International students occupy one in five places at Australian universities (Sydney Morning Herald, August 2002). Most of these students come from neighbouring Asian countries and stay in Australia between 1 to 4 years. Prior to enrolment, many international students study English at recognised intensive English language centres to achieve the required level of English. Speaking skills are considered a problem for Asian students, because o the lack of exposure to native speakers, and the emphasis on formal exam driven language courses in Asian countries. Research has indicated that language learning anxiety impacts most on speaking skills MacIntyre found foreign language anxiety is one of the strongest predictors of foreign language performance .

Early research indicated the construct of test anxiety was not applicable to language learning situations , probably because of a specificity concerning the formal evaluation component of test anxiety, and the communicative element in language learning. As a result, research into language learning anxiety has focussed on a domain specific conceptualisation of anxiety.

Language learning anxiety is defined as situation specific, , that is trait anxiety occurring in specific situations. There is some evidence in the literature for two types of anxious individuals those who experience anxiety because of a skills deficit and those who experience problems recalling previous knowledge, or interference retrieval based on Tobias' model of stages of anxiety .

Groundbreaking research into language learning anxiety was conducted by Horwitz using American university students studying various languages as a course requirement. She produced the Foreign Language Classroom Anxiety Scale (FLCAS) which has been widely used . Horwitz's research indicated that anxiety was significantly related to poor performance in the foreign language, particularly speaking skills. This research focussed on classroom anxiety. When learning a language in the target language environment learners need to study in class but they also need to use the language outside of class, so a conceptualisation of anxiety relevant to English language learners in Australia needs to take this into consideration.

To account for the in/out of class distinction a two dimensional conceptualisation of second language speaking anxiety was proposed based on communicative situations. A five point Likert type scale was designed, based on types of interactions according to setting and interlocutor. The 22 item questionnaire was piloted and revised resulting in a final instrument comprising 11 items.

The research projects posed the following research questions

1. Is conceptualisation/ and instrumentation to measure second language speaking anxiety applicable, reliable and valid?
2. Is there a relationship between speaking performance and second language speaking anxiety?
3. What are the major stressors reported by sample?

4. Do learners from different ethnic groups differ in level of second language speaking anxiety?

The participants were 275 (136 female, 139 male) advanced English for academic purposes (EAP) learners studying at accredited language centres in Australia. The ethnic background of the participants is presented in table 1. The participants studied on intensive EAP courses for an average of 20 hours per week and almost all the participants planned to study at Australian universities subsequent to completing their English course. Most of the participants were in their early 20s, and over half of the participants planned to study at postgraduate level (n=149, 64.7%). Most of the participants had been in Australia for three months or less (n=112, 44.3%) and at their centre for three months or less (n=149, 54.2%)

Table 1

All of the students completed the anxiety subscales and oral performance was assessed using a type of International English Language Testing System (IELTS) interview. This test is used to assess the level of English for university entry in Australia. To provide more in depth information about perceived stressors 47 of the participants took part in a semi-structured interview.

The instrumentation was designed for the study because the existing measures were considered inappropriate. The subscale was designed based on communicative situations taking into consideration: setting, interlocutor, level of formality and whether initiating or responding. In the initial pilot instrument, the items were designed to capture anxiety reactions. The initial factor analysis produced high error variances and some of the participants reported problems understanding the reaction words, so the items were simplified in the main study. To facilitate understanding of the concept of second language speaking anxiety a definition was given in English at the beginning of the subscale. This was translated into Chinese, Korean, Japanese, Vietnamese, Indonesian and Thai.

The descriptive statistics for the subscales are presented in table 2.

Table 2

To assess the robustness of a two-factor model of second language speaking anxiety a confirmatory factor analysis (CFA) using LISREL 8.5 was conducted. All screening procedures used to assess conformity to underlying assumptions produced satisfactory results. A two factor model corresponding to in-class and out-of-class anxiety provided a better fit to the data than a one factor model, and this indicated a moderate fit to the data. The model replicated the pilot study, however, factor loadings were much improved, possibly due to the simplification of the items. Three error covariances were included in the model to enhance fit, these corresponded to native speaker interaction, answering questions and performing in front of the class. Figure 1 illustrates the CFA model and table 3 reports the fit indices.

A model based calculation indicated good reliability for the instrumentation (in class anxiety = .89, out of class anxiety =.87).

Figure 1

Table 3

A Pearson correlation analysis was computed, using SPSS 10, of anxiety and oral performance as measured by the IELTS speaking test indicated that there was a significant negative relationship between second language speaking anxiety and oral performance. This is illustrated in table 4.

Table 4

** significant at p = <.01

To compare anxiety across cultures a multivariate analysis of variance (MANOVA) was computed. Participants from ethnic groups with less than 5 members were excluded from the analysis and there were two missing cases where participants had not entered their nationality, thus leaving a final sample of 265.

According to the descriptive statistics, performing in front of the class was rated most frequently on the in-class subscale, while talking to native speakers rated most frequently on the out-of-class subscale. In the pilot study the questionnaire contained several items referring to communication with non-native speakers of English. The respondents reported that they did not experience anxiety speaking to non-native speakers therefore these items were removed from the questionnaire for the main study.

To provide further information beyond the descriptive statistics about perceived stressors data from the semi-structured interview was used. Participants were asked in which situations they experienced anxiety when speaking English. The categories that emerged from the data supported the quantitative analysis with the highest ranking stressor relating to classroom activities related to performing in front of others.

Table 5

To assess whether ethnicity or sex were influencing factors in second language speaking anxiety a 10 (ethncity) by 2(sex) multivariate analysis of variance(MANOVA) was conducted There was no significant effect for sex (V=.005, F(2,245) =.64, p=.53), h ²

=.005. There was however a significant main effect for ethnic group (V=.15, F(18,492) =2.28, p=.01).The univariate analyses of variance (ANOVAS) indicated significant main effects of ethnic group on in class anxiety F(9, 246) = 2.69, p<.01, h ² = .09) and on out-of- class anxiety (F(9, 246) = 3.45, p<.00001, h ² = .11). The estimated means are presented in figure 2 and 3. The graphs indicate that respondents from Confucian heritage cultures (CHCs), that is China and Chinese speaking countries, Japan and Korea, experience more anxiety than respondents from Europe, and Vietnam, with Japanese and Korean participants indicating the highest levels of anxiety.

Using a significance level of .05 Tukey's HSD post hoc tests indicated significant differences between European and Korean participants on in class anxiety and Vietnamese and Korean, Vietnamese and Japanese and Vietnamese and Taiwanese on out of class anxiety. This indicates some evidence for a distinction between CHCs and other cultures.

Figure 2

Figure 3

The results show that the two dimensional classification of anxiety is reasonable for this sample -- that out-of-class anxiety is just as likely to interfere with language use as in-class anxiety. The primary stressors refer to performance like classroom activities and communicating with native speakers and there seemed to be some differences between ethnic groups.

Based on research in education there is evidence that different types of anxious learners benefit from different treatments. The skills deficit learner, experiences anxiety because he/she does not have the necessary skills and benefits from skills input, while an interference retrieval learner is unable to retrieve information due to cognitive interference and benefits from desensitising. (Naveh-Benjamin, 1991)

To reduce in-class anxiety language learning strategies can play an important role. In the case of oral presentations and performing in front of class there is an argument for reducing the number of these activities and highlighting more use of peer interaction in English. However, giving an oral presentation is an important skill for EAP students. Careful staging may reduce potential anxiety and a gradual reduction in structured to free activities.

Learners of English in Australia benefit from being immersed in the target language environment and it is desirable to provide them with strategies to access as much language as possible to facilitate their learning. Most of the respondents referred to communicating with native speakers as being a situation in which they experienced speaking anxiety. They also perceived communicating with native speakers as being a good way to learn English. Learning communication strategies that would enable learners to initiate and maintain conversations would facilitate speaking to native speakers. Out of class tasks involving interacting with native speakers, such as student devised surveys, can provide some support for learners. In addition, encountering a greater number of native speakers, for example setting up an English club with visiting speakers, excursions to the pub. in the safe environment , would be beneficial.

The results indicated an effect for ethnicity. Interestingly Vietnamese participants performed in the same way as European participants. The Vietnamese participants differed from the other participants because they were mostly scholarship students who had already gone through a rigorous selection procedure. In addition, many of these participants already had IELTS scores and guaranteed places at university. This could lead to greater self-confidence and lower speaking anxiety. However cultural differences in language learning anxiety need further exploration taking into account variables of linguistic proficiency and language learning background.

This project focussed on a narrow section of the English language learning community -- the sample were all advanced EAP learners. It would be valuable to investigate speaking anxiety with lower level general English students. In addition there is a need to further explore the effect of ethnicity on anxiety and compare this to more contextual influences such as educational background and teacher variables.