MainP11. A 8 4 a v . 1 4 1992

LINGUISTICVOLUME 14 1992

TLANTICA

Journal of the Revue de Atlantic Provinces l'association de linguistique

Linguistic Association des provinces atlantiques

LINGUISTICZ-VTLANTICAH J =s

OFFICERS OF THE ATLANTIC PROVINCES LINGUISTIC ASSOCIATION

President 0 Président

Vice-President 0 Vice-Président

Secretary 0 Secrétaire

Treasurer 0 Trésorière

Members-at-large 0 Membres élus

Editor / Rédacteur en chef Associate Editors / Rédacteurs adjoints Review Editor / Rédacteur des critiques de livres Newsletter Editor / Rédactrice du bulletin

LINGÜISTICA atlantica is a publication of the APLA published in a volume of one number annually.

SubscriptionsThe subscription rate of $15.00 entitles the subscriber to membership in the Association.

Guide to AuthorsLINGÜISTICA atlantica publishes articles on descriptive and theoretical linguistics. Manuscripts (in two copies and on diskette) should be sent to the Editor at the above address, and should conform to the Style Guide at the end of this volume

BUREAU DE L'ASSOCIATION DE LINGUISTIQUE DES

PROVINCES ATLANTIQUES

Harold PaddockMemorial University of Newfoundland

Anthony Lister University of New Brunswick

William Davey University College of Cape Breton

Aleksandra Steinbergs Memorial University of Newfoundland

Rose Mary Babitch Centre universitaire de Shippagan

Virginia Motapanyane University of New Brunswick, Saint John

Jim BlackWladyslaw Cichocki & Maurice Holder

Barry Miller Margaret Harry

LINGUISTICA atlantica est une publication annuelle de l'ALPA qui paraît en un volume d'un

numéro.

AbonnementsLe tarif d'abonnements de $15.00 permet au titulaire de devenir membre de l'Association.

Guide à l'usage des auteursLINGUISTICA atlantica publie des articles portant

sur la linguistique descriptive et théorique. Les manuscrits (en deux exemplaires et disquette, et

conformes au Guide Stylistique imprimé dans ce volume) sont à envoyer au rédacteur en chef à

l'adresse ci-dessus.

LINGUISriCA atlantica 14 1992ISSN 1188-9932

published by 0 publié par

Linguistics Department, Memorial University of Newfoundland St. John's, Newfoundland, Canada, A1B 3X9

(Numbers 1 - 1 3 were published between 1978 - 1991 as:Les numéros 1 - 1 3 ont été publiés dans la:

Journal of the Atlantic Provinces Linguistic Association Revue de l'association de linguistique des provinces atlantiques

ISSN 0706-6910)

This journal is a member of the Council of Editors of Learned Journals.

Cette revue est membre du Conseil des rédacteurs des revues savantes.

Map by Nicholas Sanson, published by Pierre Carte de Nicholas Sanson, publiée par PierreMortier, Amsterdam, c. 1696, Partie Orientale de Mortier, Amsterdam, v. 1696, Partie Orientale l ’Amérique Angloise, held at: de l ’Amérique Angloise. De la collection du:

Centre for Newfoundland Studies, Queen Elizabeth II Library, Memorial University o f Newfoundland

Articles

G

M AI

VIRGINIA M(

Vg aL0

Book Revigi Critiques de

IRENE My

LINGUISTICZ-VTLANTICAZ i .............

Articles

GISÈLE CORBEIL

MAURICE HOLDER ANNE MACIES ROLF TURNER

CHENG LUO

VIRGINIA MOTAPANYANE

MING-MING PU GARY PRIDEAUX LOIS STANFORD

Book Reviews Critiques de livres

IRENE MAZURKEWICH

JOHN HEWSON

CONTENTS / SOMMAIRE

The Role of Information Construction in Second Language Learning

La diphthongue «oi» dans le parler «brayon» d'Edmundston, Nouveau-Brunswick

On the Uncleftability of Direct Object in Chinese

Romance Clitics and the Structure of AGRP

Episodes as Memory Units in Discourse Representation: Anaphor Use in English and Mandarin

First Language Acquisition: Method, Description, and Explanation 0 David Ingram

The French Language Today o Adrian Battye & Marie-Anne Hintze

ACKNOWLEDGEMENTS

The Editor gratefully acknowledges the continuing support for L IN G U IST IC A atlantica of Dr. Michael Staveley, Dean of Arts, Memorial University of Newfoundland. Acknowledgements go to Stewart Moss, Graphic Artist, School of Business Administration, and to Joyce MacKinnon, Publications Officer, University Relations, Memorial University of Newfoundland, for assistance in the redesign of the journal.Financial support for the production and distribution of the journal was provided by the Social Sciences and Humanities Research Council of Canada, through a grant under the Program of Aid to Learned Journals.

REMERCIMENTS

Le rédacteur en chef remercie Michael Staveley, Doyen de la Faculté des Lettres de l'Université Memorial, pour son appui continu de LINGUISTICA atlantica. Pour leur assistance dans l'élaboration de la nouvelle forme de la revue, remerciments à Stewart Moss, artiste graphique, et à Joyce MacKinnon, du bureau des publications, Université Memorial.Une subvention pour la production et la distribution de la revue a été accordée par le Conseil de recherche en sciences humaines du Canada dans le cadre de son Programme d'aide aux revues savantes.

THE ROLE OF INFORMATION CONSTRUCTION IN SECOND

LANGUAGE LEARNING

Giselle Corbeil Acadia University

Ab st r a c t

This article describes the second phase o f a three-phase study designed to teach less successful second language students a set o f effective strategies used by successful students. The first phase consisted o f identifying strategies used by successful students in their response to information contained in teacher's corrective feedback. A closer analysis o f the successful students' protocols has revealed the effectiveness o f some strategies1 in terms o f learning compared to others and hinted at the feasibility o f finding a scale o f 'constructive' pro­cesses ranging from low to high levels o f information construction. This pa­per describes the various steps taken towards the elaboration o f such a scale. It is suggested that the lack o f constructive processing engaged in by average and below average students may affect their progress in attaining linguistic profi­ciency.

I. In t r o d u c t io n

Corbeil (1990) used thinking-aloud methodology to elucidate cognitive pro­cesses of more and less successful second language learners when they were provided with teacher's corrective feedback. The more successful students en­gaged in higher levels of information construction than did the less successful students.

For instance, they were found to identify elements of information, to explore related forms, to establish connections between existing and new knowledge, to resolve discrepancies between these two sources of information, and to engage in hypothesizing rules. In contrast, poor students frequently repeated the teacher's correction inaccurately, or if accurately without signs of understanding illustrated in their responses such as 'Je comprends', 'Okay!'. In analyzing stu­dents' protocols, it became obvious that not all processes were equally learning- oriented or 'constructive', that even among the processes engaged in by more successful students, some processes were more 'constructive' than others. These observations suggested the feasibility of devising a scale of qualitative learning or constructive processes.

!ln this paper, 'strategies' and 'processes' refer to the same sort of mental activities and are therefore used interchangeably.

L I M . U I S I I C A allant un 14 ( 1W 2) 1-15

2 Giselle Corbeil The Role of I'

Although attempts at investigating the quality of learning processes in a va­riety of domains date back to the 1950s, relatively little has been done in this re­spect in second language learning. In the area of problem-solving, for instance, Bloom and Broder (1950) found that successful problem-solvers read the direc­tions and set up their own hypotheses, as opposed to unsuccessful students, who lose sight of these directions and rather select answers on the basis of 'impressions' or 'feelings'. Similarly, in the area of human information pro­cessing, Schroder, Driver and Streufert (1967) found differences among students in their handling of tasks aimed at measuring differentiation, discrimination, and integrative complexity. While some students tended to process information in a superficial way, others attempted to integrate or connect new knowledge to previous knowledge. To capture the differences in information processing, the authors developed the following scale ranging from complete lack of integration to full integration of new knowledge with prior knowledge. Four levels of in­formation processing were identified:

1. Low integration index: No alternative interpretation is considered. Conflict or uncertainty is seen as unpleasant.

2. Medium low integration index: Alternative rule structures for perceiving the event are available. Alternate and different perceptions of the same event are generated simultaneously.

3. Medium high integration index: Two conflicting or different interpretations are integrated. Ability to take another person's point of view is demonstrated.

4. High integration index: Outcomes for various comparisons are considered. More 'connectedness' between alternatives is produced.

Similar results were reported by Marton and Saljo (1976a, 1976b) who made the distinction between deep processing and surface processing of information. In surface processing, the student approaches information with the main objec­tive of reproducing it; in deep processing, an effort is made to grasp the meaning of the information and to relate this to prior knowledge.

Another major investigation into the nature of qualitative processes took place in the 1980s. Based on students' observable outcomes on a variety of aca­demic tasks, Biggs and Collis (1982) developed an instrument called the Structure o f the Observed Learning Outcome (SOLO taxonomy). This scale out­lines five increasing levels of qualitative responses, ranging from illogical con­nection to full consideration of a good number of divergent alternatives. The scale corresponds to the mental hierarchy described by Piaget:

• Pre-structural: no logical relationship to the given information.

• Uni-struc butmisse oversimp'

• Multi-stn tent with

• Relationa by the ui informati

• Extended Question ConseqiK

A similar structive eff think out lo Bereiter, Bu effort were i nations: re fflfltion. For t^e effort w ty one-wort highest leve ences to pre'"formation,O ppositionProblems tc the low est 1the highestnewPerspe( . These stu i n f l a t i o n Pr°cesses inStudy> it an,Ui* m t h e *

^ couL , esults of st

1986b), f t .

The Role o f Information Construction in Second Language Learning 3

• Uni-structural: response contains one item relevant to the given information but misses others that might modify or contradict the response. Rapid closure oversimplifies the issue.

• Multi-structural: several items in the response but only those that are consis­tent with the closer conclusion are stated. Closure is selective and premature.

• Relational: most or all of the relevant data are used and conflicts are resolved by the use of a relating concept that applies to the context of the given information which leads to a firm conclusion.

• Extended abstract: the context is seen only as one instance of a general case. Questioning of basic assumptions, counterexamples and new data. Consequently a firm closure is often seen to be inappropriate.

A similar scale was designed recently for the purpose of capturing the con­structive effort engaged in by first language readers who were instructed to think out loud while reading from informative texts (Chan 1987; Scardamalia, Bereiter, Burtis, Chan, & Mumford 1988). Five possible levels of constructive effort were developed with respect to each of three specific text-processing op­erations: relating information, questioning information, and evaluating infor­mation. For instance, under relating information, the lowest level of construc­tive effort was defined by such operations as making personal suppositions cued by one-word surface features and giving fragmentary paraphrases, whereas the highest level of constructive effort involved such operations as making infer­ences to predict and to extrapolate knowledge. As for the category questioning information, the lowest level referred to asking questions based on personal supposition while the highest level consisted of formulating situation-based problems to extend knowledge. In the category called evaluating information, the lowest level indicates provision of non-analytic affective comments while the highest level refers to the evaluation of conflicting information leading to a new perspective.

These studies have demonstrated that qualitative learning or construction of information can be captured on a scale of increasingly sophisticated cognitive processes in a variety of domains. Based on the results of the first phase of the study, it appeared that a scale of a similar nature could be devised for specific use in the area of second language learning.

One could argue that a scale could have well been devised on the basis of the results of studies designed to identify and describe second language learners' learning strategies (Holec 1987; Faerch & Kasper 1980, 1983; Naiman et al., 1978; O'Malley et al. 1985; Rubin 1975, 1981,1987; Tarone 1980; Wenden 1986a, 1986b). Had that been the case, however, the resulting scale might not have had the depth and reliability with respect to cognitive processes that the present scale could offer. In the present study the concurrent thinking-aloud technique

4 Giselle Corbeil

was used: students speak their thoughts out loud at the time of participating in learning tasks. The results were therefore likely to demonstrate more reliability than techniques commonly used in second language studies such as observations, students' self-reports, students' semi-structured and structured interviews. As pointed out by other researchers (Cohen & Hosenfeld 1981; Ericsson & Simon 1984), these methods of investigation may not be representative of learners' spontaneous practices.

2. S u m m a r y o f t h e f ir s t s t u d y

2.1 Participants

The participants were 14 anglophone adult volunteers - seven men and seven women, ranging in age from 20 to 35 - enrolled in intermediate courses of French in the Continuing Education Program offered by one of the Community Colleges in Toronto. They had previously taken 2 courses - 84 hours - of French for beginners. Most of them had been exposed to French while in high school.

While half of them were successful students scoring between B+ and A, the remainder were less successful students scoring between D+ and C+. They were judged as such by their respective teachers on the basis of their previous linguis­tic proficiency scores which consisted of a global score on the ability of students to carry out various communicative tasks with accuracy and fluency.

2.2 Methodological approach

To capture students' mental operations while they were engaged in commu­nicative activities, on-line reporting through thinking aloud was considered most appropriate. Although there have been criticisms of using verbal reports as data, concurrent verbal protocols are considered to be reliable, because the thoughts are verbalized at the time the information is heeded by the central pro- j cessor (working memory) (Nisbett and Wilson 1977, Ericsson and Simon 1984).

To ensure that students understood the process called 'thinking out loud' the investigator met individually with each student who was equipped with a tape recorder and a microphone. A conversation was initiated by the investigator and | students were prompted to voice their thoughts every time the investigator de­tected a puzzled look, a pause or hesitation. The training lasted approximately 15 minutes with each student, and was interrupted on occasion to have students listen to their tapes and assess the reliability of their recordings.

Upon completion of the training, students met, after class, in groups of 2 ,3 , or4 with the investigator. They were equipped with individual tape-recorders and microphones, and were audio-taped for the duration of the sessions. One 2-hour

The Role o f Information Construction in Second Language Learning 5

meeting was held every week for a period of 5 weeks with different grouping combinations, amounting approximately to a total of 20 hours. Topics of discus­sion revolved around everyday situations. Students were reminded often through these sessions to voice their thoughts into their microphones.

2.3 Summary of results

All tapes were transcribed verbatim. An analysis of the protocols was carried out to identify the strategies used by both the successful and less successful stu­dents. Instances of these strategies are listed below; a full account can be found in Corbett's (1990) study.

Successful students' strategies:• Attempt to recognize elements of information by identification of construc­

tions.• Questioning two words similar phonetically but dissimilar semantically.• Re-use of unfamiliar material.• Attempt to identify unfamiliar elements in terms of metalinguistic catego­

rizations.• Hypothesis-testing.• Recognition of corresponding patterns in both languages and initial attempt

at making rules.• Complication of straightforward productions.

Less successful students' strategies:• Inaccurate repetition of all elements of teacher's corrections.• No repetition of teacher's correction.• Acknowledgement by means of translation into English without any attempt

at French.• Off-focus attention.

3. D ESCRIPTION OF THE PRESENT STUDY

3.1 Refinement of existing categories

Preliminary work on the data consisted of first grouping together students' protocols according to their linguistic performance and second, grouping to­gether protocols of the same strategic content and giving them a meaningful la­bel (see Corbeil 1990). It was then possible to design a scale of constructive pro­cesses. Before proceeding, however, it was necessary to establish a more refined categorization of students than the one set up in the first phase of the study which discriminated between two groups: successful (A, A-, and B+) and less suc­cessful (C+, C, C-, and D+) students. It was hypothesized that students with A and B+ standing were using more complex constructive processes than the ones

6 Giselle Corbeil | ------------------------------------------------------------------ ----------- —-------------------------------- -1

Iwith C+ standing, and similarly for students with C+ standing compared to stu- § dents with D+ standing. The three following categories were then devised: | above average (A, A-, and B+), average (C+, C, and C-) and below average | (D+). Strategies used by students belonging to different linguistic performance i categories were discerned. I

A closer look at the strategies listed under each linguistic proficiency category | revealed that a number of the same strategies were used by students at different 1 levels. For instance, strategies like the repetition of the corrected word or the | contrast between the corrected and the incorrect forms were found among | learners of various levels of linguistic proficiency.lt thus became necessary to | eliminate from a category strategies common to other ones in order to be left * with mutually exclusive categories. This was done by screening strategies of identical content across categories and by eliminating from a higher category a strategy found in a lower one. This process was helpful in singling out the i strategies that were responsible for the attainment of a specific level of linguistic performance.

3.2 Summary of results2

Identification o f Above Average Students' Processes• Exploration of related forms followed by hypothesis-making: students are

provided with a novel expression and attempt to figure out other related but novel forms.

• Elaboration of information: students who are not being addressed by the teacher try to re-activate words which have shared semantic features.

• Re-activation of contrasting but related categories: students hear a word be­longing to a category and re-activate a contrasting category. For instance, in the case of a word in the masculine form, they then attempt to retrieve the feminine form.

• Searching for other contexts of use: students are not familiar with the new information and question its use in other contexts.

• Attempt at making rules: students attempt to make rules based on a reliable sample of similar constructions.

Identification o f Average Students' Processes• Highlight of a correction made by use of a literal but unusual translation in

English: students hear a new word, notice the unusual form compared to the first language and contrast the current form in English with the unusual form through use of the literal translation of the French expression.

instances of these processes can be found in Corbeil's (1990) study.

■ i!!i_Corbeil| T/ig Role o f Information Construction in Second Language Learning 7

spared to stu-1 hen devised!-lowaverage!Perform ance!

!ncy category •s at different word or the )und among necessary to! ler to be left strategies of ;r category a ling out the

Contrast made between their incorrect response and correct response pro­vided by the instructor: students are given a correction which they contrast with the incorrect one.Search for parallel structures: students are given a correction and re-activate parallel structures.Repetition of teacher's corrections with signs of understanding: students re­peat teacher's corrections and add signs of understanding such as 'Yes', 'Je comprends', 'Okay!'.Minimal transformation of teacher's corrections: students who are not ad­dressed attend to corrections given to peers and attempt minimal transfor­mations3.

Identification o f Below Average Students' Processes• Inaccurate repetition of teacher's corrections: students are provided with a

correction but do not repeat it accurately.• Accurate repetition of new information provided to a peer: students repeat

of linguistic % accurately the correction addressed to a peer without signs of understanding.I • Attention to only 1 alternative when 2 are suggested: students are provided ii with information about two competing phrases, but pay attention to only one. D • Off-focus attention: students express irrelevant comments instead of focus- % ing on the correction.

No repetition of teacher's correction: students who are corrected by the teacher do not repeat the correction.students are |

■ related but II

ssed by the g ares. | a word be-1 instance, in I retrieve the I

ith the new

in a reliable

3.3 Design of a scale of constructive processes

A first procedure consisted of setting up two main categories to account for strategies where the construction of information and contributions to advance learning were questionable and for the more complex strategies which gave evi­dence of advancing learning. The first category was called Minimal attention to the teacher's corrections; the other was called Increased attention to the teacher's corrections.

The following processes were grouped under the category Minimal attention to the teacher's corrections:

i:mslation in jared to the msual form

3Minimal transformation consists of supplying on one's own uncomplicated gram­matical or lexical items in relation with the corrected word. For example:

Student J'ai allé à l'égliseInvestigator Tu ES ALLÉ à l'église...Student (non-addressed) (Je suis allé à mes cours...)

8 Giselle Corbeil

• No obvious response to teacher's corrections.• Simple acknowledgement of teacher's corrections.• Translation in English without any attempt at French.• Inaccurate or incomplete repetition of teacher's correction.• Mere repetition of the correction without signs of understanding.

These strategies do not assure accuracy and fluency of further production and without a more complex transformation of information, no significant progress towards linguistic proficiency is likely to be achieved.

However, a strategy which consists of repeating the teacher's correction with signs of understanding will give students more chances to produce similar ex­amples with accuracy and eventually with fluency. This strategy could then be listed under the category Increased attention to the teacher's corrections, and ranked at level 1 to account for its relatively low complexity and its small contri­bution to the learning process. This level - the first in a five-level scale of con­structive processes - is described as:

Level 1: Reconstructing all elements of the teacher's correction, with inflec­tions or signs of understanding used to suggest comprehension of critical items in the utterance. Students who are corrected by the teacher, and peers who lis­ten, repeat the information accurately and completely. They use expressions which show their understanding such as 'Okay!', T understand', 'I remember, now'. They translate into English whenever necessary, in order to remember the new material or whenever the new material does not represent a perfect map­ping with English.

Example (l)4 illustrates this level:

(1) S/A J'ai fait le badminton mercredi soir.Invest. Tu as fait DU badminton...S/A Oui.. .J'ai fait DU badminton...S/-A Okay! You played badminton/Tu as fait DU badminton)

4 Abbreviations used in the examples:S/A = Student addressed. This refers to the student who is given

the correction.S/-A = Student not addressed. This refers to the other students

who listen to the student who is given the feedback.Invest. = Investigator.CAPITALIZED w o r d s = Words pronounced with extra stress.( ) = Verbalized thoughts. This means the thoughts students

speak out loud into their microphones.... = Pauses, hesitations, etc.

The Role o f Information Construction in Second Language Learning 9

As demonstrated by the example, students who repeat a correction with signs of understanding are likely to repeat this instance accurately, but they cut them­selves short of a good deal of information necessary to move towards the at­tainment of linguistic proficiency. If there had been more complex mental pro­cesses involved such as questioning the preference given to the use of this par­ticular structure over another one or inquiring about other instances for which the same rule applies or does not apply, students would have increased their knowledge of the language.

A slightly more complex strategy might have consisted of contrasting the cor­rected version with the incorrect one. This represents level 2 in the scale of con­structive processes, described below:

Level 2: Perception of something different or unusual (on the surface).a) Students, whether addressed or not, repeat the teacher's correction and con­

trast the wrong and the correct ways of saying it, as in example (2a) below:

(2a) S/A J'ai attendé un concert hier soir....Invest. Tu as ASSISTÉ A un concert...S/-A (Tu as ASSISTÉ A un concert...pas, tu as attendé un con­

cert. . .1 understand... )b) Students who are corrected by the teacher and the peers who listen, observe

something unexpected or unusual about it.

(2b) S/A As-tu fait quelque chose intéressant hier?Invest. Quelque chose D'intéressantS/A Quelque chose D'intéressant... (something of interest­

ing.. .funny...)S/-A (Quelque chose D'intéressant... 'of interesting' instead of

just 'interesting'... special...)c) Students try to seek a closer association with the first language.

(2c) S/A Quand le professeur a marké le test...Invest. a CORRIGÉ le test...S/A a CORRIGÉ le test... (to mark... to correct... a corrigé)S/-A (Oh! he marked it...he corrected it...il a corrigé...)

Students who use strategies which highlight in some ways the peculiarity of the correct expression put themselves in a better position to remember and re­produce accurately and fluently this particular instance than had they only re­peated the correct expression as practised at level 1. However, they would have extended their knowledge of the language by asking for more information about various applications of each of the competing forms. As far as their progress to­wards linguistic proficiency is concerned, students who limit themselves to this

level may succeed in the long run in repeating accurately and fluently a few ex­pressions, but they deprive themselves of acquiring a good deal of vocabulary, and therefore do not make significant advances towards their command of the new language. They might have benefited more from the correction had they attempted to transform the information provided by the teacher in a slightly | more complex way. For example, they could have identified elements of infor- | mation in terms of metalinguistic categories as is done at level 3.

Level 3: Attempt at recognizing elements of information by metalinguistic I identification of constructions or noting English / French parallels.a) Students who are given the teacher's correction and/or the peers who listen f

to it try to identify the correction that is given to them.

(3a) S/A J'ai levé tard ce matin...Invest. Tu T'ES levé tard... Vous comprenez?...S/A Oh! Je ME SUIS levé... (Okay!...pronominal... avec

'être'...)S/-A (Se lever...to get up...passé composé... 'être' pas 'avoir'...

le pronom aussi... )b) Students correct their peer's speech in advance of the teacher's correction.

(3b) S/A J'ai allé à la discothèque hier...S/-A (Je SUIS allé à la discothèque... )Invest. Tu ES allé à la discothèque...S/-A (Okay! I was right.. .Je SUIS allé... tu ES allé...)

c) Students re-use teacher's corrections as often as they can and in as many dif­ferent contexts as they can. Students who attempt to identify elements of cor­rection in terms of metalinguistic statements, or who ask the teacher to pro­vide them with this information, are likely to remember the correction more accurately and to trigger other related expressions associated with the met­alinguistic label they have supplied. In the example given above, the recogni­tion of an instance of a specific category of verb classes may lead to the re­trieval of similar cases, resulting in a better grasp of verb classes. However, students who go beyond the identification of elements of information and try to analyze rules of usage are learning much more about the new linguistic system. This represents the next level which is explained below.

Level 4: Analysis of rules of usage and/or active search of parallels for com­plex and unfamiliar expressions. Students who are corrected by the teacher or peers who listen to the correction try to connect new information with related concepts from either the first or the second language (French). Students try as well to find out rules of usage based on connections they establish between the new information and the previous knowledge acquired from different sources.

10________________________________________________________________________ Giselle Corbeil

The Role o f Information Construction in Second Language Learning 11

(4) S/A J'ai appelé ma mère et j'ai dit: Soyez prête...Invest. Tu pourrais dire: Préparez-vous...S/A Préparez-vous...What's that? with the 'vous' after... This

is different from 'Ecoutez'...Sounds like a reflexive verb, like levez-vous?

S/-A (Préparez-vous... To get prepared... ) Has this anything to do with 'to get'.. .like 'to get u p '... 'se lever'?

S/-A (It's funny that 'Get' is not translated...It's like 'Prepare yourself...' instead...It's the same with 'Get up'...which is translated by 'Rise yourself'...)

Students are also found to re-use teacher's corrections, but unlike level 3, they transform the corrections in more complex ways. The investigation of contexts of use for a specific instance of correction will likely result in broadening the repertoire of these students and at the same time strengthening accuracy of pro­duction. Speculations about the construction of an expression and the retrieval of closely related forms for recognition purposes generate discussions that are likely to result in a greater knowledge of the workings of the new linguistic sys­tem.

A look at the example given above is informative in this respect. The student who recognizes some form of command of the phrase 'Préparez-vous' with which s/he is not too familiar, tries to come up with connections through which s/he will eventually recognize the construction. The other behaviours associated with this level, such as making rules of usage based on connections established between the new information and previous knowledge, are likely to promote greater accuracy and fluency of production compared to the mere identification of elements of a single instance of correction as done at level 3.

Students could have extended their vocabulary by searching for other ways of conveying the same meaning, for instance: 'Je te prends en passant... Tu te pré­pares... J'arrive... Fais vite... Ne me fais pas attendre...'. This intensive search for various expressions characterizes the next level.

Level 5: Active search for various structures to convey the same meaning, and/or contrasting of the current construction with currently used procedures.

(5) S/A Je manque mon ami...Invest. Tu t'ennuies de ton ami...S/-A Oh! I know 'Tu ennuies ton ami'... 'You bother your

friend'.. .but why 'Tu t'ennuies de ton ami...' You bother yourself of your boyfriend? That doesn't make sense... Would it be that this verb has more than one meaning and more than one form?

12 Giselle Corbeil

S/-A Does that mean we can't say 'Je manque mon ami... ' but could we say... 'Mon ami manque moi . . I heard some­thing. . .like this.. .like the other way around compared to English...

S/-A Could we say: 'J'ai hâte de voir mon ami...'?

At this level, students who are given a correction about one instance reacti­vate other related forms and attempt substantial transformations. Students also go beyond problem-posing and try to solve problems by accessing previous knowledge. This level is characterized by learners taking charge o f their own learning and relying on the teacher only when personal resources have been ex­hausted.

What more is to be gained in terms of learning from level 5 compared to level 4? The active search for various structures to convey the same meaning will no doubt expand learners' repertoire considerably, especially when this search is initiated and conducted by learners themselves. This practice, which necessitates the search of previously learned forms and the discrimination between forms which carry the same meaning and those which hold only remote links with the target expression, will result in a greater command of the language. Using a given structure to convey a specific meaning instead of currently used structures will help students get a better grasp of the subtleties of the language. Grasping these subtleties represents a significant step towards the mastery of a new lan­guage. As was pointed out in the description of this level, what characterizes level 5 is the self-directed nature of learning by learners themselves. Unlike level4, learners question the teacher only after their own investigation has failed to provide them with the correct answer, or when this investigation has brought up conflicting accounts between their interpretation and the teacher's. Another im­portant difference between this level and level 4 is the intensity or the magnitude of students' involvement in learning. Becoming autonomous learners will un­doubtedly make them more effective learners and thus more effective communi­cators.

As indicated in the first phase of the study, this classification does not imply that learners who are found at level 5 never engage in lower level processes. On the contrary, these students are found using a variety of processes that are not observed among less successful students. For instance, while successful students are observed using high and low levels, less successful students are seldom found engaging in high levels of cognitive processing.

The Role o f Information Construction in Second Language Learning 13

3. Co n c l u sio n

This study has shown that successful second language learners act upon new information in ways similar to successful learners from different fields of study. Based on those students' protocols, it was possible to discern the sophistication of some processes compared to others which led to the design of a scale of con­structive processes comparable, in many respects, to the scales previously men­tioned.

The next step to be undertaken was to validate the scale of constructive pro­cesses. There was evidence of the face validity of the scale given that a) level 5 was accessed only by the most successful students, never by average or less suc­cessful students; b) level 4 was often accessed by successful students, on occasion by average students and never by less successful students; and c) levels 1, 2 and 3 were accessed by both successful and average students but rarely by poor stu­dents. However, a more rigorous assessment was in order and constituted the third phase of the study.

It is hoped that the present study has brought another dimension to the al­ready invaluable information on second language learners' strategies. By adding a dimension of complexity of information construction, the scale of con­structive processes outlined in this study can serve as a complement to the cur­rent work done in second language learning in the area of learning strategies. For instance, while repetition of information is listed among those cognitive strategies which help remembering, a refined scale assigns either a low, mod­erate or high rating in terms of qualitative learning, depending on whether repetition was accompanied or not by signs of understanding, identification, transformation or expansion of elements of information. This refinement is ex­pected to provide students with a better indication of what a simple strategy like repetition might accomplish in terms of learning outcomes when operated upon in a constructive way, as opposed to being attended to only in a passive way. It is also expected that such a measure would help students assess their current state of involvement in their learning and motivate them to surpass this state in order to maximime learning.

The study has also provided positive answers to the questions raised at the beginning of this paper. According to the results of the study, it would appear that there are cognitive processes that are more conducive to learning than others and that access to high level processes, practised mainly by above- average students, might explain their higher rate of achievement among second language learners.

14 Giselle Corbeil

REFERENCES

BIGGS, J.B. & K.F. COLLIS. 1982. Evaluating the Quality o f Learning: the SOLO Taxonomy . New York: Academic Press.

B L O O M , B .S . & L.J. B R O D E R . 1950. Problem-solving Processes o f College Students. Chicago: University of Chicago Press.

CHAN, C. 1987. Levels o f Constructive Effort in Children's Learning from Texts. Unpublished master's thesis. University of Toronto. Toronto, Ontario.

COHEN, A. & C . HOSENFELD. 1981. Some uses of mentalistic data in second lan­guage research. Language Learning 31: 285-313.

CORBEIL, G. 1990. Successful and less successful second language learners: dif­ferences in how they process information. Journal o f the Atlantic Provinces Linguistic Association 12: 131-145.

ERICSSON, K.A. & H.A. SIMON. 1984. Protocol Analysis: Verbal Reports as Data. Cambridge, Mass.: MIT Press.

FAERCH , C. & G. KASPER. 1980. Processes and strategies in foreign language learning and communication. Interlanguage Studies Bulletin Utrecht 5: 47- 118.

FAERCH, C. & G. KASPER. 1983. On identifying communication strategies in interlanguage production. In C. Faerch and G. Kasper (eds.), Strategies in Interlanguage Communication, London: Longman, 201-237.

HOLEC, H. 1987. The learner as manager: managing learning or managing to learn? In A. Wenden and J. Rubin (eds.), Learner Strategies in Language Learning. London: Prentice Hall International, 145-158.

M A R T O N , F. & R. SA LJO . 1976a. On qualitative differences in learning I. Outcome and process. British Journal o f Educational Psychology 46: 4-11.

MARTON, F. & R. SALJO. 1976b. On qualitative differences in learning II. Outcome as a function of the learner's conception of the task. British Journal o f Educational Psychology 46: 115-127.

NAIMAN, N., M. FROHLICH, H.H. S te r n & A. TODESCO. 1978. The Good Language Learner. Toronto: Ontario Institute for Studies in Education.

NlSBETT, R.E. & T.D. WILSON. 1977. Telling more than we can know: verbal re­ports on mental processes. Psychological Review 84, 3: 231-259.

O'MALLEY, J.M., A.U. CHAMOT, G. STEWNER-MANZANARES, L. KUPPER & R.P. RUSSO. 1985. Learning strategies used by beginning and intermediate ESL students. Language Learning 35: 21-46.

RUBIN, J. 1975. What the 'good language learner' can teach us. TESO L Quarterly 9: 41-51.

The Role o f Information Construction in Second Language Learning 15

RUBIN, J. 1981. The study of cognitive processes in second language learning. Applied Linguistics 2: 117-131.

RUBIN, J. 1987. Learner strategies: theoretical assumptions, research history and typology. In A. Wenden and J. Rubin (eds.), Learner Strategies in Language Learning. London: Prentice-Hall International, 15-30.

SCARDAMALIA, M., C . BEREITER, J. BURTIS, C. CHAN & D. MUMFORD. 1988. A three-year longitudinal study of constructive response to learning. Working Report, Ontario Institute for Studies in Education: Toronto, Ontario.

SCH RO D ER, H.M., M.J. DRIVER & S. STREU FERT. 1967. Human Information Processing. Holt, Rinehart and Wiston, Inc.

TARONE, E. 1980. Communication strategies, foreigner talk and repair in inter­language. Language Learning 30: 417-431.

W ENDEN, A.L. 1986a. Helping language learners think about learning. ELT Journal 40: 3-12.

WENDEN, A.L. 1986b. What do second language learners know about the lan­guage learning? A second look at retrospective accounts. A pplied Linguistics 7, 2: 186-201.

1.1

IcéléconKarrégiqueRya(Kii

LA DIPHTONGUE «OI» DANS LE PARLER «BRAYON» D'EDMUNDSTON, NOUVEAU-BRUNSWICK1

Maurice Holder, St. Thomas University Anne Macies, Gouvernement du Nouveau-Brunswick

Rolf Turner, Applied Statistics Centre, University of New Brunswick

R É S U M É

La graphie «oi» représente généralement la prononciation [wa] en français standard, mais nous avons aux niveaux populaire et dialectal tout un éventail de variantes qui vont de [we] jusqu'à [wo]. La manière dont cette diphtongue se réalise est liée dans une large mesure au mot où elle se trouve.D’une part, un grand groupe lexical est caractérisé par l'emploi de la variable (w a), avec, à Edmundston, des réalisations surtout antérieures en [we, wæ, w ae, wa], et peu de réalisations postérieures en [w a]. D'autre part, un petit sous-groupe (pois, poids, bois (substantif), mois, trois, noix) est caractérisé par l'emploi de la variable (w a), avec des réalisations surtout postérieures en [w a, w d , w d ] , et peu de réalisations en [wa].Selon nos données, il y a dans le parler «brayon» d'Edmundston un effet marqué de l'âge sur la réalisation de la variable (w a), l'emploi de la variante traditionnelle acadienne [we] diminuant progressivement avec l'âge des sujets. Par contre, en ce qui concerne la variable (w a), la réalisation plus ferm ée [ w d ] est aussi courante chez les jeunes que chez les âgés, trait qui reflète peut-être l'influence du parler québécois avoisinant. La structure syllabique joue aussi un rôle important dans le cas de la variable (w a), la réalisation [we] étant plus fréquente en syllabe ferm ée qu’en syllabe ouverte.Les résultats sont moins décisifs en ce qui concerne les autres facteurs sociolinguistiques considérés, à savoir le sexe, et le contexte stylistique. Une comparaison avec d ’autres variétés d'acadien au sud-est et au nord-est du Nouveau-Brunswick, et en Nouvelle-Écosse, révèle des similitudes et des différences intéressantes.

1. INTRODUCTION

Les recherches sur le français acadien des Provinces Atlantiques vont s'ac­célérant depuis vingt ans, dans tous les domaines de la linguistique. En ce qui concerne la sociophonétique, il faut signaler d'une part les enquêtes dirigées par Karin Flikeid au nord-est du Nouveau-Brunswick (Flikeid 1984), et dans les cinq régions acadiennes de Nouvelle-Ecosse (Flikeid 1988, 1989), d'autre part l'en­quête dirigée par Ruth King et Robert Ryan dans l'île du Prince Édouard (King & Ryan 1986,1987,1988,1989), et l'étude de Ruth King sur le français terre-neuvien

1 (King 1978,1989). Pour ce qui est du système phonologique, deux ouvrages des§> ___________________________

% 1 Résultats communiqués au 15 colloque annuel de l'Association de linguistique desprovinces atlantiques tenu au Collège universitaire du Cap-Breton à Sydney,

i Nouvelle-Écosse, 8-9 novembre 1991.LiNC;uisi ICA ntlmitica 14 (1992) 17-54

18 Maurice Holder, Anne Macies, R olf Turner •

base existent: celui de Lucci sur l'acadien de la région de Moncton au Nouveau- Brunswick (Lucci 1972), et celui de Ryan sur l'acadien de la Baie Sainte-Marie en Nouvelle-Écosse (Ryan 1981). Ces études couvrent pratiquement toutes les ré­gions acadiennes des Provinces Atlantiques à une exception près, le nord-ouest du Nouveau-Brunswick, dont la ville principale est Edmundston. Le parler d'Edmundston, souvent appelé «brayon», n'a pas suscité de grand intérêt jusqu'ici. À notre connaissance, il n'y a pas d'enquêtes linguistiques effectuées ou publiées sur le parler de cette région. C'est donc pour remplir cette lacune que nous avons entrepris de relever le défi et d'effectuer une étude sociolinguistique dans une région qui semble plutôt vierge. Comme il s'agit d'une étude pilote, l'enquête se limite à Edmundston; il est bien entendu qu'une étude complète dev­rait s'étendre aux autres communautés acadiennes de la région, telles que Saint- Léonard et Grand-Sault.

2. PRÉSENTATION HISTORIQUE D'EDMUNDSTON, DU MADAWASKA,ET DU PEUPLE «BRAYON»

Il est toujours très utile de connaître les origines d'un peuple pour mieux comprendre son originalité, son caractère propre et les forces historiques qui ont pu influencer sa situation et son comportement. Par conséquent, nous présen­terons ici une brève histoire de la vie du peuple «brayon» d'Edmundston où nous avons effectué notre enquête.

L'origine du peuple «brayon» est complexe. Selon l'Abbé Thomas Albert (1920: 45), un historien sur la région, ce peuple est d'origine bretonne et normande, à la fois acadienne et canadienne. Une grande partie de ce peuple était d'ailleurs formée de Québécois, d'Anglais, d'Écossais et d'irlandais qui s'étaient installés dans la région du Madawaska après l'arrivée des premiers colons acadiens (Michaud 1980: 26). Les premiers colons acadiens sont arrivés au Madawaska en 1785, après avoir été dépossédés de leurs terres et bannis de l'Acadie. Ils consis­taient en une dizaine de familles acadiennes sous la conduite de Joseph Daigle et Louis Mercure, qui ont jeté au Madawaska les bases de leurs demeures pour toujours.

Il faut signaler que même si plusieurs «Brayons» viennent de souches aca­diennes, ils ne se croient ni acadiens ni québécois ni même canadiens ([kanajë] comme ils disent), mais des «Brayons», les citoyens d'une région qui s'appelle le Madawaska, ou même «la République du Madawaska».

On pourrait se demander d'où vient le mot «brayon». La version offerte par La Société historique du Madawaska (1982:4), prétend que les pionniers acadiens broyaient ou écrasaient jadis le lin, et que l'écrasement se faisait au moyen d'un instrument qui s'appelait braie. La braie était un peigne à filasse qu'on peut voir

encore au nuis créé un verbe l plement la for dans le dialecmandaient au>Alors, pour cei

La deuxièiï c'est que la pli Normandie er mes Bruyants,

Même les « blable. Cepen «brayon» avec tendant que o torique.

Selon les hi à l'arrivée des portante bou autrefois M a mdawes en 1, donc c'est le D'autres quai l'idée de la ri\

Le M adaw beaucoup plu jusqu'au lac 1 rain colonisé,qui prit fin enles Canadien Saint-Jean.

Edmundst le nom de P e dans le fleuv Falls en angl 1856, levisite du g 0 lr

Destinée à fait bon vivred'Edl^undstc

mi, * a diphthongue «oi» dans le parler «brayon» d'Edmunston, Nouveau-Brunsioick______ 19

eau g ncore au musée historique de Saint-Basile. Donc à partir de ce substantif, on a eer l réé un verbe brayer. Il est également possible que ce verbe représente tout sim- ;ré | 'lement la forme normande du verbe broyer, puisque «oi» correspond à «ai» nés Mans le dialecte normand. On raconte ensuite que quand les passants de- rfe i nandaient aux femmes ce qu'elles faisaient, elles répondaient: «nous brayons», éréi r 'dors, pour cette raison, on les a apparemment surnommées les «Brayonnes».;ou La deuxième version, offerte par un écrivain local, Guy Michaud (1980: 4), que < 'est que la plupart des ancêtres des Acadiens sont venus d'une petite région de que Normandie en France, appelée «le pays de Bray», dont les habitants sont nom- jte nés Brayants, Braytois, ou Brayons.ev- Même les «Brayons» ne savent pas laquelle des versions est la plus vraisem- ¡nt- )lable. Cependant, la plupart des gens d'Edmundston ont adopté ce terme

«brayon» avec une certaine fierté, bien qu'il y en ait d'autres qui l'ignorent, pré- endant que ce mot n'a qu'une valeur folklorique sans aucune signification his-

ÎA orique.Selon les historiens locaux, Thomas Albert (1920), et Alexandre Savoie (1976),

i l'arrivée des Acadiens au Madawaska en 1785, le terrain était le site d'une im- m sortante bourgade d'Amérindiens malécites. Ainsi, le mot «Madawaska», )nt mtrefois M adouesca et M adoueskak, est d'origine micmaque. M adoues ou ;n- nadawes en langue micmaque veut dire «porc-épic», et kak «endroit ou lieu», lUS ionc c'est le pays des porcs-épics souvent appelé «portipi» (Savoie 1976: 48).

D'autres quand même déclarent que ce mot «Madawaska» est plutôt associé à .'idée de la rivière qui coule tout près en se jetant dans le fleuve Saint-Jean.

Le Madawaska primitif, à l'époque où s'est établie la colonie acadienne, était 115 beaucoup plus vaste que de nos jours, s'étendant du nord de la rivière Aroostook

usqu'au lac Témiscouata (Albert 1920: 200). Étant donné que c'était le seul ter- ns rain colonisé, une lutte s'est engagée entre le Nouveau-Brunswick et le Maine, 01 qui prit fin en 1842. Par la suite, le territoire a été partagé entre les Américains et & les Canadiens, ceux-ci occupant la rive nord, ceux-là, la rive sud du fleuve et Saint-Jean.ur Edmundston est la capitale du Madawaska. Cette ville portait originalement

le nom de Petit-Sault à cause de la chute sur la rivière Madawaska qui se jette 8' dans le fleuve Saint-Jean, et qui, par comparaison avec le Grand-Sault (Grand é] Faits en anglais) est une chute beaucoup moins majestueuse. Cependant, en 1£ 1856, le Petit-Sault a adopté le nom d'Edmundston en honneur et en liesse de la

visite du gouverneur du Nouveau-Brunswick, Sir Edmund Walter Head.31 Destinée à rester la reine du Madawaska, elle est aujourd'hui une ville où il

fait bon vivre et où règne la fierté d'être madawaskayen. La population actuelle d'Edmundston est de 11 497 habitants. Voici une carte de la région:

20 Maurice Holder, Anne Macies, R olf Turner

Carte de la région

3. HYPOTHÈSES DE BASE

Nous venons de voir que les habitants d'Edmundston ne se disent plus Acadiens aujourd'hui, et se nomment plutôt «Brayons», citoyens du Madawaska. Mais certains viennent quand même de souches acadiennes car plusieurs habitants d'Edmundston sont des descendants d'Acadiens expulsés. D'autre part, étant donné la proximité du Québec avec Edmundston, nous es­sayerons de découvrir si le parler «brayon» serait influencé de quelque façon que ce soit par le parler québécois. Nos hypothèses sont donc les suivantes:

1. Certaines anciennes variantes acadiennes devraient encore exister dans la communauté d'Edmundston, surtout parmi les personnes âgées, mais peut-être à un moindre degré qu'au nord-est du Nouveau-Brunswick, qui est la région acadienne la plus importante.

2. Puisqu'Edmundston est éloigné des autres centres francophones du Nouveau-Brunswick, et situé entre les États-Unis et le Québec (carte ci- dessus), les variantes caractéristiques du parler québécois devraient influ­encer le parler d'Edmundston. Voici deux citations à cet effet:

Au Nouveau-Brunswick, dans la région du Madawaska contiguë à la province de Québec, le parler est fortement influencé par le «québécois». (Charpentier 1989: 178)

i la diphthongue «oi» dans le parler «brayon» d'Edmunston, Nouveau-Brunsivick______ 21

|| ...dans le Nord du Nouveau-Brunswick, plus on se rapproche du Québec, plus le1 parler s'éloigne des traits typiquement acadiens. (Péronnet 1989: 6)$| D'autre part, il y a interférence due à la langue anglaise, surtout au niveaui lexical: reste à savoir si tel est également le cas aux niveaux phonologique| et grammatical.

î 3. À cause de l'effet de la scolarité et des médias tels que la radio ou la télévi-1 sion, une lente évolution linguistique devrait se manifester sous forme de| diminution des anciennes variantes avec chaque génération. D'autres| facteurs sans doute pertinents seraient le contexte phonologique, le sexeri des sujets parlants, et le style du discours.

i 4. ENQUÊTEïi| Notre corpus est basé sur un échantillon de 21 informateurs, dont 17 de laï classe ouvrière, et 4 de la petite ou moyenne bourgeoisie. Ils étaient tous des lo-! cuteurs natifs du «brayon» d'Edmundston, choisis selon l'âge et le sexe, et inter-| viewés durant l'hiver de 1984-1985. On a employé un protocole conçu de façon à| faire apparaître dix variables potentielles, et à mettre en évidence la variation| en fonction du «style contextuel» (cf. Labov 1972/76: ch.3, Chambers & TrudgillI 1980, ch.5). Des dix variables, la diphtongue orthographiée «oi», et les occlusives| /t, d, k, g/ ont été retenues aux fins d'une étude approfondie (McKillop 1987 =I Macies 1987: 23). Les occlusives ont déjà fait l'objet de deux communicationsI (McKillop et Cichocki 1988,1989 = Macies et Cichocki 1988,1989). Il reste donc àI faire connaître les résultats obtenus pour la diphtongue «oi», ce qui est le but du1 présent travail. Voici des détails concernant le protocole d'enquête.| Chaque entrevue se déroulait en cinq étapes dont la première était notée parI écrit, et les quatre dernières enregistrées au magnétophone. Ces étapes étaient: 1) renseignements personnels, 2) images à identifier, 3) lecture de trois dia-

f logues, 4) lecture de listes de mots, 5) questions d'ordre personnel ou généralf, demandant des réponses libres. On trouvera à l'ANNEXE I les mots utilisés auxI étapes 2), 3), et 4). Quant à la cinquième étape (discours libre), cette partie du! corpus reste à dépouiller, et fera l'objet d'une communication ultérieure.

Des 21 informateurs, 19 étaient d'Edmundston, et 2 de Saint-Basile qui seÎ trouve à 5 kilomètres d'Edmundston. Ils étaient répartis sur trois générations:| les vieillards (58-80 ans), les jeunes (16-23 ans) et ceux d'âge moyen (36-55 ans).I Une distribution égale était observée dans la mesure du possible. Ainsi, nousf avons sept personnes dans chaque groupe d'âge. Il y a pourtant un légerI déséquilibre selon le sexe (neuf hommes, douze femmes - voir ANNEXE II), qui| tient au fait que certains enregistrements étaient inutilisables. Enfin, une étude

statistique a été effectuée par le Docteur T. Rolf Tumer du Centre de statistique appliquée de l'Université du Nouveau-Brunswick (ANNEXE m ).

5. VARIABLE PRINCIPALE ET VARIABLES SECONDAIRES

Parmi les études où se trouve une analyse des diverses réalisations de la diphtongue orthographiée «oi», il en existe au moins trois qui relèvent spécifiquement de la sociolinguistique: Flikeid (1984), Thomas (1986), et Paradis (1988) (d'autres études inédites existent, par exemple celle de Deshaies- Lafontaine sur le parler de Trois-Rivières citée par Thomas (1986)). Flikeid distingue deux variables selon la structure syllabique, une variable principale WE en syllabe fermée, et une variable secondaire WE 2 en syllabe ouverte. Cette distinction repose d'une part sur la proportion plus élevée en syllabe fermée des réalisations en [we], d'autre part sur l'existence d'un sous-groupe lexical en syllabe ouverte où [wa] prédomine. Il s'agit d'un petit nombre de vocables, six formes seulement en français canadien selon Geoffrion (1934), Gendron (1966), Picard (1974), Dumas (1987): trois, bois (substantif), pois, poids, mois, noix. Dans le corpus de Flikeid, les mots les plus fréquents dans ce sous-groupe sont bois et mois. Dans ce même corpus, pour les autres mots fréquents en syllabe ouverte, à savoir moi, fois, quoi, c'est [we] qui prédomine, sans toutefois atteindre la proportion très élevée qu'on trouve en syllabe fermée (84% pour cette dernière contre 70% en syllabe ouverte - Flikeid 1984: 449).

Les résultats de Thomas pour le franco-ontarien de Sudbury sont semblables: double division de la variable (wa), premièrement en deux classes sociophonétiques selon l'aperture (Thomas 1986: 119): (wa-0) = [wa, wa], (wa-1) = [we, we, w d , wo]; deuxièmement en deux classes selon l'antériorité (Thomas 1986:130): /wa/ avec prédominance de réalisations en [we] et [we], et /wa/ avec prédominance de réalisations en [w d ] et [wo] (les mots dans cette catégorie cités par Thomas 1986: 122, sont trois, bois, mois). Les résultats de Thomas concordent généralement avec ceux de Holder (1972) sur le parler de Sudbury - North Bay. Le français saguenéen du Québec est assez semblable, à en croire les résultats obtenus par Claude Paradis (1988).

Nous avons trouvé dans le parler «brayon» un état de choses qui se rapproche quelque peu des autres parlers mentionnés. On y relève un certain nombre de réalisations en [we] affectant la plupart des mots où se trouvent la graphie «oi»; plus, un sous-groupe lexical où on trouve surtout des réalisations postérieures. Notre corpus comprend quatre des mots appartenant à ce sous-groupe: pois, bois, mois, trois. Étant donné que les deux groupes lexicaux s'opposent nettement selon le timbre antérieur ou postérieur du noyau de la diphtongue, il nous semble légitime de considérer qu'il s'agit de deux variables distinctes, (wa)

La diphthongue «oi» dans le parler «brayon» d'Edmunston, Nouveau-Brunsivick______ 23

antérieur et (wa) postérieur. D'autres variables encore entrent en jeu, affectant des sous-groupes lexicaux peu importants (Geoffrion 1934: 386, Thomas 1986: 117, Lucci 1972: 40, 45): un sous-groupe où [wa] alterne avec [e]/[e] (nettoyer ~ netteyer, droit ~ dret, froid ~ fret etc.); un sous-groupe où [wa] alterne avec[d]/[u] (soigner ~ sougner, poignet ~ pognet, moitié ~ motié, etc.).

Afin de distinguer entre les quatre variables en question, nous proposerions une notation où (WA) servirait de dénominateur commun. Selon ce système, (WA-1) désignerait la variable principale comprenant la plupart des mots en «oi», et (WA-2,-3,-4) désigneraient les variables secondaires comprenant les petits sous-groupes lexicaux. Les différentes qualités phonétiques et alternances caractérisant ces variables se classent comme suit: (WA-1) = prédominance de réalisations antérieures (moi, soir, etc.), (WA-2) = prédominance de réalisations postérieures (pois, trois, etc.), (WA-3) = réalisation diphtonguée alternant avec[e] (droit ~ dret, froid ~ fret, etc.), (WA-4) = réalisation diphtonguée alternant avec [d] (moitié ~ motié, poignet ~ pognet, etc.). Une notation simplifiée (wa) pour (WA-1) et (wa) pour (WA-2) sera utilisée à partir de la Section 7. Après cette section, on n'aura plus l'occasion de se référer ni à (WA-3) ni à (WA-4).

Les diverses valeurs phonétiques des quatre variables, groupées selon le trait phonétique majeur qui distingue chaque variable, sont présentées dans la liste suivante (ces catégories s'entrecoupent dans une certaine mesure comme il se verra plus loin).

Valeurs phonétiques des variables

variable (WA-1) variable (WA-2) variable (WA-3) variable (WA-4 )variantes qualité variantes qualité monoph- qualité monoph- qualité

antérieures phonétique postérieures phonétique tongues phonétique tongues phonétique

(wa-1) [we] (wa-8) [wo] (wa-12) [e] (wa-14) M(wa-2) [we] (wa-9) [wo] (wa-13) [e] (wa-15) t>](wa-3) [wæ] (wa-10) [wd](wa-4) [wae] (wa-11) [wa](wa-5) [wa]

inaccentuées(wa-6) [wa](wa -7) [we]

Dans notre corpus, les mots droit et froid ne présentent que des variantes diphtonguées antérieures, et les variantes dret, fret n'apparaissent pas; «oi» dans lesdits mots se rattache donc à la variable (WA-1). Par contre, moitié et poignet ont des variantes en [d] (4 ocurrences en [d] pour moitié, 8 en [d] pour poignet, [wa] pour la majorité des autres occurrences des deux mots); par conséquent, «oi» dans moitié et poignet se rattache à la variable (WA-4). De ce

24 Maurice Holder, Anne Macies, R olf Turner ¡¡MhojW

fait les mots moitié et poignet sont exclus de l'analyse présente qui ne porte que sur les variables (WA-1) et (WA-2).

La diversité des formes en question est le résultat d'un développement historique assez compliqué que nous passons brièvement en revue dans la section suivante.

6. BREF APERÇU HISTORIQUE DES VARIANTES

Une abondante bibliographie existe sur la question de l'origine et l'évolution du digraphe «oi» en français. Nous ne citerons que quelques-uns des ouvrages qui donnent un aperçu solide de la question. Il faut mentionner en premier les manuels de phonétique historique française les mieux connus, par exemple ceux de Beaulieux (1967), de Bourciez (1958), de Fouché (1958), de Pope (1952). Nous attirons aussi l'attention au livre de Schogt (1960) consacré tout entier au phénomène, ainsi qu'à une étude récente de l'un des auteurs du travail présent, Holder (1990).

En ce qui concerne le français québécois, un ouvrage de base est celui de Gendron (1966); pour une mise au point rapide, on pourrait consulter Geoffrion (1937), Picard (1974), Reighard (1980), et Dumas (1987). Quant au français acadien, on trouvera des commentaires sur la manière de prononcer «oi» dans Geddes (1908), Poirier (1928), Massignon (1947), et Gamer (1952).

Voici un bref résumé de l'histoire du phénomène. La plupart des mots en «oi» proviennent de mots latins avec les voyelles ï, ë en syllabe tonique libre (c'est-à- dire en syllabe accentuée ouverte). Ces deux voyelles se confondent très tôt en[e], et on reconstruit généralement les étapes entre le point de départ [e], et le point d'aboutissement [wa],comme suit: e > ei > oi > oe > w e> wa. Les voyelles postérieures ont aussi participé à cette évolution, sous l'influence du yod (par exemple gaudia > joie, vôce > voix). Les prononciations [we, we] sont encore vivantes aux niveaux populaire et dialectal en France et au Canada français.

Comme signalé en haut, il existe à côté de la variable principale (WA-1) des variables secondaires (WA-2,-3,-4) qui caractérisent certains sous-groupes lexicaux comportant un nombre limité de vocables. Dans le groupe (WA-2) - pois, poids, bois, mois, noix, trois - l'ancien s final aurait joué un rôle prépondérant dans l'évolution, selon les historiens: d'après la théorie des «sifflantes médiévales» avancée par Joos (1952), et appuyée par Martinet (1955), s provenant de s, ss latins, aurait eu une influence postériorisante au moyen âge, alors que ts provenant de c palatal latin, n'aurait pas exercé une telle influence: d'où mois (< më(n)sa) avec [wa] postérieur, et fois (< *flce, latin classique vïce) avec [wa] antérieur. Selon un autre point de vue (Fouché 1958: 243-4) le facteur

«durée» est à maxima entre de l'allongen préconsonant résultats en fr; mot qui écha] attendu [wa], une série de r (WA-1) (fois, t postérieures c des deux A da

Parmi les i froid, étroit, c existent pou première, [e] forme propri l'ouest, où 1( d 'au jourd 'h i- Maury 197 prestigieux c avait cours. I changem ent l'imparfait et (P- ex. fa ib le , le suffixe -aie suffixe -ois e En bref, la d <rencontre de;

^ en vie m it ié 'Poign<Pr°bablemei semble qU'iiCas de POlgtti attendu (Cf ,H « sw’ anante |„] telles formes

AVam an “ ■nplexe,,

PPelons

La diphthongue «o i» dans le parler «brayon» d'Edmunston, Nouveau-Brunswick______ 25

«durée» est à la base du changement: suivant la tendance à la différenciation maxima entre phonèmes, la postériorisation aurait été la conséquence naturelle de l'allongement «compensatoire» entraîné par la chute de s final et s préconsonantique: d'où bas [ba:], pâte [pa:t] (= ancien français paste). Les résultats en français canadien sont probants, étant donné qu'il n'y a qu'un seul mot qui échappe à la règle, à savoir noix de nüce ([wa] au lieu du résultat attendu [wa], cf. fois prononcé avec [wa]). Les résultats en bref sont d'une part une série de mots avec des réalisations antérieures correspondant à la variable (WA-1) (fois, toi, étoile ...), d'autre part une deuxième série avec des réalisations postérieures correspondant à la variable (WA-2) (mois, pois, bois ... cf. l'histoire des deux A dans Dumas 1987).

Parmi les mots les plus fréquents avec (WA-3) = [wa ~ e] se trouvent droit, froid, étroit, croire, noyer, nettoyer, orthographiés dret, fret, etc. Deux théories existent pour rendre compte du double aboutissement [e]/[we]. Selon la première, [e] serait une forme réduite de [w e ]; selon la deuxième, [e] serait une forme proprement parisienne, parce que Paris appartenait dialectalement à l'ouest, où le passage de ei à oi n'a jamais eu lieu (dans le parler normand d'aujourd'hui, me, te, trè, vé, sè s'entendent encore pour moi, toi, trois, voir, soir- Maury 1976). On propose qu'à Paris, «oi» était un emprunt aux dialectes prestigieux du Nord (picard) et de l'Est (champenois), où cette prononciation avait cours. Pourtant, une grande partie des mots qui auraient été affectés par ce changement ont des formes avec [e] au lieu de [wa]: les terminaisons de l'imparfait et du conditionnel (-ais, -rais), quelques adjectifs et substantifs isolés (p. ex. faible, épais, taie, craie, monnaie), les verbes connaître et paraître, enfin le suffixe -ais de beaucoup de noms de peuple (anglais, français, etc.), mais ici le suffixe -ois est un concurrent sérieux (japonais, chinois etc. - voir Holder 1990). En bref, la double issue de e fermé tonique libre serait en grande partie due à la rencontre des grands courants dialectaux à Paris au Moyen Âge (Schogt 1960).

On en vient au groupe lexical avec (WA-4) = [wa ~ d], qui comprend moyen, moitié, poignet, soigner, et quelques autres mots (Geoffrion 1934). Nous sommes probablement en présence de deux processus ici: dans le cas de moyen, moitié, il semble qu'il y ait une véritable réduction de [wa] à la voyelle simple [d]. Dans le cas de poignet et les formes populaires pogner, pognon, [d] est le résultat normal attendu (cf. cognée < cüneatu, oignon < ünione). La prononciation standard en [wa] est sans doute due à l'influence analogique de poing. Il existe aussi une variante [u] (poignet [pujtE]: (Lucci 1972: 45); soigne [sug]: (Flikeid 1988: 77). De telles formes n'apparaissent pas dans notre corpus.

Ayant aperçu la richesse exceptionnelle des formes en «oi» et son histoire complexe, nous passons maintenant à l'analyse formelle de nos données. Rappelons que dorénavant (wa) sera employé pour la variable (WA-1)

26 Maurice Holder, Anne Macies, Rolf Turner

(réalisations surtout antérieures), et (wa) pour la variable (WA-2) (réalisation ! surtout postérieures). Dans la présentation des résultats, nous suivrons dans se; | grandes lignes le modèle d'analyse établi pour l'acadien par Flikeid (1984: 283 299).

7. STRUCTURE DE LA VARIABLE (wa)

Nos résultats se présentent premièrement en termes d'occurrences totales et I de répartition des variantes, deuxièmement selon les facteurs suivants: 1) contraintes phonologiques 2) âge des sujets, 3) sexe des sujets, 4) style con­textuel. Il aurait été évidemment souhaitable de considérer d'autres facteurs tels que la classe sociale, le quartier habité, l'éducation scolaire, le revenu. Cependant, les dimensions du travail exigé dépasseraient de loin le but modeste que nous nous proposons ici.

Nous avons choisi d'examiner les quatre facteurs mentionnés en haut, pour les raisons suivantes:

1) C'est généralement le cas que le contexte phonologique a un effet plus ou moins marqué sur la réalisation d'une variable. Aussi est-il nécessaire d'étudier cette dimension linguistique.

2) La variation en fonction de l'âge peut fournir des renseignements sur les changements linguistiques potentiels en cours: on aurait ainsi un aperçu en «temps apparent» des tendances diachroniques en «temps réel».

3) Quant au sexe, nous l'avons sélectionné pour voir si les théories voulant que les femmes utilisent les formes dotées de prestige plus souvent que les hommes s'appliquent aux locuteurs d'Edmundston.

4) La prononciation des mots peut varier en fonction du style contextuel, c'est-à-dire le degré d'attention porté à la parole (= style), selon la situation dans laquelle le locuteur parle (= contexte) (Labov 1972/76: ch. 3).Il importe donc de vérifier si des variations de nature stylistique résultent de l'utilisation de différentes méthodes de présenter les mots aux sujets, afin d'obtenir des prononciations (images, dialogues, listes).

7.1 Nombre total d'occurrences, et répartition des variantes

Le nombre total des occurrences de la variable principale (wa) était de 841. Ce total a été obtenu après l'élimination de quelques mots pour lesquels on n'avait eu que quelques réponses (oie, tiroir, bouilloire, poire, poivre, balançoire, armoire, présentés comme images à identifier - ANNEXE I). Également éliminée était la deuxième occurrence de certains mots qui étaient évoqués deux fois dans le corpus {poêle, poisson, voiture - ANNEXE I). Le total ainsi obtenu se répartit sur neuf variantes, groupés en six classes: 1. variante traditionnelle [we], 2.

variante inte standard [wa [W3], et varia 1.

Ce qui fra] variante star connaissance! se peut que < faudrait un c résoudre la q (contexte pl maintenant a

[we]

7,2 Bnviro

Le T able ; Verte/ 2) en s en syUabe ir Partition.

On trouv différentes.da^s la répPriantes a

qu'efermante i0 élevé e n . ,c en syl^ tà l 'u ipatroninVe:

2) (réalisation wons dans se

eid (1984:283

nces totales el urs suivants: , 4 ) style con- :s facteurs tels i , le revenu, ? but modeste

m haut, p ou ri -

effet plus ou8* -il nécessaires

ments sur les un aperçu e n ,

ories voulant uvent que les

2 contextuel,le ), selon la 72/76: ch. 3).|ue résultent s aux sujets,

variante intermédiaire [wæ], 3. variante diphtonguée [wae], 4. variante standard [wa] (voir Léon 1966); 5. variantes postérieures [wa, w d , w d ]; 6. chva [wa], et variante ouverte du chva [wç]. Ces résultats sont présentés au Tableau 1.

Ce qui frappe ici, c'est la proportion relativement élevée de réalisations de la variante standard [wa]. Il est difficile de savoir dans l'état présent de nos connaissances si cette proportion est un reflet plus ou moins fidèle de la réalité. Il se peut que cette proportion soit en partie due à la formalité de la situation. Il faudrait un corpus plus grand avec un évantail de styles plus diversifiés pour résoudre la question. Sans perdre de vue ce fait, les facteurs mentionnés en haut (contexte phonologique, âge et sexe des sujets, style contextuel) seront maintenant analysées à partir des données présentées au Tableau 1.

Tableau 1.Nombre total d'occurrences de la variable (wa) réparties sur neuf variantes: 841 occurrences.

La diphthongue «oi» dans le parler «brayon» d'Edmunston, 'Nouveau-Brunswick______ 27

[w e ] [wæ] [wae] [wa] [wa] [w d ] [w d ] [we] [wa]

15 3 2 22 17

160 84 11 527 20 39

% 19 10 1 63 2 5

jt de 841- Ce ls on n'avait

liaUnç°ire’ént éliminée ;UX fois dans

4 se répartit

7.2 Environnement phonologique

Le Tableau 2 montre la répartition des variantes 1) en syllabe accentuée ou­verte, 2) en syllabe fermée par /r/, 3) en syllabe fermée par d'autres consonnes, 4) en syllabe inaccentuée. La Figure 1 est une représentation graphique de cette ré­partition.

On trouve [w e ] dans toutes les positions examinées, dans des proportions différentes. L'environnement phonologique joue évidemment un rôle important dans la répartition des variantes (analyse statistique à l 'ANNEXE I I I ) . Les variantes antérieures [we, wæ] sont beaucoup plus nombreuses en syllabe fermée qu'en syllabe ouverte, et à l'intérieur des syllabes fermées, la consonne fermante joue un rôle différenciateur: le taux d'utilisation de [we] est moins élevé en syllabe fermée par /r/ qu'en syllabe fermée par d'autres consonnes. Quant à l'utilisation de la variante standard [wa], on observe plus ou moins le patron inverse.

28 Maurice Holder, Anne Macies, Rolf Turnt

Enfin en syllabe inaccentuée, [w e ] est pratiquement inexistant et il y a ur proportion élevée de chva par rapport à la même variante en syllabe accentué (résultat peu surprenant étant donné que la position inaccentuée est celle qui fi vorise le plus l'emploi du chva).

Tableau 2. Réalisations de (wa): Environnement phonologique

variantes [we] [wæ] [wae] [wa] [wa] [ w d ] [ w d ] [ w b ] [wa]syllabe

ouverte accentuée25 32 0 181

6 l8

1 1 2 3

°//o 10 13 0 73 3 1syllabe accentuée fermée par Ir l

49 35 6 1388 2

100 2 0

2°//o 20 15 3 57 4 1

syllabe accentuée fermée par d'autres

con son n es 82 16 5 800 0

00 1 2

3

% 44 8.5 3 43 0 1.5

1 0 1 18 13syllabe inaccentuée 4 1 0 128 2 31

°//o 2 .5 0.5 0 77 1 19

Figure 1.Répartition des réalisations de la variable (wa) selon les contextes phonologiques des syllabes.

80

70

60

50

40

30

20

10

0syllabe ouverte syllabe accentuée syllabe accentuée syllabe inaccentuée

accentuée fermée par /r/ fermée par d’autresconsonnes

H[we] l[wæ] D[wa] l[w a] B[wa]l

I m diphthongue «oi» dans le parler «brayon» d'Edmunston, Nouveau-Brunsiuick______ 29

,e 7.3 Valeur de la forme intermédiaire [wæ]îi- Il y a lieu de se demander quelle serait la valeur dans le système de la forme intermédiaire [wæ]. La réalisation [wæ] est aussi fréquente que [we] en syllabe accentuée ouverte et en syllabe accentuée fermée par /r/, mais beaucoup moins fréquente que [we] en syllabe fermée par d'autres consonnes. La loi de la position

' est évidente (voir section suivante), mais on pourrait aussi penser à la théorie selon laquelle les valeurs intermédiaires représenteraient parfois une sorte de compromis créé dans un effort pour s'adapter à un dialecte contigu ou plus prestigieux (théorie de l'accommodation - voir Trudgill 1986: 60-2).

7.4 Syllabes ouvertes/fermées: comparaison avec d'autres variétés d'acadien

Il est rapporté que dans la région de Moncton (Lucci 1972), et en Nouvelle- Écosse (Ryan 1981, Flikeid 1988), «oi» se réalise [wa] en syllabe ouverte, et [we] en syllabe fermée. Selon certaines sources, cette distribution complémentaire est typique de l'acadien (Geddes 1894, Poirier 1928 - cités dans Flikeid 1984: 444). Cependant, [we] est observé dans les deux positions syllabiques dans une des régions acadiennes les plus importantes, le nord-est du Nouveau-Brunswick (Flikeid 1984: 444-7). C'est également le cas au nord-ouest selon nos données, bien que la proportion totale de [we] soit bien moins élevée par rapport à ce qu'a trouvé Flikeid au nord-est. Pour ce qui est des réalisations en syllabe ouverte accentuée (moi, fois, quoi), il faut noter qu'à Edmundston la proportion de réalisations en [wa] dépasse de loin celles en [wa], alors que [wa] est majoritaire ailleurs (Moncton, Nouvelle-Écosse, comme noté en haut; au nord-est du Nouveau-Brunswick la réalisation très ouverte, quand elle se produit, est postérieure comme à Moncton et en Nouvelle-Écosse).

En somme, on peut dire que le nord-ouest partage avec le nord-est l'occurrence de [we] en syllabe ouverte accentuée, tandis que dans d'autres régions acadiennes, une distribution complémentaire est observée, à savoir [wa] en syllabe ouverte accentuée, [we] en syllabe fermée accentuée (sauf devant /r/ - voir paragraphe suivant). En revanche, le nord-ouest s'oppose au nord-est et au sud-est en ce qui concerne la proportion de réalisations en [wa], beaucoup plus

! élevée que [wa] en «brayon».

7.5 Syllabes fermées par /r/: comparaison avec d'autres variétés d'acadien

| Comme il a été déjà signalé, [we] semble beaucoup plus fréquent au nord-est f qu'au nord-ouest de la province. Il y a pourtant un point où les deux parlers se | rejoignent. Flikeid (1984: 295) constate un écart entre le groupe de vocables se- terminant par /r/, et le groupe où la variable apparaît devant une autre

consonne: les vocables en /r/ sont dans l'ensemble réalisés moins souvent avec

[we]. Nous avons vu que c'est également le cas en «brayon». Flikeid propose un explication selon la fréquence des mots avec /r /, comme suit:

La raison pour cette différence pourrait être que /r/ comme consonne fermante est bie: plus fréquent que l'ensemble des autres consonnes. Il se pourrait que les locuteur soient plus conscients de la variable WE dans cet environnement phonétique, qui es celui qui revient le plus régulièrement, ou encore que la correction, si elle a lieu, port< plus particulièrement sur ce groupe de vocables. (Flikeid 1984: 295)

La fréquence des mots est souvent citée comme facteur explicatif dans le diffusion lexicale des changements phonétiques, aussi bien que du remplacemeni de phonème qui caractérise les emprunts dialectaux: les mots fréquents seraieni les premiers affectés, parce qu'ils sont davantage utilisés dans des situations de contact avec le dialecte donateur (ici le français standard). Cette théorie, qui se trouve en germe chez Bloomfield (1933), a été récemment reprise par Gerritsen et Jansen 1980 (voir Holder 1990 pour une discussion plus détaillée - nous reviendrons à la question dans la section suivante).

En outre, il faut envisager la possibilité d'une action ouvrante exercée par la consonne /r/, phénomène très répandu en français canadien. On constate une telle ouverture à Moncton (Haden 1948), et en Nouvelle-Écosse (Ryan 1981: 59, 85-7: savoir [saw ar], terre [tar], mer [mar]; Flikeid 1988: 76-8: affaire [afa:r ~afa:r], paire [pD.r], guerre [d3a:r ~ d3a:r], soir [swa:r]). Au contraire, il n'y a pas d'ouverture dans le parler de la vieille génération de la région de Moncton décrit par Lucci; celui-ci déclare formellement (Lucci 1972: 56): «Le phonème /e/ se réalise comme en français.» (exemples aux pages 44-5, 58 de l'ouvrage de Lucci). Tel est aussi le cas au nord-est selon Flikeid (1984:124,184), dans le sous- groupe lexical où l'acadien a traditionnellement /e/, réalisé invariablement [e] (mer, faire). Dans notre corpus, les mots qui tombent dans cette catégorie ont également [e] (Pierre, guerre, hiver).

7.6 Diffusion lexicale

Pour revenir à la question de la diffusion lexicale, il serait intéressant de tester l'hypothèse sur la fréquence qui veut que les mots fréquents soient les premiers à être influencés par une variante prestigieuse. Pour ce faire, on devra déterminer la fréquence dans la langue française des mots utilisés dans le corpus (à l'exception des noms de personne François et Antoine, pour lesquels il n'existe pas de données sur la fréquence). Les chiffres en question sont présentés aux Tableaux 3 et 4.

Sous la rubrique fréq se trouvent deux ensembles de chiffres séparés par un trait oblique pour chaque mot. À gauche du trait oblique figure le nombre d'occurrences de la variante standard [wa], et à droite le nombre total

1 propose une

’s, Rolf Turner

rm an te est bier e le s locuteur; lé tiq u e , qui es! le a lieu , porte

ca tif dans la ^ p la cem en ! ents seraient situations de léorie, qui se >ar Gerritsen lillée - nous

<ercée par la :onstate une ;an 1981:59, ¡ f a i r e [afa:r , il n 'y a pas j e Moncton honèm e /e/ 'ouvrage de lans le sous- jb lem ent [e] îtég orie ont

C essa n t de ; ts soient les 3 re on devra \ ns le corpus § ls il n'existe ¿sentes aux |

arés pw un Ile n on *re

m bre total

d'occurrences de la variable (wa). D'après ces chiffres, on calcule la fréquence relative de [wa] en termes de pourcentage. Afin de comparer cette fréquence avec la fréquence relative des mots dans la langue, nous avons consulté deux dictionnaires de fréquence, le Frequency Dictionary o f French Words (FDFW) de Juilland et coll. (1970), et Les indices d'utilité du vocabulaire fondamental français (IUVFF) de Savard et Richards (1970). La rubrique rang indique la posi­tion de chaque mot relativement aux autres, par ordre d'usage décroissant dans le FDFW, et par ordre d'utilité décroissante dans les IUVFF (Yusage d'un mot est calculé d'après sa fréquence et sa répartition, son utilité d'après sa répar­tition, sa fréquence, sa disponibilité, et sa valence - se rapporter aux diction­naires en question pour des détails). Le rang no. 1 indique la plus haute fréquence d'utilisation, et les numéros suivants des fréquences descendantes jusqu'aux nos. 3309 et 5083, représentant les rangs les plus bas dans les IUVFF et le FDFW re­spectivement.

L'emploi de ces dictionnaires ne va pas sans difficultés, dont il faut men­tionner trois en particulier. Premièrement, certains mots du corpus n'appa­raissent pas dans les dictionnaires. Il semble raisonnable de supposer que le rang de tels mots dépasserait les totaux établis dans les dictionnaires, c'est-à-dire 3309 et 5083 comme il vient d'être noté. Nous avons décidé, de façon plus ou moins arbitraire, d'assigner le chiffre de 5500 à ces mots.

Le deuxième problème est le suivant: tous les mots-outils sont exclus du dic­tionnaire de Savard et Richards, les auteurs prenant pour acquis qu'ils sont utiles (Savard & Richards 1970: 67). Pour résoudre ce problème, nous avons simplement assigné aux mots-outils en question, moi, pourquoi, toi, quoi, avoir, le même chiffre que celui donné par le FDFW. Ces mots ont effectivement une haute fréquence d'usage selon le FDFW, ce qui correspond aux prévisions de Savard et Richards: de cette façon, nous croyons que leur position est plus ou moins respectée.

Finalement, un problème se pose au sujet des formes verbales recevoir, boit, avoir, savoir. Dans le FDFW, on donne à chaque verbe un seul rang, basé sur l'ensemble des formes verbales selon le temps, le mode, l'aspect, la voix, le nom­bre et la personne. Heureusement, la fréquence de chaque forme individuelle est donnée, ce qui nous a permis de calculer leur rang. Par contre, il n'y a pas de chiffres pour les formes verbales individuelles dans les IUVFF. Il a été donc nécessaire d'extrapoler, encore une fois, à partir des données du FDFW. Le procédé consiste à trouver dans le FDFW la proportion entre la forme indi­viduelle et l'ensemble des formes verbales, et ensuite de calculer d'après cette proportion le chiffre qu'aurait la forme individuelle dans les IUVFF.

Pour distinguer entre les rangs pris directement dans les dictionnaires, et les chiffres qui ont été crées ad hoc, les premiers sont en caractères maigres, et les

La diphthongue «oi» dans le parler «brayon» d'Edmunston, Nouveau-Brunsivick______ 31

32 Maurice Holder, Anne Macies, Rolf Turner

Tableau 3. Variable (wa): fréquence de la variante [wa] dans chaque mot-ordre descendant absolu

C O R P U S IU V F F F D F Wmot freq % freq. rang rang rang rang

moyenne moyen moyen

moi 20/21 95 99 99pourquoi 18/21 86 282 282recevoir 18/21 86 594 800croyable 18/21 86 5500 5500oiseau 16/19 84 872 629poisson 17/21 81 377 2319voyageur 17/21 81 1078 1750toi 17/21 81 592 592fois 17/21 81 53 101incroyable 17/21 81 4029 4029arrosoir 13/16 81 84% 3041 1502 5500 1964

voiture 16/21 76 38 692croix 15/20 75 1887 1546roi 15/21 72 1185 454quoi 15/21 72 220 220froid 15/21 72 203 855noire 15/21 72 323 382histoire 14/21 67 69 464foi 13/21 62 71% 280 526 706 665

doigt 11/20 55 253 1216boit 10/19 53 944 3388gloire 11/21 52 2973 1140espoir 11/21 52 2145 821avoir 11/21 52 191 191paroisse 11/21 52 5500 5500soigne 11/21 52 308 1608mouchoir 7/14 50 2221 2306poêle 10/20 50 424 5500framboise 10/20 50 52% 5500 2046 5500 2717

toilette 10/21 48 1148 1860soir 10/21 48 156 192noir 10/21 48 323 382savoir 10/21 48 93 275poil 9/21 43 685 2408mâchoire 8/21 38 2993 5500avoine 7/21 33 4999 4999droite 7/21 33 257 876étoile 5/20 25 40% 2423 1453 779 1919

La diphthongue «oi» dans le parler «brayon» d'Edmunston, Nouveau-Brunswick______ 33

deuxièmes en caractères gras italiques. En plus, les mots sont divisés en quatre groupes dans le but de faciliter la comparaison entre les fréquences dans le corpus et les rangs assignés dans les dictionnaires. Le rang moyen dans chacun de ces groupes est noté en gros caractères.

D'après le Tableau 3, il n'y a aucune structure nette qui se dégage en com­parant les données du corpus avec celles des dictionnaires de fréquence. Il semble donc que l'hypothèse sur la fréquence n'est pas soutenue selon nos données.

Toutefois, on obtient un effet significatif en groupant les mots selon l'envi­ronnement phonologique (Tableau 4 - voir aussi Tableau 2). Dans cette per­spective, la fréquence d'utilisation de la variante standard [wa] diminue dans l'ordre suivant: syllabes inaccentuées, syllabes accentuées ouvertes, syllabes ac­centuées fermées par /r/, syllabes accentuées fermées par d'autres consonnes. Or, pour les syllabes accentuées, nous avons le même ordre de fréquence descen-

!! dante d'après les dictionnaires de fréquence. L'hypothèse sur la fréquence est donc confirmée dans le cas des syllabes accentuées. Cependant, pour ce qui est des syllabes inaccentuées, les résultats montrent l'inverse de ce qui est prédit par l'hypothèse. On doit bien sûr se tenir sur la réserve, étant donné qu'il s'agit d'une question extrêmement compliquée dont nous n'avons pu qu'effleurer la surface.

| Quoi qu'il en soit, l'effet produit par l'environnement phonologique montre un | fait incontestable: c'est que la distribution au nord-ouest est analogue à ce que | Flikeid a trouvé au nord-est; nous sommes en présence d'un des traits par lequel | l'acadien semble se distinguer des autres parlers franco-canadiens. En effet, la I répartition acadienne classique, où [we] est privilégié en syllabe fermée et [wa ~ ? wa] en syllabe ouverte, se reflète dans une certaine mesure en «brayon».

7.7 Facteurs sociolinguistiques

Dans ce qui suit, nous analyserons les variables en fonction de certains fac­teurs extralinguistiques, à savoir l'âge, le sexe, et le style contextuel. On obtient les résultats attendus en ce qui concerne l'âge; en ce qui concerne le sexe, les pré­supposés ne sont que partiellement confirmés; dans l'autre dimension sociolin- guistique examinée - le style contextuel - une corrélation est moins évidente entre style et choix de variante. Ces résultats seront maintenant examinés en détail.

7.8 Age

La répartition des variantes selon l'âge montre une différence significative d'une tranche d'âge à l'autre (Tableau 5). L'analyse statistique (ANNEXE III) con­firme cette interprétation des résultats. En gros, l'emploi de la variante tradi­tionnelle acadienne [we] va diminuant en proportion avec l'âge des sujets: cette

34 Maurice Holder, Anne Macies, R olf Turner

Tableau 4.Variable (wa): fréquence de la variante [wa] dans chaque mot: fréquence déscendante selon l'environnement phonologique.

C O R P U S IU V FF F D F Wmot freq. % fréq. rang rang rang rang

moyenne moyen moyen

croyable 18/21 86 5500 5500oiseau 16/19 84 872 629poisson 17/21 81 377 2319voyageur 17/21 81 1078 1750incroyable 17/21 81 4029 4029voiture 16/21 76 38 692toilette 10/21 48 77% 1148 1863 1860 2397

moi 20/21 95 99 99pourquoi 18/21 86 282 282toi 17/21 81 592 592fois 17/21 81 53 101croix 15/20 75 1887 1546roi 15/21 72 1185 454quoi 15/21 72 220 220froid 15/21 72 203 855foi 13/21 62 280 706doigt 11/20 55 253 1216boit 10/19 53 73% 944 545 3388 860

recevoir 18/21 86 594 800arrosoir 13/16 81 3041 5500noire 15/21 72 323 382histoire 14/21 67 69 464gloire 11/21 52 2973 1140espoir 11/21 52 2145 821avoir 11/21 52 191 191m ouchoir 7/14 50 2221 2306soir 10/21 48 156 192noir 10/21 48 323 382avoir 10/21 48 93 275mâchoire 8/21 38 56% 2993 1260 5500 1496

paroisse 11/21 52 5500 5500soigne 11/21 52 308 1608poêle 10/20 50 424 5500framboise 10/20 50 5500 5500poil 9/21 43 685 2408avoine 7/21 33 4999 4999droite 7/21 33 25 7 876étoile 5/20 25 42% 2423 2360 779 3396

Ci.

la diphthongue «oi» dans le parler «brayon» d'Edmunston, Nouveau-Brunszvick______ 35

variante est fréquente chez les âgés, relativement rare chez les jeunes, et entre les deux extrêmes chez ceux d'âge moyen. On aperçoit d'une façon assez claire

I qu'un changement linguistique de la variante [we] vers la variante [wa] est en I cours à Edmundston, puisque la variante antérieure [we] décroît avec chaqueI génération. La même tendance a été observée par Flikeid (1984, 1989) au nord-* est du Nouveau-Brunswick. Il semble raisonnable de supposer, avec Flikeid, que

l'évolution vers la variante standard parmi les jeunes soit liée au degré de sco­larisation, amenant un plus grand souci de correction.

Tableau 5. Réalisations de (wa): âge

variantes [we] [wæ] [wae] [wa] [w a] [ w d ] [ w d ] [we] [wa]

âgés 94 38 1 1252 1

30 6 12

18

% 34 14 0 45 1 611 2 1 9 4

âge moyen 44 39 10 164 14 13

°//o 15.5 14 3.5 58 4.5 4.52 0 1 7 1

j eu n es 82 16 5 80 3 8

°//o 8 2.5 0 86 1 2.5

7.9 Sexe.

D'après les données combinées présentées au Tableau 6, une comparaison globale des deux groupes de sexe ne semble pas indiquer de différences ma­jeures, sauf dans le cas de la variante [we]; pour celle-ci le taux d'utilisation est nettement plus élevé chez les hommes que chez les femmes, effet qui semble significatif du point de vue statistique (ANNEXE ni, données combinées (c)).

7.10 Interaction entre âge et sexe

La répartition au Tableau 6 et à la Figure 2, illustre que le facteur sexe est significatif si on tient compte en même temps de l'âge (analyse statistique: don­nées combinées (e), (f)). À Edmundston ce sont les hommes âgés qui emploient les variantes [we, wæ] plus souvent que n'importe quel autre groupe. C'est un résultat assez commun et bien admis par la plupart des sociolinguistes, que ce sont les hommes et surtout les hommes âgés qui utilisent, en général, les

36 Maurice Holder, Anne Macies, R olf Turner

variantes dotées de moins de prestige (Labov 1972/76: ch. 9, Chambers &Trudgill 1980: 172-4). Chez nos informateurs âgés à Edmundston, l'emploi de lavariante [we] dépasse la variante [wa] par 10%, mais c'est le contraire pour nosvieilles informatrices chez qui l'utilisation de la variante [wa] dépasse lavariante [wa] par 36%. D'ailleurs, parmi les âgés, l'emploi de la varianteintermédiaire [wæ] est de 12% plus élevé chez les hommes que chez les femmes.Donc dans la génération de l'âge d'or, à Edmundston, ce sont les femmes âgéesqui mènent le changement linguistique de [we] à [wa]. Ce résultat est conformeaux observations de Trudgill (1974, 1983: 85), selon lesquelles les informatricesutilisent plus souvent que les informateurs des formes associées aux normes deprestige. I

In more than one language, therefore, women's speech is more conservative than that g of men. .. .In ail the cases so far examined, it has been shown that, allowing for other fac- | tors such as social class, ethnie group and âge, women consistently use forms which | more closely approach those of the standard variety or the prestige accent than those I used by men. (Trudgill 1974: 90-1)

Quant au groupe d'âge moyen, c'est pratiquement l'inverse qui se produit, la f proportion des variantes [we] et [wæ] étant relativement plus élevée chez les informatrices que chez les hommes du même âge, alors que la proportion de [wa] est moins élevée. On dirait que ce sont les hommes qui mènent le changement. L'influence du facteur sexe diffère donc selon l'âge. Les femmes d'âge moyen semblent moins conscientes de la signification sociale des variables linguistiques. Labov dit à ce propos (1976: 405): «On se tromperait gravement si l'on posait en principe général que ce sont toujours les femmes qui mènent le changement lin­guistique.» Les recherches les plus récentes (Eckert 1989) démontrent effective­ment qu'il n'y a pas de rapport simple et constant entre le genre (construction sociale du sexe), et la variation; l'effet du sexe sur les variables est donc loin d'être uniforme.

Chez les jeunes, l'homogénéité est inopinément extrême. Les variantes sont réparties également entre les deux sexes dans ce groupe. Chez eux, la variante [wa] prédomine beaucoup plus régulièrement que chez les deux autres généra­tions (86%). Ici, on ne peut que répéter ce qui a été dit à la Section 7.8 au sujet de l'influence de la norme à l'école. Les jeunes préfèrent employer la forme mo­derne de prestige sans doute parce qu'ils sont plus sensibles à la valeur sociale et à la façon de parler «correctement». À ce sujet, William Labov déclare (1976: 294): «On peut déterminer la conscience qu'ont les locuteurs des variantes net­tement stigmatisées au moyen de 'tests de correction scolaire' où il s'agit de cor­riger des phrases qui s'écartent des modèles académiques.»

Cham bers l'em p lo i del; 'a ire pour nci *] dépasse li e la variante ‘z les femmes femm es âgée est conforme

informatrice; Lix norm es de

a tiv e th an tk :j ig fo r o ther fat] ie fo rm s w hid j :e n t th an th o sî

eJsJ^Ü Junp

se produit, l| 2vée chez le| »rtion de [wa|; changement| d 'âg e moyer| linguistiques | ' o n posait er n g em en t lir.-. en t effective- (construction js t donc lo ir

ariantes sont x, la variant« utres généra- .8 au su je t de a form e mo- eur sociale et éclare (1 976:ra ria n te s net-s 'a g it de cor-

Tableau 6.Réalisations de (wa): âge et sexe:

(hommes à gauche du trait oblique, femmes à droite)âgés (58-80) = 4 hommes/3 femmes ; âge moyen (36-55) = 2 hommes/5 femmes; jeunes (16-23) = 3 hommes/4 femmes

La diphthongue «oi» dans le parler «brayon» d'Edmunston, Nouveau-Brunsivick 37

[w e ] [wæ] [wae] [wa] [w q ,w d ,w d ] [we, wa] Total

âgés 65/29 29/9 0/1 51/74 0/3 8/10 153/126% 43/23 19/7 0/0 33/59 0/3 5/8 100/100

âge moyen 5/39 1/38 0/10 63/101 8/6 3/10 80/204% 6/19 1/19 0/5 79/49 10/3 4/5 100/100

jeunes 10/12 3/4 0/0 99/139 0/3 4/4 116/1620 //o 9/7 3/3 0/0 85/86 0/2 3/2 100/100

données 80/80 33/51 0/11 213/314 8/12 15/24 349/492com binées% 23/16 10/10 0/3 61/64 2/2 4/5 100/100

Figure 2:Répartition des réalisations de la variable (wa) selon l'âge et le sexe.

h[we] n [ w æ ] n [ w a ] n [ w a ] s [ w d ]

hom m esâgés

fem m esâgées

hom m esd 'âg e

m oyen

fem m esd 'âg e

m oyen

jeu n eshom m es

jeu n esfem m es

7.11 Méthode de présentation

Parmi les facteurs qui peuvent influer sur le choix d'une variante, il faut en­visager un effet possible dû au contexte stylistique; dans le cas présent, les con­textes différents correspondent aux différentes manières de présenter les mots aux informateurs afin d'obtenir des prononciations. Selon Labov (1976: 143), la lecture de mots isolés constitue un contexte plus formel que la lecture de dia­logues. Cette observation est confirmée par Flikeid (1984: 296), qui signale dans

38 Maurice Holder, Anne Macies, R olf Turner

son corpus un écart entre les deux contextes de lecture: «l'utilisation de la vari ante [we] s'avère plus forte dans le contexte lecture de dialogues que lors de le lecture de listes de mots, qui constitue le contexte le plus formel». Pour la consti tution de notre corpus, nous avons eu recours aux deux contextes de lecturt ainsi qu'à une troisième technique, à savoir images à identifier (voir Section 4) Comme procédure, cette dernière se rapproche de la lecture de mots isolés, étani | donné que l'informateur cherche à produire un mot individuel qui corresponde à l'image présentée.

Tableau 7.Réalisations de (wa): Méthode de présentation.

[we] [wæ] [wae] [wa] [w q , w d ,w d ] [w e, wa]1 0 2

im ages 38 22 3 109 3 15o //o 20 12 1.5 57 1.5 8

3 1 0 7 10dialogu es 48 25 0 242 4 17o //o 14 8 0 72 1 5

11 2 0 4 3listes 74 37 8 176 13 70 //o 23.5 12 2.5 56 4 2

Figure 3.Répartition des réalisations de la variable (wa)

selon la méthode de présentation.

80 r

70

60

50

40

30

20

10

0

h [w e] n[w æ ] a[w a] i [ w a ] afw a]

images dialogues listes

la diphthotigue «oi» dans le parler «brayon» d'Edmunston, Nouveau-Brunsivick______ 39

On présente au Tableau 7 la répartition des données selon la méthode de présentation. L'analyse statistique indique ici un effet significatif. Néanmoins, il y a un certain élément de doute qui s'introduit quand on tient compte des inégal­ités de répartition au niveau de l'environnement phonologique. On aborde ce problème au paragraphe suivant.

Dans le contexte moins formel dialogues, on s'attendrait que le taux d'utili­sation de la variante standard [wa] soit moins élevé par rapport à son utilisation

| dans les deux autres contextes. L'inverse est en fait le cas, résultat qui peut s'ex- I pliquer par le fait que le nombre de mots avec «oi» en syllabe ouverte, est | nettement plus élevé dans le contexte dialogues que dans les deux autres con- | textes. Or, nous avons vu qu'en position accentuée l'environnement pho- î nologique syllabe ouverte favorise l'emploi de la variante standard beaucoup I plus que les deux autres environnements (voir Section 7.2). Étant donné cette 1 répartition inégale selon l'environnement phonologique, on se garde de tirer | quelque conclusion que ce soit à partir des données sur la méthode de5 présentation.i Nous passons maintenant à l'étude de la variable (wa), selon les mêmes i critères sociolinguistiques d'âge, de sexe, et de niveau de langue, qui ont servi à = l'analyse de (wa).

8. STRUCTURE DE LA VARIABLE (wa)

8.1 Occurrences totales et répartition des variantes

Le nombre total des occurrences de la variable secondaire (wa) est de 84. Ce total se répartit sur quatre variantes, groupées en deux classes: variante antérieure [wa]; variantes postérieures [wa, w d , w d ] (Tableau 8) . La proportion élevée de variantes postérieures (82%), et l'absence totale des variantes antérieures mi-ouvertes [we, wæ], montrent incontestablement le statut phonologique indépendant de cette variable par rapport à la variable (wa). Ici, la structure syllabique n'est pas pertinente, puisqu'une seule position, celle de syllabe ouverte, est représentée dans le sous-groupe lexical en question (pois, bois, mois, trois). Parmi les variantes postérieures, il est intéressant de constater que la variante mi-ouverte [w d ] , de caractère populaire, est presque aussi fréquente que la variante postérieure ouverte standard [wa] (alternant avec [wa], accepté aussi comme normatif). La variante [wa] est la seule qui est attestée dans des études antérieures sur l'acadien. Comme dans le cas de la variante intermédiaire [wæ] (Section 7.5. en haut), on pourrait invoquer une influence possible du parler québécois avoisinant. Une influence possible a été déjà identifiée en «brayon», à propos d'un autre trait typiquement québécois,

40___________________________________________ Maurice Holder, Anne Macies, R olf Turm

Tableau 8.Réalisations de (wa):

Occurrences totales et répartition des variantes: 84 occurrences.

[wa] [wa] [wd] [wd]15 30 14 25

% 18 36 16 30

l'assibilation de /t, d/, devant les voyelles antérieures très fermées. L'étude de c I trait constitue le deuxième grand volet de la thèse sur laquelle le présent trava I est basé (McKillop 1987 = Macies 1987, McKillop et Cichocki 1988,1989 = Macie I et Cichocki 1988,1989; voir aussi Section 3 en haut). I

8.2 Effet de l'âge, du sexe, et de la méthode de présentation

En ce qui concerne l'âge et le sexe (Tableaux 9 et 10), le fait le plus saillant es évidemment la proportion beaucoup plus élevée d'utilisation de la variante [wa chez les jeunes femmes que dans les autres groupes (jeunes femmes - 56°/c jeunes hommes - 8 %, femmes d'âge moyen - 5%, hommes d'âge moyen - 0°/c femmes âgées - 17%, hommes âgés - 12.5 %). Ce résultat va de pair avec ce qui i j été trouvé pour la variable (wa). Par contre, pour ce qui est de la variable (wa) | l'emploi de la variante extrême [w d ] est courant chez les jeunes des deux sexes alors que pour (wa) la variante traditionnelle [we] est plus rare dans cett( tranche d'âge. Il va sans dire que la variante [we] est perçue commit «stigmatisée», tandis qu'il n'en est rien de son homologue postérieur. ^

Pour ce qui est de la méthode de présentation (Tableau 11), aucun système ne \ ne semble se dégager selon lequel les différentes méthodes auraient un effet su: la réalisation de la variable (wa).

Tableau 9. Réalisations de (wa): âge

âgés 4 7 5 12% 14 25 18 43

âge moyen 1 17 8 2% 3.5 61 28.5 7

jeu n es 10 6 1 110//o 36 21.5 3.5 39

La diphthongue «oi» dans le parler «brayon» d'Edmunston, Nouveau-Brunswick______ 41

Tableau 10.Réalisations de (wa): âge et sexe

(hommes à gauche du trait oblique, femmes à droite)

[wa] [wa] [w d ] [wo] Totalâgés 2 / 2 2/5 4/1 8/4 16/12o/ /0 12.5/17 12.5/42 25/8 50/33 100/100

âge moyen 0 / 1 5/12 1/7 2/0 8/200//o 0/5 62.5/60 12.5/35 25/0 100/100

jeunes 1/9 4/2 1 / 0 6/5 12/160//o 8/56 34/13 8/0 50/31 100/100

Tableau 11.Réalisations de (wa): méthode de présentation.

images 2 8 4 7o//o 9.5 38 19 33.5

dialogues 5 8 4 40//O 24 38 19 19

listes 8 14 6 14% 19 33.5 14 33.5

I 9. RÉSUMÉ ET CONCLUSIONk

1 Dans le but de décrire la variation phonétique du parler «brayon» d'Ed- k mundston et de déterminer si un changement linguistique est en cours dans cette

communauté, nous avons effectué une enquête auprès de 21 locuteurs, d'âge et de sexe différents.

Nos hypothèses de base ont été confirmées car, premièrement, certaines an-& ciennes variantes existent encore dans la communauté d'Edmundston, surtout i parmi les personnes âgées. Les anciennes variantes acadiennes [w e ] (étude

présente), [tjl et [dj] (McKillop et Cichocki 1988,1989 = Macies et Cichocki 1988, 1989), se retrouvent dans des vocables tels que étoile [etwEll, histoire [istwEr],

1 tiens [tjë], culotte [tjybt], Dieu [àyd], guêpe [d3Ep].Deuxièmement, certaines variantes typiquement québécoises apparaissent

dans le parler «brayon» d'Edmundston. Les variantes en question sont [wo] {pois [pwD], mois [mwD 1 - voir paragraphe suivant) et les variantes assibilées [ts] et [dz], qui se trouvent dans des lexèmes tels que tire [tsir], voiture [vwatsyr],

crocodile [kr:>kDdzil], dur [dzyr] (McKillop et Cichocki 1988, 1989 = Macies e l Cichocki 1988,1989). Pour l'instant on ne peut pas dire avec certitude si ces va I riantes sont attribuables à une influence québécoise, ou s'il s'agit d 'u r l développement indépendant (cf. assibilation dans l'île du Prince Édouard) I D'autre part, l'influence de la langue anglaise se fait sentir surtout dans le do I maine du lexique (McKillop 1987 = Macies 1987, ch. 5); au niveau phonologique, I par contre, il n'y a aucune évidence jusqu'ici qui suggérerait une influence I provenant de la langue anglaise. I

Troisièmement, en voulant découvrir s'il existe une évolution linguistique! dans cette communauté, nous avons trouvé que pour la variable principale (wa), I un changement systématique a lieu: une décroissance globale de l'ancienne va-1 riante antérieure [ w e ] en marche vers la variante centrale [wa] s'observe à I Edmundston. Par contre, en ce qui concerne la variable secondaire (wa), la va-1 riante fermée [w d ] semble relativement stable, étant presque aussi fréquente I dans l'ensemble que toute autre variante. Ici, comme nous l'avons déjà fait re -1 marquer, il y a lieu de se demander si le «brayon» serait influencé par le parler I québécois voisin. On constate aussi une certaine prédominance chez les jeunes I femmes de la variante standard [wa] comme réalisation de la variable (wa): ce I résultat va de pair avec la même tendance observée pour la variable (wa). I

L'environnement phonologique n'est pas pertinent dans le cas de la variable I postérieure (wa), puisque cette dernière se trouve uniquement en syllabe ou- 1 verte. Par contre, cette dimension phonologique est très importante dans la réalisation de la variable antérieure (wa). L'environnement «syllabe fermée par des consonnes autres que /r/» favorise plus que tout autre environnement l'apparition de la variante fermée [w e ] , tandis que la variante [wa] est très élevée en syllabe ouverte accentuée et en syllabe inaccentuée. On peut ainsi parler de diffusion lexicale au niveau des grands ensembles de mots groupés î selon la structure syllabique: en position accentuée ce sont les mots avec «oi» en f syllabe ouverte qui mènent le mouvement vers la variante standard. I

On s'est donc intéressé à tester l'hypothèse selon laquelle les mots les plus I fréquents seraient les premiers à être influencés par un emprunt prestigieux. | Pour ce faire, nous avons comparé nos données avec celles des dictionnaires de I fréquence. L'hypothèse semble être confirmée quand les mots sont groupés selon i la structure syllabique, mais non quand ils sont arrangés par ordre de fréquence l descendant absolu. Dès lors, il nous paraît inutile d'essayer d'expliquer, à cette I étape préliminaire de l'investigation, ces résultats apparemment contradic- \ toires. Il reste vrai qu'il s'agit d'une question fascinante qui mérite d'être sérieusement examinée à l'avenir.

Un autre thème qui se dégage de notre recherche, c'est que les facteurs so­ciaux affectent le choix des variantes. Le mouvement [we] > [wæ] > [wa] semble

42___________________________________________ Maurice Holder, Anne Macies, R olf TurneiM

,a diphthongue «oi» dans le parler «brayon» d'Edmunston, Nouveau-Brunswick______ 43

ié à l'âge des sujets: l'emploi de [w e ] et de [wæ] décroît avec l'âge, et l'utilisation le [wa] augmente systématiquement chez les sujets plus jeunes. Ce mouvement :aractérise les deux sexes. Généralement les femmes emploient plus souvent la /ariante standard [wa], sauf dans le cas des informatrices d'âge moyen; chez :elles-ci le taux d'occurrences de la variante [we] est plus élevé que chez les lommes du même âge, et le taux d'occurrences de [wa] moins élevé chez les 'emmes que chez les hommes du même âge. En fonction du style contextuel, la /ariation est faible et très marginale, et la variante [wa] reste toujours très forte dans chaque contexte.

On peut conclure que la variable (wa) à Edmundston subit un changement lin­guistique de la façon attendue qui va de la variante ancienne antérieure [we] y ers la variante centrale [wa], et que ce changement suit la voie qu'on trouve partout ailleurs pour maintes autres variables sociolinguistiques.

Une comparaison avec les autres variétés d'acadien au Nouveau-Brunswick Bt en Nouvelle-Écosse montre des similitudes et des différences qui peuvent se résumer ainsi:

1) Variante [we] en syllabe ouverte aussi bien que fermée au nord-est et au nord-ouest du Nouveau-Brunswick; ailleurs, distribution complémentaire comme suit: [wa] en syllabe ouverte, [we] en syllabe fermée.

2) En syllabe ouverte accentuée (moi, fois, quoi), parmi les réalisations ouvertes, [wa] plus fréquent que [wa] en «brayon»; proportion inverse ailleurs.

3) Devant /r/, tendance à l'ouverture au nord-est et au nord-ouest du Nouveau-Brunswick; [we:r] intact dans le parler de la vieille génération du

✓sud-est du Nouveau-Brunswick; [wa:r] en Nouvelle-Ecosse.

10. SUGGESTIONS POUR DES RECHERCHES ULTÉRIEURES

Ayant effectué la première étude linguistique du français «brayon», nous voudrions bien qu'on continue à étudier le parler d'Edmundston ainsi que les autres parlers de la région au nord-ouest du Nouveau-Brunswick (St. Léonard, Grand-Sault... ). Les variables que nous avons dû exclure de notre recherche seront sûrement révélatrices. Quelques-unes des plus intéressantes sont exposées plus bas.

Dans la perspective de la variation phonétique, nous suggérerions une recherche sociolinguistique plus approfondie que la nTtotre, car outre l'âge et le sexe, la dimension sociale n'est pas représentée dans notre recherche. En effet, des facteurs sociaux tels que la classe sociale, le niveau d'instruction, le type de travail, la solidarité et l'identification à un milieu pourraient être des facteurs importants dans le comportement linguistique de nos locuteurs.

44 Maurice Holder, Anne Macies, R olf Tun r

Variable Variantes Vocable

(AR) [er] ~ [ar] verte [vert] ~ [vart](EU) [0] ~ [œ] tracteur [trak ter] ~ traktœ r](O) [0] ~ [D] ~ [dw] fo r t [for] ~ [for] ~ [fDwr](E) [e] ~ [e] ~ [ej] m ère [m er] ~ [mer] ~ [mejC.F. consonne finale ~ 0 sable [sabl] ~ [sab]

arbre [arbr] ~ [arb]

Du point de vue du contexte stylistique, il serait fort intéressant d'analyser ; vrai vernaculaire, les réponses libres, pour découvrir s'il y aurait plus de co * traste entre la langue parlée spontanée et la lecture à haute voix.

La variation phonétique est loin d'être le seul champ à explorer dans cet! communauté. Les recherches pourraient s'étendre à la variation syntaxique : surtout lexicale. Puisque Edmundston est une communauté bilingue, il serait tr< ; intéressant d'étudier en profondeur le lexique de cette communauté avec une e: quête appropriée qui ferait appel à ce domaine.

Notamment, puisqu'il ne semble pas y avoir en ce moment un grand intér pour étudier ce parler «brayon», nous espérons que notre étude suscitera, e définitive, de l'intérêt chez d'autres chercheurs, et qu'elle pourra servir de poil de départ aux études futures.2

2 Une deuxième enquête sur le «brayon» est actuellement en cours sous la directior de Maurice Holder. Ce travail bénificie d'une subvention accordée par le Conseil di recherches en sciences humaines du Canada, aussi bien que de fonds provenant du programme fédéral 'Défi Canada'.

î ,a diphthongue «oi» dans le parler «brayon» d'Edmunston, Nouveau-Brunswick______ 45

A N N E X E I■

Mots en «oi» évoqués par des images; mots en «oi» lus (dialogues, listes): 21 sujets

/ariable (wa)p

images no. d'occurrences dialogues no. d'occurrences listes no. d'occurrences

oiseau 19 11. voiture 21 27. soir 21I 11 " poêle 20 20. droite 21 28. noir 211 >. toilette 21 13. quoi 21 29. avoine 21n. framboise 20 14. pourquoi 21 30. roi 21

i. poisson 21 15. noire 21 31. voyageur 21».1 étoile 20 16. voisins 21 32. paroisse 21

t r arrosoir 16 17. histoire 21 33. croix 21et. boit (vb) 19 18. toi 21 34. gloire 21

doigt 20 19. fois 21 35. croyable 21' 0. i mouchoir 14 20. froid 21 36. espoir 211 21. François 21 37. foi 21

22. moi 21 38. avoir 21i 23. Antoine 21 39. savoir 21r 24. recevoir 21 40. mâchoire 21i

25. soigne 21 41. poil 211 26. incroyable 21

190 336 315fotal: 841

Le total de 841 occurrences comprend une occurrence par mot par sujet, moins -0 valeurs manquantes au niveau des images.

Variable (wa)

im ages no. d'occurrences dialogues no. d'occurrences listes no. d'occurrences

pois 21 2. mois 21 4. trois 213. bois 21

foUl: 84

Variable (wa): mots éliminés

Les mots ci-dessous ont été éliminés pour les raisons suivantes: 1) nombre in- , suffisant de réponses (moins de 14 réponses), 2) deux occurrences d'un seul et ; nême mot: deuxième occurrence éliminée, 3) variable (WA-4): seulement deux1 nots dans le corpus - moitié, poignet.

46 Maurice Holder, Anne Macies, Rolf Turne;

Selon l'étude statistique (ANNEXE III), les deux ensembles de données données complètes et données réduites - ont à peu près la même valeur en ce qi concerne l'effet qu'ont les facteurs sociolinguistiques sur la prononciation.

im ages no. d'occurrences im ages deuxièmes variable no. d'occurrencesoccurrences (WA-4)

[wa ~ d] [wa] M

1. oie 9 1. poêle 16 1. m oitié 17 42. tiroir 7 2. poisson 12 2. poignet 13 83. bouilloire 114. poire 13 dialogues5. poivre 06. balançoire 3 3. voiture 217. boîte 6 4. droite 18. armoire 39. voile 6 listes10. voilier 2 5. soir 111. voiture 2 6. gloire 112. croix 1 7. mâchoire 1

63 53 30 12Total: 158

ANNEXE II:Age et sexe des informateurs

âgés âge moyen jeu n es(58 - 80 ans) (36-- 55 ans) (16 - 23 ans)

no. âge sexe no. âge sexe no. âge sexe

1. 80 M 8. 55 F 15. 23 F2. 72 F 9. 47 F 16. 17 F3. 68 M 10. 44 F 17. 17 F4. 65 M 11. 43 F 18. 16 M5. 62 F 12. 42 M 19. 16 F6. 60 M 13. 37 F 20. 16 M7. 58 F 14. 36 M 21. 16 M

A N N E X E I I I Analyse statistique

INTRODUCTION

L'analyse statistique suivante porte uniquement sur les données obtenues pour la variable principale (wa). Dans le cas de la variable secondaire (wa), une étude statistique n'était pas justifiée en raison de l'insuffisance de l'échantillon (84 occurrences au total). Le but de l'étude était de déterminer si divers facteurs

la diphthongue «oi» dans le parler «brayon» d'Eâmunston, Nouveau-Brunswick______ 47

(contexte phonologique, âge, sexe, méthode de présenter des mots aux sujets afin d'obtenir des prononciations), avaient un effet sur la prononciation de la diphtongue «oi» de six façons différentes par les 21 sujets. Un problème parti­culier dans cette analyse était le fait qu'il y avait un certain nombre de valeurs manquantes au niveau des images, quand un sujet ne produisait pas de mot, ou fournissait un mot différent de celui qu'on cherchait (Section 7.1, et ANNEXE I pour des détails).

D a été décidé que la meilleure manière d'analyser ces données, étant donné toutes les complexités et les rapports possibles entre diverses classifications et leurs interdépendances, était d'adapter un modèle log-linéaire aux données (voir Kotz et Johnson 1982-89, vol.2: 161-171). Un tel modèle permet d'estimer les probabilités que la diphthongue «oi» soit prononcée d'une de six manières possibles étant donné diverses circonstances ou combinaisons de facteurs. Les paramètres propres à ce modèle ont été estimés en utilisant la fonction «glim» {generaliied linear interactive modelling) du logiciel statistique Splus (1990).

En premier lieu, les données ont été analysées selon chaque méthode de présentation - images, dialogues, listes. Ensuite, les trois ensembles de données ont été combinés et une autre analyse effectuée en utilisant présentation comme un autre facteur dans le modèle. Un résumé des résultats pour les données ré­duites se présente en face (à peu près le même effet pour les données complètes - voir ANNEXE I, mots éliminés).

Résu m é e t Co n c l u sio n s

La méthode de présentation semble déterminer s'il y a un effet dû au sexe. Cet effet est significatif dans le cas des images et des dialogues, mais non-significatif dans le cas des listes. Toutefois, étant donné l'âge, l'effet dû au sexe est non- significatif pour les trois méthodes de présentation. On doit conclure que l'effet dû au sexe est significatif seulement parce qu'il se confond avec celui de l'âge.

En revanche, l'effet dû à l'âge est positif, quelle que soit la méthode de présentation. Cet effet persiste même quand il est introduit dans le modèle après le sexe. Il y a une forte indication d'un effet d'interaction entre l'âge et le sexe pour deux des trois méthodes de présentation. Par conséquent, les probabilités pour les diverses combinaisons de ces effets ne peuvent pas être obtenues en multipliant les probabilités marginales.

Lorsque les trois ensembles de données sont combinés, il y a une forte indica­tion de la présence de tous les effets en question. Ceux-ci comprennent le sexe et la présentation, en plus de l'effet d'interaction entre l'âge et le sexe. Ces effets semblent persister quel que soit l'ordre dans lequel ils sont introduits dans le modèle.

48 Maurice Holder, Anne Macies, Rolf Turner

Il est à remarquer que l'effet dû au sexe est significatif lorsqu'il est ajout* après l'âge, ce qui n'était le cas pour aucune des trois méthodes de présentatior traitées séparément. Ce résultat surprenant s'expliquerait peut-être de la façor suivante: chaque ensemble de données renferme une tendance non-significative avec un patron similaire dans chaque cas; lorsque les ensembles furent combinés, les patrons similaires se sont renforcés, et un effet significatif en résulta.

Quant à l'environnement phonologique, il y a un effet marqué dans la plupart des cas, dont les plus importants figurent ci-dessous. Il n'y a pas d'évidence d'interaction avec d'autres effets, sauf une interaction avec l'âge dans le cas de la méthode de présentation listes.

R é s u l t a t s

Images(a) Il semble y avoir un effet dû à l'environnement phonologique (valeur-p = 0.0034)(b) L'effet dû à l'âge est marginalement non-significatif (valeur-p = 0.068)(c) Aucun effet dû au sexe (valeur-p = 0.57)(d) Étant donné le sexe, aucun effet dû à l'âge (valeur-p = 0.110)(e) Étant donné l'âge, aucun effet dû au sexe (valeur-p = 0.82)(f) Aucun effet d'interaction entre l'âge et le sexe (valeur-p = 0.18)

Dialogues(a)

(b)(c)(d)(e)(f)

Listes

Il semble y avoir un effet dû à l'environnement phonologique (valeur-p = 0 à 4 décimales)Il semble y avoir un effet dû à l'âge (valeur-p = 0 à 4 décimales)Aucun effet dû au sexe (valeur-p = 0.76)Étant donné le sexe, il y a encore un effet dû à l'âge (valeur-p = 0 à 4 décimales) Étant donné l'âge, aucun effet dû au sexe (valeur-p = 0.995)Il semble y avoir une interaction entre l'âge et le sexe (valeur-p = 0 à 4 décimales)

(a)

(b)(c)(d) (e)

(f)

Il semble y avoir un effet dû à l'environnement phonologique (valeur-p = 0 à 4 décimales)Il semble y avoir un effet dû à l'âge (valeur-p = 0 à 4 décimales)Il semble y avoir un effet dû au sexe (valeur-p = 0.0097)Étant donné le sexe, il y a encore un effet dû à l'âge (valeur-p = 0 à 4 décimales) Étant donné l'âge, aucun effet dû au sexe (valeur-p = 0.40) (On doit en déduire que l'effet apparent dû au sexe noté dans (c) est présent parce que l'âge et le sexe sont partiellement confondus, et ceci dû au manque d'équilibre dans le plan).Il semble y avoir une interaction entre l'âge et le sexe (valeur-p = 0.0001)

Données combinées(a) Il semble y avoir un effet dû à l'environnement phonologique (valeur-p = 0 à 4

décimales)(b) Il semble y avoir un effet dû à l'âge (valeur-p = 0 à 4 décimales)(c) Il semble y avoir un effet dû au sexe (valeur-p = 0.0047)(d) Il semble y avoir un effet dû à la méthode de présentation (valeur-p = 0 à 4

décimales)

La diphthongue «oi» dans le parler «brayon» d'Edmunston, Nouveau-Brunswick 49

(e) Étant donné le sexe, il y a encore un effet dû à l'âge (valeur-p = 0 à 4 décimales)(f) Étant donné l'âge, l'effet dû au sexe est marginalement non-significatif (valeur-p =

0.059)(g) Étant donné la méthode de présentation, il y a encore un effet dû à l'âge (valeur-p =

0 à 4 décimales)(h) Étant donné la méthode de présentation, il y a encore un effet dû au sexe (valeur-p =

0.0047); (i) Étant donné la méthode de présentation et ensuite le sexe, il y a encore un effet dû à

l'âge (valeur-p = 0 à 4 décimales)I (j) Étant donné la méthode de présentation et ensuite l'âge, l'effet dû au sexe est i marginalement non-significatif (valeur-p = 0.059)| (k) Il semble y avoir une interaction entre l'âge et le sexe (valeur-p = 0 à 4 décimales)I (1) Étant donné la méthode de présentation, il semble encore y avoir une interaction Ü entre l'âge et le sexe (valeur-p = 0 à 4 décimales)

R E M E R C I E M E N T S

i Cette étude est une version remaniée et augmentée des chapitres 1, 2, 3 et 5 de Î la thèse de maîtrise soutenue par Anne W. Macies (antérieurement McKillop), et i acceptée par l'Université du Nouveau-Brunswick, Fredericton, en octobre 1987. i Anne Macies a effectué l'enquête sur le terrain, et tient à remercier les personnesS suivantes de leur conseil, encouragement, et appui moral: Laurier Melanson,

Anne-Marie Grignon (décédée), Doris Leblanc, Anthony House (co-directeur), Wladyslaw Cichocki (directeur). Un grand merci pour Karin Flikeid qui a aidé énormément avec la construction du questionnaire et qui a montré beaucoup

È d'intérêt pour cette recherche.i Les données recueillies par Anne Macies ont été réexaminées par Maurice

Holder, et de nombreux changements et ajouts y ont été également apportés. Néanmoins, les conclusions générales restent dans l'ensemble les mêmes par rapport à celles qui avaient été proposées à l'origine par Anne Macies.

| L'analyse statistique a été effectuée par le Docteur T. Rolf Tumer du Centre ' de statistique appliquée de l'Université du Nouveau-Brunswick. Nous remer- \ cions Debbie Dupuis, Assistante au Centre de statistique, pour son aide en

E matière de terminologie statistique française.Il nous fait également plaisir de remercier le lecteur anonyme qui a bien voulu

I évaluer notre étude. Nous avons trouvé ses suggestions fort utiles, et avonsI révisé l'article en conséquence.II Enfin un grand merci du fond du coeur est offert à tous les informateurs |l «brayons» d'Edmundston de leur accueil chaleureux et de leur coopération dés- P intéressée. Ils ont bien voulu offrir toutes leurs connaissances historiques et géo- I graphiques de la région, toutes les ressources bibliographiques dont ils dispo- ! saient, et toute leur propre expérience linguistique. Cette étude n'aurait pu se | réaliser sans leur participation et nous les en remercions chaleureusement.

50 Maurice Holder, Anne Macies, R olf Turne

R É F É R E N C E S

A L B E R T , ABBÉ T h o m a s . 1920. Histoire du Madawaska. Québec: Imprimeri franciscaine missionnaire.

BEAULIEUX, CHARLES. 1967. Histoire de l'orthographe française. N ouveau tirage. Paris: Champion.

B l o o m f ie l d , Le o n a r d . 1933. Language. New York: Holt.

BOURCIEZ, ÉDOUARD. 1958. Précis historique de phonétique française. Neuvième Édition. Paris: Klincksieck.

C h a m b e r s , J. K. & Pe t e r Tr u d g il l . 1980. Dialectology. Cam bridge: Cam bridge University Press.

C h a r p e n t i e r , J e a n - M i c h e l . 1989. Le(s) parler(s) acadien(s) et le substrat du Haut-Poitou. Actes du XVe colloque international de linguistique fonctionnelle. (Moncton/Pointe-de-l'église, Canada, 18-24 août 1988). Université de Moncton: Centre de Recherche en Linguistique Appliquée, 169-186.

DUMAS, D e n i s . 1987. Nos façons de parler. Les prononciations en franco- québécois. Sillery, Québec: Presses de l'Université du Québec.

ECKERT, PENELOPE. 1989. The whole woman: sex and gender differences in variation. Language Variation and Change, I: 245-267.

FLIKEID, KARIN. 1984. La variation phonétique dans le parler acadien du nord- est du Nouveau-Brunswick. Étude sociolinguistique. New York: Peter Lang.

FLIKEID, KARIN. 1988. Unity and diversity in Acadian phonology: an overview based on comparisons among the Nova Scotia varieties. Journal o f the Atlantic Provinces Linguistic Association / Revue de l'Association de Linguistique des Provinces Atlantiques 10: 64-110.

FLIKEID, KARIN. 1989. Recherches sociolinguistiques sur les parlers acadiens du Nouveau-Brunswick et de la Nouvelle-Écosse. Dans Mougeon, Raymond& Édouard Beniak (rédacteurs), Le français canadien parlé hors Québec: aperçu sociolinguistique. Québec: Les Presses de l'Université Laval, 183- 199.

FOUCHÉ, PIERRE. 1958. Phonétique historique du français. Volume II. Les voyelles. Paris: Klincksieck.

GARNER, JOHN El b e r t . 1952. A Descriptive Study o f the Phonology o f Acadian French. Ph. D. thesis, University of Texas.

GEDDES, JAMES JR . 1908. Study o f an Acadian-French dialect spoken on the North Shore o f the Baie-des-Chaleurs. H alle: Niem eyer.

GENDRON, JEAN-DENIS. 1966. Tendances phonétiques du français parlé au Canada. Paris: Klincksieck / Québec: Presses de l'Université Laval.

GEOFFRION, LOUIS. 1934. La diphtongue OI dans le franco-canadien. Canada français 22: 384-90.

GERRITSEN, MARINEL & FRANK JANSEN. 1980. Word frequency and lexical diffusion in dialect borrowing and phonological change. Dutch Studies 4: The Hague: Martinus Nijhoff, 31-54.

[ HADEN, ERNEST F. 1948. La petite Cendrillouse: version acadienne deCendrillon. Publications de l'Université Laval: Les Archives de folklore 3: 21-34.

HOLDER, MAURICE. 1972. Le parler populaire franco-canadien. Laprononciation de quelques Canadiens français de la région de Sudbury- North Bay. Phonetica 26: 33-49.

HOLDER, MAURICE. 1990. Lexical diffusion and word frequency in phonological borrowing: o/ou and ais/ois in the history of French. PAMAPLA/ACALPA 14: 69-82. (Papers from the Annual Meeting o f the Atlantic Provinces Linguistic Association/Actes du colloque annuel de l'Association de linguistique des provinces atlantiques).

JOOS, MARTIN. 1952. The medieval sibilants. Language 28: 222-31. (Reprinted in Readings in Linguistics I. Chicago: University of Chicago Press, 1957).

J u i l l a n d , A lp h o n s e , D o r o t h y B r o d in , & C a t h e r i n e D a v i d o v i t c h . 1970. Frequency Dictionary o f French Words. Collection The Romance Languages and their Structures 3. The Hague: Mouton.

KING, R uth. 1989 Le français terre-neuvien: aperçu général. Dans Mougeon, Raymond & Édouard Beniak (rédacteurs), Le français canadien parlé hors Québec: aperçu sociolinguistique. Québec: Les Presses de l'Université Laval, 227-244.

KING, R u th & ROBERT Ryan. 1986. La phonologie des parlers acadiens de l'île- du Prince-Édouard. PAMAPLA/ACALPA 10: 95-108. (Papers from the Annual Meeting o f the Atlantic Provinces Linguistic Association/Actes du colloque annuel de l'Association de linguistique des provinces atlantiques).

KING, R uth & R obert Ryan. 1987. The construction of a rural sociolinguistic corpus: The Prince Edward Island study. In Thomas, Alan (ed.), Methods in Dialectology: Proceedings o f the Sixth International Conference held at the University College o f North Wales, 3rd -7 th August 1987. Clevedon / Philadelphia: Multilingual Matters Ltd., 95-108.

K in g , R u t h & ROBERT RYAN. 1988. The P.E.I. Acadian Project: A progressreport. PAMAPLA/ACALPA 12: 44-52. (Papers from the Annual Meeting of the Atlantic Provinces Linguistic Association/Actes du colloque annuel de l'Association de linguistique des provinces atlantiques).

! la diphthongue «oi» dans le parler «brayon» d'Edmunston, 'Nouveau-Brunswick 51

52 Maurice Holder, Anne Macies, Rolf Turnei

KING, RUTH & ROBERT RYAN. 1989. La phonologie des parlers acadiens de l'île du Prince-Édouard. Dans Mougeon, Raymond & Edouard Beniak (rédacteurs), Le français canadien parlé hors Québec: aperçu sociolinguis- tique. Québec: Les Presses de l'Université Laval, 245-259.

KOTZ, SAM UEL, & NORMAN L. JOHNSON (rédacteurs). 1982-88. Encyclopedia q Statistical Sciences. New York: John Wiley & Sons.

LABOV, W ILLIAM . 1972. Sociolinguistic Patterns. Collection Conduct and Communications 4, Philadelphia: University of Pennsylvania.

LABOV, WILLIAM. 1976. Sociolinguistique. Paris: Éditions de Minuit. (Traductior de l'anglais de Labov 1972).

LÉON, PIERRE R. 1966. Prononciation du français standard. Paris: Didier.

LUCCI, VINCENT. 1972. Phonologie de l'acadien. (Parler de la région de Moncton). Collection Studia Phonetica 7. Paris: Didier.

MACIES*, An n e W. 1987. Une étude sociolinguistique du parler «brayon» d'Edmundston au Nouveau-Brunswick. Thèse de m aîtrise inédite. The University of New Brunswick, octobre, 1987.

MACIES*, A n n e & WLADYSLAW C ic h o c k i. 1988. Variation dans la réalisation des occlusives dans le parler «brayon» du Nouveau-Brunswick. PAMAPLA/ACALPA 12: 53-58. (Papers from the Annual Meeting o f the Atlantic Provinces Linguistic Association/Actes du colloque annuel de l'Association de linguistique des provinces atlantiques).

MACIES*, ANNE & WLADYSLAW CICHOCKI. 1989. Affrication dans le parler «brayon» du Nouveau-Brunswick (Canada). Actes du XVe colloque international de linguistique fonctionnelle. (Moncton/Pointe-de-l'église, Canada, 18-24 août 1988). Université de Moncton: Centre de Recherche en Linguistique Appliquée, 229-235.

MARTINET, ANDRÉ. 1955. Économie des changements phonétiques. Berne: A. Franck.

MASSIGNON, GENEVIÈVE. 1962. Les parlers français d'Acadie: Enquête linguistique. Paris: Klincksieck, 2 vol.

MAURY, NICOLE. 1976. Le système vocalique d'un parler normand. Collection Studia phonetica 11. Paris: Didier.

MCKILLOP*, A n n e W. 1987. = MACIES 1987,MCKILLOP*, ANNE & WLADYSLAW CICHOCKI. 1988 & 1989. = MACIES &

C ic h o c k i 1988 & 1989.

GU N.-B. 1881 marquis.

MOUGEON, RA canadien ] de l'Univ

PARADIS, CLAl linguisticj

PÉRONNET, L( Éléments

PICARD, MARC

et phonolPOIRIER, PASC

ImprimePOPE, MlLDRI

Press.Reighard, j o

rules. In 4th ïn terBenjam i

RYAN, R o b e i région à Centre i

Savard, î e a :vocabulaLaval.

Savoie, a i e Presses

ScH0GT,HE] Amsteri

U So c ié t é iHistoriqMadaw

Spk kferen Tl»M A S , A i

Collecti

* ANNE MACIES, antérieurement MCKILLOP

la diphthongue «oi» dans le parler «brayon» d'Edmunslon, Nouveau-Brunswick______ 53

. MICHAUD, GUY. 1980. La Paroisse de l'Immaculée-Conception, Edmundston. N.-B. 1880-1980. Bibliothèque nationale d'Ottawa: Les Presses des ateliers marquis.

à MOUGEON, RAYMOND & ÉDOUARD BENIAK (rédacteurs). 1989. Le français I canadien parlé hors Québec: aperçu sociolinguistique. Québec: Les Presses I de l'Université Laval.

PARADIS, Cla u d e . 1988. La diphtongue /wa/ en français saguenéen. Langues et linguistique 14: 253-275.

PÉRONNET, LOUISE. 1989. Le parler acadien du Sud-Est du Nouveau-Brunswick. Éléments grammaticaux et lexicaux. New York: Peter Lang.

PICARD, M arc. 1974. La diphtongue /wa/ et ses équivalents. Cahier delinguistique N° 4: le français de la région de Montréal, aspects phonétique et phonologique. Montréal: Les Presses de l'Université du Québec.

POIRIER, PASCAL. 1928. Le parler franco-acadien et ses origines. Québec: Imprimerie franciscaine missionnaire.

POPE, MILDRED. 1952. From Latin to Modem French. Manchester University Press.

REIGHARD, JOHN. 1980. The transition problem: lexical diffusion vs. variable rules. In E. Traugott, R. Labrum & S. Shepherd (editors), Papers from the 4th International Conference on Historical Linguistics. Amsterdam: John Benjamins.

RYAN, R o b e rt. 1981. Une analyse phonologique d'un parler acadien de la région de la Baie Sainte-Marie (Nouvelle-Écosse), Canada. Québec: Centre international de recherche sur le bilinguisme, Université Laval.

SAVARD, Jean-GUY & JACK RICHARDS. 1970. Les indices d'utilité duvocabulaire fondamental français. Québec: Les Presses de l'Université Laval.

SAVOIE, ALEXANDRE-J. 1976. Un demi-siècle d’histoire acadienne. Montréal: Presses de l'imprimerie Gagné Ltée.

SCHOGT, HENRY GILIUS. 1960. La double issue de E fermé tonique libre. Amsterdam: van Oorschot.

La So cié té H isto riq u e d u M ad aw aska . 1982. Revue de la SociétéHistorique du Madawaska, Le Brayon. Vol. 10, nos. 1-2. Edmundston: Le Madawaska Ltée.

Splus Reference Manual. 1990. Seattle: Statistical Sciences, Inc.T h o m a s , A l a i n . 1986. La variation phonétique: cas du franco-ontarien.

Collection Studia Phonetica 21. Ottawa: Didier.

54 Maurice Holder, Anne Macies, R olf Turne

THOMAS, ALAN (ed.) 1987. Methods in Dialectology: Proceedings o f the Sixth International Conference held at the University College o f North Wales, 3rd _yth August 1987. Clevedon/Philadelphia: M ultilingual M atters Ltd.

TRUDGILL, PETER. 1974. Sociolinguistics: An introduction. London: Penguin Books.

TRUDGILL, PETER. 1983. On Dialect. Social and Geographical Perspectives. Washington/New York: New York University Press.

TRUDGILL, PETER. 1986. Dialects in Contact. Oxford/New York: Blackwell.

While Ket been con si parent c o i validity o obey the t AH. This from over linear on quasi-verl measure i with, and

1. PRELIMIM

In their c r proposed a Ni

(1)

SI

The AH is i 'H to w hich relativization rect object (D % t of com 0n the AH, an

The °FContinuity C

(2)

AS(

Thus lang0rtosUBandÌBSSSnjj

ON THE UNCLEFTABILITY OF DIRECT OBJECT IN CHINESE

Cheng Luo Brock University

ABSTRACT

While Keenan and Comrie's (1977) A ccessibility H ierarchy ( A H ) theory has been considered applicable to syntactic processes other than relativization, ap­parent counterevidence is not lacking which poses a potential threat to the validity o f the theory. For example, while in general Chinese cleft sentences obey the A H , the uncleftability o f the direct object (DO) poses a problem fo r the AH. This study, however, shows that DO uncleftability in Chinese follozvs not from overgeneralization o f the A H , but from a language-specific constraint on linear order between the contrastive focu s marker, or more generally the quasi-verb, and the main verb in Chinese. Thus while the A H attains some measure o f plausibility as a universal tendency, it nevertheless can interact with, and be negatively affected by, language-specific constraints.

1. PRELIMINARIES

In their crosslinguistic study on relativization, Keenan and Comrie (1977) proposed a Noun Phrase Accessibility Hierarchy (AH), reproduced here as (1):

(1) The Accessibility Hierarchy (AH):

SUB > DO > IO > OBL > GEN > OCOMP

The AH is essentially a statement of a universal tendency in language, accord­ing to which the positions higher on the AH are universally more accessible for relativization than those lower. Thus subject (SUB) is easier to relativize than di­rect object (DO), indirect object (IO), oblique object (OBL), genitive NP (GEN), and object of comparison (OComp); DO is easier to relativize than any lower position on the AH, and so on.

The operation of the AH is governed by a principle known as the Continuity Constraint, given here as (2):

(2) The Continuity Constraint:

Any relative clause-forming strategy must apply to a continuous segment of the AH, and strategies that apply at any one point of the AH may in principle cease to apply at any lower point. (Keenan & Comrie 1977: 67)

Thus languages may have relativization strategies which apply only to SUB, or to SUB and DO, or to the top three positions on the AH; but it is highly unlikely

I-INC.UISI K A al l ii i i t ica 1 4 ( 1 9 9 2 ) 5 5 - 7 1

T

56 Cheng Lut

that a language, for instance, can relativize SUB and OBL without also being abli to relativize DO and IO. Generalizations like this determine constraints on the form and substance of possible human languages.

The AH has been claimed to be universally applicable to various other syntac­tic processes such as WH-question formation, topicalization, and clefting. Sometimes, however, linguistic data from some languages seem to present counterevidence to the proposed AH. The uncleftability of DO in Chinese offers such an example which merits scrutiny. The present study, therefore, addresses the issue of DO uncleftability in Chinese, to see whether such apparent counter­evidence really invalidates the AH, and, if not, what causes DO uncleftability in Chinese.

2 . THE PROBLEM OF D O UNCLEFTABILITY

A cleft sentence is defined as a construction in which a particular constituent is marked by means of a syntactic and/or morphological device 'for the purpose of focus, contrast, or emphasis' (Teng 1979: 101). In Chinese, the cleft focus1 is marked by an immediately preceding contrastive focus marker (CFM) shi, which is identical in form to the copula verb. In addition, a modifier marker(MM)2, de (Ross 1983), occurs toward the end of the cleft sentence, though sometimes op­tionally. This pattern is formally represented as (3), where F stands for the cleft focus, and X and Y, for any variable, respectively.

(3) X shi F Y (de)

The cleft focus usually receives some extra stress, and can be almost any constituent on the AH except DO, as shown in (4).

(4) a. Yuehan wei wo John for Is

cong guowai ding le ji-piao. from abroad reserve Perf air=ticket

'John reserved an air ticket for me from abroad.'

b. Shi Yuehan weiCFM John for

WO cong Is from

guowaiabroad

dingreserve

ji-p iao de. air=ticket MM

'It is John who reserved an air ticket for me from abroad.'

!While the working definition of clefts is largely functional here, Huang (1982a) argues, with particular reference to clefting, that languages like Chinese may involve movement in its LF (Logical Form) component even if no movement is apparent in SS (Surface Structure). Therefore, the term 'cleft' is considered justifiable here.

2This modifier marker (MM) is usually in complementary distribution with the perfec­tive morpheme -le, as shown in later examples.

Unckft^

c. Yue Joh r

Iti!

d. Yu<Johi

'I t i

e. *Y i

'It!

While (4a) i the contrastivi ject (4d). Sente

DO unclefta the fact that it objects contra results in a di; egy, as in (5):

(5) SU

which is an ui DOuncleftabi] 1®', and was turn in Sectioi

There are ; jects in Qiine: °f the ah wit factors other to® the AH i kss advisabl< Chinese cleft] fo n d ly , i f ; ^at direct ot relativization term from ph tobjects

Mention

On the Uncleftability o f Direct Object in Chinese 57

c. Yuehan shi wei wo cong guowai ding ji-piao de.John CFM for Is from abroad reserve air=ticket MM

'It is for me that John reserved an air ticket from abroad.'

d. Yuehan shi cong guowai wei wo ding ji-piao de.John CFM from abroad for Is reserve air=ticket MM

'It is from abroad that John reserved an air ticket for me.'

e. *Yuehan wei wro cong guowai ding shi ji-piao de.John for Is from abroad reserve CFM air=ticket MM

'It is an air ticket that John reserved for me from abroad.'

While (4a) is a non-cleft with the basic proposition, (4b-d) respectively have as the contrastive focus the subject (4b), the indirect object (4c), and an oblique ob­ject (4d). Sentence (4e), where the direct object is in cleft focus, is ungrammatical.

DO uncleftability in Chinese as described above poses a problem for the AH: the fact that it is possible to cleft indirect objects and oblique objects but not direct objects contradicts the AH, which predicts just the opposite. This situation also results in a discontinuous segment on the AH in terms of the same clefting strat­egy, as in (5):

(5) SUB *DO IO OBL (GEN)3

which is an undesirable violation of (2) in the context of clefting. This problem of DO uncleftability was first recognized by Teng (1979:104) as 'a perplexing prob­lem', and was later on tackled briefly in Huang (1982: 291), to which we will re­turn in Section 4.

There are at least two approaches to this problem. One is to treat direct ob­jects in Chinese as inherently inaccessible to clefting and claim the inapplicability of the AH with Chinese cleft sentences as exceptional. The other is to examine factors other than the direct object and the A H , and claim that the deviation from the AH is due to the intervention of some other factor. The first approach is less advisable for two reasons. First, to claim that the AH is not applicable to Chinese clefting is too simplistic and adversely affects universality of the AH. Secondly, if we look at DO accessibility for other syntactic operations, we find that direct objects are easily accessible to processes such as topicalization (6a), relativization (6b), question formation (6c), and pseudo-clefting (6d). To use a term from phonology, by pattern congruity, it is not very plausible to claim that direct objects in Chinese are inherently inaccessible to clefting.

3Clefting of GEN NP frequently involves long distance dependency relations and pro­noun retention strategy in Chinese. Therefore, it is less relevant here.

58 Cheng Luo

(6) a. Ji-piao ta yijing ding le.air=ticket 3s already reserve Perf

T h e air ticket, he already reserved.'

b. Zhe shi ta ding de ji-piao.this is 3s reserve RM air=ticket

T h is is the air ticket he reserved.'

c. Ni ding le shenme?2s reserve Perf what

'What did you reserve?'

d. Ta ding de shi ji-piao.3s reserve N om is air=ticket

'What he reserved was an air ticket.'

Given such facts, the first approach would require an ad hoc statement on the inaccessiblity of DO to clefting and therefore is less desirable if we can find non- ad hoc alternatives. On the other hand, the second approach does not necces- sarily require any ad hoc solution or affect universality of the AH.

Therefore, instead of simply abandoning the AH, I will adopt the second approach and contend that DO uncleftability in Chinese stems not from over­generalization of the AH, but from a language-specific linear constraint in Chinese.

3. THE SOLUTION

Synchronically, Chinese has an unmarked order of SVO4 (Sun & Givon 1985, Li 1990: 23, Travis 1984), with direct objects unmarkedly occurring postverbally. Linearly, the contrastive focus marker (CFM ) shi, must invariably occur before the main verb of the sentence. In other words, the CFM shi as a rule can not occur after the main verb. This constraint, implicitly stated in Huang (1982: 291), can be formally stated as (7):

(7) *X MAIN VERB shi Y

Constraint (7) stipulates that no postverbal CFM is permitted in Chinese. This potentially conflicts with (3), which says that the CFM must immediately precede the cleft focus. Now if we consider (3) and (7) in terms of word order in Chinese, we have an account of IX) uncleftability. For a direct object, which occurs

4On the other hand, Li & Thompson (1975) and Tai (1973) hold that Chinese is changing from an SVO to an SOV language. Despite the controversial views, it is generally accepted that the direct object unmarkedly occurs after the verb.

On the Uncleftability o f Direct Object in Chinese 59

postverbally, to be clefted, the CFM must immediately precede it, resulting in a postverbal CFM shi, which violates (7).

The implications of the formulation of (7) go beyond the scope of direct ob­jects, since Y can stand for any postverbal constituent. More specifically, (7) predicts that not only direct objects, but all postverbal constituents are subject to the same linear constraint, viz., no postverbal constituents are cleftable. This is

| bome out by examples like (8), which would all be grammatical as non-clefts if | shi were omitted:

(8) a. *Yuehan zhao le ni shi liang d de.John look=for Perf 2s CFM two time MM

Tt is twice that John has looked for you.'

b. *T a huai de shi hen.3s bad Resultative CFM very

'He's indeed very bad.'

c. *T a re de shi han dou liu le chulai.3s hot Resultative CFM sweat all flow Perf come=out

'He was so hot that he was wet with sweat.'

Sentence (8a) has a postverbal quantifier phrase, (8b), a postverbal degree adverb, and (8c), a postverbal S. Like direct objects, all such postverbal con­stituents are uncleftable. Also like direct objects, they can undergo other syntactic processes; for example, the quantifier phrase in (8a), liang ci, can be fronted (9a), relativized (9b), questioned (9c), and pseudo-clef ted (9d).

(9) a. Ta liang d zhao ni ni dou bu zai.3s two tim e look=for 2s 2s all not in

'Twice when he looked for you, you were not in.'

b. Wo shengbing de na liang dIs be=sick MM that two tim e

'the two times when I was sick'

c. Ta zhao le ni ji ci?3s look=for Perf 2s how=many time

'How many times did he look for you?'

d. T a zhao ni de cishu shi liang ci ( bu shi yi ci). 3s look=for 2s MM tim e=num ber be two tim e n o tb eo n e tim e

'The number of times he looked for you is twice, not (just) once.'

Constraint (7) therefore not only accounts for uncleftability of direct objects, but also for that of all postverbal constituents.

60 Cheng Luo

There are several pieces of evidence in support of this analysis, the first of which involves preposed direct objects. Recall (6a), which has a preposed topi- calized object. Chinese has another means of preposing postverbal direct objects with definite reference: the ba construction, whereby a direct object occurs pre- verbally and is immediately preceded by a preposition-like object marker (OM) ba. In both preverbal positions, the direct object becomes more cleftable:

(10) a. Yuehan John

shiCFM

baOM

ji-piao cong air=ticket from

guowai ding abroad reserve

haoCompì.

le.Perf

Tt is the air ticket that John reserved from abroad.'

b. t Shi ji-piao Yuehan cong CFM air=ticket John from

ding hao le. reserve Compl. Perf

Tt is the air ticket that John reserved from abroad.'

guowaiabroad

In (10a), a preposed object in the ba-construction is fully cleftable. For the topicalized (10b), given appropriate stress on ji-piao 'air ticket' and a short pause after it, the sentence is much more acceptable than (4e), which has a postverbal focus. Therefore, both support (7).

The second kind of evidence has to do with indirect objects headed by gei 'to, for' and oblique objects headed by zai 'at' or dao 'to'. Such phrases can occur ei­ther preverbally or postverbally, with a slight semantic difference sometimes (Chao 1968). When such a phrase occurs preverbally, it is cleftable, as in (11a); however, when it occurs postverbally, it is not cleftable, as in (lib ):

(11) a. Yuehan shi gei ta ji xin de.John CFM to 3s m ail letter MM

Tt is to him that John mailed a letter.'

b. *Yuehan ji shi gei ta John m ail CFM to 3s

yi feng xin de.on e M(classifier) letter MM

Tt is to him that John mailed a letter.'

Cleftability in such cases depends entirely on whether the focused constituent is preverbal or postverbal, as predicted by (7).

The third piece of evidence comes from an infrequent variation of cleft sen­tences, wherein the direct object is cleftable but must receive extra stress, as in (12).

(12) Yuehan shi ji le yi feng xin.John CFM m ail Perf a M letter

'It's a letter that John mailed.'

On the Uncleftability o f Direct Object in Chinese 61

Theoretically, DO clefting brings (3) and (7) into conflict with each other, in that while (7) prohibits a postverbal CFM, the immediate precedence requirement of (3) necessitates a postverbal CFM for DO clefting. This conflict is resolved by letting (7) override (3): while (12) respects the word order constraint (7) in that the CFM occurs before the main verb, it violates (3) insofar as the cleft focus is not

I adjacent to the CFM. This violation, however, is prosodically compensated for by i the heavy stress on the focused DO. What (12) shows, then, is that DO uncleftabil­

ity is due not to the object itself, but to the word order constraint specified in (7), in the sense that as long as the CFM shi occurs before the main verb and con­straint (7) is respected, the sentence is grammatical even if the focus is discontin­uous from the CFM5.

Finally, as mentioned earlier, all postverbal constituents are uncleftable, a fact that is most elegantly accounted for by (7), but is not accounted for with an ad hoc statement about the direct object.

One may think of sentences like the folowing as counterexamples to (7):

(13) Yuehan cong guowai ding de shi ji-piao.John from abroad reserve N om Copula air=ticket

'What John reserved from abroad is an air ticket.'

In (13), it seems that shi occurs after the main verb yet the sentence is still grammatical. However, as indicated by the translation, this construction is in fact not a cleft sentence, but a pseudo-cleft sentence, where shi is used as a cop­ula equative verb, i.e., a main verb, rather than a CFM. The difference is that, with shi as a main verb, the sentence would be ungrammatical without it, whereas with shi as a CFM , the sentence would still be grammatical as a non­cleft even without it. This is seen in (14) (compared with (13)), where absence of shi results in a (verbless) non-sentence, possibly rendered as 'an air ticket that John reserved from abroad' on a different reading:

(14) *Yuehan cong guowai ding de ji-piao.

Additional evidence is seen in the occurrence of the nominalizer de (Ross 1983), which nominalizes the clause before it as a subject NP, the whole sequence before de being a headless relative clause. This being the case, the only main verb candidate in the remaining part of the sentence is shi.

5I thank one of the anonymous reviewers for pointing out an analogous structure in English where the focus is discontinuous from its marker. A sentence like 'Brian doesn't only want to satisfy Quebec', with emphatic stress on Quebec, will have the NP as the ex­clusive focus of only.

62 Cheng Lu

Another kind of apparent counterexample is one like (15), where it seems tha shi occurs after the main verb kanjian 'see', a violation of constraint (7), yet thi sentence is still grammatical:

(15) Wo kanjian ta shi zai ji tian yiqian.Is see 3s CFM at a=few day ago

'?It is a few days ago that I saw him.'

A more plausible analysis of (15), however, would be in terms of topic comment structure, treating shi, again, as the main verb of the sentence. I wiL argue that the constituent before shi in (15), wo kanjian ta, is a topical clause which represents known information, and the part that follows shi, the comment which represents new information, with shi as the equative main verb. The whole sentence should have been translated as T h at I saw him was several days ago.' The alternative analysis as suggested by the dubious translation in (15) would have wo as the subject-topic and kanjian as the main verb. Therefore, it is essential to find out whether the subject-topic is wo or wo kanjian ta. I will pro­vide arguments for the latter against the former.

As suggested by Li & Thompson (1981), two formal devices can be used to dis­tinguish a topic in Chinese: sentence-initial position, and the optional occurrence of pause or pause particles. In Chinese, a topic almost invariably occurs in sen­tence-initial position, and can be optionally separated from the comment 'by a pause or by one of the pause particles (Ptc) [a, ya, me, ne, or ba\ (p.86). With re­spect to (15), the first criterion does not apply since both wo and wo kanjian ta are sentence-initial. Applying the second criterion, we get:

(16) a. Wo kanjian ta me, shi zai ji tian yiqian.Is see 3s Ptc Copula at a=few day ago

'As for my seeing him, it was several days ago.'

b. ??Wo me, kanjian ta shi zai ji tian yiqian.

'As for me, (I) saw him several days ago.'

While the topical status of wo kanjian ta in (16a) is clear by virtue of its sounding perfectly natural, (16b) suggests the dubiousness of wo as a subject- topic. The implausibility of a pause particle between wo and kanjian ta suggests their integrity as a constituent, viz., the topic.

Morphologically, one of the topical features is that the verb in a clausal topic tends to show reduced 'verbness' by virtue of being rendered aspectless or tenseless (Givon 1984). This is exactly what we find in regard to (15). Consider(17):

Ott the Uncleftability o f Direct Object in Chinese 63

(17) *Wo kanjian le ta shi zai ji tian yiqian.Is see Perf 3s Cpl at a=few day ago

'(As for) my seeing him, it was several days ago.'

If kanjian were the main verb, it would not only allow for, but require, a per­fective marker in this context. The impossibility of the perfective le in (17) further suggests that what precedes shi in (15) is topical.

Finally, as we observed earlier, the absence of shi as a CFM will not affect grammaticality of the remaining part of the sentence as a non-cleft, whereas the absence of shi as a main verb will. Given this difference, if shi were analyzed as the CFM in (15), the sentence, without shi, would still be grammatical as a non­cleft without shi. But this is not the case, as (18) shows:

(18) *W o kanjian ta zai ji tian yiqian.Is see 3s at a=few day ago

'I saw him several days ago.'

This is because the temporal phrase as a rule should occur not after, but be­fore, a non-copula verb. In other words, for the sentence to be grammatical, shi has to occur as a copula verb (= main verb), which can then take a temporal phrase as its complement.

In sum, sentences like (13) and (15) are in fact not counterevidence to the structural constraint (7).

4. IS SHI AN AD V ERB?

So far, all the evidence has suggested that DO uncleftability in Chinese is caused not by any ad hoc constraint with respect to the object per se, but by a constraint with respect to the linear order between the CFM and the main verb. Consequently, the problem of DO uncleftability in Chinese has no bearing on the proposed AH; rather, it stems from a more superficial word order constraint in Chinese which overrides the effect of the AH. However, this linear constraint has so far only referred to the order between the CFM and the main verb. The ac­count would be more elegant if we could show that (7) is part of an indepen­dently motivated constraint and that the solution of the problem follows auto­matically from something already existing in the language without recourse to any ad hoc constraint.

One possibility, as suggested in Huang (1982b), is to treat the CFM shi as an adverb which, like most other adverbs, occurs before the main verb. However, there are some problems with this analysis, according to the criterion that, syn­tactically, members of a class are expected to behave similarly and are in general mutually exclusive. More specifically, the CFM shi shows some important syn­

64 Cheng h

tactic differences from preverbal adverbs such as jiu 'just', zhen 'really' and 2/ 'only'. First, while no adverb of the kind represented by jiu, zhen and zhi can oc cur in the A-not-A construction, a typical verbal construction6, shi can, as in (19):

(19) a. *T a zhen bu zhen qu?3s really not really go

'Is he really going?'

b. Ta shi bu shi wei Yuehan ding de ji-piao?3s CFM not CFM for John reserve MM air=ticket

'Is it for John that he reserved an air ticket?'

Clearly, the CFM shi behaves differently from the adverbs with respect to this verbal feature.

A similar difference is seen in their respective ability to occur independently as a short answer to yes/no questions. The CFM shi, like full verbs in Chinese, can stand alone as a short answer to yes/no questions, whereas the adverbs can not. Compare

(20) a. A Ni qu ma?2s go Q

'Are you going?'

B: Qu.go

'Yes.'

b. A Ta shi wei Yuehan ding de ji-p iao ma?3s CFM for John reserve MM air=ticket Q

'Is it for John that he reserved an air ticket?'

B: Shi (de).CFM MM

'Yes.'

6Again, I thank an anonymous reviewer for pointing out A-not-A as a typical main verb property. However, I share in part Li's (1990: 63) view that the A-not-A test is also a very common test for verbhood in general. In fact, I hold that A-not-A is better regarded as a test for general verbhood, for the reason that, although it represents a typical feature of the main verb, not all elements that can be used in the A-not-A construction are necessar­ily main verbs. This is seen in the use of modal verbs, which can take the A-not-A form and function as simple answers to questions (Li 1990: 149), the latter being another alleged main verb feature. It is for this reason that I have so far restrained myself from using A- not-A as an argument for or against the main verb status of certain verbal elements.

Ort the Uncleftability o f Direct Object in Chinese 65

c. A Ni zhen qu ma?2s really go Q

'Are you really going?'

B: *Zhen.really

'Yes.'

The form of the short answer is a full verb in (20a), a CFM in (20b), and an ad­verb in the ungrammatical (20c). Thus the CFM shi, again, behaves differently from adverbs by virtue of showing a higher degree of 'verbness' than the latter. Given such important syntactic differences, it is not very plausible if we treat shi as an adverb.

Another argument against treating shi as an adverb is that while the ad­verbs are mutually exclusive among themselves, they can nevertheless cooccur with shi, as in (21):

(21) a. *T a zhen jiu qu.3s really just go

'Is he really just going?'

b. Zheng / jiu shi ta right just cfm 3s

wei Yuehan ding de piao. for John reserve MM ticket

'It is (none other than) he who reserved a ticket for John.'

Since one of the important criteria for establishing membership of a morpho- syntactic class is mutual exclusiveness, or complementarity, the fact that shi cooccurs with this subclass of adverbs suggests that the former cannot be a pos­sible member of the latter. Therefore, evidence in terms of both the nature of shi and syntactic mutual exclusiveness indicates that the analysis of shi as an adverb is problematic.

5. Qu a s i-v e r b s

Alternatively, I will propose that the CFM shi belongs to a class of 'quasi­verbs' whose distribution is restricted to the preverbal position, on the basis of syntactic complementarity and the fact that they share many verbal and non­verbal properties. These quasi-verbs include modal auxiliary verbs as well as the CFM shi.7 Modal auxiliary verbs are words like neng 'can', ken 'will', keyi 'may', and ying (gai) 'should', whose preverbal distribution is exemplified in (22).

7Given that in general preposition-like co-verbs in Chinese also share the verbal and non-verbal properties to be discussed below, one would be tempted to include them as

66 Cheng Li

(22) a. Ta neng qu.3sg can go

'He can go.'

b. *Ta qu neng.

Syntactically, modal auxiliary verbs and the CFM are mutually exclusive, as ii (23):

(23) a. *W o neng shi zai jieshang kanjian ta de.Is can CFM o n street see 3s MM

'It's on the street that I could see him.'

b. *W o shi neng zai jieshang kanjian ta de.Is CFM can on street see 3s MM

The classification of modal verbs and the CFM as 'quasi-verbs' is also based on a number of verbal and non-verbal features shared by the two categories. Firstly, a verb, or sometimes the first syllable of a verb, can occur in the A-not-A structure (Huang 1988), as in (24a), as is also the case with modal verbs and the CFM ((24b) and (24c), respectively).

(24) a. A: Ni ding bu ding piao?2s reserve not reserve ticket

'Will you reserve a ticket?'

B: Bu ding, not reserve

'N o.'

b. A: Ni neng bu neng wei wo ding piao?2s can not can for Is reserve ticket

'Can you reserve a ticket for me?'

B: Bu neng.not can

'N o.'

c. A Ni shi bu shi wei wo ding de piao?2s CFM not CFM for Is reserve MM ticket

'Is it for me that you reserved the ticket?'

quasi-verbs as well. However, this would cause problems in terms of complementarity and subcategorization, problems which deserve further studies in the future. Therefore, I have excluded co-verbs from consideration, with the hope that future studies will shed light on co-verbs either as or not as quasi-verbs.

-)n the Uncleftability o f Direct Object in Chinese 67

B: Bu shi.not CFM

'No.'

B's responses in (24) also show two other verbal feature shared by modal verbs and the CFM - that they can be used as a short answer to yes/no questions, and that they can be negated by one of the negative forms bu, mei (you) or bie.

I Moreover, like full verbs, the modal auxiliary and the CFM can cooccur with a delimiting or emphasizing adverbs such as zhi 'only', zhen 'really' and jiu 'just', as in (25):

(25) a. Ta zhi /zhen wei wo ding le piao.3s only really for Is reserve Perf ticket

'He only/really reserved a ticket for me.'

b. Ta zhi /zhenneng weiwo ding piao.3s only really can for Is reserve ticket

"He can only/really reserve a ticket for me.'

c. Ta zhi / zhen shi wei wo ding de piao.3s only really CFM for Is reserve MM ticket

'It is only/really for me that he reserved a ticket.'

Apart from the above verbal features shared by modal auxiliaries and the CFM, there are also several non-verbal features shared by them. First of all, while a full verb used as the main verb in a sentence can generally take any of the aspectual morphemes such as the perfective -le, the progressive -zhe, and the experiential -guo, as can be seen in (25a), no aspectual morpheme can be used with a modal auxiliary ((26a)) or the CFM ((26b)):

(26) a. *T a zuotian neng-le wei ni ding piao.3s yesterday can-Perf for 2s reserve ticket

'He could reserve a ticket for you yesterday.'

b. *T a shi-le wei ni ding de piao.3s CFM-Perf for 2s reserve MM ticket

'It was for you that he reserved a ticket.'

Also, a full verb is morphologically reduplicatable in full or in part for various purposes, whereas neither modal auxiliaries nor CFM can be reduplicated. For example, a volitional verb may be reduplicated to mark a delimitative aspect (Li & Thompson 1981: 232-26), as in (27a), but this is not possible with modal verbs or the CFM, as in (27b) and (27c), respectively.

(27) a. Ni kankan zhe ben shu.2s look-look this M book

'You read this book (for a while).'

b. *T a nengneng kan zhe3s can-can look this

'He can read this book.'

ben shu. M shu

c. *Shishi ta kan zhe ben shu de.CFM-CFM 3s look this M book MM

'It is he who read this book.'

Finally, while the verb is usually an indispensable constituent in a grammati­cal sentence8, a modal auxiliary or the CFM is dispensable in that its absence would generally not affect grammaticality, although the resulting sentence may to some extent differ semantically from the original one. This is shown in (28):

(28) a. Tamen (neng) tan gangqin.3p can play piano

'They (can) play the piano.'

zai tan gangqin.Prog play piano

'(It is) they (who) are playing the piano.'

Table 1 summarizes the shared verbal and non-verbal features of modal auxiliaries and the CFM . Since they both show positive values for some of the parameters but negative values for the others, the term 'quasi-verb' is employed to attempt a unification between the two categories. Table 1, then, provides a valid basis for classifying modal verbs and the CFM shi as a class of 'quasi-verbs', which occur preverbally. Given these quasi-verbs, we are now able to generalize the linear constraint (7) as (29), which states that a quasi-verb must occur before the main verb in a sentence.

(Shi) tamen CFM 3p

(29) QUASI-VERB MAIN VERB

(29) exists independently as a word order constraint between a quasi-verb and the main verb9 in Chinese. Since it rules out the possibility of a quasi-verb

1. A-not-A2. Short ar3. Negatio4. Cooccui5. Cooccui6. Redupli7. Dispens

5- CONCLUSI

To sum ui

feet objects ch structure fee constrain vaìmQùiv ’¿needed to ;

ii\ conclu, avsMity <

r atweYy ai

8Like many other languages, verbless sentences exist in Chinese, especially in the col­loquial variety. However, since what we are concerned with here is whether in general a sentence should require the presence of a verb, we will not consider possible verbless sen­tences in the language. (] J\USe(* as'

9I realize that identification of the main verb in a Chinese clause is a difficult issue. U ^ ay)0*es ft (1990: 100) mentions three criteria: a) aspect marker, b) the A-not-A form, and c) simple an- 0f Cq l8 Prob swer, the last two of which have been shown in the above discussion to be more appropri- 0ver$)

ccurring postverbally, DO uncleftability is automatically accounted for, and no ther ad hoc explanation is needed.

>n the Uncleftability o f Direct Object in Chinese___________________________________ 69

Table 1:Shared verbal and non-verbal features of modal auxiliaries & the CFM in Chinese

Veib Quasi-verb

Modal C FM

1. A-not-A + + +2. Short answer + + +

3. Negation + + +

4. Cooccurrence with adverb + + +

5. Cooccurrence with aspect marker + - -

6. Reduplication + - -

7. Dispensability - + +

*. Co n c lu sio n

To sum up, DO uncleftability in Chinese as apparent counterevidence to the Accessibility Hierarchy has been shown to be due not to the inherent property of lirect objects nor to inapplicability of the AH itself, but to a constraint on the lin- ;ar structure which is in conflict with, and overrides the effect of, the AH. Since he constraint, which crucially involves the order of a quasi-verb and the main 'erb in Chinese, is independently motivated in the language, no ad hoc account s needed to account for DO uncleftability in Chinese.

In conclusion, while the Accessibility Hierarchy attains some measure of plausibility as a language universal, it nevertheless can interact with, and be legatively affected by, language specific constraints.

tfely used as tests for general verbhood rather than main verbhood. As for the first one, Li .1990) notes that some verbs can not take aspect markers. Such fuzzy areas, though not pos- ng a big problem to the present analysis, have been, and will probably remain to be, areas

controversy.

70 Cheng Lu

REFEREN CES

CHAO, Y-R. 1968. A Grammar o f Spoken Chinese. Berkeley, CA: University o California Press.

GlVÖN, T. 1984. Syntax: A Functional-Typological Introduction. Amsterdam Benjamins Publishing Co.

HSUEH, F. 1983. A note on the grammatical function of gei. Journal o f Chinest Language Teachers Association 14: 81-85.

HUANG, C-T. J. 1982a. Move WH in a language without WH m ovem ent Linguistic Review 1, 4: 369-416.

H U A N G , C-T. J. 1982b. Logical Relations in Chinese and the Theory oj Grammar. Ph.D. dissertation, MIT.

HUANG, C-T. J. 1988. Wo pao de kuai and Chinese phrase structure. Language 64, 2: 274-311.

KEENAN, E.L. 1987. Universal Grammar: 15 Essays. London: Croom Helm.

KEENAN, E.L. & COMRIE, B. 1977, Noun phrase accessibility and universal grammar. Linguistic Inquiry 8 ,1 : 63-99.

KEENAN, E.L. & COMRIE, B. 1979a, Noun phrase accessibility revisited. Language 55: 649-64.

KEENAN, E.L. & COMRIE, B. 1979b, Data on the noun phrase accessibility hierarchy. Language 55: 333-51.

Li, C. 1976 (ed.). Subject and Topic. New York: Academic Press.

L i, C. & THOMPSON, S. 1974, 'Co-verbs in M andarin Chinese: verbs or prepositions? Journal o f Chinese Linguistics 2: 257-278.

Li, C. & THOMPSON, S. 1975. The semantic function of word order: a case study in Mandarin. In Charles N. Li (ed.), Word Order and Word Order Change. Austin: University of Texas Press, 163-95.

L i, C. & THOMPSON, S. 1981. M andarin Chinese: A Functional Reference Grammar. Berkeley: University of California Press.

Li, Y-H A. 1990. Order and Constituency in M andarin Chinese. Dordrecht: Kluwer Academic Publishers.

ROSS, C. 1983. On the function of Mandarin DE. Journal o f Chinese Linguistics 11, 2: 214-46.

SUN, C. & GlVÖN, T. 1985, On the so-called SOV word order in Mandarin Chinese: a quantitative text study and its implications. Language 61, 2: 329-351.

Ort the Uncleftability o f Direct Object in Chinese 71

TAI, J. H-Y. 1973. Chinese as a SOV language. In C. Corum et al. (eds.), Papers from the Ninth Regional Meeting o f Chicago Linguistic Society. Chicago: Chicago Linguistic Society, 659-671.

TENG, S-H. 1979. Remarks on cleft sentences in Chinese. Journal o f Chinese Linguistics 7: 101-114.

TRAVIS, L. 1984. Parameters and Effects o f Word Order Variation. Unpublished Ph.D. dissertation, MIT.

ROMANCE CLITICS & THE STRUCTURE OF AGRP*

Virginia Motapanyane University of New-Brunswick, Saint John

ABSTRACT

This paper proposes a development o f the analysis in Kayne (1990), where clitic movement is related to the projection o f an abstract lN FL(ectional)-head, compatible with infinitive (i.e. non-tensed) clauses. It will be argued, on the basis o f Romanian data, that this abstract iNFL-head is present in tensed clauses as well, and that it can be defined as part o f the Agreement Phrase (AGRP). This proposal is extended to other Romance languages leading to a uniform account o f clitic movement to / N F L .

INTRODUCTION

Clitic movement has been argued to be a process of Romance syntax involving verb-movement to INFL1. The exact factors triggering clitic movement are as yet unclear. But recent studies (cf. Kayne 1989, 1990) indicate a relation between this process and the possibility of verb-movement to IN FL. More specifically, according to Kayne (1990), the specification for clitic elements belongs to the set of INFL properties, and surfaces in two ways: it is either associated with a functional head or it replaces an inflectional function. For example, in tensed clauses, the specification for clitic elements is associated with AGR, the head receiving the verb raised to INFL. In non-tensed clauses, the same specification replaces the function of T[ense], which lacks inflectional features, and turns it into an abstract head, to which only clitics adjoin. Thus, the association between clitic movement and an I-head led, in Kayne (1990), to a principled account of differences in word order, e.g., verb-clitic in Italian vs. clitic-verb in French.

* This work was supported by UNB Grant 23-85. I thank R. Kayne, L. Rizzi and Y. Roberge for comments on earlier versions of this paper.

*The objective of this paper being to apply Kayne's (1990) analysis to Romanian data, I will also have to assume, along the lines of his paper, that object clitics move in the derivation of S-Structure. Parallel analyses contradict the thesis of movement, arguing for the base generation of these elements at a certain level of INFL. See, for instance, Moore (1992), who argues for such an analysis in terms of the Empty Category Principle (ECP). However, the mechanism responsible for the placement of clitics in INFL does not concern my analysis, the purpose of which is to comment on the nature of the inflectional head hosting these elements which could be either base-generated or moved to that level.

LlNCiUlSI ICA atlantica 1 4 ( 1 9 9 2 ) 7 3 - 8 7

74 Virginia Motapanyc

The proposal in Kayne (1990) is crucial for an understanding of the relati< between clitic movement and verb movement in Romance. This paper will ado the analysis in Kayne (1990) as theoretical background (section 1.1.) and w further investigate the properties of the abstract inflectional head (henceforth head) serving as target for clitic movement. Romanian grammar (not tested Kayne (1990)) provides empirical data that will be used to support a ne perspective in the characterization of the I-head (section 1.2.). As a result, tl mechanism of clitic movement will receive a uniform description in Romanie clauses, i.e., non-distinct according to the [+/- tense] feature (section 2.1.), am in a second step, this description will be generalized to other Romane languages (sections 2.2. and 2.3.).

1. A FUNCTIONAL HEAD FOR CLITICS

1.1. Kayne's (1990) proposal

Italian infinitive clauses present the order verb-clitics, although in tense* clauses clitics precede the verb, e.g.:

(1) a. Variargli sarebbe un errore.speak-to-him will be an error

'Speaking to him would be an error.'

b. Sarebbe assurdo che tu gli parlassi, would be absurd that you to-him speak

'It would be absurd to speak to him.'

Kayne's (1990) account for the difference in word order illustrated in (1) relief on the following assumptions: (i) clitics adjoin2 to an I-head; (ii) verb movemeni is to a position higher than the respective I-head in infinitive clauses, hence the order in (la); (iii) in tensed clauses, clitics adjoin to the I-head containing the raised verb, giving the word order in (lb).

The properties of the I-head acquiring clitics in an infinitive clause can be summed up as follows:

(i) This functional head does not carry inflectional features; for example, T[ense] is inert in Italian infinitives and it functions as a host for clitics. This lack of inflectional functions justifies the fact that the verb-movement

2Clitic adjunction to INFL is taken as a particular case of 'incorporation' into a head in Baker's (1988) terms.

Romance Clitics & the Structure o f AGRP 75

process skips this head, without violating the Head Movement Constraint (HMC). The structure of an infinitive clause such as (la) would be (2):

(ii) The representation in (2) implies a distributional property of I-heads with respect to clitic movement: clitics can adjoin only to an abstract I-head, like T in(2), but not to a trace I-head (reserved for verb-movement). In tensed clauses, where T has positive features, clitics will adjoin to a lexical (therefore, non­trace) head, i.e., the I-head containing the raised verb. Thus, the structure of a tensed clause such as (lb) would be (3).

Not all Romance languages display the verb-clitic order in infinitivals. For instance, French has clitic-verb order, contrasting with the pattern in (2). In earlier work, Kayne (1989) related the contrast in word order between Italian and French to the setting for the Null Subject Parameter (NSP): only in Null Subject Languages (NSLs) does the verbal morphology have the property of triggering verb-movement to the highest INFL-Ievel, whereas non-NSLs, with a 'weaker' inflectional system, allow for a restrictive V to I movement, to a position that is lower than T. However, the other Romance languages, all NSLs, may behave either like Italian or French. Therefore, the relation between the NSP and the degree of V-movement does not consistently account for the data.

3The structures of (la) and (lb) come in a bracketed form in Kayne (1990). I reproduce them as tree-form representations, for expository purposes. The hierarchy of IP adopted in this paper conforms to the proposal in Belletti (1989).

(2) Verb

NPtel

Clitic movement in the infinitive IP in Italian3

So far, no formal explanation has been provided for the differences betwee verb-movement in Italian vs. French types of languages, although empiric, observations about systematic contrasts abound.

(3) AG RPI

AGR'

AGR T PCLITIC/VERB I

76_________________________ _________________________________________ Virginia Motapanya, |

This paper does not attempt to answer that theoretical question, although it may contribute to the understanding of the relation between the degree of verb- movement and clitic placement on one hand, and the NSP on the other. This paper has as its purpose to apply Kayne's analysis to Romanian data. The outcome will be a generalization of the abstract I-head for clitics to all types of IP, finite or non-finite. In other words, the analysis will propose a uniform treatment of the structures in (2) and (3).

1.2. THE STRUCTURE OF IP IN ROMANIAN

1.2.1 . Infinitive clauses

Word order in a Romanian infinitive clause is clitic-verb, as in tensed clauses. From this point of view, Romanian contrasts with Italian, as in (4):

(4) a . Parlar#// sarebbe un errore.sp eak-to -h im will be an error

'Speaking to him would be an error/

b. A le impune acest program ar fi o pedeapsà.to to-them impose this program would be a punishment

'It would be a punishment to impose this program on them/

The contrast in word order between the infinitive clauses in (4a) and (4b) is easily explained through differences in the inflectional morphology: in

Romance Clitics & the Structure o f AGRP 77

Romanian, the infinitive mood has a lexical marker a (functionally equivalent to English to), whereas in Italian this element does not appear. As a first step, let us suppose that (2) also underlies the infinitive structures in Romanian. In this case, the mood marker a must occupy the highest I-head, as proposed in (5):

(5) IPI

rI TPa Ito T

T IPle I

to them V

Clitic and verb movement in Romanian infinitive clauses: provisional version

The presence of a lexical element in the highest I-head blocks verb-movement to that level or higher up. Note that a, like English to, is an unbound morpheme and cannot trigger verb-raising to the respective head. According to the analysis proposed for the Italian clause in (2), T[ense] is an abstract head functioning as a host for clitic movement. Under these conditions, the verb can move only to the lowest I-head4. The hierarchy in (5) is predictable under Kayne's (1990) analysis:

^Obligatory verb movement in Romanian infinitive clauses is confirmed by tests with VP-adjoined adverbs and Floating Quantifiers (FQ), e.g.:

(i) Am inceput a nu-1 mai invita a§a des pe Ion.(I) have begun to not-him more invite so often pe-John

'I began not to invite John so often.'

(ii) Au inceput a-1 invita foji pe Ion.(They) have begun to him invite all pe-John

'They all began to invite John.'

78 Virginia Motapanyu

in this language the verb does not move higher than T[ense], and the word ord will be clitic-verb.

1.2.2. Tensed clauses

The structure in (3) represents the conditions allowing clitic m ovem ent; clauses where T[ense] has positive features and functions as an inflectional v abstract I-head. Because there is no abstract T, clitics will adjoin to a lexica head, i.e., the only non-trace I-head. This situation corresponds to the Italia sentence in (lb ) and can be generalized to all the tensed contexts in Romanc languages, where clitics always precede the verb.

Although this generalization captures an indisputable fact of Romanc grammar, the representation in (3) does not attain descriptive adequacy when : comes to Romanian tensed clauses. The empirical problem arising in thi language is that clitics and verb can be separated by a certain class of adverbs, a shown in (6):

(6) a. Ion nu o mai invita a§a des pe5 fata aceea.John not her more invite so often pe-girl that

'John no longer invites that girl so often.'

b. Ion nu o prea suporta pe Maria.John not her quite stand pe-Mary

'John can barely stand Mary.'

The elements intervening between clitics and verb belong to the class oi negative adverbs, presumably in an A-bar position within IP (but see the discussion in the following section). The word order in (6) displays the negation and the clitics at a higher level than the verb. If the clitic cluster adjoins to AGR; as in (3), the verb must occupy a lower I-head, because the adverb of negation precedes it.

It is obvious that the representation in (3) cannot capture the hierarchical difference noticed in (6). In this sense, an account of verb-movement in Romanian tensed clauses is in order. The status of the I-head acquiring clitics in this language will therefore be discussed.

In both examples, the italicized adverb or the FQ that are either adjoined to VP or in SPECVP (cf. Sportiche (1988)), follow the verb and precede the direct object. Hence, the verb must be higher than VP.

5Romanian displays the phenomenon of clitic doubling when the direct object is a [+human] NP. In (6a) the object clitic o 'doubles' the NP fata aceea 'that girl' and absorbs the verbal case (Accusative). A dummy preposition pe is inserted as case assigner to the NP fa ta aceea. For more details on clitic doubling facts see Dobrovie-Sorin (1990).

Romance Clitics & the Structure o f AGRP 79

2. THE RELEVANCE OF CLITIC PLACEMENT FOR V TO I MOVEMENT

2.1. V to I Movement in Romanian

In Romanian, the verb moves to INFL according to the pattern generally assumed for Romance languages since Belletti (1989). That is, the verb raises cyclically to T and AGR. This process is confirmed by the inflectional suffixes on the verb, marking distinctly the [tense] and [+agr] features, as in (7):

(7) Infinitive a face 'to do'

Indicative present facem = radical fa ce + -m (lpl)

Indicative imperfect faceam = radical face + a + m (lpl)

In (7) the forms of the present and imperfect indicative contrast through the presence vs. absence of the [tense] suffix a , visible between the radical of the verb and the marker for subject-agreement. The order of affixes shows that verb movement is to AGR in Romanian.

Although its inflectional morphology characterizes the verb-movement process in Romanian as similar to that of other Romance languages, empirical evidence presented in (6) indicates clear contrasts in hierarchical structure. In order to draw a comparison between the two types of structures I will give, first, the representation of a tensed IF in Romance, following Belletti (1990).

(8) Gianni non lo mangia piii.

The head AGR in (9) acquires two series of elements : the verb, moved by a substituting movement, and the clitic cluster, adjoined through an incorporation movement into a head. The group of clitic elements contains clitic pronouns and the negation. The negation follows the verb. Contrasting with the hierarchy of elements in (9), the word order in (6) shows the adverbs of negation mai, prea preceding the verb. This situation could be analysed in two ways: (i) the verb does not move to AGR, while clitics do; or (ii) the verb lands in AGR, while clitics do not.

The first possibility would violate several constraints, i.e., verbs raise obligatorily to INFL in Romanian (see footnote 3 and the word order in (6a)), and the verbal form displays marks for subject-agreement. Hence, the verb must pass through the AGR-head in Romanian too. Let us turn to the second possibility: the verb occupies AGR whereas clitics adjoin to a higher head. What would be the status of this I-head that does not trigger verb raising but allows for clitic movement?

80 Virginia Motapanyat

(9) AGRP

Spec AGR'

AGR NEGP

Gianni John

'John no longer eats it'

The structure of IP in Italian

It is obvious that this functional head must not correspond to an inflectional specification which would attract the verb. In this case, it must be an abstract I- head of the same type as T in infinitive clauses, as discussed for (2). Note that in structures like (6) T has positive features, i.e., it is a trace-head and cannot function as a host for clitic movement. Therefore, the abstract I-head receiving the clitics in (6) cannot be labelled T.

Taking into consideration that in a tensed clause both clitic and verb- movement depend on the existence of an AGR-head6, the most natural direction to follow would be to relate these processes to properties of AGR, classifiable into two series: (i) verbal morphology; (ii) nominal features7. The verbal morphology has the function of defining the verb for mood, tense, person; the set of nominal features would mirror, at the level of INFL, the thematic relations generated within VP (or Vmax). In this sense, we would expect that clitics in INFL cooccur with lexical NPs in argument position, as, for instance, in languages with clitic-doubling:

6The traditional assumption in the literature is that both clitic and verb movement occur in languages with a 'rich' AGR, e.g., in Romance languages, but not in English, with its 'weak' AGR (cf. Pollock (1989) amongst others).

7See also Borer (1989) for a characterization of AGR as a +N(ominal) element.

Romance Clitics & the Structure o f AGRP 81

(10) Maria l .a invitat pe Ion*.Mary him -has invited pe-John

'Mary invited John.'

In (10) the sole function of the object clitic adjoined to the auxiliary verb is to reflect the relation between verb and the object Ion 'John'. The same doubling process applies to indirect objects in Romanian. In other languages (e.g., Italian dialects), the clitic doubles a subject. However, the coexistence of clitics and the coresponding NP in the structure may be an option that depends on independent restrictions, according to the grammar in question. From this perspective, it is expected that there will be languages in which clitics can double only empty NPs8-

Let us then assume that in Romance languages the Inflectional Phrase has the property of reflecting the thematic relations through means of clitic elements. These clitics will occupy a specific I-head, projected in the structure to implement this function. Along these lines, I will propose a representation for the Romanian sentence in (6) so that the two sets of Inflectional features, i.e., for verbal morphology and clitics, can be distinguished at the level of AGRP.

However, before I implement this proposal, a remark is in order: the sentence under examination contains a negation, to be dealt with according to current work in the field (e.g., Belletti (1990) and Zanuttini (1990) and subsequent work). These analyses have in common the treatment of the negation (French ne, Italian non, Romanian nu) as a head-element projecting a phrasal structure, NEGP. For reasons debated at length in Zanuttini (1989), NEGP systematically selects a TP within the clause. With respect to the placement of NegP within IP and the syntactic status of the negation, I will follow Belletti (1991), who argues in favour of a characterization of non as a syntactic clitic which undergoes movement to the head hosting the clitic cluster at S-Structure. As far as the internal structure of NEGP is concerned, SPECNEGP may optionally contain adverbs, semantically compatible with negations. These adverbs acquire a negative interpretation through the agreement relation between SPECNEGP andNeg .

The approach put forth in Belletti (1990) and (1991) achieved a uniform characterization of negative sentences in Romance languages: NEGP, inserted between AGRP and TP, contains a negative adverb in SPECNEGP; the negation moves to AGR at S-Structure, to join the clitic cluster, and gives a word order where the negation precedes the verb and the adverb, as illustrated in (9) above.

8Note that an analysis of clitics along these lines would rather favour approaches similar to Moore (1992), where the argument position contains a PRO.

82 Virginia Motapanya

However, there is room for word order variation, as data from Italian sho that the negative adverb can be generated in other positions in the structure an acquires its negative interpretation through movement to SPECNEGP at the levi of Logical Form (LF)9 For the representation in (11) I will take into consideratio both general analyses of NEGP and cases of free placement of negative adverbs.

(11) AGRP

Spec a g r '

AGR2 AGRP

John not herm oreinvites so often pe girl that

'John no longer invites that girl so often'

The structure of a tensed IP in Romanian.

The analysis proposed in (11) distinguishes between two levels of AGR: a higher head with the function of receiving clitic elements (negation counting as a syntactic clitic according to Belletti (1991)); a lower head, containing the morphology for subject-agreement, to which the verb moves through a head-to- head substituting movement. The empirical reason for this split comes from the position of the negative adverb mai in (6), between the clitic cluster and the

^The examples in Belletti (1990) contain the adverb piu following the past participle in complex tenses; the author considers this adverb VP-adjoined then moved to SPECNEGP at LF. This movement must be obligatory, as the adverb does not have a negative value by itself (i.e., it can appear in positive versions).

Romance Clitics & the Structure o f AGRP 83

inflected verb. According to (11), the negative adverb must have also moved from its base position SPECNEGP, to a higher SPEC, from which it can cliticize to the verb at the level of Phonetic Form (PF).

The status of mai as clitic at the phonological level, classified as a 'short' word in traditional grammar, has been already discussed in the literature (see Dobrovie-Sorin (1990)). In this case, it could be argued that mai adjoins to the verb in whatever position it happens to be, including the position the verb could share with the clitic cluster. However, certain restrictions on word order show that the placement of mai is not so arbitrary, but rather corresponds to a determined position in negative sentences. Consider the examples in (12):

(12) a . Sa nu mai fie chemat.(he) SA not more be called

'Let him not be called any more'

b. *Sa nu fie mai chemat.

The contrast between (12a) and (12b) indicates that mai does not depend on the verb, but on a specific position within INFL. More precisely, it depends on a SPEC position related to the head that receives the verbal element inflected for subject-agreement - in (12a) the passive auxiliary f ie 'be'.

The clitic property of 'short' negative adverbs in Romanian like mai, leading to their movement to SPECAGRP, makes it possible to distinguish between two I- heads at that level. The equivalent adverbs of other Romance languages have a full lexical status and do not move at S-Structure.10 In other words, although all Romance languages display the IP-structure in (11), the data did not provide the relevant evidence.

2.2. V to I Movement in Romance

Extending the representation in (11) to all Romance languages would confirm parallel analyses of a two-headed AGRP in this group of languages, for independent reasons. Thus, Roberts (1991) proposes a two-head AGR level to characterize a 'rich' verbal morphology in Old English, contrasting with the 'weak' one-head AGR in modem English.

Taking Roberts' analysis even further, Belletti (1990) proposes a complete split between the two levels of AGR, which would adequately account for the mechanism deriving complex tenses in Romance languages. More exactly, Belletti argues that auxiliaries, inflected for subject-agreement, do not move to the same I-head as lexical verbs do, but to a higher level. This analysis finds

10But see Martineau (1992) for arguments for a similar phenomenon in Old French.

84 Virginia Motapanyaì

support in Romanian, where 'short' negative adverbs follow the auxiliary, e shown in (13):

(13) a . Maria nu 1- a mai invitât pe Ion.Mary not him-has more invited pe-Ion.

b. *Maria nu-1 mai a invitât pe Ion.

c. Maria nu-1 mai invità pe Ion.

'Mary no longer invites John.'

If m ai moves to SPECAGRP of AGRl, the order in (13a) indicates that thi auxiliary goes to a higher head, after receiving the agreement mark. Th< contrast between the level of verb movement in (13a), to AGR2, vs. (13c), t( A G R l, finds its theoretical support in the inherent differences betweei auxiliaries (as morphological tools) and lexical verbs, which serves as evidence for structural differences.

To sum up the discussion so far, I argued that clitics in Romanian occupy an I head solely specified for this function. Thus, I dissociated clitic movement froir inflectional morphology. Implicitly, I assume that the chains resulting from thf types of movement, i.e., clitic vs. V to I, obey separate conditions and do not interfere with the ECP. Negation being a syntactic clitic, its movement to the clitic cluster (as in (11)) will also be independent of the conditions restricting the movement of the verb. In a second step, I extended the characterization of Dé­structure in Romanian to other Romance languages, where clitics behave in the same way.

This new perspective on the placement of clitics within a tensed IP entails a reconsideration of the infinitive structures discussed in the first part of this paper. If clitic movement to INFL occurs independently of verbal inflection, it is expected that the same mechanism applies to infinitive structures. The prediction will be that the positive or negative value of an I-head, such as T, must be irrelevant for the process of clitic movement.

2.3. INFINITIVE CLAUSES REVISITED

In most Romance languages infinitive clauses are non-finite: the verbal form does not display marks for subject-agreement. As a result, an AGRP would also be excluded from the structure of such a clause. However, the analysis of a split AGRP as proposed in (11) gives us the possibility of differentiating between the

Romance Clitics & the Structure o f AGRP 85

(14) a. parlarglito speak-to-them

IP

Spece

V

AGRPIparlar to speak AGR

(14) b.

The structure of the infinitive IP in Italian

a le impuneto to-them impose

IP

Spece

r

iato

AGRPI

AGR'

AGR2le

to-them

TPI

V

The structure of the infinitive IP in Rom anian

86 Virginia Motapanyai■ nceClii i

morphology for verb-subject agreement and the nominal features of AGR. Froi this point of view, only AGR2 is compatible with the structure of infinitiv clauses. The schemas in (14) adapt the representations in (2) and (5) to thi analysis.

The status of T has been readjusted in (14), where this head functions as pai of the inflectional chain (a trace-head). Such a description is more adequate t< the empirical data in Romance, where infinitive clauses may have [+tense features11.

The structure of the infinitive clauses in (14) matches the hierarchy of tensec clauses (e.g. (9), (11)) where AGR is higher than T. The function of AGR2 is alsc the same in finite and non-finite contexts: it is specified for clitic movement Romanian data clearly indicate that the AGR compatible with infinitive INFLf carries only the functions of AGR2 and therefore qualifies as an abstract I-head independent of the inflectional chain. This analysis conveniently covers Italian1' or any Romance infinitive clause13.

3. Co n c l u sio n

Clitic movement in Romance is a process of incorporation into an abstract I- head, (cf. Kayne (1990)) characterized in this paper as an instantiation of AGR. The abstract AGR is proper to structures where INFL triggers verb movement: the presence of an abstract AGR implies the existence of a lower AGR, carrying inflectional affixes. All Romance languages have this double headed AGRP. From this viewpoint, clitic and verb movement are distinct processes related to a common factor - the structural possibility of a 'rich' AGRP.

REFEREN CES

BAKER, MARC. 1988. Incorporation: A Theory o f Grammatical Function Changing. Chicago: Chicago University Press.

BELLETO , ADRIANA. 1989. Agreement and case in past-participial clauses in Italian. Genève: Université de Genève (manuscript).

^T h e possibility of Nominative case lexical subjects in infinitive clauses has been related to the positive features of T. For Romanian see Dobrovie-Sorin (1990) and Motapanyane (1991).

12Determination of the exact level of verb movement in Italian clauses is not the purpose of this paper. It may be the highest I-head, corresponding to the level of the mood marker a in Romanian, or the movement may cross the clausal IP border in syntax (cf. Kayne 1990).

13Borer (1989) also argues, for independent reasons, for the existence of an AGRP in infinitive clauses.

e l e t t i , a d : G enève

ELETTI, AD G eneve

JORER, HAGl Subject

DOBROVIE-SStudies

Kayne, R ic t Saffir (

Kayn e , Ric i B .P hili the N o G radui A m h ei

Ma r t i n e a lP a p e rP .E .I .,

Moore j o tpresen

Mo t a p a min Ron

Po l l o c k , j -Lingui

Ro b e r t s , ld e G e

Ro b e r t s , d 217 .

Sp o r t ic h e ,coroll

Zanuttin]M .D (MeetiU niveMass;

Romance Clitics & the Structure o f AGRP 87

BELLETTI, ADRIANA. 1990. Generalized verb-movement. Genève: Université de Genève (manuscript).

BELLETTI, ADRIANA. 1992. Verb positions; NP positions: Evidence from Italian. Genève: Université de Genève (manuscript).

BORER, HAGIT. 1989. Anaphoric AGR. In O. Jaeggli and K. Safir (eds.), The Null Subject Parameter. Dordrecht: Kluwer Academic Publishers, 69-109.

DOBROVIE-SORIN, C A R M EN . 1990. The Syntax o f Romanian: Comparative Studies in Romance. Dordrecht: Foris.

KAYNE, RICHARD. 1989. Null subjects and clitic climbing. In O. Jaeggli and K. Saffir (eds.), The Null Subject Parameter. Dordrecht: Foris, 85-103.

KAYNE, RICHARD. 1990. Romance clitics and PRO. In J. Carter, R-M. Déchaine,B. Philip & T. Sherer (eds.), NELS 20. Proceedings of the 20th Meeting of the North Eastern Linguistics Society held at Carnegie-Mellon University. Graduate Linguistic Student Association, University of Massachusetts, Amherst, Mass., 255-302.

MARTINEAU, FRANCE. 1992. La négation en français: une question de temps? Paper presented at the Learned Societies Conference, Charlottetown, P.E.I., Canada, 1992.

MOORE, JOHN. 1992. Romance cliticization and relativized minimality. Paper presented at the Linguistic Society of America Conference, Chicago, 1992.

MOTAPANYANE, VIRGINIA. 1991. Theoretical Implications o f Complementation in Romanian. Ph.D. dissertation. Genève: Université de Genève.

POLLOCK, J-Y. 1989. Verb movement, universal grammar and the structure of IP. Linguistic Inquiry 20, 3: 365-424.

ROBERTS, IAN. (forthcoming) Verbs and Diachronic Syntax. Genève: Université de Genève (manuscript).

ROBERTS, IAN. 1991. Excorporation and minimality. Linguistic Inquiry 22,1: 209- 217.

SPORTTCHE, DOMINIQUE. 1988. A theory of floating quantifiers and itscorollaries for constituent structure. Linguistic Inquiry 19, 3: 425-449.

ZANUTTINI, RAFFAELLA. 1990. Two types of negative markers. In J. Carter, R- M. Déchaine, B. Philip & T. Sherer (eds.), NELS 20. Proceedings of the 20th Meeting of the North Eastern Linguistics Society held at Carnegie-Mellon University. Graduate Linguistic Student Association, University of Massachusetts, Amherst, Mass., 517-530.

à

EPISODES AS MEMORY UNITS IN DISCOURSE REPRESENTATION:

ANAPHOR USE IN ENGLISH AND MANDARIN

Ming-Ming Pu Gary D. Prideaux Lois M. Stanford

University of Alberta

ABSTRACT

The present study investigates the relationship between cognitive constraints and d iscourse processing. In particular, it examines how the hierarchical structure o f discourse is organized and presented, and what consequences that structure has fo r the linguistic coding em ployed during the dynam ic time course o f discourse processing. Special attention is directed to the distribution o f fu ll NPs and pronouns in discourse as a function o f the location o f episode boundaries. Results obtained from an experiment support the hypothesis that episode structure governs the alternative use o f anaphora.

1. INTRODUCTION1

The notion of episode or paragraph as a semantic unit dominated by a macroproposition has been extensively investigated and discussed both in lin­guistics (Longacre 1979, Schank and Abelson 1977, Hinds 1979, van Dijk 1982, van Dijk and Kintsch 1983, Tomlin 1987) and in psychology (Black and Bower 1979, Haberlandt, Berian and Sandson 1980). From both these perspectives, an episode (a semantic unit) is generally regarded as a set of actions described to obtain a subgoal, along with the outcome of those actions. Black and Bower (1979) demonstrated, in a psychological study of story processing, the existence of episodes as chunks in narrative memory. Their study yielded three basic re­sults. First, episodes in stories are stored as separate chunks in the memory rep­resentation of the story; second, adding related but relatively unimportant ac­tions to a story episode increases the recall probability of the important state­ments in the episode; and third, the recall probability of a general superordinate action in a story increases with the number of subordinate actions that further specify it. Similarly, R. Guindon and W. Kintsch (1982), in their experiment studying the macrostructure of texts, found that macrostructure formation ap-

!W e are indebted to R.S. Tomlin for his advice, suggestions, and encouragement. We would also like to thank J.T. Hogan, S.A. Thompson, G. Malicky and P. Schneider for their helpful comments. This work was supported in part by a dissertation fellowship from the University of Alberta.

I . I N t i U I S I I C A at ¡ init ial 1 4 ( 1 9 9 2 ) 8 9 - 1 0 2

90 Ming-Ming Pu, Gary Prideaux, Lois Stan -d as

pears to be a virtually automatic process. That is, people appear to fen macrostructure during reading and derive relevant macropropositions of a p sage as soon as possible. Their findings provided evidence for the 'episode' i the 'm acrostructure' theories of van Dijk and Kintsch (1978) and Schank a i Abelson (1977).

Other studies of story processing (Mandler and Johnson 1977, Kintsch 19 , Haberlandt, Berian and Sandson 1980) suggest that readers slow their proce - ing at or around the episode boundary more so than would be predicted on t » basis of sentence level and text-level factors. The boundary hypothesis, whi i derives from these findings, assumes that there are cognitive processes at ■ around the episode boundary which are not present inside the episode. At t ■ beginning of an episode, several overlapping cognitive processes may occi Haberlandt, Berian and Sandson (1980), who tested the boundary hypothei with reading and recall experiments, found that the encoding load was great at the boundary nodes than elsewhere, suggesting that readers are sensitive episode boundaries and use them in encoding story information.

In studies on anaphora patterning, many linguists (Tomlin 1987, Marslei Wilson, Levy and Tyler 1982, Givon 1983 a & b, Fox 1987a, van Dijk 1982, vc Dijk and Kintsch 1983, Chafe 1987, Tomlin and Pu 1991) have examined the in portance of episodes to the syntax of reference and have found a crucial reli tionship between the episode and the limited capacity of short-term memor They argue that referential choices may be regarded as stemming from sue factors such as textual representation in episodic memory, the role of episodi situations and other world knowledge, and the capacity of short-term memor} Many text-oriented studies show that full NPs may be used at the beginning c an episode to introduce or reinstate reference, while pronominal NPs tend to b used internally to maintain reference and cohesion.

The present study draws on the discussion of the prior research and attempt to examine further the relationship between the episode and its cognitive corre late, the memory chunk, and to determine how this relationship is realized ii linguistic coding. We argue that in discourse production, the speaker, con strained by short-term memory, organizes discourse into a sequence of episodes The encoding load for discourse representation is assumed to be greater at ar episode boundary than that within an episode since an episode is a memory unit which represents sustained attentional effort and endures until an episode boundary is reached. Attention shifts at the episode boundary, where new agents, places, times, objects or possible worlds are expected to be introduced, since each episode is subsumed by a different macroproposition (cf. van Dijk and Kintsch 1983).

Episodes as Memory Units in Discourse: Anaphor Use in English & Mandarin 91

Such a cognitive-based discourse organization may be reflected in anaphoric patterning in discourse production along the following lines: full NPs (more marking materials)2 are expected to be used at episode boundaries to mark the­matic discontinuity between episodes, while pronominals (less marking materi­als) are expected within the episode to maintain referential continuity. These predictions are quite congruent with Givon's general psychological principle: 'Expend as much energy on a task as is required for its performance.' (1983a: 18).

Based on this crucial relationship between cognitive activities of attention and memory and discourse organization, we propose the following hypothesis:

H ypothesis: The speaker's cognitive processes of attention and mem­ory control discourse organization, which is manifested in the alternative use of anaphors.

That is, during discourse production, the speaker is hypothesized to organize his discourse into episodes and to present the episodes by differential use of NPs and pronominals.

In order to test this hypothesis, an experiment was conducted in which sub­jects were asked to watch a video program consisting of three episodes from a storybook and then construct short narratives describing what they had seen. The experimental study is detailed in the following section.

2. THE EXPERIMENTAL STUDY

2.1. Stimulus Materials

The stimulus materials for the study consist of adaptations of three excerpts from a children's picture storybook (without a written text), Here comes Alex Pumpernickel, by Fernando Krahn (1981). The storybook is about a little boy named Alex Pumpernickel, presented by vivid and witty illustrations. The book consists of eight short stories, each of which describes the activities of a certain period of time during a day in Alex's life. Each natural episode is given a title based on the activity, with a picture-clock denoting the time of the day.

The three natural episodes adapted for the present experimental study are subtitled: 1) Alex Pumpernickel in a sticky situation, 2) Alex Pumpernickel swats, and 3) Alex Pumpernickel lends a hand. Each of the three natural episodes con­sists of 8 pictures which are presented in pairs on each page. The three episodes- twelve pairs of pictures, with the subtitles removed, were made into a black and white video program which could be viewed directly as a cartoon sequence of 24 pictures from a Macintosh screen. The resulting video program was de­

^The terminology more/less marking materials is Giv6n's (1983a).

92 Ming-Ming Pu, Gary Prideaux, Lois Stanfo,

signed to give as little presupposition as possible to subjects so that rich and ur contaminated data could be collected for the study.

2.2. Experimental Conditions

Two experimental conditions, Even and Odd, were established to test th present hypothesis. In the Even condition, the picture sequence was presented i the original pairs; that is, there are twelve picture frames, and the three origins story episode boundaries do not cut into any of the frames. The presentation o the pictures in the Even condition is illustrated in Figure 1.

"Alex Pum pernickel... in a sticky situation"

EB 1 1 2 3 4 5 6 7 8

Fram e 1 Frame 2 Fram e 3 Fram e 4

'Alex Pum pernickel... swats"

EB2 9 10 11 12 13 14 15 16

Fram e 5 Frame 6 Fram e 7 Fram e 8

i'Alex Pumpernickel lends a hand"

EB3 17 18 19 20 21 22 23 24

Fram e 9 Fram e 10 Fram e 11 Fram e 12

Figure 1: Even Condition

In the Odd condition, the first single picture of the first episode ("Alex Pumpernickel in a sticky situation') was presented alone and the rest of pictures still in pairs, with the last single picture of the last episode ('Alex Pumpernickel lends a hand') also presented alone. There are therefore thirteen picture frames in the Odd condition, and two of the three original story episode boundaries (EBs) are embedded in two of the picture frames (i.e. EB 2 cuts into Frame 5, and EB 3 cuts into Frame 9). The presentation of pictures in the Odd condition is illustrated in Figure 2.

Episodes as Memory Units in Discourse: Anaphor Use in English & Mandarin 93

"Alex Pum pernickel... in a sticky situation"

EB1 1 2 3 4 5 6 7

Fram e 1 Frame 2 Frame 3 Fram e 4

EB2 "Alex Pum pernickel... swats"

10 11 12 13 14 15

Frame 5 Frame 6 Frame 7 Fram e 8

EB3 "Alex Pumpernickel lends a hand it

16 17 18 19 20 21 22 23 24

Fram e 9 Frame 10 Frame 11 Frame 12 Fram e 13

Figure 2: Odd Condition

Since discourse organization is assumed to be controlled by cognitive pro­cesses, it was expected that subjects would response to the episode boundary in exactly the same way regardless of how the picture sequence was presented. In other words, subjects were expected to recognize and mark the episode bound­ary with full NPs regardless of whether or not it is embedded within a picture frame.

2.3. Subjects and procedures

Forty volunteers participated in the experiment. Twenty were native Chinese speakers who were all students at the University of Alberta and who speak English as a second language (though they vary a great deal in English fluency). Fifteen of the Chinese participants speak Mandarin as their first language, the other five (from the South of China) speak fluent Mandarin and communicate with other Chinese mostly in Mandarin. All Chinese subjects did the experiment in Mandarin Chinese.

All twenty English participants were native English speakers. Thirteen were students at the University of Alberta, and the other seven were university grad­uates. All forty subjects were adults (over the age of twenty); ten of the Chinese speakers and seven of the English speakers were male. All English subjects did the experiments in English. English subjects were assigned randomly to the two

English conditions, English Even and English Odd, while Mandarin speakei were randomly assigned to the two Mandarin conditions, Mandarin Even an Mandarin Odd.

There were two narrative production tasks for each subject: an on-line dc scription task and a recall task. In the on-line description task, which was peí formed first, subjects were asked to watch the video program and at the sam time produce a story based on the pictures presented on the screen. Each subjec was told to take as much time as needed for each single picture or pair of pic tures on a screen. Once finished with a screen, the subject could not see it again.

Upon completion of the on-line description, subjects were asked to recall th< entire story they had just described. There were instructed to retell as much a: possible of the story without looking at the picture sequence again on th< Macintosh screen. The recall task is the focus of the present paper. Sinc< episodes are expected to act as separate memory units/chunks, subjects shoulc be able to structure and mark such units linguistically.

In the recall task, each of the Chinese groups - M andarin Even anc Mandarin Odd - was divided into two subgroups: five of each group performec the recall task in oral form and the other five in written form. The task was sc divided for the following reason. Mandarin makes no gender distinction among third-person pronouns in oral form; all third-person singular pronouns ('he/she/it') have the same pronunciation ta. Chinese subjects thus might have to use NPs to distinguish male characters from female characters in orally retelling the story, where the pronouns would do in English. However in written Mandarin, gender distinction is present for personal pronouns, and there are three different forms for 'he', 'she', and 'it'. By performing a written recall task, subjects would be able to use disambiguating pronouns instead of NPs. Thus we might distinguish disambiguating anaphors from those sensitive to episode boundary conditions by comparison of oral and written productions for the Mandarin groups.

3. RESULTS

3.1. Psychological reality of episodes

The episode boundary results obtained from the recall task in all four groups give evidence that episodes exist as chunks in narrative memory. Subjects in each of the four groups performed the recall task immediately after the on-line description task. Although there was no written clue in either of the conditions that there were three episodes in the story, 34 out of 40 subjects (85%) recognized the three natural episodes and mentioned the fact overtly. Overall, the

94_____________________________________________ Ming-Ming Pu, Gary Prideaux, Lois Stanfo,

Episodes as Memory Units in Discourse: Anaphor Use in English & Mandarin 95

Mandarin groups3 had the same rate of episode recognition as the English groups: 85% (17 out of 20) across the languages. However, both Even conditions had higher rates of episode recognition than both Odd conditions: 90% (9 out of 10) for the English Even and 100% for the Mandarin Even; 80% (8 out of 10) for the English Odd and 70% (7 out of 10) for the Mandarin Odd. Each one of these 34 'recognizing' subjects overtly mentioned that the story consisted of three episodes. They used the phrases such as 'three short stories', 'three sections', 'three parts', 'three episodes', 'three groups/sets of pictures', etc. Such mention generally occurred at the beginning of the recall, and many subjects again started each of the three episodes with 'the first story/section ... ', 'the second story/section...', and 'the third story/section ...'.

More interestingly, some subjects' recall showed the specific monitoring role macropropositions play in discourse processing. Five of the 34 subjects who overtly mentioned the three natural episodes then failed to remember the con­tent of all three during their recall. Of the five subjects, two were in the English groups (1 in Even and 1 in Odd), and 3 in the Mandarin groups (1 in Even and 2 in Odd). The one episode they tended to forget was always the middle one ('Alex Pumpernickel swats'). The five subjects remembered and recalled the first and the third episodes first, and then commented that they remembered there was one more episode, but could not tell what it was. The way they finally recalled the second episode is also of great significance: each first recalled the paragraph level theme, or macroproposition, and then the whole episode came flowing out. Some exact wordings were: 'Well, I remembered it's the boy chasing the fly ...', 'Okay, it's about the kid swatting a fly ...', or 'Yes, it's about the boy and the fly'.

In addition to the overt mention of the three episodes, subjects consistently marked the beginning of each episode, using full NPs reinstating reference throughout their recall task. This demonstrates, as the boundary hypothesis (Mandler and Johnson 1977, Kintsch 1977, Haberlandt, Berian and Sandson 1980) claims, that cognitive processes at episode boundaries are different from those inside the episode. The subject had to devote a special effort to encoding

3Two sets of data (written and oral recall) were collected for the Mandarin recall task since the Mandarin third person pronouns have gender distinction in their written form.It was expected that Mandarin subjects would use more lexical pronouns in their written recall to differentiate between referents than in their oral recall. However comparing lexical pronouns used between the written and oral recall, there is no difference found for the human central character: for the Even condition, 27.45% in the written and 27.01% in the oral; for the Odd condition, 22.70% in the written and 22.61% in the oral. As for the human non-central characters, lexical pronouns used in the written recall are even fewer than those used in the oral recall. Since no difference is found in subjects anaphoric choice between the Mandarin written and oral recalls, the two sets of data are discussed as one in the present study.

96 Ming-Ming Pu, Gary Prideaux, Lois Stanfc

the beginning of an episode because (1) the subject tried to grasp the initiatii and topical event of the episode during the quick flow of discourse processing, ( the subject identified the protagonist of the episode and established a new men ory location for the protagonist, (3) at the beginning, the subject shifted the pe spective, breaking the sustained attentional effort for the previous episode eve when the protagonist of the episode remained the same, and (4) the subje< lacked the expectations that facilitate inference processes. Without such expe< tations, the beginning is essentially an isolated node.

3.2. Alternative Use of Anaphors

We now examine in detail how subjects' episodic organization was reflecte< in their reference management. As was shown in the present experimental re suits, NPs were used at the boundary to mark the boundary shift or thematic dis continuity, and pronouns (including zero anaphora) within the episode t< maintain referential continuity. Table 1 shows the frequencies (the second anc third columns) of full NPs and pronominals (including zero anaphora) used a the three original episode boundaries and within them, and the proportions o Hits (the final column), namely, the proportions of NPs used at episode bound aries plus pronominals used within episodes.

Table 1: Episode Boundary Results

ANAPHOR NP PRONOUN PROPORTION OFEB NEB EB NEB HITS (% )

English Even 73 125 2 340 76.48English Odd 71 131 3 360 76.46

Mandarin Even 75 139 0 353 75.49Mandarin Odd 76 143 3 294 72.71

TOTAL 295 538 8 1347 75.05

EB = at an episode boundary NEB = within an episode

Table 1 shows that the hit rates of the four groups are strikingly similar, with an average hit rate of about 75%. There is no statistically significant difference (at the 0.05 level) for the hit rates between the two conditions of the same lan­guage groups, nor is there any between the two languages. The results demon­strate that in discourse production the episodic organization which reflects at­tention/memory spans is indeed coded in the syntax of reference. It is obvious that the encoding load at the episode boundary is always greater than at the re­

Episodes as Memory Units in Discourse: Anaphor Use in English & Mandarin 97

maining part of the episode so that subjects expended more 'energy' on the task by using more marking materials.

4. DISCUSSION

Although the results of the experimental study are consistent with the present hypothesis, there still remains 25% of the overall tokens which appear to run counter to the hypothesis, i.e., intra-episode NPs (those used within the episode- about 24%), and inter-episode pronominals (those used at the boundary - about 1%). All 8 cases of inter-episode pronominals were used by the six subjects who didn't recall the story as consisting of three episodes. None of these men­tioned that there were three episodes in the story in their recall, nor did they mark the boundary as most of the others did. Since almost all counter-examples are intra-episode NPs, the following section focuses on these.

4.1. Analysis of counter-examples

A close study of the data reveals some secondary functions of NPs used within the episode. After the episode boundary is recognized and marked, other impor­tant features within the episode also requires extra attention on the part of the speaker or/and the listener, such as differentiating the protagonist and minor characters and signaling changes of events, perspectives, participants, etc. Consequently, more marking materials (full NPs) may be used in these specific places to mark the different status of characters in the story and some minor thematic discontinuity within the episode. The general patterns of NPs used within the episode reveal several distinguishing factors.

First, the human central character was almost always maintained by pronominals within episode, and human non-central characters were very fre­quently referred to by NPs within episodes (see Tables 2 and 3) even if the central and non-central characters were of different gender.

Table 2: Distribution of Anaphors in Mandarin Recall Task

CONDITIONREFERENT

NPPRON

EVEN H U M -C H U M -N C105 109 310 43

ODDH U M -C H U M -N C127 122 257 40

TOTAL 415 152 84 162

% NP 25.30 71.71 33.07 75.31

% PRON 74.70 28.29 76.93 24.69

98 Ming-Ming Pu, Gary Prideaux, Lot's St an ft ' -¿odes as

Table 3: Distribution of Anaphors in English Recall Task

CONDITION

REFERENT

N PPRON

HUM -C96273

EVEN

H UM -N C102

69

ODD

H UM -C HUM -NC97 105 290 73

TOTAL 369 171 387 178

% NP 26.02 59.65 25.06 58.99

% PRON 73.88 40.35 74.94 41.01

HUM-C = Human central character HUM-NC = Human noncentral characte

The experimental results reveal that the more central a referent, the more is attended to, the longer it stays in focus, and consequently the more frequentl the inexplicit (or attenuated) anaphors, i.e., lexical and/or zero pronouns, ar used to code and identify it. For example, Subject 5 in the English Odd recall tas produced (1).

(1) The child went outside. Just as he went around the corner, he saw his mothe coming home with some groceries. So he stopped and asked his mother what wa inside the bag. The mother then gave him one o f the bags to carry and they pro ceeded on. He was following her. Then the curiosity o f the boy got the better o him because he stopped. He had to see what was in the bag. He opened up th bag and out jumped this big lobster and scared him. So he put the lobster bad in. Because he was scared, he was crying a little bit and picked up the bag, and fol lowed his mother back into the house.

Here, lexical and zero pronouns are consistently preferred for the centra: character, with full NPs used for the non-central character even if gender could come into play. The differential use of anaphors between the central and non­central characters thus resulted in more NPs (35% of all intra-episode NPs) than expected occurring within the episode.

The second major factor to emerge in the exceptions to the hypothesis is that more NPs were used within the first episode (in which Alex appears with an­other child) than within the other two, probably because:

• At the beginning of the total recall task, subjects usually established and identified the participants with more NPs than expected.

• In the first episode, both of the participants appear in each of the eight pictures and both take part in the activities together; subjects thus tended to weigh both characters equally for centrality.

• For the English subjects (10 out of 20) who regarded the two children as being of the same gender, NPs had to be used often to distinguish between

(3)

Episodes as Memory Units in Discourse: Anaphor Use in English & Mandarin 99

the two characters. For example, Subject 4 in the English Even recall condition produced (2).

(2) This appears to be three separate stories. The first story, ( ...) and in each of these stories, the story role is taken by a fat child with a pig-tail. In the first story, the child is playing tennis outside with another friend (u m ...), another little kid. And they were playing just outside a large house. The thinner child hit the ball towards the fat child, the fat child hit it and it went back toward the thin child. The thin child tried to get it, but missed. And the ball went into a window ...

Unable to differentiate between the two children in gender for the moment, the subject used full NPs to refer to both of the children throughout the retelling of the first episode. The first episode witnesses 30% of all intra-episode NPs.

The third factor to emerge was that some minor thematic discontinuity oc­curred within episodes: there are changes of scene, changes of participant, changes of perspective or point of view within each natural episode. Subjects tended to use NPs to signal these changes, i.e., to treat them as indicating sub­episodes in the story structure. For example, Subject 9 in the English Odd recall condition produced (3).

(3) ...T h en the final, the third event, the girl is out on the street and sees this woman coming toward her. And the woman might even be her mother carrying some groceries. And she stops and enquires,( ...) she stops the woman and asks her about what bags she’s carrying. ( ...) Well I guess she offers to carry one of the bags. And so the woman gives her a bag. / And then the little girl stops and lets the woman get farther ahead and around a corner. Then when the woman goes around the corner, she stops and she is curious about what’s in the bag. She opens it up and a lobster jumps out ...

Here in the third episode, when Alex stops at the comer of the street with the woman walking out of sight, there is a shift of focus from the interaction of both characters to the action of Alex alone. This signals a sub-boundary or thematic discontinuity (as is indicated by a in the above recall data), which the subject marked with an NP.

The above discussion of the counter-examples has shown that the differential use of anaphors by subjects during their narrative production not only reflects cognitive factors of attention and memory, but also directs the listener to iden­tify uniquely the correct referent, the episode boundary, the change of partici­pant and perspective, etc. Therefore, the counter-examples also appear to be 'rule-governed' since their occurrences depend on subjects' episodic organiza­tion of the story processing, the pragmatic status of the participants, the sup­posed needs of the listener, and the contextual and discourse information at hand.

100 Ming-Ming Pu, Gary Prideaux, Lois Stanfor

5 CO N CLU SIO N

The findings of the present experimental study demonstrate first that episodic organization of narrative production has psychological content: the story wa; : memorized as a series of episodes. Second, episodes exist as memory chunks: ai I episode is remembered as a whole, and tends to be forgotten as a whole. Third i episodes are shown to be dominated by macropropositions: the paragraph leve f theme such as 'the boy swats the fly' governs the lower level propositions which f are elaborations of the macroproposition. Fourth, the fact that the initial and § the final episodes tend to be better remembered than the middle ones is also the | manifestation of the effect of cognitive constraints on discourse production: ini- | tial information must form the foundation of the mental structure of the infor- f mation being processed and it has a privileged place in a language user's mental § representation (Gemsbacher 1989), while final information also has a privileged 1 place in mental representation because of short-term memory constraints. § Because of these privileged positions, initial and final information (or episodes I in our study) are more resistant to being suppressed and they are more strongly I enhanced and remembered. Finally, not only has the present hypothesis been supported, but in addition, it has been shown that the differential use of anaphors also has some important secondary functions in discourse production.

The present study further demonstrates that while stories and texts may be * presented or produced in a linear fashion, they are nevertheless formulated and processed hierarchically as constrained by the cognitive processes of memory and attention. In this process, the episode serves as a basic unit in production as ' well as in comprehension. The different episodes of a discourse are subsumed under different macropropositions, which are properly connected by various connective devices (e.g. connectives, conjunctions, adverbs, e tc .) , and which are I coherent to the overall macrostructure of the discourse. Nevertheless, the dis­continuity between episodes is also marked in discourse processing. These re­sults also strongly suggest that the status of the episode as a basic unit in dis­course processing has some universal characteristics, since the results are so consistent across two vastly different languages. In general, the correlation \ between discourse organization and anaphoric patterning is crucial if we are to ; extend our understanding of the relationship between language and cognition. !

REFERENCES

BLACK, J.B. & G.H. BOWER. 1979. Episodes as chunks in narrative memory. Journal o f Verbal Learning and Verbal Behavior 18: 309-31.

CHAFE, W. 1987. Cognitive constraints on information flow. In R.S. Tomlin (ed.), Coherence and Grounding in D iscourse. A m sterdam : John Benjamins, 21-51.

VAN DIJK, T & W. KlNTSCH. 1978. Cognitive psychology and discourse: recalling and summarizing stories. In W.U. Dressier (ed.), Current Trends in Text Linguistics. Berlin and New York: de Gruyter, 61-80.

VAN DIJK, T. 1982. Episodes as units of discourse analysis. In D. Tannen (ed.), Analyzing discourse: Text and Talk. W ashington D.C.: Georgetown University Press, 177-95.

VAN DIJK, T. & W. KlNTSCH. 1983. Strategies in Discourse C om prehension . New York: Academic Press.

FOX, B.A. 1987a. Discourse Structure and Anaphora. Cambridge: Cambridge University Press.

FOX, B.A. 1987b. Anaphora in popular written English narratives. In R. S. Tomlin (ed.), Typological Studies in Language: Coherence and G rounding in Discourse. Amsterdam: John Benjamins, 121-67.

GERNSBACHER, M.A. 1989. Mechanisms that improve referential access. Cognition , 32: 99-156.

GlVÓN, T. (ed.) 1983. Topic Continuity in Discourse: Quantitative Cross-lan­guage Studies. Amsterdam: John Benjamins.

GlVÓN, T. 1983a. Topic continuity and word order pragmatics in Ute. In T. Givón (ed.), Topic Continuity in Discourse: Quantitative Cross-language Studies. Amsterdam: John Benjamins, 343-63.

GlVÓN, T. 1983b. Topic continuity in spoken English. In T. Givón (ed.), Topic C ontinu ity in D iscourse: Q uantitative C ross-lan gu age Studies. Amsterdam: John Benjamins, 41-58.

GUINDON, R. & W. KlNTSCH. 1982. Priming Macrostructures. Technical report, University of Colorado: Colorado.

H a b e r la n d t , K., C. B eri AN, & J. SANDSON. 1980. The episode schema in story processing. Journal o f Verbal Learning and Verbal Behavior 19: 635-51.

H in d s , J. 1979. O rganizational patterns in discourse. In T. Givón (ed .),Discourse and Syntax 12. New York: Academic Press, 135-57.

KRAHN, F. 1981. Here comes Alex Pumpernickel! Boston and Toronto: Little, Brown & Company.

Episodes as Memory Units in Discourse: Anaphor Use in English & Mandarin__________ 101

102 Ming-Ming Pu, Gary Prideaux, Lois Stanfoi

KlNTSCH, W. 1977. On comprehending stories. In M.A. Just and P. Carpente (eds.), Cognitive Processes in Comprehension. H illsdale, New JersevErlbaum, 33-62. First Langu

t o a m . C aiLi, C.N. & THOM PSON, S. 1979. Third person pronouns and zero-pronouns ii j.521-34916

Chinese discourse. In T. Givón (ed.), Discourse and Syntax 12. New YorkAcademic Press, 311-35. Ingram 1

LONGACRE, R.E. 1979. The paragraph as a grammatical unit. In T. Givór overview c(ed .),Discourse and Syntax 12. New York: Academic Press, 115-34.. phonology,

MANDLER, J. & N . JOHNSON. 1 9 7 7 . Remembrance of things parsed: stor) more b aiastructure and recall. Cognitive Psychology, 9 : 1 1 1 - 9 1 . methodolo

MARSLEN-WILSON, W., LEVY, E. & TYLER, L. 1 9 8 2 . Producing interpretable dis- developmecourse: the establishment and maintenance of reference. In R.J. Jarvella & considers tW. Klein (eds.), Speech, Place, and Action. Chichester: John Wiley and Sons utterances,Ltd., 3 3 9 - 7 8 . period. Thi

PRIDEAUX G.D. & W.J. BAKER. 1986. Strategies and Structures: Current Issues the end ofin Linguistic Theory 46. Amsterdam: John Benjamins. that h a v e

SCHANK, R.C. & ABELSON, R.P. 1977. Scripts, Plans, Goals and Understanding. considerabHillsdale, NJ: Erlbaum. Ingram

TO M LIN , R.S. 1987. Linguistic reflections of cognitive events. In R. S. Tomlin single-woi(ed.), C oherence and Grounding in D iscourse. A m sterdam : John complex ;Benjamins, 455-80. literature.

TOMLIN, R.S. & M.M. Pu. 1991. The management of reference in Mandarin dis- P onecourse. Cognitive Linguistics 2,1: 65-93. d evelop m

developm account f proposed them, and the erner^ outlined ij innate u n gram m ar

Ingrair °f langua; language account t d eterm in mentionei acquisitic ftke char,

BOOK REVIEWS 0 CRITIQUES DE LIVRES

First Language Acquisition: Method, Description, and Explanation, by David Ingram. Cambridge: Cambridge University Press, 1989, 572 pages. $24.95. ISBN 0-521-34916-8 (pbk).

Ingram has taken on a formidable task in this book; namely, to present an overview of the research on child language acquisition in the four areas of phonology, morphology, syntax and semantics. His intention is to provide a more balanced view of the theoretical assumptions that influence the methodology and the descriptive approaches to research on early language development. He restricts the descriptive review of the literature to what he considers to be representative studies in prelinguistic development, single-word utterances, first word combinations, and the simple and complex sentence period. Thus the review is limited to research which includes studies up to about the end of the child's third year. However, given the proliferation of publications that have appeared in the past several decades, this still constitutes a considerable task.

Ingram argues that the three linguistic periods of language acquisition - the single-word utterance, first word combinations, and the period of simple and complex sentences - each have their own theoretical orientation in the literature. For example, the period of single-word utterances subsumes a theory of phonemic perception and production and a theory of phonological development, as well as a theory of lexical development. Similarly, the development of inflectional morphology requires a theory of morphology to account for the data. He deals with the various theories that have been proposed for each period, discusses the data that have been offered in support of them, and sorts out the issues and implications that follow. In his discussion of the emergence of grammar, Ingram adopts the general approach of nativism outlined in Chomsky (1965,1968,1975,1981) and elsewhere which posits a set of innate universal principles that restrict the choice of what constitutes a possible grammar of any language.

Ingram is careful to distinguish the term 'theory of acquisition' from 'theory of language' or 'grammar.' As he points out, a theory of language assumes that language is leamable given the usual input requirement, but it does not take into account the stages that characterize acquisition. Rather, it concerns itself with determining the set of principles or Universal Grammar (UG) that, as mentioned above, are part of the child's genetic endowment. Chomsky describes acquisition as being 'instantaneous', but this reference is directed at the adult­like characteristics of the grammar of the child once the parameters of the

104 Book Reviews / Critiques de livr isI Lan,

language have been fixed. Ingram, on the other hand, is concerned with 'theory of acquisition as a set of principles, distinct from those of UG, th< account for the stages the child goes through to reach the adult gramma (Ingram 1989: 64). That is to say, he focuses on the process of acquisition the leads to the finite or adult-like state. He contends that it is possible to test th principled changes that the child demonstrates at each stage and his purpose i to link acquisition data with theory in order to provide an explanation for th language learning process.

Two m ajor theories of language developm ent, the nativist and th behaviourist, are constrasted by Ingram. The latter view s learning a increm ental: habits are established slow ly over time and the child' reconstructing of the grammar during this process is held to be highly restrictec and subject to environmental conditions. The inadequacies of behaviourism a an account of language learning have been outlined, most notably in Chomsky ii his review of Skinner (1959). Ingram also points out that behaviouristic theory i clearly unable to account for this development at the period of the third year o life in a child when simple and complex sentences are being produced. Assuming the nativist fram ew ork, Ingram considers two possible positions: th< maturationist and the constructionist. They differ, however, on the question o when the principles of UG become available to the child. Maturationists likt Chomsky argue that the principles become available to the child at somt genetically determined time. Constructionism on the other hand, an approach whose framework is based on the work of Piaget (1971), does not attribute changes in linguistic behaviour to maturation but accounts for all changes by the building up of structure. What occurs earlier is built upon rather than replaced b) what comes later; the focus is on determining the sequential stages oj

development.The maturationist approach is similar to constructivism in that it also allows

restructuring but, as Ingram points out, it is more problematical as it predicts two extreme possibilities. One of these, the 'strong inclusion hypothesis', states that UG appears very early so that restructuring is virtually non-existent. In this case all changes would be due to performance and not competence. The 'weak inclusion hypothesis' allows restructuring under two possible conditions: a late occurring condition, because the input the child needs in order to trigger the appropriate principle are not heard until late in the process of language development. The other possibility is that, input aside, some principles do not mature until later. As Ingram notes, child language data would be irrelevant under the maturationist view since any changes would not be attributable to the nature of the child's earlier grammar, but would be due to other factors such as

iew in p u t o r shat the trig g

Furtherm oiU s , h e a d o j ¡or tw o m ain acquisition da 5 expanded ti

But it m a ] certainly det< crucial, a n d required. W accomplishec

In g ra m b description < perception, concludes tl studies th at 1 first lin g u ist reviews stuc acquisition c the single-w system b e g i analysis o f com m only f environm en children w h what h as be comprehend m yriad of d<

The secoi are p ro d u ce in clu d es th B1o o m , Bovs ^ th e l i te r h g r a m re vsyntacticallir° le of la n g

briefly G°vemn\er recen t m o c

f

¡¡vr First Language Acquisition: Method, Description & Explanation 0 D. Ingram___________105

jth a new input or maturation. This view leaves open the question of determining (haj what the triggering experience might be and when principles mature,

jmj» Furthermore, according to Ingram, the maturationist account is not testable.Thus, he adopts the constructionist approach which he maintains is preferable

^ for two main reasons. It does make predictions that can be tested, and it allows ie jg acquisition data to play a role in linguistic theory. That is, the scope of the theory

is expanded to include generalizations about the nature of linguistic competence.But it may not be the case that the maturationist approach is untestable;

certainly determining the nature of what constitutes 'old' from 'new' input is as crucial, and a more explicit definition of what is meant by maturation is

[¿'s required. We need to know what is maturing and how that process is ,te(j accomplished before the approach can be properly evaluated.

Ingram begins his discussion of child language development with a description of the prelinguistic period. He covers the areas of infant speech perception, infant speech production and early cognitive development and

0jt concludes the chapter with a review of infant-adult interaction. The major studies that have been carried out in these areas are presented in detail. For the first linguistic period, which is characterized by the single-word utterance, he

Qf reviews studies on early word comprehension and production as well as the acquisition of semantic categories and the use of overextension strategies during

ne the single-word period. He also looks at the child's emerging phonological ^ system beginning with discussion on perceptual development followed by an [{e analysis of the phonological characteristics of the first 50 words that are ^ commonly first produced. He concludes this chapter by examining the linguistic

environment; that is, the kind of linguistic interaction that is elicited by young 0[ children when they start to produce their first words. This involves research on

what has been commonly referred to as 'motherese' in the literature. There is a comprehensive review of the research of the single-word utterance period; a myriad of detail is clearly and concisely presented.

The second period of linguistic development in which first word combinations are produced was the focus of extensive research in the 1960s and early 1970s. It includes the longitudinal studies carried out by Brown and his colleagues, Bloom, Bowerman and Braine. These studies are the ones most frequently cited in the literature and have been the source of many subsequent reanalyses. Ingram reviews the proposals on pivot grammar, semantically-based versus syntactically-based grammars, and the functionalist model that emphasizes the role of language processing in determining the form that language will take. He also briefly mentions the parameter-setting model which is incorporated into the Government Binding framework, but does not go into detail as it is a relatively recent model. This model has been adopted in much of current acquisition

as'in'is

ke

106 Book Reviews / Critiques de liv s

research. One of the first applications of this theory was Flynn (1984), wi> looked at the second language acquisition of anaphora. An interesting aspect f the Government Binding framework for developmental studies is that it may, i fact, provide an explanation for the transition from the single word stage to t. j multiword stage - a transition that has not been satisfactorily dealt with in ti t literature. The more elaborate underlying categorial structure that is posited f ■ the grammar has put a new slant on what is to be acquired and this has wid • im plications for both linguistic theory and language acquisition. See, ft* example, the development of functional categories in early child language t Radford (1987, 1988) and Guilfoyle and Noonan (1988).

A wealth of detail on the phonological acquisition of single morpheme including both aspects of phonemic perception and phonological production presented in this book. Ingram also reviews the literature on the development < word meaning from a child's second year to approximately the fourth year an briefly examines the acquisition of some syntactic structures such as passive relative clauses and pronominal reference.

This book contains an in-depth analysis of many of the major issues in th complex area of inquiry that has involved a vast number of people from man disciplines. Ingram covers the relevant research on the four major languag areas and summarizes many of the studies in chart forms. He also includes sonn research that is not commonly found in texts published in English, as illustrate in his exem plary treatm ent of the phonological aspects of early chil development. At the end of each chapter, suggestions for further readings ar offered which will be very useful for upper-level undergraduate and graduat students. This book will also be of interest to other researchers in the field as source of reference for many years to come.

REFEREN CES

CHOMSKY, N. 1959. Review of Verbal Behaviour by B.F. Skinner. Language 3Z 26-58.

CHOM SKY, N. 1965. Aspects o f the Theory o f Syntax. Cam bridge, Mass Cambridge University Press.

CH O M SKY, N. 1968. Language and M ind. New York: H arcourt Bract Jovanovich.

CHOMSKY, N. 1975. Reflections on Language. New York: Pantheon.CHOMSKY, N. 1981. Lectures on Government and Binding: the Pisa Lectures

Dordrecht: Foris.

First Language Acquisition: Method, Description & Explanation 0 D. Ingram 107

FLYNN, S. 1984. A universal in L2 acquisition based on a PBD typology. In F.R. Eckman, L.N. Bell and D. Nelson (eds.), Universals o f Second Language Acquisition, Rowley, Mass.: Newbury House, 75 - 87.

GUILFOYLE, E. & NOONAN, M. 1988. Functional categories and language acquisition. Paper presented to the 13th Annual Boston University Conference on Language Development.

PIAGET, J. 1971. Biology and Knowledge. Chicago, 111.: Chicago University Press.RADFORD, A. 1987. The acquisition of the complementiser system. Research

Papers in Linguistics 2, 55-76, University College of North Wales, Bangor.RADFORD, A. 1988. Small children's small clauses. Transactions o f the

Philological Society 86, 1: 1-43.

Irene Mazurkewich Memorial University of Newfoundland

— $ —

The French Language Today, by Adrian Battye & Marie-Anne Hintze. Routledge: London and New York, 1992. Pp x + 374. $24.95. ISBN 0-415-07814-8 (pbk).

Teachers of French language and linguistics are always searching for that elusive text that will present the fundamentals of French phonology and mor­phology without straying to the limits of dinosaurian descriptivism on the one hand or incomprehensible gobbledygook on the other. In this book, the British authors have put together five chapters, as follows:

1. External History of the French Language (1-65)2. The Sound System of French (57-146)3. French Word Structure (147-202)4. The Sentence Structure of French (203-296)5. Varieties of French (297-356)

Of these the first chapter, while it contains a valuable range of information, may have to be omitted if the course is to be completed in one semester, or may be left to the students to read on their own. In either case they should be warned that there are all kinds of inaccuracies; just to take the Canadian data on page 5, for example, we are told that the British North America Act dates from 1887 (twenty years late!), that Bill 101 made French the 'sole official language' of Quebec (should be working language, since English is still official, guaranteed by the BNA Act), and that the Meech Lake Accord was rejected 'by the provinces of Manitoba and New Brunswick', when history will also lay the blame on the in­transigence of Clyde Wells of Newfoundland.

108 Book Reviews / Critiques de livn

The second chapter on the sound system also has serious problems of inacct racy. The nasal consonants are described as stops (81), for example, because c the buccal closure. But these sounds are not released through the nose; the ai passage is not stopped, and the resulting sound is a continuant. They should b classed as sonorants (the word is not to be found) along with [1] and [r], conso nants that are missing from the classification table on p. 75. We are also told tha the apico-alveolar [r] is heard in Quebec (82), whereas the normal pronunciatioi in the Quebec City region has always been uvular, and today the rolled [r] of th< Montreal region is swiftly disappearing, replaced by the uvular approximant [k]

There is also a lack of sufficient generalisation, so that the student is giver details to learn without being aware of their coherence. There is for example i low level phonetic rule whereby vowel lengthening takes place in stressed sylla bles closed by a voiced fricative, which includes [r]. The authors simply list tht consonants concerned without any indication that these sounds form a class.

The whole question of syllable structure is dealt with in a way that takes nc account of important work done in the last twenty years. And the related ques­tion of gliding, where for example the vowel /u/ can be reduced phonetically tc [w] before another vowel, so that /lue/ (louer) may be reduced in speech to a monosyllable [lwe], is dealt with as a mass of confused detail. How can one deal with such variation and never mention the word allophone, which never ap­pears once in the whole book? In fact everything is phonetics, put in square brackets, and the word phonology is also conspicuous by its absence.

This chapter too has its inaccuracies, since the authors claim (129) 'there are no examples of words whose meaning is distinguished by the alternation of [u] and [w].' On the same page they give both trois [tRwa] and trouer [tRue]; the preterit of this latter, troua [tRua], forms of course a minimal pair with trois, contradicting the authors' claim. This pair is in fact discussed at length in Grundstrom 1983, which although it was published nine years earlier, does not appear in the authors' bibliography, and references in the text tend rather to­wards such items as Schane 1968 and Martinet 1969.

There are unnecessary confusions in the following chapter on morphology as well. The morpheme is defined (following Lyons 1968) as 'the minimum unit of grammatical analysis' (p. 150), which leaves a few unanswered questions. Then in a diagram at the bottom of the page we are informed that the gender mor­pheme is marked by -e-, and the plural morpheme marked by -s. Quite apart from the problem of dealing with marking at the level of the orthography (we have a whole page (158) to tell us why you don't make triste feminine or gris plu­ral), one would understand from this that morphemes do not mark meanings such as plural but are (?) the meanings.

The French Language Today 0 A. Battye & M-A. Hintze 109

The paradigms taken from Martinet in this chapter are quite useful, but the authors' introduction that precedes it has its problems. Their paradigms, which they claim are given 'in full' (166) consistently leave out the conditional, without comment, although it is mentioned in the list of abbreviations, and turns up, of course, in the paradigms from Martinet. Even worse, the form es composées are reduced to three (167): ai donné, avais donné, aurai donné, and there is no men­tion of the formes surcomposées, which are not introduced for another hundred pages (280) and even then in a less usual form: J'ai eu fait cela. The paradigmatic contrasts between tense, mood, and aspect, which may be found in most school grammars in France are totally ignored, and the only mention of aspect concerns the imperfect (294), which paradigmatically appears as part of the tense system, as the following paradigm (from Curat 1991: 53) shows.

Table 1

SIMPLE COMPOSÉ SURCOMPOSÉ

Infinitive aimer avoir aimé avoir eu aiméPres. Participle aimant ayant aimé ayant eu aiméPast Participle aimé (eu aimé)

Subjunctive 1 aime aie aimé aie eu aiméSubjunctive 2 aimasse eusse aime eusse eu aime

Present aime ai aimé ai eu aiméImperfect aimais avais aimé avais eu aiméPassé Simple aimai eus aimé eus eu aiméFuture aimerai aurai aimé aurai eu aimeConditional aimerais aurais aimé aurais eu aimé

This layout contrasts moods (vertical blocks), tenses (vertical sets) and aspects (horizontal), and shows how aspect typically affects every tense form, as is the norm for aspect systems.

Unfortunately in this text there is no sense of system, everything is atomistic, and when we come to examime syntax and tense usage in Chapter Four, we are greeted with such tenses as (282):

passé récent: nous venons de donnerfutur proche: nous allons donner

as standard forms in the tense system. If you do this kind of thing, of course, there is no sensible limit to the number of constructions one can describe as tenses. And if we expect these forms to behave as regular tense forms, we should be able to derive nous sommes venus de donner, nous avons été venus de don-

110 Book Reviews / Critiques de livn

ner, nous sommes allés donner, nous avons été allés donner, nonsense form which demonstrate the incoherence of this approach, and lack of any rigour tha might be described as linguistic, let alone scientific. To procédé in a coheren fashion, we let the evidence speak for itself, as in the paradigm from Curat.

There are some good sections in the chapter on syntax, however, and som< useful structures and notions introduced. But there does not seem to be a ven good sense of judgment as to what should be introduced at this level, since on p 287 we have the verbs of motion introduced, and are told that they use the auxil iary être in their compound forms: this is studied in Junior High School. The sec tion on the subjunctive is also very brief, dealing only with the generalities to bt found in any standard grammar, and on p. 293 we read '...w hether the subjunc­tive can be viewed as a meaningful verbal category or simply as an extra market of subordination in Modem French is a distinctively difficult question to decide.' The meaningfulness of the subjunctive is in fact not in question, since there exist clear cut minimal pairs with meaningful contrast, as one would hope all uni­versity students of French are taught:

Je comprends qu'il a acheté une auto. I understand he bought a car.Je comprends qu'il ait acheté une auto. I understand why he bought a car.Je suis d'avis qu'il a une auto. I agree that he has a car.Je suis d'avis qu'il ait une auto. I agree with his having a car.

The final chapter on varieties of French also has much interesting and valu­able information, but is spoiled by several weaknesses, the major one being the failure to distinguish between social and regional levels of usage. The whole section on the morphology of Canadian French on p.314, for example, lists items that are français populaire, not Canadian French, and are heard in France as well as in Canada. If we were to take this section at its face value, we would end up believing that no one speaking French in Canada ever uses ne with negative forms of the verb! Also on Canadian French, there is never any mention of the regional difference between Acadian and Québécois, and on p. 315 a passage from La Sagouine is quoted to illustrate Canadian French! T'as qu'à ouèrel

The bibliography is brief and not particularly rich on French linguistic au­thors: no Damourette et Pichon, whose seven volumes should be known to every student of French, no Imbs, no Guillaume, no Benveniste, no Pottier, no Moignet, no Culioli, nothing from Fuchs, Gaatone, Joly, Klinkenberg, Martin, Wilmet, and others who have contributed so much to the description and discus­sion of the linguistic categories of French in recent years.

The book is described on the flyleaf as 'an introduction to the techniques of linguistics as applied to the French language and a reference work for the more advanced student'. I cannot imagine a reference work with a bibliography of this

The French Language Today 0 A. Battye & M-A. Hintze m

kind, and the application of linguistic techniques is equally lacking in substance. A mountain of work would have to be done on a rewrite before it could be recom­mended for use as a university text. The authors should start by reading Grundstrom, a text in the same market, but not in their bibliography.

Finally this is a difficult book to read because it is difficult to keep it open. The binding is made of very rigid plastic, which requires both hands to be constantly occupied. Readers should let the publishers know that this kind of binding may be the latest in technology and may be inexpensive, but is certainly neither recom- mendable nor acceptable.

REFERENCES

CURAT, HERVÉ. 1991. Morphologie verbale et référence temporelle en français moderne. Genève-Paris: Droz.

GRUNDSTROM, A l l a n W. 1983. L'analyse du français. Lanham, Maryland, London, New York: University Press of America,.

John Hewson Memorial University of Newfoundland

The foli tended iencyin manusc• Fora

with (or u noti in s ir with form

* Punc quoti tuatii quoti

' Biblic ses, chapt indici

Da w E/U r

Fu k e i d ,NcÉdsoc

K i n g , r K. Pr< E d

U b o v ,of

LINGUISTIC/-\TLANTICA H L -XX

STYLE GUIDE

The following few guidelines are in­tended to ensure a degree of consis­tency in the format of submitted manuscripts:

• Forms cited as linguistic examples within the text appear in italics (or underscored if italicization is not possible). Glosses, if any, are in single quotation marks (‘ and ’), with no comma separating cited forms from glosses.

• Punctuation marks always follow quotation marks unless the punc­tuation is an integral part of the quoted matter.

• Bibliographical references for the­ses, journal articles and book chapters should follow the format indicated below.

Guide Stylistiq u e

Pour assurer une certaine uniformité dans les manuscrits, les auteurs sont priés de bien vouloir suivre les quelques normes suivantes:

• Les formes citées comme exemples linguistiques dans les textes sont en italique (ou soulignées si l'ita­lique n'est pas possible). Les tra­ductions des formes citées parais­sent entre guillem ets simples (‘ et ’), sans virgule entre formes citées et traduction.

• Les signes de ponctuation suivent les guillemets, sauf si ceux-là font partie intégrante des form es citées.

• Les références bibliographiques pour une thèse, un article dans une revue savante ou un chapitre dans un livre devraient être conformes au format ci-dessous.

REFERENCES / BIBLIOGRAPHIE

D a w e , AUDREY. 1986. Fundamentals o f Micmac H istorical M orphology. Unpublished M.A. thesis, Memorial University of Newfoundland.

FLIKEID, KARIN. 1988. Recherches sociolinguistiques sur les parlers acadiens du Nouveau-Brunswick et de la Nouvelle Écosse. In Raymond Mougeon & Édouard Beniak (eds.), Le français canadien parlé hors Québec: aperçu sociolinguistique. Québec: Presses de l'Université Laval, 183-200.

KING, R u th & ROBERT R y an . 1988. The PEI project: a progress report. In T. K. Pratt (ed.), Proceedings o f the Twelfth Annual Meeting o f the Atlantic Provinces Linguistic Association. Charlottetown: University of Prince Edward Island, 44-52.

L a b o v , W i l l i a m . 1990. The intersection of sex and social class in the course of linguistic change. Language Variation and Change 2, 2: 205-251.

CONTENTS / SOMM AIRE

Articles

GISÈLE CORBEIL The Role of Information Construction in SecondLanguage Learning

MAURICE HOLDER ANNE MACIES ROLF TURNER

La diphthongue «oi» dans le parler «brayon» d 'Edm undston, N ouveau-Brunsw ick 1 7

CHENG LUO On the Uncleftability of Direct Object in Chinese 55

VIRGINIA MOTAPANYANE Romance Clitics and the Structure of AGRP 7 3

MING-MING PU GARY PRIDEAUX LOIS STANFORD

Episodes as Memory Units in Discourse Representation: Anaphor Use in English and M andarin 8 9

Book Reviews Critiques de livres

IRENE MAZURKEWICH First Language Acquisition: Method, Description, and Explanation 0 David Ingram 1 0 3

JOHN HEWSON The French Language Today 0 A drian Battye & M arie-Anne Hintze 1 0 7

L I N G U I S T I C A a t l a n t i c a 14 ( 1992)

" l U N