foreign language listening anxiety and listening performance: conceptualizations and causal...

14
Foreign language listening anxiety and listening performance: Conceptualizations and causal relationships Xian Zhang a,b, * a The Pennsylvania State University, USA b Guangdong University of Foreign Studies, People’s Republic of China Received 22 December 2011; revised 3 January 2013; accepted 3 January 2013 Available online 4 February 2013 Abstract This study used structural equation modeling to explore the possible causal relations between foreign language (English) lis- tening anxiety and English listening performance. Three hundred participants learning English as a foreign language (FL) com- pleted the foreign language listening anxiety scale (FLLAS) and IELTS test twice with an interval of three and a half months. Exploratory causal models indicated that the FL listening anxiety could affect FL listening performance, but FL listening per- formance did not appear to affect FL listening anxiety systematically. The causal relationships may be attributed to the situation- specific nature of FL listening anxiety. Ó 2013 Elsevier Ltd. All rights reserved. Keywords: Anxiety; Listening; Structural equation modeling; Foreign language anxiety 1. Introduction Individual difference variables such as aptitude, motivation, emotion and learning strategies have been recognized as influential factors in foreign language (FL) learning (Do ¨ rnyei and Skehan, 2003). For example, emotion 1 can have facilitating or debilitating effects on learning (Dirkx, 2001) and can affect one’s motivation to learn (Woldkowski, 1999). Among all the individual difference variables, anxiety appears to be especially important (Arnold, 1999; Gregersen, 2005; MacIntyre, 1995a; Oxford, 1999). Relationships between FL anxiety and FL performance have been of interest for decades. Scovel (1978) reviewed studies from the 1970s and found mixed results in that positive, negative, or non-significant relationships were all observed. He noted that the inconsistency may be attributable to variation in the types of anxiety involved. In conceptualizing and measuring FL anxiety, Horwitz et al. (1986) developed the Foreign Language Classroom Anxiety Scale (FLCAS), which ostensibly taps communication appre- hension, test anxiety, and fear of negative evaluation. Horwitz et al. argued that FL anxiety has unique components limited to the FL context that make it different from other types of anxiety (MacIntyre and Gardner, 1991a). In short, FL anxiety is conceptualized as situation-specific. * The Pennsylvania State University, 801 Southgate Dr. #A14, State College, PA 16801, USA. Tel.: þ1 81 47532838. E-mail address: [email protected]. 1 There are two types of emotion: positive emotion (engagement and excitement) and negative emotion (e.g. boredom and isolation). 0346-251X/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.system.2013.01.004 Available online at www.sciencedirect.com System 41 (2013) 164e177 www.elsevier.com/locate/system

Upload: xian

Post on 08-Dec-2016

217 views

Category:

Documents


3 download

TRANSCRIPT

Page 1: Foreign language listening anxiety and listening performance: Conceptualizations and causal relationships

Available online at www.sciencedirect.com

System 41 (2013) 164e177www.elsevier.com/locate/system

Foreign language listening anxiety and listening performance:Conceptualizations and causal relationships

Xian Zhang a,b,*

aThe Pennsylvania State University, USAbGuangdong University of Foreign Studies, People’s Republic of China

Received 22 December 2011; revised 3 January 2013; accepted 3 January 2013

Available online 4 February 2013

Abstract

This study used structural equation modeling to explore the possible causal relations between foreign language (English) lis-tening anxiety and English listening performance. Three hundred participants learning English as a foreign language (FL) com-pleted the foreign language listening anxiety scale (FLLAS) and IELTS test twice with an interval of three and a half months.Exploratory causal models indicated that the FL listening anxiety could affect FL listening performance, but FL listening per-formance did not appear to affect FL listening anxiety systematically. The causal relationships may be attributed to the situation-specific nature of FL listening anxiety.� 2013 Elsevier Ltd. All rights reserved.

Keywords: Anxiety; Listening; Structural equation modeling; Foreign language anxiety

1. Introduction

Individual difference variables such as aptitude, motivation, emotion and learning strategies have been recognizedas influential factors in foreign language (FL) learning (Dornyei and Skehan, 2003). For example, emotion1 can havefacilitating or debilitating effects on learning (Dirkx, 2001) and can affect one’s motivation to learn (Woldkowski,1999). Among all the individual difference variables, anxiety appears to be especially important (Arnold, 1999;Gregersen, 2005; MacIntyre, 1995a; Oxford, 1999). Relationships between FL anxiety and FL performance have beenof interest for decades. Scovel (1978) reviewed studies from the 1970s and found mixed results in that positive,negative, or non-significant relationships were all observed. He noted that the inconsistency may be attributable tovariation in the types of anxiety involved. In conceptualizing and measuring FL anxiety, Horwitz et al. (1986)developed the Foreign Language Classroom Anxiety Scale (FLCAS), which ostensibly taps communication appre-hension, test anxiety, and fear of negative evaluation. Horwitz et al. argued that FL anxiety has unique componentslimited to the FL context that make it different from other types of anxiety (MacIntyre and Gardner, 1991a). In short,FL anxiety is conceptualized as situation-specific.

* The Pennsylvania State University, 801 Southgate Dr. #A14, State College, PA 16801, USA. Tel.: þ1 81 47532838.

E-mail address: [email protected] There are two types of emotion: positive emotion (engagement and excitement) and negative emotion (e.g. boredom and isolation).

0346-251X/$ - see front matter � 2013 Elsevier Ltd. All rights reserved.

http://dx.doi.org/10.1016/j.system.2013.01.004

Page 2: Foreign language listening anxiety and listening performance: Conceptualizations and causal relationships

165X. Zhang / System 41 (2013) 164e177

To understand this conceptualization of FL anxiety, it is helpful to distinguish three general categories: trait, state,and situation-specific anxiety. Trait anxiety is the continual tendency to feel anxious under a variety of circumstances(Spielberger, 1983). People with high trait anxiety are very likely to get anxious in all kinds of situations. Situation-specific anxiety is different from trait anxiety, in that it focuses on one single situation (MacIntyre and Gardner,1991a). It is generally stable over time but may vary across situations (stability of a trait is a measure of how thetrait may fluctuate and change over time; a stable trait suggests that the trait does not have strong fluctuation over timewhereas an unstable trait suggests that the trait can fluctuate and change a lot). For example, FL anxiety is situation-specific, referring to the possibility that negative emotional feelings may be aroused when learning and using an FL(MacIntyre and Gardner, 1989, 1991a). Unlike trait and situation-specific anxiety that refer to the tendency to becomeanxious (not an actual emotional feeling per se), state anxiety embodies actual feelings, such as nerves, discomfort,and uneasiness which usually surfaces at a specific point in time (Spielberger, 1983). Since state anxiety refers to theemotions, it can be unstable over time (MacIntyre and Gardner, 1991a), which makes it different from the more stabletrait/situation-specific anxiety.

State anxiety is by no means completely separable from situation-specific anxiety. When continuously experi-encing anxiety in a certain situation, state anxiety that is usually unstable can develop into situation-specific anxiety.MacIntyre and Gardner (1989) used FL anxiety as an example. A student can experience anxiety as a result of poorperformance in the FL classroom. However, when the student repeatedly experiences such state anxiety, it can besolidified into a situation-specific form, namely the FL anxiety. When FL anxiety develops, one is apt to experiencehigh levels of anxious feelings.

Identifying those with high level of anxiety and helping them to alleviate it may lead to better FL performance, inthat research over the past two decades has shown it to relate negatively to FL performance (Aida, 1994; Atashenehand Izadi, 2012; Hurd, 2007; Kitano, 2001; MacIntyre and Gardner, 1991a; Matsuda and Gobel, 2004; Phillips, 1992).Despite evidence of a negative correlation, the causal effects of FL anxiety and FL performance remain unclear(Cheng et al., 1999; Horwitz, 2001). The debate concerning whether FL anxiety is a cause of poor performance ormerely a consequence has long been in evidence. Sparks et al. hold the view that FL anxiety is merely the consequenceof low performance (Sparks and Ganschow, 1991, 1996; Sparks et al., 2000): students with high FL anxiety levels intheir body of research usually had low FL aptitude. Hence, Sparks et al. speculated that FL anxiety is only a con-sequence of cognitive deficits rather than a cause of poor performance. On the other hand, MacIntyre (1995b) andHorwitz (2000) argued that FL anxiety is a well documented phenomenon among FL learners and can be a source ofdiminished performance. MacIntyre (1995b) maintains that FL anxiety can interfere with the encoding, storage andretrieval processes, all of which can contribute to low performance. Horwitz (2000) further noted that many “advancedand successful” FL learners have claimed to have experienced high levels of anxiety when learning and using an FL.

The debates have yet to resolve the issue. Little evidence is available to help reach a conclusion about the nature ofthe relationship between FL anxiety and FL performance. A recent study by Yan and Horwitz (2008) may offer someclues. They used grounded-theory analysis to investigate how FL anxiety interacted with different variables to in-fluence language learning. Twenty-one ESL students were interviewed through a three-stage grounded-theory anal-ysis. Based on the students’ perception of and reflection on FL, a theoretical model was developed suggesting that FLclassroom anxiety could interact with motivation so as to affect FL performance. However, there was no direct ev-idence of the reverse, i.e. that FL performance could influence FL anxiety.

Yan and Horwitz’s (2008) study provided qualitative empirical evidence of a possible causal direction between FLanxiety and FL performance. The current study drew on the quantitative methods to investigate the causal relationship.But rather than investigating FL anxiety and FL performance in general, it appeared reasonable to focus on one of thefour language skills (listening, reading, writing, or speaking) and performance in the chosen skill. It is certainly thecase that performance in one type of language skill may be very different from overall performance in all four types oflanguage skills (Alderson et al., 1995). Moreover, anxiety specific to one type of language skill appears to differ fromgeneral FL anxiety (Cheng et al., 1999; Saito et al., 1999), which is often measured by Horwitz et al.’s (1986) FLCAS.

1.1. FL listening anxiety

The FLCAS is widely used to measure FL classroom anxiety, but it has a heavy load on speaking (Aida, 1994;Cheng et al., 1999). Relating FL anxiety measured by FLCAS to FL performance of language skills, such as listeningand reading, might not be adequate. Although speaking in an FL is often reported by learners to be most anxiety

Page 3: Foreign language listening anxiety and listening performance: Conceptualizations and causal relationships

166 X. Zhang / System 41 (2013) 164e177

provoking (Horwitz, 2001; Phillips, 1992; Young, 1990), FL anxiety can relate to a specific type of language skill suchas listening or reading (MacIntyre and Gardner, 1994b). Evidence supports the existence of language skill-specificanxiety, which reportedly has different constructs from the FLCAS (Cheng et al., 1999; Elkhafaifi, 2005; Saito et al.,1999). Cheng et al. (1999) maintained that measuring language skill-specific anxiety is meaningful in that theseanxiety constructs enable one to identify more precisely the source of anxiety in relation to proficiency in a particularlanguage skill. Therefore, to study the causal relations of FL anxiety and FL performance in a specific skill, measuringanxiety related to that specific skill should produce more precise relations.

Although listening to an FL can be highly anxiety provoking (Krashen, 1985), studies concerning FL listeninganxiety (FL listening anxiety refers to FL anxiety that is associated specifically with FL listening situations) are ratherlimited. Vogely (1998) attempted to identify the source of listening anxiety in the language classroom and noted thatFL listening anxiety (FLLA) appeared to be associated with the types of listening input, listening process (strategiesand time), and instructional factors, such as in-class practices and tests. Kim (2000) constructed a 33-item measure ofFL listening anxiety of 253 college EFL learners in Korea. Using both quantitative and qualitative methods, Kimuncovered a negative correlation between FL listening anxiety and FL listening performance. A more recent studytook place in the Arabic language classroom. On the basis of the FL reading anxiety scale that Saito et al. (1999)developed, Elkhafaifi (2005) constructed the FL listening anxiety scale (FLLAS) to assess FL listening anxiety of233 students learning Arabic in north American universities. The FLLAS, which has 20 items, had an estimatedinternal reliability of .96 (Cronbach’s alpha). Although FL listening anxiety shared a noticeable variance with Hor-witz’s FL classroom anxiety (Pearson r2 ¼ .44), the FLLAS had a different construct from the FLCAS. Moreover, FLlistening anxiety was negatively correlated with FL listening test scores. The correlation coefficient for FL listeninganxiety and Arabic listening test scores was higher than the coefficient for the FL classroom anxiety and Arabiclistening test scores.

1.2. FL listening performance as measured by IELTS test

Listening is the most basic communication skill and is also one of the most frequently used language skills incommunication activities (Rivers and Temperley, 1978, p. 62). It plays an important role in the language learningprocess and facilitates other language skills (Vandergrift, 1999). Given the importance and the fundamental role oflistening in language learning, the present study focuses on FL listening performance, as measured by listening testscores.

Justification for test scores as indicators of learner’s language ability requires a valid measure (Bachman andPalmer, 1996). The listening component of the IELTS test (The International English Language Testing System)which is used is to “assess the proficiency in English of candidates seeking entry to UK higher education” (Davies,2008, p. 93) served as the measure of the participants’ FL (English) listening proficiency. The test originated in 1989as a revision of the ELTS test (English Language Testing System). To revise the ELTS, a revision team performeda needs analysis (Alderson et al., 1995), in which content validation of the draft test was carried out (Alderson andClapham, 1992). The revised test, the IELTS, has four parts: listening, reading, speaking, and writing. The listen-ing component is “intended to have a useful predictive function to tell whether or not students would be able to copewith listening to lectures or instructions.” (Clapham and Alderson, 1997, p. 12) Clapham and Alderson noted that thelistening component is “of a suitable level of difficulty with a satisfactory level of reliability” (p. 12). In light of thefact that the listening protocol of the IELTS test has a good predictive power over candidates’ academic listeningproficiency and “offers a measure of language aptitude” (Davies, 2008, pp. 110e112), it is selected to measureEnglish listening proficiency.

1.3. Purpose of the study

Yan and Horwitz (2008) produced evidence of possible causal effects of FL anxiety on FL performance, but thefocus was speaking. Listening, on the other hand, is not well studied in this field. Elkhafaifi’s (2005) study focused onlistening and furthered our understanding of the relationship between FL listening anxiety and FL listening perfor-mance, but the determinant causal effects between the two factors have not yet been established. To close the gaps andmore clearly investigate the causal directions of FL listening anxiety and FL listening performance, the current study

Page 4: Foreign language listening anxiety and listening performance: Conceptualizations and causal relationships

167X. Zhang / System 41 (2013) 164e177

incorporated structural equation modeling to explore the potential causal relationship between FL listening anxietyand English listening performance. The following research questions served as the stimulus for the research:

(1) What are the underlining sub-components of FL listening anxiety?(2) Is there a causal relationship between FL listening anxiety and English listening performance?

2. Method

2.1. Participants

The participants consisted of 300 first year English majors at a university in China. Among them, 261 were femaleand 39 were male. Both the FLLAS and an IELTS listening test were administered to these participants twice. Everyparticipant had studied English for more than 6 years before participating in the current study. The first IELTS lis-tening test score was 15.4 points on average, an intermediate level of listening proficiency indicating that the par-ticipants were suitable for taking part in the current study.

2.2. Instruments

2.2.1. Foreign language listening anxiety scale (FLLAS)The FLLAS in the current study was a replication of the FLLAS by Elkhafaifi (2005). Instead of measuring FL

listening anxiety in Arabic classrooms, the version in this study tapped FL anxiety in English listening classrooms(Appendix A). Participants filled out the 20-item 5-point Likert scale FLLAS to indicate how they felt in response toeach statement. Response options ranged from “strongly disagree” to “strongly agree,” with 1 point for the former and5 points for the latter. Administration of the survey followed the guidelines from Dornyei (2003). To ensure that theparticipants could understand each item, the FLLAS was translated into Chinese by the author. Two English in-structors checked for accuracy of the translation and made appropriate changes to the translated version. A computerprogram was written to conduct the survey online. The computer version had been piloted with three first year collegestudents (not among the participants) and it was found that the computer version was reliable for conducting thesurvey online. The average time for completing the survey was approximately 5 min.

2.2.2. The IELTS listening testThe listening component of the IELTS test used in the current study was selected from the IELTS practical test,

which was an official IELTS test. It had four sessions with 40 questions, including 8 multiple choice questions and 32gap fillings. The first session was a recording of a conversation between a customer and a salesman in a car showroom.Session II was a talk given by an admissions officer from a UK university to prospective graduate students abroad.Session III was a talk between two students about a course they had taken. Session IV was a presentation given bya student about a project on household waste recycling. All instructions in the test were given in both English andChinese. Other parts of the tests were in English.

2.3. Procedure

In the second week of enrollment into the college (T1), the participants completed the FLLAS. To avoid ambiguity,the FLLAS conducted at this time has been identified as FLLAS (T1) in the following text. Right after completion ofthe FLLAS, participants took the IELTS listening test, named IELTS (T1) in the following text. Three and a halfmonths later (T2), the participants completed the FLLAS again (FLLAS (T2)) as well as the same IELTS listening test(IELTS (T2)). The procedure was the same in both interventions at T1 and T2.

The IELTS tests were marked according to the standard marking criteria provided for the official IELTS listeningtest. All tests were double marked by the author and another English instructor, who then compared their marks andchecked for scoring discrepancies. Cronbach’s alphas of the double markings were .93 and .94 for IELTS (T1), IELTS(T2), respectively. The author and the English instructor discussed the discrepancies and made any necessary cor-rections based on the consensus of the two markers.

Page 5: Foreign language listening anxiety and listening performance: Conceptualizations and causal relationships

168 X. Zhang / System 41 (2013) 164e177

2.4. Data analysis

Data analysis was conducted two stages. The first stage involved instrument analysis and validation: analyzing andvalidating the FLLAS and the IELTS listening test. The second stage involved constructing the causal models anddetermining the causal directions.

In the first stage, exploratory factor analysis (EFA) served to reveal the constructs that the FLLAS measures withemerging factors being subsequently submitted to confirmatory factor analysis (CFA) in order to test data fit and tovalidate the factor solution of the EFA.ARasch analysis2 was run on the IELTS test results for validation. Such a processin this stagewas important since valid instruments provided the foundation for the statistical analysis in the second stage.

The second stage entailed the use of structural equation models to determine the possibility of a causal relationshipbetween FL listening anxiety and the English listening performance. Structural equation modeling (SEM) rather thanmultiple regression was employed on the basis that SEM is superior in that it takes into account the measurementerrors of the independent variables and dependent variables, whereas multiple regression ignores measurement errorfor all independent variables. Ignoring measurement error can yield inaccurate results and suggest misleading con-clusions (Raykov and Marcoulides, 2000).

3. Results

3.1. Construct validation: the FLLAS

Data for FLLAS (T1) were first submitted to KaisereMeyereOlkin measure to verify sampling adequacy. TheFLLAS was initially submitted to principal component analysis. The oblique rotation (direct oblimin, with Kaisernormalization), which allow for correlation between factors, was employed for factor rotation. Both the Kaiser’s(1960) criterion of eigenvalue greater than 1 and the scree test (Cattell, 1966) were taken into account in determin-ing number of factors to be extracted. Only items with factor loadings of an absolute value greater than .4 on a givenfactor were included in interpretation (Steven, 2002).

The KaisereMeyereOlkin measure revealed a KMO ¼ .839, which indicated an adequate sample size for factoranalysis (Kaiser, 1974). The Bartlett’s Test of Sphericity yielded a chi-square value of 1454.42 (p < .01), suggestingsufficiently large correlations between items. Based on Kaiser’s criterion of eigenvalue greater than 1, the initial analysispermitted the extraction of six factors. Scree plots showed inflexion patterns that justified retaining three or six factors.

To extract optimal numbers of factors and avoid overfactoring, principal-axis factor analysis with oblique rotationwas incorporated. Factor solutions, ranging from six-factor to three-factor solutions, were all examined for parsimonyand interpretability. The three-factor model was chosen because it was the most parsimonious, and extracted factorswere easy to interpret. Finally, nine items that had no significant loading on any factors were removed. The remaining11 items underwent a maximum likelihood factor analysis. The oblique rotated pattern matrix appears in Table 1. Theresults suggested that three factors should be extracted. The total variance for which these three factors accounted was58.94%: 36.23% for Factor 1, 11.80% for Factor 2 and 10.92% for Factor 3. It shall be noted that items 12, 13 and 18required reverse scoring for consistency with other items in the factor structure. To use item 13 as an example, a highlevel of the reversed scores of 4 or 5 points meant a low level of self satisfaction/confidence.

The first latent factor comprised five indicators (observed variables) of anxiety, which reflected nervousness, upset/distress, or feeling intimidated when facing listening activities (high scores of these indicators suggested high level ofstate anxiety). Therefore, Factor 1 might be best labeled Listening Anxiety (LA). The three items representing thesecond latent factor suggested that it might be best named Self-Belief (SB), with traits of confidence and satisfactionwith one’s FL listening proficiency. Indicators of the third factor suggested that it warranted the label FL ListeningDecoding Skills (DS), which describe learners’ cognitive ability related to memory, attention and understanding (withhigh scores for these indicators suggesting low proficiency in listening).

Validation of the EFA results was performed by submitting the three-factor solution to confirmatory factor analysiswith maximum likelihood estimation on both FLLAS (T1) and FLLAS (T2), as given in Fig. 1. As maximum like-lihood estimation method assumes a normal distribution of the data, multivariate normality and linearity were tested

2 Rasch analysis is widely used in field of language assessment for item analysis.

Page 6: Foreign language listening anxiety and listening performance: Conceptualizations and causal relationships

Table 1

Oblique rotation of the three-factor solution.

Factors

1 (LA) 2 (SB) 3 (DS)

5. I am nervous when I am listening to a passage in English when I’m not familiar with the topic .79

6. I get upset whenever I hear unknown grammar while listening to English .72

4. I feel intimidated whenever I have a listening passage in English to listen to .60

1. I get upset when I’m not sure whether I understand what I’m hearing in English .54

8. It bothers me to encounter words I can’t pronounce while listening to English. .50

13. I feel confident when I am listening to English. .83

12. I enjoy listening to English. .53

18. I am satisfied with the level of listening comprehension in English that I have achieved so far. .43

10. By the time you get past the strange sounds in English, it’s hard to remember what you’re

listening to.

.71

9. I usually end up translating word by word when I’m listening to English. .52

3. When I’m listening to English, I get so confused that I can’t remember what I’ve heard. .46

Eigenvalue 3.99 1.30 1.20

% of Variance 36.23 11.80 10.92

169X. Zhang / System 41 (2013) 164e177

for each set of FLLAS results. For FLLAS (T1), two outliers surfaced and were removed from the dataset. For FLLAS(T2), no outlier was found. After outliers were removed, the data were subjected to the KolmogoroveSmirnov test ofnormality, which yielded D(df¼297) ¼ .05 (p > .05) and D(df¼299) ¼ .05 (p > .05) for FLLAS (T1) and FLLAS (T2),respectively. A non-significant value indicated that the data were normally distributed and suitable for the maximumlikelihood CFA.

Regression estimates of LA, SB, and DS as predicted by their indicators were all significantly different from zero(p< .01). Significant correlation estimates among LA, SB, and DS (all were significant at .01 level) suggested that thethree latent variables were correlated. These correlations are indicated by the two way arrows joining the three factorsin the model. Commonly used fit indices appear in Table 2. A majority of commonly used fit indices were above theacceptable fit threshold (Byrne, 2001; Tseng et al., 2006): the goodness-of-fit statistics (GFI), the comparative fit index(CFI), the adjusted goodness-of-fit index (AGFI), the TuckereLewis index (TLI), the incremental fit index (IFI), andthe root mean square error of approximation (RMSEA). The strength confirmed that the three-factor model fitted theFLLAS (T1) data well. To validate the constructs, the three-factor solution model was then run on FLLAS (T2). Gooddata fit statistics confirmed that the three-factor solution was valid.

The third factor Decoding Skills was closely related to self-rating listening proficiency, which suggests it is a latentfactor of self-assessed ability in listening. Because the three indicators of DS are items that represent self-assessmentlistening proficiency, this factor may not be suitable for inclusion as a component of FL listening anxiety. Inasmuch asself-rated listening proficiency may be closely related to FL listening performance, DS can have a strong influence onand predictive power in FL listening performance. Retaining DS as a component of FL listening anxiety in the causalmodel might produce misleading indications of causality. Because the current study concentrated on determining the

Fig. 1. The three-factor solution model. Latent variables: LA ¼ listening anxiety; SC ¼ self-confidence; DS ¼ decoding skills; indicators:

lq_1 ¼ FLLAS item 1; lq_4 ¼ FLLAS item 4; lq_5 ¼ FLLAS item 5; lq_6 ¼ FLLAS item 6; lq_8 ¼ FLLAS item 8; lq_12 ¼ FLLAS item 12;

lq_13 ¼ FLLAS item 13; lq_18 ¼ FLLAS item 18; lq_3 ¼ FLLAS item 3; lq_9 ¼ FLLAS item 9; lq_10 ¼ FLLAS item 10.

Page 7: Foreign language listening anxiety and listening performance: Conceptualizations and causal relationships

Table 2

Model fit indexes for the three-factor CFA model.

c2/df GFI AGFI CFI TLI IFI NFI RMSEA

Acceptable fit <3 >.9 >.9 >.9 >.9 >.9 >.9 <.08

FLLAS (T1) 2.35 .95 .91 .93 .90 .93 .89 .07

FLLAS (T2) 1.93 .96 .93 .95 .94 .95 .91 .06

170 X. Zhang / System 41 (2013) 164e177

possible causal influences between FL listening anxiety and FL listening performance, DS could neither be consideredas a component of FL listening anxiety in the current study nor be used in the causal models.

3.2. Construct validation: the IELTS listening comprehension test

Average scores for the IELTS (T1) and IELTS (T2) were 15.4 points (SD ¼ 4.86) and 20.8 points (SD ¼ 5.10),3

respectively. Each test result was submitted to two-faceted Rasch model for item analysis using FACETS (version 3.65).In IELTS (T1), no participant had the correct answer for both Item 1 and Item 32. The majority of items were

located in a band between þ3 and �3 logits except Item 1 and Item 32. For IELTS (T1), with extremes, the meandifficulty measure was .33. Without the extremes, Item 1 and 32, mean difficulty measure was .00. For IELTS (T2),Item 1 and Item 32 remained the most difficult items with item difficulty of þ4.96 and þ4.67 logits away from the 0.Three participants had the correct answers for Item 1 and four had the correct answers for Item 32. Infit mean squaresof the two items were in the acceptable range of .7 and 1.3 (Bond and Fox, 2007).

These results suggest that the IELTS listening test was appropriate in that it exhibited high reliability and good itemfits. Reliability statistics for the two test results were all above .95. Mean difficulty was adequate for these participants.Infit mean squares for all items (without extremes) of the two test results were in an acceptable range between .7 and1.3 (Bond and Fox, 2007). The high reliability and good Infit statistics suggested that the test was suitable formeasuring participants’ English listening proficiency.

3.3. The causal models

To assess whether causal relations might exist between FL listening anxiety and FL listening performance involvedthe use of structural equation modeling. A structural equation model generally has two parts: the measurement modeland the structural model. The measurement model defines the relationships between the observed variables (e.g.survey items) and the latent variables that are indicated by the specific observed variables. The structural model teststhe hypothesized correlation (estimated by the correlation coefficient of two latent variables) or the hypothesizedcausal effects (estimated by regression coefficient of one latent variable on another).

The current study used two causal models to estimate the causal relations. The first model used Factor 1 e theListening Anxiety (LA), and IELTS test results for exploring the possible causality. The second model included Factor2, Self-Belief (SB), as a component of listening anxiety for establishing the possible causal relationship. The purposeof employing two causal models was to examine what differences it might bring to the causal relations between FLanxiety and FL listening performance if SB was included as a component.

3.3.1. Structure of the first causal modelFig. 2 shows the first causal model, which was constructed on the basis of the assumption that FL anxiety can

deteriorate FL performance (Horwitz et al., 1986; Horwitz, 2001; MacIntyre, 1995a; MacIntyre and Gardner, 1994b),while low FL performance can provoke FL anxiety (MacIntyre and Gardner, 1994a). The structural model part of thefirst causal model was made up by two latent variables, Listening Anxiety at T1 and T2 (LA_1 and LA_2), and twoobserved variables, IELTS test scores at T1 and T2 (IELTS_1 and IELTS_2). The causal paths to be tested were given as:

(1) Listening Anxiety affects IELTS. The model assumed that FL listening anxiety could affect IELTS scores: LA atT1 influenced IELTS at T1, and LA at T2 affected IELTS at T2. These causal relationships were represented bythe single-directional paths of LA_1 / IELTS_1 and LA_2 / IELTS_2.

3 Mean scores of IELTS (T2) were significantly higher than mean scores of IELTS (T1), p < .01.

Page 8: Foreign language listening anxiety and listening performance: Conceptualizations and causal relationships

Fig. 2. The first causal model. Latent variables: LA_1 ¼ listening anxiety at T1; IELTS_1 ¼ IELTS score at T1; LA_2 ¼ listening anxiety at T2;

IELTS_2 ¼ IELTS score at T2; indicators: l1_1 ¼ FLLAS item 1 at T1; l1_4 ¼ FLLAS item 4 at T1; l1_5 ¼ FLLAS item 5 at T1; l1_6 ¼ FLLAS

item 6 at T1; l1_8 ¼ FLLAS item 8 at T1; l2_1 ¼ FLLAS item 1 at T2; l2_4 ¼ FLLAS item 4 at T2; l2_5 ¼ FLLAS item 5 at T2; l2_6 ¼ FLLAS

item 6 at T2; l2_8 ¼ FLLAS item 8 at T2.

171X. Zhang / System 41 (2013) 164e177

(2) IELTS affects Listening Anxiety. The model tested whether IELTS could affect listening anxiety. This causalrelation was given by the single-directional path: IELTS_1 / LA_2.

3.3.1.1. Model fit and model evaluation. Commonly used model fit statistics of the first model are given in Table 3.All eight indiceswere above the acceptable threshold level. They included the chi-square/df ratio, the goodness-of-fit index(GFI), the adjusted goodness-of-fit index (AGFI), the comparative fit index (CFI), the TuckereLewis index (TLI), theincremental fit index (IFI), the normed fit index (NFI) and the root mean square error of approximation (RMSEA).

In light of the strength of the fit statistics, regression coefficients could be safely interpreted (Byrne, 2001).Standardized regression coefficients appear in Fig. 2 next to the regression paths. A negative coefficient suggested thatone factor had a negative effect on another factor while a positive coefficient suggested a positive effect.

To distinguish paths with significant coefficients (p < .05) from those with non-significant coefficients (p > .05),solid and dotted paths are used in Fig. 2 with solid paths representing significant influences. Both of the two regressionpaths from LA_1/ IELTS_1 and LA_2/ IELTS_2 were significant at .01 level, which indicated that the situation-specific listening anxiety seems to have exerted a significant influence on language performance. However, significanteffects of language performance on listening anxiety did not surface as regression coefficient of IELTS_1 / LA_2was relatively small and non-significant.

3.3.2. Structure of the second causal modelAs previously discussed, FL anxiety is a multi-faceted variable (MacIntyre and Gardner, 1991a) that involves one’s

beliefs, perceptions and feelings about an FL (Horwitz et al., 1986). Therefore, the second causal model took Self-Belief

Table 3

Commonly used fix indices of the first causal model.

c2/df GFI AGFI CFI TLI IFI NFI RMSEA

Acceptable fit <3 >.9 >.9 >.9 >.9 >.9 >.9 <.05

The 1st causal model 1.43 .97 .94 .98 .98 .98 .95 .04

Page 9: Foreign language listening anxiety and listening performance: Conceptualizations and causal relationships

172 X. Zhang / System 41 (2013) 164e177

(SB) into account and examined the causal effects of FL listening anxiety on FL performance. As shown in Fig. 3, theSEM was made up by four first-order latent variables (LA_1 and LA_2, Listening Anxiety at T1 and T2), two second-order latent variables (FLLA_1 and FFLA_2, Foreign Language Listening Anxiety at T1 and T2) and the two observedvariables IELTS test scores at T1 and T2 (IELTS_1 and IELTS_2). The causal paths to be tested were given as:

(1) Foreign language listening anxiety affects IELTS. Similar to the first causal model, the second model assumedthat FL listening anxiety could affect IELTS scores: FL listening anxiety (FLLA) at T1 influenced IELTS at T1and FL listening anxiety at T2 affected IELTS at T2. These causal relations were represented by the single-directional paths of FLLA_1 / IELTS_1 and FLLA_2 / IELTS_2.

(2) IELTS affects listening anxiety. The model tested whether IELTS could affect listening anxiety. This causalrelation was given by the single-directional path: IELTS_1 / FLLA_2.

3.3.2.1. Model fit and model evaluation. Commonly used model fit statistics of the revised model are given in Table 4.Compared with the first causal model, all of the major indices dropped due to the inclusion of 3 more observedvariables under SB. Despite this, all eight indices were still above the acceptable threshold level. The predictive powerof the FL listening anxiety (FLLA) on IELTS had improved at both T1 and T2. Therefore, the second-order factorFLLA, which included Self-Belief as one component, had a stronger influence on FL performance.

4. Discussion

4.1. The latent factors (LA, FLLA) are situation-specific anxiety

Distinguishing the differences between trait anxiety, situation-specific anxiety and state anxiety is important forunderstanding FL listening anxiety. Although FL anxiety scales such as FLLAS (Elkhafaifi, 2005) are intended tomeasure anxiety specific to FL listening situations, it was found in the current study that single items in the FLLAS

Fig. 3. The second causal model. Latent variables: LA_1 ¼ listening anxiety at T1; SB_1 ¼ self-belief at T1; FLLA_1 ¼ FL listening anxiety at

T1; IELTS_1 ¼ IELTS score at T1; LA_2 ¼ listening anxiety at T2; IELTS_2 ¼ IELTS score at T2; SB_2 ¼ self-belief at T2; FLLA_2 ¼ FL

listening anxiety at T2; indicators: l1_1 ¼ FLLAS item 1 at T1; l1_4 ¼ FLLAS item 4 at T1; l1_5 ¼ FLLAS item 5 at T1; l1_6 ¼ FLLAS item 6

at T1; l1_8 ¼ FLLAS item 8 at T1; l2_1 ¼ FLLAS item 1 at T2; l2_4 ¼ FLLAS item 4 at T2; l2_5 ¼ FLLAS item 5 at T2; l2_6 ¼ FLLAS item 6

at T2; l2_8 ¼ FLLAS item 8 at T2.

Page 10: Foreign language listening anxiety and listening performance: Conceptualizations and causal relationships

Table 4

Commonly used fix statistics of the revised model.

c2/df GFI AGFI CFI TLI IFI NFI RMSEA

Acceptable fit <3 >.9 >.9 >.9 >.9 >.9 >.9 <.05

The 2nd model 1.64 .93 .90 .96 .95 .96 .90 .046

173X. Zhang / System 41 (2013) 164e177

measure anxiety are not stable over time. For example, the five survey items under the latent factor LA asked par-ticipants to rate their levels of nervousness or upset/distress when facing various English listening activities. Asindicated in the first causal model in Fig. 2, correlations of the five indicators in T1 and T2 were either low (Item 1, 8 &6) or non-significant (Item 4 & 5), which implies that their nerves, upset/distress fluctuate greatly over time. The first-order latent factor LA, subsuming the five survey items (indicators), was more stable over time: the correlation co-efficient between T1 and T2 was almost .70 (p < .01). Regression loads of five survey items on LA meant that LApredicted one’s tendency to become anxious or nervous toward listening. In conclusion, the high correlation coeffi-cient between LA at T1 and T2 suggested that LA is a stable construct.

Since it has been argued that FL anxiety should be regarded as an affective variable with complex constructsincluding not only anxious feelings but also one’s perception and belief toward an FL (Horwitz et al., 1986; MacIntyreand Gardner, 1991a), the current study included Self-Belief as a sub-component of FL listening anxiety (FLLA). In thesecond causal model, both LA and SB were subsumed under a higher order factor FLLA. The model showed that LAhad high loads on FLLA (.94 and .90 at T1 and T2 respectively), which indicated that the second-order latent factorFLLAwas in fact measuring anxiety. Since SB also had significant high loads on FLLA, FLLA should be regarded asa factor that includes an anxiety dimension and a self-belief dimension. As the correlation of FLLA (T1) and FLLA(T2) was higher than that of LA (1) and LA (2), FLLA was more consistent over time than its sub-component LA.Therefore, FLLA is regarded as situation-specific anxiety that has a multi-faceted construct comprising anxiousness,sadness, intimidation, confidence and satisfaction.

4.2. FL listening anxiety influenced FL performance

According to the first and second causal models, FL listening anxiety has a significant influence on listeningperformance. According to the first causal model, standardized path coefficients of LA/ IELTS were�.20 and�.13at T1 and T2 respectively and both coefficients were highly significant (p < .01). The second causal model observeda stronger influence on FL performance by the FLLA (�.28 and .23 at T1 and T2 respectively). According to the twomodels, influences of anxiety on performance were consistent at the two time phases. Participants with high anxiety atT1 tended to have lower performance at T1. At T2, although listening anxiety fluctuated and might be different fromT1 (LA and FLLAwere not perfectly correlated as indicated in the two causal models), the significant regression loadsfrom the two causal models indicated that influence of anxiety on performance had the same pattern for an individualparticipant: when his/her anxiety went up, his/her performance tended to drop; when his/her anxiety reduced, his/herperformance tended to improve. Since the study used longitudinal data, we can expect that if an individual has a highlevel of anxiety at an early time (T1) but low level of anxiety at a later time (T2), his/her performance would tend toimprove at T2.

Both models provided statistical evidence to support the argument that FL anxiety can be a cause of poor per-formance. Comparing the two causal models, FLLA that included SB as a component performed better than LA inpredicting FL performance. Therefore, including SB in FL listening anxiety might produce higher prediction powerover FL listening performance.

4.3. FL performance did not systematically affect FL listening anxiety

Both the first and second causal models did not find evidence supporting that IELTS test scores influenced FLlistening anxiety. As given in the two causal models, regression coefficients of IELTS (T1) on LA (T2) or on FLLA(T2) were relatively small and non-significant (p > .05), implying that the IELTS test scores did not seem to havea systematic effect on FL listening anxiety. Such results were supported by the study of Yan and Horwitz (2008), whofound FL anxiety affected FL performance while causal effects of FL performance on FL anxiety were not found.

Page 11: Foreign language listening anxiety and listening performance: Conceptualizations and causal relationships

174 X. Zhang / System 41 (2013) 164e177

The answer to why FL performance did not systematically affect FL listening performance may again rely on thenature of FL listening anxiety. FL anxiety, as a situation-specific anxiety, does not form suddenly. As MacIntyre andGardner (1991a) suggested, continuous encountering of anxious feelings in FL learning context will lead to a con-solidation of the anxious feelings (state anxiety) into the situation-specific FL anxiety. Such situation-specific anxietymeasured by LA or FLLA evidently took shape before the participants took part in the current study (all participantshad studied English for more than 8 years previously). It would be rather stable over time and might not be stronglyaffected by the performance on a single test (IELTS at T1). Nevertheless, there is no doubt that poor performance cancreate anxious feelings (Sparks and Ganschow, 1996; MacIntyre, 1995a). But these anxious feelings were stateanxiety rather than FL listening anxiety per se. The state anxiety related to the test could be unstable. It could beweakened or enhanced based on future experiences in FL activities or tests. It was unlikely that performance on onetest could have a sudden strong impact on the FL listening anxiety that was formed gradually over time. Therefore, it isnot surprising to find that the influence of listening performance on FL listening anxiety was not significant in the twocausal models.

The causal relations between FL listening anxiety and FL performance explained by the two causal modelssuggested that FL listening anxiety should be noticed and treated, as alleviating FL listening anxiety helps improvelistening performance. Several suggestions for alleviating listening anxiety are given below.

4.4. Suggestions for alleviating listening anxiety

Vocabulary can arouse listening anxiety (Vogely, 1998). FL learners may pay too much attention to new sounds,which can cause comprehension deficits while listening. Hence, appropriate treatments should be given to deal withnew vocabulary. To alleviate anxiety over new vocabulary, FL teachers can provide a list of new words with phoneticsymbols before listening and familiarize learners with the new words. Although word lists are gradually falling out offavor among FL teachers, but they can be a very useful tool for initial exposure to new words (Nation, 1982; Schmittand Schmitt, 1995). With a new word list, students can pay more attention to information rather than to the distractingnew sounds. As regards students at a higher proficiency level and with adequate background knowledge, FL teacherscan encourage them to guess meanings of new words since guessing has been commonly adopted by the commu-nicative approach (Schmitt, 1997). FL teachers may advise the messages to students that it is common to guess themeanings of new words when learning an FL and it is not embarrassing to guess incorrectly.

While new vocabulary can be a source of anxiety, listening instructions that focus too much on production andcorrect answers can also arouse anxiety (Arnold, 2000). FL teachers should avoid using only those listening pro-duction approaches that rely merely on correct answers, which can create high levels of anxiety and distract learnerattention (Arnold, 2000). Instead of asking only for correct answers, FL teachers can encourage students to take risksto make mistakes. There are two advantages in encouraging risk-taking mistakes. By sending the message that makingmistakes is a normal process of learning, FL teachers can help students reduce the concerns over “losing face,”a normal phenomenon often observed in FL classrooms (Horwitz et al., 1986; Liu, 2008; Yan and Horwitz, 2008).Another advantage is that mistakes allow the FL teachers to spot how learners make specific mistakes so that theprocess of making mistakes can be appropriately addressed (Field, 2003). Besides, FL teachers should also helplearners to build up confidence and self-esteem by providing them with encouragement, empathy and comforts(Atasheneh and Izadi, 2012).

Other ways to alleviate FL listening anxiety have been offered by Kondo and Ying-Ling (2004) and Oxford (2008).For example, Oxford suggests “indirect strategies,” which include using “progressive relaxation, deep breathing andmediation,” and using “music” and “laughter.” (pp. 164e165) Oxford also encouraged FL teachers to teach “directstrategies” such as association and elaboration (see pp. 60e69) to alleviate language anxiety. All these strategies mayhelp learners cope with listening anxiety.

5. Limitations

The language used in this study is limited to English, and we still do not have evidence to prove that the structure ofthe FLLAS can function in other types of foreign languages such as Russian, Spanish and French etc. Besides, thescope of the current study was limited to the context of China as all participants were from Chinese colleges.

Page 12: Foreign language listening anxiety and listening performance: Conceptualizations and causal relationships

175X. Zhang / System 41 (2013) 164e177

Participants with different FL learning backgrounds are needed to validate the structure of the FLLAS and assess theeffects of FL listening anxiety on FL listening performance.

6. Conclusions

To my knowledge, this is the first attempt to use quantitative methods to explore a possible causal relationshipbetween FL listening anxiety and English listening performance. The results suggested that FL listening anxiety,especially when allied to insecurity about one’s own FL listening ability, can cause performance to deteriorate, butpoor performance at a specific point in time does not necessarily accentuate anxiety because FL anxiety is a situation-specific trait that is gradually formed over time.

Nevertheless, examining the causal effects was not the only purpose in measuring FL listening anxiety. Anotherpurpose is that it allows FL teachers to identify FL learners who have a high level of foreign language listeninganxiety. These FL learners who can easily become anxious about listening activities and lose confidence may needextra attentions (Gregersen, 2003, 2005). Because anxiety may not only affect cognitive input process (MacIntyreand Gardner, 1991b) but also lead to avoidance behaviors such as skipping homework or class (Horwitz et al.,1986), identifying anxious students could allow FL teachers to pay additional attention to this group of studentsand help them overcome their discomfort or nerves and encourage them to improve satisfaction and build upconfidence.

There is one final point that is worth of mentioning. In addition to anxiety, other individual differences (IDs)factors such as self-efficacy, motivation, learning strategies and self-regulation are important in determining theoutcome of learning an FL. These factors do not act in isolation from one another. On the contrary, these IDsfactors are dynamically related and they can interact with each other during the course of learning an FL(Dornyei, 2009). For example, Mills et al. (2006) found that FL anxiety is associated with self-efficacy and bothof these two factors contribute to FL performance. Hurd and Xiao (2010) found that FL anxiety may interplaywith factors such as the context of learning (e.g. distance learning), learning strategies and job-related experi-ences. Therefore, in order to fully understand the relationship between FL anxiety and language performance,more studies are needed to look at the whole dynamic system, including anxiety, other key IDs factors, thelearning context and so on.

Acknowledgment

This study is support by the National Key Research Center for Linguistics and Applied Linguistics, GuangdongUniversity of Foreign Studies. I would like thank the editors and the anonymous reviewers for their insightfulcomments. All remaining errors are my own.

Appendix A. Foreign language listening anxiety scale

Statements (1) through (20) describe how you feel about listening to English. Please indicate whether you (1) strongly disagree, (2) disagree, (3)

neither agree nor disagree, (4) agree, or (5) strongly agree. Please read each statement carefully, give your first reaction to each statement, and

mark an answer for every statement.

1. I get upset when I’m not sure whether I understand what I’m hearing in English. 1 2 3 4 5

2. When I listen to English, I often understand the words but still can’t quite understand what the speaker is saying. 1 2 3 4 5

3. When I’m listening to English, I get so confused I can’t remember what I’ve heard. 1 2 3 4 5

4. I feel intimidated whenever I have a listening passage in English to listen to. 1 2 3 4 5

5. I am nervous when I am listening to a passage in English when I’m not familiar with the topic. 1 2 3 4 5

6. I get upset whenever I hear unknown grammar while listening to English. 1 2 3 4 5

7. When listening to English I get nervous and confused when I don’t understand every word. 1 2 3 4 5

8. It bothers me to encounter words I can’t pronounce while listening to English. 1 2 3 4 5

9. I usually end up translating word by word when I’m listening to English. 1 2 3 4 5

1. By the time you get past the strange sounds in English, it’s hard to remember what you’re listening to. 1 2 3 4 5

11. I am worried about all the new sounds you have to learn to understand spoken English. 1 2 3 4 5

12. I enjoy listening to English. 1 2 3 4 5

13. I feel confident when I am listening to English. 1 2 3 4 5

Page 13: Foreign language listening anxiety and listening performance: Conceptualizations and causal relationships

14. Once you get used to it, listening to English is not so difficult. 1 2 3 4 5

15. The hardest part of learning English is learning to understand spoken English. 1 2 3 4 5

16. I would be happy just to learn to read English rather than having to learn to understand spoken English. 1 2 3 4 5

17. I don’t mind listening to English by myself but I feel very uncomfortable when I have to listen to English in a group. 1 2 3 4 5

18. I am satisfied with the level of listening comprehension in English that I have achieved so far. 1 2 3 4 5

19. English culture and ideas seem very foreign to me. 1 2 3 4 5

20. You have to know so much about English history and culture in order to understand spoken English. 1 2 3 4 5

176 X. Zhang / System 41 (2013) 164e177

References

Aida, Y., 1994. Examination of Horwitz, Horwitz, and Cope’s construct of foreign language anxiety: the case of students of Japanese. Modern

Language Journal 78, 155e168.

Alderson, J.C., Clapham, C., 1992. Applied linguistics and language testing: a case study of the ELTS test. Applied Linguistics 13, 149e167.

Alderson, J.C., Clapham, C.M., Wall, D., 1995. Language Test Construction and Evaluation. Cambridge University Press, Cambridge, England.

Arnold, J., 1999. Affect in Language Learning. Cambridge University Press, Cambridge.

Arnold, J., 2000. Seeing through listening comprehension exam anxiety. TESOL Quarterly 34, 777e786.

Atasheneh, N., Izadi, A., 2012. The role of teachers in reducing/increasing listening comprehension test anxiety: a case of Iranian EFL learners.

English Language Teaching 5, 178e187.

Bachman, L.F., Palmer, A.S., 1996. Language Testing in Practice: Designing and Developing Useful Language Tests. Oxford University Press,

Oxford, England.

Bond, T.G., Fox, C.M., 2007. Applying the Rasch Model: Fundamental Measurement in the Human Sciences, second ed. Erlbaum, Mahwah, NJ.

Byrne, B.M., 2001. Structural Equation Modeling with AMOS: Basic Concepts, Applications and Programming. Erlbaum, London.

Cattell, R.B., 1966. The scree test for the number of factors. Multivariate Behavioral Research 1, 245e276.

Cheng, Y.S., Horwitz, E.K., Schallert, D.L., 1999. Language writing anxiety: differentiating writing and speaking components. Language Learning

49, 417e446.

Clapham, C.M., Alderson, J.C., 1997. Constructing and Trialing the IELTS Test: IELTS Research Report 3. The British Council, University of

Cambridge Local Examinations Syndicate and International Development Program of Australian University and Colleges, Cambridge,

England.

Davies, A., 2008. Assessing Academic English: Testing English Proficiency 1950e89: The IELTS Solution. Cambridge University Press,

Cambridge.

Dirkx, J.M., 2001. The power of feelings: emotion, imagination, and the construction of meaning in adult learning. New Directions for Adult and

Continuing Education 89, 63e72.Dornyei, Z., 2003. Questionnaires in Second Language Research: Construction, Administration and Processing. Erlbaum, Mahwah, NJ.

Dornyei, Z., 2009. The Psychology of Second Language Acquisition. Oxford University Press, Oxford, England.

Dornyei, Z., Skehan, P., 2003. Individual differences in second language learning. In: Doughty, C., Long, M. (Eds.), The Handbook of Second

Language Acquisition. Blackwell, Oxford, pp. 589e630.

Elkhafaifi, H., 2005. Listening comprehension and anxiety in the Arabic language classroom. Modern Language Journal 89, 206e222.

Field, J., 2003. Promoting perception: lexical segmentation in second language listening. ELT Journal 57, 325e334.

Gregersen, T.S., 2003. To err is human: a reminder to teachers of language-anxious students. Foreign Language Annals 36, 25e32.Gregersen, T.S., 2005. Nonverbal clues: clues to the detection of foreign language anxiety. Foreign Language Annals 38, 388e396.

Horwitz, E.K., 2000. It ain’t over til it’s over: on foreign language anxiety, first language deficits, and the confounding of variables. Modern

Language Journal 84, 256e259.

Horwitz, E.K., 2001. Language anxiety and achievement. Annual Review of Applied Linguistics 21, 112e126.Horwitz, E.K., Horwitz, M.B., Cope, J.A., 1986. Foreign language classroom anxiety. Modern Language Journal 70, 125e132.

Hurd, S., 2007. Anxiety and non-anxiety in a distance language learning environment: the distance factor as a modifying influence. System 35,

487e508.Hurd, S., Xiao, J., 2010. Anxiety and affective control among distance language learners in China and the UK. RELC Journal 41, 183e200.

Kaiser, H.F., 1960. The application of electronic computers to factor analysis. Educational and Psychological Measurement 20, 141e151.

Kaiser, H.F., 1974. An index of factorial simplicity. Psychometrika 39, 31e36.

Kim, J.-H., 2000. Foreign Language Listening Anxiety: A Study of Korean Students Learning English. Unpublished doctoral dissertation,

University of Texas, Austin.

Kitano, K., 2001. Anxiety in the college Japanese classroom. Modern Language Journal 85, 549e566.

Kondo, D.S., Ying-Ling, Y., 2004. Strategies for coping with language anxiety: the case of students of English in Japan. ELT Journal 58, 258e265.

Krashen, S.D., 1985. The Input Hypothesis: Issues and Implications. Longman, New York.

Liu, M., 2008. An exploration of Chinese EFL learners’ unwillingness to communicate and foreign language anxiety. Modern Language Journal

92, 71e86.

MacIntyre, P.D., 1995a. How does anxiety affect second language learning? A reply to Sparks and Ganschow. Modern Language Journal 79,

90e99.

MacIntyre, P.D., 1995b. On seeing the forest and the trees: a rejoinder to Sparks and Ganschow. Modern Language Journal 79, 245e248.

MacIntyre, P.D., Gardner, R.C., 1989. Anxiety and second language learning: toward a theoretical clarification. Language Learning 39, 251e275.

Page 14: Foreign language listening anxiety and listening performance: Conceptualizations and causal relationships

177X. Zhang / System 41 (2013) 164e177

MacIntyre, P.D., Gardner, R.C., 1991a. Methods and results in the study of anxiety and language learning: a review of the literature. Language

Learning 4, 85e117.

MacIntyre, P.D., Gardner, R.C., 1991b. Language anxiety: its relation to other anxieties and to processing in native and second languages.

Language Learning 41, 513e534.

MacIntyre, P.D., Gardner, R.C., 1994a. The effects of induced anxiety on three stages of cognitive processing in computerized vocabulary

learning. Studies in Second Language Acquisition 16, 1e17.

MacIntyre, P.D., Gardner, R.C., 1994b. The subtle effects of language anxiety on cognitive processing in the second language. Language Learning

44, 283e305.

Matsuda, S., Gobel, P., 2004. Anxiety and predictors of performance in the foreign language classroom. System 32, 21e36.

Mills, N., Pajares, F., Herron, C., 2006. A reevaluation of the role of anxiety: self-efficacy, anxiety, and their relation to reading and listening

proficiency. Foreign Language Annals 39, 276e295.

Nation, I.S.P., 1982. Beginning to learn foreign vocabulary: a review of the research. RELC Journal 13, 14e36.

Oxford, R.L., 1999. Anxiety and the language learner: new insights. In: Arnold, J. (Ed.), Affect in Language Learning. Cambridge University

Press, Cambridge, pp. 58e67.Oxford, R.L., 2008. Language Learning Strategies: What Every Teacher Should Know. Beijing World Publishing Corporation, Beijing.

Phillips, E., 1992. The effects of language anxiety on students’ oral test performance and attitudes. Modern Language Journal 76, 14e26.

Raykov, T., Marcoulides, A.M., 2000. A First Course in Structural Equation Modeling. Erlbaum, Mahwah, NJ.

Rivers, W.M., Temperley, M.S., 1978. A Practical Guide to the Teaching of English as a Second Foreign Language. Oxford University Press, New

York.

Saito, Y., Horwitz, E.K., Garza, T.J., 1999. Foreign language reading anxiety. Modern Language Journal 83, 202e218.

Schmitt, N., 1997. Vocabulary learning strategies. In: Schmitt, N., McCarthy, M. (Eds.), Vocabulary: Description, Acquisition and Pedagogy.

Cambridge University Press, Cambridge, England, pp. 199e227.

Schmitt, N., Schmitt, D., 1995. Vocabulary notebooks: theoretical underpinnings and practical suggestions. English Language Teaching Journal

49, 133e143.

Scovel, T., 1978. The effect of affect: a review of the anxiety literature. Language Learning 28, 129e142.Sparks, R., Ganschow, L., 1991. Foreign language learning differences: affective or native language aptitude differences? Modern Language

Journal 75, 3e16.

Sparks, R.J., Ganschow, L., 1996. Anxiety about foreign language learning among high school women. Modern Language Journal 80, 199e212.

Sparks, R.J., Ganschow, L., Javorsky, J., 2000. Deja vu all over again. A response to Saito, Horwitz, and Garza. Modern Language Journal 84,

251e255.

Spielberger, C.D., 1983. Manual for the State-Trait Anxiety Inventory. Consulting Psychologists Press, Palo Alto, CA.

Steven, J.P., 2002. Applied Multivariate Statistics for the Social Sciences, fourth ed. Erlbaum, Hillsdale, NJ.

Tseng, W.T., Dornyei, Z., Schmitt, N., 2006. A new approach to assessing strategic learning: the case of self-regulation in vocabulary acquisition.

Applied Linguistics 27, 78e102.

Vandergrift, L., 1999. Facilitating second language listening comprehension: acquiring successful strategies. ELT Journal 53, 168e176.

Vogely, A., 1998. Listening comprehension anxiety: students’ reported sources and solutions. Foreign Language Annals 31, 67e80.Woldkowski, R., 1999. Enhancing Adult Motivation to Learn. Jossey-Bass, San Francisco.

Yan, X., Horwitz, E.K., 2008. Learners’ perceptions of how anxiety interacts with personal and instructional factors to influence their achievement

in English: a qualitative analysis of EFL learners in China. Language Learning 58, 151e183.

Young, D.J., 1990. An investigation of students’ perspectives on anxiety and speaking. Foreign Language Annals 23, 539e553.