Learning Styles Concepts and Evidence

See discussions, stats, and author profiles for this publication at: https://www.researchgate.
net/publication/233600402
Learning Styles: Concepts and Evidence
Article in Psychological Science in the Public Interest · December 2008

DOI: 10.1111/j.1539-6053.2009.01038.x
CITATIONS READS
2,013 90,278
4 authors, including:
Mark A. Mcdaniel Doug Rohrer

Washington University in St. Louis University of South Florida
334 PUBLICATIONS 30,915 CITATIONS 53 PUBLICATIONS 9,340 CITATIONS
SEE PROFILE SEE PROFILE
Robert A. Bjork
University of California, Los Angeles
190 PUBLICATIONS 26,880 CITATIONS
SEE PROFILE
All content following this page was uploaded by Robert A. Bjork on 13 November 2017.
The user has requested enhancement of the downloaded file.

Psychological
Science in the
PUBLIC
INTEREST
CONTENTS V olume 9 Number 3 December 2008
Learning Styles: Concepts and Evidence

Harold Pashler, Mark McDaniel, Doug Rohrer, and Robert Bjork
106 Introduction
106 An Overview of Learning Styles: Doctrines and Industry
107 How Did the Learning-Styles Approach Become So Widespread and Appealing?
Origin and Popularity
Interactions of Individual Differences and Instructional Methods
108 What Evidence Is Necessary to Validate Interventions Based on Learning Styles?
Existence of Study Preferences
The Learning-Styles Hypothesis
Interactions as the Key Test of the Learning-Styles Hypothesis
Primary Mental Abilities: Relation to Learning Styles
111 Evaluation of Learning-Styles Literature
Style-by-Treatment Interactions: The Core Evidence Is Missing
Learning-Styles Studies With Appropriate Methods and Negative Results
113 Related Literatures With Appropriate Methodologies
Aptitude-by-Treatment Interactions
Personality-by-Treatment Interactions
116 Conclusions and Recommendations
Points of Clarification
Costs and Benefits of Educational Interventions
Beliefs Versus Evidence as a Foundation for Educational Practices and Policies
Everybody’s Potential to Learn
A JOURNAL OF THE ASSOCIATION FOR PSYCHOLOGICAL SCIENCE

About the Authors
Harold Pashler is Professor of Psychology and a faculty member of the Cognitive Science Program at the University of
California, San Diego. His main areas of interest are human learning and the psychology of attention. Pashler’s learning
research focuses on methods for optimizing acquisition and retention of knowledge and skills. In the field of attention,
Pashler’s work has illuminated basic attentional bottlenecks as well as the nature of visual awareness. Pashler is the author
of ThePsychologyofAttention (MIT Press, 1998) and the editor of Stevens’HandbookofExperimentalPsychology (Wiley,
2001). He received the Troland Prize from the National Academy of Sciences for his studies of human attention, and was
elected to membership in the Society of Experimental Psychologists.
Mark McDaniel is Professor of Psychology at Washington University in St. Louis, with a joint appointment in Education.
He received his PhD from the University of Colorado in 1980. His research is in the general area of human learning and
memory, with an emphasis on prospective memory, encoding and retrieval processes in episodic memory, learning of
complex concepts, and applications to educational contexts and to aging. His educationally relevant research includes
work being conducted in actual college and middle-school classrooms. This research is being sponsored by the Institute of
Educational Sciences and the James S. McDonnell Foundation, and his work is also supported by the National Institutes of
Health. McDaniel has served as Associate Editor of the Journal of Experimental Psychology: Learning, Memory, and
Cognition and Cognitive Psychology and as President of the Rocky Mountain Psychological Association, and he is a
fellow of Divisions 3 and 20 of the American Psychological Association. He has published over 200 journal articles, book
chapters, and edited books on human learning and memory, and is the coauthor, with Gilles Einstein, of two recent books:
Memory Fitness: AGuide for Successful Aging (Yale University Press, 2004) and Prospective Memory: An Overview and
SynthesisofanEmergingField (Sage, 2007).
Doug Rohrer is Professor of Psychology at the University of South Florida. He received his doctoral degree in Psychology
from the University of California, San Diego, and he was a faculty member at George Washington University before moving to
the University of South Florida. Before attending graduate school, he taught high-school mathematics for several years.
Most of his research concerns learning and memory, with a recent emphasis on learning strategies.
Robert A. Bjork is Distinguished Professor and Chair of Psychology at the University of California, Los Angeles. His
research focuses on human learning and memory and on the implications of the science of learning for instruction and
training. He has served as Editor of Memory &Cognition and PsychologicalReview (1995–2000), Coeditor of Psycholo-
gicalScienceinthePublicInterest (1998–2004), and Chair of the National Research Council’s Committee on Techniques
for the Enhancement of Human Performance. He is a past president or chair of the Association for Psychological Science
(APS), the Western Psychological Association, the Psychonomic Society, the Society of Experimental Psychologists, the
Council of Editors of the American Psychological Association (APA), and the Council of Graduate Departments of
Psychology. He is a recipient of UCLA’s Distinguished Teaching Award, the American Psychological Association’s
Distinguished Scientist Lecturer and Distinguished Service to Psychological Science Awards, and the American
Physiological Society’s Claude Bernard Distinguished Lecturership Award.
P SY CH OLOGI CA L S CIE NCE IN TH E PUB LIC INTE RES T
Learning Styles
Concepts and Evidence
Harold Pashler,1 Mark McDaniel,2 Doug Rohrer,3 and Robert Bjork4
1
University of California, San Diego, 2Washington University in St. Louis, 3University of South Florida, and 4University of
California, Los Angeles
SUMMARY—The term ‘‘learning styles’’ refers to the concept the same for all students. Finally, in order to demonstrate
that individuals differ in regard to what mode of instruc- that optimal learning requires that students receive in-
tion or study is most effective for them. Proponents of struction tailored to their putative learning style, the
learning-style assessment contend that optimal instruction experiment must reveal a specific type of interaction be-
requires diagnosing individuals’ learning style and tai- tween learning style and instructional method: Students
loring instruction accordingly. Assessments of learning with one learning style achieve the best educational
style typically ask people to evaluate what sort of infor- outcome when given an instructional method that differs
mation presentation they prefer (e.g., words versus pic- from the instructional method producing the best out-
tures versus speech) and/or what kind of mental activity come for students with a different learning style. In
they find most engaging or congenial (e.g., analysis versus other words, the instructional method that proves most
listening), although assessment instruments are extremely effective for students with one learning style is not the most
diverse. The most common—but not the only—hypothesis effective method for students with a different learning
about the instructional relevance of learning styles is the style.
meshing hypothesis, according to which instruction is best Our review of the literature disclosed ample evidence
provided in a format that matches the preferences of the that children and adults will, if asked, express preferences
learner (e.g., for a ‘‘visual learner,’’ emphasizing visual about how they prefer information to be presented to them.
presentation of information). There is also plentiful evidence arguing that people differ
The learning-styles view has acquired great influence in the degree to which they have some fairly specific apti-
within the education field, and is frequently encountered tudes for different kinds of thinking and for processing
at levels ranging from kindergarten to graduate school. different types of information. However, we found virtu-
There is a thriving industry devoted to publishing learn- ally no evidence for the interaction pattern mentioned
ing-styles tests and guidebooks for teachers, and many above, which was judged to be a precondition for vali-
organizations offer professional development workshops dating the educational applications of learning styles. Al-
for teachers and educators built around the concept of though the literature on learning styles is enormous, very
learning styles. few studies have even used an experimental methodology
The authors of the present review were charged with capable of testing the validity of learning styles applied to
determining whether these practices are supported by education. Moreover, of those that did use an appropriate
scientific evidence. We concluded that any credible vali- method, several found results that flatly contradict the
dation of learning-styles-based instruction requires robust popular meshing hypothesis.
documentation of a very particular type of experimental We conclude therefore, that at present, there is no ad-
finding with several necessary criteria. First, students equate evidence base to justify incorporating learning-
must be divided into groups on the basis of their learning styles assessments into general educational practice. Thus,
styles, and then students from each group must be ran- limited education resources would better be devoted to
domly assigned to receive one of multiple instructional adopting other educational practices that have a strong
methods. Next, students must then sit for a final test that is evidence base, of which there are an increasing number.
However, given the lack of methodologically sound studies
of learning styles, it would be an error to conclude that all
Address correspondence to Harold Pashler, Department of Psychol-
ogy 0109, University of California, San Diego, La Jolla, CA 92093; possible versions of learning styles have been tested and
e-mail: [email protected]. found wanting; many have simply not been tested at all.
Volume 9—Number 3 Copyright r 2009 Association for Psychological Science 105

Learning Styles
Further research on the use of learning-styles assessment would provide sufficient evidence for the learning-styles con-
in instruction may in some cases be warranted, but such cept, as detailed in the following sections, and then we searched
research needs to be performed appropriately. for evidence that satisfied this minimal criterion.
AN OVERVIEW OF LEARNING STYLES: DOCTRINES

INTRODUCTION AND INDUSTRY
The term learning styles refers to the view that different people As described earlier, the concept of learning styles encompasses
learn information in different ways. In recent decades, the not only a large body of written materials but also what seems to
concept of learning styles has steadily gained influence. In this be a thriving set of commercial activities. The writings that touch
article, we describe the intense interest and discussion that the on the learning-styles concept in its broadest sense include
concept of learning styles has elicited among professional ed- several thousand articles and dozens of books. These figures may
ucators at all levels of the educational system. Moreover, the seem surprisingly large, but one should keep in mind the sheer
learning-styles concept appears to have wide acceptance not number of different schemes or models of learning styles that
only among educators but also among parents and the general have been proposed over the years. For example, in a relatively
public. This acceptance is perhaps not surprising because the comprehensive review, Coffield et al. (2004) described 71
learning-styles idea is actively promoted by vendors offering different schemes, and they did not claim that their list was
many different tests, assessment devices, and online technolo- exhaustive.
gies to help educators identify their students’ learning styles and The commercial activity related to learning styles is largely
adapt their instructional approaches accordingly (examples are centered around the publishing and selling of measurement
cited later). devices to help teachers assess individual learning styles; typ-
We are cognitive psychologists with an interest both in the ically, although not always, these devices classify the learner
basic science of learning and memory and in the ways that into different style categories. Testing has been recommended
science can be developed to be more helpful to teachers and by organizations at all levels of education that might be pre-
students. We were commissioned by Psychological Science in the sumed to base their recommendations on evidence. For exam-
Public Interest to assess, as dispassionately as we could, the ple, the National Association of Secondary School principles
scientific evidence underlying practical application of learning- commissioned the construction of a learning-styles test that it
style assessment in school contexts. This task involved two distributed widely (Keefe, 1988). Similarly, the Yale Graduate
steps: (a) analyzing the concept of learning styles to determine School of Arts and Sciences (2009) currently maintains a Web
what forms of evidence would be needed to justify basing ped- site that offers advice for Yale instructors; the site informs vis-
agogical choices on assessments of students’ learning styles and itors that ‘‘college students enter our classrooms with a wide
(b) reviewing the literature to see whether this evidence exists. variety of learning styles.’’ The site goes on to recommend that
Our team began this undertaking with differing—but not pas- instructors determine their own ‘‘modality of learning’’ as well as
sionately held—opinions on learning styles as well as a shared assess their students’ learning styles and make their instruc-
desire to let the empirical evidence lead us where it would. tional choices accordingly.
We start by offering the reader a brief overview of the learning- Furthermore, the learning-styles concept is embraced in a
styles concept, including some of the publications and entre- number of current educational psychology textbooks. For in-
preneurial ventures that have been developed around the idea. stance, Omrod (2008) wrote, ‘‘Some cognitive styles and dis-
Next, we analyze the learning-styles concept from a more ab- positions do seem to influence how and what students learn. . . .
stract point of view. Here, we grapple with some potentially Some students seem to learn better when information is pre-
confusing issues of definition and logic that in our opinion re- sented through words (verbal learners), whereas others seem to
quire more careful consideration in connection with learning learn better when it’s presented through pictures (visual learn-
styles than they have so far received. We argue that this analysis ers)’’ (p. 160, italics in original). Thus, educational psychology
is a useful, and essential, prerequisite to organizing and ap- students and aspiring teachers are being taught that students
praising the evidence on learning styles. Finally, we describe the have particular learning styles and that these styles should be
results of our search of published literature, draw some con- accommodated by instruction tailored to those learning styles.
clusions, and suggest lines of future research. We should em- Some of the most popular learning-style schemes include the
phasize, however, that the present article is not a review of the Dunn and Dunn learning-styles model (e.g., Dunn, 1990), Kolb’s
literature of learning styles; indeed, several such reviews have (1984, 1985) Learning Styles Inventory, and Honey and Mum-
appeared recently (e.g., Coffield, Moseley, Hall, & Ecclestone, ford’s (1992) Learning Styles Questionnaire. The assessment
2004; Kozhevnikov, 2007; Sternberg, Grigorenko, & Zhang, devices that have been developed in relation to the model
2008). In brief, we sought to determine what kinds of findings of Dunn and Dunn are particularly popular and extensive.
106 Volume 9—Number 3

H. Pashler et al.
Customers visiting the Web site of the International Learning approximately $100.00 (as of June 2008). The Hay Group also
Styles Network (www.learningstyles.net) are advised that distributes an informational booklet called ‘‘One Style Doesn’t
Fit All: The Different Ways People Learn and Why It Matters’’
Learning style is the way in which each learner begins to con- (Hay Group, n.d.). According to the booklet, the practical ben-
centrate on, process, absorb, and retain new and difficult infor- efits of classifying individuals’ learning styles include ‘‘placing
mation (Dunn and Dunn, 1992; 1993; 1999). The interaction of them in learning and work situations with people whose learning
these elements occurs differently in everyone. Therefore, it is strengths are different from their own,’’ ‘‘improving the fit be-
necessary to determine what is most likely to trigger each student’s tween their learning style and the kind of learning experience
concentration, how to maintain it, and how to respond to his or her
they face,’’ and ‘‘practicing skills in areas that are the opposite of
natural processing style to produce long term memory and reten-
their present strengths’’ (Hay Group, n.d., p. 11).
tion. To reveal these natural tendencies and styles, it is important to
use a comprehensive model of learning style that identifies each
These three examples are merely some of the more popular
individual’s strengths and preferences across the full spectrum of and well-advertised products within the learning-styles move-
physiological, sociological, psychological, emotional, and envi- ment. Readers interested in a more comprehensive view should
ronmental elements. (International Learning Styles Network, 2008) consult Coffield et al. (2004).
As of June 2008, the company sells five different assessment HOW DID THE LEARNING-STYLES APPROACH
tools for different age groups—ranging from the Observational BECOME SO WIDESPREAD AND APPEALING?
Primary Assessment of Learning Style (OPAL) for ages 3 to 6 to
Building Excellence (BE) for ages 17 and older (at a cost of Origin and Popularity
approximately $5.00 per student for the classification instru- The popularity and prevalence of the learning-styles approach
ment). The vendor claims these assessments ‘‘measure the pat- may, of course, be a product of its success in fostering learning
terns through which learning occurs in individual students; they and instruction. Assessing the extent to which there is evidence
summarize the environmental, emotional, sociological, physio- that the approach does indeed foster learning is the primary goal
logical, and global/analytic processing preferences that a stu- of this review. However, there are reasons to suspect that other
dent has for learning’’ (International Learning Styles Network, factors—in addition to, or instead of, actual effectiveness—may
2008). A summer certification program is also offered in con- play a role in the popularity of the learning-styles approach.
nection with this approach (the basic certification program costs Most learning-styles taxonomies are ‘‘type’’ theories: That is,
$1,225 per trainee, excluding meals and lodging, with a higher they classify people into supposedly distinct groups, rather than
level certification for conducting research on learning styles also assigning people graded scores on different dimensions. One
offered for an additional $1,000). The Dunn and Dunn assess- can trace the lineage of these theories back to the first modern
ment instrument for adults asks respondents to indicate, for typological theorizing in the personality field, which was un-
example, whether they learn best when they hear a person talk dertaken by the psychiatrist and psychoanalyst C.G. Jung
about something, whether their desk is typically disorganized (1964). Jung’s ideas were explicitly incorporated into a psy-
and messy, whether they would say that they normally think in chological test developed in the United States, the Myers–Briggs
words as opposed to mental images, and whether they would Type Indicator test. This test became very popular starting in the
characterize themselves as someone who thinks intuitively or 1940s and remains widely used to this day. The Myers–Briggs
objectively (Rundle & Dunn, 2007). categorizes people into a number of groups, providing infor-
Kolb’s (1984, 1985) Learning Styles Inventory is another very mation that is said to be helpful in making occupational deci-
popular scheme, particularly within the United States. It con- sions. The assumption that people actually cluster into distinct
ceives of individuals’ learning processes as differing along two groups as measured by this test has received little support from
dimensions: preferred mode of perception (concrete to abstract) objective studies (e.g., Druckman & Porter, 1991; Stricker &
and preferred mode of processing (active experimentation to Ross, 1964), but this lack of support has done nothing to dampen
reflective observations). The Learning Styles Inventory classi- its popularity. It seems that the idea of finding out ‘‘what type of
fies individuals into four types on the basis of their position along person one is’’ has some eternal and deep appeal, and the suc-
these two dimensions: divergers (concrete, reflective), assimi- cess of the Myers–Briggs test promoted the development of type-
lators (abstract, reflective), convergers (abstract, active), and based learning-style assessments.
accommodators (concrete, active). The self-assessment requires Another, very understandable, part of the appeal of the
people to agree or disagree (on a 4-point scale) with, for ex- learning-styles idea may reflect the fact that people are con-
ample, the idea that they learn best when they listen and watch cerned that they, and their children, be seen and treated by
carefully, or that when they learn they like to analyze things and educators as unique individuals. It is also natural and appealing
to break them down into parts. to think that all people have the potential to learn effectively and
The Learning Styles Inventory is distributed by the Hay Group easily if only instruction is tailored to their individual learning
(http://www.haygroup.com) and sold in packs of 10 booklets for styles. Another related factor that may play a role in the popu-
Volume 9—Number 3 107

Learning Styles
larity of the learning-styles approach has to do with responsi- Henson & Hwang, 2002; Veres, Sims, & Shake, 1987), the ex-
bility. If a person or a person’s child is not succeeding or existence of preferences with some coherence and stability is not
celling in school, it may be more comfortable for the person to in dispute. A study by Massa and Mayer (2006), which is dis-
think that the educational system, not the person or the child cussed in more detail later, provides further evidence on this
himself or herself, is responsible. That is, rather than attribute point. Massa and Mayer developed three instruments to assess
one’s lack of success to any lack of ability or effort on one’s part, people’s preferences for receiving instruction verbally versus
it may be more appealing to think that the fault lies with in- accompanied by pictorial illustrations. Responses on these in-
struction being inadequately tailored to one’s learning style. In struments were significantly correlated with the degree to which
that respect, there may be linkages to the self-esteem movement college students chose to receive verbal elaboration versus
that became so influential, internationally, starting in the 1970s pictorial elaboration of technical terms in an electronics lesson.
(Twenge, 2006). Massa and Mayer also found significant correlations between the
instruments they used to assess people’s preference for certain
kinds of representations and the mode of elaboration people
Interactions of Individual Differences and
elected to receive in the electronics lesson. (As discussed at
Instructional Methods
more length later, however, the preference for visual versus
As we argue in the next section, credible evidence in support of
verbal information intake had little, if any, relationship to an
practices based on learning styles needs to document a specific
individual’s objectively measured specific-aptitude profile.)
type of interaction between instructional method and assess-
Having noted the reality of these preferences, we emphasize
ments of an individual’s learning style. Basically, evidence for a
that the implications of such preferences for educational prac-
learning-styles intervention needs to consist of finding that a
tices and policies are minimal. The existence of preferences says
given student’s learning is enhanced by instruction that is tai-
nothing about what these preferences might mean or imply for
lored in some way to that student’s learning style.
anything else, much less whether it is sensible for educators to
Naturally, it is undeniable that the optimal instructional
take account of these preferences. Most critically, the reality of
method will often differ between individuals in some respects. In
these preferences does not demonstrate that assessing a student’s
particular, differences in educational backgrounds can be a
learning style would be helpful in providing effective instruction
critical consideration in the optimization of instruction. New
for that student. That is, a particular student’s having a particular
learning builds on old learning, for example, so an individual
preference does not, by itself, imply that optimal instruction for
student’s prior knowledge is bound to determine what level and
the student would need to take this preference into account. In
type of instructional activities are optimal for that student. Many
brief, the existence of study preferences would not by itself
research studies (see, e.g., McNamara, Kintsch, Butler-Songer,
suggest that buying and administering learning-styles tests
& Kintsch, 1996) have demonstrated that the conditions of in-
would be a sensible use of educators’ limited time and money.
struction that are optimal differ depending on students’ prior
knowledge. Later in this review, we summarize some of the ev-
idence suggesting that aptitude measures can help predict what
The Learning-Styles Hypothesis
instructional methods are most effective.
What, then, is the version of the learning-styles hypothesis that
has practical implications for educational contexts? It is the
WHAT EVIDENCE IS NECESSARY TO VALIDATE
claim that learning will be ineffective, or at least less efficient
INTERVENTIONS BASED ON LEARNING STYLES?
than it could be, if learners receive instruction that does not take
account of their learning style, or conversely, it is the claim that
We turn now to the core of the learning-styles idea: an assess-
individualizing instruction to the learner’s style can allow peo-
ment of the degree to which it has been validated.
ple to achieve a better learning outcome.
It is important to note that there is a specific version of the
Existence of Study Preferences learning-styles hypothesis that evidently looms largest both
In reviewing the literature on learning styles and examining the within the educational literature and within the minds of most
different ways in which this term is frequently used, we make a people writing about learning styles: the idea that instruction
basic distinction between what we call the existence of study should be provided in the mode that matches the learner’s style.
preferences and what we call the learning-styles hypothesis. The For example, if the learner is a ‘‘visual learner,’’ information
existence of preferences, as we interpret it, amounts simply to should, when possible, be presented visually. We refer to this
the fact that people will, if asked, volunteer preferences about specific instance of the learning-styles hypothesis as the mesh-
their preferred mode of taking in new information and studying. ing hypothesis—the claim that presentation should mesh with
Given that learning-style questionnaires focusing on prefer- the learner’s own proclivities.
ences have at least some psychometric reliability (i.e., a person’s Most proponents of the learning-styles idea subscribe to some
score on one day predicts their score on another day; e.g., form of the meshing hypothesis, and most accounts of how in-

H. Pashler et al.
struction should be optimized assume the meshing hypothesis: Figures 1D to 1I show some hypothetical interactions that
For example, they speak of (a) tailoring teaching to ‘‘the way in would not provide support for the learning-styles hypothesis
which each learner begins to concentrate on, process, absorb, because, in each case, the same learning method provides
and retain new and difficult information’’ (Dunn & Dunn’s optimal learning for every learner. Note that these findings are
framework; International Learning Styles Network, 2008), (b) insufficient even though it is assumed that every interaction in
the learner’s preferred modes of perception and processing Figure 1 is statistically significant. It is interesting to note that
(Kolb’s, 1984, 1985, framework), or (c) ‘‘the fit between [peo- the data shown in Figures 1D and 1G do produce a crossover
ple’s] learning style and the kind of learning experience they interaction when the data are plotted so that the horizontal axis
face’’ (Hay Group, n.d., p. 11). Note that the learning-styles represents learning method, as shown in Figure 2, but this mere
hypothesis, as defined here, could be true without the meshing rearrangement of the data does not alter the fact that the same
hypothesis being true—if, for example, individuals classified as learning method maximizes performance of all subjects.1 Thus,
visual learners profited more from verbal instruction in some as noted earlier, a style-by-method crossover interaction con-
situations or if individuals classified as verbal learners profited stitutes sufficient evidence for the learning-styles hypothesis if
more from visual instruction. In our review, we searched for and only if the horizontal axis represents learning style, as in
evidence for both this broad version of the learning-styles hy- Figures 1A to 1C.
pothesis and the more specific meshing hypothesis. To provide the most liberal criterion in our search for evidence
supporting the learning-styles hypothesis, we cast the hypoth-
esis so that it requires only the style-by-method crossover in-
Interactions as the Key Test of the Learning-Styles teraction described previously. It does not require that the
Hypothesis optimal method for each group would somehow match or con-
To provide evidence for the learning-styles hypothesis—whe- form to each group’s learning style (the meshing hypothesis re-
ther it incorporates the meshing hypothesis or not—a study must ferred to earlier).
satisfy several criteria. First, on the basis of some measure or
measures of learning style, learners must be divided into two or
Primary Mental Abilities: Relation to Learning Styles
more groups (e.g., putative visual learners and auditory learn-
In our discussion of styles thus far, we have focused on prefer-
ers). Second, subjects within each learning-style group must be
ences for how information would be presented to a person rather
randomly assigned to one of at least two different learning
than on the notion of the person having different ability to pro-
methods (e.g., visual versus auditory presentation of some ma-
cess one kind of information or another. This focus is in con-
terial). Third, all subjects must be given the same test of
formity with the dictionary definition of style and matches at
achievement (if the tests are different, no support can be pro-
least the most typical usage of the term learning style within the
vided for the learning-styles hypothesis). Fourth, the results
education field. However, the notion of learning style as a set of
need to show that the learning method that optimizes test per-
preferences and the notion of learning style as a specific aptitude
formance of one learning-style group is different than the
are very closely intertwined in many discussions of learning
learning method that optimizes the test performance of a second
styles. Moreover, it is our impression that among the general
learning-style group.
public, the notion of learning styles and the notion of differential
Thus, the learning-styles hypothesis (and particular instruc-
abilities are scarcely distinguished at all. There is, after all, a
tional interventions based on learning styles) receives support if
commonsense reason why the two concepts could be conflated:
and only if an experiment reveals what is commonly known as a
Namely, different modes of instruction might be optimal for
crossover interaction between learning style and method when
different people because different modes of presentation exploit
learning style is plotted on the horizontal axis. Three such
the specific perceptual and cognitive strengths of different in-
findings are illustrated in Figures 1A to 1C. For each of these
dividuals, as suggested by the meshing hypothesis.
types of findings, the method that proves more effective for
Similar to the learning-styles hypothesis, the idea of specific
Group A is not the same as the method that proves more effective
abilities also implies a special form of crossover interaction.
for Group B. One important thing to notice about such a cross-
However, the interaction is different in kind from what was
over interaction is that it can be obtained even if every subject
outlined earlier as the key test of the learning-styles hypothesis.
within one learning-style group outscores every subject within
the other learning-style group (see Fig. 1B). Thus, it is possible 1
A reviewer of an earlier version of this article noted that the interactions
to obtain strong evidence for the utility of learning-style as- shown in Figures 1H and 1I might have potential practical importance, even in
sessments even if learning style is correlated with what might, the absence of a true crossover. If one could sort people into two groups, one of
which would benefit from an instructional manipulation and the other of which
for some purposes, be described as ability differences. More- was completely unaffected by it, it might (on some assumptions) be worthwhile
over, the necessary crossover interaction allows for the possi- doing the sorting and selectively offering the manipulation. We agree. However,
as we show later, the general conclusions reached here do not depend on this
bility that both learning-style groups could do equally well with issue because we have not found any actual interactions of the types in Figures
one of the learning methods (see Fig. 1C). 1H and 1I in the learning-styles literature.

Learning Styles
Acceptable Evidence
In examples A, B, and C, the learning method that optimized the
mean test score of one kind of learner is different from the learning method that optimized
the mean test score of the other kind of learner.
A B C
Method 2 Method 2
Test Method 2 Method 1
Score
Method 1
Method 1
A Style B Style A Style B Style A Style B Style

Learners Learners Learners Learners Learners Learners
Unacceptable Evidence
In examples D through I, the same learning method optimized the mean test score
of both kinds of learners, thereby precluding the need to customize instruction.
D E F
Method 1 Method 1 Method 1
Test
Score
Method 2
Method 2 Method 2

G H I
Method 1 Method 1 Method 1
Test Method 2
Score
Method 2
Method 2
Fig. 1. Acceptable and unacceptable evidence for the learning-styles hypothesis. In each of the hypothetical experiments, subjects have been first
classified as having Learning Style A or B and then randomly assigned to Learning Method 1 or 2. Later, all subjects have taken the same test. The
learning-styles hypothesis is supported if and only if the learning method that optimized the mean test score of one group is different from the learning
method that optimized the mean test score of the other group, as in A, B, and C. By contrast, if the same learning method optimized the mean test score of
both groups, as in D through I, the result does not provide evidence. (Note that all nine interactions are assumed to be statistically significant.) In general,
the learning-styles hypothesis is supported if and only if a study finds a crossover interaction between learning method and learning style, assuming that
the horizontal axis represents the learning-style variable. See the text for more details.
If the notion of specific aptitudes or skills is valid, one ought to There is little doubt that specific-ability differences of this
be able to divide subjects into two or more groups (e.g., Group A kind exist. The first psychologist to provide strong empirical
of learners with high auditory ability and Group B of learners evidence for the idea of specific-ability differences was Louis
with high visual ability). There should then be two tests such that Thurstone (e.g., Thurstone, 1938). Thurstone proposed seven
Group A outscores Group B on one test, whereas Group B out- ‘‘primary mental abilities’’: verbal comprehension, word fluency,
scores Group A on the other test. number facility, spatial visualization, associative memory,

H. Pashler et al.
A B
Method 1 80 80
60 Test 60
Test
Score 40 Score 40 A Style Learners
Method 2 20 20 B Style Learners
A Style B Style Method 1 Method 2

Learners Learners
Method 1 80 80
Test 60 Test 60 60 A Style Learners
Score Score
Method 2 20 20 B Style Learners
A Style B Style Method 1 Method 2

Learners Learners
Fig. 2. Examples of crossover interactions that would not validate the learning-styles hypothesis. The two hypothetical
outcomes in A are identical to the outcomes in B, and these examples demonstrate that the choice of variable for the hor-
izontal axis can affect whether an interaction appears to ‘‘cross over.’’ Regardless of appearance, though, each of the
graphs above demonstrates that the same learning method (Method 1) proved superior for all subjects. Thus, the data above
do not provide evidence for the learning-styles hypothesis. However, if the horizontal axis depicts the learning-style
variable, a crossover interaction is both sufficient and necessary to show evidence for the learning-styles hypothesis, as in
Figures 1A–1C. Note that the above two results are identical to those in Figures 1D and 1G.
perceptual speed, and reasoning. Although these abilities are not offer one kind of evidence, and one kind of evidence alone: a
completely uncorrelated (implying, to some, the idea of general crossover interaction of the form illustrated in Figures 1A to 1C.
mental ability or ‘‘g’’; see Jensen, 1998; Spearman, 1927), they On the basis of this analysis, we scoured the literature to identify
do show a moderate degree of independence (Thurstone, 1938). studies that provided such evidence. Remarkably, despite the
Although this provides evidence for specific aptitudes, it does vast size of the literature on learning styles and classroom
not show that one needs to provide different groups with different instruction, we found only one study that could be described
forms of instruction to maximize their performance on any single as even potentially meeting the criteria described earlier,2 and
outcome test. Thus, evidence for specific aptitudes does not, by as we report in the following text, even that study provided less
itself, validate the learning-styles hypothesis. than compelling evidence.
There are few data on the relationship between preferences The study in question was reported by Sternberg, Grigorenko,
and specific aptitudes. However, one recent and well-executed Ferrari, and Clinkenbeard (1999). In this study, 324 ‘‘gifted and
study, which we discuss at more length later, discloses that talented’’ high school students were given the Sternberg Triar-
preference for visual versus verbal information intake shows chic Abilities Test, which provided a rating of each student’s
hardly any relationship to an individual’s objectively measured analytical, creative, and practical ability. On the basis of this
specific-aptitude profile (Massa & Mayer, 2006). Thus, the test, the authors selected a subset of 112 subjects (35%) for
common assumption that preferences and abilities are closely whom one of these three abilities was much higher than the other
tied is open to challenge. But as we have defined the learning- two, and depending on their area of strength, these subjects were
styles hypothesis, one could find evidence for the hypothesis assigned to the high-analytical, high-creative, or high-practical
regardless of whether the style measure involved a specific ap- groups. (Another 87 students were assigned to two additional
titude, a preference, or both. groups not described here, and the remaining 125 students were
excluded from the study.) The participating subjects enrolled in
an introductory psychology summer course at Yale University,
EVALUATION OF LEARNING-STYLES LITERATURE
and each student was randomly assigned to class meetings that
emphasized analytical instruction, creative instruction, practi-
Style-by-Treatment Interactions: The Core Evidence
Is Missing 2
We also encountered one study in the domain of user information technology
For the reasons described earlier, it is our judgment that a val- training that appears to offer one interaction of the form discussed here (see
idation of an intervention based on learning styles would need to Bostrom, Olfman, & Sein, 1990).

Learning Styles
cal instruction, or memory instruction (a control condition). istered to sort visual from verbal learners in various ways. In
Their course performance was assessed by raters, and the ratings general, the results, which the researchers replicated, showed
were ‘‘subjected to principal-component analyses’’ (Sternberg no tendency for better performance for those who received help
et al., 1999, p. 7). The authors reported several analyses, and, screens matched to their preferences. Critically, Massa and
for the analysis of the interaction of interest, they compared Mayer found no support for any of these interactions despite
the course performance of matched subjects (i.e., students who exhaustive analysis of nearly 20 individual-difference measures
received instruction that matched their strongest ability) to that spanned their three proposed facets of verbalizer–visualizer
mismatched subjects. The article states that after the data were learning styles. The authors concluded that their results pro-
‘‘screened for deviant scores’’ (Sternberg et al., 1999, p. 10), vided no support for ‘‘the idea that different instructional
matched subjects reliably outscored mismatched subjects on methods should be used for visualizers and verbalizers’’ (Massa
two of the three kinds of assessments. & Mayer, 2006, pp. 333–334).
Thus, the authors reported a style-by-treatment interaction. Within a medical-education context, a recent study by Cook,
Although suggestive of an interaction of the type we have been Thompson, Thomas, and Thomas (2009) examined the hypothesis
looking for, the study has peculiar features that make us view it as that learners with a ‘‘sensing learning style’’ would do better when
providing only tenuous evidence. For one thing, the reported given instruction in which the problem was presented prior to the
interaction was found only with highly derived measures (as content information used to solve the problem, whereas ‘‘intuitive
noted above), and the untransformed outcome measures (e.g., the learners’’ would do better with the reverse. The authors noted that
mean score on each final assessment) were not reported for the this learning-styles taxonomy is similar to Kolb’s (1984, 1985)
different conditions. Furthermore, and as noted previously, only concrete–abstract dimension. Studying a sample of 123 internal
about one third of the subjects were classified into the groups that medicine residents and presenting modules on four ambulatory
produced the interaction. Finally, the interaction was achieved medicine topics, they found no support for this prediction.
only after the outliers were excluded for unspecified reasons. In Another study reaching a similar conclusion, albeit using
brief, although the article presents data that may be worth fol- tasks with less direct correspondence to real educational ac-
lowing up, it has serious methodological issues. Even for those tivities, was reported by Constantinidou and Baker (2002).
who might disagree with this judgment, the potential support that These investigators used a laboratory task to ask whether self-
this study could provide for any of the particular interventions reported preferences in information uptake predicted ability to
based on learning styles that are being marketed at the present perceive and store information in different modalities. They
time is extremely limited because the instructional manipulation examined the relationship between adults’ scores on the Visu-
does not seem to correspond to any of the more widely promoted alizer–Verbalizer Questionnaire (VVQ; Richardson, 1977)
and used learning-styles interventions. to their verbal free-recall performance on a task that presented
In summary, our efforts revealed at most one arguable piece of words through the auditory modality, the visual modality (as
evidence for the learning-styles hypothesis in general. For the line drawings of the corresponding object), or both. The VVQ
many specific assessment devices and interventions being ac- asks people a series of questions about their relative preference
tively marketed to teachers, as described earlier in this article, for taking in information through verbal versus visual means.
we were unable to find any evidence that would meet the key VVQ scores were not related in any strong or clearly interpret-
criteria discussed earlier (i.e., interactions of the form shown in able way to relative levels of free-recall performance for
Figs. 1A–1C). Moreover, we found a number of published different input modalities. Visual presentations produced better
studies that used what we have described as the appropriate free recall than did purely verbal presentations, and the authors
research design for testing the learning-styles hypothesis and reported finding ‘‘no relationship between a visual learning
found results that contradict widely held versions of the learn- style and the actual learning of verbal items that are
ing-styles hypothesis; we turn to these studies now. presented visually or auditorily’’ (Constantinidou & Baker,
2002, p. 306).
These studies, which we believe are methodologically strong,
Learning-Styles Studies With Appropriate Methods and provide no support for the learning-styles hypothesis (or its
Negative Results popular specific version, the meshing hypothesis). As mentioned
Massa and Mayer (2006) reported a particularly informative and previously, however, it would clearly be a mistake to label these
well-designed study of learning styles with a set of three ex- negative results as a conclusive refutation of the learning-styles
periments. They constructed a reasonably realistic computer- hypothesis in general. Further research modeled on the work of
based electronics lesson. Two different sorts of help screens Massa and Mayer (2006) may bring to light assessments paired
were customized for verbal or visual learners, providing either with interventions that do meet our criteria. But at present, these
supplementary printed text or carefully developed diagrams and negative results, in conjunction with the virtual absence of
illustrations, respectively. A wide variety of preference-based positive findings, lead us to conclude that any application of
and ability-based individual-difference measures were admin- learning styles in classrooms is unwarranted.

H. Pashler et al.
RELATED LITERATURES WITH APPROPRIATE discussions focusing on the literal and inferential aspects of
METHODOLOGIES discourse. This approach is assumed to place a greater burden
on the student for acquiring specific reading skills (see Freebody
Aptitude-by-Treatment Interactions & Tirre, 1985).
Although the literature on learning styles per se has paid scarce All of the sixth-grade students in a large school district who had
attention to the need for group-by-treatment interactions, there been in one of the two reading programs for 2 years or longer
has been a clear recognition of the importance of such interac- served as subjects (N 5 180, nearly equally distributed across
tions within an older educational psychology literature, going reading programs). Their aptitudes were assessed with a stan-
back to Cronbach’s (1957) appeal for research to uncover in- dardized test that included nonverbal and verbal measures of
teractions between aptitude and aspects of the instructional ability. The outcome measure was the reading test score achieved
context (termed treatments). Although the validity of aptitude- at the conclusion of the sixth-grade year. Multiple regression
by-treatment interactions (ATIs) is a separate issue from the analyses produced a significant ATI. The interpretation of the
validity of learning-style measures, which is the primary focus of interaction was based on predicted outcomes (from the regression
the current article, we describe several ATIs so that the reader equations) for particular low-ability values and particular high-
may gain an appreciation of a literature that recognizes the need ability values. These predicted outcomes indeed showed that
to demonstrate the necessary interaction. students with low ability would generally perform better on the
Initial attempts to demonstrate so-called ATIs were reviewed structured reading program (Matteson) than on the less structured
in a classic work by Cronbach and Snow (1977). According to reading program (Scott Foresman). The reverse would be pre-
Cronbach and Snow, these attempts were not highly successful dicted for the students with high ability: better performance on
because treatment durations were too brief, and aspects of the the less structured than on the more structured reading instruc-
methodologies were inadequate. After that review, significant tion method. Although suggestive, these data do not establish that
improvements were made in methodologies, with a number of students at a particular ability level (either low or high) fared
studies examining treatments implemented in classroom set- significantly better (in terms of reading outcomes) as a function of
tings for relatively long durations. the reading program in which the students were enrolled.
The kind of potential interaction that has received the most Additional direct support for the idea that learning outcomes
attention within the ATI tradition involves the degree to which for students with high and low abilities might reverse with a
the teaching approach provides ample structure or guidance for greater degree of structure embedded in instruction was re-
the learner. The primary hypothesis that has stimulated much of ported in the domain of elementary school mathematics (Cramer,
the work in this area is the idea that students with high ability Post, & Behr, 1989). Fourth graders being taught fractions were
tend to fare better in less structured learning environments than given four lessons (in six 40-minute class periods) on completing
in highly structured learning environments. By contrast, stu- rational numbers tasks that involved shading a particular frac-
dents with low ability are hypothesized to fare better with in- tional area (two thirds) of different kinds of visual figures (e.g., a
struction that is highly structured and provides explicit rectangle divided into three columns). In the high-structured
guidance than with instruction that is less structured and pro- condition, instruction was teacher centered with little student
vides little guidance (see, e.g., Snow, 1977). A variant of this choice. The teacher paced through each example in large-group
theme that also sparked interest is the idea that highly struc- lecture fashion. In the low-structured condition, the teacher
tured situations might reduce performance differences between provided an initial introduction to the problems and then stu-
students with high and low abilities (Freebody & Tirre, 1985). dents worked through examples at their own pace. The materials
As detailed in the next paragraph, two key difficulties in eval- involved leading questions to guide the learner to discovery of
uating this hypothesis are as follows: (a) The implementation of the key concepts. Both instructional conditions used identical
instructional methods that differ in structure (guidance) has examples, and both contained a 10-minute practice phase that
been quite variable, and (b) the measures used to assess student completed each 40-minute class period. Students in the higher
abilities have varied considerably. and lower ranges of cognitive restructuring ability, as measured
Freebody and Tirre (1985) reported an ATI in line with the by the Group Embedded Figures test (see Witkin & Good-
above hypothesis that involved two competing reading-in- enough, 1981), were assigned to each instructional condition. At
struction approaches. One approach, the Matteson program (see the conclusion of the lessons, the students completed a final test
Schlenker, 1978), provides a list of behavioral objectives in containing problems (rational numbers tasks) of the type taught
major reading-skill areas (e.g., word recognition, vocabulary in the lessons. For the more difficult problems—those requiring
development, literal and interpretative comprehension) com- physical restructuring of the diagrams—a crossover interaction
bined with individualized learning packages that cover these between ability and the degree of instructional structure
areas, following precisely defined sequences. The other ap- emerged. The students with high ability performed better
proach, the Scott Foresman (1972) program, is not strictly se- following low- than high-structured instruction; by contrast,
quenced and monitored. Instead, the emphasis is on frequent students with low ability performed better following high- than

Learning Styles
low-structured instruction. Particularly notable is that the main hypothesis being considered in reviewing this body of ATI
students with low ability outscored the students with high ability work, because the inquiry and public discussion methods
(at least nominally) after both received the more highly struc- encouraged learner self-direction (less structure). However,
tured instruction. This pattern thus provides evidence that Peterson et al. (1980) offered an interpretation based on the
learning is optimized when students with low ability are pro- underlying cognitive demands placed on the students by
vided with structured instruction and students with high ability the different instructional methods. They suggested that the
are provided with less structured instruction. lecture-recitation approach implemented in the study placed a
However, other studies that examined different content do- heavier burden on students’ cognitive skills than did the other
mains and used different assessment instruments did not always approaches. Specifically, students had to comprehend and
support the idea that high-ability students are better off with less attend to the lectures, take careful notes, and memorize target
structured instruction, whereas low-ability students profit more information. The idea is that students with high ability would
from higher-structured instruction. In Janicki and Peterson have the requisite skills to accomplish these challenges. Of
(1981), 117 grade school students completed a 2-week fractions course, this interpretation does not clarify why the students with
unit in a ‘‘direct’’ instructional fashion (which involved homework high ability would fare less well with the other instructional
assignments that students completed in class on their own) or in a methods, relative to the lecture-recitation method.
less structured fashion (this involved mixed-ability four-student One study activity that appears to be sensitive to individual
group seatwork with choice of homework or math games). Apti- ability differences is concept mapping (creation of diagrams that
tude, as determined by a composite measure that included Ravens show the relationship among concepts), with students with low
Progressive Matrices, did not interact with instructional method. verbal ability profiting more from concept maps (in a chemistry
Greene (1980) similarly failed to find an interaction when fifth learning activity) than students with high verbal ability
and sixth graders with high and low ability (as determined by (Stensvold & Wilson, 1990). Not surprisingly, in most studies the
Lorge-Thorndike verbal and nonverbal tests) were given high- students with higher ability outperformed the students with
structured instruction (specified sequence of workbook assign- lower ability in both instructional conditions.
ments and performance standards) or low-structured instruction However, complete crossovers have recently been reported
(choice and pacing of which exercises to do in the workbook) on with embedded-question techniques for learning from textbook
a letter-series task. The letter-series task was chosen to reflect chapters. In Callender and McDaniel (2007), the ability of in-
general problem-solving goals in education. The basic result terest was the degree to which learners can construct a coherent
was that the students with higher ability performed better than representation of presented content (either through text or lec-
the students with lower ability regardless of instruction. tures). Poor structure builders are assumed to perform relatively
In a well-conducted experiment, Peterson, Janicki, and Swing poorly at constructing a coherent representation of connected
(1980, Study 2) manipulated instruction for a 2-week ninth-grade discourse that is either read or spoken (Gernsbacher, 1990).
social studies unit across six classes (146 students). Two teachers Such comprehenders appear to construct too many substructures
taught each of three classes with one of three teaching methods. to accommodate incoming information, rather than constructing
One teaching method was a standard lecture-recitation ap- a unified integrated representation of the target material. By
proach. In the second method, termed inquiry, students re- contrast, good structure builders are able to extract coherent,
searched a historical question using primary sources. The third well-organized mental representation of the text. Accordingly,
method, public issues discussion, required students to support a Callender and McDaniel reasoned that embedding questions
position on a current public issue using primary material. Ap- into a textbook chapter would orient poor structure builders to
titude was defined as verbal ability. The outcome measure was a anchoring information around which to build a coherent repre-
test that included multiple-choice questions on historical facts sentation and therefore improve learning for students at this level
and short essay questions requiring integration and evaluation of of comprehension ability. Embedded questions might be super-
material. Critically, the test targeted readings and content com- fluous for good structure builders, however, because they are
mon to all three instructional approaches. In line with the pre- already able to construct coherent representations.
vious findings, there was no interaction between teaching method To test these predictions, Callender and McDaniel (2007) had
and ability for the essay performances, with students with higher college-age subjects read a chapter from an introductory psy-
ability performing better on the essay questions in general. chology textbook with or without embedded questions. After-
It is interesting to note that for the multiple-choice questions, ward, the subjects were given a multiple-choice test consisting
there was a significant ATI such that students with high ability of questions targeting the information featured by the embedded
performed better with the lecture-recitation teaching method questions and questions on information not targeted by the
than with the inquiry or public issues discussion methods, embedded questions. For poor structure builders, embedded
whereas students with low ability performed better when re- questions significantly improved performance on target ques-
ceiving the inquiry or public issues methods than with the lec- tions (relative to reading without embedded questions) but not
ture-recitation method. This pattern would appear to counter the performance on nontarget questions. Good structure builders

H. Pashler et al.
did not profit from embedded questions, and indeed their per- consideration is that learners with an internal locus of control
formance for nontarget information was better without embed- may fare better with less structured than with highly structured
ded questions. Note that these patterns could be considered instruction, whereas learners with an external locus of control
evidence for the general notion that more guided study activities will achieve more with highly structured than with less struc-
are preferable for comprehenders of lower ability, whereas less tured instruction.
guided presentations (no embedded questions) are preferred for Several studies have examined this hypothesis in college
comprehenders of higher ability. These patterns clearly require mathematics classes for prospective elementary school teachers.
replication, as only one chapter was considered and the subjects Horak and Horak (1982) examined two instructional methods
were in a laboratory experiment and not an actual course. Yet, during a 2-week unit on transformational geometry, with each
this finding illustrates the potential fruitfulness of attempting to method randomly assigned to a particular class section (total
link more specific cognitive processing abilities to instructional number of students was 102). In the highly guided instruction
techniques designed to dovetail with those abilities. (‘‘deductive’’), students were given rules or principles and then
In summary, ATIs evidently do occur, but it has not been easy proceeded to apply the rules to examples. In the less guided
to determine exactly when they occur. This diversity of outcomes instruction (‘‘inductive’’), students were given examples, with no
is perhaps not surprising given that available studies vary on a rule or principle stated for the students or expected from them.
number of potentially critical dimensions, including target The criterial test included questions designed to test lower
content, particular implementations of variations in instruc- levels of understanding (knowledge of terminology and repro-
tional structure, assessments used to index ability, and the kinds duction of material presented) and higher levels of under-
of criterial (outcome) tests used. In some studies, the ATIs can standing (e.g., problem solving). Marginally significant support
be reported for one type of criterial measure but not another for the predicted interaction was found for the questions testing
(e.g., see Cramer et al., 1989; Peterson, 1979; Peterson et al., lower levels of understanding: Students with an external locus of
1980). At best, then, the ATI literature provides a mixed picture. control performed better after the highly guided instruction than
A few studies are consistent with the idea that structured in- after the less guided instruction. The reverse was observed for
struction produces better learning outcomes for students of students with an internal locus of control, with performance after
lower ability (relative to less structured instruction), whereas less guided instruction exceeding performance after highly
less structured instruction produces better learning outcomes guided instruction (this also occurred with the questions tapping
for students with higher abilities (relative to structured in- higher levels of understanding).
struction). But other studies either did not obtain significant Parallel findings of marginal magnitude were reported in
ATIs involving general ability and the degree of structure in similar mathematics classes for elementary school teachers with
instruction or in some cases indicated that students with lower shorter treatment periods (McLeod & Adams, 1980/1981).
ability fared worse with structured instruction than with less Three experiments were conducted using somewhat different
structured instruction. The greater coherence of the literature instantiations of amount of guidance given during instruction
assessing structure building suggests that a more fine-grained and somewhat different target content. In only one experiment
approach that focuses on individual differences in underlying was the interaction significant (although a second experiment
cognitive processes, rather than general aptitudes, and imple- showed the same pattern): In this experiment, all students spent
ments instructional methods that target those processes may be 1 week learning about networks with an inductive set of mate-
more fruitful in producing robust interactions between learner rials (see earlier). The amount of guidance was manipulated by
ability and learner-directed activities. having students work individually on problems and encouraging
help from the instructor (high guidance, here students asked
may questions) or by having students work in groups of 4 (low
Personality-by-Treatment Interactions guidance, very few questions were posed to the instructor). On
There are also some more fragmentary but methodologically an immediate but not a delayed (given several weeks after in-
sophisticated studies documenting personality-by-instructional struction) criterial test, students with an internal locus of control
treatment effects, though these findings, like the aptitude– performed better with low guidance than with high guidance; the
treatment interactions described just above, do not speak to the reverse was found for the students with an external locus of
validity of the learning-styles hypothesis. Several studies have control. The absence of significant interactions in the other two
looked at a personality measure called locus of control, which experiments may have been a consequence of shorter treatments
refers to an individual’s belief about whether his or her suc- (75-min lesson in one experiment) or small sample size (just
cesses or failures are a consequence of internal or external under 60 students in each experiment), as the authors suggested.
factors (Rotter, 1966). An internal locus of control indicates a It is interesting to note that Janicki and Peterson’s (1981)
belief that outcomes are a consequence of one’s own actions. An study that failed to find an interaction with general ability (re-
external locus of control reflects the belief that outcomes are viewed in the preceding section) did observe a significant per-
unrelated to one’s own actions. One hypothesis that has received sonality-by-treatment interaction with a composite factor of

Learning Styles
locus of control and attitudes toward math (in teaching fractions clearly specified measures and then randomize learners to re-
to grade-school students). This composite factor of account- ceive one of several different instructional treatments. Equally
ability for learning interacted with instructional method such crucial, the interventions must be followed by a common pre-
that those students with higher accountability (more internal specified learning assessment given to all the participants in the
locus of control; 36% of the students) performed better on im- study. The paucity of studies using this methodology is the main
mediate and delayed computation and story-problem tests when factor that renders the learning-styles literature so weak and
in the less guided small group setting than when in the highly unconvincing, despite its large size.
guided direct-instruction setting. Instructional setting did not
produce differences for the students with lower accountability
Points of Clarification
(external locus of control).
Although we have argued that the extant data do not provide
In summary, there is modest evidence for the idea that stu-
support for the learning-styles hypothesis, it should be empha-
dents with an internal locus of control benefit more from less
sized that we do not claim that the same kind of instruction is
guided or structured instruction than from more guided in-
most useful in all contexts and with all learners. An obvious
struction, whereas students with an external locus of control
point is that the optimal instructional method is likely to vary
might benefit more from guided (structured) instruction than
across disciplines. For instance, the optimal curriculum for a
from less guided (structured) instruction. Previous studies re-
writing course probably includes a heavy verbal emphasis,
inforce those reviewed herein with similar patterns (Daniels &
whereas the most efficient and effective method of teaching
Stevens, 1976; Horak & Slobodzian, 1980; Parent, Forward,
geometry obviously requires visual–spatial materials. Of course,
Cantor, & Mohling, 1975; Yeany, Dost, & Mattews, 1980). The
identifying the optimal approach for each discipline is an em-
reliability and generalizability of these findings to other content
pirical question, and we espouse research using strong research
areas and to longer instructional treatments remain to be dem-
methods to identify the optimal approach for each kind of sub-
onstrated. A clear uncertainty is specifying the exact aspects of
ject matter.
instruction (group vs. individual work; density of questions di-
Furthermore, it is undoubtedly the case that a particular
rected at the instructor; homework choice vs. no choice) that are
student will sometimes benefit from having a particular kind of
interacting with locus of control.
course content presented in one way versus another. One sus-
pects that educators’ attraction to the idea of learning styles
CONCLUSIONS AND RECOMMENDATIONS
partly reflects their (correctly) noticing how often one student
may achieve enlightenment from an approach that seems useless
Our evaluation of the learning-styles concept led us to identify
for another student. There is, however, a great gap from such
the form of evidence needed to validate the use of learning-style
heterogeneous responses to instructional manipulations—
assessments in instructional settings (i.e., Figures 1A–1C). As
whose reality we do not dispute—to the notion that presently
described earlier, our search of the learning-styles literature has
available taxonomies of student types offer any valid help in
revealed only a few fragmentary and unconvincing pieces of
deciding what kind of instruction to offer each individual. Per-
evidence that meet this standard, and we therefore conclude that
haps future research may demonstrate such linkages, but at
the literature fails to provide adequate support for applying
present, we find no evidence for it.
learning-style assessments in school settings. Moreover, several
studies that used appropriate research designs found evidence
that contradicted the learning-styles hypothesis (Massa & Costs and Benefits of Educational Interventions
Mayer, 2006; Constantinidou & Baker, 2002). Finally, even if a It should also be noted that even if the evidence had convinc-
study of a particular learning-style classification and its corre- ingly documented style-by-method interactions—which we
sponding instructional methods was to reveal the necessary have concluded is scarcely the case—the interactions would
evidence, such a finding would provide support for that partic- need to be large and robust, and not just statistically significant,
ular learning-style classification only—and only then if its before the concomitant educational interventions could be
benefits surpass the high costs of student assessments and tai- recommended as cost-effective. After all, there is no doubt that
lored instruction. interventions built around learning styles will be costly. Stu-
Our conclusions have particularly clear-cut implications for dents must be assessed and grouped by learning style and then
educational researchers, in our opinion. We urge investigators given some sort of customized instruction, which, in turn, re-
examining learning-styles concepts to embrace the factorial quires additional teacher training as well as the creation and
randomized research designs described in the earlier ‘‘Inter- validation of instructional activities for each learning style.
actions as the Key Test of the Learning-Styles Hypothesis’’ Moreover, if one is to partition the children within a given
section, because these alone have the potential to provide ac- classroom and teach each subset differently, this may require
tion-relevant conclusions. The kind of research that is needed increasing the number of teachers. Ultimately, the practical
must begin by classifying learners into categories based on question will be whether the benefits of learning-styles inter-

H. Pashler et al.
ventions exceed other ways of using the time and money needed Everybody’s Potential to Learn
to incorporate these interventions. As a final comment, we feel the need to emphasize that all hu-
mans, short of being afflicted with certain types of organic
damage, are born with an astounding capacity to learn, both in
Beliefs Versus Evidence as a Foundation for Educational the amount that can be learned in one domain and in the variety
Practices and Policies and range of what can be learned. Children, unless stifled in
Basic research on human learning and memory, especially re- some way, are usually virtuosos as learners.
search on human metacognition, much of it carried out in the last As we asserted earlier, it is undeniable that the instruction
20 years or so, has demonstrated that our intuitions and beliefs that is optimal for a given student will often need to be guided by
about how we learn are often wrong in serious ways. We do not, the aptitude, prior knowledge, and cultural assumptions that
apparently, gain an understanding of the complexities of human student brings to a learning task. However, assuming that people
learning and memory from the trials and errors of everyday living are enormously heterogeneous in their instructional needs may
and learning. Many demonstrations have shown that partici- draw attention away from the body of basic and applied research
pants who are asked to predict their own future performance on learning that provides a foundation of principles and prac-
following conditions of instruction that researchers know to be tices that can upgrade everybody’s learning. For example, the
ineffective will often predict better performance under poorer finding that learners’ memory for information or procedures can
conditions of instruction than will participants provided with be directly enhanced through testing (Roediger & Karpicke,
better conditions of instruction (for a review, see Schmidt & 2006) is not something that applies to only a small subset of
Bjork, 1992). Part of the problem is that conditions that make learners but (as far as can be told) applies to all. Although
performance improve rapidly during instruction or training, performance of a student on a test will typically depend on that
such as blocking or temporal massing of practice, can fail to student’s existing knowledge, testing (when carried out appro-
support long-term retention and transfer, whereas conditions priately, which sometimes requires providing feedback) appears
that introduce difficulties for learners and appear to slow the to enhance learning at every level of prior knowledge.
learning process, such as interleaving different types of prob- Given the capacity of humans to learn, it seems especially
lems, or employing temporal spacing of practice on what is to be important to keep all avenues, options, and aspirations open for
learned, often enhance long-term retention and transfer. As our students, our children, and ourselves. Toward that end, we
learners, we can also be fooled by subjective impressions, such think the primary focus should be on identifying and introducing
as the ease or sense of familiarity we gain on reading expository the experiences, activities, and challenges that enhance
text or how readily some information comes to mind, both of everybody’s learning.
which can be products of factors unrelated to actual compre-
hension or understanding.
There is growing evidence that people hold beliefs about how SUMMARY
they learn that are faulty in various ways, which frequently lead
people to manage their own learning and teach others in non- Our review of the learning-styles literature led us to define a
optimal ways. This fact makes it clear that research—not intu- particular type of evidence that we see as a minimum precon-
ition or standard practices—needs to be the foundation for dition for validating the use of a learning-style assessment in an
upgrading teaching and learning. If education is to be trans- instructional setting. As described earlier, we have been unable
formed into an evidence-based field, it is important not only to to find any evidence that clearly meets this standard. Moreover,
identify teaching techniques that have experimental support but several studies that used the appropriate type of research design
also to identify widely held beliefs that affect the choices made found results that contradict the most widely held version of the
by educational practitioners but that lack empirical support. On learning-styles hypothesis, namely, what we have referred to as
the basis of our review, the belief that learning-style assessments the meshing hypothesis (Constantinidou & Baker, 2002; Massa
are useful in educational contexts appears to be just that—a & Mayer, 2006). The contrast between the enormous popularity
belief. Our conclusion reinforces other recent skeptical com- of the learning-styles approach within education and the lack of
mentary on the topic (e.g., Coffield et al., 2004; Curry, 1990; credible evidence for its utility is, in our opinion, striking and
Willingham, 2005, 2009). Future research may develop learn- disturbing. If classification of students’ learning styles has
ing-style measures and targeted interventions that can be shown practical utility, it remains to be demonstrated.
to work in combination, with the measures sorting individuals
into groups for which genuine group-by-treatment interactions
can be demonstrated. At present, however, such validation is Acknowledgments—We are grateful to Maxwell Moholy for
lacking, and therefore, we feel that the widespread use of assistance with the literature review. We also thank Elena L.
learning-style measures in educational settings is unwise and a Grigorenko, Roddy Roediger, and two anonymous reviewers for
wasteful use of limited resources. their comments on an earlier version of this article.

Learning Styles
REFERENCES International Learning Styles Network. (2008). About learning styles.

Retrieved August 6, 2009, from http://www.learningstyles.net/
Bostrom, R.P., Olfman, L., & Sein, M.K. (1990). The importance of index.php?option=com_content&task=view&id=20&Itemid=70
learning style in end-user training. MIS Quarterly, 14, 101–119. &lang=en
Callender, A.A., & McDaniel, M.A. (2007). The benefits of embedded Janicki, T.C., & Peterson, P.L. (1981). Aptitude–treatment interaction
question adjuncts for low and high structure builders. Journal of effects of variations in direct instruction. American Educational
Educational Psychology, 99, 339–348. Research Journal, 18, 63–82.
Coffield, F., Moseley, D., Hall, E., & Ecclestone, K. (2004). Learning Jensen, A.R. (1998). The g factor: The science of mental ability. West-
styles and pedagogy in post-16 learning. A systematic and critical port, CT: Praeger.
review. London: Learning and Skills Research Centre. Jung, C.G. (1964). Psychological types: Or, the psychology of individu-
Constantinidou, F., & Baker, S. (2002). Stimulus modality and verbal ation (H. Godwin Baynes, Trans.). New York: Pantheon Books.
learning performance in normal aging. Brain and Language, 82, Keefe, J.W. (1988). Profiling and utilizing learning style. NASSP
296–311. learning style series. Reston, VA: National Association of Sec-
Cook, D.A., Thompson, W.G., Thomas, K.G., & Thomas, M.R. (2009). ondary School Principals.
Lack of interaction between sensing-intuitive learning styles and Kolb, D.A. (1984). Experiential learning. Englewood Cliffs, NJ: Pren-
problem-first versus information-first instruction: A randomized tice-Hall.
crossover trial. Advances in Health Science Education, 14, 79–90. Kolb, D. (1985). Learning style inventory. Boston: McBer.
Cramer, K.A., Post, T.R., & Behr, M.J. (1989). Cognitive restructuring Kozhevnikov, M. (2007). Cognitive styles in the context of modern
ability, teacher guidance, and perpetual distracter tasks: An ap- psychology: Toward an integrated framework of cognitive style.
titude–treatment interaction study. Journal for Research in Psychological Bulletin, 133, 464–481.
Mathematics Education, 20, 103–110. Massa, L.J., & Mayer, R.E. (2006). Testing the ATI hypothesis: Should
Cronbach, L. (1957). The two disciplines of scientific psychology. multimedia instruction accommodate verbalizer-visualizer cog-
American Psychologist, 12, 671–684. nitive style? Learning and Individual Differences, 16, 321–336.
Cronbach, L.J., & Snow, R.E. (1977). Aptitude and instructional McLeod, D.B., & Adams, V.M. (1980/1981). Locus of control and
methods: A handbook for research on interaction. New York: Ir- mathematics instruction: Three exploratory studies. Journal of
vington. Experimental Education, 49, 94–99.
Curry, L. (1990). One critique of the research on learning styles. Ed- McNamara, D.S., Kintsch, E., Butler-Songer, N., & Kintsch, W. (1996).
ucational Leadership, 48, 50–56. Are good texts always better? Interactions of text coherence,
Daniels, R.L., & Stevens, J.P. (1976). The interaction between the in- background knowledge, and levels of understanding in learning
ternal-external locus of control and two methods of college in- from text. Cognition and Instruction, 14, 1–43.
struction. American Educational Research Journal, 13, 103–113. Ormrod, J.E. (2008). Educational psychology: Developing learners (6th
Druckman, D., & Porter, L.W. (1991). Developing careers. In D. Druckman ed.). Upper Saddle River, NJ: Pearson.
& R.A. Bjork (Eds.), In the mind’s eye: Enhancing human perfor- Parent, J., Forward, J., Cantor, R., & Mohling, J. (1975). Interactive effects
mance (pp. 80–103). Washington, DC: National Academy Press. of teaching style and personal locus of control on student performance
Dunn, R. (1990). Rita Dunn answers questions on learning styles. and satisfaction. Journal of Educational Psychology, 67, 764–769.
Educational Leadership, 48, 15–19. Peterson, P.L. (1979). Aptitude Treatment interaction effects of
Foresman, S. (1972). Windows (Teacher’s guide). Glenview, IL: Scott teacher structuring and student participation in college instruc-
Foresman. tion. Journal of Educational Psychology, 71, 521–533.
Freebody, P., & Tirre, W.C. (1985). Achievement outcomes of two Peterson, P.L., Janicki, T.C., & Swing, S.R. (1980). Aptitude–treatment
reading programmes: An instance of aptitude–treatment interac- interaction effects of three social studies teaching approaches.
tion. British Journal of Educational Psychology, 55, 53–60. American Educational Research Journal, 17, 339–360.
Gernsbacher, M.A. (1990). Language comprehension as structure Richardson, A. (1977). Verbalizer–Visualizer: A cognitive style di-
building. Hillsdale, NJ: Erlbaum. mension. Journal of Mental Imagery, 1, 109–126.
Greene, J.C. (1980). Individual and teacher/class effects in aptitude Roediger, H.L. III., & Karpicke, J.D. (2006). The power of testing
treatment studies. American Educational Research Journal, 17, memory: Basic research and implications for educational practice.
291–302. Perspectives on Psychological Science, 1, 181–210.
The Hay Group. (n.d.). One style doesn’t fit all: The different ways people Rotter, J.B. (1966). Generalized expectancies of internal versus exter-
learn and why it matters. Retrieved September 28, 2009, from http:// nal control of reinforcements. Psychological Monographs, 80,
www.haygroup.com/tl/Downloads/Why_People_ Learn.pdf (Whole no. 609).
Henson, R.K., & Hwang, D.-Y. (2002). Variability and prediction of Rundle, S., & Dunn, R. (2007). The Building Excellence Survey [self-
measurement error in Kolb’s Learning Style Inventory scores a directed learning tool]. Retrieved October 23, 2009, from http://
reliability generalization study, 62, 712–727. www.learningstyles.net/index.php?option=com_content&task=
Honey, P., & Mumford, A. (1992). The manual of learning styles. view&id=25&Itemid=78&lang=en
Berkshire, England: Honey, Ardingly House. Snow, R.E. (1977). Individual differences and instructional theory.
Horak, V.M., & Horak, W.J. (1982). The influence of student locus of Educational Researcher, 6, 11–15.
control and teaching method on mathematics achievement. Jour- Schlenker, J. (1978). Matteson Four-Dimensional Reading Project.
nal of Experimental Education, 51, 18–21. Matteson, IL: Reading Service Center.
Horak, W.J., & Slobodzian, K.A. (1980). Influence on instructional Schmidt, R.A., & Bjork, R.A. (1992). New conceptualizations
structure and locus of control on achievement of preservice ele- of practice: Common principles in three paradigms suggest
mentary science teachers. Journal of Research in Science Teach- new concepts for training. Psychological Science, 3, 207–217.
ing, 17, 213–222. Spearman, C. (1927). The abilities of man. London: Macmillan.

H. Pashler et al.
Stensvold, M.S., & Wilson, J.T. (1990). The interaction of verbal ability settings. Educational and Psychological Measurement, 4, 1127–
with concept mapping in learning from a chemistry laboratory 1133.
activity. Science Education, 74, 473–480. Willingham, D.T. (2005, Summer). Do visual, auditory, and kinesthetic
Sternberg, R.J., Grigorenko, E.L., Ferrari, M., & Clinkenbeard, P. learners need visual, auditory, and kinesthetic instruction?
(1999). A triarchic analysis of an aptitude–treatment interaction. American Educator, 29(2), 31–35.
European Journal of Psychological Assessment, 15, 1–11. Willingham, D.T. (2009). Why don’t students like school: A cognitive
Sternberg, R.J., Grigorenko, E.L., & Zhang, L. (2008). Styles of learning scientist answers questions about how the mind works and what it
and thinking matter in instruction and assessment. Perspectives on means for the classroom. San Francisco: Jossey-Bass.
Psychological Science, 3, 486–506. Witkin, H.A., & Goodenough, D.R. (1981). Cognitive styles: Essence
Stricker, L.J., & Ross, J. (1964). An assessment of some structural and origins. New York: International University Press.
properties of the Jungian personality typology. Journal of Abnor- Yale University Graduate School of Arts and Sciences. (2009). Grad-
mal and Social Psychology, 68, 62–71. uate teaching center: Teaching students with different learning
Thurstone, L.L. (1938). Primary mental abilities. Chicago: University of styles and levels of preparation. Retrieved September 28, 2009,
Chicago Press. from http://www.yale.edu/graduateschool/teaching/learning
Twenge, J. (2006). Generation me: Why today’s young Americans are styles.html
more confident, assertive, entitled—and more miserable—than ever Yeany, R.H., Dost, R.J., & Mattews, R.W. (1980). The effects of diag-
before. New York: Free Press. nostic-prescriptive instruction and locus of control on the
Veres, J.G., Sims, R.R., & Shake, L.G. (1987). The reliability and achievement and attitudes of university students. Journal of Re-
classification stability of the learning style inventory in corporate search in Science Teaching, 17, 537–546.
View publication stats

Learning Styles Concepts and Evidence

Uploaded by

Copyright:

Available Formats

Learning Styles Concepts and Evidence

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Learning Styles Concepts and Evidence

Uploaded by

Copyright:

Available Formats

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

Learning Styles: Concepts and Evidence

Article in Psychological Science in the Public Interest · December 2008

Mark A. Mcdaniel Doug Rohrer

SEE PROFILE SEE PROFILE

The user has requested enhancement of the downloaded file.

Learning Styles: Concepts and Evidence

A JOURNAL OF THE ASSOCIATION FOR PSYCHOLOGICAL SCIENCE

Volume 9—Number 3 Copyright r 2009 Association for Psychological Science 105

AN OVERVIEW OF LEARNING STYLES: DOCTRINES

106 Volume 9—Number 3

Volume 9—Number 3 107

108 Volume 9—Number 3

Volume 9—Number 3 109

A Style B Style A Style B Style A Style B Style

A Style B Style A Style B Style A Style B Style

110 Volume 9—Number 3

A Style B Style Method 1 Method 2

A Style B Style Method 1 Method 2

Volume 9—Number 3 111

112 Volume 9—Number 3

Volume 9—Number 3 113

114 Volume 9—Number 3

Volume 9—Number 3 115

116 Volume 9—Number 3

Volume 9—Number 3 117

REFERENCES International Learning Styles Network. (2008). About learning styles.

118 Volume 9—Number 3

Volume 9—Number 3 119

View publication stats

You might also like