51
Language Learning 43:4, December 1993, pp. 507-557 Processing Strategies in L2 Learners of French: The Role of Transfer L. Kathy Heilenman University of Iowa Janet L. McDonald Louisiana State University This study compared the comprehension processingstrat- egies of 15 monolingual English native speakers and 8 bilingual French native speakers to 112 secondlanguage (L2) learners of French, using stimuli containing word order and clitic pronoun (type and agreement) cues in French. Results indicated differential dependence on cue use by the two native speaker groups, with English native speakers depend- ingmore on word order for interpretation and French native speaker depending more on clitic pronoun agreement. Inter- pretations produced by L2 learners of French indicated an immediate abandonment of L1 word order strategies with a much later onset of clitic pronoun agreement strategies. Results are discussed within the framework of the Competi- tion Model. An earlier version of a portion of this paper was presented at the Second Conference on Second Language Acquisition and Foreign Language Learning held at the University of Illinois at UrbandChampaign in April 1990. Our thanks to Catherine Jolivet and Fatima Chajia for recording the French stimuli. Thanks also to two anonymous reviewers for their comments on an earlier version of'this paper. Requests for reprints may be sent to L. Kathy Heilenman, Department of French and Italian, The University of Iowa, 31 SchaefferHall, Iowa City, Iowa 52242. Telephone: (319) 335-2253. E-mail: L-HeilenmanQuiowa.edu 507

Processing Strategies in L2 Learners of French: The Role of Transfer

Embed Size (px)

Citation preview

  • Language Learning 43:4, December 1993, pp. 507-557

    Processing Strategies in L2 Learners of French: The Role of Transfer

    L. Kathy Heilenman University of Iowa

    Janet L. McDonald Louisiana State University

    This study compared the comprehension processingstrat- egies of 15 monolingual English native speakers and 8 bilingual French native speakers to 112 secondlanguage (L2) learners of French, using stimuli containing word order and clitic pronoun (type and agreement) cues in French. Results indicated differential dependence on cue use by the two native speaker groups, with English native speakers depend- ingmore on word order for interpretation and French native speaker depending more on clitic pronoun agreement. Inter- pretations produced by L2 learners of French indicated an immediate abandonment of L1 word order strategies with a much later onset of clitic pronoun agreement strategies. Results are discussed within the framework of the Competi- tion Model.

    An earlier version of a portion of this paper was presented at the Second Conference on Second Language Acquisition and Foreign Language Learning held at the University of Illinois at UrbandChampaign in April 1990. Our thanks to Catherine Jolivet and Fatima Chajia for recording the French stimuli. Thanks also to two anonymous reviewers for their comments on an earlier version of'this paper.

    Requests for reprints may be sent to L. Kathy Heilenman, Department of French and Italian, The University of Iowa, 31 Schaeffer Hall, Iowa City, Iowa 52242. Telephone: (319) 335-2253. E-mail: L-HeilenmanQuiowa.edu

    507

  • 508 Language Learning Vol. 43, No. 4

    INTRODUCTION

    The majority of research concerning transfer in second lan- guage learning has focused on evidence provided by production (e.g., Gass & Selinker, 1983; Odlin, 1989), with relatively few studies investigating transfer as in either listening or reading (e.g., Glisan, 1985; Durgunoglu & Hancin, 1992). This has un- doubtedly been due t o the fact that, from the point of view of second language teaching and research, the most obvious instances of transfer or interference have been those produced by learners either in speaking or writing. Transfer phenomena in listening and reading have not only been less salient; they have also been less likely to be perceived as language learning problems. As Lado (1957) pointed out, the effects of native language transfer in production and reception are far from identical:

    Whenhe[theL2learnerlspeakshechooseshismeaningsand then produces the forms that in the native language would signal those meanings. In listening, he hears the forms and attaches the meanings they would have in his native lan- wage. (Lado, 1957, p. 59)

    Thus, although transfer could logically occur in both production and comprehension, only in production would it be immediately obvious. Less obvious, of course, does not mean less important. From a theoretical point of view, it is important that transfer phenomena be investigated not only across all linguistic sub- systems but also across all modalities. From a more practical point of view, information concerning possible listening or syntactic processing accents (MacWhinney, 1992, p. 381) may provide suggestions toward improving second language instruction.

    Recent work on crosslinguistic sentence interpretation has provided a new impetus to the study of transfer in the receptive skills. Here, the Competition Model (Bates & MacWhinney, 1981, 1982, 1987, 19891, a psycholinguistic model of crosslinguistic processing, has provided a methodological as well as an interpre- tative framework for examining transfer phenomena across languages. Work within the Competition Model has established

  • Heilenman and McDonald 509

    that native speakers of different languages rely on different cues in sentence interpretation. Research focussing on bilinguals has sought to examine whether bilinguals process their second lan- guage with strategies appropriate to their first o r to their second language. Before discussing these results, however, we turn t o a brief presentation of the Competition Model itself.

    THE COMPETITION MODEL1

    The Competition Model is a functionalist language-process- ing model based on the concepts of form-function mapping, probabilistic processing, and language acquisition as mediated by environmental factors. The description provided here is intended only as background for the present study. Interested readers may refer t o Bates & MacWhinney (1981, 1982, 1987, 19891, MacWhinney (1987a, 1987b, 1989a, 1989b, 1989~1, and references therein for further information.

    Form- function mappings. Essentially, the Competition Model views grammars as solutions to the problem of using linear forms (e.g. , word order, morphology, intonation, lexical marking) to mark nonlinear functions (e.g., actor, agent, topic, subject). This mapping between form and function is not one-to-one but many-to- many (Bates & MacWhinney, 1982, 1989; MacWhinney, 1987b). For example, in English, several cues (forms) typically coalesce around the functions of actorlagentkopidsubject. For instance, in Example 1, the cues of preverbal word order, noun animacy, and verb agreement all point t o the boy as the actdrlagentltopiclsubject.

    1. The boy overturns the bottles.

    The use of these cues to mark such functions, however, is proba- bilistic. In Example 2, only the cue of preverbal word order exclusively favors the wind as the actor/agent/topic/subject.

    2. The wind overturns the bottle.

    Further, functions are not obliged to co-occur, but may be split, as happens when the patient rather than the agent is topicalized

  • 510 La ngu age Lea r n ing Vol. 43, No. 4

    (Example 3). Here preverbal position and verb agreement mark the bottles as the topiclsubject, whereas noun animacy and the preposition by favor the boy as actorlagent:

    3. The bottles were overturned by the boy.

    Thus, the mapping between any particular form and any particu- lar function is probabilistic and not absolute.

    Cue-basedprocessing. Given the probabilistic nature ofform- function mapping, how do people use cues t o assign functions? According to the Competition Model, sentence interpretation oc- curs via the cooperation and competition of various cues, and the strength with which these cues influence the process depends on the total form-function mapping distributions of the language involved. That is, the more informative or stronger the mapping between a particular form and function, the more strongly that form will be used in assigning that function. According to the Competition Model, this cue strength, defined as the probability or weight that the organism attaches t o a given piece of informa- tion relative to some goal o r meaning with which it is associated (Bates & MacWhinney, 1987, p. 641, is a function of cue validity. Cue validity refers to the objective, quantifiable relationship between the cue and the function it marks over the universe of sequences within a particular task domain. For example, t o determine the validity of the cue of preverbal position in marking the actorlagent function in English, we would examine a random sampling of possible sentences in English t o see how often this cue is present or available t o mark the function. In addition, we would check t o see if, when present, it was reliable in indicating the correct word as actorlagent. The product of these two factors, availability and reliability, is the overall validity of the cue (McDonald, 1986). As it turns out, in English, finite verbs are almost always preceded by nouns or pronouns (imperative se- quences comprising an exception), yielding high availability for the preverbal position cue. In addition preverbal words are very frequently actordagents, thus according this cue high reliability. The product of this high availability and high reliability yields

  • Heilenman and McDonald 51 1

    high overall validity for the cue of preverbal position in marking the actodagent role in English.

    Overall cue validity thus calculated from text counts has been shown to be predictive of child language acquisition (McDonald, 1986; Bates & MacWhinney, 1989). To account for ad*& values, however, a refined definition of cue validity, known as conflict ualidity, has been necessary (McDonald, 1986, 1987b). Conflict validity measures the relative utility of a cue in marking a function when this cue conflicts or disagrees with other cues. In Example 4, preverbal position favors the ball as the actodagent, but noun animacy favors the boy:

    4. The ball hit the boy.

    That the ball is the entity performing the action of the verb indicates that, in the case of conflict, word order is a stronger, or more valid cue to actodagenthood than is noun animacy. The transition between beginning acquisition strengths as tied to overall validities, and adult usage as tied to conflict validities has been explained through a learning-on-error model (McDonald, 1986, 19891, which initially gives equal weight t o all types of sentences, but later gives greater weight t o sentences involving conflicts among cues, as these latter sentences cause more inter- pretation errors later in learning. Details of this model can be found elsewhere (McDonald, 1986,1989; McDonald & MacWhinney, 19911, the readers of this article should remember however that the model predicts modification of cue strengths based on errors.

    One additional concept, cue cost, has been invoked t o explain the course of language acquisition in Competition Model terms. Cue cost refers to the relative detectability and assignability of cues. Cues that are difficult t o perceive (e.g., unstressed particles) and/or difficult t o assign meaning to because much material must be held in memory (e.g., agreement phenomena spanning several sentence constituents), are likely t o be abandoned when the processing system is under stress. In addition, such cues probably represent relatively late acquisitions (Bates & MacWhinney, 1987).

  • 512 Language Learning Vol. 43, No. 4

    TRANSFER AND THE COMPETITION MODEL

    Given that the form-function mappings in various languages differ, the emphasis that native speakers of different languages give diferent cues in interpreting sentences also differs. For example, English native speakers (ENS2 assign the semantic/ syntactic role of actor (i.e., who or what performs the action of the verb) in English sentences almost exclusively according t o the information provided by word order, with the information carried by animacy and subject-verb agreement given only minor consid- eration (MacWhinney, Bates, & Kliegl, 1984). Adult native speakers of French, on the other hand, appear t o pay relatively little attention t o the information carried by word order and depend instead on animacy and agreement (subject-verb and pronoun) information for the interpretation of French sentences (Kail, 1989; McDonald and Heilenman, 1991, 1992). Such processing differ- ences across languages pose an interesting question for bilinguals, namely, the type of processing strategies bilinguals develop for their first and second languages.

    Transfer, o r the influence of the knowledge of one language on the way in which another is learned o r processed (Shirai, 1992), has been a productive source of inquiry for Competition Model studies. Given that questions formulated in Competition Model terms can be framed as concrete, testable hypotheses, sentence interpretation behavior by L2 speakers of varying proficiency levels can be compared with that seen in native speakers (Kilborn, 1989). Further, the Competition Model, unlike models that as- sume the existence of either/or deterministic rules, provides a theoretical framework within which the gradual emergence of L2 capacities can be studied (Kilborn & Ito, 1989). In other words, by asking L2 learners t o interpret sentences in their L2, and then comparing their performance to that of native speakers of their L1 as well as to that of native speakers of the L2, the developmental course of L2 form-function mappings can be charted and the influence of L1 strategies on L2 interpretation can be investigated. Four possible developmental courses, suggested by Kilborn & Ito

  • Heilenman and McDonald 513

    (19891, are: (a) L1 strategies will be used to interpret both L1 and L2; (b) Separate strategies will be developed, an L1 strategy for L1 and an L2 strategy for L2; (c) L2 strategies will be used to interpret both L2 and L1; and (d) An intermediate solution somewhere between L1 and L2 will be developed.

    Thus far, studies have investigated the relationship between L1 and L2 sentence interpretation for English-French (McDonald & Heilenman, 1991, 19921, Dutch-English and English-Dutch Wlborn & Cooreman, 1987; McDonald, 1986, 1987a), German- English and English-German (Bates & MacWhinney, 198 1; Kilborn, 1989), English-Italian and Italian-English (Bates & MacWhinney, 1981; Gass, 1987; English-Chinese (Miao, 1981), Japanese-En- glish and English-Japanese (Harrington, 1987; Kilborn & Ito, 1989; Sasaki, 19911, and Spanish-English bilinguals (Wulfeck, Juarez, Bates, & Kilborn, 1986). (See Kilborn & Ito, 1989, and MacWhinney, 1992, for reviews.) In general, these studies indi- cate that L2 speakers initially tend t o transfer L1 strategies to L2 sentence interpretation. With increasing exposure t o the L2, however, adaptation of strategies appropriate to the L2 may occur.3 Mathematical models constructed by McDonald and MacWhinney (1989) and McDonald and Heilenman (199 1) have provided evidence that cue strength a t various learning stages is predicted by the cue validity measures described above, in combi- nation with the learning-on-error model developed within the Competition Model. That is, as learners gain experience with their L2, they appear t o be using the distribution of cues in incorrectly interpreted L2 sequences t o reset their L2 cue strengths to values more appropriate to L2 processing.

    The study to be reported here investigated the course of development of cue use by Anglophone L2 learners of French with between roughly 40 and 350 hours of classroom exposure to French. Although the studies cited above have investigated cue use by L2 learners with varying degrees of proficiency, the major- ity have concentrated on fairly advanced L2 learners and few have investigated a cross section of learners whose exposure to the L2 has been in a guided, classroom setting. Because previous work

  • 514 Language Learning Vol. 43, No. 4

    concerning the assignment of the actor role during sentence interpretation has shown a strong reliance on word order by ENS (Bates, McNew, MacWhinney, Devescovi, & Smith, 1982; MacWhinney et al., 1984; Harrington, 1987; Kilborn, 1989) and a strong dependence on clitic pronouns byFNS(Kail,l989; McDonald, & Heilenman, 1991; Heilenman & McDonald, 19931, the current study examined the development of the use of word order versus (clitic) pronoun type and agreement cues in the interpretation of French sentences by L2 learners. Before turning t o the experi- ment itself, however, we present information regarding the mapping between these cues and the roles of actor/agent/subject in English and in French.

    CUES IN ENGLISH AND FRENCH

    ENGLISH

    Word order. Canonical word order for English is subject-verb- object(SVO), with other word ordersfoundin limited environments (MacWhinney et al., 1984; Quirk, Greenbaum, Leech, & Svartvik 1985). Although passive constructions (e.g., The car was washed by Jill) approximate OVS structure, the actual order is better represented as (grammatical) subject-verb-prepositional phrase in which the grammatical subject represents the semantic patient and the agent is marked with a by-phrase. From a Competition Model perspective, the English passive structure embodies the breakdown of the more usual subject-actor coalition.

    Other possible orders are limited to OSV (relative clauses, interrogatives, left-dislocations), and VOS (right-dislocations, imperatives) as shown below:

    0%: Relative clause Theres the car(that) Jill washed. Interrogative Left-dislocation My first boyfriend-him they

    Which car did Jill wash?

    hated. VOS: Right-dislocation Loved cars, my brother.

    Imperatives Wash the car, Jill.

  • Heilenrnan and McDonald 515

    Other possible word orders (SOV and VS04) are extremely rare (MacWhinney et al., 1984).

    As can be seen from the examples above, English permits both left and right dislocations, although these structures are not common and these dislocations conform to the limited word orders allowed in English. Such dislocations may involve movement of a noun phrase, frequently accompanied by contrastive stress (e.g., My first boyfriend they hated; Loved cars, my brother) or they may also involve therepetition ofthe object or subject, often in pronomi- nal form (e.g., My first boyfriend-him they hated; He loved cars, my brother).

    Pronoun type. English uses pronoun morphology t o mark the subjectlactor role, with subjectdactors tending t o be in the nomi- native case (e.g., I , she, they) rather than in the oblique case (e.g., me, her, them). Of the seven unique subject pronouns in English, five (I, he, she, we, they) are distinct from the object forms. All direct and indirect object forms are, however, indistinguishable from each other. Hudson (1990) has argued that case may be a misnomer for the distinction between subject and object pronouns. He, along with Quirk et al. (1985), contends that English pronouns have only one form, the unmarked, nonsubject one (e.g., me) with so-called subject forms being used only in front of tensed verbs where a clear subject relationship is present. Thus, the use of I or me, for example, may well reflect a distributional rather than a case difference.

    Pronoun agreement. Third person singular English pronouns aremarkedforgender(e.g., he,she,it). NounsinEnglish, however, are generally not gender marked; only a small number (e.g., mother, father, widow, widower, bride, groom) contain gender information as part of their meaning. When both pronouns and coreferential gender-marked nouns are present, gender agree- ment can be used to match pronounswithnouns in either dislocated sequences, or across discourse. For example, in the dislocated sequence She watched the boy, the mother, gender agreement between the feminine pronoun she and the NP, mother, allows the assignment of the actor role to mother.

  • 516 Language Learning Vol. 43, No. 4

    Cue use by ENS. Given the limited set of word orders available t o English speakers, there should be a strong tendency for such speakers t o assign SVO interpretations t o NVN se- quences, OSV interpretations t o "V sequences, and VOS interpretations t o V" sequences. These tendencies consist basically of two substrategies: (a) interpreting preverbal nouns as subjects/actors and (b) treating postverbal nouns as objectdpa- tients (MacWhinney et al., 1984). Ifthere is no preverbal noun, the latter strategy assigns the actodagent role t o the noun that is not immediately postverbal. Thus, English speakers should show a strong first-noun bias in assigning actorhood in NVN sequences and a strong second noun bias in NNVand V" sequences, exactly the pattern found in several studies (Bates, McNew, MacWhinney, Devescovi, & Smith, 1982; MacWhinney et al., 1984; Harrington, 1987; Kilborn, 1989).

    Further, because case inflection in English is found only on pronouns, whereas word order information is present for every sequence, stronger use of word order cues as compared to pronoun case inflection cues would be expected, a prediction borne out by McDonald's (198733) study. McDonald asked ENS to interpret NVN strings containing word order, animacy, and pronoun case cues. Results indicated virtually no use of pronoun case informa- tion, with ENS relying on word order instead. Thus, for adult ENS, the assignment of the actodagent role derives from the use of a strong word order strategy that totally dominates information derived from pronominal case marking.

    In addition, because most common nouns in English do not contain gender-identifying information, one might expect weaker use of gender agreement than of word order in sentence interpre- tation. The gender agreement cue may also be comparatively weaker in English than in a language in which all nouns are marked for gender.

    FRENCH

    Word order. Canonical word order in French is SVO, al-

  • Heilenman and McDonald 51 7

    though variation is permitted, especially with the use of case marked pronouns.

    NVN (SVO)

    (OVS) interrogative

    relative clause

    NNV (SOV) object clitic

    (OSV) interrogative

    relative clause

    left dislocation

    V" WSO) interrogative

    (VOS) imperative

    Le garcon a renverse les bou-

    The boy knocked over the bottles Qu'a renverse le garcon? What-obj. knocked over the boy? Les bouteilles qu'a renverskes le

    Thebottles that-obj. knockedover

    teilles.

    garcon htaient vides.

    the boy were empty.

    Le garCon les a renvershes. The boy them knocked over Qu'est-ce que le garcon a ren-

    What-obj. the boy knocked over? Les bouteilles que le garcon a

    renvershes Btaient vides. The bottles that-obj. the boy

    knocked over were empty Et un pain cuit tu peux me

    rapporteraussi?(Citedin Pohl, 1976.)

    And a roll cooked you are able to me to bring?

    verse?

    A-t-il renverse les bouteilles? Did he knock over the bottles? Regarde le chat, mon petit. Look at the cat, my little one.

    (See Le Bidois, 1968, and Pohl, 1976, for discussion of OSV constructions without resumptive clitic pronouns.)

    Two interrelated phenomena, clitic pronouns and disloca- tion, combine t o permit relatively free word order in French. In Example 5 , the left dislocated subject NP legarcon is copied by the subject clitic pronoun il. Similarly, the right dislocated NP, les bouteilles, is copied by the object clitic pronoun les.

  • 518 Language Learning Vol. 43, No. 4

    5. Le garGon, ill lesJ renverse les bouteillesJ. The boy he them overturns the bottles.

    Such sequences are typical of informal, spoken, and even written French, allowing variation in the service of pragmatic motivation (e.g., Ashby, 1982,1988; Calv&,1982; Barnes, 1985,1990; Trevise, 1986a; Lambrecht, 1987; de Fornel, 1988; Matthews, 1989).

    Clitic pronoun type.5 As was the case for English, French clitic pronouns, but not nouns, indicate nominative and accusative case. Unlike English, however, French has two types of clitic pronouns, unstressed o r bound clitic pronouns that are found preverbally (Example 6) and stressed o r tonic pronouns that occur elsewhere (Example 7).

    6. Le gaqon les renverse. The boy them overturns.=The boy overturns them.

    (Zes=unstressed clitic pronoun) 7. Cest toi qui as renvers6 les bouteilles?

    It is you who (nom.) overturned the bottles?=Are you the one who overturned the bottles? (toi=stressed, tonic pronoun)

    The series of unstressed pronouns is usually referred t o as clitic pronouns. Morphologically, clitic pronouns are closely tied to the verb and carry number and/or gender information. Syntactically, clitic pronouns satisfy subcategorization requirements of verbs while also being in complementary distribution with noun phrases (Borer, 1986). Thus, although English pronouns and French clitic pronouns appear similar, they differ in important ways, with the English pronouns behaving more like words and the French clitic pronouns behaving more like affixes (Taylor, 1989). For example, as Borer (1986) demonstrates, English pronouns allow conjunc- tion (much like words) whereas French clitic pronouns do not.

    8. John eats the apples and the cake.

    9. Jean mange les pommes et le gateau. John eats them and it.

    *Jean les et le mange.

    For further discussion see Klavans (19851, Ossipov (1990), and Zwicky and Pullum (1983).

  • Heilenman and McDonald 519

    Lambrecht (1987, 19881, among others (e.g., Sankoff, 1982; Klavans, 1985; Harris, 1988; Gadet, 1989; Matthews, 1989; Ossipov, 1990; Roberge, 19901, contends that clitic pronouns, especially subject clitics, function as preverbal number and gender markers. Thus, in a sequence such as Example 10 below (from Ossipov, 1990, idiomatic translation ours), the clitic pronounsje and les are said t o function as affix-like clitics indicating number and gender as well as the case or argument structure of the verb.

    10. Je, sais que moii, les professeurs,, jei les, ai trouvks bien smatts. I know thatme the teachers1 them have found well cool=

    you know for me, the teachers, I thought they were real cool.

    Thus, unlike English pronouns that seem to have largely lost their case-marking value, French clitic pronouns have gained in the amount of information they carry. Roberge (1986), for ex- ample, has suggested that French, which was once a null subject language like Latin, has slowly lost its verbal inflections, com- pensating with an increase in reliance on clitic pronouns as verbal affixes. These affixes, according to Roberge, serve a purpose similar t o that served by the rich subject-verb agreement system in Italian.

    Note further that clitic subject pronouns provide an almost complete one-to-one form-function coding. That is, of the nine subject clitic forms (including on), seven (je [I], tu lyou-singular familiar], il [he], on [one, we, they], elle [she], ils [they-masculine], elles [they- feminine]) code only the subject function whereas two (nous [we], uous lyou-plural, formal]) also code direct object, indirect object, and stressed pronominal functions. Direct object clitic pronouns are more complex. Two forms (nous and uous) are homophonous with subject clitic pronouns and stressed pronouns. Two other forms (me [(to) me], and te [(to) you-singular, familiar]) are homophonous with indirect object clitic pronouns, and the remaining three forms (le [the, it, him], la [the, it, her], les [the, them]) are homophonous with definite articles.

    Clitic pronoun agreement. Third person subject and direct

  • 520 Language Learning Vol. 43, No. 4

    object clitic pronouns are marked for gender in French (e.g., iZ(s), elle(s), le, la) except for the third person plural object clitic pronoun les, which is marked only for number. In addition, nouns in French are nearly always marked by the definite o r indefinite article for grammatical gender (Ze 1 un cahier [the-masc./a-masc. notebook], la l u n e photo, [the-fem./a-fem. photo]). Thus, because a clitic pronoun and its referent must agree in gender and number for a grammatical, sensible sequence to result, gender and number agree- ment can be used to help determine the coreferent of clitic pronouns.

    Cue use by FNS. Several studies have looked at the interpre- tation of various word orders by adult French native speakers (Kail & Charvillat, 1986; Kail, 1989; McDonald & Heilenman, 1992; Vion, 1992). In general, these studies indicate that FNS prefer to assign the first noun t o the actodagent role for NVN sequences (SVO), exhibit a somewhat weaker first-noun prefer- ence for NNV sequences (SOW, and either a neutral or second-noun preference for VNN sequences (VSO or VOS).

    Previous research (Kail & Charvillat, 1988; Kail, 1989; McDonald & Heilenman, 1991, 1992; Heilenman & McDonald, 1992) has shown, however, that FNS use information provided by clitic pronouns and verb agreement much more strongly than they do information coded by word order. This is not to say that FNS ignore word order in dislocated sequences; work on dislocated sequences with various word orders and clitic pronoun cases has yielded a highly significant interaction between the two factors (Heilenman & McDonald, 1993). This interaction was the result of FNS preference for interpreting these dislocated sequences so that an SVO or SOV subsequence6 would be present. This prefer- ence for an SVO or SOV syntactic core would cause FNS to interpret a subject clitic-V-N-N sequence as sVOS rather than sVS0, as the former contains a SVO syntactic core. Previous work has also shown that FNS pay significant attention t o gender agreement cues in &ding the cox efer en% of a cL6c pxono-. The use of this agreement cue easily defeats word order biases (Kail, 1989; McDonald & Heilenman, 1991; Heilenman & McDonald, 1993).

  • Heilenman and McDonald 52 1

    In comparing the relative strength of cue use of FNS to ENS, we would predict stronger use of word order by ENS, due t o the more rigid orders seen in English, and stronger use ofcase type and gender agreement by FNS due to the greater availability and reliability of these cues in French.

    BILINGUAL ENGLISH-FRENCH CUE USE

    Research with advanced English-French bilinguals has shown that these speakers are able to master successfully the cues of clitic pronoun agreement, verb agreement, noun animacy, and word order in French, and t o do so at levels that render them indistin- guishable from FNS (McDonald & Heilenman, 1991). Thus, very advanced second language speakers of French show no reliance on English strategies in interpreting French.

    The mastery of strategies appropriate t o French, however, does not come immediately. Research on ENS who are just beginning t o learn French demonstrates the time course of mas- tery for the cues of word order, verb agreement, and noun animacy (McDonald & Heilenman, 1992). In this study, although there was no difference between ENS and FNS in reliance on noun animacy, word order strategies were stronger in English, whereas reliance on verb agreement cues was stronger in French. Second-language learners of French almost immediately abandoned English word order preferences, losing second-noun preferences in V and VNN word orders, and showing a weakened first-noun preference in NVN strings. Appropriate French word order strategies, how- ever, particularly a second-noun preference on VNN strings did not develop until substantially later. A significant use of verb agreement, as is appropriate in French, did not appear even marginally in these second-language learners until after two or more years of classroom exposure. McDonald and Heilenman (1992) point t o the role of input in explaining the word order results, citing evidence from classroom teacher talk of word order variation in French that would have been misinterpreted using English strategies. The later onset of appropriate verb agreement

  • 522 Language Learning Vol. 43, No. 4

    strategies is attributed to the relative lack of such cues, given the widespread neutralization of the third-person singular versus plural distinction in spoken French. (E.g., il mange [he eats1 and ils mangent [they eat1 are pronounced identically).

    The current study also examined the time course of cue mastery in native English speakers who are classroom learners of French. The cues explored include word order, (clitic) pronoun type, and (clitic) pronoun agreement. Clitic pronoun types include subject clitic pronouns alone, object clitic pronouns alone, and subject and object clitic pronouns together. Based on previous research, we expected t o find a rapid abandonment of English second-noun word order strategies accompanied by a relatively late development of appropriate French word order strategies. Then, given that native speakers of English show relatively little dependence on pronoun case marking, along with the fact that the French clitic pronoun system is both complex and presents prob- lems of detectability (Bates & MacWhinney, 1987; Perdue, 19911, we also expected to see a relatively late development ofreliance on clitic pronoun type and gender agreement as cues t o meaning.

    METHOD

    PARTICIPANTS

    Four groups of French learners currently enrolled in univer- sity-level French courses participated. These included 31 students in first semester (FLl), 53 students in second semester (FL2), 19 students in third semester (FL3), and 9 students enrolled in or beyond fourth semester (FL4). Of this latter group, one was enrolled in a fourth-semester French course, the remainder in the third or fourth year of French. None of the French learners had had significant experience with languages other than English or French and none had been in a French-speaking country for more than a total ofthree weeks. Thus, nearly all their French exposure occurred in the classroom. In addition t o the four French learner

  • Heilenman and McDonald 523

    groups, two control groups, one of 8 bilingual FNS and one of 15 monolingual ENS were included. Although the FNS were currently living in the United States, all reported extensive daily use of French.

    STIMULI

    A vocabulary pool representing words likely t o be known t o all students7 was established. This pool consisted of 12 animate nouns (l'homme [man], le monsieur Igentlemanl, le fr6re [brother], le cow- boy [cowboyl, le garCon [boy], lepapa [father], la femme [woman], la dame [lady], la saeur [sisterl,laprincesse [princess], lafille Igirll, la maman [mother]) and three causative faire expressions (faire bouger [to make moue], faire tomber [to makefalll, faire tourner [to make turn]). Note that the nouns were chosen to contain gender identifying information in their meaning, s o use of gender agree- ment could also be seen in the English version of the stimuli given t o ENS.

    Stimuli were devised by completely crossing three levels of word order (NVN, NNV, and V"), three levels of clitic pronouns (subject clitic pronoun only, direct object clitic pronoun only, and both subject and direct object clitic pronouns), and three levels of clitic agreement (clitic pronoun[sl agreeing in gender with both nouns, clitic pronouns[sI pointing t o the fist noun as subject, and clitic pronounlsl pointing to the second noun as subject). Two exemplars of each sequence type were formed, yielding 54 test sequences. These sequences were combined with 18 additional filler sequences not containing clitic pronouns for a total of 72 sequences. Nouns in all sequences were animate and singular; verbs were all causatives of the form faire + infinitive to allow comparison with previous studies (Kail & Charvillat, 1986,1988; Kail, 1989; McDonald & Heilenman, 1991, 1992; Heilenman & McDonald, 1993). To guard against event probabilities, vocabu- lary items were assigned t o sequence slots randomly, with the restriction that the same noun could not occur twice in the same sequence. Clitic pronouns (subject and direct object) always preceded the verb as is normal in French. Thus, the actual word

  • 524 Language Learn ing Vol. 43, No. 4

    orders were NVN: Noun, (subject clitic and/or object clitic), verb, noun; NNV: noun, noun, (subject clitic and/or object clitic), verb; VNN: (subject clitic and/or object clitic), verb, noun, noun. The test sequences were randomized and a native speaker of French re- corded each sequence followed by a short pause. A set of four practice sequences preceded the main stimuli. An equivalent tape was made in English by a native English speaker. All sequences were spoken with a similar flat intonation contour. See Appendix 1 for sample stimuli.

    In experiments designed t o test strength of cue use within the Competition Model paradigm, it is necessary t o completely cross all levels of a cue with all levels of the other cues. Given that fully grammatical sequences in which cues conAict are relatively rare, this is a necessary procedure to examine the relative importance of cues on sequences in which such cues are forced t o conflict. Although this complete crossing often yields ungrammatical se- quences, previous research has shown that grammatical and ungrammatical sequences are judged using similar strategies (MacWhinney, Pleh, & Bates, 1985). As Kail(1989) has pointed out, this does not imply that ungrammatical sequences are pro- cessed exactly the same way as are grammatical ones. Rather, processing of both types of sequences will probably be done on the basis of the same set of cues. Accordingly, the stimuli for our experiment were devised by completely crossing all variables involved in each experiment without concern for the grammaticality of the resultant sequences.s

    PROCEDURE

    Participants first completed a form dealing with their lan- guage background. Then, t o control for prior familiarity with the lexical items used, we gave the students a list of the French words used in the stimuli, together with their English equivalents. The participants listened t o a recording ofthe French vocabulary items and their English equivalents and were encouraged to circle, write down, and study any items with which they were not familiar. We

  • Heilenman and McDonald 525

    then told them that they would be listening to some simple sequences and that they were to choose the noun in each sequence that performed the action of the verb. We stressed that some of the sequences might sound odd and that there were no right or wrong answers. Instead, it was their first impression of the meaning of the sequence that was of interest. The participants were provided with answer sheets listing the two nouns contained in each sequence and were instructed t o circle the noun that was the actor/ agent. Individual particupants then listened to the tape, marking their answer for each sequence immediately after hearing it. The L2 participants could extend the time available for answering by pushing the pause button. Participants were not, however, al- lowed to hear each sequence more than once. The FNS and ENS control groups were not given the option of using the pause button.

    RESULTS

    The data for each group of speakers were analyzed in a 3 x 3 ~ 3 (word order x type of clitic pronoun x clitic pronoun agreement) within participants ANOVA. This design was extended to include a two-level between-participant factor when comparing the perfor- mance of a learner group t o that of a native speaker group. Results are, then, extremely complex, with numerous main effects and interactions. Effects in the ANOVA were considered to be signifi- cant when they fell a t or below the .05 level. In addition to reporting significance levels, the relative strength of the cues will be denoted by the percentage of variance they account for in the ANOVA. Cell means for all groups are given in Appendix 2.

    ENGLISH NATIVE SPEAKERS

    The results of the ANOVA on the ENSS performance on English sentences are given in Table 1. As indicated in this table, word order was by far the most influential cue for ENS, accounting for 38% of the total variance. Native speakers chose the first noun

  • 52 6 Language Learn i ng Vol. 43, No. 4

    85% of the time in NVN sequences, 54% of the time in NNV sequences, and 21% of the time in V" sequences. Although this generally agrees with the results of previous studies (e.g., MacWhinney et al., 19841, i t fails t o replicate the strong second- noun bias usually found for NNV strings.

    Previous studies' test items, however, contained no extra pronominal elements occurring before the verb. Thus, actual word orders in our experiment's items were N Pronoun(s1 VN; NN Pronoun(s) V, and Pronoun(s)VNN (see stimuli in Appendix 1). As noted in the discussion ofword order strategies above, ENS assign actorlagent status to the element immediately preceding the verb and not to the element immediately following it. Thus, in N Pronoun(s1 VN strings, actorlagent status would not be assigned to the noun following the verb, consistent with a first-noun bias. In NN Pronoun(s) V strings, the actorlagent would be assigned t o the pronoun preceding the verb. Given the design of this study, the pronoun would agree in gender half the time with the first noun and half the time with the second noun, resulting in no strong preference for either noun. Finally, in Pronoun(s1 V" strings, the actor/agent role would not be assigned to the noun after the verb, resulting in a second-noun preference. Thus, if pronouns are counted as main elements, the use of English word order prefer- ences found elsewhere results in exactly the interpretation pattern observed.

    Although ENS'S choice behavior did not show a main effect of pronoun type, these participants did make use of the information provided by the gender-agreement cue choosing the first noun as actorlagent 53% of the time when agreement was neutral, 59% of the time when agreement favored the first noun, and 47% of the time when agreement favored the second noun. The amount of total variance accounted for by this cue was, however, small, amounting t o only 1%.

    A significant interaction occurred between pronoun type and gender agreement (see Table 2). ENS showed the strongest use of gender agreement when the subject pronoun alone was present, and weaker use when both subject and object pronouns were

  • Heilenman and McDonald 52 7

    Table 1 ANOVA Results for English Native Speakers

    Effect df Fratio p< Variance

    Word Order 2,24 146.3 .001 38%

    Clitic Gender Agreement 2,24 5.3 .050 1%

    Word Order x Gender Agreement 4,48 3.3 .050 2% Pronoun Type x Gender Agreement 4,48 15.5 .001 12% Word Order x Pronoun Type x 8,96 2.5 .050 2%

    Pronoun Type 2,24

  • 528 Language Learning Vol. 43, No. 4

    Table 2 Pronoun Type by Gender Agreement Interaction in ENS

    Percent Choice of First Noun ~ ~

    Gender Subject Object Subject and Agreement Pronoun Pronoun Object Pronouns

    With Both Nouns 53 59 48 With First Noun 77 34 68 With Second Noun 29 70 42

    Table 3 Word Order by Gender Agreement Interaction in ENS

    Percent Choice of First Noun

    Agrees with Agrees with Agrees with Word Order Both Nouns First Noun Second Noun

    NVN NNV V"

    90 86 80 61 61 39 10 32 22

    Table 4 Word Order by Pronoun Type by

    Gender Agreement Interaction in ENS Percent Choice of First Noun

    Subject Gender Subject Object and Object

    Word Order Agreement Pronoun Pronoun Pronouns

    NVN Agrees With Both Agrees With First Agrees With Second

    NNV Agrees With Both Agrees With First Agrees With Second

    V" Agrees With Both Agrees With First Agrees With Second

    96 96 69 54 88 17 8

    46 2

    88 69 96 75 17 69 15 15 46

    85 92 75 54 77 31 6

    35 19

  • Heilenman and McDonald 529

    order, gender agreement, and pronoun type (see Table 4). This reflects the fact that the treatment of the object pronoun like a subject pronoun had a greater effect in the NNV word order than in other word orders, because first-noun choice had a wider range for this word order (from 1748%) than in others: The NVN order showed a ceiling effect, with values ranging from 69-96%; the V" word order showed floor effects with values ranging from 246%.

    FRENCH NATIVE SPEAKERSg

    The results of the ANOVA on the FNS's performance on French sentences are shown in Table 5. French Native Speakers responding to French stimuli also made significant use of word order, with this cue accounting for 18% of the total variance. French Native Speakers exhibited a first-noun preference on NVN strings (77% first-noun choice), a neutral o r slight second noun preference on NNV strings (45% first-noun choice), and a second noun preference on VNN strings (32% first-noun choice).

    Stronger than word order, however, was FNS's use of the clitic gender agreement cue, with this cue accounting for 35% ofthe total variance. When agreement was neutral, FNS chose the first noun as actorjagent 53% of the time, 83% of the time when agreement favored the first noun, and 18% of the time when agreement favored the second noun.

    Of the three interactions reported in Table 5, only one, that between word order and clitic pronoun type will be presented here."' (See Heilenman & McDonald, 1993, for discussion of other interactions.) Table 6 presents the relevant data.

    As discussed earlier, FNS interpretations depended on the presence of SVO or SOV syntactic cores in the stimuli. For the NVN word order, we found first-noun preferences for the subject clitic pronoun alone (Sm containing the underlined syntactic core sVO), and for the object clitic pronoun alone (m0). This tendency was attenuated when both pronouns were present. ( S d O and O d S interpretations both contain syntactic cores.) For the NNV word order, neutral performance was found when the

  • 530 Language Learning Vol. 43, No. 4

    Table 5 ANOVA Results for French Native Speakers

    Effect d f Fratio p c Variance

    Word Order 2, 14 17.8 Pronoun Type 2, 14 e l Clitic Gender Agreement 2, 14 45.8 Word Order x Pronoun Type 4,28 7.1 Word Order x Gender Agreement 4,28 3.3 Pronoun Type x Gender Agreement 4,28 3.7 Word Order x Pronoun Type x 8, 56 1.3

    n s=not significant

    Clitic Gender Agreement

    .001 18% ns 0%

    .001 35%

    .001 4%

    .050 2%

    .050 3% ns 1%

    Table 6 Word Order by Pronoun Type Interaction in FNS

    Percent Ch ice of First Noun

    Subject Clitic Object Clitic Subject and Object Word Order Pronoun Pronoun Pronouns

    NVN NNV V"

    84 84 62 53 32 51 23 38 35

    subject clitic pronoun alone was present (no syntactic core present in either SOsV or OSsV interpretations), a second-noun preference was manifest for the object clitic pronoun alone ( O N ) , and neutral performance occurred again when both subject and object clitic pronouns were present (SO& and O S d both contain syntactic cores). Finally, for the VNN word order, a second-noun preference was found for subject pronouns alone ( m S ) . This tendency was attenuated when either object clitic pronouns alone (no syntactic core present in either oVS0 or oVOS interpretations) or both subject and object clitic pronouns were present (&SO and &OS both having syntactic cores).

  • Heilenman and McDonald 53 1

    NATIVE SPEAKERS OF FRENCH VERSUS NATIVE SPEAKERS OF ENGLISH

    A comparison of the performance of the ENS and FNS shows differences in the cue use by the two groups of native speakers. First, these two groups differed in the use of the word order cue, F(2, 38)=4.0, p

  • 532 Language Learning Vol. 43, No. 4

    p=.05; FL4, F12,40)=3.0,p=.07). Figure 1 illustrates these differ- ences. Whereas all French L2 groups maintained a first-noun preference in the NVN word order and a neutral performance on NNV strings (with the exception of a second-noun preference in FLl), they dropped the English-based, marked second-noun pref- erence on VNN sequences. Of all four groups of French L2 learners, only the performance ofthe most inexperienced learners, FL1, differed significantly from that of FNS, F(2, 74)=4.2, pc.05. Thus, the L2 learner groups show quick abandonment of strong use of English word order strategies.

    All French L2 groups also made significant or marginal use of clitic pronoun agreement cues !see Table 7). None of the French L2 groups differed from the ENS group in the use of this cue, and all showed weaker use of the clitic pronoun agreement cue than that seen in the FNS group: FL1, F(2,74)=40.7,~

  • Heilenman and McDonald 533

    Table 7 ANOVA Results for French L2 groups

    Group and Effect df F ra t io p < Variance

    FL 1 Word Order 2, 60 24.4 .001 17%

    Clitic Gender Agreement (CGA) 2, 60 7.8 .001 1% Word Order x Pronoun Type 4,120 e l ns 0% Word Order x Gender Agreement 4,120 3.5 .010 1% Pronoun Type x Gender Agreement 4,120 9.5 .001 3% Word Order x Pronoun Type x CGA 8,240 3.2 .005 2%

    Word Order 2, 104 43.4 .001 14% Pronoun Type 2, 104 6.4 .005 0% Clitic Gender Agreement (CGA) 2, 104 19.8 .OOO 2% Word Order x Pronoun Type 4,208 2.5 .050 0%

    Pronoun Type x Gender Agreement 4,208 20.8 .001 5%

    Pronoun Type 2, 60

  • 534 Language Learning Vot. 43, No. 4

    First Noun Choice 9 0 Y T

    707-

    40%-

    25%- 20%-

    10%-

    55 55

    i I

    21

    ENS FL1 FL2 FL3 FL4 FNS Language Group

    FLl=French learners, Group 1; FLZ=French learners, Group 2; FL3=French learners, Group 3; FL4=French learners, Group 4; ENS=English Native Speak- ers; FNS=!?rench Native Speakers

    Figure 1. Use of word order cues by English native speakers, French native speakers, and French L2 learner groups.

    reversal with the object clitic pronoun than ENS: FL1, F(4,

    pc.005; FL4, F(4,80)=6.3,p

  • Heilenman and McDonald 535

    35%- 305%- 255%- 205%- 155%- 10%-

    55%- 0%

    Variance Accounted For pronoun agreement 405%- word order _ _ - -

    38\ \ \ \ \ b--27--

    1 1 2 1 ~

    Language Group

    FLl=French learners, Group 1; FL2=French learners, Group 2; FLS=French learners, Group 3; FL4=French learners, Group 4; ENSEnglish Native Speak- ers; FNS=French Native Speakers

    Figure 2. Percentage of variance accounted for by word order and (clitic) pronoun cues for English native speakers, French native speakers, and French L2 learner groups.

    syntactic cores was not seen in the performance of the French L2 learners. This resulted in a significant interaction, when the FL groups were compared t o the FNS, between language group, word order and clitic pronoun type: FLl,F(4,148)=3.7,~

  • 536 Language Learning Vol. 43, No. 4

    thereof with object clitic pronouns, in NNV word order. FL1 differed from both the ENS, F(4, 168)=5.7,~

  • Heilenman and McDonald 537

    native speakers of a language like English-in which word order is of primary importance and case information carried by pronouns is essentially ignored or even misinterpreted-fare when learning a language such as French, in which the information carried by clitic pronouns comprises an important cue for FNS whereas that provided by word order does not. McDonald and Heilenman (1992) investigated the developmental course followed by L2 learners of French when asked t o interpret sequences containing word order, noun animacy, and verb agreement information. There, as was the case here, learners rapidly abandoned English-based word order strategies, particularly in the case of the noncanonical V" word order when the second-noun preference seen in ENS changed to a slight first-noun preference for beginning learners (see Figures 1 and 2).

    This rapid decline in reliance on word order strategies by L2 learners of French can be interpreted as a response to the number of errors such a strategy would produce given the French input. McDonald (1986; 1987b; 1989) and McDonald and MacWhinney (199 1) have proposed a learning-on-error model of cue acquisition in which learners readjust cue strengths on the basis of the misinterpretations produced. l2 To investigate this possibility, transcripts of teacher-talk from French classes a t the first-semes- ter, second-semester, and third-semester levels as well as teacher-talk from an advanced French literature class were exam- ined for transitive sequences with subjects and direct objects. These were then compared to transitive, subject, and direct object sequences found in a set of three conversations among FNS as well as with data from Moreau (1987) (see Table 8).13 On average, had learners used the strict word order strategies of English (SVO, OSV, VOS) t o interpret these sequences, their error rates would have ranged from 17-28% for FNS in conversation with one another, from 20-48% for FNS instructors, and from 5-18% for nonnative speaker instructors. Thus, given the large error rates caused by use of English word order strategies, learners could reasonably be expected to reduce their dependency on them.

    Although space does not permit a full discussion of the issues

  • 538 La ng u ug e Leu m i ng Vol. 43, No. 4

    Table 8 Word Order in French

    Percent of Corpus

    Source svo ovs sov osv vso vos French Native Speakers Moreau (1987) 70 3 14 9 0.1 2 Cosnier & Kerbrat- 58 1 27 13 0.0 0

    Instructor, Semester 1 49 2 06 32 2.0 0 Instructor, Semester 2 36 0 48 16 0.0 0 Instructor, Semester 3 62 05 16 17 0.0 0

    Orecchioni (1987)

    Nonnative Speakers of French Instructor, Semester 1 71 0 05 23 0.0 1 Instructor, Semester 2 65 0 15 20 0.0 0 Instructor, Semester 3 67 0 18 16 0.0 0 Instructor, Advanced Lit. 73 01 13 13 0.0 0

    All nonnative speaking instructors were ENS but one who was a native speaker of Spanish. This speakers English, however, was near-native.

    surrounding negative versus positive evidence (e.g., Cook, 1988; Gass & Lakshmanan, 1991; Lightfoot, 1991), it should be noted here that claims of relationship, such as those espoused by the Competition Model, between language processing and samples of language (putative input) are empirical in nature. That is, if predictions derived from text counts are borne out by experimental evidence or vice versa, a relationship is established. The exact nature of the mechanisms mediating that relationship are in all likelihood quite complex as discussion of the concepts of cue validity, cue cost, and conflict validity makes clear. Moreover, the concept of error is not a simple one involving as i t does competi- tion between items within networks (MacWhinney, 1989b) and conflict internal to the speaker (McDonald, 1989) or failure to parse (MacWhinney, 1992). (For further discussion, see MacWhinney, 1987b,1989a, 1989b; McDonald and MacWhinney, 1989, Taraban, McDonald, and MacWhinney, 1989). In other

  • Heilenman and McDonald 539

    words, misinterpretation does not have to be explicit to be effec- tive. Misunderstandings, as signaled by inhrect feedback (dis- crepancies between learner interpretation and external informa- tion either from surrounding context or later discourse information) or by an internal failure t o parse, appear t o be sufficient to cause a revision of cue weights.

    Beyond the rapid abandonment of word order, a second interesting finding is the significant lag in use of clitic pronoun information by L2 learners of French as compared to FNS. Only the most advanced group of L2 learners of French began to use information coded by clitic pronoun agreement at even close t o the appropriate level. Even here, however, their use was still signifi- cantly below that of the FNS. In addition, this lag is comparable t o that found in a related study (McDonald & Heilenman,1992)14 in which L2 French learners manifested a similar lag for the use of verb agreement cues in French. In both cases, only the most advanced group of learners began to use information coded by clitic pronoun agreement or verb agreement in the same way as do FNS. Nevertheless, although the small number of advanced learners in the present study suggests some caution in interpretation, the results indicate that it is possible for L2 French learners to begin t o approach native speaker norms. Previous research has shown that the mastery of verb agreement and clitic pronoun agreement cues is possible for ENS learning French. McDonald andHeilenman (199 1) found that very advanced English-French bilinguals were able to use both a t levels indistinguishable from that of FNS.

    However, as indicated by the behavior of the L2 learners of French in this study, such mastery is far from immediate. It seems likely that, here, considerations of cue cost come into play. It will be remembered that cue cost refers t o processing limitations as reflected in the perceivability and assignability of cues. Ch ic pronouns in French are non-syllabic, unstressed, and subject to reduction (e.g., j e mange [I eat] vs. jentends [I hear]; il Ze regarde [he it Zoohs at] vs. iZ Zentend [he it hears]), as well as varying in availability in the input. Although subject clitic pronouns are ubiquitous, especially in conversation, when subject clitic pro-

  • 540 Language Learning Vol. 43, No. 4

    nouns rather than lexical nouns are the norm (Lambrecht, 1987), direct object pronouns are less so. For example, extrapolating from data reported in Di Vito (19911, only between 4-9% of all clauses in her extensive corpus of spoken and written French contained direct object clitic pronouns. Thus, for L2 learners of French, information coded by clitic pronouns must, at least initially, represent noise in the signal rather than meaningful input (cf. Perdue, 1991).

    All learner groups in our study but the most advanced appeared t o be processing French direct object clitic pronouns as if they were subject pronouns, a strategy parallel t o that found for ENS responding to English versions of the French stimuli. This result directly contradicts what would be predicted based on arguments of detectability and assignability. That is, how could first-, second-, and third-semester learners of French manage to both perceive and process clitic pronouns that they should in principle find difficult t o detect as well as difficult to process? The obvious answer is perhaps the most likely one. Kroll and Curley (19881, using a bilingual naming task with novice bilinguals, accounted for their results by suggesting that beginning language learners access L2 words through their L1. That is, they translate. Given that the learners in our study were, if anything, less experienced than the learners in Kroll and Curley's study, (maxi- mum classroom exposure of less than 11 months for this study15 as compared t o less than 30 months for Kroll and Curley), and given that the stimuli were clearly pronounced with definite word boundaries, it seems quite possible that learners in the first three groups were, perhaps to varying extents, processing French stimuli in English.

    Supporting evidence for this interpretation comes from a study by Heilenman (1981) in which low-intermediate adult Anglophone French L2 learners showed a differential perfor- mance on interpretation of items in a French listening comprehension task. When task demands were low, the students were able t o monitor morphological and inflectional information, including clitic pronouns and verb agreement, performing with a

  • Heilenman and McDonald 54 1

    high degree of accuracy. Thus, they were able to perceive and process such cues, given clear stimuli and a task that facilitated such processing. In addition, in a task asking them to match active and passive sentences to pictures, they wrote comments in English (e.g., boy by ball)-again an indication that, although they were processing the French stimuli, some percentage of them were probably doing so through English. Finally, when task demands were high, the students ability to effectively process inflectional and morphological information decreased sharply, indicating that real-time perception and use of such cues was problematic.

    It seems likely then that the apparent transfer of English word order strategies into French was due t o the translating of French stimuli into English by at least a certain percentage of the learners in our present study. Casting this in terms of the Competition Model, the learners were apparently using a one-for- one strategy whereby each French item was directly connected to its closest English equivalent and then processed using connec- tions built up within an English network. However, with experience, the French network became sufficiently developed and the more advanced learners began to rely on information generated by connections among French nodes. (See MacWhinney, 1992, for further discussion.) Thus, the lag in appropriate use of informa- tion carried by clitic pronouns may represent a rather profound restructuring of the systems learners are using to comprehend French.

    Although the similar timing of a lag in the use of clitic pronoun agreement information in o u r study and verb agreement information (McDonald & Heilenman, 1992) by L2 learners of French may simply be coincidental, it may also be indicative of the restructuring referred to above. French linguistics has a relatively long tradition of viewing clitic pronouns, especially subject clitic pronouns, as preverbal number and gender markers. Wartburg (1963, p. 681, has contended that, whereas Italian conjugates its verbs using suffixes (e.g., -0, -i, -u), French does so with prefixes (e.g.,je, tu , ill, a conclusion with which many other researchers have concurred (e.g., Sankoff, 1982; Klavans, 1985; Roberge, 1986,

  • 542 Language Learning Vol. 43, No. 4

    1990; Lambrecht 1987,1988; Gadet, 1989; Matthews, 1989; Ossipov, 1990).

    Thus, unlike English pronouns, which seem t o have largely lost their case-marking value, French clitic pronouns have over the centuries gained in the amount of information they carry, serving not only a referential but also an inflectional function. Anglophone L2 learners of French, then, may have t o learn not only the distribution of bound clitic versus unbound tonic or stressed pronouns in French (see Clitic Pronoun Type and Clitic Pronoun Agreement above); but also that clitic pronouns serve as markers of verb agreement as well as serving the more familigr referential function. (See Liceras, 1985; Haegeman, 1990 for discussion within a Universal Grammar framework. 1 In terms of the Compe- tition Model, ENS learning French will have t o discover that morphological cues (verb agreement and clitic pronoun agree- ment) and not syntactic cues (word order) provide the best road map through French sentences.

    The final area in which the L2 learners contrasted with the FNS was the lack of sensitivity of the former group to a syntactic core preference in the interpretation of dislocated sentences. Again, this fact may be tied t o L2 input, in that classroom L2 French learners appear t o lack exposure to such dislocated se- quences (Calv6, 1982; London, 1985). First, dislocations are considered highly inappropriate in monitored speech and in writ- ing (Gadet, 199 1 ) and thus are unlikely to be considered the variety of French appropriate for the classroom (cf. Valdman, 1989). Sankoff (19821, for example, found clitic-copied suhect NPs (e.g., Moi j e [Me I]) t o be present for 35% of the utterances produced by a Montreal woman in an informal context, but virtually absent within the more formal context of a classroom lecture. Further, such sequences are conditioned by context and require more than the sentence-level discourse usually found in first- and second- year French as a foreign language classrooms. (See Ashby, 1982; Barnes, 1985, 1990; Lambrecht, 1987, 1988, for details on the pragmatics of dislocation in French.) In fact, inspection of the teacher-talk transcripts referred to above revealed relatively few

  • Heilenman and McDonald 543

    instances of dislocated sequences. If dislocations produced by one instructor as part of a grammar exercise dealing with object clitic pronouns are excluded, only 30 clear cases of dislocated subject or direct object NPs with resumptive clitic pronouns are found for about seven hours of teacher-talk, with more dislocations being produced by nonnative speakers than native speakers in a ratio of about 2:l. This compares with 27 cases found for less than one hour ofconversation between FNS (Cosnier & Kerbrat-Orecchioni, 1987). Thus, for various reasons, including setting, number of interlocuters, and classroom conventions, dislocated sequences are less likely t o be part of the input for classroom learners of French than for naturalistic learners. To the extent that such input facilitates the acquisition of a clitic pronoun agreement strategy in sentence processing, classroom learners will be disad- vantaged.

    Evidence from studies of French produced by adult classroom learners of L2 French may also be relevant here. According to Trbvise (1986a), dislocations are very infrequenP in the speech of relatively advanced Anglophone L2 French learners, and, to our knowledge they are also quite rare in the French produced by beginning and intermediate Anglophone classroom learners of French.I7 This is apparently not the case for naturalistic learners of French (e.g., Perdue & Deulofeu, 1986; Perdue, 1990, 19911, leading t o the speculation that dislocated sequences in the input may actually facilitate acquisition of the French clitic pronoun system for such learners (Trbvise, 1986b; Perdue, 1991).

    Finally, the results of our study, when taken in conjunction with others done within the framework of the Competition Model, suggest several conclusions about the role of transfer in the listening strategies employed by adult L2 learners. First, there is clear evidence that L2 learners attempt t o transfer L1 strategies, in many cases achieving a reasonable fit t o the input with which they are faced. In some cases, this fit is apparently sufficient t o encourage learners t o remain with the L1 strategies. Such has been the case, for example, for native speakers of languages in which noun animacy plays an important role in sentence process-

  • 544 Language Learning Vol. 43, No. 4

    ing. Gass (19871, for example, found that native speakers ofItalian learning English tended to continue to rely on animacy cues when interpreting English. As demonstrated by McDonald and Heilenman (1992), such a strategy, when combined with a slight SVO bias, does indeed lead t o relatively few errors. When, however, L1 strategies prove counterproductive in dealing with L2 input, learners seem t o be able t o jettison them rather rapidly. The pattern of early loss of English-based word order strategies seen in the present study has also been found for L2 learners of Italian (Gass, 19871, Japanese (Kilborn & Ito, 1989; Sasaki, 19911, and Dutch (McDonald, 1987a).

    The acquisition of morphological L2 strategies, however, appears problematic, especially for L1 speakers of a language such as English in which morphology plays such a minor role. There are several possible reasons. First, strategies grounded in morphol- ogy may exact a greater cue cost from learners than do strategies grounded in word order. To the extent that morphological cues are less salient, learners will have difficulty perceiving and using them. In addition, morphological cues, particularly bound mor- phemes, are known to be resistant t o transfer (Odlin, 1989), probably due to the fact that bound morphemes are so tightly interwoven into a languages structure.

    Further, clitic pronouns, which serve both anaphoric and morphological functions (cf. Matthews, 1989) may also serve t o mark argument roles connected with individual verbs (e.g., Tesniere, 1959; MacWhinney, 198713; Robinson, 1990). Blanche- Benveniste, Deulofeu, Stefanini, and Van den Eynde (1987) have suggested that the basic structure in French is the lexical unit consisting of the verb and its clitic paradigm. That is, instead of assuming that pronouns replace nouns, they assume that nouns are lexicalized representations of pronouns. This fits in well with MacWhinneys (1987b) characterization of grammatical roles as form-function relationships in which forms are the surface word order patterns and morphological markings that cue particular relations (p. 260) and functions are the underlying meaningful relations without which semantic interpretation could not pro-

  • Heilenman and McDonald 545

    ceed (p. 260). The suggestion here is that the verb-clitic pronoun links in French represent structural frames from which depen- dency structures are built (see Robinson, 1990). The learners task, then, is t o establish and strengthen connections between particular verbs and the roles they specify (MacWhinney, 1987b1, coded as clitic pronouns. Such learning no doubt requires time as learners extract the relevant relationships.

    Assuming that the above interpretations are accepted as reasonable, the results of this study have implications for the L2 classroom. If adoption of appropriate L2 strategies is indeed dependent on input, and particularly on learning-on-error, input could be tailored t o maximize the rate of learning. For example, exposure t o the freer word order of French apparently led L1 speakers of English t o speedily abandon their English word order strategies. It is reasonable t o suggest, then, that these students might have been aided in their acquisition of appropriate clitic pronoun agreement and syntactic core strategies by the inclusion of more naturalistic spoken language, including dislocated se- quences, in the classroom. As Hatch, Shirai, and Fantuzzi (1990) have pointed out, connectionist models allow questions about scope and sequence in regard to L2 learning t o once more be taken seriously. In addition, such models allow for the principled testing of such sequencing. This by no means implies that the relationship between input and L2 acquisition is best represented by a straight- forward what goes in equals what comes out concept of language learning. Although that possibility is not necessarily excluded, we consider it quite unlikely. Here work similar to that conducted in L1 acquisition on the effects of recasts and continuations on the acquisition of specific structures (e.g., Nelson, Denninger, Bonvillian, Kaplan, & Baker, 1984) may be one route toward a better understanding of the role of input in L2 acquisition.

    Revised version received on 4 January 1993

  • 54 6 Language Learning Vol. 43, No. 4

    NOTES

    The Competition Model referred to here is quite different from the Competi- tion Model proposed by Felix (e.g., 1985). The latter is concerned with competition between ageneral problem-solvingmodule and a language-specific system in adult language acquisition, rather than with competition among lexical items in a general theory of language processing. The following terminology is adopted here. (L)NS refers to native speakers of a language (e.g., FNS=French native speakers). When referring to bilinguals, the first language will be the L1, the second the L2. Thus, French-English bilinguals indicates people whose L1 is French and whose L2 is English. here appears to be some asymmetry in the development of appropriate L2 strategies. Work done with native speakers of Dutch, French, German, Italian, Japanese, and Spanish with L2 English, has indicated a tendency to transfer L1 strategies (e.g., animacy, case marking, verb agreement) to English L2 processing. Native speakers of English, however, with L2 Dutch, French, Italian, and Japanese, have shown a tendency toward early abandonment of English word order strategies and a slow adaptation toward appropriate L2 strategies(seeKi1born & Ito, 1989; MacWhinney, 1992,forreviews). McDonald and Heilenman (1991,1992) have attributed these effects to characteristics of L2 input, speculating that L2 learners are resetting their L1 cue strengths to LZlevelsbasedon misinterpretationscausedby theuseofL1 cue strengthswith L2 input. Because use of animacy and agreement strategies in English do not cause a large number ofmisinterpretations, L2 speakers of English may retain these strategies. However, the use of English word order strategies in other languages often results in high error rates, which would cause native English speakers to discard these strategies and eventually adopt ones appropriate to the L2. SOV and VSO sequences can be found in archaidpoetic language as in: (SOV) Pensive poets painful vigils keep (Pope, cited in Crystal, 1987, p. 98) and (VSO) Govern thou my song (Milton, cited in Crystal, 1987, p. 98). Clitic pronouns, within the context of this paper, refers to what are commonly known as personal subject pronouns (je, tu , it, elle, on, nous, uvus, ils, elles). We exclude ce (it, that, this) as well as pronominaVreflexive forms from consider- ation. It will be remembered that, although word order in French is basically SVO, the presence ofpreverbal clitic object pronouns produces an SOVpattern. Thus, both these patterns are highly frequent and easily interpreted in French. Nine ofthe 12 nouns were in the first sevenlessons of the textbook then in use. The remaining three (cow-boy, dame, princesse) have closely related English equivalents. The verbfaireispresentedin Lesson 7, tomberinLesson 8. Bouger and tourner both have English cognates (budge and turn). Classes typically completed 12 lessons of the textbook during the first semester. Because all testing was done near the end of the semester, it can safely be assumed that all students in the first semester group had completed at least Lesson 7. Although we agree with McLaughlin and Harrington (1989) who suggest the

  • Heilenman and McDonald 54 7

    careful examination of methodologiesinvolving interpretations ofagrammatical sequences, we are satisfied that the participantsin thisexperiment wereindeed processing the stimuli using linguistic rather than general problem-solving strategies. See Gass (1987), Bates and MacWhinney (1989, pp. 43-46) for discussion of this issue. This section summarizes information found in Heilenman and McDonald (1993). Readers are referred there for further detail and discussion. Stimuli containing the clitic object pronoun la could possibly have been interpreted by participants as past rather than present tense sequences (P. M. Lightbown, personal communication, 1992) as in the two examples below.

    1. la rnaman la fait tourner lhomme (present) 2. la rnarnan la fait tourner lhomme (pass6 compost5)

    If this were the case, then the agreement information carried by the clitic object pronoun would be neutralized. Inspection of the data, however, indicated that such past tense interpretations were unlikely. of the 18 pairs involved (in which sequences withle contrastedwith sequenceswithla), 6 hadcliticpronoun agreement indicatingboth nouns and 8 had such high N1 or N2 choice that any variation due tooneitemofthe pair wasprecluded, leavingonly4pairsin which any effect would have been evident. Visual inspection of these paired items indicated that only one of the four pairs fell in the direction indicated by a past interpretation. Thus, there is no consistent evidence that participants were interpreting la as la . Note also that object pronouns are found preverbally in subject position in coordinated subjects (Hudson, 1990) (e.g., Me and him are leaving), in child English (e.g., Huxley, 1970), and in so-called Tarzan-talk (Me Tarzan, you J a ne) . Note that the learning-on-error model does not require all misinterpretations to be detected nor doesit require that such misinterpetations be signaled to the learner by an interlocutor. Misinterpretations may also be internal to the learner (e.g., a discrepancy between the learners interpretation and informa- tion from the environment). I3The relatively large number of OVS sequences for FNS instructor, Semester 1 was due to the use of inverted questions (e.g., quel 6ge a-t-il? [what age has he?]) whereas the large number of SOV sequences for FNS instructor, Semester 2 is due to the fact that this lesson concerned direct object clitic pronouns, thus resulting in frequent examples such as il me regarde (he me looks at). The increased number of OSV sequences for classroom talk as compared to conver- sation is due to the greater number ofquestions (e.g.,quest-ceque tu fais? [what is it that you do?]) in the former context. I4Although data from the ENS and FNS reported here a re from the same speakers as in McDonald & Heilenman (1992), different learners of French served as participants for the two studies. 15As is frequently the case, the students in the first semester were not necessarily complete beginners, with many having already studied French in high school. Given that the effect in question, however, persists through the third semester of study, we doubt that classroom exposure prior to beginning

  • 548 Language Learning Vol. 43, No. 4

    first semester university-level French had any discernible influence on the results. TrBvise (1986a) did find frequent examples of anaphoric chst (e.g., Legroupe ici chst tr& diuisB [The group here its uery diuided]). 17Barnes (1990, p. 106, note 5) reported fairly frequent and appropriate use of dislocations in the oral production of advanced (undergraduate major/minor) Anglophone L2 French learners. She also reported frequent and appropriate use of such structures in the French of L2 learners who had spent significant time in a Francophone environment. Although Barnes does not specify, i t seems likely that her counts included anaphoric ckst (see note 161, which has been found to be frequent in Anglophone L2 French (TrPrBvise, 1986a).

    REFERENCES

    Ashby, W. J. (1982). The drift of French syntax. Lingua, 5 7 , 2 9 4 6 Ashby, W. J. (1988). The syntax, pragmatics and sociolinguistics of left- and

    right-dislocations in French. Lingua, 75,203-229. Ball, R. V. (1983). Noun, pronoun and discourse structure in French. Modern

    Languages, 64,229-239. Barnes, B. K. (1985). The pragmatics of left detachment in spoken standard

    French. Amsterdam: John Benjamins. Barnes, B. K. (1990). Apports de lanalyse du discours B lenseignement de la

    langue [Contributions of discourse analysis to language teaching]. French Review, 64,95-107.

    Bates, E., McNew, S., MacWhinney, B., Devescovi, A., & Smith, S. (1982). Functional constraints on sentence processing: A cross-linguistic study. Cognition, 11, 245-299.

    Bates, E. & MacWhinney, B. (1981). Second-language acquisition from a functionalist perspective: Pragmatic, semantic, and perceptual strategies. In H. Winitz (Ed.), Native language and foreign language acquisition (pp. 190-214). New York: New York Academy of Sciences.

    Bates, E. & MacWhinney, B. (1982). Functionalist approaches to grammar. In E. Wanner & L. R. Gleitman (Eds.), Language acquisition: TF state of the art (pp. 173-218). Cambridge: Cambridge University Press.

    Bates, E. & MacWhinney, B. (1987). Competition, variation and language learning. In B. MacWhinney (Ed.), Mechanisms of language acquisition (pp. 157-193). Hillsdale, New Jersey: Lawrence Erlbaum Associates.

    Bates, E. & MacWhinney, B. (1989). Functionalism and the Competition Model. In B. MacWhinney & E. Bates (Eds.), The crosslinguistic study of sentence processing (pp. 3-73). Cambridge: Cambridge University Press.

    Blanche-Benveniste, C., Deulofeu, J., Stkfanini, J., &Van den Eynde, K. (1987). Pronom et syntaxe: Lapproche pronominale et son application au franqais

  • Heilenman and McDonald 549

    [Pronoun and syntax: The pronomial approach and its application to French]. Paris: Societ4 #Etudes Linguistiqueet Anthropologique deFrance.

    Borer, H. (1986). Introduction. In H. Borer (Ed.), The syntax ofpronominal clitics (pp. 1-1 1). Orlando, Florida: Academic Press.

    Calve, P. (1982). Un trait du fi-anquais par16 authentique: La dislocation [A feature of authentic spoken French: Dislocation]. Bulletin de IACLA, 4,25- 44.

    Charvillat, A. & Kail, M. (1991). The status of canonical SVO sentences in French: A developmental study of the on-line processing of dislocated sentences. Journal of Child Language, 18,591-608.

    Clark, E. (1985). The acquisition of Romance, with special reference to French. In D. I. Slobin (Ed.), The crosslinguistic study of language acquisition: Vol. 1. The data (pp. 687-782). Hillsdale, New Jersey: Erlbaum.

    Cook, V. J. (1988). Chomskys universal grammar: An introduction. Oxford: Basil Blackwell.

    Cosnier, J. & Kerbrat-Orecchioni, C. (1987).Dbcrire la conversation [Describing conversation]. Lyon: Presses Universitaires de Lyon.

    Crystal, D. (1987). The Cambridge encyclopedia of language. New York: Cambridge University Press.

    de Fornel, M. (1988). Constructions disloqukes, mouvement thkmatique et organisation preferentielle dans la conversation [Dislocated constructions, thematic movement, and preferred organization in conversation]. Langue Franqaise, 78, 101-128.

    Di Vito, N. 0. (1991). Incorporating native speaker norms in second language materials. Applied Linguistics, 12, 383-396.

    Dixon, R. M. W. (1991). A new approach to English grammar, on semantic principles. Oxford: Claren don.

    Durgunoglu, A. Y., & Hancin, B. J. (1992). An overview of cross-language transfer in bilingual reading. In R. J. Harris (Ed.), Cognitive processing in bilinguals (pp. 391-411). Amsterdam: North-Holland.

    Felix, S. W. (1985). More evidence on competing cognitive systems. Second Language Research, 1,47-72.

    Gadet, F. (1989). Lefran~aisordinaire[OrdinaryFrench]. Paris: Armand Colin. Gadet, F. (1991). Le parle code dans lkcrit: Le traitement du dktachement par

    les grammairiens du XXE sihcle [Spoken language in the written: Treatment of detachment by 20th century grammarians]. Langue FranGaise, 89,110- 124.

    Gass, S. M. (1987). The resolution of conficts among competing systems: A bidirectional perspective. Applied Psycholinguistics, 8, 329-350.

    Gass, S. M. (1989). How do learners resolve linguistic conflicts? In S. M. Gass & J. E. Schachter (Eds.), Linguistic perspectives on second language acqui- sition (pp. 183-199). Cambridge: Cambridge University Press.

  • 550 La nguage Lea m ing Vol. 43, No. 4

    Gass, S. M. & Selinker, L. (Eds.). (1983). Language transfer i n language learning. Rowley, Massachusetts: Newbury House Publishers.

    Gass, S. M. & Lakshmanan, U. (1991). Accounting for interlanguage subject pronouns. Second Language Research, 7,181-203.

    Gasser, M. (1990). Connectionism and universals of second language acquisi- tion. Studies in Second Language Acquisition, 12, 179-199.

    Glisnn, E. W. (1985). The effect of word order on listening comprehension and pattern retention: An experiment in Spanish as a foreign language. Lan- guage Learning, 35,443-469.

    Haegeman, L. (1990). Current linguistic theory and L2 learning: Some propos- als for research. Studia Germanica Gandensia, 20,9-51.

    Harrington, M. (1987). Processing transfer: Language-specific processing strat- egies as a source of interlanguage variation. Applied Psycholinguistics, 8, 351-377.

    Harris, M. (1988). French. In M. Harris & N. Vincent (Eds.), The romance languages (pp. 209-245). London: Croom Helm.

    Hatch, E., Shirai, Y., & Fantuzzi, C. (1990). The need for a n integrated theory: Connecting modules. TESOL Quarterly, 24,697-716.

    Heilenman, L. K. (1981). Do morphemes mature? The relationship between cognitive maturation and linguistic development in children and adults. Language Learning, 31,51-65.

    Heilenman, L. K. & McDonald, J. L. (1993). Dislocated sequences and word order in French: A processing approach. Journal of French Language Studies, 3, 165-190.

    Hudson, R. (1990). English word grammar. Cambridge, Massachusetts: Basil Blackwell.

    Huxley, R. (1970). The development of the correct use of subject personal pronouns in two children. In G. B. Flores d'Arcais & W. J. M. Levelt (Eds.), Advances in psycholinguistics: Research papers presented at the Bressanone Conference ofPsycholinguistics, summer courses of the University ofPadova, Jury 1969 (pp. 141-165). Amsterdam: North-Holland.

    Kail, M. (1989). Cue validity, cue cost, and processing types in sentence comprehension in French and Spanish. In B. MacWhinney & E. Bates(Eds.1, The crosslinguistic study of sentence processing (pp. 77-117). Cambridge: Cambridge University Press.

    Kail, M., & Charvillat, A. (1986). Linguistic cues in sentence processing in French chi1 dren and adults from a crosslinguistic perspective. In I. Kurcz, G. W. Shugar & J. H. Danks (Eds.), Knowledge and language (pp. 349-374). Amsterdam: North-Holland.

    Kail, M. & Charvillat, A. (1988). Local and topological processing in sentence comprehension by French and Spanish children. Journal of Child Lan- guage, 15,631-662.

  • Heilenman and McDonald 551

    Kilborn, K. W. (1987). Sentence processing in a second language: Seeking a performance definition of fluency. Unpublished doctoral dissertation, Uni- versity of California, San Diego.

    Kilborn, K. W. (1989). Sentence processingin a secondlanguage: The timing of transfer. Language & Speech, 32, 1-23.

    Kilborn, K. W., & Cooreman, A. (1987). Sentence interpretation strategies in adult Dutch-English bilinguals. Applied Psycholinguistics, 8, 415-431.

    Kilborn, K. W., & Ito, T. (1989). Sentence processing strategies in adult bilinguals. In B. MacWhinney & E. Bates (Eds.), The crosslinguistic study of sentence processing (pp. 257-291). Cambridge: Cambridge University Press.

    Klavans, J. (1985). The independence of syntax and phonology in cliticization. Language, 61,95-120.

    Kroll, J. F. & Curley, J. (1988). Lexical memory in novice bilinguals: The role of' concepts in retrieving second language words. In M. M. Gruneberg, P. E. Morris, & R. N. Sykes (Eds.), Practical aspects of memory: Current research and issues: Vol. 2. Clinical and educational implications (pp. 389-395). New York: Wiley.

    Lado, R. ( 1957). Linguistics across cultures: Applied linguistics for language teachers. Ann Arbor: University of Michigan Press.

    Lambrecht, K. (1987). On the status of SVO sentences in French discourse. In R. S. Tomlin (Ed.), Coherence and grounding in discourse: Outcome of a sumposium, Eugene, Oregon, June 1984 (pp. 217-261). Amsterdam: John Benjamins.

    Lambrecht, K. (1988). Presentational cleft constructions in spoken French. In J. Haiman & S. A. Thompson (Eds.), Clause combining in grammar and discourse (pp. 135-179). Amsterdam: John Benjamins.

    Liceras, J. M. (1985). The value of clitics in non-native Spanish. Second Language Research, 1,151-168.

    Lightfoot, D. (1991). How to set parameters: Arguments from language change. Cambridge, Massachusetts: MIT Press.

    London, D. (1985). Le pronom tanique etlacommunication [Tonic pronouns and communication]. Canadian Modern Language Review, 42,91-92.

    MacWhinney, B. (1987a). Applying the Competition Model to bilingualism. Applied Psycholinguistics, 8, 315-327.

    MacWhinney, B. (1987b). The Competition Model. In B. MacWhinney (Ed.), Mechanisms of languageacquisition (pp. 249-308). Hillsdale, NJ: Erlbaum.

    MacWhinney, B. (19894. Competition and connectionism. In B. MacWhinney & E. Bates (Eds.), The crosslinguistic study of sentence processing (pp. 422- 457). Cambridge: Cambridge University Press.

    MacWhinney, B. (1989b). Competition andlexical categorization. In R. Conigan, F. R. Eckman, & M. P. Noonan (Eds.), Linguistic categorization (pp. 195- 241). Philadelphia: John Benjamins.

  • 552 Language Learning V O l . 43, No. 4

    MacWhinney, B. (1989~). Competition and teachability. In M. L. Rice & R. L. Schiefelbusch (Eds.), The teachability of language (pp. 63-104). Baltimore: Paul H. Brookes.

    MacWhinney, B. (1992). Transfer and competitionin secondlanguagelearning. In R. J. Harris (Ed.), Cognitive processing in bilinguals (pp. 371-390). Amsterdam: North-Holland.

    MacWhinney, B., Bates, E., & Kliegl, R. (1984). Cue validity and sentence interpretation in English, German and Italian. Journal of Verbal Learning and Verbal Behavior, 23, 127-150.

    MacWhinney, B., PlBh, C., & Bates, E. (1985). The development of sentence interpretation in Hungarian. Cognitive Psychology, 17, 178-209.

    Matthews, S. J. (1989). French in flux: 5pological shift and sociolinguistic variation. In T. J. Walsh (Ed.), Georgetown University Round Table on Languagesand Linguistics 1988 (pp. 188-203). Washington,DC:Georgetown University Press.

    McDonald, J. L. (1986). The development ofsentence comprehension strategies in English and Dutch. Journal ofExperimenta1 Child Psychology, 41,317- 335.

    McDonald, J. L. (1987a). Sentence interpretation in bilingual speakers of English and Dutch. Applied Psycholinguistics, 8 ,379413 .

    McDonald, J. L. (1987b). Assigning linguisticroles: The influence ofconflicting cues. Journal of Memory & Language, 26, 10Cb117.

    McDonald, J. L. (1989). The acquisition of cue-