Professional Documents
Culture Documents
2009 - Vandewaetere - Measuring Students Attitude Towards CALL
2009 - Vandewaetere - Measuring Students Attitude Towards CALL
To cite this article: M. Vandewaetere & P. Desmet (2009): Introducing psychometrical validation of
questionnaires in CALL research: the case of measuring attitude towards CALL, Computer Assisted
Language Learning, 22:4, 349-380
This article may be used for research, teaching, and private study purposes. Any
substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,
systematic supply, or distribution in any form to anyone is expressly forbidden.
The publisher does not give any warranty express or implied or make any representation
that the contents will be complete or accurate or up to date. The accuracy of any
instructions, formulae, and drug doses should be independently verified with primary
sources. The publisher shall not be liable for any loss, actions, claims, proceedings,
demand, or costs or damages whatsoever or howsoever caused arising directly or
indirectly in connection with or arising out of the use of this material.
Computer Assisted Language Learning
Vol. 22, No. 4, October 2009, 349–380
Introduction
More and more CALL applications are already implemented in many classrooms or
language labs. This evolution increases the concern that the effectiveness of CALL is
studied and corroborated. Recently, some ‘state of the art’ studies were published
that give a critical overview of research and practice in CALL in recent years, such as
the work of Beatty (2003), Chambers and Davies (2001), Felix (2005) or Levy and
Stockwell (2006). In her paper ‘The unreasonable effectiveness of CALL: What have
we learned in two decades of research?’ Felix (2005) presents a thorough discussion
on strengths and weaknesses in CALL research. A strength is that learners’ overall
perceptions of CALL are positive if they are provided with stable technologies and
receive good support in order to deal with the technology (Felix, 2005). Moreover,
the use of technologies may also positively affect learners’ attitudes and
participation. Summarizing, we can state that there is growing evidence for the
effectiveness of CALL, and more specifically on the domains spelling, reading and
writing (Felix, 2005). However, research on the effectiveness of CALL faces some
problems which have been summarized by Felix (2008): misleading titles; poor
description of the research design; failure to investigate previous research; poor
choice of variables to be investigated; and overambitious reporting of results. Besides
the research related problems there are also concerns about technical difficulties
interfering with the learning process: older students not feeling comfortable with
computers, younger students not possessing the necessary metaskills for coping with
the new learning environments, training needs in computer literacy for both students
and teachers (Felix, 2004). All these factors might heavily influence CALL
effectiveness, without being accounted for.
Notwithstanding the above-mentioned critical considerations, interest in
CALL research as such still increases and focuses on several fields. Chapelle
Downloaded by [National Cheng Kung University] at 19:45 23 April 2013
studies that have investigated learners’ attitudes and perceptions towards CALL
applications, have focused on the short-term use of synchronous and asynchronous
tools developed specifically for a particular study. This has resulted in different
methods that are handled in CALL research. For a comprehensive overview of these
studies, we refer the reader to the study of Felix (2004).
This approach lead Laurillard (2001) to expound that, more and more, we need
to drop the naı̈ve belief in the surplus value of new technologies in learning and the
naı̈ve conclusion that ICT has a positive impact on motivation, simply because there
is a large discrepancy between the specific questions asked in evaluation studies in
new technology, and the often predictable conclusions they come to. Stating that
positive perceptions of CALL correlate with higher learning outcomes in CALL is
nihil novo sub solem, especially since most research reporting such correlation has
Downloaded by [National Cheng Kung University] at 19:45 23 April 2013
Defining attitude
Although the construct of attitude towards computers has gained recognition as a
critical factor in the use and acceptance of information technologies, there is no
single clearly defined definition of computer-related attitude (Liaw, 2002). Several
researchers tried to define and validate the construct of attitude, and most of them
subscribe to the viewpoint of the tripartite model, suggesting that attitudes can be
decomposed into three major components: cognitive, affective and behavioral (Liaw,
2002; Smith, 1971; Wenden, 1991). The cognitive component involves beliefs or
perceptions about the objects or situations related to the attitude. The affective
component expresses the feelings that arise about the cognitive element and the
appraisal (good or bad) of these feelings. Finally, the evaluation of the affect is
translated into a behavioral component that gives utterance to the attitude and
certain attitudes tend to prompt learners to adopt particular learning beha-
viors. More general foreign language attitude questionnaires also adopt this
352 M. Vandewaetere and P. Desmet
Conceptual development
Attitude towards CALL (Computer-Assisted Language Learning)
Several researchers already have measured one or more of the three components of
attitude. Items that have been used in other research were extracted and adjusted for
CALL context. The cognitive component consists of items related to intelligence
(Lightbrown & Spada, 1999) and foreign language aptitude (Horwitz, 1988; Nikitina
& Furuoka, 2007) (for instance: ‘I think I would do well when learning a foreign
language with computer assistance’ beliefs or preconceptions about CALL compared
to traditional language learning as a component of metacognitive knowledge (Bernat
& Gvozdenko, 2005; Lightbrown & Spada, 1999), for instance: ‘Computer-based
Computer Assisted Language Learning 353
tests can never be as good as paper-and-pencil tests’). Three subsets of items form the
affective/evaluative component: integrative/instrumental orientation and motivation
(Gardner, 2001; Horwitz, 1988), for instance: ‘I (would) like learning a new language
by computer’; teacher influence (Cotterall, 1995), for instance: ‘Teachers’ proficiency
of using computers in language learning largely defines my attitude towards
computer use in language learning’; specific beliefs about CALL (Bax, 2003) and
trust in CALL, for instance: ‘I have faith in computer-based language tests’. The
third component, behavior/personality, comprises the subset inhibition/exhibition
(Sil International, 1999), for instance: ‘In a face-to-face learning situation
(classroom), I often experience anxiety when speaking in the foreign language’.
Downloaded by [National Cheng Kung University] at 19:45 23 April 2013
remedial learning, distance learning, blended learning, etc.) and comes in many
applications (e.g., learning platform, computer-mediated communication, course-
ware, etc.), we want to present a preliminary attempt to an operational definition of
attitude towards CALL, CAL and FLL, making use by the presented framework.
The operational definition of attitude towards CALL, CAL and FLL is disconnected
from specific learning contexts or experiments and is instead more focusing on the
learners’ attitudes towards several faces of CALL (Gan, 2004; Sengupta, 2001).
Having defined the construct of attitude in the introduction, in a next part of this
research we focus on the measurement of attitude towards CALL, CAL and FLL. In
order to let the reader interpret the results of the analysis we present a short
introduction to some frequently used statistics in research on questionnaire
validation. For each construct a questionnaire was developed and the results, as
Downloaded by [National Cheng Kung University] at 19:45 23 April 2013
Method
Participants
Participants were recruited via e-mail. Four populations were contacted:
Measures
All participants completed a set of items measuring attitudes in the context of
computer-assisted (language) learning and general language learning. In the final
Computer Assisted Language Learning 355
Table 1. Frequencies (N), percentages (%) of foreign language knowledge (at least notions).
N (%)
1. Assignment from teacher to student (via electronic learning 80 (33.5)
environment, weblog, . . .) when learning a foreign language
2. Interaction between teacher and student (via forum, chat, . . .) 44 (18.4)
when learning a foreign language
3. Mutual interaction between students (via forum, chat, . . .) when 38 (15.9)
learning a foreign language
4. Interaction between students and native speakers (via chat, 33 (13.8)
weblog, forum, . . .) when learning a foreign language
5. Doing language exercises via computer 159 (66.5)
6. Doing language tests via computer 86 (36.0)
7. None of the above 57 (23.9)
Validation procedure
For the validation of each questionnaire, a four-step strategy was used to investigate
the construct validity of the questionnaire. Confirmatory factor analyses were
performed by using AMOS 16.0 (Arbuckle, 1995–2007; Byrne, 2001). This software
program implements the structural equation modeling (SEM) approach or analysis
of covariance structures, and, as a specific case of SEM, confirmatory factor analysis.
For the other analyses, we used SPSS 16.0.
(1) First, the descriptive statistics, the internal consistency (Pearson intercorrela-
tions) and reliability (Cronbach’s a) were computed per subscale on the entire
sample.
(2) Second, exploratory factor analysis (EFA) was conducted on the entire
Downloaded by [National Cheng Kung University] at 19:45 23 April 2013
Fit indices
In step three of the validation procedure, we test how well the constructed item set
represents the underlying construct of attitude. To indicate the model fit, several fit
indices were used. We briefly discuss these indices.
Computer Assisted Language Learning 357
Results
Data preparation
Since maximum likelihood estimation assumes multivariate normality of the data,
this was tested (DeCarlo, 1997) for all items. In the item set that represents the
A-CALL,five observations with a high Mahalanobis D2-value were examined. Only
one observation was considered as outlier and was removed from further analyses
related to the A-CALL. The D’Agostino-Pearson K2 tests and Jarque-Bera LM tests
358 M. Vandewaetere and P. Desmet
showed multivariate normality for all variables in the analyses, except for the items
‘CALL is a valuable extension of the classical learning methods’, ‘I have faith in
computer-based language tests’ and ‘I have faith in computer-based language
exercises’, having a rather positive skewed distribution.
For the item sets that represent the A-CAL and the A-FLL, one observation
could be labeled as outlier, having an extremely high Mahalanobis D2-value. This
observation was removed from subsequent analysis related to A-CAL and A-FLL.
The D’Agostino-Pearson K2 tests and Jarque-Bera LM tests demonstrated
multivariate normality for only a minor part of the selected variables. Variables in
this item set showed a high degree of kurtosis (degree to which the data are peaked or
flat relative to a normal distribution) and skewness (degree to which the data are
asymmetric or skewed to the left or right, relative to a normal distribution). Most
Downloaded by [National Cheng Kung University] at 19:45 23 April 2013
items have been answered affirmatively and mean scores of the items have a tendency
towards right (i.e., totally agree with statement). In view of the Likert-scale property
of the questionnaires presented here and because of the fact that the assumption of
multivariate normality is violated in our data, an alternative estimation method is
often recommended. Instead of using the maximum likelihood (ML) estimation
method that is highly influenced by non-normal data, it is suggested that the
asymptotically distribution free (ADF) method or weighted least squares (WLS)
method is used to estimate model parameters (Browne, 1984). However, simulation
studies suggest that ML estimation in general tends to be more stable, but also shows
higher accuracy in terms of model fit compared to the other estimators, such as GLS
or WLS (Olsson, Foss, Troye, & Howel, 2000). Therefore, ML estimation was used
to analyze the data.
A-CALL questionnaire
STEP 1: internal consistency and reliability of draft questionnaire
The descriptive statistics, the internal consistency and reliability (Cronbach’s a)
were computed for each subscale (the three components of attitude). Internal
consistency values for the subscales are .659 (cognition), .840 (evaluative-affective)
and .536 (behavioral), indicating that only the subscale ‘evaluative-affective’ is
internally consistent. A closer look to the item structure and interrelations
between the items is needed in order to explore fully the factor structure of the
questionnaire.
Factor IV was labeled ‘teacher influence’ and three items summarize the influence of
the teacher on the attitude and motivation towards CALL (e.g., ‘Teacher’s
proficiency of using computers in language learning largely defines my attitude
towards computer use in language learning’). Factor V consists of three items and
was labeled ‘degree of exhibition to CALL’ and captures the extent of inhibition/
exhibition that subjects experience with CALL (e.g., ‘I feel less inhibited when
communicating in the foreign language via computer (chat) than in a face-to-face
situation’).
(n1 ¼ 120) resulting in an RMSEA ¼ 0.106, indicating a rather bad fit. Four items
(‘Computer-based language tests are easier than paper-and-pencil tests’, ‘CALL
requires a certain degree of computer proficiency’, ‘CALL leads to a heavier
workload’ and ‘I prefer the classical way of learning languages’) had low factor
loadings or loadings on more than one factor on the factor structure as revealed by
EFA. These items were removed from further analyses, leaving 20 items in the item
set. Appendix 1 contains the English translation of the final version of the
questionnaire. Factor I (effectiveness of CALL vs. non CALL) has four items (2, 3, 4
and 5, all reversed), factor II (surplus value of CALL) has 10 items (1, 6, 7, 8, 9, 10,
11, 12, 16 and 17), factor III (teacher influence) has three items (13, 14 and 15) and
factor IV (degree of exhibition to CALL) consists of four items (18, 19 and 20, with
items 19 and 20 reversed).
The modification indices of the tested model suggested correlated residuals
between items 6 (‘Computer-assisted language learning is a valuable extension of the
classical learning methods’) and 9 (‘Computer-assisted language learning can stand
alone’), items 6 and 7 (‘Computer-assisted language learning gives more flexibility to
language learning’) and items 16 (‘I have faith in computer-based language tests’)
and 17 (‘I have faith in computer-based language exercises’). Probably the correlated
residuals are caused by a content overlap between the items. On the other hand,
items 6, 16 and 17 did not meet the assumptions of normality and followed a
positively skewed distribution. The model was refitted, with free estimation of the
error covariance for the three pairs of items with correlated residuals. This resulted
in a significant improvement of the model fit. Again, modification indices suggested
slight changes in the model construction, but because of model parsimony it was
decided not to include these changes.
Using the validation sample (n2 ¼ 120) the construct structure of the final model
was confirmed. Table 3 presents the major goodness-of-fit indices for the final four-
factor solution for the calibration and validation sample. The values of the indices do
Table 3. A-CALL: Goodness-of-fit indices for the four-factor structure tested with CFA.
*significant, a 5 .01.
360 M. Vandewaetere and P. Desmet
not allow to conclude for a reasonable or good fit. Values for w2 indicate a significant
deviation of the postulated structure and the obtained dataset. Also, the values for
CMIN/DF, ideally close to 1, indicate a discrepancy between our theoretical model
and the data. Values for RMSEA, GFI, AGFI, CFI and NNFI indicate a mediocre
model fit.
Figure 1 displays the standardized factor loadings of the four-factor model for
both samples.
Downloaded by [National Cheng Kung University] at 19:45 23 April 2013
Figure 1. A-CALL: Standardized factor loadings of the four-factor model as obtained with
confirmatory factor analysis for the calibration sample and the validation sample (between
parentheses). Numbering of items is in accordance with the numbering in Appendix 1.
Computer Assisted Language Learning 361
‘Teacher influence’ does not correlate with ‘degree of exhibition to CALL’ (r ¼ .03).
The positive correlations between the subscales and between the subscales and the
total scale score point to the logical consistency of the questionnaire and indicate
that the A-CALL questionnaire is a feasible working instrument but should be the
subject of more confirmatory research.
A-CAL questionnaire
STEP 1: internal consistency and reliability of draft questionnaire
The value for Cronbach’s a is .818, indicating that the internal consistency and
reliability of the entire scale (15 items) is good. Exploratory factor analysis will give
more details on the underlying factor structure of the questionnaire.
Table 4. A-CALL: Means (M), standard deviations (SD), reliability (a) and Pearson
correlations between the A-CALL subscales (entire sample).
M (SD) a 2 3 4 5
1. Effectiveness of CALL vs. 17.15 (4.13) .80 .42* .18* 7.08 .59**
non CALL (4 items)
2. Surplus value of CALL (10 items) 45.51 (8.36) .86 – .38* .30** .92**
3. Teacher influence (3 items) 13.97 (2.92) .91 – – .03 .53**
4. Degree of exhibition to CALL 12.54 (3.51) .74 – – – .43**
(3 items)
*p 5 .01 (two-tailed).
362 M. Vandewaetere and P. Desmet
Using the validation sample (n2 ¼ 120), the structure of the final model was
confirmed, but not as strong as was expected. Table 5 presents the major goodness-
of-fit indices for the final two-factor solution for the calibration and validation
sample. The values of the indices do not allow to conclude for adequate or good fit.
Values for w2 indicate a significant deviation between the postulated structure and
the obtained dataset. Also, the values for CMIN/DF, ideally close to 1, indicate a
discrepancy between our theoretical model and the data. Values for GFI, AGFI,
CFI and NNFI indicate a mediocre to good model fit.
Figure 2 displays the standardized factor loadings of the two-factor model for
both samples. As can be seen in this figure, standardized regression coefficients are
high, pointing to high internal consistency.
A-FLL questionnaire
In line with previous research on attitude, the construct of attitude was split into
three subcomponents: cognitive, affective-evaluative and behavioral. Each
Table 5. A-CAL: Goodness-of-fit indices for the two-factor structure tested with CFA.
*significant, a 5 .01.
Computer Assisted Language Learning 363
Downloaded by [National Cheng Kung University] at 19:45 23 April 2013
Figure 2. A-CAL: Standardized factor loadings of the two-factor model as obtained with
confirmatory factor analysis for the calibration sample and the validation sample (between
parentheses). Numbering of items is in accordance with the numbering in Appendix 2.
Table 6. A-CAL: Means (M), standard deviations (SD), reliability (a) and Pearson
correlations between the A-CAL subscales (entire sample).
M (SD) a 2 3
1. Computer proficiency (5 items) 22.41 (6.44) .92 .59* .95*
2. Computer integration (4 items) 21.47 (3.66) .80 _ .82*
*p 5 .01 (two-tailed).
accounting for 35% of the variance. Factor I was labeled ‘language learning
proficiency’. Six items capture the self-evaluated proficiency of subjects with
language learning (e.g., ‘I’m efficient in learning a language’). Factor II was labeled
‘language learning beliefs’. Four items define the beliefs subjects have related to
language learning (e.g., ‘People with a high IQ are good at learning a language’).
Seven items did not load substantially on one of the two factors and they did not
form separate factors either. These items were related to specific properties of
language learning (e.g., ‘According to me, learning French is easier than learning
English’ and ‘The most important part of language learning is vocabulary’) and were
removed from further analyses.
In a next step, a confirmatory factor analysis was executed on the two-factor
structure with ten items. What strikes one the most is that none of the items that
Downloaded by [National Cheng Kung University] at 19:45 23 April 2013
This is probably due to the wording of the items (conditional sentences, e.g., ‘If I
would have more time, then I would learn a new language’).
Following the exploratory factor analysis, a confirmatory factor analysis was
executed on the three-factor structure with 18 items. Items that did not significantly
contribute (i.e., no significant regression weights) were removed from the analysis.
The residuals of the items 9 (‘In a country with two or more official languages, it is
important to speak at least two of the languages’) and 10 (‘I consider it very logical
that I learn a language that is spoken in my country’) were allowed to covary,
because of similar wording. Finally, a model was obtained with three factors
(intrinsic motivation, extrinsic motivation and teacher influence), with fit indices
showing a good to very good model fit.
Because of the preliminary and demonstrative nature of this part of the
Downloaded by [National Cheng Kung University] at 19:45 23 April 2013
acceptable fit. Since this research does not want to present a final validated series
of questionnaires and hence is preliminary in nature, no validation analyses were
done and results must be interpreted with caution.
Figure 3 (a, b, c) displays the standardized factor loadings of the three-
component model of attitude towards foreign language learning. As can be
seen in this figure, standardized regression coefficients or loadings of the items on
the cognitive component are very high, pointing to high internal
consistency. Related to the subscales of the affective/evaluative component, factor
loadings are varying between acceptable and high. The loadings of the items
associated with the subscales of the behavioral/personality component are
acceptable but need further exploration in terms of wording or underlying
construct representation.
Downloaded by [National Cheng Kung University] at 19:45 23 April 2013
Table 7. A-FLL: Goodness-of-fit indices for the subcomponents’ structure tested with CFA
(only calibration sample).
*significant, a 5 .01.
Figure 3a. A-FLL – Cognitive component: Standardized factor loadings of the one-factor
model as obtained with confirmatory factor analysis for the calibration sample. Numbering of
items is in accordance with the numbering in Appendix 3.
Computer Assisted Language Learning 367
Downloaded by [National Cheng Kung University] at 19:45 23 April 2013
Table 8. A-FLL – all components: Means (M), standard deviations (SD), reliability (a) and
Pearson correlations for the cognitive, affective/evaluative and behavioral/personality subscale
(entire sample).
M (SD) a 2 3 4 5 6 7
Cognitive (6 items) 28.37 (7.74) .94
Affective-evaluative (13 items)
1. Extrinsic motivation – 7.66 (3.52) .85 28* .06
reversed (3 items)
2. Intrinsic motivation (7 items) 40.18 (5.00) .79 _ .39*
3. Teacher influence (3 items) 16.89 (3.09) .92 _ _
Behavioral/personality (13 items)
4. Inhibition – reversed (2 items) 8.49 (3.02) .85 7.08 .27* .25*
5. Exhibition (3 items) 17.98 (2.12) .64 _ .45* .01
Downloaded by [National Cheng Kung University] at 19:45 23 April 2013
*p 5 .01 (two-tailed).
Table 9. A-FLL – all components: Pearson intercorrelations for all subscales (entire sample).
2 3 4 5 6 7 8 9 10
1. Cognitive component .44* .53* .31* .01 .63* .20* .43* .58* .41*
2. Affective component – .61* .86* .63* .43* .04 .62* .52* .09
3. Affective – extrinsic – – .28* .06 .56* .13* .40* .44* .43*
motivation
4. Affective – intrinsic – – – .39* .29* .01 .56* .43* 7.05
motivation
5. Affective – teacher influence – – – – .05 7.06 .30* .21* –
6. Behavioral component – – – – – .58* .41* .73* .18*
7. Behavioral – inhibition – – – – – – – .27* .72*
8. Behavioral – exhibition – – – – – – .08* .45* .25*
9. Behavioral – tolerance – – – – – – – – .01
of ambiguity
10. Behavioral – learning effort – – – – – – – .22*
– –
*p 5 .01 (two-tailed).
Discussion
Based on the items of several existing, but not validated questionnaires about
attitude towards foreign language learning and computer-assisted learning we
composed a set of items, resulting in a first version of three questionnaires:
attitude towards computer-assisted language learning (A-CALL), attitude towards
computer-assisted learning (A-CAL) and attitude towards foreign language
learning (A-FLL). These questionnaires were designed to assess the attitude and
its subcomponents towards CALL, CAL and FLL. A cognitive component, an
affective/evaluative component and a behavioral/personality component are
generally agreed to be the three subcomponents of attitude.
Analysis of internal consistency of the preliminary A-CALL questionnaire
indicated that the tripartite model of attitude could roughly be corroborated in our
370 M. Vandewaetere and P. Desmet
between the several subscales show several relationships. The more people are
anxious about learning a new language (degree of inhibition), the less tolerant
they are when involved in ambiguous (language learning) situations and the more
they consider the language learning effort (too) high. Also, the more people like
to be exposed to a foreign language, the less they consider the learning effort as
(too) high.
Conclusion
The influence of learners’ attitudes towards language learning and, more
specifically, computer-assisted language learning, is one of the most frequently
investigated topics in the debate on the effectiveness of CALL. However, too
often in CALL research, a set of questions is considered as a reliable and
representative measurement of what researchers want to measure. Also, questions
are developed in line with software that is used, in line with specific goals of the
studies, or in line with participant characteristics. Since the construct of attitude
cannot be formulated unidimensionally, the need for an operational and
multifaceted definition of the attitude-construct emerged. This research described
in this paper aimed at introducing a standard procedure for constructing and
validating measurement for attitude towards computer-assisted language learning
(CALL), computer-assisted learning (CAL) and foreign language learning (FLL).
More generally, this procedure can also be adopted for the validation of other
CALL-related unobservable constructs that we want to measure. Rather than
focusing on the results of the statistical analyses, we wanted to present the reader
with a clearly formulated and stepwise approach to the several types of statistical
validation of a questionnaire in this research field. The focus of this research was
on the process of developing and validating unobservable or latent constructs in
CALL research.
There are some limitations to this study. First, we intend the A-CALL, A-CAL
and A-FLL to be used in a wide range of situations and for diverse groups of
subjects. For this study however, the validation of the three questionnaires was
limited to a sample that largely consisted of subjects with a certain background in
language learning (e.g., 78% of the participants had foreign language training in
school for at least five hours a week and about two thirds of all participants with
CALL-experience had experience with making CALL-exercises), suggesting that
previous foreign language learning behavior or prior experiences with CALL might
have influenced attitude towards CALL and therefore might have caused some
372 M. Vandewaetere and P. Desmet
confirmatory factor analyses did not show good fit. It is highly probably that this is
due to a small sample size with regard to the heterogeneity of our sample. More
research with more diverse types of learners and learners’ backgrounds is needed to
elaborate the findings in this research and to confirm or disconfirm the structure of
the models developed in this research.
A final limitation of this study is that it is based on the self-report of attitude.
Relating the scores of subjects on the A-CALL to the information gathered through
tracking and logging data will give a more detailed view of influences of attitude
towards CALL on the effectiveness of CALL for individual learners or groups of
learners.
Future research should focus on multi-group analyses including different profiles
or clusters that exist in (language) learners. As Huskinson and Haddock (2006)
suggest, individuals with highly structured attitudes may have more accessible
cognitions and effects, resulting in stronger influences of these components on the
behavior of individuals. Furthermore, by using a time series design, eventually
accompanied by the inclusion of experimental conditions, attitude towards CALL,
CAL and FLL can be modeled in function of the progress or decline of learners.
Learners’ attitudes towards CALL could be related to the collective intelligence of
learning environments, representing learners’ behavior in an electronic learning
environment, collected by tracking and logging data. By repeatedly analyzing the
combination of this collective intelligence representing learners’ behavior and more
personal user characteristics (such as attitude towards CALL), it is expected to
elaborate and discuss the conclusions of this and other CALL research.
The results of this research fit in the more general conclusion that is drawn from
most CALL-research: a lot remains to be done (Felix, 2008). Our results indicate
support for a multifaceted structure of attitude towards CALL, including cognitions,
affects and behavior. Using the A-CALL and the other questionnaires as a stand-
alone instrument for measuring one’s attitude towards CALL would not fully
express the possibilities of this questionnaire. The development of these ques-
tionnaires was only a first and preliminary step towards the positioning of attitude
towards CALL into a broader framework of encouraging or discouraging factors
that influence the success or otherwise of CALL, such as attitude towards computer-
assisted learning in general and general attitude towards language learning. Gardner
and colleagues (Gardner, 2005; Masgoret & Gardner, 2003) already instigated the
construction of larger models that place second language learning in relation to other
constructs. The construction of larger integrative models also fits in the recent call of
Hubbard (2008) to create more native CALL theories. In his research he reported on
Computer Assisted Language Learning 373
Acknowledgements
The authors want to express their appreciation to the two anonymous reviewers for their
valuable comments.
Notes on contributors
Mieke Vandewaetere is a PhD candidate in instructional design and computer-assisted
learning. She received her MA in experimental and theoretical psychology and also followed a
MSc in statistical data-analysis. Her primary interests lie in the area of cognitive psychology,
statistics and data-analysis and computer-assisted learning. Other interests are instructional
design, educational data-mining and intelligent tutoring systems.
Piet Desmet is full professor of French and applied linguistics at the Linguistics Department of
KU Leuven and KU Leuven Campus Kortrijk. His research mainly focuses on computer-
assisted language learning (adaptivity, e-testing, corpusCALL, digital libraries, etc.). He is one
of the board members of the research team itec (interdisciplinary research on technology,
education and communication) of KU Leuven Campus Kortrijk.
References
Arbuckle, J.L. (1995–1999). Amos 16.0. User’s guide. Chicago, IL: SPSS.
Ayres, R. (2002). Learner attitudes towards the use of CALL. Computer-Assisted Language
Learning Journal, 15(3), 241–249.
Bax, S. (2003). CALL – Past, present and future. System, 31, 13–28.
Beatty, K. (2003). Teaching and researching computer assisted language learning. Applied
Linguistics in Action Series. Harlow: Pearson Education.
Bentler, P.M. (1990). Comparative fit indexes in structural models. Psychological Bulletin, 107,
238–246.
Bernat, E., & Gvozdenko, I. (2005). Beliefs about language learning: Current knowledge,
Pedagogical implications, and new research directions. Teaching English as a Second or
Foreign Language, 9(1), 1–21.
Bollen, K.A., & Long, J.S. (1993). Testing structural equation models. Newbury Park, CA:
Sage.
Browne, M.W. (1984). Asymptotic distribution free methods in the analysis of covariance
structures. British Journal of Mathematical and Statistical Psychology, 37, 62–83.
Browne, M.W., & Cuddeck, R. (1993). Alternative ways of assessing model fit. In K. Bollen &
J. Long (Eds.), Testing structural equation models (pp. 136–162). Newbury Park, CA:
Sage.
Byrne, B.M. (2001). Structural equation modelling with AMOS: basis concepts, applications and
programming. New Jersey, NJ: Lawrence Erlbaum Associates, Inc.
374 M. Vandewaetere and P. Desmet
Cattell, R.B. (1978). The scientific use of factor analysis in behavioral and life sciences. New
York: Plenum.
Chambers, A., & Davies, G. (2001). New technologies and language learning: A European
perspective. Lisse: Swets & Zeitlinger.
Chapelle, C.A. (2003). English language learning and technology: Lectures on applied linguistics
in the age of information and communication technology. Amsterdam: John Benjamins
Publishing.
Cotterall, S. (1995). Readiness for autonomy: Investigating learner beliefs. System, 23, 195–
205.
DeCarlo, L.T. (1997). On the meaning and use of kurtosis. Psychological Methods, 2, 292–307.
Desmet, P. (2007). L’apport des TIC à la mise en place d’un dispositif d’apprentissage des
langues centré sur l’apprenant. ITL. International Journal of Applied Linguistics, 154, 91–
110.
Desmet, P., & Héroguel, A. (2005). Les enjeux de la création d’un environnement
d’apprentissage électroniques axé sur la compréhension orale à l’aide du système auteur
Downloaded by [National Cheng Kung University] at 19:45 23 April 2013
Lightbrown, P., & Spada, F. (1999). How languages are learned (rev. ed.). Oxford: Oxford
University Press.
Masgoret, A.-M., & Gardner, R.C. (2003). Attitudes, motivation, and second language
learning: A meta-analysis of studies conducted by Gardner and associates. Language
Learning, 53, 123–163.
Merisuo-Storm, T. (2007). Pupils’ attitudes toward foreign-language learning and the
development of literacy skills in bilingual education. Teaching and Teacher Education,
23(2), 226–235.
Myung, I.J. (2003). Tutorial on maximum likelihood estimation. Journal of Mathematical
Psychology, 47, 90–100.
Nikitina, L., & Furuoka, F. (2007). Beliefs about language learning: A comparison between
novice and intermediate level students learning Russian at a Malaysian university. The
Linguistics Journal, 2(1), 7–27.
Olsson, U.H., Foss, T., Troye, S.V., & Howel, R.D. (2000). The performance of ML, GLS and
WLS estimation in structural equation modeling under conditions of misspecification and
Downloaded by [National Cheng Kung University] at 19:45 23 April 2013
totally totally
disagree agree
1. My language learning will 1 2 3 4 5 6 7
proceed more when this is
assisted by a computer.
Downloaded by [National Cheng Kung University] at 19:45 23 April 2013
(continued)
Computer Assisted Language Learning 377
Appendix 1. (Continued).
totally totally
disagree agree
15. Teacher’s proficiency of using 1 2 3 4 5 6 7
computers in language learning
largely defines my attitude
towards computer use in
language learning.
16. I have faith in computer-based 1 2 3 4 5 6 7
language tests.
17. I have faith in computer-based 1 2 3 4 5 6 7
language exercises.
18. I feel less inhibited when 1 2 3 4 5 6 7
Downloaded by [National Cheng Kung University] at 19:45 23 April 2013
Scoring:
. Effectiveness of CALL vs. non-CALL: sum score of items 2, 3, 4 and 5 (all reversed).
. Teacher influence: sum score of items 13, 14 and 15.
. Degree of exhibition to CALL: sum score of items 18, 19 and 20.
378 M. Vandewaetere and P. Desmet
totally totally
disagree agree
1. I can work well with computers. 1 2 3 4 5 6 7
2. I have an advanced knowledge of 1 2 3 4 5 6 7
computers.
Downloaded by [National Cheng Kung University] at 19:45 23 April 2013
Scoring:
. Computer proficiency: sum score of items 1 to 5.
. Computer integration: sum score of items 6 to 9.
Computer Assisted Language Learning 379
totally totally
disagree agree
COGNITIVE COMPONENT
1. I’m efficient in language learning 1 2 3 4 5 6 7
Downloaded by [National Cheng Kung University] at 19:45 23 April 2013
(continued)
380 M. Vandewaetere and P. Desmet