JVER v25n4 - Learning Styles in Vocational Work Experience

Learning Styles in Vocational Work Experience

José Hermanussen Regional Center for Vocational Education, Tilburg, The Netherlands	Ronny F. A. Wierstra Utrecht University, The Netherlands
Jan A. de Jong Utrecht University, The Netherlands	Jo G. L. Thijssen Utrecht University, The Netherlands

Abstract

A critical discussion of existing instruments for measuring learning styles in work-based learning situations resulted in a new instrument: the Questionnaire Practice oriented Learning (QPL). It consists of bipolar items, intended to measure five bipolar constructs: immersion, reflection, conceptualization, experimentation, and regulation. In a pilot study testing the usefulness of the instrument, data were gathered about the work-based learning of 407 students of a Dutch vocational school. A cluster analysis, followed by a discriminant analysis, resulted in three well interpretable work-based learning styles: 1. 'Focused on doing with incidental learning', 2. 'Learning on the basis of external regulation', and 3. 'Self-regulated learning on the basis of theory and reflection'. Some suggestions for further research on the instrument and the learning style model behind the instrument are offered in conclusion.

Introduction

An important component of vocational education programs is the work experience component. In many countries vocational education contains both a school component and a field component. In senior secondary vocational education in The Netherlands, the field component covers at least 20% of the curriculum. Students can choose for either a school-based program (with a maximum of 60% field experience) or a work-based program (with over 60% field experience). Little is known about the way students learn from their field experiences. For an empirical investigation into these learning processes and in order to develop procedures to support these learning processes, a descriptive model of learning-from-work-experiences is needed, plus an instrument to measure individual differences in learning-from-work-experiences. In particular, we are interested in identifying work-based learning styles. Generally speaking a learning style is defined as a coherently used combination of learning activities that a student usually employs in a particular type of teaching-learning situation, a combination that is characteristic of him/her in a certain period (compare Wierstra et al., 2000 ; Vermunt & Verloop, 1999 , Slaats, Lodewijks & Van der Sanden, 1999 ). For the purpose of our research we confine the teaching-learning situation to work-based learning.

Mainstream educational theory with regard to learning is biased towards formal theoretical learning in classrooms and individual study. A large category of learning style research (e.g. Biggs, 1987 ; Marton, Hounsel & Entwistle, 1988 ; Schmeck, 1988 ; Vermunt, 1998 ) is, implicitly or explicitly, focused at this type of learning. Slaats ( 1999, p. 174 ) concludes that the dominant learning models and the learning inventories departing from these models are more closely related to theoretical learning settings than to practical settings. In a sense this is understandable, because many studies of learning styles have been conducted in the fields of general and higher education, types of education that are typically theoretical and highly abstract in nature. Slaats argues that further research is needed in order to gain insight in the learning processes that occur in practical settings ( p.177 ).

Researchers and educators interested in on-the-job learning often resort to the experiential learning style model of David Kolb (e.g. Kolb, 1984 ) and his followers, since this model is more geared to learning in practical settings, although claimed to be generally applicable. It is a comprehensive model of learning-from-experience, integrating a variety of literature on learning, cognitive development, and cognitive styles. This model depicts experiential learning as a process alternately involving four basic modes of learning. According to Kolb, students usually have preference for some modes above others. Such a combination of preferred learning modes constitutes the person's learning style. Kolb ( 1976 , 1985 ) developed two versions of an instrument to measure experiential learning style. From a test theoretical point of view, both instruments contain major flaws, which will be discussed below. Although some investigators have developed alternative instruments based upon Kolb's experiential learning model, trying to correct some of these flaws, the results are still not satisfactory.

Not only are the instruments based upon Kolb's model problematic, but also the model itself has some weak spots. Its dimensionality can be criticized, for both theoretical and empirical reasons, as will be demonstrated. Nevertheless, several elements in the model have considerable face value, and are well rooted in the literature on experiential learning and cognitive development. In our study, we developed an instrument inspired by the experiential learning theoreticians as for the identification of four learning modes, but without incorporating Kolb's assumptions with respect to the dimensionality of these learning modes and with correction of some of the test-theoretical weaknesses found in the existing instruments.

Kolb's Experiential Learning Theory

The following four propositions summarize Kolb's experiential learning theory (see also De Ciantis & Kirton, 1996 ):

1. Experiential learning involves four distinct learning modes: concrete experience CE ('feeling'), reflective observation RO ('watching'), abstract conceptualization AC ('thinking'), active experimentation AE ('doing').
2. The four learning modes represent four stages in experiential learning. Experiential learning is a cyclical process in the sequence: CE, RO, AC, AE and so on in a new cycle.
3. The four learning modes represent two dimensions: CE and AC are supposed to be poles on a dimension 'prehension', and RO and AE are poles on a dimension 'transformation'.
4. Although normal adults possess and use all the four learning modes, there are differences between individuals in preference patterns or 'strengths and weaknesses'. These preferences can be assessed on the basis of the two dimensions. That means that patterns of preferences can be characterized in terms of the following learning styles: diverger (CE preferred to AC, and RO preferred to AE), assimilator (AC preferred to CE, and RO preferred to AE), converger (AC preferred to CE, and AE preferred to RO), accommodator (CE preferred to AC, and AE preferred to RO).

The Kolb model may be considered as a learning stage model (focusing on element 2) or as a learning style model (focusing on elements 3 and 4). Within the framework of this article we are interested in the learning style model and not in the learning stage model. However, the question may be raised whether the learning style model is not conflicting in several aspects with the learning stage model. As indicated by De Ciantis & Kirton ( 1996, p. 809, 810 ), Kolb, in this interpretation of learning style, 'has inadvertently conflated three theoretically unrelated cognitive elements - style, level (abilities or capacity), and process'. We agree with this criticism, and in addition we find insufficient empirical evidence of the existence of the two dimensions mentioned in proposition 3, as will be elaborated upon in the next section. Thus the foundations for Kolb's learning style assessment as described in proposition 4 are lacking. Therefore, we ignore this supposed dimensionality, and restrict ourselves to the first proposition as point of departure for the development of our instrument.

There are several other obstacles one comes across when trying to apply Kolb's model to work-based learning in (particularly) vocational education. One of these problems is that experiential learning is a less isolated phenomenon than suggested by the model. Sensory experience is not the only input the student receives. Since the student acts in an educational context, other inputs will consist of concepts, theories, models and strategies explicitly or implicitly taught by others. And also, other people (like teachers and practicum supervisors) may prescribe (internal and external) activities to be performed. A descriptive model of experiential learning should not just describe the way students go through an 'autistic' learning cycle, but should also provide categories to describe the way students respond to external instructions and prescriptions.

Measuring Work-Based Learning

Based upon the experiential learning model, described above, Kolb developed an instrument to measure learning style: the Learning Style Inventory or LSI ( Kolb, 1976 ), which he later revised ( Kolb, 1985 ). As explained, we needed an instrument to measure the experiential learning constructs concrete experience, reflective observation, abstract conceptualization, and active experimentation. In deviation from Kolb, we wanted to measure these learning modes as they are manifested in vocational field experiences in particular. In addition, we wished to avoid some test theoretical pitfalls identified in evaluation studies on the LSI.

Evaluation of the LSI

We limit our discussion to the 1985 version of the LSI (the LSI-1985). The instrument consists of twelve short statements concerning learning situations, and the respondents are asked to rank-order four sentence endings that are supposed to represent the four learning modes (CE, RO, AC, AE). One of the items for example is 'I learn by feeling /watching /thinking /doing'. In each item the respondent is asked to rank the four sentence endings from 4 to 1, to the extent each mode applies to him. The individual learning style is determined in two steps. First, the four scale scores CE, RO, AC and AE are computed through adding for each learning mode the corresponding rank numbers across the 12 items. In the second step the difference scores (AC - CE) and (AE - RO) are computed. The two difference scores determine the learning style, which can be either 'diverger' (dominance of CE and RO), 'assimilator' (dominance of AC and RO), 'converger' (dominance of AC and AE), or 'accomodator' (dominance of CE and AE). Both the test format (forcing the respondent to choose between four learning modes, and repeating this twelve times) and the graphic representation of the results (in quadrants) are appealing features of the test, which may explain its popularity in the field. The instrument is considered to possess strong face validity and intuitive appeal ( Cornwell, Manfredo & Dunlap, 1991 ; Veres III, Sims & Locklear, 1991 ). The internal-consistency (coefficient alpha) of the scales is reported to be acceptable ( Sims et al., 1986 ; Veres, Sims & Shake, 1987 ; Willcoxson & Prosser, 1996 ). Yet the LSI has been heavily criticized, for a series of reasons, some of a test-theoretical nature, and others of an empirical nature, for example:

• Ambiguity of the test content. As Willcoxson and Prosser ( 1996, p. 248 ) observe, "the accompanying instructions do not specify that the person completing the inventory should think of a given learning context when filling it out. Thus the response of someone focussing upon learning preferences in the context of acquiring driving skills might be quite different from the responses recorded by that same person when focussing upon the study of English Literature in an academic context..." For the aims of our research the context should be specified as vocational work experience.
• Forced ranking of the four learning modes, causing a built-in interdependence of the four learning mode scores. The variables are 'ipsative'. As a consequence of the method of ranking the alternatives, the sum of the CE-, RO-, AC- and AE-scores is constant for each item and each person, namely 1+2+3+4=10. Accordingly, the four scale scores are (linearly) dependent on each other. This built-in dependency of the learning modes obscures the real relation between them. It will cause the correlations between the variables to shift in a negative direction and the artificial and the real relations between the variables cannot be separated ( Loo, 1999, 1996 ; Cornwell & Manfredo, 1994 ; Geiger, Boyle & Pinto, 1992 ).
• Inconclusive evidence of the existence of the two dimensions 'prehension' and 'transformation'. Although Yahya ( 1998 ) does report these dimensions, not all factorial studies find them. Some studies report a different factorial structure, involving the dimensions (AC-AE) and (CE-RO) ( Cornwell, Manfredo & Dunlap, 1991 ; Geiger, Boyle & Pinto, 1992 ). Factor solutions that did support the proposed dimensions accounted only for a small percentage of the variance ( 32. 1%, reported by Loo, 1996 ).

Adaptations to the LSI

There have been several attempts to improve the LSI, in particular its disputed use of multiple ranking. Marshall and Merritt tried out different formats to present the answering alternatives: a (four point) Likert-type normative format ( Merritt & Marshall, 1984 ), and a semantic differential format ( Marshall and Merritt, 1985 ). Their conclusion is that "these results suggest that valid normative forms for the LSI can be developed and that the structure provided through the use of semantic differential format can improve scale internal consistency as compared to that of a Likert-type of normative format" ( Marshall & Merritt, 1985, p. 936 ). Honey and Mumford ( 1982 ) developed a scale with a normative answering format as well, although with only two alternatives per item (agree/ not agree). They also extended the number of items. In our opinion, the normative format has the disadvantage of being very susceptible to acquiescence set and social desirability. We agree with Marshall & Merritt as far as their preference for a semantic differential format is concerned, but we think that presenting the choice between the two modes it is incompatible with the unsupportive empirical results regarding the existence of these two dimensions.

A New Instrument

On the basis of the preceding arguments, a new instrument should be developed to measure the way students learn from work experiences. This is also the opinion of others, who are conducting research into learning strategies in vocational education. Slaats ( 1999 ) asserts: "further research might produce a more suitable questionnaire than Kolb's LSI for measuring learning styles in practical settings" ( p. 177 ). In our opinion, this instrument should focus explicitly upon work-based learning, it should measure clearly identifiable components of this learning, which can manifest themselves in inter-individually different patterns, and it should use a semantic differential format. Theoretically, either of four procedures can be used to develop such an instrument:

1. Start with exploratory phenomenological research, interviewing students about their learning activities in the work experience program, and categorizing their answers. Formulate items, based upon the resulting categories. Determine dimensions by way of factor analysis.
2. Start with a review of empirical literature on work-based learning and experiential learning. Identify variables describing aspects of this type of learning, plus directions for the measurement of these variables. Construct a test battery containing measurements for each of the variables identified. Determine dimensions by way of factor analysis.
3. Start with a general theory of learning (styles). Develop a more specified theory of work-based learning styles, based upon that general theory. Operationalize the concepts constituting the theory.
4. Start with the four learning modes assumed by current experiential learning models. Develop alternative items for each of the four modes, tailored to learning from work experience, and using a semantic differential format.

Although each of these procedures has its merits, we employed procedure number 4, supplemented with procedure number 2. The most creative part of it was the determination of the opposites of each of the learning modes as formulated by Kolb, without involving one of the other modes. When doing this, we were inspired by concepts and dimensions reported in other literature about work-based and experiential learning. Our efforts resulted in the following bipolar constructs (Table 1):

Table 1
Bipolar Constructs Related to Experiential Learning Modes
Experiential learning mode	Construct	First pole	Second pole (opposite)
Concrete experience (CE)	Immersion	Immersed (emotionally/intellectually)	Detached (emotionally/intellectually)
Reflective observation(RO)	Reflection	Insight-oriented (orlearning-oriented)	Results-oriented (or performance-oriented)
Abstract Conceptualization (AC)	Conceptualization	Generic (or strategic)	Idiosyncratic (or pragmatic)
Active experimentation(AE)	Experimentation	Inquiring (or experimenting)	Prescription oriented

The first construct (Immersion: immersed versus detached) resembles the active-reflective dimension found by De Ciantis & Kirton ( 1996 ), in their psychometric reexamination of the four experiential learning modes (using the instrument of Honey & Mumford), with concrete experience at one pole, and reflective observation at the other pole. It is about the amount of distance the student tends to take from the ongoing events. As the literature about on-the-job learning shows, learning by immersion is common practice (see for example McCall, Lombardo & Morrison, 1988 ), although often criticized (see for example Jacobs & Jones, 1995 ).

The second construct (Reflection: insight-oriented versus results-oriented) resembles the goal orientation dimension reported in psychological literature ( Dweck, 1986 ; Porter & Tansky, 1996 ), ranging from performance orientation to mastery or learning orientation. Students with a learning orientation attempt to understand their tasks, and learn from them; students with a performance orientation strive to succeed with little effort, and are satisfied with success, even if they do not understand how it was acquired. Many authors recommend the fostering of a more reflective attitude than often encountered in on-the-job learning (for example Hart-Landesberg, Braunger & Reder, 1992 ).

The third construct (Conceptualization: generic versus idiosyncratic) resembles the theorist-pragmatist dimension found by De Ciantis & Kirton ( 1996, p. 817 ), which "seems to describe the mode of evaluation in decision making, ranging between 'considering many sources, angles and data before making a decision' to 'earlier closure to make a decision more expediently and efficiently'.". . It is about theory-inspired action versus pragmatic action. According to Garrick ( 1998, p. 108 ), students tend to be less interested in truth than in usefulness.

The fourth construct (Experimentation: inquiring versus prescription-oriented) resembles psychological constructs like curiosity ( Reio & Wiswell, 2000 ; Spielberger & Starr, 1994 ) and openness to experience ( Digman, 1990 ). On-the-job training programs differ widely in the extent to which they appeal to either prescription or inquiry ( De Jong & Versloot, 1999 ). In summary, the first and fourth constructs are concerned with depth and width of involvement respectively, and the second and third with post-active reflection and pre-active reflection respectively.

As we observed earlier, the Kolb model does not recognize external social influences; it treats the student as an isolated individual processing individual experiences. Since in vocational education teachers and tutors are considered to exercise at least some influence upon the learning process, we introduced a fifth construct, which we called 'Regulation': the student's preference for either internal regulation or external regulation. This construct resembles constructs like metacognitive activity ( Ford et al., 1998 ) and self-directed learning Confessore & Kops, 1998), and more in particular the regulation strategies 'internal regulation' and external regulation, as described by Vermunt ( 1998 ) and Slaats, Lodewijks & Van der Sanden ( 1999 ), referring to self-initiated strategies versus dependency upon external resources for regulation of the learning process. As observed by Poell ( 1998, p. 5 ), "Work-related learning is usually referred to in terms of the activities of trainers, consultants , or HRD staff", and "Employees are not regarded as crucial learning actors, who have their own theories and interests as to what they should learn, for what purpose, and in what way". Yet, modern work conditions seem to demand 'free agent learners' ( Marsick et al., 2000 ).

Research Problem

Our aim was to construct a first version of an instrument (the Questionnaire Practice oriented Learning or QPL) for measuring learning strategies in vocational work experience. The instrument should be based upon theories and instruments of the experiential learning theoreticians, but overcome the drawbacks mentioned in the previous sections. It had to be an easy-to-administer questionnaire, and scale scores should be calculated by averaging the item scores per scale.

Research question 1, then, is: How well can the learning modes immersion, reflection, conceptualization, experimentation and regulation be measured with the QPL, and how are these modes interrelated?

But these learning modes alone yield no learning styles. Research question 1 is about clustering of items and clustering of variables (scales). Learning styles, on the other hand refer to the clustering of students. In order to determine the usefulness of the questionnaire for the determination of learning styles, we wanted to examine whether the QPL can reveal well interpretable clusters of students with common patterns of learning mode scores. Thus, research question 2 can be formulated as: What work-based learning styles can be found, using the QPL, among students in a modal school for senior secondary education in The Netherlands?

Method

Instrument

For measuring the five constructs representing aspects of work-based learning, a new measuring instrument, the Questionnaire Practice oriented Learning (QPL) was developed. Each QPL-item contains two opposite statements (comparable with a semantic differential), out of which the respondent should make a choice on a scale from 1 to 5. An example of an item in the Conceptualization scale is: 'When I get stuck in a task I consider the possibility to find a solution with help of the theory / When I get stuck in a task the theory is not very helpful for me, because practice is often very different'. After testing a trial version, the first version of the QPL was developed. This first version consisted of sixty-eight items (13 or 14 items per scale).

Procedure

The investigation was carried out in the health care and engineering departments of a school for senior secondary vocational education in The Netherlands. With help of school managers and teachers, students who had finished a work experience program were asked to fill in (anonymously) the QPL while they were back in school for a seminar. They were asked to respond to the items with reference to their most recent work experience program.

Sample

Of the 450 QPL questionnaires, 407 were filled in. The average age of the students was 21.6 years; 105 of them were male and 302 female. The number of students from the health care department was 313, and from the engineering department 94. Students were almost equally divided over the school-based program (217) and the work-based program (190).

Data Analyses

Related to research question 1 we performed classical analyses on items (factor analysis and Cronbach alpha) and on scales (correlations and factor analysis). Related to research question 2 we performed a cluster analysis and a discriminant analysis. Cluster analysis is the appropriate method for identifying subgroups containing individuals with similar attributes ( Wierstra & Beerends, 1996 ; Vermetten, 1999, p. 98/99 ). Discriminant analysis is used to describe and interpret the resulting clusters.

Results

Reliability of QPL Scales

On the sixty-eight QPL items factor analyses (principal factors) and reliability analyses were conducted. These analyses were directed on identification of the five scales. Thus, in the factor analyses five factors were looked for, after a varimax rotation of the first five principal factors. The factor analysis served as a selection procedure for determining which items should be included in a particular scale, after which the Cronbach alpha reliability of the scale was computed (scale score = average of item scores). We used the following strategy for item selection and item validation (convergent and discriminant validation). If an item turned out to be hardly related to the construct at which it was aimed (low loading on the relevant factor and low item-scale correlation), or if an item seems to measure more than one construct, the item was removed. After removal of thirteen items the five scales of Table 2 were generated. The reliability of the scales is not high, but yet suitable for research purposes ( Nunnally, 1978 ).

Table 2
Cronbach Alpha Reliability & Exemplary Items of the QPL Scales (N = 407)
Scale	Cronbach	Sample items
		Positive pole	Negative pole
Immersion (13 items)	.63	When I have to take a decision, I play it by ear. I don't mind if I am expected to act differently from what theory tells.	I consult the theory when I have to make a decision. It disturbs me when I am expected to act differently from what theory teaches me to do.
Reflection (12 items)	.62	When I have finished a task I ask myself: what have I learned from it? In post hoc discussions I try to phrase what went right and what went wrong.	When I have finished a task I go on with the next assignment. In post hoc discussions I seldom try to phrase what went right and what went wrong.
Conceptualization (12 items)	.70	I try to find out how field assignments relate to theory. Before starting with an assignment, I consider what problems I can expect.	I don't bother, because practice is too remote from theory. I just start with an assignment; in the process of working at it I find out what may go wrong.
Experimenta-tion (8 items)	.69	I like to get the freedom to try out how I can best handle thing. I like to make decisions about my way of doing things.	I prefer well delineated assignments, so I know what is expected from me. I appreciate to get indicated what I might better do or leave off.
Regulation (10 items)	.64	If I don't understand something, I try to find an explanation by myself. After I finish an assignment, I can judge by myself whether I did it right.	I often ask my supervisor explanations about things that I do not understand. I really know whether I did it right only after my supervisor says so.

Correlations between the QPL Scales

The correlations between the QPL scales are given in Table 3. This table shows that the divergent validity of the scales is satisfactory.

Table 3
Correlations between the QPL Scales (N = 407)
	Immersion	Reflection Conceptualization	Experimentation	Regulation
Immersion		-.06	-.30***	.00	-.15*
Reflection			.51***	.07	.23***
Conceptualization				.07	.23***
Experimentation					.50***
*: p<.05; ** p<.01; ***p<.001(two-sided).

The significant negative correlation between Conceptualization and Immersion (-.30) and the relatively high positive correlation between Conceptualization and Reflection (.51) are quite understandable. The same is true for the relatively high correlation between Experimentation and Regulation (.50). The relations between the five scales may be summarized by the results in Table 4 of a principal factor analysis on the scales, followed by a varimax rotation.

Table 4
Factor Loadings of five QPL Scales on three Varimax Rotated Factors
	Factor 1	Factor 2	Factor 3
Immersion	-.08	.01	.97
Reflection	.90	.09	.11
Conceptualization	.81	.07	-.32
Experimentation	-.03	.89	.07
Regulation	.21	.83	-.14
eigenvalue	1.90	1.26	.95
% explained variance	38.00	25.20	10.02

On the first factor, high loadings are found for Reflection and Conceptualization. Experimentation and Self Regulation have high loadings on the second factor. On the third factor only Immersion shows a high loading.

Cluster Analysis

Whereas factor analysis results in clustering of variables, cluster analysis groups cases (i.e. persons). In order to find learning styles in the practical phase a cluster analysis (K-means cluster program of SPSS 7.5 for Windows) on the sample of 407 students was conducted, on the basis of the five QPL scale scores. We tried several numbers of clusters, with a maximum of four. The best solution (in terms of discrimination and interpretation) was found for three clusters consisting of 117, 176 and 114 students, respectively. For interpreting the clusters we used discriminant analysis.

Discriminant Analysis

On the three clusters a discriminant analysis with two discriminant functions was conducted. The discriminant analysis is used for descriptive purposes; our only purpose is to be able to interpret the differences found between the clusters. A discriminant function is (like a factor in factor analysis) a latent variable, this is a particular weighed addition of the manifest variables, in this case the five scale scores. The weights of the scale scores are determined by the computer program in such a way that the new overarching variable - the latent variable - differentiates maximally between the clusters, while within a cluster the differences between the students on the discriminant function are minimal. The two discriminant functions turned out significant (however, significance of these functions is not our main concern, but the descriptive power of the functions). Table 5 summarizes the results of the discriminant analysis. The structure coefficients depicted in table 5 are the correlations (pooled within groups) between the scale scores and the discriminant functions.

Table 5
Structure Coefficients, Eigenvalues and % Explained Variance for two Discriminant Functions
	Function 1	Function 2
Immersion	-.39	.55
Reflection	.47	.13
Conceptualization	.67	-.10
Experimentation	.18	.71
Regulation	.31	.58
eigenvalue	2.07	.55
% explained variance	80%	20%

On the basis of the two discriminant functions, 97.1% of the students could be classified correctly in one of the three clusters. The first discriminant function can be interpreted as Analysis: taking distance, conceptualize, reflect. The second discriminant function can be interpreted as Initiative: immersion, experimentation, regulation. The cluster means of the discriminant functions are depicted in Table 6.

Table 6
Cluster Means ('Centroïds') of two Discriminant Functions
Clusters	Discriminant function 1 (Analysis)	Discriminant function 2 (Initiative)
1	-1.83	-.68
2	-.05	-.85
3	1.97	.61

It appears from table 6 that the first discriminant function (Analysis) describes a contrast between all three clusters, but especially between the clusters 1 and 3. On the second discriminant function (Initiative) especially type 3 distinguishes itself from type 1 and 2.

Scores of the Three Clusters on the QPL Scales

In Figure 1 the profiles of the three clusters on the five scales are shown. On the x-axis the scales are represented and on the y-axis the mean item score for each scale.

Figure 1
Profiles of QPL-learning style clusters

For each of the five scale scores we tested by means of an ANOVA whether the three cluster means differed significantly. As Table 7 shows all F-ratios are significant.

Table 7
Means and Standard Deviations of QPL Scales for three Clusters; ANOVA F-, and p-Values
	cluster 1	cluster 2	cluster 3	F
Immersion	3.90 (.33)	3.34 (.37)	3.34 (.38)	98.31***
Reflection	3.09 (.40)	3.34 (.43)	3.80 (.35)	93.15***
Conceptualization	2.90 (.35)	3.40 (.36)	3.88 (.42)	138.88***
Experimentation	3.10 (.59)	2.66 (.53)	3.47 (.59)	72.30***
Regulation	2.85 (.45)	2.70 (.46)	3.38 (.47)	78.18***
N	117	176	114
* p<.05; **p<.01; ***p<.001(two-sided).

Next we investigated by means of Tukey's HSD-test whether pairs of clusters differ significantly. The following results were found:

• Immersion: Cluster 1 scores significantly higher than the clusters 2 and 3.
• Reflection: Cluster 1 scores significantly lower than the clusters 2 and 3. Cluster 3 scores significantly higher than the clusters 1 and 2.
• Conceptualization: Cluster 1 scores significantly lower than the clusters 2 and 3. Cluster 3 scores significantly higher than the clusters 1 and 2.
• Experimentation: Cluster 2 scores significantly lower than the clusters 1 and 3. Cluster 3 scores significantly higher than the clusters 1 and 2.
• Regulation: Cluster 3 scores significantly higher than the clusters 1 and 2. Cluster 2 scores significantly lower than the clusters 1 and 3.

Work-Based Learning Styles

On the basis of the preceding information (results of cluster analysis, discriminant analysis, ANOVA and Figure 1 ) the three learning style clusters for practical learning could be described as follows:

Cluster 1: Focused on Doing with Incidental Learning (117 students). Students with this learning style do not learn very intentionally. The student aims largely at being busy during the practical work and at immersing in experiences. They approach the practical tasks on the basis of intuition and feeling and they hardly relate these to the school theory. Because they don't conceptualize nor reflect much, their experiences are not integrated with theoretical knowledge.

Cluster 2: Learning on the basis of External Regulation (176 students). The learning activities of students with this learning style are in the practical phase mainly externally regulated. Reflecting and relating practical activities to school theory occurs on a moderate level.

Cluster 3: Self-regulated Learning on the basis of Theory and Reflection (114 students). Students with this learning style incorporate the practical work experiences into a theoretical framework. During the practical work they regulate their learning activities themselves, using theoretical concepts learned at school and concepts formed by themselves. They reflect much and they prefer learning by 'trying out', not in the sense of 'trial and error', but by a thoughtful way of experimenting.

Conclusion

We tested a first version of a new instrument, the Questionnaire Practice oriented Learning (QPL), using a sample of 407 students in senior secondary vocational education. The QPL measures five learning modes derived from literature on work based and experiential learning: immersion, reflection, conceptualization, experimentation, and regulation. Its reliability turns out to be modest but sufficient for research purposes (scales Cronbach alpha between .62 and .70). The scales Reflection and Conceptualization are positively correlated, as is also the case for the scales Experimentation and Regulation. Accordingly, two or three factors ('Analysis', 'Initiative', and 'Immersion') cover most of the variance in the data. Cluster analysis results in three well interpretable clusters, representing three distinct learning styles. Two of these learning styles seem to be opposites, namely 'Self-regulated Learning on the basis of Theory and Reflection' and 'Focused on Doing with Incidental Learning' (for this last learning style the work activities are largely intentional, but the learning is largely 'incidental'). The third learning style (denoting the largest group of students), is characterized by dependence of External Regulation, and takes an intermediate position with regard to the other learning style elements. These are clearly interpretable learning styles, which corroborates the validity of the constructs we developed to describe work-based learning.

These results may be integrated in a theory-of-work-based-learning, which should be further developed on the basis of both theoretical and empirical future studies. A first next step is an investigation of the relations between aspects of work-based learning as measured by the QPL and learning strategies for school learning. Another next step is the investigation of the relations of QPL scores with study discipline and type of work experience program. Also, the QPL may be improved by developing new items related to the constructs measured (thus attempting to raise the scale Cronbach alpha reliabilities), and maybe also by adding new constructs based upon a further study of the literature on work-based learning. A thus improved test can be of good use as a diagnostic tool in schools for vocational education. It may also serve as a measurement instrument in research studies which should uncover how the learning effects of field experiences depend upon the interaction of student learning styles and characteristics of the work place as a learning environment.

References

Biggs , J. B. (1987). Student approaches to learning and studying Hawthorn , Victoria: Australian Council for Educational Research.

Boekaerts , M. (1997). Self-regulated learning: a new concept embraced by researchers, policy makers, educators, teachers, and students. Learning and Instructio , 7 (2), 161-186.

Confessore , S. J., & Kops, W. J. (1998). Self-directed learning and the learning organization: examining the connection between the individual and the learning environment. Human Resource Development Quarterly , 9 (4), 365-375.

Cornwell , J. M., Manfredo, P. A., & Dunlap, W. P. (1991). Factor analysis of the 1985 revision of Kolb's Learning Style Inventory. Educational and Psychological Measurement , 51 (2), 455-462.

Cornwell , J. M., & Manfredo, P. A. (1994). Kolb's learning styletheory revisited. Educational and Psychological Measurement , 54 (2), 317-327.

De Ciantis, S. M., & Kirton, M. J. (1996) A psychometric reexamination of Kolb's experiential learning cycle construct: a separation of level, style and process. Educational and Psychological Measurement , 56 (5), 809-820.

De Jong, J. A., & Versloot, A. M. (1999). Job Instruction: its premises and its alternatives. Human Resource Development International , 2 (4), 391-404.

Digman , J. M. (1990). Personality structure: emergence of the five-factor model. Annual Review of Psychology , 41, 417-440.

Dweck , C. S. (1986). Motivational processes affecting learning. American Psychologist , 41 (10), 1040-1048.

Ford , J. K., Smith, E. M., Weissbein, D. A., & Gully, S. M.(1998). Relationships of goal orientation, metacognitive activity, and practice strategies with learning outcomes and transfer. Journal of Applied Psychology , 83 (2), 218-233.

Garrick , J. (1998). Informal learning in the workplace: unmasking human resource development . London/New York: Routledge.

Geiger , M. A., Boyle, E. J., & Pinto, J. (1992). A factor analysis of Kolb's revised Learning Style Inventory. Educational and Psychological Measurement , 52, 753-759.

Hart -Landesberg, S., Braunger, J., & Reder, S. (1992). Learning the ropes: the social construction of work-based learning . Berkeley, CA: National Center for Research in Vocational Education.

Honey , P., & Mumford, A. (1982). The manual of learning styles . Maidenhead: Peter Honey.

Jacobs , R. L., & Jones, M. J. (1995). Structured on-the-job training: unleashing employee expertise in the workplace . San Francisco: Berrett-Koehler.

Kolb , D. A. (1976). Learning style inventory: technical manual . Boston: McBer & Company

Kolb , D. A. (1984). Experiential learning: experience as the source of learning and development . Englewood Cliffs, NJ: Prentice Hall.

Kolb , D. A. (1985). Learning Style Inventory 1985: Technical Manual. Boston: McBer & Company.

Loo , R. (1996). Construct validity and classification stability of the revised learning style inventory. Educational and Psychological Measurement , 56 (3), 529-536.

Loo , R. (1999). Confirmatory factor analysis of Kolb's Learning Style Inventory (LSI-1985). British Journal of Educational Psychology , 69 (2), 213-219.

Marshall , J. C., & Merritt, S. L. (1985). Reliability and construct validity of alternate forms of the Learning Style Inventory. Educational and Psychological Measurement , 45 (2), 931-937.

Marsick , V. J., Brooks, A., Cseh, M., Keelor Lovin, B., Vernon,S., Watkins, K. E., & Ziegler, M. (2000). Meeting the informal learning challenges of the free agent learner: drawing insights from research-based lessons learned. In K. P. Kuchinke (Ed.), Proceedings 2000 AHRD Conference Raleigh-Durham , NC, March 8-12, 2000, 149-165.

Marton , F, Hounsel, D., & Entwistle, N. (Eds.) (1988). The experience of learning . Edinburgh: Scottish Academic Press Ltd.

McCall , M. W. Jr., Lombardo, M., & Morrison, A. M. (1988). The lessons of experience . Lexington (Mass.): Heath & Company.

Merritt , S. L., & Marshall, J. C. (1984). Reliability and construct validity of ipsative and normative forms of the Learning Style Inventory. Educational and Psychological Measurement , 44 (2), 463-472.

Nunnally , J. C. (1978). Psychometric Theory . New York: MCGraw Hill.

Poell , R. F. (1998). Organizing work-related learning projects: a network approach (doctoral thesis). Nijmegen (NL): Rob Poell.

Porter , G., & Tansky, J. W. (1996). Learning orientation of employees: moving toward organization-based assessment. Human Resource Development Quarterly , 7 (2), 165-178.

Reio R. T. G., & Wiswell, A. (2000). Field investigation of the relationship among adult curiosity, workplace learning, and jobperformance. Human Resource Development Quarterly , 11 (1), 5-30.

Schmeck , R. R. (1988). Learning strategies and learning styles . New York: Plenum Press.

Sims , R. R., Veres, J. G., Watson, P., & Buckner, K. (1986). The reliability and classification stability of the Learning Style Inventory. Educational and Psychological Measurement , 46 (2), 753-761.

Slaats , J. A. M. H. (1999). Reproduceren & construeren: leerstijlen van leerlingen in het middelbaar beroepsonderwijs (Reproducing & constructing: learning styles of students in senior secondary vocational education) . Doctoral dissertation. Tilburg NL: Katholieke Universiteit Brabant.

Slaats , A., Lodewijks, H. G. L. C., & Van der Sanden, J. M. M. (1999). Learning styles in secondary vocational education: disciplinary differences. Learning and Instruction , 9 (5), 475-492.

Spielberger , C. D., & Starr, L. M. (1994). Curiosity and exploratory behavior. In H. F. O'Neil Jr., & M. Dillings (Eds.), Motivation: theory and research (pp. 221-243). Hillsdale, NJ: Erlbaum.

Veres , J. G., Sims, R. R., & Shake, L. G. (1987). The reliability and classification stability of the Learning Style Inventory in corporate settings. Educational and Psychological Measurement , 47 (2), 1127-1133.

Veres III, J. G., Sims, R. R., & Locklear, T. S. (1991). Improving the reliability of Kolb's Revised Learning Style Inventory. Educational and Psychological Measurement , 51 (1), 143-150.

Vermetten , Y. J. M. (1999). Consistency and variability of student learning in higher education (doctoral thesis) . Tilburg NL: Tilburg University.

Vermunt , J. D. H. M. (1998). The regulation of constructive learning processes. British Journal of Educational Psychology , 68 (1), 149-171.

Vermunt , J. D., & Verloop, N. (1999). Congruence and friction between learning and teaching. Learning and Instruction , 9 (3), 257-280.

Wierstra , R. F. A., & Beerends, E. P. M. (1996). Leeromgevingspercepties en leerstrategieën van eerstejaars studenten sociale wetenschappen (Learning environment perceptions and learning strategies of freshman social science students). Tijdschrift voor Onderwijsresearch , 21 (4), 306-322.

Wierstra , R. F. A., Kanselaar, G., Van der Linden, J. L., Lodewijks, H. G., & Vermunt, J. D. (2000). The impact of the university context on European students' learning strategies and learning environment preferences . Manuscript submitted for Higher Education.

Willcoxson , L., & Prosser, M. (1996). Kolb's Learning Style Inventory (1985): review and further study of validity and reliability. British Journal of Educational Psychology , 66 (2), 247-257.

Yahya , I. (1998). Willcoxson and Prosser's factor analyses on Kolb's (1985) LSI data: reflections and re-analyses. British Journal of Educational Psychology , 68 (2), 281-286.

Authors

JOSË HERMANUSSEN, Regional Center for Vocational Education, Wandelboslaan 30, 5042 PD Tilburg, The Netherlands. [E-Mail: Jhermanussen@rocmb.nl ], tel. 31-13-5904666. As of August, 2001, José Hermanussen will be Researcher at the Institute of Applied Social Sciences (ITS) of the Catholic University of Nijmegen (KUN) in Nijmegen. Her research focuses on vocational education.

RONNY F. A. WIERSTRA, Assistant Professor, School of Educational Sciences, Utrecht University, P.O. Box 80.140, 3508 TC Utrecht, The Netherlands. [E-Mail: R.Wierstra@fss.uu.nl ], tel. 31-30-2534799. Dr. Wierstra's research interests are learning environments and learning strategies.

JAN A. DE JONG, Assistant Professor, School of Educational Sciences, Utrecht University P.O. Box 80.140, 3508 TC Utrecht, The Netherlands. [E-Mail: J.dejong@fss.uu.nl ], tel. 31-30-6990588. Dr. de Jong's research is in work-based learning.

JO G. L. THIJSSEN, Professor, School of Educational Sciences, Utrecht University, P.O. Box 80.140, 3508 TC Utrecht, The Netherlands.[E-Mail: j.g.l.thijssen@tref.nl ], tel. 31-30-2534940. Dr. Thijssen's research focuses on human resource development.