Classroom discourse in lower secondary French-as-a-foreign- language classes in Norway: Amounts and contexts of first and target language use

This article investigates amounts and contexts of target language (TL) use by teachers and pupils in French-as-a-foreign-language classrooms in Norway, an educational context in which a communicative approach to language teaching is prescribed but target-languageonly teaching has not been the prevailing idea. A total of 45 video-recorded French lessons from the last two years of lower secondary school were analyzed for language use and instructional activities. In most classrooms, the L1 was the language of instruction. TL use was restricted to speaking exercises, greetings, and vocabulary instruction and did not seem to increase with class level. The findings diverge in many respects from those in previous research conducted in other educational contexts. By adding data from an underresearched context to the existing body of knowledge, the present study nuances the debate on L1 use in second and foreign language learning and thus the recommendations for classroom practice commonly found in the literature. It also outlines a detailed methodological framework that can be adopted in future studies to obtain data that are comparable across contexts. © 2020 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).


Introduction
The balance between target language (TL) and first language (L1) use in second and foreign language classrooms is a debated issue (Butzkamm, 2003;Ellis & Shintani, 2014;Turnbull & Dailey-O'Cain, 2009). Following the rise of communicative language teaching (CLT) and immersion techniques in the 1970s, the TL-only approachdwhich excludes pupils' L1 from the classroomdwas long considered the ideal in many contexts and even prescribed in the 1990s by official policies in several countries, especially English-speaking ones like England and Wales (Macaro, 2001), and Canada (McMillian & Turnbull, 2009). The TL-only approach has also been dominant in the context of English teaching in former British colonies (Philipson, 2016). Most previous research on TL use was set in contexts where the principle of TL-only prevailed. The recommendations for classroom practice currently found in the literature are thus based primarily on findings from classrooms in which the TL is the main language medium (Shin et al., 2019). These recommendations include what is often called 'judicious use of L1', which most current researchers believe can facilitate TL learning and fulfil important functions in second and foreign language learning (Cook, 2001;Ellis & Shintani, 2014;Shin et al., 2019;Swain & Lapkin, 2000). review, we focus on studies that address L1 and TL use in secondary education. We also focus on recent studies (from 2000 onwards). For a review of studies from before 2000, see Turnbull and Arnett (2002).

Amount of L1 and TL use
Previous research indicates that there is considerable variation in L1 and TL use in SL and FL classrooms. Macaro (2001) video-recorded six student teachers of French in lower secondary schools in the UK to determine the amount of teacher L1 use. He found L1 use to be very limited: on average, L1 was used for 6.1% of the total speaking time. A larger survey study from Australia (Crawford, 2004) showed considerably higher levels of L1 use. Teachers of seven different languages in primary and secondary education (n ¼ 581) were asked to estimate their amount of TL use. The majority placed themselves in the group using TL less than 40% of the time. It is not clear whether these percentages were related to lesson or (teacher) speaking time, but the proportion of L1 use was much higher than in Macaro's study. Littlewood and Yu (2011) showed that TL use varies considerably between contexts. Fifty tertiary students in Hong Kong and Mainland China were asked to recall the percentage of classroom time their teachers spent using L1 in the secondary school English classroom. There was a considerable difference between the two contexts: 43% of the Chinese students estimated that their teacher used L1 more than 75% of the time, compared to only 5% of the Hong Kong students. Although estimates based on memory must be interpreted with caution, the findings indicate that the two contexts have very different language teaching traditions.
Several studies also indicate considerable individual variation in TL use. Bateman's (2008) study of US student teachers' attitudes towards TL use employed questionnaires, journal entries, and observations to estimate TL use by ten future Spanish teachers. The estimated time of TL use varied from 5% to 95%, with the most common answer being approximately 80%. Although it is unclear whether the estimates related to the entire lesson or to the teachers' speaking time, it is clear that, even among participants in the same course, TL use can vary considerably. Kim and Elder (2008) used audio recordings and interviews to study the practices and perceptions concerning language use of two native speaker teachers of French and Korean, respectively, who taught at the beginner level in secondary schools in New Zealand. The amount of TL use was measured as the number of speech units in a few selected segments of the audio-recorded lessons. The French teacher used far more TL than the Korean teacher (approximately 80% vs 20% for the analyzed segments), a difference which was attributed to language status and distance between the TL and the L1. Liu et al. (2004) used audio recordings and questionnaires to determine South Korean high school English teachers' use of L1 and TL in class. The amount of each language was measured using word count. The authors found considerable individual variation, but the average amount of TL use was rather low (32%). Riordan (2018) used a similar methodology to determine the amount of TL use by non-native-speaking German teachers in Irish secondary schools. She also found considerable individual variation but higher average TL use (46%). She also observed that the teachers tended to over-estimate their amount of TL use.
In a Nordic context, inspectorate reports indicate that TL use in FL classrooms is limited (Skolinspektionen, 2010). Research studies that actually measure TL use are scarce. Stoltz (2011) is a notable exception. He used audio recordings to investigate interactional patterns of L1 and TL use in two French L3 classrooms in upper secondary schools in Sweden. Unlike many other studies, which focused on the teacher's choice of language, Stoltz was interested in both the teacher's and the pupils' language use. He found that the proportion of TL use varied with instructional patterns, from approximately 56% in plenary teaching (measured in speech units) to more than 90% in conversations between Swedish pupils and visiting French-speaking exchange pupils. In group work (with or without teacher assistance), around 70% of the speech units were in French. However, Stoltz also measured the amount of L1 and TL by the number of words. Measuring words instead of speech units radically changed the picture, as the pupils' speech units in French were very short (typically around three words).
Stoltz's study is the only one to our knowledge that quantified TL use in L3 teaching in Nordic countries. Brevik & Rindal (2020) quantified TL use in English L2 classrooms in Norway, finding that English was the predominant language of communication. However, as noted above, English has a very different status in Norwegian schools than the L3s, and TL use in the classrooms is therefore probably also different.

Functions of L1 and factors determining language choice
There seems to be less variation in the functions of L1 than in the amount. Previous research shows that teachers use L1 to teach grammar and culture, deal with behavioural issues, give complex instructions and build relations (Aberdeen, 2018;Askland, 2018;Bateman, 2008;Crawford, 2004;Heimark, 2013;Littlewood & Yu, 2011;Macaro, 2001). Subject areas and language functions thus clearly influence teachers' language choice. Stoltz (2011) found that activity or task type also played a major role in the learners' language choice.
Learners' TL proficiency is often considered the most influential factor in the teacher's amount of TL use (Hall & Cook, 2012). However, several studies question this idea. Macaro's (2001) study included first-, second-and third-year French classrooms, but no significant link was found between the level of the class and the amount of L1 use. Meiring and Norman (2002) used questionnaires (N ¼ 46) to investigate estimated TL use among secondary school FL teachers in the UK. The teachers reported using more TL with beginners than with pupils who had studied the language one year longer. Although class level does not equal proficiency level, it is to be expected that pupils in higher classes have a higher mastery of the TL on average than pupils in lower classes.
Studies from a Nordic context also suggest that learners' TL learning experience does not necessarily have an impact on the teachers' TL use. Heimark (2013) and Askland (2018) set their studies in a secondary school context in Norway. Heimark (2013) used a survey (n ¼ 85), interviews (n ¼ 6) and observations (n ¼ 6 Â 2) to investigate French L3 teachers' beliefs about CLT, whereas Askland (2018) studied opinions and practices regarding grammar teaching among ten teachers of Spanish L3, using interviews and observations. Although neither study quantified TL use, they did not indicate any increase of TL use with class level. The teachers in Heimark's (2013) study identified extensive TL use as the core of CLT, but judging from the observations their actual use was limited at all class levels, with the exception of one teacher. The teachers in Askland's (2018) study chose to teach grammar in the L1 but claimed to use more Spanish for other activities, a claim that the observations could not confirm, except for one teacher. According to the teachers in both studies, the discrepancy between beliefs and actual practice was due to the low TL proficiency of the pupils; they feared the pupils would feel lost if surrounded by the TL. This concern does find some support in the literature. Macaro and Lee (2013) found that teacher TL use could lead to cognitive overload and frustration for some learners. However, their study addressed a TL-only context. Pupils in contexts where the TL-only approach does not prevail often express a desire for more TL use in class (Fern andez & Andersen, 2019).
The impact of the teachers' proficiency level is also subject to debate. Several studies emphasized the impact of this factor on TL use (Bateman, 2008;Chambless, 2012;Crawford, 2004;Liu et al., 2004;Riordan, 2018). Crawford (2004) found that teachers who rated themselves as very proficient in the TL reported using it more than those who rated themselves as less proficient. Other studies attached less importance to the teachers' proficiency level (Askland, 2018;Kim & Elder, 2008). In Askland's (2018) study, for instance, teachers with low TL proficiency made little use of the TL, but this was also the case for native-speaking teachers.
Other factors that might influence the amount of TL use are traditions and attitudes as well as local teaching cultures (Littlewood & Yu, 2011), language distance and status (Kim & Elder, 2008), programme goals (Crawford, 2004), teacher fatigue, and available preparation time (Bateman, 2008;Heimark, 2013). Bateman emphasized the time issue and argued that even teachers with a high level of TL proficiency require more time and preparation to teach in it. Riordan (2018) made a similar point when arguing that even teachers who are highly proficient TL speakers might lack the language necessary to perform functions that are specific to the language classroom, as they have never learned such context-specific vocabulary and structures.

Methodological concerns
This overview of previous research shows that TL use in FL contexts varies considerably, as do the methods for measuring it. Different measurement techniques can yield significantly different results (Stoltz, 2011). Moreover, much of the research on the amount of L1/TL use is based on self-reports and estimates by teachers or learners (typically through questionnaires). Several studies combining self-reports with measures of actual language use through audio or video recordings have shown that there are often notable gaps between perceived and actual practice (Aberdeen, 2018;Duff & Polio, 1990;Hall & Cook, 2012;Riordan, 2018). As for studies that measure actual language use, the findings are rarely directly comparable, either because the measurement techniques differ or because the studies do not provide enough details about the methodological procedures. For example, the reference point for measuring TL or L1 use is not always clear in the previous literature; that is, whether it relates to lesson or speaking time. Macaro (2001) and Stoltz (2011) presented detailed procedures for measuring language use. However, the participants in these studies might not be typical representatives of their profession. Macaro's (2001) study included participants who were all successful students of the researcher, and it is unclear whether their extensive use of the TL mirrors the common practice in UK French classrooms. The two teachers who participated in Stoltz's (2011) study both volunteered to join, and they and their pupils were aware of the topic of the study, which might have inspired them to maximize TL use. Regarding the pupils' TL proficiency level and its impact on teacher TL use, studies investigating this issue have been cross-sectional and thus cannot determine whether teachers adjust their TL use in line with the development of their pupils' knowledge and skills.
This study adds to the existing research by measuring actual TL use in a rarely studied context: L3 instruction in Nordic countries. It differs from other studies that measured actual L1/TL use in secondary education in that it includes randomly selected teachers. It is longitudinal in the sense that the same classrooms were recorded in grades 9 and 10. We can thus follow L1/TL use from one year to another. The study was guided by the following research questions: 1. How much L1 and TL is used in Norwegian lower secondary French L3 classrooms? 2. Does TL use increase with class level? 3. In what situations and for what purposes are the two languages used?

Methodology
This study is part of the Linking Instruction and Student Experiences (LISE) project at the University of Oslo. During 2015e2018, the LISE researchers systematically collected video recordings of instruction in six lower secondary schools in Oslo and the surrounding areas. 3 The LISE project recorded instruction in six subjects (including French) over two school years (grades 9 and 10). To ensure variation among the schools, LISE selected urban, suburban and rural schools (two of each) from areas that differed in terms of socio-economic status. Invitations to participate were sent to the head of the school. After receiving his or her approval, an information meeting was organized with the schoolteachers.
Participating teachers, pupils, and pupils' parents all signed written consent forms. The study was approved by the Norwegian Centre for Research Data. Pupils who did not wish to be filmed were placed in a spot outside the view of the camera.

Participants
Six French-as-a-foreign-language teachers and 85 pupils participated in the study. The teachers were randomly selected, as they did not volunteer to join but simply happened to work at the invited schools and were encouraged by the school principal to join, along with other subject teachers at the same school.
The classes varied in size (see Table 1). The pupils (age 14e16) were in their second and third years of beginner French studies and received 2e3 teaching hours of French a week. The teachers were experienced, apart from teacher F, who was relatively newly qualified. She was also the only native French speaker. The teachers had varying amounts of ECTS in French and varied backgrounds from French-speaking areas. Table 1 provides an overview of the participants' backgrounds.

Data collection
Two palm-sized, wall-mounted cameras simultaneously recorded the same lesson: one faced the class, and the other faced the board. The teacher wore one microphone while another was placed in the middle of the room to capture the pupils' voices. This design captured whole-class discussion reasonably well, although in some classrooms the voices of pupils sitting far from the microphone were inaudible. Teacherepupil interaction was always captured because of the teacher's microphone. A member of the research team was present during all lessons but did not intervene in the instruction.
Previous classroom studies suggest that four consecutive lessons provide sufficient information to obtain an overview of teaching practice (Klette et al., 2017). Accordingly, we recorded four consecutive French lessons in each school during each year, apart from School A, which decided to leave the project after the first year. For this school, we have video material from grade 9 only. Altogether, 45 French lessons were filmed: 25 in grade 9 (2015e2016) and 20 in grade 10 (2016e2017).
The duration of the lessons varied between schools (45e70 min), and some schools had double units (i.e. two lessons in a row without a break). Table 2 gives an overview of the video material, including the number of lessons recorded for each school and each grade (with the distribution of single and double units in parentheses) and the number of minutes filmed.

Data analysis
The analysis of the video material occurred in two stages: language coding and activity coding. For both coding procedures, the coder started and ended the coding when the teacher opened and closed the class. The total length of the videos (39h15min) is somewhat longer than the total coded time (38h9min).
First, to answer RQ1 and 2, all speaking time was coded as L1 (Norwegian), TL (French) or L1eTL (mixture). The language coding was performed by an experienced research assistant using Interact software to time-stamp the onset and offset of L1 and TL use by teachers and pupils: A language code was activated when either the teacher or a pupil spoke and deactivated a Small ¼ less than 10 pupils, medium ¼ 10e20 pupils, and large ¼ more than 20 pupils.
3 These six schools were recruited from a sample of 50 schools that participated in the nationwide and large-scale Linking Instruction and Student Achievement (LISA) project or its pilot. The LISA project focused on mathematics and language arts instruction in grade 8. Also a seventh school participated in the LISE project, but this school did not offer French and is therefore not included in the present study.
when they stopped speaking. Since the audio quality was insufficient to capture all pupil-to-pupil talk, we followed the procedure used in Brevik & Rindal (2020) and coded the teacher's speech, pupils' speech to the teacher, and pupilepupil interaction that was captured by the audio equipment. The code L1eTL captured situations in which there were frequent shifts between short utterances (less than 3 seconds) in each language, and this switching of codes lasted for 3 seconds or more. The transcript in Appendix 1 illustrates the L1eTL code. During some activities, neither the teacher nor the pupils spoke. These sequences, which include silent individual work and time spent watching videos or listening to sound files, were not assigned any language code. Additionally, sequences for which it was impossible to identify which language was used did not receive any language code (e.g. when many pupils spoke simultaneously). The amount of time not coded for language use ranges from 1% to 26% of a lesson.
To check the reliability of the language coding, the second author coded one lesson from each school for language use, applying the same principles as the research assistant. The respective percentages of L1/TL use are given in Appendix 2. Based on these numbers, language coding reliability was deemed sufficient.
Second, the videos were coded for type of activity to answer RQ3, which asked in what situations each language was used. The activity codes were created inductively based on what happened in the videos. The first author initially watched the videos and noted the different activities that occurred, the approximate time spent on them and the language used. This resulted in a preliminary set of codes, which were used for a more thorough analysis in which both authors independently rewatched the video with the highest percentage of TL use and used the coding scheme to time-stamp the different activity sequences in that particular lesson. During this process, we frequently compared and discussed our coding and the categories. Table 2 Overview of the video data.

School
No of lessons Minutes  Consequently, the initial coding scheme was refined, and a few overlapping codes were merged. The final set of codes and their descriptions are listed in Table 3. Activities were divided into two main categories: those related to subject matter content and those linked to classroom management and organizational aspects. The categories mirror Ellis' (1994) distinction between medium-oriented goals and framework goals.
A sequence could receive codes from both main categories. For example, a sequence that was assigned the subject content code 'speaking exercise' combined with the code 'tutoring' from the classroom management category depicted a sequence in which the teacher provided individual or group guidance during a speaking exercise, typically by walking from group to group listening and giving advice. Codes from the same main category were mutually exclusive. An advantage of the timestamping in Interact is that it allows for frequent shifts between codes. An instruction sequence on 'francophonie' might, for instance, include parts devoted to vocabulary learning or to text reading related to a specific French-speaking area. Such a sequence would be coded 'culture and society', but interrupted by shorter sequences of 'vocabulary' and 'reading'. We prioritized the part of the subject content that was dominant in the sequence in question. For example, teacher F gave an introduction to Edith Piaf and then played one of her songs without providing any accompanying listening comprehension activity. Both sequences were thus coded as 'culture & society'.
During tutoring sequences, we used the subject content code that described what occurred in the class as a whole. If the pupils were writing texts, the sequence was coded as 'written activity', even if the teacher occasionally explained grammar points or discussed vocabulary while moving among the students. Occasionally, there was station teaching, in which case we coded the sequence according to where the teacher focused his or her attention.
When the final coding scheme was defined and the principles of coding outlined above were established, both authors independently coded one lesson per school. For this analysis, we selected from each school the lesson with the highest percentage of TL use. We used the kappa statistic in Interact to calculate inter-rater reliability. Since time-stamping unavoidably entails differences between coders regarding the exact moment of clicking the button, we allowed for a time tolerance of 3 seconds. We obtained a kappa of 0.86 for subject content codes and of 0.72 for classroom management codes. Inter-rater reliability was thus considered to be high, and the remaining videos were coded by the second author alone. The two Interact files (one with time-stamping of language use and one of activities) were then merged and subsequently exported to an Excel file to facilitate sorting and calculation of the data.

The amount of L1 and TL in lower secondary French L3 classrooms
The quantitative analysis of the video-recorded material showed that most of the speaking that occurred in class (32 h) was in the L1. In total (grades 9 and 10 together), more than 25 h (1514 min) were coded as L1, whereas only 6 h (357 min) were coded as TL. Approximately 50 min were coded as L1eTL. Figs. 1 and 2 show the distribution of languages used in relation to speaking time and lesson time, respectively. Fig. 2 shows that, for teachers and pupils together, only 16% of all teaching time was used for French speaking. However, there is considerable variation among the schools (see Fig. 3). School A has a different language profile than the others in that L1 and TL were used in approximately equal amounts. This school also had the second highest percentage for the code L1eTL. At the other end of the scale, school C demonstrated very limited use of the TL: less than 10% of the speaking time. The other schools all used the TL for about 20% of the speaking time. Fig. 4 shows the range of TL use for each classroom. School C has a low and stable use of the TL, while for other classrooms there are notable differences between lessons. Such a difference is especially salient for schools D and F, which exhibit a TL use below 20%, except for one lesson in each school with a notably higher percentage. In both cases, these lessons were largely dedicated to speaking exercises. TL use thus seemed to be dependent on the type of task introduced in class. The language of instruction in these two schools is Norwegian, but the classes occasionally perform speaking exercises in French, which brings the overall TL percentage up to 20%.
Our findings show considerably lower amounts of TL use than most previous studies (e.g. Bateman, 2008;Macaro, 2001;Riordan, 2018;Stoltz, 2011). 4 The discrepancy between the present study's findings and those in Stoltz' (2011) study might seem particularly surprising since both addressed French L3 teaching in a Nordic context. However, in contrast to the participants in Stoltz' study, our teachers were randomly selected and did not know about the purpose of the study. There is thus no reason to believe that they were particularly confident in their professional practice or that they would maximize their TL use to make a good example. Moreover, our study addressed the beginner level, whereas Stoltz's (2011) study included pupils who had learned French for 2e3 years longer than the pupils in our study.
The differences with other studies could be related to the setting in which this study was conducted. Unlike much previous research, the current study was set in a context where the use of L1 has remained the practice (Hall & Cook, 2012). The situation is quite different for SL teaching, where maximizing TL use is common practice (Brevik & Rindal, 2020). This difference in teaching cultures within the same region is probably related to the different statuses of the languages (Kim & Elder,  2008). Pupils are generally motivated to learn English, and they are used to hearing it, whereas they find it more difficult to see the immediate use of the L3s (Bardel et al., 2019), which appear as truly foreign, and not familiar, languages. Kim and Elder (2008) also highlighted the factor of distance between the TL and the L1. Although French might be typologically somewhat more distant from Norwegian than English is, the distance in terms of out-of-school exposure is probably more important.

TL use and grade levels
The use of the TL did not seem to increase from grade 9 to 10. Only in school F was such an increase observed, but it appeared to have more to do with the class activities than with grade level (see below). In schools D and E, the TL use dropped from grade 9 to 10. Moreover, TL use did not seem to correlate with the teachers' proficiency level. Although we did not measure the teachers TL proficiency level in this study, we have an indication of it based on the amount of time each teacher has spent in a French-speaking area (see Table 1) and on their utterances in French in class. Teacher C, who had the lowest amount of TL use, has lived in a French-speaking country for many years and seemed to be a highly proficient French speaker. Teacher F is a native speaker but did not use the TL much. Meanwhile, teacher A, who used equal amounts of L1 and TL, also seemed to be a highly proficient French speaker.
Learners' proficiency level has been identified in previous research as one of the most influential factors in teachers' choice of language use (Hall & Cook, 2012), yet several studies suggest that this factor is less important than previously assumed (Askland, 2018;Heimark, 2013;Meiring & Norman, 2002). Although we cannot claim that the pupils necessarily had a higher proficiency level in grade 10 than in grade 9, our data do not support the idea of TL proficiency as a highly influential factor, as there was no increase in TL use from grade 9 to 10. The findings suggest that our teachers' TL use and their facilitation of pupils' TL use is rather static and does not vary according to learners' needs or competence levels. The stable TL use might signal a lack of critical reflection (cf. Ellis & Shintani, 2014) on the part of the teachers concerning their own TL use. However, in order to answer the question more fully, we would have needed to study the same teachers' instruction at more advanced levels since the development in TL proficiency from grade 9 to 10 might not be significant enough to make a difference. Fig. 5 gives an overview of the subject content activities that occurred in each school when French was used, while Fig. 6 gives a similar overview for activities that occurred when Norwegian was used. 5 Speaking exercises account for a large part of the French use in these schools. Overall, approximately 1.5 h of the time coded for TL is speaking exercises, followed by vocabulary work (approximately 1 h). In school E, reading aloud accounted for most of the French use, but this activity was not prominent in the other schools. Other activities were rarely conducted in French, except for in school A.

Situations in which TL and L1 are used
When Norwegian was used, much time was spent on written activities, grammar and vocabulary work. Overall, subject content matter was taught in the L1, except in school A, where both languages were widely used.
Turning to the category Classroom management (Fig. 7), we see that tutoring accounts for most of the TL use. The teacher in school A also uses French for other classroom management purposes, such as instructions, transitions, group organization and behavioural management. The other teachers rarely use French for these purposes. However, in all classrooms, pupils and teachers greet each other in French.
The corresponding figure for L1 use (Fig. 8) shows that all these classroom management issues, apart from greetings, occur in Norwegian more frequently than in French, except in school A.
To sum up, in our data TL is used for greetings, speaking exercises, vocabulary work, and in individual or group tutoring. The fact that much of the French use occurs in tutoring situations corresponds with Stoltz's (2011) finding that TL use is higher in group work than in plenary sessions. Our findings also support Stoltz's (2011) conclusion that activity type heavily  influences the choice of language. Some activities, like speaking exercises and reading aloud, inevitably trigger the use of the TL, whereas others do not. Previous research has identified grammar and culture as two areas that are often taught in the L1 (Aberdeen, 2018;Askland, 2018;Bateman, 2008;Crawford, 2004;Heimark, 2013;Littlewood & Yu, 2011;Macaro, 2001). In our data as well, these areas are taught almost exclusively in Norwegian, but the same applies to all other activities that do not automatically imply TL use.
School A is a clear exception to the pattern described above. The teacher speaks primarily in French, and both L1 and TL are used for all types of activities. This classroom is distinctly bilingual: the teacher and pupils choose the language that fits the situation best. Askland (2018) and Heimark (2013) both found the same pattern in their studies of Spanish (n ¼ 10) and French (n ¼ 6) teachers, respectively. L1 was the language of instruction, except for in one classroom in each study. Taken together, these small-scale studies indicate that there is a teaching culture shared by most L3 teachers in Norway (Llovet Vil a, 2018), which implies restricted TL use. However, in a notable minority of classrooms, the teacher implements a truly communicative approach.

Discussion
The findings show that the ideals of CLT, which include extensive use of the TL, have generally not been implemented in Norwegian French-as-a-foreign-language classrooms. Although L1 use is not prohibited in the context of the present study, TL use is surprisingly low considering the fact that CLT is the framework in force.
A few precautions are needed, however. First, this study examined only six classrooms, and the findings are thus not generalizable. However, they do concur with previous research from the same context (Askland, 2018;Heimark, 2013). Second, audio files and video clips were not counted as language use in this study. Thus, the findings do not tell the whole truth about TL exposure. Nevertheless, the activity coding shows that listening comprehension and culture (the two categories in which audio and video clips would be included) account for a small proportion of the teaching time. TL practice and exposure thus seem to be limited, and pupils have few opportunities to develop their TL proficiency. Consequently, an important conclusion from this study is that it is paramount to find strategies to enhance TL use in FL teaching. We return to this issue in the section on implications for practice.
Another significant contribution of this study, is the methodological framework that it describes. In the present study, we measured TL use in duration instead of the more widely used method of word count (Shin et al., 2019). Word count does not necessarily give justice to the effort behind TL use, as people speak faster and more in their L1 than in an L3. Moreover, the methodology of measuring duration can easily be applied across contexts, while the use of word count is problematic when the L1 and the TL are typologically very different (Rolin-Ianziti & Brownlie, 2002).
Another methodological choice concerns the reference point in relation to which TL and L1 use is measured: the entire lesson or speaking time only. With the time-stamping method that we used, percentages can be easily calculated for both. Time-stamping was also applied to classroom activities, which facilitated the analysis of how and when different languages were used. Previous research into this issue was often based on small-scale interactional analyses of excerpts from audio recordings, while the language and activity coding of entire lessons applied in this study provided a broad picture of connections between language use and classroom activities.
A third case in point concerns the speakers in focus. Most previous studies focused on teacher TL use, although from a modern CLT approach pupils' output is also considered essential for their learning (Swain, 1985). The choice of speaker(s) in focus can impact the findings considerably. If we had studied teacher discourse exclusively, teacher A would have obtained an even higher score on TL use, as the pupils account for most of the L1 use in her classroom. Teachers D and F would have obtained a lower percentage, since most of the L3 use in their classrooms stemmed from pupils' speaking exercises in which the teacher did not actively participate. In other words, focusing on the teacher alone would enhance the TL percentage in some classrooms and reduce it in others.

Implications for educational practice
This study demonstrates a need for change in the prevailing teaching culture. There is a need to move away from extensive use of L1 as the instructional language. Instead of arguing for 'judicious use of the L1' (Hall & Cook, 2012) d a concept that implies 'moderate use' and has as a premise that the TL is the main language of instruction d we call for judicious alternation between L1 and TL. The question that we examine in this section is how such judicious alternation can best be achieved through joint efforts by school owners and leaders, teacher education, and teachers.

Implications for school owners and leaders
School leaders and owners can facilitate a change in teaching culture by providing language teachers with the necessary time and opportunities for professional development. This study did not investigate teachers' rationale for using or not using the TL, but one factor frequently mentioned in previous studies is teachers' TL proficiency. Although the present study and others showed that high teacher TL proficiency does not necessarily result in high amounts of TL use, it is a necessary condition for using the TL in class. TL use in the classroom places high demands on the teacher (Fern andez & Andersen, 2019), and teachers who are not proficient in the TL will not be able to use it as the language of instruction (Liu et al., 2004;Riordan, 2018). As we have argued elsewhere (Vold, 2017), language teachers therefore need to be given the opportunity to maintain and develop their oral skills d a desire also expressed by low-proficiency teachers (Askland, 2018). This could be accomplished, for example, through regular language courses abroad or online conversation courses. In any case, teachers need professional and financial support from their leaders.
Another factor that influences TL use is the time available for lesson preparation (Bateman, 2008;Heimark, 2013). In FL lessons, teachers not only prepare content but also the means by which the content is mediated. They therefore need more time than teachers of other subjects. However, this particular need is rarely taken into account by school leaders.

Implications for pre-and in-service teacher education
It is essential that language teacher education programmes provide opportunities for students' TL skills development. In addition, a change in attitudes and beliefs is needed. Teachers need to experience that TL use can be beneficial for learning at beginner levels and that it does not necessarily inhibit motivation. In order to create such experiences, pre-and in-service teacher training needs to include examples and models of good practice with which teachers can identify. A fruitful path to pursue in in-service teacher education could be to let teachers who implement a truly communicative approach, such as teacher A in the present study, serve as peer mentors in collective professional development projects. In pre-service teacher education on campus, video recordings could be used. The models should not represent out-of-reach TL-only ideals but illustrate strategies and techniques used by successful teachers.
Crichton (2009) claimed that, in language teacher education, students are often encouraged to use the TL but are not trained in how to do so. The use of models (whether live or through video recordings) provides students with the possibility to observe in practice the use of strategies that are otherwise just briefly mentioned. These include repetitions and reformulations of TL discourse, adjustment of speech rate, and the use of body language and gestures as well as visual aids, such as pictures, graphs, or concrete objects. Another strategy is to regularly ask pupils to summarize in L1 in order to track their understanding and force them to concentrate. Yet another strategy is the establishment of a joint agreement in the FL class, implying that pupils are aware that they are not expected to understand all that is being said, and that being exposed to unfamiliar words and structures is a necessary condition for learning.
Another way to train teacher students in using the TL is to provide them with lists of classroom-relevant, useful phrases (Willis, 1990). Offering such lists helps student teachers acquire the necessary domain-specific language of their profession. Riordan (2018) even suggested establishing courses in Language for Teaching Purposes as a sub-discipline within the area of Language for Specific Purposes (LSP) to give non-native speaker language teachers the possibility to acquire classroomspecific language. Such a line of thought acknowledges the fact that the ability to use classroom-specific TL does not automatically follow from a general or academic mastery of the TL.
The issue of TL use and how to succeed with it clearly merits more attention in language teacher education. Macaro & Mutton (2004) also emphasized the importance of support during the first years in the teaching profession, observing that for novice teachers, the need to maintain discipline in class tended to outweigh subject-specific considerations. Novice teachers need guidance to set realistic subject-specific goals, which they can gradually move towards through practice and observation of good models.

Implications for teachers
Ultimately, it is the teacher who is responsible for how the TL is used in class. Teachers should familiarize themselves with strategies to enhance TL use, such as those mentioned above, and try them out in their own classrooms. In addition, it is paramount for professional development that teachers reflect on their practices and develop a critical perspective on their own language use (Edstrom, 2009;Ellis & Shintani, 2014). Edstrom (2009) suggested using recordings, journals and learner feedback as a basis for systematic reflection and language use analysis. Only through gaining insight into their own language use can teachers set attainable goals for themselves and their pupils.
For several of the teachers in the present study, one example of such a goal could be to actively participate in the oral exercises offered to pupils. Instead of walking around and just listening, the teacher could be an interlocutor, which would give him or her just as much, if not more, information about the pupils' TL proficiency. By taking such a small step, the teacher would function as a more competent other, serve as a model for the pupils, and expose them to more input. The teachers could also make it a habit to extend the greetings sessions that open every lesson by adding some small-talk. If provided with some useful phrases (Willis, 1990), this should be feasible also for teachers with low TL proficiency.
Another strategy to enhance TL exposure is the use of audio files or videos. Even for teachers with high TL proficiency, it can be challenging to maintain communication in the TL throughout a lesson (Edstrom, 2009). Extensive use of authentic oral material could give the teacher welcome breaks and reduce teacher fatigue while pupils still receive TL input.
How much TL use should there be in a classroom in a context such as the one examined in this study? Like Cook (2001), we will not specify a particular amount. We believe that every individual teacher should use reflective practice to determine stepby-step sustainable amounts for himself or herself and his or her class and to choose appropriate strategies to attain that goal. What constitutes judicious alternation between L1 and TL will vary between classrooms. For a teacher who is accustomed to teaching in L1, conducting one-fourth of a lesson in the TL might represent an ambitious first step.

Concluding remarks
This article provided an overview of the main findings concerning TL use and activity patterns in six French L3 classrooms in Norway. Stressing the importance of explicitness and detail in the methodological descriptions of measurements of TL use, we adopted a methodological framework based on Macaro (2001) and Brevik and Rindal (2020) to accurately measure time use. We found that, in five out of six classrooms, the language of instruction was Norwegian. French was used only for speaking exercises, greetings and some vocabulary work. Although we looked at differences in the amount of TL use and activity patterns between the six schools, it was beyond our scope to zoom in on individual classrooms to give more detailed accounts of what occurs within them. This is an interesting path for future research. The content taught in each of these classrooms varied considerably, although all teachers were guided by the same curriculum. We also did not look into sequences where teachers or pupils code-switched. In future research, qualitative interactional analyses of selected sequences could be performed to investigate the interactional patterns that occur. Moreover, pupils' opportunities to use the TL in the classroom merit more attention. A few studies have addressed this issue (Crichton, 2009;Stoltz, 2011), but it is much less studied than teacher TL use.
Finally, we emphasize the importance of future research focusing on a broad variety of educational contexts to avoid simplified generalizations and biased conclusions. So far, previous research into language use in SL or FL classrooms has mostly been set in genuine CLT contexts, and recommendations for practice based on such contexts can represent out-ofreach ideals for teachers accustomed to using L1 as the main language medium. In many contexts, FLs are still taught in the L1. The language teaching contexts in the greatest need to move forward might thus not have received sufficient attention.

Funding
This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. In this example, the teacher (T) speaks primarily in French (FR), and the pupil (P) answers primarily in Norwegian (NO). The first utterance from the teacher is coded as TL use, with preceding utterances in French. The L1-TL code starts with the pupil's first utterance. An English translation of all utterances follows in brackets.

Declaration of competing interest
T