The two faces of Janus? Anxiety and enjoyment in the foreign language classroom

The present study investigates Foreign Language Enjoyment (FLE) and Foreign Language Classroom Anxiety (FLCA) in the classroom. Participants were 1746 current FL learners from around the world. We used a measure of FLE, based on Likert scale ratings of 21 items (Dewaele & MacIntyre, 2014), and a measure of FLCA based on 8 items extracted from the FLCAS (Horwitz, Horwitz, & Cope, 1986). Statistical analyses revealed that levels of FLE were significantly higher than those of FLCA. FLE and FLCA were linked to a number of independent variables: participants’ perception of their relative level of proficiency within the FL classroom, number of languages known, education level, number of FLs under study, age group and general level of the FL (ranging from lower-intermediate to advanced). Female participants reported both more FLE and more FLCA. Cultural background of participants also had a significant effect on their scores. Participants’ views on episodes of enjoyment in the FL class revealed the importance of teachers’ professional and emotional skills and of a supportive peer group. Many participants mentioned the moment at which they realised that their long effort in mastering an aspect of the FL paid off.


Introduction
In Roman mythology, Janus is the god of beginnings and transitions. In ancient times, foreign language learners might have found great comfort in Janus, as the learning of a foreign language (FL) often feels likes an endless beginning followed by a very long succession of transitions. Janus is traditionally depicted with two faces, one looking to the future, the other to the past. In the present research we are concerned with both directions of the vision, as they reflect emotions in FL learning. For learners, having an anxious negative face is a common experience, but so is the face of positive enjoyment.
It is fair to say that SLA researchers have mostly focused on the negative face of emotion, with anxiety being the most studied topic (Dewaele, 2007;Horwitz, 2001;Lu & Liu, 2011;MacIntyre, 1999;MacIntyre & Gardner, 1991c;Saito, Garza, & Horwitz, 1999;Scovel, 1978; and for a review, see Horwitz's (2010) research timeline). Foreign Language Anxiety (FLA) can be defined as "the worry and negative emotional reaction aroused when learning or using a second language" (MacIntyre, 1999, p. 27).
FLA has been linked to a number of learner variables. Lower levels of FLA have been linked to a younger age of onset of acquisition, authentic use of the FL during the learning phase, frequent current use of the FL, a high degree of FL socialisation, and a large network of people with whom the FL is used (Dewaele, 2013b). Dewaele (2013b) found minimal differences in FLA linked to gender. However, older participants and more highly educated participants scored significantly lower on FLA in their various languages. The number of languages previously learned tends to be associated with lower FLA in native and FLs (Dewaele 2010(Dewaele , 2013bDewaele, Petrides, & Furnham, 2008;Thompson & Lee, 2012). Levels of FLA have also been found to increase significantly, and linearly, from the L1 to the L5 of pentalinguals (Dewaele, 2013b). In addition to the number of languages known, the typological distance between the target language and existing languages in the learner's repertoire influences FLA. If the target language belongs to a familiar linguistic family, anxiety tends to be significantly lower (Dewaele, 2010).
The effects of FLA can be quite insidious. Indeed, they can compromise the learning potential of the FL learner, wreck the best teaching techniques and render the most attractive material inadequate (Arnold & Brown, 1999, p. 2). Not surprisingly, FLA has been described as one of the strongest predictors of success or failures in FL learning (MacIntyre, 1999). FLA can disrupt the optimal functioning of the prefrontal lobe of the brain, which affects memory and thus reduces learning capacity (Arnold & Brown, 1999, p. 2). MacIntyre and Gardner (1994aGardner ( , 1994b demonstrated that anxiety can disrupt language ac-quisition and production at the input, processing and output stages of the learning process. Studies have consistently shown that FLA negatively affects learning and communication, and those effects are both complex and multidimensional (Dewaele, 2002;Horwitz, 2010). Taking a micro-perspective, MacIntrye and Serroul (2015) used a dynamic approach to link anxiety experiences to difficulties in vocabulary retrieval in real time during communication, resulting in spikes and dips in perceived competence and willingness to communicate. Using a similar methodology, Gregersen, MacIntyre and Meza (2014) demonstrate links between perceived anxiety scores and fluctuations in heart rate in real time, showing the physiological responses that partially underlie the anxious state. Although there has been some suggestion that a facilitating form of FLA is possible (Hewitt & Stephenson, 2012;Kleinmann, 1977), such effects have been inconsistent and difficult to find in the research literature (Phillips, 1992) leading Horwitz (1990) to question whether the facilitating effects of FLA even exist.
There is a long list of potential sources of FLA, including harsh error correction (Gregersen, 2003;Young, 1991), self-presentation concerns (Cohen & Norst, 1989), competitiveness among learners (Bailey, 1983), incompatibility between teacher and student , personality traits such as neuroticism (Dewaele, 2002(Dewaele, , 2013a, perfectionism (Dewaele, Finney, Kubota & Almutlaq, 2014;Gregersen & Horwitz, 2002), tolerance of ambiguity in the second language (Dewaele & Shan Ip, 2013), and many other factors (Horwitz, 2010). Indeed, language learning is a process that is particularly prone to anxiety-arousal, in part because error correction is both an indispensible part of learning and a key source of anxiety (Gregersen, 2003). It is therefore ironic that anxiety arousal tends to increase the number of errors a person is likely to make (MacIntyre & Gardner, 1994a). The vicious cycle of anxiety-arousal will be familiar to teachers and learners alike. Over repeated occurrences, the expectation of anxious feelings becomes firmly established in the learners' minds; it is the expectation of similar future anxious episodes that are at the core of the worry component of FLA (MacIntyre & Gardner, 1989).
FLA most often has been defined and measured as a stable characteristic that a learner brings from one language learning situation to another (Horwitz, 2010). Measures of anxiety that are focussed on FL situations produced much more consistent and meaningful results than concepts such as manifest anxiety, test anxiety or trait anxiety (Horwitz, 2010;MacIntyre & Gardner, 1989, 1991c. The first measures of language-specific anxiety were contained in Gardner's (1985Gardner's ( , 2010 Attitude-Motivation Test Battery (AMTB) that included measures of French classroom and French use anxiety. Although the data collected by Gardner and associates often showed strong correlations between language anxiety and second language proficiency, the concept of language anxiety was not as prominent in the socio-educational model as was the role of motivation, so the anxiety-related contributions have been underestimated over the years (MacIntyre & Gardner, 1991c;MacIntyre 2010). Other measures of language anxiety are available that focus on the stages of the learning process (MacIntyre & Gardner, 1994b) and on specific skill areas, such as listening (Elkhafaifi, 2005;Rivers, 1981;Scarcella & Oxford, 1992), reading (Saito et al., 1999) and writing (Cheng, 2002;Cheng, Horwitz, & Shallert, 1999). The most frequently used measure in research on FLA is the well-established Foreign Language Classroom Anxiety Scale (Horwitz, Horwitz, & Cope, 1986). The scale includes 33 items such as "I get nervous and confused when I am speaking in my FL class" and tends to produce highly reliable scores that correlate well with other FLA measures (MacIntyre & Gardner, 1991b;Pae, 2013).
In approaching the present study, we wondered whether a focus on anxiety's negative effects is dealing with only half of the issue. Given the consistency of the negative correlations between anxiety and a variety of language-related factors, recommendations for dealing with language anxiety typically focus on reducing the negativity of the experience, dealing with the unpleasant feelings, and ameliorating its disruptive effects (Gregersen & Mac-Intyre, 2014;Young, 1991). The literature overwhelmingly has concentrated on negative emotion, with positive FL emotion not being as widely studied (Bown & White, 2010;Imai, 2010). Arnold and Brown (1999) already pointed out that "much more attention is given to the question of negative emotions . . . (one) should not lose sight of the importance of developing the positive" (p. 2).
The existing research that has considered emotions has often been embedded in a broader study of motivation. For example, Gardner (1985Gardner ( , 2010 proposed a collection of affective variables that define the motivated learner, including positive attitudes toward speakers of the target language and emotions such as interest in FLs and desire to learn the FL. A related approach, Clément's (1980Clément's ( , 1986 socio-contextual model, proposed two key motivational processes, the second of which (self-confidence) reflects positive perceptions of competence and low levels of anxiety. Both of these models emphasize positive affect, but with a focus on motivation. Nevertheless, each model proposes a prominent role for positive emotions.
There is good reason to believe that studying positive emotion in greater detail will produce a novel understanding of the processes involved, and interest in positive emotions has been increasing in recent years. The impetus for more detailed consideration of the role of positive emotion in second and foreign language acquisition has been supported by developments in positive psychology (see Lake, 2013;MacIntyre & Mercer, this volume;Peterson, 2006;Seligman & Csikszentmihalyi, 2000). Similar to the SLA research on emotion, much of psychology has been focused on unpleasant, negative experiences: clinical depression, phobias, traumatic stress, prejudice, discrimination, personality disorders, and the like (Seligman & Csikszentmihalyi, 2000). But times are changing in psychology, in particular when it comes to understanding the nature and function of positive emotions.
Leading the theoretical development in the area of positive emotion is Barbara Fredrickson, whose influential work on the broaden-and-build theory has argued for a clear differentiation between positive and negative emotions (Fredrickson, 2001(Fredrickson, , 2003(Fredrickson, , 2006. According to Fredrickson, specific negative emotions each tend to be associated with a specific action tendency, a compulsion toward a specific type of behavior (see Reeve, 2005). For example, anger leads to the urge to destroy obstacles in one's path, fear leads to protective behaviours, and disgust leads to rejection as in quickly spitting out spoiled food. Fredrickson's research proposes that positive emotions produce a different type of response.
[The broaden and build] theory states that certain discrete positive emotionsincluding joy, interest, contentment, pride, and love-although phenomenologically distinct, all share the ability to broaden people's momentary thought-action repertoires and build their enduring personal resources, ranging from physical and intellectual resources to social and psychological resources. (Fredrickson, 2003, p. 219) Positive emotion can help dissipate the lingering effects of negative emotional arousal, helping to promote personal resiliency in the face of difficulties. Positive emotions also facilitate exploration and play, leading to the opportunity to have new experiences and learn in an efficient way; this is the "broaden" side of Fredrickson's theory. The social dimension of positive emotions is closely connected with the "build" side of the theory. Because people tend to be attracted to others with positive emotions, and positivity engenders both goodwill and social bonds, positive emotions help a person build resources that collectively might be considered social capital (Adler & Kwon, 2002). Given the dependence that learners have on speakers of the language (parents, teachers, native speakers, advanced learners and near peers), the presence of other people offers numerous resources that facilitate learning (Gardner, 1985;MacIntyre, Baker, Clément, & Conrod, 2001).
The differentiation of positive emotion from negative emotion in the psychology literature raises the interesting question of the relationship between positive and negative emotions in FL learning in particular: Are they like the two faces of Janus, or are they different gods/concepts altogether? MacIntyre and Gregersen (2012) argued in favour of the latter position: "Positive emotion has a different function from negative emotion; they are not opposite ends of the same spectrum" (p. 193). They argue that learners' imagination has positive-broadening power, a perspective that is consistent with Fredrickson's (2001) description of positive emotions as actively promoting health and well-being and not simply being the absence of negativity.
Research on several of the positive emotions has yet to be undertaken in SLA. In the present study, we sought a positive emotion that was parallel to the well-studied negative emotion of FLA. We decided to use enjoyment for several reasons. Theoretically, enjoyment is a defining component of Csikszentmihalyi's (1990) concept of flow, a positive state where challenges and skills to meet them are aligned well. Enjoyment is indicative of a state in which psychological needs are being met. Most emotion theorists list joy as a basic emotion that arises in situations where people experience ". . . desirable outcomes related to personal success and interpersonal relatedness" (Reeve, 2005, p. Csikszentmihalyi (1990) notes that experiencing enjoyment involves having a chance to complete a task, concentration, clear goals, and immediate feedback. Language students often will comment on how much (or how little) they enjoyed a particular language learning activity, class session, and/or course. On a daily basis, the process of language learning will implicate the two key sources of enjoyment: developing interpersonal relationships and making progress toward a goal. In an early study of flow, Csikszentmihalyi (1975) described how a challenging activity that well exceeds a person's skill level can lead to worry and anxiety, but as skills come into line with the degree of challenge, enjoyment and flow experiences can emerge. If skill far exceeds the challenge, however, Csikszentmihalyi would suggest apathy and boredom would set in. It is therefore important to know whether students are enjoying the course and its activities, whether they are feeling anxious, or whether they are not feeling much of anything.
Examining both positive and negative emotions in the same study is an advisable practice. The absence of emotion is more difficult to interpret. Dewaele and MacIntyre (2014) have argued that a lack of enjoyment, by itself, can be ambiguous; does it reflect boredom, hating the subject, the perceived irrelevance of the teacher and the peers, or something else? In a similar way, the absence of FLA might be ambiguous. Does low anxiety reflect confidence in one's ability to perform well or a lack of engagement in the learning process (Csikszentmihalyi, 1990;Egbert, 2004;Tardy & Snyder, 2004)? What seems clear is that positive emotions such as "interest-enjoyment" are associated with better learning, while negative emotions are negatively related to it (Ryan‚ Connell, & Plant‚ 1990, p. 14). Ryan et al. found that positive and negative emotions were related directly to text comprehension in the L1, which in turn directly predicted subsequent recall (p. 14). The effect of emotions was maintained even when variation in verbal ability was taken into account (p. 14).
The present study investigates both Foreign Language Enjoyment (FLE) and FLCA, respectively a positive and negative emotion, using an internet-based survey. If FLE and FLCA correlate to a large extent, it might be possible to claim that FLE and FLA are indeed a Janus-faced concept. If they show clear separation, we will have support for the independence of positive and negative emotions. The use of an internet-based survey allowed for a wide variety of participants from all over the world. The sample was obtained by snowball sampling that involved sending email requests to various lists of language educators and asking them to forward the survey to interested colleagues. The software used allowed for data collection on both Likert-scale and open-ended items.
Before beginning this project we developed a new FL enjoyment scale. We used Ryan et al.'s (1990) Interest/Enjoyment subscale as the kernel of the new scale: Their 7 items related to enjoyment, fun, interest, boredom were specifically adapted to the FL environment. They were rephrased so that they refer not to a single activity at one point in time but elicit a more global judgment of past FL classes. We added items relating to dealing with FL mistakes made in public, identity, improvement in using the FL, pride in one's own performance, group membership, the social environment and cohesiveness, attitudes towards the learning of the FL, the presence of laughter, and judgments about peers and teachers. Nine items were first-person singular ("I"), statements, 3 items were first person plural ("we") statements and the remaining 9 statements were third person singular or plural statements ("it," "the peers," "the teacher"). The final scale contains 21 items reflecting various facets of FLE (creativity, pride, interest, fun) and a positive environment in the FL class (teacher and peers). In addition, we chose to use 8 FLCAS items reflecting physical symptoms of anxiety, nervousness and lack of confidence (Horwitz et al., 1986). The items were chosen to capture the reliability of the original scale without sacrificing the reliability of the measurement (MacIntyre, 1992). We then investigated whether the independent variables that are linked to FLE have the same effect on FLCA. Finally, we took an emic perspective (Dewaele, 2009), looking at what participants had to say, in their own words, about episodes in their FL that they really enjoyed.
The following were our research questions: 1. Is there more enjoyment in FL classes than anxiety? 2. How strongly linked are FLE and FLCA?
3. What is the effect of number of languages known, number of FLs studied, mastery in the FL, relative standing in the group, education level, age group, gender and global-regional group on FLE and FLCA? 4. What do respondents say characterizes positive episodes of enjoyment in the FL class?

Gathering the sample
The sample was obtained by convenience sampling. Some data was collected by a volunteer who coordinated testing of the youngest participants. The majority of the data was collected through snowball sampling.

Participants and demographics
A total of 1,746 multilinguals filled out the web-based questionnaire. From the demographics section of the instrument, we extracted a series of independent variables that were used to create groups for data analysis.

Sex
The majority of the sample (n = 1287, 73.7%) was female; approximately one third of the respondents were male (n = 449). This sort of distribution, with a majority of female participants, is quite typical in web-based language questionnaires (Wilson & Dewaele, 2010).

Education
Participants came from all levels of education, with 91 having an intermediate high school diploma, 113 a high school diploma, 994 a Bachelor's degree, 450 a Master's degree and 94 a PhD. Overall, the sample is well-educated.

Languages being studied
Close to half of the participants reported that they studied English as a FL (n = 855, 49%), followed by French (n = 276, 15.8%), Spanish (n = 218, 12.5%), Dutch (n = 157, 9%) and German (n = 139, 8%). The remaining 100 participants (5.7%) studied a variety of other languages. Only a small minority (n = 34) was not currently involved in FL learning but reported recent FL experience. A majority (n = 1322) was studying one FL, with smaller numbers studying two (n = 215) or three (n = 37) FLs. A quarter of participants had been studying a FL for 5 years or less, another quarter had studied it between 6 and 9 years, a third quarter had studied a FL between 10 and 12 years, with the remaining participants having studied a FL between 13 and 50 years.

FL mastery
Asked about how they would describe their mastery of the FL in general, 5 described themselves as beginners (

Relative standing
Asked about how they would describe their relative standing in the group of FL learners, 42 described themselves as far below average (2.4%), 171 as below average (9.8%), 772 as average (44.2%), 638 as above average (36.5%) and 116 as far above average (6.6%).

The instrument
The questionnaire started with a demographics section containing items as described immediately above. The next section asked participants to indicate the extent of their agreement with 29 items describing the FL class, of which 8 were extracted from the FLCAS and 21 items written to reflect FLE. Responses were given on standard 5-point Likert scales with the anchors absolutely disagree = 1, disagree = 2, neither agree nor disagree = 3, agree = 4, strongly agree = 5. All items referring to FLE were positively phrased. Consistent with the original FLCAS, 2 items were phrased to indicate low anxiety and 6 items were phrased to reflect high anxiety (see Appendix A and B). The two low anxiety items were reverse-coded so that high scores on this measure reflect high anxiety. A final open-ended question asked for a focussed essay (MacIntyre, Burns, & Jessome, 2011) on a specific event that the respondents enjoyed. The instructions were as follows: "Describe one specific event or episode in your FL class that you really enjoyed, and describe your feeling in as much detail as possible." The questionnaire was pilot-tested with 15 participants. This led to the deletion of some items and the reformulation of others. The research design and questionnaire obtained approval from the Ethics Committee of Birkbeck College, University of London. The final version of the questionnaire was posted on-line using Googledocs and an open call was addressed to multilinguals, including those who had participated in previous studies, asking them to forward the call to friends, teachers or students.

Data analysis
Average scores on the 5-point scale were calculated for FLE (N = 1746, M = 3.82, SD = .46). Scores ranged from 1.29 to the maximum 5.0. Internal consistency of the 21 items, as measured by Cronbach alpha coefficient, was high (.86). A one-sample Kolmorogov-Smirnov test revealed that the distribution was not normal (Z = 2.01, p < .05), but the Z-result was low. A look at the distribution ( Figure 1) with a trendline reflecting moving averages shows a Bell curve with a skew toward scores on the positive end of the distribution, with the tail in the low enjoyment section.
Average scores on the 5-point scale were calculated for FLCA items (N = 1746, M = 2.75, SD = .83). Scores ranged from 1 to the maximum 5. Internal consistency of the 8 items, as measured by Cronbach alpha coefficient, was as high as for FLE (.86). Here again a one-sample Kolmorogov-Smirnov test revealed that the distribution was not normal (Z = 2.84, p < .05), with a higher value for Z than observed for the FLE scale. A look at the distribution ( Figure 1) with a trendline reflecting moving averages shows a wider curve with a slight skew toward scores on the high end of the distribution. Given the results of the normality tests, we exercised caution with the data analysis, testing the relationships reported below with both parametric and non-parametric approaches. The pattern of results showed no difference between the two approaches in the significance tests of the independent variables. One-way ANOVAs and t tests tolerate moderate violations to their normality assumption (skewed distributions) rather well and allow for more detailed post hoc tests; therefore, we will present the parametric statistics (Rosenkrantz, 2008, p. 478). Results from the non-parametric analyses are available from the authors.

Figure 1 Distribution of FLE and FLCA scores
One thousand and seventy-six participants (61% of the total number) answered the open question, producing 52,471 words; on average the responses approach 50 words in length (M = 48.8 words). Dörnyei (2007, p. 243) points out that initial qualitative datasets can be messy and chaotic; ours is no exception. Even though the internet survey questions were written in English and Chinese, participants also responded in French, Spanish, German and Dutch. The variety of languages used by the respondents made the use of Nvivo software impossible; as a consequence the data was coded manually. Some respondents chose to describe specific events (as the question requested) but others described a pattern of events or their perceptions of their experience rather than the experience itself. Given the exploratory nature of this research, the coding process used all available data to generate the categories reported below. Finally, the respondents spontaneously drew links between enjoyment and other positive emotions, such as pride and satisfaction.

Quantitative data
A paired t test revealed that our participants reported significantly more enjoyment than anxiety in their FL classes (df = 1745, t = -41.1, p < .0001). Figure  2 shows also that there was almost twice as much dispersion around the mean for FLCA compared to FLE. In other words, FLE scores were distributed more tightly around the mean, compared with FLCA that had more variability in its scores, more fully covering the range of possible scores.

Figure 2 A comparison of mean FLE and FLCA scores (with SD)
A Pearson correlation revealed a significant negative relationship between FLE and FLCA (r(1746) = -.36, p < .0001). Participants with higher scores on FLE showed lower scores on FLCA. Looking at the strength of this relation- ship, these two variables share only 12.9% of variance, which is considered a very small effect size (Cohen, 1992). 2 A series of one-way ANOVAs were used to examine the effects of the independent (demographic) variables on FLE and FLCA (see Table 1). We found that those who knew more languages, had reached a higher level in the FL, felt that they were above average in their group of peers, were more advanced in their education and were older reported significantly less FLCA and scored significantly higher on FLE. Studying more FLs was associated with more FLE but not with FLCA. There was some variation in the strength of the effect: The number of languages known, the general and relative standing had a stronger effect on FLCA than on FLE. Number of FLs under study, education level and age had a stronger effect on FLE (see Table 1). Each of these independent variables will be examined in more detail below. 38.0*** 63.9*** 3.8** 5.1*** 7.3*** FLE 9.9*** 3.5* 22.7*** 27.2*** 7.3*** 7.7*** 8.6*** * p < .05 ** p < .01 *** p < .0001 Figure 3 shows that knowing more languages is associated with a linear increase in FLE, and with a linear decrease in FLCA until the fifth language, after which it seemed to stabilise. Although the effect is significant, the effect size is small (FLE: eta 2 = .022; FLCA: eta 2 = .032). Tukey post hoc tests revealed that the bilinguals reported significantly lower mean FLE scores than all other groups, trilinguals to sextalinguals (p < .0001). Further, the pentalinguals and sextalinguals showed mean enjoyment scores that were significantly higher than trilinguals (p < .05). As a general trend, FLE tended to increase with the number of languages acquired.

Number of languages acquired
A similar pattern emerged for FLCA, with significant differences (p < .001) between the bilinguals and the other groups. The scores of trilinguals were significantly higher than those from the pentalinguals and sextalinguals (p < .001), and the scores of the quadrilinguals were significantly higher than those of the pentalinguals (p < .05). In general, anxiety scores decline with an increasing number of languages acquired. Figure 4 shows the effect of number of FLs participants were studying on their FLE and FLCA. The effect was significant on FLE (p < .03, with a very small effect size: eta 2 = .004) but not on FLCA (p = .11). The FLE values for those studying one and two languages were very similar, and only the values for those studying three languages differed. Tukey post hoc tests showed that the group studying three languages had significantly higher FLE scores compared with those studying one (p < .023) or two languages (p < .034). Although the FLCA scores appear to decline in the three-languages group, the difference was not statistically significant.

FL mastery
The general level of mastery of FLs, ranging from lower-intermediate to advanced, is significantly linked to both FLE and FLCA (both p < .0001, with effect sizes of eta 2 = .062 and eta 2 = .099 respectively). Interestingly the difference between lower-intermediate and intermediate was minimal, but at higher general levels we notice a surge in FLE and a drop in FLCA (see Figure 5)

Relative standing
A similar pattern emerged for the effect of relative standing in the group of FL learners (see Figure 6). Again, the effect is highly significant for both FLE and FLCA (p < .0001, with effect sizes of eta 2 = .059 and eta 2 = .128 respectively).

FLCA FLE
The values of FLE and FLCA were very close among those who fell far below the average level in their peer group. These values diverged linearly for FLE, and close to linearly for FLCA. Tukey post hoc tests showed that there was no significant difference between the "far below average" and "below average" groups for both FLE and FLCA, but there is significant difference between pairwise comparisons of the other groups (at least p < .003), with the exception of the difference between the "above average" and "far above average" groups. As respondents reported feeling higher in the hierarchy of their peer group, they also reported increasingly high levels of FLE and lower levels of FLCA, with the slope somewhat stronger in the case of declining anxiety.

Figure 6
The effect of relative standing in a group of learners on FLE and FLCA

Education
Grouping the respondents according to their level of education produces a pattern with a shallow slope. As can be seen in Figure 7, FLE levels increase slightly but steadily and significantly across education levels (p < .0001, with eta 2 = .017), but FLCA decreases only at the two highest levels of education (p < .005, with eta 2 = .009). Tukey post hoc tests showed no significant difference in FLE between the mid high school and high school groups, but the mid high school group scored significantly lower than those with BAs, MAs and PhDs (all p < .0001). The differences in FLE between the other groups are not significant, with only a marginally significant difference between the high school group and the PhD group (p = .06). The pattern is slightly different for FLCA with a non-significant increase in values between the mid high school, high school and BA groups after which values start to fall, with the MA and PhD groups

FLCA FLE
scoring significantly lower than the BA group (p < .033 and p < .022 respectively). The difference in FLCA between the MA and PhD groups is not significant.

Age
The effect of age group is strong and significant both on FLE and FLCA (p < .0001, with effect sizes of eta 2 = .026 and eta 2 = .017 respectively), but with differently shaped patterns. FLE shows a steady increase from pre-teens to those in their thirties (see Figure 8). Tukey post hoc tests showed that teenagers scored significantly lower on FLE than those in their twenties (p < .013), thirties (p < .0001) and forties (p < .0001). Those in their forties have significantly higher FLE scores than all younger groups, after which FLE stabilises and does not further significantly increase. Generally, enjoyment tends to rise for those in older learner groups (Figure 8). The pattern is quite different for FLCA, where teenagers score the highest mean anxiety levels (significantly higher than those in their thirties and forties, both p < .004), followed by those in their twenties (who score significantly higher than those in their thirties and forties (p < .044 and p < .025 respectively), after which FLCA levels drop back to pre-teen levels and are no longer significantly different from each other. With this pattern, we see a bump in anxiety during the teens and twenties that drops off in older groups (Figure 8).

Global-regional differences
Global-regional group has a significant effect on FLE and FLCA (both p < .0001 with effect sizes of eta 2 = .024 and eta 2 = .021 respectively). Tukey post hoc tests showed that North American participants reported the highest levels of FLE and lowest levels of FLCA while the Asian participants had the lowest levels of FLE and the highest levels of FLCA. The FLE scores of the North Americans were significantly higher than those of the Arabs (p < .043), the Asians (p < .0001), and the Europeans (p < .0001). The South Americans also scored significantly higher than the Asians (p < .005) on FLE. The differences in FLE between South Americans, Arabs and Europeans were not as pronounced ( Figure 9). The FLCA scores of the Asian group were significantly higher (all p < .0001) than those of all other groups which did not differ among themselves.

Qualitative data: Participants' views on enjoyable episodes in the FL class
The range of what participants understood to be enjoyable episodes in the FL classroom was vast. We thus focused on coding data into main themes. To do this, pre-teens teenagers twenties thirties forties fifties sixties+

FLCA FLE
we used content analysis: "The qualitative categories used in content analysis are not predetermined but are derived inductively from the data analysed" (Dörnyei, 2007, p. 245). The categories thus emerged through close reading of the responses to the open question. Participants' description of enjoyable episodes often included descriptions of specific FL classroom activities. Some episodes were tales of personal triumph in the classroom, either because the teacher had publicly recognised the participant's good performance or given them a high mark on some piece of work, or because the peers had shown appreciation for the performance through their interest or laughter. There could be a slight overlap between the appreciation of the teacher and that of the peers. Other episodes dealt with instances of private personal triumph, such as the sudden realisation of having done something well and having progressed in the mastery of the FL.

Figure 9
The effect of global-regional group on FLE and FLCA Participants varied in their choice of temporal focus: While most chose a time frame ranging from a couple of minutes to an hour, some described episodes of enjoyment spread over a longer period of time. Enjoyment over longer timescales could be linked to qualitatively different interpretations of enjoyment. A recurrent theme was the appreciation for the skills of good FL teachers. Some episodes were situated outside the classroom and involved the authentic use of the FL as a tool for communication. Finally, a category was created for all comments that did not fit any of the previous categories: These included observations about language learning in general, or hating FLs, or the lack of appreciation of the FL teacher and peers, or anecdotes about some funny incident in the FL class that did not belong in any of the categories.

FLCA FLE
The 1076 episodes were assigned to one of the following categories: (a) Authentic Use of FL, (B) Classroom activity, (C) Other, (D) Peer recognition, (E) Realisation of progress, (F) Teacher recognition, and (G) Teacher skills. Major themes are presented in Table 2, ranked in descending order of frequency and relative proportion. One category stands out, Classroom activity, to which two fifths of episodes were assigned. The two most frequent activities were role-plays and debates. The other categories (Peer recognition, Other, Teacher recognition, Realisation of progress, Teacher skills, Authentic use of FL) each represent around 10% of the total number of episodes. We chose data extracts that were most representative of the category, most poignant and most interesting. We also included some extracts in other languages, to give voice to those who did not respond in English. The extracts give a good idea of the participant-in-context, with the enjoyment arising from a unique confluence of internal and external factors.

Classroom activities
We do not aim to report exhaustively on any of the sub-categories. Space permits only a sampling of the responses that were offered. Participants mentioned role-plays, debates, film-making, games, singing, preparing group presentations and other activities that they had enjoyed in the FL classroom. One learner, who we shall call XX, 3 remembers an unusual activity involving group work that helped students' understanding of prefixes and boosted her confidence:

When working on prefixes which indicate the direction of movement (returning, just leaving, leaving) we used little people dolls and toy cars to create scenarios in which people moved by vehicle or by foot and then created sentences which used the pre-fixes and described what was going on. This was really effective because it was active and got us involved in something fun which mitigated some of the stress sometimes associated with plane verb conjugation. I made a lot of mistakes and that did make me a bit anxious (though there is less anxiety associated with small group/pair work rather than large group/full class exercises), but it was still fun and
there was a lot of laughter at the juvenileness of the dolls. (XX,female,24) Another type of group activity, role plays, are appreciated as they allow a certain freedom in the exchange, as LE experienced: Es gracioso cuando tenemos que hacer diálogos con los compañeros, y más cuando uno el otro responde cosas equivocadas. o cuando en los diálogos una realmente imita a el personaje que le corresponde. [It's nice when we have dialogues with our peers, and when the interlocutor answers something wrong, or when in the dialogues one can really imitate the corresponding character.] (LE,female,24) More serious group activities can be the source of enjoyment for other students. For example, debates can be intellectually stimulating in the FL class, as KX remembers: During one of my FL classes we organized a debate about different kinds of media manipulations. It was really stimulating as far as the topic and the language skills utilized are concerned. I really enjoyed myself and actually lost track of time. As both student and language teacher, I believe my teacher achieved the highest form of educational success possible-teach and let the students enjoy at the same time. (KX,female,25) Finally, activities that evoke creativity also are well received. Students, as HX explained, love to be creative in the FL class, which boosts their FLE: Last year my classmate and I made a short French film about the dangers of too much partying in first year and it was hilarious to make and show to the class, I enjoyed the freedom and creative aspect of it. (HX,female,19) It is important not to assume that classroom activities are universally enjoyed or anxiety-provoking. Rather, a person-in-context approach suggests that it is the match between the challenge of the activity and the skill of the learner that form a foundation for enjoyment.

Peer recognition
Some participants reported that activities such as speaking in front of peers can have a liberating, even cathartic effect on those who normally avoid being at the centre of attention: I'm a really really shy person in general, but in the last few years, that has changed a lot, mostly because of FL classes where you are "forced" to speak. Last week, we were asked to describe family relationships, which is a very hard question for me to answer. But since the atmosphere in the group is really nice, I thought for once I would try telling the truth, which is that my family and I don't get along because I'm gay and they don't like it (to put things mildly). I was surprised at the overall positive reaction I got from everyone. They were very encouraging which was really really nice. Their reaction was so unexpectedly nice that I cried. (SV,female,24) Peer recognition can also result from clever teacher intervention. Participants reported how lightening the mood in the FL class can help ward off FLCA by lessening the anxiety of peers' reaction:

Teacher recognition
This category contains extracts where participants mentioned public or private praise from the teacher. The praise could be in the form of a good mark for a piece of work or a compliment in front of the peers. MG was really happy with getting recognition from her teacher, especially because she had worked hard.
I love it when I do a writing activity to the best of my ability, using time phrases and specialist vocabulary and I get a great comment and mark, it makes me feel like I'm good at a skill that I enjoy, which makes me very happy. I also love it when I can understand complex sentences. (MG,female,16) Encouragement by the teacher at the first steps was greatly appreciated by MJ (female, 18): "Being complimented on my pronunciation despite feeling as though I struggle slightly -I felt encouraged."

Realisation of progress
Many participants, like XX2, mentioned the moment in their FL development where they suddenly realised how far they had progressed and how the in-vestment of time and effort paid off. This realisation was usually a private moment, but in some cases it was triggered by some classroom activity: During my written class I realised I could write a lot more Japanese than I initially thought, which was quite amazing. It was exciting to see that I ran out of pages. I felt surprised and accomplished, like my learning was paying off, and that I am actually ABLE to use the language, finally, to express things. (XX2,female,25) In the case of JO (female, 14) it was simply the sudden understanding of an interaction in the FL: "I was able to fully understand a learning assistant who was talking to my friend and I and I felt satisfied and proud." Both quotes show that brief episodes of enjoyment triggered a broader, long-lasting positive emotion.
In other cases it's simply the memory of having done something well in the FL, as XC (male, 24), a Chinese participant, remembers wistfully: "The time I spoke fluently."

Teacher skills
This category contains observations about how the participant's teacher created an enjoyable atmosphere in the FL class. Most of the episodes have a longer time-frame than those in the other categories, that is, they describe the behaviour of the teacher over a period of time rather than at any specific point. MF (female, 18) especially appreciated the kindness of her teacher and the respect she has for students: "I like my class of English . . . My teacher is really kind and she really help us and she respects the students." Some participants, like GB (female, 21), enjoy teachers who expect learners to be active and independent members of the class community: "I really like the participation part where the teacher is asking our opinion. Being able to express our thoughts and making students active rather than passive members is going to aid the learning process." A good teacher who can create a positive atmosphere boosts learners' willingness to improve their skills and is greatly appreciated:

I'd say that I really enjoyed my Modern English classes in Year
One. We had the most amazing professor; we had everything organised, taught and assessed perfectly and we worked until the very last minute. I tend to get a bit lazy, but with this professor and the atmosphere that she had been creating, I never lost concentration even when we had 4 hours in a row (of the same subject). (SD,female,20) Participants mentioned both actions by teachers and their positive attitude: I just generally liked the lessons because of the lesson slideshows and the teacher's happy attitude towards everything. She would make sure everyone is very involved all the time. I don't have a specific moment because I enjoy all of it. (TC,female,13) Enjoyment of the FL class can also depend on the organisation of the teaching, more specifically the number of students in class and the use of native speaker teachers: C'est une petite classe de 10 personnes, cela permet à tout le monde de se sentir à l'aise et ce que je trouve génial c'est que ce sont des personnes dont l'espagnol est la langue maternelle qui nous enseignent cette langue. [It's a small class of 10 students, it allows us to feel comfortable, and I love it that the teachers are all native speakers of Spanish.] (LM, female, 21)

Authentic use of FL
This is the category with the smallest frequency of episodes, possibly because authentic use of the FL typically happens outside the FL classroom. Some participants pointed out that the authentic use of the FL can boost FLE. EX enjoyed using her German on a game site away from the prying eyes of her peers and teacher. This use of the FL as a tool for communication can be crucial at all levels and ages: In French we pretended we were in a sandwich shop and had to order our own sandwich. I really enjoyed it as it was a good way to practice our skills. I felt really confident doing this and in the end I got a nice sandwich for my efforts! (SX, female, age 11) When we went on a German games site we could play games and really enjoyed playing a game with my friend where we could both join in at the same time. It was good because you were not being watched by the rest of the class and it was more fun because it didn't matter if you got it wrong, you just learnt the correct answer and it was more of a joke than feeling embarrassed or shy. (EX,female,11) CR (male, 20) seeks opportunities to use his FL out in the street in authentic interactions: "What I like is to use the language, to find someone in the street and to talk to him."

Other
This category was quite heterogeneous. Some participants, like FS, reported enjoying things in the FL class that were probably not intended by the teacher. For example, when class work is boring the only relief comes from chatting with peers. It is not clear if FS (female, 15) talks to her friends in the FL: "I usually only enjoy the lessons when I'm talking with friends. I find the topics very boring and repetitive and so don't really have an episode in the class that I really enjoyed."

Discussion
The results of this study help to shed light on the different patterns of positive and negative emotions among FL learners. Overall, participants reported significantly higher levels of FLE than of FLCA. Moreover, the distributions of scores for FLE indicated consistently high levels of enjoyment with less variation around the mean than for FLCA, which showed a dispersion of scores across the full range of the scale. These results reveal that our FL learners tended to be consistent in enjoying FL classes but showed diversity in FLCA levels. Although there was a significant negative correlation between FLE and FLCA, they shared only 12.9% of their variance (a small effect size; Cohen, 1992) and the two distributions of scores were quite different. We can therefore claim that these two dimensions are related but that enjoyment and anxiety appear to be independent emotions, and not opposite ends of the same dimension. This result indicates that the absence of enjoyment does not automatically imply a high level of FLCA, and an absence of FLCA does not mean a presence of FLE. In terms of emotional dynamics, it is easy to imagine a person who is enjoying a language class or native speaker conversation and who experiences some anxiety from time to time (see MacIntyre & Legatto, 2011), or a disinterested student with both low enjoyment and low anxiety. The excerpt from XX in which she mentions group work with toys describes anxiety and enjoyment occurring together.
The pattern of the results, specifically the higher score for the positive emotion (enjoyment) and lower score for the negative emotion (anxiety), will be encouraging to many language professionals. In terms of the theory of positive emotion, enjoyment is associated with the urge to play, to be creative and to expand one's limits (Fredrickson, 2001). Fredrickson's ideas suggest that experiencing enjoyment and playfulness in language might be an especially facilitating experience for language learners, as play has been associated with fostering social bonds and brain development. Even among adults, enjoyment and its associated playful behaviours offer a safe psychological base from which to explore an unfamiliar linguistic and cultural world. Enjoyment might be the emotional key to unlocking the language learning potential of adults and children alike; if a teacher, parent, friend, or mentor creates an enjoyable context, they likely have gone a long way towards facilitating learning. As Fredrickson (2001Fredrickson ( , 2013 suggests, the broadly-defined resources that are built during play, such as specific skills and relationship bonds, endure long after the joyful feeling has dissipated; therein lies the power of a positive-broadening emotion (MacIntyre & Gregersen, 2012).
Anxiety, on the other hand, always has had a somewhat conflicted status in the literature. The vast majority of studies show a consistent, negative correlation between language anxiety and FL proficiency scores, including course grades, standardized tests, and other measures (Horwitz, 2010). In terms of emotional effects, strong anxiety feelings are no doubt disruptive to behaviour, interfering with interpersonal communication, cognition and learning (Eysenck, 1979;Gregersen & Horwitz, 2002;MacIntyre & Gardner, 1994a, 1994b. But consistent with the adaptive nature of emotion in general, and the narrowing effects of negative emotion and anxiety in particular, it is likely that enjoyment and anxiety will cooperate from time to time, enjoyment encouraging playful exploration and anxiety generating focus on the need to take specific action (Arnold & Brown, 1999). Conceptualized as two separate dimensions, the question becomes one of describing a constructive balance between enjoyment and anxiety, rather than implicitly taking them as opposite ends of the same dimension. Our data suggests that, among the active and successful learners in our sample, the balance is tipped in favour of enjoyment, with some anxiety present as part of the emotional mix. Therefore, consistent with Fredrickson's (2013) updated thinking on positive emotion, the ratio of positive to negative emotion warrants close scrutiny.
We do need to exercise some caution in interpreting these findings. Although our sample is very large, it also is mostly well educated and self-selected. The sample likely does not represent the general population of FL learners, if such a sample actually could be defined in a meaningful way. A positive bias towards FL learning is likely to be present among our respondents: FL learners who really hate FLs would probably not want to spend 15 minutes filling out an online questionnaire on FL learning, even if they were still taking a language course. Moreover, it is probable that only the most enthusiastic FL learners keep studying FLs until they graduate. Nevertheless, the wide range of the present sample offers at least some protection in the sampling procedures.
The third set of research questions dealt with the effect of a series of demographic, sociobiographical and FL-related variables. A series of ANOVAs allowed us to calculate the effect of number of languages known, number of FLs studied, mastery of the FL, relative standing in the group, education level, age group, gender and global-regional group on FLE and FLCA. The findings probably will not surprise most readers: 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 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.
Significant cultural group differences emerged for FLE and FLCA. North American participants seemed to have the most enjoyment and the lowest anxiety, while Asians reported least joy and highest anxiety, with South Americans, Arabs and Europeans scoring somewhere in the middle. It is crucial not to over-interpret or over-generalize these differences, given that the number of respondents in each of the groups differs greatly and there is considerable variation within groups as well. Over four-fifths of the sample consists of Europeans, which means more diversity in nationalities, age range, proficiency levels, teaching methods, exposure to the FL than in the smaller global-regional groups. Among the smaller groups, a strong proportion of students in one school or university might affect the results for the whole region.
Of special note is that the sex of the respondent showed significant differences on both FLE and FLCA: The female participants scored higher on FLCA and on FLE. The female participants thus experienced both more negative and positive emotions in the FL class. Indeed, stereotypically, females are often credited with being the better FL learners as well as more emotionally intense than males (Chaplin & Aldao, 2013). The fact that our female participants scored higher on both FLCA and FLE is intriguing because it is consistent with the notion that stronger positive and negative emotions co-occurring in FL learning might be preferable to weaker emotions overall. If both positive and negative emotions are inherently adaptive, then stronger emotional experiences, if they occur at optimal times, have some potential to facilitate language learning, even if the emotions include ambivalent or negative feelings (MacIntyre, 2007). Fredrickson (2013) has suggested that the ratio of positive to negative emotions might be more important than simply the absence of negative emotion. In Figure 6 above, we see that the ratio of positive to negative emotion in the most advanced group of learners is approximately 2:1, then 1½:1 in the intermediate group, and finally 1:1 in the group self-described as performing far below average. This suggests the intriguing possibility that future research might examine not simply the scores for individual emotional responses but rather the ratio of positive to negative emotions. Further, examining the ways in which emotions interact with each other, and with various other factors, dynamically as they occur, will help shape our understanding of this intriguing area of SLA MacIntyre & Serroul, 2015).
The final research question dealt with the themes that emerged in episodes of enjoyment in the FL class. We used the qualitative data of the 1,076 participants who offered a response to the open-ended question on enjoyment to complement the previous studies that have examined learners' descriptions of their experience with anxiety (Cohen & Norst, 1989;MacIntyre, Burns & Jessome, 2011;Price, 1991). The responses showed that specific positive events can shape the development of enjoyment in the FL. 4 FL learners enjoyed novel activities, such as using dolls and toy cars, making a short video, or preparing a group presentation. These were typically activities that empowered student choice, such as a topic of discussion or debate that was relevant to their concerns and interests. Learners described not wanting to be treated like passive recipients of knowledge, but enjoyed being allowed to have a sense of autonomy and to use their imagination to make progress in the FL (MacIntyre & Gregersen, 2012). As Arnold (2011) pointed out, in general, choice is motivating. Teachers who build a comfortable degree of choice into their classes can motivate learners. Group projects are particularly rich in potential choices because they can require different kinds of skills and offer a range of options (Arnold, 2011).
The classroom environment can facilitate enjoyment. In particular, teachers who were positive, humorous, happy, well-organised, respectful of students, and praised them for good performance were appreciated by their students. Respondents forgave teachers for gentle teasing and for laughing together when they made mistakes. Laughter that occurs when things do not go as planned can have a healthy effect on learners, taking the negative emotional tension out of the room. The qualitative data suggests that teachers are closely connected to the FLE of their students (Arnold, 2011). Peers also can facilitate enjoyment. Perhaps it helps not to have too many peers in the FL class: Smaller groups are more conducive to closer social bonds, a positive informal atmosphere, and to more frequent use of the FL.
Our participants' views confirmed Arnold's (2011) observation that experiencing real achievement is the surest route to self-esteem and enjoyment. The enjoyable events that participants described frequently were the result of a long period of hard work toward a goal, such as speaking fluently or communicating with native speakers (Csikszentmihalyi, 1990). Participants interpreted these events as recognition of their effort. Of course, teachers need to deal with negative emotions and self-beliefs students may have, as these can inhibit their progress and encourage realistically positive views of themselves (Arnold, 2011). Side conversations with peers might be an enjoyable relief from monotonous lectures but can also be distracting in a classroom. Ideally, the FL classroom is an environment that is both stimulating and non-threatening (Arnold, 2011;Dewaele, 2005Dewaele, , 2011MacIntyre, 1999;Nelson & Murphey, 2011). It is difficult to over-estimate the impact of classroom dynamics. Dörnyei and Murphey (2003, pp. 3-4) pointed out that in ". . . a 'good' group, the L2 classroom can turn out to be such a pleasant and inspiring environment that the time spent there is a constant source of success and satisfaction for teachers and learners alike." These findings reflect many of the observations in  on ways for FL teachers to reduce FLCA and to boost enjoyment in the FL.
As a caveat, it is important not to overgeneralize these descriptions or to prematurely fit them into "laws" of effective teaching and learning. Each respondent is positioned within a unique context and found enjoyment in the activities above as they took place in that context. As  noted in their comparison of situations producing high and low willingness to communicate, the differences in situations that promote a positive or negative emotional reaction often are quite subtle, with the potential to go either way at any moment. Teasing by the teacher is a clear example of an action that easily might be misinterpreted with only subtle variations in the classroom situation. We wish to emphasize that the same event might be enjoymentinducing or anxiety-provoking depending on how it is interpreted by the learner in the moment. This was nicely illustrated in the extract by NM where she explained how embarrassed she was by her teacher's scolding but how this repeated scolding turned into a running joke as the year progressed.

Conclusion
Can Foreign Language Enjoyment and Foreign Language Classroom Anxiety be compared to the two faces of Janus; are they the two sides of the same coin? Although a significant negative correlation was found between FLE and FLCA, the amount of shared variance is relatively small, allowing us to conclude that they cannot be compared to two faces of Janus. Perhaps it is better if we represent FLCA by Phobus, the Greek god of fear (that has no Roman equivalent), and FLE by Laetitia, the Roman goddess of joy (Dillaway, 1833). It seems that FL learners turn away from Phobus towards Laetitia as they progress.
This study presents statistical evidence to support the hypothesis that FLE and FLCA are different dimensions and not two sides of the same coin. Our participants reported significantly more FLE than FLCA, a pattern that seemed more striking at higher levels of self-perceived general and relative standing in the FL. The pattern of results is consistent with the suggestion that the ratio of positive to negative emotion might be more important than the presence or absence of either type of emotion. Self-selection bias in our large sample makes it difficult to generalize to all contexts and these results are part of an emerging series of studies on positive emotion in SLA. Narrative material on episodes of enjoyment in the FL class showed a diversity of elements that contributed to participants' FLE, ranging from teachers' professional and emotional skills to a kind and supportive peer group. Many participants seemed aware of the maxim "we have to learn to walk before we can run" and their episodes focused on their first suc-cessful steps or their first convincing run in the FL which filled them with pride and a sense of achievement. Savouring small successes is good advice for the long term health of both teachers and learners (Seligman, 2006).