Content uploaded by Geoffrey Bird
Author content
All content in this area was uploaded by Geoffrey Bird on Sep 18, 2019
Content may be subject to copyright.
!
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
!
1&
Accepted at Journal of Experimental Psychology: General on 17th Sept 2019
Understanding How Minds Vary Relates to Skill in Inferring
Mental States, Personality, and Intelligence.
Jane R. Conway1*, Michel-Pierre Coll2, Hélio Clemente Cuve2, Sofia Koletsi3,
Nicholas Bronitt3, Caroline Catmur3 & Geoffrey Bird1,2
1 MRC Social, Genetic & Developmental Psychiatry Centre, Institute of Psychiatry,
Psychology & Neuroscience, King’s College London.
2 Department of Experimental Psychology, University of Oxford.
3 Department of Psychology, Institute of Psychiatry, Psychology & Neuroscience,
King’s College London.
*Correspondence: jane_rebecca.conway@kcl.ac.uk (J.R. Conway) or
geoff.bird@psy.ox.ac.uk (G. Bird)
Word count: 8535.
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
2&
Abstract
Using a ‘theory of mind’ allows us to explain and predict others’ behaviour in terms
of their mental states, yet individual differences in the accuracy of mental state
inferences are not well understood. We hypothesised that the accuracy of mental state
inferences can be explained by the ability to characterise the mind giving rise to the
mental state. Under this proposal, individuals differentiate between minds by
representing them in ‘Mind-space’ – a multidimensional space where dimensions
reflect any characteristic of minds that allows them to be individuated. Individual
differences in the representation of minds and the accuracy of mental state inferences
are explained by one’s model of how minds can vary (Mind-space), and ability to
locate an individual mind within this space. We measured the accuracy of
participants’ model of the covariance between dimensions in Mind-space that
represent personality traits, and found this was associated with the accuracy of mental
state inference (Experiment 1). Mind-space accuracy also predicted the ability to
locate others within Mind-space on dimensions of personality and intelligence
(Experiment 2). Direct evidence for the representation of minds in mental state
inference was obtained by showing that the location of others in Mind-space affects
the probability of particular mental states being ascribed to them (Experiment 3). This
latter effect extended to mental states dependent upon representation of trait
covariation (Experiment 4). Results support the claim that mental state inference
varies according to location in Mind-space, and therefore that adopting the Mind-
space framework can explain some of the individual differences in theory of mind.
Keywords:
theory of mind; individual differences; personality; social cognition; Mind-space.
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
3&
Introduction
When trying to understand other people’s behaviour, our explanations are
greatly enriched by referring to their mental states, such as what they believe, know,
desire or intend. This ‘theory of mind’ (ToM) ability is considered crucial in social
interactions, from everyday relationships to political negotiations and criminal trials.
The scientific study of ToM has spanned 40 years (Premack & Woodruff, 1978) and
multiple disciplines, including developmental, socio-cognitive, clinical, and
comparative psychology, artificial intelligence, and neuroscience (Gallagher & Frith,
2003; Happé, 1994; Heyes, 2015; Rabinowitz et al., 2018). However, a fundamental
challenge in the ToM literature persists: what is it that makes some people better at
inferring mental states than others (see Repacholi & Slaughter, 2003, for discussion)?
There are two main reasons why individual differences in ToM have been
difficult to explicate. First, empirical measurement of unobservable mental states is
difficult, necessitating that for most tasks the ‘correct’ and ‘incorrect’ mental state
inferences are predetermined by the authors based on rationality and logic (Baron-
Cohen, Leslie, & Frith, 1985) or by consensus (Dziobek et al., 2006). With such task
designs, performance does not reflect the accuracy of mental state inference, but
instead how rational, or how typical, mental state inferences are. Even when task
performance has the potential to reflect the accuracy rather than rationality/typicality
of the participant’s mental state inference (e.g. the ‘Beauty Contest’, Nagel, 1995),
results provide little insight into individual variance in the representational or
inferential processes by which that inference was derived (Heyes, 2014). Second, due
to these difficulties measuring the accuracy of mental state inferences, individual
differences in performance on ToM tasks have typically been attributed to domain-
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
4&
general abilities (Devine & Hughes, 2014; Milligan, Astington, & Dack, 2014;
Sabbagh, Xu, Carlson, Moses, & Lee, 2006) rather than domain-specific processes or
representational structures. Verbal skills, memory, or inhibitory control contribute to
performance on ToM tasks that demand those abilities, but cannot explain variance
unique to mental state inference.
Previous work describing improvements in ToM from early to late childhood
and into adulthood has revealed continuing improvements in mental state inference
(so-called ‘advanced ToM’, e.g. Osterhaus, Koerber & Sodian, 2016). This work
details how, during development, individuals gradually incorporate additional sources
of information into their mental state inferences, and therefore provides one
framework within which to understand individual differences in ToM. For example,
as social and emotional understanding becomes (1) increasingly more sophisticated,
and (2) integrated into mental state inferences (e.g. Baron-Cohen, O’Riordan, Stone,
Jones, & Plaisted, 1999; Burnett, Bird, Moll, Frith, & Blakemore, 2009), individual
differences in either the degree of social/emotional understanding or its integration
into mental state reasoning could explain individual differences in the accuracy of
mental state inferences.
The work presented here is concerned with a second way in which individual
differences in the accuracy of mental state inference can be understood: the
representation of others’ minds. Crucially, minds moderate the link between
situational contexts and the mental states they evoke: two different target minds in the
same situation may generate completely different mental states. The accuracy with
which those target minds can be represented, therefore, is likely to contribute to
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
5&
accuracy in inferring the target’s mental states. Thus, the experiments reported here
address how individual differences in mind representation may give rise to individual
differences in the accuracy of mental state inference. The work is based on the
hypothesis that a major source of naturalistic variance in the probability of others
having particular mental states is variability in the people in one’s environment.
Mental states are the product of a specific individual mind, and therefore accurate
representation of how minds vary likely affects the accuracy of any mental state
inference (Conway et al., 2019).
Empirical work suggests that representation of minds, and the processes
occurring within minds, are initially not explicitly integrated with mental state
inference, but become so as children develop. For example, Ruffman (1996) found
that until 7 years of age children often find it easier to attribute an incorrect false
belief than a correct true belief, when attributing a true belief would require the child
to understand the distinction between knowledge states in an individual’s mind (i.e.
they may be ignorant about X but know Y). Instead, young children applied a simple
rule of the form “if a person didn’t see something then they cannot know it”. Thus, for
children below 7 years of age, in at least some situations, mental state inference is
determined by the situation an individual is in, not by a model of how minds, and the
processes within minds, inform mental states.
Older children slowly begin to understand explicitly the link between minds
and mental state inferences. This is most clearly demonstrated by the work on
‘interpretive theory of mind’, the understanding that two individuals can be exposed
to exactly the same information and yet draw different conclusions. For example,
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
6&
children above 7 years of age are able to understand that two individuals who are
shown the same small portion of a picture can make different inferences about the
picture as a whole (Lalonde & Chandler, 2002). Around the age of 10, children can
understand that it is impossible to know which of two percepts will be formed by an
unknown individual when they perceive an ambiguous figure which affords two
distinct percepts (such as a visual illusion; Osterhaus et al., 2016).
With respect to an implicit understanding of the link between minds and
mental states, a rudimentary understanding may be gained in childhood and is
certainly present during adolescence and adulthood. For example, during
stereotyping, individuals decide that minds of a certain type (e.g. those belonging to
out-groups) are more likely to hold particular beliefs or to have certain intentions than
minds of another type (e.g. those of the in-group). To illustrate, work on Fiske,
Cuddy, Xu, & Glick’s (2002) Stereotype Content Model has shown that the two
dimensions characterising stereotype content (warmth and competence) are associated
with changes in the frequency of inferred mental states. For example, the warmth
dimension changes the inferred intentions of the stereotyped individual, such that
groups associated with high warmth are expected to hold positive intentions towards
the self, while those associated with low warmth are expected to hold negative
intentions towards the self (see Fiske et al., 2002). While these mental states are broad
and non-specific, they may be operationalised in very specific ways in particular
contexts. For example, during a sales negotiation, a member of a group stereotyped as
warm may be thought to favour fairness over profit, while a member of a group
stereotyped as cold might be expected to favour profit over fairness. Even children of
between 3-5 years of age show a rudimentary understanding of gender stereotypes,
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
7&
and use them to determine what males and females are likely to desire (Aboud, 1988;
Wellman & Liu, 2004). Thus, from relatively early in development, judgements of the
probability of particular mental states are altered on the basis of the type of mind
giving rise to them (although this link may not be explicitly represented until late
childhood).
The preceding work demonstrates therefore, that at least by older childhood or
adolescence, a target’s mind is explicitly represented in order to infer the probability
of particular mental states. The experiments reported here build on this work to test
the hypothesis that individual differences in mind representation may explain
individual differences in the accuracy of mental state inferences. Specifically, we
hypothesised that minds may be represented as locations within a multidimensional
space (‘Mind-space’) in which dimensions reflect any discriminable aspect of minds,
such as their cognitive abilities (e.g. intelligence) and behavioural tendencies (e.g.
personality traits; Conway et al., 2019). As such, Mind-space is similar to the idea of
Face-space (Valentine, 1991; Valentine, Lewis, & Hills, 2016), which is theorised to
be a multi-dimensional space where dimensions represent ways in which faces can be
discriminated. Once formed, individual faces are thought to be represented as points
within this multi-dimensional space. Mind-space may be thought of as analogous to
Face-space. For example, target minds A and B may be represented in a 3-
dimensional Mind-space with dimensions of working memory, extraversion, and
conscientiousness, but each target is located at a different point within the space
according to their characteristics. One benefit of representing minds within a multi-
dimensional space is that covariance between dimensions can be more easily
represented and utilised to make mental state inferences. Locating a mind within
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
8&
Mind-space could permit accurate mental state inference because the target’s mental
states are, in part, dependent on their location in the space. For example, if I can
accurately place targets A and B along the extraversion dimension, I could better
predict their respective attitudes (i.e. mental states) towards attending a party. A
person is therefore more likely to be accurate at inferring a target’s mental states if:
(1) the person represents the relevant dimensions and any covariance between
dimensions; (2) they can accurately locate a mind in Mind-space based on samples of
behaviour; (3) they use a target’s location in Mind-space combined with situational
factors when generating mental state inferences. (See Figure 1 for a full example.)
Figure 1. Schematic illustration of how the Mind-space framework can be used to
explain individual differences in Theory of Mind (ToM). The Mind-space framework
suggests that individual differences in ToM are due to: (1) The accuracy of the
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
9&
representation of the dimensions within which minds vary and the relationship
between these dimensions (i.e. Mind-space); (2) The ability to locate a target mind
within Mind-space; (3) The ability to combine diagnostic information about the
situation the target is in with the target’s position in Mind-space to accurately infer
their mental state; (4) The propensity to consider position in Mind-space before
making a mental state inference (not illustrated). Person 1 and Person 2 are asked to
estimate the attitude of two targets (A and B) towards parties on weekends and
weekdays based on how extraverted they appear. Person 1 can accurately locate the
targets on the extraversion dimension, but Person 2 cannot. Person 1’s Mind-space
accurately reflects the positive correlation between conscientiousness and
extraversion whereas Person 2’s does not. Due to Person 1’s accurate representation
of Mind-space, only Person 1 can infer the targets’ degree of conscientiousness on the
basis of their degree of extraversion. This enables Person 1 to infer that because
Target A is more extravert than B, Target A is also more conscientious than B, and so
Person 1 can predict that Target A will more likely have diverging attitudes to parties
on the weekend vs. a weekday. Person 2 has no basis to predict differential attitudes
to parties based on the day of the week, and this is furthered compounded by their
failure to locate the targets accurately within their Mind-space. As a result, Person 1
makes more accurate mental state inferences than Person 2.
We aimed to measure the accuracy of the covariance between dimensions that
represent personality traits in an individual’s Mind-space. Personality is particularly
apt for this first test of the Mind-space theory because factor analyses have
established that traits can be represented using five (Goldberg, 1990) or six (Ashton &
Lee, 2007) dimensions. Although each dimension is distinct there is some degree of
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
10&
correlation between them, thus the existing personality literature provides ground
truth values for the average covariance between traits in the population (or at least
ground truth values for the population completing a particular personality test at a
particular moment in history). The presence of covariation across a number of
dimensions would be most efficiently represented in a multi-dimensional space such
as Mind-space. We therefore developed the ‘Personality Pairs Task’ which asks
participants to estimate the average correlations between traits on six personality
dimensions (Ashton & Lee, 2009). These estimated correlations can then be compared
to ground truth values from a similar population to determine the accuracy of an
individual’s Mind-space. If there exists a relationship between the representation of
minds and the inference of mental states, we hypothesised that performance on a ToM
task would be associated with Mind-space accuracy (Experiment 1).
In Experiment 2, we sought to test whether Mind-space accuracy predicts the
ability to locate a target mind within Mind-space. Accordingly, participants in
Experiment 2 completed the Personality Pairs Task and were asked to estimate the
personality and intelligence of a number of targets on the basis of video-recorded
‘thin-slices’ of behaviour. Such thin-slices provide minimal experience of a target yet
can result in surprisingly accurate predictions of their traits and abilities (Borkenau,
Mauer, Riemann, Spinath, & Angleitner, 2004; Carney, Colvin, & Hall, 2007).
Participants were asked to locate each target on personality and intelligence
dimensions and their estimates were compared to ground truth values we collected for
each target. If Mind-space accuracy predicts the ability to locate an individual within
Mind-space, scores on the Personality Pairs Task should predict the accuracy of
participants’ target location estimates. The design of Experiment 2 also allowed us to
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
11&
assess if similarity in personality between the participant and the target affects the
accuracy of trait judgements. Higher accuracy for targets similar to the self may
reflect an egocentric bias whereby participants anchor their judgements of the targets’
traits on their own traits (Epley, Keysar, Van Boven, & Gilovich, 2004), and such
egocentricity would result in more accurate judgements when the target is similar, but
less accurate judgements for dissimilar targets. Under the Mind-space framework,
providing one can accurately locate oneself within Mind-space, similarity effects
would be due to increased experience of the mapping between one’s position in Mind-
space and behaviour across situations. This greater experience would enable a target’s
position in Mind-space to be derived from behaviour more accurately, and across a
greater number of situations, if the target occupied a similar position as the self within
Mind-space (Conway et al., 2019). Under either account, if similarity in personality
between the participant and the target affects the accuracy of trait judgements, then
we should observe higher accuracy on the thin-slice location task for targets that are
similar to the participant compared to those who differ.
Even if results in accordance with the predictions of the Mind-space
framework are observed in Experiments 1 and 2, it could be argued that they do not
provide a direct test of the Mind-space framework itself. They are not designed to
provide evidence that participants incorporate the position of a target mind within
Mind-space when inferring the content of their mental states. Accordingly, in
Experiment 3, we investigated how the position of targets in Mind-space, combined
with situational information, affects the probability of particular mental states being
inferred. This work builds on, but goes beyond, previous demonstrations that older
children recognise that two minds may produce different mental states when exposed
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
12&
to the same information (Lalonde & Chandler, 2002), or that different types of minds
may be associated with different probabilities of generally positive or negative
intentions towards the in-group (Fiske et al., 2002), by showing quantitatively the
degree to which the probability of certain mental states is updated as target minds
move through Mind-space, and as other minds move through the target’s Mind-space.
Participants in Experiment 3 were presented with a series of vignettes based
on the Sally-Anne False Belief Task (Baron-Cohen et al., 1985). In this task, Sally
places a marble in her basket and leaves the scene; while she is away Anne takes the
marble from Sally’s basket and puts it in her own box. The critical test question asks:
where will Sally look for the marble on her return? The ability to ascribe a false belief
to Sally – that she will look for the marble in the location where she left it (her basket)
rather than where it really is (Anne’s box) – is considered a litmus test of theory of
mind (Dennett, 1978; Wimmer & Perner, 1983). False belief tasks involving an
unseen change-of-location have been used extensively to test the theory of mind
ability of human infants (Baillargeon, Scott, & He, 2010), children (Kulke, Reiß,
Krist, & Rakoczy, 2017), people with autism (Happé, 1994), non-human primates
(Heyes, 2017), and artificial agents (Rabinowitz et al., 2018). However, these tasks do
not take into account the representation of the particular minds of Sally and Anne; in
the task they are merely anonymous protagonists (Conway et al., 2019). We presented
participants with vignettes in which the Sally character varied across four levels of
paranoia, and the Anne character across four levels of dishonesty. We predicted that
the mental state attributed to Sally by the participant would vary as a function of
where Sally was in the participant’s Mind-space, and where the participant believed
Anne to be in Sally’s Mind-space; specifically that at higher levels of paranoia and
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
13&
dishonesty, participants would be less likely to infer that Sally would look in her
basket where she left her marble, and be more likely to infer that Anne has stolen the
marble and hidden it in her own box. If this prediction is supported, it would provide
direct evidence for the incorporation of position in Mind-space when inferring mental
states.
Experiment 3 has the potential to show that a characteristic of the target mind
is represented and used to inform mental state inferences for which it is relevant. It
does not, however, have the potential to show that the target mind is represented
within a multi-dimensional space. Experiment 4 therefore used the same basic design
as Experiment 3, but tested the following prediction: that providing a participant with
information about a target mind’s location on certain ‘source’ dimensions should
allow that target’s mind to be located on other dimensions, to the extent that those
other dimensions covary with the source dimension within that participant’s Mind-
space. Accordingly, Experiment 4 asked participants to complete the same false belief
vignettes as in Experiment 3, for a number of Sally characters that varied on source
dimensions which a validation study suggested to be associated with paranoia in the
general population. If varying the position of the Sally character on the source
dimensions changes the mental state attributed to her, and crucially if it does so to the
degree that the participant believes each source trait covaries with paranoia, then this
would provide stronger evidence for the idea that target minds are located within a
multi-dimensional space, and that target location in Mind-space is used in mental state
inference.
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
14&
Collectively, the four experiments were designed to provide complementary
tests of the Mind-space theory. As detailed above, Experiments 3 and 4 account for
variability in the minds available for representation and how the location of a mind in
Mind-space affects the probability of which mental state is attributed to that mind.
Experiment 2 examines the ability to locate a specific mind in Mind-space and how
this relates to Mind-space accuracy. First, in Experiment 1, we test for a relationship
between the accuracy of Mind-space and the accuracy of mental state inferences. If
the accuracy of mental state inference is indeed determined by the accuracy of Mind-
space, then those individuals who have a more accurate representation of how minds
vary, in this case operationalised as the covariance between personality dimensions,
should also make more accurate mental state inferences.
Experiment 1
Method
Participants. Sixty adults volunteered to take part in this experiment in return
for a small monetary sum or undergraduate research participation credits. Participants
(48 female) were aged between 18 and 55 years old (M = 23.62, SD = 6.21). An a
priori power calculation using the pwr package in R (Champely et al., 2018) indicated
that for Cohen’s f2 = .15 and α= .05, a sample size of 58 would provide 80% power
for the main hypothesis being tested (with two predictor variables). The local
Research Ethics Committee approved the study.
Measures.
Personality Pairs Task. The Personality Pairs Task (PPT) comprised 72
questions. Each question included a pair of items measuring traits on the HEXACO
personality inventory (Ashton & Lee, 2009). The HEXACO-60 is a 60-item
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
15&
questionnaire that captures six personality dimensions. Five of these are similar to
those captured in five-factor personality models: Emotionality (E), similar to
Neuroticism; Extraversion (X); Agreeableness (A); Conscientiousness (C); and
Openness to Experience (O). Honesty-Humility (H) represents a sixth dimension not
captured within the five-factor models (Ashton & Lee, 2007), and reflects traits of
sincerity, fairness, greed-avoidance, and modesty. On each trial of the Personality
Pairs Task, participants were asked to rate how likely, on average, is it that someone
who has one trait would also have the other. For example: “On average, how likely is
it that someone who people think of as having a quick temper, would also make
decisions based on the feeling of the moment rather than on careful thought?”
Participants responded using a sliding scale from ‘Extremely Unlikely’ (-100) to
‘Neither Likely Nor Unlikely’ (0), to ‘Extremely Likely’ (+100), and this response
was divided by 100 to give a negative or positive estimated correlation coefficient.
There were two pairs of traits presented for every combination of the six HEXACO
personality dimensions. The actual inter-trait correlation values for the population
were obtained from a sample (N = 2,868) collected by Lee and Aston (Lee & Ashton,
2016). Participants’ accuracy was computed by taking the absolute difference score
between the population correlation and their estimated correlation between the traits,
and calculating the mean difference score across the 72 trials. Smaller difference
scores indicate higher accuracy at predicting the actual population correlation values,
and therefore a more accurate Mind-space.
Movie for the Assessment of Social Cognition (MASC). The MASC
(Dziobek et al., 2006) is a naturalistic theory of mind task, which requires participants
to watch a 15-minute video of four characters having dinner together. After each
video segment, a multiple-choice question with four possible responses is asked.
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
16&
There are 45 mental state questions and 21 control questions (Santiesteban, Banissy,
Catmur, & Bird, 2015). The control questions do not require any mental state
representation and account for non-mentalistic factors that may affect performance,
e.g. memory, attention, verbal comprehension, or motivation. For the mental state
questions, the multiple-choice options reflect four response types: no mental state
inference; insufficient mental state inference; correct mental state inference; and
excessive mental state inference. Participants’ scores were computed as the
percentage of correct responses on the mental state and control questions respectively;
and for each of the three incorrect response types to mental state questions, the sum
score of the number of errors was also computed (i.e. no mental state inference;
insufficient mental state inference; and excessive mental state inference).
Procedure. Participants completed the study individually on a computer in a
testing room in a single session of approximately one hour. The measures were
presented in the following order: Personality Pairs Task [36 trials]; MASC;
Personality Pairs Task [36 trials].
Statistical Analyses. Multiple regression models were performed using the lm
function in R. To assess whether non-normality of residuals affected the models,
robust regression models were also performed using the boot package in R (Canty &
Ripley, 2017) to provide bootstrapped 95% confidence intervals of regression
coefficients based on 2000 bootstrap samples. A close resemblance between the
bootstrapped coefficients and the original coefficients indicated that non-normal
distributions did not affect the model. The data for this study are available at
https://doi.org/10.17605/OSF.IO/4K9HS.
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
17&
Results
Descriptive statistics for all variables are presented in Table 1. To investigate
whether Mind-space accuracy is associated with the accuracy of mental state
inference after controlling for non-mentalistic reasoning ability, a multiple regression
model was performed with PPT difference score as the outcome variable and
percentage correct scores on the MASC mental state and control questions as the
predictor variables (Table 2, Model 1.A: PPT mean difference score ~ MASC Mental
State % correct + MASC Control % correct). The model explained a significant
proportion of the variance in PPT scores, R2 =0.13, F (2, 57) = 4.20, p = .02. As
shown in Table 2 (Model 1.A), only performance on the MASC mental state
questions significantly predicted accuracy on the PPT. Performance on the MASC
control questions did not predict accuracy on the PPT. This suggests that those
participants who performed better on a theory of mind task had a more accurate
Mind-space, as indicated by lower difference scores on the PPT. That the relationship
was observed for the mental state questions only, not the control questions, suggests
that it is specific to theory of mind and not attributable to variance in other cognitive
domains such as memory, attention, or verbal ability.
To further assess which type of theory of mind errors were associated with
poorer Mind-space accuracy, a second multiple regression model was performed with
PPT difference score as the outcome variable and error type sum scores on the MASC
mental state questions as the predictor variables (Model 1.B: PPT mean difference
score ~ MASC no mental state inference + MASC insufficient mental state inference
+ MASC excessive mental state inference). The model explained a significant
proportion of the variance in PPT scores, R2 = 0.22, F (3, 56) = 5.17, p = .003. Only
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
18&
errors indicating no mental state inference significantly predicted performance on the
PPT (Table 2, Model 1.B). Errors indicating insufficient or excessive mental state
inference did not predict PPT performance. These results show that those who failed
to make any mental state inference had a less accurate Mind-space, as indicated by
higher difference scores on the PPT.
Table 1
Descriptive Statistics for Experiment 1
Variable
Mean
SD
Range
PPT Difference Score
0.37
0.13
0.15 – 0.70
Mental State (MS) Qs % Correct
77.55
11
40 – 93.33
Control Qs % Correct
90.79
6.84
71.43 – 100
Errors: No MS Inference
1.58
1.61
0 – 6
Errors: Insufficient MS Inference
3.58
3.14
0 – 17
Errors: Excessive MS Inference
4.93
2.58
0 – 11
Note. PPT = Personality Pairs Task. MS = Mental State. Qs = Questions.
!
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
!
19&
Table 2
Experiment 1 Regression Analyses: Predictors of Performance on the Personality Pairs Task
Predictor
B
SE
95% CI
Bootstrap 95% CI
β
t
p
Model 1.A
Mental State Qs % Correct
-0.004
0.002
[-0.007, -0.001]
[-0.007, -0.001]
-0.31
-2.30
.03*
Control Qs % Correct
-0.002
0.002
[-0.007, 0.003]
[-0.007, 0.003]
-0.11
-0.81
.42
Model 1.B
Errors: No MS Inference
0.033
0.010
[0.013, 0.053]
[0.014, 0.053]
0.41
3.27
.002**
Errors: Insufficient MS Inference
-0.001
0.005
[-0.011, 0.009]
[-0.011, 0.008]
-0.02
-0.16
.88
Errors: Excessive MS Inference
0.009
0.006
[-0.002, 0.021]
[-0.005, 0.022]
0.19
1.61
.11
Note. Qs = Questions. MS = Mental State. * p < .05. ** p < .01.
!
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
!
20&
Discussion
As predicted, Experiment 1 demonstrated that performance on a ToM task was
associated with Mind-space accuracy as measured by the Personality Pairs Task. A
relationship was observed both for overall ToM accuracy and for errors indicating a
failure to infer any mental state. Building on previous evidence that adults represent
others’ minds when inferring mental states (e.g. Fiske et al., 2002), these results
provide evidence for the relationship between the accuracy of mind representation and
the accuracy of mental state inference.
In Experiment 2, we tested the following predictions: that those with a more
accurate Mind-space would be better able to locate specific targets within Mind-
space; and that similarity in personality to the target will affect the accuracy with
which they do so (Conway et al., 2019). The accuracy of Mind-space was again
measured using the Personality Pairs Task. The ability to locate individuals within
Mind-space accurately was assessed using a thin-slice procedure in which participants
watched short video-recordings of a number of targets reciting a simple sentence.
They were asked to estimate the personality and intelligence of each target based on
this ‘thin-slice’ of their behaviour, and participant estimates were compared to the
target’s actual personality and IQ scores as a measure of their accuracy. If results are
as predicted, then participants who have a more accurate Mind-space as measured by
the Personality Pairs Task should also be more accurate when locating individuals
within Mind-space on the basis of thin-slices of their behaviour.
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
21&
Experiment 2
Method
Participants. Sixty-eight adults that did not take part in Experiment 1
volunteered to take part in this experiment in return for a small monetary sum or
undergraduate research participation credits. Participants (58 female) were aged
between 18 and 57 years old (M = 23.76, SD = 7.52). An a priori power calculation
indicated that for Cohen’s f2 = .15 and α= .05, a sample size of 66 would provide 80%
power for the hypotheses being tested (with three predictor variables). The local
Research Ethics Committee approved the study.
Measures.
Behavioural samples of targets: thin-slice video stimuli. ‘Thin-slices’ of
targets’ behaviour were presented to participants via video stimuli. Ten males and ten
females were recruited to feature as targets in the thin-slicing video stimuli. Each
target was filmed from the chest up against a white background (See Supplemental
Materials Video S.1, or https://doi.org/10.17605/OSF.IO/4K9HS) saying the phrase
“Hi, I am a participant in this study and my ID number is xxxx”. Each target was
given a unique four-digit ID number to say. Video duration was between six and nine
seconds (depending on the rate of the target’s speech). Targets completed the self-
report HEXACO-60 personality inventory, and the observer-report HEXACO-60
(Ashton & Lee, 2009) was completed by someone who knew them well. This
procedure provided a mean self-reported score and observer-reported score for each
target on each of the six dimensions on the HEXACO. The Matrix Reasoning and
Vocabulary sub-scales of the Wechsler Abbreviated Scale of Intelligence 2nd edition
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
22&
(Wechsler, 2011) were administered to targets, from which the target’s Intelligence
Quotient percentile rank was obtained.
Ratings of behavioural samples of targets. For the personality ratings,
participants were first given a description of the HEXACO personality inventory and
the meaning of the six dimensions. They were provided with descriptions of all six
dimensions and all statements one would agree and disagree with if one scored highly
on each dimension. (Note that this task was performed after the participants
completed the HEXACO in relation to their own personality and thus could not have
affected their scores on this measure; see Procedure below for task order.) After the
target’s video was presented, participants were asked to rate that target’s personality
on each of the six dimensions on a sliding scale ranging from the ‘lowest’ to ‘highest’
possible score. These ratings provided a response between 1 and 5 that allowed for
comparison with the target’s mean on each dimension. Participant accuracy was
computed by taking the absolute difference score on each dimension between (a) the
target’s self-reported mean and the participant’s estimated mean, and (b) the target’s
observer-reported mean and the participant’s estimated mean. Smaller difference
scores indicate higher accuracy at predicting the target’s personality.
For the intelligence ratings, as for personality, participants were first given
instructions on how intelligence is defined and how to rate the target’s intelligence
compared to the general population where responses indicate the target’s percentile
rank (e.g. On this scale, ‘average’ means that if you chose a group of 100 at random,
half (50%) of them would be more intelligent and half (50%) of them would be less
intelligent than the person you are rating; ‘Top 25%’ means that 75 people would
be less intelligent than the person you are rating; ‘Bottom 25%’ means that 75 people
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
23&
would be more intelligent than the person you are rating.). After viewing the target’s
video, participants were asked to rate them on how intelligent they are compared to
the general population on a scale from 0% to 100% with markers at ‘Bottom 25%’,
‘Average’, and ‘Top 25%’. This allowed for comparison with the target’s actual IQ
percentile rank by taking the absolute difference score between the target’s rank and
the participant’s estimate. As before, smaller difference scores indicate higher
accuracy at predicting the target’s IQ.
Personality Pairs Task. As described in Experiment 1.
Participant-Target similarity in personality. Participants completed the self-
report HEXACO-60 personality inventory. Participants were asked to respond to
statements on a 5-point Likert scale from ‘Strongly Disagree’ to ‘Strongly Agree’. A
mean score was computed for each of the six dimensions (minimum score = 1,
maximum = 5). We then computed absolute difference scores between each
participant and target by subtracting the participant’s score for each of the six
dimensions from the target’s self-reported HEXACO scores. Smaller difference
scores indicate more similarity between the participant and target.
Procedure. Participants completed the study individually on a computer in a
testing room in a single session of approximately one hour. The measures were
presented in the following order: Personality Pairs Task [72 trials]; Self-report
HEXACO; Ratings of behavioural samples of targets from thin-slicing video stimuli
[20 trials].
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
24&
Statistical analyses. The statistical analyses were as for Experiment 1 with
the addition of random effects to the linear models to take into account the variance
across participants, targets and HEXACO personality dimensions. Analyses were
performed using the lmer package (Bates et al., 2018). The data for this study are
available at https://doi.org/10.17605/OSF.IO/4K9HS.
Results
Descriptive statistics for all variables are presented in Table 3. To investigate
whether those with a more accurate Mind-space were better able to locate specific
targets within Mind-space, mixed models were performed. The outcome variable for
Model 2.A was the difference between the target’s self-reported score and the
participant’s estimate of it for each of the six HEXACO dimensions (‘SRH difference
score’). Model 2.B was similar except it used the target’s observer-reported score
(‘ORH difference score’). Both models 2.A and 2.B had PPT difference score as the
fixed effect, and participants (68) target (20) and personality dimensions (6) as
random effects allowing for random intercepts. The outcome variable for Model 2.C
was the difference between the target’s IQ percentile and the participant’s estimate of
it (‘IQ difference score’), with PPT difference score as the fixed effect and target (20)
as the random effect. Additional information on the distribution of personality trait
scores and their contribution to the accuracy of personality estimates is presented in
Supplemental Materials (Fig S1 and Table S1).&
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
25&
Table 3
Descriptive Statistics for Experiment 2
Variable
Mean
SD
Range
PPT Difference Score
0.37
0.11
0.15 - 0.67
SRH Difference Score
0.83
0.62
0 - 3.60
ORH Difference Score
0.78
0.59
0 - 3.60
IQ Difference Score
20.58
14.05
0 - 71
Note. PPT = Personality Pairs Task. SRH = Self-report HEXACO. ORH = Observer-
report HEXACO. IQ = Intelligence.
As shown in Table 4, performance on the PPT significantly predicted SRH
difference scores (Model 2.A), ORH difference scores (Model 2.B), and IQ difference
scores (Model 2.C). As hypothesised, those participants with a more accurate Mind-
space, as indicated by lower difference scores on the PPT, were more accurate at
estimating the target’s self- and observer- reported scores on the HEXACO and the
target’s IQ percentile rank, thus supporting the prediction that they would more
accurately locate targets in Mind-space based on a minimal sample of behaviour.
To investigate whether similarity in personality between the participant and
the target was associated with the accuracy of trait judgements, we ran the same
models as previously except now the fixed effect was the participant-target similarity
score (Model 2.D: outcome variable = SRH; Model 2.E: outcome variable = ORH;
Model 2.F: outcome variable = IQ). As shown in Table 5, degree of similarity
significantly predicted SRH difference scores (Model 2.D) and ORH difference scores
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
26&
(Model 2.E), but not IQ difference scores (Model 2.F). Participants who were more
similar in personality to targets were more accurate at estimating the target’s self-
reported scores and observer-reported scores on the HEXACO personality measure,
but personality similarity had no effect on estimates of the target’s IQ.
!
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
!
27&
Table 4
Experiment 2: Regression Analyses
Predictor
Random Effects
Outcome
B
SE
95% CI
Bootstrap 95% CI
t
p
Model 2.A
PPT
Target; Personality Trait;
Participant
SRH
0.51
0.12
[0.27, 0.75]
[0.26, 0.76]
4.12
<.001**
Model 2.B
PPT
Target; Personality Trait;
Participant
ORH
0.56
0.14
[0.29, 0.83]
[0.28, 0.84]
4.00
<.001**
Model 2.C
PPT
Target
IQ
5.80
2.70
[0.52, 11.09]
[0.51, 11.03]
2.15
0.03*
Note. PPT = Personality Pairs Task. SRH = Self-report HEXACO. ORH = Observer-report HEXACO. IQ = Intelligence. For the random effects,
there were 20 targets, six personality traits and 68 participants. * p < .05. ** p < .001.
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
28&
Table 5
Experiment 2: Regression Analyses
Predictor
Random Effects
Outcome
B
SE
95% CI
Bootstrap 95%
CI
t
p
Model 2.D
Similarity
Target; Personality Trait; Participant
SRH
0.17
0.01
[0.15, 0.19]
[0.15, 0.19]
16.34
<.001**
Model 2.E
Similarity
Target; Personality Trait; Participant
ORH
0.05
0.01
[0.03, 0.07]
[0.03, 0.07]
5.21
<.001**
Model 2.F
Similarity
Target; Personality Trait; Participant
IQ
0.15
0.19
[-0.22, 0.52]
[-0.69, 0.25]
0.81
0.42
Note. Similarity = Difference in personality between targets and participant. SRH = Self-report HEXACO. ORH = Observer-report HEXACO.
IQ = Intelligence. For the random effects, there were 20 targets, six personality traits and 68 participants. ** p < .001.
!
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
!
29&
Discussion
As predicted, Experiment 2 demonstrated that those with a more accurate
Mind-space were better able to locate specific targets within Mind-space.
Furthermore, similarity in personality to the target affected the accuracy of estimates
of personality traits, but not IQ.
In Experiment 3, we sought quantitative evidence that the location of a target
mind in Mind-space affects the probability of specific mental states being attributed to
that target mind. Arguably, this has not been demonstrated in Experiments 1 and 2;
for example, although Experiment 1 demonstrated an association between the
accuracy of Mind-space and the accuracy of mental state inference (an association
that was specific to mental state inference and therefore unlikely to be a product of
domain-general individual differences in, for example, inferential ability or
motivation), this association could be caused by individual differences in social-
specific factors, such as social attention, which independently influence the accuracy
of Mind-space and mental state inference, rather than the accuracy of Mind-space
directly influencing the accuracy of mental state inference. Accordingly, Experiment
3 used a variant of the Sally-Anne task to vary the position of one character (Sally)
within the participant’s Mind-space, and the other character (Anne) within Sally’s
Mind-space. It was predicted that movement of a target mind along dimensions of
Mind-space would alter the probability of specific mental states being attributed if
they are dependent upon those dimensions given a specific situation.
The classic false belief unseen change-of-location task used in this experiment
(the ‘Sally-Anne’ task) is a staple of ToM research (e.g. Baillargeon, Scott, & He,
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
30&
2010; Kulke, Reiß, Krist, & Rakoczy, 2017; Happé, 1994; Rabinowitz et al., 2018).
Experiment 3 modifies this simple task such that participants have to remember a
personality feature for both characters and make a probabilistic judgement about one
character’s behaviour. Due to the additional working memory requirements
introduced by the requirement to hold in mind the personality of the characters the use
of a simple task was preferred, although the simplicity may limit the size of any effect
observed.
Experiment 3
Method
Participants. Sixty-three adults volunteered to take part in this experiment in
return for a small monetary sum or undergraduate research participation credits.
Participants (51 female) were aged between 17 and 59 years old (M = 25.08, SD =
0.95). An a priori power calculation using G*Power (Faul, Erdfelder, Buchner, &
Lang, 2009) indicated that for a medium effect size and α = .05, a sample size of 24
would provide 80% power for the main hypotheses being tested (without covariates).
The local Research Ethics Committee approved the study.
Measures.
Mental State Stories. Thirty-two vignettes were presented to participants.
Each vignette featured two characters and an unseen change-of-location as in the
Sally-Anne False Belief task (Baron-Cohen et al., 1985). In each vignette: the ‘Sally’
character puts an object in a location; then leaves the scene during which time the
‘Anne’ character moves the object to a different location; ‘Sally’ later returns looking
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
31&
for her object. There were four Sally characters (Emily, Ben, Amelia, George) and
four Anne characters (Jessica, Oliver, Isabella, Jack). They are described as having
been “work colleagues for many years, so they all know one another very well”. Two
vignettes were presented for every combination of Sally and Anne characters.
Paranoia manipulation. The Sally characters were designed to vary across
four levels of paranoia. Participants were told that these characters completed a
questionnaire and were shown the questionnaire items and the characters’ scores. The
questionnaire items were three items taken from the Paranoia Scale (Fenigstein &
Vanable, 1992), a 20-item measure of paranoia for use in non-clinical populations.
The items were: It is safer to trust no one; I tend to be on my guard with people who
are somewhat more friendly than I expected; Some people have tried to steal my ideas
and take credit for them. Participants were told that the characters could score
anywhere between 0 and 4 on each statement, and therefore between 0 and 12 in total,
with higher scores indicating higher levels of agreement with the statements. Before
each set of stories for each combination of Sally and Anne characters, participants
were reminded of the items and the character’s score. The four levels of paranoia
corresponded to total scores of 0, 4, 8, and 12.
Dishonesty manipulation. The Anne characters were manipulated across four
levels of dishonesty using the same approach as for the Sally characters. The
questionnaire items were three items taken from the Honesty-Humility dimension of
the HEXACO personality inventory (Ashton & Lee, 2009). The items were: If I knew
I could never get caught, I would be willing to steal a million dollars; I’d be tempted
to use counterfeit money, if I were sure I could get away with it; If I want something
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
32&
from someone, I will laugh at that person’s worst jokes. The four levels of dishonesty
corresponded to total scores of 0, 4, 8, and 12.
Mental State Inference. After each mental state story, participants were asked
to respond on a sliding scale with the extremes of the scale labelled with two response
options. The options represented the two locations that in traditional unseen change-
of-location tasks with binary measures reflect a false or true belief (i.e. respectively,
where Sally knew the object to be last vs. where the object has been moved to by
Anne). Participants were asked to move the slider so that it represents the probability
that Sally will look in one of the two response locations. False and true belief options
were counterbalanced across the right and left ends of the scale. Responses were
coded so that a rating of 50 indicated neither location was more likely, ratings closer
to 100 indicated greater probability of the false belief location, and ratings closer to 0
indicated greater probability of the true belief location.
Manipulation check. After participants had completed all 32 mental state
stories, they were shown the trials again with the Sally and Anne characters’ scores
and vignettes, but without the mental state inference response scale. Instead, they
were asked to report, using a four-point Likert scale (from ‘not at all’ to ‘highly’):
How paranoid do you (the participant) think Sally is; How paranoid does Anne think
Sally is; How honest do you (the participant) think Anne is; How honest does Sally
think Anne is? This provided first and second-order inferences of the characters’
traits.
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
33&
Self-report Measures. Participants also completed the full Paranoia scale
(Fenigstein & Vanable, 1992); the Honesty-Humility subscale of the HEXACO
(Ashton & Lee, 2009); the Autism Spectrum Quotient 10 (AQ10; Baron-Cohen,
Wheelwright, Skinner, Martin, & Clubley, 2001), a measure of autistic traits (e.g.
attention to detail or others’ intentions); and the Perspective Taking Scale of the
Interpersonal Reactivity Index (IRI PT; Davis, 1983), a measure of the tendency to
consider another person’s point of view.
Procedure. Participants completed the study individually on a computer in a
testing room in a single session of approximately one hour. The measures were
presented in the following order: Mental State Stories [32 trials]; Manipulation
Check; AQ10; IRI PT; Paranoia Scale; Honesty-Humility HEXACO Scale.
Statistical analyses. The statistical analyses were conducted using a Repeated
Measures Analysis of Variance in SPSS (v24, IBM, Armonk, NY, USA) with
Paranoia (4 levels) and Dishonesty (4 levels) as the within-subject factors and the four
self-report measures as covariates. The dependent variable was the probability rating
on the mental state inference measure, which was the average rating of the two trials
for each combination of the factor levels. Where assumptions of sphericity were
violated, Greenhouse-Geisser corrected values are reported. Bonferroni corrections
were used to adjust the alpha level when conducting post-hoc multiple comparisons.
The data for this study are available at https://doi.org/10.17605/OSF.IO/4K9HS.
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
34&
Results
Descriptive statistics for all variables are presented in Table 6. There were no
significant effects of any of the covariates, and they were dropped from further
models (note this did not affect the pattern of results). The lack of any effect of the
covariates indicates that there was no relationship between participants’ traits and the
probability of their mental state inferences. There was a significant main effect of the
Sally character’s level of paranoia on the probability of the mental state inferred, F
(2.20, 136.11) = 57.96, p < .001, ηp2 = .48. There was also a significant main effect of
the Anne character’s level of dishonesty on the probability of the mental state
inferred, F (3, 186) = 15.93, p < .001, ηp2 = .20. These main effects were characterised
by a significant negative linear trend indicating a reduction in the probability ratings
of the Sally character looking in the location corresponding to a false belief, for both
Paranoia, F (1, 62) = 99.31, p < .001, ηp2 = .62, and Dishonesty, F (1, 62) = 32.12, p <
.001, ηp2 = .34 (full contrasts are shown in Table S2). The variables were not normally
distributed and the robustness of ANOVA to departures of normality is debated
(Glass, Peckham, & Sanders, 1972; Lix, Keselman, & Keselman, 1996), therefore two
Robust Repeated Measures One-way ANOVA with 4 Factor Levels using 2000
bootstrap samples in the WRS2 package in R (Mair & Wilcox, 2018) were also
carried out, and confirmed the results (Paranoia: F = 57.06, Fcrit = 2.95, p < .05;
Dishonesty: F = 14.58, Fcrit = 2.81, p < .05; Post hoc comparisons shown in Table
S3). The effects of paranoia and dishonesty on the probability of mental state
inferences are shown in Figure 2.
There was a significant interaction effect between Sally’s levels of paranoia
and Anne’s levels of dishonesty, F (7.13, 441.79) = 8.82, p < .001, ηp2 = .12. A simple
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
35&
effects analysis showed that Sally’s paranoia had an effect at all levels of Anne’s
dishonesty: Level 1: V = 0.70, F (3, 60) = 47.27, p < .001; Level 2: V = 0.47, F (3, 60)
= 17.86, p < .001; Level 3: V = 0.58, F (3, 60) = 27.62, p < .001; Level 4: V = 0.33, F
(3, 60) = 9.64, p < .001. Similarly, Anne’s dishonesty had an effect at all levels of
Sally’s paranoia: Level 1: V = 0.39, F (3, 60) = 12.58, p < .001; Level 2: V = 0.22, F
(3, 60) = 5.65, p = .002; Level 3: V = 0.51, F (3, 60) = 21.07, p < .001; Level 4: V =
0.15, F (3, 60) = 3.56, p = .019. Post hoc contrasts with corrections for multiple
testing are shown in Table S3. The interaction was mainly driven by differences
between levels 1 and 4 of Paranoia, with levels of Dishonesty having strongly
different effects at level 1 of Paranoia but more similar effects at level 4.
The ratings of the characters’ traits in the manipulation check are shown in
Tables S4 and S5. Overall, they show that participants correctly inferred the
characters’ levels of paranoia or dishonesty from the information provided about their
scores on the respective questionnaires.
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
36&
Table 6
Descriptive Statistics for Experiment 3
Variable
Mean
SD
Range
Mental State Probability:
Paranoia Level 1
77.74
23.98
0 - 100
Paranoia Level 2
72.22
21.46
0 - 100
Paranoia Level 3
57.47
23.02
6.5 - 100
Paranoia Level 4
48.94
26.04
0 - 100
Dishonesty Level 1
69.41
27.50
0 - 100
Dishonesty Level 2
65.76
24.11
0 - 100
Dishonesty Level 3
61.33
24.93
0 - 100
Dishonesty Level 4
59.87
27.49
0 - 100
Honesty-Humility
3.59
0.62
1.88 - 4.88
Perspective Taking
17.49
5.17
7 - 28
Autism Quotient
2.73
1.79
0 - 8
Paranoia
39.92
14.54
20 - 85
Note. Higher values on Mental State Probability indicate a higher probability of the
false belief location. SD = Standard Deviation.
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
37&
Figure 2. The effect of targets’ locations in Mind-space on the probability of the
mental state inferred. Note that higher values on the ‘False Belief Probability’ axis
indicate higher probabilities of searching in the ‘false belief’ location, that is, where
the Sally character left her object. Error bars show within-subject 95% confidence
intervals around the means (Morey, 2008).
Discussion
The results of Experiment 3 are consistent with the idea that participants
locate a target’s mind within Mind-space before inferring the target’s mental state,
and that the location of the target mind within Mind-space is used to infer the
probability of particular mental states. Specifically, the more paranoid that Sally was,
and the more dishonest that Sally thought Anne was, the less likely participants were
to predict that Sally would look in the location in which she left her object.
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
38&
It is interesting to note that although the probability of ascribing a false belief
to Sally decreased as paranoia and dishonesty increased, the probability ratings tended
not to dip below 50%. This indicates that Sally was not likely to look in the false
belief location, where she had left her object, but also not likely to look in the true
belief location, where Anne had moved her object. This is most probably attributable
to an aspect of the study design: although the stories mentioned only two locations as
in the original task (Baron-Cohen et al., 1985), participants may have inferred that
although Sally suspected her object had been moved, she did not know the exact
location it had been moved to by Anne. Future studies may find increased true belief
ratings by constraining the situational information further using pictorial stimuli
rather than vignettes.
Although the task used was relatively simple, one can see large effects of
changing the protagonists’ position in Mind-space, and the position of the other
character in the protagonist’s Mind-space. Given that there is no objectively correct
answer on this task, these results highlight the ambiguity in interpreting ‘failures to
represent the protagonist’s false belief’ in the standard version of the unseen change-
of-location task without further interrogation of participants’ reasoning. If the
participant attributes paranoia/distrust to others in the absence of a cue to do so, they
may respond in a manner which is typically interpreted as a failure to represent false
belief (Happé & Frith, 1996).
While the results of Experiment 3 are consistent with one of the central tenets
of the Mind-space theory - that the accuracy of mental state inference depends on the
accuracy of characterising the target mind – Experiment 3 was not designed to show
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
39&
that target minds are represented within a multi-dimensional space. Experiment 4
built upon the design of Experiment 3 in order to provide a more specific test of this
aspect of the Mind-space theory. Accordingly, participants completed the same false
belief vignettes task as used in Experiment 3 with a range of Sally characters.
However, in Experiment 4, participants were given information about the Sally
characters’ scores on a range of traits (not including paranoia) which were selected on
the basis of a validation study to covary with paranoia in the minds of a similar
population to that which participants in Experiment 4 were drawn from. If participants
represent minds within a multi-dimensional space in which covariances between
dimensions are also represented, and use target locations within Mind-space to inform
mental state inferences, then moving the Sally character on traits associated with
paranoia should result in modified mental state inferences. Crucially, the size of the
effect on mental state inference should vary for each participant as a function of the
degree to which each trait is associated with paranoia within that participant’s Mind-
space.
Experiment 4
Methods
Participants. 55 participants (24 female) took part in an online task (built
using the Gorilla Experiment Builder; Anwyl-Irvine, Massonnié, Flitton, Kirkham &
Evershed, 2018) of approximately 20 minutes for monetary compensation.
Participants were aged between 18 to 59 years old (M = 31.35, SD = 11.99), were
residing in the UK, and reported English as their first language. Five participants were
excluded prior to analysis after reporting mental health conditions in a screening
questionnaire. The sample size for Study 4 was calculated a priori using simulations
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
40&
(DeBruine & Barr, 2019; Brysbaert & Stevens, 2018) based on parameter estimates
from Study 3. The results of these simulations indicate that with N=28 there is more
than 80% power to detect an effect of magnitude similar to that observed in
Experiment 3 with an alpha of .05. Twenty-eight was therefore set as the minimum
sample size, but all participants volunteering to participate within the recruitment
window were tested. The local Research Ethics Committee approved the study.
Measures
Mental State Stories.
The same 32 vignettes used in Experiment 3 were also used in this
experiment.
Stimulus Validation Study
A validation study using an analogous format to the Personality Pairs Task
was devised in order to identify traits commonly associated with paranoia. In this
study, 50 participants were asked to rate the association between 102 traits and
paranoia using the same visual analogue scale as used in the Personality Pairs Task.
The validation study was conducted online with participants resident in the UK who
reported English as their first language. The results of this task were used to identify
words which were commonly associated with paranoia (both negatively and
positively) across participants (see Supplemental Materials Figure S2). Care was
taken to ensure that the selected traits were not mere synonyms or antonyms of
paranoia by cross-checking thesaurus entries (Thesaurus.com, Oxford English
Thesaurus). In addition, words were excluded using OpenMeaning
(http://www.openmeaning.org/viz/), an online platform which allows for the
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
41&
visualization of semantic spaces and provides a ranking of words of interest based on
their semantic relatedness to a target word (in this case paranoia). None of the
selected traits from the validation study appeared as one of the top 50 words
semantically related to paranoia. Following this process, the final traits used in the
experiment (known as ‘source traits’ hereafter) were carefree, rational, and trusting,
which are negatively correlated with paranoia, and superstitious, pessimistic and
cautious, which are positively correlated with paranoia.
Paranoia Manipulation: Study 4
As in Study 3, the ‘Sally’ characters were designed to vary across four levels
of paranoia. However, in Study 4 paranoia was manipulated using the source traits
which, on the basis of the validation study, were expected to result in Sally being
placed at different positions along the paranoia dimension within Mind-space if
covariation between traits is represented. Participants were told that the characters
completed a questionnaire where they responded to a number of questions of the
form: "Please rate the degree to which you would describe yourself as:" and then
each of the six source traits was presented. Participants were told that the characters
answered by choosing one of the following four options: Not at All, A Little Bit,
Somewhat, and Very Much. At Paranoia Level 1, the Sally character responded ‘Very
Much’ to the three traits negatively correlated with paranoia, and ‘Not at All’ to the
three traits positively correlated with paranoia; at Level 2, the responses were ‘A Little
Bit’ to the positive traits and ‘Somewhat’ to the negative traits; at Level 3, the
responses were ‘Somewhat’ to the positive traits and ‘A Little Bit’’ to the negative
traits; and at Level 4, the responses were ‘Very Much’ to the positive traits and ‘Not at
All’ to the negative traits. These responses were designed to allow participants to infer
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
42&
low paranoia at level 1 to high paranoia at level 4. Unlike Experiment 3, Study 4 did
not include any dishonesty manipulation for the Anne character.
Mental State Inference. Apart from the changes described above, the mental
state inference task was the same as in Experiment 3.
Explicit Paranoia and Association Ratings. After participants had completed
all 32 mental state inference trials, they were shown each Sally character’s
questionnaire responses again and asked to report, using a four-point Likert scale
(from ‘not at all’ to ‘highly’): “How paranoid do you think ‘Sally’ is?”. Following the
paranoia ratings, participants were asked to estimate the association between paranoia
and the six source traits used to manipulate Sally’s paranoia using the same method as
used in the Personality Pairs Task (see Table S7).
Statistical Analyses. Statistical analysis was conducted using Linear Mixed
Models implemented in the lme4 package (Bates, Maechler, Bolker & Walker, 2014)
in R. Experiment 4 is designed to test the predictions that: (1) participants locate
minds within Mind-space based on information they are given about particular source
traits; (2) they use that information to locate those minds on dimensions they believe
to be correlated with the source traits; and (3) they use the location of minds within
Mind-space to predict the probability of particular mental states. For these predictions
to be supported, the data must show that each participant locates a particular Sally
along the paranoia dimension according to the degree to which they believe the source
traits are correlated with paranoia, and that this affects the mental states they attribute
to that Sally character. Thus, a predicted relative paranoia score, for each participant
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
43&
and each Sally, was derived by multiplying Sally’s score on each source trait by the
degree to which that participant thought that source trait was associated with paranoia
(from the paranoia association ratings), and then summing across source traits. This
final Mean Predicted Relative Paranoia (MPRP) score represents where the
participant would locate each Sally on the paranoia dimension if Sally’s scores on the
source traits cause the participant to locate Sally on the paranoia dimension at a
location in accordance with the participant’s estimated correlation between the source
traits and paranoia.
MPRP was included as a fixed effect to predict the False Belief Probability
while controlling for trial and participant random intercepts (False Belief Probability
~ MPRP + (1 | trial) + (1 | participant). It was hypothesised that the higher the MPRP
(i.e. the more paranoid Sally was thought to be), the less likely it would be for
participants to attribute a false belief to Sally’s character.
Results
Descriptive statistics for the estimated probability of the ‘false belief’ location
as a function of Sally’s scores on the source traits are presented in Table S6. As
predicted, the model results show a significant effect of MPRP on the False Belief
Probability attribution (
b
= - 8.85, 95% CI [-10.65, -7.03], p < .001, see Figure 3 and
Table S8). Crucially, a model comparison including the MPRP model, a model with
the Sally source traits (unweighted by their correlation with paranoia) as a fixed
effect, and a null model, with all models carrying the same random effects structure,
was also performed. The results indicated the MPRP model was significantly better
than the null and the unweighted Sally source trait models (
c
2(1) = 83.4, p < .001, see
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
44&
Table S9). Examination of the AIC and BIC values also showed that the MPRP
model outperformed the Sally source traits model (ΔAIC = 32, ΔBIC = 42, where
differences of 6 are generally considered to be non-negligible (Burnham & Anderson,
1998)). Thus, results suggest that participants (1) use their estimate of the correlation
between the source traits and paranoia to estimate Sally’s location on the paranoia
dimension, and (2) use this information to inform their estimates of the probability of
Sally’s mental states.
As a manipulation check, we computed a slope that represents the change in
explicit paranoia ratings across levels of Sally’s scores on the source traits. This was
achieved by calculating, for each participant, the mean explicit paranoia rating, and
then mean-correcting each rating. Linear weights were then assigned for each level of
Sally source traits and the weighted sum of the explicit paranoia ratings computed (all
values for these computations are provided in the data file for this study in the OSF
archive https://doi.org/10.17605/OSF.IO/4K9HS). These slope values represent the
degree to which changing scores on the source traits (across Sally characters)
produces changes in explicit paranoia ratings for each participant. When tested
against zero using a one-sample t-test, the slopes were found to be significantly
different from zero (indicating that changing the Sally character’s scores on the
source traits caused explicit paranoia ratings to change; M = 8.27, 95% CI [7.52 –
9.01], t(48) = 22.22, p < .001. The same procedure was repeated on the MPRP data to
derive slope values that reflect the degree to which paranoia ratings would change as
a function of changing scores for the Sally character on the source traits, if
participants based the paranoia ratings on their estimated correlations between source
traits and paranoia. As expected, we found a significant positive correlation between
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
45&
the explicit paranoia judgement slopes and the MPRP slopes (r(47) = .40, p = .005).
Thus, the degree to which participants adapted their explicit paranoia judgements as a
function of Sally’s scores on the source traits, corresponded with the MPRP
calculated on the basis of participants’ judgements of the correlation between the
source traits and paranoia.
&
Figure 3. Effect of Mean Predicted Relative Paranoia (MPRP) score on ‘False
Belief’ Attribution. Shaded area represents the 95% confidence interval. MPRP is
calculated on the basis of the Sally character’s scores on various traits and the degree
to which each participant believes those traits to be associated with paranoia.
Discussion
Experiment 4 demonstrates that when provided with information about a
target’s mind that allows it to be located on a number of source dimensions,
participants use that information to extrapolate the location of the target mind on
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
46&
dimensions they believe covary with the source dimensions, and they do so in a
manner which reflects the degree of estimated covariation. Furthermore, they use the
estimated location on the new dimensions to make inferences about the target’s
mental states where relevant. This pattern of data is consistent with predictions from
the Mind-space theory, and also with previous demonstrations that, for example,
individuals are thought to have different mental states depending on their locations on
dimensions of warmth and competence (Fiske et al., 2002).
General Discussion
We sought to understand individual differences in theory of mind by testing a
theory in which other minds are represented in a multidimensional space. Within this
framework the position of a target mind within Mind-space is combined with
information about the situation the target is in, in order to infer the probability of the
target having particular mental states. Accordingly, individual differences in the
accuracy of mental state inferences may be explained by factors including the
accuracy of an individual’s Mind-space (i.e. the degree to which their Mind-space
accurately captures variance in other minds), and the ability to locate a target mind
accurately within Mind-space. Experiment 1 demonstrated that variance in ToM
ability (i.e. the accuracy of mental state inference) was associated with how
accurately the covariance between personality dimensions was represented within
Mind-space. Experiment 2 showed that the accuracy of Mind-space was associated
with the ability to locate another person within Mind-space, on dimensions relating to
personality traits and intelligence, based on a minimal sample of their behaviour. The
results obtained in Experiment 3 support the prediction that the location of a target
mind in Mind-space affects the probability of particular mental states being attributed
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
47&
to that target given the situation they are in. Experiment 4 extended this result to show
the dimensional nature of mind representation. Participants extrapolated from the
location of a target’s mind on source dimensions to estimate the target’s location on
novel dimensions of mind, and used this estimate to infer the probability of mental
states.
The results of Experiment 1 demonstrate a relationship between understanding
the structure of personality in the general population and the ability to make accurate
mental state inferences about particular characters. In designing the MASC task, the
authors ensured that each character had distinctive traits (e.g. outgoing vs. shy;
Dziobek et al., 2006). Implicit in this task is the relationship between the characters’
traits and the kind of mental states they generate, yet how traits and mental states
relate to one another has rarely been addressed, particularly in adulthood. It should be
acknowledged, however, that several trait theories of mind (person) representation
exist, and some of these theories specify that traits may be associated with differential
probabilities of particular mental states being inferred (for example the work on
stereotyping by Fiske et al., 2002; for a full discussion of such theories and their
relationship to Mind-space see Conway et al., 2019, ‘Relationship to existing
theories’, p.805). Of particular relevance is the work of Tamir and Thornton (Tamir &
Thornton, 2018; Tamir, Thornton, Contreras, & Mitchell, 2016), who argue that traits
are represented in a 3-dimensional space, and that traits can be used to infer the
probability of types of mental states (e.g. beliefs vs desires) and states of mind (e.g.
fatigued vs invigorated), which can also be represented in a 3-dimensional space.
Neuroimaging work has identified where in the brain traits and mental states may be
represented: activation in the temporo-parietal junction tends to occur when
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
48&
representing others’ thoughts or beliefs when they differ from one’s own (Saxe &
Powell, 2006; Koster-Hale, Richardson, Velez, Asaba, Young & Saxe, 2017),
whereas activation in the medial prefrontal cortex is thought to reflect representations
of specific people and their enduring social traits (Hassabis et al., 2014; Mitchell,
Cloutier, Banaji, & Macrae, 2006; Tamir et al, 2016; although see Cook, 2014).
However, the demonstration that there is brain activation specific to mental states vs.
traits does not provide a psychological account of how such information is used. The
Mind-space framework attempts to provide a model to link representation of a
particular mind and its qualities to inference of the mental states that this mind holds.
The findings of Experiment 1 support the idea that the quality of mind representation
may be a determinant of individual differences in theory of mind.
The results of Experiments 3 and 4 support the contention that mental state
inference is a process in which the probability of a particular mental state in a given
individual is inferred based on the learned probability of observing that mental state
given the context and the individual’s position in Mind-space (see Figure 1).
Accordingly, in addition to the factors studied in the current experiments, the
accuracy of mental state inferences is likely to be a product of two further factors: the
accuracy with which position in Mind-space is mapped to the probability of particular
mental states given a specific situation; and one’s propensity to consider the position
of the target mind in Mind-space before making a judgement as to the target’s mental
state. The finding that a less accurate Mind-space was associated with a lack of
mental state inference (Experiment 1) may be especially relevant to this last factor.
We speculated that an association between the accuracy of Mind-space and the ability
to locate a target mind within Mind-space may be due to common effects of social
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
49&
motivation, social attention, or social learning (Conway et al., 2019). Decreased social
motivation in particular may explain why an individual may form inaccurate models
of how minds vary, have a worse ability to locate minds within Mind-space, and be
less likely to make mental state attributions.
With respect to the finding that the accuracy of Mind-space predicts the ability
to locate others within Mind-space (Experiment 2), it is important to note that
participants were not highly accurate in their estimates. This inaccuracy is likely
attributable to the minimal exposure to the targets in the thin-slice videos. Predictive
accuracy has been shown to improve when thin-slices are extended for some traits, for
instance Carney, Colvin and Hall (2007) found good accuracy for judgements of
extraversion, conscientiousness, and intelligence after 5 seconds, whereas longer
exposure was required for neuroticism, openness to experience and agreeableness.
Whether the accuracy of an individual’s Mind-space predicts their ability to locate an
individual within Mind-space after longer exposure, or predicts their ability to
increase the accuracy with which they locate an individual after increased exposure,
remains to be determined. It should also be acknowledged that these results may hold
for only a small portion of Mind-space relating to personality. Personality represented
a good initial test of the Mind-space theory as there is a wealth of data available on
personality trait covariance, meaning that the accuracy of an individual’s model of
personality covariance can be established. However, whether these results would also
be found for other aspects of Mind-space with little or no relation to personality (e.g.
the factor structure of intelligence), also remains to be seen.
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
50&
One possibility suggested by these data is that individuals may not have a
unitary theory of mind ability, but rather that accuracy in the inference of mental
states, and in locating another mind within Mind-space, may depend upon the
particular mind to be modelled and its relationship to the kinds of minds one has
previously encountered which have shaped one’s Mind-space. This is supported by
the finding that greater similarity between participants and targets resulted in more
accurate trait judgements (Experiment 2), and that individuals use trait judgments
when inferring mental states (Experiments 3 and 4). Therefore, individuals who are
more typical of the population being represented (i.e. have average trait scores
themselves) are more likely to make accurate inferences about the minds and mental
states of others; both on average across inferences made for specific targets in the
population, and for targets about whom nothing is known where the optimal strategy
is to attribute average trait values to them.
Intriguingly, previous research on implicit personality theory indirectly
supports the contention that those who have typical trait covariances across a number
of dimensions make more accurate mental state inferences, but only if one accepts as
true the hypothesis that the accuracy of mental state inference depends upon the
accuracy of mind representation. Specifically, it has been demonstrated that an
individual’s model of personality is partly built upon their view of their own
personality: if they have a causal model explaining the patterning of traits in their own
personality (e.g. I am optimistic because I am intelligent and have always succeeded)
they are likely to assume the same patterning of traits in the general population (i.e.
that optimism is typically associated with intelligence; Critcher, & Dunning, 2009;
Critcher, Rom, & Dunning, 2015). Individuals with trait covariance typical of the
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
51&
population would therefore have a more accurate Mind-space if they based their
population model on their own personality; and if accuracy of mind representation
determines accuracy of mental state inference, they would make more accurate mental
state inferences as a result.
The idea that one’s theory of mind ability may depend on the target mind to be
represented has interesting implications for atypical groups. Neurotypical participants
may perform well on existing theory of mind tasks in which the ‘correct’ answers are
derived by neurotypical consensus (e.g. Dziobek et al., 2006), as their own mind is
similar to the average. Conversely, neurotypical participants may also have minds that
are particularly easy to represent by the majority of the population. In contrast, those
who have atypical minds may find it harder to represent the minds of neurotypical
individuals, and in turn, be harder for neurotypical individuals to represent (Edey,
Cook, Brewer, Johnson, Bird, & Press, 2016; Brewer et al., 2016). The same loss of
accuracy is likely to occur when we need to represent the minds and mental states of
out-groups (Sasson, Faso, Nugent, Lovell, Kennedy, & Grossman, 2017; Bruneau &
Saxe, 2012).
Related suggestions have been made previously; for instance, Happé and Frith
(1996) suggested that children diagnosed with Conduct Disorder may have a ‘theory
of nasty minds’, that may be adaptive to aversive developmental environments and an
accurate reflection, based on their prior experience, of how others think and behave.
In their study of mental state inference, children with Conduct Disorder performed
less well than typically developing children but better than those with Autism
Spectrum Disorder, and showed a particular ability for mental state inference in
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
52&
antisocial situations, such as bullying. Therefore, even in the absence of explicit
information about others’ traits, children with Conduct Disorder may ascribe more
negative mental states than the typical population due to inaccurately locating others
in Mind-space, and/or atypical mappings between locations in Mind-space and mental
states.
In sum, these studies try to account for variance in the ability of humans to
infer accurately the mental states of others. The empirical support for Mind-space
presented here highlights the importance of modelling minds when considering
individual differences in the representation and inference of others’ mental states,
personality, and intelligence.
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
53&
Author Contributions
J.R. Conway, C. Catmur, and G. Bird developed the study concept and design. Data
collection and stimulus development was performed by J.R.C., H.C. Cuve, S. Koletsi,
N. Bronitt, with additional assistance from M. Fagundez, and K. Overall. J.R.C. and
H.C.C. (Expt. 4) performed the data analysis and interpretation under the supervision
of M.P. Coll, C.C., and G.B. J.R.C. drafted the manuscript, J.R.C, H.C.C., M.P.C.,
C.C., and G.B. provided critical revisions, and all authors approved the final version
of the manuscript for submission.
Acknowledgments
This work was supported by an Economic and Social Research Council studentship
[Ref: 1413340] awarded to J.R. Conway.
Context
The current paper is the first empirical test of a new theoretical framework advanced
by the authors (Conway, Catmur, & Bird, 2019) that aims to explain individual
differences in the accuracy of mental state inferences (‘mentalizing’ or ‘theory of
mind’). This paper reports four studies testing the predictions of a new mechanistic
model of mentalizing – the ‘Mind-space’ model – which suggests that minds are
represented within a multidimensional space, much as faces are thought to be
represented within Face-space. This model recognizes that mental states are a product
of, and dependent upon, the specific mind that gives rise to them. Under this model,
therefore, individual differences in mentalizing ability can be explained by individual
differences in the ability to represent variance in minds, and in the ability to determine
the characteristics of another’s mind when attempting to infer their mental states. The
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
54&
Mind-space model presents a framework to understand variance in mentalizing
ability, which has implications for the study of this ability in clinical groups (most
notably Autism Spectrum Disorder), across childhood development, and its
implementation in artificial agents.
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
55&
References
Aboud, F. E. (1988). Children and prejudice. B. Blackwell.
Anwyl-Irvine, A. L., Massonnié, J., Flitton, A., Kirkham, N., & Evershed, J. K.
(2019). Gorilla in our Midst: An online behavioral experiment builder. Behavior
research methods, 1-20.
Ashton, M. C., & Lee, K. (2007). Empirical, Theoretical, and Practical Advantages of
the HEXACO Model of Personality Structure. Personality and Social Psychology
Review, 11(2), 150–166. https://doi.org/10.1177/1088868306294907
Ashton, M. C., & Lee, K. (2009). The HEXACO-60: A short measure of the major
dimensions of personality. Journal of Personality Assessment, 91(4), 340–345.
https://doi.org/10.1080/00223890902935878
Baillargeon, R., Scott, R. M., & He, Z. (2010). False-belief understanding in infants.
Trends in Cognitive Sciences, 14(3), 110–118.
https://doi.org/10.1016/j.tics.2009.12.006
Bates, D., Maechler, M., Bolker, B., Walker, S., Christensen, R.H.B., Singmann,
H., …Green, P. (2018). Linear Mixed-Effects Models using 'Eigen' and S4.
Retrieved from https://github.com/lme4/lme4/ http://lme4.r-forge.r-project.org/
Baron-Cohen, S., Leslie, A. M., & Frith, U. (1985). Does the autistic child have a
'theory of mind'? Cognition, 21(21), 37–46.
https://doi.org/https://doi.org/10.1016/0010-0277(85)90022-8
Baron-Cohen, S., O'riordan, M., Stone, V., Jones, R., & Plaisted, K. (1999).
Recognition of faux pas by normally developing children and children with
Asperger syndrome or high-functioning autism. Journal of autism and
developmental disorders, 29(5), 407-418.
Baron-Cohen, S., Wheelwright, S., Skinner, R., Martin, J., & Clubley, E. (2001). The
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
56&
Autism-spectrum Quotient (AQ): Evidence from Asperger Syndrome/High-
Functioning Autism, Males and Females, Scientists and Mathematicians. Journal
of Autism and Developmental Disorders, 31(1), 5-17.
Borkenau, P., Mauer, N., Riemann, R., Spinath, F. M., & Angleitner, A. (2004). Thin
Slices of Behavior as Cues of Personality and Intelligence. Journal of Personality
and Social Psychology, 86(4), 599–614. https://doi.org/10.1037/0022-
3514.86.4.599
Brewer, R., Biotti, F., Catmur, C., Press, C., Happe, F., Cook, R., & Bird, G. (2016).
Can Neurotypical Individuals Read Autistic Facial Expressions? Atypical
Production of Emotional Facial Expressions in Autism Spectrum Disorders.
Autism Research, 9(2), 262–271. https://doi.org/10.1002/aur.1508
Bruneau, E. G., & Saxe, R. (2012). The power of being heard: The benefits of
‘perspective-giving’ in the context of intergroup conflict. Journal of
Experimental Social Psychology, 48(4), 855–866.
https://doi.org/10.1016/j.jesp.2012.02.017
Brysbaert, M., & Stevens, M. (2018). Power analysis and effect size in mixed effects
models: A tutorial. Journal of Cognition, 1(1).
Burnett, S., Bird, G., Moll, J., Frith, C., & Blakemore, S. J. (2009). Development
during adolescence of the neural processing of social emotion. Journal of
cognitive neuroscience, 21(9), 1736-1750.
Burnham, K. P., & Anderson, D. R. (1998). Practical use of the information-theoretic
approach. In Model Selection and Inference (pp. 75-117). Springer, New York,
NY.
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
57&
Canty, A., & Ripley, B. (2017). Bootstrap Functions. Retrieved from https://cran.r-
project.org/web/packages/boot/boot.pdf
Carney, D. R., Colvin, C. R., & Hall, J. A. (2007). A thin slice perspective on the
accuracy of first impressions. Journal of Research in Personality, 41(5), 1054–
1072. https://doi.org/10.1016/j.jrp.2007.01.004
Champely, S., Ekstrom, C., Dalgaard, P., Gill, J., Weibelzahl, S., Anandkumar, A., …
Volcic, R., De Rosario, H. (2018). Basic Functions for Power Analysis. Retrieved
from https://github.com/heliosdrm/pwr
Conway, J.R., Catmur, C., & Bird, G. (2019). Understanding Individual Differences
in Theory of Mind via Representation of Minds, Not Mental States. Psychonomic
Bulletin and Review, https://doi.org/10.3758/s13423-018-1559-x
Cook, J. L. (2014). Task-relevance dependent gradients in medial prefrontal and
temporoparietal cortices suggest solutions to paradoxes concerning self/other
control. Neuroscience and Biobehavioral Reviews, 42, 298–302.
https://doi.org/10.1016/j.neubiorev.2014.02.007
Critcher, C. R., & Dunning, D. (2009). Egocentric pattern projection: How implicit
personality theories recapitulate the geography of the self. Journal of personality
and social psychology, 97(1), 1.
Critcher, C. R., Dunning, D., & Rom, S. C. (2015). Causal trait theories: A new form
of person knowledge that explains egocentric pattern projection. Journal of
personality and social psychology, 108(3), 400.
Davis, M. H. (1983). Measuring Individual Differences in Empathy: Evidence for a
Multidimensional Approach. Journal of Personality and Social Psychology,
44(1), 113–126. https://doi.org/10.1037/0022-3514.44.1.113
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
58&
DeBruine, L. M., & Barr, D. J. (2019, June 2). Understanding mixed effects models
through data simulation. https://doi.org/10.31234/osf.io/xp5cy
Dennett, D. C. (1978). Beliefs about beliefs. Behavioral and Brain Sciences, 4, 568–
570.
Devine, R. T., & Hughes, C. (2014). Relations between false belief understanding and
executive function in early childhood: A meta-analysis. Child Development,
85(5), 1777–1794. https://doi.org/10.1111/cdev.12237
Dziobek, I., Fleck, S., Kalbe, E., Rogers, K., Hassenstab, J., Brand, M., … Convit, A.
(2006). Introducing MASC: A movie for the assessment of social cognition.
Journal of Autism and Developmental Disorders, 36(5), 623–636.
https://doi.org/10.1007/s10803-006-0107-0
Edey, R., Cook, J., Brewer, R., Johnson, M. H., Bird, G., & Press, C. (2016).
Interaction takes two: Typical adults exhibit mind-blindness towards those with
autism spectrum disorder. Journal of Abnormal Psychology, 125(7), 879–885.
https://doi.org/10.1037/abn0000199
Epley, N., Keysar, B., Van Boven, L., & Gilovich, T. (2004). Perspective taking as
egocentric anchoring and adjustment. Journal of Personality and Social
Psychology, 87(3), 327–339. https://doi.org/10.1037/0022-3514.87.3.327
Faul, F., Erdfelder, E., Buchner, A., & Lang, A. G. (2009). Statistical power analyses
using G* Power 3.1: Tests for correlation and regression analyses. Behavior
research methods, 41(4), 1149-1160.
Fenigstein, A., & Vanable, P. A. (1992). Paranoia and Self-Consciousness. Journal of
Personality and Social Psychology, 62(1), 129–138.
https://doi.org/10.1037//0022-3514.62.1.129
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
59&
Fiske, S. T., Cuddy, A. J., Glick, P., & Xu, J. (2002). A model of (often mixed)
stereotype content: competence and warmth respectively follow from perceived
status and competition. Journal of personality and social psychology, 82(6), 878.
Gallagher, H. L., & Frith, C. D. (2003). Functional imaging of “theory of mind.”
Trends in Cognitive Sciences, 7(2), 77–83. Retrieved from
http://www.ncbi.nlm.nih.gov/pubmed/12584026
Glass, G. V., Peckham, P. D., & Sanders, J. R. (1972). Consequences of failure to
meet assumptions underlying the fixed effects analyses of variance and
covariance. Review of Educational Research, 42(3), 237-288.
Goldberg, L. R. (1990). An Alternative ‘Description of Personality’: The Big Five
Factor Structure. Journal of Psychology and Social Psychology, 59(6), 1216–
1229. https://doi.org/10.1037//0022-3514.59.6.1216
Happé, F. G. (1994). An advanced test of theory of mind: understanding of story
characters’ thoughts and feelings by able autistic, mentally handicapped, and
normal children and adults. Journal of Autism and Developmental Disorders,
24(2), 129–154. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/8040158
Happé, F. G., & Frith, U. (1996). Theory of mind and social impairment in children
with conduct disorder. British Journal of Developmental Psychology, 14, 385–
398.
Hassabis, D., Spreng, R. N., Rusu, A. A., Robbins, C. A., Mar, R. A., & Schacter, D.
L. (2014). Imagine all the people: How the brain creates and uses personality
models to predict behavior. Cerebral Cortex, 24(8), 1979–1987.
Heyes, C. (2014). Submentalizing: I am not really reading your mind. Perspectives on
Psychological Science, 9(2), 131-143.
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
60&
Heyes, C. (2015). Animal mindreading: what’s the problem? Psychonomic Bulletin &
Review, 22(2), 313-327.
Heyes, C. (2017). Apes Submentalise. Trends in Cognitive Sciences, 21(1), 1–2.
Koster-Hale, J., Richardson, H., Velez, N., Asaba, M., Young, L., & Saxe, R. (2017)
Mentalizing regions represent distributed, continuous, and abstract dimensions of
others’ beliefs. Neuroimage, 161, 9-18.
Kulke, L., Reiß, M., Krist, H., & Rakoczy, H. (2017). How robust are anticipatory
looking measures of Theory of Mind? Replication attempts across the life span.
Cognitive Development, (August), 0–1.
https://doi.org/10.1016/j.cogdev.2017.09.001
Lalonde, C. E., & Chandler, M. J. (2002). Children's understanding of interpretation.
New Ideas in Psychology, 20(2-3), 163-198.
Lee, K., & Ashton, M. C. (2016). Psychometric Properties of the HEXACO-100.
Assessment, 25(5), 543-556. https://doi.org/10.1177/1073191116659134
Lix, L. M., Keselman, J. C., & Keselman, H. J. (1996). Consequences of assumption
violations revisited: A quantitative review of alternatives to the one-way analysis
of variance F test. Review of Educational Research, 66(4), 579-619.
Mair, P., & Wilcox, R. (2018). WRS2: A Collection of Robust Statistical Methods.
Retrieved from: https://r-forge.r-project.org/projects/psychor/
Milligan, K., Astington, J. W., & Dack, L. A. (2014). Language and Theory of Mind :
Meta-Analysis of the Relation Between Language Ability and False-belief
Understanding. Child Development, 78(2), 622–646.
https://doi.org/10.1111/j.1467-8624.2007.01018.x
Mitchell, J. P., Cloutier, J., Banaji, M. R., & Macrae, C. N. (2006). Medial prefrontal
dissociations during processing of trait diagnostic and nondiagnostic person
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
61&
information. Social Cognitive and Affective Neuroscience, 1(1), 49–55.
https://doi.org/10.1093/scan/nsl007
Morey, R. D. (2008). Confidence Intervals from Normalized Data; A correction to
Cousineau (2005). Tutorials in Quantitative Methods for Psychology, 4(2), 61-64.
Nagel, R. (1995). Unraveling in guessing games: An experimental study. The
American Economic Review, 85(5), 1313-1326.
Osterhaus, C., Koerber, S., & Sodian, B. (2016). Scaling of advanced theory‐of‐mind
tasks. Child development, 87(6), 1971-1991.
Premack, D., & Woodruff, G. (1978). Does the chimpanzee have a theory of mind?
Behavioral and Brain Sciences, 4, 515–526.
Rabinowitz, N., Perbet, F., Song, H.F., Zhang, C., Eslami, S.M.A., & Botvinick, M.
(2018). Machine Theory of Mind. Retrieved from arXiv:1802.07740v2
Repacholi, B. & Slaughter, V. (Eds.) (2003). Individual differences in
theory of mind. Implications for typical and atypical development. New York:
Psychology Press.
Ruffman, T. (1996). Do children understand the mind by means of simulation or a
theory? Evidence from their understanding of inference. Mind & Language,
11(4), 388-414.
Sabbagh, M. A, Xu, F., Carlson, S. M., Moses, L. J., & Lee, K. (2006). The
Development of Executive Functioning and Theory of Mind. Psychological
Science, 17(1), 74–81. https://doi.org/10.1111/j.1467-9280.2005.01667.x
Santiesteban, I., Banissy, M. J., Catmur, C., & Bird, G. (2015). Functional
Lateralization of Temporoparietal Junction: Imitation Inhibition, Visual
Perspective Taking and Theory of Mind. European Journal of Neuroscience,
42(8), 2527-2533. https://doi.org/10.1093/biostatistics/manuscript-acf-v5
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
62&
Sasson, N. J., Faso, D. J., Nugent, J., Lovell, S., Kennedy, D. P., & Grossman, R. B.
(2017). Neurotypical Peers are Less Willing to Interact with Those with Autism
based on Thin Slice Judgments. Scientific Reports, 7(October 2016), 1–10.
https://doi.org/10.1038/srep40700
Saxe, R., & Powell, L. (2006). It’s the thought that counts: Specific brain regions for
one component of theory of mind. Psychological Science, 17(8), 692–699.
Tamir, D. I., & Thornton, M. A. (2018). Modeling the predictive social mind. Trends
in Cognitive Sciences, 22(3), 201–212.
Tamir, D. I., Thornton, M. A., Contreras, J. M., & Mitchell, J. P. (2016). Neural
evidence that three dimensions organize mental state representation: Rationality,
social impact, and valence. Proceedings of the National Academy of Sciences,
113(1), 194–199.
Valentine, T. (1991). A unified account of the effects of distinctiveness, inversion,
and race in face recognition. The Quarterly Journal of Experimental Psychology
Section A, 43(2), 161-204.
Valentine, T., Lewis, M. B., & Hills, P. J. (2016). Face-space: A unifying concept in
face recognition research. The Quarterly Journal of Experimental Psychology,
69(10), 1996-2019.
Wechsler, D. (2011). Wechsler Abbreviated Scale of Intelligence - Second Edition.
San Antonio, TX: NCS Pearson.
Wellman, H. M., & Liu, D. (2004). Scaling of theory of mind tasks. Child
development, 75(2), 523-541.
Wimmer, H., & Perner, J. (1983). Beliefs about beliefs: Representation and
constraining function of wrong beliefs in young children’s understanding of
deception. Cognition, 13, 103–128.
Running&Head:&UNDERSTANDING&HOW&MINDS&VARY&
63&
A preview of this full-text is provided by American Psychological Association.
Content available from Journal of Experimental Psychology General
This content is subject to copyright. Terms and conditions apply.
