ArticlePDF Available

Abstract and Figures

Researchers have warned that causal illusions are at the root of many superstitious beliefs and fuel many people's faith in pseudoscience, thus generating significant suffering in modern society. Therefore, it is critical that we understand the mechanisms by which these illusions develop and persist. A vast amount of research in psychology has investigated these mechanisms, but little work has been done on the extent to which it is possible to debias individuals against causal illusions. We present an intervention in which a sample of adolescents was introduced to the concept of experimental control, focusing on the need to consider the base rate of the outcome variable in order to determine if a causal relationship exists. The effectiveness of the intervention was measured using a standard contingency learning task that involved fake medicines that typically produce causal illusions. Half of the participants performed the contingency learning task before participating in the educational intervention (the control group), and the other half performed the task after they had completed the intervention (the experimental group). The participants in the experimental group made more realistic causal judgments than did those in the control group, which served as a baseline. To the best of our knowledge, this is the first evidence-based educational intervention that could be easily implemented to reduce causal illusions and the many problems associated with them, such as superstitions and belief in pseudoscience.
Content may be subject to copyright.
Implementation and Assessment of an Intervention to
Debias Adolescents against Causal Illusions
Itxaso Barberia*, Fernando Blanco, Carmelo P. Cubillas, Helena Matute
Departamento de Fundamentos y Me
todos de la Psicologı
a, Universidad de Deusto, Bilbao, Spain
Researchers have warned that causal illusions are at the root of many superstitious beliefs and fuel many people’s faith in
pseudoscience, thus generating significant suffering in modern society. Therefore, it is critical that we understand the
mechanisms by which these illusions develop and persist. A vast amount of research in psychology has investigated these
mechanisms, but little work has been done on the extent to which it is possible to debias individuals against causal illusions.
We present an intervention in which a sample of adolescents was introduced to the concept of experimental control,
focusing on the need to consider the base rate of the outcome variable in order to determine if a causal relationship exists.
The effectiveness of the intervention was measured using a standard contingency learning task that involved fake
medicines that typically produce causal illusions. Half of the participants performed the contingency learning task before
participating in the educational intervention (the control group), and the other half performed the task after they had
completed the intervention (the experimental group). The participants in the experimental group made more realistic
causal judgments than did those in the control group, which served as a baseline. To the best of our knowledge, this is the
first evidence-based educational intervention that could be easily implemented to reduce causal illusions and the many
problems associated with them, such as superstitions and belief in pseudoscience.
Citation: Barberia I, Blanco F, Cubillas CP, Matute H (2013) Implementation and Assessment of an Intervention to Debias Adolescents against Causal
Illusions. PLoS ONE 8(8): e71303. doi:10.1371/journal.pone.0071303
Editor: Jose
sar Perales, Universidad de Granada, Spain
Received March 22, 2013; Accepted June 25, 2013; Published August 14, 2013
Copyright: ß 2013 Barberia et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits
unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: Support for this research was provided by Grant 2011-26965 from Direccio
n General de Investigacio
n of the Spanish Government, and Grant IT363-10
from Departamento de Educacio
n, Universidades e Investigacio
n of the Basque Government. The funders had no role in study design, data collection and analysis,
decision to publish, or preparation of the manuscript.
Competing Interests: The authors have declared that no competing interests exist.
* E-mail:
Despite the exponential development of scientific research in
recent decades, the sad truth is that many people still hold a vast
number of unrealistic and irrational beliefs about the functioning
of the world. Some of these beliefs are clearly eccentric and openly
violate our present knowledge about the laws of nature, including
superstitions related to supernatural forces such as omens,
witchcraft, astrology, and psychic powers. Moore [1] reported
that belief in psychics and paranormal activity is worryingly
prevalent in the American population. For example, 32% of the
people interviewed in 2005 believed in ghosts, 37% thought that
houses can be haunted, and 21% believed that witches exist (see
also [2]).
Misbeliefs also underlie many pseudoscientific practices, which
are especially dangerous because they ‘‘possess the superficial
appearance of science but lack its substance’’ ([3] p. 1216). These
practices are intentionally presented as scientific, even if they do
not meet the minimum acceptable standards for science. This is
the case for many so-called ‘‘alternative medicines’’, such as
homeopathy. According to the Special Eurobarometer on Science
and Technology [4], 34% of Europeans consider homeopathy to
be ‘‘scientific’’. Meanwhile, the purported therapeutic mechanism
of homeopathic products is implausible, and research shows that
their alleged healing effects may be attributable solely to the
placebo effect [5]. Nevertheless, many people use these products,
sometimes substituting them for treatments with demonstrated
efficacy [6]. It should be obvious that superstitious and pseudo-
scientific beliefs become extremely worrisome when they begin to
drive people’s decisions about many important areas of their daily
lives. The effects range from the expenses paid to fortune-tellers or
clairvoyants to health risks of ineffective treatments for a variety of
Interestingly, one phenomenon that is central to these
unrealistic beliefs is the fact that people sometimes develop
illusions of causality, that is, they perceive the existence of causal
links between events that are actually uncorrelated [7], [8]. Our
cognitive system has evolved to sensitively detect causal relation-
ships in the environment, as this ability is fundamental to predict
future events and adjust our behavior accordingly. However,
under certain conditions, the very same cognitive architecture that
encourages us to search for causal patterns may lead us to
erroneously perceive causal links that do not actually exist. These
false perceptions of causality may be the mechanism underlying
the emergence and maintenance of many types of irrational
beliefs, such as superstitions and belief in pseudoscience. These
illusions could also be the basis of many types of group stereotypes
[9] and may promote ideological extremism [10] hence contrib-
uting to intergroup conflict and suffering throughout the world.
A fundamental difficulty of causal induction, as acknowledged
by philosophers such as Hume [11], is that causality is not directly
observable and must therefore be inferred from empirical
evidence. One of the primary empirical inputs for causality is
the correlation or contingency between the potential cause and the
PLOS ONE | 1 August 2013 | Volume 8 | Issue 8 | e71303
outcome of interest. Given one binary cause and one binary
outcome, the contingency can be easily formalized by using the Dp
index [12], [13], which is the difference between the probability of
the outcome given the presence of the potential cause and the
probability of the outcome given its absence: Dp = P(Outcome|-
Cause) - P(Outcome|,Cause). The result of this subtraction
determines the contribution of the potential cause to the
occurrence of the outcome, providing information about the
generative, preventive, or null influence of the former on the latter.
Although more sophisticated rules than Dp have been formulated
as a normative standard for assessing causal strength [14], they
usually include the same contrast between the two mentioned
conditional probabilities. Furthermore, most normative indexes
yield the same result when assessing null contingencies, that is,
situations in which P(Outcome|Cause) = P(Outcome|,Cause),
which are the major focus of this paper.
Much as it could happen in real world, the participants in a
typical causal learning experiment are initially presented with two
events that might be causally related (the target cause and the
outcome). Imagine that the hypothesized relationship involves a
certain herb as a potentially effective remedy for headaches. Next,
the participants are sequentially exposed to several observations (in
this example, several fictitious patients) in which the presence and
the absence of the potential cause and the outcome are combined
to generate four different possible situations: (a) the herb (i.e., the
potential cause) is ingested, and relief (i.e., the outcome) occurs; (b)
the herb is taken, and relief is not observed; (c) the herb is not
ingested, but relief occurs; and (d) the herb is not taken, and relief
is not observed. The contingency between the potential cause and
the outcome is experimentally manipulated by determining the
frequency of each of these four types of event. After being exposed
to a number of these fictitious cases during the training phase, the
participants are asked to evaluate the extent to which the herb is
actually able to produce relief (i.e., to judge the strength of the
causal relationship between the potential cause and the outcome
under study). Although some studies have demonstrated that
participants’ causal perception is in fact sensitive to the actual
contingency [15], [16], [17], systematic biases have been detected
as well.
One extensively studied bias is the illusion of causality or causal
illusion cited above (note that we will use the general term illusion of
causality or causal illusion to refer to what some authors have called
the illusion of control. We assume that the latter is simply a specific
case of the former). This bias refers to the perception of a causal
relationship when the target event and the outcome occur together
by mere chance, but are actually independent of each other. That
is, the probability of the outcome is the same regardless of the
presence or absence of the potential cause, P(Outcome|Cause) = -
P(Outcome|,Cause), and, therefore the contingency is zero. One
fundamental factor that modulates the magnitude of this illusion is
the probability of the occurrence of the outcome, P(Outcome).
When the outcome occurs frequently, the illusory perception of a
causal link is facilitated [18], [19], [20], [21], [22], [23].
Relevant to the present work, the probability of the potential
cause, P(Cause) also influences the illusion of causality. More
specifically, as P(Cause) increases, so it does the misperception of
the effectiveness of the potential cause, especially when the
outcome also occurs relatively frequently [20], [24], [25].
Significantly, in those experiments, the contingency is always set
to zero; therefore, the probability of relief will be similar regardless
of whether the herb is administered or not. However, if the
P(Outcome) is high, many patients will recover from their
headaches independent of the administration of the herb. In these
circumstances, any potential treatment (such as the herb in our
example) will have a strong chance of co-occurring (or occurring
close in time) with healing. Moreover, the number of co-
occurrences will increase as P(Cause) increases. These coincidenc-
es may be largely responsible for the increased perception of a
causal relationship. In fact, as noted by many researchers [26],
[27], [28], [29], [30], [31], [32], causal impressions may be guided
by heuristics that give more weight to cause-outcome coincidences
(i.e., instances that seem to confirm the relationship) than to
outcome occurrences that could not have been generated by the
potential cause (but see [33] for an explanation of how, under
certain assumptions, it could be normative to weight coincidences
more heavily than the rest of event types).
On this basis, one way to diminish causal illusions may be to
decrease the number of accidental cause-outcome coincidences to
which people are exposed. This goal might be reached by
decreasing either the P(Outcome) or the P(Cause). Unfortunately,
the P(Outcome) is usually out of individual and institutional
control in real life (e.g., it is not possible to change the prevalence
of conditions such as headaches). However, the P(Cause) is
frequently subject to individual decisions and can therefore be
modified. In this sense, previous research on causal illusions has
sometimes employed contingency learning tasks in which the
participants are allowed to decide, for each observation, whether
they want to introduce the target cause or not, and subsequently
observe if the outcome occurs [18], [24], [25]. In these active tasks
in which the participant’s behavior directly determines the
proportion of trials in which the cause is present, instructional
manipulations can reduce the illusion of causality by encouraging
the participant to introduce the target cause in approximately half
of the trials [20], [25], [34]. However, when the instructions
simply request that the participants try to obtain the outcome as
often as possible [24], [25], the participants’ natural tendency
seems to be to introduce the potential cause a relatively high
number of times. For example, Blanco et al. [24] found that the
participants in their study introduced the potential cause in more
than half of the trials and that this behavior became more marked
as the experiment progressed. Although Blanco et al. [24] did not
report this specific statistical analysis, we analyzed their data and
observed that their participants administered the potential cause
more often than that expected by chance. A comparison against a
theoretical P(Cause) of 0.5 showed that the difference was
statistically significant, t(81) = 3.86, p,.001, for Experiment 1
and t(91) = 7.63, p,.001, for Experiment 2.
This tendency to focus on the instances of the reality in which
the potential cause is present might be related to a more general
bias in people’s information sampling strategies. To this respect,
several authors have called attention to an effect that is sometimes
called the positivity bias or positive testing strategy [29], [35], [36]. This
effect refers to the finding that, when testing the validity of a
hypothesis, people predominately sample cases that would
produce a positive outcome if the hypothesis were correct [37],
[38], [39], [40], [41] (note that even though some of these findings
were initially interpreted as indicating a confirmation bias, Klayman
and Ha [38] subsequently argued that the preference for positive
tests does not necessarily represent a bias towards confirmation,
see also [37], [42]). For example, when testing the hypothesis that
a person is an extrovert, people tend to ask more questions that
refer to extraversion than questions that refer to introversion [40].
In other words, people preferentially ask questions that would
receive an affirmative answer if the person were indeed an
extravert. We propose that a similar hypothesis-testing strategy
might operate in a typical causal learning experiment. If we are
presented with the goal of determining whether an herb is an
effective remedy for eliminating headaches, a positive testing
Debiasing Adolescents against Causal Illusions
PLOS ONE | 2 August 2013 | Volume 8 | Issue 8 | e71303
strategy might involve predominately choosing to observe cases in
which the herb is taken rather than cases in which the herb is not
taken. A high P(Cause) value is the logical consequence of a
positive testing strategy. If relief from headaches occurs at a high
rate [i.e., if P(Outcome) is high] and we frequently ingest the herb
when we experience the early symptoms of a headache [i.e., if
P(Cause) is high], then recovery will frequently follow herb
ingestion. Because we persevere in this behavior, we tend not to be
exposed (or at least we are less often exposed) to information
related to the actual rate of spontaneous remission. As a result, the
impression that the herb is effective will persist.
A straightforward method of overcoming this natural tendency
might involve encouraging people to recognize the importance of
searching and considering information about the state of facts
when the potential cause is absent. This is in some sense analogous
to teaching people about the logic of experimentation in science, in
which experimental and control groups are meticulously designed
to assess the influence of the factor that is being tested,
independent of any other confounding variables. Although this
idea underlies many scientific educational programs [43], [44],
[45], we know of no program that has tested the efficacy of this
training in reducing subsequent causal illusions. Based on this idea,
the goal of the present study was to develop and test an
educational intervention that would reduce the tendency to
illusory perceptions of causality by encouraging people to
understand the importance of exposing themselves to more
cause-absent control observations. The targets of the intervention
were secondary school students. The decision of addressing the
intervention to this population was both practical (i.e., students at
that level are still immersed in educational contexts in which it is
easier to intervene) and theoretical, as adult resistance to scientific
thinking may arise soon in life [46].
Overview of the Intervention
The intervention introduced the participants to the concept of
contingency described above. In the intervention, the participants
learned that comparing the probability of an outcome in the
presence and the absence of the potential cause is the normative
manner of assessing the empirical evidence for a hypothesized
causal link. In this sense, the emphasis was placed on the idea that
the rate of cause-outcome co-occurrence, or P(Outcome|Cause), is
necessary to infer a generative causal relationship between a
potential cause and an outcome but is certainly not sufficient. The
base rate of the outcome, or P(Outcome|,Cause), is also
important to consider if one is to reach an appropriate conclusion.
Understanding the necessity of considering this latter piece of
information should encourage people to expose themselves to
more instances in which the cause is absent. That is, to expose
themselves to a lower P(Cause), which should, in turn, diminish the
tendency to develop causal illusions.
We know of no other educational intervention that was
designed to debias people against causal illusions. Indeed,
relatively little work has been done to investigate the extent to
which debiasing against a variety of cognitive biases is possible,
and the results have been mixed (see [10], [47], [48] for reviews).
Lilienfeld et al. [10] eloquently noted several potential barriers to
successful debiasing interventions that should not be ignored. First,
they suggested that people might not debias because they tend not
to accept that their perspective is biased; this effect is called the
‘‘bias blind spot’’ or the ‘‘not me’’ bias [49], [50], [51]. Second,
they alerted that interventions are more effective when people
perceive that the bias to be corrected is relevant to their daily lives
[47], [52]. Following Lilienfeld et al.’s [10] recommendations, we
expected that we might improve the effectiveness of our debiasing
intervention by first demonstrating to the participants how easily
they might arrive at biased conclusions and how relevant these
conclusions might be to their daily lives. To achieve this goal, the
intervention included an initial section in which we tried to induce
a causal illusion in our participants before proceeding to explain
how to approach causal induction normatively.
The effectiveness of the intervention was measured by exposing
the participants to a typical contingency learning task, as described
in the Introduction. The participants were asked to determine the
strength of a cause-outcome relationship through successive
observations in which they could decide if they wanted to
introduce the potential cause and then subsequently observed
whether the outcome occurred. Half of the participants (the
control group) performed the contingency learning task before the
intervention was conducted, whereas the other half (the experi-
mental group) performed the same task after the intervention was
completed. Our hypothesis was that the intervention would
improve the ability of the experimental participants to gather
and evaluate empirical evidence for a potential causal relationship.
More specifically, we expected that the participants who had been
exposed to the intervention would show a decreased tendency to
choose to observe cause-present observations [i.e., they would
generate lower P(Cause) values] and would exhibit weaker
causality illusions than the participants who performed the task
before the intervention.
The computer task involved sequentially investigating two
potential causal relationships. That is, the participants were asked
to evaluate the effectiveness of a medicine not only in a zero
contingency situation in which P(Outcome|Cause) = P(Outco-
me|,Cause), but also in a second situation in which there was
a generative relationship between the potential cause and the
outcome, P(Outcome|Cause) .P(Outcome|,Cause). The posi-
tive contingency condition served to control for the possibility that
the intervention made the participants generally more skeptical
about any potential cause-effect relationship, rather than improv-
ing their specific understanding of how causal effectiveness should
be assessed and interpreted.
Ethics Statement
The ethical review committee of the University of Deusto
approved this study. The intervention was offered to the Faculty of
Engineering of the University of Deusto as a workshop on critical
thinking that could be implemented as part of their larger set of
activities in a summer camp on technology and robotics. The
Faculty of Engineering was provided with detailed written
information on the purpose, methods, and data treatment of the
study before they invited us to participate in their programs. Only
those minors whose parents, next of kin, guardians, or caretakers
specifically requested in their general application to the summer
camp to participate in our workshop (including its evaluation and
statistical treatment and potential publication of the data)
participated in the workshop and the present evaluation of its
efficacy. The ethical review committee of the University of Deusto
considered that the experimenters did not need to obtain an
additional direct consent of the participants’ parents, but only that
of the summer camp organizers, which was obtained verbally in
agreement with the ethical review committee advice.
Sixty-two secondary school students participated in the study.
The control group and the experimental group both included 31
participants. Data were not recorded for two participants in the
Debiasing Adolescents against Causal Illusions
PLOS ONE | 3 August 2013 | Volume 8 | Issue 8 | e71303
control group; therefore, they are not included in the analyses.
Thus, the final N was 60. In addition, two participants in the
control group failed to provide their ages. Among the participants
who reported their ages, the average ages were 14.26 (SEM = 0.30)
and 14.84 (SEM = 0.29) in the control and experimental groups,
respectively. There were no significant age differences between the
groups: t(56) = 1.38, p=.17.
The same intervention was repeated in four separate sessions,
each involving different participants (a minimum of 12 participants
and a maximum of 20 per session). The participants were
randomly assigned to the control and experimental groups at the
beginning of each session, and each group was placed in a different
classroom. The only variation between the groups was the time at
which the participants performed the contingency learning task:
the control group performed the task before the intervention, and
the experimental group performed the task after the intervention.
The instructions for the contingency learning task were provided
by the same experimenter for both groups.
Intervention. The intervention was divided into two phases.
The first phase involved staging. The second phase involved an
explanation of the appropriate manner of approaching everyday
causal inference. Both phases were completed in a single session of
approximately 80 minutes.
Phase 1. The goal of this phase was to generate a situation in
which participants might be inclined to form a biased impression
of the effectiveness of a target product. The product was a small
rectangular piece of regular ferrite. However, the participants were
told that the product was made of a brand new material that had
recently been developed by a group of researchers. They were told
that upon contact with human skin, the product stimulates the
nervous system, improving both the physical and the intellectual
abilities of its carrier. Mimicking the strategy used in pseudosci-
ence [3], [49], we offered an explanation that was intentionally
hyper-technical (i.e., we employed scientific jargon, using words
such as electromagnetism, cell, atom, nervous system, and activation).
Once the alleged properties of the ferrite bar had been
explained, the experimenter attached the product to the wrist of
each of the participants. The participants were then asked to
perform a series of tasks to experience the benefits of the product.
First, they had to complete several paper-and-pencil tasks (e.g.,
solving mazes or crossing out all the consonants from a matrix of
letters and numbers) as quickly as possible. Significantly, the
participants always performed these tasks while wearing the ferrite
bar. Therefore, they lacked a decisive control condition to which
compare their performance. However, we tried to influence the
participants’ perceptions regarding the effectiveness of the product
by telling them after each exercise that the people who wore the
product in previous tests reported that they had felt that they had
performed the tasks especially well (e.g., when solving the mazes,
they could determine the solution very quickly, as if their minds
were faster than their hands).
In a second series of activities, the participants were presented
with several physical exercises involving strength, stability and
flexibility. The exercises were similar to those advertised on the
webpages of popular performance-enhancing bracelets that have
been proven to be bogus, such as Power BalanceH (www., as cited by Porcari et al. [53]; this
video has now been removed). Using a procedure similar to the
procedure presented in these videos, we encouraged the
participants to perform each exercise first without wearing the
ferrite bar and immediately afterwards holding the product in one
hand. Therefore, in this second series of activities, we did provide
the participants with a control condition (i.e., they could compare
their performance with and without the product). However, the
control condition was intentionally suboptimal because the
influence of the ferrite bar was confounded with the potential
influence of warming up and learning from practice (i.e., the test
with the product was always performed second). In fact, previous
research has shown that the alleged effects of holographic bracelets
disappear when the effect of order is controlled for [53], [54].
Phase 2. After the first staging phase and before we informed
the participants that the product was fake, we introduced the
participants to the concept of contingency as the correct way to
infer causality from empirical information. Before introducing the
idea of contingency, we suggested some examples in which the
target cause was frequently followed by the outcome (see [55] for
the additive benefit of employing both formal rules and examples
in statistical reasoning training). For instance, we presented a
situation similar to the herb-relief example described in the
introduction and asked questions such as ‘‘If 80% of people who
take the herb feel better, does this represent proof of the herb’s
effectiveness?’’ These examples were used to emphasize the idea
that even a high rate of cause-outcome coincidences does not
guarantee the existence of a causal link. Moreover, we focused on
the fact that if people experienced relief soon after taking the herb,
they might feel inclined to continue using the herb in the future,
thus depriving themselves of the possibility of observing whether
relief was likely to occur spontaneously (i.e., without their taking
the herb).
The participants were also briefly introduced to the importance
of choosing a good control condition when testing causal links. We
suggested that it is fundamental to compare the probability of the
outcome in the presence and the absence of the potential cause
while independently controlling for all of the other factors that
could also affect the outcome. For example, the participants were
invited to consider the case of a plant fertilizer that is tested on a
farm in a rainy location. If the growth observed at this farm is
compared with that observed at a farm that did not receive the
fertilizer but is located in a drier place, we will not be able to
determine if the superior growth of the plants that received the
fertilizer is due to the fertilizer or to the differences in the climate.
This example is analogous to the unclear boundary between the
influence of the product and the influence of the practice in the
physical exercises performed in the first phase of the intervention.
At the end of this phase, the participants were asked to judge,
given the information they received, whether the method of testing
the ferrite bar that was used in the first phase was adequate. After
a discussion about the problems related to the cognitive tasks (i.e.,
a discussion of the lack of a control condition) and the physical
activities (i.e., a discussion of the inadequacy of the control
condition), we revealed the truth to the participants about the
ineffectiveness of the ferrite bar.
Measurement (contingency learning task). Participants in
neither group were given any information about the purpose of the
contingency learning task. In the case of the control group, they
were told that, before starting with the workshop, they would be
playing a computer game. Similarly, participants in the experi-
mental group were told that the workshop was finished and that
they would then play a computer game.
The procedure was similar to the conventional contingency
learning paradigm used in the literature (e.g., [24]). It consisted of
two different stages. In the first stage, the participants were asked
to imagine that they were medical doctors. They were told that a
fictitious medicine, Batatrim, could potentially provide relief for
patients suffering from a fictitious illness called the Lindsay
syndrome. They were asked to determine the extent to which the
Debiasing Adolescents against Causal Illusions
PLOS ONE | 4 August 2013 | Volume 8 | Issue 8 | e71303
medicine was effective. To do so, the participants sequentially
observed the records of 40 fictitious patients suffering from the
syndrome and decided whether they wanted to administer the
medicine to each patient. In each trial, after making the decision,
they observed whether the patient was cured. After all 40 patients
had been observed, the participants were asked to evaluate the
effectiveness of the medicine on a scale ranging from 0 (ineffective)
to 100 (entirely effective). In the second stage, the participants
were presented with a second medicine, Dugetil, as a potential
treatment for a second syndrome, Hamkaoman. The procedure
was exactly the same as the procedure for the first medicine.
One of the medicines was presented in a zero contingency
condition, P(Outcome|Cause) = P(Outcome|,Cause), whereas
the other was presented in a positive contingency condition,
P(Outcome|Cause) .P(Outcome|,Cause). In the zero contin-
gency condition, 6 out of every 8 patients felt relief independent of
the participants’ decision to administer the medicine (i.e., the
probability of relief was.75 with or without the medicine). In the
positive contingency condition, only 1 out of every 8 patients who
did not receive the medicine felt relief, whereas 6 out of every
8 felt relief after taking the medicine (i.e., the probability of relief
was.75 with the medicine and.125 without it). Therefore, the
objective contingency was.625 in the positive contingency condi-
tion. Half of the participants from each group were exposed first to
the zero contingency condition and then to the positive
contingency condition. This order was reversed for the other half
of the participants.
It is plausible to assume that the participants in the experimental
group, who had just been deceived by the experimenters, might
have been especially likely to feel suspicious when performing this
task and might therefore have been less prone to assume the
existence of a causal relationship, regardless of the evidence they
encountered. The positive contingency condition was introduced
to control for this potential confounding effect. We expected the
participants in the experimental group to be more realistic in the
zero contingency condition while retaining the ability to reach an
accurate conclusion about the positive contingency condition.
The top panel of Figure 1 shows the proportion of trials in
which the participants from each group chose to administer the
medicine [i.e., this figure represents the P(Cause) values]. The
mean P(Cause) in the zero contingency condition was.75
(SEM = .04) for the control group and.56 (SEM = .05) for the
experimental group, whereas the mean P(Cause) in the positive
contingency condition was.86 (SEM = .02) for the control group
and.66 (SEM = .04) for the experimental group. The participants
in the control group generated a higher P(Cause) than the
participants in the experimental group. There was also an increase
in the P(Cause) for the positive contingency condition relative to
the zero contingency condition. A 2 (experimental vs. control
groups) 6 2 (zero contingency vs. positive contingency) ANOVA
for the P(Cause) showed the significant main effects of Group, F(1,
58) = 17.88, MSE = .07, p,.001, g
= .24, and Contingency, F(1,
58) = 15.08, MSE = 0.02, p,.001, g
= .21. The interaction was
not significant, F ,1. As expected, the participants that were
exposed to the intervention generated a lower P(Cause) than
participants in the control group. There was also an unexpected
main effect of Contingency, showing that the participants
introduced the target cause in a higher proportion of trials in
the positive contingency condition than in the zero contingency
condition. We could hypothesize that this effect might be related
to the fact that administering the medicine produced a higher
relief rate than not administering the medicine in the positive
contingency condition. Even though the participants were
instructed to try to determine whether the medicine was effective
(instead of producing the relief in as many patients as possible),
given that relief can be considered a desirable outcome, they may
have found it difficult to refrain from acting in a way that
increased their chances of curing the fictitious patients.
The bottom panel of Figure 1 shows the causal judgments made
by the participants at the end of the task. The values were high
and similar for the two groups in the positive contingency
condition: the mean judgment was 70.28 (SEM = 2.77) for the
control group and 71.81 (SEM = 2.59) for the experimental group.
By contrast, in the zero contingency condition, the judgments
made by the participants in the experimental group were closer to
zero (i.e., they were more accurate) than those made by the
participants in the control group: the mean judgment was 61.69
(SEM = 3.35) for the control group and 30.55 (SEM = 5.59) for the
experimental group. Therefore, the intervention reduced the
illusory perception of causality, leading the participants in the
experimental group to make more realistic judgments than the
control participants. This conclusion is supported by the informa-
tion conveyed by the histograms in Figure 2: the intervention led
many participants in the experimental group to give a judgment of
zero in the zero contingency condition. A 2 (experimental vs.
Figure 1. Mean P(Cause) (top panel) and mean causal
judgments (bottom panel). The top panel represents the mean
proportion of trials in which the participants from the control and
experimental groups decided to administer the medicine in the zero
contingency and positive contingency conditions. The bottom panel
represents the mean judgments regarding the effectiveness of the
medicines in the zero contingency and positive contingency conditions.
The filled bars refer to the participants in the control group, and the
striped bars refer to the participants in the experimental group. The
error bars represent the standard error of the mean.
Debiasing Adolescents against Causal Illusions
PLOS ONE | 5 August 2013 | Volume 8 | Issue 8 | e71303
control groups) 6 2 (zero contingency condition vs. the positive
contingency condition) ANOVA on the judgments of causality
showed significant main effects of Group, F(1, 58) = 16.34,
MSE = 401.98, p,.001, g
= .22, and Contingency, F(1,
58) = 39.55, MSE = 470.58, p,.001, g
= .41. More importantly,
there was a significant interaction between Group and Contin-
gency, F(1, 58) = 16.99, MSE = 470.58, p,.001, g
= .23. The
causal judgments did not differ significantly between the exper-
imental group and the control group for the positive contingency,
t(58) = 0.40, p=.69, but they did differ for the zero contingency,
t(58) = 4.70, p,.001. Thus, the participants in the experimental
group were able to detect the existence of a causal relationship
when there was good evidence for it and to detect its absence when
there was no evidence for it.
In addition, analyses were conducted to determine whether the
effect of the intervention on the judgments in the zero contingency
condition was direct or whether it was mediated by the differences
in P(cause) between the two groups. Thus, we conducted a
mediation analysis [56] to isolate the direct effect of Group (i.e.,
the intervention) on judgments while partialling out the effect of
the P(Cause). In other words, we can determine the amount of
variance in causal judgments between the experimental and
control groups that can be attributed to the intervention producing
differences in the P(Cause) between the two groups, which, in turn,
produced the differential perceptions of causality (see [34] for a
similar strategy).
The mediation analysis procedure described by Baron and
Kenny [56] consists of three consecutive steps that reveal three
pieces of information: the total effect of Group on judgments (path
c in Figure 3), the indirect effect explained by the mediation of
P(Cause) (paths a and b in Figure 3 ), and the direct effect of
Group that remains after the indirect effect has been partialled out
(path c’ in Figure 3). Following this procedure, we first assessed the
total effect of Group on the judgments (by regressing the
judgments onto the Group), b = .52, t(58) = 4.70, p,.001. Next,
we assessed the indirect effect of Group on judgments mediated by
the P(Cause). This involves two requirements: (a) ensuring that the
participants in the experimental group, compared with those in
the control group, did generate a lower P(Cause) [i.e., the P(Cause)
was regressed onto the Group], b = .38, t(58) = 3.09, p,.01; and
(b) showing that the P(Cause) had a positive impact on judgments
while controlling for the effect of Group [this was done by
conducting a multiple regression analysis on the judgments with
P(Cause) and Group as predictors], b = .64, t(57) = 7.32, p,.001.
Ultimately, the same multiple regression model [with P(Cause) and
Group as predictors of judgments] revealed that a significant direct
effect of Group on judgments remained after the indirect effect
mediated by the P(Cause) had been partialled out, b = .28,
t(57) = 3.26, p,.01. The Sobel test [57] indicated that, although it
was significant, the direct effect of Group on judgments was
significantly smaller than the total effect reported in the first step in
which the effect of P(Cause) was not partialled out, z = 22.85,
p,.01. This partial mediation suggests that the impact of the
intervention on causal judgments in a zero contingency situation
occurred for two reasons: first, the intervention directly produced
more realistic judgments in the experimental group; second, the
intervention also affected the judgments indirectly by decreasing
the P(Cause) that the participants generated.
It could be argued that, since the participants decided the
number of trials in which the potential medicine was administered,
the actual contingency experienced by participants could slightly
depart from the programmed contingency. However, these
variations were minimal: the mean contingency to which the
participants were exposed in the zero contingency condition
was.04 (SEM = .04) for the control group and.02 (SEM = .02) for
the experimental group, whereas the mean contingency to which
the participants were exposed in the positive contingency
condition was.63 (SEM = .04) for the control group and.64
(SEM = .01) for the experimental group. Nevertheless, and in
order to ensure that these variations did not affect the results, we
Figure 2. Histogram depicting the distribution of the judgments in the zero contingency condition, for the experimental group (top
panel) and the control group (bottom panel).
Debiasing Adolescents against Causal Illusions
PLOS ONE | 6 August 2013 | Volume 8 | Issue 8 | e71303
repeated all the previous statistical analyses on accuracy scores
instead of causal judgments, finding the same conclusions.
Accuracy was computed as the absolute difference between the
causal judgment (rescaled to range between 0 and 1) and the actual
contingency each participant was exposed to.
In the Introduction, we suggested that causal illusions underlie
many of the superstitious and pseudoscientific beliefs that prevail
in our society [7]. Because unrealistic beliefs can be extremely
harmful, we believe that teaching people how to evaluate causal
hypotheses more accurately has broad implications for the effort to
develop a knowledge-based society. The present intervention
constitutes an initial effort in this direction.
We found that training a group of adolescents in the rational
manner of making inferences about cause-outcome relationships
decreased their illusory perceptions of causality in a subsequent
non-contingent situation. Moreover, including a control condition
in the positive contingency scenario allowed us to conclude that
the lower causal ratings observed in the experimental group could
not be solely explained by a general increase in suspicion in this
group. Rather, the group specifically made more realistic
judgments in the null contingency condition while preserving an
accurate view of the positive contingency condition.
In addition, a mediation analysis showed that one of the reasons
for this decrease in the illusion of causality was that the
intervention helped the participants to diminish their exposure
to the potential cause. As noted in the Introduction, our
spontaneous tendency is to expose ourselves to more cause-present
observations than cause-absent observations, which strongly
contributes to the development of the illusion [24]. As shown in
Figure 1, the behavior of the participants in the control group
supported the idea that this spontaneous tendency is the default
strategy: the participants in this group generated an average
P(Cause) of.75 (i.e., they chose to administer the medicine in
approximately 3 of 4 observations) when the situation involved a
zero cause-outcome contingency. In contrast, the behavior of the
participants in the experimental group suggests that they
internalized the importance of experimental control because they
tended to generate more cause-absent observations than did those
in the control group. The average P(Cause) in the experimental
group in the zero contingency condition was only .56.
In addition to the differences between the information sampling
strategies, the causal judgments reported by the participants in the
experimental group were lower than those reported by the control
group. Although the mediation analysis reported above showed
that the effect of the intervention on causal judgments was partially
mediated by differential exposure to cause-present trials, the
intervention still had a significant, direct effect on the final causal
judgments that was independent of the P(Cause) that the
participants generated. This effect is consistent with the idea that
people tend to spontaneously weight cause-present information
(and especially cause-outcome coincidences) more heavily than
they do cause-absent information [30]. On this basis, it seems that
the intervention most likely affected causal judgments in two
different ways: first, by diminishing the P(Cause) generated by the
participants, and second, by encouraging the participants to pay
more attention to cause-absent information, P(Outcome|,Cause),
or to weight that information more heavily. Future studies should
measure the effectiveness of the debiasing intervention while
controlling the subsequent exposure to the P(Cause). This could be
easily done by employing observational contingency learning tasks
(e.g., [7]) in which participants cannot decide if the cause is
introduced or not but they merely observe whether the cause is
present and whether the outcome occurs [i.e., the P(Cause) is set
up by the experimenters]. This would allow us to isolate the
individual influence of the intervention in the evidence sampling
strategies from its influence in the interpretation of this evidence.
Our approach to measuring the effectiveness of the intervention
was fairly different from the conventional strategy that is used in
the statistical reasoning literature [55], [58], [59]. These studies
typically measure the internalization of statistical concepts by
presenting the participants with verbal descriptions of everyday
situations that involve applying these concepts. For example, Fong
et al. [55] trained a group of participants in the ‘‘law of large
numbers’’ and found improvement when the participants were
asked to reason about verbal descriptions of new examples to
which the law was relevant. The analogous strategy in our domain
could be presenting participants with information about the
relevant probabilities for calculating the contingency, P(Out-
come|Cause) and P[Outcome|,Cause), in new scenarios and
then determining whether they could infer that there was (or that
there was no) good empirical evidence for assuming a causal
relationship. Instead, in the present work, the effectiveness of the
intervention was measured using a transfer task that required the
practical application of the principles underlying the concept of
contingency in a trial-by-trial setting. Our study suggested that
understanding the necessity of considering the P(Outcome|,-
Cause) value led to a change in the decision-making process on a
trial-by-trial basis. We believe that this strategy is meaningful
because the contingency learning task entails a context that
resembles more closely the process of causal induction in many
real life situations in which people (and other animals) learn by
sequential observations of the presence and absence of the
potential cause and the outcome over time, instead of encounter-
ing the covariational information in a summarized format. The
use of a sequential presentation format also has theoretical
implications; it has been suggested that the cognitive mechanisms
that operate when dealing with trial-by-trial information (such as
the information in our study) are not necessarily the same as those
that are activated when this information is received in a
summarized form [60]. Finally, an additional advantage of using
a contingency learning task to assess the intervention is that this is
a standard procedure in the study of human judgment and
decision-making, and thus, an extensive body of relevant
theoretical and experimental literature is available [61]. Never-
Figure 3. Mediational structure underlying the experimental
manipulation in the zero contingency condition. The total effect
of the intervention on the causal judgments, depicted as path c, is
divided into two components: one indirect effect (paths a and b)
through the P(Cause), and one direct effect (path c’), which is the result
of discounting the indirect effect. The mediation analysis reveals that
the intervention affected the judgments both directly and indirectly, via
the P(Cause).
Debiasing Adolescents against Causal Illusions
PLOS ONE | 7 August 2013 | Volume 8 | Issue 8 | e71303
theless, in our modern society people frequently have to deal with
summarized information (e.g., statistics that appear in the
newspapers) and, therefore, future studies might explore if the
present results prevail when employing a summarized version of
the contingency learning task, in which participants would receive
verbal descriptions of the probabilities involved in the situation.
We noted in the Introduction that the spontaneous tendency to
generate a high P(Cause) could be related to a more general
information sampling bias that some authors have called a positive
testing strategy [35]. For a hypothesis involving a potential cause-
outcome relationship between two events, this strategy may
involve preferentially testing the hypothesis by searching for cases
in which the outcome would occur if the hypothesis were correct
(i.e., situations in which the potential cause is present). A positive
test strategy can be categorized as a form of ‘‘strategy-based error’’
[47], [48], a type of error that is relatively conscious or deliberative
[48], [62]. Because of its deliberative nature, people who fall prey
to this type of error might be especially good candidates for
debiasing when the benefits of accuracy are stressed or when the
stakes for correctness are high enough [47]. Thus, the success of
our debiasing intervention could at least be partially attributable to
our emphasis on the benefits of accurate causal estimation. To
emphasize these benefits, we employed an initial phase in which
the participants were deceived in a manner that could also occur
in their daily lives (i.e., the presentation and demonstration of a
‘‘miracle product’’). This first phase was also intended to help us
avoid a potential problem acknowledged by some authors, such as
Lilienfeld et al. [10], who have stated that participants might not
benefit from debiasing efforts because they have trouble perceiving
their own biases [50] and therefore fail to realize that they need a
remedy. Thus, the initial staging phase was also aimed to increase
the participants’ awareness of how easily causal illusions can
develop. Although we did not formally assess the success of this
specific aspect of the intervention, when the participants were
asked at the end of the session, the majority of them expressed that
they were cheated in the first phase. However, at present, we
cannot elucidate whether the intervention would have been
equally effective if we had simply described the abstract concept of
contingency or, more generally, whether omitting some compo-
nents of the intervention would have affected the results. Future
studies should also explore which features of the intervention are
crucial or contribute to its effectiveness to a greater extent.
To this respect, the reason for using a between-subjects instead
of a within-subjects design (the latter would involve measuring the
same participants’ performance before and after the intervention)
was to unambiguously determine the influence of the intervention
apart from the potential effect of the practice or familiarity with
the same contingency learning task (involving the same contin-
gencies). We consider that this is an appropriate starting point for
the first attempt to debias adolescents against causal illusions.
Nevertheless, future interventions should also include within-
subjects measures of the effectiveness of the intervention (i.e.,
pretest-posttest designs), together with active control groups that
would require subjects to engage in alternative causal inference
tasks different from the intervention presented here.
As far as we know, the present study is the first serious effort to
implement and experimentally evaluate an educational interven-
tion to prevent the formation of causal illusions. We believe that
the potential benefits of this type of intervention are considerable.
Causal illusions may underlie many harmful beliefs that eventually
generate important societal issues (e.g., the use of pseudo-
medicines, racism, economic collapse) and too often guide political
decisions. Unsurprisingly, many governmental and scientific
organizations are now committed to advancing scientific reasoning
and literacy at school [63], [64]. Because our current study was
conducted in a classroom in a school context, we feel confident
that this type of evidence-based intervention could be easily
implemented in an educational program. In addition, the fact that
our evaluation was based on a standardized method used in
experimental psychology makes it an ideal tool for assessing and
comparing different debiasing strategies in different schools.
Indeed, developing a standardized method that can be used for
comparison purposes will be the first step toward a truly evidence-
based debiasing program worldwide.
As stated by Lilienfeld [65], ‘‘…the most important psycholog-
ical experiment never done would […] begin with the construction
of a comprehensive evidence-based educational programme of
debiasing children and adolescents in multiple countries against
malignant biases…’’, and ‘‘…launching such an endeavour by
conducting small-scale pilot studies would seem to be a worthwhile
starting point". To the best of our knowledge, the ‘‘most important
experiment’’ for ‘‘sav[ing] the world’’ (as Lilienfeld phrases it)
remains to be conducted, but we believe that the intervention and
the assessment procedure presented here constitute a promising
advance in this direction.
We thank Pablo Garaizar, Cristina Gime´nez, Andrea Jorge, In˜aki Lecue,
Cristina Orgaz, Nerea Ortega-Castro, Lucı
aOn˜ate, Miguel A. Vadillo,
and Gustavo A. Va´zquez, for their assistance when conducting the
intervention here described.
Author Contributions
Conceived and designed the experiments: IB FB CPC HM. Performed the
experiments: IB FB CPC. Analyzed the data: IB FB. Wrote the paper: IB
1. Moore DW (2005) Three in four Americans believe in paranormal. Princeton:
Gallup News Service. Available:
four-americans-believe-paranormal.aspx. Accessed 2013 Jul 8.
2. Newport F, Strausberg M (2001) Americans’ belief in psychic and paranormal
phenomena is up over last decade. Princeton: Gallup News Service. Available:
phenomena-over-last-decade.aspx. Accessed 2013 Jul 8.
3. Lilienfeld SO, Landfield K (2008) Science and Pseudoscience in law
Enforcement: A User-Friendly Primer. Crim Justice Behav 35: 1215–1230.
4. European Commission (2005) Special Eurobarometer 224: Europeans, science
and technology. Brussels: Author.
5. Shang A, Huwiler-Mu¨ntener K, Nartey L, Ju¨ni P, Do¨rig S, et al. (2005) Are the
clinical effects of homeopathy placebo effects? Comparative study of placebo-
controlled trials of homeopathy and allopathy. The Lancet 366: 726–732.
6. Freckelton I (2012) Death by homeopathy: issues for civil, criminal and coronial
law and for health service policy. J Law Med 19: 454–478.
7. Matute H, Yarritu I, Vadillo MA (2011) Illusions of causality at the heart of
pseudoscience. Brit J Psychol 102: 392–405.
8. Shermer M (1997) Why people believe weird things: Pseudoscience, superstition
and other confusions of our time. New York: W. H. Freeman & Co.
9. Hamilton DL, Gifford RK (1976) Illusory correlation in interpersonal
perception: A cognitive basis of stereotypic judgments. J Exp Soc Psychol 12:
10. Lilienfeld SO, Ammirati R, Landfield K (2009) Giving debiasing away: Can
psychological research on correcting cognitive errors promote human welfare?
Perspect Psychol Sci 4: 390–398.
11. Hume D (1987) A treatise of human nature (2nd ed.). Oxford: Clarendon Press.
(Original work published 1739).
12. Allan LG (1980) A note on measurement of contingency between two binary
variables in judgment tasks. Bull Psychon Soc 15: 147–149.
13. Jenkins HM, Ward WC (1965) Judgment of contingency between responses and
outcomes. Psychol Monogr 79: 1–17.
Debiasing Adolescents against Causal Illusions
PLOS ONE | 8 August 2013 | Volume 8 | Issue 8 | e71303
14. Cheng P (1997) From covariation to causation: A causal power theory. Psychol
Rev 104: 367–405.
15. Shanks DR, Dickinson A (1987) Associative accounts of causality judgment. In
GH Bower (Ed.), The psychology of learning and motivation. San Diego, CA.:
Academic Press. Vol. 21 pp. 229–261.
16. Wasserman EA (1990) Detecting response-outcome relations: Toward an
understanding of the causal texture of the environment. In GH Bower (Ed.),
The psychology of learning and motivation. San Diego, CA: Academic Press.
Vol. 26 pp. 27–82.
17. Wasserman EA, Elek SM, Chatlosh DL, Baker AG (1993) Rating causal
relations: Role of probability in judgments of response-outcome contingency.
J Exp Psychol Learn Mem Cogn 19: 174–188.
18. Alloy LB, Abramson LY (1979) Judgements of contingency in depressed and
nondepressed students: Sadder but wiser? J Exp Psychol Gen 108: 441–485.
19. Buehner MJ, Cheng PW, Clifford D (2003) From covariation to causation: A test
of the assumption of causal power. J Exp Psychol Learn Mem Cogn 29: 1119–
20. Hannah S, Beneteau JL (2009) Just tell me what to do: Bringing back
experimenter control in active contingency tasks with the command-perfor-
mance procedure and finding cue density effects along the way. Can J Exp
Psychol 63: 59–73.
21. Matute H (1995) Human reactions to uncontrollable outcomes: Further
evidence for superstitions rather than helplessness. Q J Exp Psychol 48: 142–157.
22. Msetfi RM, Murphy RA, Simpson J, Kornbrot DE (2005) Depressive realism
and outcome density bias in contingency judgments: the effect of the context and
inter-trial interval. J Exp Psychol Gen 134: 10–22.
23. Valle´e-Tourangeau F, Murphy RA, Baker AG (2005) Contiguity and the
outcome density bias in action-outcome contingency judgements. Q J Exp
Psychol 58: 177–192.
24. Blanco F, Matute H, Vadillo MA (2011) Making the uncontrollable seem
controllable: The role of action in the illusion of control. Q J Exp Psychol 64:
25. Matute H (1996) Illusion of control: Detecting response-outcome independence
in analytic but not in naturalistic conditions. Psychol Sci 7: 289–293.
26. Anderson JR, Sheu C (1995) Causal inferences as perceptual judgments. Mem
Cognit 23: 510–524.
27. Catena A, Maldonado A, Ca´ndido A (1998) The effect of frequency of judgment
and the type of trials on covariation learning. J Exp Psychol Hum Percept
Perform 24: 481–495.
28. Kao S-F, Wasserman EA (1993) Assessment of an information integration
account of contingency judgment with examination of subjective cell importance
and method of information presentation. J Exp Psychol Learn Mem Cogn 19:
29. Mandel DR, Lehman DR (1998) Integration of Contingency Information in
Judgments of Cause, Covariation, and Probability. J Exp Psychol Gen 3: 269–
30. Perales JC, Shanks DR (2007) Models of covariation-based causal judgment: A
review and synthesis. Psychon Bull Rev 14: 577–596.
31. Schustack MW, Sternberg RJ (1981) Evaluation of evidence in causal inference.
J Exp Psychol Gen 110: 101–120.
32. White PA (2003) Making causal judgments from the proportion of confirming
instances: The pCI rule. J Exp Psychol Learn Mem Cogn 29: 710–727.
33. McKenzie CRM, Mikkelsen L A (2007) A Bayesian view of covariation
assessment. Cognitive Psychol 54: 33–61.
34. Blanco F, Matute H, Vadillo MA (2012) Mediating Role of Activity Level in the
Depressive Realism Effect. PLoS One 7(9): e46203.
35. Klayman J, Ha Y-W (1987) Confirmation, disconfirmation, and information in
hypothesis testing. Psychol Rev 94: 211–228.
36. Nickerson RS (1998) Confirmation bias: A ubiquitous phenomenon in many
guises. Rev Gen Psychol 2: 175–220.
37. Fiedler K, Walther E, Nickel S (1999) The Auto-Verification of Social
Hypotheses: Stereotyping and the Power of Sample Size. J Pers Soc Psychol
77: 5–18.
38. Klayman J, Ha Y-W (1989) Hypothesis Testing in Rule Discovery: Strategy,
Structure, and Content. J Exp Psychol Learn Mem Cogn 15: 596–604.
39. Mynatt CR, Doherty ME, Tweney RD (1977) Confirmation bias in a simulated
research environment: An experimental study of scientific inferences. Q J Exp
Psychol 29: 85–95.
40. Snyder M, Swann WB (1978) Hypothesis testing processes in social interaction.
J Pers Soc Psychol 36: 1202–1212.
41. Wason PC (1960) On the failure to eliminate hypotheses in a conceptual task.
Q J Exp Psychol 12: 129–140.
42. McKenzie CRM (2005) Judgment and decision making. In K Lamberts & RL
Goldstone (Eds.), Handbook of cognition. London: Sage. 321–338 pp.
43. McLean CP, Miller NA (2010) Changes in critical thinking following a course on
science and pseudoscience: A quasi-experimental study. Teach Psychol 37: 85–
44. Morier D, Keeports D (1994) Normal science and the paranormal: The effect of
a scientific method course on students’ beliefs in the paranormal. Res High Educ
35: 443–453.
45. Wesp R, Montgomery K (1998) Developing critical thinking through the study
of paranormal phenomena. Teach Psychol 25: 275–278.
46. Bloom P, Weisberg DS (2007) Childhood origins of adult resistance to science.
Science 316: 996–997.
47. Arkes H (1991) Costs and benefits of judgment errors: Implications for debiasing.
Psychol Bull 110: 486–498.
48. Larrick RP (2004) Debiasing. In DJ Koehler & N Harvey (Eds.), Blackwell
handbook of judgment and decision making. Oxford, United Kingdom:
Blackwell. 316–338 pp.
49. Lilienfeld SO, Ammirati R, David M (2012) Distinguishing science from
pseudoscience in school psychology: Science and scientific thinking as safeguards
against human error. J Sch Psychol 50: 7–36.
50. Pronin E, Gilovich T, Ross L (2004) Objectivity in the eye of the beholder:
Divergent perceptions of bias in self versus others. Psychol Rev 111: 781–799.
51. Pronin E, Lin DY, Ross L (2002) The bias blind spot: Perceptions of bias in self
versus others. Pers Soc Psychol Bull 28: 369–381.
52. Harkness AR, DeBono KG, Borgida E (1985) Personal involvement and
strategies for making contingency judgments: A stake in the dating game makes a
difference. J Pers Soc Psycho 49: 22–32.
53. Porcari J, Hazuga R, Foster C, Doberstein S, Becker J, et al. (2011) Can the
power balanceH bracelet improve balance, flexibility, strength, and power?.
J Sports Sci Med 10: 230–231.
54. Brice SR, Jarosz BS, Ames RA, Baglin J, Da Costa C (2011) The effect of close
proximity holographic wristbands on human balance and limits of stability: A
randomized, placebo-controlled trial. J Bodyw Mov Ther 15: 298–303.
55. Fong GT, Krantz DH, Nisbett RE (1986) The effects of statistical training on
thinking about everyday problems, Cogn Psychol 18: 253–92.
56. Baron RM, Kenny DA (1986) The moderator-mediator variable distinction in
social psychological research: Conceptual, strategic and statistical considerations.
J Pers Soc Psychol 51: 1173–1182.
57. Sobel ME (1982) Asymptotic confidence intervals for indirect effects in structural
equation models. In S Leinhardt (Ed.). Sociological Methodology. Washington
DC: American Sociological Association. 290–312 pp.
58. Lehman DR, Nisbett RE (1990) A longitudinal study of the effects of
undergraduate education on reasoning. Dev Psychol 26: 952–60.
59. Lehman DR, Lempert RO, Nisbett RE (1988) The effects of graduate training
on reasoning: Formal discipline and thinking about everyday-life events. Am
Psychol 43: 431–442.
60. Shanks DR (1991) On similarities between causal judgments in experienced and
described situations. Psychol Sci 2: 341–350.
61. Beckers T, De Houwer J, Matute H (2007) Editorial: Human contingency
learning. Q J Exp Psychol 60: 289–290.
62. Kahneman D (2003) A perspective on judgment and choice: Mapping bounded
rationality. Am Psychol 58: 697–720.
63. American Association for the Advancement of Science, AAAS (n/d). Project
2061: A long-term AAAS initiative to advance literacy in Science, Mathematics,
and Technology.Available: 2013 Jul 8.
64. Confederacio´n de Sociedades Cientı
ficas Espan˜olas, COSCE (n/d). Proyecto
ENCIENDE: ENsen˜anza de las CIENcias en la Dida´ctica Escolar.Available: 2013 Jul 8.
65. Lilienfeld SO (2008) Can psychology save the world? British Psychological
Society Research Digest. Available:
2007/09/can-psychology-save-world.html. Accessed 2013 Jul 8.
Debiasing Adolescents against Causal Illusions
PLOS ONE | 9 August 2013 | Volume 8 | Issue 8 | e71303
... The purpose of evaluating hallucinations refers to identifying the psychotic phenotype that would allow estimating and preventing future psychotic crises (e.g., Oliver et al., 2019 ). (2) They have also been used to quantify perceptual distortions or alterations and cognitive biases (e.g., Barberia et al., 2013 ;Barberia et al., 2018 ;Groome et al., 2019 ;Khun et al., 2016 ). This use comes from the semiotic model of perception and from cognitive psychology, which determine a very clear difference with respect to hallucinations: in perceptual distortions, a provocative sensory object exists, but it is perceived in an altered way (e.g., Belloch et al., 1995 ). ...
... This does not contradict the hypothesis of Irwin (2009) , but it does warn that other psychological mechanisms could intervene and generate these differences. Taking into account the studies published by several professionals, it is possible that the variables responsible for these differences are related not only to beliefs but also to cognitive or causal learning mechanisms (e.g., Barberia et al., 2013 ;Barberia et al., 2018 ;Groome et al., 2019 ). ...
Full-text available
Anomalous phenomena are human experiences that are characterized by challenging the foundations of current scientific ontology (i.e., psi phenomena). The problem lies in the fact that some studies have obtained significant results that support the existential validity of psi phenomena. This fact calls into question the role of psychology-and specifically that of psychological assessment-in scientifically justifying and objectively evaluating this type of behavior. This work examines the construct validity and reliability of the Multivariable Multiaxial Suggestibility Inventory-2 (MMSI-2), a psychometric test that measures both anomalous phenomena and the main psychological predictive variables that could generate them. The study included 804 participants without psychiatric history. The participants were evenly distributed into two groups: participants who believe in the existence of the paranor-mal and participants who are non-believers. Confirmatory factor analysis was applied, factorial invariance between both groups was examined, and Cronbach's alpha and Omega reliability coefficients were calculated. The results allowed accepting the 'strong factorial invariance' for the internal structure of the MMSI-2. In parallel, latent means analysis indicated that believers had higher scores than non-believers in the 4 latent variables of the test. Regression models indicated that the Clinical Personality Tendencies (CPT), Incoherent Manipulations (IMA) and Altered States of Consciousness (ASC) scales predicted 51.2% of anomalous phenomena. It is concluded that the MMSI-2, with its 174 items and 20 scales, is a valid and reliable psychometric instrument. This research is a continuation of the Escolà-Gascón (2020) report, in which the first psychometric properties of the MMSI-2 were published.
... In addition, these causal illusions have been explored using procedures that usually include motivational components or additional goals which may affect information gathering and causal inferences. For example, participants may be required to evaluate the effects of a fictitious medicine while at the same time trying to heal as many patients as possible [48,49]. It is not clear whether a causal illusion can be detected in an information sampling setting when motivational components or secondary goals are removed, or when sampling strategies are guided by the effect rather than by the cause. ...
... Laboratory-based research on contingency learning has been shown to be a successful approach to real-life problems because of its ability to detect relevant factors that may contribute to causal inferences, but also because it has been the foundation for designing and testing evidence-based interventions that have proven to be effective in improving critical thinking skills, and therefore, at reducing potentially harmful causal misconceptions in real contexts [48,49]. Future research may extend our results to real-world contexts in order to design interventions aimed at protecting users when using the internet. ...
Background The internet is a relevant source of health-related information. The huge amount of information available on the internet forces users to engage in an active process of information selection. Previous research conducted in the field of experimental psychology showed that information selection itself may promote the development of erroneous beliefs, even if the information collected does not. Objective The aim of this study was to assess the relationship between information searching strategy (ie, which cues are used to guide information retrieval) and causal inferences about health while controlling for the effect of additional information features. Methods We adapted a standard laboratory task that has previously been used in research on contingency learning to mimic an information searching situation. Participants (N=193) were asked to gather information to determine whether a fictitious drug caused an allergic reaction. They collected individual pieces of evidence in order to support or reject the causal relationship between the two events by inspecting individual cases in which the drug was or was not used or in which the allergic reaction appeared or not. Thus, one group (cause group, n=105) was allowed to sample information based on the potential cause, whereas a second group (effect group, n=88) was allowed to sample information based on the effect. Although participants could select which medical records they wanted to check—cases in which the medicine was used or not (in the cause group) or cases in which the effect appeared or not (in the effect group)—they all received similar evidence that indicated the absence of a causal link between the drug and the reaction. After observing 40 cases, they estimated the drug–allergic reaction causal relationship. Results Participants used different strategies for collecting information. In some cases, participants displayed a biased sampling strategy compatible with positive testing, that is, they required a high proportion of evidence in which the drug was administered (in the cause group) or in which the allergic reaction appeared (in the effect group). Biased strategies produced an overrepresentation of certain pieces of evidence at the detriment of the representation of others, which was associated with the accuracy of causal inferences. Thus, how the information was collected (sampling strategy) demonstrated a significant effect on causal inferences (F1,185=32.53, P<.001, η2p=0.15) suggesting that inferences of the causal relationship between events are related to how the information is gathered. Conclusions Mistaken beliefs about health may arise from accurate pieces of information partially because of the way in which information is collected. Patient or person autonomy in gathering health information through the internet, for instance, may contribute to the development of false beliefs from accurate pieces of information because search strategies can be biased.
... The results obtained in the present study show that understanding and predicting beliefs in pseudoscience is a complex process, as recently suggested (Fasce & Pic o, 2019b). Although our results need to be replicated and studied in-depth, the analytical approaches provided in this study could be useful for any type of proposal or intervention aiming at reducing pseudoscience acceptance (Barberia et al., 2013(Barberia et al., , 2018Lilienfeld et al., 2012Lilienfeld et al., , 2014Matute et al., 2015;McLean & Miller, 2010;Wilson, 2018). ...
Full-text available
The spread of pseudoscience (PS) is a worrying problem worldwide. The study of pseudoscience beliefs and their associated predictors have been conducted in the context of isolated pseudoscience topics (e.g., Complementary and alternative medicine). Here, we combined individual differences (IIDD) measures (e.g., Personality traits, thinking styles) with measures related with the information received about PS: familiarity and disproving information (DI) in order to explore potential differences among pseudoscience topics in terms of their associated variables. These topics differed in their familiarity, their belief rating and their associated predictors. Critically, our results not only show that DI is negatively associated with pseudoscience beliefs but that the effect of various IIDD predictors (e.g., Analytic thinking) depends on whether DI had been received. This study highlights the need to control for variables related to information received about pseudoscientific claims to better understand the effect of other predictors on different pseudoscience beliefs topics. This article is protected by copyright. All rights reserved.
... Understanding whether subjective contingency estimation influences pseudoscientific health beliefs has important implications for the strategies we adopt to overcome them. If people's pseudoscientific health beliefs are related to the perceived contingency between events, then a potentially effective strategy for correcting these false beliefs is through scientific education-in particular, using base-rate comparison, that is comparing to the frequency of the outcome without intervention [10], to determine the unique influence of the putative cause (e.g., recovery from the common cold in the absence of Echinacea use). If pseudoscientific health beliefs are not based on perceived contingency, then other factors should predict them, such as proximity to others who hold such beliefs or other personal factors, and therefore strategies such as base-rate instruction may not be effective in correcting pseudoscientific thinking. ...
Full-text available
Beliefs about cause and effect, including health beliefs, are thought to be related to the frequency of the target outcome (e.g., health recovery) occurring when the putative cause is present and when it is absent (treatment administered vs. no treatment); this is known as contingency learning. However, it is unclear whether unvalidated health beliefs, where there is no evidence of cause–effect contingency, are also influenced by the subjective perception of a meaningful contingency between events. In a survey, respondents were asked to judge a range of health beliefs and estimate the probability of the target outcome occurring with and without the putative cause present. Overall, we found evidence that causal beliefs are related to perceived cause–effect contingency. Interestingly, beliefs that were not predicted by perceived contingency were meaningfully related to scores on the paranormal belief scale. These findings suggest heterogeneity in pseudoscientific health beliefs and the need to tailor intervention strategies according to underlying causes.
... Although both perceptual alterations present within the spectrum of psychoses, they differ from hallucinations since they require a sensory triggering object (e.g., Jaspers, 1993;El-Mallakh and Walker, 2010). However, perceptual deception does not usually have a psychopathological origin (e.g., Parker, 2006), which is why they are also called "illusions" or perceptual biases (e.g., Barberia et al., 2013Barberia et al., , 2018. Examples include pareidolia and the Barnum effect and affective (e.g., Belloch et al., 1995;Shermer, 2011). ...
Full-text available
This paper presents the English adaptation of the Multivariable Multiaxial Suggestibility Inventory-2 (MMSI-2), a questionnaire developed specifically for psychological assessment and prediction of anomalous phenomena. The sample consisted of 613 respondents from England (47.6% were women and 52.4% men). All of them were of legal age (mean = 34.5; standard deviation = 8.15). An exploratory factor analysis was applied, and three confirmatory factor models were adjusted. Omega coefficients and test-retest designs were used for reliability analysis. The MMSI-2 has a valid internal structure consisting of five macrofactors: Clinical Personality Tendencies (CPT), Anomalous Perceived Phenomena (APP), Incoherent Manipulations (IMA), Altered States of Consciousness (ASC), and Openness (OP). Omega coefficients for CPT and OP factors were low but acceptable. Furthermore, test-retest trials were excellent for all scales and factors. The psychological factors CPT, IMA, and ASC predicted 18.3% of the variance of anomalous experiences (APP). The authors concluded the English MMSI-2 was a valid and reliable test for the evaluation of anomalous phenomena but recommend that subsequent research reviews the predictive quality of the underlying model.
... Comprender los motivos por los que las personas, después de largos periodos formativos, siguen teniendo visiones distorsionadas de las realidades cotidianas, es un buen comienzo para entender cómo se genera y adquiere el conocimiento. Trabajos previos han mostrado que los sesgos cognitivos pueden reducirse a través de intervenciones educativas que enseñan a escolares técnicas para obtener una visión más realista de lo que les muestran las informaciones recibidas desde los diferentes medios (Barberia, Blanco, Cubillas, & Matute, 2013;Barberia, Tubau, Matute, & Rodríguez-Ferreiro, 2018). Esta idea coincide con los trabajos de Shirky (2010de Shirky ( , 2011 que pone el acento en la dificultad de realizar unos filtrados de la información que sean realmente efectivos. ...
Full-text available
RESUMEN El presente trabajo analiza el nivel de conocimiento que sobre el mundo actual tienen los futuros docentes que cursan estudios de Grado y Postgrado en Educación en la Universidad del País Vasco. En un contexto de hiperconectividad donde comprobar la veracidad de las informaciones se hace cada vez más complejo, es de gran importancia analizar la percepción que el futuro profesorado tiene sobre el estado actual del mundo. Este conocimiento permite proponer medidas de actuación que ayuden a los docentes a comprender cómo es la adquisición y generación de conocimiento, a través de unos medios de comunicación cada vez más digitalizados. Para ello, se ha utilizado el cuestionario ideado por Rosling et. al, que consta de trece preguntas con tres posibles respuestas cada una. Los resultados obtenidos muestran un porcentaje de aciertos del 21%, una media que no llega a los tres aciertos por persona, y una moda que se sitúa en los dos aciertos. Estos datos son muy similares a los obtenidos por las personas de países socio-económicamente avanzados. Igualmente, existe gran coincidencia en los ítems en los que se obtienen mejores y peores niveles de acierto. El paso por los estudios de educación parece no tener incidencia en el desarrollo de esquemas mentales que ayuden a tener visiones más acertadas de la realidad, que puedan prevenir tanto entre el alumnado como el profesorado la prevalencia de creencias y falacias infundadas.
... However, these overestimations of contingency have also been shown responsible for serious health problems, sometimes even death (Freckelton, 2012). Thus, it is important to understand how they work in order to develop interventions that might help increase or reduce their impact (see Barberia, Blanco, Cubillas, & Matute, 2013;Barberia, Tubau, Matute, & Rodríguez-Ferreiro, 2018;MacFarlane et al., 2018). ...
Many experiments have shown that humans and other animals can detect contingency between events accurately. This learning is used to make predictions and to infer causal relationships, both of which are critical for survival. Under certain conditions, however, people tend to overestimate a null contingency. We argue that a successful theory of contingency learning should explain both results. The main purpose of the present review is to assess whether cue-outcome associations might provide the common underlying mechanism that would allow us to explain both accurate and biased contingency learning. In addition, we discuss whether associations can also account for causal learning. After providing a brief description on both accurate and biased contingency judgments, we elaborate on the main predictions of associative models and describe some supporting evidence. Then, we discuss a number of findings in the literature that, although conducted with a different purpose and in different areas of research, can also be regarded as supportive of the associative framework. Finally, we discuss some problems with the associative view and discuss some alternative proposals as well as some of the areas of current debate. (PsycINFO Database Record (c) 2019 APA, all rights reserved).
... On the other hand, certain types of biases, including the causal illusion, seem to be associated to optimism and positive attitude: for example, causal illusions are apparently weaker in depressed people [61,62], although see [63,64]. Therefore, it would be desirable to design ways to bring the causal illusion under our control, to prevent it when it leads to harmful behaviors [65], and to promote it when it fosters wellbeing [66]. In this regard, our two experiments illustrate a means to either prevent causal illusions (Experiment 1) or to enhance them (Experiment 2) by shaping the assumptions that people make before collecting the contingency information. ...
Full-text available
Previous research revealed that people’s judgments of causality between a target cause and an outcome in null contingency settings can be biased by various factors, leading to causal illusions (i.e., incorrectly reporting a causal relationship where there is none). In two experiments, we examined whether this causal illusion is sensitive to prior expectations about base-rates. Thus, we pretrained participants to expect either a high outcome base-rate (Experiment 1) or a low outcome base-rate (Experiment 2). This pretraining was followed by a standard contingency task in which the target cause and the outcome were not contingent with each other (i.e., there was no causal relation between them). Subsequent causal judgments were affected by the pretraining: When the outcome base-rate was expected to be high, the causal illusion was reduced, and the opposite was observed when the outcome base-rate was expected to be low. The results are discussed in the light of several explanatory accounts (associative and computational). A rational account of contingency learning based on the evidential value of information can predict our findings.
Background We have previously presented two educational interventions aimed to diminish causal illusions and promote critical thinking. In both cases, these interventions reduced causal illusions developed in response to active contingency learning tasks, in which participants were able to decide whether to introduce the potential cause in each of the learning trials. The reduction of causal judgments appeared to be influenced by differences in the frequency with which the participants decided to apply the potential cause, hence indicating that the intervention affected their information sampling strategies. Objective In the present study, we investigated whether one of these interventions also reduces causal illusions when covariation information is acquired passively. Method Forty-one psychology undergraduates received our debiasing intervention, while 31 students were assigned to a control condition. All participants completed a passive contingency learning task. Results We found weaker causal illusions in students that participated in the debiasing intervention, compared to the control group. Conclusion The intervention affects not only the way the participants look for new evidence, but also the way they interpret given information. Teaching implications Our data extending previous results regarding evidence-based educational interventions aimed to promote critical thinking to situations in which we act as mere observers.
Objective: Fraudulent health claims-false or misleading claims used to promote health remedies that are untested, ineffective, and often harmful-cause extensive and persistent harm to consumers. To address this problem, novel interventions are needed that address the underlying cognitive mechanisms that render consumers susceptible to fraudulent health claims. However, there is currently no single framework of relevant psychological insights to design interventions for this purpose. The current review aims to address this gap. Method: An integrative theoretical review was conducted across several relevant disciplines including criminology; behavioural economics; and cognitive, health, and social psychology. Results: The current review presents a novel taxonomy that aims to serve as an agenda for future research to systematically design and compare interventions based on empirical evidence. Specifically, this taxonomy identifies (i) the psychological drivers that make consumers susceptible to fraudulent health claims, (ii) the psychological barriers that may prevent successful application of interventions, and (iii) proposes evidence-informed treatments to overcome those barriers. Conclusions: The resulting framework integrates behavioural insights from several hitherto distinct disciplines and structures promising interventions according to five underlying psychological drivers: Visceral influence, Affect, Nescience, Misinformation, and Norms (VANMaN). The taxonomy presents an integrative and accessible theoretical framework for designing evidence-informed interventions to protect consumers from fraudulent health claims. This review has broad implications for numerous topical issues including the design and evaluation of anti-fraud campaigns, efforts to address the growing problem of health-related misinformation, and for countering the polarisation of politically sensitive health issues.
Full-text available
This study showed that accuracy of the estimated relationship between a fictitious symptom and a disease depends on the interaction between the frequency of judgment and the last trial type. This effect appeared both in positive and zero contingencies (Experiment 1), and judgments were less accurate as frequency increased (Experiment 2). The effect can be explained neither by interference of previous judgments or memory demands (Experiment 3), nor by the perceptual characteristics of the stimuli (Experiments 4 and 5), and instructions intended to alter processing strategies do not produce any reliable effect. The interaction between frequency and trial type on covariation judgment is not predicted by any model (either statistical or associative) currently used to explain performance in covariation detection. The authors propose a belief-revision model to explain this effect as an important response mode variable on covariation learning.
Full-text available
In this article, we attempt to distinguish between the properties of moderator and mediator variables at a number of levels. First, we seek to make theorists and researchers aware of the importance of not using the terms moderator and mediator interchangeably by carefully elaborating, both conceptually and strategically, the many ways in which moderators and mediators differ. We then go beyond this largely pedagogical function and delineate the conceptual and strategic implications of making use of such distinctions with regard to a wide range of phenomena, including control and stress, attitudes, and personality traits. We also provide a specific compendium of analytic procedures appropriate for making the most effective use of the moderator and mediator distinction, both separately and in terms of a broader causal system that includes both moderators and mediators. (46 ref) (PsycINFO Database Record (c) 2012 APA, all rights reserved)
Full-text available
Confirmation bias, as the term is typically used in the psychological literature, connotes the seeking or interpreting of evidence in ways that are partial to existing beliefs, expectations, or a hypothesis in hand. The author reviews evidence of such a bias in a variety of guises and gives examples of its operation in several practical contexts. Possible explanations are considered, and the question of its utility or disutility is discussed.
The theory of formal discipline-that is, the view that instruction in abstract rule systems can affect reasoning about everyday-life events-has been rejected by 20th century psychologists on the basis of rather scant evidence. We examined the effects of graduate training in law, medicine, psychology, and chemistry on statistical reasoning, methodological reasoning about confounded variables, and reasoning about problems in the logic of the conditional. Both psychology and medical training produced large effects on statistical and methodological reasoning, and psychology, medical, and law training produced effects on ability to reason about problems in the logic of the conditional. Chemistry training had no effect on any type of reasoning studied. These results seem well understood in terms of the rule systems taught by the various fields and indicate that a version of the formal discipline hypothesis is correct.