Oral english performance in Danish primary school children An interactional usage-based approach

Following the call in Sandlund, Sundqvist, and Nyroos (2016) for incorporating discursive approaches into the field of oral second language (L2) testing, this paper proposes an interactional usage-based approach to the analysis of oral L2 performance. Based on Eskildsen (2018a), we combine analytic tools from usage-based linguistics and conversation analysis. We draw on usage-based linguistics to analyze performance in terms of test-takers’ inventories of linguistic constructions and on conversation analysis to understand their interactional competence in terms of the relation between the linguistic constructions and the actions they are used to accomplish. Performance assessment is thus constructional and interactional. Participants in this pilot study were two Danish primary school children who performed two consecutive oral tasks: a semi-guided interview and a picture-elicited narrative task. Data were analyzed by means of cross-child comparisons and cross-task comparisons within each child. Our data confirm the observation from previous research that simple question-answer(-assessment) sequences dominate oral test for-mats, but also that the format is sometimes abandoned, which allows for the accomplishment of new social actions. Moreover, the picture-description task affords a different speech exchange system with the interviewer participating more as an active listener when the children do not voluntarily carry out the requested task.


Introduction
This paper discusses the inclusion of interactional competence in the assessment of second language (L2) oral skills (Roever & Kasper, 2018;Sandlund, Sundqvist, & Nyroos, 2016). Since the early days of communicative competence in second language acquisition (SLA) research (Canale & Swain, 1980) and language testing (e.g., Bachman, 1990), there has been a growing interest in the situated and interactional aspects of the test situation (e.g., He & Young, 1998;Kasper & Ross, 2013;MacNamara & Roever, 2006;Roever & Kasper, 2018;van Compernolle, 2011). Traditional approaches to testing have emphasized test-takers' individual linguistic performance, typically through measures of complexity, accuracy and fluency (Housen & Kuiken, 2009;Skehan, 2009). This perspective taps into the test-takers' spoken language without much regard for their interactional abilities, as pointed out by Roever and Kasper (2018), who propose instead to use the concept of interactional competence as an assessment tool.
Building further on this, we aim to develop an interactional usage-based approach to the analysis of oral L2 performance, that is, to provide a new method for assessing oral skills in a bottom-up, usage-based fashion. This method, in addition to capturing linguistic skills, encompasses test-takers' interactional competence, that is, test-takers' methods of accomplishing specific actions through semiotic resources, including language (Pekarek Doehler, 2018). Based on Eskildsen (2018aEskildsen ( , 2018b, this implies combining analytic tools from usage-based linguistics (UBL) and conversation analysis (CA). We draw on UBL to analyze performance in terms of test-takers' inventories of linguistic constructions and on CA to understand their interactional competence in terms of the relation between the linguistic constructions and the actions they are used to achieve. Performance assessment is thus constructional and interactional, bridging the gap between psycholinguistic and sociolinguistic approaches to assessment (cf. discussion in Roever & Kasper, 2018). It should be noted, however, that we are not interested in testing per se but in understanding our participants' oral linguistic and interactional skills. Our data come from oral interviews with young learners of English in Denmark. These interviews are an elicitation instrument rather than a test, and 525 we use the data to: a) assess, that is, understand and describe, our test-takers' L2 oral skills; and b) develop an interactional usage-based way to do it.

Assessing young learners' proficiency
Over the last two decades (e.g., McKay, 2006;Nikolov, 2016a;Rea-Dickins, 2000), the assessment of young learners' proficiency has become a central issue in language learning research. This assessment is usually concerned with how children progress over time in their L2 and what levels of proficiency they achieve in one or more of the four language skills (i.e., listening, speaking, reading, writing) by the end of certain periods (Nikolov, 2016b). Nikolov (2016b) provides a useful overview of the different kinds of assessments used for young language learners and divides these into three categories according to the purpose they were developed for. The first category covers international research projects like the ELLiE project (Enever, 2011), which used the same achievement tests for learners over a period of three years in seven different countries. A single task was used for assessing each language skill. The second category involves international examinations developed for young learners to be certified in their English proficiency, including the Cambridge Young Learners English Tests, the Pearson Test of English Young Learners and the TOEFL Primary Test. The final category involves what is referred to as assessment for learning, that is, ongoing teacher-based assessment which can be used to promote children's L2 development in the classroom (Sternberg & Grigorenko, 2002). The purpose of our oral task differs from the above in that it was deployed as an elicitation tool to provide data that could be used to analyze, understand and describe our testtakers' linguistic and interactional performance. This will lay foundations for subsequent large-scale comparisons between our test-takers with a specific focus on investigating whether and to what extent there are age-related differences in their linguistic and interactional competence.
While previous research on young learners' language has predominantly examined linguistic performance alone, the approach that we adopt here involves the co-investigation of interactional competence and linguistic repertoires. That is, we look at the actions accomplished, and the linguistic resources used to accomplish them. Our approach to the analysis of the linguistic repertoires, UBL, implies taking a construction-based view that does not compartmentalize language (cf. Eskildsen & Cadierno, 2007). This will be explained in the next section.

Usage-based linguistics
UBL is a cover-term for a range of linguistic theories that, in brief terms, unite in abolishing the syntax-lexis distinction and the competence-performance dichotomy, and instead insist that all linguistic units are meaningful, and that language is learnable on the basis of experience. UBL is concerned with the semiotic nature of language, that is, the form-meaning pairings which language is seen to consist of. These form-meaning pairings, also called symbolic units or constructions, are described along a continuum of specificity (i.e., from fixed formulas to abstract schematic templates which in turn sanction the single instantiations) and complexity (i.e., from morphemes to full utterances). Language knowledge, in this conception, is a structured inventory of these constructions.
Usage-based first language (L1) research has revealed how the linguistic inventory develops from recurring linguistic material in use (e.g., Ellis, 2002;Lieven, Salomo, & Tomasello, 2009;Tomasello, 2003), following a trajectory from specific multi-word expressions to partially fixed, partially schematic utterance schemas, to increasingly schematic constructions based on systematic commonalities among patterns -for example, shifting from Where's the ball? to Where's the X?, and eventually to Where COPULA NP?. Similar learning trajectories have been observed in adult L2 learning in larger corpus analyses of L2 English as well as in case studies of L2 use and learning over time (e.g., Ellis & Ferreira-Junior, 2009;Eskildsen, 2012Eskildsen, , 2015Roehr-Brackin, 2014;Tode & Sakai, 2016). Usage-based research on children learning an L2 is virtually non-existent, with the exception of Karrebaek (2011), who used a construction-based apparatus to analyze the development of the majority language in a bilingual child, Suliman, in a Danish kindergarten. Her study focused on Suliman's socialization into the kindergarten, but Karrebaek also analyzed Suliman's linguistic utterances and found a high degree of repetition and reliance on recurring multiword expressions in line with usage-based predictions.

Conversation analysis
Recently, it has been argued that UBL needs to draw on a theory of social action, that is, CA, to capture how social practices are constructed and made visible for people to learn (from) them (Eskildsen, 2018a;Eskildsen & Kasper, 2019). Crucial to an understanding of CA is the idea that when an action is produced, the next relevant action is occasioned, and this next action gives meaning to the prior one. In this view, the ascribing of functions to linguistic expressions is done by people in situ rather than a priori (cf. Levinson, 2013). In other words, by providing an answer to a question, accepting an invitation, or mitigating and producing an objection to a comment or assessment, people show their understanding of what their co-participant has just said, thus ensuring constant construction and maintenance of intersubjectivity. If intersubjectivity is challenged, people can initiate repair and work through the challenge to restore intersubjectivity (for further detail on CA, see e.g., Hutchby & Wooffitt, 2008;Schegloff, 2007). It should be observed, however, that CA is not solely concerned with the modality of talk but with all interactional behavior, including embodied actions such as gesture, gaze, and body posture (Nevile, 2015). This is also brought to bear on the analyses below.

Combining UBL and CA
UBL shares with CA the core concept that (L2) learning derives from observable phenomena in the environment (Eskildsen & Cadierno, 2015;Kasper & Wagner, 2014). UBL makes it possible to investigate the fabric of linguistic-semiotic resources that transcend lexical specificity and contextual borders, allowing explorations of constructions as cognitive routines toward which CA takes an agnostic stance (Burch, 2014). CA, on the other hand, throws light on the situated specifics of the interactions in which people put their linguistic repertoires to use to accomplish social actions. A branch of conversation analytic SLA (CA-SLA), which is of direct relevance to this paper, has explored L2 speakers' interactional competence -that is, socially shared methods of accomplishing particular actions, such as repair, turn openings and closings, story-telling, dispreferred responses, and how those methods change over time (for a recent overview, see Salaberry & Kunitz, 2019).
Our perspective here is an "interactional usage-based approach" (Pekarek Doehler, 2018), which draws on the methodological combination of UBL and CA (Eskildsen, 2011, among others). We apply this two-pronged methodology to two oral tasks given to two young learners of English to propose a way forward for formulating an interactional, usage-based approach to assessing L2 proficiency. We explore how interactional competence is displayed in the oral tasks and what linguistic material is used to perform the social actions identified. Therefore, the focus will be on an empirical description of the sequential structure of the interviews, the social actions found therein, and the linguistic resources deployed to accomplish them. This description concerns the relationship between social action and linguistic expression in situ, adding to the current research on the interface between "form-meaning patterns" and "construction-action relations" (Eskildsen & Kasper, 2019). The analyses of social actions and linguistic expressions will include both cross-child comparisons and cross-task comparisons within each individual child. Of particular interest in relation to the latter is examining possible differences in children's performance on tasks that are more or less constrained in nature and thus allow for more or less spontaneous talk. 528 3. Method

Participants
The participants in the study were two Danish primary school children who had started learning English either in the first grade (age 7; Nicoline) or the third grade (age 9; Bo). At the time of data collection, the children had received English instruction for two years and were thus attending the third and fifth grades, respectively (age 9 and 11). The two children attended two elementary schools in Odense, Denmark, which followed the national guidelines on English language teaching (Danish Ministry of Education, 2019). Applying to all learners irrespective of their starting age, the guidelines establish what children should be able to do after finishing the fourth, seventh, and ninth grade. After the fourth grade, the relevant benchmark point in this case, children should be able to participate in short and simple conversations as well as understand frequent words, expressions and short texts on everyday topics in English.
The participants were part of a larger-scale research project investigating the role of age of onset, that is, the age factor, and a range of contextual factors (i.e., the quantity and quality of exposure to English inside and outside the classroom) and socio-affective factors (e.g., children's motivation and attitudes towards English) in children's rate of L2 learning and short-term language proficiency (see Cadierno & Eskildsen, 2018, for details). While the total number of participants in the large-scale project is 276 children, the participants for the present study were selected out of the 36 children whose oral data were transcribed. Selection took place through a purposeful sampling technique aimed at choosing two children with very different degrees of language proficiency. The reason for this criterion is that we wanted two very different children in order to test the scope and validity of our analytical apparatus. Bo, the more proficient one, was a late starter and Nicoline, the less proficient one, was an early starter. The fact that there was one from each group is a by-product of the first selection criterion: The early starters, as we know from the quantitative studies in the large-scale project (Cadierno & Eskildsen, 2018;Cadierno et al., 2020), are generally much less proficient than the late starters.

Instrument and procedure
Children performed two consecutive oral tasks within one single session, which constituted our elicitation instrument. The choice of tasks was inspired by the BAF project (Muñoz, 2006). Both tasks were conducted on an individual basis in a face-to-face situation with a native speaker (NS) of English who had very little Danish proficiency. The first one was a semi-guided interview where the NS asked each child a series of personal questions related to their age, their family and friends, and their free time activities, followed by a few questions about what they did the weekend before and what they would do the weekend after. Finally, they were given the choice of asking the interviewer questions about her or her family. Even though the interviewer had an interview guide with a series of fixed questions, there was room for extra questions and follow-ups on child-initiated topics.
After completing the interview, children performed the picture-elicited narrative task. Children were shown a series of six pictures that form a story (the so-called "dog story;" Heaton, 1966) and were asked to re-tell the story depicted by the pictures. They had the pictures in front of them when retelling the story. A series of prompts were used if the children did not engage in re-telling the story. The prompts inquired about what was going on in several of the pictures. After re-telling the story, children were asked a series of personal questions related to the story. Before moving on to our analyses of our two focal children's oral performance, we present an outline of the sequential structures found in the data. This is necessary for the discussion of the children's interactional competence in the results section.

Results: Sequential structures of the tasks
Participants in interaction, irrespective of the type of interaction, must collaborate to distribute turns-at-talk and design their turns as actions so that they build adjacency pairs and sequences (question-answer, invitation-acceptance, assessment-agreement, etc.) as well as potentially larger interactional structures, such as story-telling and descriptions. Interlocutors may also engage in sequentially organized repair work in order to achieve and maintain intersubjectivity, that is, on-going mutual understanding (Schegloff, 2007). The ability to engage in interaction, among other things along these lines, is what the concept of interactional competence captures.
The first part of the task, the semi-guided interview, is characterized by recurring sequential environments: question-answer sequences with a possible sequence-closing third turn (4.1), insertion sequences (4.2), and multi-turn responses (4.3). The second part, the transitory sequence between the semiguided interview and the picture-description task, is dominated by question-answer sequences with a possible sequence-closing third turn but with the child asking the questions. The third part, the picture-description task, is sequentially different with the interviewer playing a less dominant, yet active, role as she displays listenership through nods and acknowledgment tokens, does confirmations and encouragements, and elicits further descriptions that then predominantly follow the question-answer (-assessment) format.

Question-answer sequences and a sequence-closing third turn
The predominant sequential environment is that of question-answer pairs with the possibility of sequence-closing third turn, that is, question-answer-assessment sequences (Schegloff, 2007), in which the third turn functions as an evaluation of the appropriateness and relevance of the answer (van Compernolle, 2011). In most cases, these pairs and sequences center on requests for information (from Karen, the interviewer) and the children providing the information in their responses. The second question in the interview guide, concerning the children's age, will serve as an example (note that Karen's turn at Line 3 is the sequentially optional assessment; transcription conventions can be found in the appendix):

Insertion sequences
The participants may also engage in repair work/confirmation request sequences that are insertion sequences in these formats. Insertion sequences work to put the current sequential progression on hold until some elaboration or information is provided, or the trouble has been resolved (Schegloff, 2007). The example, Extract 2, shows Bo's ability to initiate repair. Karen is asking a follow-up question to what Bo plays on the computer (Line 1). Instead of responding immediately, Bo initiates repair as he asks for confirmation that Karen is referring to a specific Pokemon game (Line 2). Given confirmation (Line 3), Bo then responds that he plays it on his phone (Line 4) and Karen acknowledges receipt of the response before continuing with a new follow-up question (not shown): Extract 2. Insertion sequence.

Multi-turn responses
Lastly, some questions potentially invite responses that run over several turns, before a next question is asked. These are typically the more open-ended ones such as Can you tell me about x?, which result in reports or "mini-story-tellings". Extract 3 is an example where Karen is asking Bo about the most recent movie he saw (Line 1). Bo embarks on a telling, spanning two turns with an acknowledgment token from Karen in between (Lines 2-5). We note how Bo's intonation patterns may also work to signal that he is not done after the first turn at Line 2 (rising intonation) and that his story is complete at Line 5 (falling intonation). This is also how Karen treats his turn-design, as seen in her acknowledgment token (Line 3) and follow-up question (Line 6): >WHAT'S IT CALLED< 05 =e::h fat. 06 K: .hh okay ah hah hah hah .hhh was it a funny movie?

Results: The two children's performance
In this section we present the results of the two children's linguistic and interactional repertoires. For each child we follow the same structure. We first provide an overview of the child's linguistic inventory, extracted from the data and describe it in terms of concrete instantiation and assumed schematicity, following general UBL principles. We then couple the linguistic repertoire with the social actions accomplished by the child. In a subsequent summary section, we outline the commonalities and differences between the children's achievements of the actions found in the data with respect to both linguistic repertoires and interactional competence. Table 1 below provides an overview of Bo's linguistic repertoire. In the table we specify the task type (i.e., interview vs. picture-based narration), the linguistic expressions used by the child (i.e., instantiations), and the type of schema that sanctions the linguistic expressions. As shown in Table 1, Bo's linguistic inventory predominantly consists of repeated patterns centered on five verbs and one auxiliary, all high-frequency verbs in English, namely the copula, have, like, play, see and can. In addition, there are prepositional phrases, some of which are stand-alone uses, drawing on to, about, with and on as well as a few instances of other verbs (go, buy) and what is probably an instance of cross-linguistic transfer, winking, essentially an Anglified version of the Danish word for wave, namely vinke, coerced into the progressive form. In addition, Bo's linguistic inventory consists of constructions of varying degrees of specificity and schematicity: seemingly fixed multiword expressions with a coherent pragmatic function such as you're welcome; thank you; have a nice day, constructions consisting of fixed parts and open slots (e.g., I can see X; there is a NP) and more schematic constructions such as the copula construction (e.g., The pokemon is nice), the intransitive construction (e.g., I swim, I sleep with my friends), the transitive construction (he like candy; I have a big sister), interrogative questions with wh-words (how old are you?; what do you do in Denmark?), motion constructions (e.g., he's came to prison; the dog is climbing in the X), the future construction with the verb gonna, and coordinated and subordinated clauses. We note that the transitivity schemas in Danish are syntactically identical to the English ones. The only notable difference in Bo's production across the two tasks is the heavy repetition of the I can see X-pattern in the narration task.

Bo: Embodied interactional competence
The point of this section is to outline how Bo uses his linguistic repertoire, mapped out above, and to empirically illustrate the aspects of his interactional competence that transcend the accomplishment of the basic sequential structure. The purpose is to understand how the linguistic repertoire plays into his interactional competence. As mentioned above, question-answer(-assessment) sequences dominate the interview. In this part of the data, Bo answers relevantly and appropriately in English to all Karen's questions -with, perhaps, one exception when he replies to her nice to meet you at the beginning with a thank you. From a usage-based emergentist perspective, it is interesting to note that his responses often carry traces of the questions (e.g.,

what's her name -her name is; who are your best friends -O is my best friend and S is my best friend; what's that about -about a rich man; what do you play on the computer -I play CS Go)
, which underlines the pervasive nature of recurrence in discourse (cf. Bates & MacWhinney, 1988;Hopper, 1998). In terms of interactional competence, then, the bulk of the interview-part of the data concerns the children's ability to respond appropriately to information-seeking questions, and one way to accomplish this 534 is through reuse of components of the question. This indicates an interrelationship, worthy of more future research, between linguistic resources, interactional competence and the affordances brought about by the contents of the interview itself.
There are also situations where Bo draws on other aspects of interactional competence. We already saw an example of Bo initiating repair in Extract 2. Next, in Extract 4, which comes from the beginning of the picture-description task, we see an example of a word search. A word search is initiated when a current speaker runs into production trouble. In the extract, Bo is answering Karen's question about what he sees in the pictures. He then encounters a problem and begins the word search (Line 2). The typical resources people use to signal that they are searching for a word include pauses, speech perturbations, prolonged vowels and displays of "doing thinking" (Brouwer, 2003;Goodwin & Goodwin, 1986). Here we see Bo pausing and producing speech perturbations and a signal that he is currently thinking of a way to formulate (hvad hedder det nu 'what's it called'). At Line 4 he finally produces make sandwiches with rising intonation, which receives a nod and an acknowledgment token from Karen (Line 5). Bo thus provides a candidate solution to the word search himself but also seeks and gets confirmation from Karen: >WHAT'S IT CALLED< 03 K: [nods several times 04 B: =e::h make sandwich?es looks up at K 05 K: nods mhm:?
In the next example, we see that Bo has linguistic resources to do turn-allocation. This extract, 5, follows immediately Extract 4, hence the beginning at Line 6. Karen's turn at Line 5 (Extract 4) also works as an invitation to Bo to continue and he does so, adding a dog to the list he has begun creating. Karen produces an acknowledgment token and following a pause Bo again continues with yet another observation and a mother (Lines 7-9). Then a long pause ensues (Line 11). Such pauses usually happen when trouble occurs, and here it seems that it is not clear whose turn it is. Bo's yes-token with falling intonation (Line 12) finally allocates the turn to Karen who asks him about the next picture (Line 13): Extracts 5 and 6 will also be used to discuss some differences in the sequential structures and Bo's interactional accomplishment. These extracts, taken from the narrative task, show how Bo is providing the description over multiple turns (2,4,6,9,12) and how Karen is contributing to the co-construction of this sequential structure by displaying listenership through short acknowledgment tokens and head-nods (3,5,7,10). Resembling how people do story-telling, this sequential structure is sustained by the participants' actions and might be expected to characterize the narrative task across the children. Therefore, it calls not only on other linguistic expressions (e.g., I can see X) but also on other interactional skills such as signaling turn-completion and turn-allocation. Such skills may also be found in the interview part of the data, but to a lesser degree, namely in cases where Karen is asking questions of a more open-ended nature, as seen in Extract 3.
The last example is of Bo giving an account. Extract 6 comes from the part in the interview where Karen asks the children if they have any questions for her. Bo responds by asking Karen how old she is (Line 1). The question itself is probably a bi-product of the interview situation: The children have been asked the same question and now that it is their turn to ask, it may seem like a natural place to begin (other children do the same). Bo receives Karen's provided information with surprise (oh in Line 3). Karen treats this as a need for confirmation (Line 4), and Bo then upgrades his surprise with another verbal token and a facial expression of skepticism with lowered inner corners of the eyebrows ("the not face," recently found to be a straightforwardly recognizable facial expression, perhaps universally, cf. Benitoz-Quieroz, Wilbur, & Martinez, 2016). The disbelief on the part of Bo continues until he finally produces an account at Line 13 -he claims to think Karen was younger, which she meets with laughter and praise. She verbally thanks him -which he responds to with a you're welcome, seemingly aware of the thank you-you're welcome adjacency pair:  Table 2 below provides an overview of Nicoline's linguistic repertoire. It is clear that Nicoline's linguistic inventory predominantly consists of one-word constructions comprised of three main word types: nouns (e.g., dog, girl, boy, cow), numerals (i.e., nine, one) and adverbials (i.e., yes, no). There is only instance of a transitive construction with the verb like (i.e., I like swim), where the preposition to is omitted. In addition, Nicoline uses several Danish expressions that can consist of: (a) single words, that is, numerals like fem 'five' and seks 'six;' nouns like børn 'children' and pige 'girl' and a single verb vinker 'wave;' (b) phrases (e.g., en kurv 'a basket;' efterskole 'after school'); and (c) clauses (i.e., det ved jeg ikke 'I don't know;' spiller rundbold 'sandlot baseball;' er det mom? 'is it mum?'). There does not seem to be any difference in Nicoline's linguistic inventory across the two tasks. She mostly uses one-word expressions in both tasks.

Nicoline: Linguistic repertoire
In the narration task these are mainly nouns whereas in the interview task other forms of short responses are employed.

Nicoline: Embodied interactional competence
It is clear from Table 2 that Nicoline does not produce much English. However, many of the basic components in the interview can actually be accomplished with very limited means. The information-seeking questions can, in many cases, be answered appropriately with one-word answers, such as name or age. It does seem, however, that her limited linguistic repertoire also has a bearing on her interactional performance. We will use the third question, do you have any brothers and sisters?, as illustration. Prima facie, this is a polar question that can be answered with a yes or no (the direct speech act), but it can also be interpreted as a request for information about what siblings the children have (the indirect speech act). Whereas Bo seemed to operate on the latter interpretation as he responded with I have a big sister and I have a little brother, Nicoline orients solely to the question as a polar question as she produces a yes in response. Now, according to the interview guidelines, the next thing is for Karen to ask about the sibling's age. In Bo's interview, this is a straightforward enterprise as Karen asks about his sister's age, but in Nicoline's case, the interactional trajectory is different as Karen proceeds to ask follow-up questions that will allow her to ask about siblings' ages. Table 3 displays the transcription and the actions in the unfolding question-answer sequences. There are also cases where Nicoline does not seem to be able to deliver a response. In these cases, she may respond with speech perturbations (ehm, eh, m::), laughter tokens, and gazes away from Karen, initiation of repair in Danish (hvad? 'what?'), or claims of insufficient knowledge, either in Danish (det ved jeg ikke 'I don't know') or through embodied conduct (arms open, palms upward, headshakes) or a combination of these (on embodied claims of insufficient knowledge, see e.g., Sert & Walsh, 2013). The question about Nicoline's favorite TV-show will serve as example. Here, Karen makes three attempts to elicit a response (Lines 1, 3, 5) before Nicoline delivers her embodied claim of insufficient knowledge (Line 6): There is one instance where Nicoline produces more than a one-word response in English (Extract 9). Karen asks Nicoline what she likes to do (in her free time), and when she receives a claim of insufficient knowledge (Lines 1-3), she gives Nicoline two embodied examples and repeats the question (Lines 4-6), following which Nicoline reproduces a reduced version of Karen's first example, I like to swim (Line 8). Overall, then, Nicoline responds to Karen's interview questions with minimal responses, claims of insufficient knowledge or an open-class repair initiator (hvad?) in Danish. This repair-initiator is referred to as open-class because the speaker of the first turn cannot know what parts of her turn caused trouble, or even if the trouble was a hearing issue. Karen's reaction in the two cases where Nicoline produces hvad? is to respond with an okay, suggesting either that she simply does not understand the Danish word and closes the sequence, or that she interprets it as a token of lack of comprehension and lets Nicoline know that she need not worry about not having understood.
In the picture-description task, Nicoline does not produce any lengthy talk either. Her linguistic performance is so limited that virtually no analyzable English is present (cf. Table 2). Her interactional competence is displayed through understanding of and responses to some of Karen's interview questions and repair-initiations and various instances of claims of insufficient knowledge. Much of this, however, is carried out in Danish or through embodied resources. Nicoline represents a case of limited L2 performance and interactional competence.

Discussion
The data have shown that the two tasks, with Bo and Nicoline respectively, run off in very different ways. For example, the two children's different responses to the question about siblings are indicative of their interactional competence: Bo responds to the indirect speech act of the question (and by implication to the direct speech act, too), whereas Nicoline only responds to the direct speech act. These differences in their interactional competence have a fundamental impact on the interactional trajectories of the interviews. One might speculate that this is a matter of perspective on the task on the part of the children, but given Nicoline's displayed trouble in understanding some of Karen's questions and her apparent reluctance to say much in English, this probably has more to do with comprehension and production difficulties in English. This is supported by Table  2 outlining Nicoline's limited linguistic inventory.
However, the question-answer(-assessment) sequences and the information-seeking nature of the questions make it difficult to ascertain the assumed co-dependency of linguistic repertoire and interactional competence, because the children's turn in the sequence, the answer, is designed to provide information, and this can be accomplished in various ways, linguistically, without consequences for the interactional accomplishment. For example, in the case of How old are you?, the responses found in the interviews we have looked at so far are age (e.g., nine), "I'm age," "I'm age years old," and "age years old." All answers are apt, interactionally and linguistically, and one variety does not point to greater L2 proficiency than the others. In this case, longer and more complex responses are not necessarily indicative of a higher degree of interactional competence.
Aside from providing information when requested, the participants are performing other social actions through their turns-at-talk and embodied conduct. Bo, for example, engages in repair sequences, he successfully elicits confirmations from Karen in word searches by producing candidates with rising intonation, and he does "thinking faces" and gazes that normatively preempt interruptions from co-participants. Nicoline, on the other hand, only initiates open-class repair in Danish and the trouble is never resolved in these cases, or she produces embodied claims of insufficient knowledge, in which case Karen either closes the sequence (Extract 9) or pursues the answer by elaborating on the question (Extract 10).
As may be inferred from the preceding discussion, we investigate actions that are targeted by the oral tasks and hence predictably found in the data as well as actions that are occasioned by the on-going talk and thus not targeted by the tasks. Those targeted include providing information, asking for information, and giving a description, whereas the more emergent skills, taking Bo's interview as our starting point, include accomplishing repair and word searches and contributing to turn-taking organization. Bo's varied resources, including embodied conduct, to accomplish these social actions, targeted by the task or not, all illustrate his interactional competence. These resources and the social actions they are used to accomplish are essential to the in-situ co-construction of meaning in the interviews, even if they are not targeted by the task, so in order to gain holistic insight into the children's proficiency it is important that they be taken into consideration in L2 proficiency assessment and testing (cf. Roever & Kasper, 2018).
Our data allow us to map out and compare the children's linguistic repertoires, on the one hand, and how they engage in providing answers to information-seeking questions, on the other. Moreover, we can also compare how they accomplish the other more emergent social actions to the extent that these interactional environments are found across children. Other test designs are aimed at investigating test-takers' accomplishment of the same social actions, for example as found in the work by Youn (2015). Her test-design is a scenario in which students are instructed to make a request a professor for a recommendation letter. Both participants are provided with instructions of what to do but not what to say. This design allows a comparison between test-takers' methods to accomplish the act of requesting. The downside to this approach is that it only tests one particular aspect of interactional competence. A combination of test designs that, on the one hand, target how L2 speakers accomplish particular and specific social actions and, on the other, allow investigations of emergent, locally occasioned aspects of interactional competence may be a way ahead for future developments in L2 assessment.

Conclusion
The differences between Bo and Nicoline, both with respect to linguistic repertoires and interactional accomplishment, can be fairly straightforwardly drawn out. This pilot study has helped us identify both linguistic patterns and interactional practices that we can expect to find (or be absent) in the remaining data, thus guiding us empirically to possible points of comparison across all our informants. The nut to crack in terms of L2 assessment is then to provide a solid account of the relationship between the children's linguistic repertoires and their accomplishment of social actions. We have argued here that our combination of UBL and CA provides a fruitful methodology for doing precisely that.