LEAIND-01120; No of Pages 4

Learning and Individual Differences xxx (2015) xxxxxx

Contents lists available at ScienceDirect

Learning and Individual Differences

journal homepage: www.elsevier.com/locate/lindif

Cognitive styles and performance on complex, structured tasks

yvind L. Martinsen a,, Adrian Furnham b
Norwegian Business School, Oslo, Norway
University College London, London, UK

a r t i c l e

i n f o

Article history:
Received 19 May 2014
Received in revised form 3 April 2015
Accepted 24 July 2015
Available online xxxx
Keywords:
Cognitive style
Assimilator
Explorer
Structured tasks
Motivation

a b s t r a c t
This study looks at the conditional relationship between the AssimilatorExplorer (AE) cognitive styles and performance on complex, structured tasks. We predicted that the achievement motive should moderate the style
performance relationship. Eighty-three participants with mean age 18.5 years completed a cognitive style test,
a measure of the achievement motive, and a uid intelligence test (control variable). The dependent variable
was scored on a remote association test (RAT) with items based on complex word pairs. Results showed that Explorers performed better when they scored higher on the achievement motive. Assimilators performed better
when they scored lower on the achievement motive. The idea of optimal motivation was supported by the pattern of interaction.
2015 Elsevier Inc. All rights reserved.

1. Introduction
The AssimilatorExplorer styles (AE styles) describe individual differences in cognitive strategy preferences. The AE styles can be seen as
an important moderator in research on creativity and problem solving
and the purpose of the present paper is to further investigate how
these styles may moderate problem solving performance. The theory
posits that the strategy preferences associated with the AE styles describe competence implications for different types of tasks along a structuredunstructured (high novelty) dimension (Kaufmann, 1979). More
recent positions argue that the task competence described by the AE
styles will be moderated by the achievement motive (Martinsen,
1994). However, while previous studies have emphasized performance
on complex, unstructured tasks, we presently focus on complex, structured tasks. This represents a further test of the interactive nature of
the AE styles.
The theory of the AssimilatorExplorer styles was proposed by
Kaufmann (1979, 1995) and later elaborated by Martinsen (1993,
1994, 1995a,b). It has been placed in the WholistAnalyst category of
style constructs (Riding & Rayner, 1998). Assimilators are more rulebound in problem solving behavior and inclined to interpret new events
in terms of existing knowledge. Explorers seek novelty, which manifests
as a search for new types of solutions and new ways of solving problems
even without external pressure to do so.
The AE style construct has been operationalized as a continuum
where higher scores describe Explorers and lower scores describe
Corresponding author at: Department for Leadership and Organizational Behavior,
Norwegian Business School, Nydalsveien 37, 0484 Oslo, Norway.
E-mail address: oyvind.martinsen@bi.no (.L. Martinsen).

Assimilators. The measure of AE styles (Kaufmann & Martinsen,

1992) includes three correlated sub-factors that dene the secondorder AE style construct. (Martinsen & Diseth, 2011). The main AE
style construct has several correlations with measures of personality,
where the Explorer end of the continuum is associated with lower
scores on Neuroticism, Agreeableness, and Conscientiousness, and
higher scores on Extroversion and Openness compared with the Assimilator end of the continuum (Martinsen & Diseth, 2011). Moreover, the
AE continuum is not correlated with general intelligence (Martinsen
& Kaufmann, 2000). In the latter study, the AE style dimension also
correlated signicantly with a motivational factor. Signicant differences between students have been found in different types of educational settings; art students show the highest scores (more
explorative). A signicant correlation between the AE styles and scores
on a creative activities checklist (Explorers having higher creativity
scores) also exists, but there is only a weak, yet still signicant, correlation between these styles and measures of verbal and gural uency
(Explorers have higher divergent thinking scores). Taken together, the
correspondence between the AE theory and the pattern of empirical
ndings hitherto is consistent with the basic denitions of cognitive
style. This implies a strong relationship between personality and style,
no or low correlations with abilities, and differential implications for
performance (Martinsen, Kaufmann, & Furnham, 2011).
Previous studies have shown that the task-specic strategies described by the AE styles do not necessarily lead to superior performance. The degree of compatibility between stylistic dispositions and
task characteristics affect performance mainly in combination with
other inuences. Interactive effects between stylistic dispositions, degree of task-relevant experience (Martinsen, 1993, 1995a), cues in the
situation, such as solution hints (Martinsen, Furnham, & Hrem in

http://dx.doi.org/10.1016/j.lindif.2015.07.013
1041-6080/ 2015 Elsevier Inc. All rights reserved.

Please cite this article as: Martinsen, .L., & Furnham, A., Cognitive styles and performance on complex, structured tasks, Learning and Individual
Differences (2015), http://dx.doi.org/10.1016/j.lindif.2015.07.013

.L. Martinsen, A. Furnham / Learning and Individual Differences xxx (2015) xxxxxx

preparation), strength of the achievement motive (Martinsen, 1994), as

well as positive mood (Kaufmann & Martinsen, 2006) together may inuence performance. Several aspects of this dynamic interplay may be
explained by a theory of optimal motivation for the task depending on
task complexity (Atkinson, 1974; Yerkes & Dodson, 1908) and where
also the type of task (structured vs. unstructured), position on the AE
style inventory, and the strength of the achievement motive (or other
positive affect arousing conditions) together determine the quality of
performance. In Atkinson's (1974) account of achievement motivation,
resultant task motivation is mainly a function of perceived competence
for the task and the strength of the achievement motives. The performance effect of resultant task motivation, however, is posited to be dependent on the structure or complexity of the task. On complex
problem solving tasks the prediction is that performance should be impaired when task competence is high and achievement motivation is
also high. In our study we use the AE styles as a proxy for task competence and this is further outlined below.
In previous studies we have emphasized performance on unstructured tasks. Thus, in the present context, we found it important to further investigate the styleperformance relationship on tasks that favor
Assimilators' strategy dispositions. It seems that on complex tasks
with solution cues, which were seen as inuential search constraints
by Kaplan and Simon (1990), Assimilators should perform better at
the outset. However, according to the theory above, this styleperformance relationship should also be moderated by the strength of the
achievement motive.
Previous studies (Martinsen, 1994) have used classic insight problems that can be described as unstructured, ill dened and difcult because they demand restructuring. A related type of task can be
constructed based on principles from the Remote Associates Test
(RAT; Mednick, 1962). Such tasks have been used as insight problems
in the line of research by Bowden, Jung-Beeman, Fleck, and Kounios
(2005). Items in RAT present the subject with three stimuli words that
are supposed to elicit one another remotely associated word. Such a
task presents the individual with a large problem space, especially if
one or several word associations are uncommon (remote). In such
cases, the task would favor the Explorer style. However, these tasks
may also be constructed so they favor the Assimilator type of strategy
disposition. Using complex word pairs, rather than distantly associated
words, can moderate the structure of such tasks since word pair-based
tasks implicitly present the problem-solver with a rule to follow. Such
a rule can be made salient through task instructions or by presenting
sample items with solutions. When word pairs are used as the basis
for RAT items, the task should theoretically align with Assimilators'
competence rather than Explorers' competence because they include a
general problem-solving rule. Thus a constraint may limit search in
large problem spaces, but theory argues that people with different styles
are differently inclined to utilize such constraints. Based on this, we expect that Assimilators should be more competent on the present RAT
tasks, but also that strength of the achievement motive should moderate the performance of styleperformance relationships. Compared
with previous studies (Martinsen, 1994), we expect a reverse pattern
of interaction and put forward the following hypothesis:
H1. The relationship between the AE styles and performance on structured, complex RAT tasks should be moderated by scores on the
achievement motive.
The pattern of the posited interaction should indicate that Assimilators perform better when scores on the Motive to Approach Success are
in the lower range and worse when the scores on the Motive to Approach Success are in the higher range. Explorers should perform better
when scores on the Motive to Approach Success are in the higher range
and worse when the scores are in the lower range.
Finally, because some style constructs were contaminated by intelligence (Martinsen, 1997) in previous style research, it is important to

control for intelligence to provide evidence for the validity of style constructs. We included a measure of uid intelligence, which has been
found to have a high loading on the G-factor (Undheim, 1981).
2. Method
2.1. Sample
Eighty-three students (54 males and 27 females) from a Norwegian
upper high school participated. Their mean age was 18.5 years. Upper
high school in Norway lasts three years and includes students between
the ages of 1617 and 1819 years.
2.2. Measures
2.2.1. AE inventory
The revised AE inventory (Kaufmann & Martinsen, 1992) was used.
The scale is continuous and Explorers have higher scores and Assimilators lower scores on the inventory. Each item has a ve-point response
scale and the present version of the inventory has 30 items. The inventory has been validated in previous studies (Martinsen & Diseth, 2011;
Martinsen & Kaufmann, 2000). Alpha for the AE inventory was .91 in
the present study.

2.2.2. Achievement motive

A short form of the Achievement Motives Scale (AMS; Gjesme &
Nygrd, 1970; Nygrd & Gjesme, 1973; Lang & Fries, 2006) was used.
The present version of this scale (Nygrd, 1997 [personal communication]) includes a brief vignette where the participant is asked to think
of a problem-solving situation in which there is uncertainty about the
outcome. Following this, the participant was asked to respond to 20 adjectives describing achievement affects. Ten adjectives described positive affects (e.g., engaging) that indicated motivation to achieve
(motive for success: Ms), and 10 adjectives described negative affects
(e.g., unpleasant) that indicated motivation to avoid failure (motive to
avoid failure: Mf). We used a ve-point response scale for each item
and two summed scores. Cronbach's alpha reliabilities were .87 for Ms
and .90 for Mf. We presently use Mf as a control variable based on previous studies (Martinsen, 1994).
2.2.3. Verbal analogies
A brief and standard test of verbal analogies was taken from
Mnnesland (1985) and used as a measure of uid intelligence. This
test has 20 items and participants were given 6 min to complete the test.

2.2.4. Remote Associates Test (RAT)

Fifteen items were constructed using the procedure described by
Bowden et al. (2005) but adapted so that word pairs became the main
basis for item construction. Since translations from Norwegian to English may not be illustrative, we cite examples provided by Bowden
et al. (2005, p. 324). These include French, car, shoe (correctly associated word: horn), and Boot, summer, ground (correctly associated
word: camp). Participants were given 20 min on this test, and its
alpha reliability was .74.
2.3. Procedure
Participants were encouraged to volunteer by the school staff. They
were told that they were participating in research on problem-solving
and were debriefed after the study was completed, which took
60 min. Participants rst completed the AE inventory, the short AMS
scale, the verbal analogies test, and nally the present version of RAT.

Please cite this article as: Martinsen, .L., & Furnham, A., Cognitive styles and performance on complex, structured tasks, Learning and Individual
Differences (2015), http://dx.doi.org/10.1016/j.lindif.2015.07.013

.L. Martinsen, A. Furnham / Learning and Individual Differences xxx (2015) xxxxxx

Table 1
Correlations, means, and standard deviations between predictors, control variables, and product terms. For gender: males = 1 and females = 2.

1. Gender
2. Verbal analogies
3. Mf
4. Ms (C)
5. AE (C)
6. Ms (C) Ms (C)
7. Ms (C) AE (C)
8. RAT

Mean

Std
Dev

10.57
23.82
.00
.01
39.73
14.04
4.59

4.47
7.96
6.34
12.48
40.84
80.24
2.66

.189
.081
.043
.040
.010
.020
.281

.106
.103
.246
.024
.074
.154

.598
.192
.177
.019
.056

.183
.087
.105
.191

.084
.232
.155

.372
.292

.122

Note. (C) centered variables.

Correlation is signicant at the 0.05 level (2-tailed).
Correlation is signicant at the 0.01 level (2-tailed).

3. Results
As can be seen in Table 1, there are negative correlations between A
E, verbal analogies, and RAT, however, only the rst of these is signicant. The correlations between the AE styles and the two motives are
in the lower range and insignicant, but similar to the results of previous studies (Martinsen, 1994).
In order to test the posited (H1) interactive effect between the AE
styles and Ms on RAT, scores on AE and Ms were centered and multiplied according to the procedure outlined by Aiken and West (1991).
We used a hierarchical regression analysis, which emphasizes the signicance of the R square increment for the product term as evidence
for interaction. Two persons had missing data on one of the included
variables. In these analyses, the initial test of the Ms by AE interactive
effect was not signicant. The data were therefore investigated further
and revealed a signicant inverted U relationship between Ms and
RAT. Such a curvilinear effect can in itself support the hypothesis of optimal motivation for the task (Yerkes & Dodson, 1908), where complex
tasks are better solved with moderate motive strength.
When controlling for this curvilinear relationship in our further
analyses, the postulated joint effect between the AE styles and Ms
was clearly signicant and in support of our hypothesis. The results
can be seen in Table 2.
As shown in Table 2, the interactive effect of the AE styles and Ms
explains 4.2% of the variance and the full model explains nearly 20% of
the variance in the dependent variable.
When inserting our control variables for gender, verbal analogies,
and Mf rst in the equation, both the Ms Ms curvilinear effect and
the Ms AE interactive effects were still signicant. The increment
in R2, because of Ms AE, was now 5.5% (p = .016). The full model
(gender, verbal analogies, Mf, AE, Ms, Ms Ms, Ms AE) explained
33% of the variance (27% adjusted) in the RAT scores.

Table 2
Summary of hierarchical regression analysis to test the interaction between the AE styles
and the Motive to Approach Success (MS) on the Remote Associates Test (RAT). AE and
Ms were centered. N = 81.

Ms
AE
Ms * Ms
Ms * AE
Rsq
F
Rsq
F
Note:
p b .05.
p b .01.

Step 1

Step 2

Step 3

.167
.132

.191
.154
.322

.182
.110
.402
.227
.042
4.06
.198
4.74

.053
2.22

.102
9.44
.155
4.79

The pattern of the posited joint effect is shown in Fig. 1. Here it is evident that the shape of Ms is well tted to an inverted U at various levels
of the AE styles. It is also clear that Assimilators perform better at lower
levels of Ms and Explorers perform better at higher levels of Ms.
To further test the interaction, we did simple slope analyses. A significant interaction term in a two-way interaction design signies a difference between the slopes for a predictor at different levels of a
moderator. In the present case, this means that the slopes for Ms were
signicantly different at higher and lower levels of the AE strategy dispositions. When testing the simple slopes for the Ms at one standard deviation below the mean on the AE scale, following the recommended
procedure by Aiken and West (1991), the simple slope for Ms was negative and signicant (simple slope = .17, SE = 4.77, t (77) = 2.88, p =
.002 (one tailed)). The simple slope for Ms at one standard deviation
above the mean on the AE scale was, however, not signicant. When
including the control variables: verbal analogies, gender, and Mf, the
simple slopes for Ms were signicant at two standard deviations
above the mean (explorer strategy dispositions) on the AE inventory
(simple slope = .16, SE = .091, t (73) = 1.79, p = .037 (one tailed)).
The simple slopes for Ms were negative and signicant at one standard
deviation below the mean (assimilator strategy dispositions) on the AE
inventory. These results imply that arousal of Ms generally impaired
performance for Assimilators, while only high-scoring Explorers improved their performance because of strong Ms scores.

4. Discussion
The present theory posits that the performance on complex tasks
should be inuenced by the existence of search constraints (task structure), general heuristics associated with cognitive style, and motive
arousal. Within this framework our study was constructed to expand
upon previous ndings by Martinsen (1994), where insight tasks without salient search constraints were included. In light of the theory, those
tasks clearly favored participants with the Explorer style. In contrast to
these studies, the present study included adapted RAT tasks designed
to favor the heuristic competence of Assimilators through providing
cues in the problem situation and in the problem instructions.
As posited, the results showed a reversed pattern of interaction compared to what we have found in our previous studies and were in line
with the theory's prediction. Both Explorers and Assimilators tended
to perform well on the dependent tasks, but their success depended
on their scores on the motive to approach success. Consequently, our
idea of optimal motivation as dened by task competence and motive
strength in interplay was supported by the pattern of interaction. In
order to detect the predicted interaction, it was necessary to control
for a curvilinear relationship between Ms and task performance. This
inverted U relationship can clearly be attributed to mechanisms related
to optimal arousal dating back to Yerkes and Dodson (1908).

Please cite this article as: Martinsen, .L., & Furnham, A., Cognitive styles and performance on complex, structured tasks, Learning and Individual
Differences (2015), http://dx.doi.org/10.1016/j.lindif.2015.07.013

.L. Martinsen, A. Furnham / Learning and Individual Differences xxx (2015) xxxxxx

Fig. 1. Three-dimensional plot of the interaction between the AE styles and the Motive to
Approach Success (Ms). Scores on the vertical (z-) axis represent scores on the dependent
variable while scores on Ms are located on the x-axis and the scores on AE are located on
the y-axis.

4.1. Conclusion
The present study provided further evidence that cognitive style
may be a valid predictor of task performance but only in a complex interplay with other inuences.

References
Aiken, L. S., & West, S. G. (1991). Multiple regression: Testing and interpreting interactions.
London: Sage.
Atkinson, J. W. (1974). Strength of motivation and efciency of performance. In J. W.
Atkinson, & J. O. Raynor (Eds.), Motivation and achievement (pp. 193218). New
York: Wiley.
Bowden, E. M., Jung-Beeman, M., Fleck, J., & Kounios, J. (2005). New approaches to
demystifying insight. Trends in Cognitive Science, 9, 322328.

Gjesme, T. & Nygrd, R. (1970). Achievement-related motives: Theoretical considerations

and construction of a measuring instrument. Unpublished manuscript. Educational
Research Institute, University of Oslo, Norway.
Kaplan, C. K., & Simon, H. A. (1990). In search of insight. Cognitive Psychology, 22, 374419.
Kaufmann, G. (1979). The Explorer and the Assimilator: A cognitive style distinction and
its potential implications for innovative problem solving. Scandinavian Journal of
Educational Research, 23, 101108.
Kaufmann, G. (1995). A theory of cognitive strategy preferences in problem solving. In G.
Kaufmann, K. H. Teigen, & T. Helstrup (Eds.), Problem solving and cognitive processes.
Essays in honor of Kjell Raaheim. (pp. 4576). Bergen: Fagbokforlaget.
Kaufmann, G., & Martinsen, . (1992). The AE scale revised. Bergen: University of Bergen,
Department of General Psychology.
Kaufmann, G., & Martinsen, . (2006). Style, strategy, and mood. Paper presented at the
47th Annual Meeting for the Psychonomic Society. Houston, Texas.
Lang, J. W. B., & Fries, S. (2006). A revised 10-item version of the Achievement Motives
Scale. Psychometric properties in German-speaking samples. European Journal of Psychological Assessment, 22, 216224.
Martinsen, . L. (1993). Insight problems revisited: The inuence of cognitive styles and
experience on creative problem solving. Creativity Research Journal, 6, 435449.
Martinsen, . L. (1994). The effect of individual differences in cognitive style and motives
in solving insight problems. Scandinavian Journal of Educational Research, 38, 8396.
Martinsen, . L. (1995a). Cognitive styles and experience in solving insight problems:
Replication and extension. Creativity Research Journal, 8, 291298.
Martinsen, . L. (1995b). Insight with style: Cognitive style and its function in solving insight problems. In G. Kaufmann, K. H. Teigen, & T. Helstrup (Eds.), Problem solving and
cognitive processes. Essays in honor of Kjell Raaheim. (pp. 77119). Bergen:
Fagbokforlaget.
Martinsen, . L. (1997). The construct of cognitive style and its relationship to creativity.
High Ability Studies, 8, 135158.
Martinsen, . L., & Diseth, . (2011). The factor structure of the AssimilatorExplorer
Inventory-revised. Creativity Research Journal, 23, 273283.
Martinsen, . L., Furnham, A., & Hrem, T. (2015). An Integrated Perspective on Insight:
How Cognitive Style and Motivation Moderate Effects of Search Constraints on Insight
Problems. (in preparation).
Martinsen, . L., & Kaufmann, G. (2000). The AssimilatorExplorer cognitive styles and
their relationship to affectivemotivational orientations and cognitive performances.
In R. Riding, & S. Raynor (Eds.), International perspectives on individual differences. New
developments in learning/cognitive styles, 1. (pp. 341). Stamford, CT: Ablex Publishing
Corporation.
Martinsen, . L., Kaufmann, G., & Furnham, A. (2011). Cognitive style and creativity. (2nd
ed.). In M. A. Runco, & S. R. Pritzker (Eds.), Encyclopedia of Creativity, Vol. 1. (pp.
214221). Academic PressSan Diego.
Mednick, S. A. (1962). The associative basis of the creative process. Psychological Review,
69, 220232.
Mnnesland, K. (1985). Intelligensprver for Voksne Gruppeprve serie 3 for alderen over 14
r [Intelligence tests for the age above 14 years]. Oslo: Tano.
Nygrd, R. (1997; personal communication). A short form of the Achievement Motive
Scale. Unpublished test. Oslo: Department for Educational Research, University of
Oslo.
Nygrd, R., & Gjesme, T. (1973). Assessment of achievement motives. Scandinavian
Journal of Educational Research, 17, 3946.
Riding, R. J., & Rayner, S. G. (1998). Cognitive styles and learning strategies. London: David
Fulton Publishers.
Undheim, J. O. (1981). On intelligence II: A neo-Spearman model to replace Catell's theory
of uid and crystallized intelligence. Scandinavian Journal of Psychology, 22, 181187.
Yerkes, R. M., & Dodson, J. D. (1908). The relation of strength of stimulus to rapidity of
habit formation. Journal of Comparative Neurology and Psychology, 18, 459482.

Please cite this article as: Martinsen, .L., & Furnham, A., Cognitive styles and performance on complex, structured tasks, Learning and Individual
Differences (2015), http://dx.doi.org/10.1016/j.lindif.2015.07.013