The Acquisition of Verbal Epistemic Stance Marking during Study Abroad: The Case of je pense in L2 French

Abstract

Epistemic stance markers, such as je pense in French, take on a variety of discursive functions, ranging from asserting an opinion, indicating the source of information, and mitigating a claim, to pragmatic functions, such as gaining time for discourse planning. Previous research suggests that the epistemic marker je pense is mostly used in French conversation to express opinions and can be used as an indicator of the development of a learner’s assertiveness and pragmatic competence during a study abroad period. Using a functional approach, this paper seeks to find out the extent to which study abroad fosters the development of assertiveness and pragmalinguistic competence among L2 learners, through an analysis of stance marking in interview data from 26 Anglophone learners of French, who spent nine months abroad in a French-speaking country, and 10 French native speakers. The results show that learners are globally less assertive in their use of je pense than native speakers, particularly prior to departure, and that they develop pragmatic uses of je pense, thereby showing a development in their interactional strategies. Finally, the high inter-variability in the way learners used je pense suggests the importance of personal style.

1. Introduction

Expressing stance is inherent in human speech: in conversation, speakers do not merely communicate factual information, they also indicate how they relate to a particular state of affairs, and they align with their interlocutors (Du Bois 2007). Among the most frequent expressions of subjectivity, epistemic stance (i.e., to what extent a speaker is certain about a given claim) is intimately linked to speaker style and personality, as it allows them to take a position and express their level of confidence towards a given piece of information (Bucholtz and Hall 2005; Baumgarten and House 2010). Epistemic stance can be conveyed through a variety of linguistic means, including modal verbs (il doit faire noir; “it must be dark”), but also adverbs (peut-être; “maybe”), and verbal expressions, such as je pense—”I think”; j’imagine—”I imagine”; and il me semble—”it seems to me”. According to Gablasova et al. (2017, p. 614), epistemic stance marking “fulfils three major interconnected functions in the interaction: (i) expressing opinion, (ii) maintaining relations between the interlocutors, and (iii) discourse organization.” Learning how to express epistemic stance in conversation through native-like form–function mappings is therefore a crucial challenge for second language (L2) learners, especially during a study abroad (SA) period, when ability to communicate efficiently with locals is a key competence. While a few studies have focused on the acquisition of epistemic stance in L2 English in a guided context (Gablasova et al. 2017; Kirkham 2011), there is very little work on the development of stance taking in L2 French, nor on the acquisition of stance taking in an SA context. The latter is often considered to foster linguistic development and improve students’ self-confidence and autonomy (see Benson et al. (2012); Milstein (2005); Papatsiba (2005) on self-perceived increase in autonomy and self-efficacy; Grieve (2015) on length of stay and expression of identity; and Compiegne (2020) on SA and the transition to adulthood).

The marking of epistemic stance forms part of the domain of modality. Research on the L2 acquisition of modality shows that at lower proficiency levels, epistemic stance is more often expressed with cognitive verbs and adverbials than with modal verbs, generally considered to be acquired late (Salsbury and Bardovi-Harlig 2000; Fordyce 2009, 2014; Leclercq and Edmonds 2017). Expressing epistemic stance also taps into pragmatic ability, as some stance markers fulfil interactional management functions (Kärkkäinen 2003; Aijmer 2009; Gablasova et al. 2017).

Set within a functional approach, this paper focuses on je pense, one of the most frequent markers of verbal epistemic stance in Hexagonal French,1 and on its development in L2 French during SA. Indeed, learners of French face challenges related to the learning of the tense and aspect morphological features associated to this stance marker, its syntactic placement, and its discursive values. In the functionalist tradition, linguists seek to study human communication by analyzing the interaction between the different levels of linguistic organization (phonology, morphosyntax, lexicon) and contexts of use (Lenart and Perdue 2004, p. 2). By connecting the syntactic and pragmatic choices learners make relative to the use of je pense, we provide an ad hoc contribution to this Special Issue.

The current analysis examines whether learners evolve towards native-like patterns of use of this stance marker during a prolonged stay abroad period. We use data from the LANGSNAP project2 (Mitchell et al. 2017), which is a longitudinal study spanning over 21 months, including semi-guided interview data by French native speakers and Anglophone learners of French, recorded before, during, and after a nine-month stay abroad period in France. Our contribution aims to shed light on the impact of SA on the way Anglophone learners of French develop pragmalinguistic competence, and on the expression of epistemic stance in conversation, a domain which is still largely unexplored.

2. Theoretical Considerations

2.1. Expressing Stance

Stance research is interested in how speakers express their subjectivity, particularly through positioning themselves relative to propositional content (Englebretson 2007). Stance can be expressed through a variety of linguistic means, including adjectives, adverbs, and verbal expressions. Stance marking is inherent in human communication: utterances are always anchored in a given context, and there is no such thing as a purely objective statement. In the words of Biber et al. (1999, p. 971, quoted by Kärkkäinen 2003, p. 68), “In most cases, speakers and authors first identify their personal perspective—their attitude towards the proposition, the perspective that it is true from, or the extent to which the information is reliable—thereby encouraging listeners and readers to process the following propositional information from the same perspective.” The focus on the way the addressee interprets the speaker’s perspective highlights the importance of studying stance taking in interactional discourse, a point of view also supported by Du Bois (2007) in his paper on the stance triangle.

In fact, Kärkkäinen (2003, p. 20) suggests that subjectivity is a central feature in language, which takes on an organizing principle in discourse. She quotes Scheibman (2002), whose study shows that American English interactional discourse teems with “those subject-predicate combinations that permit speakers to personalize their contributions, index attitude and situation, evaluate, and negotiate empathetically with other participants.” In other words, expressions such as “I think”, “I believe”, and “I don’t know” take on a variety of discursive functions, including discourse organization (signaling alignment with what has been said, or retaining speaker turn); assessing a propositional content (epistemic values); or providing the source of information (evidential values). Kärkkäinen (2003, p. 47) even suggests, based on Holmes (1988, p. 25), that the prevalence of such verbal expressions in spoken English, compared with a relatively lower use of modals and quasi-modals, is due to speakers’ marked preference for “explicitly personalized markers of stance” in the form of a first-person pronoun, followed by a verb.

Such personalization is also common in spoken French, where speakers often start their speaking turn with moi je “me I” (see Béal 2010, p. 107), and express explicitly the relation they hold towards the propositional content by introducing the constituents in a concentrical order as in (1):

(1) J’ai ma sœur qui a quitté Paris; “I have my sister who left Paris”

This example starts with a localization of the agent (ma soeur; “my sister”) relative to the speaker (je; “I”), then moves on to the theme (a quitté Paris; “left Paris”), as shown by Culioli (1999, p. 105). In Culioli’s enunciation theory, the speaker’s subjectivity takes on a paramount role in the structuring and indexing of information; it is not only expressed through word order and syntactic structure, but also through the use of modality and deictics. Along the same lines, Biber et al. (1999, p. 971) observe that in English, stance markers often precede the clause containing new information. They consider that this initial position reflects the fact that stance markers’ main function is to create an interpretative framework for the proposition. Kärkkäinen (2003, p. 68), based on her own English data, confirms that in interactional discourse, stance marking often occurs before the propositional content, and she also observes that the most prevalent pattern is to express stance at the beginning of an intonation unit (which often, but not always, corresponds to a clause unit). She proposes the following default pattern for epistemic intonation units:

(2) (connective) + stance marker + proposition/clause fragment/prepositional phrase/noun phrase

She suggests that the reason for this archetypal utterance organization is that “recipients actively attend to the emerging structure of the speaker’s utterance”, so as to prepare their own response. In other words, stance markers at the beginning of an intonation unit function as an interactional signal that the speaker is expressing a subjective position, which the interlocutor will then use as a basis for alignment. Kärkkäinen also observes that initial stance markers often coincide with the start of turn and suggests that when an epistemic stance marker appears in another position (medial or final), it may have a specific interactive function, which she does not specify (but see Mullan 2010, 2012 for a discussion on the role of intonation to tease apart the different discursive functions of discourse markers).

We will now focus on research on “I think” and je pense to find out how these verbal stance markers are used in English and French, in an attempt to understand the constructs that are available to Anglophone learners of French, before offering a brief literature review of relevant L2 research.

2.2. Expressing Stance with Cognitive Verbs “I Think” and je pense

“I think” is generally considered to be one of the most frequent epistemic stance markers in English, particularly in a conversation context (Thompson 2002, p. 138; Baumgarten and House 2010; Aijmer 2009; Mullan 2010, 2012). According to Kärkkäinen (2006, p. 700), it can be viewed as a prototype of subjective expression, as it contains the first-person pronoun, “I”, and refers to cogitation (its basic semantic value).

When used as a stance marker, it takes on a variety of semantic values, that can be ordered along an assertiveness continuum, as follows in Figure 1:

According to Mullan (2010), the same semantic values can be attributed to je pense, although French and English native speakers’ usage might differ.3 Syntactically speaking, “I think” and je pense may both be used as independent clauses (IC), either as a short answer, or as a full clause, as in (10); they may also be used, as in (3), as matrix clauses (MC) introducing a content subclause; or they can be used as a parenthetical (P), either in medial or final position, as in (5).

The different stance values are exemplified in examples (3) to (10). Examples in French are taken from the interview data of the LANGSNAP corpus, while examples in English are taken from the PROLINGSA corpus4 (Leclercq et al. 2021), which is a longitudinal database consisting of a series of five interviews performed by two investigators (one French, the other American) with five French Erasmus students before, during, and after an academic stay abroad in a British or Irish university.5 The transcription conventions adopted in all examples were taken from the CHAT transcription manual (MacWhinney 2000).

2.2.1. Hedge

In (3), je pense is used as a hedge,6 or a mitigator by learner 111 to express tentativeness (Schneider 2007). The successive reformulations (<je voulais> [//] je pense que je voulais; “I wanted I think I wanted”) and the combination of je pense with a modal verb in the imperfect (a non-target-like form in this context, used instead of the conditional) suggest a lack of assertiveness about what the speaker really wanted. Although je pense acts as a mitigator to je voulais, it could also serve a pragmatic purpose of speech management (keeping the speaking turn and gaining time to formulate the utterance).

(3) FrL2, LANGSNAP, VA3, *111

*INT: et si tu pouvais changer quelque chose de ton année à l’ étranger

  “and if you could change something from your year abroad”

*INT: qu’est+ce+que ce serait? “what would it be? ”

*111: (…) hum <je voulais> [//] je pense que je voulais <d’ avoir plus des a(mis)> (hedge, MC)

“hum I wanted I think I wanted to have more friends”

*111: [//] de faire plus des efforts <de fait> [//] de faire les amis français

“to make more efforts to made to make more French friends”

*111: parce+que c’ est difficile.

“because it’s difficult”

(4) EngL1, PROLINGSA, CT2

*INT2: so is it student accommodation.

*INT2: or is this?

*C: yeah yeah in Eden Hall.

*INT2: &ah ok so i think. (hedge, MC)

*INT2: that’s.

*INT2: where melissa is too no?

In (4), the American interviewer, INT2, chooses not to appear too assertive. Although she knows Melissa lives in Eden Hall, she uses “I think” as a hedge to make her interlocutor, participant C, feel more authoritative about matters related to daily life on campus.

2.2.2. Epistemic Values

Epistemicity pertains to the domain of evaluation of a propositional content. In (5a), prior to departure, speaker 131 uses je pense in a matrix clause to evaluate content located in the future or counterfactual. Learner 137 uses it as a parenthetical in (5b), to express their opinion. Of course, it is difficult to clearly separate an epistemic value and an opinion value, as both are evaluative; however, when assessing in an irrealis context, as in (5a), speakers appear slightly less certain of their judgement. When je pense is placed next to an assertion in the present tense, as is the case in (5b), it is interpreted as more assertive and indicative of an opinion.

(5a) FrL1, LANGSNAP, Pre-test, *131

*131: et du coup je pense que ça va être assez difficile. “and so I think that is going to be quite difficult” (epistemic, MC)

(5b) FrL2, LANGSNAP, Pre-test, *137

*137: et c’ est une belle opportunité je pense. “and that’s a beautiful opportunity I think” (opinion, P)

The values of “I think” are quite similar: in (6) “I think” takes on an epistemic value of evaluating the chances of realization of the following propositional content., i.e., the speaker assesses the probability of her forgetting something as quite high.

(6) EngL1, PROLINGSA, AT2

*INT2: like i think. (epistemic, MC)

*INT2: i would forget.

*INT2: you know.

*INT2: what i mean?

As for (7), it illustrates the use of “I think” with an epistemic or evidential value: in this instance of reported speech, INT2, the American interviewer, quotes Pascale’s assessment of Caroline’s state of mind regarding her return to France. Based on her previous conversation with Caroline, Pascale inferred that she did not want to come back to France and would prefer to stay in Ireland.

(7) EngL1, PROLINGSA, CT5

*INT2: pascale told me.

*INT2: the last time she saw you.

*INT2: she said.

*INT2: i think caroline. (epistemic/evidential, MC)

*INT2: does not want to come back.

2.2.3. Evidential

As for evidential values, they are exemplified in French (8) and in English (9). In (8), je pense indicates that speaker 130 uses his or her recollections to answer the interviewer.

(8) FrL1, LANGSNAP, PT, *130

INT: ok et décris moi un événement significatif soit positif soit négatif qui s’ est passé pendant ton séjour. “ok and describe a significant event which might be positive or negative which took place during your stay”.

*130: euh (...) euh le jour <où j’ ai rentré> [//] où j’ ai rencontré mes deux coloc(ataire)s je pense. “er (…) er the day where I entered where I met my two flatmates I think” (evidential, P)

Finally, (9) provides a combination of the expression of opinion and inference. INT2, based on her visual perception of her interlocutor’s small room, inferred that the latter would not spend much time packing her things at the end of the academic year.

(9) EngL1, PROLINGSA, AT5

*INT2: i think. (opinion + evidential, MC)

*INT2: packing should be easy.

*INT2: i saw your room with only nine square meters

2.2.4. Opinion

As illustrated before, it is often difficult to make a clear-cut distinction between the different stance values (hedge, epistemic, evidential, or opinion); however, (10) features two clear instances of the expression of opinion. It also illustrates the interactional dynamic between the interlocutor, who provides an interactional cue (tu penses “you think”), and the uptake by the learner. The latter uses je pense twice, first as a short positive answer, and second to propose a more developed version of her views. This also permits her to align with her interlocutor.

(10) FrL2, LANGSNAP, PT, *130

*INT: oh d’accord et est-ce que tu penses que cette expérience va t’aider euh pour ton année en Angleterre? “Oh ok and do you think this experience will help you er for your year in England?” (cue)

*130: euh oui je pense “er yes I think” (opinion, IC, uptake)

*130: parce que euh je pense que ça sera plus facile de parler en anglais. “because er I think that it will be easier to speak in English” (opinion, MC, uptake)

2.2.5. Pragmatic Functions

In conversation, je pense and “I think” are also used as pragmatic markers, to signal the beginning of a speaking turn, align with a previous intervention, maintain the speaking turn or gain time to plan the rest of the message (see examples (11), (12), and (13)).

(11) FrL2, PT, *114

FrL1, LANGSNAP, *130

INT: d’ accord bon et pour finir euh si tu avais choisi d’ étudier aux Etats Unis au lieu de l’ Angleterre comment est+ce que tu penses que ta vie serait différente?(cue) “So well and to finish er if you had chosen to study in the US instead of England well do you think your life would be different?”

*130: euh (.) je sais pas trop. (…) “er I don’t know”

*130: il n’ y a pas vraiment de +... “there is not really…”

*130: c’ est que des jeunes. “it’s only young people”

*130: je pense que +... “I think that +…” (pragmatic, MC, uptake)

*130: je ne sais pas. “I don’t know”

(12) FrL2, PROLINGSA, AT1

*INT1: so let’s start with the personal basis.

*INT1: what do you think. (cue)

*INT1: this &heu stay abroad experience is going to bring to you?

*A: &heu i mean.

*A: i still live &heu with &heu my mother.

*A: so i think. (opinion, MC, uptake)

*A: it will be a good experience.

*A: to live alone.

(13) EngL1, PROLINGSA, AT2

*A: It’s ok for me.

*INT2: OK well I thi(nk) yeah I think it’s ok for us. (pragmatic, IC, alignment)

As illustrated in the previous examples, stance and pragmatic functions often overlap (see also Kärkkäinen 2003; Du Bois 2007; Baumgarten and House 2010; Mullan 2010), and it is often difficult to draw the line between the expression of stance and conversational strategies, such as expressing alignment with the interlocutor, or gaining time for discursive planning.

2.3. Use in English and French Conversation

While “I think” has been widely described as one of the most frequent stance markers in English conversation, there is very little data on the use of je pense. In one of the few contrastive studies on opinion markers in French and English, Mullan (2010, pp. 119–23) claims that “I think” and je pense take on similar interactional functions, in fairly similar proportions. Based on a dataset of ten hours of French and English conversations, audio recorded in Australia between August 2000 and September 2002, Mullan found that two thirds of her occurrences were discourse markers with a pragmatic function, while the remaining occurrences were epistemic instances, mostly to express speaker opinion. She also suggests that “I think” is regularly used with evidential values, while she does not observe such use with je pense. She attributes this to cultural differences: the French conversation style is seen as confrontational, with French speakers granting particular importance to the expression of opinions and feelings; while Australian speakers are believed to prefer to give fact-based information about events and to restrain from expressing an opinion (Mullan 2010, pp. 19, 149–50). Mullan (2012, p. 203) proposes that native Australian speakers use “I think” mostly as a hedge, “to positively assert or claim a particular viewpoint towards a proposition while disclaiming knowledge of it”. She considers that je pense is used by French speakers to express “an opinion based on reflection, where the speaker wishes to positively assert or claim a particular stance towards a proposition”.

In the next subsection, we present an overview of relevant studies on the L2 acquisition of epistemic stance markers to find out what they can contribute to the larger SLA debate.

2.4. L2 Acquisition of Epistemic Stance Markers

As mentioned earlier, there is very little research on the acquisition of epistemic stance in French or English, and even less work on the use of je pense by learners of French, whatever the context of acquisition (SA vs. guided).

2.4.1. Discursive Functions and Development of Pragmatic Competence

Pragmatic competence is generally considered to cover two subcomponents: pragmalinguistics and sociopragmatics (Thomas 1983; Xiao 2015; Pérez Vidal and Shively 2019). The former focuses on the speaker’s knowledge of the pragmatic function of a linguistic form in interaction, while the latter focuses on the ability to express, linguistically, the taking into account of context, relationships between the participants, their degree of shared knowledge, and social rules and norms (Delahaie 2015, p. 4). In a panoramic paper on the impact of SA on L2 acquisition, Howard (2021) highlights the fact that one of the areas where learners progress most is sociopragmatic competence. Based on an extensive literature review, Howard concludes that progress in the acquisition of L2 grammatical features does not appear to be systematic, and is largely dependent on learners’ proficiency before departure, as well as on their motivation, their capacity to develop social networks, their length of stay, and the intensity of contact with L2 input.

As for the development of the pragmalinguistic competence, it is also variable, as exemplified in Aijmer (2009), a comparative study on how Swedish learners and native speakers of English use “I don’t know” in interviews. This paper showed that learners predominantly used “I don’t know” as a speech management tool, while native speakers used it to ask indirect questions as a way to express politeness. Using a longitudinal design, Grieve (2013) studied the development of the use of the discourse marker “like” by teenage German students spending five to ten months in a host family in Australia. She concluded that length of stay positively impacted the adoption of native-like patterns of use, because the participants who stayed longer were more committed to integrating into the native-speaker community and were therefore more likely to adopt their language use.

Regarding the use of “I think”, Salsbury and Bardovi-Harlig (2000), in their paper on the acquisition of modality in L2 English, situated within an interlanguage pragmatics approach, presented a twelve month longitudinal study in which eight learners of English (all beginners at the onset of the project, and with a variety of different source languages) were enrolled in an intensive English program at Indiana University. The database consists of various types of recorded interactions: learner to learner and learner to native speakers. The authors analyzed instances of oppositional talks in conversation extracts. Their results highlight the fact that low-level and more advanced learners make a frequent use of “I think” and “maybe” to mitigate their disagreements in oppositional talk, rather than modal verbs. Such frequent use of “I think” was also identified by Fordyce (2009) in his paper on the use of epistemic forms by 41 elementary and intermediate Japanese learners of English to express opinion, in written (essays) and conversational tasks. Fordyce (2014, p. 8) argues that such predominance of epistemic stance marker “I think” at lower levels of acquisition could be due to its lack of semantic ambiguity, and its syntactic flexibility (it can be placed in an initial, medial, or final position in the utterance).

Using conversation data between dyads of native Australian and French speakers, Mullan (2012) analyzed the use of “I think”, je pense, je trouve, and je crois by native speakers and learners of French and English. Her results highlighted the fact that Australian English speakers used “I think” predominantly (66%) for organizational functions (alignment, online planning, or marking a boundary in discourse …), even if they also used it to express their opinion (33.6%). French native speakers made a more balanced use of organizational (53%) and opinion (46%) functions for je pense. As for learners of French, they mainly used je pense for organizational functions (87%), and almost never to express their opinion (13%), which, according to Mullan (2012, p. 216), suggests a “transference” from L1 patterns.

Overall, L2 studies on the development of the use of cognitive verbs show that learners and native speakers do not always use cognitive verbs with the same discursive functions, even after SA. Under certain conditions, SA nevertheless seems to provide favorable conditions for the development of the pragmalinguistic competence, although individual factors, such as learners’ identity, motivation for social integration, age, goals, or living arrangements, may affect the learning outcome (see Xiao 2015 and Pérez Vidal and Shively 2019 for a review).

2.4.2. Inter-Individual Variation and SA

While SLA as a field has often been more interested in finding common acquisitional paths and patterns than in highlighting individual variation, recent research has sought to make up for this through research protocols taking into account individual variation and style. In their seminal 2017 paper, Gablasova and colleagues present a corpus-based, cross-sectional analysis of the speech of 132 advanced L2 speakers of English from a variety of linguistic backgrounds. The speakers performed one monologic and three interactional tasks, and the authors analyzed the way they used adverbial, adjectival, and verbal epistemic forms (including “I think”) in the different types of tasks. Their results found marked differences in the use of epistemic stance markers between monologic and dialogic tasks, as well as “considerable inter-speaker variation, indicating the existence of individual style in the use of epistemic markers” (Gablasova et al. 2017, p. 613). The question of style is also at the heart of Liao (2009), a mixed method variationist study of the use of discourse markers in interaction by Chinese L2 speakers of English in a SA context. The results stress the variability in the use of discourse markers, which the author attributes to stylistic preferences. Liao (2009, p. 1) also defends the view that the way L2 speakers use discourse markers is often linked to their sociolinguistic competence and their “multi-faceted social identities”, a view echoed by Grieve (2013, 2015). Individual variation was also seen as a key element in the linguistic outcome of a SA period in Dewaele and Dewaele (2021): their paper compared actual and self-perceived proficiency gains of Anglophone learners of L2 French after an Erasmus stay in a francophone country. They concluded that there were significative gains in receptive and productive skills, including grammar and vocabulary knowledge, but the participants showed no significant change in pronunciation after SA. The authors nonetheless emphasized the “considerable variation between students in terms of how much progress they made, with some not making any measurable improvement, while others made considerable gains” (p. 13). The development of a strong francophone social network was seen as a key factor in language development.

Finally, Leclercq (Forthcoming), in a study on the use of “I think” by French learners of English, based on data from the PROLINGSA project (a longitudinal study of five Erasmus students followed over the course of an academic year abroad in the United Kingdom or Ireland) revealed that the five learners displayed distinct stylistic preferences in the use of “I think” (with two main profiles: moderate and intensive users) and that their use of such a marker evolved over their nine month SA period. A qualitative analysis of the oral production of two intensive users highlighted the connection between the discursive functions with which “I think” was used and the evolution of “their degree of self-confidence, and the way they positioned themselves in discourse relative to others.” In other words, this paper illustrated how the use of epistemic stance marker “I think” revealed the increasing self-confidence of the learner and their increased ability to use epistemic markers with a pragmatic value.

2.4.3. Summary of L2 Research

To sum up, previous research suggests the following:

(1) That the use of epistemic stance markers is variable according to the type of discourse (narrative or conversation) (Gablasova et al. 2017; Leclercq Forthcoming).

(2) That “I think” is found in conversation data with evidential, opinion, and organizational values, while je pense is mostly used to express opinions (Mullan 2010, 2012).

(3) That SA immersion favors the development of pragmatic competence (Xiao 2015; Pérez Vidal and Shively 2019) and a gain in self-confidence (Compiegne 2020; Leclercq Forthcoming).

(4) That there is high interpersonal variability in the way these markers are used, thereby suggesting the importance of personal style (Liao 2009; Grieve 2013, 2015; Gablasova et al. 2017; Dewaele and Dewaele 2021; Leclercq Forthcoming).

Against such a backdrop, we wish to investigate to what extent L2 French learners’ use of the stance marker je pense, reflecting a more or less assertive stance, evolves over the course of an academic year abroad.

More specifically, we seek to find out whether immersion favors the adoption of native-like patterns of use regarding the expression of epistemic stance in French L2 conversation, and the development of pragmalinguistic competence.

2.5. Research Questions

(1)

To what extent does the use of je pense by Anglophone learners of French reflect an evolution in the way they mark stance in conversation?

(2)

To what extent do the pragmatic uses of je pense reflect learners’ gain in pragmalinguistic competence?

(3)

To what extent is learner’s use of je pense a matter of style?

3. Materials and Methods

3.1. Dataset

We used data from the LANGSNAP project, which is a longitudinal study spanning over 21 months, including 9 months abroad, during which the researchers tracked the linguistic development of 29 learners of French. Our database includes 10 French native speakers and 26 Anglophone learners of French, with pre-test (PT) and Visit Abroad 3 (VA3) semi-guided interview data (we excluded participant 100, a highly proficient French/English bilingual, and participants 108 and 126, as their mother tongues (L1s) were Finnish and Spanish, respectively). All learner participants were adult students enrolled in a British university to study languages, with a special focus on French. As part of their undergraduate curriculum, they had to spend the third or fourth year of their programme abroad in a French-speaking country. Among the 26 learners of French, six were Erasmus exchange students, 15 were teaching assistants in a French secondary school, and five were workplace interns. All report having studied French for a long time, stretching between 8 and 15 years, in a variety of contexts, including French instruction in secondary school. The mean age for this group was 20 at PT, and the group included 2 males and 24 females.

The French control group (n = 10) were all Erasmus students spending an academic year in a British university as part of their undergraduate curriculum. They were recorded shortly after their arrival on campus. Their mean age at PT was 20, and the group included 3 males and 7 females.

3.2. Transcription and Coding

The recordings were transcribed on CLAN by the LANGSNAP team. The transcriptions are available from the online LANGSNAP repository. We then identified the occurrences of je pense produced by the target participants (we excluded occurrences produced by the interviewers from the analysis) and coded them for discursive function (stance marker, pragmatic marker, or cogitation verb). Stance markers were further subdivided according to their semantic values, along an assertiveness continuum: opinion, evidential, epistemic, hedge. We then coded for the type of clause, as follows:

• Independent clauses (IC): short answers and IC.
• Parentheticals (P): Je pense in middle or final position.
• Matrix clause (MC): je pense as a matrix clause followed by complement pronoun que introducing a complement clause.

Finally, we coded for interactional cue. When a question asked by the interviewer contained tu penses—“you think”—it was coded as “cue”, and the subsequent je pense, produced by the participants, were labeled “uptake”, so as to identify the occurrences which were prompted by the interlocutor, as in (14).

(14) FrL1, LANGSNAP, PT, *130

*INT: oh d’accord et est-ce que tu penses que cette expérience va t’aider euh pour ton année en Angleterre? “oh, OK, and do you think that this experience will help you er for your year in England?” (cue)

*130: euh oui je pense. “er yes I think” (uptake)

*130: parce que euh je pense que ça sera plus facile de parler en anglais. “because er I think it will be easier to speak in English” (uptake)

3.3. Data Analysis

We provide a quantitative analysis of the dataset, starting a description of “je pense” items produced by native speakers and learners (PT and VA3), starting with its frequency of use, its syntactic distribution, and its occurrences in an interactional context, before analyzing its discursive functions. We then provide a more detailed analysis of individual variation, and examine interviewer questions, so as to answer our research questions.

4. Results

4.1. Quantitative Analysis

Table 1. Data description.

Group	Length of Productions (# Token Items)	je pense # %
FrL1 n = 10	M = 1491.7 SD = 515.83 Range 779–2581	68 4.56
FrL2 PT n = 26	M = 1261 SD = 398.97 Range = 679–2440	316 25.05
FrL2 VA3 n = 26	M = 1262.15 SD = 601.10 Range = 600–3398	228 18.06

provides information relative to the language use of the participants, including their use of je pense. Length of production (in number of token items) was measured using the freq command of the CLAN program, which lists and counts the different types and tokens used in the language produced by each of the participants (FrL1, FrL2 PT, and FrL2 VA3); interviewers’ interventions were excluded from the token count. The 10 French native speakers produced a total of 14,917 tokens. Native speakers appear to have globally produced longer conversations than learners, but all groups display individual differences in production length, as illustrated by SD and range figures. A look at PT and VA3 indicates that, although mean length remains the same, SD figures increase and range widens between PT and VA3. Inter-individual variation in conversation length were indeed bigger at VA3 than at PT, suggesting an impact of SA on the development of individual conversational style, with some individuals becoming more talkative at VA3.

Regarding frequency of use, the percentage values in Table 1 highlight the gap between the relatively scarce use of je pense by French native speakers (4.56%) and the learners, who use this stance marker at a rate of 25.05% (PT) and 18.06% (VA3) of the mean number of words in their productions. This could be due to the fact that “I think” is widely used by native speakers of English in similar contexts. Positive transfer, therefore, may have occurred even before the learners embarked on their SA. Such overuse is in line with the findings of (Fordyce 2009), especially at the earlier stages of L2 development.

A statistical test to examine the relationship between group membership and the frequency of use of je pense was conducted in the three subgroups of the dataset (FrL1, FrL2 PT, and FrL2 VA3). Because this dataset was significantly skewed, we conducted a Kruskal–Wallis test in order to determine whether group membership was significantly related to the percentage of occurrences of je pense produced. The results showed a significant difference between the mean ranks of the FrL1 and FrL2PT groups—H(3) = 7.896, p = 0.01929. This suggests that at PT level, learners used je pense differently from native speakers, while percentage of use at VA3 shows no significant difference with either FrL1 (H(3) = 0.605, p = 0.4364) or FrL2PT (H(3) = 4.1007, p = 0.04287) groups. The fact that there is no significant difference between PT and VA3 suggests no change over time. However, the fact that the learners are significantly different from the NSs at PT but not VA3 suggests some change.

We will now have a closer look at the syntactic position and discursive functions that je pense takes on in the oral production of L1 and L2 French speakers.

Table 2. Syntactic distribution of je pense in the database—descriptive statistics.

	FrL1	FrL2PT	FrL2VA3
IC
#	10	33	27
%	14.70	10.44	11.84
M	1	1.27	1.04
SD	1.33	1.31	1.15
Range	0–4	0–4	0–4
# speakers	5/10	15/26	15/26
MC
#	52	252	168
%	76.47	79.75	73.68
M	5.2	9.69	6.46
SD	2.94	6.53	7.59
Range	0–9	0–23	0–28
# speakers	9/10	23/26	20/26
P
#	6	31	33
%	8.82	9.81	14.47
M	0.6	1.19	1.27
SD	0.84	1.72	1.64
Range	0–2	0–7	0–5
# speakers	4/10	16/26	12/26
Total #	68	316	228

presents the syntactic distribution of all occurrences of je pense in the dataset. It clearly shows that native speakers and learners, irrespective of the time spent abroad, favor its use as a matrix clause (je pense que). Native speakers’ second choice is its use as an IC, either as a short answer, as in (14), or as the main verb of a full IC. In those two categories (IC and MC), je pense features in initial position in the clause, sometimes after a “connective”, as proposed by Kärkkäinen (2003, p. 47) (see (2)). Learners’ use of IC is relatively stable from PT (10.44%) to VA3 (11.84%). Parentheticals present a different picture. They are scarcely used by native speakers (6 occurrences, 8.82%), while learners use them increasingly between PT (9.81%) and VA3 (14.47%). A closer look at the data reveals variability in the number of participants’ use of P. Finally, while most speakers use je pense as an MC at least once in their conversation, range figures indicate that there is variation in the quantitative way in which speakers and learners (especially at VA3) use je pense as an MC.

Another notable result is provided by the analysis of the number of je pense occurrences triggered by a conversational cue (the interviewer asking the participant ‘tu penses que …’), see

Table 3. Occurrences of je pense following an interactional cue.

	FrL1	FrL2PT	FrL2VA3
Cues
#	26	56	29
Uptake
#	20	91	29
Cue to uptake ratio	1.3	0.61	1

.

The cue to uptake ratio indicates that at PT, learners do not systematically rely on interviewer cues to use je pense. At VA3, the ratio indicates that each time the interviewer used tu penses, uptake occurred in at least one utterance of the answer. This suggests progress at VA3 towards more native-like patterns of alignment, and thereby a gain in pragmalinguistic competence. These results point to divergent conversational strategies by learners and native speakers and tend to indicate an impact of SA on the conversational use of je pense.

4.2. Discursive Analysis

In this section, we focus on the discursive functions adopted by our two groups of participants (FrL1, FrL2 at PT and VA3), in an attempt to find out to what extent je pense was used in an assertive way to express opinion and what its other uses were.

4.2.1. Discursive Functions

Figure 2 shows that French native speakers mainly use je pense as a means to express stance (opinion 75%, epistemic meanings 14.7%, evidentiality 4.4%, total 94.1%). Pragmatic uses and expression of cogitation are marginal (2.9% each). This is in line with the findings of Mullan (2012).

As for learners, they offer a different picture: at PT, stance meanings (epistemic, evidential, and hedge) cover 78.2% of all occurrences, with a tiny proportion of evidentials (3.2%), while pragmatic uses correspond to 20.3% of all occurrences. At VA3, the proportion of stance markers decreases (70.6%), while the proportion of pragmatic uses increases to 28.9%. A closer look at the stance data reveals that learners appear to take a less assertive stance than native speakers, even if the expression of opinion develops slightly between PT and VA3 (opinion: FrL2PT 50.3%, FrL2VA3 56.1%). Development of pragmatic uses suggest learners employ je pense as a speech-management tool, to gain time to plan the rest of the utterance, as in (11). This could be due to the difficulties learners face to express themselves in a foreign language. Because learners may lack fully automatized access to French linguistic resources, they may develop discursive strategies to keep their speaking turn while retrieving the linguistic means to formulate their thoughts. It might also be a transfer effect from their L1 patterns of use (Mullan 2012 found that organizational uses dominated in the speech of native Australian speakers).

4.2.2. Individual Variation in Pragmatic Uses

There is some variation in the development of pragmatic uses, as observed in Leclercq (Forthcoming), where the author found that only 1 of the 5 Erasmus students of the PROLINGSA dataset had developed pragmatic uses of “I think” over the course of the SA, at a rate of 12.5% of all occurrences of “I think”. The other four participants used this marker with stance meanings (evidential, hedge, opinion) only. The author concluded that individual preferences may vary and that learners developed their own style over SA. In the LANGSNAP data, the picture is quite different: two of the native speakers of FrL1 use je pense just once with a pragmatic meaning, against 21 of the 26 L2 French learners at PT, and 12 of the learners at VA3. At PT, these 21 learners all make a relatively moderate use of the device (M = 3.05, SD = 2.29, range 1–8), while VA3 sees the emergence of a different pattern, with fewer users but a more intensive use (M = 5.5, SD = 4.79, range 1–17). There are 2 profiles that can be distinguished: 10 moderate users (1–9 times), and 2 intensive users (12–17 times). In other words, while the global percentage of pragmatic uses increases from PT (20.3%) to VA3 (28.9%), individual usages evolve differently, with a lot of moderate users at PT, and a few more intensive users at VA3.

Regarding epistemic uses, learners use them considerably more at PT than at VA3 (respectively, 24.7%, with 22 out of 26 users, range of use 1–11; and 3.1%, corresponding to only 5 users, range of use 1–2) and also much more than native speakers (14.7%, 5 out of 10 users, range 1–4). This could be linked to the development of stylistic preferences on the part of learners over the course of their SA period, but it could also be a consequence of the type of questions asked at PT, when students are still invited by the interviewer to project themselves towards future events, as in (15), where the interviewer asks learner 101 about her plans for her year in France:

(15) FrL2, LANGSNAP, *101, PT

*LGR: Très bien, qu’est-ce que tu vas faire l’année prochaine? “Very well, what are you going to do next year?”

*101: ehm j’ai décidé de choisi d’être ehm une assistant dans une école élémentaire (…) donc je pense que euh je pourrais ehm euh aider aider les enfants les jeunes pour ehm euh pour apprendre apprendre des des nouvelles ehm euh les moyennes nouvelles pour apprendre une langue une langue. “ehm I decided to choosen to be ehm an assistant in an elementary school (…) so I think that er I could ehm er help children the young ones to ehm er to learn learn some some new ehm er new means to learn a language language” (epistemic)

In (15), 101’s French is hesitant, with numerous dysfluency markers (filled pauses, self-repetitions, and reformulations). The first part of her answer is assertive as it starts with verbs expressing volition and decision (j’ai décidé/choisi). The second part is more tentative: 101 uses je pense with an epistemic value, followed by a modal in the conditional mode (je pourrais), thereby indicating the counterfactual value of the proposition, and the lesser degree of certainty of the speaker. Such lack of assertiveness is also reflected in the numerous dysfluency markers, which could be due to a difficulty in planning the answer.

While having a different set of questions at PT and VA3 provides rich, ecological data and allows for fine-grained analyses of individual trajectories, it leads to data elicited at those two points in time being difficult to compare. Our results should be interpreted with this limitation in mind.

4.2.3. Variation in Forms

In Table 2, we subsumed all the different versions of je pense under the same label to provide a broad-range picture of the phenomenon. However, as

Table 4. Type of je pense expressions in the database.

Types of je pense	FrL1	FrL2PT	FrL2VA3
j’ai pensé—“I have thought” (passé composé)	0	0	1
je dois penser—“I must think” (present)	0	1	0
je ne pense pas—”I don’t think” (present)	0	10	0
je pensais—“I thought” (imperfect)	0	0	2
je pense—“I think” (present)	68	296	214
je pense pas—”I don’t think” (present)	0	7	8
je peux pas penser—”I can’t think” (present)	0	0	1
je penserai—”I will think” (future)	0	1	0
je veux penser—”I want to think” (present)	0	0	1
*j’en pense—“I object pronoun ‘en’ think” (present)	0	0	1
*ø pense—”think”	0	1	0
TOTAL	68	316	228

illustrates, FrL1 speakers only use je pense, while learners use a variety of je pense-based expressions, including forms with a modal verb (dois, peux, veux) and inflected forms (passé composé, imperfect, future).

The use of a diversity of forms (even if je pense largely dominates at PT and VA3 alike) suggests that some learners have not fully grasped the formulaic nature of je pense in the French input, even after a lengthy SA.

4.2.4. Analysis of the Role of Interviewer Questions in the Use of je pense

While it has been shown that native speakers and learners display different semantic preferences regarding the use of je pense, it is not clear to what extent this is due to crosslinguistic factors,7 to proficiency issues (triggering a greater use of pragmatic meanings), or to conversation factors, and particularly to the role of interviewer questions on the type of discursive function attributed to je pense. In this section, we focus on the interview script the interviewers relied on in the LANGSNAP project.

The same interview scripts were used for all PT and VA3 interviews, although there are some differences in formulation and in the total quantity of questions asked by the team of six interviewers to the 37 participants (native speakers and learners. We focus on the PT and VA3 interviews of learner 101 to analyze the interview scripts.

At PT, learner 101 is asked 29 questions, including 7 past-oriented ones in the passé composé: (“alors pourquoi as tu choisi d’ étudier les langues vivantes?”—”so why did you choose to study modern languages?”) and 12 future-oriented ones (“à quelles differences est-ce que tu t’attends?”—”what kind of differences do you expect?”; “qu’est-ce que tu feras pour te détendre?”—”what will you do to relax?”; “qu’est-ce que tu vas faire l’année prochaine?”—”what are you going to do next year?”).

At VA3, 17 questions were asked to 101, including 6 in the passé composé, 2 in the conditional (“Est-ce-qu’il y a des choses que tu aurais voulu savoir avant d’arriver en France de la vie française?”—”Are there things that you would have liked to know before arriving in France about French life?”; “Si tu pouvais changer quelque chose de ton année à l’étranger qu’est-ce que ce serait?”—”If you could change something to your year abroad what would it be?”), and two future-oriented (“As-tu des buts pour les prochains jours prochaines semaines?”—”Do you have goals for the next days next weeks?”; “Qu’est-ce qui te manquera le plus de vivre à City?”—”What will you miss most about living in City?”)

The higher proportion of future-oriented questions in the PT interview might have triggered a higher number of epistemic occurrences, where the speaker assesses the chances of realization of some events. At VA3, the learner was invited by the interviewer to take a retrospective glance on the SA period. This brought about a series of recollections on the part of the speaker, who often used je pense with an evidential value to express the recollection process and the inferences based on recollections, as in (8).

The analysis of the interview script for 101 PT and VA3 interviews, in relation with Figure 1, invites us to consider that there is a link between the type of questions asked and the type of stance value attributed to je pense by learners. But as the high percentage of pragmatic values shows, je pense also appears as a widespread tool to retain the speaking turn, to gain time to plan the rest of the utterance, and to align with the interviewer.

5. Discussion and Conclusions

In this paper, we sought to find out how the use of the stance marker je pense by Anglophone learners of French evolved over a period of nine months abroad in a French-speaking country. In particular, we wanted to uncover whether the use of this marker was a matter of individual style, whether it reflected changes in the way learners express stance (in a more or less assertive way), and whether it demonstrated a gain in pragmalinguistic competence.

The first question addressed the following: to what extent the use of je pense by Anglophone learners of French reflected an evolution in the way they marked stance in conversation. We analyzed the syntactic choices made by our participants, before focusing on the discursive functions they attributed to je pense.

Regarding syntax, the prevalence of MCs among native speakers and learners confirms Kärkkäinen’s (2003) default syntactic pattern (see (2)), where the privileged position for a stance marker is at the beginning of the intonation unit or the utterance. It constitutes both an expression of the speaker’s perspective and an interactional cue. From a functional perspective, making adequate morphosyntactic choices in an L2 is a key communicative competence as it makes the clause content interpretable by the interlocutor. However, learners also need to learn what the native patterns of use are regarding discursive values. In other words, the morphosyntactic competence is not sufficient to guarantee learners will communicate in a fully efficient way. They also need to develop pragmalinguistic competence, with SA globally appearing as a favorable context of development (Grieve 2013, 2015).

Regarding the acquisition of pragmalinguistic competence, our results show that learners have understood how central the expression of subjectivity in conversation is but they are globally less assertive than native speakers. The marking of opinion is the main function of je pense in all groups, even if learners and native speakers also use this verbal expression with other discursive functions, particularly epistemic assessment, in relation with future-oriented interviewer questions. Learners display little evolution regarding the expression of opinion between PT and VA3, thereby indicating that a lengthy stay abroad period did not foster a boost in assertiveness. The greater assertiveness shown by French native speakers, who mostly use je pense to express their opinions, is in line with the observations of Béal (2010) and Mullan (2010, 2012). Mullan finds French speakers to be more assertive and straightforward in the expression of their opinions (she even uses the word “confrontational” to describe the typical French attitude in conversation) than Australian English speakers, who tend to tone down their opinions to avoid clashes with their interlocutors. It is an observation that Gąsior (2015) holds to be true of Irish English speakers as well. In our dataset, Anglophone learners of French also appear less assertive than French speakers, as illustrated by the greater use of epistemic uses by learners of French at PT. This might reflect a transfer from native patterns of use, as suggested by Mullan (2012), since learners use je pense in a more tentative way than French natives; nine months abroad may not have been sufficient for learners to modify their deeply engrained cultural scripts regarding the expression of opinion (Mullan 2010, 2012; Gąsior 2015). Future comparative research on the expression of opinion by learners of different source languages might shed new light on the possibility to reconceptualize such form–function mappings over a prolonged SA period.

Next, we examined the extent to which the pragmatic uses of je pense reflect learners’ gains in pragmalinguistic competence. All in all, we found interactional effects (a link between the type of question and the type of discursive function of je pense, and the alignment with the interviewer) for all groups of participants. Yet, although we observe a clear development of pragmatic uses of je pense by Anglophone learners of French, they do not behave like French native speakers. While native speakers and learners both use je pense in conversation as an interactional device, signaling some sort of alignment with the interviewer, learners also use pragmatic je pense as a strategy to retain their speaking turn and gain time to plan their discourse. This may be due to a lack of fully automatized access to French linguistic means. It is not clear whether the increase in percentage of pragmatic use between PT and VA3 reflects a gain in pragmalinguistic competence (as illustrated with the more target-like cue to uptake ratio at VA3), or to new learner communication strategies. The development of pragmatic uses could be linked to the development of fluency, a phenomenon which was observed for L2 English by Aijmer (2009), in her study of “I don’t know”, and by Liao (2009, p. 1320), in her analysis of discourse marker “yeah” for self-repair after a disfluency episode. This is a phenomenon which we would like to explore in future research.

Our last question focused on the extent to which learners’ use of je pense is a matter of individual choice. Our results contribute new insights into epistemic stance marking in French, highlighting the high degree of inter-individual variation in the use of je pense, especially by learners. Some individuals never use it at all, while others seem to make it a regular feature of their conversation. Some use this marker at PT but not at VA3, and vice versa. This emphasizes the importance of personal style in the use of epistemic stance markers (Kirkham 2011) in relation with its development during SA, in line with Gablasova et al. (2017) and Leclercq (Forthcoming), and invites future case study research to investigate the complex factors at stake. In particular, we intend to study the role of intonation in teasing apart the different discursive functions of epistemic markers je pense and “I think”, as advocated by Bybee and Scheibman (1999), Mullan (2010, 2012), or Edmonds et al. (2021).

To conclude, our study suggests that, in spite of an extensive SA, learners have not fully acquired the favorite patterns of use of French native speakers, but they seem to have gained new interactional strategies to retain their speaking turn and gain time to formulate their thoughts, while developing their own style to express their stance.