and Anthony P Monaco. 2001. A forkheaddomain gene is mutated in a severe speech and language disorder. Nature 413: 519-523. Lardiere, Donna. 2000. Mapping features to forms in second language acquisition. In Second Language Acquisition and Linguistic Theory, ed. John Archibald. Malden: Blackwell. Lardiere, Donna. 2005. On Morphological Competence. In Proceedings of the 7th Generative Approaches to Second Language Acquisition Conference (GASLA 2004), ed. Laurent Dekydtspotter, 178-192. Somerville, MA: Cascadilla Proceedings Project. Lasnik, Howard, and Nicholas Sobin. 2000. The who/whom puzzle: on the preservation of an archaic feature. Natural Language and Linguistic Theory: 343-371. Lightfoot , David. 1999. The Development of Language: Acquisition, Change, and Evolution. Oxford: Blackwell. Lippi-Green, Rosina. 1997. English with an Accent: Language, Ideology, and Discrimination in the United States. London: Routledge. Loving, Aretta. 1990. Whatever happened to me? (An Objective case study) Or, The Groanings of a grammarian. Notes on Linguistics 48: 21-24, 51. Lust, Barbara. 1977. Conjunction reduction in child language. Journal of Child Language 4: 257-287. Lust, Barbara, and Cynthia A. Mervis. 1980. Development of coordination in the natural speech of young children. Journal of Child Language 7: 279-304. MacWhinney, Brian. 1992. The CHILDES Database, 2nd Ed. Dublin, OH: Discovery Systems. Marantz, Alec. 1997. No escape from syntax: Don't try morphological analysis in the privacy of your own lexicon. In Proceedings of the 21st Penn Linguistics Colloquium, eds. A. Dimitriadis et al., 201-225. Philadelphia: University of Pennsylvania Working Papers in Linguistics. Marantz, Alec. 2005. Generative linguistics within the cognitive neuroscience of language. The Linguistic Review 22: 429-445.
317
Markson, Lori, and Paul Bloom. 1997. Evidence Against a Dedicated System for Word Learning in Children. Nature February 27: 813. Martin, Stefan, and Walt Wolfram. 1998. The Sentence in African-American Vernacular English. In AfricanAmerican English, eds. Salikoko S. Mufwene et al., 1136. New York: Routledge. Matushansky, Ora. 2006. Head movement in linguistic theory. Linguistic Inquiry 37: 69-109. Mayr, Ernst. 1982. The Growth of Biological Thought. Cambridge, MA and London, England: The Belknap Press of Harvard University Press. Mayr, Ernst. 1997. This is Biology: The Science of the Living World. Cambridge, MA: The Belknap Press of Harvard University Press. Mayr, Ernst. 2001. What Evolution Is. New York, NY: Basic Books. McFadden, Thomas. 2004. The Position of Morphological Case in the Derivation: A Study on the Syntax-Morphology Interface. Doctoral Dissertation, University of Pennsylvania. McFadden, Thomas. 2006. Default case and the status of compound categories in Distributed Morphology. Paper presented at Penn Linguistics Colloquium (PLC) 30, University of Pennsylvania. McFadden, Thomas. to appear. Default case and the status of compound categories in Distributed Morphology. In Proceedings of the 30th Annual Penn Linguistics Colloquium. Philadelphia: University of Pennsylvania Working Papers in Linguistics. Meechan, Marjory, and Michele Foley. 1994. On resolving disagreement: Linguistic theory and variation-"there's bridges". Language Variation and Change 6: 63-86. Mittelstaedt, Jennifer. 2006. Auxiliary Leveling and Morphological Theory: The Case of Smith Island English. Doctoral Dissertation, Georgetown University. Mittelstaedt, Jennifer. in progress. Auxiliary Leveling and Morphological Theory: The Case of Smith Island English. Doctoral Dissertation, Georgetown University. Mittelstaedt, Jennifer. to appear. Apparent-time change in the Smith Island auxiliary verb system. In Diversity in Language: Perspectives and Implications, eds. Matsumoto Yoshiko et al. Stanford University: CSLI. Mittelstaedt, Jennifer, and Jeffrey K. Parrott. 2002. A Distributed Morphology account of weren't leveling. Paper presented at NWAV 31, Stanford University.
318
Moro, Andrea. 2000. Dynamic Antisymmetry. Cambridge, MA: MIT Press. Munn, Alan. 1994. Topics in the Syntax and Semantics of Coordinate Structures. Doctoral Dissertation, University of Maryland. Nevins, Andrew. 2006. The Representation of Third Person and its Consequences for Person-Case Effects. Ms. Harvard University. Nevins, Andrew. to appear. Cross-modular parallels in the study of Phon and Phi. In Phi Theory, eds. David Adger, Daniel Harbour and Susana Bejar. Oxford: Oxford University Press. Nevins, Andrew, and Jeffrey K. Parrott. in progress. A Variable Impoverishment analysis of leveling variation in English be and -n't. Ms. Harvard University and Georgetown University. Newmeyer, Frederick J. 2003. Grammar is grammar and usage is usage. Language 79: 682-707. O'Conner, Patricia T. 1996a. Woe is I: The Grammarphobe's Guide to Better English in Plain English. New York: Berkley Publishing Group. O'Conner, Patricia T. 1996b. Like I said, don't worry. Newsweek, December 9, page 12 Parrott, Jeffrey K. 2000a. A Partial solution to the theoretical paradox posed by Smith Island weak expletive it. Ms. Georgetown University. Parrott, Jeffrey K. 2000b. Negative Inversion in African American Vernacular English: A case of optional movement? In Proceedings of the 28th Western Conference on Linguistics (WECOL), Vol. 11, eds. Nancy Mae Antrim et al., 414-427. California State University, Fresno: Department of Linguistics. Parrott, Jeffrey K. 2001a. A Morphological theory of head movement. Ms. Georgetown University. Parrott, Jeffrey K. 2001b. Explaining linear ordering asymmetries in English: A morphological alternative. Paper presented at Asymmetry Conference, Université du Québec à Montréal. Parrott, Jeffrey K. 2001c. A Preliminary sketch of variation in the verbal agreement system of Smith Island English. Ms. Georgetown University. Parrott, Jeffrey K. 2002. Dialect death and morphosyntactic change: Smith Island weak expletive it. In University of Pennsylvania Working Papers in Linguistics 8.3: Papers from NWAV 30, eds. Daniel Ezra Johnson and Tara Sanchez, 175-189. Philadelphia:
319
University of Pennsylvania Working Papers in Linguistics. Parrott, Jeffrey K. 2006. Beyond you and I: Distributed Morphological mechanisms of pronoun-case variation. In PWPL Vol. 12.2: Papers from NWAV 34. Philadelphia: University of Pennsylvania Working Papers in Linguistics. Parrott, Jeffrey K. to appear-a. Review of Cornips and Corrigan, eds. Syntax and Variation: Reconciling the Biological and the Social. 2005, John Benjamins. English Language and Linguistics. Parrott, Jeffrey K. to appear-b. Distributed Morphological mechanisms of Smith Island weren't leveling. In Proceedings of the 30th Penn Linguistics Colloquium. Philadelphia: University of Pennsylvania Working Papers in Linguistics. Pesetsky, David, and Esther Torrego. to appear. The Syntax of valuation and the interpretability of features. In Clever and Right: A Festschrift for Joe Emonds, eds. S. Karimi, V. Samiian and W. Wilkins. Phillips, Colin. 2004. The immediacy of grammar. Paper presented at Georgetown University Round Table (GURT) 2004, Georgetown University, Washington D.C. Phillips, Colin, Nina Kazanina, and Shani Abada. 2005. ERP effects of the processing of syntactic long-distance dependencies. Cognitive Brain Research 22: 407-428. Pinker, Steven. 1994. The Language Instinct. New York, NY: W. Morrow and Co. Pinker, Steven. 2003. Language as an adaptation to the cognitive niche. In Language Evolution: States of the Art, eds. M. Christiansen and S. Kirby. New York: Oxford University Press. Pinker, Steven, and Paul Bloom. 1990. Natural language and natural selection. Behavioral and Brain Sciences 13: 707-784. Pinker, Steven, and Ray Jackendoff. 2005. The faculty of language: What's special about it? Cognition 95: 201236. Poeppel, David, and Kenneth Wexler. 1993. The full competence hypothesis of clause structure in early German. Language 69: 1-33. Poeppel, David, and David Embick. 2005. Defining the relation between linguistics and neuroscience. In Twenty-First Century Psycholinguistics: Four Cornerstones, ed. A. Cutler: Lawrence Erlbaum.
320
Pollard, Carl, and Ivan A. Sag. 1994. Head-driven Phrase Structure Grammar. Chicago: University of Chicago Press. Poplack, Shana, and Sali Tagliamonte. 1991. African American English in the diaspora: Evidence from oldline Nova Scotians. Language Variation and Change 3: 301-340. Poplack, Shana ed. 2000. The English History of African American English. Malden, MA: Blackwell. Prince, Alan, and Paul Smolensky. 1993. Optimality Theory: Constraint interaction in generative grammar. Technical Report TR-2, Rutgers University Cognitive Science Center, New Brunswick, NJ: To appear, MIT Press. Quattlebaum, Judith A. 1994. A study of case assignment in coordinate noun phrases. Language Quarterly 32: 131147. Quinn, Heidi. 2005. The Distribution of Pronoun Case Forms in English. Amsterdam: John Benjamins. Quirk, Randolph, and Sidney Greenbaum. 1973. A Concise Grammar of Contemporary English. San Diego: Harcourt Brace Jovanovich. Ramus, Franck, Marc D. Hauser, Cory Miller, Dylan Morris, and Jacques Mehler. 2000. Language Discrimination by Human Newborns and by Cotton-Top Tamarin Monkeys. Science 288: 349-351. Redfern, Richard K. 1994. Is between you and I good English? In Centennial Usage Studies. Publications of the American Dialect Society 78, eds. Greta D. Little and Micheal Montgomery, 187-193. Tuscaloosa: University of Alabama Press. Rickford, John R. 1998. The Creole Origins of AfricanAmerican English: Evidence from Copula Absence. In African American English: Structure, History and Use, eds. Salikoko S. Mufwene et al., 154-200. London: Routledge. Rickford, John R., Thomas A. Wasow, Norma Mendoza-Denton, and Juli Espinoza. 1995. Syntactic variation and change in progress: Loss of the verbal coda in topirestricting as far as constructions. Language 71: 102131. Rispoli, Matthew. 1994. Pronoun case overextensions and paradigm building. Journal of Child Language 21(1): 157-172. Rispoli, Matthew. 1995. Factors contributing to the frequency of pronoun case overextension. In The Proceedings of the 27th annual Child Language Research
321
Forum, ed. E. V. Clark. New York, NY: Cambridge University Press. Roberts, Ian. 2001. Head movement. In The Handbook of Contemporary Syntactic Theory, eds. Mark Baltin and Chris Collins. Malden, MA: Blackwell. Rupp, Laura. 2005. Constraints on non-standard -s in expletive there sentences: A Generative-variationist perspective. English Language and Linguistics 9: 225288. Schilling-Estes, Natalie. 1997. Accommodation vs. concentration: Dialect death in two post-insular island communities. American Speech 72: 12-32. Schilling-Estes, Natalie. 2000. Exploring morphological change: The was/weren't pattern in Smith Island English. Paper presented at New Ways of Analyzing Variation (NWAV) 29, Michigan State University. Schilling-Estes, Natalie. 2001. Variation study and discourse analysis: The sociolinguistic interview as a situated speech event. Paper presented at NWAV 30, North Carolina State University. Schilling-Estes, Natalie. 2002a. Introduction: Linguistic Structure. In The Handbook of Language Variation and Change, eds. Jack K. Chambers, Peter Trudgill and Natalie Schilling-Estes. Malden, MA: Blackwell. Schilling-Estes, Natalie. 2002b. Investigating stylistic variation. In The Handbook of Language Variation and Change, eds. Jack K. Chambers, Peter Trudgill and Natalie Schilling-Estes, 375-401. Malden, MA: Blackwell. Schilling-Estes, Natalie. 2004. "We ain’t got really many words; we just say ‘em backwards": Investigating ‘backwards talk’ in Smith Island, MD. Paper presented at New Ways of Analyzing Variation (NWAV) 33, University of Michigan, Ann Arbor, MI. Schilling-Estes, Natalie. 2005. Language change in apparent and real time: The Community and the individual. In University of Pennsylvania Working Papers in Linguistics 10.2: Papers from NWAV 32, eds. Keelan Evans et al. Philadelphia: University of Pennsylvania Working Papers in Linguistics. Schilling-Estes, Natalie, and Walt Wolfram. 1994. Convergent explanation and alternative regularization patterns: Were/weren't leveling in a vernacular English variety. Language Variation and Change 6: 273302.
322
Schilling-Estes, Natalie, and Walt Wolfram. 1999. Alternative models of dialect death: Dissipation vs. concentration. Language 75: 486-521. Schilling-Estes, Natalie, and Walt Wolfram. 2003. Language change in "conservative" dialects: The Case of past tense be in Southern enclave communities. American Speech 78: 209-228. Schütze, Carson T. 1996. The Empirical Base of Linguistics: Grammaticality Judgments and Linguistic Methodology. Chicago: The University of Chicago Press. Schütze, Carson T. 1997. Infl in child and adult language: Agreement, case and licensing. Doctoral Dissertation, MIT. Schütze, Carson T. 1999. English expletive constructions are not infected. Linguistic Inquiry 30: 467-484. Schütze, Carson T. 2001. On the nature of default case. Syntax 4: 205-238. Schwartz, B. D. 1985. Case and conjunction. Southern California Occasional Papers in Linguistics 10: 161186. Senghas, Ann, Sotaro Kita, and Asli Ozyurek. 2004. Children Creating Core Properties of Language: Evidence from an Emerging Sign Language in Nicaragua. Science 305: 1779-1782. Sheenan, Kevin. 1994. Order and disorder on Smith Island. Raritan 14: 109-134. Shu, Weiguo, Julie Y. Cho, Yuhui Jiang, Minhua Zhang, Donald Weisz, Gregory A. Elder, James Schmeidler, Rita De Gasperi, Miguel A. Gama Sosa, Donald Rabidou, Anthony C. Santucci, Daniel Perl, Edward Morrisey, and Joseph D. Buxbaum. 2005. Altered ultrasonic vocalization in mice with a disruption in the Foxp2 gene. PNAS 102: 9643-9648. Sigurðsson, Halldór Á. 2006. PF is more ‘syntactic’ than often assumed. Paper presented at InterPhases: A Conference on Interfaces in Current Syntactic Theory, University of Cyprus, Nicosia, Cyprus. Smith, Jennifer. 2000. Synchrony and Diachrony in the Evolution of English: Evidence from Scotland. Doctoral Dissertation, University of York, England. Sobin, Nicholas. 1994a. An acceptable ungrammatical construction. In The Reality of Linguistic Rules, eds. Susan D. Lima, Roberta L. Corrigan and Gregory K. Iverson. Amsterdam: John Benjamins. Sobin, Nicholas. 1994b. Why we should reconsider teaching grammar: a response to the romantic position. Language Quarterly 32: 51-74.
323
Sobin, Nicholas. 1997. Agreement, default rules, and grammatical viruses. Linguistic Inquiry 28: 318-343. Sorace, Antonella. 2005. Selective optionality in language development. In Syntax and Variation: Reconciling the Biological and the Social, eds. Leonie Cornips and Karen P. Corrigan. Amsterdam: John Benjamins. Spencer, Andrew, and Arnold M. Zwicky eds. 1998. The Handbook of Morphology. Oxford: Blackwell. Stroik, Thomas S., and Michael T. Putnam. 2006. Surviving Reconstruction. Paper presented at InterPhases, University of Cyprus, Nicosia, Cyprus. Tagliamonte, Sali, and Jennifer Smith. 2000. Old was, new ecology: Viewing English through the sociolinguistic filter. In The English History of African American English, ed. Shana Poplack, 141-171. Malden, MA: Blackwell. Terrace, H.S. et al. 1979. Can an ape create a sentence? Science 206: 891-902. Tinbergen, N. 1952. Derived activities: Their causation, biological significance, origin and emancipation during evolution. Quarterly Review of Biology 27: 132. Tortora, Christina. 2006. The Case of Appalachian expletive they. American Speech 81: 266-296. Trester, Anna Marie. 2003. Dialect loss in Smith Island English: A Look at non-standard subject-verb concord. Paper presented at New Ways of Analyzing Variation (NWAV) 32, University of Pennsylvania. Trester, Anna Marie. 2006. Subject Verb Concord in Smith Island English. Ms. Georgetown University. Ullman, M. T. 2004. Contributions of neural memory circuits to language: The declarative/procedural model. Cognition 92: 231-270. Uriagereka, Juan. 1999. Multiple Spell Out. In Working Minimalism, eds. Samuel David Epstein and Norbert Hornstein. Cambridge, MA: MIT Press. Vargha-Khadem, F, K Watkins, K Alcock, P Fletcher, and R Passingham. 1995. Praxic and Nonverbal Cognitive Deficits in a Large Family with a Genetically Transmitted Speech and Language Disorder. PNAS 92: 930-933. Vargha-Khadem, F., K. E. Watkins, C. J. Price, J. Ashburner, K. J. Alcock, A. Connelly, R. S. J. Frackowiak, K. J. Friston, M. E. Pembrey, M. Mishkin, D. G. Gadian, and R. E. Passingham. 1998. Neural basis of an inherited speech and language disorder. PNAS 95: 12695-12700.
324
Weinreich, Uriel, William Labov, and Marvin Herzog. 1968. Empirical foundations for a theory of language change. In Directions for historical linguistics, eds. W. P. Lehmann and Yakov Malkiel. Austin: University of Texas Press. Wennersten, John R. 1992. Maryland's Eastern Shore: A Journey in Time and Place. Centerville, MD: Tidewater Publishers. Wexler, Kenneth. 1994. Optional infinitives, head movement and the economy of derivations. In Verb Movement, eds. David Lightfoot and Norbert Hornstein. Cambridge: Cambridge University Press. Wexler, Kenneth. 1998. Very early parameter setting and the unique checking constraint: a new explanation of the optional infinitive stage. Lingua: 23-79. White, Lydia. 2003. Second Language Acquisition and Universal Grammar. Cambridge, UK: Cambridge University Press. Wiklund, Anna-Lena. 2001. Dressing up for Vocabulary Insertion: The parasitic supine. Natural Language and Linguistic Theory 19: 199-228. Wilson, John, and Alison Henry. 1998. Parameter setting within a socially realistic linguistics. Language in Society 27: 1-21. Winford, Donald. 1996. The problem of syntactic variation. In Sociolinguistic Variation: Data, Theory, and Analysis: Selected Papers from NWAV 23 at Stanford, eds. Jennifer Arnold et al. Stanford: CSLI. Wolfram, Walt. 2002. Language death and dying. In The Handbook of Language Variation and Change, eds. Jack K. Chambers, Peter Trudgill and Natalie SchillingEstes. Malden, MA: Blackwell. Wolfram, Walt, and Natalie Schilling-Estes. 1995. Moribund dialects and the language endangerment canon: The case of the Ocracoke brogue. Language 71. Wolfram, Walt, and Natalie Schilling-Estes. 1998. American English: Dialects and Variation. Malden, MA: Blackwell. Wolfram, Walt, and Natalie Schilling-Estes. 2003. Parallel development and alternative restructuring: The Case of weren't intensification. In Sociolinguistic Dialectology: Papers in Honour of Peter Trudgill, eds. Jenny Cheshire and David Britain, 133–153. Amsterdam: John Benjamins. Yang, Charles. 2004. Toward a theory of language growth. In Variation and Universals in Biolinguistics, ed. Lyle Jenkins. Amsterdam: Elsevier.
325
Zanuttini, Raffaella. 1997. Negation and Clausal Structure. Oxford: Oxford University Press. Zanuttini, Raffaella. 2001. Sentential negation. In The Handbook of Contemporary Syntactic Theory, eds. Mark Baltin and Chris Collins. Malden, MA: Blackwell. Zoerner, Ed. 1996. The Case of &'-adjunction to VP. In Configurations, ed. Anna-Maria Di Scullio. Somerville: Cascadilla Press. Zwicky, Arnold M., and Geoffrey K. Pullum. 1983. Cliticization vs. inflection: English n't. Language 59: 502-513.
326
