Foreign Language Classroom Anxiety of Arab learners of English: The effect of personality, linguistic and sociobiographical variables

The present study focuses on the link between psychological, sociobiographical and linguistic variables and Foreign Language Classroom Anxiety of 348 Arabic learners of English (250 females, 98 males). Data were collected using the Arabic Foreign Language Anxiety Questionnaire (AFLAQ; Al-Saraj, 2011, 2014) and an Arabic version of the Multicultural Personality Questionnaire-Short Form (MPQ-SF; van der Zee, van Oudenhoven, Ponterotto & Fietzer, 2013). Multiple regression analyses revealed that self-perceived proficiency in oral English and frequency of use of English explained over a third of variance in FLCA: More proficient and frequent users felt less anxious. Two personality traits, Emotional Stability and Social Initiative explained a further fifth of variance in FLCA, with emotionally stable and more extraverted participants scoring lower on FLCA. Age was the final predictor of a small amount of variance, with older participants feeling less anxious. Degree of multilingualism, sex and education level had no effect on FLCA.


Introduction
The number of studies on Foreign Language Classroom Anxiety (FLCA) has grown substantially in the last few years, with researchers investigating both internal and external sources of FLCA (Dewaele, 2012;. The puzzle about the characteristics of anxious foreign language learners is still incomplete. While FLCA has been linked to various psychological variables, no study has, to our knowledge, considered the relationship between FLCA and all high-order personality traits. It thus remains unclear to what extent FLCA is linked to a FL learner's complete personality profile, in addition to well-researched sociobiographical and linguistic variables. This is a highly relevant question because of the relative stability of personality traits, and the relative stability of sociobiographical and linguistic background variables. If a strong relationship exists between FLCA, psychological, sociobiographical and linguistic background variables, then FL teachers and researchers should be aware that FLCA cannot simply be flushed out of the FL learner. What teachers can envisage are ways of allowing learners to handle this negative emotion and boost positive emotions instead, or at least develop a productive interaction between both (Dewaele & MacIntyre, 2014;Gregersen & MacIntyre, 2014). Horwitz (2010) pointed out that "the concept of anxiety is . . . multi-faceted, and psychologists have differentiated a number of types of anxiety including trait anxiety, state anxiety, achievement anxiety, and facilitative-debilitative anxiety" (p 145). She argued that the profusion of anxiety types could explain why "so many early studies on the relationship between 'anxiety' and achievement provided mixed and confusing results" (p 145). Horwitz, Horwitz and Cope (1986) emphasised the multi-faceted nature of FLCA, defining it as "a distinct complex of self-perceptions, beliefs, feelings and behaviours related to classroom learning arising from the uniqueness of the language learning process" (p. 128). FLCA develops over time. It typically starts as an undifferentiated, negative affective response to some experience in the FL class (MacIntyre & Gardner, 1991b). Through repeated experience this anxiety becomes firmly associated with the FL class. It thus becomes a situationspecific anxiety, though it can spill over into other situations where the FL needs to be used. Gardner (1991a, 1991b) noted that FLCA can be exacerbated by excessive self-evaluation, worries over potential failure, and concern over the opinion of peers. All this leads learners to waste precious cognitive energy, disrupting information processing, and as a result hindering FL performance and acquisition . Levels of FLCA have been linked to a range of sociobiographical, social and educational variables, but relatively less research has focused on the effect of personality traits on FLCA (Horwitz, 2010;Shao, Yu, & Ji, 2013).
The present study focuses on Arabic learners of English. It stems from our concern that despite the fact that English has become the most widely taught foreign language in Arab countries, relatively little research has focused on Arabic students' experiences of FLCA (see however Elkhafaifi, 2005;Melouah, 2013). Arabic students' FLCA in English may be exacerbated by the switch from primarily rote methods of instruction in the high school system to Western-influenced teaching methods in university English classes, requiring more independent work, more group work and a focus on oral communication (Al-Saraj, 2014). We also argue that it is useful to consider FLCA within one cultural context, and hence that the instrument to measure FLCA should be tailored to the specific culture (Al-Saraj, 2011). This allows us to avoid the "vexing problems . . . on whether personality trait scales possess conceptual and functional equivalence across culture" (p. 175). It would make sense, then, to see to what extent a personality questionnaire specifically adapted for Arabs such as the Multicultural Personality Questionnaire (MPQ; van der Zee & van Oudenhoven, 2001) can predict the FLCA of Arabs in English FL classes, in addition to the usual sociobiographical, social and educational variables. It is not our aim, in the present study, to compare the findings of our Arabic learners with any other group, we are only interested in within-group differences and relationships.
We are aware that the present study fits in the classic, modular individual differences paradigm, which has come under increasing criticism lately (Dörnyei, MacIntyre, & Henry, 2014;Dörnyei & Ryan, 2015). Indeed, Dörnyei and Ryan point out that the identification of small and discrete components of learner psychology, measured within well-selected learner samples in order to predict the effectiveness of second language acquisition (SLA) is a seemingly logical approach that is in fact an illusion. Adopting a dynamic systems approach, Dörnyei, MacIntyre and Henry (2014) argue that since change in SLA is typically nonlinear, it is very difficult to make predictions. They also insist on looking at "the whole system and the interaction of the parts, rather than focusing on specific units (e.g., variables) within it" (p. 2). As a consequence, they propose to find "alternatives to conventional research methodologies that, by and large, relied on statistical procedures to examine linear rather than nonlinear relationships" (p. 2). They thus issue a call to SLA researchers to go and explore "unchartered territories" (p. 2). We would like to argue that while it is excellent to boldly go where no man (or woman) has been before, it is still important to continue the exploration of supposedly "charted territories." Just as artists can revisit common themes with striking originality, researchers can cast fresh light on supposedly familiar topics. Indeed, we feel that establishing links between small and discrete components of learner psychology and linguistic profiles, within specific learner samples, can help us gain a better understanding of the dynamic and unique connections of FL learner emotions and their selfperceived FL performance.

Literature review
Echoing what Dörnyei et al. (2014) and Dörnyei and Ryan (2015) said about change in SLA and the importance of looking at the whole system, the interaction of the parts, and avoiding microscopic analyses, we argue that the same is true for the area of Foreign Language Anxiety (FLA) itself. The field is developing quickly, is in a constant state of flux, with some parts developing faster than others, and links being established both within and outside the traditional area of FLA research, with increased geographical coverage in terms of populations. It is thus important to take stock of where FLA research stands today, and see in what direction it may evolve.

Proficiency in the FL
Self-perceived proficiency in the FL and self-reported course grades are usually found to be the strongest predictors of FLA or FLCA (Arnaiz & Guillén, 2012;Horwitz, 2001;Liu, 2006;Liu & Chen, 2013;MacIntyre & Gardner, 1994;Shao, Yu & Ji, 2013;Sparks & Ganschow, 2007;Thompson & Lee, 2013, 2014. Participants who feel proficient and/or get good grades typically suffer less from FLA/FLCA. High levels of self-perceived proficiency in the FL and good course grades are typically also linked to more positive attitudes towards the FL (Liu & Chen, 2013). However, Onwuegbuzie, Bailey, and Daley (1999) found that students with high levels of university academic achievement reported more FLCA. Similarly, Saito and Samimy (1996) found that the advanced Japanese students of English suffered more from FLCA than intermediate peers. Also, Marcos-Llinás and Juan-Garau (2009) reported that students with high levels of FLCA did not score significantly lower on course achievement, which could be an indirect indicator of proficiency. The proportion of anxious speakers in English classes among Chinese undergraduate students at three different proficiency levels remained quite stable (Liu, 2006).
The independence of the measures have also been questioned, as selfreported FL proficiency scores may themselves already be biased because of anxiety. MacIntyre, Noels and Clément (1997) showed that after controlling for actual proficiency in the French L2 of 37 Anglo-Canadian students, the anxious students underestimated their proficiency relative to less anxious students, who tended to overestimate their proficiency (p. 265).
The image that emerges from the research is not clear-cut: The wide variety of proficiency or self-proficiency measures, of FLA and FLCA measures, as well as the complex interactions between learner characteristics and situational variables, obscure any panoramic view.

Frequency of use of the FL
How frequently learners are exposed to a FL and how frequently they use it typically determines how proficient an individual will feel in that language and how much anxiety that individual will feel when using that language (Dewaele, 2013b;Liu, 2006). A study of North American students spending a year abroad in France showed that those who used French broadly and frequently made much more progress in French than those who only used their French in limited service encounters (Kinginger, 2011). Dewaele (2013b) found a highly significant effect of frequency of use of a FL on FLA in that language: A higher general frequency of use of the L2, L3, L4 or L5 corresponded to lower levels of FLA experienced by 1453 adult multilinguals in these languages with various types of interlocutors in various situations. Thompson and Lee (2014) also reported that 148 Korean EFL students who had studied abroad were less anxious in the English FL classroom after their return. Frequency of use and self-perceived proficiency are different measures, but they are probably influencing each other. Indeed as FL users gain mastery in the FL, they are likely to seek more opportunities to use the FL which will further boost their proficiency in the FL.

Gender
The effect of gender on FLA/FLCA is not clear-cut: In some studies female students score higher on FLA/FLCA than their male counterparts (Arnaiz & Guillén, 2012;Dewaele, 2007). In other studies female learners reported lower levels of FLCA (MacIntyre, Baker, Clément, & Donovan, 2002). Sometimes researchers found no gender effect at all (Aida, 1994;Dewaele, Petrides, & Furnham, 2008) or limited effects (Donovan & MacIntyre, 2005). Donovan and MacIntyre (2005) did not find gender differences in FLA among Canadian junior high and high school students, but female university students reported higher levels of FLA than their male peers. Generally, it seems that the effect size associated with gender is very small (MacIntyre et al., 2002, p. 558). In recent research on a large internet-based sample, Dewaele and MacIntyre (2014) found that the 1278 female participants reported significantly more FL enjoyment and FLCA than the 449 male participants, although the effect size was small. A follow-up study of specific items of the FLCA scale showed that female participants experienced significantly more mild forms of FLCA: They worried significantly more than male peers about their mistakes and were less confident in using the FL, again with a small effect size, but no gender difference emerged for items linked to paralysing effects of FLCA (Dewaele, MacIntyre, & Boudreau, 2015). Gender thus seems to have very different effects on FLA/FLCA in various studies.

Age
The effects of age are equally variable, possibly linked to age groups of participants (typically teens and young adults at university). Bailey, Onwuegbuzie and Daley (2000), Dewaele (2007) and Donovan and MacIntyre (2005) reported that older participants suffered more from FLCA than younger ones. However, Arnaiz and Guillén (2012) found the opposite pattern, namely older adult multilinguals reporting less FLA than younger adults in their different languages. A study into a large sample of adult multilinguals showed that FLA seems to peak for participants in their twenties, after which it drops consistently across age groups (Dewaele, 2010b). Dewaele and MacIntyre (2014) found that teenagers scored highest on FLCA, followed by participants in their twenties, with a steady drop in FLCA among older age groups.

Linguistic profile
Multilingualism and specific language profiles have been linked to FLA/FLCA. Knowing more languages has been linked to lower levels of anxiety in all languages (Dewaele, 2007(Dewaele, , 2010a(Dewaele, , 2010bDewaele & MacIntyre, 2014;Dewaele et al., 2008;Thompson & Lee, 2013). Typological distance between known languages and the target language also affect levels of FLA. Learners who already know a language belonging to the same linguistic family as the target language typically suffer less from FLA/FLCA than peers who have no such affordances (Dewaele, 2010a). Dewaele (2010a) found that learners of French who already knew one or more other Romance languages reported lower levels of FLA in French compared to participants who knew languages belonging to other families.
Another factor affecting FLA is the order of acquisition of FLs. Levels of FLA have been found to increase significantly, and linearly, among pentalinguals in FLs learnt later in life. These later FLs were typically not mastered to the same degree as those acquired earlier in life (Dewaele, 2013b). Other linguistic profile variables linked to FLA were age of onset of acquisition, with higher age correlating with higher levels of FLA in different situations; context in which an FL had been acquired, with formally instructed participants feeling significantly more anxious than mixed and naturalistic learners; socialisation in the FL and size of FL networks were also linked to the FLA: More socialised participants with larger networks of interlocutors in a FL were less anxious across situations (Dewaele, 2013b).

Personality traits
Personality traits "refer to consistent patterns in the way individuals behave, feel and think" (Pervin & Cervone, 2010, p. 228). They thus "summarize a person's typical behavior" (2010, p. 229). The currently dominant taxonomy of personality traits is the so-called Big Five: five broad, bipolar dimensions, namely Openness, Conscientiousness, Extraversion, Agreeableness and Neuroticism (2010, p. 228). The Big Five are situated at the summit of the hierarchy; there are a large number of narrower facets, "lower-order" personality traits, that are often correlated with Big Five traits but also explain unique variance.

The multicultural personality dimensions
One Big Five questionnaire is the Multicultural Personality Questionnaire (MPQ; van der Zee & van Oudenhoven, 2000, 2001, which was developed to assess five traits that are key to multicultural effectiveness, cultural adaptability and psychological well-being in a foreign environment. As van der Zee, Zaal and Piekstra (2003) pointed out: "Even the MPQ scales that closely correspond with Big Five scales are designed to cover more narrowly those aspects of the broader trait that are of relevance to multicultural success" (p. 78). These five traits have "demonstrated incremental validity over broad personality measures such as the Big Five in predicting criteria such as students' international orientation" (van der Zee et al., 2013, p. 118).
The five dimensions used in the MPQ are Cultural Empathy, which "refers to empathizing with the feelings, thoughts, and behaviours of culturally diverse individuals" (p. 118); Openmindedness, which "reflects an open and unprejudiced attitude toward cultural differences" (p. 118); Social Initiative, which "refers to actively approaching social situations and demonstrating initiative in these interactions" (p. 118); Emotional Stability, which "reflects an ability to stay calm under novel and stressful conditions" (p. 118); and Flexibility, which "refers to interpreting novel situations as a positive challenge and adapting to these situations accordingly" (p. 118).
Cultural Empathy is most strongly related to the Big Five dimension of Agreeableness (van der Zee et al., 2003, p. 79); Openmindedness is related to Openness-to-experience (p. 79); Emotional Stability and Social Initiative are strongly correlated with Neuroticism and Extraversion (p. 80). Flexibility is an addition to the Big Five dimensions reflecting intellectual flexibility, that is, a lack of rigidity (p. 80). van der Zee et al. (2003) did find a link between three MPQ dimensions and one verbal dimension: Cultural Empathy, Openmindedness, and Flexibility correlated positively with verbal ability. No research has yet, to our knowledge, combined these dimensions with FLA/FLCA scores, although there has been previous research on specific Big Five dimensions using other instruments.

Neuroticism
Dewaele (2002) found a significant relationship between Neuroticism (used as proxy for anxiety, and the opposite of Emotional Stability) and 100 Flemish students' levels of FLA in their English L3 (p. 31). A further study (Dewaele, 2013a) revealed a significant positive link between Neuroticism and the L2 FLCA of 86 mature multilingual students who were enrolled at the University of London. A similar relationship emerged in the L3 of 66 students (p. 678). A comparable pattern was found in a second group of 62 students from University of Les Iles Balears in Mallorca, Spain. The relationship between Neuroticism and FLCA was significant in the students' L2, L3, and L4 (p. 678). It is likely that participants with higher scores on the Neuroticism scale were more worried about how their linguistic FL competence would be judged.

Extraversion
MacIntyre and Charos (1996) reported a significant negative relationship between Extraversion and French L2 anxiety among Anglo-Canadian students (p. 19). Dewaele (2002) also found a significant negative relationship between Extraversion and 100 Flemish students' levels of FLA in their English L3 (p. 31). The same pattern emerged in Dewaele (2013a), with more extravert English L2 learners from Mallorca reporting significantly lower levels of FLCA (p. 678). This finding was linked to facets of Extraversion such as risk-taking and optimism.

Psychoticism
Psychoticism, which has been described as tough-mindedness, aggressive behaviour and coldness (Eysenck, Eysenck, & Barrett, 1985), appears to be inversely linked to FLA/FLCA. A significant negative relationship emerged between Psychoticism and 100 Flemish students' levels of FLA in their English L3 (Dewaele, 2002, p. 31). A similar pattern emerged in Dewaele (2013a), with English L2 learners from Mallorca who scored higher on Psychoticism having significantly lower levels of FLCA (p. 678). One possible explanation is that high-psychoticism participants feel less anxious because they are less preoccupied about the perception their interlocutors have of them.

Trait Emotional Intelligence
A number of low-order personality traits have also been found to predict levels of FLA/FLCA. Dewaele et al. (2008) considered the link between Emotional Intelligence (EI) and sociobiographical variables on Communicative Anxiety (CA) in the L1, and FLA in the L2, L3, and L4 of 464 adult multilinguals in different situations. Trait Emotional Intelligence was defined as the "the extent to which individuals attend to, process, and utilize affect-laden information of an intrapersonal (managing one's own emotions) or interpersonal (managing others' emotions) nature" (Petrides & Furnham, 2003, p. 39). Participants with higher levels of trait EI reported significantly lower CA and FLA in all their languages, possibly because they are better able to read the emotional state of mind of their interlocutors, thereby controlling their CA/FLA. Shao, Yu and Ji (2013) reported similar patterns among 510 Chinese non-English-major first year students. A significant negative association emerged between trait EI and English classroom learning anxiety (p. 921). The authors found that FLA had a significant mediating effect on EI in predicting students' English achievement and FLA also mediated the relationship between EI and self-rated English proficiency. The authors argue that FLA could be seen as an emotion, one which students who score high on EI might be better able to control: "These students would draw on their emotional competence (e.g., by stimulating their motivational resources or using moderation techniques to reduce their anxiety in English class)" (p. 924). Dewaele and Shan Ip (2013) investigated the relationship between FLCA and Second Language Tolerance of Ambiguity (SLTA) among 73 secondary school students in Hong Kong. SLTA was defined as feeling comfortable with uncertainty and displaying a willingness to try out guesses (Rubin, 1975, p. 45). The study supported Gudykunst's anxiety/uncertainty management theory (Gudykunst, 2005, p. 298), and, more specifically, the axiom that more tolerance of ambiguity is linked to less anxiety. FLCA and SLTA were found to be inversely related and sharing over half of their variance (Dewaele & Shan Ip, 2013, p. 57). In other words, those among the 73 Hong Kong EFL learners who were more tolerant of ambiguity in their English classes were significantly less anxious in using the language. In a similar vein, Thompson and Lee (2013) reported that a previously unreported fourth factor emerged from their factor analysis of FLCA scores from 148 Korean EFL students. They labelled the fourth factor "fear of ambiguity in English" (the other factors were English class performance anxiety, lack of self-confidence in English and confidence with native speakers of English) (p. 10).

Perfectionism
Higher levels of Perfectionism have been linked to higher levels of FLCA. Gregersen and Horwitz (2002) interviewed a small number of specifically selected highly anxious learners (based on FLCA scores) and found that these learners were more perfectionist, setting themselves higher personal performance standards, procrastinating more, being more fearful of evaluation and concerned about errors. Using three datasets through online questionnaires, Dewaele (in press) found significant positive correlations between FLA/FLCA and various measures of perfectionism among three groups of FL learners: 58 adult English L2 users, and 323 Japanese and 69 Saudi students learning English. This seems to suggest that the link between Perfectionism and FLCA is stable across different populations of FL learners.

Some concluding remarks
To summarise, a range of variables that are linked directly or indirectly to FLA/FLCA have been identified in the research literature. Although some researchers have started including psychological variables in their design, they were typically few in number. Indeed, the main variables to have been investigated were sociobiographical in nature or linked to FL instruction and use. We argue that it is necessary to include personality traits in research designs on FLA/FLCA, preferably measured with instruments specifically designed for multilingual and multicultural individuals, in order to see to what extent these traits predict FLA/FLCA, in addition to the more traditional sociobiographical variables and factors linked to FL learning and use.

Research questions and hypotheses
The present study explores the links between higher-order personality traits, sociobiographical variables, linguistic history, current use of English and FLCA in English among Arabic learners. We will address the following research questions: 1. Are there effects of gender, age and education level on FLCA? We hypothesise that female, younger and less educated participants will score higher on FLCA. 2. Is there an effect of self-perceived proficiency and frequency of use of English? We hypothesise that participants who feel proficient and use English frequently will suffer less from FLCA. 3. Is there an effect of the number of languages known on FLCA? We expect that participants who know more languages will feel less anxious in their English foreign language class. 4. Is there an effect of personality on FLCA? We expect that participants who score high on Cultural Empathy, Flexibility, Social Initiative, Openmindedness and Emotional Stability will suffer less from FLCA.

Participants
Potential participants were invited in Arabic to complete the online questionnaire through email as well as notifications on social media including Facebook, LinkedIn, and Twitter. Only Arabic native speakers who were studying English, or had recently completed English classes, qualified. Anyone interested in the study was encouraged to also send the information to friends and family, and colleagues were asked to circulate the invitation. A notice of the study was posted to an online message board at the Saudi Ministry of Education Presidency for Girls. In addition, invitations to participate in the study were circulated to universities in Saudi Arabia, the United Arab Emirates, Oman, Algeria, Tunisia and Egypt. A total of 348 participants (250 females, 98 males) filled out the three parts of the questionnaire. Their ages ranged from 14 to 65 years (M = 27, SD = 10). A quarter of participants had a high school diploma or less, close to half had a bachelor's degree (n = 163), less than fifth had a master's degree (n = 63) and close to 10% had a PhD (n = 33). All were native speakers of Arabic but some had grown up with multiple languages. A majority of participants were bilinguals (n = 155), followed by trilinguals (n = 123), and these were followed by smaller groups of quadrilinguals (n = 19), pentalinguals (n = 27), sextalinguals (n = 8), septalinguals (n = 7), octalinguals (n = 4), and nonalinguals (n = 2), with 2 participants reporting 11 languages and 1 participant reporting 13 languages. We created one group for all those knowing four or more languages (4+).
Nearly 80% of participants were originally from Saudi Arabia (n = 277), with smaller groups from Egypt (n = 9), Yemen (n = 8), the United Arab Emirates (n = 7), and the remaining 47 participants were from 18 different countries. All were studying English or had recently finished studying English in Arabic institutions. We are also aware that Arab culture is relatively heterogeneous, and we do not claim to have a representative nor a very homogenous group of Arab learners and users of English. This will be kept in mind when interpreting our results.
Participants reported frequent use of English (M = 4.1, SD = 1.0 on a 5point Likert scale with the following descriptors: 1 = once a year, 2 = once a month, 3 = once a week, 4 = up to 3 hours a day, 5 = more than 3 hours a day). They judged their oral proficiency in English to be quite high (M = 3.5, SD = 1.1 on a 5-point Likert scale ranging from minimal to maximal proficiency). Self-reported oral proficiency in English and frequency of use of English were positively correlated (Pearson r = .47, p < .0001, N = 348). This suggests that both measures have 22% of shared variance, which Cohen (1992) would categorise as a medium effect size.

Instruments
The anonymous questionnaire was administered online. It had three sections: (a) demographic information (e.g., participants' age, sex) and demographic information pertinent to language learning and cultural background (e.g., country of origin, languages known to participants and order of acquisition), (b) the AFLAQ, and (c) the Arabic MPQ-SF.
We used the Arabic Foreign Language Anxiety Questionnaire (AFLAQ; Al-Saraj, 2011, 2014, specifically designed to assess the anxiety of native speakers of Arabic when using and learning English. The AFLAQ mirrors the Foreign Language Classroom Anxiety Scale (FLCAS; Horwitz et al., 1986) that is commonly used. However, the proportion of items addressing each topic and the particular issues on the AFLAQ are tailored to the Arab student population. As a result, the questionnaire has a particular focus on prompts of anxiety and stressors prevalent in this population (Al-Saraj, 2011, 2014Dewaele & Al-Saraj, 2013). Seventeen items from the original FLCAS (Horwitz et al., 1986) were retained with minimal modification or only slight rewording. For example, FLCAS item 4, "it frightens me when I don't understand what the teacher is saying in the foreign language" (Horwitz et al., 1986), was reformulated to avoid connotations with the Arabic word that would translate to "frightens." The modification is small but crucial, and the new item reads simply (translated from Arabic), "I feel anxious when I don't understand what the teacher is saying in the foreign language." Five other items were more heavily adapted to make them clearer to the Arab students while maintaining the general content. New items were created based on what Saudi students reported in a questionnaire (Al-Saraj, 2011) to be anxiety-provoking situations in the class, namely practiced and spontaneous public speaking, listening and comprehension, and fear of being negatively evaluated. The AFLAQ, which is included in the Appendix, contains 33 items, three of which are reverse-coded, and utilises a five-point Likert scale ranging from 1 (strongly disagree) to 5 (strongly agree). The AFLAQ yields a single overall score indicative of FLCA. Cumulative scores on the AFLAQ range from 33 (a 1 on each of the AFLAQ's 33 items, after re-coding of reverse-coded items) to 165 (a 5 on each of the AFLAQ's 33 items, after re-coding). Higher scores indicate a higher level of FLCA. All 348 participants completed the questionnaire. Scores ranged from 35 to 161 (M = 94. 5,SD = 25.2). A Cronbach's alpha analysis of the 33 items revealed a very high level of internal consistency (α = .96). A one-sample Kolmogorov-Smirnov test revealed that the AFLAQ scores are normally distributed (Kolmogorov-Smirnov Z = 0.76, p = ns).
The MPQ has been used around the world and has been applied to various groups: students of several age levels, local employees, expatriate employees, their spouses and children, general citizens and refugees (van der Zee et al., 2013, p. 118). In all cases, the scales proved to be reliable and showed consistent patterns of correlations with related variables. Van Oudenhoven, Timmerman and van der Zee (2007) tested the assumption of cross-cultural equivalence of the scales of the MPQ. A multigroup common factor analysis yielded a satisfactory level of scale equivalence for the Dutch, Italian, German, and Australian-English versions of the MPQ.
It makes perfect sense to use the MPQ because FL is by definition a first step in developing not only multilingualism but also a certain degree of multiculturalism. We used the recent 40-item version of the questionnaire, the Multicultural Personality Questionnaire-Short Form (MPQ-SF; van der Zee et al., 2013). An Arabic version of the original English form of the MPQ-SF was created in collaboration with Jan Pieter van Oudenhoven, one of the questionnaire's authors. The MPQ-SF utilises a 5-point Likert scale, ranging from 1 (totally not applicable) to 5 (completely applicable).
Each dimension was composed of eight items with possible response values ranging from 1 to 5, and so total scores on each factor ranged from 8 (a 1 on each of the factor's 8 items after re-coding the 17 negatively-oriented items) to 40 (a score of 5 on each of the factor's 8 items after re-coding of negatively-oriented items). High scores indicate a high level of the trait (e.g., high Emotional Stability). Cronbach's alpha analyses revealed that the five scales have sufficient internal consistency, as shown in Table 1. Dörnyei and Taguchi (2009) argue that applied linguists should aim "at reliability coefficients in excess of .70; if the Cronbach alpha of a scale does not reach .60 this should sound alarm bells" (p. 95). The first version of the questionnaire was pilot-tested with 12 participants. This led to the reformulation of some items. The research design and questionnaire obtained approval from the Ethics Committee of the School of Social Sciences, History and Philosophy at Birkbeck College, University of London, UK.

Statistical considerations
The sample size is sufficient for multiple regression analyses (Tabachnick & Fidell, 2001). Although the data are normally distributed, we performed a number of supplementary analyses to make sure no major assumptions for multiple regression had been violated. Collinearity diagnostics showed that tolerance values (1 -R 2 ) were sufficiently above zero, which means that this assumption has not been violated. Residual scatterplots and normal probability plots showed no deviations from normality (Tabachnik & Fidell, 2001).

Results
Pearson correlation analyses revealed that FLCA was significantly and negatively correlated with four personality traits: Cultural Empathy, Social Initiative, Openmindedness and Emotional Stability, as shown in Table 2    -.13* -.09 -.38*** -.36*** -.46*** Notes. *p < .05, **p < .01, ***p < .0001 A significant negative Pearson correlation was found between self-rated oral proficiency in English and FLCA, as shown in Table 3 and Figure 3. A slightly weaker negative relationship was found between frequency of use of English and FLCA (Table 3). The number of languages known by participants was not linked to their FLCA (Table 3). Age was negatively linked to FLCA (Table 3).

Figure 3
The link between self-perceived oral proficiency and FLCA Table 3 Pearson correlation analyses between FLCA, knowledge and use of English, multilingualism, and age Self-rated oral proficiency in English Frequency of use of English

Number of languages known
Age FLCA -.57*** -.36*** -.08 -.15** Notes. *p < .05, **p < .01, ***p < .0001 An independent t test showed that there were no gender differences for FLCA scores (t(348) = 0.091, p = ns). A one-way ANOVA revealed that level of education was unrelated to FLCA (F = 1.4, df = 3, p = ns). Multiple stepwise linear regressions were performed for FLCA in order to determine the amount of unique variance explained by the variables that correlated significantly with FLCA, namely Cultural Empathy, Social Initiative, Openmindeness, Emotional Stability, self-rated oral proficiency in English, frequency of use of English and age.

Discussion
To summarise, the findings reported here allow us to partially reject Hypothesis 1 on the relationship between FLCA and sociobiographical variables: Age was a small but significant predictor of FLCA, with older participants reporting lower levels of FLCA. However, gender and education level were unrelated to FLCA. Hypothesis 2 is confirmed: Self-perceived proficiency turned out to be the strongest predictor of FLCA, and frequency of use of English was also a small but significant predictor of FLCA. In other words, participants who felt proficient in oral English and used the language more frequently were much less likely to suffer from FLCA in English. Hypothesis 3 is rejected as the knowledge of more languages was not linked to lower levels of FLCA. Hypothesis 4 is partially confirmed: Participants who were highly emotionally stable were least likely to suffer from FLCA in English. Emotional Stability may serve as a protective factor against anxiety in the face of foreign language classroom stressors. Social Initiative was also a significant predictor of FLCA, with participants scoring high on this dimension suffering less from FLCA. Two other dimensions, Cultural Empathy and Openmindedness showed the expected negative relationship in the correlation analyses, but turned out not to be significant predictors in the multiple regression analyses.
The weak relationships between sociobiographical variables and FLCA were not unexpected. Previous findings indicated scattered effects in different directions (Arnaiz & Guillén, 2012). This is possibly due to the type and age of FL learners. Age differences seem more salient when comparing younger populations (MacIntyre et al., 2002). When a more heterogeneous population is considered in terms of age, differences tend to shift in the other direction, with older participants reporting lower levels of FLA (Dewaele, in press;Dewaele et al., 2008). We were, however, surprised that the number of languages known by participants was not linked to FLCA as previous research has shown that the knowledge of more languages is linked to lower levels of FLA/FLCA across languages (Dewaele, 2010a(Dewaele, , 2010bDewaele et al., 2008;Thompson & Lee, 2013). The main difference between samples used in previous studies and the present participants is that they all have Arabic as an L1. It is unclear how this could somehow be linked to a lack of relationship with multilingualism.
The finding that self-perceived proficiency appeared as the strongest predictor of FLCA is not entirely unexpected (Arnaiz & Guillén, 2012;Liu & Chen, 2013;MacIntyre & Gardner, 1994;Sparks & Ganschow, 2007), but the amount of variance explained (over a third) is much higher than in Dewaele and Shan Ip (2013), where it amounted to a mere 6.6% of variance. The frequency of use of English explained a modest additional 1.5% of unique variance in the present study. This demonstrates that self-perceived proficiency and frequency of use behave as separate dimensions, despite being correlated.
The personality traits Emotional Stability and Social Initiative explained a further 18.5% of variance in FLCA, which is in line with the findings for Neuroticism and Extraversion in Dewaele (2013a). It thus seems that FL learners who are by nature inclined to "actively approach social situations and demonstrate initiative in these interactions" (van der Zee et al., 2013, p. 118), especially in the FL classroom, are less likely to suffer from FLCA. Moreover, students who can stay calm under "novel and stressful conditions" (van der Zee et al., 2013, p. 118), which, transposed to an FL classroom means having to express themselves in a language they have not yet entirely mastered with peers and teacher listening and judging their performance, will also suffer less from FLCA. The finding that Openmindedness and Cultural Empathy were significantly negatively correlated with FLCA was not unexpected, as Openmindedness shares characteristics with Tolerance of Ambiguity (Dewaele & Shan Ip, 2013) and Cultural Empathy shares similarities with Emotional Intelligence, which is linked to FLA (Dewaele et al., 2008, Shao, Yu & Ji, 2013. Van der Zee et al. (2013) report a significant correlation between Cultural Empathy and Emotional Intelligence and between Openmindedness and Emotional Intelligence (both p < .01) (p. 123). However, these two dimensions were not found to explain unique variance in the regression analyses.
An important aspect of the present study is the use of the AFLAQ and the Arabic version of the MPQ-SF. Al-Saraj (2011, 2014 has argued that a mere translation of instruments might result in culturally inappropriate or irrelevant instruments, hence the need not just to translate but also to adapt questionnaires with a specific population in mind. The AFLAQ and the Arabic version of the MPQ-SF were specifically tailored towards Arabic learners and users of English. We explained earlier that our aim was not to compare the results of our participants with other groups but rather to try and find relatively stable relationships between independent and dependent variables within this relatively homogeneous group of Arabic learners. Further research could explore potential differences between specific nationalities within the group. We mentioned earlier that Dörnyei and Ryan (2015) felt of the modular individual differences approach that "while yielding valuable insights, this seemingly logical approach proved to be an illusion as a whole" (p. 21). We respectfully beg to differ. While we do not deny the fact that systems behave dynamically and that nonlinearity makes longitudinal SLA research particularly challenging, we feel that a focus on specific variables, within a specific, relatively homogeneous leaner sample, using a conventional statistical approach can produce satisfying and original results, thus expanding the frontiers of the so-called "charted territories." We do not deny that other unseen variables may have played a role in the background and that complex interactions exist between all variables, but with this caveat in mind, we feel that the relationships we have uncovered are essential in understanding the bigger picture of FLA/FLCA. It seems to us that for the health of our discipline, it is crucial that researchers be encouraged to sail everywhere, including against the tide into charted waters.

Conclusion
The present study investigated the link between self-perceived oral proficiency in English (a FL), frequency of use of English, sociobiographical variables and five multicultural personality traits on FLCA of Arabic learners of English using instruments in Arabic (AFLAQ; Al-Saraj, 2011, 2014 and an Arabic version of the MPQ-SF; van der Zee et al., 2013). The results suggest that participants who felt more proficient in oral English and used the language frequently were significantly less likely to suffer from FLCA in English. These two linguistic variables explained over a third of variance in FLCA. Participants' personality was also linked to their levels of FLCA with two personality traits, Emotional Stability and Social Initiative, explaining a further 18.5% of variance in FLCA. These two personality traits have been found to have similar relationships in previous research (Dewaele, 2013a). Age was the only sociobiographical variable to explain a further 0.7% of variance. Number of languages known, gender and education level were unrelated to FLCA. It thus seems that the English FLCA of our Arab learners is predicted mainly by proficiency and frequency of use of English, but that two psychological traits also determine levels of FLCA.