Patterns of interaction in a paired speaking test: comparing L1 and L2 interactions

Abstract

The shift of focus from a cognitive to social view has drawn the attention of second language assessment researchers to delve into interactional competence as an intriguing topic in paired oral tests. To fill the gap on how interactional patterns compare across L1 and L2 interactions this study was specifically set to explore and compare 45 paired interactions of L1 English speakers, L1 Persian speakers, and L2 English learners. While the findings indicated that the collaborative pattern was the most common pattern in the three groups less than half of the interactions in total were collaborative. Furthermore, not all the interactional patterns were found in all the groups. The frequency and prominence of the common patterns also varied. The findings could be explained by cultural norms of speech (e.g., in taking turns) and interlocutor effects (e.g., dominating the conversation or adopting a passive role). The findings imply challenges for fair assessment when using paired tasks as assessment tools. Furthermore, the findings call for rubric development and rater training programs to include description and discussion of interactional patterns to help raters improve their understanding and rating of paired interactions.

Introduction

The shift from a cognitive view in language testing where language competence is considered an individual ability to a social view in which communicative language ability is manifested in a social context through a joint attempt at co-constructing a particular discursive practice has brought about some changes to language proficiency tests. From the time ‘interactional competence’ (IC) was introduced (Mehan, 1979, and later on Kramsch, 1986) based on this social view, researchers have focused on tasks that can elicit IC more effectively. In line with this, paired speaking tasks have been used in some high-stakes and classroom assessments (e.g., Cambridge English exams). Compared to the traditional oral interview which has been the dominant test format for a long time, the paired oral test has a positive washback (Ducasse & Brown, 2009; Taylor, 2001), is time and cost-efficient (Ducasse & Brown, 2009), elicits a wide range of interactional skills (Brooks, 2009; Galaczi, 2014; Nakatsuhara, 2004; Plough et al., 2018; Taylor, 2001; Vo, 2020), and is preferred by language learners (Egyud & Glover, 2001; May, 2000; Taylor, 2001).

Despite all the advantages stated for the paired oral test, concerns remain about how to assess the co-construction of discourse in this test (see e.g., Fulcher, 2003; May, 2009, 2011; van Batenburg et al., 2018; Youn & Chen, 2021) which is realized in four patterns of interaction including collaborative, parallel, asymmetric and blended (Galaczi; 2008; Storch, 2002). These interactional patterns pose challenges for raters to assess the joint discourse. It has been argued that asymmetric patterns are more difficult for raters to rate (Galaczi, 2004; May, 2009). Also, test takers are more likely to receive higher scores when interacting in collaborative patterns (Galaczi, 2008; Isaacs, 2013; Storch, 2002). As such, raters have a difficult task deciding about the appropriate score for each interlocutor in paired interactions. The difficulty increases as the interactions may be influenced by interlocutor effects (Csepes, 2009; Davis, 2009; Foot, 1999; May, 2009; Norton, 2005; Roever & Kasper, 2018). That is, an interlocutor may unfairly influence a test taker’s performance and make scores biased. For example, in studying the effect of interlocutor proficiency in a paired test, Davis (2009) found that lower-level test takers generated more language (words) when interacting with a higher-level interlocutor.

The literature still suffers from a paucity of research on how the interactional patterns are affected by various factors. Furthermore, not much is known about whether such interactional patterns are language and culture-specific or may similarly be found in social interactions in both L1 and L2 contexts. There have been arguments in the literature that certain aspects of interactions could be language or culture-specific (e.g., Fulcher, 2003; May, 2009; Taguchi, 2014), but very few studies have been designed to specifically investigate this issue (e.g., Isaacs, 2013). For example, Fulcher (2003) argues that turn-taking conventions may vary from culture to culture. Similarly, May (2009, p. 21) argues that personality- and culture-related factors “might determine, or at least contribute to” the paired interactional patterns. The raters in May's study referred to body language and assertiveness as important features of IC. However, these features "could be seen as aspects of culture and L1 usage. What constitutes assertiveness in communication to the raters might be interpreted as aggressive and inappropriate communication in some cultures" (p. 21). However, not much is known about how native speakers of different languages co-construct the discourse in their L1 interactions, or how L1 interactions are different from L2 interactions. These are important questions that can help define the construct of IC and determine whether IC can be conceptualized similarly across L1 and L2 interactions or depending on the language of interaction the conceptualization should differ. Furthermore, investigating such questions can be an important step in creating fairness in performance-based assessment by informing the development of assessment rubrics and enhancing raters’ assessment literacy. Additionally, test users are benefited as they can more effectively interpret and use the scores coming from paired oral tests. The current study was, therefore, set to investigate L1 and L2 interactional patterns. More specifically, paired interactions of L1 English speakers, L1 Persian speakers, and L2 English learners were explored. The following research question was, therefore, put forward in this study:

How do the paired interactions of L1 English speakers, L1 Persian speakers and L2 English learners compare in interactional patterns?

Literature review

The last two decades have displayed IC as an intriguing topic for research in Second Language Acquisition (e.g., Cekaite, 2007; Çimenli et al., 2022; Dings, 2014; Doehler & Berger, 2018; Masuda, 2011; Nguyen, 2018; Taguchi, 2014; Tecedor, 2016) and Second Language Assessment (e.g., Ducasse & Brown, 2009; Galaczi, 2004, 2014; Lam, 2019; May, 2006, 2009, 2011; Roever & Ikeda, 2022; Sato, 2012). Various factors have been brought to attention in the co-construction of discourse in paired and group interactions helping improve understanding of the IC construct. The factors studied include proficiency level (Artunç & Hart, 2020; Youn, 2020), task effect (Author et al., 2016; Brooks, 2009; Vo, 2019), interlocutor effect (Davis, 2009; O’Louglin, 2002; O’Sullivan, 2002), and planning time (Lam, 2019) among others. On the other hand, a few studies have focused on exploring patterns of interactions that emerge in a social context under the effect of various factors. These studies have specified a number of interactional patterns for IC. Instead of focusing on micro-features of IC, such studies have explored how interactions at a more macrolevel are patterned. As such, they have focused on individuals’ collaboration and engagement in interactions to classify different patterns of interaction. To this end, in a longitudinal classroom-based study, Storch (2002) investigated the nature of interaction in an ESL class including 10 pairs of adult students. She found four patterns of dyadic interaction based on the two criteria of mutuality (engagement with the interlocutor’s contribution) and equality (degree of control over the interaction): collaborative, dominant/dominant, dominant/passive, and expert/novice.

Galaczi’s (2008) analysis of the discourse in interactions of 30 pairs of First Certificate in English (FCE) candidates led to a similar categorization of interactional patterns. She categorized the interactional patterns into collaborative, parallel (dominant/dominant in Storch’s study), and asymmetric (dominant/passive and expert/novice in Storch’s study), and added another pattern called blended (a combination of two patterns in the same dyadic interaction). The study further provided evidence for a relationship between the interactional patterns and IC scores in a way that higher scores were given to test takers in collaborative patterns.

However, there is paucity of research on whether such L2 interactional patterns are similarly found in L1 interactions of different cultures. While extensive research has examined interactional competence in Second Language Acquisition and Assessment, there is a notable gap in the exploration of how interactional patterns manifest in interactions of different linguistic and cultural backgrounds in both L1 and L2 contexts. In Isaacs’s (2013) study, evidence was found for the role of linguistic or cultural background in interactions. For instance, while “Mandarin speakers accounted for 27% of participants in the study, they were represented in 58% of the cases of individuals classified as passive within the asymmetric group”. Also, speakers of Indic languages accounting for 27% of participants “were overrepresented in the asymmetric dominant group at 83%” (p. 238). Similar findings were previously reported by Lu (2010) who found that Chinese EFL test takers had less elaboration and were less interactive in their spoken discourse than the Italian EFL test-takers, a finding that was explained by the taciturn culture of the Chinese in preferring to be brief and concise in response to elders and authorities to show respect. However, these studies are limited to L2 learners and could not provide a clear picture of how L1 and L2 interactions compare. In fact, existing studies have predominantly focused on L2 learners, leaving unanswered questions about the comparability of interactional patterns between L1 and L2 contexts and across different cultural and linguistic groups. To the best of the researcher’s knowledge, no study has been conducted to investigate how native speakers of different languages co-construct discourse in their L1 interactions. Nor has been any study to explore how such interactions are different from L2 interactions. As such, and in response to the call for cross-cultural research on interaction (e.g., Galaczi & Taylor, 2018; VO, 2020) to help define the construct of IC, the current study was conducted to specifically fill the gap on patterns of interaction. The study could be significant as this is the first study to explore interactional patterns in relation to the language of interaction which encompasses linguistic and cultural background. The study was specifically set to explore and compare the interactional patterns in paired interactions of L1 English speakers, L1 Persian speakers, and L2 English learners.

Methods

Participants

Ninety university students (27 males and 63 females) from three different language backgrounds participated in the study (Table 1). Convenient sampling was employed to select the participants in all the three groups; that is, the participants were selected based on their availability and willingness to participate in the study. The first group, L1 Persian speakers, included 30 native speakers of Persian. At the time of the study, they were either university students or had just finished a degree at university. They had majored in different fields and were selected from different universities in Shiraz and Tehran, two major cities in Iran. They were monolinguals of Persian since they had either no or a low level knowledge of any other languages as expressed by themselves before being included in the study. The second group, L1 English speakers, consisted of 30 native speakers of English. They were all American students who spoke English as their L1. Like the first group, this group only included participants who were basically monolinguals with no or a low level knowledge of other languages. They were majoring in different fields of study and were selected from different universities and colleges in Stillwater in Oklahoma and Princeton in New Jersey of the USA. The third group, L2 English learners, included 30 EFL learners who were selected from different universities in Shiraz, Iran. They were Iranian native speakers of Persian. Proficiency was not a variable of interest in this study; however, to make sure the participants were proficient enough (intermediate and above) to take part in paired interactions, they attended an interview with the researcher and research assistant. All the three groups of participants were provided with some general information about the study. They were also ensured that their data would only be employed for research purposes, that they would remain anonymous, and that they could withdraw from the study at any time.

Table 1 Participants’ demographic information

Instruments

The literature has discussed the advantages of paired test over the traditionally dominant interviews and monologues (e.g., Brooks, 2009; Ducasse & Brown, 2009; Galaczi, 2014; Nakatsuhara, 2004; Plough et al., 2018; Taylor, 2001; Vo, 2020). A paired oral test was used in this study to collect the data on interactional patterns. The participants had one minute for preparation and five minutes for discussion. The participants were supposed to discuss whether physical punishment should be allowed at schools or not. The three groups of participants sat the same oral test but interacted in different languages. The native speakers interacted in their L1s, therefore, group I interacted in Persian and group II in English. The L2 learners interacted in L2; that is, English. The participants of each group were randomly assigned to pairs. The interactions were audio-recorded with the participants’ consent for later analysis. Unless deemed necessary, the researcher and research assistants did not interfere with the interactions and functioned as silent observers.

Data collection

The data were collected in two steps. First, paired interactions were collected from different contexts in line with the purpose of the study. The L1 English interactions were collected with the help of two research assistants living in America. Second, all the recorded interactions were transcribed using conventions adapted from Atkinson and Heritage (1984; see the Appendix). To ensure accuracy, the data were transcribed by two different individuals and then were checked by the researcher. The L1 English and L2 English interactions were first transcribed by a BA graduate of ELT who had experience transcribing English interactions. The transcripts were then given to a PhD candidate of ELT to check and make corrections if necessary. After that, about 30% of the samples were randomly selected and checked by the researcher to ensure accuracy. Very few corrections were made by the researcher which showed the accuracy of the transcriptions by the two individuals. A similar procedure was followed for the L1 Persian interactions. First, as an L1 Persian speaker, the same BA graduate who had transcribed English interactions transcribed the data. Then, another L1 Persian speaker with an MA degree in Teaching Persian to Speakers of Other Languages, checked the transcription, and finally the researcher checked 30% of the data and made a few minor corrections.

Data analysis

The transcripts were analyzed qualitatively for interactional patterns using Galaczi’s (2008) and Storch’s (2002) typology of interactional patterns. To determine the patterns, the transcription of each paired interaction was first read to get an overall impression of the interaction. Then, the interaction was read again and checked against the detailed characteristics of each interactional pattern provided in Galaczi (2008) and Storch (2002). The interlocutors’ roles were carefully studied on a turn-by-turn basis. The iterative reading and rereading of the interactions were followed in this stage. The patterns were distinguished based on the two dimensions of mutuality and equality (Storch, 2002). To ensure the reliability of the general coding of the patterns, 33.33% of the interactions were randomly selected from the three groups (five samples from each group) and checked by a PhD candidate in ELT who was doing her PhD on IC. As she was familiar with interactional patterns and relevant studies on IC, only one brief session of training was provided to her explaining the purpose of the analysis and the distinguishing characteristics of different patterns. She was then asked to analyze and code the 15 random samples. The intercoder agreement turned out to be 70%. Disagreements were mostly related to the blended pattern. This was not unexpected as this pattern is “a blend of two interactional styles” (Galaczi, 2008, p.97), and in cases it proved difficult to determine whether a second style was obviously present in the interactions or not. As such, it was decided to consider a pattern blended when there was a clear shift in interactional roles which was maintained in several consecutive turns, otherwise the interaction was labeled by the dominant pattern of the interaction. The problematic cases were discussed and resolved accordingly. The rest of the data was then coded by the researcher. The following section provides a more detailed explanation of the analysis of patterns. Exemplar excerpts are used to illustrate the patterns.

Results

Interactional patterns

Following Storch (2002) and Galaczi (2008), interactions that exhibited moderate to high mutuality and moderate to high engagement were considered collaborative. In such interactions, both interlocutors actively contributed to the task developing their own topic (self-focused) and contributing to the development of their interlocutors’ topic (other-focused). Such interactions were highly cohesive and coherent with interlocutors completing each other’s utterances or incorporating or repeating each other’s’ utterances and then extending on them. Such interactions could further be characterized by short gaps between turns, overlaps in turns, frequent acknowledgements, questions, requests for clarification, and positive and negative feedback. The following excerpt which is part of an exchange between two L1 English speakers, a man and a woman, illustrates the collaborative pattern.

Excerpt 1

• 1 A: I don’t think that anyone should discipline your chi:ld except fo:r the parents. So: it's my personal opinion on that.

• 2 B: exactly

• 3 A: you know

• 4 B: why is that though?

• 5 A: why is that? (.) Because that’s your child.

• 6 B: yeah, but it’s like (.) punish your child at home, so (.) why shouldn't (.) your child be punished at school? (3)

• 7 A: good point, (.) but physically?

• 8 B: (.) physically

• 9 A: no

• 10 B: you never (.), you would never physically punish your child?

• 11 A: no

• 12 B: not at all?

• 13 A: would I? No. [So] I wouldn't want someone else to do it. (?) (3) Do you not agree with me?

• 14 B: [really]? I mean, I, I do.

• 15 A: I can't. So, that's all I can say about it. No, I don’t agree with it.

• 16 B: I don’t agree with it either. But (.) I mean I came up (.) with being punished physically punished when I did something wrong, so

• 17 A: physically punished?

• 18 B: physically punished like getting a whipping, yeah.

In this interaction, both participants jointly contribute to the discourse and engage with each other’s ideas. Their speech is characterized by features of a highly-involved interaction (Galaczi, 2008; Tannen, 1981; Van Lier, 1996) such as short turns, rapid speaker change, avoidance of gaps between turns, supportive interruptions (e.g., turns 14 & 17), frequent questions (turns 4, 6, 7, 10, 12, 13 & 17) and listener support in the form of agreement (turns, 2 & 16). Cohesion is created as the interlocutors repeat or incorporate each other’s ideas (e.g., turn 5), and extend on or complete each other’s ideas (e.g., turn 7). Further characteristics of this interaction which makes it collaborative is the balanced quantity of talk between the two participants, and also balance in initiating and expanding topics (Galaczi, 2008). Both participants easily shift between their roles as speakers and listeners. A salient feature of this interaction is the frequent use of questions. By raising questions interlocutors impose conversational limitations on each other’s contribution as they have no choice but to take the floor and speak in a relevant manner (Galaczi, 2008). In this excerpt, both interlocutors use questions to extend the topic over several turns and shift between their conversational roles as speakers and listeners.

Parallel (similar to Storch’s dominant-dominant) interactions were characterized by a moderate to high level of equality but a moderate to low level of mutuality. This means that while both interlocutors actively contributed to the task, they lacked willingness or ability to fully engage with each other’s contribution (Storch, 2002). As such, they mostly developed their own topic and rarely tended to develop the topic initiated by their interlocutors. Unlike the collaborative interactions which showed a balance in speaker and listener roles, parallel interactions illustrated a much stronger role for the speaker than the listener (Galaczi, 2008). Such patterns were further characterized by limited listener support for example in the form of acknowledgement tokens, and rare syntactic and lexical cohesion between turns. The two excerpts below are used to illustrate the parallel interaction. The first excerpt is part of an interaction between two female L2 English learners.

Excerpt 2

• 1 A: I think mm (.) the biggest impact that physical punishment ha:s on children (.) is fear. (.) And when you have fear you have no: confidence. (.) You have no: self-respect. You have no: self-esteem. And this fear ruins their, mm their spirit, their soul (.) and they cannot be (.) successful (.) in in their future.

• 2 B: yeah, and also when the adult punish the children (.) they feel that the adult are in the position of the power. (.) [Now] when children find the position of power, when they are older than other children (.), they just react the same as their (.) parents, their teachers, their, (.) their adults.

• 3 A: [Ok] exactly, that, that was what I was thinking. It's like (.) and also: (.) revenge [is] another concept. (.) When they have the power (.) they want to get this revenge from others, [from] their children, from their students, if, (.) and it's (.) I think, not good at all.

• 4 B: [yeah] and also, when we talk about punishment we (.) want the idea of (.) lea:rning and (.) tea:ching something. Children should learn and (.) adult should teach them something to children. But with physical punishment we don't tea:ch anything to them. (.) We just teach them how to hate, how to detest, how to have fear, how to (.) have low confidence and other things.

• 5 A: yeah, I do agree. And (.) I think physical punishment should be replaced by ignorance, [reward] and ignorance. For example, if a child (.) has done something wrong, (.) I think we should mm, ignore them (.), not pay attention to them and replace this attention with something good or something better. [In this] way this bad behavior is forgotten. Or (.) I don't know, it’s (.) replaced by something good.

• 6 B: [yeah] [yeah] and also, (.) I think, again I wanna say that it's very important to teach ss- (.) to teach something when we ah punish children. (.) And if we show the consequences of the bad acts that children did in the past (.) they understand it, o:k, I did it, so this is the result. (.) And they, step by step, they will understand it.

Although both participants contribute equally to this interaction there are no collaborative completions, therefore, there is a low level of mutuality or engagement with each other’s ideas. This excerpt involves six turns with as many as six different topics (fear, ignorance, power, revenge, learning and teaching, consequences (teaching). Almost none of the topics gets extended over two turns except for some extension in turns 3 and 6. Each participant initiates and extends their own topic without developing their interlocutor’s topic. While there exist some acknowledgment tokens (turns 2, 3, 5, & 6) they are rather superficial agreements because unlike collaborative patterns these agreements here are followed by very little or no extension of the other’s ideas. That is, after a minimal response (e.g., yeah), the interlocutor initiates and extends a new topic ignoring what was stated in the previous turn by the other interlocutor. Such a minimal response is “just a turn-taking device, which is then often followed by a topic shift” (Glaczi, 2008, p. 106). It creates a pro forma cohesion but actually functions as a topical disengagement (Jefferson, 1993). Also, there are few cohesive links between turns (Galaczi, 2008). Another salient feature of this interaction is the long turns taken by each of the interlocutors. That is, unlike the collaborative patterns which were characterized by more short turns indicative of engaged talk, here the turns are longer and therefore fewer. The interlocutors keep the floor longer to be able to extend the topic in a turn assuming they have other topics to initiate and develop in other turns.

The second excerpt below like the previous interaction shows a parallel interaction with a high level of equality and a low level of engagement; that is, both interlocutors actively contribute to the task but have little engagement with each other’s ideas. Each interlocutor initiates a topic and extends it in a single turn with limited links between the turns. However, this excerpt, happening between two L1 Persian speakers, an Iranian woman and an Iranian man, illustrates a rather specific type of parallel interaction. The whole interaction here including 593 words is completed in five very long turns (for space limitations, only the beginning of each turn showing the topic is presented here). Long turns were also observed in the previous interaction from the L2 English group, however, there the interaction with almost the same number of words (including 604 words) was extended over 15 turns. But here in this interaction, there are as few as five turns and the turns turn out to be roughly three times longer. The speakers do not compete for the floor at all. They patiently wait for their interlocutor to finish their turn and signal that their turn is over so that they can start their turn. Because of this, there is not even a single example of overlapping or interrupted turns. The other parallel interactions in the L1 Persian group had even fewer turns.

Excerpt 3

• 1 A: tænbihe bædæni kolæn (.) tu mædɒres modʒɒz næbɒjæd bɒʃe. (.)be xɒtere ʔinke: (.)tænbihe bædæni bɒʔes miʃe ke dɒneʃʔɒmuz (.) bæʔd ʔæz modæti ʔɒdæt kone (.). va bærɒʃ ʔɒdi miʃe. …

Physical punishment should not be allowed at schools at all, because physical punishment makes the student get used to it after a while, and it becomes usual for them. …

• 2 B: bæ:le (.) be næzære mæn hæm tænbihe bædæni ʔæslæn xub nist. bætʃehɒ bædtær gostɒx miʃæn. … (4)

Yes, in my opinion too, physical punishment is not good at all. Children become ruder.

• 3 A: mæn ʔælʔɒn (.) tæGribæn (.) sijose sijotʃɒhɒr sɒle moʔælem hæstæm dɒneʃʔɒmuzɒni ke ʔun ʔævajele xedmætæm (.) mididæm bɒ ʔælʔɒn xejli færG kærdæn. yæʔni ʔælʔɒn (.) dɒneʃʔɒmuz (.) ræftɒre moʔælem kerdɒre moʔælem ro bebine ʔæslæn xodeʃ (.) motævædʒeh miʃe moʔælem ʔæzæʃ tʃi mixɒd. …

I have been a teacher for almost 33, 34 years now. The students I used to see at the beginning of my service are very different now; that is, now when a student pays attention to the teacher’s behavior, the teacher’s actions, they themselves will understand what the teacher wants from them. …

• 4 B: ʔæslæn tænbihe bædæni dʒævɒb nemide. (.) ʔægær dʒævɒb midɒd xejli ʔæz ræftɒrhɒ bɒjæd tæGir mikærd tu bætʃehɒ. (.) be Gejr ʔæz ʔun hæm væGti tu gofteguhɒje xɒnevɒdegi ʔɒdæm miʃine sohbæt mikone hætɒ bɒ pedær bærɒdær jɒ xɒhær bærɒdærɒje bozorgtæri ke (.) ʔæz næslhɒje gozæʃte budæn (.) hi:tʃ kodumeʃun mm xɒtereje xu:bi tæʔrif nemikonænd. (.) hæmiʃe ʔæz ʔun dowrɒneʃun be ʔonvɒne (.) bædtærin dowrɒn jɒd mikonæn….

Physical punishment does not work at all. If it did, many children’s behaviors would have to change. Apart from that, when in family talks you speak to the father, sister or elder sisters and brothers from previous generations, none has a good memory. They always remember that time as the worst time.

• 5 A: dɒneʃʔɒmuzɒji ke vaGeʔæn tænbihe bædæni miʃodænd sɒbe:G (.) jek (.) xɒtereje xejli bædi dɒræn. …

Students who were really physically punished in the past, have a very bad memory.

The following excerpt illustrates the least frequent pattern in the data; asymmetric pattern. This pattern overall includes two sub-patterns: the dominant-passive pattern demonstrates moderate to low levels of both equality and mutuality and the expert-novice pattern shows moderate to low equality but moderate to high mutuality. In the first one, one person takes the floor and dominates the conversation and the other interlocutor adopts “a more passive subservient role” (Storch, 2002, p. 129). In the expert-novice pattern, the dominant person as an expert encourages the other person to get engaged. The following excerpt between two female L2 English learners, illustrates the expert-novice pattern. Due to space limitations, parts of the long monologues are omitted and replaced by three dots.

Excerpt 4

• 1 A: In my opinion, mm physical punishment should be mm should be rem- removed from the schools er because I think mm it will have poor (.) I think it will have bad effect and (.) on the on the children. They will be you know their mm their potential, mm mm their potential skills may be mm may be repressed, instead of, instead of being flourished. … What's your opinion about it?

• 2 B: Ok, it's very bad punishment. Physical punishment is very bad because the students are mm very mm angry and don't like er this it. And (.) I think mm when we want punish mm the students (.) for example mm mm their notes again and again write notes and [mm] [yes] instead of physical punishment

• 3 A: [writing notes]? [instead of physical punishment]? But in my opinion it's better to talk to these kinds of students (.) in order to understand the the main reason of not (.) not obeying mm their (.) the order of their teachers. …

• 4 B: yes, and mm talk to mm [their] [family] yes and mm (5) their family mm talk to her par mm her children. And (.) if may be better than the students.

• 5 A: [parents]? [Their parents?] It will have better results.

• 6 B: yes

• 7 A: I have the same opinion as you. (3) and (5)

• 8 B: and if the students improve her lessons, mm ah we mm take some (?) gifts for the, for them and (3)

• 9 A: you know besides physical punishment mm we have some mm some bad res- some bad effects on the kids [yes] on them, on the children. You know, they may feel ashamed. They may feel shy in front of their [peers], their students, their classmates and mm they will be repressed instead of flourish. And I [think] yeah. It will (.) it will make them alone. … (5)

• 10 B: [yes] [students] [and maybe depressed]

• 11 A: is that finished?

• 12 B: yes

In this excerpt, A is clearly dominating the interaction by initiating more topics and developing them. Most of the topic initiation and extension is done by this speaker. She speaks more and takes longer turns. The second speaker is passive both because of her interlocutor’s dominating behavior and because of her own failure to take the floor. For example, in turn 10, after having a very long turn, the dominating speaker has a long pause to give the floor to her interlocutor, but she fails to say anything, so the dominating speaker raises a question which is only answered by a minimal response ‘yes’. The dominating speaker, however, functioning as an expert tries to involve her interlocutor in the interaction and invites her participation by raising questions (e.g., turns 1 & 11) and by having a long pause to let her take the turn (turns 7 & 9). In turn 7, the dominating speaker who has received a minimal response in the previous turn from her interlocutor, provides an agreement followed by a 3-second pause to let the passive speaker continue with and elaborate on her ideas presented in turn 4. But after receiving no contribution she says ‘and’ has a long 5-second pause to encourage her passive participant to contribute to the task, a strategy which works as the passive participants contributes in turn 8. Like a teacher she encourages her interlocutor to participate (Storch, 2002). She is authoritative but not authoritarian (van Lier, 1996). She further tries to get engaged with her interlocutor’s contribution by completing or extending on her ideas (turns 3 & 5) and agreeing to her ideas (turn 7).

Frequency and prominence of interactional patterns

Table 2 illustrates the frequency and prominence of interactional patterns for the three groups. As depicted, the collaborative pattern was found to be the most frequent interactional pattern in the L1 English group dominating more than half of the interactions (53.33%), and in the L1 Persian group dominating 40% of the interactions. In the L2 English group, the frequency of this pattern drops to only one third of the interactions and equals the frequency of the blended pattern. It is further indicated that the asymmetric pattern is the least frequent pattern of interaction in the L1 Persian and L2 English groups with a low frequency. The L1 English interactions did not display this pattern at all. Also, more blended and asymmetric patterns were found in L2 than L1 interactions

Table 2 Interactional patterns of different language groups

Further analysis indicated considerable variability in the L1 Persian interactions. The parallel and asymmetric interactions in the L1 Persian group included fewer turns and therefore were less interactive. Actually, in two of the parallel interactions and in the only asymmetric pattern of the L1 Persian group, there were as few as three turns (parallel patterns at most included five turns). These patterns were vividly different from the other L1 Persian interactions and also interactions in the L1 English and L2 English groups in terms of the number of turns. In the asymmetric pattern, the dominant speaker initiated the interaction with a turn that lasted for four minutes. During this time, the other interlocutor was just quietly listening and waiting for her turn, seeming satisfied with this waiting. She did not show any sign of competing for the floor or taking the turn until the dominant speaker was done with his speech and gave her a signal to take the turn. Then, she took the turn which lasted for 20 seconds. Finally, the first speaker took the turn again and completed his talk. In this interaction, 80% of the discourse was created by the first speaker in one turn. The whole interaction looked like a monologue which included a 20-second turn by the second speaker. In the two parallel interactions, each speaker initiated and presented their ideas independently and in detail in one or two turns with a very low level of mutuality or engagement.

Discussion

The purpose of this study was to investigate whether IC is displayed differently in interactions of different language groups. To this end, paired interactions of L1 English speakers, L1 Persian speakers, and L2 English learners were compared based on the interactional patterns. The findings indicated that the collaborative pattern was the dominant pattern in all the language groups. Less than half of the interactions in total were collaborative, and the blended patterns also mostly included the collaborative pattern. However, differences were also found among the three groups. Not all the patterns were found in all the groups. Similar patterns also varied in their frequency and prominence (rank-ordering of the patterns). Previous research (Galaczi, 2008; Isaacs, 2013; Storch, 2002) has reported differences in the frequency of interactional patterns. While Galaczi (2008) reports a frequency of 30% for each of the collaborative, parallel, and blended patterns, and 10% for the asymmetric pattern, Storch (2002) reports 50% for collaborative, 20% for parallel, and 30% for asymmetric. Also, Isaacs (2013) reports 36% for collaborative, 38% for parallel, and 26% for asymmetric. Since these studies are specifically conducted on L2 interactions, they have linked the findings to the proficiency level of interlocutors with lower levels indicating more parallel or asymmetric patterns and higher levels showing more collaborative styles of interaction. While proficiency can be a logical reasoning for why some of the L2 interactions are not collaborative it cannot explain why native speakers’ interactions are not collaborative. Thus, it can be argued, based on the findings of the current study, that the linguistic and cultural background of interlocutors can play a role in interactional patterns as many of the L1 interactions in the two native groups were not collaborative. This finding is in line with L2 interactional studies (Fulcher, 2003; Galaczi, 2008; Isaacs, 2013; Lu, 2010; May, 2009; Ross, 1998, 2018; Young, 1995) which had previously argued for the effect of linguistic and cultural backgrounds on L2 interactions. Such studies, however, did not include L1 interactions to be able to specifically illuminate the effect.

Another justification for the findings of this study may come from the interlocutor effect in pair work. It has been discussed in the literature that using paired tasks does not necessarily lead to collaboration in L2 interactions, (Foot, 1999; Galaczi, 2008; Norton, 2005; Storch, 2002; Taylor, 2001; Teasley & Roschelle, 1993). The findings of the current study lent support to this idea by indicating that not only in L2 interactions but also in L1 interactions, the use of paired tasks does not necessarily lead to collaboration. This may mean that interactional patterns also depend on the pairing (who you are paired with in an interaction) (Artunç & Hart, 2020; Isaacs, 2013; Kim & McDonough 2008; May, 2009). In Artunç and Hart’s (2020) study, the interlocutor effect on patterns of interaction was confirmed in L2 interactions. The most frequent patterns in high-high pairs, low-low pairs and low-high pairs were collaborative, parallel and asymmetric respectively. Furthermore, no collaborative pattern was found in low-low pairs.

Finally, the findings about the parallel and asymmetric patterns in the L1 Persian group with very few turns can further highlight the complexity of the way the linguistic and cultural background and the interlocuter (pairing) effect function in tandem to create very specific and unique patterns of interaction. The fact that such patterns of interaction only appeared in the L1 Persian group can be justified by the cultural norms of speech. Longer turns may mean the speakers do not consider interrupting a partner’s turn acceptable and wait for their turns (Isaacs, 2013); in other words, they expect “to be invited to speak” (Galaczi, 2008, p.110). Furthermore, the fact that such patterns appeared only in a fraction of the L1 Persian interactions can highlight the interlocutor (pairing) effect.

Conclusions and implications

This study explored the interactional patterns in L1 and L2 contexts. The findings provided insights into the role of linguistic and cultural background in the patterns of interaction in paired oral tests. Currently, high-stakes tests are criticized for their failure to include appropriate interaction-based tasks (so, the construct of IC is underrepresented in these tests) or for their failure to consider IC-specific scoring criteria even when such tasks are included (e.g., Malone & Montee, 2014; Roever & Dai, 2021; Roever & Ikeda, 2022; Roever & Kasper, 2018; Seedhouse & Nakatsuhara, 2018). Thus, high-stakes tests are highly encouraged to include in their test batteries interaction-based tasks specifically targeting IC and to develop appropriate rubrics to measure IC. “In the overwhelmingly typical case where test-takers’ real-world language use involves interacting with others, it seems clear that their ability to do so should be a core part of the information gathered on their language ability” in language tests (Roever & Dai, 2021, p. 24). The challenge, however, is the role of cultural norms of speech in the final product of such interaction-based tasks. The findings of this study indicated that cultural norms along with the interlocutor effect may play their role in the pattern of interaction created in paired tasks. The problem arises when we consider that the type of pattern affects the scores assigned to the interlocutors (Galaczi, 2004, 2008; Isaacs, 2013; May, 2009; Storch, 2002). This can threaten the validity of the decisions made based on such scores and may have detrimental consequences for test takers.

The findings may have several implications. First, the literature has suggested the collaborative pattern as the most effective pattern in helping students learn a second language (Kim & McDonough, 2008; Storch, 2002). “Learners might benefit from explicit teaching of collaborative dyadic interaction” (Galaczi, 2008, p.114). They could receive higher scores (e.g., Davis, 2009; Galaczi, 2008; Isaacs, 2013; Storch, 2002) and scaffold peers (Storch, 2002) when interacting in a collaborative talk. In Isaacs’s (2013) study, interlocutors in the collaborative group had the highest satisfaction of their interaction among all the groups. They also had the fewest problems in pronunciation and communication and favored more the idea of collaborating with their partner in the future. To benefit from the advantages of this collaboration, paired tasks are suggested as they are assumed to be interactive by nature. However, findings of the current study indicated that paired interactions of L1 and L2 are not necessarily collaborative. Therefore, teachers should know that using pair work in their classes is not enough for their students to become collaborative. They need to be well equipped with other strategies of making students collaborative. Furthermore, “the question remains as to whether collaborative interactional patterns should be uncritically positioned as the ‘gold standard’ of communication” (May, 2009, p. 418) when many of L1 and L2 interactions are non-collaborative. This is a question which calls for further research.

Moreover, as discussed previously, cultural norms of speech and interlocutor effects may impact upon interactions and create specific patterns which are not at all collaborative. This can imply a big challenge for fair assessment when using paired tasks as assessment tools. How can we make sure that test takers receive fair scores when they do not have equal opportunities to display their speaking ability in different interactional patterns?

The last implication is related to scoring paired speaking interactions. Rater training programs are recommended to include detailed discussions and rating of interactional patterns with a focus on non-collaborative patterns to help raters improve their understanding and rating of paired interactions. Also, rubric developers are recommended to include in their rubrics descriptors of different interactional patterns to help raters with fairer scoring of paired interactions.

The study may have some limitations. First, the findings of the study may have been affected by the fact that demographic variables in the pairs were not controlled. Further explorations are therefore suggested on the effect of demographic variables on interactions in pair work. Second, the findings may have been influenced by regional variations as the sample was limited to specific locations in two countries.

Appendix

Transcription conventions

[ ]overlapping utterance

?rising intonation

.falling intonation

…omitted parts

boldfacewords pronounced with stress

-abruptly ended sound

: lengthening of vowel sound

(.)pause shorter than 3 seconds

(5)length of pause if ≥ 3 seconds

(( ))nonverbal sounds (e.g., laughter)

(?)incomprehensible utterance

Adapted from Atkinson & Heritage (1984)