Studies in Second Language Learning and Teaching Are foreign language learners’ enjoyment and anxiety specific to the teacher? An investigation into the dynamics of learners’ classroom emotions

Previous research has considered fluctuations in students’ foreign language enjoyment (FLE) and foreign language classroom anxiety (FLCA) over months or years (Dewaele & MacIntyre, 2014, 2016). However, there has been no investigation of the effect of the teacher on these emotions at a single point in time. In this study, we investigate the question whether FL learners experience similar levels of FLE and FLCA in the same language if they have two different teachers. Participants were 40 London-based secondary school students studying modern languages with one Main Teacher and one Second Teacher. Statistical analysis revealed that while FLCA was constant with both teachers, FLE was significantly higher with the Main Teacher. Predictors of FLE such as attitudes towards the teacher, the teacher’s frequency of use of the target language in class and unpredictability were also significantly more positive for the Main Teacher. Item-level analysis revealed that the teacher creat-ing a positive emotional atmosphere in class contributed to the higher FLE score. Items that reflected more stable personal and group characteristics varied less between the two teachers. The findings suggest that FLE is more teacher-de-pendent than FLCA, which is more stable across teachers.


Introduction
In their introductions to special issues on emotions in second language acquisition (SLA), Li (2018, 2020) celebrated the advent of a more holistic approach of learner emotions, triggered by positive psychology, the empirical study of how people thrive and flourish (MacIntyre, Gregersen, & Mercer, 2016. Positive psychology rejects the palliative approach in general psychology, with its focus on abnormalities, disorders, mental illness, and the development of ways to reduce pain and learn to cope with negative experiences, in favor of the development of tools to build positive emotions, foster greater engagement, and boost the appreciation of meaning in life and its activities (MacIntyre & Gregersen, 2012;MacIntyre, Gregersen, & Mercer, 2016;MacIntyre & Mercer, 2014). MacIntyre et al. (2019) point out that the emergence of positive psychology in applied linguistics fits in with "the zeitgeist of general education which now sees the promotion of well-being as both desirable and necessary" (p. 9).
The emergence of positive psychology in SLA has caused a move away from an exclusive focus on negative emotions, such as foreign language classroom anxiety (FLCA), in favor of a broader range of learner emotions, including positive emotions, such as foreign language enjoyment (FLE) (Dewaele, Chen, Padilla, & Lake, 2019). The first study to juxtapose FLCA and FLE in the same research design was the study by Dewaele and MacIntyre (2014). It was based on quantitative and qualitative data from 1746 FL learners of all ages and from all over the world. It showed that levels of FLE and FLCA that participants reported experiencing in their foreign language (FL) classes were linked to a range of learner-internal and learner-external variables. Qualitative data indicated that while FLE and FLCA were relatively stable over time, they could occasionally peak, or drop, depending on the task at hand or comments from the teacher and peers. Speaking in front of peers was frequently mentioned as both highly anxiety-provoking and enjoyable. Further research focused on the exact effect that learner-internal and learner-external variables had on FLE and FLCA (Dewaele & MacIntyre, 2019;Dewaele, Witney, Saito, & Dewaele, 2018). FLE in particular seems to be linked more strongly to the teacher, while FLCA seems to be less context-dependent. The research design of the current study, which draws on a subsample of the participants in , will allow us to find out to what extent FLE and FLCA are linked to individual teachers among learners who had two teachers for the same FL. It will also allow us to determine just how teacher-specific FLE and FLCA are in the specific context of two British schools.
2. Literature review 2.1. Foreign language classroom anxiety (FLCA): From 1986 to today Horwitz, Horwitz, and Cope's (1986) study heralded a new era in research on FLCA, the so-called specialized approach (MacIntyre, 2017). Horwitz et al. (1986) used feedback from foreign language (FL) learners about triggers in their FL classes that made them anxious, which included the fear of not doing well on tests, of speaking up in public, and the possible negative reactions by peers and teachers. All these sources of FL anxiety coalesce into a "conceptually distinct variable in FL learning" (p. 125), namely FLCA, which was defined as "a distinct complex of self-perceptions, beliefs, feelings, and behaviors related to classroom learning arising from the uniqueness of the language learning process" (Horwitz et al., 1986, p. 128). Horwitz later insisted that FLCA is "multi-faceted" (2010, p. 145). Indeed, anxious learners have the tendency to become anxious in specific situations: "When individuals experience Language Anxiety, they have the trait of feeling state anxiety when participating in language learning and/or use" (italics in original, Horwitz, 2017, p. 33). Horwitz adds that anxiety may simply emerge from thinking about FL situations. Many applied linguists have explored language anxiety since 1986, seeking to identify its relatively stable sources as well as its relationship to student performance and achievement (see Dewaele, 2017;Horwitz, 2010). Researchers who focused on instructed learning contexts have investigated learners' FLCA within classrooms but there has also been research on anxiety experienced outside the classroom where the social stakes are higher (Dewaele, Petrides, & Furnham, 2008;Ross & Rivers, 2018). Sevinç and Dewaele (2018), for example, found that Turkish immigrants in the Netherlands suffer from heritage language anxiety in interactions with members of the Turkish community and from majority language anxiety in interactions with Dutch speakers, fearing in both cases that disfluencies or errors would mark them as outsiders.
Recently researchers have become increasingly interested in the dynamic nature of anxiety over different time frames, ranging from months or years, to seconds in the idiodynamic approach (Boudreau, MacIntyre, & Dewaele, 2018;Gregersen, MacIntyre, & Meza, 2014). Gregersen et al. (2014) investigated variation in heart rate and ratings of state language anxiety over a three-minute period using an idiodynamic approach. Participants were three high-and three low-anxiety American students who were enrolled in a Spanish class and had to give a short oral presentation in front of their peers. An analysis of the values showed that all participants spent some time in the high anxiety zone and that most also had moments in the low anxiety zone. High anxiety participants reported in subsequent interviews that word searches caused spikes in anxiety while low anxiety participants used strategies to mitigate word retrieval problems. Even during tasks that the learner appreciates, anxiety can be triggered by some innocuous obstacle, which sets off a chain reaction, the so-called self-exacerbating syndrome (Jussim & Eccles, 1995). The learner may feel a sudden pang of anxiety and begin to worry that peers or teacher will notice it, leading to even more anxiety. In other words, learners may be more or less prone to experiencing anxiety, but all will experience it momentarily and it will dissipate at different rates. Dewaele (2002) argued that FLCA is not a stable psychological dimension, as it varies across the FLs at the same point in time. His 100 Flemish students were found to report significantly more FLCA in their second language (L2) French than in their third language (L3) English, and that FLCA in French (but not in English) was linked to students' social class and the political tensions between language communities in Belgium. An individual's anxiety is thus linked to triggers that arise in the micro-context (i.e., a specific sentence or task), in the meso-context (i.e., anxiety about reactions of teachers or peers in the classroom or in some group) and in the macro-context (i.e., political and historical context). One could therefore argue that anxiety is at the confluence of both learner-internal variables and learner-external variables (Dewaele, 2017). While the number of learner-internal variables having an effect on anxiety could be described as being more or less finite, the number of external variables is much larger and possible interactions between micro-, meso-and macro-levels push potential causes towards infinity. Of course, by controlling a number of crucial learner-external variables, such as age and educational context, it might be easier to judge the effect of a number of external variables such as the teacher, and teacher practices, combined with learnerinternal variables (cf. Dewaele, Magdalena Franco & Saito, 2019;Saito, Dewaele, Abe & In'nami, 2018).

Positive psychology and the broadening of the emotional spectrum in FL research
MacIntyre and Gregersen (2012) pointed to the crucial role of all emotions in the FL classroom, not just the negative ones. Referring to positive psychology, they argued that effects of positive emotions go beyond pleasant feelings since they enhance learners' ability to notice things in a classroom environment and strengthen their awareness of language input. Students in a positive emotional state are better able to absorb the FL. Positive emotions also help counteract the damage caused by negative emotions. MacIntyre and Gregersen (2012) explained that students who are consumed by negative emotions are more likely to have a narrow focus, which restricts the intake of language input, an idea at the basis of Krashen's (1981) affective filter hypothesis. Positive emotions contribute to students' longer-term resiliency and hardiness. A final benefit of positive emotion in the classroom, according to MacIntyre and Gregersen (2012), is that it allows learners to feel safe as they explore the FL and play with it, boosting social cohesion with their peers as they do so.
Dewaele and MacIntyre (2014) developed a study design that allowed them to consider variation in FLCA and FLE simultaneously. They devised a FLE scale, developing and adapting the interest/enjoyment subscale from Ryan, Connell, and Plant (1990) to a FL context. The new FLE scale consisted of 21 items with Likert-scale ratings about the emotions towards the FL learning, the peers and the teacher. It was complemented with a FLCA scale of 8 items extracted from Horwitz et al. (1986). FLE was defined as "a complex emotion, capturing interacting dimensions of challenge and perceived ability" (Dewaele & Mac-Intyre, 2016, p. 216). A total of 1746 FL learners from around the world participated. FLE was found to be moderately negatively correlated with FLCA (r = -.36). Statistical analyses allowed the identification of learner-internal and externalvariables that had a significant relationship with levels of FLE and FLCA: Participants who already mastered several languages, who had reached a higher level in the FL, who felt more proficient than their peers, who had reached a higher level of education and who were older reported significantly more FLE and significantly lower levels of FLCA. In addition, those studying more FLs also scored significantly higher on FLE, where FLCA was not associated with studying more FLs. (Dewaele & MacIntyre, 2016. p. 262) An analysis of qualitative feedback from 1076 participants about enjoyable episodes in the FL class showed that certain classroom activities boosted learners' levels of FLE. These were typically unusual activities that gave students a degree of autonomy. Additionally, teachers were typically central in the stories because their attitude, humor, organization, respect, and praise for outstanding performance were frequently mentioned. Based on these findings, Dewaele (2015) argued that teachers need to have an emotional thermostat, allowing them to keep their students in an optimally positive emotional mood.

The effect of the teacher on FLE and FLCA
Shifting the focus to the effect of teacher-centered variables,  collected data from 189 secondary school FL learners in London, UK. A moderate negative correlation emerged between FLE and FLCA (r = -.194). Predictors of FLE and FLCA turned out to be very different: Higher levels of FLE were linked to more positive attitudes toward the FL (eta 2 = .29), the FL teacher (eta 2 = .27), frequent use of the FL by the teacher (eta 2 = .12), the amount of time students spent speaking the FL (eta 2 = .08), and teacher unpredictability (eta 2 = .06). Levels of FLCA were linked to being more advanced in the FL (eta 2 = .17), higher relative standing among peers in the FL (eta 2 = .08), and positive attitudes toward the FL (eta 2 = .07). Pursuing this avenue of research,  collected data from 210 former and current Spanish EFL learners. Some teacher characteristics (i.e., teacher's friendliness and foreign accent) predicted close to 20% of variance in FLE but others (i.e., strictness, FL use in class and teacher age) explained only 8% of variance in FLCA.
In a mixed-methods study on the nature of FLE and FLCA and the role of classroom-specific and psychological predictors, Dewaele and MacIntyre (2019) collected feedback from 750 FL learners from around the world. Statistical analyses showed that FLE was mostly predicted by attitudes towards the teacher (24% of variance) as well as the personality trait of cultural empathy (8% of variance) while FLCA was mostly predicted by the personality trait of emotional stability (30% of variance), followed by relative standing (9% of variance) and by social initiative (2.4% of variance). Analysis of participants' own attribution of their FLE and FLCA confirmed the statistical findings as FLE was most often attributed to the teacher while FLCA experiences were most frequently linked to the self. The strong effect of the teacher on FLE also emerged in the Chinese EFL context (e.g., Li et al., 2018;Li et al., 2019;Jiang & Dewaele, 2019;Jin & Zhang, 2018) and in the Kazakh context with Turkish as a target language   Boudreau, MacIntyre, and Dewaele (2018) used the idiodynamic approach to examine the highly unstable relationship between FLE and FLCA in FL speech, on a very short timescale. Ten Anglo-Canadian students completed oral tasks in their French L2 after which they viewed a video recording of their performance and provided per second measurements of their FLE and FLCA. They were then interviewed about the reasons for fluctuations in their FLE and FLCA. Correlation analyses of FLE and FLCA values of the same participant revealed highly dynamic patterns veering from positive to negative and then to zero. In other words, sometimes high FLE corresponded with low FLCA, but a few seconds later the relationship could shift completely and unpredictably. Feedback from participants suggested that a variety of causes were linked to the fluctuations, ranging from difficulties in vocabulary searches, to using strategies to control FLCA, to enjoyment in discussing a particular topic.

Time scales for variation in FLE and FLCA
Given the difficulty of following a cohort of FL learners through their whole secondary education, Dewaele and Dewaele (2017) adopted a pseudo-longitudinal design 1 to look at variation in FLE and FLCA over time. Participants were the same 189 London-based FL learners that were used in . Mean levels of FLE, FLCA and their relationships with independent variables were calculated for three age groups of students within the same two London schools. A comparison of the 12-13 year olds (age group 1), 14-15 year olds (age group 2) and 16-18 year olds (age group 3) showed that FLCA did not change over time, and that FLE was the highest for age group 3, with a dip for age group 2. Multiple regression analyses revealed that different independent variables predicted FLE and FLCA in the three age groups. In age group 1, FLE was linked to relative standing in the group of peers (28% of variance) and FLCA to the amount of FL knowledge already acquired (25% of variance), with the effect sizes being moderate. In age group 2, attitude towards the FL was the strongest predictor of FLE (30% of variance), followed by attitude towards the teacher, teacher predictability (a negative predictor) and the number of languages in the learner's repertoire, explaining all together a total of 44.5% of variance (large effect sizes). Relative standing in the group of peers and FL knowledge predicted less than half the amount of variance in FLCA (17% and 3% respectively, which are medium effect sizes). In age group 3, attitude towards the teacher was the only predictor of FLE (explaining 44.5%, a large effect size), while relative standing in the group (a negative predictor) and teacher predictability were the stronger predictors of FLCA (explaining 21% and 8% of variance respectively, a medium effect size) (Dewaele & Dewaele, 2017, p. 18). The authors thus concluded that despite relatively modest changes in mean levels of FLE and the lack of change in FLCA, "various psychological and sociobiographical variables were in a tug of war over pupils ' emotions" (p. 20). In other words, the causes of FLE and FLCA are highly dynamic and change over time.
Adopting a dynamic systems theory approach, Dewaele and Pavelescu (2019) collected qualitative data from two Romanian FL learners during one school semester to investigate how variation in the learners' FLE and FLCA affected their willingness to communicate (WTC) in English. Material included lesson observations, a written task and semi-structured interviews that probed into the learners' FLE and FLCA when they started learning the language. The learners' levels of FLE and FLCA were found to fluctuate sharply in the short term and more slowly over the longer term, affecting their WTC. The fluctuations in FLE and FLCA could be attributed to the learners' previous experience with the FL, their use of the FL outside of the classroom, their personality, but also to classroom-specific factors, such as the lack of interest in a film or conversation topic, a dislike for a replacement teacher or the seating arrangement in the classroom.
What emerges from the literature review is that a well-established interest in the sources of FLCA has expanded to include positive emotions such as FLE. FLE and FLCA have been shown to be independent dimensions linked to different learnerinternal and learner-external variables. Teachers have a much stronger effect on their students' FLE than on their FLCA. No study so far has, to our knowledge, investigated whether learners who have two different teachers for the same FL might experience similar levels of FLE and FLCA with both. This is the aim of the present study, namely an investigation into the teacher-specificity of FLE and FLCA.

Research questions
1. Are learners' FLE and FLCA similar in the classes of two different teachers teaching the same FL to the same class? 2. Are there differences in the teacher-related predictor variables for FLE identified in  between the two teachers (i.e., attitude toward the teacher, frequency of use of the FL by the teacher, amount of time students spent speaking the FL and teacher unpredictability)? 3. What does an item-level analysis of the FLE scale reveal about reasons behind significant differences?

Participants and demographics
The present sample is a subsample of 40 high school students (17 females, 23 males) extracted from the full sample of 189 participants in . They were selected because they reported having two FL teachers for the same FL, compared to the other 149 who had just one teacher for the FL. Participants' age ranged from 13 to 18 (M = 16.6, SD = 1.3). They came from two schools in Greater London: 17 students were from Dame Alice Owen's, a selective state school in Potters Bar, and 23 students were from Westminster School, a private school in Central London, which is selective and fee-paying. Both schools are amongst the top performing schools in the UK. 2 Most participants were studying French as a FL (N = 24), while the others were studying Spanish (N = 8), German (N = 8), with smaller numbers studying Hindi, Italian, Japanese, Mandarin, Polish, Portuguese, and Russian. A large majority of participants were British (N = 30) and had English as an L1 (N = 38). Other L1s (often acquired simultaneously from birth) included Bulgarian, Cantonese, Dutch, Farsi, French, Hungarian, Punjabi and Russian. The sample consisted of one self-reported bilingual (2.5%), 10 trilinguals (25%), 10 quadrilinguals (25%), 12 pentalinguals (30%), and 5 sextalinguals (12.5%). The two remaining participants reported knowledge of 7 and 8 languages respectively. A fifth of participants (N = 8 or 20%) reported growing up with more than one language from birth.
Students were asked to rate themselves compared to their FL learning peers, ranging from far below average, below average, average, above average and far above average. Only 3 students rated themselves as below average (7.5%), with higher numbers choosing average (N = 10, 25%), above average (N = 20, 50%) and far above average (N = 7, 17.5%). These ratings were non-significantly positively correlated with self-reported results on their last major FL test (r (39) = .31, p = .053). These test scores ranged from 69% to 100%, with the mean of 89.5% (SD = 7.5). In other words, these were very good FL students.

Instrument
The questionnaire started with a demographics section from which the above information was retrieved. Following this, participants were asked whether they had just one or two FL teachers for their first foreign language. These teachers did not have the same status, as the more senior teacher typically had more contact time with the students. Although no reference was made in the questionnaire to the status of the teacher, participants automatically labelled the teacher with whom they spent most time as Main Teacher and the other teacher as Second Teacher. No information was collected about how many people were included in both groups.
One item enquired about the frequency of use of the FL in class by the Main Teacher and the Second Teacher (ranging from "hardly ever" -1, "not very often" -2, "sometimes" -3, "usually" -4, and "all the time" -5). The next four questions inquired about the average proportion of time spent on writing, reading, listening and speaking by the Main Teacher and the Second Teacher: the options ranged from 0-10% (1) to 90-100% of the time (10). The following question asked how predictable the classes were with the Main Teacher and the Second Teacher ("very predictable" -1, "predictable" -2, "variable" -3, "not predictable" -4, "very unpredictable" -5). Attitudes towards the Main Teacher and the Second Teacher were collected using a 5-point Likert scale (ranging from "very unfavorable" -1, "unfavorable" -2, "neutral" -3, "favorable" -4, to "very favorable" -5).
Students then completed 10 items, which were extracted from the FLE questionnaire (see Table 2), and reflected on the original structure of the FLE scale (Dewaele & MacIntyre, 2014), about how much enjoyment they experienced overall in the classes of the Main Teacher and the Second Teacher. They were based on standard 5-point Likert scales with the anchors "absolutely disagree" = 1, "disagree" = 2, "neither agree nor disagree" = 3, "agree" = 4, and "strongly agree" = 5. All the items were positively phrased. The analysis revealed high internal consistency reliability for the Main Teacher (Cronbach's alpha = .91) and the Second Teacher (Cronbach's alpha = .92).
Another 8 items were extracted from the FLCAS and reflected physical symptoms of anxiety, nervousness and lack of confidence (Horwitz et al., 1986) and were used in MacIntyre (1992). They also captured the reliability of the original scale (Dewaele & MacIntyre, 2014). Two FLCA items were phrased to indicate low anxiety and six were phrased to indicate high anxiety. The low anxiety items were reversecoded so that high scores reflected high anxiety for all items on this measure. Scale analysis revealed high internal consistency for the Main Teacher (Cronbach's alpha = .90) and the Second Teacher (Cronbach's alpha = .92) (see the list of items in the Appendix). Botes, Dewaele and Greiff (2020) used exploratory and confirmatory factor analysis of the FLE and the FLCA items and found good fit statistics.
A one-sample Kolmorogov-Smirnov test revealed that the normality of distribution of the items was narrowly rejected for three out of the four dependent variables (FLE Main Teacher: KS = .18, p < .002; FLE Second Teacher: KS = .12, p = .18; FLCA Main Teacher: KS = .14, p < .034; FLCA Second Teacher: KS = .15, p < .026). A look at the distribution shows a skew towards higher values for FLE and lower values for FLCA. We ran both non-parametric and parametric statistics and found very similar results (available from the corresponding author). We opted for the more powerful parametric statistics, taking into account that t-tests tolerate moderate violations to their normality assumption (Rosenkrantz, 2008, p. 478).
The online questionnaire was completely anonymous since no names of participants nor their teachers were collected. Students filled out the questionnaire once and were asked to reflect on their FL classes since the start of the academic year (approximately 6 months). The research design and questionnaire obtained approval from the Ethics Committee of the first author's institution, the headmasters of Westminster School and Dame Alice Owen's School, as well as the FL teachers at both institutions. Consent was obtained in two stages. Parents of about 600 students were informed that their children would be contacted to participate in a survey on affective variables in the foreign language classroom. They were invited to contact the researcher to obtain extra information. Some parents did so and none opted out of the survey. Next, students were invited by their school to participate in the study, and their individual consent was obtained at the start of the survey. The questionnaire was posted online using Googledocs and remained accessible for a month.

Results
A first series of paired t-tests showed that FLE was significantly higher with the Main Teacher than with the Second Teacher (t(39) = 3.01, p < .005), with Cohen's d = .360, which approaches a small effect size (see Plonsky & Oswald, 2014). No differences emerged in FLCA in the classes of the Main Teacher and the Second Teacher (t(39) = .61, p = .55, Cohen's d = .04) (see Figure 1).

Figure 1 Mean scores (and SD) for FLE and FLCA with the Main Teacher and the Second Teacher
A second series of paired t-tests revealed that the Main Teacher scored significantly higher than the Second Teacher on three out of the four predictor variables for FLE identified in  (see Table 1 and Figure 2). The effect size is situated between small and medium.  In order to answer the final research question, paired t-tests were run for the 10 items of the FLE scale in order to identify the ones where the difference between the Main Teacher and the Second Teacher were strongest (see Table  2). Differences were significant for five items (see Figure 3), with small effect sizes (Plonsky & Oswald, 2014).  Five items were responsible for the higher FLE score with the Main Teacher (see Table 2). Not surprisingly, they related to the positive emotional atmosphere that the teacher could create where students were excited, enjoyed the class, had fun and shared humor. Participants felt that their Second Teacher did not create an equally positive emotional atmosphere. The five items that did not show significant differences were more independent of the classroom interactions and reflected more stable characteristics such as pride in accomplishments in the FL, seeing the mastery of a FL as something cool, enjoying group membership and good relations with nice peers.

Discussion
The answer to the first question was unexpected, as levels of FLE were significantly higher with the Main Teacher than with the Second Teacher, while levels of FLCA remained similar with both teachers. This confirms previous research showing that variation in FLE is strongly related to the teacher Jiang & Dewaele, 2020;Li et al., 2018Li et al., , 2019, whereas FLCA is more strongly predicted by learner-internal variables (Dewaele & MacIntyre, 2019). Thus, FLCA is slightly more trait-like whereas FLE is more state-like. This finding has important pedagogical implications because it means that teachers have to work hard to create the optimal emotional climate in their classrooms to allow learners to enjoy the class, and not all teachers manage to do that (Li et al., , 2019. The fact that no difference in learners' FLCA emerged between the two teachers confirms previous findings that teacher characteristics are weak predictors of FLCA Dewaele & MacIntyre, 2019). Considering the finding that Teacher 2 levels of FLE were higher with the Main Teacher than with the Second Teacher, it is hardly surprising that our participants' attitudes towards their Main Teacher were significantly more positive than towards the Second Teacher. Given that the Main Teacher is typically the more experienced teacher, he or she may be better equipped to create enjoyable lessons than his or her less experienced colleagues. The second research question focused on differences in students' ratings of teacher-related predictor variables for FLE for their two teachers (i.e., attitude toward the teacher, frequency of use of the FL by the teacher, and the amount of time students spent speaking the FL and teacher predictability). Significant differences emerged for three out of the four predictor variables for FLE. Students had significantly more positive attitudes towards their Main Teacher. They also reported that the Main Teacher used the FL more frequently in class and was less predictable than the Second Teacher. This finding strengthens the suggestion made previously about the Main Teacher eliciting more FLE among learners because they had more teaching experience. Frequent use of the FL in class suggests a high level of expertise in the FL and confidence in using it in class. Lower predictability suggests that the teacher dares to deviate from the rigid guidelines on what and how to teach, and strives to be more creative (see Dewaele, Gkonou, & Mercer, 2018). It is likely that the self-selected participants in the survey were good FL learners, in two elite schools, who particularly enjoyed the challenge of communicating in the FL and who appreciated not being exposed to the same routines in every class. It can be assumed that these two factors contributed to a favorable attitude towards the teacher, which in turn fed into the learners' FLE.
The third research question focused on the analysis of 10 individual items in the FLE scale in order to gain a better understanding about the specific reasons behind the significant differences in FLE between the Main Teacher and the Second Teacher. The values of half of the items were significantly higher with the Main Teacher than with the Second Teacher (see Table 2). What these items had in common was that they directly reflected how students felt in a specific classroom (e.g., excitement, enjoyment, fun, shared humor and a positive sense of community). This sits well with the observation in Dewaele and MacIntyre (2014) that specific classroom activities, expertly led by the teacher, figured most frequently in participants' descriptions of a highly enjoyable episode in their FL classroom. Similar findings emerged in the Chinese EFL context (e.g., Li et al., 2018Li et al., , 2019. The FLE items that varied less between the Main Teacher and the Second Teacher were less linked to what happened in the classroom as they reflected longer-term, more stable characteristics of learners and the group. Pride, attitudes toward FL learning, toward the group membership and the peers are less likely to vary much with a different teacher. The current study is not without limitations. Participant self-selection means that participants were more likely to be enthusiastic FL students, a common problem in applied linguistics research (Dewaele, 2018). Another limitation is linked to the delicate nature of assessing teacher "quality." Approval of our ethics application, and obtaining the agreement of the two schools, depended on full and total anonymity of participants and their teachers. This meant that we were incapable of collecting the names of the teachers, which would have allowed more precise analyses rather than the rather generic categories of "Main Teacher" and "Second Teacher". We could have established whether the same students experienced more or less FLE and FLCA with the Main Teacher than with the Second Teacher. This limitation might also be a strength, as having had teacher names might have led to too much fragmentation within an already small sample and would have complicated statistical analyses. Qualitative data would also have offered the possibility of getting a better insight into the causes of FLE and FLCA with various teachers. However, this would have led to a violation of the conditions for ethics approval by the research institution and by the schools, as teacher names would have likely cropped up during interviews or open questions.

Conclusion
The originality of the present study lies in the analysis of data on classroom emotions from the same cohort of learners studying the same set of foreign languages with two different teachers. Any variation in FLE and FLCA could hence be attributed to the teacher. The finding that FLCA remained similar with both teachers but that FLE was higher with the Main Teacher confirmed previous work on the nature of these two emotions. FLCA is more strongly linked to learner-internal variables, including personality traits such as emotional stability and social initiative, while FLE is more dependent on learner-external variables such as the teacher and the peers (Dewaele & MacIntyre, 2019). What happens in the classroom is thus more likely to cause ups and downs in FLE. One possible cause is that teachers differ in their ability to regulate the emotional temperature in their classroom and, by extension, the FLE and the longer-term motivation of their students. This ability has been linked not just to experience but to trait emotional intelligence and happiness (Dewaele, 2015;Dewaele, Gkonou & Mercer, 2018;Moskowitz & Dewaele, 2019). The fact that learners reported most FLE with the more experienced Main Teacher should therefore not discourage novice teachers. As Arnold (2020) points out in the preface to The Emotional Rollercoaster of Language Teaching: "Teachers who are highly motivated will work to find the best ways to teach the language and to relate to their students. When teachers show interest in their students, have empathy with them, reduce their anxiety, and contribute to their feeling of confidence, they will have more positive results with their teaching" (p. xx).