See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/262525037

Foreign Language Classroom Anxiety Scale: A Comparison of Three Models

Article · December 2012

CITATIONS READS

7 39,890

2 authors:

Rokiah Paee Jecky Misieng

University Malaysia Sarawak 27 PUBLICATIONS 210 CITATIONS
1 PUBLICATION 7 CITATIONS
SEE PROFILE
SEE PROFILE

All content following this page was uploaded by Jecky Misieng on 29 January 2016.

The user has requested enhancement of the downloaded file.

 第 9 回マレーシア日本語教育研究発表会 口頭発表

外国語教室不安尺度－3 つのモデルの比較－
Foreign Language Classroom Anxiety Scale: A Comparison of Three Models

Rokiah Bt Paee・Jecky Misieng

Universiti Malaysia Sarawak

Abstract
Foreign language learning anxiety has long been recognized as one of the factors affecting the effectiveness
of language learning but research findings have shown conflicting results. Horwitz et al. (1986) proposed a theory
that predicts learners’ foreign language anxiety in the classroom and developed the Foreign Language Classroom
Anxiety Scale (FLCAS) that was hypothesized to include three domains: communication apprehension, test anxiety
and fear of negative evaluation. However Aida’s (1994) study revealed FLCAS is a four factor model: speech
anxiety and fear of negative evaluation, fear of failing the Japanese class, degree of comfort when speaking with
native speakers of Japanese and negative attitudes towards the Japanese class. Another four factor model was
proposed by Zhao (2007) who reconstructed Horwitz et al.'s three factor model into four domains: communication
apprehension, test anxiety, fear of negative evaluation and anxiety of foreign language class. In the current study,
Horwitz’s three factor model, Aida’s and Zhao’s four factor model of the FLCAS were revisited and compared to
see which one has a better fit for the Malaysian Japanese language learners. The FLCAS was administered to 328
beginning learners of Japanese in a Malaysian university. The internal consistency coefficient of the instrument was
Cronbach’s alpha = .896 (m = 98.0 and SD = 15.17) which shows a reasonably high internal consistency.
Confirmatory Factor Analysis (CFA) were carried using SPSS AMOS and Aida’s four factor model shows a better
fit to the Malaysian data.

外国語学習不安は学習効果に大きな影響をもたらしているひとつの要因として認識されてきた。しかし、
これまでのところ研究結果はさまざまである。Horwitz et. al. （1986） は教室内での学習者の学習不安を
推測する構成概念を提唱し、外国語を学習する際に持つ学習不安の程度を測定する外国語学習不安尺度
（FLCAS）を開発した。FLCASはコミュニケーション不安、テスト不安、否定的評価に対する不安とい
う3つの不安因子から構成されている。しかしAida （1994）の研究では、FLCASは「発話不安と否定的評
価に対する不安」「落第の恐れ」「日本語母語話者と話すときの快適さ」「日本語クラスに対する否定的態
度」という4 因子から構成されていることが判明された。Zhao （2007） はHorwitz のFLCASを再構築し、
4因子に分けなおした。この研究ではHorwitz et. al.（1986）の3因子モデルとAida（1994）及びZhao（2007）
の4因子モデルのどれがマレーシアの日本語初級者に当てはまるかを調査した。マレーシアにある国立大
学の328人の日本語学習者にFLCASを配った。FLCASのCronbach's alpha = .896 (m = 98.0 and SD = 15.17)と
いう高い値を得られた。そしてSPSS AMOS でConfirmatory Factor Analysis （CFA）を行った結果、Aida
（1994） の4因子モデルが最もマレーシア日本語学習者に当てはまるということがわかった。

Keyword: Foreign language anxiety, three factor model, four factor model, confirmatory factor
analysis

- 42 -
 第 9 回マレーシア日本語教育研究発表会 口頭発表

1. Introduction
Foreign language learning anxiety has long been recognized and identified by instructors
as one of the factors affecting the effectiveness of language learning. Defined as the dread or
uneasiness that appears in a learner when attempting to use a foreign language by Gardner and
MacIntyre (1993), it can also be viewed as a negative emotional response such as worry or
anxious while in the process of learning and using a foreign language (MacIntyre, 1999). A
large number of researchers from the field of language education and psychology have
investigated the effects of anxiety on language education (Horwitz et. al., 1986; MacIntyre &
Gardner, 1989; Skehan, 1989; Young, 1991) and their findings proved that some form of
anxiety exist that inhibit learning of language especially a foreign language.

To date, many studies have shown a negative correlation between language anxiety and
learner performance (Aida, 1994; Cheng et. al, 1999; Horwitz, 2001; MacIntyre & Gardner
1989; Saito & Samimy, 1996; Spielman & Radnofsky, 2001) with recent findings showing that
anxiety can differentially affect how learners are able to process input and subsequently
produce output (Sheen, 2008). Even though some studies have found that anxiety can have a
facilitating effect on learning outcomes (Spielman & Radnofsky, 2001), most of these studies
reported unaccommodating effect (Cheng at. al., 1999; Horwitz, 2001).

In explaining the varied conclusions reported by many researchers, Horwitz et. al. (1986)
attributed them to the insufficient conceptualizations of the construct of anxiety made worse by
the fact that there is a lack of valid and reliable anxiety measure specific to FL learning.
Consequently, she and her colleagues developed an instrument called FLCAS to measure the
level of learners’ anxiety. Horwitz et. al. (1986) then came up with the definition of language
learning anxiety as an amalgamation of various incapacitating psychological as well as
behavioral factors that go with language learning situations influenced by the unique process
which is inherent in language learning. Based on this observation, three types of performance
anxieties are proposed and labeled as communication apprehension, test anxiety and fear of
negative evaluation. It is believed that these three anxieties form an integral part of FLA and
together they conspire to inhibit learning as the learner attempts to learn and use a language.

It would seem that researchers have their finger on the crux of the matter pertaining to
language learning anxiety with the advent of the FLCAS. However, a number of studies began
to appear revealing inconsistent results with the sub categorization of language anxiety as well
as the number of factors that made up the FLCAS. Exploratory studies that were carried out
using the FLCAS have put forth findings that are both varied and inconsistent in the way the
FLCAS measure language anxiety (Aida, 1994; Yuan, 2011; Zhao, 2007). Each study proposed

- 43 -
 第 9 回マレーシア日本語教育研究発表会 口頭発表

different models to capture the fundamental nature of language anxiety as proposed by

Horwitz and her colleagues. One study (Gregersen, 2006) has shown that cultural differences
may play a role in these mixed findings due to the fact that the FLCAS were used across
language and cultural diversities. Therefore, this current study was an investigation to ascertain
which of these proposed models can adequately explain FLA as defined by Horwitz et. al.
(1986) based on data collected from students learning a foreign language in their native
country.

2. Literature Review
The growing awareness experienced by many language educators on the debilitating
effect of language anxiety has caused a significant rise in the number of studies regarding this
phenomenon (Gregersen, 2006) both by language educators and psychologists. With the
introduction of instruments such as the French Class Anxiety Scale and the French Use
Anxiety Scale (Gardner, 1985) and the more popular FLCAS as developed by Horwitz et. al.
(1986), interest in language anxiety research has continued to rise (Kitano, 1998).

Horwitz et. al. (1986) characterized language learning anxiety 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’. In their conception, language learning
anxiety, can be subcategorized into three distinct forms of performance anxieties namely
communication anxiety, test anxiety and fear of negative evaluation. They argued that these
three forms of anxiety are fundamental to the concept of FLA and together caused language
learning to be a daunting task for a learner.

With this view of language anxiety, Horwitz and her colleagues came up with an
instrument called the FLCAS. It was constructed based on self-reports from students, their own
clinical experiences as well as evidence culled from reviews of similar instruments. The
finalized version of the FLCAS contained 33 items which employs 5-point Likert-type scales
with selections ranging from “strongly agree” to “strongly disagree.” This instrument was
intended to measure foreign language learners’ level of anxiety while learning a language in
the classroom. A higher score obtained by the instrument would indicate a higher level of FLA.
Reliability of the scores obtained from the instrument based on data collected from 108
respondents was quite high with Cronbach’s alpha of .93 (Horwitz, 1986). In addition,
test-retest reliability carried out with a sample of 78 participants over a period of eight weeks
was ascertained to be r=.83 (p < .01) (Horwitz, 1991).

- 44 -
 第 9 回マレーシア日本語教育研究発表会 口頭発表

Test Anxiety
Item 3, 5, 6, 8, 10, 11,
12, 16, 17, 20, 21, 22,
25, 26, 28
Communication
Fear of Negative
Apprehension
Evaluation
Item 1, 4, 9, 14, 15,
Item 2, 7, 13, 19,
18, 24, 27, 29, 30,
23, 31, 33
32

FLCAS

Figure 1. Model 1 FLCAS (Horwitz et. al, 1986)

Aida (1994) decided to review Horwitz et. al.’s three factor model of FLA by validating
an adapted FLCAS based on students of Japanese to attempt to find out the underlying
structure of the FLCAS. Assessment of reliability and the relationship of students’ anxiety
levels to their performance in Japanese were carried out concurrently. The study reported that
the adapted FLCAS was able to produce scores that are highly reliable at measuring anxiety
levels of students learning Japanese. Aida established four factors instead of three: speech
anxiety and fear of negative evaluation, fear of failing the Japanese class, degree of comfort
when speaking with native speakers of Japanese, and negative attitudes towards the Japanese
class. Six items (items 2, 6, 15, 19, 28 and 30) were removed from the final model because
they failed to load on any of the four factors. Incidentally, the finding in this study was
consistent with other studies based on western languages where they all concluded that
language anxiety was negatively related to students’ performance in the language they were
learning.

Fear of Failing Comfortableness

Item 10, 22, 25, with Japanese
26 Item 11, 14, 32

Speech Anxiety
Item 1, 3, 4, 7, 8, Negative Attitude
9, 12, 13, 16, 18,
20, 21, 23, 24, 27, Item 5, 17
29, 31, 33
FLCAS

Figure 2. Model 2 FLCAS (Aida, 1994)

- 45 -
 第 9 回マレーシア日本語教育研究発表会 口頭発表

Another effort to re-examine Horwitz’s FLCAS factor structure was done by Zhao
(2007) who used a Chinese version and established that there are also four factors namely
communication anxiety, test anxiety, fear of negative evaluation and anxiety of English classes.
The instrument was administered to a group of 115 second year high school students learning
English as a foreign language in China. Unlike Aida’s four factor model, all 33 items from the
original instrument were used in the final model proposed. However, Zhao did not report any
reliability measures nor how the items load on their respective factors.

Fear of Negative
Test Anxiety
Evaluation
Item 2, 8, 10, 19,
Item 3, 7, 13, 15,
21
20, 23, 25, 31, 33

Communication Anxiety of English

Apprehension Classes
Item 1, 9, 14, 18, Item 4, 5, 6, 11, 12,
24, 27, 29, 32 16, 17, 22, 26, 28,
FLCAS 30

Figure 3. Model 3 FLCAS (Zhao, 2007)

Another researcher (Yuan, 2011) compared Zhao’s (2007) four factor model to that of
Horwitz and colleague’s three factor model based on a sample of 300 Chinese college students
learning English as a foreign language making use of the Chinese version of the FLCAS. In
Yuan’s study, both models were reported to show all 33 items were significant indicators for
their respective factors (p>.0001). The correlations between the four latent constructs for
Horwitz’s model ranged from .91 to 1.0 while Zhao’s model has correlations ranging from .97
to 1.0. However, when the models were analyzed for fit, Horwitz and colleague’s model
(X2=2169.18, RMSEA=.07, AIC=7.72, SBC=8.57, BCCVI=7.78) was reported to have a much
better fit for Chinese learners as compared to the Zhao’s model (X2=2211.19, RMSEA=.07,
AIC=7.88, SBC=8.77, BCCVI=7.93).

In Saito and Samimy’s study (1996), it was observed that language anxiety affected
students differently in accordance to the level of instruction they were experiencing. They
examined the role of language learner anxiety of Japanese language students’ as a predictor of
performance at three different instructional levels. Among other predictor variables used was
Year in College which was reported to be the best predictor of performance for beginning level

- 46 -
 第 9 回マレーシア日本語教育研究発表会 口頭発表

students. However, language anxiety came up on top as the best predictor of performance for
both intermediate and advanced level students. The study proved conclusively that the
predictive variable of the students of Japanese performance differs according to instructional
levels and that the impact of FLA has a positive correlation with Japanese language learners'
instructional levels.

Wan Zarina et. al. (2007) conducted a study on two groups of students in a Malaysian
public university learning the Japanese language to determine and compare their perceptions
towards their feelings of anxiety they experienced when learning the foreign language.
Numerous variables that may affect the level of anxiety such as the student’s course, gender
and race were studied. The findings of the study indicate that male students experienced higher
level anxiety in communication apprehension compared to female students whereas in fear of
negative evaluation, both gender experience the same level of anxiety. For the overall anxiety,
Chinese and Malay students showed a higher level of anxiety compared to Indian students and
students of other races.

In other Asian contexts, similar results were also obtained in studies involving both high
school students and college students. In Taiwan, Cheng (1994) found out that scores obtained
from the FLCAS was the best predictor of senior high school students’ English language
proficiency levels. Meanwhile, in Korea, another researcher found FLCAS scores were
significantly negatively correlated with final grades obtained by college English as Foreign
Language learners in both reading and conversation classes (Kim, 1998)

3. Purpose of the Study

This study revisited Horwitz’s et. al. (1986) three factor model and Aida’s (1994) and
Zhao’s (2007) four factor model of the FLCAS and compared it to see which model has a
better fit for the Malaysian Japanese language learner.

4. Methodology
4.1 Participants
The participants of the present study were 328 beginner learners of Japanese at one
public university in Malaysia. Japanese language course is offered as an elective subject and
the students have to sit for four oral tests and one written test in order to pass the course. They
were from various faculties, academic majors and ethnic groups.

- 47 -
 第 9 回マレーシア日本語教育研究発表会 口頭発表

4.2 Procedures
The questionnaire was distributed during the last week of the semester. At this point, all
students had undergone the four oral tests. Therefore, learners’ FLCAS scores reflect their
anxiety during the whole semester.

4.3 Instrument
Aida(1994) adapted version of Horwitz et. al. FLCAS where the term ‘foreign language’
in the original FLCAS were replaced with ‘Japanese language’ was used. The questionnaire
consisted of two parts. Part one was intended to collect personal information of the participants,
such as their name, age, gender, academic major and ethnicity and the participant’s first
language. The second part consists of the 33 statements. The respondents were asked to rate
each item on a five-point Likert scale ranging from 1 (strongly disagree) to 5 (strongly agree).
When statements of the FLCAS were negatively worded, response were reversed and recoded.

4.4 Data Analyses

The reliability was tested using Cronbach’s Alpha and later confirmatory factor analysis
(CFA) was done using SPSS AMOS version 20. Brown (2006) pointed out that in CFA, the
number of factors and the pattern of indicator-factor loadings are specified in advance on the
basis of strong empirical knowledge or theory. For the purpose of this study, three models form
the basis the empirical knowledge and a comparison was carried out in terms of
goodness-of-fit (GOF) statistics and size of the factor loadings to see which of the competing
models have the best fit. The better fit model was determined by comparing the fit indices of
the models produced using the value of Chi-square (X2), Degree of Freedom (df), Root Mean
Square Error Approximation (RMSEA), Akaike Information Criterion (AIC), Standardized
Root Mean Residual (SRMR), Comparative Fit Index (CFI) and Parsimony Comparative Fit
Index (PCFI).

For Chi-square goodness of fit index, the best fit is usually represented by small numbers,
large p-values so as not to reject the null hypothesis of no significant difference between our
observed and specified model. Other categories of fit indices are often recommended by
experts to accompany the X2 index and df to ensure accuracy in our interpretation. They are
either absolute or incremental fit indices with 1.0 as perfect fit and values above .95 as
acceptable fit. Examples of absolute and incremental fit indices are CFI and PCFI Indices that
are based on the discrepancy between observed and predicted covariance matrices of our
models like the RMSEA and SRMR states that the value of 0 means perfect fit and values
below .05 is considered adequate fit. The AIC, on the other hand, states that the model that
gives small values close to 0 is more parsimonious and therefore, a better fitting model.

- 48 -
 第 9 回マレーシア日本語教育研究発表会 口頭発表

5. Results
The internal consistency coefficient of the instrument was Cronbach’s alpha = .896 (m =
98.0 and SD = 15.17) which indicates a reasonably high internal consistency.

Table 1
A Comparison of the Three Models
Model X2 df RMSEA SRMR AIC CFI PCFI
Aida (1994) 792.056 318 .068 .656 966.056 .790 .716
Horwitz et. al. (1986) 1175.428 492 .065 .692 1379.425 .752 .700
Zhao (2007) 1177.382 489 .066 .691 1387.382 .750 .695

Table 1 shows the fit statistics for all the three models. Aida’s model which shows both
the CFI and PCFI indices as having higher values closest to the recommended .95 when
compared to the Horwitz et. al.’s and Zhao’s models is the model of choice here. In terms of
discrepancy between the specified and the observed models, all models show poor fit but
according to the SRMR fit index, Aida’s model is not as bad as the other two models. However,
based on the RMSEA, it is the worst fitting model. Based on X2 values and df to account for
errors, Aida’s model is again on top of the other two models with less values and more degrees
of freedom. In fact with a normed Chi-square (NC) ratio where the X2 is divided by its df) of
2.49 indicative of a parsimonious good fit, it is far better than Horwitz et. al.’s and Zhao’s
models with NC’s of 2.38 and 2.40 respectively. This is further supported by the smallest AIC
value that the model obtained as compared to the other two models. Therefore, all fit indices
except the RMSEA reveals that Aida’s model is the best fitting model.

6. Discussion and Conclusion

FLA has become the focus of many language learning research and findings have shown
conflicting result. Horwitz et. al. (1986) hypothesized a theory that predicts learners’ FLA in
the classroom setting and came up with a 33 item instrument called the FLCAS to measure it.
Based on the instrument, they proposed a three factor model that includes the three domains of
communication apprehension, test anxiety and fear of negative evaluation. Using an Asian
language to tests Horwitz et. al.’s conception of FLA. Aida conceptualized a four factor model
but did not include six items due to weak factor loadings on their respective factors. The
domains of Aida’s (1994) model include speech anxiety and fear of negative evaluation, fear of
failing the class, comfortableness in speaking with native Japanese and negative attitudes
towards the Japanese class. Zhao (2007), on the other hand, retained all the original 33 items in
his four factor model which he renamed as communication anxiety, test anxiety, fear of
negative evaluation and anxiety of English classes. All the researchers reported high

- 49 -
 第 9 回マレーシア日本語教育研究発表会 口頭発表

reliabilities in the scores obtained from their instruments.

Based on the confirmatory factor analyses carried out on all the three models, Aida’s
four factors model is discovered to be a more viable model in accounting for FLA among
Malaysian students studying Japanese as a foreign language. The best possible explanation for
this is that unlike the other two models, in Aida’s model, six items that were not loading
particularly well on their factors were removed from the analysis. Upon closer examination,
many of the items were either too similar or redundant. Examples of the items included are
item 19, “I am afraid that my Japanese teacher is ready to correct every mistake I make” and
item 15, “ I get upset when I don’t understand what the teacher is correcting”. Furthermore, the
factor of Test Anxiety which was present in both Horwitz et. al.’s and Zhao’s model but not
found in Aida’s model could be another possible explanation for this result. This seem to
confirm the finding of MacIntyre & Gardner (1989) that test anxiety contributed to the general
anxiety factor and not to the communicative anxiety factor which seems to indicate that test
anxiety is a more generic predicament faced by students instead of a specific indicator of FLA.

A follow up study is recommended to test more models that have been proposed to seek
the best model of all that can best explained this complex phenomenon.

References
Aida, Y. (1994). Examination of Horwitz, Horwitz, and Cope’s construct of foreign language
anxiety : The case of students of Japanese. The Modern Language Journal, 78, 155-168.
Brown, T. A. (2006), Confirmatory Factor Analysis for Applied Research. New York: Guilford
Press.
Cheng, Y. K. (1994). The effects of attitudinal, motivational, and anxiety factors on the English
language proficiency of Taiwanese senior high school students. Unpublished doctoral
dissertation, University of Georgia, Athens.
Cheng, Y.S., Horwitz, E. K., & Schallert, D. (1999). Language anxiety: Differentiating writing
and speaking components. Language Learning, 49, 417–446.
Gardner, R.C. (1985). Social psychology and second language learning: The role of attitudes
and motivation. London: Edward Arnold.
Gregersen, T. (2006). A cross-cultural comparison of nonverbal teacher immediacy and foreign
language anxiety in Chilean and Russian English language classrooms. Revista Signos
39(62)Valparaiso
<http://www.scielo.cl/scielo.php?script=sci_arttext&pid=S0718-09342006000300004&l
ng=en&nrm=iso&ignore=.html> Accessed on 5 Sept. 2012

- 50 -
 第 9 回マレーシア日本語教育研究発表会 口頭発表

Horwitz, E. K. (1991). Preliminary Evidence for the reliability and validity of a foreign
language anxiety scale. In E. Horwitz & D. Young (Eds.), Language anxiety: From theory
and research to classroom implications (pp. 141-150). Englewood Cliffs, NJ: Prentice
Hall.
Horwitz, E. K. (2001). Language anxiety and achievement. Annual Review of Applied
Linguistics, 21, 112-126.
Horwitz, E. K., Horwitz, M.B. & Cope, J. (1986). Foreign language classroom anxiety. The
Modern Language Journal, 70(2): 125-132.
Kim, S. Y. (1998). Affective experiences of Korean college students in different instructional
contexts: Anxiety and motivation in reading and conversation courses. Unpublished
doctoral dissertation, The University of Texas, Austin
Kitano, K. (2001). Anxiety in the college Japanese language classroom. The Modern Language
Journal, 85(4), 549-566
MacIntyre, P. D. (1999). Language anxiety: A review of the research for language teachers. In
D. J. Young (Ed.), Affect in foreign language and second language learning: A practical
guide to creating a low-anxiety classroom atmosphere (pp. 24–45). Toronto:
McGraw-Hill.
MacIntyre, P. D., & Gardner, R. C. (1989). Anxiety and second language learning: Toward
theoretical clarification. Language Learning, 39, 251-275
――――――――――――――. (1991). Methods and result in the study of anxiety and
language learning: A review of the literature. Language Learning, 41, 85-117.
Saito, Y., Samimy K. K. (1996). Foreign language anxiety and language performance: A study
of learner anxiety in beginning, intermediate, and advanced-level college students of
Japanese. Foreign Language Annals, 29(2), 239-249.
Sheen, Y. (2008). Recasts, language anxiety, modified output, and L2 learning. Language
Learning, 58(4), 835-874.
Spielman, G. & M. Radnofsky. 2001. ‘Learning language under tension: New directions from a
qualitative Study’. The Modern Language Journal 85(2), 259–278
Young, D. J. (1991). Creating a low-anxiety classroom environment: What does language
anxiety research suggest? Modern Language Journal, 75, 426-436.
Wan Zarina, W. Z., Azizah, R., Norfazila, S., Nur Al-Huda, H. (2007) Kajian tahap
keresahan di kalangan pelajar bahasa asing terhadap mata pelajaran bahasa asing.
<http://eprints.utm.my/5784/1/75061.pdf> Accessed on 3 Aug 2012
Cao, Y. (2011). Comparison of Two Models of Foreign Language Classroom Anxiety Scale,
Phillipine ESL Journal, 7, 73-93.
Na, Z. (2007). A Study of High School Students’ English Learning Anxiety. Asian EFL
Journal, 9(3), 22-34.

- 51 -

View publication stats