A la recherche des clitiques perdus:The dictogloss as a measure of the comprehension of y and en by L2 learners of French
In an effort to ascertain whether the paucity of object clitics in L2 production documented in the extant research may reflect comprehension difficulties, this article reports on the use of a dictogloss task to determine the degree to which intermediate-level L2 learners of French (N = 110) were able to process and reproduce the meaning of the clitics y and en. An analysis of the reconstructed texts revealed the presence of competing interlanguage forms. Overall, deleted objects, strong pronouns, and lexical noun phrases were used with greater frequency than the target forms. Errors related to animacy, argument structure, and referent constituted the primary source of non-target-like usage. Given the learners' frequent use of animate forms in lieu of y and en, it is suggested that teachers might do well to provide explicit instruction on the animacy distinction in prescriptive French.
L'objectif de la recherche rapportée dans cet article était de déterminer si le peu de clitiques objets documentés dans les tâches de production en langue seconde (L2) dans les recherches antérieures sont le reflet de difficultés de compréhension chez les apprenants. À cette fin, une tâche de dictogloss a été soumise à 110 apprenants de français L2 de niveau intermédiaire, pour déterminer dans quelle mesure ces apprenants étaient capables de comprendre la signification des clitiques y et en et de les utiliser. Une analyse des textes restitués a révélé la présence de diverses formes en concurrence dans l'interlangue des participants. En général, les apprenants ont plus fréquemment éliminé des objets ou employé d'autres pronoms ou des groupes nominaux plutô t que les formes ciblées. Les erreurs reliées à l'animéité, à la structure argumentative et aux référents constituaient la cause principale d'emploi de tournures différentes des formes ciblées. Étant donné la fréquence avec laquelle les apprenants emploient des formes animées au lieu des formes y et en, il serait opportun que les enseignants fournissent des informations explicites aux étudiants concernant le juste emploi de l'animéité en français. [End Page 471]
French as a second language (FSL), pronouns, clitics, listening, comprehension, language processing
français langue seconde (FLS), pronoms, clitiques, compréhension orale, traitement langagier
The French pronominal system presents a considerable challenge for second language (L2) learners. Pre-verbal object clitics are particularly fragile grammatical elements whose usage L2 learners fail to master, despite their ubiquity in both spoken and written language. Clitic forms are present in instructional materials and are the object of form-focused instruction in the classroom setting. Despite this, I have observed that instructed L2 learners in Canada, of varying ages and proficiency levels, make limited use of object clitics in their oral and written production. When they do use these forms, they strugglewith morpho-semantic distinctions (gender, number, and person), word order, and verb argument structure. These difficulties are further compounded by competing weak/strong pronominal paradigms that offer an apparent choice between J'y vais and Je vais là ('I go there'), for example. Given the presence of dual pronominal systems in French, the lack of one-to-one mappings for many object clitics,1 their complex grammatical behaviour, their phonological form, and their low saliency in the speech stream, in addition to their diminished rates of production by L2 learners (e.g., Harley, 1986; White, 1996), is it possible that object clitics also pose difficulties in aural comprehension?
In the present study, a dictogloss task was used to determine to what degree university-levelL2 learners of French are able toprocess andreproduce the clitics y and en. Following from VanPatten (2004), processing involves the formation of (partial or complete) form-meaning connections during online listening comprehension. The dictogloss (Wajnryb, 1990) is a pedagogical task in which L2 learners reconstruct a written version of an orally delivered text. Traditionally, researchers have used the dictogloss as a means of studying students' interactions during form-focused language-production tasks (for a review see Fortune, 2005). In this study the dictogloss served as an exploratory tool to provide insight into learners' processing of the clitics y and en and the competing representations of these forms in their interlanguage (IL).
Strong object pronouns and (weak) object clitics
As Table 1 shows, each French pronoun has both a strong and a weak form. The pronominal system also includes indirect y/à ça, locative y/là, and the genitive en/de ça.2 Strong pronouns are placed post-verbally, in the same slot as full determiner phrases. In contrast, clitic pronouns [End Page 472] are verbal affixes (bound forms unable to stand apart from their hosts). French object clitics cannot be modified, conjoined, stressed, or separated from their verbal hosts. These forms can undergo liaison/elision and compete with strong object pronouns (Kayne, 1975, pp. 66-83).
Object clitics and strong object pronouns in French
The clitic forms targeted in the current study, y and en, replace nominal clauses or prepositional phrases that expand either a verb or a noun (Dubois, 1965, p. 137). Two strong pronouns alternate with the clitic y in its indirect-object and locative functions:
Bruno n'arrêtait pas d'y penser (indirect object clitic) vs. Bruno n'arrêtait pas de penser à ça (strong indirect object): 'Bruno didn't stop thinking about it.'
Bruno y allait tous les jours après le travail pour prendre une bière (locative clitic) vs. Bruno allait là tous les jours après le travail pour prendre une bière (strong locative): 'Bruno went there every day after work for a beer.'
En typically tends to replace a noun or a noun phrase (NP) that is preceded by the preposition de. The current study focuses on en in its referential and partitive functions, which alternate with two types of strong pronouns:
. . . et leur demander ce qu'elles en pensaient (indirect object clitic) vs. . . . et leur demander ce qu'elles pensaientde cela (strong indirect object): '. . . and ask them what they thought about it.'
. . . il en buvait quand elle a attiré son attention (partitive clitic) vs. . . . il buvait de ça quand elle a attiré son attention (strong partitive): 'He was drinking (some of) it when she caught his attention.' [End Page 473]
VanPatten (1996, 2002, 2004) has proposed and refined a model of input processing based on principles and corollaries describing what happens in a learner's working memory (WM) when he or she is engaged in input processing and detailing how intake is derived from input. VanPatten asserts that L2 learners have a limited capacity for processing input. Capacity limitations necessitate a careful allocation of attention, such that learners preferentially derive meaning from the lexis rather than from grammatical form. VanPatten's principles also address how factors such as redundancy, meaningfulness, and placement influence processing. By reference to VanPatten's processing principles (2004) and his communicative value construct (1985), whereby the value attributed to any given form is a function of how much it contributes to the overall meaning of a sentence (+/2 semantic information; +/2redundancy), French object clitics can be analyzed for their overall processibility. Using VanPatten's 'meaning-fulness' criteria, and for the purposes of the present study, y and en are classified as having a semantic value (+S), as they provide cues to animacy and argument structure by their very nature (-animate, + indirect).3 These characteristics facilitate co-referencing with intra-or inter-sentential antecedents, making y and en redundant (+R). Given their (+S, +R) classification, object clitics could be tentatively classified as having an intermediate-level communicative value. Analysis of these forms under VanPatten's Sentence Location Principle (2004), however, would indicate that object clitics are poor candidates for processing, given their sentence-medial placement. While the semantic weight and redundancy of clitics appear to counteract each other to some extent, the lack of acoustic saliency and medial position of these forms should not particularly facilitate their perception, noticing, and subsequent processing.
Research on the acquisition of pronominalization in French
The acquisition of French clitics by first language (L1), L2, bilingual L1, and specifically language impaired populations has been widely studied. The following characteristics are common to all learner populations:
1. A delay in the appearance of object clitics (Hamann, Rizzi, & Frauenfelder, 1996; Herschensohn, 2004)
2. A tendency to overgeneralize subject clitics to object contexts (Jakubowicz, 1991; Naiman, 1974; Selinker, Swain, & Dumas, 1975) [End Page 474]
3. Underdeveloped knowledge of verb-argument structure and diffi-culty using double object-clitic constructions (Kenemer, 1982; Connors, Nuckle, & Greene, 1981)
4. The use of null pronouns, strong pronouns, and lexical objects in pronominalization contexts (Chillier et al., 2001; Paradis, 2004; Schlyter, 1997)
Research suggests that these four developmental characteristics are universal. However, their frequency and the length of time for which they persist in learner production differ greatly as a function of the context of acquisition. For example, while normally developing mono-lingual L1 learners are able to consistently produce syntactically correct object clitics from the age of four onward (e.g., Chillier et al., 2001), L2 learners may never show mastery of productive clitic use (e.g., Herschensohn, 2004).
Word order also poses acquisitional difficulties for L2 learners, whereas their monolingual and bilingual counterparts make virtually no object-placement errors in SOV or SVO contexts (Chillier et al., 2001; Jakubowicz & Rigaut, 2000). SOV word order, however, is particularly challenging for L2 learners of varying L1 backgrounds, who often go through a phase of non-target-like postposed clitic placement (Schlyter, 1997; Towell & Hawkins, 1994).
Although many studies have examined the production and comprehension of object clitics by monolingual and bilingual child learners, L2 learners have been investigated to a much lesser extent (Wust, 2006). One particularly underrepresented area is aural comprehension of clitics. Two studies tested the aural comprehension of object clitics by L2 learners using elicited imitation measures in which the learners repeated utterances they had heard. Naiman's (1974) work with Grade 1 and 2 French immersion students revealed that both direct and indirect lexical objects were more easily processed than their clitic counterparts. In a study by Gundel, Stenson, and Tarone (1984), adult L2 learners of French made no use of lexical NPs in their repetitions but frequently deleted objects. The learners who had studied abroad made the most target-like use of object clitics. Although admittedly limited in scope, these two studies would tentatively suggest that neither child nor adult L2 learners of French are particularly successful at comprehending and reproducing clitics.
While language comprehension and production draw on a common grammar (Chomsky, 1972), this does not preclude the existence of two differentiated systems involved in language [End Page 475] comprehension (e.g., Bishop, 1997) and production (e.g., Levelt, 1989). Ellis (1994, p. 47) asserts that 'there is, in fact, a fundamental difference between comprehension and production in processing terms,' the former carrying a lighter cognitive load. Given the assumption of a single grammar, however, could there then be a link between the comprehension and the production of clitics? If L2 learners are not producing object clitics, or are doing so only sporadically, perhaps it is because they are incapable of syntactically parsing and representing these forms in comprehension.4 In order to determine the validity of this assumption, more in-depth research beyond simple imitation measures at the sentence-level must be undertaken to assess L2 learners' ability to correctly interpret the meaning of object clitics.
The current study proposes to fill a clear gap in the research on clitics by addressing the following research questions:
1. Do L2 learners produce clitics, strong (i.e., freestanding) pronouns, null objects, and lexical NPs in y and en pronominalization contexts?
2. Does a learner's ability to replicate the meaning of y and en vary depending on L2 proficiency level (low vs. intermediate vs. high), as determined by a placement test?
3. Does a L2 learner's ability to replicate the meaning of y and en vary depending on the total amount of exposure to French (French immersion vs. core French)?
Method
Participants
The participants in the current study were 110 students from the overall population enrolled in French 150 at the University of Alberta. Nine classes volunteered to participate: three low-, three intermediate-, and three advanced-level sections of French 150, for a total of 34 low-, 32 intermediate-, and 44 advanced-level students. Four adult native speakers (NSs) of French from France served as a control group. Because of the heterogeneous nature of the population of students enrolled in the course, on the first day of the fall semester participants took a pen-and-paper placement test, consisting of seven distinct parts, intended to assess performance on the primary grammatical structures to be covered in the eight-month course: agreements, [End Page 476] indicative mood (past, present, and future tenses), subjunctive mood, personal pronouns, strong pronouns, relative pronouns, possessive pronouns, and if-clause constructions. Students also completed a general measure of knowledge of French syntax, in which they were required to use a series of supplied items (e.g., sentence builders) to construct complete and meaningful sentences, and responded to an essay prompt in French that was designed to elicit past-tense indicative forms as well as present and past conditional forms. Each of the nine sections of the course met during one of three time slots; the placement tests for each time slot (09:00, 11:00, or 14:00) were graded by the instructors, who then prepared frequency charts of the results. The course instructors, in consultation with the language program coordinator, analyzed the distribution of marks and assigned students to low, intermediate, or advanced sections of the course. Neither the reliability nor the content and construct validity of the placement test was assessed. Cut-off points for section assignment were not established empirically.
For research purposes, the 110 participants were divided into 55 dyads, 11 of them consisting of post-French immersion (FI) students and the other 44 consisting of post-core French (CF) students. Student responses to a background questionnaire revealed that they had had anywhere from three to 13 years of French instruction in school and that 79.1% of them came from monolingual anglophone language backgrounds. The other 20.9% came from homes where one or more other languages in addition to English were spoken to varying degrees; 10% of these students indicated that French was spoken in their homes. The dyads included more female than male students (78.2% and 21.8% respectively), and 92.5% of the participants fell into the 18- to 25-year-old age category. Participants reported never or rarely using written (68%) or spoken (59%) French media outside the classroom when not specifically asked to do so by their instructor. The majority of students had experienced short stays in a francophone environment: 42.1% reported having spent less than 24 hours, and an additional 17.7% having stayed for less than seven days. The 37.3% of students who stayed with francophone or French-speaking friends and the 15.3% who lived with a host family as part of a home-stay program indicated that they had, for the most part, made frequent use of French in their interactions. Finally, 55.5% of students reported making an effort outside regularly scheduled class time to improve their French, primarily through oral interactions with friends and family members or with customers or colleagues at their place of employment. [End Page 477]
Procedure
Testing took place during a regularly scheduled 50-minute class period. After signing a consent form and completing a background questionnaire, participants received an activity packet containing three tasks, which were administered in the following order: an aural L2-L1 translation, a dictogloss, and a written cloze. Only the dictogloss task data are examined here.
The dictogloss task
The dictogloss (Wajnryb, 1990) is a form-focused activity that allows learners to reflect on their own L2 written output. Learners are read a short, dense text (typically four to five sentences) that allows for practice of specific grammatical constructions. During the first reading, pauses between sentences allow students time to jot down notes. Following a second reading, learners work in pairs to re-create the text from their shared resources, such that the reconstruction reflects the information contained in the original and follows the conventions of the target language. The reconstructed texts are analyzed and compared in a whole-class setting. Successful dictogloss completion requires learners to (1) process the input; (2) hold the forms in memory; (3) generate, discuss, and negotiate the utterances that will form the reconstructed text; and (4) create a written version.
A modified dictogloss technique was used in this study; the modifications were (1) a division of the text into two separate passages; (2) an increase from two to three passage readings; (3) a lack of provision for note-taking; (4) an explicit emphasis on the number of sentences required in students' reconstructions; and (5) omission of the whole-class comparison/analysis phase. The stimulus text for the dictogloss task, a story about a man and a woman who catch each other's eyes at a bar, included at least one token of each of the six clitic pronoun types (le, l,' lui, leur, y, and en) and contained specific vocabulary items and characters from the course textbook. Participants were told to focus on listening to the text because an ideal reconstruction would involve exactly the same number of sentences as the original. Students were asked not to take any notes during the listening phase. The researcher then read the first part of the passage three times, in immediate succession, at a normal rate of speech. This initial passage consisted of six sentences containing a total of six weak pronominal forms (indicated in boldface type): [End Page 478]
1. Bruno et Marianne se sont vus pour la première fois dans un bar.
2. Bruno y allait tous les jours après le travail pour prendre une bière.
3. En effet, il en buvait une quand elle a attiré son attention.
4. Marianne était tellement belle qu'il a eu un coup de foudre pour elle.
5. Il a eu envie de le lui dire et de l'embrasser sur-le-champ.
6. Mais Bruno n'avait pas le courage de lui parler ce soir-là.
Following the presentation of the first part of the listening passage, each of the 55 pre-assigned dyads had approximately 10 minutes to create a single co-constructed written rendering of what they had heard. Exactly the same process was repeated for the second part of the passage, which consisted of five sentences containing a total of eight weak pronominal forms:
7. Cependant, en sortant du bar, Marianne lui a fait au revoir de la main.
8. Bruno en était ravi.
9. Une fois rentré à la maison, Bruno n'arrêtait pas d'y penser et n'a pas pu s'empêcher d'en parler.
10. Et Marianne, pour sa part, a appelé ses meilleures amies pour leur raconter son histoire et leur demander ce qu'elles en pensaient.
11. Est-ce que cet homme du bar pourrait l'aimer ?
Coding
The greatest shortcoming of research on French pronominalization acquisition is the lack of systematic delineation of the contexts in which the variables can occur. In order to identify the range of forms produced, the researcher must examine learner-generated discourse and identify each context in which object pronominalization would be permissible. Rather than doing this, however, some researchers have simply reported the relative frequencies of object clitics and null objects (e.g., Schlyter, 1997; White, 1996; Zobl, 1980). The current study controlled for pronominalization contexts by identifying all target-like and target-deviant object-clitic forms produced by the student dyads in conjunction with transitive verbs. Target-like forms in pronominalization contexts included the clitics y and en, lexical NPs, and their strong pronoun counterparts là, à ça/cela, and de ça/ cela. While clitics are clearly the most discourse-appropriate forms in these contexts, the primary goal of the present study was to determine the extent to which university-level learners of L2 French are able to process and reproduce the meaning of clitics; achieving this goal [End Page 479] required including lexical NPs and strong pronouns in the 'target-like' category. Target-deviant forms included clitic omissions as well as clitics, NPs, and strong pronoun forms referring to animate referents in the dictogloss.
For coding purposes, each rewritten text was divided into 14 sentences or clauses. For each individual target context, all occurrences of verbs (both infinitive and conjugated forms) were identified. Each individual verb was matched to its counterpart in the stimulus dictogloss text and coded for the form used to represent the argument (argument omission, clitic, strong pronoun, or lexical NP). These forms were then classified as either rendering the original meaning conveyed in the dictogloss or referring instead to the human protagonists. Instances in which there was no matching verb in student dictoglosses were coded as 'No equivalent sentence.' Frequencies and percentages were subsequently calculated for the various forms used in each pronominalization context.
The coding system representing the categories of forms used by the dyads in pronominalization contexts allowed for their subsequent quantification, which is presented in the following section.
Results and discussion
Quantitative analysis
Although the dictogloss text contains six distinct direct and indirect object forms (le, l,' lui, leur, y, and en), the discussion here focuses on y and en to allow for comparison of the processing and replication of these two inanimate clitic forms with multiple form-to-meaning mappings. Neither animacy nor multiple mappings has been investigated in the literature on the acquisition of clitics in L2 French (with the exception of animacy in Andersen, 1986). Table 2 synthesizes the numbers and percentages of correct clitics, strong forms, and NPs, as well as of deletions, supplied by the student dyads. It also shows the total number of each form type produced across the six target contexts and details the extent to which target sentences were not reproduced by students.
Table 3 synthesizes the numbers and percentages of additional forms produced by the students that were not counted among the correct forms categorized in Table 2 (e.g., other clitics, strong forms, and NPs). For comparative purposes, the total number of each form type produced across the y and en contexts is included. [End Page 480]
Form usage in pronominalization contexts
Table 4 gives an overall picture of the linguistic forms supplied that replicate the meaning of y and en in the original aural text, showing a wide range of learner performance (from 1.8% on indirect en in sentence 9 to 76.4% on locative y in sentence 2).
Interlanguage form usage in pronominalization contexts
[End Page 481]
Replication of the meaning of y and en
The findings presented in Tables 2 and 3 suggest that the dictogloss task was challenging for participants, who reproduced only 196 of a possible 330 verbal hosts in pronominalization contexts. A more global examination of the linguistic forms supplied shows that clitics (n = 84) were favoured over null pronouns (n = 46), lexical NPs (n = 38), and strong pronouns (n = 20). While these forms did not necessarily replicate the meaning of y and en (as Table 3 shows), theywere present in these contexts. This overarching tendency to use clitics in pronominalization contexts differs from the oral-production data of early-childhood monolingual and bilingual learners, who have high rates of lexical complements and object omissions up to the end of their third year (e.g., Hamann et al., 1996; Müller, Crysmann, & Kaiser, 1996; Hulk, 2000). It also contrasts with data from child L2 learners, who prefer lexical complements, strong objects, and null pronouns in clitic contexts, albeit to varying degrees (Adiv, 1984; White, 1996).5
Although object cliticswere the preferred formoverall, thiswas not the case on an item-by-item basis. Clitics were the most frequently supplied form in the locative y, indirect y, and indirect en (sentence 9) contexts, but null pronounswere favoured in the indirect en contexts in sentences 8 and 10, and an NP was most frequently provided in the partitive en context.
The findings presented in
The impact of proficiency
The second research question posited a possible difference in learner performance depending on L2 proficiency level, as measured by a placement test on the first day of classes. It was predicted that the processing and replication of y and en would increase as a function of language proficiency. Table 5 synthesizes the raw mean scores for correct clitics, strong forms, and NPs supplied by the student dyads. Differences were found in low-, intermediate-, and advanced-level students' ability to replicate the meaning of y using a clitic (M = 0.53 vs. 0.63 vs. 1.09); the meaning of y using an appropriate clitic, strong pronoun, or NP (M = 0.82 vs. 0.81 vs. 1.14); the meaning of en using a clitic (M = 0 vs. 0 vs. 0.68); and the meaning of en using an appropriate clitic, strong pronoun, or NP (M = 0.35 vs. 0.75 vs. 1.09). There was an increase in accurate replication of the meaning of the clitics y and en as a function of proficiency, with two notable exceptions. First, dyads in the low-level sections performed similarly to their intermediate-level peers on the replication of the meaning of y using any appropriate form. Second, not a single dyad from the low- or intermediate-level sections used the clitic en (in either its partitive or its indirect function) in their recreations of the dictogloss task, which may indicate that this is a late-emerging form in L2 production as well as among L1 learners of French (e.g., Hamann, Rizzi, & Frauenfelder, 1995).
Raw mean scores for the replication of the meaning of y and en as a function of proficiency
[End Page 483]
The impact of exposure
The third research question posited a possible difference in learner performance as a function of educational background. It was predicted that post-FI students, because of the much higher number of hours of classroom-based instruction in French they had received, would outperform their post-CF counterparts. Table 6 synthesizes the raw mean scores for correct clitics, strong forms, and NPs supplied by the dyads. Post-CF and post-FI students performed similarly in their replication of both the meaning of y using a clitic (M = 0.77 vs. 0.82) and the meaning of y using an appropriate clitic, strong pronoun, or NP (M = 0.95 vs. 0.91). A difference was found, however, in their ability to replicate the meaning of en using a clitic (M = 0.18 vs. 0.73) and the meaning of en using an appropriate clitic, strong pronoun, or NP (M = 0.61 vs. 1.36). This apparent similarity in performance is deceiving, however, as the post-CF learners used y only in the locative context with the formulaic sequence J'y vais ('I go there'), while the post-FI learners also used y in indirect constructions.
Qualitative analysis: The clitic y
Raw mean scores for the replication of the meaning of y and en as a function of educational
This section examines the learners' dictogloss recreations from a qualitative perspective. The stimulus sentences are not sequenced chronologically; rather, they are presented according to individual clitic form and function to facilitate the continued examination of the asymmetry between locative: indirect y and partitive: indirect en. Following from Pirvulescu (2006), learner production of clitics, null objects, strong pronouns, and lexical NPs is discussed in relation to the preferred [End Page 484] forms used by the two NS dyads in y and en contexts in their recreations of the same aural text.
Dictogloss sentence #2: Bruno y allait tous les jours après le travail pour prendre une bière
As Table 2 shows, the stimulus sentence was reproduced in 60.0% of the dictogloss texts, while an additional 5.5% of the texts contained the equivalent strong form là. Students also rendered the meaning of locative y using the NP au bar ('at the bar') in 10.9% of recreations and deleted the locative in 14.5% of cases. Argument deletion could be considered acceptable, as the verb aller can be used intransitively in standard French, although in this context it seems most appropriate to classify it as an illicit locative omission with respect to the two NS control dyads, which both reproduced an overt locative clitic.6
The types of forms supplied by students in the locative pronominalization context show the hypothesized range of forms. The preference for locative y in conjunction with the verb aller ('to go') could be attributed to a frequency effect due to the prior learning and storage of a particular chunk in long-term memory (LTM), in this case J'y vais ('I go there'). It stands to reason that the majority of the dyads were able to process and reproduce this sentence efficiently precisely because they did not need to carry out multiple searches in LTM for individual, unrelated items. Rather, their ability to chunk contributed to the accurate reproduction of this sentence by increasing the number of items that could be simultaneously stored.
Dictogloss sentence #9: Bruno n'arrêtait pas d'y penser
The second token of y occurs in the ninth sentence of the dictogloss, which more than one-quarter (27.3%) of the dyads did not recreate. The absence of this sentence may be more attributable to its placement than to its complexity, as L2 memory research shows that beginning and concluding sentences are most salient for learners and are more accessible for recall purposes than medial ones (Ervin-Tripp, 1974). The form most commonly provided was the indirect clitic y (18.2%), identical to what the students had heard in the original dictogloss passage and to the form used by both NS dyads. Students deleted the indirect object in 5.5% of re-creations. Indirect y and null objects account for only 23.7% of the data for this particular target sentence, raising the question of what forms were used to represent the original referent y. Student versions of this sentence included the hypothesized [End Page 485] range of forms: (a) object clitics, as in *lui penser (12.7%), *la penser (7.3%), and en penser (3.6%); (b) strong postposed forms, as in penser à elle (5.5%), penser d'elle (7.3%), and penser à lui (3.6%); and (c) lexical NPs, as in penser de Marianne (3.6%).7 Given the occurrence of feminine forms such as *la penser, penser à elle, penser d'elle, and penser de Marianne, one possible interpretation of these student reconstructions is that Bruno was thinking not of the moment when Marianne waved goodbye to him but, rather, of Marianne herself. Such an interpretation could indicate that learners were performing only partial parses of this target sentence, relying primarily upon meaning-based comprehension heuristics. In other words, perhaps what the learners perceived as a strongly plausible animate antecedent (Marianne) led their parsers to pursue an incorrect syntactic analysis. Alternatively, participants may have correctly interpreted d'y penser as 'thinking of it' but been unable to access the equivalent L2 forms (y or à ça/cela/ceci) for pronominalizing an inanimate referent during the productive phase of the task. This interpretation takes into consideration the fact that, immediately before participating in the dictogloss task, the students had taken part in an aural French-English translation task. Given the immediate succession of the two listening tasks, non-intentional 'priming' may have occurred, producing an association between L2 clitics and their L1 equivalents. From the data, it is clear that the verb penser à used in combination with an indirect object referring to an abstract concept posed difficulties for the learners, although it is impossible to determine whether the difficulties occurred during comprehension, during production, or both.
Why were these learners better able to represent the meaning of locative y than to represent that of indirect y? Perhaps verb frequency served as a mediating factor in the learning of their form-meaning mappings. Although both aller and penser are classified as high-frequency verbs, aller is the much more frequent of the two (aller and penser are ranked 34th and 154th, respectively, in Gougenheim, 1964). Solely on the basis that higher-frequency forms are more accessible from the standpoint of storage and retrieval, locative y should be more easily processed. In addition to frequency, collocation may also factor into student performance. While the verb aller typically collocates with à in a locative function, the verb penser collocates with both à and de. In other words, from a cue-based-processing perspective, it is likely that students have formed a strong mapping between y and the verb aller, whereas y and en are competing arguments for the verb penser, which makes the syntactic calculation involved in argument [End Page 486] selection more complex. Moreover, although the clitic y expresses both a locative and an indirect function, the meaning of the indirect function could also constitute a source of difficulty for L2 learners because of its abstractness.
Summary
The types of forms supplied by the students in the two pronominalization contexts associated with locative and indirect y show the exact range of forms hypothesized: the clitic y, strong pronouns, null objects, and lexical NPs. Clitics were preferred in locative and indirect contexts by both learner dyads and NS dyads. Despite the preference for clitic usage, the meaning of locative y was better replicated than that of indirect y. Locative y usage was attributed to verb frequency and chunk learning, while a variety of explanations were offered to account for the use of forms conveying meanings other than those expressed in the stimulus text in the indirect context. These explanations include the difficulty of choosing between the competing arguments à and de, the abstractness of indirect y, and an inability to access L2 equivalents of the expression 'thinking of it' if indirect y had been stored as L1 forms during the comprehension phase of the task.
Qualitative analysis: The clitic en
Dictogloss sentence #8: Bruno en était ravi
The first example of indirect en occurred in the eighth sentence of the dictogloss. A very high percentage (52.7%) of student versions of the dictogloss contained no equivalent sentence, which is not surprising in light of the inaccurate renderings of this sentence produced by dyads that attempted to reproduce it and in light of its omission by one of the NS dyads. The NS dyad that reproduced this sentence used the target form en, as did 5.5% of the student dyads. Although the students made minimal usage of indirect en, they successfully rendered the essence of the sentence through their production of Bruno était ravi (30.9%) and *Bruno se ravi (1.8%).
Dictogloss sentence #9: [Bruno] n'a pas pu s'empêcher d'en parler
A second example of the clitic en in its indirect function occurs in the ninth sentence of the dictogloss. One of the NS dyads and 69.1% of [End Page 487] the learner dyads failed to reproduce this sentence. Perhaps the length of the sentence, combined with the inclusion of three object clitics, was a factor in its low reproduction rate. Given that one of the NS dyads did not reproduce sentence 9, the question arises whether the L2 dyads exhibited a similar pattern of behaviour. Some of the L2 dyads reproduced this entire sentence (including both object clitics), while others reconstructed it into two sentences to accommodate the two ideas being expressed (y penser and en parler) and still others provided only partial reconstructions.
Only one of the L2 dyads provided the anticipated clitic en, as did the second NS dyad. Learner production shows a clear preference (yet again) for animate indirect objects, as is illustrated by the usage of animate feminine object forms. Such a strong preference for feminine forms, appearing to refer to Marianne as the only named female character in the dictogloss, is striking. This favouring of feminine forms might indicate that very few of the students actively processed en and so most of them were incapable of correct co-indexing.
Dictogloss sentence #10: . . . et leur demander ce qu'elles en pensaient
Dictogloss sentence #10 includes a third example of the indirect clitic en. The prepositional verb penser de was less accurately processed and reproduced than its counterpart penser à (sentence 9): only 5.5% of dyads supplied the target form, and the object was deleted in 27.3% of re-created texts. The deletion rate may be a reflection of the input to which these L2 learners had been exposed, given the frequent intransitive usage of the verb penser in normative French. Only one sentence in the entire task was reproduced less often than this one, which was absent from almost half (43.6%) of the student dictoglosses and from the production of one of the NS dyads. Given the complexity of the sentence, which is due to its length coupled with the multiplicity of arguments represented pronominally, listeners who were unable to reproduce it likely allocated their precious attentional resources such that (almost) all their processing space was used for input processing and very little remained for storage purposes. The complexity of the sentence would only further compound other documented learner-internal factors that play into syntactic processing, such as low L2 WM capacity, poor L2 decoding, and inadequate L2 processing speed (McDonald, 2006).
In their re-creations of this sentence students did exhibit all four of the most common form types favoured by L2 learners of French in [End Page 488] obligatory pronominalization contexts in free production, with a preference for null pronouns. Once again, however, their form choice often did not appear to refer to the referent in the stimulus text. As was the case with y penser, students may have stored en penser in memory as 'thinking about it.' In their IL this notion was, perhaps, rendered as pensaient de lui, *lui pensaient, *la pensaient, or y pensaient.8 Students did, however, use the correct argument (de) in association with animate co-referents in the form of lexical NPs, as in et demander leur opinion de cet homme (7.3%), and of strong pronouns, as in et demander ce qu'elles pensaient de lui (7.3%). Only a few student sentences showed difficulties with the argument structure associated with the verb penser, as in et demander ce qu'elles y pensaient (1.8%).
The pattern, established in the sentences discussed earlier, of learners using animate objects in inanimate contexts continues. This pattern may indicate shallow parses on the part of the learners, because the clitic y itself provides an animacy cue (y = inanimate). Is it possible that intermediate-level learners of L2 French are not particularly sensitive to animacy cues during sentence processing? Previous research indicates that adult NSs of French use animacy and agreement cues to interpret sentences in their L1 (McDonald & Heilenman, 1992). In contrast, beginning- and intermediate-level L2 learners of French (whose L1 is English) rely upon clitic case inflection, verb agreement, and noun animacy to a much lesser extent in sentence interpretation; instead, these learners favour effective L1 processing strategies, such as using word order to assign agent and patient roles (Heilenman & McDonald, 1993). From a competition-model perspective, y and en are costly cues to interpret to begin with, because they are unstressed forms that are difficult to perceive and because assignment of meaning is complicated by often distant agreement cues (e.g., spanning a number of sentence constituents or even sentence boundaries). High cue cost, coupled with reliance upon L1 sentence-interpretation strategies, may mean that animacy cues are disregarded and that the human actors in the dictogloss are more salient referents than the actions, facts, locations, or events to which y and en refer. The more salient animate referents may consequently be more readily available during the reconstruction phase of the dictogloss and might then plausibly be selected and used in y and en contexts. There is clearly a pattern of animate feminine forms being used to reproduce the meaning of inanimate y and en, which seems to point to a breakdown in comprehension. However, given that the dyads were not asked to perform concurrent recalls during the task, one cannot conclusively [End Page 489] rule out the possibility that learners did understand the meaning of these clitics and that the breakdown occurred in the productive phase of the task.
Dictogloss sentence #3: En effet, il en buvait une quand elle a attiré son attention
In the third sentence of the dictogloss, en functions as a quantified direct object. For this partitive context, a substantial number of dyads (41.8%) did not reproduce an equivalent sentence in their dictogloss texts, and one dyad (1.8%) used the verb boire ('to drink') without a complement. In 16.4% of cases, students provided the clitic en, just as in the original listening passage. Although this replication rate is not particularly high, it is three or more times as great as the use of en in its indirect function in the contexts discussed above. This is also the only target context in which a lexical NP was the favoured representation of the original form: in 38.2% of instances, students did not replace the quantified direct object with en but provided a corresponding lexical NP - une bière ('a beer') - instead, indicating successful comprehension. The use of NPs in clitic-permissible contexts is well attested in the L1 and bilingual acquisition literature in French and has been characterized as a stage through which learners pass on the way to clitic mastery (see Wust, 2006). In this particular sentence, however, use of an NP seems very natural and could certainly not be construed as discourse inappropriate. In fact, one of the NS dyads that completed this same task reproduced une bière while the other dyad used the partitive en, indicating that, in this context, preference for one form over the other is a choice made by the speaker.
Summary
The learner dyads were more able to process and accurately reproduce the verb/complement combination penser à than to process and accurately reproduce penser de. Indirect en was also highly problematic in the target sentences Bruno n'a pas pu s'empêcher d'en parler and Bruno en était ravi, where it was used in a mere 1.8% and 5.5% of student dictoglosses, respectively. The students primarily used feminine animate strong pronouns, clitics, and lexical NPs to replace indirect en. The dyads' processing and reproduction of the meaning of partitive en was determined to be more accurate than their processing and reproduction of the meaning of indirect en. [End Page 490]
Conclusions of this study
The primary objective of the present study was to investigate the extent to which university-level learners of French are able to process and reproduce the clitic forms y and en on a dictogloss task. There are, however, two inherent weaknesses of the dictogloss task in its current application as a research tool. First, although the dictogloss provides rich insights into learners' ILs, it also conflates comprehension and production, because students are engaged in many operations: listening to an oral text, collaborating with peers, and re-creating an agreed-upon version of the text in writing; as a result, it is impossible to differentiate with complete certainty problems relating to the reception of clitics from those related to their reproduction. A second dictogloss-related problem is the interpretation of instances in which learner dyads do not reconstruct target sentences or provide reconstructions containing alternative meanings to those conveyed by y and en in the original text. In the absence of concurrent think-aloud data, one can only speculate as to whether (elements of) these sentences were (a) not perceived and parsed, (b) not remembered as a result of memory overload, or (c) not accessible to the students for graphic representation.
The findings of the study are significant because they help to characterize the underlying difficulty experienced by L2 learners of French in replicating the meaning of the clitics y and en in pronominalization contexts and because they allow the identification of pertinent learner-related factors. While university-level students from different educational backgrounds have a clitic slot in their IL syntax, post-FI students are better able than their post-CF peers to realize these forms in production. The processing and production of these forms also appears to be developmentally constrained, with higher-proficiency learners outperforming their lower-proficiency classmates. From a linguistic standpoint, the results indicate that not all clitic forms are created equal. When a single form stands for different meanings, there is a clear asymmetry in learner production. Learners reproduced the meanings of locative y and partitive en more accurately than those of their indirect counterparts. The high percentage of sentences not represented in the student-generated dictoglosses may indicate problems relating to comprehension, production, or a combination of the two. If one focuses instead on what the dyads were capable of producing, it is remarkable that those who attempted to recreate the target text often produced sentences containing the appropriate content words but inaccurate representations of the original argument [End Page 491] structure and pronominal references (which may indicate an inability to make connections at the discourse level).
A more fine-grained analysis of the data shows results consistent with previous research on the acquisition of pronominalization by learners of French in a variety of contexts. While the learners did make some use of y and en in object pronominalization contexts, these forms were not the preferred type of object representation (with the exception of locative y in sentence #2). Compared to the adult learners of French in Gundel et al. (1984), the university students in the current study deleted object pronouns to a much lesser extent. The null-object rate of 14% in the current study contrasts with rates of 29% and 45%, respectively, for learners who had studied abroad and those who had not; these students also showed a higher rate of usage of lexical NPs. As in Gundel et al.'s study, time spent in a francophone environment led to more frequent use of clitics among the students discussed here.
If only forms that refer to the original meaning portrayed by y and en are considered, the results show that, overall, students made greater use of null objects, strong pronouns, and lexical NPs than of object clitics on a context-by-context basis. This does not mean that the students shied away from using object-clitic forms. An analysis of their reconstructions shows an attempt to use clitics, which, for the most part, were correctly placed in pre-verbal position, indicating that these students are quite advanced in terms of the four-stage clitic placement acquisition sequence that has been documented among L2 French learners (e.g., Herschensohn, 2004; Schlyter, 1997). However, the students made form choices that did not match those presented in the target text, primarily involving the use of an animate object clitic to refer to an inanimate referent. These clitic forms indicate that students may not have been using appropriate sentence-processing cues, such as animacy and agreement, to derive meaning from syntax, instead choosing an object clitic that would refer to the most accessible referent: primarily Marianne, but sometimes Bruno (in sentences of which he was not the subject).
Overall, the IL analysis of the student reconstructions suggests a general lack of sensitivity to the clitics y and en during the parsing phase of this receptive task, and the frequent representation of y and en by animate forms may indicate that participants were relying on top-down processing to decode this aural text. Since the love story was so clearly about Bruno and Marianne, the students may have been expecting these referents to resurface continually and may, as a result, have focused their attention on any cues that were consistent with this pattern, instead of processing the actual object-clitic forms. [End Page 492] This substitution pattern was particularly striking in conjunction with verbs, such as parler and penser, that take both animate and inanimate complements but can also appear without an overt object. An alternative explanation for the data, presented in the results section, is that while some learners were able to understand the inanimate (and often abstract) antecedents of indirect y and en, they were unable to reproduce these meanings because of the way in which these forms were stored in memory. For example, y or en might be stored as the English equivalent 'it'; it is possible that, in these learners' IL systems, 'it' was in fact equivalent to another clitic, such as le, la, or les. A more definitive answer to this question could be provided by future research involving the use of concurrent think-alouds.
Pedagogical implications and directions for future research
Previous research on the acquisition of pronominalization indicates that object clitics are difficult to master in both monolingual and bilingual childhood language acquisition, although this developmental issue is resolved relatively quickly (e.g., Jakubowicz, 1991; Belletti & Hamann, 2004). However, the current study underlines the fact that object clitics also represent a major obstacle for intermediate-level L2 adult learners (whether in terms of comprehension, production, or both). In fact, L2 learners may never gain mastery over object-clitic forms (Herschensohn, 2004; Kenemer, 1982). Paradis (2004, p. 79) asserts that French clitic acquisition poses problems for L2 learners not for transfer-based reasons but, rather, because it is 'challenging in its own right.' Teachers must not resign themselves to such an explanation; instead they should explore new pedagogical techniques that may prove more amenable to acquisition. In light of the results reported here, together with the inherent complexity of the French pronominal system, instructed L2 learners cannot be left to their own devices. Anglophone students, in particular, may very well be listening for object clitics in all the wrong places (i.e., post-verbally). Or perhaps these students are not listening for object clitics at all, unless specifically asked to do so, given the purposeful nature of listening. For these reasons, the object-clitic instructional sequence presented in prototypical communicative textbooks (presentation of the clitic paradigm, followed by mechanical, meaningful, and communicative practice) should be expanded to include activities that (1) sensitize students to the phonological forms of object clitics and (2) train them to listen for these forms pre-verbally to facilitate auditory detection. [End Page 493] Moreover, given that both the current study and Andersen (1986) indicate the problematic nature of +/2 human distinctions for L2 learners in non-subject clitic production, these sensitization activities should be accompanied by explicit instruction on the animacy distinction in prescriptive French.
The dictogloss reconstructions produced in the current study seem to indicate that many learners recognized that the target verbs were complement taking but, at the same time, either missed, ignored, or failed to process the actual complements (y and en). One hypothetical explanation for the findings is that inaccurate processing may have stemmed from both an over-reliance on context and an under-reliance on animacy and agreement cues, such that the original inanimate clitic forms were reproduced as animate clitics, strong pronouns, and NPs that designated one or the other of the main protagonists in the orally presented dictogloss text. Future research on object-clitic processing could provide evidence to either support or refute this claim through the use of a dictogloss task accompanied by both concurrent and retrospective verbal reports. Following from Camps (2003), in the concurrent reports learners would provide insight into their online cognitive processes as actual form-meaning connections were being made. During retrospective interviews, the researcher would ask pointed questions about specific clitic forms mentioned during the concurrent report; any learner-generated references to clitic traits such as gender, number, person, animacy, and case (during language-related episodes); and the actual reproduction of these forms during the written task.
While further research involving verbal reports is needed to determine the validity of the provisional explanation offered in the current study (whereby inaccurate processing is the root cause of learners' failure to replicate the meaning of y and en), based on my practical experience I would suggest that increasing L2 learners' awareness of the morphophonology of French object clitics and giving them strategies to detect these forms in the speech stream is only part of the solution. Teachers would also do well to experiment with techniques associated with VanPatten's PI in order to 'manipulate learner attention during IP and/or manipulate input data so that more and better form-meaning connections are made' (2002, p. 763). A PI-inspired lesson on object clitics thus includes a brief introduction to the actual forms and provides basic metalinguistic information about their meaning, placement, and function. Learners are told that, while in English it is natural to see or hear sentences in which the order is subject-verb-object, in French it is important to be careful not to [End Page 494] 'bypass' the object pronoun, which is often placed before the verb. In addition to learning processing strategies to facilitate location of clitics, learners are also given tips on how to pay attention to the gender, number, person, and animacy cues inherent to clitic forms in order to co-index them with the appropriate co-referents. These morphological features are addressed one at a time and are accompanied by oral and written activities that push students to make correct grammatical/ semantic role assignments. Given the success of PI in helping students to better interpret and produce clitic forms in L2 Spanish (for detailed discussion see VanPatten, 2004), similar processing-oriented sequences may also help learners of French increase their accuracy with respect to clitics.
Both inside and outside the classroom, knowing how to use and to interpret object clitics is an integral part of communicative interactions in French. Learners who overuse strong forms, null pronouns, and lexical NPs rather than using weak clitics produce language that is markedly non-standard. But what is the best way to help L2 learners of French increase their use of clitics? The solution must depend on a deeper understanding of the nature of learners' difficulties in interpreting and reproducing the meanings of these forms, information that could be provided by concurrent and retrospective verbal protocols during and immediately following processing tasks. While the findings of the present study serve to shed some light on where these problems may lie, they also underline the need for further research that will clarify whether low rates of replication of the meaning of y and en are indeed attributable to a breakdown in comprehension, production, or both.
Acknowledgements
I gratefully acknowledge the funding from the Social Sciences and Humanities Research Council of Canada (Award #752-2003-1243) that supported this research. I thank the French 150 instructors and students at the University of Alberta who were involved in this study, and I am indebted to Leila Ranta and four anonymous CMLR reviewers for their valuable comments and suggestions. [End Page 495]
Notes
1. Lack of one-to-one mapping means that a single form stands for multiple meanings, as is the case for te, nous, vous, elle(s), y, and en. As an illustrative example, vous can be the second person plural subject, the second person singular formal subject, the direct object, the indirect object, or the strong postposed pronoun.
2. The term 'indirect' will henceforth be used to refer, more generally, to the clitics y and en.
3. While for the purposes of this study y and en are classified as having semantic value, this value is reduced compared to that of the NPs to which they refer. While y and en provide cues to animacy and argument structure by their very nature (-animate, +indirect), one must still look to the verbal host to determine the function: locative, indirect, or partitive.
4. Just as Fodor (1998, p. 342) asserts that what cannot be parsed cannot be learned, since 'grammar acquisition is at the mercy of the information provided by the parser,' I amsuggesting that grammatical forms that cannot be parsed may very well remain incomprehensible.
5. White (1996) found that in the first two years of L2 acquisition, object clitics and null pronouns were rarely used; lexical complements and the strong pronoun ça were preferred.
6. Pirvulescu (2006, p. 227) uses the notion of the 'clitic-context,' wherein 'the referent is definite, it is the topic of the discussion; and it is contained in the immediately preceding discourse.' She examines object-clitic omission in early monolingual learners of French by comparing their object-omission rates in clitic-context with those of adults.
7. A reviewer suggested that not all occurrences of y penser in student dictoglosses necessarily imply that the intended meaning was 'thinking of it.' Perhaps some students did indeed understand y penser as 'thinking of her' but, as a result of exposure to vernacular varieties of French in which y can refer to animate referents (e.g., J'y parle for Je lui parle), had assimilated this particular usage to the extent that the standard distinction whereby y cannot refer to animate referents was overridden.
8. These IL forms are indicative of the responses provided by eight of the 55 learner dyads.
