YNIMG-08352; No. of pages: 11; 4C:
NeuroImage xxx (2011) xxx–xxx
Contents lists available at ScienceDirect
NeuroImage
j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / y n i m g
Review
Individual differences in cognitive style and strategy predict similarities in the
patterns of brain activity between individuals
Michael B. Miller ⁎, Christa-Lynn Donovan, Craig M. Bennett, Elissa M. Aminoff, Richard E. Mayer
University of California, Santa Barbara, CA, USA
a r t i c l e
i n f o
Article history:
Received 15 March 2011
Revised 18 May 2011
Accepted 19 May 2011
Available online xxxx
Keywords:
fMRI
Inter-individual variability
Cognitive style
Encoding strategies
Memory retrieval
a b s t r a c t
Neuroimaging is being used increasingly to make inferences about an individual. Yet, those inferences are
often confounded by the fact that topographical patterns of task-related brain activity can vary greatly from
person to person. This study examined two factors that may contribute to the variability across individuals in
a memory retrieval task: individual differences in cognitive style and individual differences in encoding
strategy. Cognitive style was probed using a battery of assessments focused on the individual's tendency to
visualize or verbalize written material. Encoding strategy was probed using a series of questions designed to
assess typical strategies that an individual might utilize when trying to remember a list of words. Similarity in
brain activity was assessed by cross-correlating individual t-statistic maps contrasting the BOLD response
during retrieval to the BOLD response during fixation. Individual differences in cognitive style and encoding
strategy accounted for a significant portion of the variance in similarity. This was true above and beyond
individual differences in anatomy and memory performance. These results demonstrate the need for a
multidimensional approach in the use of fMRI to make inferences about an individual.
© 2011 Elsevier Inc. All rights reserved.
Contents
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . .
Methods. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Participants . . . . . . . . . . . . . . . . . . . . . . . . .
Stimuli . . . . . . . . . . . . . . . . . . . . . . . . . . .
Behavioral paradigm . . . . . . . . . . . . . . . . . . . . .
Assessments of cognitive style and strategy . . . . . . . . . .
Functional MRI data acquisition . . . . . . . . . . . . . . . .
Functional MRI data analysis . . . . . . . . . . . . . . . . .
Diffusion tensor imaging (DTI) analysis . . . . . . . . . . . .
Individual variability analysis . . . . . . . . . . . . . . . . .
Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Behavioral analysis . . . . . . . . . . . . . . . . . . . . . .
Individual variability analysis . . . . . . . . . . . . . . . . .
Between-subject variability . . . . . . . . . . . . . .
Hierarchical regression analyses—whole brain . . . . . .
Hierarchical regression analyses—separate brain regions .
Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . .
Appendix A. Supplementary data . . . . . . . . . . . . . . . . .
References . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
Introduction
⁎ Corresponding author at: Department of Psychology, University of California,
Santa Barbara, CA 93106-9660, USA. Fax: +1 805 893 4303.
E-mail address: [email protected] (M.B. Miller).
One of the goals of neuroergonomics is to use neuroscientific
tools to understand the individual mind at work in a naturalistic
1053-8119/$ – see front matter © 2011 Elsevier Inc. All rights reserved.
doi:10.1016/j.neuroimage.2011.05.060
Please cite this article as: Miller, M.B., et al., Individual differences in cognitive style and strategy predict similarities in the patterns of brain
activity between individuals, NeuroImage (2011), doi:10.1016/j.neuroimage.2011.05.060
2
M.B. Miller et al. / NeuroImage xxx (2011) xxx–xxx
environment (Parasuraman and Wilson, 2008). But the use of
functional magnetic resonance imaging (fMRI) to make an inference
about an individual is particularly challenging (see Parasuraman and
Jiang, this issue). Imagine a situation in which an institution (e.g., a
court of law) seeks to use neuroimaging to establish the veracity of an
individual's memory. They have a template based on hundreds of
individuals for what that individual's pattern of brain activity should
look like for a true memory. Yet, the individual in question has a very
different pattern of brain activity. Are they to conclude that the
individual must not be experiencing a true memory? The veracity of
memory is only one dimension that may explain why this individual's
pattern of brain activity is so different from the other individuals. For
example, there may have been fundamental differences in the
individual's cognitive style that significantly affected their pattern of
activity as well. If neuroimaging will be used to make an inference
about an individual, then multiple dimensions in which individuals
may differ must be considered.
We have shown previously that the topographical pattern of brain
activity underlying a memory retrieval task can vary greatly from
individual to individual, sometimes with little to no overlap in
significant activations between individuals (Miller et al., 2002; Miller
et al., 2009). However, despite that extensive variability, the
individual patterns of brain activity are relatively consistent over
time (Miller et al., 2002; Miller et al., 2009; for review of fMRI
reliability, see Bennett and Miller, 2010), suggesting that differences
in the patterns of brain activity are due to systematic differences in
individual characteristics and are not due to random measurement
error (see Fig. 1). In order to effectively use fMRI to infer unique
aspects of the individual mind, it is necessary to untangle the critical
factors that can vary the individual patterns of brain activity. In this
study, we examine whether individual differences in strategy and
cognitive style can account for the degree of similarity between any
two individual patterns of brain activity during a retrieval task.
Using neuroscience methods to gain insight into the individual
mind is a common goal among many institutions, including education
(Byrnes, 2001; Posner and Rothbart, 2005), the military (National
Research Council, 2008), and courts of law (Brown and Murphy,
2010). Implicit in the goals of these institutions is the ability to make
judgments about an individual based on neuroscientific data collected
from a group of individuals. This goal is often incompatible with
the general scientific goal to make an inference about a general
phenomenon that applies to a population by averaging data across
individuals (Faigman, 2010). For fMRI in particular, this demand to
average data across individuals is compounded by the fact that the
signal-to-noise ratio (SNR) of the BOLD signal is very low (Friston
et al., 1999) and that false positives due to the low SNR are far too
common (Bennett et al., 2010). Yet, as acquisition devices and
analytical tools for fMRI become more and more sophisticated,
increasing effort is being made to infer unique aspects of the
individual based on the group data.
A critical question remains within functional neuroimaging: do the
results of a group analysis accurately represent the individuals that
make up that group? Many studies have concluded that it does not
(Heun et al., 2000; Machielsen et al., 2000; McGonigle et al., 2000;
Miller et al., 2002; Feredoes and Postle, 2007; Seghier et al., 2008;
Miller et al., 2009; Seghier and Price, 2009; Parasuraman and Jiang,
this issue). For example, we found that the observed variations in
functional brain activity across the whole brain during a simple
recognition task were extensive, with some individuals activating
mostly prefrontal regions while others activated mostly parietal
regions (Miller et al., 2002; Miller et al., 2009). This was in contrast to
the group analysis, which prominently showed both regions to be
equally active. Are there quantifiable factors that might help to
explain this effect? We found indirect evidence in a recent study that
some variations in the degree of brain activity similarity between
individuals during a memory retrieval task may be due to individual
Fig. 1. Two participants from two scanning sessions held months apart representing the cross-correlation of unthresholded t-maps (retrieval N baseline) across subjects (average
inter-individual correlation across all subjects = .224) and sessions (average intra-individual correlation across all subjects = .482). Each r-value represents the degree of similarity
between individual t-statistic map volumes, i.e., the higher the r, the more similar the pattern of brain activity. The depiction of t-statistic maps is thresholded (p b .001 uncorrected)
for visualization purposes only (adapted from Miller et al., 2009).
Please cite this article as: Miller, M.B., et al., Individual differences in cognitive style and strategy predict similarities in the patterns of brain
activity between individuals, NeuroImage (2011), doi:10.1016/j.neuroimage.2011.05.060
M.B. Miller et al. / NeuroImage xxx (2011) xxx–xxx
differences in strategy (Miller et al., 2009). That is, the larger the
difference in decision criterion the more dissimilar the two patterns of
brain activity. However, a criterion measure is not a direct measure of
memory strategy. Therefore, in this study we directly measure
individual differences in cognitive style and strategy and test whether
or not these differences can account for a significant portion of the
variance in the similarity of brain activity between individuals.
Brain activity during a recognition memory task is a useful
platform to study individual variability for two reasons: (1) the
tremendous amount of previous research conducted on the relationship between recognition memory and fronto-parietal regions
and (2) the widespread and distributed nature of the activity underlying the task (Miller and Van Horn, 2007). Many of the frontoparietal regions implicated in recognition memory are known to
underlie cognitive processes peripheral to the actual retrieval process
(Shimamura, 1995; Fletcher et al., 1998; Moscovitch and Winocur,
2002). One potential implication of this distributed architecture is
that one and the same behavioral outcome—such as an “old” response
on a recognition test—could be the result of information processing
and neural circuitry that are distinct to the individual but vary across
individuals.
There is considerable evidence that individuals engage in a variety
of strategies during the encoding phase of a standard recognition
memory task (Stoff and Eagle, 1971; Battig, 1975; Weinstein et al.,
1979; Paivio, 1983; Reder, 1987; Graf and Birt, 1996) and that individual differences in strategy can modulate the BOLD activity in
certain brain regions (Savage et al., 2001; Casasanto et al., 2002; Speer
et al., 2003; Kondo et al., 2005; Tsukiura et al., 2005). One notable
study by Kirchhoff and Buckner (2006) identified the various
strategies people adopt during the unconstrained encoding of an
unrelated pairs of pictures. They found that verbal elaboration
correlated with activity in prefrontal regions associated with
controlled verbal processing, while visual inspection correlated with
activity in the extrastriate cortex known to be involved in higherorder visual processing. These findings suggest that different encoding strategies may recruit different brain regions, even though the
memory performance is similar between the two strategies. While the
Kirchhoff and Buckner (2006) study showed how different encoding
strategies can recruit different brain regions during the encoding
phase, we investigated how different encoding strategies may account
for the similarity in brain activity during the retrieval phase. One of
the classic principles of episodic memory, the encoding specificity
principle, states that encoding operations are directly related to
retrieval operations (Tulving and Thomson, 1973). This relationship is
illustrated by the fact that differences in encoding strategies can have
a direct effect on the brain activity that occurs during retrieval
(Raposo et al., 2009; Kirchhoff, 2009). From this evidence, we
hypothesize that individual differences in encoding strategy will
account for a significant portion of the variance that is observed in the
similarity of brain activity between individuals during a retrieval task.
Aside from the particular strategy that an individual may choose
to engage in during a memory task, individuals may also have a
particular style of thinking or a preferred set of cognitive operations
that could affect their pattern of brain activity. For example, the
visualizer–verbalizer dimension of cognitive style is based on the idea
that some people (who could be called visualizers) are better at
processing visual material, whereas other people (who could be called
verbalizers) are better at processing verbal material (Mayer and
Massa, 2003; Massa and Mayer, 2006). Although individual differences in the tendency to engage in visual or verbal processing has
little relationship to memory recall (Richardson, 1978, 1998) and the
search for research-based attributes that may interact with instructional methods has had a somewhat disappointing history (Cronbach
and Snow, 1977; Pashler et al., 2009; Sternberg and Zhang, 2001;
Zhang and Sternberg, 2009), differences in cognitive style may still be
pertinent to the issue of individual differences in brain activity during
3
a retrieval task. The likelihood of involvement of any particular
process and its underlying brain region may depend a great deal on
the processing style of the individual. In other words, a visualizer and
verbalizer may not necessarily differ in performance on learning
outcome tests but may have highly distinct patterns of regional brain
activity during learning (i.e., encoding) and retrieval that achieve the
same level of performance. There is evidence suggesting that this
may well be the case. Two recent studies by Kozhevnikov et al.
(2002)) and Kozhevnikov et al. (2005) attempted to clarify and revise
the visualizer–verbalizer dimension and subsequently introduced two
subtypes of visualizers: object visualizers who tended to focus on
object properties such as shape and color and spatial visualizers who
tended to focus on spatial properties such as location and spatial
relations. A later fMRI study revealed that object visualizers activated
more ventral regions of the visual processing stream while spatial
visualizers activated more dorsal regions of the visual processing
stream (Motes and Kozhevnikov, 2006). We hypothesize that differences in cognitive style may also be a significant contributor to the
variability of the patterns of brain activity during a retrieval task.
In this study we test the hypothesis that individual differences in
strategy and cognitive style will account for a significant portion of the
variance in the similarity in the patterns of whole-brain activity
between individuals during a memory retrieval task. To measure the
variability in the patterns of brain activity across individuals we crosscorrelate the unthresholded t-statistic maps contributed by each
individual from the retrieval task (Miller et al., 2002; Miller et al.,
2009). The more similar two individual patterns of brain activity, the
higher the correlation value (see Fig. 1). Individuals were assessed on
strategies and cognitive style after the fMRI scanning session using a
battery of tests. In addition, strategy was implicitly manipulated
within participants by varying the imageability of the word stimuli. If
strategies and style depend to some degree on visualizing the word
stimuli, then manipulating the imageability of the words should affect
the predictability of those factors. We predict that individual
differences in strategy and cognitive style will significantly account
for inter-individual differences in brain activity above and beyond any
factors attributable to individual differences in anatomy and memory
performance.
Methods
Participants
A group of 50 participants (age 18–55, M = 25.8) were recruited
from the undergraduate and graduate student population at University of California, Santa Barbara (UCSB) and were paid for their
participation. Data from three participants was excluded due to
excessive motion (1), scanner malfunction (1), or withdrawal from
the study (1). Two more participants were excluded because they
were left-handed. The remaining 45 participants were comprised of
22 men and 23 women. All participants gave informed consent as
approved by the UCSB Institutional Review Board.
Stimuli
Participants completed two study–test sessions, the order of which
was counterbalanced across participants. In one session, participants
were presented lists of highly imageable, concrete nouns, while the
other session presented lowly imageable, abstract nouns. Words were
chosen to be 4–12 letters in length. High-imageability words had
imageability scores greater than one standard deviation above the
mean on the MRC Psycholinguistic Database Imageability rating
(Coltheart, 1981) while low-imageability words were chosen to be
greater than 1 SD below the mean. Within each study-test session the
word order was randomized. Whether the word was old or new was
also counterbalanced across participants.
Please cite this article as: Miller, M.B., et al., Individual differences in cognitive style and strategy predict similarities in the patterns of brain
activity between individuals, NeuroImage (2011), doi:10.1016/j.neuroimage.2011.05.060
4
M.B. Miller et al. / NeuroImage xxx (2011) xxx–xxx
Behavioral paradigm
This experiment employed an event-related fMRI design and
included two study–test sessions—one for the low-imageability words
and the other for the high-imageability words. Each study session
consisted of 106 words that were presented for 2 s each. These study
word trials were randomly intermixed with eighty 2-s fixation trials
(‘+’). During the study session participants were simply instructed to
try and learn the words for a later memory test. There was a 10-min
interval between study and test phases. During this time, structural
scans (first high-resolution T1 MPRAGE and then diffusion tensor
imaging scans) were collected. The test sessions consisted of 318
events: 212 test words (50% studied and 50% new) and 106 fixation
trials, again randomly intermixed. Participants were instructed to
respond with their index finger if an item was studied (i.e., “old”) and
with middle finger if the item was new. Responses were collected
using a MR-compatible button box held in the right hand. Once the
scanning portion of the experiment was complete, participants
completed an assessment of their cognitive style and strategies.
Assessments of cognitive style and strategy
Cognitive style was assessed using a visualizer–verbalizer test
battery. This consisted of (1) a participant questionnaire on Verbal–
Spatial Ability Rating and Verbal–Visual Style Rating (Mayer and
Massa, 2003); (2) the Card Rotation Test (Ekstrom et al., 1976);
(3) the Paper Folding Test (Ekstrom et al., 1976); (4) a Vocabulary
Test (18 items adapted from the Vocabulary scale of the Armed
Services Vocational Aptitude Battery (ASVAB) as selected from a
test preparation book (Hogan and Cannon, 2007); (5) the Verbalizer–
Visualizer Questionnaire (VVQ; Richardson, 1977); (6) the Santa
Barbara Learning Style Questionnaire (Mayer and Massa, 2003); and
(7) the Object–Spatial Imagery Questionnaire (Blajenkova et al.,
2006). The results from the Visualizer–Verbalizer test battery data
were submitted to a factor analysis using Principle Components
Analysis and VARIMAX rotation to yield four orthogonal factors that
are related to visualizing and verbalizing tendencies. Factors were
interpreted and named based on the weightings of each of the
measures. Prior research has been conducted showing the relationship of these scales to visualizing and verbalizing cognitive styles
(Mayer and Massa, 2003; Massa and Mayer, 2006). Factor scores were
then computed for each participant from the linear combinations of
scale scores.
Cognitive encoding strategies were measured using a strategy
questionnaire adapted from Kirchhoff and Buckner (2006). Although
this questionnaire is meant to probe the strategies used during the
study phase of the experiment, those encoding strategies would be
related to the strategies and processes utilized during retrieval as well
(Tulving and Thomson, 1973). This questionnaire consisted of 10
strategy descriptions (e.g., “Repeated the words to yourself” or
“Formed a picture of the word in your mind”) and instructed the
participant to rate on a 5-point scale how often they used each
strategy during the study phase (see Appendix for a sample of the
Questionnaire). A separate questionnaire was used for the set of highimageability words and the set of low-imageability words. As in the
analysis above, the ratings were submitted to a Principle Components
Analysis and VARIMAX rotation to yield four components with
eigenvalues greater than 1.0.
After the scanning was complete, in addition to the questionnaire,
cognitive strategy was probed by simply asking the participants their
memory strategy. Questions were asked regarding the first set of
words, the second set of words, whether there was a difference when
studying the two sets of words, and how they made a decision as to
whether the word was studied or not. The participants answered
these questions through free response. These responses were then
categorized as either verbal or visual by three blind raters (50%
agreement across all three raters and 87% agreement across two
raters). One of the authors then mediated any disagreements using
the general rule that any indication of visual imagery would be coded
as visual.
Functional MRI data acquisition
Functional images were acquired with gradient-recalled echoplanar imaging (TR = 2000 ms, TE = 30 ms, RF flip angle = 90, gradient-echo pulse sequence, 33 contiguous axial slices at 3.0 mm thick
with a 0.5 mm slice gap, and an in-plane resolution of 64 × 64 pixels in
a FOV of 192 cm, producing voxels of 3 mm × 3 mm × 3 mm) on a 3 T
Siemens Trio MRI scanner equipped with high-performance gradients.
Each BOLD run was preceded by four scans to allow for steady-state
magnetization to be approached. In addition to the functional scans,
a high-resolution T1-weighted structural image was acquired using a
3-D SPGR pulse sequence (TR = 25 ms, TE = 6 ms, RF flip angle = 25°,
bandwidth = 15.6 kHz, voxel size = .9375 mm × 1.25 mm × 1.2 mm).
Diffusion-weighted MRI data were acquired using a diffusion weighted,
single-shot spin-echo, echo-planar sequence (TR = 9022 msec;
TE=91 msec; flip angle=90°; slice thickness=2.0 mm; number of
slices=70 (axial); FOV=240 mm; matrix size=128×128; acquisition
time = 5:24 min). Diffusion weighting (b-value= 1000 s/mm 2) was
applied along 32 directions with one additional reference image
acquired having no diffusion weighting (b-value= 0 s/mm2). Foam
padding was used to minimize head motion.
Functional MRI data analysis
Initial data preprocessing utilized SPM5 (Wellcome Department of
Cognitive Neurology, London, UK) for slice acquisition correction,
motion correction, coregistration, and spatial normalization. The
spatially normalized scans were then smoothed with an 8 mm FWHM
isotropic Gaussian kernel to accommodate anatomical differences
across participants. Statistical analysis was conducted using customized
Matlab scripts implementing a standard least-squares voxel-wise
general linear model, as described in Guerin and Miller (2009). There
were 10 separate parameters to model each 2-s post-stimulus time
point up to 20 s (Ollinger et al., 2001). This unique set of 10 parameters
was utilized for each trial type: old-high imageable, new-high imageable, old-low imageable, new-low imageable. To contrast responses
between trial types, we calculated the mean of the 2nd, 3rd, and 4th
time points of the estimated event-related response. In addition to the
parameters already discussed, parameters were included to model
linear drift within each session and the session-specific means. The
critical contrast used for the individual variability analysis was retrieval
greater than fixation for both the high-imageability condition and the
low-imageability condition. The t-statistic maps from these contrasts
that were input into the cross-correlation analysis were unthresholded
so that a particular result would not be determined by the setting of an
arbitrary threshold (see Miller et al., 2009). The thresholds used for Fig. 2
are for visualization purposes only.
Diffusion tensor imaging (DTI) analysis
All analyses were carried out using the SPM Tools diffusion
toolbox as implemented in SPM5 (http://sourceforge.net/apps/trac/
spmtools/). For preprocessing, diffusion-weighted images were motioncorrected and coregistered to the high-resolution T1-weighted image,
which we spatially normalized to the MNI template brain. The resulting
normalization parameters were subsequently applied to the diffusionweighted images, reorienting the gradient directions accordingly.
Following preprocessing, second-order diffusion tensors and fractional
anisotropy (FA) values were established using the standard multiple
regression approach. Individual FA images reflect the coherence of water
diffusion on a voxel-by-voxel basis.
Please cite this article as: Miller, M.B., et al., Individual differences in cognitive style and strategy predict similarities in the patterns of brain
activity between individuals, NeuroImage (2011), doi:10.1016/j.neuroimage.2011.05.060
M.B. Miller et al. / NeuroImage xxx (2011) xxx–xxx
5
Fig. 2. Display of widespread inter-individual variability across the retrieval condition with high-imageability words. The large rendering in the top left represents the results of a
second-level group analysis of the data. The remaining 45 renderings each represent the results from one subject for the retrieval(high) N baseline contrast. The threshold for all
images was identical and set to p b 0.005 (uncorrected) for purposes of illustration. The mean inter-individual correlation value across the entire dataset was 0.35.
Individual variability analysis
To quantify the variability in individual patterns of brain activity
across the whole brain or across particular regions the unthresholded
t-statistic map image volumes were cross-correlated across participants. This was done separately for each imageability condition. If one
takes a three-dimensional volume of continuous values in each voxel
and correlates that volume with another three-dimensional volume of
continuous values in the same voxel matrix and atlas space then the
result will be a single correlation value that represents the similarity
between two volumes (see Miller et al., 2009, for a discussion of the
method). A mask generated during the group analysis of all 45
participants was used so that only voxels with a signal intensity value
for every participant were included in the analysis. The crosscorrelation of 45 individual t-statistic maps produced 990 observations of unique pairings between different individuals. The resulting
correlation values were then submitted to a multivariate hierarchical
regression analysis with 60 predictor variables. Each factor was
entered into the regression equation as part of a set of factors based on
their theoretical relevance (see Table 1). Each set of factors is
described below in the order that they were entered into the
regression equation. It was critical to enter cognitive style and
strategy after the other variables so that any relationship between
them and the correlation values could not be attributed to differences
in another characteristic.
1. Procedural was a single factor representing the presentation order
(high- or low-imageability blocks first), coded as 0 for same and 1
2.
3.
4.
5.
for different. This was considered a nuisance variable and therefore
it was entered first.
Demographics included variables for the similarity in age and sex.
Differences in the similarity of brain activity could be attributed to
differences in age and/or sex of the individuals, so this factor was
entered next.
Anatomy included a cross-correlation of each individual's normalized high-resolution anatomical (mprage) image and of each
individual's normalized diffusion tensor image (FA). Similar to the
demographics factors, any differences in the similarity of brain
activity could be attributed to structural differences in anatomy as
measured by a high-resolution T1 scan or by DTI. Therefore, this
factor was entered next.
Performance was represented by the difference between the
individuals' measures in d´ and in criterion. Any of the predicted
differences in similarity of brain activity that could be accounted
for by differences in cognitive style and strategy could be attributed
instead to differences in memory performance. Therefore, it was
necessary to account for these factors prior to entering in the sets
having to do with cognitive style and strategy.
Encoding Strategy was represented by five variables. The first
variable represented the categorization of the individual's response
to the open-ended strategy question, coded as 0 for same and 1 for
different across the two study sessions. The prediction was that
different strategies will have lower correlations. Strategy was
further represented by the four factor scores from the strategy
questionnaire. The four factors represented four different strategies
for high-imageability words (verbal, visual, categorization, and
Please cite this article as: Miller, M.B., et al., Individual differences in cognitive style and strategy predict similarities in the patterns of brain
activity between individuals, NeuroImage (2011), doi:10.1016/j.neuroimage.2011.05.060
6
M.B. Miller et al. / NeuroImage xxx (2011) xxx–xxx
Table 1
The unique variance (standardized betas) contributed by each factor in the prediction of the degree of similarity in the patterns of brain activity between individuals during a
memory retrieval task.
High imageability
Procedural
Demographics
Anatomy
Performance
Encodingstrategya
Cognitivestyle
Order
Age
Sex
MPRAGE
DTI (FA)
Memory (d´)
Criterion
Open-ended
Verbal
Visual
Categorical
Rote/Elaborate
Verbal
Visual
Imagery
Spatial
Low imageability
Whole
Frontal
Parietal
Tempo.
Occip.
Whole
Frontal
Parietal
Tempo.
Occip.
−.015
−.042
.019
.100
.281*
−.006
−.060
−.009
−.100
−.044
−.065
.106*
.016
−.139*
−.081
−.025
−.002
.080
−.009
.010
.205*
−.010
−.004
.005
−.073
.024
−.141*
.129*
−.029
−.129*
−.016
−.032
−.022
−.016
.037
.017
.152*
−.013
.082
−.043
−.066
.011
−.036
.070
.033
−.131*
−.158*
−.055
.003
−.120*
.030
.032
.284*
−.015
−.115*
−.018
−.147*
.027
.055
.064
.047
−.174*
−.114*
−.020
−.005
−.004
.013
.100
.256*
.020
−.023
−.002
−.061
−.161*
−.115*
.048
.100
−.024
−.019
.040
−.028
−.100*
.005
.123*
.407*
.005
−.037
.130*
−.068
−.171*
−.176*
.107*
.052
−.020
−.112*
−.040
−.022
−.116*
.001
.017
.351*
.039
−.052
.168*
−.106*
−.180*
−.205*
.096
.005
−.004
.023
−.010
−.029
−.078
.057
.058
.247*
.020
.067
.112*
−.091
−.199*
−.181*
.105
.046
.000
−.105*
−.067
−.031
−.012
.014
.080
.412*
−.052
−.063
.029
−.013
−.059
−.102*
.092
.065
−.057
−.114*
−.076
−.005
−.118*
−.016
.118*
.170*
−.032
−.011
.071
−.005
−.160*
−.072
.033
.159*
.027
−.130*
.002
Beta values in bold type are significant at the p b .01 level, and those with an * are significant at the p b .001 level. aThe five factors were determined separately for the high and lowimageability conditions, but those factors with a similar intuitive label were included in the same row (see Supplementary Materials).
rote memorization) and four different strategies for low-imageability words (verbal, visual, categorization, and elaboration) (see
Supplemental Material). Only the factor scores relevant to the
high-imageability words were entered into the equation with brain
activity associated with the high-imageability condition, and vice
versa. The prediction was the bigger the difference in factor scores,
the lower the correlation.
6. Cognitive Style was represented by the four factors derived from the
battery of visualize–verbalize tests. The four factors represented
four different cognitive styles: verbal, visual, imagery, and spatial
(see Supplemental Material). The prediction was the bigger the
difference in factor scores, the lower the correlation. The set of
cognitive style factors was entered after the set of strategy factors
for arbitrary reasons, and the order of these two sets had little
effect on the results (indicating little shared variance between the
two sets).
7. Individual Residuals represented 44 dummy variables, each one
representing whether or not a particular individual was a part of
the correlation pair. As reported in Miller et al. (2009), some
individuals are more deviant from the group than others in their
patterns of brain activity. It was necessary to enter these variables
last because some of the differences between individuals were
expected to be captured by the variables previously entered, but
there could be differences in certain characteristics that we have
not captured. These variables were meant to illustrate unique
aspects of individuals that have yet to be represented.
Results
Behavioral analysis
A signal detection analysis was separately conducted on the
results of the recognition test for the high-imageability words and
the low-imageability words. Using a repeated-measures ANOVA we
found that memory accuracy, or d´, was significantly higher for the
high-imageability words (1.51) than the low-imageability words
(0.84) (F(1,44) = 81.25, MSE = .123, p b .001), while criterion, C, was
not significantly different between the high-imageability words
(0.18) than the low-imageability words (0.13) (F(1,44) = 1.80,
MSE = .029, n.s.).
Details of the results of the factor analysis for both cognitive style
and strategy are contained in the Supplemental Material. In brief, the
results from the battery of visualizing and verbalizing tests loaded onto
four style factors that we labeled based on an intuitive evaluation of
the component scores as verbal (Factor 1), visual (Factor 2), imagery
(Factor 3), and spatial (Factor 4). The results from the strategy
questionnaire similarly loaded onto four factors for the both the highimageability and the low-imageability words labeled as verbal
(Factor 1), visual (Factor 3 for high imageability and Factor 2 for the
low imageability), categorical (Factor 2 for high imageability and
Factor 3 for the low imageability), rote memory for high-imageability
words (Factor 4) and elaboration for low-imageability words (Factor 4).
Correlation tests (with alpha at .05) showed that individual
differences in the visualizing and verbalizing style factor scores had no
significant relationship with memory (d´) in either the high imagery
cognition or the low imagery condition. Some of the strategy factors
did have a significant relationship to d´, but only in the high imagery
condition. The visual strategy factor (Factor 3) in the high imagery
condition strongly correlated with memory performance (d′):
r = .446, p b .002. In contrast, the rote strategy factor (Factor 4) had
a negative effect on memory performance: r = −.407, p b .006. Since
the words used in the high imagery condition were those that were
highly imageable, it makes sense that a visual strategy may benefit
long-term memory performance.
Individual variability analysis 1
Between-subject variability
The extreme difference in the patterns of brain activity between
individuals is evident in Figs. 2 and 3. The findings replicate our
previous studies (Miller et al., 2002; Miller et al., 2009) in that the
average correlation between high-imageability and low-imageability
words within the same participant (.626) was significantly higher
than the average correlation between high-imageability and lowimageability words between different participants (.341) (F(1,2023) =
312.15, MSE= .011, p b .001). There was also a difference in variability
1
Although the focus of this study was on the variability of brain activity during the
retrieval phase of the task, we did have data on the brain activity during the encoding
phase as well. However, the fMRI data for the encoding phase turned out to possess a
much lower SNR relative to the retrieval phase (the average correlation between
participants for encoding was .185 compared to .358 for retrieval). This was most
likely due to the fact that only half as many trials were collected during the study
phase. Any inferences that could be drawn about the differences in variability between
encoding and retrieval (e.g., differences in encoding strategy were a significant factor
for the variability in brain activity during retrieval but not during encoding) would be
confounded by the differences in the number of trials.
Please cite this article as: Miller, M.B., et al., Individual differences in cognitive style and strategy predict similarities in the patterns of brain
activity between individuals, NeuroImage (2011), doi:10.1016/j.neuroimage.2011.05.060
M.B. Miller et al. / NeuroImage xxx (2011) xxx–xxx
7
Fig. 3. Display of widespread inter-individual variability across the retrieval condition with low-imageability words. The large rendering in the top left represents the results of a
second-level group analysis of the data. The remaining 45 renderings each represent the results from one subject for the retrieval(low) N baseline contrast. The threshold for all
images was identical and set to p b 0.005 (uncorrected) for purposes of illustration. The mean inter- individual correlation value across the entire dataset was 0.37.
between the imageability conditions, with the average correlation of
low-imageability words (.371) being significantly higher than the
average correlation of high-imageability words (.346) (F(1,990) =
56.96, MSE = .005, p b .001).
Hierarchical regression analyses—whole brain
The results are summarized in Fig. 4 (which collapses the results
across imageability conditions) and Table 1 (which lists the
standardized betas for each factor). For high-imageability words,
the full model accounted for 79.3% of the variance in correlation
values across pairs of whole-brain patterns of activity. The model
first takes into account the differences in the order of the conditions (procedural), which was not significant (R 2 change = .000,
F(1,988) = 0.05, n.s.); then differences in demographics (R2 change=
.006, F(2,986) = 3.09, p = .046); differences in anatomy (R2 change=
.108, F(2,984) = 59.83, p b .001); and differences in performance (R2
change= .006, F(2,982)= 2.46, n.s.). Above and beyond those differences, differences in strategy were significantly related to the variability
(R 2 change = .023, F(5,977) = 5.16, p b .001). Above and beyond
differences in strategy and the other characteristics, differences in
cognitive style were also significantly related to the variability (R2
change= .025, F(4,973) = 7.25, p b .001). The shared variance between
the set of cognitive style factors and the set of strategy factors was less
than 2%. Finally, another 62.7% of the variance in correlation values could
be accounted for by the dummy variables coding for each individual (R 2
change= .627, F(44,929) = 63.81, p b .001). For the low-imageability
condition, the full model accounted for 82.3% of the variance in
correlation values. The results were similar to the high-imageability
condition except that anatomy and strategy differences accounted for a
much larger portion of the variance in the correlation values. Again,
the model first takes into account the differences in the order of
the conditions (procedural), which was not significant (R2 change=
.002, F(1,988) = 1.63, n.s.); then differences in demographics (R 2
change= .012, F(2,986) = 6.13, p = .002); differences in anatomy (R2
change= .252, F(2,984) = 168.72, p b .001); and differences in performance (R 2 change = .002, F(2,982)= 1.22, n.s.). Above and beyond
those differences, differences in strategy were significantly related to
the variability (R2 change= .094, F(5,977) = 28.75, p b .001). Above and
beyond differences in strategy and other characteristics, differences in
cognitive style were also significantly related to the variability (R2
change= .017, F(4,973) = 6.63, p b .001). Finally, another 44.5% of the
variance in correlation values could be accounted for by the dummy
variables coding for each individual (R 2 change= .445, F(44,929) =
53.23, p b .001).
Hierarchical regression analyses—separate brain regions
An identical analysis as the one used for the whole brain was
conducted separately for the cortices of each of the lobes. Masks for
each lobe were generated using the Wake Forest Pickatlas Matlab
toolbox (Maldjian et al., 2003). The results for each lobe were very
similar to the whole brain, with the differences summarized in Fig. 4
and Table 1 with the standardized betas for each factor listed. One of
the notable differences between the lobes was in the low-imageability
condition, in which strategy differences (i.e., differences in the
particular strategies that individuals chose to utilize for that particular
set of words) had a much stronger relationship to variability in the
Please cite this article as: Miller, M.B., et al., Individual differences in cognitive style and strategy predict similarities in the patterns of brain
activity between individuals, NeuroImage (2011), doi:10.1016/j.neuroimage.2011.05.060
8
M.B. Miller et al. / NeuroImage xxx (2011) xxx–xxx
Fig. 4. Results of the hierarchical regression analysis predicting the similarity in the patterns of brain activity between individuals collapsed across the high and low-imageability
conditions. Each set of factors was entered in the order depicted in the legend. The far left chart represents the results for the whole brain, while the other four represent the results
for the cortex of each of the four lobes. The number in parentheses is the average correlation between individuals for that brain region.
frontal and parietal lobes than in the temporal and occipital lobes.
Another notable difference between the lobes was in the highimageability condition, in which cognitive style differences (i.e.,
differences in the general tendency of cognitive style regardless of the
particular set of words) had a stronger relationship to variability in
the parietal and temporal lobes than in the frontal and occipital lobes.
Most of the direct effects reported in Table 1 were in the expected
direction. The direct effects represent the unique variance of each of
the factors derived from the regression equation prior to entering the
individual residuals. Some of the notable relationships include the
following: the bigger the difference in age, the lower the correlation
values (i.e., the more variable), particularly in the low-imageability
condition, while the difference in sex had no relationship to
correlation values in either condition. As for anatomy, differences in
white matter integrity as measured by the FA maps were a much
stronger factor than overall differences in anatomy as measured by
the high-resolution T1 MPRAGE. The more similar the white matter
maps, the more similar the functional maps for every lobe in both
conditions. This was clearly the strongest factor in every analysis.
Differences in performance had little relationship to variability, except
for criterion differences in the temporal lobe in the high-imageability
condition.
The critical factors were in the set labeled cognitive style or the set
labeled strategy. While most of the direct effects for these factors were
in the predicted direction, a few factors were surprising. For example,
three factors had significant effects in the direction opposite from the
expected one (positive betas): differences in verbal tendency in the
occipital lobe, differences in the open-ended query about strategy in
the low-imageability condition in the frontal and parietal lobe, and
differences in the fourth strategy factor across several regions. In these
cases, the standardized beta was positive indicating that the bigger
the difference between the two individuals the more similar their
patterns of brain activity. However, this may be due to the imbalance
in the number of individuals within each of these categories. For
example, in the low-imageability condition, most of the participants
utilized a verbal strategy. Only five individuals maintained a visual
strategy. Yet, for all five of these participants, the average correlation
between them and the rest of the participants was higher than the
overall average, i.e., these five individuals happened to be more like
the group than most of the other individuals. Therefore, due to the low
number of individuals and the characteristics of this group, the
difference in strategy factor is biased towards being more similar to
the group than maybe would be represented if we had a larger
sample.
Most of the significant factors in these sets were in the predicted
direction (negative betas). That is, the more two individuals were
different in their strategy and cognitive style, the more the individuals
were different in their patterns of brain activity. One critical factor was
the tendency to visualize, which was the main target in manipulating
the imageability of the words. As shown in Table 1, the bigger the
difference in the tendency to visualize the lower the correlation value,
but only in the high-imageability condition and not in the lowimageability condition as predicted. While, in general, cognitive style
factors seemed to play a slightly larger role in the high-imageability
condition, strategy factors clearly played a larger role in the lowimageability condition and most notably in the frontal and parietal
regions. Differences in the two strongest strategy factors (verbal and
visual strategies) were strongly related to variability in the lowimageability condition in the frontal and parietal lobes but not in the
high-imageability condition. Interestingly, differences in the visual
strategy factor scores were significantly related to memory performance (d´) in the high imagery condition but not the low imagery
condition, while differences in the same factor scores were significantly related to similarity in brain activity in the low imagery
condition but not the high imagery condition.
Discussion
What makes the pattern of brain activity so similar for two
individuals and, yet, so different for another two individuals? We
found, as predicted, that individual differences in cognitive style and
encoding strategy during a memory retrieval task were significant
Please cite this article as: Miller, M.B., et al., Individual differences in cognitive style and strategy predict similarities in the patterns of brain
activity between individuals, NeuroImage (2011), doi:10.1016/j.neuroimage.2011.05.060
M.B. Miller et al. / NeuroImage xxx (2011) xxx–xxx
factors in explaining this variability. For example, the more different
two individuals were in their tendency to visualize highly imageable
word stimuli, the more different their two patterns of brain activity
were across the whole brain. Further, we found that the relationship
between differences in cognitive style and strategy and differences in
the similarity of brain activity was significantly affected by implicit
manipulations of strategy, in this case, manipulating the imageability
of the word stimuli.
We defined strategy as a specific set of cognitive operations that an
individual utilized to better encode certain word stimuli, such as
forming an image of the word in their mind or using the word in a
sentence. We found participants used a combination of operations or
strategies during a particular block of trials. Thus, certain strategies
tended to group together, depending on the imageability condition and
that those groupings (factors) could be intuitively identified as visual,
verbal, or otherwise (see Supplemental Material). Likewise, we defined
cognitive style as a specific set of cognitive processes that may be similar
to the cognitive operations engaged by specific strategies, but we
defined them separately because they represent general processing
tendencies by the individual that are not necessarily a direct or explicit
strategy, like the tendency to visualize written material. A number of
visualizing–verbalizing test batteries were completed by the participants and the results of those tests also tended to group together into
intuitively identifiable factors, like visual tendency or verbal tendency.
While the two sets of factors, cognitive style and strategy, may have
engaged similar processes, they accounted for unique proportions of the
variance in the similarity of brain activity. In the analysis of the whole
brain, both sets of factors together accounted for 8.0% of the variance in
the similarity of brain activity, while less than 0.3% of the variance was
shared between the two sets. Each set of factors was a significant and
unique predictor of similarity in brain activity.
It should be noted that not just any difference between individuals
could be related to the similarity in their patterns of brain activity. For
example, whether or not the two individuals were the same sex had
no effect on their brain similarity. Also, the difference in age between
the two individuals had no relationship to similarity in brain activity
in the high-imageability condition, but it did have a significant
relationship in the low-imageability condition. We also found that
differences in memory performance had little relationship to
differences in the similarity of brain activity. Even though differences
in memorability (d´) are known to modulate the activity in key brain
regions (Nyberg et al., 1996; Tulving et al., 1999; Kirchhoff, 2009), it
appears to have little relationship to differences in the topography of
brain activity across large regions.
In our previous study we found that correlations of spatially
normalized high-resolution anatomical images were not related to
correlations of functional brain activity (Miller et al., 2009). We
surmised, though, that finer measures of anatomical differences might
eventually prove to be related. In the current study, we found that the
correlations of the high-resolution MPRAGE images were significantly
related to the correlations of functional brain activity across the whole
brain, but the effect was inconsistent across different cortical regions.
It was significant in the occipital lobe, but not in the frontal, parietal or
temporal lobes.
Still, not all differences in anatomy were so weakly related to
differences in functional brain activity. Clearly one of the strongest
factors overall was the difference in white matter integrity as
measured by fractional anisotropy. The effect was significant across
every region of the brain, but it was strongest in the temporal cortex
(with standardized betas as high as .412). The study of individual
differences in white matter connectivity has been of great interest
across neuroscience (Wolbers et al., 2006; Boorman et al., 2007; Niogi
et al., 2010; Tomassini et al., 2011) and this finding is consistent with
that literature demonstrating the greater the difference in white
matter integrity between two individuals the bigger the difference in
task-related functional brain activity. Although the degree of
9
similarity between individual DTI images seemed to have little
relationship to differences in actual memory performance, it clearly
had a significant impact on the similarity of brain activity underlying
memory performance. This would suggest that similar memory
performance on a recognition test could be associated with very
different patterns of brain activity, and that individual differences in
white matter connectivity plays a major role in that variance.
The most striking factor, though, as depicted in Fig. 4, was the
leftover amount of variability that could be attributed to specific
individuals (some individuals were more deviant in their pattern of
brain activity than others). For example, in the whole-brain analysis,
the full model accounted for 81% of the variance in the similarity of
brain activity, but only 27% of the variance could be attributed to
unique characteristics of the individuals that we were able to identify,
with the differences in cognitive style and strategy being two of those
factors. That means that there was another 54% of the variance that
was attributable to specific individual traits but that we have not been
able to identify yet. This variability could conceivably be due to
differences in cognition but also due to differences in personality,
physiology, state of mind, and/or other factors, such as caffeine intake.
Though a large amount of variance remains unexplained, it is clear
that differences in cognitive style and strategy are significant predictive
factors. In previous studies and reports on brain activity associated with
memory retrieval (Miller and Van Horn, 2007; Miller et al., 2009) we
have proposed that differences in strategy could account for some of
that variability. Episodic memory retrieval engages widespread regions
throughout the cortex, and includes cognitive processes that might be
peripheral to the actual retrieval of episodic information (Shimamura,
1995; Fletcher et al., 1998; Moscovitch and Winocur, 2002). A single
“old” response on a recognition test could be accomplished along
multiple processing routes through the brain, many of which may
depend on the strategy of the individual. The Encoding Specificity
Principle states that memory performance relies on the relationship
between processes engaged during retrieval and the processes engaged
during encoding (Tulving and Thomson, 1973). Previously, it has been
shown that differences in encoding processes can have a direct effect on
the brain activity that occurs during retrieval (Raposo et al., 2009;
Kirchhoff, 2009). Indeed, we found that differences in encoding
strategies had a significant relationship to the similarity in brain
activity. This was particularly pronounced in the low-imageability
condition relative to the high-imageability condition, which may be due
to the increased demands on strategy for words that do not easily evoke
a visual image. We also found that strategy differences had a larger effect
in frontal and parietal regions than in temporal and occipital regions,
which would be consistent with literature on memory strategies and
those brain regions (Gershberg and Shimamura, 1995; Davachi et al.,
2001; Savage et al., 2001; Crescentini et al., 2010). For example, previous
studies with aging populations have suggested that as memory
performance declines prefrontal and parietal cortex activations become
more variable, suggesting that older individuals compensate for reduced
memory by engaging in more strategic processing (Cabeza et al., 2002;
Spreng et al., 2010).
Individual differences in cognitive style go back as far as the work of
Paivio (1971, 1986), including a distinction between visualizers and
verbalizers. Since that time, a great deal of effort has been put into
assessing and categorizing those differences (Mayer and Massa, 2003;
Sternberg and Zhang, 2001; Zhang and Sternberg, 2009). Yet, the
practical effect of those distinctions on memory (Richardson, 1978,
1998) and on other research-based interventions has been largely
absent (Massa and Mayer, 2006; Pashler et al., 2009). However, our
results suggest that those distinctions in cognitive style are not
without consequences. Individual differences in cognitive style can
have a significant effect on the similarity of brain activity. In particular,
we found that differences in visual style had a significant effect on
similarity throughout the frontal, parietal and temporal lobes. As
predicted, we found that this effect was evident for highly imageable
Please cite this article as: Miller, M.B., et al., Individual differences in cognitive style and strategy predict similarities in the patterns of brain
activity between individuals, NeuroImage (2011), doi:10.1016/j.neuroimage.2011.05.060
10
M.B. Miller et al. / NeuroImage xxx (2011) xxx–xxx
words but not for lowly imageable words. This suggests that
knowledge about an individual's cognitive style may be necessary
when evaluating the individual's pattern of brain activity, depending
on the type of stimuli.
There are many methods to probe individual differences in brain
activity, like correlating brain activity for a particular region with
individual differences in behavior or separating individuals into
distinct groups and comparing the activity between the two groups
(for a brief review, see Miller and Van Horn, 2007). While these
methods are useful and easy to visualize, they are examining correlations only along a single dimension. Our method explores multiple
dimensions by utilizing a multivariate regression analysis to predict
the correlation of functional t-statistic maps (i.e., the similarity in brain
activity). We found that multiple factors predict similarity in brain
activity during a memory retrieval task, including differences in age,
normalized white matter integrity, and cognitive style and strategy.
The multidimensional quality of individual differences in brain
activity could be a critical factor in achieving the goal of neuroergonomics to use neuroimaging to understand unique aspects the
individual mind in the work place. In the Introduction, we presented
an example in which the veracity of an individual's memory was
questioned by the fact that the individual's pattern of brain activity
was different than a typical pattern of activity for a true memory.
However, the difference may have been due to other factors, like the
individual's cognitive style or the strategy the individual used to
retrieve the episode. In some extreme cases, the failure to consider the
multiple factors that can drive differences in brain activity between
individuals can have life or death consequences. For example, a recent
court case grappled with the use of functional brain images as evidence
of a defendant's past mental state (Brown and Murphy, 2010). In that
case, a neuroscientist testified on the brain images of convicted
child rapist and murderer Brian Dugan during the sentencing phase,
arguing that the scans of Dugan were different enough from nonpsychopath individuals and that this dissimilarity indicated that
Dugan lacked the capacity to make normal moral decisions. This
argument was intended to persuade jurors that Dugan's mental defect
precluded the use of the death penalty. However, the difference in
activity that was evident in Dugan's brain image may not be the result
of diminished moral reasoning, but could be attributed to other
potential differences, such as the difference in white matter integrity
or cognitive style or personality. Our study indicates that any inference
that can be made about an individual based on their pattern of brain
activity must take into consideration the multiple factors that can
contribute to the variations in that activity.
Variability is often considered a nuisance, leaving the study of
individual variability in brain activity to the dustbin of statistical
averaging. But the systematic nature of variations in brain activity
between individuals offers a unique opportunity to fulfill one of the
goals of neuroergonomics; to use neuroscience to understand unique
aspects of the individual mind.
Acknowledgments
This work was funded by an Institute for Collaborative Biotechnologies grant to MBM through contract number W911NF-09-D0001 from the U.S. Army Research Office.
Appendix A. Supplementary data
Supplementary data to this article can be found online at doi:10.
1016/j.neuroimage.2011.05.060.
References
Battig, W., 1975. Within-individual differences in ‘cognitive’ processes. In: Solso, R.L.
(Ed.), Information processing and cognition. Erlbaum, Hillsdale, NJ, pp. 195–228.
Bennett, C.M., Miller, M.B., 2010. How reliable are the results from functional magnetic
resonance imaging? Annals of the New York Academy of Sciences 1191, 133–155.
Bennett, C.M., Baird, A.A., Miller, M.B., Wolford, G.L., 2010. Neural correlates of
interspecies perspective taking in the post-mortem Atlantic Salmon: an argument
for multiple comparisons correction. Journal of Serendipitous and Unexpected
Results 1 (1), 1–5.
Blajenkova, O., Kozhevnikov, M., Motes, M., 2006. Object–Spatial imagery: a new selfreport imagery questionnaire. Applied Cognitive Psychology 20, 239–265.
Boorman, E.D., O'Shea, J., Sebastian, C., Rushworth, M.F., Johansen-Berg, H., 2007.
Individual differences in white-matter microstructure reflect variation in functional connectivity during choice. Current Biology 17 (16), 1426–1431.
Brown, T., Murphy, E., 2010. Through a scanner darkly: functional neuroimaging as
evidence of criminal defendant's past mental states. Stanford Law Review 62, 1119.
Byrnes, J.P., 2001. Minds, Brains, And Learning. Guilford, New York.
Cabeza, R., Anderson, N.D., Locantore, J.K., McIntosh, A.R., 2002. Aging gracefully:
compensatory brain activity in high-performing older adults. NeuroImage 17,
1394–1402.
Casasanto, D.J., Killgore, W.D.S., Maldjian, J.A., Glosser, G., Alsop, D.C., Cooke, A.M.,
Grossman, M., Detre, J.A., 2002. Neural correlates of successful and unsuccessful
verbal memory encoding. Brain and Language 80, 287–295.
Coltheart, M., 1981. The MRC psycholinguistic database. Quarterly Journal of
Experimental Psychology Section A-Human Experimental Psychology 33, 497–505.
Crescentini, C., Shallice, T., Del Missier, F., Macaluso, E., 2010. Neural correlates of
episodic retrieval: An fMRI study of the part-list cueing effect. NeuroImage 50,
678–692.
Cronbach, L., Snow, R., 1977. Aptitudes and Instructional Methods: A Handbook for
Research on Interactions. Irvington Publishers, New York.
Davachi, L., Maril, A., Wagner, A., 2001. When keeping in mind supports later bringing
to mind: neural markers of phonological rehearsal predict subsequent remembering. Journal of Cognitive Neuroscience 13, 1059–1070.
Ekstrom, R.B., French, J.W., Harman, H.H., 1976. Manual for kit of factor-referenced
cognitive tests. Educational Testing Service, Princeton.
Faigman, D.L., 2010. Evidentiary incommensurability: a preliminary exploration of the
problem of reasoning from general scientific data to individualized legal decisionmaking. Brooklyn Law Review 75, 1137.
Feredoes, E., Postle, B.R., 2007. Localization of load sensitivity of working memory
storage: quantitatively and qualitatively discrepant results yielded by singlesubject and group-averaged approaches to fMRI group analysis. NeuroImage 35,
881–903.
Fletcher, P.C., Shallice, T., Frith, C.D., Frackowiak, R.S.J., Dolan, R.J., 1998. The functional
roles of prefrontal cortex in episodic memory: II. Retrieval. Brain 121, 1249–1256.
Friston, K.J., Holmes, A.P., Worsley, K.J., 1999. How many participants constitute a
study? NeuroImage 10, 1–5.
Gershberg, F., Shimamura, A., 1995. Impaired use of organizational strategies in free
recall following frontal lobe damage. Neuropsychologia 13, 1305–1333.
Graf, P., Birt, A.R., 1996. Explicit and implicit memory retrieval: Intentions and
strategies. In: Reder, L.M. (Ed.), Implicit Memory and Metacognition. Erlbaum,
Mahwah, NJ, pp. 25–44.
Guerin, S.A., Miller, M.B., 2009. Lateralization of the parietal old/new effect: an eventrelated fMRI study comparing recognition memory for words and faces. NeuroImage 44 (1), 232–242.
Heun, R., Jessen, F., Klose, U., Erb, M., Granath, D.O., Freymann, N., Grodd, W., 2000.
Interindividual variation of cerebral activation during encoding and retrieval of
words. European Psychiatry 15, 470–479.
Hogan, T.P., Cannon, B., 2007. Psychological Testing: A Practical Introduction. John
Wiley & Sons, Hoboken (NJ).
Kirchhoff, B.A., 2009. Individual differences in episodic memory: the role of selfinitiated encoding strategies. The Neuroscientist 15 (2), 166–180.
Kirchhoff, B.A., Buckner, R.L., 2006. Functional-anatomic correlates of individual
differences in memory. Neuron 51, 263–274.
Kondo, Y., Suzuki, M., Mugikura, S., Abe, N., Takahashi, S., Iijima, T., Fujii, T., 2005.
Changes in brain activation associated with use of a memory strategy: a functional
MRI study. NeuroImage 24, 1154–1163.
Kozhevnikov, M., Hegarty, M., Mayer, R.E., 2002. Revising the visualizer–verbalizer
dimension: evidence for two types of visualizers. Cognition and Instruction 20,
47–77.
Kozhevnikov, M., Kosslyn, S.M., Shephard, J., 2005. Spatial versus object visualizers: a
new characterization of visual cognitive style. Memory & Cognition 33, 710–726.
Machielsen, W.C.M., Rombouts, S.A.R.B., Barkhof, F., Scheltens, P., Witter, M.P., 2000.
FMRI of visual encoding: reproducibility of activation. Human Brain Mapping 9,
156–164.
Maldjian, J.A., Laurienti, P.J., Burdette, J.B., Kraft, R.A., 2003. An automated method for
neuroanatomic and cytoarchitectomic atlas-based interrogation of fMRI data sets.
NeuroImage 19, 1233–1239.
Massa, L.J., Mayer, R.E., 2006. Testing the ATI hypothesis: should multimedia instruction
accommodate verbalizer–visualizer cognitive style. Learning and Individual
Differences 16, 321–335.
Mayer, R.E., Massa, L.J., 2003. Three facets of visual and verbal learners: cognitive ability,
cognitive style, and learning preference. Journal of Educational Psychology 95,
833–846.
McGonigle, D.J., Howseman, A.M., Athwal, B.S., Friston, K.J., Frackowiak, R.S.J., Holmes,
A.P., 2000. Variability in fMRI: an examination of intersession differences.
NeuroImage 11, 708–734.
Miller, M.B., Van Horn, J.D., 2007. Individual variability in brain activations associated
with episodic retrieval: a role for large-scale databases. International Journal of
Psychophysiology 63 (2), 205–213.
Please cite this article as: Miller, M.B., et al., Individual differences in cognitive style and strategy predict similarities in the patterns of brain
activity between individuals, NeuroImage (2011), doi:10.1016/j.neuroimage.2011.05.060
M.B. Miller et al. / NeuroImage xxx (2011) xxx–xxx
Miller, M.B., Van Horn, J., Wolford, G.L., Handy, T.C., Valsangkar-Smyth, M., Inati, S.,
Grafton, S., Gazzaniga, M.S., 2002. Extensive individual differences in brain
activations during episodic retrieval are reliable over time. Journal of Cognitive
Neuroscience 14, 1200–1214.
Miller, M.B., Donovan, C.L., Sokol-Hessner, P., German, E., Van Horn, J.D., Wolford, G.L.,
2009. Unique and persistent individual patterns of brain activity across different
memory retrieval tasks. NeuroImage 48, 625–635.
Moscovitch, M., Winocur, G., 2002. The frontal cortex and working with memory. In:
Stuss, D.T., Knight's, R.T. (Eds.), The Frontal Lobes. Oxford University Press, Oxford.
Motes, M., Kozhevnikov, M., 2006. Cognitive style and individual differences in
imagery: perspectives from neuroscience [Technical report]. Office of Naval
Research, Arlington, VA.
National Research Council, 2008. Report of the NRC Committee on Opportunities in
Neuroscience for Future Army Applications. The National Academies Press,
Washington, DC.
Niogi, S.N., Mukherjee, P., Ghajar, J., McCandliss, B.D., 2010. Individual differences in
distinct components of attention are linked to anatomical variations in distinct
white matter tracts. Frontiers in Neuroanatomy 4 (1), 1–12.
Nyberg, L., McIntosh, A.R., Houle, S., Nilsson, L.G., Tulving, E., 1996. Activation of medial
temporal structures during episodic memory retrieval. Nature 380, 715–717.
Ollinger, J.M., Shulman, G.L., Corbetta, M., 2001. Separating processes within a trial in
event-related functional MRI. NeuroImage 13, 210–217.
Paivio, A., 1971. Imagery and Verbal Processes. Holt, Rinehart, & Winston, New York.
Paivio, A., 1983. The empirical case for dual coding. In: Yuille, J.C. (Ed.), Imagery,
Memory, and Cognition: Essays in Honor of Allan Paivio. Erlbaum, Hillsdale, NJ,
pp. 39–63.
Paivio, A., 1986. Mental Representations: A Dual Coding Approach. Oxford University
Press, New York.
Parasuraman, R., & Jiang, Y. (2011). Individual differences in cognition, affect, and
performance: behaivioral, neuroimaging, and molecular genetic approaches.
Neuroimage, this issue.
Parasuraman, R., Wilson, G.F., 2008. Putting the brain to work: neuroergonomics past,
present, and future. Human Factors 50, 468–474.
Pashler, H., McDaniel, M., Rohrer, D., Bjork, R., 2009. Learning styles: concepts and
evidence. Psychological Science in the Public Interest 9, 105–119.
Posner, M.I., Rothbart, M.K., 2005. Influencing brain networks: implications for
education. Trends in Cognitive Science 9 (3), 99–103.
Raposo, A., Han, S., Dobbins, I.G., 2009. Ventrolateral prefrontal cortex and self-initiated
semantic elaboration during memory retrieval. NeuroImage 47, 2261–2271.
Reder, L.M., 1987. Beyond associations: strategic components in memory retrieval. In:
Gorfein, D.S., Hoffman, R.R. (Eds.), Memory and learning: The Ebbinghaus Centennial
Conference. Lawrence Erlbaum Associates, Inc., Hillsdale, NJ., pp. 203–220.
Richardson, A., 1977. Verbalizer–visualizer: a cognitive style dimension. Journal of
Mental Imagery 1, 109–126.
Richardson, J.T.E., 1978. Reported mediators and individual differences in mental
imagery. Memory and Cognition 6, 376–378.
11
Richardson, J.T.E., 1998. The availability and effectiveness of reported mediators in
associative learning: a historical review and an experimental investigation.
Psychonomic Bulletin & Review 5, 597–614.
Savage, C.R., Deckersbach, T., Heckers, S., Wagner, A.D., Schacter, D.L., Alpert, N.M.,
Fischman, A.J., Rauch, S.L., 2001. Prefrontal regions supporting spontaneous and
directed application of verbal learning strategies: evidence from PET. Brain 124,
219–231.
Seghier, M.L., Price, C.J., 2009. Dissociating functional brain networks by decoding the
between-subject variability. NeuroImage 45, 349–359.
Seghier, M.L., Lazeyras, F., Pegna, A.J., Annoni, J.-M., Khateb, A., 2008. Group analysis and
the subject factor in functional magnetic resonance imaging: analysis of fifty
righthanded healthy subjects in a semantic language task. Human Brain Mapping
29, 461–477.
Shimamura, A.P., 1995. Memory and frontal lobe function. In: Gazzaniga, M.S., et al.
(Ed.), The Cognitive Neurosciences. MIT Press, Cambridge, MA, pp. 803–813.
Speer, N., Jacoby, L., Braver, T., 2003. Strategy-dependent changes in memory: effects
on behavior and brain activity. Cognitive, Affective & Behavioral Neuroscience 3,
155–167.
Spreng, R.N., Wojtowicz, M., Grady, C.L., 2010. Reliable differences in brain activity
between young and old adults: a quantitative meta-analysis across multiple
cognitive domains. Neuroscience and Biobehavioral ReviewsI 34, 1178–1194.
Sternberg, R.J., Zhang, L. (Eds.), 2001. Perspectives on Thinking, Learning, and Cognitive
Styles. Erlbaum, Mahwah, NJ.
Stoff, D.M., Eagle, M.N., 1971. The relationship among reported strategies, presentation
rate, and verbal ability and their effects on free recall learning. Journal of
Experimental Psychology 87, 423–428.
Tomassini, V., Jbabdi, S., Kincses, Z.T., Bosnell, R., Douaud, G., Pozzilli, C., Matthews, P.M.,
Johansen-Berg, H., 2011. Structural and functional bases for individual differences
in motor learning. Human Brain Mapping 32 (3), 494–508.
Tsukiura, T., Mochizuki-Kawai, H., Fujii, T., 2005. The effect of encoding strategies on
medial temporal lobe activations during the recognition of words: An event-related
fMRI study. NeuroImage 25, 452–461.
Tulving, E., Thomson, D.M., 1973. Encoding specificity and retrieval processes in
episodic memory. Psychological Review 80, 352–373.
Tulving, E., Habib, R., Nyberg, L., Lepage, M., McIntosh, A.R., 1999. Positron emission
tomography correlations in and beyond medial temporal lobes. Hippocampus 9,
71–82.
Weinstein, C.E., Underwood, V.L., Wicker, F.W., Cubberly, W.E., 1979. Cognitive
learning strategies: verbal and imaginal elaboration. In: O'Neil, H.F., Spielberger,
C.D. (Eds.), Cognitive and Affective Learning Strategies. Academic Press, New York,
pp. 45–75.
Wolbers, T., Schoell, E.D., Büchel, C., 2006. The predictive value of white matter
organization in posterior parietal cortex for spatial visualization ability. NeuroImage 32 (3), 1450–1455.
Zhang, L., & Sternberg, R. J. (Eds). (2009). Perspectives on the Nature of Intellectual
Styles. New York: Springer.
Please cite this article as: Miller, M.B., et al., Individual differences in cognitive style and strategy predict similarities in the patterns of brain
activity between individuals, NeuroImage (2011), doi:10.1016/j.neuroimage.2011.05.060