Table 3
Componential Structures of the FLCAS as Published in Previous Studies
Aida (1994), similarly to Horwitz, et al. (1991) in her validation study of the instrument also found four components with a varied number of indicators for each.
The first component was interestingly the sum of fear of negative evaluation and Horwitz,
Horwitz, and Cope (1991)
Aida (1994) Cheng, Horwitz, and Schallert (1999)
Pérez-Paredes and Martínez-Sanchez (2001)
Tóth (2008)
Speech anxiety
Communication apprehension
Fear of negative evaluation
Test anxiety
Speech anxiety and fear of negative evaluation 18 items
Fear of failing the class
4 items
Comfortableness in speaking with native speakers
3 items
Negative
attitudes towards the foreign language class 2 items
Low self-confidence in speaking English
10 items
General English classroom performance anxiety
10 items
Communication apprehension
10 items
Anxiety about foreign language learning
processes and situations 8 items
Comfortableness of using English inside and outside the classroom 3 items
Negative
attitudes toward learning English
2 items
Global FLA
30 items
Fear of inadequate performance in English class
1 item
Attitudes to English classes
1 item
Teacher related anxieties
1 item
speech anxiety with 18 indicators. Fear of failing the class, comfortableness in speaking with native speakers and negative attitudes towards the foreign language class were the other three dimensions the author identified, with four, three and two indicators loading onto each respectively.
Cheng, Horwitz and Schallert (1999) in their study on reading anxiety and foreign language classroom anxiety reported two components with ten items loading onto each. The first component was identified as low self-confidence and the other as general English classroom performance anxiety. This componential structure, however, does not seem to stand firm ground as it seems that foreign language classroom anxiety is viewed as a composite of itself (English classroom performance anxiety) and another factor.
In the Spanish context of learning English as a foreign language, Pérez-Paredes and Martínez-Sanchez (2001), similarly to Horwitz et al. (1991), mention a four-component structure of the FLCAS. These dimensions are the following:
communication apprehension, anxiety about foreign language learning processes and situations, comfortableness of using English inside and outside the classroom, and negative attitudes toward learning English. These factors were measured by ten, eight, three and two items respectively, which also calls for caution in terms of disproportional representation of the dimensions and with regards identifying a component with the help of only two indicators (Szokolszky, 2004).
In the Hungarian setting, Tóth (2003, 2008) translated and validated the FLCAS. As a step to ensure validity; she also embarked on confirming the factorial structure of the instrument. Her results consisted of three identified factors: the first component of global foreign language anxiety was represented by 30 items, whereas the other three components with one item each. The reliability of the one-item
components is psychometrically questionable, and the content validity of the 30-item component may also be a source of concern. Given the diversity of factorial solutions available, the need to investigate the componential structure of the Hungarian version of the FLCAS adapted to the secondary school context seems justified.
5.1.2 Results of Principal Components Analyses with the Present Sample
The descriptive statistics show that the scores on the FLCAS of the present sample of 214 learners are close to being normally distributed (see Figure 3 below), with scores ranging from 35 to 149, M = 75.50, SD = 22.19. In Table 3 it can be seen that the sample size employed in this study falls along the middle in comparison with the other studies which range from 96 (Aida, 1994) to 423 (Cheng, Horwitz, &
Schallert, 1999). However, the range of scores in the present study (35-149) proved to be the largest among the six pieces of research listed, which probably accounts for the lowest mean (M = 75.50).
Figure 3. The normal distribution of FLCAS scores.
Table 4
Comparison of Descriptive Statistics of FLCAS Studies
Horwitz, Horwitz, and Cope
(1991)
Aida (1994)
Cheng, Horwitz, and Schallert (1999)
Pérez-Paredes and
Martínez-Sanchez (2001)
Tóth (2008)
present study
Sample N = 108 1st year university Ss in the US
Spanish beginners
N = 96 1st year university Ss in the US
Japanese beginners
N = 423 English majors in Taiwan
English
N = 198 14-65years in Spain
English post beginners
N = 117 1st year English majors in Hungary English advanced
N = 214 tenth grade high school Ss
in Hungary English various levels
Instrument α = .93 range: 45-147 M = 94.5 SD = 21.4
α = .94 range: 47-146 M = 96.7 SD = 22.1
α = .95 NA NA NA
α = .89 range: 49-140 M = 89.07 SD = 18.98
α = .93 M = 84.36 SD = 19.26
α = .94 range: 35-149 M = 75.50 SD = 22.19
Also evident from Table 4 are the Cronbach's alpha values indicating the internal consistency of the questionnaire. This too with the value of α = .94 for the present sample is in line with the Cronbach's alpha values of previous studies using the English (α = .93 (Horwitz, Horwitz, & Cope, 1991); α = .94 (Aida, 1994)), Spanish (α = .89) (Pérez-Paredes & Martínez-Sanchez, 2001), and Hungarian (α = .93) (Tóth, 2008) versions of the FLCAS. It must be noted here, however, that calculating Cronbach's alpha values for the complete instrument may not necessarily
be meaningful as all the items in the scale do not seem to measure the same construct component (see discussion above of the instrument’s componential structure).
The results of the principal component analysis conducted on the data from the sample of 214 learners depicted the three components as defined by Horwitz, Horwitz, and Cope (1991), albeit the distribution of the items cannot be considered ideal from a psychometric perspective, especially in the case of test anxiety, where only two items were found to load onto the scale. 15 items loaded on the communication apprehension scale (5 of them with negative loadings) accounting for 46.22% of the total variance.
A statistic that is used in determining the adequacy of a component is the Kaiser-Meyer-Olkin (KMO) indicator, which reflects the strength of the partial correlations among the variables that are entered in the principal component analysis (Székelyi & Barna, 2002). The goal is to avoid including pairs of variables in the set that highly correlate with one another, and seek KMO values that are 0.5 or higher accompanied by a low correlation matrix determinant. In the present case, the KMO = .923, which is considered to be adequately high and a correlation matrix determinant Det < .001 acceptably low. The third indicator to consider in principal component analysis is the significance of the Bartlett’s test of sphericity which should be p < .05.
In this analysis it was p < .001 (Székelyi & Barna, 2005) (see Table 5).
The component of fear of negative evaluation consisted of 10 items that explained 49.30% of the variance with Det = .010, KMO = .882, Bartlett’s test significant at p < .001. The item loadings can be seen listed below in Table 6.
Table 5
Loadings of FLCAS Items on the Component of Communication Apprehension Item Factor loading
flcas1 .647
flcas3 .716
flcas4 .597
flcas9 .741
flcas11 -.606
flcas12 .663
flcas14 -.560
flcas18 -.736
flcas20 .669
flcas26 .723
flcas27 .825
flcas28 -.796
flcas30 .641
flcas32 -.526
flcas33 .678
The third element of the theoretical construct upon which the instrument was based was test anxiety with two items defining it. Although the loadings of the items are high and together they explain 70.45% of the variance, the determinant (Det = .816) and KMO (KMO = .500) statistics suggest that the correlation of the two items is too high, therefore they do not explain the component sufficiently.
Table 6
Loadings of FLCAS Items on the Component of Fear of Negative Evaluation
Item Loading
flcas7 .787
flcas10 .688
flcas13 .646
flcas16 .689
flcas17 .535
flcas19 .619
flcas23 .842
flcas24 .772
flcas25 .628
flcas31 .760
When issues of content validity of the above components were considered, based on theory and results of previous studies, a more adequate and detailed componential structure emerged. The dimension of communication apprehension could be broken down into three components that have been defined in previous studies:
• participating in speaking tasks in a FL class (cf. Aida, 1994) - referring to the apprehension experienced when speaking in front of peers and the instructor
• general negative attitude towards the FL class (cf. Aida, 1994; Pérez-Paredes
& Martínez-Sanchez, 2001) - expressing discomfort of attending language classes
• lack of self-confidence (cf. Cheng, Horwitz, & Schallert, 1999; Clément, 1980) - indicating negative self-perceptions
In accordance with studies that have found evidence to both the instructor and peers playing an important role in learners’ language anxiety, the items pertaining to the
dimension of fear of negative evaluation lent themselves to be grouped according to the person or group that seems to provoke anxiety in the language learner:
• peers (cf. Horwitz, Horwitz, & Cope, 1991)
• teacher (including evaluation of test performance)(cf. Price, 1991)
This is in line with Higgins’ (1987) discrepancy theory, where one of the sources of anxiety is thought to be the ‘significant other’, a person or group of people who are important to the learner. In case of sixteen-year-old language learners in the context of the foreign language classroom, the ‘significant others’ are evidently their peers and their instructor.
Based on these theoretical considerations, with the help of principal component analysis, the existence of these dimensions was examined. The five components and the item loadings are summarized in Table 7 below. All item loadings are acceptable with respect to conserving the majority of the information pertaining to each item and in case of four components they explain more than 50% of the variance. The KMO values suggest that the partial correlations are low and principal component analysis can be computed with meaningful results. The Cronbach's alpha values also imply that the items determine scales whose reliability is acceptable as all approach a value of .70 or above (Szokolszky, 2004).
Table 7
Componential Structure of the FLCAS Name of
component
Item loadings Variance explained
Det. KMO Bartlett’s test significance Participating in
speaking tasks
flcas1 .645 flcas3 .738 flcas9 .777 flcas20 .681 flcas27 .860 flcas30 .691 flcas33 .727
53.90% .071 .886 .001
General negative attitude towards FL class
flcas6 .620 flcas12 .624 flcas5 -.646 flcas26 .719 flcas17 .763
45.82% .408 .709 .001
Self-confidence flcas11 .672 flcas14 .728 flcas18 .794 flcas28 .770 flcas32 .749
55.32% .196 .747 .001
Fear of negative evaluation of peers
flcas13 .675 flcas16 .710 flcas23 .838 flcas24 .821 flcas25 .628 flcas31 .796
56.08% .110 .874 .001
Fear of negative evaluation of teacher
flcas10 .781 flcas19 .671 flcas21 .687 flcas8 -.753 flcas4 .739
52.91% .302 .816 .001
The descriptive statistics of the scales (Table 8) show that the learners tend to have a positive self-confidence with reference to using English as a foreign language. A relatively large value for standard deviation can be observed in the case of negative
attitudes towards language classes, which means that there is a greater variance in how learners feel about foreign language classes.
Table 8
Descriptive Statistics of the FLCAS Scales
Scales Item
M SD
Item N
Reliability α
Participating in speaking tasks in a FL class
2.11 .37 7 .85
General negative attitude towards FL class
2.18 .53 5 .69
Self-confidence 3.26 .21 5 .79
Fear of negative evaluation of peers
2.11 .20 6 .84
Fear of negative evaluation of teacher
2.16 .27 5 .78
It must be noted here that these five components represented by the five scales are not to be interpreted as the latent dimension of the FLCAS but rather as groups of items that are related to one another preserving most of the heterogeneity of the original items of the scale (Székelyi & Barna, 2005). As these scales (with the exception for attitudes to foreign language class) were found to possess psychometrically acceptable properties, their means as aggregate scores were later used in later analysis as observed variables for determining the relationship of language anxiety and other variables.
In conclusion, the principal component analysis of the FLCAS using the present data resulted in a five componential structure, as opposed to the previously theorized three-component view of language anxiety. Out of the five components, four proved to have acceptable reliability: fear of participating in speaking task, fear
of negative evaluation of peers and teachers, and lack of self-confidence. These components were included in further analyses.