A psychometric evaluation of the Learning Styles Questionnaire: 40-item version

Sixty‐six English‐speaking postgraduate distance‐education medical students completed the Learning Styles Questionnaire (LSQ: 40‐item version). This was completed while attending a residential workshop at the beginning of the semester, and 44 of these students completed the same LSQ questionnaire 5 months later at the completion of the semester. The psychometric properties of the LSQ were assessed using Cronbach's alpha (internal consistency), test‐retest, correlational analyses and factor analysis. The results indicated that the LSQ (40‐item version) has poor reliability and validity, and therefore requires further development and psychometric evaluation.

In the past 25 years the theory of learning styles has generated considerable interest and discourse ([3]). A learning style has been defined as a "description of the attitudes and behaviour which determine an individual's preferred way of learning" ([11], p.1). Competing ideas about learning have led to a growth of terms and concepts, many of which are used interchangeably in learning styles research (eg, learning styles, learning strategies, and terms such as cognitive styles and thinking styles are also used). Different learning styles have been identified, and a large number of learning style measures are now available (see [1]; [4]).

Despite the intuitive appeal of the idea that people have individual preferences and learning styles, there is a host of conceptual and empirical problems. [4]) commented that the learning styles domain is disjointed and that different learning style inventories have been largely unexamined, in that many measures have been accepted without question.

Two of the most widely known measures of learning styles are Kolb's Learning Style Inventory (LSI), which is based on [15], [16] structural model of experiential learning (1976, 1984), and the Learning Style Questionnaire (LSQ: [12], [11]). The LSI was first developed in 1976 and revised in 1985 (LSI‐II), and was designed to measure the degree to which individuals display one of the four learning styles (ie, diverger, assimilator, accommodator and converger). It has been widely used to measure student learning style.

The revised LSI (LSI‐II, 1985) is a 12‐item questionnaire that [17] claims assesses the individual's learning mode preference in the learning cycle. Specifically, it aims to categorise and thereby identifies individual learning style type as either a diverger (dominant learning abilities of feeling and watching), assimilator (dominant learning abilities of thinking and watching), accommodator (dominant learning abilities of feeling and doing) or converger (dominant learning abilities of thinking and doing). Each item requires respondents to rank in order four sentence endings that correspond to the four learning modes described in the [15] Experiential Learning Model (1976). The inventory claims to offer individuals an understanding of how they solve problems, deal with new situations and manage others ([16]). However, major criticisms of Kolb's work have been raised regarding his method of measuring learning styles and on the psychometric properties of the LSI and LSI‐II ([20]; [22]; [27]; [28]).

[11]) and [2]) suggested that the LSQ may be an alternative to the LSI, because the LSQ offers a measurement of learning styles with normative scoring and appears to have higher face validity because it is directly related to behaviours, attitudes and preferences, as most of the items in the LSQ are behaviourally framed (ie, they ask the respondent to agree or disagree with an action that they may engage in). The LSQ is an 80‐item self‐report measure; a 40‐item version is also available. The 40‐item version has become available more recently and was designed for young people, or for people who were not in managerial roles and who would therefore find some of the questions difficult to answer. It was reported that the shorter version serves as a valid alternative to the 80‐item version (Suzanne Hill, personal correspondence).

Each item is a statement referring to a context‐related behaviour, attitude or intention, with which respondents are required to indicate their broad agreement or disagreement. Four separate scores are obtained, one for each of the four learning styles. The four learning styles are activist (eg, seeks out novel perspectives, thinks on their feet), theorist (eg, analytical, pays attention to detail), reflector (eg, listens, observes and reflects before acting) and pragmatist (eg, practical, enjoys experimenting to solve problems). These scales are believed to be equivalent to Kolb's four stages in his learning style: concrete‐experience, abstract‐conceptualisation, reflective‐observation and active experimentation respectively.

Like the LSI, the LSQ (80‐item) has come under close scrutiny recently in terms of its psychometric properties. Nonetheless, the LSQ has been applied mostly within management development and training ([7]), although it has been used regularly in educational settings (eg, [2]; [6]; [8]). The major criticisms include design problems (eg, the LSQ forces someone to agree or disagree with its items and no negatively coded reversed scored responses are included), low internal consistency ([2]; [6]; [8]‐shortened version), some items are low in content validity ([2]; [25]) and concurrent validity ([8]‐shortened version) (See Table 1 for a summary). If the LSQ is to be useful for educational research, its psychometric value must first be established.

1 Summary of psychometrics findings for the Learning Styles Questionnaire (LSQ)

1 a No publication date and psychometrics reported in the accompanying 40‐item LSQ manual

• 2 b No standard deviations provided in guide
• 3 c Authors did not report significance
• 4 d Test‐retest based on a 2‐week interval using Spearman's correlation coefficient (n = 19)
• 5 * indicates significance to the 0.05 level;
• 6 ** indicates significance to the 0.01 level;
• 7 *** indicates significance to the 0.001 level;
• 8 **** indicates significance to the 0.0001 level
• 9 A, activist; R, reflector; T, theorist; P, pragmatist; N/P, not provided; GP, general practice

Recently, [4]) categorised 13 different models of learning styles and evaluated the psychometric properties of the questionnaires based on these models. The LSQ was categorized as a 'flexibly stable' learning style instrument, and psychometrically, it met only one criterion out of the four. The LSQ was deemed to have acceptable psychometric value based on test‐retest, but not on any other marker of reliability or validity. They concluded by stating that perhaps too much is expected from what is a rather simple, self‐report questionnaire.

In this study, we investigated the psychometric properties of the LSQ (40‐item version), as there appears to be no research undertaken to investigate the psychometric properties of this commercially available measure. Although [8]) used a 40‐item LSQ, they reported that they selected these 40‐items by themselves using specified criteria instead of the official Honey and Mumford 40‐item version. However, like [8]), we used the LSQ with Asian students, as cultural factors have not been explored in great depth when assessing the learning styles of students.

This is of great relevance in Australia and Asia, as international student enrolment at Australian higher education institutions has grown significantly over the past 10 years and most international students who study for an Australian qualification originate from Asia ([14]). Therefore, it is important to investigate the psychometric properties of learning style questionnaires from both the Western and Eastern regions of the world, especially as more and more students from Eastern regions are obtaining further education qualifications from Western educational institutions.

Thirty‐seven women and 29 men (N = 66), with ages ranging from 25 to 39 years (M = 27.63, standard deviation [SD] = 2.58), participated in the study. All participants were English‐speaking postgraduate distance‐education medical students from Monash University who were based in Hong Kong; the vast majority of which had graduated either from the Chinese University of Hong Kong (n = 43) or from Hong Kong University (n = 17), and all participants had graduated between 1990 and 2002. The bulk of the participants were undertaking the Graduate Diploma of Family Medicine (n = 55) course. This particular distance‐education course aims to foster the critical appraisal skills of general practitioners in reference to the nature of the discipline, its daily practice and its literature and research base.

LSQ—brief version (LSQ: Honey and Mumford, unspecified).

The LSQ consists of 40 items requiring a true/false response. There are four hypothesised scales (activist, reflector, theorist and pragmatist) corresponding to the four stages of Honey and Mumford's learning cycle. No publication date was specified with the Helpers Guide accompanying the questionnaire. Respondents are instructed to either place a tick or a cross in the box next to each statement, indicating whether they agree with more than disagree with the statement (tick) or whether they disagree with more than they agree with the statement (cross). Each of the four scales (activist, reflector, theorist and pragmatist) is made up of 10 items.

The data was collected in Hong Kong when students attended residential workshops in March 2004 at the beginning of the unit, and again in August 2004, following the completion of the unit. In March, 98 general practice (GP) students were invited to participate and 66 consented (response rate 67%). In August, 42 GP students completed the LSQ again (response rate is 43%).

The means and SDs for each of the four LSQ learning style scales are presented in Table 2.

2 Mean scores and SDs of the four LSQ learning styles scales

10 SD, standard deviation; LSQ, Learning Styles Questionnaire

Internal consistency coefficients (Cronbach's alpha) for all the LSQ learning style scales are presented in Table 3. As shown, the alphas across the scales ranged between 0.25 and 0.50. Cronbach's alpha scores below 0.70 are conventionally considered unreliable and inadequate for research applications ([21]). Therefore, as a result of the low alphas obtained in this study, the use of the 40‐item LSQ for postgraduate medical students may be unsuitable for future research applications.

3 Internal consistency coefficients (Cronbach's alpha) and test‐retest Spearman correlations for the four learning styles scales of the LSQ

• 11 * indicates significance to the 0.05 level;
• 12 ** **indicates significance to the.01 level
• 13 LSQ, Learning Styles Questionnaire

A subsample (n = 42) completed the LSQ twice with a 5‐month interval between administrations. Spearman correlations were used to measure temporal stability between time 1 and time 2: activist 0.45 ( p = 0.003); reflector 0.38 ( p = 0.013); theorist 0.25 ( p = 0.116); and pragmatist 0.36 ( p = 0.021) (see Table 3). These coefficients generally suggest that the Activist, Reflector and Pragmatist scales show significant, yet low, stability over a 5‐month‐long time interval (correlation coefficients greater than 0.50 indicates reasonable temporal stability) ([21]).

As can be seen in Table 4, two pairs of scales were significantly correlated to one another. These figures suggest a moderate positive correlation between activist and pragmatist scales, and the reflector and the theorist scales.

4 Intercorrelations and p‐values for the scales of the LSQ (N = 66)

• 14 * indicates significance to the 0.05 level
• 15 PCC, Pearson's Correlation Coefficient; LSQ, Learning Styles Questionnaire

The 40‐item short version of the LSQ was factor‐analysed using principal component analysis, followed by oblimin rotations as conducted by [8]), to allow for correlations among scales. Oblimin rotations failed to converge in 25 iterations (Convergence = 0.001). Principal component analysis extracted 15 factors with eigenvalues greater than one, which together accounted for 73.26% of the total variance. A four‐factor solution accounted for 31.90% of the total variance. An examination of the factors matrix revealed that items for the four scales did not load coherently nor consistently on the factors as expected. Furthermore, items within the same scale very often had opposite loadings on another factor. These results were very similar to those found by [8]) and do not support a four‐factor structure.

The coefficient alpha values are low on all of the four scales, and especially low on the Theorist and Pragmatist scales. The alpha coefficients found in this study are below the minimum 0.8, specified by [21]) as being the minimum acceptable level for commercial tests. Although our scores are all generally lower than those reported in previous research that used the 80‐item version (eg, [2]; [6]; [26]), the alpha coefficients reported in these studies were, in general, moderate to low. Our internal consistency results were, however, very similar to the study by [8]), which also used a 40‐item version of the LSQ (although they selected the 40 items by themselves). These results therefore suggest that as psychometric theory predicts, as the numbers of items per scale decreases so does the internal consistency of the questionnaire.

Spearman correlations coefficients on a subsample (n = 42) over a 5‐month interval suggested that the activist and pragmatist scales showed low consistency over this time interval, and the reflector and theorist scales did not have any significant temporal reliability.

Lastly, based on factor analysis, we found that the four scales did not load reasonably nor dependably on the factors as expected. These results were similar to [8]) factor exploration of their 40‐item LSQ version, in which they selected the items by themselves.

This study was limited by its sample size (although other larger studies have found similar results) and the fact that the respondents have English as their second language (although each student is required to have a high level of written and oral command of English [TOEFL scores] to gain entry into the course, as it is delivered in the English language). These two factors may have impacted on the results to some extent, and therefore need to be taken in consideration.

Based on our findings, and taking into account the previous studies that have investigated the psychometric properties of the LSQ, we recommend that the LSQ (40‐item version) should be used with caution for research purposes, at least until further psychometric studies are completed on a refinement to the current measure (eg, deletion of poor items, reversed scored items and the use of a Likert scale instead of a dichotomous true/false format). We also question the use of such an instrument when decisions about individuals are being made when English is the person's second language.

This study was funded by the Monash University, Office of the Deputy Vice Chancellor (Academic), Victoria, Australia. The authors also wish to acknowledge the contribution by all the postgraduate diploma students who took part in this study.