Phonelics, Univenity of Helsinki 39, 129-136. PRODUCTIVITY AND STEREOTYPY IN LANGUAGE DISORDERS. Mick Perkins. Speech Science Unit, University ...
t29
I
I
Proccedings oJ thc Thitd Cotgress of,he er&olional Cli6ical Phonelics atd Linguittics AtsociotioL 9-l August 193, Eelsinki, (R. AulanLo & A.-M. Korpijaakko-Huutl€, eds.), Publicarioos of tbe Depadmenr of Phonelics, Univenity of Helsinki 39, 129-136.
PRODUCTIVITY AND STEREOTYPY IN LANGUAGE DISORDERS Mick Perkins Speech Science Unit, University of Sheffield
Abstract A common feature of
a wide range of language disorders is the high incidence of stereotyped and nonproductive language. A brief overview is given of the different ways in which these phenomena are manifested in disordered language, and their nature is discussed in terms of ihe reciprocal relationship between frequency of occunence and productivity. It is argued that
measures such as lexical type-token ratio provide only a partial picture
of nonproductive
language, and that linguistic units larger than the word also need to be taken into consideration. An analytical framework incorporating such units is presented which is able to characterize and quantify the extent to which a sample of language is productive independendy of its aetiology. Profiles of four different patients derived by the procedure are discussed.
T. INTRODUCTION a wide range of language disorders is the high incidence of slereotyped and nonproductivel language- Examples range from cases of severe aphasia where the same
A common feature of
expression is produced every time the patient attempts speech, to disorders such as schizophrenia and autism where a lack of linguistic variety may simply reflect what is sometimes called a 'poveny of ideas' or an 'insistence on samensss'. Both the extent of such rigidity in language as well as its underlying cause may thereforc vary considerably. In this paper I will briefly consider several differtnt types of nonproductive language, and will present a conceptual and analytical framework which is able to characterize and quantify the extent to which a sample of language is stereotyped or nonproductive independently of its aetiology-2
2
EXAMPLES OF NONPRODUCTIVITY IN LANGUAGE DISORDERS
Most studies of lack of productivity in the language pathology literature focus on lexis. Single words cenainly seem to be the most often repeated units in aphasia, though the frequency of fixed structures such as I can't say and I don't lnaw is sometimes commented on l?roductivity' refers to fte variability of linguistic paBeming, or more generdlly "tbat property of lhe language syste[r which enables native speakers to construct and undersland an indelinilely large number of utterances" (Lyons, 1977:76). Lack of Foductivity has beeD rcfen€d to variously as 'automaiic' (Jackor\ 1932), 'fomulaic' (Nida 1966), holistiC (Bever, 1975), 'idiornatiC (Weitueich,.l969), tlolwrcpositi@al (Van Lancker, 1987) and 'ready-made' (Lyons, 1968).
2A morc extensive account of the procedure outlined bere is provided in Perkins (forthcoming)
r30 (Buckingham, Aviakan-Whitaker and Whitaker, 1975t Nespoulous and Lecours, 1990). Repeated phrases and sentences also are found across a range of disorders such as frontal lobe damage, schizophrenia and autism, though size of unit is rarely the focus of research interest. One problem with describing such unis is that it is not always clear whether a repeated phrase or sentence is actually being used as such, or whether it is merely a 'stereotype' or rot€-learned
'chunk', more on a par with a single word- This is a crucial point and
will
be discussed in
detail below2.1. Perseveration
Lack of productivity due to the rcpetition of linguistic units has received considerable attention (see, for example, Wallesch (1990)). One of the most common types of repetitive linguistic behaviour is perseveration, defined by Buckingham, Aviakan-Whitaker and Whitaker (1979:329) as "the recurrence, out of cont€xt and in the absence of the original stimulus, of some behavioural act'. Perseveration may be immediate and continuous, as in palilalia where the final part of an utterance is repeated several times. One example from Ibayashi, Tanaka, Peng, Joanette and l-€cours (1992:243) is "This is August, August, August,
August, August, ... ". Perseveration does not always involve immediate repetition of a stimulus. 'Recurrent' perseveration, defined by Sandson and Albert (1984) as "the unintentional r€petition, afler cessation, of a previously emitted response to a subsequent stimulus", is sommon in a number of disorders including left hemisphere lesions, dementia and schizophrenia. The following example from Santo Pietro and Rigrodsky (1986) shows recunent perseveritive responses (in italics) to a picture naming task:
Stirmrlus
Resporrse
BOOK
ECG
book apple book money book
IRON
irish
BUTTER
book
APPLE PEN
MONEY
22. Speech Automatisms One of the most extreme forms of linguistic nonproductivity is the automatic rep€tition of a single word or phrase found in some types of aphasi& Some examples from Code (1982) are: sister sister; Wednesday; I want to one two one two. Such 'speech automatisms' or 'real-word recurring utterances' (ibid.) are involuntary, contextually inappropriate and produced every time, or almost every time, that speech is att€mpt€d. In severe cases, the patient is described as being unaware of the exprcssion's inappropriat€ness and as making no attempt !o suppress it, though in less severe cases there may be awareness of inappropriateness and attempts at suppression (Code, 1989). 2.3. Echolalia
One rype of nonproductivity which differs from those discussed so far is echolalia-the repetition or 'echoing' by a speaker of an utterance addressed to him-in that it consists of other- rather than self-repetition. Echolalia is found in transcortical aphasias (Caplan, 1987) and Alzheimer's disease (Cummings, Benson, Hill and Read, 1985), and in more than thrce quarters of all speaking autistic children (Frith, 1989:123). A number of studies have shown that echolalic utterances in mental deficiency (Campbelt and Grieve, 1978) and in autism (Schuler and Prizant, 1985) may have a range of communicative functions, and comparisons have been drawn with the kinds of echo repetitions found in normal language acquisition,
131
though there are generally felt to be both qualititative and quantitative differences (Schuler and Prizant, ibid.). 2.4. Topic Bias
A further source of nonproductivity in language which is often less immediately obvious and dramatic than those alrcady mentioned is 'topic bias'-ie the tendency to continually revert to the same resricted set of conversational topics. In principle, such repetitiveness may be regarded not as a primarily linguistic phenomenon, but as an indirect consequence of a more general cognitive inflexibility. Such inflexibility is a common feature of autism, and Frith (1989:l3ltnotes that "highly verbal autistic people often say'by the way -.- ', 'talking of ... ', 'well, anyway ..- ', when they are in fact not introducing a new topic". Topic bias has also been noted in leaning-disabled adults (e.9. Rein and Keman (1989) wherc it is confusingly referred to as 'verbal perseveration'), and closed head injury.
3. PRODUCTIVITY, STEREOTYPY, REPETITIVENESS AND FREQUENCY We have seen so far that nonproductivity in language disorders appears to be closely rclated to repetitiveness. One simple but fundamental point which is rarely directly addressed in discussions of repetitiveness is that in order for an instance of repetition to count as such, the repeated items must be identica.l, or at least very similar. Sameness and repetitiveness are, in fait, rwo sides of the same coin. In order to draw out the implications of this observation I will now directly consider the phenomenon of 'st€reotyped' language, a calegory of nonproductive linguistic behaviour not yet discussed. Stereotyped language ranges from the holistic roie leamed 'chunks' typical of children with autism noted by Prizant (1983) and others, to the 'nonpropositional' speech that oftsn survives in cases of aphasia where there is virtually no productive language left (Jackson, 1932; Van Lancker, 1987). Unfortunately, however, the terms 'stereotyPe, 'stereotypy'and 'stereotyped' are not always used consistendy. Some choose to emphasize the repetitiveness of stereotypes while others emphasize their 'sameness'. Frith and Done (1990:233), for example, purposeless movement or act that occurs deline stereotyped behaviour generally as "any rstereotypyr very narrowly to refer to recurring repeatedly", and Lebrun (1986) uses the term spbech automatisms. Linguists, on the other hand, tend to be more interested in the lack of stmctural variability of st€reotypes-ie their 'sameness'-than in tleir tendency to recur- Crystal (1991:326), for example, defines a steleotype as "a sequence of words which resembles a productive grammatical stucture but which in fact has been leamed as a single unit and has little or no productivity". This works well enough for idioms, proverbs, song lyrics and clich6s which are easily recognizable, but if one is trying to idendry such a nonproductive sequence in a sample of disordered language or child language, often the only clue one has to its status as an invariant structure is the frequency with which the form is repeated. This is an important poinq and as we shall see later on, in practice it will be necessary to take account both of frequency of repetition as well as degree of sameness in order to properly charact€rize degree of productivity in a language sample.
-
4. TYPE.TOKEN RATIO The most common procedure for estimating the degree of nonproductivity of a language sample is to calculate the lexical type-token ratio (LTTR) by dividing the number of different worils (types) by the total number of words (tokens). This provides a dire.ct measure of repetitiveness at word level. For example, if there arc fifty different words in a hundred word sample, this gives an LTTR of 5CVl00 = 0.5. Generally speaking, a higher LTTR implies a greater diversity in vocabulary and a lower LTTR implies a more repetitive or stereotyped vocabulary. The LTTR measure is not without shortcomings, the main one being that LTTR varies as a function of sample length. However, three fairly recent critical studies of the measure (Hess et al., 1986; Hess Haug and Landry, 1989; Richards, 1987) sugge.st that the
r32 basic LTTR is a valid comparative tool for language samples provided that a) rhe language samples to be compared are the same sizel b) sample size is measured in number of tokens; and c) the minimal sample size is somewhere between 250 and 350 tokens.
5. PHRASE AND SENTENCE STEMS One thing that LTTR does not do-and which no other analytical procedure attempts to do, as far as I am aware-is analyse degree of linguistic variability above the level of the word. We
have seen, however, that phrases and sentences may also be repeared, either wholly or partially, and although this will be indirectly reflected in the LTTR in that the same words will be used on each occasion, it provides no direct information on repetition of multiword strings. Thus there appears !o be no procedure at present for measuring syntactic as opposed io lexical variability in language samples. One problem in devising such a procedure is deciding on what to count as one's urit of analysis. Insofar as a word string lacks productivity-ie its lexical content and syntactic structure are the same whenever it is used-it would seem appropriate to regard it as a lexical unir This is the way that idioms are usually treated by linguists. The other exteme of a word string with full productivity would be an instance where a given syntactic structure was never used with the same words more than once, apart from instances of pure coincidence. This view of language whereby every word string is either a lexical unit or an instantiation of a fully productive syntactic structurc is implicit in classical generative grammar (eg Chomsky, 1965). However, it has been pointed out by a number of linguists that normal-not to mention disordered-language contains many forms which are neither completely invariant nor completely productive. Lyons (1968:178) notes that sentences li&.e What's the use of worrying? and Wat's the use of getrtng everything ready the night b$ore? sharc L common invariant 'schema' containing two variable 'slots' which can be shown x Wat's the use of -ing ...? Pawley and Syder (1983) have argued that in addition to a generative grammar and a lexicon, a speaker probably has several hundred thousand sentence 'stems' which have to be learned and stored separately in the same way that individual words are. For example, although the meaning of I'm really glad you could come could also be expressed as I'm in a yery BLad state a.s a result of your coming or Your coming has brought me real gl.adness or in any of a virtually infinit€ variety of other ways, only the former would commonly be used by native speakers of English. However the senlence itself is not unproductive. Therc are many possible variants such as I|e're really ghd you could bing Harry or She was really glad I could help her. What is invariant is the underlying sent€nce stem 'NP - BE+tense - really glad (that) - NP VP'. Pawley and Syder (ibid.) also provide evidence to suggest that there is a positive conelation between a speaker's fluency on a given occasion and the number of nonproductive and semiproductive forms used. Fillmore (1979) similarly suggests that the proportion of nonproductive and semiproductive forms can account for individual differences in speaker style. The degr€e of productivity of a wordstring is hard to establish c pnan- Fraser (1970) proposes a hierarchy of 'frozenness' in terms of the range of transformations that can apply to a given form. For example, What's the use of --in9...2 has no related declarative form The we of --ing is ... which could be given in response. This is a valid analytical approach if one is in a position to elicit speaker intuitions from an informant. It has severe limitations, however, if one's informant is language disabled, or if one is analysing a transcript of language data-ie one is looking at language performance rather than directly at competence. In such circumstances, a better approach is to consider the productivity of a string as a function of its frequency of occurrence. This reflects a general principle implicit in the discussion of ster€otypy above that the more frequendy an expression occurs, the less productive it is likely to be. Such a principle has also been observed in discussions of LTTR referred to above, and in discussions of idioms (eg Chafe, 1968) where it has been noi€d that for idiomatic structures such as tic& the bucket and can of worms which have a literal counterpart, the text frequency of the literal counterpart is much lower than that of the idiomatic fonn. In the analytical framework proposed below, therefore, the units of analyis are the sent€nce or phrase 'stem'-or'stem type -and its linguistic realizations--or 'stem tokens'.
133
6. OUTLINE OF A PROCEDURE FOR ANALYSING NONPRODUCTIVITY The framework proposed in this section is an illustrative outline of an analytical procedure which is cunently being automated to enable it to be run on a microcomputer (Perkins & ElGihani, forthcoming). Although one or two points of detail have still to be fixed and some revisions are likely as a result of the automation process, the overall conceptual shape is clear. The procedure, which involves the measurement of phrase- or sentence-'stem typetoken ratio' (STTR), is intended to give a more comprehensive representation of what is perceived as 'degree of sameness', 'ster€otypedness', 'rigidity' or 'repetitiveness' in a language sample than is cunendy possible merely through the calculation of LTTR. LTTR provides a measure of range and productivity of vocabulary, and STIR pmvides an additional measure of lexico-grammatical variability in terms of the frequency of occurrence of phrase and sentence stems. In addition ro this, a 'variability quotient' (SVQ) is calculated for each stem, and subsequently for the whole sample, which indicates the extent to which stems are always realized by the same form, as opposed to a range of forms. Finally, a bontiguity rating' (CR) provides a measure of perseveration and echolalia. The procedure includes the following measures:
6.1. Lexical Type-Token Ratio (LTTR) This provides a measure of how frequendy the same words occur. It is calculat€d by dividing the number of different words (types) by the total number of words (tokens). Compared samples should have the same number of tokens, and minimum size should be 350 tokens, as not€d above. 6.2. Stem Type-Tok€n Ratio (STTR)
The STTR indicates how frequently the same phrase and sentence stems occur. It includes the
following three steps:
A- Identifv all the'stem tokens'in
a languaSe sample
A string of words counts as a stem token if it is at least 2 words long, and a proportion of betwesn 50% and 100% of its component morphemes are identical to those of at least one other string in the sample. A stem (or 'stem type') is the set of morphemes which remains invariant across the range of stem tokens. (Optimum sample size in terms of number of stem tokens still r€mains to be established.) For example, the following utterances : fracturpd my skull fracturing my skull fracturgd his skull would count as tokens of the same stem since all of their morphemes, apart from those underlined, are identical. The common stem (st€m type) could be shown as:
fractur* * skull
above stem type in a) has three tokens, which gives
al STTR of
1/3 = 0.33
STTR, the more repetitive the whole language sample is.
t34 6.3 Stem Variability Quorient (SvQ) This measures the extent to which stems are realized by the same form as opposed to a range of forms-in other words, how productive they are. It is calculated as follows:
a stem has 2 different variants for 5 tokens, its SVQ will be 2l5 = 0.4. However, if all its 5 tokens are identical, its SVQ will be ll5 = O.2.In the example used in B the SVQ is 3/3 = 1. The lower the Variability Quotient, the morc stereotyped the €xpression
is.
the lower the degree
of
6.4 Conriguity Rating (CR) This is a measure of the proportion of stem tokens which are 'contiguous' (ie which occur immedialely next to a token of the same sbm with no intervening utterance). It is a measure of the degree of perseveration in a sample. In dialogue, the final utterance of the previous speaker is also taLen into account in order to measure the degree of echolalia. Both lexical contiguity (I-CR) and stem contiguity (SCR) can be calculated alriough only the latter has been included in the examples below. Thus for any language sample, application of the above procedure will provide an overall prohle of its degree of stereotypy/productivity/repetitiveness in terms of its LTTR, STTR, mean SVQ and CR.
In order to give some idea of how the procedure can be used as a comparative !ool, I will briefly comparc the profiles of four patienB. Figure 1 shows the repetitiveness profiles derived from analysing four samples of comparable data of the same size3-
0.9 o.8 o.7
Vdc€ DiEord€r€d
-'-A +B
o.6
H€ad lnjurcd
0.5 Fluent Al'hasic
o.4
I
- -C ---+--- D Ru€nt Aphasic 2
0.3 o-2 0.1 0 LTTR
Flgurc
1
STIR
SO
A comparison of four petisnt profflas
3Eacl sunpte consisted of 3l claNes, ard for the m€asurement of LTTR the first 1?l word td(etrs were us€d-
135
A is a patient with a psychogenic voice disorder but with normal language; B is the headinjured patient described above; C and D are both fluent aphasics. The first thing !o note is that the language-disordered patients B, C and D all show a greater degrce of lexical and stem repetition than the non-language disordered patient A- The value of the STTR m€asure is evident in the fact that it can show such a range of variation when the LTTR shows so little. Although C's lexical repetitiveness (LTTR) is vinually identical to that of ihe two other language-disordered patients, his relatively high STTR score indicates that the structur€s in which the repeat€d words are used are considerably varied. The fact that STTR (frequency) does not co-vary with SVQ @roductivity) shows that both measures are necessary to give an overall picturc of degree of repetitiveness. Finally, the CR of patient D indicates that the major difference between her profile and that of patient B is the amount of perseveration.
7. CONCLUSION The analytical procedure outlined here provides an objective means of quantifying the extent to which a sanple of language is nonproductive, repetitive or stercotyped by focussing on the reciprocal relationship between the frequency of occurrence and the degree of productivity of its component elements. The novelty of the approach lies in its identification of units of analysis larger than the word, and in the way in which rlrcir variability is calculated. It should be possible to use the procedure to establish norms of repetitiveness in different types of language sample, to assist in rhe differential diagnosis of a range of language disorders in both adults and children, and to monitor remediational and developmental language change in individuals- It is also envisaged that the procedure could prove useful in the contrastive analysis of language samples of many kinds in areas ranging from sociolinguistics m literary stylistics.
E. REFERENCES Bever, T.G. (1975) Cerebral asymmetries in humans are due to the differentiation of two incompatible processes: Holistic and analylic- Annals of the New York Acadenry of Science,263,251-262, Buckingham, H.W., Aviakan-Whitaker, H. and Whitaker, H. (1975) Linguistic structures in stercotyped aphasic sph- Linguistics, 154/155, 5-13. Buckingham, H.W., Whitaker, H. and Whitaker, H.A. (1979) On linguistic perseveration. In Whitaker, H. and Whitaker, H.A. (Eds) Studies in Neurolinguistics, Volume 4. (New
York Academic Press).
Campbell, B. and Grieve, R. (1978) Social and attentional aspecls of echolalia in highly echolalic mentally retarded persons. Am ertcan Joumal of Mental Diseases,82, 414-416. Caplan, D. (1987) NeurolinSuistics and Linguistic Aphasiology (Cambridge: Cambridge University Press) Chafe, W. (1968) Idionaticity as an anomaly in the Chomskyan paradigm. Foundations of
language,4,
109.127
.
Chomsky, N. (1965) Aspects of the Theory of Syntax. (Cambridge, MA: MIT Press) Code, C. (1982) Neurolinguistic analysis of recurrent utterances in aphasia. Corter,18,141152.
Code, C. (1989) Speech automatisms and recuring utterances. In C. Code (Ed) The Characterisrtcs ofAphasia (London: Taylor & Francis). Crystal, D. (1991) A Dictionary of Linguistics and Phonetics (3rd edition) (Oxford: Blackwell). Cummings, J.L., Benson, D-F., Hill, M.A. and Read, S. (1985) Aphasia in dementia od the Alzheimer type. Neurology, 35, 3W39. Fillmore, C.J. (1979) On fluency. In C.J. Fillmore, D. Kempler and W,S-Y.Wang (Eds) Individual Differences in Language Ability and language Beheviour (New York: Academic Press).
t36 Language' 6' 22Fraser, B- (1970) Idioms within a transformational grammar ' Foundttions of 42.
Done. D.J' (1990) Stereotyped -behaviour in madness and in health' In S'F' Frith:'C.D. '^'"&;;; and ;; ai il;th i ids) The N;urobiotogv of Behaviourat stereorypv (oxford" Oxford University Press). e.irrt.'iJ. riS89l eu tiim: Exploinins the EniSma (Oxtotd: Blackwell)' il iiil 6.*- 5Jri"-i.-k-.'iur ino lan?rv, R.C: (198b) Sample size and type-token ratios for oral ;iild;;. i;lu ar iI speicn aid Hearins Riiearch'2e' r2e't34' . ,"Jlu"atv, R-c. (i9i9) The reliabilitv of tvpe-token ratios for th.e s"rtoot ug" cttii iten. Joumal of Speech and Hearing Research' 32' 536-
'-f;';il;L;T;;';i;;i H*;:E.w:i;i;,-i.i. "ruJil;;&;";f 540.
f., fanata, R.' Peng, F.C'C.' Joanette, Y' and L€cours, AR' (1992) Palilalia $.tfr tlavas"li, --Gurnutii intracerebral heirorrhage (A case repon). loumal of Neurolinquistics,7, 241249.
fu"fson, f.ff- $932) Selected Writings of John Hughlings lackson' (Loldon: Hodder and Stoushton).
lof,r'i""ifi.'iiSaa)
Studies in language behavior:
I. A program of
research. Psychological
Monoeraphs,56, l-15.
L"d;; f.
(
i9li6i eptrasia witt recunent utterance: t rcview ' British Joumnl of Disorfurs of
Conuru"nication, 21, 1 l-2O.
Ly;;,i:Ti;6st-L;;ar"tioi
to Theoretical Linguisticr (cambrid8e: Cambridge universitv
Press)-
I-"on..l-'.'i tSUZI Semantics. (Cambridge: Cambridge University Press)'
A.R. ( l-990) Pourquoi I'aphasique peut-il dire "Je ne peux.pas i-L. -d t Nil;l;;;: ' 'peut pas li chariter"?: De l'ini€r6t iles dissociations verbales dans ne"ouit r'""jiJiLi'o"i;ul id';; ;; r;t";- i"rt'".idriJ Cahiers du centre Interdisciplinaire des Sciences du
Unsape- (forlouse: Universid
de Toulouse-le
Mirail)'
Mouton)' Nid;;"i:aid66t, Sinopsis of Enelkh Sqtarc (2nd ed) CIhe Hague: nativelike selection and puzzles for linguistic theorv: i.ft. 3"d"i i,;;L;;.;;d itdtll"r*o ^ -'tiiir*ili" j.c.' (Eds) and Language R.W. schmidi and Richirds n'.,iriv. iir' (London: Longman). Communication. -i?;itit"o.itgl -. Ripetitiveness in language disorders: A new analvtical P".k"#; fi:il. nrocedure.
p"rfi"iilii. e El-Gihani, A. (forthcoming) The computational natural language.
prir*i, s,M. if963) Language ^ ';;"di;ti:d;aild 307.
analysis of repetitiveness in
acquisition and communicative behaviour in autism: toward 296o-r-iiie 'vTttote''or it. lournal of Speech and Hearing Disorders '
8'
n"i". i.p. anO Keman, K.T. (1989) The funcrional use of verbal perseverations-by qd^ults who '*';;;;;trll;;wiLa. ilicatio" and Trainins in Menal Remrdaion' 24' 38I-389' -, 'iilirbaT)-itpa"k;; ratios: whal"do thev reallv tett us? Joumal of Child -
ni"fr'rlid., Ianpuase,14,201-209.
s-orofi. i.'""a ArU"n, M.L.
(1984) Varieties of perseveration' Neuropsycholttgia,22, 715-
732.
o, M.J. and Rigrodsky, S. (1986) Pattems of oral-verbal perseveration in adult lttnguage'Z9' l-17. In E. Schopler and c. B. Mesibov (Eds), s"tiiei,"n.'""a iriianr, e.:M. itsas) Echolalia. (New York: Plenum)' Commttnication Problems in Autism V"J;;6;'5ii987) Nonpropositionai speech: neurolinguistic- studies' In A' W' Ellis (Ed)
S*iJii"t
aohasics. Brain and
'
- p-iriii'i, the Psv;holoiv ;f Innguagi volume 3 (London: Lawrence Erlbaum) w"ir"i;r,,- c.-w. (r'sso) Hedetiti;e ierbal behaviour: functional and neurological considerations. Aph asiology, 4, 133-154-
the analysis of-idioms. In J. Puhvel (Ed) Substance and (Berkeley: University ol Califomia Press)' ktnguage oi Structure
w"i;r;i.h,-U. dgOSi iioUld;i in
PUBLICATIONS OF THE DEPARTMENT OF PHONETICS
EN JULKAISUJA
r39*
PROCEEDINGS OF THE THIRD CONGRESS OF THE INTERNATIONAL CLINICAL PHONETICS AND LINGUISTICS ASSOCIATION, 9-11 August 1993, Helsinki
REIJO AULANKO
&
Edited by ANNA-MAIJA KORPIJAAKKO-HUUHKA
FEBRUARY 1994
