Managing the Flow of Talk: A Longitudinal Case Study of the Multiword Expression det er sant in L2 Norwegian Interactions

Abstract

Drawing on recent scholarship integrating usage-based linguistics (UBL) and conversation analysis (CA) in the investigation of second language development, this paper reports on a microanalysis tracing one adult learner’s recurring, increasingly frequent, and diverse use of the multiword expression det er sant (it/that is true) (DES) in L2 Norwegian interactions over a time span of four months. While the use of DES did not undergo syntactic changes, the analysis revealed that the functions of the expression diversified as DES became more frequent in use. Initially functioning primarily as a means of expressing agreement, DES was increasingly used to preface disagreement and indicate the closing of self-initiated repair sequences. This functional diversification was accompanied by increasing morphophonological reduction indicating an ongoing process of routinization. We argue that these changes, also when non-linear and not complying with the target norm, allowed the learner to participate more actively in managing the flow of talk. The study provides insight into how interaction shapes L2 development and the role of education in providing access to situations that foster increased membership in the speech community.

Introduction

Over the past few decades, several calls have been made for researchers to bridge the gap between social and cognitive approaches to second language learning (e.g. Firth and Wagner 1997; De Bot et al. 2007; Eskildsen 2009; Hulstijn et al. 2014; Douglas Fir Group 2016). A recent research proposal responding to this call is usage-based CA-SLA.¹ Combining insights from UBL, second language acquisition (SLA), and conversation analysis (CA), usage-based CA-SLA offers both an overall theoretical understanding of language and language development and a methodology allowing language to be studied longitudinally at the microlevel as a socially situated phenomenon (e.g. Eskildsen 2012; Pekarek Doehler and Eskildsen 2022).²

Usage-based SLA and CA-SLA share the view that language learning is dependent on and shaped by social activity (Beckner et al. 2009; Pekarek Doehler and Eskildsen 2022). Primarily attending to grammatical development, UBL presumes a developmental trajectory from lexically fixed multiword expressions through semi-fixed patterns to fully productive constructions (Langacker 2000), known as the exemplar-based learning trajectory (e.g. Pekarek Doehler and Eskildsen 2022). In CA-SLA, the accomplishment of social actions is promoted as a key research objective. While investigating language use from a genuinely emic perspective (Kasper and Wagner 2011), longitudinal CA focuses on how interaction fosters change in the way linguistic resources are used to perform social actions. Recent research has proposed a distinct developmental trajectory for the routinization of linguistic material to achieve interactional goals in an L2 grammar-for-interaction (Pekarek Doehler and Skogmyr Marian 2022).

Drawing on recent scholarship integrating UBL and CA-SLA, in particular Pekarek Doehler and Skogmyr Marian (2022), the present study tracks one adult learner, Linda, and her use of the multiword expression (MWE)³det er sant (DES; it/that is true) in biweekly conversations over the course of 17 weeks as she grows more proficient in L2 Norwegian conversations. At the time of data collection, Linda was an intermediate-level learner permanently settled in Norway, following a class designed for progressing from the A2 level to B2, an advancement characterized by increased lexical variation and idiomaticity, as well as grammatical complexity and interactional fluency, including the management of conversation and turn-taking practices (Council of Europe 2020). Throughout the period, DES was applied with growing token frequency, yet according to our data, the MWE never varied in terms of modifiers, tense, or word order (Horbowicz et al. 2020). Low type frequency may signify a lack of grammatical development; yet, combined with an increase in token frequency implying entrenchment and growing automaticity, DES attracted our attention as it seemed to constitute an important interactive device in Linda’s L2.

In this paper, we explore how the interactional functions, as well as the sound production of DES, change over time in Linda’s L2 interactions. Interactional functions involve building a relationship with the interlocutor (such as showing affiliation), while changes in an expression’s functional range, that is, functional diversification, may be seen as part of the learner’s emerging L2 grammar-for-interaction, explained as ‘patterns of language use that serve to build social actions and manage the organizational infrastructure of communicative interaction’ (Pekarek Doehler and Skogmyr Marian 2022: 24). We formulated the following research questions (RQs):

1. In which interactional contexts does Linda use DES in her L2 Norwegian?

2. To what extent does Linda’s use of DES change or expand its functional range and formal realization over time?

The study is inspired by Pekarek Doehler and Skogmyr Marian’s (2022) recent contribution outlining a developmental trajectory for the L2 grammar-for-interaction. It takes a form-to-function approach, that is, the point of departure is the linguistic resource in interaction rather than a specific context or social action (see also Deppermann and Pekarek Doehler 2021). In the analysis, we break down the data into two phases marked by an abrupt shift in frequency in the middle of the period. A sudden rise in frequency may signal a transition point in a learner’s L2 system that possibly precedes qualitative changes in form–function mappings (e.g. Verspoor et al. 2008).

The article is structured as follows. First, we outline the syntactic structure and meaning of the focal construction det er sant. Second, we review the literature and situate the study within the theoretical landscape of usage-based SLA and CA-SLA. Finally, we move on to present the data and analytical procedure before embarking on the analysis.

Background

Focal construction det er sant (DES)

DES, treated here as an MWE, belongs to a group of syntactically similar expressions designating epistemic meaning in Norwegian. The expression consists of three elements: the third person pronoun det, the finite present tense form of the verb være (be), er, and the epistemic adjective sant, inflected for neuter gender. The semantic value of the adjective is ‘true’, but depending on the context, it may also be translated as ‘right’. In the typical Eastern Norwegian dialect spoken in the region where Linda resides, a word-by-word pronunciation of DES would be /de ær sɑnt/. However, following general pronunciation patterns in spoken Norwegian, the present tense/r/ is normally deleted in unstressed syllables (Kristoffersen 2000: 311), and depending on the stress pattern and context, the vowels in det er may also be reduced, resulting in/dæ/. When the vowels are reduced, <rs > may additionally merge into /ʃ/. Hence, the most common realizations of det er sant include /deæˈsɑnt/, /dæˈsɑnt/, and /dæˈʃɑnt/.

The first two words in the string, det er (it is), constitute a highly frequent collocation in Scandinavian languages (Bartning and Hammarberg 2007). The pronoun det can refer anaphorically to preceding discourse (corresponding to the English ‘that’), but it can also act as a formal subject (corresponding to ‘it’), requiring a finite clause complement. This formal ambiguity is paralleled in other realizations following the same pattern—that is, det er + adjectives with epistemic meaning, such as mulig ‘possible’, sannsynlig ‘likely’, riktig ‘right’, or klart ‘clear’ (Faarlund et al. 1997: 347). However, the different lexical realizations of the pattern vary in frequency. A search in NoTa⁴ reveals that klart and sant are by far the most commonly used complements for the construction. The corpus data also reveal that 79 per cent (MWE: det er sant) and 86 per cent (MWE: det er klart) of all occurrences appear unmodified, which indicates the pattern’s frequent use as one unit in spoken Norwegian. This does not imply that DES and similar expressions cannot be modified or expanded. Among the most common modifications for both MWEs are the maximizing adverb helt (completely) and the particle jo (after all).

With one exception, DES was the only realization of the pattern det er + epistemic adjective appearing in Linda’s L2 data. Linda’s use of DES was further restricted to unmodified uses. To our knowledge, no study to date has addressed det er sant in L1 or L2 Norwegian.

Toward usage-based CA-SLA

Usage-based approaches to language and language learning refer to a family of theories that share the view that language is learned from usage and that linguistic knowledge is not compartmentalized but part of general cognition (Langacker 2000; Bybee 2006; Ellis 2006; Goldberg 2006; Beckner et al. 2009; Pekarek Doehler and Eskildsen 2022). In UBL, language is an inventory of constructions (Ellis 2006; Goldberg 2006), that is, form—meaning mappings at various levels of complexity and abstraction, learned through participation in social events. Acquiring a new language is thus a matter of gradually establishing new form—meaning mappings while adapting those that already exist, based on experience.

Two basic processes permit language to be used with increasing ease and creativity: entrenchment and schematization (Langacker 2000; see also Lesonen et al. 2020). While schematization involves gradual abstraction of commonalities among specific realizations of a construction, entrenchment refers to how recurring uses of a specific expression result in routinization that makes it ‘a well-rehearsed routine that is easily elicited and reliably executed … no longer requiring conscious attention to its parts or their arrangement’ (Langacker 2000: 3). Automatization hinges on entrenchment. Formulaic expressions hold a high degree of entrenchment, but they may also be the first seeds for the reanalysis and extraction of a more abstract pattern. However, Ellis noted that ‘highly frequent and prototypically functional phrases ( … ) [that] are broad ranging and safe’ (Ellis 2012: 29) may additionally serve as phrasal teddy bears equipping learners with effective and cost-saving tools in interaction. In other words, their functional scope may broaden, while the syntactic pattern itself does not show signs of schematization. Hence, formulas are not only stepping-stones for grammatical development but may themselves be targets of L2 learning.

Within SLA, entrenchment and schematization have primarily been approached as cognitive phenomena. However, over the past two decades, ethnomethodological CA has contributed to an ongoing shift in conceptualizing language development from the internalization of linguistic knowledge to the ‘continuous adaptation of linguistic and other semiotic resources in response to locally emergent communicative needs’ (Pekarek Doehler 2010: 106). From this perspective, grammar is seen as co-constructed emergent practices rather than an abstract representation of linguistic knowledge (Markee 2008). The speaker’s knowledge of grammar influences the shape of a turn-constructional unit (TCU)—a sentence, a clause, a phrase, or a lexical item constituting an entity on the level of grammar, prosody, and social action—a basic building block of a turn in conversation. Likewise, interaction shapes and sharpens grammar that becomes applicable in specific contexts due to the configuration of the contingencies of situated language use (Pekarek Doehler and Eskildsen 2022). Aligning with a usage-based model of language, grammar and interaction influence one another in a circular relationship (Markee 2008: 407).

Research conducted in CA-SLA has explored how learners accomplish specific actions, such as turn design (Nguyen 2020), storytelling (Pekarek Doehler and Berger 2018), requests (Theodórsdóttir and Eskildsen 2022), topic management (Lee and Hellermann 2014), and word searches (Pekarek Doehler and Berger 2019). Based on longitudinal data, Pekarek Doehler and Berger (2018) reported on one L2 French user’s growing sensitivity toward the interactional context of narrative practice. Over the course of nine months, the storytelling sequences were delivered with increasingly extensive prefatory work that secured recipiency and acceptance from the co-participants of the conversation. Similarly, Pekarek Doehler and Berger (2019) focused on an intermediate L2 user’s word searches in ‘the wild’ over a period of 10 months. The analysis uncovered an increasing orientation toward the progressivity of the talk while maintaining intersubjectivity. Subsequently, changes in word-search practices reflected an increase in epistemic authority and independence on the part of the L2 user (Pekarek Doehler and Berger 2019: 72).

Studies integrating perspectives from UBL and CA have noted that interaction may foster both formal and functional changes in the L2, which manifest themselves through more productive uses of a given pattern (Eskildsen 2012; Hauser 2013). Theodórsdóttir and Eskildsen (2022) traced the development of expressions involving the verb ætla (would like) in one L2 Icelandic learner. Their findings show how the routinization of a specific MWE employed in a specific social context (ég ætla að fá, ‘I would like to get’) can be a precursor to both formal and functional diversification, allowing for different formal realizations and broadening its contextual scope from requests in service encounters to other environments (such as accounts or information requests).

However, pattern expansion is not the only sign of L2 progress; development can also be visible in the sequential position and speech production of a given structure. Pekarek Doehler (2018) focused on the developmental trajectory of the clausal connector parce que (because) and the MWE je sais pas (I don’t know) over a period of 10 months. The study showed how the turn position of parce que changed from turn-medial to turn-initial to perform a specific social action: seeking recipiency and affiliation. As a result, parce que functioned as a means of organizing interactions. Je sais pas, on the other hand, was observed to undergo both semantic and phonological reduction while changing its role in interaction from an epistemic disclaimer to a discourse-marking function. Moreover, Pekarek Doehler and Berger (2019) demonstrated how the function of the multiword expression comment on dit (how do you say) was altered from an explicit question to a marker of cognitive search. While the syntactic form remained stable, the construction over time exhibited prosodic backgrounding and morphophonological reduction, thereby acting as an interactional resource for holding the floor.

Similar findings reflecting the development of an L2 grammar-for-interaction are reported in Pekarek Doehler and Skogmyr Marian’s (2022) study of an L2 French learner’s increasingly routinized and diverse use of comment on dit over three semesters. Over time, the learner’s use of comment on dit developed from primarily being used in a literal sense (often involving a complement) into discourse-marker-like uses indicating cognitive search while functioning as a floor-holding device. In a study spanning four years, Pekarek Doehler and Balaman (2021) documented how the MWE let me check was routinized as an alert practice in L2 video-mediated interaction. Interestingly, the routinization of let me check for this specific social action emerged from a variety of constructions previously applied in similar contexts, indicating—somewhat contrary to the predictions of the exemplar-based learning trajectory—a developmental trajectory from more to less grammatical productivity.

Interaction may also contribute to L2 competence without involving observable formal changes through the routinization of a particular expression for social action. In a study of an L1 Korean shopkeeper’s use of one formula, ‘we limit ten under twenty centu charge okey?’, Kim (2019) provided evidence of how the idiosyncratic (i.e. non-target-like) expression was routinized to perform a specific social action of directing customers’ attention to the store’s policy. These findings show that routinization does not necessarily mean target-language compliance, and that communicative success is an important factor influencing the format of the intended social action. In line with these findings, we expected L2 development to be reflected in a variety of formal and functional changes that may be subtle and not easily observable without detailed microanalysis.

The present study

Pekarek Doehler and Eskildsen (2022) observed that UBL offers a method for studying exemplar-based language development longitudinally, while CA-SLA contributes a way of targeting ‘the local-social specifics and the dimension of action accomplishment in L2 learning’ (2022: 6). At an overall level, the advantage of combining the two resides in the possibility of achieving a more advanced understanding of how language use shapes L2 development. More specifically, the goal of the present study aligns with previous research exploring longitudinally how the routinization and diversification of a particular MWE manifest themselves and are strengthened through interaction. According to Pekarek Doehler and Skogmyr Marian (2022: 25), language is ‘the product of socio-communicative routines’ that may become routinized through repeated use in repeated actions.

As noted above, frequency is a driving force in routinization and subsequent entrenchment. Haiman (1994) argued that routinization also involves semantic erosion and prosodic reduction, as well as changes in grammatical structure. As such, routinization can be regarded as a form of grammaticization (and vice versa) that can be inspected through changes in linguistic features (Haiman 1994). Drawing on the work of Haiman (1994) and UBL, as well as previous evidence from longitudinal CA (Pekarek Doehler 2018; Pekarek Doehler and Balaman 2021), Pekarek Doehler and Skogmyr Marian called to attention the developmental trajectory of L2 grammar-for-interaction, in that ‘frequent combinations in use may lead to routinization and ultimately grammaticization of constructions’ (2022: 25). For some MWEs, the developmental trajectory of L2 grammar-for-interaction involves not only an increase in MWE frequency but also a development from literal to more discourse-marker-like uses accompanied by semantic bleaching and prosodic and grammatical reductions. Functional diversification is embedded in the process of routinization.

The point of departure for the present study is the trajectory of one MWE used repeatedly and with increasing frequency in a series of dyadic interactions rather than a specific social action. By approaching the data from this perspective, the developmental trajectory of the focal construction is foregrounded in a way that allows us to inspect changes in learners’ linguistic repertoires over time and with interactional experience. While previous longitudinal studies have tracked the L2 grammar-for-interaction developmental trajectory in learners’ interactions for longer stretches of time (e.g. Pekarek Doehler 2018; Pekarek Doehler and Balaman 2021; Pekarek Doehler and Skogmyr Marian 2022), this study pursues Linda’s uses of DES over a shorter but relatively fine-grained timescale. In the present approach, we focus on development within this specific time window by conducting a sequential analysis that maps the interactional contexts and social actions accomplished by Linda’s DES over time (RQ1) as well as changes in the functional range and the formal realization of the DES (RQ2). The overall aim is to investigate the extent to which processes of routinization and diversification can be identified. As such, the present study ties into recent developments within the field of usage-based CA-SLA.

Data

The data for the analysis consist of extracts including DES (N = 42) from eight audio recordings between Linda, a woman in her thirties raised multilingually in an English-speaking country reporting English to be her first language, and her former teacher Pau, who is also one of the researchers.⁵ The total duration of the recordings amounts to 106 minutes (Supplementary Appendix 1). The data cover a period spanning 17 weeks and were originally collected as part of a larger research initiative following a group of four learners of Norwegian through their second semester of Norwegian classes (Horbowicz et al. 2020; Horbowicz and Nordanger 2021), a design that allowed us to observe their intermediate L2 development. These learners participated in an intensive language course at an academic institution in Norway with the goal of achieving the B1/B2 level (Council of Europe 2020). The conversations concerned various preplanned everyday topics (Supplementary Appendix 1) but could also take new directions according to the participants’ associations and interests.

All conversations took place in Pau’s office, which was located close to the language classroom. The room had a large window and contained two chairs, a desk with a computer screen, a few plants, and bookshelves covering two walls. The desk was arranged in a way that allowed face-to-face interactions. Although the recordings took place in the teachers’ hall, the spatial arrangement, ensuring both privacy and accessibility, was intended to facilitate participation and encourage everyday talk in a relaxed atmosphere. The conversations did not occur spontaneously but were prearranged to elicit L2 data and provide the course participants with supplementary conversation practice. Nonetheless, we regard them as meaningful additions to Linda’s ‘biography of learning in social encounters’ (Brouwer and Wagner 2004: 42), and part of her L2 development.⁶ After all, such ‘conversations for learning’ can be seen as a realistic and well-rehearsed social practice for a language student. When the data collection began, Linda had been living in Norway with a Norwegian-speaking partner for almost two years, and the class she attended had just completed an A2-level exam. By the end, she was preparing for her final exam at the B2 level.

Procedure

As a first step in the analysis, all instances of DES were extracted from the data in their immediate context. Subsequently, the extracts were subjected to sequential microanalysis using CA as a toolkit. All extracts were transcribed following Jeffersonian conventions (after Schegloff 2007, see Supplementary Appendix 2), whereas the sound production of DES was measured in terms of phonetic realization, intonation (rising/falling), main stress, pace, and volume (see also Pekarek Doehler and Skogmyr Marian 2022) with the digital tool Praat (Boersma and Weenink 2023). To ensure comparability across time, all conversations followed a pattern of initial small talk before the prescribed topic was introduced. Even though the conversational topics varied between data points, the maintenance of the same participants and speech-exchange system (conversation-for-learning) ensured consistency in the sequential environments and conduct of similar social actions across data points (see Deppermann and Pekarek Doehler 2021).

Table 1 depicts the number of DES tokens and the DES frequency in Linda’s recordings over time. Frequency is accounted for as the number of occurrences per 100 finite verb contexts. We counted finite verb contexts since Linda did not always realize finite verbs as normally required in Norwegian (she sometimes omitted verbs or verbal inflection). Turning to the distribution of DES tokens and frequency across time, as shown in Table 1, we see that DES was used rather infrequently during the first half of the data collection. The frequency of occurrences begins increasing from Project Week 9 and reaches its peak in Project Week 17. The rise in the absolute number of occurrences also reflects a substantial increase in DES frequency: from Project Week 9, Linda applied DES close to four times as often as in Week 7. Marked by this jump in frequency, the period was broken down into two phases: Weeks 1–7 (Phase 1) and Weeks 9–17 (Phase 2). Below, we analyze and compare instances of DES employed in the two phases.

Analysis

The sequential analysis reported below is organized chronologically in accordance with the two phases. The analytical focus is the identification of social actions accomplished by DES, changes in functional scope, and sound production.

Phase 1

In Phase 1, spanning Weeks 1–7 and covering the first half of the data collection period, Linda’s use of DES is infrequent, as the MWE appears once or twice per conversation. Throughout the period, the social action performed by DES is expressing agreement, and DES is consistently produced in a slightly reduced form as/deæˈsɑnt/, exhibiting/r/ deletion, and with a flat and finalizing intonation.

DES as expressing agreement

In this section, we discuss two extracts. Extract 1 is drawn from the first week of data collection and illustrates the social action performed by DES in Phase 1. In Extract 1, where Linda and Pau are talking about Christmas holidays, DES appears in line 11 as a single TCU expressing Linda’s agreement with an assessment made by Pau. DES is sequentially placed following Pau’s turn in lines 9–10, reformulating Linda’s narrative in general terms (note the use of the impersonal pronoun man ‘one’ in line 9). In this turn, Pau is showing affiliation with Linda while producing an assessment that she can claim access to (Stivers 2008). Pau’s turn in lines 9–10 can be seen not as asserting but as soliciting confirmation (Stivers et al. 2011: 14). Linda’s agreement with DES in line 11 can thus be considered an adequate second-pair part in this exchange.

(1) Christmas weather, Week 1⁷

Although not the main focus of the present analysis, we also note that in line 9, Pau uses ikke sant (right) as an affiliation token (Svennevig 2007). Thus, Linda’s application of DES in line 11 may be said to be afforded by the interlocutor. Afforded uses have been taken as evidence for the role of interaction in L2 development by contributing to frequency and subsequent entrenchment (Eskildsen 2012).

A similar pattern as in Extract 1 is identified in Extract 2, recorded in Week 5: Linda’s DES appears following a personal narrative and an assessment delivered by Pau.

(2) See other places, Week 5

In lines 1–5, Linda provides an account of her preferred spare-time activity. In lines 6–8, Pau laughingly suggests that this activity can be difficult to achieve, as the place in question is too small to be explored for a long time. This turn may be considered an assessment of the feasibility of Linda’s plans. In line 9, Linda seems to initially agree (her DES expresses agreement), but the use of så (so), pronounced with a rising intonation, projects continuation of the turn. Next, Linda repeats DES, this time followed by men (but), which suggests an upcoming contrasting move (cf. Pomerantz 1984: 72). While Linda’s turns in lines 10 and 12 are an almost exact repetition of the statement in lines 1–2 and 5, they may be interpreted as contesting Pau’s assessment. Concurrently, Linda ‘exerts’ her epistemic rights to the discussed topic by repeating her prior turns, upon which the topic comes to a close in line 12.

Lines 6–10 in Extract 2 show the features of what has been defined as the cardinal concessive pattern (Couper-Kuhlen and Thompson 2000), that is, a dyadic sequence of conversational moves, allowing the speakers to express a potential contrast. The sequence consists of three moves: in the first move, the first speaker makes a point, the validity of which is acknowledged by the second speaker in the second move, while the third move offers a potentially contrasting point. One of the functions ascribed to the pattern is ‘creating “agreement” by acknowledging the simultaneous existence of two potentially incompatible perspectives’ (Couper-Kuhlen and Thompson 2000: 405). As such, the cardinal concessive pattern can be applied to develop the talk-in-progress. In Extract 2, however, rather typical of Linda’s DES use in Phase 1, the turns following DES do not develop the topic further.

Phase 2

In Phase 2, we observe a notable rise in the frequency of DES (see Table 1). DES is used repeatedly, and often in close vicinity to other instances. The prosodic features of DES do not exhibit consistent changes, although we observe growing instability in terms of pace and further consonant reduction from Week 10 onwards. The vowel production remains unaffected. These changes will be commented on below.

Linda’s use of DES in Phase 2 is somewhat consistent with Phase 1, as the action of expressing agreement with the interlocutor’s display of understanding, as exhibited in Extract 1, is still attested. We also find DES employed in the cardinal concessive pattern, as exhibited in Extract 2, but in Phase 2, Linda uses DES to preface turns that contribute to the topical development of the conversation more often than before (see Extract 3). A second social action first observed in Phase 2 is the use of DES as a stepping stone into a disagreeing turn (shown in Extracts 4 and 5). Next, we distinguish a third social action accomplished by DES, in which Linda uses DES to close self-initiated repair sequences (see Extracts 6–7) and finally, we show how DES is used to perform an explicit confirmation request (Extract 8).

The following extract demonstrates the continued use of DES as an expression of agreement in Phase 2. Similar to Extracts 1 and 2, the topic of the conversation concerns facts known only to Linda (her educational choices) and not to Pau, which gives Linda epistemic rights over her interlocutor. Prior to Extract 3, Linda delivers a narrative about initially studying law before she changed her mind and completed a degree in international development. Pau’s turn in lines 1 and 3 is a general assessment of the usefulness of choosing a career as a lawyer, delivered with strong mitigation using epistemic verbs tro (think) and vite (know). In line 4, Linda uses DES to express agreement with Pau’s assessment, but here, her DES may be interpreted as a preface to disagreement (Couper-Kuhlen and Thompson 2000), also encouraged by Pau’s mid-telling response token in line 5 (Stivers 2008), as she explains in line 6 that there is a common denominator for her educational choices—her interest in human rights. Thus, in line 6, Linda accounts for her career path choices while at the same time she rejects Pau’s suggestion that her initial choice of law studies was, like for many other law students, a matter of choosing a safe course (there will always be a need for lawyers, line 1).

(3) Educational choices, Week 9

While used to express agreement, DES as applied in Extract 3 also functions as a preface to Linda contributing to the topical talk in the following lines. As such, DES is used similarly to Extract 2, but here, the contribution to the topical sequence is more substantial, although we note that the turns in lines 6–8 are promoted by Pau’s turn in line 5 (ja?). Nonetheless, this use of DES shows that Linda is exerting a growing influence over the flow of talk. As such, it can be considered a precursor to the DES actions discussed below.

DES as prefacing disagreement

In the following extract from Week 12, DES is applied to evaluate and disagree with Pau’s original account (not a reformulation of Linda’s prior turn, in contrast to previous examples). In this stretch of talk, where Pau and Linda are talking about travelling in Europe, a new function emerges: DES as a preface to disagreement.

(4) Travelling in Europe, Week 12

In lines 2–5, Linda says that she has traveled a lot in Western Europe but is not as familiar with the eastern part, something she wants to change. Pau’s turns in lines 8–10 are an assessment, suggesting that it is normally the case that Eastern Europe is not so popular with tourists, before she delivers a mitigated account of the reasons for this (the Eastern bloc has been so destroyed in the war, lines 12–13). In line 14, Linda’s DES functions as an affiliation token, evaluating the prior turn, yet as it is delivered outside transition relevance place (TRP) signaled normally by the final intonation contour (note the continuing intonation in Pau’s turn in line 13), it can also be seen as disaligning with the in-progress action by treating Pau’s turn as complete when it is not (Stivers 2008: 36). Moreover, in line 14, DES is delivered together with men (but) as one turn with no prosodic boundary, suggesting that a contrastive move is upcoming. While Linda agrees with Pau’s assessment that the Eastern bloc was badly damaged by the war, she expresses a differing opinion that it has changed now (it is better now, line 14). Subsequently, Linda delivers several turns to provide the basis for her assessment (lines 16–20) before returning to the personally framed topic by mentioning the specific countries she wants to visit (line 22). Extract 4 demonstrates the way in which Linda employs DES to close off some sequences to pursue others, and thus to position herself as a knowledgeable and agentive participant in the conversation. It is worth mentioning that Pau herself is from a country in Eastern Europe, implying that she also has epistemic rights to this topic, even though her turn in lines 12–13 is delivered with mitigation. In response, Linda explicitly refers to the source of her knowledge (lines 16–20), asserting her epistemic access to the topic (Stivers et al. 2011).

In Extract 5 below, drawn from Week 10, DES is once again used as a preface to introducing her own view during the interlocutor’s narrative. In the beginning, Pau delivers her opinion on the topic of meeting new people at a certain age. Her turn in line 7 has not reached a TRP, so Linda’s turn in line 8 can be seen as disaligning with the prior assessment (Stivers 2008: 36). The turn is initiated by an affiliative DES, following which Linda produces same agreement by repeating Pau’s assessment in almost exactly the same words (man må prøve, ‘one has to try’, line 7 > du må prøve, ‘you have to try’, line 9). According to Pomerantz (1984: 72), same agreements constitute weak agreement and may preface disagreement. Linda’s turns in lines 11–12, in which she introduces additional variables to the topic (changing place of living), suggest that she does not fully agree that the difficulties in meeting new people are only, or primarily, a matter of age. Her initial weak agreement in lines 8–9 can therefore be considered a first step into presenting a differing view in lines 11–12.

(5) Meeting new people, Week 10

In extracts 4 and 5, Linda uses DES as a preface to presenting her own, potentially contrastive, view on the topics introduced by her conversational partner, thus asserting epistemic access to the topic (Stivers et al. 2011). Concurrently, as a collateral effect of its sequential position, DES may be seen as taking a discourse-regulating function as a turn-entry device. This secondary effect is especially well illustrated in Extract 5, where DES is delivered in a turn initial position and at a considerably faster pace than most other occurrences in the data (marked with ><, see also Extract 6). The phonetic realization is also different as the final consonant is reduced, resulting in /deæˈsɑn/, an assumingly idiosyncratic realization. As such, the growing variability in DES sound production co-occurs with the emergence of more discourse-marker-like uses, but the variability does not seem to be attached to specific actions. We return to this when summarizing the findings.

DES as closing self-initiated repair sequences

One highly specific function that emerges in Phase 2 is the use of DES to close repair sequences. Repair is a prevalent feature of all conversational language, not only typical of L2 use, and there is a general preference for self-initiated repair (Schegloff et al. 1977). Repair sequences are typically marked by speech perturbations, such as cut-offs or hesitation markers, which are visible in Extracts 6 and 7.

In Extract 6 from Week 12, Linda argues for the benefits of holidays. Her turns in lines 1–4 are marked with several signs of trouble, such as sound stretches, pauses, repetitions, and hesitation markers, as well as a marker of cognitive search, jeg vet ikke hva det heter (I don’t know what it’s called) (line 2). In line 6, Pau delivers a response token, and in line 8, she provides a candidate solution, which is immediately repeated by Linda in line 9 and assessed using DES, again delivered at a heightened pace, upon which she immediately incorporates the solution into her continued talk-in-progress.

(6) Effectivity, Week 12

Although DES as used by Linda in Extract 6 represents a distinct function, there are certain resemblances between this use and the prior uses of DES expressing agreement (Extracts 1 and 2): DES appears as a third-turn device, following Linda’s narrative and Pau’s short, mitigated turn, which can be seen as inviting confirmation (note the rising intonation in Pau’s turn in line 8). Furthermore, the use of DES observed in Extract 6 functionally resembles the instances of DES prefacing disagreement attested in Extracts 4 and 5, as in both uses DES constitutes a stepping stone to providing a substantial contribution to the talk and to holding the floor by delivering several turns. As such, Linda’s DES development can be seen as a continuing line, where new functions build upon previous ones.

The use of DES in the conclusion of a word search can also be observed in the last recordings (Week 17), illustrated by Extract 7. Here, Linda is searching for a word for ‘trial exams’ used by the interlocutor in her previous turn (lines 2–3). The repair sequence is signaled by the long pause of 1.7 seconds and the marker of cognitive search, hva heter det (what’s it called)⁸ in line 6. In line 8, Pau submits a candidate solution, delivered with a try-marking intonation, and the word is immediately picked up by Linda and accepted using DES in line 9. Almost the whole turn in line 9 is delivered with a laughing voice, which together with a cut-off explanation, jeg er litt nørv- (I am a bit nerv-) (which notably is reiterated from earlier on in the conversation, not rendered here) functions as a face-saving device (Yu 2013). In line 11, Linda provides a response to Pau’s question in line 1, assessing her performance in class and continuing the topic suggested by her interlocutor.⁹

(7) Trial exams, Week 17

As also attested in Extract 5, in Extract 7, DES is delivered as /deæˈsɑn/, showing additional consonant reduction. While in Extract 5, the reduction could be accounted for by the heightening of pace, the pace of speech is now regular. The deletion of /t/, marking neuter gender inflection, may indicate that the adjective is losing significance to the MWE as a whole. But as /t/ reduction is only observed in two occurrences of DES, we cannot draw definite conclusions about whether DES will eventually stabilize in this form in Linda’s Norwegian. Notably, the appropriateness of DES in contexts such as the one shown in Extracts 6 and 7 is debatable from a target language perspective, as, for instance, det er riktig (that is correct) or det stemmer (that’s right) would be more common realizations. In Norwegian, sant cannot take the reading of ‘correct’. As such, Linda’s current use may imply that DES is becoming less semantically specific in her L2. Nonetheless, this use of DES is not challenged by the interlocutor and may therefore be considered communicatively successful. Pau’s lack of reaction to the use of DES observed in Extracts 6 and 7, showing acceptability of this interactional conduct, may further affect the frequency with which DES is used to perform exiting repair sequences (Deppermann and Pekarek Doehler 2021).

DES as an explicit call for confirmation

Finally, Linda applies DES to elicit confirmation regarding the lexical correctness of the delivered turn. This use of DES is attested only once in the data, and we cannot say whether it constitutes a lasting addition to Linda’s repertoire. In Extract 8, Linda’s DES in line 3, delivered with a rising intonation, is a request for confirmation of whether the chosen word is correct, despite the fact that Pau has already signaled understanding with what can be considered a change-of-state token (okei) in line 2 (Heritage 1984). In fact, Linda pronounces the lexeme with an additional phoneme [l] (saying skulespill instead of skuespill), yet the resemblance is so close that it does not influence understanding. Pau recasts the word pronouncing skuespill in line 4, and additionally stresses the first syllable in which Linda’s mispronunciation occurred. In line 5, Linda repeats the lexeme twice in two different variants, and the word confirmation sequence comes to a close.

(8) Easter plays, Week 12

Like the two previous examples, DES used to perform an explicit call for confirmation emerges in the context of a word-search and demonstrates similar semantic idiosyncrasy as in Extracts 6 and 7. It also represents a single-attested use in the data. At this point of development, we therefore treat this use as afforded by the uses observed in exiting self-initiated repair sequences, rather than as a distinct function emerging in Linda’s DES repertoire. Note that Linda does not alter the word order to create a question (i.e. er det sant, ‘is it true/right’), possibly indicating that DES is stored as a fixed lexical unit in Linda’s L2.

Summary

The analysis depicts how Linda’s use of DES in talk changed over time. Three distinct social actions accomplished by DES are identified: expressing agreement, prefacing disagreement, and performing repair sequences, the latter encompassing both closing self-initiated repairs and requesting confirmation of the choice of words. Furthermore, a growing variability in phonetic realization in terms of pace and consonant reduction is observed in Phase 2. In the incipient stages, Linda applies DES to agree with her interlocutor’s assessments, which are delivered in response to Linda’s personal accounts. In Phase 2, DES is increasingly used to preface disagreement with her interlocutor’s assessments and to close self-initiated repair sequences to return to the talk-in-progress. A collateral of these diversifications is the use of DES as a turn-entry device allowing Linda to provide longer turns and thereby contribute more substantially to the talk.

The new social actions observed in Phase 2 can be best accounted for in terms of expansion of resources, where newly emerging functions come to accompany existing ones (the basic DES function of expressing agreement is also demonstrated in Phase 2, including lasting elaborations of this function, as shown in Extract 3). This implies that even though distinct social actions are identified, Linda’s evolving DES resource also includes ‘intermediate’ or transitional uses, indicating that the emergence of new functions is closely intertwined with already established ones. Transitions are thus not clear-cut but gradual.

Overall, the analysis reveals a developmental line of functional diversification co-occurring with an increasing morphophonological reduction. Changes in both areas coincide with a jump in frequency from Week 9, but neither can be said to be straightforwardly linear. We argue that Linda’s DES diversification in Phase 2 signals her increasing ability to participate more actively in topical talk and control the flow of discourse. In the next section, we discuss the meaningfulness and implications of the findings with regard to L2 functional development and routinization, and the study of languages in educational contexts more generally.

Discussion

Initially, we formulated two research questions: (i) In which interactional contexts does Linda use DES in her L2 Norwegian? and (ii) To what extent does Linda’s use of DES change or expand its functional range and formal realization over time? The analysis reveals that DES was used in several contexts, ranging from the basic function of expressing agreement to prefacing disagreement and closing repair sequences. Even though these three functions can be considered distinct, in Phase 2, they overlap in time, and the transitions between them seem gradual: the two latter grow from but also co-exist with the former (Deppermann and Pekarek Doehler 2021). We set out to explore to what extent Linda’s DES trajectory followed previous findings depicting the development of the L2 grammar-for-interaction as the functional diversification and routinization of ‘precise linguistic resources for accomplishing precise actions’ (Pekarek Doehler and Skogmyr Marian 2022: 25). Linda’s DES use certainly became more frequent within the time we followed her, and DES was applied in a broader range of social actions. In line with previous reports by Pekarek Doehler (2018), Pekarek Doehler and Balaman (2021), and Pekarek Doehler and Skogmyr Marian (2022), functional changes in the use of DES were observed, from more literal to more discourse-marker-like uses. DES, used as a stepping stone to a disagreeing turn and a resource to exit repair sequences in Extracts 4–7, as well as an explicit call for confirmation observed in Extract 8, performs discourse-regulating functions, allowing Linda to indicate a growing focus on the progressivity of the talk (Pekarek Doehler and Berger 2019) and to influence the development of topical sequences in conversation.

In L2 grammar-for-interaction, linguistic resources are routinized to perform specific social actions. However, some MWEs may also become entrenched as broad-ranging and cost-efficient tools in conversation: ‘teddy bears’ in the words of Ellis (2012). Rather than considering this a paradox, we regard it as two sides of the same coin: on the one hand, Linda’s use of DES becomes more specified as it diversifies for specific actions, and on the other hand, DES expands in range and thereby broadens. In sum, given the rise in frequency accompanying the functional diversification, the MWE appears to function as a teddy bear in Linda’s L2 at this stage of development.

While the functional diversification of Linda’s DES seems undisputable, the signs of routinization are not entirely conclusive. As noted above, routinization follows frequent use (Langacker 2000), but it also involves semantic and phonological reduction (Haiman 1994). The changes in Linda’s DES use between Phases 1 and 2 are accompanied by increased phonetic variability, potentially marking an increased tendency for morphophonological reduction. Linda’s DES already exhibits /r/ reduction in the early stages of the data collection. The variability arising in later stages, including /t/ reduction, results in additional simplification of the linguistic form. Since it co-occurs with increased functional diversification, it may be taken as the first indication of added economization and growing routinization (Deppermann and Pekarek Doehler 2021) in a developing L2 system. As noted by Pekarek Doehler and Skogmyr Marian (2022), the discourse-marker-like use of comment on dit seemed to be first established when the learner reached the B2 level. These authors reason that as proficiency and participation increases, a need for developing efficient tools for managing the flow of conversation arises. Linda was at the threshold between the B1 and B2 levels when we completed the data collection, and even though her development seems to follow a similar trajectory, it is possible that she had not reached a developmental point where full routinization could be expected.

L2 development within longitudinal CA is also often conceptualized as the gradual process of becoming a full member of a speech community (see Deppermann and Pekarek Doehler 2021). However, acknowledging the non-linearity of development, Deppermann and Pekarek Doehler emphasized how ‘a given practice or resource may be used, then recede, and then reappear over time’ (2021: 134). The present study identifies non-linearity first and foremost in Linda’s variable/t/-reduction and shifts in DES pace but also in how her DES follows a developmental trajectory, showcasing changes that do not necessarily meet the ‘related normative expectations’ (Deppermann and Pekarek Doehler 2021: 138). While Pekarek Doehler and Berger (2019) and Pekarek Doehler and Skogmyr Marian (2022) described a development trajectory toward the target language norm, the changes in Linda’s DES use involved new form–meaning mappings (DES as used in Extracts 7–9) and phonetic reductions that do not comply with conventionalized patterns in Norwegian. Linda’s DES trajectory reveals that both diversification and routinization might take unexpected forms in an L2, and that non-target-like uses may undergo routinization and subsequently become entrenched. On a general level, target-like production may be less important to routinization than communicative success. As indicated by Kim (2019), repeated successful uses of a potentially idiosyncratic MWE may lead to entrenchment, rendering the MWE a useful tool for achieving communicative goals. Our data thus indicate that non-target-like material may become subject to routinization, also when the idiosyncrasy relates to the functional level.

The significance of non-linearity in L2 development has repeatedly been pointed out in research (e.g. De Bot et al. 2007; Beckner et al. 2009). While the observed non-linearity may be hinging on the studied timescale and time window, the present study underscores the presence of variability, including progress and regress, in the developmental process of becoming a ‘competent member’ (Goodwin 2018, cited in Deppermann and Pekarek Doehler 2021). Throughout the period, Linda is recalibrating her resources to become a more proficient L2 speaker. This process, requiring recurrent reorganization of resources across linguistic and interactional levels, is conducted in a back-and-forth fashion in conversations-for-learning, which as such, become an important arena for negotiating and readjusting her L2 grammar-for-interaction. The diversification and growing variability we observe accompanying frequency imply that substantial changes contributing to Linda’s interactional competence are taking place in her L2 system. With time, DES seems to develop into an important pattern ‘to build social actions and manage [ … ] interaction’ (Pekarek Doehler and Skogmyr Marian 2022: 24), also when usen in a non-target like manner. In other words, DES is an important addition to Linda’s developing L2 grammar-for-interaction.

Conclusions and implications

In his 2009 paper, Eskildsen encouraged future research to ‘investigate in detail the link between local interactional phenomena and linguistic development’ (2009: 354) to understand second languages development. By tracing the development of one interactional resource over time, the present study contributes to a growing body of evidence for L2 development in learners’ socially situated interactions. The study produced two main outcomes. First, in line with previous research, functional diversification and beginning routinization may take place even when syntactic changes are not observed. Second, the development of interactional competence may also involve the routinization of non-target like material. As such, they further underscore that developing L2 interactional skills and fluency involves adopting form–meaning patterns that do not necessarily comply with target language norms.

Access to communicative situations is key to the development of the L2 grammar-for-interaction, and education plays a major role in providing settings where learners can participate in conversation practices that allow them to recalibrate and expand their linguistic resources. However, as such practices foster increased membership in the speech community, they may also nurture stabilization of non-target like L2 patterns accepted by the conversational partner to promote fluent communication. Future research should not stop at studying how grammar is adopted in educational settings but explore how and when changes happen across a range of settings and timescales in order to understand how interaction shapes L2 development.

Notes on Contributors

Paulina Horbowicz, Ph.D., is associate professor at the Department of Scandinavian Studies at Adam Mickiewicz University in Poznań, Poland. Her research interests include Norwegian second language development from a usage-based and dynamic systems perspective, cognitive models of first and second language competence, and conversation analysis. Address for correspondence: Department of Scandinavian Studies, Adam Mickiewicz University, al. Niepodleglosci 4, 61-874 Poznań, Poland. <[email protected]>

Marte Nordanger, Ph.D., is currently a post doc.-researcher in IMPECT-project realized at Western Norway University of Applied Sciences in Bergen, Norway, and an associate professor at the Department of Scandinavian Languages and Literature at Inland Norway University of Applied Scienecs. Her research interests include second language acquisition from a usage-based and dynamic systems perspective, and the impact of language policy and testing on second language learning and teaching.

Acknowledgements

We would like to thank Ann-Kristin Helland Gujord, Ingvild Nistov, and Aïda Leistad Thomassen for their comments on earlier drafts of this paper. Also, we express our gratitude to the anonymous reviewers for their insightful feedback and helpful suggestions which have greatly contributed to the final shape of the paper. Any remaining errors are our own. The authors state that there is no conflict of interest to disclose. Both authors have contributed equally to this work.

References

Footnotes