Multilingualism and foreign language anxiety: the case of Saudi EFL learners

The present study focuses on the link between foreign language anxiety (FLA), self-perceived proficiency, and multilingualism in the under-explored English as a Foreign Language (EFL) context of Saudi Arabia. Ninety-six Arabic undergraduate college-level EFL students (56 males, 40 females) answered the Arabic version of the Foreign Language Classroom Anxiety Scale (FLCAS – Horwitz, Horwitz, & Cope, 1986). The analyses revealed that Saudi multilinguals suffered from low to moderate levels of FLA with female participants experiencing more anxiety than their male counterparts. Multiple regression analyses revealed that gender and self-perceived proficiency explained over a quarter of variance in FLA. Furthermore, the study did not find any role of experience abroad in predicting FLA.


Introduction
Research has indicated that learners suffer from anxiety in foreign language classes more than in other classes (Horwitz et al., 1986;MacIntyre & Gardner, 1994). MacIntyre and Gardner (1994) defined the construct of foreign language anxiety (FLA) as "the feeling of tension and apprehension specifically associated with second language contexts including speaking, listening, and learning" (p.284). Horwitz et al. (1986) defined FLA as "a distinct complex of self-perceptions, beliefs, feelings, and behaviors related to classroom language learning arising from the uniqueness of the language learning process" (p. 128). FLA could be one of the affective factors that most severely inhibits the second language learning process (Arnold & Brown, 1999;Hu & Wang, 2014) and language performance (e.g., Horwitz, 2010;Liu & Jackson, 2008;MacIntyre & Gregersen, 2012;Gkonou, Daubney, & Dewaele, 2017). Therefore, a large number of studies have been conducted focusing on the impact of anxiety on second/foreign language learners.
While the impact of FLA on bilingual learners has been researched extensively (e.g., Arnaiz & Guillén, 2012;Bailey, 1983;Horwitz et al., 1986;MacIntyre & Gardner, 1994;Ohata, 2005;Pappamihiel, 2002;Williams & Andrade, 2008;Young, 1991;Zhang & Zhong, 2012) including the English as a foreign language (EFL) context in Saudi Arabia (e.g. Alrabai, 2014;Alshahrani, 2016;Ashahrani & Alshahrani, 2015;Al-Saraj, 2013Hamouda, 2013), the relationship between anxiety, self-perceived proficiency, and multilingualism -defined broadly as "any experience with a third language" (Thompson & Khawaja, 2015, p.1) -has not been researched enough in second language learning (Thompson & Lee, 2012) especially in the Arabian Gulf region. Therefore, this study fills the gap in the current second/foreign language learning literature by examining FLA with multilingual students learning EFL in the under-represented context of Saudi Arabia. Specifically, this study first measures the FLA levels of the multilingual participants. Second, it attempts to increase current understanding of the factors precipitating FLA by determining whether background variables (gender, experience abroad, and English self-perceived proficiency) influence anxiety among multilinguals. MacIntyre and Gardner (1991) argue that FLA is unique and specific to the foreign language context. Therefore, it should be distinguished from other types of anxiety such as general anxiety and state anxiety. The complexity of FLA has made it difficult for researchers to reach a consensus about the exact causes of this phenomenon. Some researchers argue that lack of adequate language skills provoke anxiety (Sparks & Ganschow, 2007;Sparks, Ganschow & Javorsky, 2000). MacIntyre and Gardner (1994), however, attribute FLA to factors other than linguistic abilities. In their experiment with French language learners, they found that students who were filmed while completing vocabulary activities exhibited higher levels of anxiety than their peers who were not exposed to camera filming. The exposure to video camera negatively affected their performance as participants started forgetting vocabulary that they had learned.

Foreign language anxiety
Another factor that may cause language anxiety has to do with the role played by the teacher in the classroom. Abu-Rabia (2004) explored the relationship between teachers' attitudes and EFL students' language anxiety. He concluded that teachers' supportive attitude helped diminish students' levels of anxiety. Similarly, Ewald (2007) reported that teachers' supportive attitudes helped alleviate students' anxiety. The teacher's use of unfamiliar teaching methods can also impact students' anxiety (e.g., Brantmeier, 2005). This finding is especially true in the case of expatriate teachers who may use teaching practices not expected among students in the local culture (Al-Saraj, 2014). Yan and Horwitz (2008) observed that foreign teachers occasionally shocked Chinese EFL students by establishing an informal student-teacher relationship, which is a deviation from the cultural norms in China. This deviation caused anxiety among Chinese students. In her study that involved Saudi female college students, Al-Saraj (2013) cited teacher-student interactions and teacher behavior as some of the major sources of anxiety among all participants.

Variables associated with FLA
Researchers have identified several variables linked to FLA including age, gender, experience abroad, and language proficiency.

FLA and age
The studies that explored the correlation between age and FLA have not yielded consistent results. Donovan and MacIntyre (2005) found that Anglo-Canadian university students studying French had higher levels of FLA than junior high and high school students. The same results were reported by Dewaele (2007a) in his study involving adult language learners. He found that younger learners had lower levels of FLA across languages. However, other studies found opposite results. Dewaele et al. (2008), for example, found significant negative correlations between age of adult multilinguals and their FLA scores. Older participants experienced lower levels of FLA. MacIntyre, Baker, Clément, and Donovan (2002) also found that younger (grade 8) Anglo-Canadians reported higher levels of anxiety than grade 9 pupils at a junior high school in French immersion programs. In a more recent study, Dewaele and MacIntyre (2014) reported that teenage participants suffered more from FLA than participants in their twenties.

FLA and gender
Previous research that examined the relationship between FLA and gender has yielded inconclusive results. Some studies reported male learners experiencing higher levels of anxiety than their female peers (Campbell & Shaw, 1994;Kitano, 2001). MacIntyre, Baker, Clément, and Donovan (2002), for example, reported that male 7th to 9th grade students in a junior high school enrolled in a French immersion program have maintained the same levels of anxiety, while their female counterparts witnessed a decrease in anxiety levels from 8th to 9th grade. In a more recent study involving 64 Indonesian learners of English in Indonesia and Australia, Hasan and Fatimah (2014) reported that male students exhibited more anxiety than their female peers. These results contradict the findings reported by other studies which found that females tended to suffer from higher levels of FLA (e.g., Arnaiz & Guillén, 2012;Donovan & Maclntyre, 2005;Elkhafaifi, 2005;Machida, 2001;Park & French, 2013). In the South Korean context, Park and French (2013) cite sociocultural factors as the main reason for higher levels of FLA among university female learners. They aver that in a male-dominated society where females are expected to take a submissive role, females tend experience stress when asked to express themselves. However, some studies did not find significant gender differences in FLA (e.g., Aida, 1994;Dewaele & Ip, 2013;Dewaele, Petrides, & Furnham, 2008;Matsuda & Gobel, 2004). In a study involving English-major students in, Matsuda and Gobel (2004) report that male and female university students majoring in English in Japan exhibited similar levels of FLA. In a more recent study, Bensalem (2017) reported no correlation between gender and FLA among college-level EFL students in Tunisia. Thompson and Lee (2012) argue that experience abroad refers to the involvement in any formal or informal exchange program regardless of its duration. There is a dearth of research regarding FLA and its relationship with experience abroad. The very few studies that examined the correlation between FLA and experience abroad found that experience abroad decreases the level of FLA (e.g., Allen & Herron, 2003;Coleman, 1997;Shapson, Kaufman, & Day, 1981;Thompson & Lee, 2012). Thompson and Lee (2012) contend, however, that experience abroad may not play any role in decreasing FLA unless the learner reaches a certain level of proficiency in the target language. Allen and Herron (2003) posit that the relationship between study abroad and affective variables, such as anxiety, is complex. They maintain that research has not reached conclusive results regarding the long term impact of study abroad on affective variables.

FLA, multilingualism, and proficiency
In this study, multilingualism refers to "any experience with a third language" (Thompson & Khawaja, 2015, p.1). There has been little research to examine the relationship between FLA, proficiency, and multilingualism (Thompson & Lee, 2012). The few studies that examined FLA among multilinguals, have reported that multilinguals tend to experience lower levels of FLA (Dewaele, 2007a(Dewaele, , 2010Dewaele et al., 2008;Thompson & Lee, 2012;Thompson & Khawaja, 2015). Santos, Cenoz, and Gorter (2015) and Cenoz (2013) argue that multilinguals experience less anxiety than bilinguals because they are more experienced language learners. Within the same vein, Kemp (2001) asserts that learners who know more than two languages tend to develop more grammar learning strategies, and thus become better language learners. Consequently, they will experience reduced levels of anxiety. One of the seminal studies that reported the effect of additional languages on FLA was conducted by Dewaele (2007a). He found that trilinguals and quadrilinguals tended to experience lower levels of anxiety in their second language than bilinguals. However, when speaking their third language, they demonstrated the same levels of anxiety. These outcomes were attributed to the fact that "trilinguals and quadrilinguals have become better communicators as a result of their multilingualism and that their self-confidence, as well as their self-perceived competence has grown as a result" (p. 404).
Other studies conducted in different linguistic contexts provided evidence for the positive effect of multilingualism on reducing anxiety. Thompson and Lee (2012), for instance, reported that students learning English in Korean universities who exhibited higher levels of proficiency in a third language (French or Chinese) had lower levels of anxiety in English than students with lower levels of third language proficiency. These results were echoed by Thompson and Khawaja's (2015) study which reported that university level Turkish multilingual learners of English had lower levels of anxiety than their bilingual counterparts.
In a more recent study, Phongsa, Mohamed Ismail and Low (2017) compared the foreign language anxiety experienced by monolingual and bilingual tertiary students who were learning EFL in the Lao People's Democratic Republic. Results showed that bilinguals felt more comfortable when interacting with native speakers of English, and witnessed an increase in their self-confidence in using English. The researchers suggested that multilingualism could be useful in reducing FLA.

FLA and self-perceived proficiency
A few studies established a relationship between FLA among multilinguals, and self-perceived proficiency. In a study involving secondary school and university students learning French, Donovan and MacIntyre (2005) found a negative correlation between self-perceived proficiency and FLA. Participants with higher self-perceived proficiency exhibited lower levels of FLA. The researchers concluded that self-perceived proficiency was a significant predictor of FLA. The same findings were corroborated by Dewaele (2007b) who reported that self-perceived oral proficiency in the learners' second, third, fourth and fifth language predicted their levels of FLA.
Even though research has shown high correlation between self-report measures of proficiency and linguistic measures of proficiency (MacIntyre et al., 1997), some researchers have doubted the validity of self-perceived competence measures (DeKeyser, 2006). There is concern about the tendency of some learners to have a biased perception of their real proficiency due to anxiety. Maclntyre et al. (1997) aver that language learners who demonstrate high anxiety could underestimate their level of proficiency. Conversely, learners who have lower levels of anxiety could overestimate their true proficiency of the target language. The relationship between anxiety, self-perceived proficiency, and multilingualism is an underresearched area in second language learning (Thompson & Lee, 2012). To the researcher's best knowledge, no study has explored the issue of anxiety among Saudi multilinguals. Therefore, this study is an attempt to investigate the correlation between anxiety, self-perceived proficiency, and multilingualism among Saudi multilinguals learning EFL. It also seeks to increase our understanding of the impact of background variables (gender, experience abroad, and English self-perceived proficiency) on students' anxiety levels.
The following research questions were addressed: 1. What is the level of FLA among Saudi multilinguals? Do male and female students experience similar anxiety levels? 2. To what extent do sociobiographical variables (gender, experience abroad, and English selfperceived proficiency) predict FLA?  Table 1 provides basic information about the participants. Their ages ranged from 19 to 29 years (M = 22.11, SD = 2.19). Fifty-six (58%) of the participants were male and 40 (42%) were female. All participants were majoring in languages and translation, and had studied French during their freshman and sophomore year at college. Participants self-rated their proficiency in English on a scale from 1 to 10 for listening, speaking, reading and writing (see Table 2). The same scale was used as in previous studies (e.g., Dewaele et al., 2008;Santos, Cenoz, & Gorter, 2015;Thompson & Lee, 2012).

Background questionnaire
This questionnaire was designed to elicit participants' information about age, gender, course level, major, experience of a third language and whether they had a chance to study abroad. It included selfratings of proficiency in English.

Anxiety scale
An adapted version of the Foreign Language Classroom Anxiety Scale (FLCAS) was used to measure anxiety levels among participants, replacing the phrase "foreign language" with "English language" to focus on this aspect of students' emotions. FLCAS is a self-reported measure of learners' anxiety in the foreign language classroom designed by Horwitz et al. (1986). It consists of 33 statements. Each item on the scale is rated on a five-point Likert scale ranging from 1 (strongly agree) to 5 (strongly disagree). The mean scores in the FLCAS range from 33 to 165, with lower scores indicating lower anxiety while higher scores indicate higher anxiety. Twenty-four of the items are positively worded; nine are negatively worded. The scale had a high rate of reliability with an alpha coefficient of .93 (Horwitz et al., 1986). In the present investigation, the FLCAS had a Cronbach's alpha coefficient reliability index of .88. The survey measure was pilot-tested prior to the onset of this study experiment.

Procedure
Data were collected by means of an online questionnaire. The measure was pilot tested among 24 multilingual Saudi students who met the participation criteria. After getting the permission of the Dean of Faculty in each of the two institutions, participants who agreed to answer the survey were briefed about the purpose of the study. They were assured of the confidentiality of all the background information they shared. Completion of the questionnaire was anonymous. Participants had the option to choose between Arabic and English as the survey language. The decision to include an Arabic version was due to the fact that the target participants were not accustomed to responding to surveys in English. The researcher thought that an Arabic version would increase the participation rate as students were more comfortable with Arabic. The survey took between 10 and 15 minutes to complete.

Data analysis
In order to analyze the collected data, descriptive statistics (i.e., means and standard deviations) were used to summarize participants' responses. The eight items (2,5,11,14,18,22,28 and 32) that were negatively worded were reverse-coded so that a high score indicated high anxiety. An independent ttest was conducted to examine any statistically significant difference between the FLCAS scores of male and female students. Furthermore, in order to assess the effect of the learners' background variables simultaneously and determine their relative contribution to the prediction of FLA, multiple regression analysis was performed of FLA.

Findings
This section presents the results for each of the research questions.

Level of FLA among Saudi multilinguals (male and female)
To measure the level of FLA among multilinguals, means and standard deviations for participants' responses to each FLCAS item were calculated (see Table 3). The survey uses a five-point Likert scale with a possible score ranging between 33 to 165. The mean language anxiety score for the 96 participants was 88.22 (SD = 20.36). As displayed in Table 4, the range of scores in the present study was 51-132. Following Arnaiz and Guillén's (2012) scale, participants had three levels of anxiety. Table  4 summarizes the different levels of anxiety experienced by participants. The overwhelming majority of students (88.54%) experienced low to average levels of FLA. Approximately 11% of students suffered from a high level of anxiety. As reported in Table 5, results of the two-independent samples t-test show that mean FLCAS scores differ between males (M = 79.98, SD = 15.33, n = 56) and females (M = 99.75, SD = 21.08, n = 40) at the .05 level of significance (t = 5.05, df = 67, p < .05). On average, females tend to have higher levels of FLA than males.

Socio-biographical variables and FLA
In order to examine the simultaneous effect of the background variables (gender, experience abroad, and English self-perceived proficiency) on the participants' FLA level, multiple regression analysis was conducted. Results (see Table 6) show that the presented model is significant as there was a significant relationship between the background variables and anxiety (p = .00; R² = 0.279; R = 0.528). The R value for the regression model indicates a medium linear relationship between the three learner variables and FLA. As shown in Table 6, R² was 0.28, which means gender and English self-perceived proficiency combined to explain 28% of the variance in first year English majors' anxiety. According to Cohen's (1988) criteria for assessing the predictive power of a set of independent variables, this indicates a medium effect size.
Two of the background variables, gender and English self-perceived proficiency, explained 27.9 % of the variance of anxiety in students. Among the predictor variables, gender had the highest relative impact on anxiety with t value of 8.014, while English self-perceived proficiency had the lowest relative impact with t value of -.88. Experience abroad was not a predictor of FLA.

Discussion
This section is devoted to the discussion of the results of each of the research questions. The first research question examined the anxiety levels experienced by male and female Saudi multilinguals.
The results show that overall, this sample of multilinguals exhibited low to average levels of FLA. These findings lend further support to previous research that examined the impact of FLA on EFL multilinguals and reported similar levels of FLA (Arnaiz & Guillén, 2012;Bensalem, 2017;Phongsa et al., 2017). All these studies used FLCAS, which may explain the similarity of their findings. However, in a recent study involving bilingual and multilingual Turkish learners, Thompson and Khawaja (2015) reported higher levels of FLA.
Furthermore, the current study provides empirical evidence that the female participants experienced higher levels of FLA than their male counterparts. These findings are aligned with the outcomes reported by previous research (e.g., Abu-Rabia, 2004;Arnaiz & Guillén, 2012;Clarck & Trafford, 1996;Dewaele, 2007a;Donovan & MacIntyre, 2005;Park & French, 2013). In the case of Korean EFL students, Park and French (2013) attribute the higher levels of FLA among female learners to sociocultural factors. Park and French (2013) assert that females in Korea tend to feel stressed when asked to express their opinions because they are expected by their society to embrace a submissive role in a society that is dominated by males. However, the current study's findings are not consistent with a recent study that involved Arab multilingual EFL learners conducted by Bensalem (2017), who did not find any association between gender and FLA among multilingual EFL students in Tunisia. The author attributes these results to the special status of females in Tunisia. Tunisian women have equal opportunities to compete with jobs from diverse sectors. The empowerment of women has boosted their self-confidence and thus reflects on their equal level of FLA with their male peers (Bensalem, 2017). A plausible explanation for the outcomes of the current study is provided by Song (2018) who argues that Saudi females' deference to keep an ideal image of womanhood in the eyes of Saudi culture makes shyness part of their gender identity. This explains their tendency to not participate in the classroom and experiencing higher levels of anxiety. Obviously, teachers play a key role in reducing anxiety as they need to create a friendly and supportive learning environment in which students are encouraged to be actively involved in learning.
The second research question aims to determine which of the learners' background variables (gender, study abroad experience, and English self-perceived proficiency) are predictors of FLA. Gender and English self-perceived proficiency explained 28% of the variance of anxiety among students. Gender has been found to be the highest predictor for FLA. The significant role that gender plays in FLA is aligned with a number of previous studies (e.g., Arnaiz & Guillén, 2012;Stephenson, 2007). However, the current study contradicts the findings in a recent experiment involving Arab EFL students where gender was not a significant predictor of FLA (Bensalem, 2017).
The current study lends support to the role played by self-perceived proficiency in predicting FLA as it was found in the literature (Arnaiz & Guillén, 2012;Dewaele, 2007b;Dewaele & Al Saraj, 2015;Sparks & Ganschow, 2007;Tóth, 2007). Learners who highly rate their proficiency in the target language tend to experience lower levels of FLA.
One finding of interest in this study is that experience abroad is not a predictor of FLA among Saudi multilinguals. A plausible reason for this outcome is that the length of experience abroad may not have been long enough to have a significant impact on levels of FLA. Most of the participants did not stay abroad longer than a month. There may be other affective variables involved as Allen and Herron (2003) argued. Another possible explanation is provided by Thompson and Lee (2012), who argue that learners who have experience abroad may not witness lower levels of FLA unless they reach a certain level of proficiency. Most of the participants of this study had an intermediate level of English. This study could have yielded a different outcome if students had a higher level of proficiency.

Conclusion, limitations and future research
This study explored FLA among multilingual students learning EFL in Saudi Arabia, which is an underexplored context. The findings showed participants exhibited low to average levels of FLA. Female participants experienced higher levels of FLA than their male peers. The study found that gender and self-perceived proficiency are predictors of FLA since they accounted for about a quarter of the variance of anxiety among participants. More proficient learners experienced less anxiety than their less proficient peers. Similarly, female multilinguals suffered higher levels of anxiety than their male peers. Gender and self-perceived proficiency have been found to have the same impact as in previous research (Arnaiz & Guillén, 2012). Additionally, the finding in this study did not find any role played by experience abroad in predicting students' levels of FLA.
The study has limitations that must be addressed. Firstly, this study had access to a relatively small number of participants so the findings should be interpreted with caution. A larger sample size may have yielded different outcomes. Secondly, the data were collected using only quantitative analysis.
Adding qualitative methods such as interviews could have yielded more insights regarding students' anxiety. Furthermore, the study did not measure the levels of proficiency in French among participants and whether they had a pleasant experience learning the language. Such information could help identify potential factors causing FLA. Future research should examine how multilingual students' anxiety levels may differ based on their proficiency in their third language. Studies involving learners with knowledge of a fourth language could help educators to develop a better understanding of the phenomenon of anxiety among Saudi multilinguals.