Content uploaded by Diego Fernandez-Duque
Author content
All content in this area was uploaded by Diego Fernandez-Duque on May 03, 2015
Content may be subject to copyright.
Superfluous Neuroscience Information Makes
Explanations of Psychological Phenomena
More Appealing
Diego Fernandez-Duque
1
, Jessica Evans
1
, Colton Christian
2
, and Sara D. Hodges
2
Abstract
■Does the presence of irrelevant neuroscience information
make explanations of psychological phenomena more appealing?
Do fMRI pictures further increase that allure? To help answer
these questions, 385 college students in four experiments read
brief descriptions of psychological phenomena, each one accom-
panied by an explanation of varying quality (good vs. circular) and
followed by superfluous information of various types. Ancillary
measures assessed participants’analytical thinking, beliefs on
dualism and free will, and admiration for different sciences. In
Experiment 1, superfluous neuroscience information increased
the judged quality of the argument for both good and bad expla-
nations, whereas accompanying fMRI pictures had no impact
above and beyond the neuroscience text, suggesting a bias that
is conceptual rather than pictorial. Superfluous neuroscience
information was more alluring than social science information
(Experiment 2) and more alluring than information from prestigious
“hard sciences”(Experiments 3 and 4). Analytical thinking did not
protect against the neuroscience bias, nor did a belief in dualism
or free will. We conclude that the “allure of neuroscience”bias is
conceptual, specific to neuroscience, and not easily accounted for
by the prestige of the discipline. It may stem from the lay belief
that the brain is the best explanans for mental phenomena. ■
In 1995, a group of graduate students had just completed
the 8th Cognitive Neuroscience Summer Institute at
University of California, Davis, and thus, it was time to
get one of the commemorative t-shirts. They read “Image
is Everything,”a double entendre linking a hot new
methodology to an older adage about the importance
of strategic self-presentation. Nonetheless, we suspect
that in a broader sense the attendants to that Summer
Institute wore the motto in all earnestness, awed at possi-
bilities of the nascent field of neuroimaging. Twenty years
later, it seems uncontroversial that neuroimaging has led
to a redescription of mental processes in ways that have
often enriched our understanding of the psychology.
When done well, cognitive neuroscience provides addi-
tional explanatory power to the mechanisms that underlie
psychological processes. Alas, cognitive neuroscience—
like any science—is not always reported well. Sometimes,
superfluous information is added that does not provide
additional insight. In this study, we ask whether such super-
fluous neuroscience information increases the perceived
quality of psychological explanations and begin to explore
the possible mechanisms underlying this effect.
One possible reason why some neuroscience infor-
mation may bias judgments is perceptual: The ease of pro-
cessing of brain pictures might make the argument seem
more compelling ( perceptual processing hypothesis).
Consistent with this view, perceptual qualities of brain pic-
tures, such as their three dimensionality, tend to increase
the perceived scientific quality of the accompanying text
(Keehner, Mayberry, & Fischer, 2011; Reber & Schwarz,
1999). Another possible reason for the neuroscience bias
is the prestige of the “hard”sciences ( prestige of science
hypothesis). People believe that biological explanations
are more complex and more scientific than psychologi-
cal explanations. This bias toward the natural sciences
emerges as early as kindergarten, and vestiges of it can
be observed in adulthood (Keil, Lockhart, & Schlegel,
2010). Thus, explanations that invoke neuroscience
may be viewed as reflecting greater expertise. In a re-
lated vein, the use of jargon in describing neuroscience in-
formation might also cue expertise. For example, the mere
presence of a nonsense math equation increases the per-
ceived quality of a scientific abstract (Eriksson, 2012). Finally
and most intriguingly, superfluous neuroscience informa-
tion might increase the perceived scientific quality of expla-
nations if people’s lay theories of the mind embrace the
idea that the brain is the best explanans of mental phenom-
ena (i.e., a brain-as-engine-of-mind hypothesis). If so,
superfluous explanations should fool participants into
seeing the explanations as informative, but giving the
superfluous information a “neuro”flavor would be essential;
this hypothesis predicts that other jargon or scientific cues
would not work as effectively.
1
Villanova University,
2
University of Oregon
© 2015 Massachusetts Institute of Technology Journal of Cognitive Neuroscience 27:5, pp. 926–944
doi:10.1162/jocn_a_00750
Past research has explored whetherthepresenceofabrain
picture increases the perceived quality of a neuroscience
explanation. Although initial findings supported the hy-
pothesis, more recent studies have failed to replicate those
findings (Hook & Farah, 2013; Michael, Newman, Vuorre,
Cumming, & Garry, 2013; Schweitzer, Baker, & Risko,
2013; Gruber & Dickerson, 2012). In 2008, McCabe and
Castel gave their participants a one-page summary of a
cognitive neuroscience finding written for the popular
press. This baseline condition was compared with experi-
mental conditions in which the same neuroscience in-
formation was accompanied by either an fMRI or a bar
chart. Participants in this study rated the scientific reason-
ing most highly when the neuroscience explanation was
paired with the fMRI, leading the authors to conclude that
brain images conferred credibility to the neuroscience
explanations. However, more recent studies using a similar
methodology have failed to replicate these findings. In
2013, Hook and Farah showed their participants short
descriptions of fictional neuroscientific research (e.g.,
“when viewing images of food, obese participants had re-
duced activation in brain areas related to self-control”).
These vignettes were paired with either a brain picture
or a bar chart; the presence of a brain picture did little
to modify the perceived quality of a neuroscientific explana-
tion. That same year, a comprehensive study by Michael
and collaborators, comprising 10 experiments and almost
2000 participants, reached the same conclusion.
As these recent studies clearly show, the addition of a
brain picture does little to increase the perceived quality
of a neuroscientific explanation (Hook & Farah, 2013;
Michael et al., 2013; Schweitzer et al., 2013; Gruber &
Dickerson, 2012). However, none of those studies have
tested whether superfluous neuroscience information
(either pictorial or textual) has an influence on an other-
wise non-neuroscientific explanation. This omission leaves
open the possibility that neuroscience may exert an undue
influence in judgments of research quality. Furthermore,
this omission is theoretically important because it raises
the possibility that the effect would be driven not by the
perceptual properties of brain images but rather by con-
ceptual properties of neuroscience, such as its status as a
hard science and/or its role as the “engine of the mind”in
current lay theories of the mind. In other words, it remains
a possibility that neuroscience would influence people’s
judgment in profound ways.
Does the glow of neuroscience seem to illuminate the
explanation of psychological phenomena? Addressing
this question requires a design comparing the impact
of superfluous neuroscience explanations to the impact
of non-neuroscience explanations. To our knowledge,
there is only one study that addresses this question. In
2008, Weisberg and collaborators asked participants
to read vignettes about well-established psychological
phenomena and their possible explanations ( Weisberg,
Keil, Goodstein, Rawson, & Gray, 2008). One vignette,
for example, described the “curse of knowledge,”the
phenomenon in which participants who do know a fact
report that an inaccurately large percentage of others
would know it too. Although the description of the phe-
nomena was always accurate, the quality of the explana-
tion was variable: sometimes the arguments were good,
and other times they were circular. For the curse of knowl-
edge, the circular statement was “The researchers claim
that this ‘curse’happens because subjects make more
mistakes when they have to judge the knowledge of
others.”
1
A second factor provided the critical manipula-
tion: Half of the vignettes included superfluous neuro-
science sentences, whereas the other half did not. For
the curse of knowledge, the superfluous information
made the circular explanation read as follows: “Brain
scans indicate that this ‘curse’happens because of the
frontal lobe brain circuitry known to be involved in self-
knowledge. Subjects make more mistakes when they
have to judge the knowledge of others.”Superfluous sen-
tences proved effective, increasing the perceived quality of
circular explanations.
Weisberg et al.’s work was novel and insightful, but the
findings raised a number of new questions. Conceptually,
the study did not address whether the effect was specific
to the addition of neuroscience information or if the
addition of any science information would work. The
neuroscience condition did not display any brain images
(it was solely words) and thus did not address whether
pictorial information, such as brain images, had an effect
above and beyond that of written explanations. In addition,
there were some methodological limitations to the study.
Because the neuroscience information condition was
compared with a condition devoid of any explanation,
length of explanation was a possible confounding reason
why the neuroscience information condition boosted
ratings (i.e., a length of explanation account). To further
complicate matters, the neuroscience information was
inserted in the middle of the vignette, interrupting the
flow of reading. Irrespective of content, such an inter-
ruption could have obscured the circularity of the argument.
Furthermore, the superfluous information modified the
sentence structure, possibly masking its circularity. In
the absence of superfluous information, the circularity of
the argument was obvious: “This curse happens because
subjects make more mistakes when they have to judge
the knowledge of others.”In other words, the subordinate
clause was a restatement of the phenomenon, making it
very easy to detect the circularity. When superfluous neuro-
science information was included, it read “this curse hap-
pens because of the frontal lobe brain circuitry known to
be involved in self-knowledge. Subjects make more mis-
takes when they have to judge the knowledge of others.”
In this condition, the restatement of the phenomenon
appeared not as a subordinate clause but as a separate
sentence. As a consequence, the circularity might not have
been so easy to detect. This could explain the “beneficial”
effect of superfluous information without the need for a
rich interpretation based on neuroscience. Consistent with
Fernandez-Duque et al. 927
this account, superfluous information was beneficial only for
circular explanations and not for good explanations. These
limitations notwithstanding, Weisberg et al.’s work has often
been cited along McCabe and Castel (2008) as evidence sup-
porting the distorting influence of neuroscience, and after
failures to replicate McCabe and Castel’s initial findings, it
currently stands as the single piece of evidence in support
of the allure of neuroscience.
In four experiments, we addressed the methodological
limitations of Weisberg et al.’s work and began to inves-
tigatethepossiblemechanismsbywhichtheallureof
neuroscience may operate. To facilitate comprehension
and the flow of reading, we used streamlined vignettes and
placed the superfluous explanation at the very end. To
assess the perceptual fluency hypothesis, we added
a condition in which brain images supplemented the neuro-
science information (Experiment 1). To rule out a length of
explanation hypothesis, we included a condition with
superfluous social science information (Experiment 2); this
also allowed us to test whether the bias was limited to
neuroscience or instead extended to the social sciences.
Finally, we added a condition in which the information was
devoid of any neural references but still contained infor-
mation coming from the “hard”sciences (Experiment 3).
This allowed us to contrast the prestige of science hypothe-
sis, which predicted that the effect would generalize to any
natural science pseudo explanation, against the brain-as-
engine-of-mind hypothesis, which predicted that the
effectiveness of superfluous information would be specific
to neuroscience information.
We also used a variety of ancillary measures to test
several additional hypotheses. To assess whether general
reasoning skills protect against the neuroscience bias, we
included three measures of analytical thinking (the Cog-
nitive Reflection Task [CRT], a set of Concrete Syllogisms,
and a set of Abstract Syllogisms). To assess possible mod-
ulations of the effect by lay theories of mind, we included
surveys on dualism, free will, determinism, and religious
beliefs. Finally, to assess thepossibleroleofscientific
prestige in the neuroscience bias, we collected infor-
mation about participants’attitudes toward neuroscience
and other natural sciences and toward psychology and
other social sciences.
EXPERIMENT 1: NEUROSCIENCE VERSUS
NEUROSCIENCE WITH BRAIN IMAGES
In Experiment 1, we aimed to replicate the original find-
ings of Weisberg et al.’s work, showing that superfluous
neuroscience information increases the perceived quality
of psychological explanations. More importantly, we also
tested whether including brain pictures had an influence
above and beyond that of written text referencing neuro-
science (i.e., perceptual fluency hypothesis). Finally, we
explored whether participants’analytical skills were
negatively correlated with the bias, consistent with a pro-
cessing limitation rather than a conceptual limitation.
Methods
Disclosure of Research Conduct
We report all measures collected, all data exclusions (if
any), and all manipulations, in all the studies conducted
for this project.
Participants
A total of 91 Villanova undergraduates participated in
the task. Sixty-four of them were students in an introduc-
tory psychology course and participated for course credit
(32 women, 27 men, and 5 who did not report gender).
Another 27 participants were students in a physiological
psychology course who participated as part of a class-
room activity (82% of students in the class were women).
The introductory psychology course serves students who
have diverse academic interests and majors, including
business (approximately 25%), science (25%), nursing
(15%), social studies (8%), and psychology (6%). The phys-
iological psychology course serves mostly students major-
ing in psychology (>90%).
Stimuli
General instructions. The following instructions were
presented at the beginning of a paper questionnaire
distributed to participants:
You will read descriptions of various scientific
findings. All the findings come from solid, replicable
research; they are the kind of material you would
encounter in an introductory psychology textbook.
You will also read an explanation of each finding.
Unlike the findings themselves, the explanations
of the findings range in quality: Some explanations
are better than others, they are more logically
sound. Your job is to judge the quality of such
explanations, which could range from a very poor
explanation (−3) to a very good explanation (+3).
Vignettes. There were 18 vignettes adapted from
Weisberg et al.’s work. Each vignette described a unique
research topic and was displayed on a separate page with
the topic title at the top (e.g., “1. Curse of Knowledge”)
and the instruction reminder at the bottom. The relevant
information was broken down into “method,”“findings,”
and “explanation.”The explanations presented arguments
of varying quality. The good quality argument was, in most
cases, the genuine explanation that researchers typically
give for each phenomenon. For the curse of knowledge,
it stated that it “happens because subjects have trouble
switching their point of view to consider what someone
else might know, mistakenly projecting their own knowl-
edge onto others.”The bad quality argument was a cir-
cular restatement of the phenomenon, and therefore, it
928 Journal of Cognitive Neuroscience Volume 27, Number 5
was nonexplanatory. For the curse of knowledge, it stated
that it “happens because subjects make more mis-
takes when they have to judge the knowledge of others.
People are much better at judging what they themselves
know.”
Following the argument, superfluous neuroscience
information was added for the neuroscience condition.
For the curse of knowledge vignette, it stated that “stud-
ies indicate that self-knowledge engages a neural circuitry
in the frontal lobes of the brain.”The superfluous in-
formation always appeared after the argument. The infor-
mation was similar to the information used by Weisberg
et al., albeit slightly modified so that the information
could be presented as a stand-alone sentence at the end
of the argument. In the neuroscience +brain image con-
dition, a brain image and a figure caption were displayed
in addition to, and just below, the explanation. (For a
complete example, see Appendix A; for a complete list of
superfluous neuroscience information, see Appendix B.)
Images. Brain images were selected by the first author
from cognitive neuroscience journal articles. The informa-
tion depicted in the image was always consistent with
the information offered in the superfluous neuroscience
explanation; for example, if the superfluous information
referred to activation in the right pFC, the selected brain
image similarly depicted activation in the right pFC. Selec-
tion was limited to high-quality, high-contrast images.
These images included a variety of formats and orienta-
tions. Six images depicted the lateral view of an “inflated”
brain (i.e., with visible sulci pushed outward), a seventh
image depicted a lateral view of a “standard”brain (i.e.,
not inflated), and an eighth image depicted a ventral view.
Nine other images were 2-D cross-sections of the brain
(two showed a coronal cut, two showed a sagittal cut, and
five others showed more than one cut). Finally, one vignette
was paired up with a 3-D rendition of a head with a horizon-
tal and a sagittal plane exposed. The image sizes ranged
from 4.3 × 4.1 cm to 6.6 × 9.1 cm and were printed on
white paper by a high-definition color printer. Areas of brain
activation were always clearly visible; in some cases, they
were further highlighted by a cross hair or an arrow.
Image captions. For each brain image, we included a
caption that accurately described the image but provided
no additional information. For example, for the vignette with
the neuroscience explanation “As we get older, the memory
centers in the medial-temporal lobe change in specific ways
(see Figure below),”the image caption was “Age-related
brain changes, including medial-temporal lobe (circled).”
Design
Two factors were crossed in a 2 × 3 within-subject facto-
rial design: Quality of argument (good; circular) and Type
of superfluous information (no information; neuro-
science; neuroscience + brain image). Crossing these
two factors created six within-subject conditions. Every
vignette was doctored so that it could fit in any of the six
conditions. Participants read three vignettes in each of
the six conditions, for a total of 18 vignettes. Participants
saw one of six different forms of the 18 vignettes, with
each form created by changing subsets of three vignettes
from one condition to another. That is, vignettes moved
in groups of three, such that three vignettes that were
in the same condition on one form (e.g., Vignettes 2,
8, 18) would be in a different condition on one of the
other forms (still in Positions 2, 8, 18). For example, if
a participant receiving the first form saw a group of
vignettes in the “good quality, neuroscience informa-
tion”condition, then a participant receiving the second
form would see that same group of vignettes in the “bad
quality, no information”condition. Across the six forms,
each vignette appeared in each condition once. The
vignettes were shown in fixed order, with condition order
varying across forms.
Ancillary Measures
We asked participants to complete three individual dif-
ference measures of analytical thinking: the CRT, a set
of Concrete Syllogisms, and a set of Abstract Syllogisms.
These measures could be used to see whether susceptibil-
ity to being biased by neuroscience information is more
likely in those who are less able to analyze quality of
arguments and thus may rely more heavily on superficial
argument cues.
All but one participant completed the CRT, a three-
item task that measures one’s tendency to respond
with deliberate responses rather than intuitive ones
(Frederick, 2005). An example of one of the items from
this task is: “A bat and a ball cost $1.10. The bat costs
$1.00morethantheball.Howmuchdoestheball
cost?”The intuitive response (10 cents) is incorrect,
and arriving at the correct answer (5 cents) requires
deliberate thought. The total number of correct answers
for the three items (0, 1, 2, or 3) serves as the score for
this task.
Most participants were also asked to assess the logic
of 16 syllogistic arguments adapted from a study by
Markovits and Nantel (1989). Each syllogistic argument
took the form of three statements, and participants were
asked to answer whether or not the third statement
could be concluded based on the information in the
first two statements. Participants first assessed eight con-
crete syllogistic arguments—that is, the arguments used
familiar categories of items, such as:
Premise 1: All things that are smoked are good for the
health.
Premise 2: Cigarettes are smoked.
Conclusion: Cigarettes are good for the health.
Fernandez-Duque et al. 929
Participants then completed eight abstract syllogistic
arguments, which were made abstract by using made-
up words for categories, such as:
Premise 1: All animals love water.
Premise 2: Selacians do not like water.
Conclusion: Selacians are not animals.
The concrete and abstract syllogistic arguments were
identical in structure, differing only in terms of the content
words. However, for all of the concrete syllogisms, when
the syllogistic arguments were valid, the logical conclusion
was inconsistent with common knowledge (e.g., smoking
example). When the syllogistic arguments were invalid, the
(illogical) conclusion was consistent with common
knowledge (e.g., Premise 1: All unemployed people are
poor; Premise 2: Bill Gates is not unemployed; Conclusion:
Bill Gates is not poor). The syllogisms were distributed as a
single packet. A total of 66 participants completed both
versions of the tasks. Because of a researcher’smistake,the
other 25 participants completed only one version of the task
(13 completed the concrete version, and 12 answered the
abstract version).
Results
The approach to the data analysis was similar in all four
experiments. Data for argument quality ratings were re-
coded to go from 1 (very poor)to7(very good ). Next,
for each participant, responses to the three vignettes per
condition were averaged to obtain mean scores in each
of the six conditions. Those mean scores were entered
in a preliminary analysis to assess whether the course
in which participants were enrolled (introductory or
physiological psychology) interacted with the factors of
interest. Because there were no interactions between
course and our variables of interest (in Study 1 or any
of the other experiments), course was not included as
a factor in reported analyses.
Data were submitted to a 2 × 3 within-subject ANOVA
with Quality of argument (circular; good) and Type of
superfluous information (no information; neuroscience;
neuroscience + brain image) as factors. Significant main
effects were followed up by simple contrasts.
Figure 1 depicts the main findings of Experiment 1. Not
surprisingly, good arguments were deemed more com-
pelling than circular arguments, increasing the quality
judgment by 0.70 of a point, F(1, 90) = 43.80, p<.001,
η
p
2
= .33. More importantly, there was a main effect of
Information type, F(2, 180) = 37.67, p< .001, η
p
2
=.30.
Information type and Argument quality did not interact,
F(2, 180) = 0.69, p= .5. A simple contrast showed that
superfluous neuroscience information made the argu-
ment more compelling than no information, increasing
the quality judgment by 0.88 of a point, F(1, 90) = 58.69,
p< .001, η
p
2
= .395. The inclusion of brain pictures did
not provide any additional benefit, increasing the quality
judgment of the neuroscience explanation by a negligible
amount (.05 of a point), F(1, 90) = 0.21, p=.64.
Ancillary Measures
Table 1 shows that, as expected, all three measures of
analytical thinking (CRT, Concrete Syllogisms, and Abstract
Syllogisms) were positively correlated. The influence of
neuroscience information was calculated as the difference
score between the neuroscience condition and the no
information condition. The influence of brain images
was calculated as the difference score between the neuro-
science +brain image conditions and the neuroscience
condition. Neither of these two scores showed a negative
correlation with any of the measures of analytical thinking.
Thus, we found no evidence that increased analytical mind-
set protected participants from the neuroscience bias. This
null effect should be interpreted cautiously, as we also
failed to find a correlation between the ability to discrimi-
nate between good and circular arguments and any of the
measures of analytical thinking.
Discussion
Superfluous neuroscience information increased the
judged quality of the scientific argument, thus replicating
Figure 1. Top: Judged quality of arguments for Experiment 1, plotted with
95% within-subject confidence interval (Masson & Loftus, 2003). Bottom:
Effect sizes of the neuroscience information against no information
contrast and neuroscience + brain image against no information contrast.
930 Journal of Cognitive Neuroscience Volume 27, Number 5
the main finding of Weisberg et al. (2008) with a more
rigorous methodology in which the superfluous informa-
tion did not interrupt the flow of the argument. The only
substantial difference was that Weisberg et al. found the
effect only for circular explanations, although we found it
for good explanations, too. This finding suggests that the
addition of neuroscience information does not just make
bad explanations better; even if a good explanation—one
that is not lacking in explanatory power—is provided, the
effect still persists.
Importantly, brain images did not have an impact
above and beyond the written text that mentioned neuro-
science. These results suggest that the effect of neuro-
science on judgments of argument quality cannot be
attributed to superficial properties of the stimulus, such
as the perceptual quality of the picture. Furthermore, we
did not find any evidence that more analytical aptitude
protected participants from the bias. Thus, it seems more
likely that the effect of neuroscience is conceptual, driven
by lay theories on the role of neuroscience, or possibly,
by a reverence for the natural sciences in general. Alter-
natively, it is possible that the effect is simply an artifact
of neuroscience explanation being longer than explana-
tions with no superfluous information. We address these
possibilities in the experiments described below.
EXPERIMENT 2: NEUROSCIENCE VERSUS
SOCIAL SCIENCES
In Experiment 2, we replaced the brain image condition
of Experiment 1 with a condition that contained super-
fluous social science information. This allowed an assess-
ment of the length of explanation hypothesis, which
predicts that the effect would be absent when comparing
the neuroscience condition to a condition of similar
length. The comparison against the social science condi-
tion also allowed us to test the specificity of the effect: Is
the benefit of superfluous information specific to neuro-
science, or would it generalize to social science infor-
mation? We predicted that the effect would be stronger
for neuroscience, possibly because of the higher perceived
status of the natural sciences. To explore this inter-
pretation, we also collected information on participants’
attitudes toward the natural and social sciences.
Methods
Participants
A total of 90 Villanova undergraduates participated in the
study as to match the number of participants in Experi-
ment 1. Sixty-four of them were students (both psychology
majors and non-majors) in an introductory psychology
course and participated for course credit (33 women,
31 men). Another 26 participants (22 women) were
students in a physiological psychology course required
for psychology majors—they participated as part of a
classroom activity.
Stimuli
In Experiment 2, participants were exposed to three
different types of superfluous information (neuroscience;
social science; and none). For example, for the vignette
describing mental rotation, the neuroscience information
read: “Brain scans of these subjects show activation of a
neural circuit in parietal and occipital lobes.”The social
information for that same vignette read: “Cross-cultural
research shows that perception of complex objects is
influenced by the amount of industrialization in the
society.”The neuroscience information was largely the
same as in Experiment 1, although we did introduce
some slight changes in the wording, to ensure that no
sentence would carry any explanatory power. For exam-
ple, the attentional blink sentence was shortened to
“areas in the frontal lobe were active in response to the
stimuli”instead of “areas in the frontal lobe previously
showntomediateattentionfunctionedinresponseto
the stimuli.”Other changes included the choice of more
obvious anatomical correlates (e.g., right parietal lobe
instead of premotor cortex for the vignette on spatial
reasoning), the elimination of jargon (e.g., deleting “CA3”
from the sentence “CA3 brain cells of the hippocampus”),
and some additional streamlining (see Appendix B).
Finally, the formatting of Experiment 2 was different from
Experiment 1 in that the information for all three parts of
Table 1. Correlations between Measures of Analytical Thinking and the Allure of Neuroscience Bias in Experiment 1
Measure 1 2 3 4 5 6 M SD
1. CRT –.48 .48 .10 .12 .05 1.24 1.09
2. Syllogisms (concrete) –.78 .09 .08 .02 6.07 1.82
3. Syllogisms (abstract) –−.08 .11 .11 6.19 1.74
4. Neuroscience vs. No Info –−.20 −.01 ––
5. Neuro + Image vs. Neuroscience –−.14 ––
6. Good vs. Circular –––
p< .05 in bold font.
Fernandez-Duque et al. 931
each vignette (method, findings, explanation) was further
broken down into bullet points to facilitate their reading
and comprehension.
Design
As in Experiment 1, two factors were crossed in a 2 × 3
within-subject factorial design: Quality of argument (good;
circular) and Type of superfluous information (no informa-
tion; neuroscience; social science). We excluded the data
from one question from one form from analysis because
by mistake that question in that form had both the circular
and the good explanations. We also recoded, in one form,
two questions in which the quality of the explanation was
flipped from the intended design. Other aspects of the
design remained the same as in Experiment 1.
Ancillary Measures
We included the same measures of analytic thinking that
were used in Experiment 1. In addition to these measures,
we included an assessment of attitudes toward science
and questionnaires regarding beliefs in dualism, free will,
genetic determinism, and religion. These ancillary mea-
sures were presented in fixed order following the main task.
Attitudes toward science. Participants were asked to use
a 10-point scale to rate nine different disciplines in terms
of their scientific rigor, their societal prestige, and the
knowledge gap between an expert in the field and a
layperson. Besides Neuroscience, Psychology, and Social
Psychology, we included three natural sciences (Biology,
Chemistry, Genetics) and three social sciences (Sociology,
Cultural Anthropology, Political Science). (For exact word-
ing, see Appendix C.) Next, participants read contrasting
definitions of Behavioral Neuroscience (the study of the
brain and its contributions to thinking and behavior) and
Cultural Psychology (the study of culture and its contribu-
tion to thinking and behavior) and then matched dif-
ferent personality traits to the practitioners the trait best
described, either behavioral neuroscientists or cultural
psychologists. There were eight competence traits (deter-
mined, persistent, scientific, serious, skillful, intelligent;
creative, imaginative) and four warmth traits (sincere,
social, tolerant, warm) adapted from the work of Fiske
et al. (Fiske, Cuddy, & Glick, 2007; Rosenberg, Nelson, &
Vivekananthan, 1968).
Questionnaires on determinism, free will, dualism, and
religious beliefs. A 24-item questionnaire probed par-
ticipants’beliefs about genetic determinism, scientific
determinism, free will, mind/brain dualism, and religion
(for exact wording, see Appendix D).
2
We found no corre-
lations between these measures and neuroscience allure,
which we operationalized for these analyses as the in-
crease in argument quality brought about by superfluous
neuroscience information (vs. superfluous social informa-
tion). The same was true for the subsequent experiments.
We do not discuss these questionnaires any further.
Results
The data were submitted to a 2 × 3 ANOVA with Quality of
argument (circular; good) and Type of superfluous
information (no information; neuroscience; social science)
as within-subject factors. Significant main effects were
followed up by simple contrasts.
Figure2depictsthemainfindingsofExperiment2.
Once again and not surprisingly, good arguments were
deemed more compelling than circular ones, in this study
by 0.80 of a point, F(1, 89) = 45.47, MSE =1.91,p<.001,
η
p
2
= .34. Consistent with our hypothesis, there was a main
effect of Information type, F(2, 178) = 8.08, p<.001,
MSE = 1.23, η
p
2
= .09. Information type and Argument
quality did not interact, F(2, 178) = 0.32, p= .73.
A simple contrast analysis showed that superfluous neuro-
science information made the arguments more compelling
than no information, increasing the quality of arguments
by 0.48 of a point, F(1, 89) = 13.54, MSE =2.8,p< .001,
η
p
2
= .13. This result replicates the finding of Experiment 1.
Figure 2. Top: Judged quality of arguments for Experiment 2, plotted
with 95% within-subject confidence interval. Bottom: Effect sizes of the
neuroscience information against no information contrast and social
science against no information contrast.
932 Journal of Cognitive Neuroscience Volume 27, Number 5
More importantly, superfluous neuroscience information
was also more compelling than social science information,
increasing the quality of arguments by 0.31 of a point,
F(1, 89) = 8.6, MSE =1.1,p< .004, η
p
2
=.08.Finally,argu-
ments adorned with superfluous social science information
were deemed only slightly better than those devoid of it,
increasing the quality of arguments by 0.18 of a point, an
effect that failed to reach statistical significance, F(1, 89) =
2.5, MSE =1.14,p=.11,η
p
2
= .027.
Ancillary Measures
Analytical thinking. Table 2 shows that all three mea-
sures of analytical thinking (CRT, Concrete Syllogisms,
and Abstract Syllogisms) were positively correlated, as in
Experiment 1. Table 2 also shows that all three measures
of analytical thinking were positively correlated with the
ability to discriminate between good and circular explana-
tions (rs > .23, ps < .01). The influence of neuroscience
information was computed in two different ways. One
was the difference score between the neuroscience con-
dition and the no information condition, and the other
was the difference score between the neuroscience con-
dition and the social science condition. Crossing the
three measures of analytical thinking with the two mea-
sures of neuroscience influence allowed us to test six
possible correlations; only one of them was statistically
significant and even then it was a weak correlation at
r=−.24. Thus, we found little evidence to support the
hypothesis that increased analytical mindset protected
participants from the neuroscience bias.
Attitudes toward science. For each of the nine disci-
plines, participants used a 10-point scale (1–10) to make
three related judgments: one on scientific rigor, another
on the knowledge gap between experts and laypeople,
and a third one on the societal prestige of the discipline.
The three judgments were averaged to compute an Ad-
miration Score (see Figure 3). Admiration scores for
Biology, Chemistry, and Genetics were averaged into a
Natural Sciences mean score, and admiration scores for
Sociology, Anthropology, and Political Science were aver-
aged into a Social Sciences score. Admiration scores for
neuroscience, psychology, and social psychology were
kept separate. Pairwise comparisons with Bonferroni
adjusted alpha levels of .0025 per test (.05/20) showed
significant differences in all the contrasts. Neuroscience
was the most admired discipline, the natural sciences were
more admired than the social sciences, and psychology and
social psychology lay in the middle.
Trait ratings for the sciences. Participants rated the
sciences on eight traits that have been linked to com-
petence in previous work (determined, persistent, scientific,
serious, skillful, intelligent; creative, imaginative) and four
traits that have been linked to warmth (sincere, social, toler-
ant, warm). Each trait was attributed in a two-alternative
forced-choice to either Behavioral Neuroscientists or Cul-
tural Psychologists. Behavioral Neuroscientists were
deemed more competent than Cultural Psychologists, re-
ceiving endorsements of competence traits 69.7% of the
time (95% CI [66.5, 72.9]), one-sample ttest against chance,
t(89) = 12.39, p< .001. The percentage rose to 78.6% when
the least prototypical items (creative, imaginative) were
filtered out. Cultural Psychologists were deemed higher
in warmth than Behavioral Neuroscientists (82.7%), 95% CI
[77.9, 87.5], t(89) = 13.47, p< .001.
Discussion
Experiment 2 showed that there was an effect of super-
fluous information that was specific to neuroscience,
above and beyond any bias caused by the social sciences.
Table 2. Correlations between Measures of Analytical Thinking and the Allure of Neuroscience Bias in Experiment 2
Measure 12 3 4 5 6 M SD
1. CRT –.32 .27 .04 −.04 .23 1.31 1.05
2. Syllogisms (concrete) –.83 −.11 −.14 .27 6.17 1.81
3. Syllogisms (abstract) –−.24 −.17 .29 6.37 1.68
4. Neuroscience vs. No Info –.57 −.02 ––
5. Neuroscience vs. Social –−.12 ––
6. Good vs. Circular –––
p< .05 in bold font.
Figure 3. Participants’attitudes toward different sciences in
Experiment 2, plotted with 95% within-subject confidence interval.
Fernandez-Duque et al. 933
Thus, the neuroscience effect is not just because of longer
explanations, as the neuroscience and the social science
information were comparable in length. As in Experiment 1,
the neuroscience effect was present for both circular
and good explanations. Although, as predicted, analytical
thinking helped participants to distinguish between good
and bad explanations, analytical thinking did little to help
participants to avoid the neuroscience bias. Thus, as in
Experiment 1, we found no support for a performance
explanation of the neuroscience bias. Instead, the bias
seems more likely to be conceptual, driven by lay theories
about the role of neuroscience, or possibly, by a reverence
for the natural sciences in general. Consistent with the
prestige of science hypothesis, neuroscience and the
other natural sciences were held in higher regard than
the social sciences, and neuroscience was slightly more
admired than the other natural sciences. However, it is
also possible that scientific prestige was a coincidental
feature, rather than the driving force behind the effect, a
possibility we test in Experiment 3.
EXPERIMENT 3: NEUROSCIENCE VERSUS
HARD SCIENCES
If the effect of superfluous neuroscience information
depends on the scientific prestige of neuroscience, the ef-
fect should generalize to other “hard”sciences. In contrast,
a failure to generalize to other “hard”sciences would be
consistent with a brain-as-engine-of-mind hypothesis.
According to the brain-as-engine-of-mind hypothesis,
the brain is assigned a privileged role in folk theory of
mind. Such a privileged role would make neuroscience—
but not other natural sciences—the most alluring ex-
planation of psychological phenomena. Thus, in Ex-
periment 3, we tested the competing predictions of the
prestige of science hypothesis and the brain-as-engine-
of-mind hypothesis by adding a condition with super-
fluous information from the hard sciences that was devoid
of neuroscience information. Those non-neuroscience
explanations used jargon from a variety of fields, including
genetics, math, computer science, physics, biology, and
biochemistry.
Methods
Participants
A total of 96 Villanova undergraduates participated in the
task as to approximately match the number of partici-
pants in Experiments 1 and 2. Sixty-nine of these partic-
ipants were students in an introduction to psychology
course and participated for course credit (47 women,
22 men). Another 27 participants (22 women, 5 men) were
students in a physiological psychology course required
for psychology majors. They participated as part of a
classroom activity. The average age was 19.8 years (SD =
1.9, range = 18–32).
Stimuli
In Study 3, the no information condition of Study 2 was
replaced with a hard science condition that provided
superfluous information from the natural sciences. We
drafted explanations that included information from a di-
verse set of disciplines, including genetics (n= 6); biology
and biochemistry (n= 6); and math, physics, and com-
puter science (n= 6). Thus, there were three different
types of superfluous information (hard science; neuro-
science; and social science). For example, in a vignette
explaining the phenomenon of “confirmatory bias,”the
hard science explanation read: “The response to emotional
messages is modulated by the gene expression of the glu-
cocorticoid receptor involved in the physiology of stress.”
For the same vignette, the neuroscience explanation
read, “A brain structure called the amygdala known to be
involved in emotional processing is activated by this task.”
For the social explanation, the vignette read, “Social fac-
tors, such as the number of people present in the room
have an influence in the magnitude of the effect.”(See
Appendix B for complete list of explanations.) Finally, as
in Experiment 2, participants completed assessments of
analytical thinking (CRT, syllogisms) and attitudes toward
science, as well as a 24-item questionnaire on determinism,
free will, dualism, and religious beliefs.
Results
The data were submitted to a 2 × 3 ANOVA with Quality
of argument (circular; good) and Type of superfluous
information (hard science; neuroscience; social science)
as within-subject factors. Significant main effects were
followed up by simple contrasts.
Figure 4 depicts the main findings of Experiment 3. As
in the two previous experiments, good arguments were
deemed more compelling than circular ones, increasing
the quality of arguments by 0.79 of a point, F(1, 95) =
82.81, MSE =1.09,p< .001, η
p
2
= .47. Also as in the pre-
vious experiments, there was a main effect of Information
type, F(2, 190) 12.47, p< .001, MSE = 1.09, η
p
2
= .12.
Once again, Information type and Argument quality did
not interact, F(2, 190) = 0.03, p=.97.
A simple contrast analysis showed that superfluous
neuroscience information made the arguments more
compelling than hard science information, increasing
the quality of arguments by 0.33 of a point, F(1, 95) =
8.80, MSE =1.15,p=.004,η
p
2
= .08. Superfluous neuro-
science information was also more compelling than social
science information, increasing the quality of arguments
by 0.53 of a point, F(1, 95) = 20.77, MSE =1.3,p=
.001, η
p
2
= .18. Arguments adorned with hard science
information were deemed somewhat better than those
adorned with social science information, with hard sci-
ence information increasing the quality of arguments
by 0.20 of a point, F(1, 89) = 4.75, MSE =0.83,p= .03,
η
p
2
=.05.
934 Journal of Cognitive Neuroscience Volume 27, Number 5
Ancillary Measures
Analytical thinking. Table 3 shows that all three mea-
sures of analytical thinking (CRT, Concrete Syllogisms,
and Abstract Syllogisms) were positively intercorrelated,
as in Experiments 1 and 2. Furthermore, two of the three
measures of analytical thinking were positively correlated
with the ability to discriminate between good and circular
explanations. The influence of neuroscience information
was computed in two different ways. One was the dif-
ference score between the neuroscience condition and
the hard science condition, and the other was the differ-
ence score between the neuroscience condition and the
social science condition. Crossing the three measures of
analytical thinking with the two measures of neuroscience
influence allowed us to test six possible correlations: They
were all uncorrelated (rs between −.045 and .086). In
other words, we found no support for the hypothesis that
increased analytical mindset protected participants from
the neuroscience bias.
Attitudes toward science and trait ratings for the sciences.
As in Experiment 2, neuroscience was the most admired
discipline, the natural sciences were admired more than
the social sciences, with psychology and social psychology
lying in the middle (see Figure 5); pairwise comparisons
with Bonferroni-adjusted alpha levels of .0025 per test
(.05/20) showed significant differences in all the contrasts.
Behavioral Neuroscientists were deemed more compe-
tent than Cultural Psychologists, receiving endorsements
of competence traits 70.0% of the time (95% CI [68%,
72%]), one-sample ttest against chance, t(95) = 19.95,
p< .001. The percentage rose to 81.3% (95% CI [77.3%,
85.3%]) when the least prototypical competence traits
(creative, imaginative) were filtered out. Cultural Psycholo-
gists were deemed higher in warmth, being attributed
warmth traits 85.9% of the time (95% CI [82.1%, 89.7%]),
t(95) = 18.74, p<.001.
Table 3. Correlations between Measures of Analytical Thinking and the Allure of Neuroscience Bias in Experiment 3
Measure 12 3 4 5 6 M SD
1. CRT –.36 .34 .09 −.01 .22 1.44 1.04
2. Syllogisms (concrete) –.80 −.002 .05 .20 6.32 1.74
3. Syllogisms (abstract) –−.05 .06 .14 6.36 1.61
4. Neuroscience vs. Hard Science –.66 −.04 ––
5. Neuroscience vs. Social –−.08 ––
6. Good vs. Circular –––
p< .05 in bold font.
Figure 4. Top: Judged quality of arguments for Experiment 3, plotted
with 95% within-subject confidence interval. Bottom: Effect sizes
of the contrast of neuroscience information against social science
information and the contrast of hard science against social science.
Figure 5. Participants’attitudes toward different sciences in
Experiment 3, plotted with 95% within-subject confidence interval.
Fernandez-Duque et al. 935
Discussion
Superfluous neuroscience information more effectively
raised perceived explanation quality than superfluous
information from the other hard sciences. This was the
case for both circular and good explanations and occurred
despite the relative similarity in these disciplines in terms
of their perceived rigor and prestige. Indeed, in the
absence of neuroscience jargon, there was only a modest
effect of hard sciences information relative to social sci-
ences information. This pattern of results argues against
the prestige of science hypothesis and favors instead the
brain-as-engine-of-mind hypothesis.
EXPERIMENT 4: A REPLICATION OF
NEUROSCIENCE VERSUS HARD SCIENCES
In Experiment 4, we aimed to replicate the findings of
Experiment 3 using a different site (a large public univer-
sity on the west coast, as opposed to a smaller private
universityinthenortheast)andadifferentmethodof
data collection (online and remotely accessed instead of
paper-and-pencil completed on site). On the basis of the
results of Experiment 3, we predicted that superfluous
information would be most beneficial in the neuroscience
condition. Less importantly, we also predicted that par-
ticipants’attitudes would be more positive toward neuro-
science and the other natural sciences than toward the
social sciences.
Methods
Participants
A total of 141 university student participants signed up
to participate in an online study in exchange for par-
tial fulfillment of a psychology class requirement. Four
participants withdrew without providing data. Another
29 participants were excluded because they failed an
Instructional Manipulation Check twice in a row. This
check was included to ensure that participants were read-
ing the instructions thoroughly (Oppenheimer, Meyvis, &
Davidenko, 2009). Thus, we were left with 108 partici-
pants (75 women and 31 men). When participants were
asked their major, answers clustered around natural sci-
ences (23%), psychology (23%), business (17%), and other
majors, such as social studies and journalism (26%).
Stimuli and Procedure
Participants who signed up to do the study were presented
with a series of questions via Qualtrics survey software.
After consenting to participation and passing the Instruc-
tional Manipulation Check, participants completed the
main task.
3
The stimulus materials were the same as in
Experiment 3, with the information broken down into
“method,”“findings,”and “explanation,”with information
displayed in bullet-points, as in Experiments 2 and 3.
Unique to Experiment 4, each of the sections and bullet
points was displayed sequentially at a self-paced rate, with
the preceding information remaining in display until the
vignette was complete and a response was made. Also,
the main findings in each vignette were initially displayed
in bold font. After each vignette, a message appeared
reminding participants of the task instructions. All of these
modifications were aimed at facilitating comprehension
and encouraging careful reading of the vignettes. At the
end of the main task, participants completed the same
questionnaires as in Experiment 3 (except for the religio-
sity questions for which we did not have human partici-
pants compliance approval at this site). Two participants
did not provide data on trait ratings for the sciences.
Results and Discussion
Figure 6 depicts the main findings of Experiment 4. Once
again, good arguments were deemed more compelling
than circular ones, increasing the quality of arguments
by 0.54 of a point, F(1, 107) = 37.83, MSE = 0.85, p<
.001, η
p
2
= .26. As in all previous experiments, there was
Figure 6. Top: Judged quality of arguments for Experiment 4, plotted
with 95% within-subject confidence interval. Bottom: Effect sizes of the
contrast of neuroscience information against social science information
and the contrast of hard science information against social science
information.
936 Journal of Cognitive Neuroscience Volume 27, Number 5
a main effect of Information type, F(2, 214) = 4.48, MSE =
0.92, p<.001,η
p
2
= .07. Once again, the two factors did
not interact, F(2, 214) = 0.69, p= .50.
As in Experiment 3, superfluous neuroscience infor-
mation made the arguments more compelling than hard
science information, increasing the quality of arguments
by 0.36 of a point, F(1, 107) = 14.75, MSE = 0.94, p<
.001, η
p
2
= .12.
4
Superfluous neuroscience information
was also more compelling than social information, in-
creasing the quality of arguments by 0.29 of a point,
F(1, 107) = 9.90, MSE =0.90,p= .002, η
p
2
= .085.
The slight benefit of hard science over social science find-
ing observed in Experiment 3 was not replicated in
Experiment 4. If anything, hard science information de-
creased the quality of arguments by a negligible amount
of 0.07 of a point, in favor of the social science informa-
tion, F(1, 107) = 0.61, p= .43, η
p
2
= .006.
Ancillary Measures
Analytical thinking. Table 4 shows that the analytical
thinking measures (CRT, Concrete Syllogisms, and Ab-
stract Syllogisms) were positively correlated among them-
selves. Furthermore, performance in concrete syllogisms,
a measure of analytical thinking, was correlated with the
ability to distinguish between good and circular argu-
ments. In contrast and consistent with Experiments 1–3,
there were no significant negative correlations between
analytical thinking and the allure of neuroscience bias
(rs between −.05 and .10).
Attitudes toward science and trait ratings for the sciences.
The data on participants’attitudes toward different dis-
ciplines replicated the pattern obtained in Experiments 2
and 3, with the only exception being that the difference
between Social Psychology and the other Social Sciences
failed to reach statistical significance. As in previous exper-
iments, Behavioral Neuroscientists were deemed more
competent than Cultural Psychologists, receiving endorse-
ments of competence traits 63.4% of the time, one-way
ttest against chance, t(105) = 7.4, p< .001. The per-
centage rose to 73.0% when the least prototypical items
(creative, imaginative) were filtered out. Cultural Psycholo-
gists were deemed higher in warmth than Behavioral
Neuroscientists (78%), t(104) = 12.2, p< .001 (Figure 7).
A possible alternative explanation in Study 4 is that
explanation quality was affected by sentences providing
the superfluous information in the different domains. In
other words, the hard science explanations might have
been less effective than the neuroscience explanations
simply because the sentences detailing the hard science
superfluous information were viewed as inferior to the
sentences detailing the neuroscience superfluous infor-
mation. To explore this, we asked a new group of students
in a physiological psychology course to judge the perceived
quality of the superfluous information sentences. Par-
ticipants were randomly assigned to judge hard science
(n= 12), neuroscience (n= 13), or social science (n=
14) superfluous sentences, which were displayed in isola-
tion. They were told that for each sentence they had to
“judge its quality; that is, decide whether it is the kind of
sentence that may be published in an academic publica-
tion or college textbook.”The sentences were printed in
a fixed order, without the titles or the other parts of the
vignettes. Participants answered on a 7-point scale (1 =
not at all,7=absolutely). A one-way ANOVA showed a
main effect of Domain, F(2, 36) = 3.55, p= .04, with Tukey
post hoc tests showing that the hard science sentences
were deemed of better quality than the social science sen-
tences (M
hard
= 5.03; M
social
= 4.36). Most importantly,
the comparison between hard science sentences and the
Figure 7. Participants’attitudes toward different sciences in
Experiment 4, plotted with 95% within-subject confidence interval.
Table 4. Correlations between Measures of Analytical Thinking and the Allure of Neuroscience Bias in Experiment 4
Measure 12 3 4 5 6 M SD
1. CRT –.48 .45 .06 .10 .01 .79 1.04
2. Syllogisms (concrete) –.80 −.05 .03 .20 4.06 2.09
3. Syllogisms (abstract) –−.07 .04 .16 4.51 1.70
4. Neuroscience vs. Hard Science –.50 .18 ––
5. Neuroscience vs. Social –.03 ––
6. Good vs. Circular –– –
p< .05 in bold font.
Fernandez-Duque et al. 937
neuroscience sentences failed to reach statistical signifi-
cance, and to the extent that there was a difference, it
was in the direction opposite to the one supporting a qual-
ity confound artifact (M
hard
=5.03;M
neuro
= 4.57). Thus,
we found no evidence to support the contention that the
neuroscience bias was driven by differences in quality of
the superfluous information provided.
GENERAL DISCUSSION
Across four experiments, the presence of irrelevant neuro-
science information made arguments more compelling.
Experiment 1 documented the main finding for circular
explanations as well as for good explanations and showed
that brain images did not have an impact above and
beyond written text. Experiment 2 ruled out length of
explanation as a possible artifact and showed that the
effect was specific to neuroscience, as it did not generalize
to social science information. Experiments 3 and 4 further
documented the specificity of the neuroscience effect,
above and beyond the effect of other “hard”sciences.
The absence of a neuroimage effect, above and beyond
written neuroscience information (Experiment 1), is con-
sistent with recent findings from the literature (Hook &
Farah, 2013; Michael et al., 2013; Schweitzer et al., 2013).
This absence suggests that the allure of neuroscience
information does not stem from processing differences
related to the stimulus’s perceptual quality ( perceptual
processing hypothesis) but rather from a deeper embrace
of neuroscience as an explanatory mechanism. Consistent
with this interpretation, we found little evidence that ana-
lytical thinking aptitude protected against the neuro-
science effect, despite evidence that analytical thinking
sometimes helped participants to distinguish between
good and bad explanations.
Nonetheless, our conclusion that the allure of neuro-
science is conceptual rather than pictorial should be fur-
ther explored in future studies. A promising approach
would be to directly compare a neuroscience image to
a non-neuroscience image. Assuming that perceptual fea-
tures of the stimuli are well matched across conditions,
this comparison could show more conclusively that brain
images have an effect because of the conceptual infor-
mation they provide and not because of their perceptual
properties. Future studies should also explore whether
the effect of the neuroscience text is fully redundant with
the effect of the neuroscience image. If so, a brain image
without accompanying text should be sufficient to cause
a maximum effect. In the absence of a contribution from
perceptual fluency (an absence we documented in Ex-
periment 1), we predict that an isolated brain image will
be less effective than neuroscience text. On the other
hand, studies on moral judgment have sometimes shown
that the presence of a brain image is effective even when
the image is not explicitly related to the case with textual
elaboration. For example, in one such a study, the in-
clusion of a brain image nudged participants to endorse
deterministic explanations of criminal behavior and mini-
mize moral condemnation (Beall et al., 2013). A potential
limitation of the current study stems from the stimuli
differences across domains. In developing the materials,
we aimed for sentences that were of the same quality in
the non-neuroscience domains as in the neuroscience
domain. We succeeded in that aim for the hard science
disciplines: Participants judged the additional informa-
tion provided in the hard science explanations as equally
likely to appear in an academic journal as the additional
information provided in the neuroscience explanations.
Nonetheless, participants attributed more explanatory
power to the neuroscience explanations than to hard
science explanations. Thus, it appears to be something
specific to neuroscience, not mere information quality,
that drove the effect. It does remain a possibility (despite
what is indicated by our follow-up data) that the addi-
tional information provided differed in some other way
across domain. However, the fact that neuroscience infor-
mation was favored over information from a diverse group
of hard science disciplines, as well as favored over infor-
mation from the social sciences, suggests that there is a
unique advantage when using neuroscience as an explana-
tion for psychological phenomena. If people conceptual-
ize the brain as the engine of the mind, as our study
suggests, then invariably they will see the neuroscience
information as more relevant than hard science counter-
parts. With our follow-up data, we did our best to dis-
entangle this relevance effect from one based on more
overall argument quality by asking participants to evaluate
the domain-specific information separately from the phe-
nomena being explained. However, relevance could be
rightly viewed as an element of argument quality—just
one that is contextually dependent.
Our study was limited to American college students—
and college students taking psychology courses at that.
Although that narrows our sample, it is worth noting that
we found the same effects among students taking Intro-
ductory Psychology (many of whom were not psychology
majors) and students in an upper level physiological psy-
chology course (almost all of whom were psychology
majors). This diversity makes it unlikely that our findings
stemmed from any unique peculiarity of students major-
ing in psychology. Nor did we find differences between
students attending a northeastern private university and
those attending a public institution in the northwest. None-
theless, clearly future studies would benefit by investigat-
ing whether our results generalize to a broader range
of educational levels and backgrounds and also to cul-
tures that hold different lay theories about the relation-
ship between mind and brain than the participants in our
experiments.
Our findings were mostly consistent with the current
literature. First, we found no effect of brain images above
and beyond the effect of neuroscience, a result that rep-
licates the work of Michael and others (Hook & Farah,
938 Journal of Cognitive Neuroscience Volume 27, Number 5
2013; Michael et al., 2013; Schweitzer et al., 2013; Gruber
& Dickerson, 2012). Also, we found no correlation be-
tween the “allure of neuroscience”and beliefs in dualism,
a result consistent with the work of Hook and Farah
(2013). Most importantly, we found that people’srea-
soning about psychological phenomena was biased by the
presence of irrelevant neuroscience information, a finding
that replicates the work of Weisberg and collaborators
(2008). One difference, however, is that in Weisberg’s
study the effect of neuroscience was limited to circular
explanations, whereas in our study it also extended to
good explanations. An explanation of this inconsistency
awaits future studies.
Our results raise the question of why people (or at
least our participants) thought that neuroscience explana-
tions were good explanations, even when they were not
(i.e., even when the explanations were circular). One pos-
sibility that we entertained was that neuroscience carried
an aura of real science that was lacking in the social sci-
ences (Knobe, Prasada, & Newman, 2013). Experiment 2
showed that indeed our participants had a high regard
for neuroscience and that superfluous social science infor-
mation was ineffective in affecting the perceived quality of
the arguments. Nevertheless, Experiments 3 and 4 made
clear that the scientific prestige alone was not the driving
force behind the effect, but rather a coincidental feature,
as neuroscience information was more compelling than
other highly respected disciplines, such as chemistry,
biology, genetics, physics, and math. Furthermore, the
natural sciences weren’t any more compelling than the
social sciences despite their much higher prestige. Thus,
the most plausible mechanism for the allure of neuro-
science may be the brain-as-engine-of-mind hypothesis.
According to this view, people assign to neuroscience
a privileged role in explaining psychological phenomena
not just because neuroscience is a “real”science but be-
cause it is the most pertinent science for explaining the
mind. Absent of a neuroscience link, “hard science”expla-
nations may be too detached from the psychological phe-
nomenon to effectively count as explanations (Yopchick
& Kim, 2009). It remains an open question as to whether
the inclusion of superfluous psychological material would
be as effective as neuroscience material. Psychology is
an interesting domain because it is less prestigious than
neuroscience but equally pertinent—if not more—to
explanations of the mind. As such, it could help to test
the relative contributions of scientific prestige and prox-
imity in determining the explanatory power of various
explanations of mental phenomena.
The domain of psychology may also help explain the
lack of correlation between holding dualist beliefs and
being swayed by the allure of neuroscience. This null re-
sult runs counter to the brain-as-engine-of-mind hypoth-
esis but is consistent with the results of Hook and Farah’s
(2013) study, which was explicitly designed to test such
a correlation. Future studies should assess whether super-
fluous information from psychological science would be
particularly compelling to those who endorse dualist
beliefs.
Part of neuroscience’s allure may be that it allows the
neat, tidy attribution to one causal source: the brain.
Future research should examine how neuroscience expla-
nations stand up to other powerful “single source”expla-
nations, such as genetic explanations. Thus, for example,
Huntington’s disease can be traced to a mutation on the
HD gene (Bates, 2005)—as such, a genetic explanation
provides a very high level of explanatory power for this
disorder. A strong test of the allure of neuroscience bias
would be to contrast the genetic explanations of dis-
eases with neuroscientific explanations. However, a host
of mental phenomena, including symptoms of mental
disorders (Ahn, Proctor, & Flanagan, 2009), affective
states (Lunt, 1991), and moral transgressions (Monterosso,
Royzman, & Schwartz, 2005) are multiply determined—in
fact, single causes are probably the exception rather than
theruleforthephenomenawecareaboutmost.As
such, infatuation with any single source explanation—
whether it is the brain or something else—may impede
humans’progress to find and accept more complete
explanations.
Fernandez-Duque et al. 939
APPENDIX A. SAMPLE STIMULUS USED
IN EXPERIMENT 1
Face Recognition
The Method
Participants sat in front of a computer screen and
watched a rapidly presented series of pictures, half of
which were faces and half of which were places. Half of
the participants had to press a button each time they saw
a face, and the other half of the participants pressed a
button for places.
The Findings
Researchers analyzed the response times and discovered
that the patterns of response times were different for
faces than for places.
The Explanation
This happens because the participants’responses were
contingent on whether they saw a face or a place on
the screen. An examination of the activation patterns
in participants’brains shows that the extrastriate cortex
(known to be involved in processing complex visual stim-
uli) is activated by pictures of faces and places (see Figure
below).
Figure. fMRI of brain, showing in blue the activation of extrastriate
cortex in response to visual stimulus.
This explanation is:
−3−2−1 0 +1 +2 +3
VERY POOR VERY GOOD
APPENDIX B. Superfluous Information—Neuroscience Information Used in Experiments 1–4,
Social Science Information Used in Experiments 2 and 3, and Hard Science Information Used
in Experiments 3 and 4
Topic Neuroscience Social Science Hard Science
Curse of knowledge Studies indicate that
self-knowledge engages
a neural circuitry in the
frontal lobes of the brain.
Studies indicate that
self-knowledge develops
jointly for individuals raised
in interdependent societies.
Studies indicate that
self-knowledge is linked
to epigenetic changes in
the structure of DNA.
Accessibility Heuristic Information about animals is
stored by brain cells of the
hippocampus, which have
been shown to contribute
to memory.
Information about animals is
stored best in hunter–gatherer
societies, which depend on
such knowledge for survival.
Information about animals is
regulated by the expression
of genes important for
long-term memory.
Infant math Scans of babies’brains show
activation in parietal lobe,
an area important for the
integration of information
across sensory modalities.
The size of the family that babies
are raised in determines the
babies’expectations about
the stimulus on this task.
Scans of the babies’eye
movements show a pattern
consistent with Fitts’law,
T=a+blog2 (1 + D/S),
so that the time required to
reach a target (T) is a function
of the target’s distance (D)
and size (S).
Goal-directed actions Brain areas of the visual system
known to process visual
motion become activated
in this task.
The level of involvement in
cultural games and pastimes,
such as “Cops and Robbers,”
“Hide and Seek,”and
“Tag,”influences people’s
descriptions in this task.
The random motion video was
implemented as a continuous
time stochastic process.
940 Journal of Cognitive Neuroscience Volume 27, Number 5
Topic Neuroscience Social Science Hard Science
Face recognition Neuroscientists have shown that
the extrastriate cortex—an
area of the brain known to
be involved in processing
complex visual stimuli—is
activated by pictures of
faces and places.
Sociologists have shown that
faces and places are among
the most important stimulus
for social interactions.
Computational scientists have
used spectrograms to show
that pictures of faces and
places convey a range of
spatial frequencies.
Object labeling The brain’s language areas in
the left temporal lobe are
activated by object labeling
tasks such as this.
This pattern of development
is found across cultures
throughout the world.
This pattern of language
development is disrupted in
children with specific language
impairment due to a mutation
in gene forkhead box P2
(FOXP2) on chromosome 7q.
Mental rotation Brain scans of these participants
show activation of a neural
circuit in parietal and
occipital lobes.
Cross-cultural research shows
that perception of complex
objects is influenced by the
amount of industrialization
in the society.
Genetic studies show that
polymorphism in the ApoE
gene correlates with
performance in this task.
Confirmatory bias A brain structure called the
amygdala known to be
involved in emotional
processing is activated
by this task.
Social factors, such as the
number of people present in
the room have an influence in
the magnitude of the effect.
The response to emotional
messages is modulated by
the gene expression of the
glucocorticoid receptor involved
in the physiology of stress.
Spatial memory As we get older, the memory
centers in the medial-temporal
lobe change in specific ways.
As we get older, the ability
to remember cultural
information remains intact
whereas the ability to
remember visuospatial
information begins to
deteriorate.
As we get older, the expression
of the APOE4 gene becomes
more pronounced, leading to
memory effects.
Other-race effect Scans of the participants’
brains show that the fusiform
gyrus in the brain responds
to faces.
Sociological studies show that
race plays an important role
in people’s perceptions of
each other.
Facial recognition relies on
algorithms based on PCA, a
mathematical procedure that
converts a set of observations
into values of linearly
uncorrelated variables.
Attentional blink Areas in the frontal lobe were
active in response to the
stimuli.
The effect was reduced in
countries where multitasking
is more common.
Biologists have found the same
pattern in other primates.
Religious beliefs The right pFC, a brain area
that contributes to abstract
thinking, is important for
this task.
Symbols, an aspect of culture
that contributes to abstract
thinking, are important for
this task.
Quantum mechanics, an area of
theoretical physics sometimes
used for describing abstract
thinking, provides a unique
framework for interpreting
findings in this area.
Word associations Broca’s area, a part of the
brain’s language system, is
of importance to this task.
The ability to verbally
communicate meaningful
information is a basic human
function that is found in every
society and is of importance
to this task.
Allelic variations in FOXP2, a
gene important for the
language system, have been
associated with performance
in this task.
APPENDIX B. (continued)
Fernandez-Duque et al. 941
APPENDIX C. ATTITUDE TOWARD SCIENCES
1. Each of the three questions was presented in a separate
page, in fixed order, with the nine disciplines listed in
acolumninthefollowingfixedorder(Biology,Psy-
chology, Sociology, Chemistry, Cultural Anthropology,
Political Science, Neuroscience, Genetics, Social Psychol-
ogy). Participants responded by writing a number (1–10)
next to each item
a) Rate the scientific rigor of each of these disciplines
(that is, how close its practitioners adhere to the
scientific method)? (Use a 10 point scale: 1 = not
at all,10=absolutely.)
b) How big is the knowledge gap between what an
expert in this field knows about it and what the
average person knows about it? (1 = no gap,10=
enormous gap)
c) As viewed by society, how prestigious is this disci-
pline? (Now, we are not interested in your personal
beliefs but rather in how society sees it). (1 = not
at all,10=extremely)
2. BEHAVIORAL NEUROSCIENCE (BN) is the study of the
brain and its contributions to thinking and behavior.
CULTURAL PSYCHOLOGY (CP) is the study of culture
and its contribution to thinking and behavior. Match
each trait below to the discipline practitioners that it best
describes (BN = behavioral neuroscience, CP = cultural
psychology). For each trait you should select one and
only one of the two disciplines. BN or CP? Which group
is more Creative, Determined, Sociable, Tolerant, Scien-
tific, Persistent, Sincere, Skillful, Intelligent, Serious,
Warm, Imaginative?
APPENDIX D. DETERMINISM, FREE WILL,
DUALISM, RELIGION
INSTRUCTIONS. This is a questionnaire in which we are
trying to assess how people think about mind, brain, and
behavior. There are no right or wrong answers. Answer
in a 7-point scale from strongly disagree (1) to strongly
agree (7).
1. I think the chief reason why parents and children are
so alike in behavior and character is that they possess a
shared genetic inheritance. _____
2. I believe that many talents that individuals possess can
be attributed to genetic causes. _____
Topic Neuroscience Social Science Hard Science
Spatial abilities In this type of spatial reasoning
task, the right parietal lobe of
the brain, an area important
for spatial processing and
mental rotation, is often
activated.
In this type of spatial reasoning
task, performance differences
are observed between cultures
that speak a language in
which the spatial frame of
reference is egocentrically
defined (left–right) and
those in which the frame
is geocentrically defined
(north, south).
In this type of spatial reasoning
task, testosterone levels are
aligned with group differences
in performance (high in men,
low in women).
Emotional states Parts of the brain involved in
emotion processing, such
as the amygdala, become
active when children
experience distress.
The attachment relation between
mother and child, which
varies across countries,
influences the amount of
distress children experience.
Biochemical assays show that
cortisol levels rise when
children experience distress.
Mental imagery The task led to activation of
large regions the visual cortex
in the lateral occipital lobe.
This pattern of results occurs
even in cultures that
discourage fantasy and
make-believe.
Mathematical models show that
response time is best fit as
a linear function of distance
[T=a+bD].
Binocular rivalry In this task, there is activation
of V1, the primary visual
area of the brain.
It is a phenomenon that occurs
across cultures throughout
the world.
Biologists have found the
same pattern of results in
New World primates.
Memory for vocal patterns Information is stored in
and retrieved from the
hippocampus, the primary
brain area involved in
memory.
Information is stored in and
retrieved from a network
of personal knowledge
embedded within a wider
knowledge of one’s
community and history.
Biochemical studies have shown
that an increase in blood
levels of adrenaline facilitates
memorization during the
studying phase.
APPENDIX B. (continued)
942 Journal of Cognitive Neuroscience Volume 27, Number 5
3. I think that genetic predispositions have little influ-
ence on a person’s personality characteristics ____
4. I am of the opinion that intelligence is a trait that is
strongly determined by genetic predispositions____
5. I think that the upbringing by parents and the social
environment have far greater significance for the de-
velopment of abilities and personal traits than genetic
predispositions. ____
6. Your genes determine your future. ____
7. Bad behavior is caused by bad life circumstances. ____
8. People’s biological makeup influences their talents and
personality. ____
9. People have complete control over the decisions they
make. ____
10. People must take full responsibility for any bad choices
they make. ____
11. People can overcome obstacles if they truly want to.
____
12. The mind is fundamentally physical. ____
13. Some spiritual part of us survives after death. ____
14. Each of us has a soul that is separate from the body.
____
15. All mental processes are the result of activity in the
nervous system. ____
16. The mind is a nonmaterial substance that interacts
with the brain to determine behavior. ____
17. Minds are inside brains but not the same as brains.
____
18. Some nonmaterial part of me (my mind, soul, or
spirit) determines my behavior. ____
19. My mind (consciousness, memory, will) is an emer-
gent property of my brain and cannot be separated
from it. ____
20. The mind and the brain are the same thing. When I
use the word “mind,”it is just a shorthand term for
the things my brain does. ____
21. Minds are in principle independent of bodies, to
which they are only temporarily attached. ____
22. For each thought that I have, there exists a certain
state that my brain is. ____
Religion
23. What are your religious beliefs? (0 = not religious,
7=deeply religious) ____
24. Please indicate the extent to which you believe in the
existence of a Supreme Being or God (or supreme
beings/Gods). (0 = not at all,7=completely) _____
Reprint requests should be sent to Diego Fernandez-Duque,
Psychology Department, Villanova University, 800 Lancaster
Ave., Villanova, PA 19085, or via e-mail: diego.fernandezduque@
villanova.edu.
Notes
1. The good explanation said that “The researchers claim that
this ‘curse’happens because subjects have trouble switching
their point of view to consider what someone else might know,
mistakenly projecting their own knowledge onto others.”
2. For genetic determinism, five items were taken from Keller’s
(2005) work (e.g., “I think the chief reason why parents
and children are so alike in behavior and character is that they
possess a shared genetic inheritance.”). For scientific determin-
ism, three items were taken from Paulhus and Carey’s(2011)work
(e.g., “People’s biological makeup influences their talents and
personality.”). For free will, another three items were taken from
Paulhus and Carey’s(2011)work(e.g.,“Bad behavior is caused by
bad life circumstances.”). For dualism, 11 items were taken from
work conducted by Hook and Farah (2013; e.g., “Some non-
material part of me (my mind, soul, or spirit) determines my
behavior”) and Stanovich (1989; e.g., “Minds are inside brains
but not the same as brains.”). Finally, for religion, we used two
items inquiring about religiosity and belief in a supreme being
(e.g., “What are your religious beliefs (not religious versus deeply
religious)?”).
3. Of the participants who passed the Instructional Manip-
ulation Check, 25 passed it in the first try. Another 83 partici-
pants failed the Instructional Manipulation Check the first time
but passed it after receiving a warning. Another 29 failed even
that second time and, therefore, were excluded from the data
analysis.
4. Analyses are averaged across items. As an exploratory anal-
ysis, we also looked to see if type of superfluous informa-
tion exerted a bias at the level of individual items. For most
items (16 of 18 for the good explanations; 12 of 18 for the cir-
cular explanations), the neuroscience information was rated
higher than the hard science information. Thus, the effect did
not appear to be driven by a few outliers. Next, we assessed
possible differences within the hard sciences information, given
that the superfluous information drew from a diverse set of dis-
ciplines, including genetics (n= 6); biology and biochemistry
(n= 6); and math, physics, and computer science (n= 6). We
entered the data from the item analysis into a 2 × 2 × 3 mixed
ANOVA, with Quality of argument (circular, good) and Domain
(neuro, hard science) as within-item factors and “Item set”as a
between-item factor. Each set had six items grouped according
to the nature of the information provided in the hard science
condition. For example, “curse of knowledge”was in Set 1 be-
cause the hard science information provided for it was about
DNA and epigenetics, whereas “binocular rivalry”was in Set 2
because the hard science information about it drew on biology
and “mental imagery”was in a Set 3 because the hard science
information about it was a mathematical formula. This analysis
showed the expected main effects of Argument quality, F(1, 15) =
13.7, p= .002, and Domain, F(1, 15) = 9.1, p= .008. Most
importantly, it showed no interaction between Set and Domain
(p= .32). For all three disciplines, the bias was in the expected
direction (neuro > hard) [genetics: 4.76 vs. 4.53; bio: 4.81 vs. 4.1;
non-bio: 4.4 vs. 4.1]. In summary, the effect of neuroscience
wasnotlimitedtoacomparison with a single hard science
discipline, but rather it generalized across all three.
REFERENCES
Ahn, W. K., Proctor, C. C., & Flanagan, E. H. (2009). Mental
health clinicians’beliefs about the biological, psychological,
and environmental bases of mental disorders. Cognitive
Science, 33, 147–182.
Bates, G. P. (2005). History of genetic disease: The molecular
Huntington disease: A history. Nature Reviews Genetics, 6,
766–773.
Beall, E., Meindl, P., Iyer, R., Monterosso, J., Graham, J., &
Schwarz, B. (2013). Biology or bad character? Brain
scan images increase determinism and decrease moral
attributions for criminal acts. Paper presented at the
Fernandez-Duque et al. 943
Annual Conference of the Society for Personality and
Social Psychology, New Orleans, LA.
Eriksson, K. (2012). The nonsense math effect. Judgment
and Decision Making, 7, 746–749.
Fiske, S. T., Cuddy, A. J. C., & Glick, P. (2007). Universal
dimensions of social cognition: Warmth and competence.
Trends in Cognitive Sciences, 11, 77–83.
Frederick, S. (2005). Cognitive reflection and decision making.
Journal of Economic Perspectives, 19, 25–42.
Gruber, D., & Dickerson, J. A. (2012). Persuasive images in
popular science: Testing judgments of scientific reasoning
and credibility. Public Understanding of Science, 21,
938–948.
Hook, C. J., & Farah, M. J. (2013). Look again: Effects of
brain images and mind-brain dualism on lay evaluations
of research. Journal of Cognitive Neuroscience, 25,
1397–1405.
Keehner, M., Mayberry, L., & Fischer, M. H. (2011). Different
clues from different views: The role of image format in public
perceptions of neuroimaging results. Psychonomic Bulletin
& Review, 18, 422–428.
Keil, F. C., Lockhart, K. L., & Schlegel, E. (2010). A bump
on a bump? Emerging intuitions concerning the relative
difficulty of the sciences. Journal of Experimental
Psychology: General, 139, 1–15.
Keller, J. (2005). In genes we trust: The biological component of
psychological essentialism and its relationship to
mechanisms of motivated social cognition. Journal of
Personality and Social Psychology, 88, 686–702.
Knobe, J., Prasada, S., & Newman, G. E. (2013). Dual character
concepts and the normative dimension of conceptual
representation. Cognition, 127, 242–257.
Lunt, P. K. (1991). The perceived causal-structure of loneliness.
Journal of Personality and Social Psychology, 61, 26–34.
Markovits, H., & Nantel, G. (1989). The belief-bias effect in
the production and evaluation of logical conclusions.
Memory & Cognition, 17, 11–17.
Masson, M. E. J., & Loftus, G. R. (2003). Using confidence
intervals for graphically based data interpretation.
Canadian Journal of Experimental Psychology-Revue
Canadienne De Psychologie Experimentale, 57, 203–220.
McCabe, D. P., & Castel, A. D. (2008). Seeing is believing: The
effect of brain images on judgments of scientific reasoning.
Cognition, 107, 343–352.
Michael, R. B., Newman, E. J., Vuorre, M., Cumming, G., &
Garry, M. (2013). On the (non) persuasive power of a brain
image. Psychonomic Bulletin & Review, 20, 720–725.
Monterosso, J., Royzman, E. B., & Schwartz, B. (2005).
Explaining away responsibility: Effects of scientific
explanation on perceived culpability. Ethics & Behavior,
15, 139–158.
Oppenheimer, D. M., Meyvis, T., & Davidenko, N. (2009).
Instructional manipulation checks: Detecting satisficing
to increase statistical power. Journal of Experimental
Social Psychology, 45, 867–872.
Paulhus, D. L., & Carey, J. M. (2011). The FAD-Plus: Measuring
lay beliefs regarding free will and related constructs.
Journal of Personality Assessment, 93, 96–104.
Reber, R., & Schwarz, N. (1999). Effects of perceptual fluency
on judgments of truth. Consciousness and Cognition, 8,
338–342.
Rosenberg, S., Nelson, C., & Vivekananthan, P. S. (1968). A
multidimensional approach to the structure of personality
impressions. Journal of Personality and Social Psychology,
9, 283–294.
Schweitzer, N. J., Baker, D. A., & Risko, E. F. (2013). Fooled
by the brain: Re-examining the influence of neuroimages.
Cognition, 129, 501–511.
Stanovich, K. E. (1989). Implicit philosophies of mind: The
dualism scale and its relationships with religiosity and belief
in extrasensory perception. Journal of Psychology, 123, 5–23.
Weisberg, D. S., Keil, F. C., Goodstein, J., Rawson, E., &
Gray, J. R. (2008). The seductive allure of neuroscience
explanations. Journal of Cognitive Neuroscience, 20,
470–477.
Yopchick, J. E., & Kim, N. S. (2009). The influence of causal
information on judgments of treatment efficacy. Memory &
Cognition, 37, 29–41.
944 Journal of Cognitive Neuroscience Volume 27, Number 5
A preview of this full-text is provided by The MIT Press.
Content available from Journal of Cognitive Neuroscience
This content is subject to copyright.
