She Told Me About a Singing Cactus: Counterintuitive Concepts Are More Accurately Attributed to Their Speakers Than Ordinary Concepts

Abstract

Communication is central to human life, yet it leaves humans vulnerable to misinformation and manipulation. Humans have therefore evolved a suite of psychological mechanisms for the evaluation of speakers and their messages. Here, we test a key hypothesized function of these "epistemic vigilance" mechanisms: the selective remembering of links between speakers and messages that are inconsistent with preexisting beliefs. Across four experiments, participants (N = 707) read stories associated with different contexts, with each story containing concepts that violate core knowledge intuitions ("counterintuitive concepts") and ordinary concepts. Exp. 1 revealed that after a brief distractor (2 minutes) participants more accurately attributed counterintuitive concepts to their speakers than ordinary concepts. Exp. 2a-b replicated this finding and found that this attribution accuracy advantage also extended to counterintuitive versus ordinary concepts associated with other contextual details-places and dates. Exp. 3 then tested whether this attribution accuracy advantage was more stable over time for speakers than for places. After a short distractor (20-minutes), there was a counterintuitive versus ordinary concept attribution accuracy advantage for both speakers and places. However, when participants were tested again after a long delay (48-hours), this attribution accuracy advantage more than doubled for speakers but disappeared entirely for places. We discuss the implications of these findings to the set of psychological mechanisms theorized to monitor and evaluate communication to guard our database of beliefs about the world.

Full text

SPEAKERS OF COUNTERINTUITIVE CONCEPTS 1
She told me about a singing cactus: Counterintuitive concepts are more accurately attributed to
their speakers than ordinary concepts
Spencer Mermelstein *
Department of Psychological & Brain Sciences
University of California, Santa Barbara
Michael Barlev
Department of Psychology
Arizona State University
&
Tamsin C. German
Department of Psychological & Brain Sciences
University of California, Santa Barbara
Main text word count: 6500
* Correspondence concerning this article should be addressed to Spencer Mermelstein,
Department of Psychological and Brain Sciences, University of California, Santa Barbara, Santa
Barbara, CA 93106. Email: mermelstein@psych.ucsb.edu
All data and materials have been made publicly available via the Open Science
Framework and can be accessed at https://osf.io/x5k2u/.
Previous drafts of this article are available online at https://psyarxiv.com/6cp8e. The data
reported here were presented in a talk at the 2019 meeting of the Human Behavior and Evolution
Society in Boston, MA.
© 2020, American Psychological Association. This paper is not the copy of record
and may not exactly replicate the final, authoritative version of the article published in the
Journal of Experimental Psychology: General. Please do not copy without authors'
permission. The final article will be available, upon publication, via its DOI:
10.1037/xge0000987

SPEAKERS OF COUNTERINTUITIVE CONCEPTS 2
Abstract
Communication is central to human life, yet it leaves humans vulnerable to misinformation and
manipulation. Humans have therefore evolved a suite of psychological mechanisms for the
evaluation of speakers and their messages. Here, we test a key hypothesized function of these
“epistemic vigilance” mechanisms: the selective remembering of links between speakers and
messages that are inconsistent with preexisting beliefs. Across four experiments, participants (N
= 707) read stories associated with different contexts, with each story containing concepts that
violate core knowledge intuitions (“counterintuitive concepts”) and ordinary concepts. Exp. 1
revealed that after a brief distractor (2 minutes) participants more accurately attributed
counterintuitive concepts to their speakers than ordinary concepts. Exp. 2a-b replicated this
finding and found that this attribution accuracy advantage also extended to counterintuitive
versus ordinary concepts associated with other contextual details – places and dates. Exp. 3 then
tested whether this attribution accuracy advantage was more stable over time for speakers than
for places. After a short distractor (20-minutes), there was a counterintuitive versus ordinary
concept attribution accuracy advantage for both speakers and places. However, when participants
were tested again after a long delay (48-hours), this attribution accuracy advantage more than
doubled for speakers but disappeared entirely for places. We discuss the implications of these
findings to the set of psychological mechanisms theorized to monitor and evaluate
communication to guard our database of beliefs about the world.
Keywords: epistemic vigilance, communication, counterintuitive concepts,
misinformation, source memory

SPEAKERS OF COUNTERINTUITIVE CONCEPTS 3
1. Introduction
Communication is central to human life. Human social living, from coordinating
collective action to negotiating reciprocal exchange to social learning, is made possible by
communication. Indeed, by allowing access to information stored in the minds of others,
communication is an engine behind the evolution of cumulative cultural adaptations that cannot
be discovered individually, from the processing of toxic plants for safe consumption or the
Inuit’s cold weather clothing (Henrich & McElreath, 2003), to the printing press, electricity, or
semiconductors. Moreover, in human evolutionary history, communication obviated the need for
individual trial-and-error learning in domains where errors can be catastrophic, such as learning
which animals are dangerous (Barrett & Broesch, 2012) or which plants are edible (Wertz &
Wynn, 2014). In short, communication is a key human adaptation that underlies our species’
success across diverse ecologies.
However, because people vary in both their competence and trustworthiness, relying on
communication opens listeners up to being misinformed or deceived (Sperber et al., 2010).
Humans have therefore evolved a suite of psychological adaptations for “epistemic vigilance” or
the evaluation of speakers and their messages (Sperber et al., 2010). Collectively, epistemic
vigilance mechanisms guard our database of beliefs about the world.
The origins of such psychological adaptations have been observed in young children. As
early as 2 years of age, toddlers not only update their beliefs in light of testimony from adults
(Harris & Lane, 2014), but they also show sensitivity to the logical structure of arguments. For
instance, toddlers are less likely to accept an argument backed by circular logic (e.g., “It’s a fish,
because I saw that it is a fish”) compared to one based on evidence (e.g., “It’s a fish, because I

SPEAKERS OF COUNTERINTUITIVE CONCEPTS 4
saw it swimming in the water”; Castelain, Bernard, & Mercier, 2018). By 5 years of age,
children adjust their acceptance of messages on the basis of characteristics of speakers such as
whether their previous testimony turned out to be true or false (e.g., Jaswal & Neely, 2006;
Koenig & Harris, 2005), whether they were nice or mean to others in the past (e.g., Landrum,
Mills, & Johnston, 2013; Mascaro & Sperber, 2009), and whether their previous testimony
conformed with or dissented from a group consensus (e.g., Corriveau, Fusaro, & Harris, 2009;
for a review see Harris, Koenig, Corriveau, & Jaswal, 2018).
The study of epistemic vigilance mechanisms takes on particular urgency in the present
day as “fake news”, political disinformation, and conspiracy theories proliferate on online
platforms and elsewhere in a heretofore unprecedented scale (Lazer et al., 2018). For instance,
long-impugned claims about a link between vaccinations and autism spectrum disorders fuel an
“Anti-Vax” movement responsible for a worldwide re-emergence of life-threatening infectious
diseases (e.g., Larson, Cooper, Eskola, Katz, & Ratzan, 2011; Poland & Spier, 2010), and
misinformation about anthropocentric global climate change reduces public support for
mitigation efforts (e.g., Cook, Ellerton, & Kinkead, 2017; van der Linden, Leiserowitz,
Rosenthal, & Maibach, 2017). A more thorough understanding of how epistemic vigilance
mechanisms function could inform efforts to combat the proliferation and impact of such
messages.
Here, we test experimentally a key hypothesized function of epistemic vigilance
mechanisms: the selective remembering of links between speakers and messages that are
inconsistent with preexisting beliefs (Sperber, 1997; Sperber et al., 2010; see also Mercier,
2017). By linking messages that violate preexisting beliefs to their speakers, listeners may
continue evaluating these messages in light of new information about their speakers, as well as

SPEAKERS OF COUNTERINTUITIVE CONCEPTS 5
update their judgements about the speakers given new information about their messages. As an
example, consider a scientist who makes a surprising claim about the dangers of a new vaccine
that you believed to be safe. You view this scientist as trustworthy and competent, so you
tentatively accept his claim. But your epistemic vigilance mechanisms might link the claim (“this
new vaccine is dangerous”) with its speaker (this scientist) for further evaluation of both the
claim and the speaker. Should, in the future, factual errors with this claim be found, you may re-
evaluate its truth value (you might now reject the claim that this new vaccine is dangerous) as
well as update your judgement about its speaker (you might now view the scientist as less
competent and/or trustworthy). On this account, we expect a particularly robust link between
speakers and, not everything they say, but specifically their messages that are inconsistent with
preexisting beliefs.
As a secondary prediction, we test whether epistemic vigilance mechanisms monitor
additional contextual details, or “meta-data”, surrounding the acquisition of messages that violate
prior beliefs (Cosmides & Tooby, 2000; Johnson, Hashtroudi, and Lindsay, 1993; Mahr &
Csibra, 2018). For instance, Mahr and Csibra (2018) recently articulated a functional view of
episodic memory wherein social interactions, in particular communicative exchanges, are
remembered along with a set of contextual details such as the social background of the
interaction (e.g., whether it happened in front of a group), when it happened, including relative to
other events, and where it happened. Memory of such contextual details may further facilitate the
ongoing evaluation of messages that violate preexisting beliefs. Returning to the above example,
should we learn that the scientist who made this claim was at the time on the payroll of a rival
vaccine company, we might doubt the accuracy of his claim more so than if the scientist had
started working for this rival company a while after he made this claim.

SPEAKERS OF COUNTERINTUITIVE CONCEPTS 6
However, two considerations suggest that links between messages that violate prior
beliefs and their speakers, more so than links with other contextual details, should be of
particular relevance to epistemic vigilance mechanisms. First, the truth value assigned to a
message greatly depends on information about its speaker, generally more so than on other
contextual details like the place or time of communication. For example, whether a message is
accepted or rejected can entirely depend on whether its speaker is trustworthy or not. Second,
messages reveal important information about their speakers such that linking messages to their
speakers also allows listeners to update their judgement of these speakers should new
information about their messages come to light. Thus, links between messages that violate
preexisting beliefs and their speakers may be especially memorable as compared to such links
with other contextual details.
We chose “counterintuitive concepts” as our test case of messages that violate preexisting
beliefs. Counterintuitive concepts violate intuitions such as about folk physics, biology, and
psychology (so-called “core knowledge” intuitions). For example, beliefs about people that can
walk through walls violate intuitions about the solidity and spatio-temporal continuity of bodies
(Boyer, 2001). As core knowledge intuitions reliably develop and are universally held (e.g.,
Carey, 2009), counterintuitive concepts are one class of communicated information that should
be flagged by the epistemic vigilance mechanisms of listeners broadly as requiring further
monitoring and evaluation.
In sum, we predicted better memory for the links between speakers, and potentially also
other associated contextual details, and counterintuitive concepts as compared to ordinary
concepts (those concepts consistent with prior beliefs).

SPEAKERS OF COUNTERINTUITIVE CONCEPTS 7
1.1. Previous research
Two literatures inform the present investigation. First, an extensive literature finds
memory advantages for information that is inconsistent with preexisting beliefs (Hunt, 1995; von
Restorff, 1933), such as information that violates stereotypes about social groups (e.g., Stangor
& McMillian, 1992), schematic expectations (e.g., Hirshman, Whelley, & Palij, 1989; McDaniel
& Einstein, 1986), and core knowledge intuitions (e.g., Banerjee, Haque, & Spelke, 2013; Barrett
& Nyhof, 2001; Boyer & Ramble, 2001; Norenzayan, Atran, Faulkner, & Schaller, 2006).
Erdfelder and Bredenkamp (1998) suggest that belief- or expectation-violating information
differentially recruits attention, and as such undergoes more elaborate encoding that facilitates its
later retrieval.
Second, studies from the literature on source memory find that schematic expectations or
stereotypes bias memory judgements about the speakers or other contextual details associated
with particular messages (Bayen, Nakamura, Dupuis, & Yang, 2000; Kuhlmann, Vaterrodt, &
Bayen, 2012; Marsh, Cook, & Hicks, 2006; Mather, Johnson, & De Leonardis, 1999). For
example, Bayen et al. (2000) found that utterances characteristic of medical doctors (e.g., “We
are ready to run some tests”), yet spoken by a lawyer, were later misattributed to a doctor;
Mather et al. (1999) found that utterances characteristic of Democrats (e.g., “I am pro-choice”),
yet spoken by a Republican, were later misidentified as having been spoken by a Democrat. It
has been suggested that such misattributions are a result of schema-based guessing biases: when
participants cannot remember the speaker or other contextual details of a particular message,
they select those that are schematically most likely to have been associated with it (e.g.,
Kuhlmann et al., 2012).

SPEAKERS OF COUNTERINTUITIVE CONCEPTS 8
Nonetheless, other studies find that stimuli and the contextual details associated with
them are better remembered when the stimuli are paired with an unexpected versus an expected
context (Bell, Buchner, Kroneisen, & Giang, 2012; Ehrenberg & Klauer, 2005; Küppers &
Bayen, 2014). For example, Küppers and Bayen (2014) presented participants with a word
describing a particular location (e.g., “kitchen” or “bathroom”) followed by items that were
either schematically expected or unexpected of that location (e.g., “oven” or “toothbrush”).
During a later memory task, participants were presented with the previously shown items and
were asked to identify the location each item was paired with. Participants in this study were
better at recalling locations that were unexpected for the items (e.g., “toothbrush” paired with
“kitchen”) compared to those that were expected for the items (e.g., “oven” paired with
“kitchen”), which suggests that a violation of an expectation about the context with which an
item is typically associated may enhance memory for that item-context pair.
Although previous studies have investigated memory for stimuli that violate prior beliefs
(e.g., Boyer & Ramble, 2001) and memory for stimuli and their associated contexts when the
pairing violates expectations (e.g., a toothbrush paired with a kitchen context; Küppers & Bayen,
2014), the present study is the first to explore memory for links between stimuli that by
themselves violate preexisting beliefs and their associated contexts. With such a design we test a
key hypothesized function of epistemic vigilance mechanisms concerning the “meta-data” (such
as the associated speakers, places, and times) stored along with messages that violate preexisting
beliefs, independent of any expectations about links between such messages and their speakers or
when or where the information was communicated.
1.2. Counterintuitive concepts

SPEAKERS OF COUNTERINTUITIVE CONCEPTS 9
The human conceptual repertoire is founded in part on species-typical, reliably
developing core knowledge mechanisms that are specialized for representing concepts from
domains such as physical objects and their spatio-temporal properties and mechanics (“folk
physics”), human-made artifacts including tools, animals and their biology (“folk biology”),
plants, and persons and their mental states (“folk psychology”) (e.g., Baillargeon, Scott, & Bian,
2016; Barrett et al., 2013; Carey, 2009; German & Barrett, 2005; Inagaki & Hatano, 2002;
Spelke, 1990; Spelke & Kinzler, 2007; Wertz, 2019). For example, infants understand that
objects are cohesive and bounded wholes that neither separate nor coalesce, and that objects only
move on contact (Baillargeon, 2004; Spelke, Breinlinger, Macomber, & Jacobson, 1992). Infants
also interpret and predict the behavior of persons in terms of internal mental states, understand
that beliefs are linked to perceptions, and that people can have beliefs that are false (Baillargeon
et al., 2016; Onishi & Baillargeon, 2005).
The human mind is also capable of representing concepts that violate core knowledge
intuitions – indeed, such “counterintuitive” concepts are widespread in science (Shtulman, 2017)
and religion (Boyer, 1994, 2001, 2003). For example, the concept of heritable genetic mutations,
a fundamental principle of the theory of evolution by natural selection, violates folk biological
intuitions about the immutability of animal “essences”; the concept of a statue that can hear
prayers violates folk physical intuitions by transferring a psychological property to a human-
made artifact. Although counterintuitive concepts violate universal intuitions, people nonetheless
come to believe in many such concepts and may even hold in high esteem those with expertise
about these concepts (e.g., scientists and religious specialists).

SPEAKERS OF COUNTERINTUITIVE CONCEPTS 10
Barlev and colleagues (Barlev, Mermelstein, & German, 2017, 2018; Barlev,
Mermelstein, Cohen, & German, 2019) recently presented empirical evidence that even though
counterintuitive concepts are widely believed in, they cannot be fully reconciled with the core
knowledge intuitions with which they conflict (also see Barrett, 1998; Barrett & Keil, 1996). The
God concept in Christianity, for example, is initially built by co-opting the person “template,” a
set of core intuitions about the physical, biological, and psychological properties of people.
Accordingly, young children reason that God is capable of having beliefs that are false just like
persons, and it is only later that children come to view God (but not ordinary people) as infallible
(Lane, Wellman, & Evans, 2010). Barlev and colleagues used a statement verification task where
adult Christian religious adherents evaluated as “true” or “false” statements that were consistent
or inconsistent between core intuitions about persons and acquired theology about God. As
predicted, participants were slower and less accurate at verifying inconsistent statements as
compared to consistent statements, suggesting that core knowledge intuitions about the
psychology (Barlev et al., 2017, 2018) and physicality (Barlev et al., 2019) of persons coexist
and interfere with acquired beliefs about God (e.g., infallibility).
Thus, counterintuitive concepts are an ideal case for testing predictions about the
functioning of epistemic vigilance mechanisms: because counterintuitive concepts violate, and
cannot be reconciled with, universally held core knowledge intuitions, they should be flagged by
the epistemic vigilance mechanisms of listeners broadly as warranting further monitoring and
evaluation.
1.3. The current study

SPEAKERS OF COUNTERINTUITIVE CONCEPTS 11
Across four experiments, participants read a series of short stories, with each story
containing counterintuitive and ordinary concepts, and each story transmitted by different
persons (Exp. 1-3) or at different places (Exp. 2a, 3) or on different dates (Exp. 2b). Then,
following a delay, participants were asked to attribute each concept to its associated context.
Given our goal of investigating memory for links between messages that violate prior beliefs and
the context of their acquisition, it was critical that we presented each speaker, place, or date with
an equal number of counterintuitive and ordinary concepts. Without this design feature,
attributions made during the task might be regulated not by remembered links between specific
concepts and their speakers, but by general associations formed between some speakers with
counterintuitive concepts and other speakers with ordinary concepts.
Exp. 1 tested the prediction that counterintuitive concepts are more accurately attributed
to their speakers than ordinary concepts. Exp. 2a-b replicated this and tested whether
counterintuitive concepts associated with different contextual details, places (2a), and times (2b),
also exhibit a counterintuitive versus ordinary concepts attribution accuracy advantage. Lastly,
Exp. 3 used longer periods of delay, with a first attribution phase after a 20-minute delay and a
second attribution phase after a 48-hour delay, to examine the relative stability of the links
between concepts and their associated contextual details. We predicted that counterintuitive
versus ordinary concepts would exhibit an attribution accuracy advantage, and that this effect
would be more stable over time for speakers than for other contextual details.
2. Experiment 1

SPEAKERS OF COUNTERINTUITIVE CONCEPTS 12
The goal of Exp. 1 was to test whether counterintuitive concepts are more accurately
attributed to their speakers than ordinary concepts.
2.1. Method
2.1.1. Participants
A-priori power analyses were computed for all experiments reported here (see
Supplementary Materials). Participants (N = 107; 66% female) were undergraduates at the
University of California, Santa Barbara (UCSB) (Mage = 19.4; SD = 2.27), who in this and all
other experiments reported here received course credit for their participation. Participants
identified as East, South, or Southeast Asian (35%), White (32%), Hispanic or Latino (22%), or
as another ethnic/racial background (11%). All experiments in this article were approved by
UCSB’s IRB (protocol #23-18-0027) and informed consent was obtained from all participants.
2.1.2. Design
The independent variable was Concept (Counterintuitive vs. Ordinary), presented within-
subjects. The dependent variables were the proportion of counterintuitive and ordinary concepts
correctly attributed to their speaker.
2.1.3. Materials and procedure

SPEAKERS OF COUNTERINTUITIVE CONCEPTS 13
Materials were adapted from Banerjee et al. (2013) and consist of four 340-word stories,
each associated with a different speaker, and each containing three counterintuitive and three
ordinary concepts (for a total of 24 concepts across the four stories). The concepts were created
as follows. Three pairs of nouns (e.g., Cat / Dog) were generated in each of the following
domains: animals, plants, non-living natural objects, and human-made artifacts. Each noun was
embedded in a descriptor composed of two adjectival clauses: a first clause that is consistent
with the domain and a second clause that is either also consistent (forming an ordinary concept)
or contains a violation of a physical, biological, or psychological core knowledge intuition held
about the domain (forming a counterintuitive concept). For example, the noun “Cat” was paired
with either the ordinary descriptor “had soft fur and liked to play with toys” or the
counterintuitive descriptor “had brown spots and could walk through solid walls” (a violation of
intuitive physics). The two variants of each concept (Cat/Dog + ordinary descriptor and Cat/Dog
+ counterintuitive descriptor) were controlled for number of words per sentence and were
balanced in terms of overall sentence structure and complexity. See Table 1 for sample concepts.
See Supplemental Materials for all concepts.

SPEAKERS OF COUNTERINTUITIVE CONCEPTS 14
Two lists of concept stimuli were created by varying which descriptor (counterintuitive or
ordinary) was linked with which noun in a pair. For example, in list 1 “Cat” was paired with the
counterintuitive descriptor (and “Dog” was paired with the ordinary descriptor) whereas in list 2
“Cat” was paired with the ordinary descriptor (and “Dog” was paired with the counterintuitive
descriptor). Participants were randomly assigned one of the two concept stimuli lists such that,
between lists, the descriptors remained fixed but the noun that they were paired with was varied.
In this way, we could verify that attribution accuracy was a function of the type of descriptor
(counterintuitive or ordinary), rather than a property of particular noun-descriptor pairings.
Participants were asked to imagine that they frequently go camping with four close
friends named Miguel, Joanna, Sam, and Ariel, and that during one of these trips, each friend
took a turn telling the participant one of the four short stories. Critically, to prevent participants
from broadly associating certain types of concepts to certain speakers, three ordinary and three
Table 1
Example counterintuitive and ordinary concepts
Noun Pairs Domain
Cat / Dog Animal
Shrub / Cactus Plant
Branch / Rock
Object
Table / Chair
Artifact
Noun Pairs Domain
Cat / Dog Animal
Shrub / Cactus Plant
Branch / Rock
Object
Table / Chair
Artifact
that is thick and hard and looks shiny in the sunlight
that is firm to the touch and can hold lots of weight
Note. Counterintuitive descriptors contain violations of core knowledge intuitions. Concepts are modified from Banerjee et al. (2013).
that is big and often floats in midair
Ordinary Descriptor
that has soft fur and likes to play with toys
that is dark green and is growing next to a stream
Counterintuitive Descriptor
that has brown spots and can walk through solid walls
that is small in size and likes to sing loudly
that feels cold to the touch and can speak in French

SPEAKERS OF COUNTERINTUITIVE CONCEPTS 15
counterintuitive concepts were randomly distributed throughout the middle of each story, such
that each friend was associated with an equal number of both types of concepts.
Finally, participants were randomly assigned to receive one of four different versions of
the task, created by varying which person was associated with which story, such that each person
was associated with each story across the different versions. Task versions 1 and 2 used stimuli
list 1 and task versions 3 and 4 used stimuli list 2. See below for an example of one of the short
stories and Supplementary Materials for all stories.
[Miguel / Joanna / Sam / Ariel] tells you the following story:
A brother and a sister moved with their parents to a new house on a new street
that they had never seen before. The new house was in a neighborhood several
miles away from where they used to live. The brother and sister were excited to
explore their new home and to learn more about the neighborhood. As soon as
their boxes were unpacked, the brother and sister decided to go see what they
could find in and around their new home.
First, they climbed up a staircase and went into the attic, where they saw a lizard
on the floor. This was a lizard that had a long, thin tail and could never die no
matter what happened to it. The kids left the attic and wandered to their parent’s
bedroom. In the bedroom, they saw a hammer lying on the carpet. The hammer
had a wooden handle and needed food every day to stay strong. After leaving the
bedroom, the kids continued on into the basement, where they noticed a shovel on
top of a table. The shovel felt heavy to hold and was a light brown in color.
Growing bored of the house, the kids went outdoors into their new backyard.
They looked up and saw a rainbow. This rainbow was high in the sky and could
be seen from the ground. The kids skipped down the street and came across a
garden that had a single rose in it. The rose swayed in the wind and could be in
two different parts of the world at the exact same time. The kids finally reached
the front yard of their closest neighbor’s house. On the lawn, the kids spotted a
rat. The rat ate insects off the ground and moved around quickly on all four of its
feet.
Satisfied with what they had seen, the kids went back inside thinking that their
new home was going to be a very interesting place to live.

SPEAKERS OF COUNTERINTUITIVE CONCEPTS 16
Participants were tested in groups of up to 8 in semi-private computer workstations.
Qualtrics software was used to administer all experiments. Data were analyzed using R 3.5.1 and
JASP 0.9. Qualtrics scripts, data, and R code are available at https://osf.io/x5k2u/. Participants
were instructed to “pay particularly careful attention to the person who is telling you the story
and what happens in the story” and that they would “need to remember this information for a
memory test that will occur later in the study.” During the encoding phase, the stories were
presented one at a time and in a random order. Each story was “locked” on the screen for 90
seconds (estimated as the average reading time across the four stories), after which participants
were allowed to continue whenever they were ready; this was done to make sure participants did
not speed through the stories. After reading each story, as a check that they have read that story
and to verify that they encoded the person associated with the story, participants were asked to
identify the friend who told them that story in a forced choice question. During the distractor
phase – lasting 2 minutes in this experiment – participants were shown a blank map of the
United States and were asked to type the names of as many states as they could. Last, during the
attribution phase, participants were presented with the 24 concepts they read during the
encoding phase, one at a time and in randomized order, along with the names of the four friends
with whom the concepts were associated. Participants were instructed to “identify, as accurately
as possible, which of your friends was the one who told you each statement.” The entire study
took approximately 20 minutes. Fig. 1 summarizes this procedure.

SPEAKERS OF COUNTERINTUITIVE CONCEPTS 17
Fig 1. Summary of the procedure used in Exp. 1-3. Participants read four 340-word stories, each
containing three counterintuitive and three ordinary concepts, and each associated with a
different speaker or other contextual information (places or dates). After reading each story,
participants completed an attention check to verify they read and remembered the speaker or
other contextual information. In Exp. 1 and 2a-b there was a distractor phase lasting 2 minutes
before the attribution phase, where participants were asked to attribute each concept to the
speaker or context with which it was associated. In Exp. 3 there were two attribution phases, one
after a distractor phase lasting 20 minutes, and another after a 48 hours delay.
2.2. Results
In this and all other experiments reported in this article there were no statistically
significant differences between stimuli lists or task versions. A paired-samples t-test revealed, as
predicted, that counterintuitive concepts were more accurately attributed to their speakers than
ordinary concepts, t(106) = 5.05, p < .001, d = 0.49, 95% CI = [0.29, 0.69]. See Fig. 2 for a
pirate plot.
Time
ATTRIBUTION
PHASE
DISTRACTOR
PHASE
ENCODING
PHASE
STORY ATTENTION
CHECK X 4
X 24 trials

SPEAKERS OF COUNTERINTUITIVE CONCEPTS 18
Fig 2. Pirate plot of mean attribution accuracy (%) for counterintuitive and ordinary concepts in
Exp. 1. Inference bands correspond to 95% within-subjects CIs. The dotted line at 25% indicates
chance performance.
3. Experiment 2a
As predicted, Exp. 1 found that after a brief delay counterintuitive concepts were more
accurately attributed to their speakers than ordinary concepts. The goal of Exp. 2a-b was to
investigate whether other contextual details also show an attribution accuracy advantage for
counterintuitive versus ordinary concepts. In doing so, we tested between two alternative
possibilities of what “meta-data” is linked to messages that violate preexisting beliefs. As we

SPEAKERS OF COUNTERINTUITIVE CONCEPTS 19
argued above, links between messages that violate preexisting beliefs and their speakers are
plausibly more relevant to epistemic vigilance mechanisms than links between such messages
and other contextual details. Thus, one possibility is that the attribution accuracy advantage for
contextual details like places and dates would be smaller as compared to persons. Alternatively,
it is nonetheless possible that a broad variety of meta-data remains linked to messages that
violate preexisting beliefs. On this account, after a brief delay, speakers and contextual details
such as where or when a message was acquired will show a similar counterintuitive versus
ordinary concepts attribution accuracy advantage. Exp. 2a-b tested between these two accounts
by comparing the attribution accuracy of counterintuitive versus ordinary concepts linked with
speakers versus places (Exp. 2a) and speakers versus dates (Exp. 2b).
3.1. Method
3.1.1. Participants
Participants were N = 200 (64% female) UCSB undergraduates (Mage = 18.9; SD = 1.23).
Participants identified as White (40%), East, South, or Southeast Asian (30%), Hispanic or
Latino (20%), or as another ethnic/racial background (10%).
3.1.2. Design
This study used a 2 (Concept: Counterintuitive vs. Ordinary) x 2 (Condition: Person vs.
Place) design with repeated measures on the first factor. The dependent variables were the

SPEAKERS OF COUNTERINTUITIVE CONCEPTS 20
proportions of counterintuitive and ordinary concepts correctly attributed to their associated
person or place.
3.1.3. Materials and procedure
Participants were randomly assigned to the Person or Place condition. The Person
condition was identical to Exp. 1. In the Place condition, instead of information about a speaker,
each story began with information about a national park where the story was told (“While you
are camping in [Mammoth / Big Sur / Joshua Tree / Sequoia] you hear the following story”). The
rest of the procedure was the same as in Experiment 1 except that participants in the Place
condition were asked to attribute each concept to the place where they were told about it.
3.2. Results
Attribution accuracy means were entered into a 2 (Concept: Counterintuitive vs.
Ordinary) x 2 (Condition: Person vs. Place) mixed ANOVA with repeated measures on the first
factor. Results revealed a main effect of Concept [F(1, 198) = 29.36, p < .001, η²p = .13], no
main effect of Condition [F(1, 198) < 1.0, p > .250], and no interaction between the two [F(1,
198) = 1.25, p > .250]. After a brief delay, counterintuitive concepts were more accurately
attributed to their associated persons or places than ordinary concepts, and this effect was not
statistically different for persons as compared to places. See Fig. 3 for pirate plots.
4. Experiment 2b

SPEAKERS OF COUNTERINTUITIVE CONCEPTS 21
4.1. Method
4.1.1. Participants
Participants were N = 188 (78% female) UCSB undergraduates (Mage = 18.9; SD = 1.13).
Participants identified as East, South, or Southeast Asian (36%), White (29%), Hispanic or
Latino (25%), or as another ethnic/racial background (10%).
4.1.2. Design
This study used a 2 (Concept: Counterintuitive vs. Ordinary) x 2 (Condition: Person vs.
Date) design with repeated measures on the first factor. The dependent variables were the
proportions of counterintuitive and ordinary concepts correctly attributed to their associated
persons or dates.
4.1.3. Materials and procedure
Participants were randomly assigned to the Person or Date condition. The Person
condition was identical to Exp. 1. In the Date condition, instead of information about a speaker,
each story began with information about a date on which the story was told (“On [April 7 / April
12 / April 19 / April 26] a friend tells you the following story”). The rest of the procedure was

SPEAKERS OF COUNTERINTUITIVE CONCEPTS 22
the same as in Experiment 1 except that participants in the Date condition were asked to attribute
each concept to the date on which they were told about it.
4.2. Results
Attribution accuracy means were entered into a 2 (Concept: Counterintuitive vs.
Ordinary) x 2 (Condition: Person vs. Date) mixed ANOVA with repeated measures on the first
factor. Results revealed a main effect of Concept [F(1, 186) = 24.59, p < .001, η²p = .12], no
main effect of Condition [F(1, 186) = 2.44, p = .120], and no interaction between the two [F(1,
186) < 1.0, p > .250]. After a brief delay, counterintuitive concepts were more accurately
attributed to their associated speakers or dates than ordinary concepts, and this effect was not
statistically different for persons as compared to dates. See Fig. 3 for pirate plots.
Experiment 2a (2-min delay) Experiment 2b (2-min delay)

SPEAKERS OF COUNTERINTUITIVE CONCEPTS 23
Fig 3. Pirate plots of mean attribution accuracy (%) for counterintuitive (CI) and ordinary (OR)
concepts in Exp. 2a and 2b. Inference bands correspond to 95% within-subjects CIs. The dotted
line at 25% indicates chance performance.
5. Experiment 3
Exp. 2a-b replicated and extended Exp. 1 by demonstrating that after a brief delay
counterintuitive versus ordinary concepts were more accurately attributed not only to their
speakers, but also to other contextual details: their places and times of acquisition. The
attribution accuracy advantage for counterintuitive versus ordinary concepts in these experiments
was not statistically different for speakers as compared places or dates, suggesting that epistemic
vigilance mechanisms may initially flag a variety of contextual details surrounding the
acquisition of messages that violate preexisting beliefs. We next explored the stability over time
of links between such messages and their associated contextual details.
In Exp. 3, participants completed the attribution task twice, once after a short distractor
phase (20 minutes) and again after a 48 hours delay. We predicted that counterintuitive concepts
would be more accurately attributed to the contexts of their acquisition than ordinary concepts,
and that this advantage would be more stable over time for speakers as compared to other
contextual details.
5.1. Method
5.1.1. Participants

SPEAKERS OF COUNTERINTUITIVE CONCEPTS 24
Participants were N = 212 (73% female) UCSB undergraduates (Mage = 18.7; SD = 1.09).
Participants identified as White (40%), East, South, or Southeast Asian (28%), Hispanic or
Latino (24%), or as another ethnic/racial background (8%). Of these, n = 194 (92%) returned for
the second session. We report results from participants who completed both sessions only. The
pattern of results for the first session remains the same if we analyze data from the full sample.
5.1.2. Design
This study used a (Concept: Counterintuitive vs. Ordinary) x 2 (Delay: 20-minutes vs.
48-hours) x 2 (Condition: Person vs. Place) design with repeated measures on the first two
factors. The dependent variables were the proportions of counterintuitive and ordinary concepts
correctly attributed to their associated persons or places.
5.1.3. Materials and procedure
The procedure was identical to that in Exp. 2a, except that after the encoding task,
participants completed a 20-minutes (rather than a 2-minutes) battery of distractor tasks before
the first attribution task. After 48 hours, participants then returned for a second testing session to
complete the attribution task again. Although participants knew there would be a second session,
they were not told they would be tested for their memory of the first session stimuli again.
5.2. Results

SPEAKERS OF COUNTERINTUITIVE CONCEPTS 25
Attribution accuracy means were entered into a 2 (Concept: Counterintuitive vs.
Ordinary) x 2 (Delay: 20-minutes vs. 48-hours) x 2 (Condition: Person vs. Place) mixed
ANOVA with repeated measures on the first two factors. Results revealed a main effect of
Concept [F(1, 192) = 24.69, p < .001, η²p = 0.11], a main effect of Delay [F(1, 192) = 69.39, p <
.001, η²p = 0.27], and no main effect of Condition [F(1, 192) < 1.0, p > .250]. There were no
two-way interactions: Concept x Delay [F(1, 192) < 1.0, p > .250], Condition x Delay [F(1, 192)
< 1.0, p > .250], and Condition x Concept [F(1, 192) = 3.67, p = .057]. Critically, there was a
three-way Concept x Delay x Condition interaction [F(1, 192) = 10.36, p = .002, η²p = 0.05]. We
unpack this three-way interaction below. See Figure 4 for pirate plots and Supplemental
Materials for an alternative analytic approach using difference scores. Both approaches yielded
the same conclusions.
5.2.1. Attribution accuracy advantage for persons versus places after 20-minutes and 48-
hours.
After a 20-minute delay, the counterintuitive versus ordinary concepts attribution
accuracy advantage did not statistically differ between Persons and Places, t(192) < 1.0, p > .250,
thereby replicating the findings of Exp. 2a. However, after a 48-hour delay, this attribution
accuracy advantage was significantly greater for Persons as compared to Places, t(192) = 3.46, p
< .001, d = .50, 95% CI = [0.21, 0.78].
5.2.2. Change in attribution accuracy over time for persons and places.

SPEAKERS OF COUNTERINTUITIVE CONCEPTS 26
Simple main effect analyses evaluated attribution accuracy for counterintuitive versus
ordinary concepts over time, separately in the Person and Place conditions. In the Person
condition, the attribution accuracy advantage for counterintuitive versus ordinary concepts more
than doubled with time: after 20-minutes, t(99) = 2.54, p = .012, d = 0.26, 95% CI = [0.05, 0.45];
after 48-hours, t(99) = 5.68, p < .001, d = 0.57, 95% CI = [0.36, 0.78]. In the Place condition, the
attribution accuracy advantage for counterintuitive versus ordinary concepts disappeared entirely
with time: after 20-minutes, t(93) = 2.92, p = .004, d = 0.30, 95% CI = [0.09, 0.51]; after 48-
hours, t(93) < 1.0, p > .250.
5.2.3. Comparing rates of decline in attribution accuracy over time.
Attribution accuracy for person-counterintuitive concepts (CI) pairs started higher than
that for person-ordinary concepts (OR) pairs (MCI = 55.7% vs. MOR = 50.4%) and was more
stable over time (Mdifference = -5.2%, SEdifference = 1.9% vs. Mdifference = -10.7%, SEdifference = 1.6%,
respectively; t(99) = 2.55, p = .012, d = 0.26, 95% CI = [0.06, 0.45]). Attribution accuracy for
person-CI pairs started about the same as for place-CI pairs (MCI = 54.1%) but was more stable
than it over time (Mdifference = -5.2%, SEdifference = 1.9% vs. Mdifference = -11.3%, SEdifference = 1.9%,
respectively; t(192) = 2.27, p = .025, d = 0.33, 95% CI = [0.04, 0.61]). On the other hand,
attribution accuracy for place-CI pairs started higher than for place-OR pairs (MCI = 54.1% vs.
MOR = 42.7%) but was less stable over time (Mdifference = -11.3%, SEdifference = 1.9% vs. Mdifference =
-6.5%, SEdifference = 2.0%, respectively; t(93) = 2.03, p = .045, d = 0.21, 95% CI = [0.004, 0.41]).

SPEAKERS OF COUNTERINTUITIVE CONCEPTS 27
There was no significant difference in attribution accuracy over time for person-OR versus
place-OR pairs, t(192) = -1.70, p = .090, d = -0.25, 95% CI = [-0.53, 0.04].
In sum, after a 20-minute delay, there was an attribution accuracy advantage for
counterintuitive versus ordinary concepts associated with persons or with places, and the two did
not statistically differ. However, after 48-hours, this attribution accuracy advantage more than
doubled in size for persons; this was due to the relative stability of attribution accuracy for
person-CI links over time as compared to person-OR links. Conversely, the counterintuitive
versus ordinary concepts attribution accuracy advantage for places disappeared entirely after 48-
hours; this was due to a relatively rapid decline of attribution accuracy over time for place-CI
links as compared to place-OR links.
20-minute delay 48-hour delay

SPEAKERS OF COUNTERINTUITIVE CONCEPTS 28
Fig. 4. Pirate plots of mean attribution accuracy (%) for counterintuitive (CI) and ordinary (OR)
concepts after 20 minutes and 48 hours. Inference bands correspond to 95% within-subjects CIs.
The dotted line at 25% indicates chance performance.
6. General discussion
Communication is central to human life. Yet communication leaves listeners vulnerable
to misinformation and manipulation. As a consequence, it has been proposed that humans
evolved a suite of adaptations – collectively termed “epistemic vigilance” mechanisms – to
mitigate such threats by monitoring and evaluating communication (Sperber et al., 2010). Here,
we tested the hypothesis that epistemic vigilance mechanisms selectively remember the links
between speakers and messages that are inconsistent with preexisting beliefs (Sperber 1997;
Sperber et al., 2010; see also Cosmides & Tooby, 2000; Johnson et al., 1993). We tested this
hypothesis using the case study of concepts that violate core knowledge intuitions about folk
physics, biology, and psychology (“counterintuitive concepts”; e.g., Boyer, 2001). Across four
experiments, participants read stories containing counterintuitive concepts (e.g., “a cat that has
brown spots and can walk through solid walls”) and ordinary concepts (e.g., “a dog that has soft
fur and likes to play with toys”) that were associated with persons or with other contextual
details (places or times). After a delay, participants were asked to attribute these concepts to the
context of their acquisition.
As predicted, Exp. 1 found that after a brief delay (2 minutes) participants were better at
attributing counterintuitive than ordinary concepts to their speakers. Exp. 2a-b replicated these
findings and further found that this attribution accuracy advantage for counterintuitive versus

SPEAKERS OF COUNTERINTUITIVE CONCEPTS 29
ordinary concepts extended to other contextual details: places (Exp. 2a) and dates (Exp. 2b).
Thus, after a brief delay, a broad variety of contextual details are differentially linked in memory
to messages that violate preexisting beliefs (e.g., Cosmides & Tooby, 2000; Johnson et al.,
1993).
We hypothesized, however, that it may be especially relevant for epistemic vigilance
mechanisms to remember who told you a message that is inconsistent with your preexisting
beliefs, more so than where or when you heard this message. Given this, we explored the
possibility that the links between messages that violate preexisting beliefs and their speakers are
especially stable over time compared to links between such messages and other contextual
details.
Exp. 3 tested this using repeated attribution tests. After a short distractor phase (20-
minutes), participants were better at attributing counterintuitive than ordinary concepts to their
associated contextual details, and this memory advantage did not differ for speakers versus
places. After a 48-hour delay, however, participants no longer showed an attribution accuracy
advantage for counterintuitive versus ordinary concepts and their associated places. In contrast,
participants were not only still better at attributing counterintuitive versus ordinary concepts to
their speakers, but this effect more than doubled.
The current study advances our understanding of how epistemic vigilance mechanisms
monitor and evaluate communication. Epistemic vigilance mechanisms detect inconsistencies
between acquired messages and preexisting beliefs, and selectively remember contextual details
surrounding the acquisition of such messages, with memory for links between such messages and
their speakers being especially stable over time. The linking of messages that violate preexisting
beliefs with such “meta-data” is a key function of epistemic vigilance mechanisms, as they are

SPEAKERS OF COUNTERINTUITIVE CONCEPTS 30
then able to continue evaluating these messages should new information about the competence or
trustworthiness of their speakers come to light, as well as continue evaluating speakers given
new information about their messages.
Indeed, linking messages to meta-data about their speakers is a plausible step toward
developing profiles of our social partners as sources of information. Messages that are at odds
with preexisting beliefs are particularly informative in this regard, as these could reveal that their
speakers have information that we do not, or that they are incompetent or even deceptive. For
instance, should one friend spread negative rumors that are at odds with your positive opinion of
a mutual friend, your epistemic vigilance mechanisms might associate this claim with its
speaker, and you might be motivated to search for additional information about the claim and/or
its speaker as you attempt to reconcile the claim with your preexisting beliefs. Whether you
subsequently accept or reject the claim, remembering the link between the claim and its speaker
might still be advantageous, as it can influence your decisions on whether to believe future things
that speaker says.
Moreover, our findings add to a growing literature (e.g., Mayo, 2019; Mercier, 2017,
2020) suggesting that, contrary to previous accounts, humans are not unduly gullible. Believing
misinformation such as “fake news,” political propaganda, or conspiracies may instead mainly be
a function of its fit (or lack thereof) with preexisting beliefs and motivations. Thus, as
recommended by Lewandowsky, Ecker, Seifert, Schwarz, and Cook (2012), for example,
targeting factors such as an audience’s preexisting beliefs may be a productive starting point in
combating the spread of misinformation.
The findings reported here are also relevant to the source memory literature. In contrast
to past studies on source memory that leveraged violations of expectations about the pairing of

SPEAKERS OF COUNTERINTUITIVE CONCEPTS 31
stimuli and their associated contexts (e.g., a toothbrush paired with a kitchen setting versus a
bathroom setting; Küppers & Bayen, 2014), the studies reported here demonstrate that stimuli
that violate preexisting beliefs by themselves are selectively linked to their contextual details.
Future research is needed to shed light on the exact mechanism by which links between
messages that violate prior beliefs and their associated contexts are remembered. We consider it
possible that epistemic vigilance mechanisms store such messages in a “meta-representational”
data structure that is specialized for linking messages to their meta-data (Cosmides & Tooby,
2000). As suggested by Leslie (1987), meta-representation constitutes the minimal cognitive
architecture needed to decouple representations from one’s existing database of beliefs, including
representations of the mental states of others (mentalizing) and counterfactuals (e.g., pretend
play). As meta-representations, messages that violate preexisting beliefs are hypothesized to
remain quarantined, along with meta-data about the context of their acquisition, pending further
evaluation (Mercier, 2017; Sperber, 1997; Sperber et al., 2010).
Alternatively, the mind might use other mechanisms to link messages that violate prior
beliefs with their meta-data. For instance, on recall people may reconstruct the links between
speakers and their messages. In the experiments reported here, participants could have
remembered who the speaker of, say, the first story presented was, and, independent of this,
remembered the concepts that were in that first story, thereby allowing them to identify the
speaker of the concepts in the first story. In other words, rather than a direct speaker-concept
link, participants could have formed speaker-story and concept-story links that allowed them to
reconstruct the speaker-concept link.
1
Regardless of the exact mechanism by which meta-data
about messages that violate prior beliefs is stored, the selective remembering of meta-data
1
The authors thank Karen J. Mitchell for raising this possibility.

SPEAKERS OF COUNTERINTUITIVE CONCEPTS 32
surrounding the acquisition of such messages, as demonstrated in the experiments reported here,
is a key predicted function of epistemic vigilance mechanisms, and facilitates their capacity for
monitoring and evaluating communication.
Future research may also investigate the broader range of messages that trigger epistemic
vigilance mechanisms. For example, messages that are improbable but not impossible, such as
“there are alligators in the New York City sewers”, are likely to be subjected to epistemic
scrutiny by adults and also by children, who seem to have a weaker grasp of the improbable
versus impossible distinction (Shtulman & Carey, 2007). Moreover, Sperber et al. (2010) suggest
that epistemic vigilance mechanisms are sensitive to the personal relevance of a message. Thus,
one might be more likely to scrutinize a claim about the existence of alligators in the NYC
sewers if she lives in NYC as compared to California.
In conclusion, we demonstrated that people selectively remember the links between
messages that violate preexisting beliefs and their contextual details, especially their speakers.
Memory for the context in which messages that violate preexisting beliefs are shared may be
crucial to the ongoing monitoring and evaluation of such messages, particularly the
differentiation of beneficial from harmful messages, and to constructing profiles of our social
partners. Human social living is made possible by communication, and in turn, it is
psychological mechanisms like those studied here that safeguard us against misinformation and
make communication advantageous.
Context of the research
Communication is central to human social life, yet it exposes listeners to misinformation and
manipulation. Here, we study the cognitive mechanisms that are theorized to have evolved to
keep communication advantageous. We focus on one hypothesized function of these “epistemic

SPEAKERS OF COUNTERINTUITIVE CONCEPTS 33
vigilance” mechanisms: the representation of the contextual details - such as the speakers - of
information that is inconsistent with preexisting beliefs. Our study was inspired by Sperber et al.
(2010), who articulated the theoretical logic of these epistemic vigilance adaptations, and
Sperber (1997), who suggested that messages that are inconsistent with preexisting beliefs are
stored along with speaker tags. We were also inspired by Tooby and Cosmides (2000) who
articulated a broader theoretical model of the "meta-data" stored along with messages. In our
prior research (Barlev et al., 2017, 2018, 2019), we found that concepts that conflict with
universally-held core knowledge intuitions (“counterintuitive” concepts) do not revise those
intuitions but co-exist alongside them. We have therefore used counterintuitive concepts as our
test case here. The present study is part of a broader research program into the functions of
epistemic vigilance mechanisms and how they are used to critically evaluate communication and
thereby guard our database of beliefs from misinformation.

SPEAKERS OF COUNTERINTUITIVE CONCEPTS 34
References
Baillargeon, R. (2004). Infants' physical world. Current Directions in Psychological Science,
13(3), 89-94. doi: 10.1111/j.0963-7214.2004.00281.x
Baillargeon, R., Scott, R. M., & Bian, L. (2016). Psychological reasoning in infancy. Annual
Review of Psychology, 67, 159–186. doi:10.1146/annurev-psych-010213-115033
Banerjee, K., Haque, O. S., & Spelke, E. S. (2013). Melting lizards and crying mailboxes:
Children’s preferential recall of minimally counterintuitive concepts. Cognitive Science,
37(7), 1251–1289. doi: 10.1111/cogs.12037
Barlev, M., Mermelstein, S., & German, T. C. (2017). Core intuitions about persons coexist
and interfere with acquired Christian beliefs about God. Cognitive Science, 41(S3), 425–
454. doi: 10.1111/cogs.12435
Barlev, M., Mermelstein, S., & German, T. C. (2018). Representational co-existence in the God
concept: Core knowledge intuitions of God as a person are not revised by Christian
theology despite lifelong experience. Psychonomic Bulletin and Review, 25, 2330–2338.
doi: 10.3758/s13423-017-1421-6

SPEAKERS OF COUNTERINTUITIVE CONCEPTS 35
Barlev, M., Mermelstein, S., Cohen, A. S., & German, T. C. (2019). The embodied God: Core
intuitions about person physicality coexist and interfere with acquired Christian beliefs
about God, the Holy Spirit, and Jesus. Cognitive Science, 43, e12784.
doi:10.1111/cogs.12784
Barrett, H. C., & Broesch, J. (2012). Prepared social learning about dangerous animals in
children. Evolution and Human Behavior, 33(5), 499–508. doi:
10.1016/j.evolhumbehav.2012.01.003
Barrett, H. C., Broesch, T., Scott, R. M., He, Z., Baillargeon, R., Wu, D., ... & Laurence, S.
(2013). Early false-belief understanding in traditional non-Western societies. Proceedings
of the Royal Society B: Biological Sciences, 280(1755), 20122654. doi:
10.1098/rspb.2012.2654
Barrett, J. L. (1998). Cognitive constraints on Hindu concepts of the divine. Journal for the
Scientific Study of Religion, 608–619. doi: 10.2307/1388144
Barrett, J. L., & Keil, F. C. (1996). Conceptualizing a nonnatural entity: Anthropomorphism in
God concepts. Cognitive Psychology, 31(3), 219–247. doi: 10.1006/cogp.1996.0017
Barrett, J. L., & Nyhof, M. A. (2001). Spreading non-natural concepts: The role of intuitive
conceptual structures in memory and transmission of cultural materials. Journal of
Cognition and Culture, 1(1), 69-100. doi: 10.1163/156853701300063589

SPEAKERS OF COUNTERINTUITIVE CONCEPTS 36
Bayen, U. J., Nakamura, G. V., Dupuis, S. E., & Yang, C.-L. (2000). The use of schematic
knowledge about sources in source monitoring. Memory and Cognition, 28(3), 480–500.
doi: 10.3758/BF03198562
Bell, R., Buchner, A., Kroneisen, M., & Giang, T. (2012). On the flexibility of social source
memory: A test of the emotional incongruity hypothesis. Journal of Experimental
Psychology: Learning, Memory, and Cognition, 38(6), 1512-1529. doi:
10.1037/a0028219
Boyer, P. (1994). The naturalness of religious ideas: A cognitive theory of religion. Berkeley,
CA: University of California Press.
Boyer, P. (2001). Religion explained: The evolutionary origins of religious thought. New York:
Basic Books.
Boyer, P. (2003). Religious thought and behaviour as by-products of brain function. Trends in
Cognitive Sciences, 7(3), 119–124. doi: 10.1016/S1364-6613(03)00031-7
Boyer, P., & Ramble, C. (2001). Cognitive templates for religious concepts: Cross-cultural
evidence for recall of counter-intuitive representations. Cognitive Science, 25, 535–564.
doi: 10.1016/S0364-0213(01)00045-3

SPEAKERS OF COUNTERINTUITIVE CONCEPTS 37
Castelain, T., Bernard, S., & Mercier, H. (2018). Evidence that two‐year‐old children are
sensitive to information presented in arguments. Infancy, 23(1), 124-135. doi:
10.1111/infa.12202
Carey, S. (2009). The origin of concepts. New York: Oxford University Press.
Cook, J., Ellerton, P., & Kinkead, D. (2018). Deconstructing climate misinformation to identify
reasoning errors. Environmental Research Letters, 13(2), 024018. doi: 10.1088/1748-
9326/aaa49f
Cosmides, L., & Tooby, J. (2000). Consider the source: The evolution of adaptations for
decoupling and metarepresentation. In D. Sperber (Ed.), Metarepresentations: A
multidisciplinary perspective (pp. 53–115). New York: Oxford University Press.
Corriveau, K. H., Fusaro, M., & Harris, P. L. (2009). Going with the flow: Preschoolers prefer
non-dissenters as informants. Psychological Science, 20(3), 372-377. doi:
10.1111/j.1467-9280.2009.02291.x
Ehrenberg, K., & Klauer, K. C. (2005). Flexible use of source information: Processing
components of the inconsistency effect in person memory. Journal of Experimental
Social Psychology, 41(4), 369-387. doi: 10.1016/j.jesp.2004.08.001

SPEAKERS OF COUNTERINTUITIVE CONCEPTS 38
Erdfelder, E., & Bredenkamp, J. (1998). Recognition of script-typical versus script-atypical
information: Effects of cognitive elaboration. Memory and Cognition, 26(5), 922–938.
doi:10.3758/BF03201173
Faul, F., Erdfelder, E., Lang, A. G., & Buchner, A. (2007). G* Power 3: A flexible statistical
power analysis program for the social, behavioral, and biomedical sciences. Behavior
Research Methods, 39(2), 175-191. doi: 10.3758/BF03193146
German, T. P., & Barrett, H. C. (2005). Functional fixedness in a technologically sparse culture.
Psychological Science, 16(1), 1-5. doi: 10.1111/j.0956-7976.2005.00771.x
Harris, P. L., & Lane, J. D. (2014). Infants understand how testimony works. Topoi, 33(2), 443-
458. doi: 10.1007/s11245-013-9180-0
Harris, P. L., Koenig, M. A., Corriveau, K. H., & Jaswal, V. K. (2018). Cognitive foundations of
learning from testimony. Annual Review of Psychology, 69, 251-273. doi:
10.1146/annurev-psych-122216-011710
Henrich, J., & McElreath, R. (2003). The evolution of cultural evolution. Evolutionary
Anthropology, 12(3), 123–135. doi: 10.1002/evan.10110

SPEAKERS OF COUNTERINTUITIVE CONCEPTS 39
Hirshman, E., Whelley, M. M., & Palij, M. (1989). An investigation of paradoxical memory
effects. Journal of Memory and Language, 28(5), 594-609. doi: 10.1016/0749-
596X(89)90015-6
Hunt, R.R. (1995). The subtlety of distinctiveness: What von Restorff really did. Psychonomic
Bulletin & Review, 2, 105–112. doi:10.3758/BF03214414
Inagaki, K., & Hatano, G. (2002). Young children’s naive thinking about the biological world.
New York, NY: Psychology Press.
JASP Team (2017). JASP (Version 0.9)[Computer software].
Jaswal, V. K., & Neely, L. A. (2006). Adults don’t always know best: Preschoolers use past
reliability over age when learning new words. Psychological Science, 17(9), 757–758.
doi: 10.1111/j.1467-9280.2006.01778.x
Johnson, M. K., Hashtroudi, S., & Lindsay, D. S. (1993). Source monitoring. Psychological
Bulletin, 114, 3–28.
Koenig, M. A., & Harris, P. L. (2005). Preschoolers mistrust ignorant and inaccurate
speakers. Child Development, 76(6), 1261–1277. doi: 10.1111/j.1467-8624.2005.00849.x

SPEAKERS OF COUNTERINTUITIVE CONCEPTS 40
Kuhlmann, B. G., Vaterrodt, B., & Bayen, U. J. (2012). Schema bias in source monitoring varies
with encoding conditions: Support for a probability-matching account. Journal of
Experimental Psychology: Learning, Memory, and Cognition, 38(5), 1365. doi:
10.1037/a0028147
Küppers, V., & Bayen, U. J. (2014). Inconsistency effects in source memory and compensatory
schema-consistent guessing. The Quarterly Journal of Experimental Psychology, 67(10),
2042-2059. doi: 10.1080/17470218.2014.904914
Lakens, D. (2013). Calculating and reporting effect sizes to facilitate cumulative science: A
practical primer for t-tests and ANOVAs. Frontiers in Psychology, 4, 863. doi:
10.3389/fpsyg.2013.00863
Landrum, A. R., Mills, C. M., & Johnston, A. M. (2013). When do children trust the expert?
Benevolence information influences children's trust more than expertise. Developmental
Science, 16(4), 622-638. doi: 10.1111/desc.12059
Lane, J. D., Wellman, H. M., & Evans, E. M. (2010). Children’s understanding of ordinary and
extraordinary minds. Child Development, 81(5), 1475-1489.
doi: 10.1111/j.1467-8624.2010.01486.x
Larson, H. J., Cooper, L. Z., Eskola, J., Katz, S. L., & Ratzan, S. (2011). Addressing the vaccine
confidence gap. The Lancet, 378(9790), 526-535. doi: 10.1016/S0140-6736(11)60678-8

SPEAKERS OF COUNTERINTUITIVE CONCEPTS 41
Lazer, D. M., Baum, M. A., Benkler, Y., Berinsky, A. J., Greenhill, K. M., Menczer, F., . . . &
Schudson, M. (2018). The science of fake news. Science, 359(6380), 1094–1096. doi:
10.1126/science.aao2998
Lewandowsky, S., Ecker, U. K., Seifert, C. M., Schwarz, N., & Cook, J. (2012). Misinformation
and its correction: Continued influence and successful debiasing. Psychological Science
in the Public Interest, 13(3), 106–131. doi: 10.1177/1529100612451018
Mahr, J. B., & Csibra, G. (2018). Why do we remember? The communicative function of
episodic memory. Behavioral and Brain Sciences, 41, E1. doi:
10.1017/S0140525X17000012
Marsh, R., Cook, G., & Hicks, J. (2006). Gender and orientation stereotypes bias source-
monitoring attributions. Memory, 14(2), 148-160. doi: 10.1080/09658210544000015
Mascaro, O., & Sperber, D. (2009). The moral, epistemic, and mindreading components of
children’s vigilance towards deception. Cognition, 112(3), 367-380. doi:
10.1016/j.cognition.2009.05.012
Mather, M., Johnson, M. K., & De Leonardis, D. M. (1999). Stereotype reliance in source
monitoring: Age differences and neuropsychological test correlates. Cognitive
Neuropsychology, 16(3-5), 437-458. doi: 10.1080/026432999380870

SPEAKERS OF COUNTERINTUITIVE CONCEPTS 42
Mayo, R. (2019). Knowledge and distrust may go a long way in the battle with disinformation:
Mental processes of spontaneous disbelief. Current Directions in Psychological Science,
409-414. doi: 10.1177/0963721419847998
McDaniel, M. A., & Einstein, G. O. (1986). Bizarre imagery as an effective memory aid: The
importance of distinctiveness. Journal of Experimental Psychology: Learning, Memory,
and Cognition, 12(1), 54-65. doi: 10.1037/0278-7393.12.1.54
Mercier, H. (2017). How gullible are we? A review of the evidence from psychology and social
science. Review of General Psychology, 21(2), 103–122. doi: 10.1037/gpr0000111
Mercier, H. (2020). Not born yesterday: The science of who we trust and what we believe.
Princeton University Press.
Mermelstein, S., Barlev, M., & German, T. C. (2020). She told me about a singing cactus:
Enhanced memory for the speakers of counterintuitive versus ordinary concepts
[Supplemental material]. doi: 10.17605/OSF.IO/X5K2U
Norenzayan, A., Atran, S., Faulkner, J., & Schaller, M. (2006). Memory and mystery: The
cultural selection of minimally counterintuitive narratives. Cognitive Science, 30(3), 531–
553. doi: 10.1207/s15516709cog0000_68

SPEAKERS OF COUNTERINTUITIVE CONCEPTS 43
Onishi, K. H., & Baillargeon, R. (2005). Do 15-month-old infants understand false beliefs?
Science, 308(5719), 255-258. doi: 10.1126/science.1107621
Poland, G. A., & Spier, R. (2010). Fear, misinformation, and innumerates: how the Wakefield
paper, the press, and advocacy groups damaged the public health. Vaccine, 28(12), 2361.
doi: 10.1016/j.vaccine.2010.02.052
Shtulman, A. (2017). Scienceblind: Why our intuitive theories about the world are so often
wrong. New York: Basic Books.
Shtulman, A., & Carey, S. (2007). Improbable or impossible? How children reason about the
possibility of extraordinary events. Child Development, 78(3), 1015-1032. doi:
10.1111/j.1467-8624.2007.01047.x
Spelke, E. S. (1990). Principles of object perception. Cognitive Science, 14, 29–56.
doi: 10.1016/0364-0213(90)90025-R
Spelke, E. S., & Kinzler, K. D. (2007). Core knowledge. Developmental Science, 10(1), 89-96.
doi: 10.1111/j.1467-7687.2007.00569.x
Spelke, E. S., Breinlinger, K., Macomber, J., & Jacobson, K. (1992). Origins of knowledge.
Psychological Review, 99(4), 605. doi: 10.1037/0033-295X.99.4.605

SPEAKERS OF COUNTERINTUITIVE CONCEPTS 44
Sperber, D. (1997). Intuitive and reflective beliefs. Mind and Language, 12(1), 67–83. doi:
10.1111/j.1468-0017.1997.tb00062.x
Sperber, D., Clément, F., Heintz, C., Mascaro, O., Mercier, H., Origgi, G., & Wilson, D. (2010).
Epistemic vigilance. Mind and Language, 25(4), 359–393. doi: 10.1111/j.1468-
0017.2010.01394.x
Stangor, C., & McMillan, D. (1992). Memory for expectancy-congruent and expectancy-
incongruent information: A review of the social and social developmental literatures.
Psychological Bulletin, 111(1), 42-61. doi: 10.1037/0033-2909.111.1.42
van der Linden, S., Leiserowitz, A., Rosenthal, S., & Maibach, E. (2017). Inoculating the public
against misinformation about climate change. Global Challenges, 1(2), 1600008. doi:
10.1002/gch2.201600008
von Restorff, H. (1933). Über die wirkung von bereichsbildungen im spurenfeld. Psychologische
Forschung, 18(1), 299-342. doi: 10.1007/BF02409636
Wertz, A. E. (2019). How plants shape the mind. Trends in Cognitive Sciences, 23(7), 528-531.
doi: 10.1016/j.tics.2019.04.009
Wertz, A. E., & Wynn, K. (2014). Selective social learning of plant edibility in 6-and 18-month-
old infants. Psychological Science, 25(4), 874–882. doi: 10.1177/0956797613516145