RESEARCH ESSAY
HOW TO CONDUCT A FUNCTIONAL MAGNETIC RESONANCE
(FMRI) STUDY IN SOCIAL SCIENCE RESEARCH1
Angelika Dimoka
Fox School of Business, Temple University,
Philadelphia, PA 19122 U.S.A. {angelika@temple.edu) }
This research essay outlines a set of guidelines for conducting functional Magnetic Resonance Imaging (fMRI)
studies in social science research in general and also, accordingly, in Information Systems research. Given
the increased interest in using neuroimaging tools across the social sciences, this study aims at specifying the
key steps needed to conduct an fMRI study while ensuring that enough detail is provided to evaluate the
methods and results. The outline of an fMRI study consists of four key steps: (1) formulating the research
question, (2) designing the fMRI protocol, (3) analyzing fMRI data, and (4) interpreting and reporting fMRI
results. These steps are described with an illustrative example of a published fMRI study on trust and distrust
in this journal (Dimoka 2010). The paper contributes to the methodological literature by (1) providing a set
of guidelines for designing and conducting fMRI studies, (2) specifying methodological details that should be
included in fMRI studies in academic venues, and (3) illustrating these practices with an exemplar fMRI study.
Future directions for conducting high-quality fMRI studies in the social sciences are discussed.
Keywords: fMRI, decision neuroscience, neuroIS, brain imaging
Introduction1
The ability of researchers to measure and link brain activity
to human processes, constructs, decisions, and behaviors has
demonstrated the potential of brain research for the social
sciences (e.g., Cacioppo et al. 2008; Camerer 2003; Dimoka
et al. 2011; Lee et al. 2006). By allowing for a close surrogate of brain activity, functional neuroimaging tools,
particularly fMRI (functional Magnetic Resonance Imaging),
have informed many unanswered questions in economics
(e.g., Camerer et al. 2004; Glimcher and Rustichini 2004;
King-Casas et al. 2005), marketing (e.g., Dietvost et al. 2009;
Huettel and Payne 2009; Yoon et al.2009; Yoon et al. 2006),
psychology (e.g., Crockett et al. 2008; Damasio 1994; Pessoa
1
Detmar Straub was the accepting senior editor for this paper. Eric Walden
served as the associate editor.
The appendices for this paper are located in the “Online Supplements”
section of the MIS Quarterly’s website (http://www.misq.org).
2008), and IS (e.g., Dimoka 2010, 2011; Dimoka et al. 2007),
in addition to the cognitive sciences in general (e.g., Decety
et al. 2004; Dulebohn et al. 2009). Accordingly, there has
been a proliferation of fMRI studies in social sciences
research during the last few years.
Despite the increased interest in neuroimaging tools and
specifically fMRI in the social sciences, functional neuroimaging for many social scientists is still overwhelmingly
complex (Huettel et al. 2008). Furthermore, becoming skilled
at neuroimaging tools, such as fMRI, will require substantial
time and effort (e.g., Culham 2006; Song et al. 2006). Thus,
the purpose of this paper is to integrate the neuroscience
literature outlining the process of designing, conducting, and
evaluating an fMRI study with practical guidelines on how to
undertake fMRI studies. These guidelines focus on four key
steps typically used in an fMRI study: (1) formulating the
research question, (2) designing the fMRI protocol, (3) analyzing fMRI data, and (4) interpreting and reporting fMRI
results. These four steps are illustrated with a running
MIS Quarterly Vol. 36 No. 3, pp. 811-840/September 2012
811
Dimoka/Conducting an fMRI Study in Social Science Research
example of a published fMRI study on trust and distrust in
this journal (Dimoka 2010). In addition to this running
example, other studies are employed to show how the proposed steps play out in research practice.
The paper proceeds as follows: The next section provides a
review of the basics of neuroscience and fMRI. The following section describes how to conduct an fMRI study
(formulate research questions, design fMRI protocol, analyze
fMRI data, and interpret and present fMRI results) using an
illustrative example of an fMRI study. The contributions of
the study and how to establish guidelines for conducting and
evaluating fMRI studies in the social sciences are then
discussed.
Basic Neuroscience Foundations
fMRI has become a method of choice in the decision neuroscience literature because it has superior spatial resolution and
it is a noninvasive approach that is able to precisely localize
a subject’s activated brain areas. The MRI scanner produces
cross sectional images of the brain, exploiting resonance
signals that are emitted by tissue water enveloped within a
strong magnetic field and excited by a high frequency electromagnetic pulse (Belliveau et al. 1991).2 Specifically, while
the magnetic field aligns the ions in the tissue water, the radio
frequency pulses alter the alignment in a manner that produces a rotating magnetic field, which is captured in an MRI
scanner as a three-dimensional anatomical brain image.
Using the magnetic properties of blood (Ogawa et al. 1990),
fMRI also captures localized neural activity in response to
decision and other mental processes by measuring changes
during scanning in blood flow or oxygenation (proxies for
brain activation). Neural activity increases blood oxygenation
(termed hemodynamic response) (e.g., Logothetis et al. 2001),
which is captured with an fMRI scanner that keys off of the
magnetic properties of blood (Logothetis and Wandell 2004;
Song et al. 2006). Increased brain activity results in a higher
concentration of deoxyhemoglobin, which emits a signal that
can be detected with the fMRI magnet. Signal intensity
depends on the blood’s magnetic properties, which is termed
blood oxygen level dependent (BOLD) signal (Ogawa et al.
1992). The signal increase (BOLD effect) is proportional to
brain activity (Bandettini and Ungerleider 2001), and through
this mechanism, the BOLD signal can detect increases in
neural activation in a specific brain area after a subject is
stimulated. Specifically, the BOLD signal is a reasonable
2
For more details on the basics of fMRI, see http://cognet.mit.edu/library/
erefs/cabeza/c002/section1.html.
812
MIS Quarterly Vol. 36 No. 3/September 2012
proxy for blood flow and volume in that specific area of the
brain responsible for undertaking the tasks related to a mental
process (e.g., Huettel et al. 2008). Accordingly, the BOLD
signal is used as an indirect measure or proxy of brain activity
(Logothetis et al. 2001).
As discussed in detail later in this paper (“Analyzing fMRI
Data”), brain activity can be statistically inferred via time
series data of the activated voxels (volumetric pixels) in threedimensional space using multivariate statistics, such as the
general linear model (GLM). Statistical parametric activation
maps (SPMs) are then created based on statistical analyses of
time-series images that compare the intensity of BOLD
signals relative to a baseline condition, such as the brain at
rest (e.g., Friston, Firth et al. 1995; Huettel et al. 2008).3
In 3Tesla fMRI scanners,4 functional images have a spatial
resolution of ~2–3 mm3 voxels and temporal resolution of 2–4
seconds. The fMRI scanner also acquires a single highresolution structural image of the brain. fMRI does not
capture absolute levels of blood oxygenation but the relative
intensity of BOLD signals across different conditions or
“contrasts” (Buckner and Logan 2006), as determined by the
fMRI protocol (see “Designing the fMRI Protocol”). Besides
being noninvasive and having fine-grained spatial resolution,
other advantages of fMRI include strong signal fidelity,
reproducibility, and consistency (e.g., Belliveau et al. 1991).
Because of the pinpoint localization of brain activity coupled
with fine-grained spatial resolution (2–3 mm3) and good
temporal resolution (a few seconds), fMRI is currently the
most widely used neuroimaging tool in the neuroscience
literature (e.g., Dimoka et al. 2011; Huesing et al. 2006).
Using fMRI, the neuroscience literature has linked constructs
and mental processes to specific brain areas, termed neural
correlates (Camerer et al. 2004; Lieberman 2007; see Appendix B). This is based on the assumption in the literature that
all mental processes have their origins in the brain (neural
correlates) (Glimcher and Rustichini 2004). While it is often
assumed that there is a simple one-to-one mapping between
human processes and brain areas (Lorig 2009), the true picture is more complex with essentially many-to-many brain
mapping among brain activations and mental processes and
3
SPMs are set of images with voxel values that are distributed according to
a known probability density function, usually the subject’s T or F distributions, under the null hypothesis. Those SPMs are referred to as the T- or
F-maps.
4
“3 Tesla” refers to a certain level of magnetic field to which current
commercial fMRI scanners adhere.
Dimoka/Conducting an fMRI Study in Social Science Research
constructs (Poldrack 2006).5 Hence, a complex construct
would typically map onto more than one brain areas and one
brain area can map to many constructs. For a detailed review
of the neuroscience literature using fMRI, see Dimoka et al.
(2011).
The emerging fMRI literature focuses on the localization and
functionality of brain areas that underlie mental processes.
For example, the mapping of decision processes to brain areas
is accomplished by engaging subjects with specific decision
tasks and observing their corresponding brain activations
when making decisions. A basic assumption of the neuroscience literature is that there is similarity across subjects in
the anatomy and functionality of the human brain (e.g.,
Aguirre et al. 1998), and localized activation in a certain brain
area is likely to generalize to the broader population of
healthy (right-handed) adults.6 For example, certain brain
areas are usually associated with certain mental processes, as
demonstrated in meta-analyses of how processes are mapped
on the brain (e.g., Krain et al. 2006; Phan et al. 2002; Phelps
2006). For a review of mental processes and their corresponding neural correlates, see Appendix B. As brain functionality is similar across people, the goal of fMRI studies is
to derive generalizable inferences about the population by
consciously modeling intra-subject and intersubject variability
(e.g., Holmes and Friston 1998; Mechelli et al. 2002). While
the analysis of fMRI data starts at the individual level, the
goal is to produce aggregate results with group-level analyses
(e.g., Thirion et al. 2007), as described in the section on
“Analyzing fMRI Data.”
While fMRI makes inferences about which brain area is
implicated when subjects are engaged by a mental process,
inferring that a certain mental process is solely engaged
(based on activation in a brain area) is unwarranted (termed
reverse inference; see Appendix A) (Poldrack 2006). As
discussed above, brain activity involves complex many-tomany relationship (Price and Friston 2005), and thus inferring
that activity in a brain area necessitates the existence of a
certain mental process is fraught with logical problems.
While reverse inference is not “deductively valid” (Poldrack
5
While all mental processes are assumed to have their origins in the brain, we
herein focus on research constructs as the manifestation of these processes
and their treatment in the literature as theoretical concepts. For example, the
mental processes of trusting and distrusting an entity would correspond to the
research constructs of trust and distrust.
6
While brain activations are similar across normal, right-handed individuals,
severe brain injury may cause plasticity (reorganization of the brain’s
functionality) (Price and Friston 2005). This may result in differing
functionality for injured patients since the healthy brain areas assume the
functionality of the injured ones (termed degeneracy).
2006, p. 59), it often helps generate exploratory hypotheses
(Cacioppo et al. 2008), which can later be tested empirically.
Besides, it may also be used in conjunction with other studies
to explain ex post some fMRI results. Still, reverse inference
must be used very cautiously to avoid over-interpretation of
fMRI data, as discussed in the exemplar study (Dimoka
2010).
How to Conduct an fMRI Study
To provide a roadmap for conducting an fMRI study, the
various steps are categorized into four key phases: (1) formulating the research questions, (2) designing the fMRI protocol,
(3) analyzing fMRI data, and (4) interpreting and reporting
fMRI results, as graphically illustrated in Figure 1.
Formulating Research Questions
for an fMRI Study
When formulating research questions for an fMRI study, there
are several important considerations: (1) obtaining unique
data that extend existing sources of data, (2) leveraging the
properties of brain data, (3) developing theories and hypotheses that correspond to the brain’s functionality by integrating
social science theories with neuroscience theories, while
(4) accounting for the idiosyncrasies of the fMRI
environment.
Obtaining Unique Data that Extend
Existing Sources of Data
First, the fMRI data should provide unique input in order to
complement existing data. Simply put, fMRI data are
valuable when they go above and beyond what existing
sources of data can already provide. These include constructs
that cannot be easily measured with existing methods, such
sensitive issues as gender, race, religion, hidden emotions
(like guilt, fear, anger), automated processes (e.g., automaticity), complex processes (cognitive overload), and moral
issues (for instance, ethics) (Dimoka et al. 2011). Masked
stimuli, which are not consciously observed but are still likely
to trigger brain activity, might be captured with fMRI (e.g.,
Vuilleumier et al. 2001) but may not be captured otherwise.
fMRI data are also valuable for constructs that may be subject
to measurement biases. For example, with self-reports, subjectivity bias degrades the perfect measurement of emotions
(e.g., LeDoux 2003). Moreover, social desirability may bias
MIS Quarterly Vol. 36 No. 3/September 2012
813
Dimoka/Conducting an fMRI Study in Social Science Research
Figure 1. The Four Basic Steps for Conducting an fMRI Study
truthful self-reported responses on sensitive issues (Fossati et
al. 2003), while brain responses are, arguably, not as susceptible to deliberate manipulation by subjects (Dimoka et al.
2011). Thus, fMRI data may be able to provide physiological
responses to different traits that subjects may struggle to selfreport due to knowledge bias (subjects may lack knowledge
on construct score) and rating bias (subjects cannot provide a
good estimate of the true construct score) (Burton-Jones
2009). For example, cognitive overload, utility, or emotions
are responses that are arguably difficult to measure with self
reports, but fMRI can provide an alternative form of measurement. Demand effects, characteristics, or artifacts may also
be overcome with brain data that is less affected by conscious
manipulation. By seeking to capture constructs that are difficult to measure otherwise, fMRI data might also reduce
omitted variable bias.
In sum, fMRI data should be used when it is challenging to
use existing sources of data to capture mental processes or
social science constructs. In such cases, complementing traditional and new sources of data with fMRI data allows triangulation across data sources and improves the overall quality of
measurement in the social sciences. Hence, fMRI data should
be unique in the sense that they supplement existing sources
of data by offering a different measurement perspective.
Leveraging the Properties of Brain Data
Second, because fMRI data, in response to experimental
stimuli, can be obtained essentially in near real-time
(accounting for the few seconds delay in the hemodynamic
814
MIS Quarterly Vol. 36 No. 3/September 2012
response), temporal and even causal (i.e., temporal precedence) relationships among constructs can also be readily
inferred with fMRI data; this is achieved by observing the
temporal order of brain activations associated with mental
processes and constructs. Notably, fMRI data can be captured
nearly simultaneously to when subjects observe a stimulus,
thus measuring brain activity immediately after observation,
decisions, and behaviors take place, rather than waiting to
capture reactions long after subjects have observed a stimulus.
By matching a task or stimulus to brain data with temporal
precision in real time, fMRI studies can also be used to infer
temporal or even causal relationships among processes,
constructs, decisions, and behaviors. Since temporal precedence greatly helps to infer causality (Cook and Campbell
1979; Zheng and Pavlou 2010), fMRI data can also help to
verify causal relationships among constructs. Further,
because experience can play a role in how the human brain
works, fMRI studies may be useful in examining whether
brain activity that corresponds to various constructs differs as
subjects vary by level of experience or gain experience over
time. Moreover, automated or routine activities performed by
experts may be linked to distinct brain activations. For
example, fMRI studies can compare experts versus novices or
track novices over time as they become experts. Tracking
computer users over time or comparing expert versus novice
system users could give insights on how system use evolves
over time with experience and habituation. Finally, fMRI
data could also help predict decisions and behaviors (e.g.,
Bernheim 2009; Lo and Repin 2002), especially in situations
where questionnaires, verbal reports, and interviews are not
good predictors (e.g., Mast and Zaltman 2005). By acquiring
behavioral data together with fMRI data, virtually in real-
Dimoka/Conducting an fMRI Study in Social Science Research
time, it may be possible to link brain activity with behavioral
responses and thereby to predict behavior in situations where
alternative means have failed.
Developing Social Science Theories and Hypotheses
to Correspond to Brain’s Functionality
Third, fMRI data can be used to challenge assumptions or test
competing theories. For example, Smith et al. (2002) challenged a long-held assumption in the economics literature that
payoffs and outcomes are independent. By showing that a
certain brain area (such as the insular cortex, which is usually
activated by fear of loss; Wicker et al. 2003) is only activated
in response to a certain stimulus (ambiguous gambles) but not
to a different stimulus (uncertain gambles), Hsu et al. (2005)
helped explain why people avoid gambles with ambiguous
probabilities and prefer gambles with uncertain probabilities.
Moreover, fMRI data can help identify the most likely
mediators, thus guiding theory selection to better correspond
to brain functionality. Notably, Dimoka, Pavlou, Benbasat,
and Qiu (2012) used fMRI data to show that similarityattraction is a more likely explanation for why people prefer
similar avatars for male users while dissimilarity-aversion is
a more likely explanation for female users. fMRI data can
therefore guide the selection of constructs for social science
theories by identifying whether or not there is brain activity
linked to certain constructs (thus suggesting that a construct
should be included in a model). For example, Deppe et al.
(2005) used fMRI to show that a consumer’s first brand
choice activates an area linked to social decision making
while the consumer’s second brand choice activates an area
linked to cognitive decision making. These results imply that
social considerations guide impulse purchases while cognitive
considerations guide planned purchases. Hence, fMRI data
should ideally be used to challenge existing social science
theories that are less neurologically plausible and do not
readily correspond to brain functionality, and the fMRI data
should be used to develop new social science theories that are
more in keeping with how the brain actually works.
Neuroscience theories, moreover, can be used to guide
biologically plausible hypotheses. Having specified how
fMRI data sheds light on phenomena in the social sciences, it
is useful to draw upon this literature to develop hypotheses
that link specific brain activations to social science constructs
or behaviors (Hedgcock and Rao 2009; Huettel and Payne
2009; Yoon et al. 2009). Ideally hypotheses should have
some tension in the sense that not all hypotheses are straightforward and support the same theory. fMRI studies can be
open to non-hypothesized results by allowing the brain data
to infer surprising findings that call for new theories.
This notwithstanding, in proposing testable hypotheses, it is
often necessary to assume that certain brain areas are associated with certain social science constructs (termed earlier as
reverse inference). Drawing upon extant neuroscience
literature, reverse inference may be tentatively used to generate hypotheses for exploratory testing (Cacioppo et al. 2008).
Nonetheless, scholars should be extremely wary of using
reverse inference for fMRI studies when there is minimal or
no relevant literature, either in the social science or the fMRI
literature. In such cases, two common mistakes are possible:
(1) either proposing hypotheses that relate to miniscule
behavioral differences that are difficult to infer neurologically
(particularly if no guidance is provided by the fMRI literature), or (2) proposing effects that spawn activations throughout the brain, making it difficult to pinpoint specific areas of
interest. Therefore, the author recommends heavy reliance on
relevant literature to specify expected areas of brain activation
and only developing tentative hypotheses about differences in
brain activations across different conditions.
fMRI data ideally should be used to test theories or hypotheses that are hard to address with existing tools. Thus, simply
replicating an existing behavioral study without expecting to
see some neurological findings that would help illuminate the
phenomenon may be unwarranted. Localization of neural
correlates is seldom useful unless it informs social sciences
theories. Specifically, localization of the neural correlates of
many constructs can shed light on the underlying nature or
dimensionality by drawing upon the properties of these brain
areas from the neuroscience literature. For example, while
perceived usefulness is generally viewed as a cognitive
construct, Dimoka et al. (2011) found its neural correlates to
reside in brain areas also linked to emotions.
Moreover, fMRI data can give guidance on whether constructs are distinct if their neural correlates are functionally
disassociated from each other in the brain, or whether they are
similar if they share very similar or the same neural correlates.
For example, Yoon et al. (2006) showed that person and
brand judgments have distinct neural correlates, implying that
they may be distinct constructs. Furthermore, it may be
informative to examine whether two constructs involve the
same or a different set of brain areas. For example, eye
movement and attention span the same brain areas (Corbetta
et al. 1998), implying that these two processes may be very
similar to each other, at least at a neurological level. Other
studies have shown that processes such as vision for perception and vision for action reside in clearly distinct brain areas
(e.g., Culham et al. 2003). Similarly, Kuhnen and Knutson
(2005) showed that the prospect of a monetary gain activated
a different brain area compared to the prospect of a monetary
loss of equal magnitude (neurologically supporting prospect
theory). McClure et al. (2004) found immediate (impulse pur-
MIS Quarterly Vol. 36 No. 3/September 2012
815
Dimoka/Conducting an fMRI Study in Social Science Research
chases) versus delayed (planned purchase) rewards to activate
distinct brain areas, implying a distinction. Conscious and
unconscious visual stimuli possess distinct neural correlates
(e.g., Rees et al. 2002). In sum, brain data could be used to
offer complementary data, which, when coupled with existing
knowledge from the neuroscience literature, can help inform
and ideally extend existing social science theories.
Accounting for the Idiosyncrasies
of the fMRI Environment
Last but not least, researchers need to be able to test research
questions and hypotheses within the constraints of the fMRI
scanner. Because the fMRI scanner is a full-body circular
tube (Figure 2), researchers have to design the fMRI protocol
so that subjects can perform the experimental tasks while
lying down flat. In the fMRI scanner, subjects can view
stimuli projected through goggles on a computer screen or
projected on a screen through a mirror. Subjects provide
feedback in several ways, such as a mouse or keyboard, using
hands, or verbal responses, using a microphone. Furthermore,
given that the head must stay physically still during data
collection, subjects are required to avoid excessive movement.
Finally, the fMRI scanner is noisy, and subjects not wearing
ear plugs can be distracted and, hence, yield poor responses.
In sum, fMRI studies are currently restricted to relatively
simple protocols to account for the inverse prone position, the
projected stimuli, and response option. They should be relatively short (~45–60 seconds) to account for subject fatigue
from the constrained movement and scanner noise.7
In sum, the fMRI environment is unique because of the
idiosyncrasies of the fMRI scanner, and not all studies can be
adapted for the fMRI context. Experience with fMRI studies
can lead to more complicated fMRI protocols, particularly
given the emergence of sophisticated fMRI-compatible (fiber
optic) input and output devices that enable subjects to better
view more complex visual stimuli (e.g., high resolution
goggles). Increasingly today, scholars are capturing subject
responses through more advanced output devices (e.g.,
joystick, trackball). Nonetheless, researchers will likely have
to compare the behavioral responses obtained within the fMRI
scanner to traditional lab experiments and ensure that the
idiosyncrasies of the fMRI environment do not result in any
significant behavioral or perceptual (self-reported) contradictions. Inventive researchers can work around these constraints to fit their particular needs, always keeping in mind
that (1) some experimental protocols may not be amenable to
fMRI, (2) each fMRI protocol is unique as a result of its study
design and research hypotheses, and (3) there is no single,
perfect recipe for all fMRI studies. Taken together, while
fMRI experiments share many similarities with traditional lab
experiments, researchers must carefully account for the idiosyncratic fMRI environment.
Illustrative Exemplar: Neural Correlates of Trust and
Distrust
To illustrate these guidelines, we use throughout the following section the Dimoka (2010) study as a running example
that examined the neural correlates of trust and distrust and
shed light on whether these are distinct constructs or the
opposite ends of the same continuum. Using fMRI to complement psychometric measurement scales of trust and distrust,
Dimoka captured the location, timing, and level of brain
activity associated with trust and distrust when subjects
interacted with four manipulated online seller profiles. These
seller profiles differed on their level of trust and distrust in the
context of online auctions.
The fMRI data showed that trust and distrust activate different
areas of the brain. Specifically, the neural correlates of trust
were shown to be the caudate nucleus (consistent with confident expectations about anticipated positive rewards from
trust) (King-Casas et al. 2005), the anterior paracingulate
cortex (predicting how the trustee will act in the future)
(McCabe et al. 2001) and the orbitofrontal cortex (uncertainty
from the trustor’s willingness to be vulnerable) (e.g., Krain et
al. 2006). Distrust was shown to be linked to brain areas
linked to intense negative emotions (amygdala) and fear of
loss (insular cortex). Dimoka also showed a clear distinction
in the brain areas associated with the corresponding dimensions of trust and distrust, with credibility and discredibility
mostly associated with the brain’s more cognitive areas
whereas benevolence and malevolence were mostly associated with the brain’s more emotional areas (amygdala and
insular cortex), differences that were more exacerbated in
women than in men.8
Dimoka’s study helps address the proposed guidelines for
conducting fMRI studies (Table 1). First, in terms of obtaining unique data, fMRI data supplemented self-reported
data on trust and distrust and used both sources of data to test
whether trust and distrust are distinct constructs. As noted by
Benbasat et al. (2010), trust and particularly distrust are difficult to measure with existing methods, such as self-reports
because they are subject to measurement biases. Dimoka com-
7
These are some current technological limitations that will, hopefully, be
overcome in the future. At that point, it may be possible to scan subjects
while they are in more natural working and thinking positions.
816
MIS Quarterly Vol. 36 No. 3/September 2012
8
Because of limitations of sample size, this inference was tentative.
Dimoka/Conducting an fMRI Study in Social Science Research
Figure 2. Full Body fMRI Scanner
pared self-reported data on trust and distrust with corresponding fMRI data to identify differences and similarities.
While the self reports did not reliably distinguish between
trust and distrust, the fMRI data showed trust and distrust to
have distinct neural correlates that span different brain
networks, showing their functional distinction at the neurological level.
Second, in terms of leveraging the properties of brain data,
Dimoka measured trust and distrust in the brain in near realtime while subjects observed trust/distrust stimuli, thus
showing that the neural correlates of distrust precede those of
trust. The neural correlates of trust and distrust were shown
to better predict price premiums relative to the corresponding
self-reported data on trust and distrust perceptions, with the
neural correlates of distrust (amygdala and insular cortex)
having stronger relative effects.
Third, in terms of guiding social science theories so that they
accurately reflect the brain’s functionality, Dimoka challenged social science theories that have suggested that trust
and distrust are merely the opposite ends of the same continuum. In terms of integrating social science with neuroscience theories, Dimoka explicitly hypothesized the neural
correlates of trust and distrust and their dimensions, thus
contributing to our understanding of the dimensionality of
trust and distrust. The study showed that brain activity associated with trust and distrust is a better predictor of economic
outcomes (price premiums) than the corresponding selfreported psychometric scales of trust and distrust, thereby
extending Pavlou and Dimoka’s (2006) work on the role of
trust on price premiums and the neuroscience literature on
comparing the predictive power of neurological versus selfreported data.9 Finally, it contributed by showing that trust is
more cognitive and calculative in nature while distrust is more
emotional in nature. Overall, Dimoka’s paper called for
future reexamination of the nature, dimensionality, distinction, and relative effects of social science constructs.
Fourth, in terms of accounting for the constraints of the fMRI
environment, Dimoka asked subjects to view and read traditional Likert-type measurement items projected onto a computer screen. Subjects used a custom-made fMRI-compatible
device (fiber-optic mouse) to respond their choices. The
device required minimal movement and was accommodated
in the fMRI scanner. The fMRI protocol was relatively
simple, as required by fMRI constrained settings, and it was
pretested with a behavioral study, which showed some very
similar responses to the self-reported measures of trust and
distrust within and outside the fMRI scanner.
Designing the fMRI Protocol
The fundamental issue to consider when designing an fMRI
protocol is that researchers should assess brain activation rela9
This logic depends on the assumption that there is a theoretical basis for why
trust and distrust predict price premiums. Given this assumption, brain data
better predicted price premiums than self-reported data.
MIS Quarterly Vol. 36 No. 3/September 2012
817
Dimoka/Conducting an fMRI Study in Social Science Research
Table 1. Key Guidelines When Formulating Research Questions for fMRI Studies
Guideline
Running Example (Dimoka 2010)
Obtaining unique data that extend beyond
existing sources of data
•
Overcoming potential measurement biases
(e.g., subjectivity bias, social desirability
bias)
•
Examining data differences and similarities
among different sources of data
Complementing perceptual (self-reported) measures of trust and
distrust with brain (fMRI) data
•
Measuring trust and distrust is subject to several measurement
biases
•
Comparing self-reported measures of trust and distrust with
corresponding fMRI data for differences and similarities
Leveraging the unique properties of brain
data
•
Making causal inferences across
constructs
•
Capturing brain data in near real-time
•
Predicting decisions and behavior
Conducting real-time measurement of trust and distrust and
predicting price premiums
•
Examining the timing of the neural correlates of trust and distrust
•
Comparing brain scan and psychometric measures of trust and
distrust with respect to their ability to predict a common dependent
variable (price premiums)
•
Predicting relative effects of trust and distrust on price premiums
Developing social science theories to
correspond to brain’s functionality
•
Using fMRI data to test competing theories,
guide theory development, and develop
new social science theories
•
Challenging neurally implausible theories
that do not correspond to brain’s
functionality and building new social
science theories that are consistent with
the brain’s functionality
•
Developing testable hypotheses about
which specific brain areas correspond to
constructs by integrating social science
theories with neuroscience theories
Developing trust and distrust theories to correspond to brain’s
functionality
•
Examining whether trust/distrust have the same neural correlates
and whether they span similar brain networks and challenging
theories that suggest that trust and distrust are the opposite ends of
the same (continuous) construct
•
Theories on the nature, dimensionality, distinction, and relationship
between trust and distrust should be guided by the premise that trust
and distrust have distinct neural networks
•
Integrating social science theories with neuroscience literature to
hypothesize the neural correlates of trust (i.e., caudate nucleus,
orbitofrontal cortex, anterior paracingulate cortex) and of distrust
(i.e., amygdala and insular cortex)
Accounting for fMRI environment
•
Designing protocols that can be undertaken while subjects are lying down flat
•
Offering responses using custom fiber
optic devices that can be used in fMRI
scanner
•
Avoiding the need for excessive movement
•
Relatively short experimental protocols
Designing the fMRI protocol to account for fMRI constraints
•
The protocol had subjects read trust and distrust stimuli on computer
screen through fiber-optic goggles
•
Responding to trust and distrust stimuli using a custom fiber-optic
device required the minimal movement of a single finger.
•
Completing the experimental protocol in 30–60 minutes to avoid
subject fatigue
tive to a baseline condition because absolute levels of brain
activation are non-interpretable (Song et al. 2006). The
baseline condition is captured while the subject is in a state of
relative rest. Therefore, the design of an fMRI protocol
should subtract the baseline from the experimental condition,
and researchers should focus on selecting a proper baseline
condition in order to partial out activations that are not part of
the experimental task (Friston 2002). This subtraction logic
is based on the idea that the difference between two tasks can
result in a difference in hemodynamic response to identify a
functionally active area (Culham 2006).
818
MIS Quarterly Vol. 36 No. 3/September 2012
Selecting Appropriate Subjects
In fMRI studies, it is important to explain the rules for
including or excluding subjects, attributes such as age, sex,
handedness, or medical history. Subjects should not be using
psychotropic medications, and they need to have normal or
corrected-normal visual acuity. Subjects should be recused
from fMRI studies for excessive head motion, structural
problems, or failure to follow the fMRI protocol. Incorporating behavioral performance criteria in the protocol
ensures that subjects in the fMRI data analysis properly
Dimoka/Conducting an fMRI Study in Social Science Research
performed the experimental tasks and their behavioral
responses are reasonable. If any subjects should be excluded
from the fMRI data analysis, the reasons for exclusion should
be clearly stated in the scientific reporting.
Calculating Power
Specifying sample size and statistical power in fMRI studies
is challenging. Power calculation is complex because the
fMRI images have tens of thousands of correlated voxels that
raise the number of time-series data (Friston et al. 1996;
Friston, Holmes, and Worsley 1999). Researchers should
balance the need for a large number of subjects to detect true
insignificant effects while at the same time minimizing study
costs in that fMRI studies are still generally high. The
number of subjects is chosen to ensure adequate power of
analysis for obtaining statistically significant activations at the
individual (first-level analysis) and, with more difficulty,
group (second-level analysis) levels. There are many guidelines to calculate the number of subjects (e.g., Desmond and
Glover 2002; Mumford and Nichols 2008; Murphy and
Garavan 2004). Within-subjects (repeated measures) designs
are recommended for fMRI studies to allow comparison
across the same brains. Within-subject designs also help
increase the power of analysis for comparison with other
sources of data. Besides, the statistical power (a priori
estimated or post hoc detected) analysis can be enhanced
through separation of the trial events and orthogonality of the
design parameters. Table 2 shows how to estimate power
(and adequate sample sizes), assuming certain means and
standard deviations a priori.10
analysis11 (assuming mean difference # .5 and STD # .75). In
fact, 15 subjects (6 of whom were women) participated in the
study, all being recruited from a major university in a
metropolitan area using an open ad. Subjects were paid U.S.
$35. All subjects were right-handed and were prescreened for
fMRI safety issues (no medical implants, metal piercings, or
medical problems). All ex post tests showed that no subjects
should have been excluded from the study (e.g., abnormalities
or excessive movement). A detailed source for calculating the
number of subjects required for an fMRI study is Desmond
and Glover (2002).
Selecting Trial Design
The fMRI protocol should also clearly specify the trial design,
which includes the number of trials per condition, trial duration, and intervals between trials (Figure 3). The time
between two trials is referred to as the inter-stimulus interval
(ISI). There are two major types of trial designs (Figure 4),
blocked (fixed sequential order of presentation of the
experimental conditions using extended time intervals) and
event-related (randomized order of the presentation of
experimental conditions in relatively short time intervals).
First, blocked designs present the stimuli sequentially within
each condition and alternating the conditions as a “boxcar” of
distinct blocks (Caplan 2009). The main advantage of
blocked designs is high power of detection due to the repetitive stimuli that creates an additive effect on the resulting
brain activations.
Despite concerns for Type II errors (false negatives) due to
small sample sizes, Type I errors (false positives) are more
common in fMRI studies in that they produce over 30,000
scans of the brain, thus having to conduct thousands of significance tests across numerous voxels. To control for false
positives there is a need to correct for family-wise error, and
adjustment methods, such as Bonferroni correction or false
discovery rate (FDR) should be used to account for the
multiple comparisons across voxels (Bhatt and Camerer
2003).
Second, event-related designs present the stimuli in a randomized order for a simultaneous presentation of many conditions
(Dale 1999). Simply put, event-related designs study brain
activation when a particular stimulus is present (in a random
order) versus not present. By allowing randomization across
conditions, event-related designs minimize anticipation, habituation effects, and order effects; however, this benefit comes
at a cost of lower detection power than blocked designs (and
difficulty in testing confounding from order effects). Thus, to
increase the detection power of event-related designs, temporal randomization of the intervals among successive presen-
In the Dimoka study, for a threshold of p < .05 with .80 power
for a Level 2 analysis, N = 12 was needed for the first level
11
10
For post hoc power calculations, researchers should use the realized
explained variances in research findings as a substitute for the estimated
effect sizes with a power calculator like http://danielsoper.com/statcalc3/
default.aspx.
The second level analysis in Dimoka would have required the use of nonparametric statistics since one cell had only six subjects (women) (Chi-square
tests require a minimum of five units of analysis per cell; Siegel and Castellan
1988). Nevertheless, because of the more stringent sample size requirements
of parametric tests that would have had to be met, Dimoka (p. 388) adopted
the scientifically conservative approach of interpreting only mean differences
so as not to venture into more definitive statistical inferences about differences between women and men.
MIS Quarterly Vol. 36 No. 3/September 2012
819
Dimoka/Conducting an fMRI Study in Social Science Research
Table 2. Power Calculations for fMRI Analysis
Analysis
Level
First
Test by Design
Type
Within-subjects
(i.e., repeated
measures)
Description
Comparison of each
subject’s voxels
between a resting and
stimulated condition
(across many trials of
each condition)
A-priori Power Estimates*
Assumptions**
.80 with 11-12 subjects
[STD (between subjects) = .5; STD
(within subjects) = .75; α = .05]
Mean difference = .50
.80 with 6 subjects
[STD (between subjects)=.5; STD
(within subjects) = .75; α = .05]
Mean difference = .75
.80 with 26-27 subjects
[STD (between subjects) = .5; STD
(within subjects) = .75; α = .05]
Mean difference = .25
Second
Between subjects
(i.e., between
groups)
Summation across trials
for subjects, then
comparison between
groupings of subjects
Use standard power calculations
based on total number of subjects
(post hoc) and effect sizes in the
literature (a priori)
Third
Between subjects
(i.e., between
studies)
Comparison of the
summative findings of
one study versus those
of another
Use standard power calculations
based on the number of subjects
across studies (post hoc) and
effect sizes in the literature (a priori)
*First level based on Desmond and Glover’s (2011) simulations.
**According to Desmond and Glover (2001, p. 121):, “With a muD and sigmaB of 0.5, 11-12 subjects are needed to achieve 80% at a 0.05,
assuming a value of 0.75 for sigmaW, typically observed in spatially smoothed data, and 100 time points per condition (n). Note that at muD /0.75,
approximately six subjects are needed to achieve 80% power; a decrease of 0.25 in muD (from 0.75 to 0.5) requires an additional 5 /6 subjects
to maintain 80% power, whereas an additional decrease in muD of 0.25 (from 0.5 to 0.25), requires over 20 more subjects to maintain 80% power.”
Figure 3. Graphical Representation of a Trial Design of an fMRI Protocol
Figure 4. Graphical Representation of a Block and Event Related Design of an fMRI Protocol
820
MIS Quarterly Vol. 36 No. 3/September 2012
Dimoka/Conducting an fMRI Study in Social Science Research
tations must be used (Ollinger et al. 2001). Event-related
designs are more complex than blocked designs, and there are
techniques to counter-balance across events by using fMRI
protocols with irregular intervals that vary (jitter).12 To create
and subsequently capture the cognitive strategies and/or
strategy types used by a subject to perform a given task,
researchers are advised to use a blocked design since such
design offers the repeated opportunity to develop (and thus
capture) the corresponding brain activity. When researchers
want to create and measure distinct conditions that may be
contrasted to observe differential brain activity, event-related
designs are preferred. Thus, both trial designs have advantages and disadvantages.
In the Dimoka study, an event-related design was used. First,
a randomly selected stimulus (seller profile) was presented to
the subjects for 3 seconds. This was followed by a randomly
selected measurement item for a randomly-selected construct
(trust or distrust) for each seller. Each item was shown for 5
seconds. Then, a seven-point Likert-type scale appeared.
After the subjects made their choice, a new randomly selected
seller profile was shown followed by a randomly selected
item. This procedure was repeated for four sellers and ten
measurement items for each seller in a random order. The
reason an event-related design was used was to compare trust
and distrust across the four sellers that were manipulated to
vary in their trust/distrust, and the measurement items on trust
and distrust were used to trigger trust/distrust perceptions
across sellers.
For a detailed example on how to select a trial design for
fMRI studies, see Decety et al. (2004) in the context of
cooperation/competition and Yoon et al. (2006) in the context
of person and brand judgments. An example of an eventrelated trial design is offered by Bhatt and Camerer (2005)
and Ferstl et al. (2005), while examples of a blocked design
are given by Barch et al. (1999) and Soltysik and Hyde
(2008). A detailed comparison between event-related and
blocked trial designs is given by Chee et al. (2003).
cognitive, emotional, or social processes while the corresponding brain activations are recorded within an fMRI
scanner. Relative to lab experiments, experimental tasks in
fMRI studies must be simpler to enable a straightforward link
between the experimental tasks and the observed brain
activations.
The experimental tasks should also be broken down into
experimental conditions, procedures intended to create variation in the independent variables of the study (Mandeville and
Rosen 2002). The experimental conditions are governed by
the trial design that specifies the experimental conditions
(Figure 3), with each trial consisting of one or more experimental condition that aims to either ask subjects to actively
perform a task or passively activate the brain using a visual,
auditory, or other type of stimulus (e.g., Huettel and Payne
2009; Yoon et al. 2009). The basic paradigm is to contrast
brain activity between an experimental task versus a baseline
(control) task (Huettel et al. 2006). The baseline task is used
to cancel out spurious brain activation due to visual stimuli,
movement, and other sources of noise, and thus isolate brain
activation associated only with the experimental stimuli (Song
et al. 2006). It is important to specify the nature and number
of the experimental stimuli and conditions, the number of
trials per condition, the duration of each trial, and the interstimulus interval among trials (Birn et al. 2002).
In the Dimoka study, the experimental conditions consisted of
the two dimensions of trust and distrust for four seller profiles
(24 = 16 conditions). These four seller profiles differed on the
dimensions of trust and distrust (high/low trust/distrust). Nine
trials, using one measurement item per trial, per condition
were used. Manipulation checks showed that the four sellers
differed on their degree of trust and distrust. The use of measurement items was similar to psychometric scales in lab
experiments.13 It is also consistent with the neuroscience
literature (Winston et al. 2002) that used numerical scales to
13
Specifying Experimental Tasks
The basic objective of an fMRI study is to pinpoint the exact
area of brain activation in response to a particular experimental task. Hence, an fMRI protocol should engage subjects
in a set of experimental tasks, such as viewing a visual
stimulus, making a decision, responding to a question, or
engaging in a behavior, that aims at manipulating certain
12
Jittering refers to using varying delays between the beginning of sampling
of brain images relative to the start of the stimulus presentation to the subject.
There are multiple forms of experimental stimuli that are used in the
neuroscience literature to elicit brain activation. The use of existing, wellvalidated measurement items, which capture a particular construct with
Likert-type scales, was proposed by Dimoka (2011) by integrating the theory
of measurement in the social sciences with the neuroscience literature to
localize the neural correlates of theoretical constructs with psychometric
scales with measurement items as stimuli to elicit activation in the brain areas
responsible for these constructs. The multi-item scales serve as the stimuli
that trigger brain activation while subjects read the measurement items and
process the information that pertains to a particular construct, such as trust
and distrust. This method is based on the logic that subjects engage the brain
area that corresponds to the construct measured by the psychometric
measurement items that were specifically developed to capture the construct,
thus eliciting activation specifically in the brain areas that relate to the focal
construct.
MIS Quarterly Vol. 36 No. 3/September 2012
821
Dimoka/Conducting an fMRI Study in Social Science Research
Figure 5. Graphical Representation of fMRI Experimental Procedure
rate the trustworthiness of faces and used the numerical
ratings as parametric covariates to contrast with the level of
brain activity. The trial was repeated for all four constructs,
four seller profiles, and nine measurement items (Figure 5).
In addition to the Dimoka study, other fMRI studies that offer
detailed information on the nature and sequencing of the
experimental tasks include Decety et al. (2004), Hsu et al.
(2005), and Yoon et al. (2006).
Designing Appropriate Contrasts
As explained earlier, fMRI data do not capture absolute levels
(blood always flows in the brain), but the relative intensity of
blood flow (BOLD signal) across conditions. Since fMRI is
based on a subtraction logic between an experimental and a
baseline condition, the difference in the BOLD signal is a
function of the contrasts across conditions. fMRI experiments
are similar in spirit to behavioral studies that must assure that
the observed measures (brain activations) are due to the
experimental conditions and not due to any confounding or
spurious factors that threaten internal validity (Song et al.
2006). Accordingly, the baseline condition must subtract all
brain activation other than the process of interest (Mandeville
and Rosen 2002). fMRI protocols depend on an appropriate
contrast between either the experimental and baseline condition or between two experimental conditions (e.g., high and
low values of a stimulus or independent variable). Also, since
any experimental task, such as moving a finger, viewing an
image, or hearing a sound may spawn brain activation (e.g.,
motor, visual, and auditory cortex), it is necessary to create a
contrast with the experimental condition that differs from
spurious activation. Spurious activations are highly problematic because they may raise the total level of activity in the
brain, thereby potentially statistically suppressing true brain
activations.
822
MIS Quarterly Vol. 36 No. 3/September 2012
In the Dimoka study, a contrast was created by subtracting the
study experimental primary condition (reading and responding
to the measurement items for trust and distrust) from the
baseline condition (a set of statements that resembled study
measurement items in terms of format type and number of
words, which the subjects read, processed, and were asked to
choose by pressing one of the seven buttons via a fiber-optic
mouse. The contrast for each condition was created between
the measurement items for each of the four constructs (dimensions of trust and distrust) and the corresponding control
statements with the same numerical score. To prevent
spurious brain activation in the measurement items of the
focal constructs, the complexity, format, and length of the
measurement items of these constructs were very similar
across the four sellers. A detailed example of how contrasts
should be designed can be found in Phan et al. (2004).
Repetition and Duration of fMRI Studies
fMRI studies require repetition (multiple trials) to help obtain
statistically significant brain activity since repetition helps
overcome changes in cerebral blood flow across experimental
tasks to reduce standard error. While fMRI scanning could
theoretically last as long as necessary, the fMRI protocol
should ideally constrain the number of tasks to reduce fatigue
and prevent subjects from becoming disengaged from the
experiment. The experimental tasks should not be too mundane to maintain similar brain activation throughout the study.
Typically fMRI protocols are about 30 to 60 minutes
depending on the nature of the experimental tasks.
In the Dimoka study, a set of nine similar, yet not identical,
measurement items based on the literature were used to both
measure the two dimensions of trust and distrust across four
seller profiles and serve as repetitive stimuli to spawn brain
activation associated with the two dimensions of trust and
Dimoka/Conducting an fMRI Study in Social Science Research
distrust. The total duration of the fMRI study was 6 × 9 × 4
× 10 = 2,160 seconds or about 36 minutes (6 conditions (5
constructs + 1 baseline) × 9 measurement items for each
experimental condition × 4 sellers × about 10–12 seconds for
each cycle). Along with 3 to 4 minutes for placing the subject
in the scanner and obtaining anatomical (nonfunctional) brain
images at the beginning of the fMRI session (which are used
to superimpose the functional images onto, as discussed in the
section “Acquiring fMRI Images”), the 2-second ISI, the total
time subjects spent in the fMRI scanner was about 45
minutes. No evidence of fatigue or disengagement was
reported by subjects.
of subjects, approximately three to five subjects. Besides fine
tuning the experimental tasks and potentially identifying
problems in the experimental stimuli or procedures, the
analysis of the pilot fMRI data can reveal potential problems
that would necessitate refining or even revising the fMRI
protocol. Furthermore, such pilot studies may include additional conditions to test whether it is possible to uncover any
subtler brain effects. Indeed, the pilot can also be used to
calculate the minimum amount of subjects and number of
repetitions needed to extract the hypothesized effects
(Calhoun 2006).
Pretesting fMRI Protocol with a Behavioral Study
In Dimoka, a pilot fMRI study with five subjects was conducted before the full study with 15 subjects. An example of
an fMRI study with an extensive pretest is offered by
Kriegeskorte et al. (2006).
Before executing the fMRI protocol in the fMRI scanner, the
author advises researchers to pretest a behavioral study with
a similar protocol to ensure that subjects correctly perceive
the study tasks, understand the manipulations, and perform
the tasks properly (relative to the baseline or control condition). This helps ensure that the experimental protocol is
clear to the subjects during the fMRI study. The behavioral
study should include manipulation checks for the experimental conditions. The experimental protocol should then be
modified for the fMRI environment with as little deviation as
possible to allow for comparison between any behavioral data
collected during the lab and the fMRI study, either with the
same set of subjects (to train subjects for the fMRI study) or
with a different set of subjects (for between-subjects comparison). Another purpose of running the study both inside
and outside the fMRI scanner is to compare the behavioral
data within and outside the fMRI scanner to ensure study
validity and address a potential concern that the fMRI scanner
may have altered the subject responses. Notably, Yoon et al.
(2009, p. 19) advises that “researchers should seek convergent
validity by linking fMRI data to other behavioral measures.”
Therefore, researchers are advised to undertake a behavioral
study outside the fMRI scanner with the same procedures as
the fMRI protocol to test the veracity of the experimental
tasks inside and outside the scanning environment.
In Dimoka, the behavioral data on the study’s constructs were
similar across the traditional lab and fMRI studies (p < .01),
implying that the fMRI context did not affect subjects’
behavioral responses.
Specifying Procedures Before fMRI Scan
The fMRI protocol should also include the procedures the
subjects perform before the fMRI session, including the study
instructions and any self-reported behavioral data obtained
outside the fMRI scanner. Given the time constraints of the
fMRI in terms of cost and lack of movement, researchers
should provide, according to the author’s experience, all relevant pertinent information about the experimental procedures,
and describe the familiarization exercises for subjects before
participating in the study, and they should elaborate on any
activities that require close inspection prior to the subject
entering the scanner.
In the Dimoka study, before the fMRI session, subjects were
instructed to buy an MP3 player (IPod Nano) from four madeup electronics auction sellers in eBay’s marketplace, which
served as the study context. The seller profiles differed on
trust and distrust by manipulating feedback text comments
and ratings. Subjects were given ample time to browse the
four seller profiles, depicted as the candidates from whom to
buy the IPod Nano, and they spent about 20 minutes browsing
the sellers preparatory to the fMRI session. Other examples
where additional behavioral data were collected and analyzed
before the fMRI session as a means of informing the subsequent fMRI study include Linden et al. (2003), McCabe et
al. (2001), and Sanfey et al. (2005).
Conducting a Small-Scale fMRI Study with
Small Number of Subjects
Specifying Procedures During fMRI Scan
Before undertaking the full fMRI study with all subjects, it is
strongly advised to conduct pilot studies with a small number
During an fMRI session, subjects lay on their back in an MRI
scanner. During the first 3 to 5 minutes, anatomical (structural) images of the brain are acquired while subjects lay still,
MIS Quarterly Vol. 36 No. 3/September 2012
823
Dimoka/Conducting an fMRI Study in Social Science Research
not perform any activities. The structural images serve both
to provide a high resolution image of a specific brain’s
anatomy on which the fMRI data can be overlaid, and also to
specify where the fMRI data should be collected. Functional
data are then collected while subjects respond to experimental
stimuli, such as visual, auditory, or tactile stimuli, all the
while the fMRI scanner records the BOLD signal throughout
the brain in approximately 2-second intervals. Behavioral
responses during the fMRI session can also be used for interpretation and comparison with the fMRI data (which is a
common empirical goal). The fMRI procedures need to be
described in adequate detail to allow other researchers to
replicate the experimental procedure, particularly during the
fMRI session where virtually all stimuli and their timing may
affect the brain activations given the idiosyncrasies of the
fMRI environment (see the section “Formulating Research
Questions for an fMRI Study”).
the fMRI session, and either with the same or different
subjects, depending on the study procedures.
In the Dimoka study, subjects entered the fMRI scanner lying
on their back, and visual stimuli were presented to them
through fiber-optic goggles connected to a computer. After
obtaining anatomical images, subjects undertook the experimental tasks following the trial design shown in Figure 5.
Brain activations for comparison across conditions were
obtained during the 5-second period while the subjects were
reading and processing each measurement item (before
posting their response). This 5-second period was selected to
both assure temporal separation between brain activity
between reading the measurement item and posting the
behavioral response and to avoid brain activity due to hand
movement when responding to the Likert-type scales.
Another detailed example of fMRI procedures is offered by
Decety et al. (2004).
In fMRI studies, it is necessary to describe the MRI scanner
in terms of fabricator (e.g., Siemens, GE) and field strength
(in Tesla). The higher the magnetic field strength, the better
the resolution of the image. To scan (or acquire fMRI images
from) the whole brain, about 30 slices are needed, which are
typically collected every 2 to 3 seconds (termed repetition
time or TR), depending on the study’s needs for temporal
resolution. Depending on the format and length of the fMRI
protocol, 15,000–45,000 functional brain images are usually
collected,14 each one being subdivided into small cubes called
voxels (volumetric or 3D pixels). Voxels are typically
3–5mm, with 25,000–50,000 voxels required to cover the
whole brain. Data from a single voxel over the course of an
fMRI study constitute a time series of BOLD signals that
amount to 500–1500 brain images. The fMRI scanner captures two types of images with distinct acquisition parameters:
structural images for the brain’s anatomical structure and
functional images that show changes in the BOLD signal
(Figure 6).15 Usually in a form of an appendix, it is necessary
to report the key parameters of data acquisition, including
whole or partial brain data acquisition, number of volumes
acquired per session, pulse sequence type (e.g., gradient, spin
echo), slice thickness, TE/TR/flip angle, sequential or inter-
Specifying Procedures After fMRI Scan
After the fMRI scanning session, subjects may be asked to
perform additional tasks to complement or check the manipulations of the fMRI study. For example, subjects may be
asked to engage in a task that is not constrained by the fMRI
scanner, but may be needed to complete the required set of
experimental procedures, such as making a selection, coming
to an economic decision, or performing an actual behavior.
Specifically, a dependent variable that entails an actual
physical behavior may be captured after the fMRI session.
Moreover, some experimental tasks may be performed ex post
because they require temporal separation from the experimental tasks during the fMRI session. Finally, as noted
earlier, it is strongly recommended that researchers replicate
the fMRI protocol in a traditional lab to be able to compare
the corresponding behavioral data within and outside the
fMRI scanner (e.g., Yoon et al. 2009), either before or after
824
MIS Quarterly Vol. 36 No. 3/September 2012
In the Dimoka study, after the fMRI session, subjects were
asked to give the price they would give each seller as part of
the measurement of behavioral responses later to be predicted
by fMRI data. This was used to create temporal separation
between the fMRI data and the endogenous variable (price) of
the study. Because the replication of the fMRI procedures in
a traditional lab setting was performed before the fMRI study
with a different set of subjects (which showed that all subjects
had similar behavioral responses within and outside the fMRI
scanner), subjects did not repeat any of the fMRI procedures
after the fMRI session. Finally, subjects were debriefed,
offered answers to their questions, thanked, and dismissed.
Acquiring fMRI Images
14
Most fMRI scanners collect dummy volumes while the scanner is warming
up, which are automatically removed by the scanner data acquisition
software. If not, they should be manually excluded from subsequent
statistical analysis.
15
These include the pulse sequence type (gradient/spin echo, EPI/spiral),
matrix size, field of view (FOV), acquisitions (NEX), flip angle, echo time
(TE), slice thickness, acquisition orientation (axial, sagittal, coronal, oblique),
and order of acquisition of slices (sequential or interleaved). See Huettel et
al. (2004) for more details.
Dimoka/Conducting an fMRI Study in Social Science Research
+
Structural Image
=
Functional Image
Superimposed Image
Figure 6. Superimposing the Functional Images on the Structural Brain Images
leaved order of acquisition, and the orientation of acquisition
(e.g., axial, sagittal).16
The Dimoka study reported that whole-brain fMRI data were
acquired in a time-series of about 20 minutes with a Siemens
3Tesla whole-body fMRI scanner with a standard head coil.
Voxel size was 3.33 × 3.33 × 5mm. Other studies that
reported the fMRI acquisition procedures and fMRI data
analysis parameters in detail are Decety et al. (2004), Delgado
et al. (2005), Deppe et al. (2005), and Phan et al. (2002).
Obtaining IRB Approval
fMRI studies require institutional review board (IRB) approval, but similar to most traditional lab experiments and
surveys, they most often qualify for an expedited review. A
trained lab technician is often present during fMRI studies,
and institutional IRB guidelines specify whether a radiologist
should review all fMRI scans for potential brain abnormalities. It is useful for researchers to state the IRB protocol
number and authorizing body, usually a university IRB board,
in all publications resulting from the IRB protocol.17
16
Subjects were scanned with a contiguous (no gap) 5mm axial highresolution T1-weighted anatomical images (matrix = 256 × 256; TR = 600;
TE = 15 ms; FOV = 21cm; NEX = 1; slice thickness = 5 mm). Functional
images were acquired with a gradient-echo echo planar free induction decay
(EPI-FID) sequence (T2*weighted: matrix = 128 × 128; FOV = 21cm; slice
thickness = 5mm; TR = 2s; TE = 30ms, number of slices = 28) in the same
plane as the structural images. Depending on the fMRI scanner and the type
of the anatomical images (T1 or T2), the duration of acquiring anatomical
images may vary, but it is typically a few minutes at the beginning of the
scanning session.
17
While most institutions clearly distinguish fMRI as a nonintrusive technique that qualifies for expedited review, a few may incorrectly classify fMRI
studies as an intrusive technique that uses radioactive isotopes, such as PET
scans and x-rays. It is useful for researchers to clarify to IRB committees that
fMRI does not emit any harmful rays, and it is a noninvasive technique that
simply records changes in the magnetic properties emitted by the body’s
natural brain activity.
In the Dimoka study, the fMRI protocol was reviewed and
approved by the university’s local IRB committee via an
expedited review. Subjects were given a written informed
consent to sign before the study. There may be some rare
occasions in fMRI studies when a researcher may observe
brain abnormalities, either at the structural (e.g., tumor) or the
functional (e.g., connectivity/integrity) level. Many IRBs
have a standard procedure that should be followed in these
rare instances, such as consulting a trained radiologist or
notifying the subject. However, fMRI is not a diagnostic tool,
and social scientists should refrain from making medical diagnoses or draw any medical inferences based on fMRI data.
The key steps for designing an fMRI protocol are summarized
in Table 3. For more details, see Amaro and Barker (2006),
Henson (2005), Logothetis (2008), and Poldrack (2006,
2008).
Analyzing fMRI Data
The analysis of fMRI data aims mainly at identifying the
location and level of functional activation of the brain areas
activated by the experimental stimuli specified in the fMRI
protocol. The functional images are then analyzed to identify
brain areas that are significantly more active during the
experiment relative to the baseline stimuli. The following
steps should be followed when analyzing fMRI data: (1) preprocessing of fMRI data and (2) statistically analyzing fMRI
data. Because fMRI data are massive by their very nature and
also having a complex spatial and temporal noise structure,
several assumptions and approximations need to be made for
model efficiency, as presented below.
Statistical Tools
Statistical tools for analyzing fMRI data include SPM
(www.fil.ion.ucl.ac.uk/spm), FSL (www.fmrib.ox.ac.uk/fsl),
MIS Quarterly Vol. 36 No. 3/September 2012
825
Dimoka/Conducting an fMRI Study in Social Science Research
Table 3. Key Guidelines when Designing fMRI Protocol
Step
Guidelines
Selecting appropriate subjects
Select subjects who satisfy fMRI compatibility criteria
Calculating statistical power
Specify number of subjects to provide adequate statistical power
Selecting trial design
Select blocked versus event related trial design, trial duration, and ISI
Specifying experimental tasks
Determine the tasks that subjects will perform during the experiment
Designing contrasts and controls
Design contrasts with appropriate controls to cancel spurious activity
Repetition and duration
Select repetitions based on fMRI procedures relative to study duration
Pretesting fMRI protocol
Conduct a behavioral study with the same fMRI protocol for pretesting
Conducting a small-scale fMRI study
Conduct a small-scale fMRI study with a small number of subjects
Procedures before fMRI session
Specify pre-fMRI instructions and what subjects are asked to do
Procedures during fMRI session
Obtain anatomical images and specify what subjects do in fMRI scanner
Procedures after fMRI session
Collect behavioral data, conduct manipulation checks, debrief subjects
Acquiring fMRI images
Specify focal brain areas and level of detail for acquiring fMRI images
Obtaining IRB approval
Propose protocol to conform to institution’s IRB guidelines and rules
AFNI (afni.nimh.nih.gov/afni) and Brain Voyager (www.
brainvoyager.com). SPM is freeware (Wellcome Department
of Cognitive Neurology, University College of London, UK),
run under Matlab® (The Mathworks, Inc., Natick, MA). A
primer on SPM is offered by Flandin and Friston (2008). FSL
is also freeware, written mainly by members of the Analysis
Group (FMRIB, Oxford, UK); it runs on both Apples and PCs
(Linux and Windows). AFNI is also freeware (NIMH,
Bethesda, MD, USA); it is based on the C language. Brain
Voyager is commercial (paid) software and a tutorial is available at www.brainvoyager.com/resources/documentation/
documentation.html. The software used for analysis should
be reported in the study, as often required by software
licenses.
In the Dimoka study, data analysis was performed with
SPM8, a software package that offers an affordable, simple,
one-stop solution for processing, analyzing, and visualizing
fMRI data; it also has good technical support.
Preprocessing of fMRI Data
Before fMRI data are ready for statistical analysis, they
should be preprocessed to remove noise, increase the signalto-noise ratio, and allow comparisons across subjects’
anatomically different brains. Preprocessing of fMRI data
includes six major steps—slice timing correction, realignment, spatial co-registration, segmentation, normalization, and
smoothing—as described in greater detail below.
Slice timing correction: Since images (slices) within a single
brain volume are obtained sequentially and not at exactly at
826
MIS Quarterly Vol. 36 No. 3/September 2012
the same time, a slice timing correction should be performed
to compensate for delays associated with acquisition time
differences among images during the sequential collections of
functional fMRI images (Smith 2001). For example, if TR is
set at 2 seconds, then there is a certain time lag between the
first and the final images. All of the fMRI data analysis software (e.g., SPM, FSL, AFNI, and Brain Voyager) offer tools
that correct for slice timing, and they should be used to
correct any time lags in image acquisitions. In Dimoka, slice
timing correction was performed using the standard procedure
embedded in SPM8.
Realignment: During an fMRI study, subjects spend 20 to 40
minutes inside the fMRI scanner, and some head movement
is likely to occur, resulting in spatial changes in terms of
where specific voxels correspond (Brammer 2001). Head motion can result in noticeable changes in signal intensity across
voxels over time (Friston 2002). Accurate movement corrections are very important in fMRI studies because even small
movements may result in systematic effects, effects that could
spawn false positive brain activations if accumulated over
many images (Hajnal et al. 1994). Without proper motion
realignment, head movement may result as erroneous brain
activity (Forman et al. 1995). During the realignment process, the time-series of functional images acquired from the
same subject should be realigned by specifying a set of realignment parameters to be used as statistical controls during
the statistical analysis of fMRI data (Figure 7). Realignment
is usually conducted by estimating the parameters of an affine
rigid-body transformation that minimizes the sum of squared
differences among each scan and a reference scan using the
transformation by resampling the data with a certain interpolation (e.g., trilinear, sinc, or cubic spline) (Friston 2002).
Dimoka/Conducting an fMRI Study in Social Science Research
Realignment
Movement
Movement
Figure 7. Realignment of the Brain Images to Remove Movement
In the Dimoka study, images were realigned with SPM8
within each run to correct for motion artifacts. Realignment
was conducted by estimating six parameters of an affine rigidbody transformation that minimized the sum of squared
difference between each scan and the mean of all scans in the
time series from each subject and applying the transformation
by resampling the data using a spline interpolation.
white matter, and cerebrospinal fluid. Segmentation should
be used to assign the probability that each voxel belongs to
each tissue type based on combining the likelihood for
belonging to that tissue type and the prior probability that is
derived from maps from a large number of subjects (Ashburner and Friston 2005). In Dimoka, segmentation was performed using the SPM8 procedure.
Spatial Co-Registration: During the spatial co-registration
process, the whole-brain images should be realigned to each
other because there may be systematic differences in the
images across whole-brain scans (Jenkinson 2001). Also
images that are in different modalities (functional and anatomical images) should be aligned to each other (Figure 8).
First, the images should be aligned to each other by aligning
all subsequent images of each brain volume to the first image
of the brain volume (Ashburner and Friston 2003). Second,
all images in each brain volume of images should be aligned
to the first image of the brain volume (Collignon et al. 1995).
Parameter estimation for spatial co-registration should be
accomplished with a transformation matrix that specifies the
higher resolution anatomical images and applies the transformation parameters to all fMRI images (e.g., Ashburner and
Friston 2003). The spatial normalization method and the
transformation type selected should be clearly reported. In
the Dimoka study, SPM8 was used to create a six-parameter
rigid body affine transformation (three translations in x, y, z
direction and three rotations in x, y, z axes) to correct for head
movement, and all co-registered functional images were
aligned to the first image on a voxel-by-voxel basis.
Normalization: Since the brains of various people differ in
size and shape, to compare brain activations across subjects,
their brains should be spatially normalized to a template
(average) brain to account for structural differences in eacj
subject’s brain.18 Normalization refers to scaling the data to
a standard template brain to allow intersubject comparison.
Normalization is a key step because it allows group analysis
and generalizations of fMRI results, and fMRI studies should
specify the template to which images are matched. There are
two common standard stereotaxic coordinate spaces, Talairach and Tournoux (1988)19 and MNI.20 While these differ
from one another, but they can be easily converted to the
other (Brett et al. 2002). There are online resources (www.
Segmentation: Healthy brain tissue can generally be classified into three broad types using MRI images: grey matter,
20
18
Since group-level analyses of fMRI data require normalization and spatial
smoothing that reduce spatial resolution (Price and Friston 2005), there are
guidelines for accounting for intersubject variability (e.g., Mechelli et al.
2002) and enhancing the statistical validity of group-level analysis (e.g.,
Roche et al. 2007; Thirion et al. 2007).
19
Talairach and Tournoux introduced the Talairach coordinates where any
point in the brain can be specified relative to a three-dimensional (X, Y, Z)
space, as shown in Table C1 in Appendix C.
The Montreal Neurological Institute (MNI) coordinates are based on the
average of many normal brains.
MIS Quarterly Vol. 36 No. 3/September 2012
827
Dimoka/Conducting an fMRI Study in Social Science Research
Co-Registration
Figure 8. Co-Registration of the Functional and Anatomical Brain Images
Figure 9. Normalization of the Brain Images to a Template Brain
talairach.org) that automatically convert fMRI data into
various coordinates and provide a functional explanation of
the identified brain areas.
The statistical tools used for brain imaging analysis use one
or the other stereotaxic coordinate space. Therefore, it should
be reported which has been used. Moreover, it is also possible to compute a template (average) brain model for a studyspecific set of subjects and accordingly normalize the images
on this custom-created template brain of each study subject
(Figure 9).
In the Dimoka study, using the normalization procedure
embedded in SPM8, all functional and anatomical images
were normalized into a standard stereotaxic space with MNI
coordinates (Ashburner and Friston 1999).
Smoothing: Since functional anatomy (the location of brain
functionality) may differ across subjects, the functional
828
MIS Quarterly Vol. 36 No. 3/September 2012
images should be smoothed to overcome spatial variance
(e.g., Poldrack et al. 2008). Smoothing also increases the
signal-to-noise ratio by removing high spatial frequencies
(Smith 2001). Smoothing is typically performed by replacing
the value of each voxel with a weighted value of its own value
and those of its neighboring voxels by averaging each voxel
with its neighbors, weighted by a Gaussian function that falls
with distance (Smith 2001). Smoothing is a standard procedure embedded in most statistical tools for brain imaging
analysis. In the Dimoka study, the fMRI images were
smoothed by convolving them with a Gaussian kernel of 8
cubic mm full width at half maximum (FWHM).
For data preprocessing, as reported Dimoka, the fMRI data
were realigned, co-registered, and normalized to a standard
stereotaxic space based on Talairach and Tournoux. All functional and anatomical volumes were then transformed into a
standard anatomical space using the T2 EPI template and the
SPM8 normalization procedure. The data were spatially
Dimoka/Conducting an fMRI Study in Social Science Research
Table 4. Key Guidelines for Preprocessing fMRI Data
Step
Slice-timing correction
Guidelines
Compensate for delays from time differences among fMRI images
Realignment
Specifying realignment parameters to be used as confounds in analysis
Spatial co-registration
Aligning all fMRI images to first image to account for subject movement
Segmentation
Enhancing probability that voxels belong to appropriate brain tissue
Normalization
Normalizing all subjects to a template or “average” brain for group analysis
Smoothing
Smoothing images to account for spatial and temporal variation
smoothed with a Gaussian kernel, and they were realigned to
correct for motion artifacts, increase signal-to-noise ratio, and
account for intersession variation. All brain volumes then
underwent spatial smoothing by convolving a Gaussian kernel
to increase the signal-to-noise ratio and account for residual
intersession differences. Last, the brain images created a
single image of mean activation per trial type and subject
(Worsley 2001), and they were deemed ready for data
analysis. For a detailed analysis on preprocessing fMRI data,
see Culham et al. (2006) and Knutson et al. (2007).
The key steps required for preprocessing fMRI data are summarized in Table 4.
Data Analysis
Data Analysis Philosophy: fMRI results are usually reported
in terms of activation maps, images that are obtained by
conducting statistical tests at each individual voxel level for
each condition and for each subject. This tests whether the
time-series activation of each voxel corresponds to the experimental condition relative to the control (or baseline) condition. Spatial maps are then created using statistical tools. The
most common fMRI data analysis approach is a parametric
univariate analysis implemented using the general linear
model (GLM) (Poldrack et al. 2008; Worsley and Friston
1995), and most fMRI studies employ GLM. However, other
fMRI data analysis approaches, such as independent components analysis (ICA) and its variants (probabilistic and temporal ICA), are available; these use entropy maximization to
iteratively specify independent spatial modes (McKeown et
al. 1998). Such approaches sometimes uncover more brain
activations than GLM approaches (Calhoun and Pekar 2000),
and researchers could replicate the analysis of their fMRI data
with several approaches to ensure the robustness of fMRI
results.
GLM fits a reference hemodynamic response function (hrf) to
the brain data. The hrf describes how the BOLD signal
evolves over time in response to changes in brain activity.
GLM models the BOLD response, and it determines whether
an increase in the BOLD response occurs in response to the
experimental versus the control condition by testing the
probability that the relative BOLD effect significantly differs
from zero (Kiebel and Holmes 2003). The significance of any
given voxel and its level of activation reveal the strength of
the experimental condition. After the analysis is conducted at
each individual voxel to minimize error, the value of each
voxel is a statistical quantity (often expressed in T-values or
Z-values), formally displayed in a statistical parametric map
(SPM). The T-statistic [SPM(t)] is transformed to a normal
distribution [SPM(z)]. SPM is a voxel-based approach that
makes area-specific statistical inferences to experimental
(versus control) conditions (Friston 2002). Accordingly,
SPMs should be reported relative to a selected threshold at
each individual voxel represented by a desired alpha protection level (Worsley 2001). The resulting SPM is a functional
brain image with all T-values of each individual voxel, a
process that must account for standard errors through
repetition (Lange 2001). Finally, the reported SPMs should
be over-laid on the structural brain images to graphically
display the specific areas of brain activity. This extensive
procedure was undertaken in Dimoka through the facility of
the SPM8 software.
Design Matrix: In fMRI implementations of GLM, a matrix
is needed to define the experimental design and the nature of
the hypothesis testing. This matrix, called the design matrix,
should be created by the researcher who selects and organizes
the images acquired in each run and groups them as a function
of the condition from which they were taken. Creating a
correct design matrix is one of the most important aspects of
fMRI data analysis and all statistical packages have specific
guidelines on how to create a design matrix. In most fMRI
data analyses such as Dimoka, the design matrix contains the
indicator variables (parametric covariates or regressors) that
encode the experimental tasks. These are formally identical
to classical analysis of (co)variance (i.e., ANCOVA) models.
MIS Quarterly Vol. 36 No. 3/September 2012
829
Dimoka/Conducting an fMRI Study in Social Science Research
Specifically, each column of the design matrix should
correspond to an effect that is built into the fMRI protocol to
detect confounds. These are referred to as explanatory
variables (or covariates or regressors). Their effects on the
response variable should be modeled in terms of functions of
the presence of these conditions (i.e., events smoothed with a
hemodynamic response function), and they should constitute
the first columns of the design matrix. Then, there should be
a series of terms designed to remove or model low frequency
variations in signal due to noise artifacts. The final column is
whole-brain activity. The relative contribution of each of
these columns should be assessed using standard least
squares, and inferences about these contributions should be
made using standard T or F statistics, depending on whether
the researcher is looking at a particular linear combination
(e.g., a subtraction), or all combinations simultaneously.
Finally, the rows of the design matrix should correspond to
each of the scans. This procedure was followed by Dimoka.
Another detailed implementation of the design matrix is
provided by Dale (1999).
Individual and Group Analysis: Data analysis should first be
conducted separately for each subject to obtain individual
brain activation maps (Genovese et al. 2002) (termed firstlevel analysis). The first-order analysis results in SPMs for
each subject. The individual activations are then aggregated
by combining the normalized brain activations of all subjects
in the sample (e.g., Ashburner and Friston 1999; Smith 2001)
(termed second-level analysis) to generalize the results to the
entire population (e.g., Frackowiak et al. 2004). The secondlevel analysis aggregates the effects of the first-level analysis
to infer whether the results are stable and common across the
population. The second-level analysis across subjects is
undertaken using the identified T-levels from the first-level
analysis. This is because the absolute comparative levels are
much different across subjects because of intersubject physiological variations, and thus not readily comparable. The firstlevel analysis models the sources of individual variation and
data noise as confounds for the second-level analysis (Holmes
and Friston 1998). Such intersubject physiological differences, albeit nonsystematic, are difficult to overcome due to
the relatively small sample size of fMRI studies. Therefore,
individual-level data are often analyzed with fixed-effects
models (Friston, Holmes et al. 1995), and group data with
random-effects models to account for intersubject variability
(Holmes and Friston 1998); mixed-effects models are also
common.
The first-level analysis uses a large number of autocorrelated
data, while the second-level analysis assumes a small number
of independent and identically distributed (IID) data. The
level of brain activation should be obtained by the value of
830
MIS Quarterly Vol. 36 No. 3/September 2012
the T-test at each individual voxel, which should reflect the
intensity of the BOLD signal. The aggregate of all brain
activity is then captured with z-tests, tests that correspond to
that unit normal distribution that renders the same p-values as
the T-statistic, and quantified by continuous measures (termed
z-scores) (Thirion et al. 2007). This analysis results in a
thresholded map with a continuous intensity level (measured
with z-scores) at each voxel (Genovese et al. 2002). This map
can be used subsequently in correlation or regression-type
models, either to compare the level of brain activations or to
compare the level of brain activations with non-fMRI data,
such as psychometric or behavioral data. Finally, the peak
activation (measured using a z-value) in a given brain area is
considered to be a good proxy for the overall strength of all
adjacent voxels in that brain area (e.g., Delgado et al. 2005;
Rilling et al. 2002).
In the Dimoka study, the analysis aimed at identifying the
neural correlates of trust and distrust, first at the individual
level for each subject separately and then at the group level by
aggregating the individual data from all subjects to obtain
group results. The fMRI images were first analyzed with a
first-level analysis, obtaining significant activations for each
subject. Then, second-level analyses were performed on the
aggregate data for the group (second-level) analyses. For the
second-level analysis, ROI analysis was used, as explained
below.
Region of Interest (ROI) Analysis: ROI analysis offers a
more precise analysis by focusing on certain brain areas, thus
not being affected by spurious activations in unrelated brain
areas (e.g., Poldrack 2007). The ROI should be defined in
terms of either anatomical or functional criteria (or both)
(Poldrack et al. 2008). Anatomical criteria should be used to
represent the specific brain area studied based on standard
anatomical coordinates (e.g., MNI or Talairach), while functional criteria should select voxels based on the observed
brain activation patterns. ROI analysis is used to limit corrections for multiple tests to a smaller set of voxels in a known
brain area as opposed the whole brain, thus enabling small
volume correction (Culham 2006). ROI can help test hypotheses about the existence of significant activations in expected
brain areas (Calhoun 2006).
In the Dimoka study, SPM8 was used to create the design
matrix using GLM and to identify statistically significant
voxels by setting a threshold based on spatial extent using t $
1.96 (Worsley 2001), a cluster probability of an uncorrected
alpha of 0.05 (Forman et al. 1995), and eight voxels as a
minimum size (Friston, Holmes et al. 1995). For the ROI
analysis, Dimoka used the WFU_PickAtlas freeware, which
specifies anatomical criteria using standard MNI coordinates
Dimoka/Conducting an fMRI Study in Social Science Research
Table 5. Key Guidelines for Analyzing fMRI Data
Step
Guidelines
Data analysis approach
Select data analysis method, usually general linear model (GLM)
Design matrix
Select and group fMRI images as function of the experimental condition
Individual and group analysis
Perform analysis for each subject, then proceed with aggregate analysis
Error correction
Account for Type I error by adjusting for multiple statistical tests
Region of interest analysis
Focus on particular brain area to identify significant brain activations
(http://fmri.wfubmc.edu/cms/software#PickAtlas). However,
no functional (i.e., based on brain activity) criteria were used,
given the exploratory nature of the study. Other notable
examples that have used ROI analysis are King-Casas et al.
(2005) and McClure et al. (2003).
posing a greater emphasis on meta-analysis and replication to
balance between these two types of errors.
Error Correction: Since there are thousands of voxels in the
whole brain, there is a risk of Type I error due to the many
statistical tests for each voxel. Correcting for multiple comparisons is important because a standard fMRI analysis
includes individual statistical tests at each voxel, which may
amount to approximately 10,000 tests for the whole-brain
volume. Therefore, the significance threshold should go
through a correction procedure (such as Bonferroni correction) to mitigate the number of false-positive voxels (Ramsey
et al. 2002). An alpha threshold of p < .05 is deemed to be
quite stringent, especially when coupled with a threshold of
clusters of activated voxels (cluster-size thresholding). There
are other ways in which a correction can be applied, such as
using experiment-wise error rate or false discovery rate
methods and non-parametric procedures (Poldrack et al.
2008). Besides tests at the voxel level, a cluster criterion of
adjacent voxels that exceed a certain threshold is useful, and
a cluster of about 10 voxels is generally deemed adequate to
reduce spurious false positive activations. Authors should
report how the error covariance is modeled (Poldrack et al.
2008); this could be temporal auto-correlation in the fMRI
time-series, for example. In the Dimoka study, a Bonferroni
correction at an alpha protection level of p < .05 was used
because it was deemed conservative. For other examples on
applying error correction, see Dulebohn et al. (2009).
For more details on analyzing fMRI data, see Frackowiak et
al. (2004) and Friston (2004); for a detailed example of fMRI
data analysis, see Knutson et al. (2007) and Kuhnen and
Knutson (2007).
Much of fMRI research has focused on reducing Type I errors
and the false discovery rate, thereby increasing Type II errors
(Lieberman and Cunningham 2009). This makes it difficult
to detect smaller but still important effects. The p-value
threshold of brain activity and cluster size is an unresolved
issue in the literature, typically specified at the p < 0.05 level
and 20 voxels. However, to achieve a balance between Type
I and Type II errors, Lieberman and Cunningham recommend
other parameters (i.e., p < .005 with 10 voxels), while pro-
The key steps required for the data analysis of fMRI data are
summarized in Table 5.
Interpreting and Reporting fMRI Results
Interpreting fMRI Results
In terms of interpreting fMRI results, given that the neuroscience literature has already identified the neural correlates
of many social processes (Appendix B), the first step would
be to interpret fMRI results in view of related studies in the
literature. Given the wealth of knowledge on the functionality
of brain areas associated with many processes, researchers are
advised to compare their results with related fMRI studies to
examine differences and similarities with the literature. By
comparing findings with functional and neurological findings
from the literature, researchers can enrich their interpretation
of the fMRI results. Therefore, the interpretation of fMRI
results should first be performed relative to the neuroscience
literature and the similarities and differences from the literature should be a prominent issue in the interpretation.
Second, it is important to avoid over-interpretation of brain
activations, particularly the phenomenon of reverse inference
discussed earlier. While activation in a certain brain area can
be linked to the particular process used to trigger the brain
activation in the focal study (Poldrack 2006), a brain activation is not necessary for a particular task or associated construct to be present (Van Horn and Poldrack 2009). Accordingly, causally backward inferences in the form of “brain
area X is activated, thereby prior process Y must be involved”
should be avoided, particularly when drawing inferences in a
MIS Quarterly Vol. 36 No. 3/September 2012
831
Dimoka/Conducting an fMRI Study in Social Science Research
confirmatory manner. For example, in Dimoka, activation in
the caudate nucleus does not necessarily suggest the existence
of trust; there is evidence that this reward area of the brain can
be activated by many other processes and constructs. Simply
put, activation in the caudate nucleus is associated with the
particular trust-building stimulus in the Dimoka study; still,
one cannot infer that every time we observe activation in the
caudate nucleus, this is necessarily trust. Therefore, as noted
several times earlier, it is important for social scientists to
appreciate that there is a many-to-many relationship among
brain areas and human processes (e.g., Miller 2008; Poldrack
2006; Price and Friston 2005). Reverse inference should be
used very selectively and with extreme care to interpret some
findings ex post, such as “brain area X is activated in our
study, perhaps implying that process Y may be involved given
the frequent involvement of brain area X into process Y in the
literature.” Such reverse inference should be backed up by
similar findings in the literature that link similar human
processes to the focal brain areas (Cacioppo et al. 2008).
Third, despite the high spatial resolution of fMRI, it is important to realize that voxel-wise analysis is still statistical in
nature and so too-definitive statements about which specific
brain areas are activated should be avoided. Besides, given
the multiple statistical steps required to ensure proper comparisons across subjects, such as realignment, normalization,
and co-registration (see “Analyzing fMRI Data”), it is not
possible to make strict inferences about the exact localization
of brain activity. The observed brain activations should be
interpreted relative to similar studies in the literature before
specifying the exact brain areas implicated in a process.
Finally, it is important to refrain from strong causal interpretation of fMRI results. Brain activations are statistical in
nature (as described in the section “Analyzing fMRI Data”),
and causality should be inferred judiciously when all criteria
are satisfied (covariation, temporal precedence, ruling out
rival explanations, and active manipulation of the independent
variable—the experimental stimulus) (Cook and Campbell
1979). Similar to behavioral lab experiments, the actively
manipulated experimental stimuli are the likely covariates of
the observed brain activations preceded temporally by the
stimuli. Nonetheless, given our relatively weak knowledge of
how the brain works and the nature of the BOLD effect upon
which the fMRI data are based, we cannot rule out all alternative possible explanations and threats to validity (Cook and
Campbell 1979). Thus, researchers should refrain from
making strong causal interpretations with fMRI data.
In summary, researchers should view fMRI results as complementary findings to tentatively support theoretical conjectures
rather than unequivocal and indisputable.
832
MIS Quarterly Vol. 36 No. 3/September 2012
Reporting fMRI Results
As described earlier (see Figure 6), the reported output of
fMRI studies consists of two types of images: (1) the anatomical images that show the brain’s anatomical structure and
(2) the functional images that show the level of BOLD signal
in the form of voxels in color-coded SPMs, superimposed on
the anatomical image for better visualization. Reporting of
functional images typically involves presentation of figures
with thresholded color-coded SPMs (as shown in Appendix C), and tables that show the locations where significant
activations are observed, as shown in Appendix C (for a
review, see Poldrack et al. 2008).
In general, fMRI results should be presented in figures and
tables, as described in more detail below.
Figures: fMRI results are usually reported in the form of
figures that graphically show the activated brain areas using
a thresholded color-coded statistical map that specifies the
intensity of brain activations and illustrates the number of
activated voxels. Additional details include the view of the
anatomical image (e.g., axial, coronal, saggital) and any
details needed to precisely localize the areas of brain activation, both being provided in a figure caption or in the text. It
is important to note that the thresholded color-coded images
are statistical maps that imply a direct statistical comparison
of the baseline and experimental conditions and that brain
activation is present (statistically) in one area and absent
(statistically) from another (Henson 2005). Researchers
should explicitly state that there is a statistically significant
effect across two contrasts (experimental versus control, for
example), which produces significant brain activations to
overcome concerns that fMRI figures are somehow real world
“pictures” of brain activity. Examples of representative information to include in figures with fMRI results are shown in
Appendix C (Figure C1), a similar figure was presented in
Dimoka. Other representative examples include Rilling et al.
(2004).
Tables: fMRI results should be reported in tables to precisely
specify the location of brain activations in stereotactic space
(showing the exact coordinates of all brain activations in
either MNI or Talairach and Tournoux (1989) space), number
of voxels in each activated area, activation statistics of each
brain area (e.g., peak activation), along with anatomical
labels. Examples of useful information to include in tables
are shown in Appendix C (Table C1), while a detailed
example is offered by Decety et al. (2004) and Rilling et al.
(2004).
In the Dimoka study, the functional brain activations (ROI
analysis) associated with trust and distrust were presented in
Dimoka/Conducting an fMRI Study in Social Science Research
both figures and tables. Specifically, Appendix C shows the
fMRI results for trust and distrust for the high trust/low
distrust (HL) and low trust/high distrust (LH) sellers using a
simplified version of Dimoka’s results. Appendix C depicts
the exact localization of the neural correlates of trust and
distrust.21 Besides Dimoka, other examples of presenting
fMRI results in figures are King-Casas et al. (2005) and
Sanfey et al. (2005). A more detailed illustration of fMRI
results is given by Bhatt and Camerer (2005).
Drawing Theoretical Implications: As in all studies in the
social sciences, fMRI results should be used to draw theoretical implications about existing theories in the social
sciences. As elaborated earlier in the section “Formulating
Research Questions for an fMRI Study,” stating the research
question to be answered through fMRI data ensures that some
meaningful implications for theory are derived, creating novel
insights that may guide social science theories to be more
consistent with the functioning of the brain. Indeed, many
more social scientists can certainly contribute to the emerging
neuroscience literature by adding knowledge about the neural
correlates of new, interesting constructs.
In the Dimoka study, whether trust and distrust are distinct
constructs or the ends of the same continuum has been a key
question for several disciplines in the social sciences. Therefore, offering brain evidence that trust and distrust are
neurologically distinct constructs with clearly distinct neural
correlates may be viewed as having useful theoretical implications for the social sciences. Thus, the results may guide
future research on trust and distrust to treat these two constructs as distinct variables with different antecedents and
effects. Dimoka’s findings contributed to the neuroscience
literature by specifying the neural correlates of these two
dimensions of trust (credibility and benevolence) and distrust
(discredibility and malevolence). In doing so, the study
showed that there is no single brain area dedicated to trust or
distrust, but rather a network of brain areas comprised of
dedicated areas related to the dimensions of trust and distrust
jointly form the constructs of trust and distrust. Similar to
trust and distrust, most complex theoretical constructs are
21
In Dimoka, the neural correlates of trust (specifically the caudate nucleus
and putamen) were believed to capture the “trustor’s confident expectations”
about anticipated rewards (e.g., King-Casas et al. 2005); the anterior
paracingulate cortex was shown to capture the “trustor’s prediction of how
the trustee will act” (McCabe et al. 2001); the orbitofrontal cortex was explained to capture the uncertainty due to the “trustor’s willingness to be
vulnerable” (e.g., Krain et al. 2006). In terms of the neural correlates of
distrust, the insular cortex were interpreted to capture the “trustor’s fear of
loss” (Todorov 2008), while the observed amygdala activation was thought
to capture the “trustor’s intense negative emotions of vigilance and betrayal”
(consistent with Winston et al. 2002).
likely to be associated with several brain areas. Accordingly,
researchers should not focus on simplistic one-to-one correlations between brain areas and constructs but rather the
network of brain areas that underlie these constructs and
explain their subprocesses.
In terms of interpreting and reporting fMRI results, it is
important to avoid over-interpretation by closely reflecting on
the neuroscience literature, avoiding liberal “reverse” or
“forward” inferences (Poldrack 2006). fMRI data also can be
used for statistical analysis by comparing with other types of
data, such as behavioral or secondary data. Moreover, fMRI
results should be used to draw managerial implications, as
noted below.
Drawing Managerial Implications: fMRI findings could
have broader implications for managerial practice. Hence,
social scientists, particularly in applied sciences, should consider reporting the managerial implications of fMRI results.
Implications could be in the form of deep knowledge about
the nature of human processes that could guide the design of
specific interventions for particular segments of the population to selectively manipulate these processes. Or they could
be in more specific terms, such as designing firm web sites in
such a manner that customer trust is maximized and distrust
is minimized.
Dimoka offered several managerial implications from the
distinction between trust and distrust. As trust and distrust
reside in distinct brain areas with distrust being more affective, abrupt, and impulsive, and trust being more cognitive,
calculative, and long-lived, managers could design distinct
strategies for building trust through deliberate and long-term
efforts that rely on cognitive calculation while preventing
distrust by carefully avoiding situations that could trigger the
abrupt and affective nature of distrust.
Potential Comparisons between fMRI Data and Other Data:
Given the continuous nature of fMRI data, it is possible to use
them in statistical analyses with other data, such as psychometric and behavioral data. Specific guidelines for integrating
fMRI with other sources of data are offered by Huettel and
Payne (2009) and Yoon et al. (2009); a note of caution is
offered by Vul et al. (2009). Potential statistical analyses
include correlation and regression. fMRI can be also combined with other neurophysiological measures (Dimoka,
Banker et al. 2012), such as EEG for enhancing the temporal
resolution of fMRI (e.g., Menon and Crottaz-Herbette 2005)
and TMS (transcranial magnetic stimulation ), a technique
that temporarily and noninvasively suppresses certain brain
areas (e.g., Knoch et al. 2006; Miller 2008). Such approaches
can strengthen causal inferences.
MIS Quarterly Vol. 36 No. 3/September 2012
833
Dimoka/Conducting an fMRI Study in Social Science Research
In the Dimoka study, statistical comparisons, including correlation and regression analyses, contrasted fMRI data on
trust and distrust with the corresponding psychometric data on
trust and distrust. Other examples of fMRI studies that integrate fMRI with either behavioral or psychometric data are
Delgado et al. (2005), Hsu et al. (2005), and Linden et al.
(2003). fMRI data can also be used to predict behavioral
responses (e.g., Paulus et al. 2003). Finally, examples of
meta-analyses of fMRI studies that aim to specify the neural
correlates of well-studied processes are Krain et al. (2006)
and Phan et al. (2002).
In sum, general advice for reporting fMRI results is to provide
adequate detail so the fMRI study can be initially reproduced
and interpreted by others. For more details on interpreting
and reporting fMRI results, see Poldrack et al. (2008) and
Ramsey et al. (2002).
Concluding Remarks
This study has implications for conducting high-quality fMRI
studies in social science research by providing a set of guidelines for fMRI studies and specifying methodological details
that could be included in academic papers. Coupled with
earlier work on “why to” conduct fMRI studies (Dimoka et al.
2011) and “what to” examine with fMRI (Dimoka, Banker et
al. 2012), this paper on “how to” conduct an fMRI study aims
to provide a detailed blueprint for why, what, and how to
conduct fMRI studies in the social sciences.
fMRI data are particularly valuable when existing sources of
data cannot offer reliable data, and the goal of fMRI data
should be to create novel insights beyond existing sources of
data. Simply conducting an fMRI study with the sole aim to
identify neural correlates may often not be enough. The
author advises researchers to rely on the neuroscience
literature to propose hypotheses that can be tested with fMRI
data, thus not flying blind and relying on existing knowledge,
whenever possible. While fMRI studies should have compelling motivations and be guided by the literature, exploratory work is a well-accepted practice in the neuroscience
literature by allowing brain data to speak for itself.
Given that formulating interesting research questions is a
challenging task, especially for fMRI studies, there are several
studies that have attempted to offer guidance on formulating
research questions for fMRI. Dimoka et al. (2011) offered a
set of seven opportunities that focus on (1) localizing the
neural correlates of constructs, (2) capturing hidden mental
processes, (3) complementing existing sources of data with
834
MIS Quarterly Vol. 36 No. 3/September 2012
brain data, (4) identifying antecedents of constructs,
(5) testing consequences of constructs, (6) inferring the
temporal ordering among constructs, and (7) challenging
assumptions and enhancing existing theories. Dimoka,
Banker et al. (2012) proposed several examples of research
questions that could benefit from the use of neurophysiological tools in the context of Information Systems research,
specifically in three major areas of IS research: (1) development and use of systems, (2) IS strategy and business outcomes, and (3) group work and decision support. Riedl et al.
(2010) also reflected on the discussions of a retreat on
NeuroIS to propose another set of suggestions for IS research.
In sum, there are several sources that help guide research
questions that could be potentially answered with fMRI
studies in the social sciences, and this paper focuses on how
to appropriately undertake such fMRI studies.
It is important to maintain that these suggested guidelines
should not be viewed as strict rules for how to conduct fMRI
studies but rather a flexible set of tenets that are likely to
evolve as the discipline matures. Moreover, there are technical books that provide extensive details about how to conduct an fMRI study (e.g., Buxton 2002; Faro and Mohamed
2005; Friston 2006; Huettel et al. 2005; Jezzard et al. 2001,
Poldrack 2006). Social scientists should remain current to the
rapid advancement of recent extensions of fMRI methods
(e.g., Bandettini 2009), such as diffusion spectrum imaging
that examines interconnectivity and temporal sequencing
among brain areas (e.g., Hagmann et al. 2008), pattern
classification methods that can identify patterns of brain
activations (e.g., Kriegeskorte et al. 2006), and clustering
tools that help identify functionally connected brain areas
(Stanberry et al. 2008). It is expected that major advances in
neuroimaging tools will take place in the next few years, and
social scientists who conduct fMRI studies should stay
attuned to these developments.
Reasonable Expectations for
Reviewers and Editors
Journal reviewers and editors should consider two critical
points before hastily dismissing fMRI studies as not meeting
traditional standards for traditional experiments. First, subject
sample sizes, for the interim, will tend to be small. While this
is currently a function of high cost, and the time involved to
collect and analyze data per subject, fMRI scholars can still
make legitimate empirical inferences and test theories. While
researchers typically have 10 to 20 subjects, there are tens of
thousands of observations per subject. fMRI studies will continue to report very small subject sample sizes, but with
Dimoka/Conducting an fMRI Study in Social Science Research
family-wise adjustments that can correct for errors in statistical inference (as pointed out earlier in this paper), we can
interpret the huge number of voxels in the data and draw
reasonable scientific conclusions about brain activity responding to highly specific stimuli. Second, with particular
regard to IS studies, the short history of fMRI studies means
that there will be much less prior literature using neurophysiological methods. While self-correcting over time,
reviewers and editors will have to give degrees of freedom to
authors in their referring to analogs to studies in other
domains and disciplines. As fMRI and other neurophysiological studies increasingly find their way into the IS literature,
this problem will recede in significance, but, as with any other
new areas, forward-looking journal editors and reviewers will
see the long term value of opening up new channels of
exploration and taking more risk in trusting results from
scientific methods that are new to the discipline.
Looking Toward the Future
With the improvement of fMRI scanners that permit faster
and higher resolution images with less noise and at a lower
cost, fMRI is likely to continue becoming a method of choice
for measuring brain activity in the social sciences. For
example, human–computer interaction research is already
looking into fMRI to evaluate system designs (e.g., Minnery
and Fine 2009). As with all new measurement tools, they are
most useful when they are complemented by existing methods
and data collection tools (Cacioppo et al. 2008), such as field
studies, field and lab experiments, research instruments,
interviews, archival data, and simulation models. In fact,
fMRI can benefit from other brain imaging tools, such as
EEG, near-infrared spectroscopy (NIRS), and transcranial
magnetic stimulation (TMS), to enhance its temporal resolution and help infer causality (Dimoka et al. 2011; Riedl et al.
2010). Similar to other methodological paradigms in the
social sciences, such as experiments (Jarvenpaa et al. 1985),
case studies (Benbasat et al. 1987), and theory of measurement (e.g., Burton-Jones 2009; Burton-Jones and Straub
2006; Straub et al. 2004), it is necessary for new paradigms,
such as fMRI, to establish their own practical guidelines,
quality controls, and best practices. Accordingly, other methodological essays can direct researchers on how to conduct
studies with other neurophysiological tools, such as EEG,
MEG, NIRS, TMS, eye tracking, and skin conductance
response, which can complement both fMRI data and existing
sources of data. Thus, we hope to entice social scientists to
conduct studies with neurophysiological tools that can complement existing sources of data and hopefully spawn some
interesting insights that cannot be inferred with existing
methods.
References
Aguirre, G. K., Zarahn, E., and D’Esposito, M. 1998. “The Variability of Human, BOLD Hemodynamic Responses,” NeuroImage
(8), pp. 360-369.
Amaro, E., and Barker, G. J. 2006. “Study Design in MRI: Basic
Principles,” Brain and Cognition (60), pp. 220-232.
Ashburner, J., and Friston, K. J. 1999. “Nonlinear Spatial Normalization Using Basis Functions,” Human Brain Mapping (7), pp.
254-266.
Ashburner, J., and Friston, K. J. 2003. “Rigid Body Registration,”
Chapter 2 in Human Brain Function (2nd ed.), S. J. Frackowiak,
K. J. Friston, C. Frith, R. Dolan, C. J. Price, S. Zeki, J.
Ashburner, and W. D. Penny (eds.), San Diego, CA: Academic
Press.
Ashburner, J., and Friston, K. J. 2005. “Unified Segmentation,”
NeuroImage (26), pp. 839-851.
Ba, S., and Pavlou, P. A. 2002. “Evidence of the Effect of Trust
Building Technology in Electronic Markets: Price Premium and
Buyer Behavior,” MIS Quarterly (26:3), pp. 243-268.
Bandettini, P. A. 2009. “What’s New in Neuroimaging Methods?,”
Annals of the NY Academy of Sciences: The Year in Cognitive
Neuroscience (1156:1), pp. 260-293.
Bandettini, P. A., and Ungerleider, L. G. 2001. “From Neuron to
BOLD: New Connections,” Nature Neuroscience (4), pp.
864-866.
Barch, D. M., Sabb, F. W., Carter, C. S., Braver, T. S., Noll, D. C.,
and Cohen, J. D. 1999. “Overt Verbal Responding During fMRI
Scanning: Empirical Investigations of Problems and Potential
Solutions,” NeuroImage (10), pp. 642-657
Belliveau, J., Kennedy, D., McKinstry, R., Buchbinder, B.,
Weisskoff, R., Cohen, M., Vevea, J., Brady, T., and Rosen, B.
1991. “Functional Mapping of the Human Visual Cortex by
Magnetic Resonance Imaging,” Science (254:17), pp. 716-719.
Benbasat, I., Gefen, D., and Pavlou, P. A. 2010. “Introduction to
the Special Issue on Novel Perspectives on Trust in Information
Systems,” MIS Quarterly (34:2), pp. 367-371.
Benbasat, I., Goldstein, D., and Mead, M. 1987. “The Case
Research Strategy in Studies of Information Systems,” MIS
Quarterly (11:3), pp. 368-386.
Bernheim, B. D. 2009. “The Psychology and Neurobiology of
Judgment and Decision Making: Wha’ts in it for Economists?,”
in Neuroeconomics: Decision Making and the Brain, P. W.
Glimcher, C. F. Camerer, E. Fehr, and R. A. Poldrack (eds.),
Amsterdam: Academic Press, pp. 115-126.
Bhatt, M., and Camerer, C. F. 2005. “Self-Referential Thinking
and Equilibrium as States of Mind in Games: fMRI Evidence,”
Games and Economic Behavior (52), pp. 424-459.
Birn, R. M., Cox, R. W., and Bandettini, P. A. 2002. “Detection
Versus Estimation in Event-Related fMRI: Choosing the
Optimal Stimulus Timing,” NeuroImage (15), pp. 252-264.
Brammer, M. J. 2001. “Head Motion and its Correction,” in Functional MRI: An Introduction to Methods, P. Jezzard, P. M.
Matthews, and S. M. Smith (eds.), New York: Oxford University
Press, pp. 243-250.
MIS Quarterly Vol. 36 No. 3/September 2012
835
Dimoka/Conducting an fMRI Study in Social Science Research
Brett, M., Johnsrude, J. S., and Owen, A. M. 2002. “The Problem
of Functional Localization in the Human Brain,” Nature Reviews
Neuroscience (3), pp. 243-249.
Buckner, R. L., and Logan, J. M. 2006. “Functional Neuroimaging
Methods: PET and fMRI,” in Handbook of Functional Neuroimaging of Cognition (2nd ed.), R. Cabeza and A. Kingstone
(eds.), Cambridge, MA: MIT Press, pp. 27-48.
Burton-Jones, A. 2009. “Minimizing Method Bias Through Programmatic Research,” MIS Quarterly (33:3), pp. 445-471.
Burton-Jones, A., and Straub, D. 2006. “Reconceptualizing System
Usage: An Approach and Empirical Test,” Information Systems
Research (17:3), pp. 228-246.
Buxton, R. B. 2005. An Introduction to Functional Magnetic
Resonance Imaging: Principles and Techniques, New York:
Cambridge University Press.
Cacioppo, J. T., Bernston, G. G., and Nusbaum, H. C. 2009.
“Neuroimaging as a New Tool in the Toolbox of Psychological
Science,” Current Directions in Psychological Science (17:2), pp.
62-67.
Calhoun, V. D. 2006. Unmixing fMRI with Independent
Component Analysis,” Engineering in Medicine and Biology
Magazine, IEEE (25:2), pp. 79-90.
Calhoun, V., and Pekar, J. 2000. “Where and Where Are Components Independent? On the Applicability of Spatial- and
Temporal- ICA to Functional MRI Data, NeuroImage (11), p.
S682.
Camerer, C. F. 2003. “Strategizing in the Brain,” Science (300),
pp. 1673-1675.
Camerer, C. F., Lowenstein, G., Prelec, D. 2004. “Neuroeconomics: Why Economics Needs Brains,” Scandinavian
Journal of Economics (106:3), pp. 555-579.
Caplan, D. 2009. “Experimental Design and Interpretation of
Functional Neuroimaging Studies of Cognitive Processes,”
Human Brain Mapping (30), pp. 59-77.
Chee, M. W. L., Venkatraman, V., Westphal, C., and Soon, C. S.
2003. “Comparison of Block and Event-Related fMRI Designs
in Evaluating the Word Frequency Effect,” Human Brain
Mapping (18:3), pp. 186-193.
Collignon, A., Maes, F., Delaere, D., Vandermeulen, D., Suetens, P.,
and Marchal, G. 1995. “Automated Multi-Modality Image
Registration Based on Information Theory,” in Information
Processing in Medical Imaging, Y. Bizais, C. Barillot, and R.
DiPaola (eds.), New York: Springer, pp. 263-274.
Cook, T. D., and Campbell, D. T. 1979. Quasi-Experimentation:
Design and Analysis Issues for Field Settings, Boston: Houghton
Mifflin.
Corbetta, M., Akbudak, E., Conturo, T. E., Snyder, A. Z., Ollinger,
J. M., Drury, H. A., Linenweber, M. R., Petersen, S. E., Raichle,
M. E., Van Essen, D. C., and Shulman, G. L. 1998. “A Common
Network of Functional Areas for Attention and Eye Movements,”
Neuron (21), pp. 761-773.
Crockett, M. J., Clark, L., Tabibnia, G., Lieberman, M. D., and
Robbins, T. W. 2008. “Serotonin Modulates Behavioral Reactions to Unfairness,” Science (320:5884), p. 1739.
836
MIS Quarterly Vol. 36 No. 3/September 2012
Culham, J. C. 2006. “Functional Neuroimaging: Experimental
Design and Analysis,” in Handbook of Functional Neuroimaging
of Cognition, R. Cabeza and A. Kingstone (eds.), Cambridge,
MA: MIT Press, pp. 53-82.
Culham, J. C., Danckert, S. L., DeSouza, J. F., Gati, J. S., Menon,
R. S., and Goodale, M. A. 2003. “Visually Guided Grasping
Produces fMRI Activation in Dorsal But Not Ventral Stream
Brain Areas,” Journal of Neurophysiology (81), pp. 180-189.
Dale, A. M. 1999. “Optimal Experimental Design for EventRelated fMRI,” Human Brain Mapping (8:1), pp. 109-114.
Damasio, A. R. 1994. Descartes’ Error: Emotion, Reason, and
the Human Brain,New York: Putnam.
Decety, J., Jackson, P. L., Sommerville, J. A., Chaminade, T., and
Meltzoff, A. N. 2004. “The Neural Bases of Cooperation and
Competition: An fMRI Investigation,” NeuroImage (23:2), pp.
744-751.
Delgado, M. R., Miller, M. M., Inati, S., and Phelps, E. A. 2005.
“An fMRI Study of Reward-related Probability Learning,”
NeuroImage (24), pp. 862-873.
Desmond, J. E., and Glover, G. H. 2002. “Estimating Sample Size
in Functional MRI (fMRI) Neuroimaging Studies: Statistical
Power Analyses,” Journal of Neuroscience Methods (118:1), pp.
115-128.
Deppe, M., Schwindt, W., Kugel, H., Plassmann, H., and Kenning,
P. 2005. “Nonlinear Responses Within the Medial Prefrontal
Cortex Reveal When Specific Implicit Information Influences
Economic Decision Making,” Journal of Neuroimaging (15:2),
pp. 171-182.
Dietvorst, R. C., Verbeke, W. J. M. I., Bagozzi, R. P., Yoon, C.,
Smits, M., and van der Lugt, A. 2009. “A Sales Force-Specific
Theory-of-Mind Scale: Tests of its Validity by Classical
Methods and Functional Magnetic Resonance Imaging,” Journal
of Marketing Research (46:5), pp. 653-668.
Dimoka, A. 2010. “What Does the Brain Tell Us Sbout Trust and
Distrust? Evidence from a Functional Neuroimaging Study,”
MIS Quarterly (34:2), pp. 373-396.
Dimoka, A. 2011. “Brain Mapping of Psychological Processes with
Psychometric Scales: An fMRI Method for Social Neuroscience,” NeuroImage (54), pp. S263-S271.
Dimoka, A., Banker, R. D., Benbasat, I., Davis, F. D., Dennis, A. R.,
Gefen, D., Gupta, A., Ischebeck, A., Kenning, P., Pavlou, P. A.,
Müller-Putz, G., Riedl, R., vom Brocke, J., and Weber, B. 2012.
“On the Use of Neurophysiological Tools in IS Research:
Developing a Research Agenda for NeuroIS,” MIS Quarterly
(36:3), pp. 679-702.
Dimoka, A., Pavlou, P. A., Benbasat I., and Qiu, L. 2012. “Why
Do Consumers Choose Similar Avatars in Online Shopping
Environments? A Neuroimaging Study of Similarity-Attraction
and Dissimilarity-Repulsion,” Working Paper, Temple
University.
Dimoka A., Pavlou, P. A., and Davis, F. 2007. “NeuroIS: The
Potential of Cognitive Neuroscience for Information Systems
Research,” in Proceedings of the 28th International Conference
on Information Systems, Montreal, Canada, December 11-14.
Dimoka/Conducting an fMRI Study in Social Science Research
Dimoka, A., Pavlou, P. A., and Davis, F. D. 2011. “NeuroIS: The
Potential of Cognitive Neuroscience for Information Systems
Research,” Information Systems Research (22:4), pp. 687-702.
Dulebohn, J. H., Conlon, D. E., Sarinopoulos, I., Davison, R. B., and
McNamara, G. 2009. “The Biological Bases of Unfairness:
Neuroimaging Evidence for the Distinctiveness of Procedural and
Distributive Justice,” Organizational Behavior and Human
Decision Processes (110), pp. 140-151.
Faro, S. H., and Mohamed, F. B. 2005. Functional MRI: Basic
Principles and Clinical Applications, New York: SpringerVerlag.
Ferstl, C., Rinck, M., and von Cramon, D. Y. 2005. “Emotional
and Temporal Aspects of Situation Model Processing during Text
Comprehension: An Event-Related fMRI Study,” Cognitive
Neuroscience (17:5), pp. 724-739.
Flandin, G., and Friston, K .J. 2008. “Statistical Parametric
Mapping,” Scholarpedia (3:4), pp. 6232.
Forman, S. D., Cohen, J. D., Fitzgerald, M., Eddy, W. F., Mintun,
M. A., and Noll, D. C. 1995. “Improved Assessment of Significant Activation in Functional Magnetic Resonance Imaging
(fMRI): Use of Cluster-Size Threshold,” Magnetic Resonance
Medicine (33), pp. 636-647.
Fossati, P., Hevenor, S. J., Graham, S. J., Grady, C., Keightley,
M. L., Craik, F., and Mayberg, H. 2003. “In Search of the
Emotional Self: An fMRI Study Using Positive and Negative
Emotional Words,” American Journal of Psychiatry (160:10), pp.
1938-1945.
Frackowiak, R. S., Friston, K. J., Frith, C. D., Dolan, R. J., Price,
C. J., Zeki, S., Ashburner, J., and Penny, W. 2004. Human
Brain Function, Oxford, UK: Academic Press.
Friston, K. J. 2002. “Statistics I: Experimental Design and Statistical Parametric Mapping,” in Brain Mapping: The Methods (2nd
ed.), A. W. Toga and J. C. Mazziota (eds.), San Diego, CA:
Elsevier, pp. 315-349.
Friston, K. J. 2004. “Experimental Design and Statistical Parametric Mapping,” in Human Brain Function, R. S. Frackowiak, K.
J. Friston, C. D. Frith, R. J. Dolan, C. J. Price, S. Zeki, J.
Ashburner, and W. Penny (eds.), Oxford, UK: Academic Press,
pp. 599-632.
Friston, K. J. 2006. Statistical Parametric Mapping: The Analysis
of Functional Brain Images, New York: Academic Press.
Friston, K. J., Frith, C., Frackowiak, R. S. J., and Turner, R. 1995.
“Characterizing Dynamic Brain Responses with fMRI,”
NeuroImage (2), pp. 166-172.
Friston, K. J., Holmes, A. P., Poline, J. B., Price, C. J., and Frith,
C. D. 1996. “Detecting Activations in PET and fMRI: Levels
of Inference and Power,” Neuroimage (4), pp. 223-35.
Friston, K. J., Holmes, A. P., and Worsley, K. J. 1999. “How Many
Subjects Constitute a Study?,” Neuroimage (10), pp. 1-5
Friston, K. J., Holmes, A. P., Worsley, K. J., Poline, J.-P., Frith, C.
D., and Frackowiak, R. S. J. 1995. “Statistical Parametric Maps
in Functional Imaging: A General Linear Approach,” Human
Brain Mapping (2:2), pp. 189- 210.
Genovese, C. R., Lazar, N. A., and Nichols, T. 2002. “Thresholding of Statistical Maps in Functional Neuroimaging Using the
False Discovery Rate,” NeuroImage (15), pp. 870-878.
Glimcher, P., and Rustichini, A. 2004. “Neuroeconomics: The
Consilience of Brain and Decision,” Science (306:5695), pp.
447-452.
Hagmann, P., Cammoun, L., Gigandet, X., Meuli, R., Honey, C. J.,
Wedeen, V. J., and Sporns, O. 2008. “Mapping the Structural
Core of the Human Cerebral Cortex,” PLOS Biology (6:7), pp.
1479-1493.
Hajnal, J. V., Myers, R., Oatridge, A., Schwieso, J. E., Young, I. R.,
and Bydder, G. M. 1994. “Artifacts Due to Stimulus Correlated
Motion in Functional Imaging of the Brain,” Magnetic
Resonance Medicine (31), pp. 283-91.
Hedgcock, W., and Rao, A. R. 2009. “Trade-Off Aversion as an
Explanation for the Attraction Effect: A Functional Magnetic
Resonance Imaging Study,” Journal of Marketing Research
(46:1), pp. 1-13.
Henson, R. 2005. “What Can Functional Neuroimaging Tell the
Experimental Psychologist?,” Quarterly Journal of Experimental
Psychology, (58:1), pp. 193-233.
Holmes, A. P., and Friston, K. J. 1998. “Generalisability, Random
Effects and Population Inference,” NeuroImage, (7), pp. S754.
Hsu, M., Bhatt, M., Adolphs, R., Tranel, D., and Camerer, C. F.
2005. “Neural Systems Responding to Degrees of Uncertainty in
Human Decision-Making,” Science (310:5754), pp. 1680-1683.
Huesing, B., Jancke, L., and Tag, B. 2006. Impact Assessment of
Neuroimaging, Zurich: Hochschulverlag.
Huettel, S. A., and Payne, J. W. 2009. “Integrating Neural and
Decision Sciences: Convergence and Constraints,” Journal of
Marketing Research (46:1), pp. 14-24.
Huettel, S. A., Song, A. W., and McCarthy, G. 2004. Functional
Magnetic Resonance Imaging, Sunderland, MA: Sinauer
Associates.
Huettel, S. A., Song, A. W., and McCarthy, G. 2005. “Decisions
Under Uncertainty: Probabilistic Content Influences Activation
of Prefrontal and Parietal Cortices,” Journal of Neuroscience
(25:13), pp. 3304-3311.
Huettel, S. A., Song, A. W., and McCarthy, G. 2008. Functional
Magnetic Resonance Imaging (2nd ed.), Sunderland, MA: Sinauer
Associates.
Huettel, S. A., Stowe, C. J. Gordon, E. M., Warner, B. T., and Platt,
M. L. 2006. “Neural Signatures of Economic Preferences for
Risk and Ambiguity,” Neuron (49:5), pp. 765-775.
Jarvenpaa, S. L., Dickson, G. W., and DeSanctis, G. 1985. “Methodological Issues in Experimental IS Research: Experiences and
Recommendations,” MIS Quarterly (9:2), pp. 141-156.
Jenkinson, M. 2001. “Registration, Atlases and Cortical Flattening,” in Functional MRI: An Introduction to Methods, P.
Jezzard, P. M. Matthews, and S. M. Smith (eds.), New York:
Oxford University Press, pp. 271-293.
Jezzard, P., Matthews, P. M., and Smith, S. M. 2001. Functional
MRI: An Introduction to Methods, New York: Oxford University Press.
Kiebel, S., and Holmes, A. 2003. Human Brain Function, New
York: Academic Press.
King-Casas, B., Tomlin, D., Anen, C., Camerer, C. F., and Quartz,
S. R. 2005. “Getting to Know You: Reputation and Trust in a
Two-Person Economic Exchange,” Science (308:5718), pp.
78-83.
MIS Quarterly Vol. 36 No. 3/September 2012
837
Dimoka/Conducting an fMRI Study in Social Science Research
Knoch, D., Treyer, D., Regard, M., Müri, R. M., Buck, A., and
Weber, B. 2006. “Lateralized and Frequency-Dependent Effects
of Prefrontal rTMS on Regional Cerebral Blood Flow,”
NeuroImage (31:2), pp. 641-648.
Knutson, B. Rick, S., Wimmer, G. E., Prelec, D., and Loewenstein, G. 2007. “Neural Predictors of Purchases,” Neuron (53),
pp. 147-156.
Krain, A., Wilson, A. M., Arbuckle, R., Castellanos, F. X. Milham,
M. P. 2006. “Distinct Neural Mechanisms of Risk and Ambiguity: A Meta-Analysis of Decision-Making,” NeuroImage
(32:1), pp. 477-484.
Kriegeskorte, N., Goebel, R., and Bandettini, P.
2006.
“Information-Based Functional Brain Mapping,” Proceedings of
the National Academy of Sciences of the United States of America
(103), pp. 3863-68.
Kuhnen, C., and Knutson, B. 2005. “The Neural Basis of Financial
Risk Taking,” Neuron (47:5), pp. 763-770.
Lange, N. 2001. “Statistical Procedures for Functional MRI,” in
Functional MRI, C. T. W. Moonen and P. A. Bandettini (eds.),
Heidelberg: Springer-Verlag Berlin, pp. 301-335.
LeDoux, J. 2003. “The Emotional Brain, Fear, and Amygdala,”
Cellular & Molecular NeuroBiology (23), pp. 727-38.
Lee, N., Broderick, A. J., and Chamberlain, L. 2007. “What Is
‘Neuromarketing’? A Discussion and Agenda for Future
Research,” International Journal of Psychophysiology (63:2), pp.
199-204.
Lieberman, M. D. 2007. “Social Cognitive Neuroscience: A
Review of Core Processes,” Annual Review of Psychology (58),
pp. 259-289.
Lieberman, M. D., and Cunningham, W. A. 2009. “Type I and
Type II Error Concerns in fMRI Research: Re-Balancing the
Scale,” SCAN (4:4), pp. 423-442.
Linden, D. E., Bittner, R. A., Muckly, L., Waltz, J. A., Kriegeskorte,
N., Goebel, R., Singer, W., and Munk, M. H. 2003. “Cortical
Capacity Constraints for Visual Working Memory: Dissociation
of fMRI Load Effects in a Fronto-Parietal Network,” NeuroImage
(20), pp. 1518-1530.
Logothetis N. K. 2008. “What We Can Do and What We Cannot
Do with fMRI,” Nature (453), pp. 869-878.
Logothetis, N. K., and Wandell B. A. 2004. “Interpreting the
BOLD Signal,” Annual Reviews Physiology (66), pp. 735-69
Logothetis, N. K., Pauls, J., Augath, M., Trinath T., and
Oeltermann, A. 2001. “Neurophysiological Investigation of the
Basis of the fMRI Signal,” Nature (412), pp. 150-157.
Lo, A. W., and Repin, D. V. 2002. “The Psychophysiology of
Real-Time Financial Risk Processing,” Journal of Cognitive
Neuroscience (14:3), pp. 323-339.
Lorig, T. S. 2009. “What Was the Question? FMRI and Inference
in Psychophysiology,” International Journal of Psychophysiology (73), pp. 17-21.
Mandeville, J. B., and Rosen, B. R. 2002. “Functional MRI,” in
Brain Mapping: The Methods (2nd ed.), A. W. Toga and J. C.
Mazziota (eds.), San Diego, CA: Elsevier, pp. 605-631.
838
MIS Quarterly Vol. 36 No. 3/September 2012
Mast, F. W., and Zaltman, G. 2005. “A Behavioral Window on the
Mind of the Market: An Application of the Response Time
Paradigm,” Brain Research Bulletin (67:5), pp. 422-427.
McCabe, K., Houser, D., Lee, R., Smith, V. L., and Trouard, T.
2001. “A Functional Imaging Study of Cooperation in TwoPerson Reciprocal Exchange,” Proceedings of the National
Academy of Sciences of the United States of America (98), pp.
11832-11835.
McClure, S. M., Berns, G. S., and Montague, P. R. 2003. “Temporal Prediction Errors in a Passive Learning Task Activate
Human Striatum,” Neuron (38:2), pp. 339-346.
McClure, S. M., Laibson, D. I., Loewenstein, G., and Cohen, J. D.
2004. “Separate Neural Systems Value Immediate and Delayed
Monetary Rewards,” Science (306:5695), pp. 503-507.
McKeown, M., Jung, T.-P., Makeig, S., Brown, G., Kinderman, S.,
Lee, T.-W., and Sejnowski, T. 1998. “Spatially Independent
Activity Patterns in Functional MRI Data During the Stroop
Color Naming Task,” Proceedings of the National Academy of
Sciences of the United States of America (95), pp. 803-810.
Mechelli, A., Penny, W. D., Price, C. J., Gitelman, D., and Friston,
K. J. 2002. “Effective Connectivity and Intersubject Variability:
Using a Multisubject Network to Test Differences and
Commonalities,” NeuroImage (17), pp. 1459-1469.
Menon, V., and Crottaz-Herbette, S. 2005. “Combined EEG and
fMRI Studies of Human Brain Function,” International Review
of Neurobiology (66), pp. 291-321.
Miller, G. 2008. “Growing Pains for fMRI,” Science (320:5882),
pp. 1412-1414.
Minnery, B. S., and Fine, M. S. 2009. “Neuroscience and the
Future of Human Computer Interaction,” Interactions (16:2), pp.
70-75.
Mumford, J. A., and Nichols, T. E. 2008. “Power Calculation for
Group fMRI Studies Accounting for Arbitrary Design and
Temporal Autocorrelation” NeuroImage (39:1), pp. 261-268.
Murphy, K., and Garavan, H. 2004. “An Empirical Investigation
into the Number of Subjects Required for an Event-Related fMRI
Study,” Neuroimaging (22), pp., 879-885.
Ogawa, S., Lee, T.-M., Nayak, A. S., and Glynn, P. 1990.
“Oxygenation-Sensitive Contrast in Magnetic Resonance Image
of Rodent Brain at High Magnetic Fields,” Magnetic Resonance
Medicine (14), pp. 68-78.
Ogawa, S., Tank, D. W., Menon, R., Ellermann, J. M., Kim, S.-G.,
Merkle, H., and Ugurbil, K. 1992. “Intrinsic Signal Changes
Accompanying Sensory Stimulation: Functional Brain Mapping
with Magnetic Resonance Imaging,” Proceedings of the National
Academy of Sciences of the United States of America (89), pp.
5951-5955.
Ollinger, J. M., Corbetta, M., and Shulman, G. L. 2001.
“Separating Processes Within a Trial in Event-Related Functional
MRI: II. Analysis,” NeuroImage (13:1), pp. 218-229.
Paulus, M. P., Rogalsky, C., Simmons, A., Feinstein, J. S., and
Stein, M. B. 2003. “Increased Activation in the Right Insula
During Risk-Taking Decision Making Is Related to Harm
Avoidance and Neuroticism,” NeuroImage (19), pp. 1439-1448.
Dimoka/Conducting an fMRI Study in Social Science Research
Pavlou, P. A., and Dimoka, A. 2006. “The Nature and Role of
Feedback Text Comments in Online Marketplaces: Implications
for Trust Building, Price Premiums, and Seller Differentiation,”
Information Systems Research (17:4), pp. 391-412.
Pessoa, L. 2008. “On the Relationship Between Emotion and
Cognition,” Nature (9), pp. 148-159.
Phan, K. L., Taylor, S. F., Welsh, R. C., Ho, S., Britton, J. C., and
Liberzon, I. 2004. “Neural Correlates of Individual Ratings of
Emotional Salience: A Trial-Related fMRI Study,” NeuroImage
(21), pp. 768-780.
Phan, K. L., Wager, T., Taylor, S. F., and Liberzon, I. 2002.
“Functional Neuroanatomy of Emotion: A Meta-Analysis of
Emotion Activation Studies in PET and fMRI,” NeuroImage
(16), pp. 331-348.
Phelps, E. A. 2006. “Emotion and Cognition: Insights from
Studies of the Human Amygdala,” Annual Review of Psychology
(57:1), pp. 27-53.
Poldrack, R. A. 2006. “Can Cognitive Processes be Inferred from
Neuroimaging Data?,” Trends in Cognitive Sciences (10:2), pp.
59-63.
Poldrack, R. A. 2007. “Region of Interest Analysis for fMRI,”
Social Cognitive Affective Neuroscience (2), pp. 67-70.
Poldrack, R. A. 2008. “The Role of fMRI in Cognitive Neuroscience: Where Do We Stand?,” Current Opinion in Neurobiology (18), pp. 223-227.
Poldrack, R. A, Fletcher, P. C., Henson, R. N., Worsley, K. J., Brett,
M., and Nichols, T. E. 2008. “Guidelines for Reporting an fMRI
Study,” NeuroImage, (40:2), pp. 409-414.
Price, C. J., and Friston, K. J. 2005. “Functional Ontologies for
Cognition: The Systematic Definition of Structure and
Function,” Cognitive Neuropsychology (22), pp. 262-275.
Ramsey, N. F., Hoogduin, H., and Jansma, J. M. 2002. “Functional
MRI Experiments: Acquisition, Analysis, and Interpretation of
Data,” European Neuropsychopharmacology (12), pp. 517-526.
Rees, G., Keiman, G., and Koch, C. 2002. “Neural Correlates of
Consciousness in Humans,” Nature Reviews: Neuroscience (3),
pp. 261-270.
Riedl, R., Banker, R., Benbasat, I., Davis, F. D., Dennis, A.,
Dimoka, A., Gefen, D., Gupta, A., Ischebek, A., Kenning, P.,
Pavlou, P. A., Muller-Putz, G., Straub, D., van Brocke, J., and
Weber, B. 2010. “On the Foundations of NeuroIS: Reflections
on the Gmunden Retreat 2009,” Communications of the Association for Information Systems (27:15), pp. 243-264.
Riedl, R., Hubert, M., and Kenning, P. 2010. “Are There Neural
Gender Differences in Online Trust? An fMRI Study on the
Trustworthiness of eBay Offers,” MIS Quarterly (34:2), pp.
397-428.
Rilling, J. K., Gutman, D. A., Zeh, T. R., Pagnoni, G., Berns, G. S.,
and Kilts, C. D. 2002. “A Neural Basis for Social Cooperation,”
Neuron (35:2), pp. 395-405.
Rilling, J. K., Sanfey, A. G., Aronson, J. A., Nystrom, L. E. , and
Cohen, J. D. 2004. “The Neural Correlates of Theory of Mind
within Interpersonal Interactions, NeuroImage (22:4), pp.
1694-1703.
Roche, A., Mériaux, S., Keller, M., and Thirion, B. 2007. “MixedEffect Statistics for Group Analysis in fMRI: A Nonparametric
Maximum Likelihood Approach,” NeuroImage (38:3), pp.
501-510.
Sanfey, A. G., Rilling, J. K., Aronson, J. A., Nystrom. L. E., and
Cohen, J. D. 2003. “The Neural Basis of Economic DecisionMaking in the Ultimatum Game,” Science (300:5626), pp.
1755-1758.
Siegel, S., and Castellan, J. 1988. Nonparametric Statistics for the
Social Sciences (2nd ed.), New York: McGraw-Hill.
Smith, K., Dickhaut, J., McCabe, K., and Pardo, J. V. 2002.
“Neuronal Substrates for Choice under Ambiguity, Risk, Gains,
and Losses,” Management Science (48:6), pp. 711-718.
Smith, S. M. 2001. “Preparing fMRI Data for Statistical Analysis,”
in Functional MRI: An Introduction to Methods, P. Jezzard, P.
M. Matthews, and S. M. Smith (eds.), New York: Oxford
University Press, pp. 243-250.
Soltysik, D. A., and Hyde, J. S. 2008. “High Spatial Resolution
Increases the Specificity of Block-Design BOLD fMRI Studies
of Overt Vowel Production,” NeuroImage (41), pp. 389-397.
Song, A. W., Huettel, S. A., and McCarthy, G. 2006. “Functional
Neuroimaging: Basic Principles of Functional MRI,” in Handbook of Functional Neuroimaging of Cognition, R. Cabeza and
A. Kingstone (eds.), Cambridge, MA: MIT Press, pp. 21-52.
Stanberry, L., Murua, A., and Cordes, D. 2008. “Functional Connectivity Mapping Using the Ferromagnetic Potts Spin Model,”
Human Brain Mapping (29), pp. 422-440.
Straub, D. W., Boudreau, M. C., and Gefen, D. 2004. “Validation
Guidelines for IS Positivist Research,” Communications of the
Association for Information Systems (13), pp. 380-427.
Talairach, J., and Tournoux, P. 1988. Co-Planar Stereotaxic Atlas
of the Human Brain: 3-Dimensional Proportional System: An
Approach to Cerebral Mapping, New York: Thieme.
Thirion, B., Pinel, P., Mériaux, S., Roche, A., Dehaene, S., and
Poline, J-B. 2007. “Analysis of a Large fMRI Cohort: Statistical and Methodological Issues for Group Analyses,”
NeuroImage (36:10), pp. 105-120.
Todorov, A. 2008. “Evaluating Faces on Trustworthiness: An
Extension of Systems for Recognition of Emotions Signaling
Approach/Avoidance Behaviors,” Annals of the New York
Academy of Sciences (1124), pp. 208-224.
Van Horn, J. D., and Poldrack, R. A. 2009. “Functional MRI at the
Crossroads,” International Journal of Psychophysiology (73:1),
pp. 3-9.
Vuilleumier, P., Armony, J. L., Driver, J., and Dolan, R. J. 2001.
“Effects of Attention and Emotion on Face Processing in the
Human Brain,” Neuron (30:3), pp. 829-841.
Vul, E., Harris, C., Winkielman, P., and Pashler, H. 2009.
“Puzzlingly High Correlations in fMRI Studies of Emotion,
Personality, and Social Cognition,” Perspectives on Psychological Science (4), pp. 274-290.
Wicker, B., Keysers, C., Plailly, J., Royet, J., Gallese, V., and
Rizzolatti, G. 2003. “Both of Us Disgusted in My Insula: The
Common Neural Basis of Seeing and Feeling Disgust,” Neuron
(40), pp. 655-665.
Winston, J. S., Strange, B. A., O’Doherty, J., and Dolan, R. J.
2002. “Automatic and Intentional Brain Responses During
Evaluation of Trustworthiness of Faces,” Nature Neuroscience
(5), pp. 277-283.
MIS Quarterly Vol. 36 No. 3/September 2012
839
Dimoka/Conducting an fMRI Study in Social Science Research
Worsley, K. J. 2001. “Statistical Analysis of Activation Images,”
in Functional MRI: An Introduction to Methods, P. Jezzard, P.
M. Matthews, and S. M. Smith (eds.), New York: Oxford
University Press, pp. 251-270.
Worsley, K. J., and Friston, K. J. 1995. “Analysis of fMRI TimeSeries Revisited—Again,” NeuroImage (2), pp. 173-181.
Yoon, C., Gonzales, R., and Bettman, J. R. 2009. “Using fMRI to
Inform Marketing Research: Challenges and Opportunities,”
Journal of Marketing Research (46:1), pp. 17-19.
Yoon, C., Gutchess, A. H., Feinberg, F., and Polk, T. A. 2006. “A
Functional Magnetic Resonance Imaging Study of Neural
Dissociations between Brand and Person Judgments,” Journal of
Consumer Research (33:1), pp. 31-40.
Zheng, E., and Pavlou, P. A. 2010. “Toward a Causal Interpretation
for Structural Models: A New Bayesian Networks Method for
Observational Data with Latent Variables,” Information Systems
Research (21:2), pp. 365-391.
840
MIS Quarterly Vol. 36 No. 3/September 2012
About the Author
Angelika Dimoka is an associate professor of Marketing and
Management of Information Systems at the Fox School of Business,
Temple University. She is also director of the Center for Neural
Decision Making. She holds a Ph.D. from the Viterbi School of
Engineering (specialization in Neuroscience and Brain Imaging)
with a minor in MIS from the Marshall School of Business at the
University of Southern California. Angelika’s research interests are
in the areas of cognitive neuroscience and functional neuroimaging
in marketing and MIS (Neuromarketing and NeuroIS), quantitative
analysis of uncertainty in online marketplaces, and modeling of
information pathways in the brain. Her research has appeared in
Information Systems Research, MIS Quarterly, NeuroImage, IEEE
Transactions in Biomedical Engineering, Annals of Biomedical
Engineering, IEEE in Biology and Medicine, and Neuroscience
Methods.
RESEARCH ESSAY
HOW TO CONDUCT A FUNCTIONAL MAGNETIC RESONANCE
(FMRI) STUDY IN SOCIAL SCIENCE RESEARCH1
Angelika Dimoka
Fox School of Business, Temple University,
Philadelphia, PA 19122 U.S.A. {angelika@temple.edu) }
Appendix A
Glossary
Term
Arteries
Explanation
Blood vessels that carry oxygenated blood from the heart to the rest of the body.
Axial
A horizontal view of the brain (along the x–y plane in MRI)
B0
The strong static magnetic field generated by an MRI scanner.
Basis functions
A set of functions whose linear combination can take on a wide range of functional forms. In
fMRI analyses, researchers often replace a single hemodynamic response function with basis
functions, to improve the flexibility of their design matrices.
Between-subjects
A manipulation in which different conditions are assigned to different subject groups.
Block
A time interval that contains trials from one condition.
Blocked design
The separation of experimental conditions into distinct blocks, so that each condition is
presented for an extended period of time.
Blood-oxygenation-leveldependent (BOLD) contrast
The difference in signal on T2*-weighted images as a function of the amount of deoxygenated
hemoglobin.
Bonferroni correction
A stringent form of family-wise error rate correction for multiple comparisons in which the
alpha value is decreased proportionally to the number of independent statistical tests.
Brodmann areas
Divisions of the brain based on the influential cyto-architectonic criteria initially proposed by
Korbinian Brodmann.
Central nervous system
(CNS)
The system composed by the brain and the spinal cord.
Cerebrospinal fluid (CSF)
A colorless liquid that surrounds the brain and spinal cord and fills the ventricles within the
brain.
Consent form
An abstract concept that explains behavior but which itself is not directly observable.
Attention is an example of a psychological construct.
Contrast
(1) The intensity difference between different quantities being measured by an imaging
system. It also can refer to the physical quantity being measured (e.g., T1 contrast). (2) A
statistical comparison of the activation evoked by two (or more) experimental conditions, in
order to test a research hypothesis.
MIS Quarterly Vol. 36 No. 3—Appendices/September 2012
A1
Dimoka/Conducting an fMRI Study in Social Science Research
Term
Control block
Control condition
Converging operations
Co-registration
Coronal
Correlation analysis
Deactivations
Deoxygenated Hemoglobin
(dHb)
Design Matrix
Diamagnetic
Dorsal
Event
Event-related design
False Discovery Rate
Field map
Field of view (FOV)
Field potentials
Field strength
Flip angle
Frontal lobe
Functional magnetic
resonance imaging (fMRI)
General linear model
(GLM)
Gradient coils
Hemodynamic response
(HDR)
Independent components
analysis (ICA)
Informed consent
Institutional Review Board
(IRB)
Interleaved slice acquisition
A2
Explanation
A time interval that contains trials of the control condition.
A condition that provides a standard to which the experimental condition(s) can be compared.
Also called the baseline condition or non-task condition.
The use of two or more techniques to provide complementary evidence used to test an
experimental hypothesis or scientific theory.
The spatial alignment of two images or image volumes.
A frontal view of the brain (along the /x–z/ plane in MRI).
A type of statistical test that evaluates the strength of the relation between two variables. For
fMRI studies, correlation analyses typically evaluate the correspondence between a predicted
hemodynamic response and the observed data.
Decreases in BOLD activation during task blocks compared with non-task blocks.
Hemoglobin without attached oxygen; it is paramagnetic.
In fMRI implementations of the general linear model, the specification of how the model
factors change over time.
Having the property of a weak repulsion from a magnetic field.
Toward the top of the brain.
A single instance of the experimental manipulation. Also known as a trial.
Presentation of discrete, short-duration events with randomized timing and order.
Statistical method to correct for multiple comparisons in many hypotheses testing.
An image of the intensity of the magnetic field across space.
The total extent of an image along a spatial dimension.
Changes in electrical potential over space associated with postsynaptic neuronal activity.
The magnitude of the static magnetic field generated by a scanner, typically expressed in
units of Tesla.
The change in the precession angle of the net magnetization following excitation.
The most anterior lobe of the cerebrum; it is important for executive processing, motor
control, memory, and many other functions.
A neuroimaging technique that uses standard MRI scanners to investigate changes in brain
function over time.
A class of statistical tests that assume that the experimental data are composed of the linear
combination of different model factors,
Electromagnetic coils that create controlled spatial variation in the strength of the magnetic
field.
The change in MR signal on T_2 * images following local neuronal activity. The
hemodynamic response results from a decrease in the amount of deoxygenated hemoglobin
present within a voxel.
An important class of data-driven analyses that identify stationary sets of voxels whose
activations vary together over time and are maximally distinguishable from other sets.
The process in which a potential subject voluntarily agrees to participate in a research study,
after learning about the procedures, risks, and benefits of the study.
An independent group of individuals who evaluate the ethical practice of research conducted
at their institution. Researchers who wish to conduct research with human subjects should
generally receive the approval of their local IRB.
The collection of data in an alternating order, so that data are first acquired from the oddnumbered slices and then from the even-numbered slices, to minimize the influence of
excitation pulses upon adjacent slices.
MIS Quarterly Vol. 36 No. 3—Appendices/September 2012
Dimoka/Conducting an fMRI Study in Social Science Research
Term
Interstimulus interval (ISI)
Intersubject correlations
Jittering
Magnetic resonance
MNI space
MR signal
Neural Correlates
Normalization
Parameter matrix
Peak
Pixel
Preprocessing
Pulse sequence
Random-effects analysis
Reaction time
Reference volume
Region-of-interest (ROI)
analyses
Relaxation
Repetition time (TR)
Response time
Reverse inference
Right-hand rule
Rigid-body transformation
Rostral
Sagittal
Segmentation
Explanation
The separation in time between successive stimuli. Usually refers to the time between the
end of one stimulus and the onset of the next, with the term “stimulus-onset asynchrony”
(SOA) used to define the time between successive onsets.
Functional MRI time courses that are shared by different individuals while performing the
same experimental tasks or experiencing the same stimuli.
Randomizing the intervals between successive stimulus events over some range.
Absorption of energy from a magnetic field that oscillates at a particular frequency.
A commonly used space for normalization of fMRI data; its coordinates are derived from an
average of MRI structural images from several hundred individuals.
The current measured in a detector coil following excitation and reception.
Brain areas activated by a certain construct
The transformation of MRI data from an individual subject to match the spatial properties of a
standardized image, such as an averaged brain derived from a sample of many individuals.
A matrix that describes the relative contributions of each model to the observed data for each
voxel.
The maximal amplitude of the hemodynamic response, occurring typically about 4 to 6 s
following a short-duration event.
A two-dimensional picture element.
Computational procedures that are applied to fMRI data following image reconstruction but
before statistical analysis. Preprocessing steps are intended to reduce the variability in the
data that is not associated with the experimental task, and to prepare the data for statistical
testing.
A series of changing magnetic field gradients and oscillating electromagnetic fields that
allows the MRI scanner to create images sensitive to a particular physical property.
Intersubject analysis that treats the effect of the experimental manipulation as variable across
subjects, so that it could have a different effect on different subjects.
The time required for someone to make a simple motor response to the presentation of a
visual stimulus. Note that this is distinct from response time, which applies to situations in
which someone should choose between two or more possible responses.
A target image volume to which other image volumes are to be aligned.
Evaluations of hypotheses about the functional properties of brain regions (i.e., aggregated
over a pre-determined set of voxels), often chosen to reflect a priori anatomical distinctions
within the brain.
A change in net magnetization over time.
The time interval between successive excitation pulses, usually expressed in seconds.
The time required for someone to execute a choice between two or more possible responses.
Note that this is distinct from reaction time, which applies to situations when only one
possible response is present.
Reasoning from the outcome of a dependent variable to infer the state of an independent
variable (or an intervening unobservable variable).
A method used to determine the direction of a magnetic moment generated by a moving
charge or electrical current. If the fingers of the right hand are curled around the direction of
spin, then the magnetic moment will be in the direction indicated by the thumb.
A spatial transformation that does not change the size or shape of an object; it has three
translational parameters and three rotational parameters.
Toward the front of the brain.
A side view of the brain (along the /y/–/z/ plane in MRI).
The process of partitioning an image into constituent parts, typically types of tissue (e.g., gray
matter, white matter) or topographical divisions (e.g., different structural regions like
Brodmann areas).
MIS Quarterly Vol. 36 No. 3—Appendices/September 2012
A3
Dimoka/Conducting an fMRI Study in Social Science Research
Term
Explanation
Signal
Meaningful changes in some quantity. For fMRI, an important class of signals includes
changes in intensity associated with the BOLD response.
Signal-to-noise ratio (SNR)
The relative strength of a signal compared with other sources of variability in the data.
Slice
A single slab of an imaging volume. The thickness of the slice is defined by the strength of
the gradient and the bandwidth of the electromagnetic pulse used to select it.
Slice selection
The combined use of a spatial magnetic field gradient and an electromagnetic pulse to excite
spins within a slice.
Small-volume correction
The restriction of analyses to specific regions-of-interest, defined a priori, to reduce the total
number of statistical tests and thus allow for a more liberal significance threshold.
Smoothness
The degree to which the time courses of nearby voxels are temporally correlated.
Spatial smoothing
The blurring of fMRI data across adjacent voxels to improve the validity of statistical testing
and maximize functional SNR, at a cost of spatial resolution.
Static magnetic field
The strong magnetic field at the center of the MRI scanner whose strength does not change
over time. The strength of the static magnetic field is measured in Tesla.
Statistical map
(or Statistical parameter
map)
In fMRI, the labeling of all voxels within the image according to the outcome of a statistical
test.
Stereotaxic space
A precise mapping system (e.g., of the brain) using 3-D coordinates.
Subtraction
In experimental design, the direct comparison of two conditions that are assumed to differ
only in one property, the independent variable.
Superposition
A principle of linear systems that states that the total response to a set of inputs is equivalent
to the summation of the independent responses to the inputs.
Talairach space
A commonly used space for normalization of fMRI data; its coordinates are based on
measurements from a single post-mortem human brain, as published in an atlas by Talairach
and Tournoux.
Time course
The change in MR signal over a series of fMRI images.
Time series
A large number of fMRI images collected at different points in time.
Translation
The movement of an object along an axis in space (in the absence of rotation).
Trial
A single instance of the experimental manipulation, such as stimulus presentation.
Ventral
Toward the bottom of the brain.
Volume (brain)
The collection of all images of the brain, consisting of multiple slices and voxels.
Volumetric
Relating to the measurement of volume.
Voxel
A three-dimensional volume element.
Voxel-wise analysis
Evaluation/testing of hypotheses about functional properties of individual voxels (or small
clusters of voxels), often throughout the entire brain volume.
A4
MIS Quarterly Vol. 36 No. 3—Appendices/September 2012
Dimoka/Conducting an fMRI Study in Social Science Research
Appendix B
Review of Neuroscience Literature Related to Constructs
of Interest to the Social Sciences
Construct/Process
Sample Brain Areas*
Ambiguity
Anger
Anxiety
Insular cortex, Parietal cortex
Lateral Orbitofrontal cortex
Amygdala–prefrontal circuitry, Inferior Frontal gyrus
(Brodman Area 45), Ventromedial Prefrontal cortex
Attention
Automaticity
Right Frontal and Parietal cortices and Thalamus
Frontal and Striatal cortex, Parietal lobe (Deactivation)
Calculation
Anterior Cingulate cortex, Prefrontal cortex limbic system
(mainly anterior cingulate cortex and amygdala)
Dorsolateral prefrontal cortex, Parietal cortex
Cognitive Effort
Competition
Consciousness
Inferior parietal cortex, Medial Prefrontal cortex
Parietal and Dorsal Prefrontal cortex, Striate cortex,
Extrastriate cortex
Cooperation
Disgust
Orbitofrontal cortex
Insular cortex
Displeasure
Amygdala, Hippocampus, Insular cortex, Superior Temporal
gyrus
Amygdala, Insular cortex
Distrust
Emotion in Moral Judgment
Emotional Processing
Medial Prefrontal cortex, Posterior Cingulate, and angular
gyrus
Anterior Cingulate cortex, Medial Prefrontal cortex (Emotional
Information- dorsal frontomedial cortex)
Envy
Fear
Anterior Cingulate cortex
Amygdala
Flow
Dorsomedial Prefrontal cortex, Medial Parietal cortex
Frustration
Habit
Right Anterior Insula, Right Ventral Prefrontal cortex
Basal Ganglia, Medial Prefrontal cortex, Medial Temporal
lobe
Basal Ganglia (Ventral Striatum and Putamen)
Happiness
Hate
Medial Frontal gyrus, Right Putamen, Bilaterally in Premotor
cortex, Frontal Pole and bilateral Medial Insula, Right Insula,
Right Premotor cortex , Right Fronto-Medial gyrus
Key References
Krain et al. 2006
Murphy et al. 2003
Bishop 2007
Etkin and Wager 2007
Mujica-Parodi 2007
Coull et al. 1998
Kubler et al. 2006
Poldrack et al. 2005
Ernst and Paulus 2005
McClure et al. 2004
Linden et al. 2003
Owen et el. 2005
Decety et al. 2004
Rees, Kreiman, and Koch
2002
Rees, Wojciulik et al. 2002
Rilling, Gutman et al. 2002
Britton et al. 2006
Lane et al. 1997
Murphy et al. 2003
Phan et al. 2002
Britton et al. 2006
Casacchia et al. 2009
Dimoka 2010
Winston et al. 2002
Greene et al. 2001
Damasio 1996
Ferstl et al. 2005
Phan et al. 2002
Takahashi et al. 2009
LeDoux 2003
Murphy et al. 2003
Phan et al. 2002
Iacobini et al. 2004
Katayose 2006
Abler et al. 2005
Graybiel 2008
Salat et al. 2006
Murphy et al. 2003
Phan et al. 2002
Zeki and Romaya 2008
MIS Quarterly Vol. 36 No. 3—Appendices/September 2012
A5
Dimoka/Conducting an fMRI Study in Social Science Research
Construct/Process
Information Processing
Intentions
Jealousy
Language function
Loss
Love (maternal)
Love (overlap of maternal
and romantic)
Love (romantic)
Moral Judgments
Moral Sensitivity
Motor Intentions
Multi-Tasking
Optimism
Person Recognition
Pleasure/Enjoyment
Sample Brain Areas*
Anterior Frontal cortex, Lateral Prefrontal cortex, Medial
Orbitofrontal cortex Hippocampus, Amygdala (Emotional
Information- Dorsal Frontomedial cortex)
Ventrolateral Prefrontal cortex, Brodmann Area 47
Left Prefrontal cortex
Broca’s area
Insular cortex
Ventral part of Anterior Cingulate cortex
Striatum (Putamen, Globus Pallidus, Caudate Nucleus),
Middle Insula and Dorsal Anterior Cingulate cortex
Dentate gyrus/Hippocampus, Hypothalamus, Ventral
Tegmental area
Frontopolar cortex (Brodmann Area 10), Posterior Superior
Temporal Sulcus
Amygdala, Thalamus, Upper Midbrain, Medial Orbitofrontal
cortex, Medial Prefrontal cortex, Superior Temporal Sulcus
Premotor and Parietal cortex
Fronto-polar cortex (Brodman Area 10)
Rostral Anterior Cingulate cortex, Amygdala
Left Hippocampus, Left Middle Temporal gyrus, Left Insula,
and Bilateral Cerebellum
Anterior Cingulate cortex, Putamen, Medial Prefrontal cortex
,Nucleus Accumbens
Priming
Parietal cortex, Middle Temporal cortex, Posterior Superior
cortex
Rewards and Utility
Anterior Cingulate cortex, Caudate Nucleus, Nucleus
Accumbens, Putamen
Risk
Sadness
Nucleus Accumbens
Subcallosal Cingulate cortex
Self-reflection
Self-regulation of emotion
Social Cognition
Medial Prefrontal cortex, Posterior Cingulate
Amygdala, Dorsolateral Prefrontal cortex, Hypothalamus
Amygdala, Cingulate cortex, Temporal lobe, Orbitofrontal
cortex, Right Somatosensory cortex, Ventromedial Frontal
cortex
Amygdala, Orbitofrontal cortex, Dorsolateral Prefrontal
cortex
Hippocampus, Medial Temporal Lobe
Social Cooperation
Spatial Cognition
Sympathy
A6
Anterior Superior Frontal gyrus, Inferior Frontal gyrus,
Temporal pole, Amygdala, Left Central Sulcus, Right Dorsal
Premotor cortex, Dorsomedial prefrontal cortex, pre-SMA,
and Inferior Parietal lobule
MIS Quarterly Vol. 36 No. 3—Appendices/September 2012
Key References
Dimoka and Davis 2008
Elliot et al. 1997
Ferstl et al. 2005
Dove et al. 2008
Okuda et al. 1998
Harmon-Jones et al. 2009
McDermott et al. 2003
Paulus and Frank 2003
Bartels and Zeki 2004
Bartels and Zeki 2004
Bartels and Zeki 2004
Borg et al. 2006
Moll et al. 2005
Moll et al. 2002
Desmurget et al. 2009
Lau et al. 2007
Dreher et al. 2008
Sharot et al. 2007
Paller et al. 2003
Klasen et al. 2008
McLean et al. 2009
Sabatinelli et al. 2008
Naccache and Dehaene
2001
Wible et al. 2006
Bush et al. 2002
Delgado et al. 2005
McClure et al. 2004
Knutson et al. 2001
Murphy et al. 2003
Phan et al. 2002
Johnson et al. 2002
Beauregard et al. 2001
Adolphs 1999, 2001
Rilling, Glenn et al. 2007
Moser et al. 2008
Shrager et al. 2008
Decety et al. 2004
Dimoka/Conducting an fMRI Study in Social Science Research
Construct/Process
Sample Brain Areas*
Task Intentions
Anterior Cingulate cortex, Medial and Lateral Prefrontal
Theory of Mind
Trust
Anterior Paracingulate cortex, Medial Prefrontal cortex
Anterior Paracingulate cortex, Caudate Nucleus, Putamen
Uncertainty
Orbitofrontal and Parietal cortex
Unfairness
Dorsolateral Prefrontal cortex, Ventrolateral Prefrontal cortex,
Superior Temporal Sulcus
Dorsal Lateral Geniculate nucleus to the temporal lobe
Anterior Cingulate, Dorsolateral Prefrontal cortex, Orbital
Cortical cortex
Visual Perception
Working Memory
Key References
Haynes et al. 2007
Winterer et al. 2002
McCabe et al. 2001
Dimoka 2010
King-Casas et al. 2005
Huettel et al. 2005
Krain et al. 2006
Dulebohn et al. 2009
Pollen 1999
Braver et al. 1997
Callicott et al. 1999
Cohen et al. 1997
*As discussed in the paper in more detail (“Basic Neuroscience Foundations), while it is possible to make a (forward) inference that a specific
construct is associated with certain brain areas, caution should be paid when trying to infer a construct based on the existence of activation in
certain brain areas (termed reverse inference). Given that the same brain areas are activated in response to several constructs (as clearly noted
in Appendix B), such over-interpretation is problematic. For more details on forward and reverse inference, see Lieberman (2007) and Poldrack
(2008).
References
Abler, B., Walter, H., and Erk, S. 2005. “Neural Correlates of Frustration,” Neuroreport (16:7), pp. 669-672.
Adolphs, R. 1999. “Social Cognition and the Brain,” Trends in Cognitive Science (3:12), pp. 469-479.
Adolphs, R. 2001. “The Neurobiology of Social Cognition,” Current Opinion in Neurobiology (11), pp. 231-239.
Bartels, A., and Zeki, S. 2004. “The Neural Correlates of Maternal and Romantic Move,” NeuroImage (21:3), pp. 1155-1166.
Beauregard, M., Le´ vesque, J., and Bourgouin, P. 2001. “Neural Correlates of Conscious Self-Regulation of Emotion,” The Journal of
Neuroscience (21), pp. 1-6.
Bishop, S. J. 2007. “Neurocognitive Mechanisms of Anxiety: An Integrative Account,” Trends in Cognitive Sciences (11:7), pp. 307-316.
Borg, J. S., Hynes C., Van Horn, J., Grafton, S., and Sinnott-Armstrong, W. 2006. “Consequences, Action, and Intention as Factors in Moral
Judgments: An fMRI Investigation,” Journal of Cognitive Neuroscience (18:5), pp. 803-817.
Braver, T. S., Cohen, J. D., Jonides, J., Smith, E. E., and Noll, D. C. 1997. “A Parametric Study of Prefrontal Cortex Involvement in Human
Working Memory,” NeuroImage (5:1), pp. 49-62.
Britton, J. C., Phan, K. L., Taylor, S. F., Welsh, R. C., Berridge, K. C., and Liberzon, I. 2006. “Neural Correlates of Social and Nonsocial
Emotions: An fMRI Study,” NeuroImage (31), pp. 397-409.
Bush, G., Vogt, B. A., Holmes, J., Dale, A. M., Greve, D., Jenike, M. A., and Rose, B. R. 2002. “Dorsal Anterior Cingulate Cortex: A Role
in Reward-Based Decision Making,” Proceedings of the National Academy of Sciences of the United States of America (99:1), pp.
523-328.
Callicott, J. H., Mattay, V. S., Bertolino, A., Finn, K., Coppola, R., Frank, J. A., Goldberg, T. E., and Weinberger, D. R. 1999. “Physiological
Characteristics of Capacity Constraints in Working Memory as Revealed by Functional MRI,” Cerebral Cortex (9), 1999, pp. 20-26.
Casacchia, M., Mazza, M., Catalucci, A., Pollice, R., Gallucci, M., and Roncone, R. 2009. “Abnormal Emotional Responses to Pleasant and
Unpleasant Visual Stimuli in First Episode Schizophrenia: f-MRI Investigation” European Psychiatry (24:1), pp. S700.
Cohen, J. D., Perlstein, W. M., Braver, T. S., Nystrom, L. E., Noll, D. C., Jonides, J. and Smith, E. E. 1997. “Temporal Dynamics of Brain
Activation during a Working Memory Task,” Nature (386), pp. 604-608.
Coull, J. T. 1998. “Neural Correlates of Attention and Arousal: Insights from Electrophysiology, Functional Neuroimaging and Psychopharmacology,” Progress in Neurobiology (55:4), pp. 343-361.
Damasio, A. R., 1996. “The Somatic Marker Hypothesis and the Possible Functions of the Prefrontal Cortex,” Philosophical Transactions of
the Royal Society of London, Series B: Biological Sciences (351), pp. 1413-1420.
Decety, J., and Chaminade, T. 2003. “Neural Correlates of Feeling Sympathy,” Neuropsychologia (41:2), pp. 127-138.
Decety, J., Jackson, P. L., Sommerville, J. A., Chaminade, T., and Meltzoff, A. N. 2004. “The Neural Bases of Cooperation and Competition:
An fMRI Investigation,” NeuroImage (23:2), pp. 744-751.
Delgado, M. R., Miller, M. M., Inati, S., and Phelps, E. A. 2005. “An fMRI Study of Reward-related Probability Learning,” NeuroImage (24),
pp. 862-873.
MIS Quarterly Vol. 36 No. 3—Appendices/September 2012
A7
Dimoka/Conducting an fMRI Study in Social Science Research
Desmurget, M. Reilly, K. T., Richard, N., Szathmari, A., Mottolese, C., and Sirigu, A. 2009. “Movement Intention after Parietal Cortex
Stimulation in Humans,” Science (324:5298), pp. 811-813.
Dimoka, A. 2010. “What Does the Brain Tell Us About Trust and Distrust? Evidence from a Functional Neuroimaging Study,” MIS Quarterly
(34:2), pp. 373-396.
Dimoka, A., and Davis, F. D. 2008. “Where Does TAM Reside in the Brain? The Neural Mechanisms Underlying Technology Adoption,”
in Proceedings of the 29th International Conference on Information Systems, Paris.
Dove, A., Manly, T., Epstein, R., and Owen, A. M. 2008. “The Engagement of Mid-Ventrolateral Prefrontal Cortex and Posterior Brain
Regions in Intentional Cognitive Activity,” Human Brain Mapping (29), pp. 107-119.
Dreher, J. C., Koechlin, E., Tierney, M., and Grafman, J. 2008. “Damage to the Fronto-Polar Cortex Is Associated with Impaired
Multitasking,” PLoS ONE (3:9), e3227 (http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2528949/?tool=pubmed).
Dulebohn, J. H., Conlon, D. E., Sarinopoulos, I., Davison, R. B., and McNamara, G. 2009. “The Biological Bases of Unfairness:
Neuroimaging Evidence for the Distinctiveness of Procedural and Distributive Justice,” Organizational Behavior and Human Decision
Processes (110:2), pp. 140-151.
Elliott, R., Frith, C. D., and Dolan, R. J. 1997. “Differential Neural Response to Positive and Negative Feedback in Planning and Guessing
Tasks” Neuropsychologia (35:10), pp. 1395-1404.
Ernst, M., and Paulus, M. P. 2005. “Neurobiology of Decision Making: A Selective Review from a Neurocognitive and Clinical Perspective,”
Biological Psychiatry (58), pp. 597-604.
Etkin, A., and Wager, T. D. 2007. “Functional Neuroimaging of Anxiety: A Meta-Analysis of Emotional Processing in PTSD, Social Anxiety
Disorder, and Specific Phobia,” American Journal of Psychiatry (164), pp. 1476-1488.
Ferstl, C., Rinck, M., and von Cramon, D. Y. 2005. “Emotional and Temporal Aspects of Situation Model Processing During Text
Comprehension: An Event-Related fMRI Study,” Cognitive Neuroscience (17:5), pp. 724-739.
Graybiel, A. M. 2008. “Habits, Rituals, and the Evaluative Brain,” Annual Review of Neuroscience (31), pp. 359-387.
Greene, J. D., Sommerville, R. B., Nystrom, L. E., Darley, J. M., and Cohen, J. D. 2001. “An fMRI Investigation of Emotional Engagement
in Moral Judgment,” Science (293:5537), pp. 2105-2108.
Harmon-Jones, E., Peterson, C. K., and Harris, C. R. 2009. “Jealousy: Novel Methods and Neural Correlates,” Emotion (9:1), pp. 113-117.
Haynes, J-D, Sakai, K., Rees, G., Gilbert, S., Frith, C., and Passingham, R. E. 2007. “Reading Hidden Intentions in the Human Brain,” Current
Biology (17), pp. 323-328.
Huettel, S. A., Song, A. W., and McCarthy, G. 2005. “Decisions Under Uncertainty: Probabilistic Context Influences Activation of Prefrontal
and Parietal Cortices,” Journal of Neuroscience (25:13), pp. 3304-3311.
Iacoboni, M., Lieberman, M. D., Knowlton, B. J., Molnar-Szakacs, I., Moritz, M., Throop, J. C., and Fiske, A. P. 2004. “Watching Social
Interactions Produces Dorsomedial Prefrontal and Medial Parietal BOLD fMRI Signal Increases Compared to a Resting Baseline,”
NeuroImage (21), pp. 1167- 1173.
Johnson, S. C., Baxter, L. C., Wilder, L. S., Pipe, J. G., Heiserman, J. E., and Prigatano, G. P. 2002. “Neural Correlates of Self-Reflection,”
Brain (125:8), pp. 1808-1814.
Katayose, H., Nagata, N., and Kazai, K. 2006. “Investigation of Brain Activation While Listening to and Playing Music Using fNIRS,” Alma
Mater Studiorum, University of Bologna, August 22-26, pp. 107-114.
King-Casas, B., Tomlin, D., Anen, C., Camerer, C. F., Quartz, S. R., and Montague, P. R. 2005. “Getting to Know You: Reputation and Trust
in a Two-Person Economic Exchange,” Science (308:5718), pp. 78-83.
Klasen, M., Zvyagintsev, M., Weber, R., Mathiak, K. A., and Mathiak, K. 2008. “Think Aloud During fMRI: Neuronal Correlates of
Subjective Experience in Video Games,” in Fun and Games, P. Markopoulos, B. de Ruyter, W. Ijsselsteijn, and D. Rowland (eds.),
Heidelberg: Springer Berin, pp. 132-138.
Knutson, B., Fong, G. W., Adams, C. M., Varner, J. L., and Hommer, D. 2001. “Dissociation of Reward Anticipation and Outcome with
Event-Related fMRI,” Neuroreport (12), pp. 3683-3687.
Krain, A., Wilson, A. M., Arbuckle, R., Castellanos, F. X., and Milham, M. P. 2006. “Distinct Neural Mechanisms of Risk and Ambiguity:
A Meta-Analysis of Decision-Making,” NeuroImage (32:1), pp. 477-484.
Kubler, A., Dixon, V., and Garavan, H. 2005. “Automaticity and Reestablishment of Executive Control—An fMRI Study,” Journal of
Cognitive Neuroscience (18:8), pp. 1331-1342.
Lane, R. D., Reiman, E. M., Ahern, G. L., Schwartz, G. E., and Davidson, R. J. 1997. “Neuroanatomical Correlates of Happiness, Sadness,
and Disgust,” American Journal of Psychiatry (154), pp. 926-933.
Lau, H. C., Rogers, R. D., and Passingham, R. E. 2007. “Manipulating the Experienced Onset of Intention after Action Execution, Journal
of Cognitive Neuroscience (19:1), pp. 1-10.
LeDoux, J. 2003. “The Emotional Brain, Fear, and Amygdala,” Cellular & Molecular NeuroBiology (23), 2003, pp. 727-38.
Lieberman, M. D. 2007. “Social Cognitive Neuroscience: A Review of Core Processes,” Annual Review of Psychology (58), pp. 259-289.
Linden, D. E., Bittner, R. A., Muckli, L., Waltz, J. A., Kriegeskorte, N., Goebel, R., Singer, W., and Munk, M. H. J. 2003. “Cortical Capacity
Constraints for Visual Working Memory: Dissociation of fMRI Load Effects in a Fronto-Parietal Network,” NeuroImage (20), pp.
1518-1530.
A8
MIS Quarterly Vol. 36 No. 3—Appendices/September 2012
Dimoka/Conducting an fMRI Study in Social Science Research
McCabe, K., Houser, D., Ryan, L., Smith, V., and Trouard, T. 2001. “A Functional Imaging Study of Cooperation in Two-Person Reciprocal
Exchange,” Proceedings of the National Academy of Sciences of the United States of America (98), pp. 11832-11835.
McClure, S. M., Laibson, D. I., Loewenstein, G., and Cohen, J. D. 2004. “Separate Neural Systems Value Immediate and Delayed Monetary
Rewards,” Science (306:5695), pp. 503-507.
McDermott, K. B., Petersen, S. E., Jason M., Watson, J. M., and Ojemann, J. G. 2003. “A Procedure for Identifying Regions Preferentially
Activated by Attention to Semantic and Phonological Relations Using Functional Magnetic Resonance Imaging,” Neuropsychologia
(41:3), pp. 293-303.
McLean, J., Brennan, D., Wyper, D., Condon, B., Hadley, D., and Cavanagh, D. 2009. “Localisation of Regions of Intense Pleasure Response
Evoked by Soccer Goals,” Psychiatry Research: Neuroimaging (171), pp. 33-43.
Moll, J., de Oliveira-Souza, R., Eslinger, P. J., Ivanei, E., Bramati, I. E., Mourão-Miranda, J., Pedro Angelo Andreiuolo, P. A., and Luiz
Pessoa, L. 2002. “The Neural Correlates of Moral Sensitivity: a Functional Magnetic Resonance Imaging Investigation of Basic and
Moral Emotions,” Journal of Neuroscience (22:7), pp. 2730-2736
Moll, J., Zahn, R., de Oliveira-Souza, R., Krueger, F., and Grafman, J. 2005. “The Neural Basis of Human Moral Cognition,” Nature Review
of Neuroscience (6:10), pp. 799-809.
Moser, E. I., Kropff, E., and Moser, M-B. 2008. “Place Cells, Grid Cells, and the Brain’s Spatial Representation System,” Annual Review
of Neuroscience (31), pp. 69-89.
Mujica-Parodi, L. R., Korgaonkar, M., Ravindranath, B., Greenberg, B., Tomasi, D., Wagshul, M., Ardekani, B., Guilfoyle, D., Khan, S.,
Zhong, Y., Chon, K., and Malaspina, D. 2007. “Limbic Dysregulation Is Associated with Lowered Heart Rate Variability and Increased
Trait Anxiety in Healthy Adults,” Human Brain Mapping (30:1), pp. 47-58.
Murphy, F. C., Nimmo-Smith, I., and Lawrence, A. D. 2003. “Functional Neuroanatomy of Emotions: A Meta-Analysis,” Cognitive,
Affective, and Behavioral Neuroscience (3:3), pp. 207-233.
Naccache, L., and Dehaene, S. 2001. “The Priming Method: Imaging Unconscious Repetition Priming Reveals an Abstract Representation
of Number in the Parietal Lobes,” Cerebral Cortex (11:10), pp. 966-974.
Okuda, J., Toshikatsu, F., Yamadori, A., Kawashima, R., Tsukiura, T., Fukatsu, R., Suzuki, K., Ito, M., and Fukuda, H. 1998. “Participation
of the Prefrontal Cortices in Prospective Memory: Evidence from a PET study in Humans,” Neuroscience Letters (253), pp. 127-130.
Owen, A. M., McMillan, K. M., Laird, A. R., and Bullmore, E. 2005. “N-Back Working Memory Paradigm: A Meta-Analysis of Normative
Functional Neuroimaging Studies,” Human Brain Mapping (25), pp. 46-59.
Paller, K. A., Ranganath, C., Gonsalves, B., LaBar, K. S., Parrish, T. B., Gitelman, D. R., Mesulam, M. M., and Reber, P. J. 2003. “Neural
Correlates of Person Recognition,” Learning & Memory (10:4), pp. 253-260 (http://faculty.wcas.northwestern.edu/~paller/L&M03.pdf).
Paulus, M. P. and Frank, L. R. 2003. “Ventromedial Prefrontal Cortex Activation is Critical for Preference Judgments,” Neuroreport (14:10),
pp. 1311-1315.
Phan, K. L., Wager, T., Taylor, S. F., and Liberzon, I. 2002. “Functional Neuroanatomy of Emotion: A Meta-Analysis of Emotion Activation
Studies in PET and fMRI,” NeuroImage (16), pp. 331-348.
Poldrack, R. A. 2008. “The Role of fMRI in Cognitive Neuroscience: Where Do We Stand?,” Current Opinion in Neurobiology (18), pp.
223-227.
Poldrack, R. A., Sabb, F. W., Foerde, K., Tom, S. M., Asarnow, R. M., Bookheimer, S. Y., and Knowlton, B. J. 2005. “The Neural Correlates
of Motor Skill Automaticity,” Journal of Neuroscience (25:22), pp. 5356-5364.
Pollen, D. 1999. “On the Neural Correlates of Visual Perception,” Cerebral Cortex (9:1), pp. 4-19.
Rees, G., Kreiman, G., and Koch, C. 2002. “Neural Correlates of Consciousness in Humans,” Nature Reviews Neuroscience (3:4), pp.
261-270.
Rees, G., Wojciulik, E., Clarke, K. Husain M., Frith, C., and Jon Driver, J. 2002. “Neural Correlates of Conscious and Unconscious Vision
in Parietal Extinction,” Neurocase (8), pp. 387-393.
Rilling, J. K., Glenn, A. L., Jairam, M. R., Pagnoni, G., Goldsmith, D. R., Elfenbein, H. A., and Lilienfeld, S. O. 2007. “Neural Correlates
of Social Cooperation and Non-Cooperation as a Function of Psychopathy,” Biological Psychiatry (61:11), pp. 1260-1271.
Rilling, J. K., Gutman, D. A., Zeh, T. R., Pagnoni, G., Berns, G. S., and Kilts, C. D. 2002. “A Neural Basis for Social Cooperation,” Neuron
(35:2), pp. 395-405.
Sabatinelli, D., Bradley, M. M., Lang, P. J., Costa, V. D., and Versace, F. 2007. “Pleasure Rather than Salience Activates Human Nucleus
Accumbens and Medial Prefrontal Cortex,” Journal of Neurophysiology (98), pp. 1374-1379.
Salat, D. H., van der Kouwe, A. J., Tuch, D. S., Quinn, B. T., and Fischl, B. 2006. “Neuroimaging H.M.: A 10-Year Follow-Up Examination,”
Hippocampus (16), pp. 936-945.
Sharot, T., Riccardi, A., Raio, C., and Phelps, E. 2007. “Neural Mechanisms Mediating Optimism Bias,” Nature (450:7166), pp. 102-106.
Shrager, Y., Kirwan, C. B., and Squire, L. R. 2008. “Neural Basis of the Cognitive Map: Path Integration Does Not Require Hippocampus
or Entorhinal Cortex,” Proceedings of the National Academy of Sciences of the United States of America (105:33), pp. 12034-12038.
Takahashi, H., Kato, M., Matsuura, M., Mobbs, D., Suhara, T., and Okubo, Y. 2009. “When Your Gain Is My Pain and Your Pain Is My Gain:
Neural Correlates of Envy and Schadenfreude,” Science (323:5916), pp. 937-939.
MIS Quarterly Vol. 36 No. 3—Appendices/September 2012
A9
Dimoka/Conducting an fMRI Study in Social Science Research
Wible, C. G., Han, S. D., Spencer, M. H., Kubicki, M., Niznikiewicz, M. H., Jolesz, F. A., McCarley, R. W., and Nestor, P. 2006.
“Connectivity Among Semantic Associates: An fMRI Study of Semantic Priming,” Brain and Language (97), pp. 294-305.
Winston, J. S., Strange, B. A., O’Doherty, J., and Dolan, R. J. 2002. “Automatic and Intentional Brain Responses During Evaluation of
Trustworthiness of Faces,” Nature Neuroscience (5), pp. 277-283.
Winterer, G., Adams, C. M., Jones, D. W., and Knutson, B. 2002. “Volition to Action—An Event-Related fMRI Study,” NeuroImage (17),
pp. 851-858.
Zeki, S., and Romaya, J. P. 2008. “Neural Correlates of Hate,” PLoS ONE (3:10), e3556 (doi:10.1371/journal.pone.0003556).
Appendix C
Example Results
TRUST
Caudate Nucleus
Orbitofrontal
Caudate Nucleus
Anterior PCC
Putamen
LH
Putamen
Orbitofrontal
Anterior PCC
TRUST
HL
DISTRUST
LH
Insular Cortex
Amygdala
DISTRUST
Insular Cortex
Amygdala
TRUST
HL
DISTRUST
Figure C1. The Neural Correlates of Trust and Distrust (Simplified Version of Figure 3 (p. 386) from
Dimoka 2010)
A10
MIS Quarterly Vol. 36 No. 3—Appendices/September 2012
Dimoka/Conducting an fMRI Study in Social Science Research
Table C1. Coordinates of Neural Correlates of the Dimensions of Trust and Distrust
Construct
Credibility
Benevolence
Discredibility
Malevolence
Seller
Brain Area
High-Trust/Low-Distrust
Caudate Nucleus
High-Trust/Low-Distrust
Low-Trust/High-Distrust
High-Trust/Low-Distrust
Coordinates (x,y,z)
Activation (max)
Left: -12,18,4
Right: 12,18,4
z = 3.78, p < .01
Putamen
24, 6, 0
z = 3.05, p < .01
Orbitofrontal cortex
26, 44, -12
z = 3.54, p < .01
Caudate Nucleus
Left: -8, 14, 0
Right: 10, 18, 0
z = 2.65, p < .05
High-Trust/Low-Distrust
Putamen
22, 14, -2
z = 2.43, p < .05
High-Trust/Low-Distrust
Paracingulate cortex
6, 58, 2
z = 3.15, p < .01
High- Distrust /Low-Trust
Insular cortex
Left: -32, 22, -4
Right: 38, 18, -4
z = 3.79, p < .01
High- Distrust /Low-Trust
Amygdala
Left: -20, -4, -18
Right: 22, -2, -18
z = 4.05, p < .01
High- Distrust /Low-Trust
Insular cortex
Left: -32, 22, -6
Right: 38, 20, -6
z = 1.80, p < .10
(non-significant at p < .05)
Reference
Dimoka, A. 2010. “What Does the Brain Tell Us About Trust and Distrust? Evidence from a Functional Neuroimaging Study,” MIS Quarterly
(34:2), pp. 373-396.
MIS Quarterly Vol. 36 No. 3—Appendices/September 2012
A11
Copyright of MIS Quarterly is the property of MIS Quarterly & The Society for Information Management and
its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's
express written permission. However, users may print, download, or email articles for individual use.