Académique Documents
Professionnel Documents
Culture Documents
available at www.sciencedirect.com
article info
abstract
Article history:
Impaired ability to correctly infer others mental states (Theory of Mind ToM) and the ten-
mental tasks that provided indices of ToM and apophenia. The first task (Contingency) re-
quired participants to rate the strength of relationship between the movements of two
shapes; in some conditions the movements were unrelated, in others the movement of
one shape was contingent on that of the other. In the second task (Stories), participants an-
Keywords:
swered questions on short stories with random, physical or ToM themes. The third task
Theory of mind
(Triangles) required participants to view and then describe the movement of triangles on
Schizotypy
Psychosis-proneness
Delusional thinking
and delusional thinking style (Peters et al. Delusions Inventory PDI-21). Little evidence
Apophenia
was found on any of the tasks for an association between a specific deficit in ToM accuracy
and any aspect of psychosis-proneness. On the Triangles task, schizotypal and delusionprone participants spontaneously reported perceiving a greater number of associations in
the random condition (apophenia). On both the Triangles and Contingency tasks, overmentalizing was found to be associated with delusional thinking but not with general schizotypy. We speculate that over-mentalizing may be underpinned by a hyper-associative
cognitive style, linked to an exaggeration of the normal human tendency to attribute mental
states, rather than to dysfunction in the assessment of causality. Perceiving meaning in randomness and, more particularly, attributing mental states where none are indicated, may
be important factors in the formation of paranormal and delusional beliefs.
2008 Elsevier Srl. All rights reserved.
1.
Introduction
Studying social cognition can inform understanding of individual differences relevant to both psychopathological symptoms, such as those of schizophrenia, and unusual
phenomena in healthy individuals, such as paranormal or
* Corresponding author. Sub-Department of Clinical Health Psychology, University College London, Gower Street, London WC1E 6BT, UK.
E-mail address: g.pickup@ucl.ac.uk (G.J. Pickup).
0010-9452/$ see front matter 2008 Elsevier Srl. All rights reserved.
doi:10.1016/j.cortex.2007.07.009
1317
1318
2.
Method
2.3.
Experimental tasks
2.1.
Participants
2.3.1.
2.2.
Assessment instruments
2.2.1.
The STA of the STQ was used. This 37-item self-report questionnaire is based on DSM-III (American Psychiatric Association, 1980) criteria for Schizotypal Personality Disorder.
Participants answer yes or no to questions such as, Do
you believe in telepathy? and Do you often have vivid
dreams that disturb your sleep? The STA is an extensively
used scale with good criterion validity and testretest reliability (Jackson and Claridge, 1991). Factor analytic studies of the
STA have usually reported three factors in general population
samples, namely magical thinking, unusual perceptual experiences and paranoid ideation and suspiciousness (Hewitt
and Claridge, 1989; Joseph and Peters, 1995).
2.2.2.
2004)
which moves across the screen at a constant speed, and a Reactive Mover, whose movement varies according to four conditions (each consisting of five animations). In the AnimateContingent (AC) condition the Reactive Mover rotates and appears to see (animacy cue) the Prime Mover as it passes the
gap in a wall. In the Animate Non-contingent (AN) condition
the Prime Mover is out of sight (animacy cue) of the Reactive
Mover when the Reactive Mover moves. In the Inanimate Contingent (IC) condition the Prime Mover collides with the Reactive Mover, which appears to be launched into motion (no
animacy cue). In the Inanimate Non-contingent (IN) condition,
the Prime Mover moves across the screen, passing the Reactive Mover, which remains stationary (no animacy cue).
Participants were initially shown an example from each
condition. Each test stimulus was then presented twice (items
were presented in a pseudo-random order). After the first presentation of each stimulus, participants were asked to watch
the film again and respond Yes or No to the question,
In your opinion, was there a relationship direct or indirect
between the movements of the two shapes? After the second
presentation, participants responded to this question and
were then asked to rate the strength of relationship between
the movements of the two shapes on a scale from 0 to 10, to
provide a contingency rating variable. Finally, participants
were asked, What do you think was happening in the animation you just saw? In an addition to the methodology used by
Blakemore et al. (2003), responses to this final question were
transcribed verbatim and an over-mentalizing score was calculated for each condition (AC, AN, IC, IN). Participants scored
1 for a given condition if they elaborated their answer to any of
the five items with an attribution of agency and/or intention
(e.g., it twisted in preparation to block the other shape from
getting through the gap) or animacy (e.g., the green shape
moved across the blue shapes path of vision so the blue one
turned to look at it). If none of their answers were elaborated
in this way, they scored 0 for that condition. Scoring was carried out blind to ratings on all other measures. A second rater,
blind to the study hypotheses, scored a sub-set of 20 data sets;
agreement between raters was good (90%).
2.3.2.
This task comprises 16 stories, eight ToM and eight physical, each followed by a question. The ToM stories describe
scenarios including double bluff, white lies and persuasion,
in which participants must infer the mental state of a story
character to answer the question correctly. Happe (1994)
found these stories to discriminate between high-functioning
autistic participants (who passed standard false-belief tasks)
and non-autistic controls. The physical stories were matched
for difficulty to the ToM stories in a non-clinical adult sample
(Fletcher et al., 1995). Like the ToM stories, they involve people, and participants are required to integrate story information. However, correct responding requires only inference
about physical causation rather than mentalizing. An additional eight random stories were developed for the purpose
of this study (a full list is available from the authors). These
were matched in length to the original stories, were based
around comparable themes and involved the same number
of story characters. Each passage described a series of unrelated events and was followed by a question (e.g., Why did
1319
2.3.3.
This task was designed to measure on-line mental state attributions, and has potentially greater ecological validity than
stories tasks, which, in addition to ToM, require intact memory and other cognitive functions for correct responding.
The task has been shown to differentiate high-functioning autistic children who passed standard false-belief tasks from
non-clinical controls and participants with general intellectual impairment (Abell et al., 2000). It comprises 12 short, computer-presented animations showing one large, red and one
small, blue triangle moving around the screen. There are three
conditions: random, goal-directed and ToM. In the random
animations, the triangles move about purposelessly and do
not interact with each other. The goal-directed animations
involve one triangle responding to the others behaviour;
there is no implication that either triangle is reading the others mind and the animations are likely to evoke direct descriptions of interaction (e.g., fighting or dancing). The ToM
sequences, in contrast, show one triangle reacting to the others mental state (e.g., trying to persuade the other to let it
free) and are designed to provoke explanations that refer to
these mental states. For the purposes of this study, the
goal-directed condition will be termed the physical condition as it is broadly analogous to that in the Stories task.
Participants were shown two practice items, one physical
and one ToM, followed by the test animations (in a pseudorandom order). After viewing each test item, participants
described what they perceived to be happening in the animation. Responses were recorded onto audiotape, transcribed,
and scored for accuracy according to standard criteria (maximum score 8 for each animation type; Abell et al., 2000). An
apophenia score was calculated by summing the number of
connections and elaborations given in responses to random
items. Any statement that referred to a connection between
1320
2.4.
Control tasks
2.4.1.
2.4.2.
2.5.
Procedure
3.
Results
3.1.
3.1.1.
STA scores
3.1.2.
PDI-21 scores
3.2.
Experimental tasks
On the Stories and Triangles tasks, most of the dependent variables were normally distributed with no outliers. On the Stories
task, there was one low outlier (defined as a score greater than 3
SDs from the mean) in the ToM accuracy data for the high STA
and high PDI groups; the outlying score was changed to one
lower than the next lowest score (Howell, 1997). On the Stories
and Triangles tasks, there was one high outlier in the apophenia
scores for the high STA and high PDI groups; these were
changed to one higher than the next highest score. The distributions of apophenia scores on the Triangles task were positively skewed; a square root transformation was applied to
increase the normality of these distributions for subsequent
parametric analyses. All Stories and Triangles data were analysed parametrically, but the dependent variables on the Contingency task required non-parametric treatment. All reported
p values are two-tailed.
3.2.1.
1321
Table 1 Descriptive statistics for all measures by median split on STA scores
High STA
Low STA
Mean (SD)
Range
Mean (SD)
Range
25.61 (5.07)
93.61 (39.36)
1736
16177
9.07 (4.43)
36.20 (26.57)
016
0104
Significance
of group
difference
p < .001
p < .001
5.20
9.32
3.59
.88
1.26
1.16
(2.12)
(1.25)
(2.24)
(2.08)
(.44)
(.37)
1.69.2
4.810.0
08.8
010.0
1.02.0
1.02.0
3.91
9.21
3.11
.57
1.17
1.20
(2.13)
(1.10)
(2.29)
(1.40)
(.38)
(.41)
010.0
5.810.0
010.0
05.6
1.02.0
1.02.0
12.93
11.45
16.10
3.29
(2.43)
(2.94)
(10.84)
(1.68)
3.016.0
5.016.0
5.057.0
07.0
13.70
12.03
12.77
2.77
(1.64)
(2.27)
(6.24)
(1.68)
1016.0
8.016.0
1.027.0
06.0
p .340,
p .558,
p .205,
p .256,
5.29
6.52
5.39
1.42
(1.44)
(1.18)
(5.90)
(1.23)
2.08.0
3.08.0
025.0
04.0
5.67
6.93
2.83
1.10
(1.52)
(.87)
(4.15)
(1.54)
3.08.0
5.08.0
014.0
04.0
p .169, ns
p .058, ns
p .010
p .189, ns
108.68 (4.85)
109.48 (4.28)
6.29 (1.01)
99117
100117
39
111.20 (5.02)
111.73 (4.50)
6.67 (1.09)
97119
99118
49
p .05
p .05
p .17, ns
ns
ns
ns
ns
Table 2 Descriptive statistics for all measures by median split on PDI-21 scores
High PDI-21
Low PDI-21
Mean (SD)
Range
Mean (SD)
Range
24.16 (6.61)
100.26 (32.55)
836
65177
10.57 (6.92)
29.33 (17.34)
028
059
Significance
of group
difference
p < .001
p < .001
5.25
9.35
3.66
.74
1.32
1.23
(2.01)
(1.21)
(2.23)
(1.91)
(.48)
(.43)
1.69.2
4.810.0
08.8
010.0
1.02.0
1.02.0
3.86
9.17
3.05
.71
1.10
1.13
(2.21)
(1.14)
(2.29)
(1.65)
(.31)
(.35)
010.0
5.810.0
010.0
05.6
1.02.0
1.02.0
13.0
12.32
15.03
3.26
(2.42)
(2.37)
(10.36)
(1.75)
3.016.0
7.016.0
1.057.0
07.0
13.63
11.13
13.87
2.80
(1.69)
(2.78)
(7.39)
(1.61)
1016.0
5.016.0
3.036.0
06.0
p .321,
p .077,
p .793,
p .292,
5.39
6.55
6.42
1.81
(1.43)
(1.18)
(5.94)
(1.40)
2.08.0
3.08.0
025.0
04.0
5.57
6.90
1.77
.70
(1.55)
(.88)
(2.97)
(1.15)
3.08.0
5.08.0
011.0
04.0
p .639, ns
p .194, ns
p < .001
p .001
109.19 (5.22)
109.94 (4.57)
6.29 (1.13)
99119
100118
39
110.67 (4.86)
111.27 (4.39)
6.67 (.96)
97117
99117
59
p .26,ns
p .25, ns
p .17, ns
ns
ns
ns
ns
1322
were investigated using a mixed 2 2 analysis of covariance (ANCOVA), with VIQ and PIQ as covariates. There
was no main effect of subject group [F(1, 57) .96, p
.332], no main effect of story type [F(1, 57) .02, p .897],
no subject group story type interaction [F(1, 57) 0,
p .977] and no significant effects of IQ. Group differences
between high and low PDI participants on the same tasks
were investigated using a mixed 2 2 analysis of variance
(ANOVA). This revealed no main effect of subject group
[F(1, 59) .78, p .381]. However there was a significant
main effect of story type [F(1, 59) 21.16, p < .001], reflecting
poorer performance of participants in the physical condition than the ToM condition. There was a significant
group story type interaction [F(1, 59) 4.94, p .030],
which was due to the low PDI group performing worse on
the physical stories than on the ToM stories [t(29) 4.46,
p < .001], whereas the high PDI group performed comparably on both story types [t(30) 1.84, p .076].
3.2.1.2. TRIANGLES
3.2.2.
3.2.2.3.2. Animate contingency ratings. Contrary to predictions, high STA and high PDI groups rated the relationship between the shapes movement in the AC condition as
significantly stronger than in the AN condition [high STA:
Z(31) 3.75, p < .001; high PDI: Z(31) 3.61, p < .001]. The
low STA and low PDI groups also rated the contingency of
AC as greater than AN, although unexpectedly there was
a smaller effect size here [low STA: Z(30) 1.98, p .047;
low PDI: Z(30) 1.93, p .053].
3.2.2.3.3. Comparison between groups. Following Blakemore
et al. (2003), the differences (AC AN) and (IC IN) were calculated for each participant, and a score of 0 or 1 was given to
each participant for the animate and inanimate conditions
according to whether or not they made a normal contingency
judgement (i.e., AC > AN or IC > IN). Scores were then compared for high and low STA, and high and low PDI groups using c2 tests. In all cases the group differences were not
statistically significant, reflecting the fact that the groups did
not differ in levels of normal responding [animate STA:
c2(1) 2.08, p .150; animate PDI: c2(1) .81, p .367; inanimate STA: c2(1) .98, p .321; inanimate PDI: c2(1) .98,
p .321].
3.2.3.
3.2.3.1. RANDOM STORIES. ANCOVA with VIQ and PIQ as covariates was used to investigate group differences in over-mentalizing between high and low STA participants. There was no
significant main effect of group [F(1, 57) 1.32, p .256], and
no significant effects of IQ [VIQ: F(1, 57) .09, p .766; PIQ:
F(1, 57) .09, p .763]. No significant difference in over-mentalizing between high and low PDI participants was found
[t(59) 1.06, p .292].
4.
Discussion
1323
1324
see how the tendency to see meaningful explanations for random triangle movements might relate to both an ability to
generate more ideas and a tendency to impute an association
that is not obviously indicated. Although no data on
functional hemispheric differences are available for the present tasks, we would hypothesise a RH basis for the apophenia
findings consistent with the results of Folley and Park (2005)
and Brugger and Taylor (2003). It is important to note in this
context that we were not asking participants to make a cognitive effort to generate ideas, merely to report what spontaneously occurred to them perhaps tapping more genuine
apophenia than tasks with instructions to make associations.
The relationship between implicit and explicit measures of
apophenia could be tested in future studies by administering
the Triangles task and other tests of creative association and
pattern detection to a single sample of participants.
The predicted association between high delusion proneness and a tendency to over-mentalize in random situations
was supported on the Triangles and Contingency tasks, although no association was found between general schizotypy
and over-mentalizing. In the random condition of the Triangles task and the AN condition of the Contingency task, high
PDI scores were associated with more over-mentalizing. This
finding provides support for suggestions made by Abu-Akel
(1999), Abu-Akel and Bailey (2000) and Frith (2004) that the
positive symptoms of schizophrenia are associated with
over-mentalizing. An obvious next step would be to administer the tasks used in the present study to a sample of patients
with delusions, and to analyse responses for agency attribution in the same way as was done here. In a recent study using the Triangles task, Russell et al. (2006) found that
a paranoid subgroup of patients with schizophrenia tended
to use more mentalizing terms than other patient subgroups
and controls in the random condition of the task, a result
which is clearly consistent with our findings. Functional imaging studies of mentalizing abilities in healthy individuals have
shown that three brain areas are activated when ToM tasks
are performed: the anterior paracingulate cortex, the superior
temporal sulci and the temporal poles bilaterally (Gallagher
and Frith, 2003). It has been suggested that the anterior paracingulate cortex in particular is involved in the attribution of
mental states to others (Gallagher and Frith, 2003), and one
possibility is that over-mentalizing involves overactivation
of this medial prefrontal region. The basis of our over-mentalizing findings remains speculative, and begs the question of
whether over-mentalizing is a consequence of a creative
hyper-associative style in the context of intact causal reasoning, or whether it involves some dysfunction in the assessment of causality. Evidence from a recent brain imaging
study (Satpute et al., 2005) suggests that associative and
causal processing involve both shared and distinct areas of
the prefrontal cortex. Left prefrontal cortical areas were implicated in causal processing, and the right superior temporal gyrus in associative processing. We would speculate that the
over-mentalizing evidenced in the present study is more likely
to relate to hyper-associative processing and apophenia (and
thus to the RH), than to a dysfunction in causal processing
(with its basis in the LH) per se, although causal attributions
may also be affected. In this sense, what we have termed
over-mentalizing should not be taken to necessarily imply
Acknowledgements
This research was funded by the Central Research Fund of the
University of London, and the Sub-Department of Clinical
Health Psychology, University College London. We are grateful
to Dr Sarah-Jayne Blakemore for allowing us to use the Contingency task, and to Dr Francesca Happe for providing us with
the Stories task and its scoring criteria. We are grateful to all
participants in the study.
references
1325