Academic Development Unit
La Trobe University
Reporting results of research undertaken at Monash University, Melbourne, Australia
Kolb (1985) responded to these concerns by publishing a revised version of the Learning Style Inventory (LSI2) for which he claimed improved reliability but to date, despite the fact that the LSI2 is widely used by teachers and trainers, there have been few studies investigating its reliability or validity. In this paper we report the results of research designed to address these issues as well as to provide data directly related to the Australian, which complements that from the US American or British, context.
It could be expected on the basis of Kolb's (1984) theory that learning style preferences would relate to career choice in that, for example, a computer scientist is required to establish a dynamic interplay between conceptual knowledge and experimentation in order to develop software. It should be noted also, however, that although a computer scientist with an innate preference for a convergent learning style would probably find it easier to meet daily work demands and thus individuals with such a preference are probably more likely to be attracted to and remain in the profession, not all computer scientists are likely to have this innate preference.
Kolb's (1981,1984) acknowledgement and Talbot's (1985) demonstration of the influence of long-term environmental or short-term situational factors upon learning mode implies not only that professional or academic demands may temporarily affect or permanently alter learning style preferences but also that any individual will respond to the demands of different learning contexts by utilising to differing degrees, as perceived to be appropriate, concrete, abstract, active or reflective learning strategies.
In the light of this, it is important to note that although the Learning Style Inventory (Kolb, 1976, 1985) assesses both learning style preference and the relative strength of preference for each learning mode - by requiring the ranking of either 36 single words describing learning strategies (comprising nine sets of four response alternatives - LSI1) or 48 short sentences about learning (comprising twelve sets of four response alternatives - LSI2) - the inventory does not specify reference on the part of the respondent to a particular learning context. Thus the responses of a given individual when focussing upon learning preferences related to acquiring driving skills might be quite different to the responses recorded when focussing upon the study of English Literature in an academic context. Similarly, a computer scientist with a general preference for a divergent learning style (CE/RO) might record a preference for a convergent learning style (AC/AE) if, at the time of taking the test, s/he is asked to focus upon learning in the context of a computer science course or happens to do so.
Table 1. Learning style preferences, by discipline or profession
Social work graduates
|Occupational therapy (Katz, 1988)
Social work academics
(Kruzich et al, 1986)
(Reading Brown et al ,1989)
Physical sciences (Kolb, 1984)
(Kolb, 1976, 1984)
Social sciences (Kolb, 1984)
Of the above studies, only those of Kolb (1976), Kruzich, Friesen, and van Soest (1986) and Katz (1988) have examined the issue of gender difference in preferred learning styles. Although Kruzich et al (1986) found no significant difference in preferences between males and females, Kolb (1976) reported a tendency for females to emphasise concrete experience and males to emphasise abstraction. This finding is supported by Katz (1988) who, in her study of Israeli university students, found an interaction between gender and career choice in that the engineering students who scored high on the abstract dimension were mainly men while all her occupational therapy students who scored higher on the concrete dimension were women.
In summary, the findings related to disciplinary differences and gender suggest either that academic discipline influences preferred learning styles or that individuals tend to cluster in disciplines where the tasks and learning demands match their innate learning style preferences. However, all these findings relate to the LSI1, and confirmation of these findings using the LSI2 is needed.
Of these studies, several find two bipolar factors but argue that this is attributable to instrument bias (Freedman and Strumpf, 1978, 1980; Certo and Lamb, 1980; Newstead, 1992). Certo and Lamb (1980) were unable to isolate bipolar factors when their subjects rated the LSI items individually using a Likert scale, while Wunderlich and Gjerde (1978), who found support only for existence of the Reflective Observation dimension, point to the difficulties inherent in comparing four words representing pairs from two different dimensions and argue that a factor analysis reveals.
Support for the validity of the LSI1 comes from studies by Merritt and Marshall (1984) using both the original form and an alternative modified LSI1. Ferrell (1983) also found four factors paralleling those conceptualised by Kolb (1984) as did Katz (1988). Further analysis conducted by Katz (1988) provided additional strong support for Kolb's experiential learning theory in that it yielded the arrangement of LSI1 items that would be predicted by the theory.
Moore and Sellers (1982) found no clear relationship between teaching styles and learning preferences and Fox (1984) found no relationship between learning style and instructional preferences, but in each case no validity or reliability data were presented for the instructional or teaching style instruments.
Using more established instrumentation, Highhouse and Doverspike (1987) found the predicted relationship between career and learning preference dimensions for most occupations, but found no relationship between the LSI and field independence. These researchers, like West (1982) conclude that the LSI measures learning preferences rather than cognitive style, which is not at variance with Kolb's (1984) notion of adaptation to the context of learning.
Of the four studies of the LSI2 located, two (Sims, Veres, Watson and Buckner, 1986; Veres, Sims and Shake,1987) report the high internal consistency found by Kolb (1985), although Sims et al (1986) suggest that this may be attributable to response sets, given that in the revised version of the instrument the item relating to each learning dimension appears in the same position in every set of response alternatives. Both these studies report low to moderate test-retest reliability, as is also demonstrated in the research using the LSI2 conducted by Atkinson (1988, 1989). These findings in relation to the LSI2's test-retest reliability parallel those for the LSI1 in that low to moderate test-retest reliability is reported by Freedman and Strumpf (1978), Geller (1979) and Wilson (1986), but while Freedman and Strumpf (1978), Geller (1979) and Wilson (1986) also report moderate to high internal consistency for the LSI1, Moore and Sellers (1982) and Newstead (1992) report only low internal consistency.
Despite the existence of value differences between Australians and US Americans in general (Renwick, Smart and Henderson, 1991), it was anticipated on the basis of the considerable similarities in educational practice and the desired outcomes of education that the Australian data would largely reflect that obtained in the US. That is, - consistent with Table 1 - it was predicted that undergraduate arts students (comprising, in this Australian university, those majoring in history, English, sociology, anthropology, linguistics, foreign languages and politics) would display a preference for learning through reflective observation combined with concrete experience or abstract conceptualisation (ie. be placed in the two right-hand quadrants of Table 1). It was further predicted that students of the sciences (comprising, in this case, those majoring in mathematics, psychology, computer science, engineering, chemistry, physics, biochemistry and physiology) would display a preference for abstract conceptualisation combined with either active experimentation or reflective observation (ie. be placed in the two bottom quadrants of Table 1). It was further predicted that, if males and females were equally represented in the samples from each of the faculties, there would be no significant difference found on the basis of gender alone.
In line with the method used in previous studies, subjects were not given any instructions regarding the learning context they were to consider when completing the LSI.
An analysis of the scale intercorrelations using Pearson product moment correlations, followed by a principle components factor analysis of the four scales for Science and Arts subsamples and the full sample, was used to further investigate the relationships between the scales. The factor analysis used an eigenvalue cut-off of 1, followed by a varimax rotation.
Finally, the scale mean scores were broken down by gender, discipline and whether the majority of primary and secondary schooling was completed in Australia or in Asia, using an independent samples t-test.
Table 2. Coefficient alpha reliabilities and scale intercorrelations
Table 2 also shows the correlations between the scales. Consistent with the hypothesis that there are two bipolar dimensions, the Abstract Conceptualisation scale is substantially and statistically significantly negatively correlated with the Concrete Experience scale, as is the Active Experimentation scale with the Reflective Observation scale. Both these correlations are substantially larger than the other correlations between the scales. The matrix also shows that the two bipolar dimensions are essentially independent of each other, as expected from the theory.
However, the factor analyses of the scales shown in Table 3 are not entirely consistent with this interpretation. They show that for the two sub-samples and for the total sample, the bipolar dimension linking Reflective Observation and Active Experimentation are consistent with the theory (Factor 1 for Science, Factor 3 for Arts and Factor 3 for the Total). However, only in the Science sub-sample do Concrete Experience and Abstract Conceptualisation form a bipolar dimension. In the Arts sub-sample, Active Experimentation seems to form bipolar dimensions with each of the other three scales, suggesting that in the Arts sub-sample Concrete Experience and Abstract Conceptualisation may not be as well contrasted as in the Science sub-sample.
Table 3. Factor Analysis of the LSI2 Scales
Tables 3, 4 and 5 show mean scores of each scale broken down, repectively, by gender, faculty of enrolment, and whether the majority of primarily and secondary schooling was completed in Australia or in Asia. As Tables 3 and 4 show, there is little or no difference between the sexes on any of the scales, but Arts students had a statistically significant higher average score on Concrete Experience than did the Science students while Science students had a statistically significant higher mean score on Active Experimentation than the Arts students. As well, the Science students had a statistically significant higher mean score on the AE-RO dimension, probably as a result of the higher mean score on Active Experimentation. Finally, Table 5 shows that students who had completed the majority of their primary and secondary education in Asia had a statistically significant higher mean score on Concrete Experience than Australian-educated students, while Australian-educated students had a higher mean score on the AC-CE dimension than the students who had been educated primarily in Asia.
Table 4. Mean Scores on LSI2 Scales broken down by Gender
ms Marginally significant; p < .10
* p < .05;, **p < .01
Table 4. Mean Scores on LSI2 Scales broken down by Faculty
ms Marginally significant; p < .10
* p < .05;, **p < .01
Table 5. Mean Scores on LSI2 Scales broken down by Educational Background
ms Marginally significant; p < .10
However, in this study discipline variation has been found in that for the Arts sub-sample and for the sample as a whole, Reflective Observation and Active Experimentation were found to form a bipolar dimension but Active Experimentation was also found to form bipolar dimensions with both Concrete Experience and Abstract Conceptualisation.
The results obtained for the Arts sub-sample (which are reflected in the sample as a whole) may be attributable to the fact that in disciplines that focus largely on human experience and interaction, concepts are developed at least in part from the base of one's own personal experience or feelings. It is arguable that there is in such disciplines possibly more of an interdependence between Concrete Experience and Abstract Conceptualisation than there is in the Sciences, where concepts are more likely to be developed in response to reflection upon active experimentation.
In the Arts as defined in this sample - history, English, sociology, anthropology, linguistics, foreign languages and politics - the contrast is more likely to be of Abstract Conceptualisation, Concrete Experience or Reflective Observation with Active Experimentation because in the disciplines specified the opportunity for actual hands-on, practical experimentation is extremely limited. Where experimentation does occur, as in the learning of a foreign language, it consists of the oral or written testing of hypotheses. That is, reflection upon personal experience (upon responses to one's use of language) leads to concept development (the formation of new hypotheses about construction of the language) and the testing of the concepts developed, again through personal experience.
Mean scores indicate no significant differences on LSI2 scales between male and female students, which is not unexpected in the light of previous mixed research findings in this area. Similarly, the mean score results for Arts and Science students are in line with expectations in that Arts students have a significantly higher mean score on Concrete Experience and Science students have a significantly higher mean on Active Experimentation.
For students who had completed the majority of their primary and secondary education in Asia, the higher mean score on Concrete Experience and lower mean score on the AC-CE dimension may reflect a culturally-based tendency to emphasise feelings associated with the learning environment and personal affective involvement in learning. While no studies using the LSI to compare learning styles across cultures have been found, support for such an inference comes from the noted inclination of Asian-born students to work and study together in friendship groupings, unlike Australian-born students who tend to study individually in relative isolation from others.
In conclusion, the results of this study suggest that the LSI2
is an instrument of high reliabity in terms of its internal consistency.
There is also some evidence of validity but variation has been
found on the basis of discipline. It is hypothesised that this
variation reflects differing bases of concept formation in the
Arts and the Sciences, but this hypothesis needs to be tested.
Further research might also explore the possibility that preferred
learning styles are influenced by cultural background.
Atkinson, G. (1989). Kolb's learning style inventory - 1985: test-retest déjà vu. Psychological Reports, 64, 991-995.
Baker, J.. Wallace, C., Bryans, M & Klapthor, L. (1985). Analysis of learning style. Southern Medical Journal, 78,1494-1497.
Certo S. & Lamb, S. (1980). An investigation of bias within the learning styles inventory through factor analysis. Journal of Experiential Learning and Simulation, 2, 1-7.
Ferrell, B. (1983). A factor analytic comparison of four learning-styles instruments. Journal of Educational Psychology, 75, 33-39
Freedman R. & Strumpf, S. (1978). What can one learn from the learning style inventory? Academy of Management Journal, 21, 275-282.
Freedman R. & Strumpf, S. (1980). Learning style theory: less than meets the eye. Academy of Management Review, 5, 445-447.
Fox, R. (1984). Learning styles and instructional preferences in continuing education for health professional: a validity study of the LSI. Adult Education Quarterly, 3, 72-85.
Geller, L. (1979). Reliability of the learning style inventory. Psychological Reports, 44, 555-561.
Highhouse, S. & Doverspike, D. (1987). The validity of the learning style inventory 1985 as a predictor of cognitive style and occupational preference. Educational and Psychological Measurement, 47, 749-753.
Katz, N. (1988). Individual learning style: Israeli norms and cross-cultural equivalence of Kolb's Learning Style Inventory. Journal of Cross-cultural Psychology, 19, 361-379.
Kolb, D. (1976). Learning style inventory. Boston: McBer & Co.
Kolb, D. (1981). Learning styles and disciplinary differences. In Chickering, A (ed.), The Modern American College. San Francisco: Jossey-Bass.
Kolb, D. (1984). Experiential learning: experience as the source of learning and development. New York: Prentice-Hall.
Kolb, D. (1985). Learning style inventory. Boston: McBer & Co.
Kruzich, J. Friesen, B & Van Soest, D. (1986). Assessment of student and faculty learning styles: research and application.Journal of Social Work Education, 22, 22-30.
Merritt, S. & Marshall, J. (1984). Reliability and construct validity of ipsative and normative forms of the learning style inventory. Educational and Psychological Measurement, 44, 463-472.
Moore, A. & Sellers, R. (1982). Pilot test results of two instruments for assessing the learning and teaching style of adult education teachers. Adult Literacy and Basic Education, 6, 226-237.
Newland, P., Powell, J. & Creed, C. (1987). Understanding architectural designers selective information handling. Design Studies, 8, 2-16.
Newstead, S. (1992). A study of two "quick-and-easy" methods of assessing individual differences in student learning. British Journal of Educational Psychology, 62, 299-312.
Plovnick, M. (1975). Primary care career choices and medical student learning styles. Journal of Medical Education, 50, 849-855.
Reading-Brown, M. & Hayden, R. (1989). Learning styles - liberal arts and technical training: what's the difference? Psychological Reports, 64, 507-518.
Sadler, G., Plovnick, M., & Snope, F. (1978). Learning styles and teaching implications. Journal of Medical Education, 53, 847-849.
Sims, R., Veres, J., Watson, P., & Buckner, K. (1986). The reliability and classification stability of the learning style inventory. Educational and Psychological Measurement, 46, 753-760.
Talbot, R. (1985). Situational influences on learning styles. Industrial and Commercial Training, 17, 19-28.
Veres, J., Sims, R., & Shake, L. (1987). The reliability and classification stability of the learning style inventory in corporate settings. Educational and Psychological Measurement, 47, 1127-1133.
West, R. (1982) A construct validity study of Kolb's learning style types in medical education. Journal of Medical Education, 57, 794-796.
Wilson, D. (1986). An investigation of the properties of Kolb's learning style inventory. Leadership and Organization Development Journal, 7, 3-15.
Wunderlich, R. & Gjerde, C. (1978). Another look at learning style inventory and medical career choice. Journal of Medical Education, 53, 45-54.
|Please cite this paper as: Willcoxson, L. and Prosser, M. (1996). Kolb's learning style inventory (1985): Review and further study of validity and reliability. British Journal of Educational Psychology, 66, 251-261. Reprint at:
By permission, British Psychological Society Journals Office.