Psychiatry Research 196 (2012) 207–213

Contents lists available at SciVerse ScienceDirect

Psychiatry Research
journal homepage: www.elsevier.com/locate/psychres

Mentalizing in first-episode psychosis

Amélie M. Achim a, b,⁎, Rosalie Ouellet a, c, Marc-André Roy a, b, Philip L. Jackson a, c
a
Centre de Recherche de l'Institut Universitaire en Santé Mentale de Québec, Québec, QC, Canada
b
Département de Psychiatrie et Neurosciences, Université Laval, Québec, QC, Canada
c
École de Psychologie, Université Laval, Québec, QC, Canada

a r t i c l e i n f o a b s t r a c t

Article history: Mentalizing deficits have often been observed in people with schizophrenia and a few recent studies suggest
Received 9 December 2010 that such deficits are also present in patients with first episode psychosis (FEP). It is not clear, however,
Received in revised form 7 October 2011 whether these mentalizing deficits in FEP can be accounted for by underlying processes such as social cue
Accepted 11 October 2011
recognition, social knowledge and general reasoning. In this study, we assessed mentalizing abilities in 31
people with FEP and 31 matched controls using a novel, comprehensive mentalizing task validated through
Keywords:
Recent-onset
the present study. We also assessed social cue recognition, social knowledge and non-social (or general) rea-
Schizophrenia soning performance in the same participants in order to determine if the mentalizing deficits in FEP can be at
Mentalizing least partly explained by performance in these three underlying processes. Overall, the mentalizing task
Theory of mind revealed the greatest impairment in FEP, an impairment that remained significant even after controlling
Social cognition for social cue recognition, social knowledge and non-social reasoning performance. Interestingly, non-
Emotion recognition social reasoning and social knowledge were both shown to contribute to mentalizing performance. In addi-
Social knowledge tion, social cognition measures were linked to social functioning in the FEP group, with the strongest corre-
lation observed with mentalizing performance. Taken together, these results show that mentalizing is an
aspect of social cognition that is particularly affected in FEP and might contribute to functional impairments
in these patients. These deficits could be a prime target for cognitive remediation in FEP, and our results sug-
gest that this could be done either directly or through improvement of related social and non-social cognitive
skills such as social knowledge and general reasoning.
© 2012 Elsevier Ireland Ltd. All rights reserved.

1. Introduction these deficits have a significant negative impact on social functioning

and the quality of life of patients with SZ (Green, 1996; Green et al.,
Social cognition can be defined as a group of processes that allow 2000; Couture et al., 2006). However, most studies of social cognition
people to understand and interact with each other. The ability to at- in SZ have included patients in a chronic stage of illness and/or during
tribute mental states to others, often referred to as mentalizing or active psychotic episodes (Sprong et al., 2007; Bora et al., 2009), and
theory of mind, is an important aspect of social cognition. In mentaliz- the mentalizing deficits observed in these patients could thus be
ing tasks as in real life, mental states such as intentions, beliefs, linked to biases towards the recruitment of patients with a more
knowledge or emotions are attributed based on all available sources chronic course of illness and a poorer outcome, effects of long-term
of information about the person to whom these mental states are to medication use, transient perturbation due to positive symptoms,
be attributed and about the context in which that person evolves. etc. In an attempt to address some of these biases, a few studies
Mentalizing judgments thus involve integrating several pieces of in- have measured mentalizing in remitted SZ patients, again revealing
formation in order to infer the appropriate mental state. significant deficits in these patients, though of a lesser magnitude
People with schizophrenia (SZ) generally present with important than those observed in acutely psychotic patients according to a re-
impairments in their ability to mentalize, i.e., to infer the mental cent meta-analysis (Bora et al., 2009). These studies in remitted pa-
state of a character presented in a given situation. Mentalizing deficits tients suggest that mentalizing deficits represent a trait of SZ,
have been repeatedly reported in people with SZ (Sprong et al., 2007; instead of being fully linked to symptomatic states. Studies in remit-
Bora et al., 2009), along with other neurocognitive impairments ted patients (e.g. Herold et al., 2002) however cannot account for
(Heinrichs and Zakzanis, 1998), and it is now well established that the effect of long-term medication use or the long duration of psycho-
sis and it also remains unclear when in the illness process mentalizing
impairments occur. One approach to minimize the impact of illness
⁎ Corresponding author at: Centre de Recherche de l'Institut Universitaire en Santé
Mentale de Québec (F-4500), 2601 de la Canardière, Québec, QC, Canada, G1J 2G3.
duration, get a sample representative of diverse future outcomes
Tel.: + 1 418 663 5741, + 1 418 663 9540. and address the issue of when the deficits occur is to recruit patients
E-mail address: [email protected] (A.M. Achim). early after the onset of psychosis. This is precisely what several

0165-1781/$ – see front matter © 2012 Elsevier Ireland Ltd. All rights reserved.
doi:10.1016/j.psychres.2011.10.011
208 A.M. Achim et al. / Psychiatry Research 196 (2012) 207–213

studies have done and these studies have also reported mentalizing risperidone either in oral (mean dose = 2.3 mg/day [three patients]) or intra-
muscular long-acting form (mean dose = 25 mg/2 weeks [three patients]; one patient
deficits in people with first-episode psychosis (FEP) (Inoue et al.,
was taking both formulations) and three taking a combination of quetiapine and an-
2006; Bertrand et al., 2007), in line with the suggestion that these other antipsychotic medication. Treatment had been initiated on average 20.9 months
deficits could be a general characteristic of patients with SZ. The prior to the study (median = 13.3 months, range = 1 to 57). Though we favored pa-
first FEP study on mentalizing by Inoue et al. (2006) relied on a single tients with short illness duration (less than 24 months), we also included patients
cartoon story in which three mentalizing questions were asked. This who had been followed for up to 60 months in order to get a more important sample
size. Symptoms were assessed with the Positive and Negative Syndrome Scale
simple task revealed a greater percentage of patients than controls (PANSS) (Kay et al., 1987) and level of social functioning was assessed with the Social
who responded incorrectly to one of the three questions. In a subse- and Occupational Functioning Assessment Scale (SOFAS) (American Psychiatric
quent study (Bertrand et al., 2007), a mentalizing deficit in FEP was Association, 1994).
again observed using the Hinting task (Corcoran et al., 1995), which Thirty-one control participants were recruited from the community through adds
in local stores, local media and through word of mouth. The exclusion criteria for the
includes ten short stories from which participants have to infer a
control group were the same as those of the FEP group, with the additional require-
character's intentions. Other more global measures of social cognition ment of not presenting with a psychotic disorder or a cluster A personality disorder
that likely involve mentalizing abilities have also revealed significant (as assessed through a clinical interview based on the Structured Clinical Interview
deficits in FEP relative to healthy controls (Bertrand et al., 2007; for DSM-IV (SCID) (First et al., 1998)), and not having a first-degree relative with a psy-
Koelkebeck et al., 2010). With the recent focus on early intervention, chotic disorder. Because we wanted to recruit control participants that were otherwise
as representative as possible of the general population, we however did not exclude
and given the relationship between social cognition and social func- controls that our SCID-based assessment allowed us to identify as presenting with
tioning, identifying and treating these mentalizing deficits early other axis I diagnoses. The control group thus included three participants that we iden-
could have a significant positive impact on the functional outcome tified as meeting the criteria for mild to moderate substance abuse or dependence and
of people with SZ. two participants identified as meeting the criteria for a specific phobia. None of the
control subjects had received a diagnosis and none reported taking a psychoactive
The mentalizing deficits observed in people with SZ could howev-
medication to control these symptoms.
er result from difficulties at different points in the information pro- The groups were matched in terms of age, gender and parental socio-economic
cessing chain that leads to mental states attributions (see Bless et background (Miller, 1991). There was no significant difference in estimated IQ be-
al., 2004; Brunet-Gouet et al., 2011), including the ability to recognize tween the two groups (see Table 1). After a complete description of the study, all par-
social stimuli (social cue recognition) (Edwards et al., 2001; ticipants signed a consent form in accordance with the local ethics committee
requirements.
Kucharska-Pietura et al., 2005; Addington et al., 2006a; Kohler et al.,
2010), the ability to construct and retrieve social representations (so-
2.2. Material
cial knowledge/memory) (Cutting and Murphy, 1990; Addington et
al., 2006b; Kee et al., 2009) and/or general reasoning/inferential 2.2.1. Mentalizing and non-social reasoning assessment
mechanisms (Young and Bentall, 1997). The aims of the current For the measure of mentalizing abilities, no standardized or validated tests are yet
study were thus 1) to replicate the results of mentalizing impair- available and to the best of our knowledge, very few of the most widely used tests have
been translated into French. Moreover, most of the available measures include a limit-
ments in people with a FEP relative to control subjects using a
ed number of items and suffer from ceiling effects at least in the control group, which
novel, comprehensive mentalizing task; 2) to assess distinctively so- can be problematic when contrasting with the performance of a patient group. For
cial cue recognition, social knowledge and general reasoning perfor- these reasons and to increase the sensitivity of our measure, we herein used a combi-
mance in the same two groups of participants; 3) to determine if nation of mentalizing stories that included items translated and adapted from several
mentalizing performance/deficits can be at least partly explained by previous tests that are all well regarded and have been used often in the experimental
psychology literature. These include the ‘Hinting task’ (Corcoran et al., 1995), the ‘False
performance in these underlying processes. These more elementary Belief task’ (Baron-Cohen, 1989; Frith and Corcoran, 1996), the ‘Faux-pas test’ (Baron-
processes have not been concurrently examined in previous studies Cohen et al., 1999) and the ‘Strange Stories Test’ (Happe, 1994). These tests were tar-
on mentalizing abilities in FEP and their assessment could provide geted so as to have items that cover attributions of a full range of mental states,
valuable information about the pathways leading to mentalizing im-
pairments in these patients. Since there are no standardized menta-
lizing tests as of yet and given that previous measures have often Table 1
Demographic and clinical data.
presented with ceiling effects and lack of sensitivity (e.g. Herold et
al., 2002; Inoue et al., 2006), mentalizing was measured here with a FEP Controls pa
comprehensive task developed for the purpose of this study, i.e. the Demographic data
Combined stories test, for which we also present initial psychometric n 31 31
properties. Gender (men/women) 26/5 26/5
Age (mean, S.D.) 24.9 (4.5) 25.2 (4.2) NS
SES score (mean, S.D.) 50.7 (18.2) 48.6 (14.2) NS
2. Methods
SES category (mean, S.D.) 3.7 (1.2) 3.6 (1.0) NS
Estimated IQ (mean, S.D.) 100.4 (15.1) 101.8 (10.5) NS
2.1. Participants
Education categoryb 4.0 (1.1) 3.3 (1.2) 0.03

Thirty-one participants with a FEP were recruited from the Clinique Notre-Dame-
Clinical data
des-Victoires, a specialized outpatient clinic that offers comprehensive evaluation and
PANSS positive (mean, S.D.) 15.1 (4.8)
treatment for young adults (18 to 35 years old) who are in the early stages of a psycho-
PANSS negative (mean, S.D.) 16.0 (5.9)
sis. All patients presented with a DSM-IV SZ spectrum psychosis diagnosis (American
PANSS general (mean, S.D.) 32.0 (7.0)
Psychiatric Association, 1994) including SZ (n = 23), schizoaffective disorder (n = 2),
SOFAS (mean, S.D.) 58.1 (12.1)
delusional disorder (n = 4), and psychosis not otherwise specified (n = 2). Our deci-
Patient status (outpatient/inpatient)c 28/3
sion to include patients with this range of diagnoses was based on our objective to in-
Duration of illness in months (mean/median) 20.9/13.3
clude a sample representative of all patients with a SZ spectrum psychotic disorder and
on previous reports that these diagnoses fall within the SZ spectrum when diagnoses S.D. = standard deviation.
are reassessed later in the course of the illness (Schimmelmann et al., 2005; Malla et SES = socio-economic status.
al., 2006) or based on family studies (Kendler et al., 1995; Schimmelmann et al., PANSS = Positive and Negative Syndrome Scale.
2005; Malla et al., 2006). Patients were excluded if they had a history of neurological SOFAS = Social and Occupational Functioning Assessment Scale.
a
disorder, if they presented an estimated IQ under 70 (based on the Wechsler Adult In- The demographic variables were contrasted between groups using bilateral two
telligence Scale Third Edition [WAIS-III] Vocabulary and Block Design dyad; (Ringe et sample t-tests.
b
al., 2002)) or if they did not have an adequate understanding of French (having com- According to Hollingshead's categories, adapted for Quebec: 1 = postgraduate, 2 =
pleted most schooling in French was considered as leading to adequate understanding bachelor degree, 3 = CEGEP degree, 4 = high school or equivalent, 5 to 7 = decreasing
whenever French was not the first language). All patients were taking a second- levels of partial high school. Note that this information was available for all control sub-
generation antipsychotic as their primary medication, with 17 taking quetiapine jects but for only 21 of the FEP patients.
c
(mean dose = 770.6 mg), four taking olanzapine (mean dose = 15 mg), seven taking The inpatients were stabilized and about to be discharged at the time of testing.
 A.M. Achim et al. / Psychiatry Research 196 (2012) 207–213 209

including attributions of beliefs/knowledge, of intentions/desires and of emotions. using the same item structure). In these stories a character acts on a false belief
Each item consisted of a short story depicting at least two characters evolving in a spe- about the state of the physical world. These questions do not involve interactions be-
cific situation. The ensuing task (called the “Combined stories test”) took approximate- tween the characters and require simpler answers. They were thus scored 1 point for
ly 30 min to complete. a correct answer or 0 point for an incorrect answer. The rationale for scoring these
A total of 20 second-order mentalizing stories were included in this Combined items independently from the other mentalizing items is that these First-order items
stories test; each of these items involved interactions between at least two characters. might require less integration of information and could simply rely on the knowledge
Second-order mentalizing was here defined as the attribution of a mental state about that people act according to what they know or do not know. In addition, it is worth
another character's mental state, in contrast to first-order mentalizing, which involves noting that since there were very few mistakes on these items in either group (see
a mental state about the state of the physical world. Our second-order mentalizing Results section and Table 2), including them in the mentalizing score would not have
items involved one character having a belief/knowledge, intention/desire or emotion had a noticeable impact on the statistical tests. For these reasons, and to keep the men-
in relation to another character's belief/knowledge, intention/desire, emotion or ac- talizing score representative of the ability to infer mental states in the context of social
tion/verbalization. For example, the items derived from the ‘False Belief task’ involved interactions (rather than also including a character's non-social interactions with the
a character having a (false) belief about another character's belief or action (e.g. X physical world), we chose to report these items independently from the second-
thinks Y did something), the items from the hinting task involved a character's inten- order mentalizing questions. Finally, the attention/memory questions simply inquired
tion about another character's action (e.g. X wants Y to do something), the items from about a detail of each story and were scored 1 point for a correct answer and 0 points
the faux-pas tasks involved a character's emotion about another character's action/ver- for a wrong answer. These questions were not meant to measure memory or attention
balization (e.g. X is upset about something that Y said), etc. These second-order men- capacities per see, but rather to verify that our procedure (stories remained in front of
talizing items were purposely selected so to include a variety of mental states the participant) was successful at minimizing the effects of potential attention or
combinations. Of these combinations, only the ‘belief about belief’ items (i.e. the memory deficits on mentalizing performance. These items were always given after
False Belief items) could be considered as purely cognitive mentalizing, whereas all the mentalizing, the non-social reasoning or the First-order inference questions, to
other combinations included emotions and/or intentions/desires in at least one of make sure that the participants had paid attention and were able to remember and/
the characters, and could thus be viewed as more affective mentalizing. All original or seek for details of the story.
stories were translated in French, and when necessary adapted to the reality of Except for the attention/memory questions, which were asked for every story, the
young adults (e.g., a faux-pas story originally taking place at the playground was pre- items/stories from the other categories (i.e., the different types of mentalizing stories,
sented as happening in the kitchen, while the rest of the situation, including the verba- the non-social reasoning stories and the first-order inference stories) were intermixed
lizations, was consistent with the original). The items that were included in our test throughout the test. A practice story was also presented before the test, and a menta-
were thus selected by prioritizing those that were the easiest to adapt to the socio- lizing question and a control question were then asked and participants were given
cultural reality of our participants (French-Canadian young adults). The final version feedback on their answers for that story only. This was done to familiarize them with
of the Combined stories test included six second-order mentalizing stories from each the material and to make sure that they understood the expected type of answer.
of the original test sources, except for the ‘False Belief task’ for which only two The answer to this question was not included in the test score.
second-order false belief stories were available. This restricted number of belief In sum, the Combined stories test produces four measures, including the main
items was thus compensated by also including the belief questions from ‘Faux-pas mentalizing score (second-order mentalizing) as well as three control measures,
test’ (Baron-Cohen et al., 1999). namely non-social reasoning, attention/memory and first-order inference scores.
For each item, participants were asked to read the story aloud and were then asked Because our mentalizing test represents a new measure, even if it was created
open questions that required making inferences about the characters' mental states based on existing material, we also administered another well-known measure of
(mentalizing questions). The text remained in front of the participant during question- mentalizing, namely Sarfati's cartoon task (Sarfati et al., 2003) to assess convergent va-
ing and participants were encouraged to go back to the text whenever they felt the lidity for our test. Sarfati's cartoon task mainly assesses a character's intentions (and
need (to minimize the memory load of the task). For six of the 20 mentalizing stories rarely involves a character's belief or emotion), and seems limited by ceiling effects
two mentalizing questions were asked, whereas only one mentalizing question was in the control group, but it was considered to be a good option as a validating criterion
asked for the other 14 stories, for a total of 26 answers. The answers were taken verba- because it has repeatedly been used in SZ and it does not involve verbal material which
tim and then scored 2, 1 or 0 point according to a pre-determined correction grid. A would have required a translation that could have affected the validity of the measure.
correct, complete answer was attributed 2 points, an incomplete answer was allowed In addition to assessing our test's convergent validity, internal consistency and inter-
1 point and an incorrect answer was allowed 0 point. rater reliability were also considered (see results below).
Several control conditions were also integrated in the test: 1) items to test for gen-
eral reasoning abilities (non-social reasoning questions); 2) items to test for the gener- 2.2.2. Social cue recognition and social knowledge
al capacities to link a mental state to a behavior (first-order inference questions); and Social cue recognition and social knowledge tests were also administered to deter-
3) questions to control for attention and memory effects (attention/memory ques- mine whether or not our FEP patients presented with impairments in these aspects of
tions). The items testing for general reasoning capacities focused on six non-social rea- social cognition and to determine if such impairments can partly explain the mentaliz-
soning stories (4 adapted from Happe et al. (Happe, 1994) and 2 created by us) and ing deficits previously reported in these patients (Inoue et al., 2006; Bertrand et al.,
required drawing inferences about physical causalities. Similarly to the mentalizing 2007).
questions, these items were scored 2, 1 or 0 point according to a pre-determined cor- Social cue recognition was assessed by presenting standardized emotional facial
rection grid. The first-order inference items testing for the general capacity to link a stimuli from the Ekman and Friesen series (Ekman and Friesen, 1976) and asking par-
mental state to a behavior were three simple first-order false belief stories (two from ticipants to select the corresponding emotion from a given list of labels (i.e., happy,
the literature (Baron-Cohen, 1989; Frith and Corcoran, 1996) and one created by us surprise, sad, angry, disgust, fear or neutral). There were a total of 14 items, scored 1

Table 2
Social cognition performance.

FEP Controls Group

difference
Mean (S.D.) Range Mean (S.D.) Range

Combined stories: mentalizing (/52) 38.9 (6.8) 25–48 43.6 (4.3) 34–50 0.001
Combined stories: non-social reasoning (/12) 10.5 (1.5) 7–12 11.2 (1.1) 7–12 0.050a
Combined stories: first-order (/3) 2.9 (0.3) 2–3 2.8 (0.4) 2–3 NSa
Combined stories: att./mem. (/29) 28.7 (0.5) 27–29 28.8 (0.4) 28–29 NSa
Social knowledge test (/14) 10.4 (1.4) 7–13 11.0 (1.8) 8–14 0.14
Social cue recognition test (/14) 10.6 (1.8) 7–13 11.0 (1.5) 8–14 NS
Sarfati's cartoon task (/28) 25.9 (1.9) 21–28 26.5 (1.3) 23–28 NSa

Combined stories: subscores from different original references

False beliefs (/4) 3.1 (1.2) 3.7 (0.7) 0.018a

Faux-pas false beliefs (/12) 7.9 (3.3) 9.7 (2.2) 0.005a
Faux-pas identification (/12) 6.7 (3.0) 8.8 (2.6) 0.043a
Hinting (/12) 9.6 (3.0) 11.45 (1.0) 0.151a
Strange stories (/12) 9.0 (3.0) 10.1 (1.6) 0.001a

NS = not significant.
a
From non-parametric Mann–Whitney U–tests (rank sum tests) instead of standard t-tests due to non-normal distributions and/or lack of variability.
210 A.M. Achim et al. / Psychiatry Research 196 (2012) 207–213

point for a correct answer or 0 points for an incorrect answer. Deficits have previously relationship could be detected in the control group (r(31) = 0.08,
been reported in FEP for similar social cue recognition tasks (Addington et al., 2006a).
p = 0.65). Given that the absence of a correlation in the control
Because we could not find measures of social knowledge about mental states in the
literature, the task used for this purpose was developed in our lab. The task assessed group was more likely linked to the restricted distribution of scores
social knowledge by presenting hypothetical situations (these original situations on Sarfati's task (i.e., ceiling effect), the overall pattern of results sug-
were inspired by the social themes listed in Blair and Cipolotti (2000)) and then asking gests adequate convergent validity for our task.
participants how people in general would feel in the situation. For example, one item A third step was to examine the inter-rater reliability of our Com-
was someone who learns he (or she: the way it is formulated in French does not sug-
gest a specific gender) has been lied to. After the experimenter read the situation, par-
bined stories test. After the initial scoring (performed by AMA), we
ticipants gave open responses (most often a single word) that were later scored 1 or asked a research assistant that had not been previously involved in
0 points according to a pre-determined correction grid. The correction grid was the study and was blind to initial scoring to independently re-score
based on a pre-test with ten healthy subjects and supplemented with several examples each item of the test for all 31 patients using the same standardized
of answers (mostly from the current control group) that were considered correct and
procedure, i.e. based on the verbatim of the responses and on the in-
incorrect. There were a total of 14 items.
Several responses could be considered a correct answer but these were all related. Al- formation provided in the correction grid. Convergent validity was
though the task was pre-tested in 10 healthy young adults, its psychometric properties not examined in the control group because the predetermined correc-
had not been established prior to this study.. It however represented an important step to- tion grid was supplemented with examples of verbal answers from
wards the measure of a construct that, though recognized as being an important aspect of that group. The protocols for 10 control subjects were nonetheless
social cognition processing (Bless et al., 2004; Brunet-Gouet et al., 2011; Green et al.,
2008), is often neglected in studies of social cognition in SZ or other disorders.
intermixed with those of the patients to keep the assistant blind to di-
agnosis while she scored the protocols. This step of examining the
2.3. Statistical analyses
inter-rated reliability for our 31 patients, which was pivotal to deter-
mine whether our correction grid was detailed enough to allow con-
The psychometric properties of the Combined stories test were first assessed. After sistent correction of the items based on the participants open
examining the normality of the distributions to identify eventual ceiling effects, the fol- verbalizations, yielded excellent inter-rater reliability (r(31) = 0.98,
lowing properties were determined: 1) convergent validity through a correlation of
p b 0.001).
our mentalizing score with that of Sarfati's cartoon task; 2) inter-rater reliability
through a correlation between initial scoring and scoring by an independent rater; 3) Lastly, internal consistency of the test was assessed to determine
internal consistency through a Cronbach's alpha test performed on the scores from whether items borrowed from the different tasks yielded significantly
the different initial test sources. heterogeneous performances. Across all participants, a good internal
Some psychometric properties of the Social knowledge test were also assessed, in- consistency between the sources was observed with a Cronbach's
cluding 1) convergent validity with our mentalizing task; 2) inter-rater reliability
through a correlation between initial scoring and scoring by an independent rater; 3)
alpha of 0.81.
internal consistency through a KR-20 test (for dichotomous variables) performed on
the scores of all 14 tests items. 3.1.2. Social knowledge test
Then, t-tests were used to contrast the two groups, FEP and control, for the nor-
The scores were also normally distributed in both groups for the
mally distributed social cognition measures, whereas Mann–Whitney rank sum tests
were used for the measures not meeting that condition. The between-group difference social knowledge test. Convergent validity was examined via the
on the Combined stories test's mentalizing scores was thereafter reassessed, this time expected relationship between social knowledge and mentalizing,
including non-social reasoning, social knowledge and social cue recognition as covari- through a partial correlation controlling for group. This association
ates in the analysis. Effects reaching an alpha level of 0.05, two-tailed, were considered
was found to be significant (r = 0.44, p b 0.001) and the regression
significant.
Finally, we also explored the correlation pattern between social cognition perfor-
slopes did not significantly differ between the two groups.
mance on our three social cognition tasks in the FEP group and social functioning as Inter-rater reliability was excellent (r = 0.97, p b 0.001), with only
assessed with the SOFAS, as well as with duration of illness. Because these were explor- one patient for which one item was re-scored differently, attesting
atory analyses, correlations meeting an alpha level of 0.05 were considered significant, that our scoring grid allowed consistent scoring of the participant's
but these results from multiple exploratory comparisons should be interpreted with
open responses.
caution.
Internal consistency as assessed through a KR-20 test showed a
relatively modest stability between the different test items (KR-
3. Results 20 = 0.24). This result is not too surprising given the all-or-none scor-
ing scheme of our test, but could also reflect that our test measures
3.1. Validation of the Combined stories test and the Social knowledge test several distinct components of social knowledge (ex: different mental
states). Using a lager participant sample could eventually help to de-
3.1.1. Combined stories test termine whether the current test items cluster into a certain number
A first step for the validation consisted of examining the nor- of components that each show higher internal consistency.
mality of the score distributions in the FEP group and the control
group for our Combined stories test and for Sarfati's cartoon task.
3.2. Social cognition in first-episode psychosis patients versus controls
There was no evidence of significant skewness for our Combined
stories test but significantly skewed distributions were observed in
Performance and between group statistics for each tasks are
both groups for Sarfati's cartoon task even after removing one out-
reported in Table 2. When contrasting the performance of the two
lier in the FEP group (skewness z = −2.2 in FEP and z = −2.6 in
groups separately for each test, a highly significant between group ef-
controls). The pattern of results for Sarfati's cartoon task suggested
fect was observed for performance on the mentalizing questions
a ceiling effect for that task, which was particularly prominent in
(t(60) = 3.28, p = 0.002, effect size r (ESr) 1 = 0.39), and a significant
the control group with two-thirds of the group presenting with
between group effect was also observed for the non-social reasoning
zero or a single error.
questions (U = 348.0, Z = 1.96, p = 0.050, ESr = 0.25). No significant
In a second step, we examined the convergent validity of our Com-
between group differences were found for the social knowledge test
bined stories test by assessing its correlation with Sarfati's cartoon
(t(60) = 1.49, p = 0.143, ESr = 0.19), the social cue recognition task
task, which revealed a significant association when assessed across
(t(60) = 0.99, p = 0.329, ESr = 0.13), Sarfati's cartoon task
both groups while controlling for between group effects through a
(U = 389.5, Z = 1.13, p = 0.258, ESr = 0.14) or the other two control
partial correlation (r(58) = 0.424, p = 0.001). Since the regression
slopes were significantly different between groups (F(1,57) = 8.44, 1
Note that effect sizes r vary between − 1 and 1. An r = 0.50 is considered as a large
p = 0.005), correlations were thereafter assessed separately in each effect size (ES) large and corresponds to a Cohen's d = 1.15. An r = 0.30 is considered a
group. In the FEP group, a significant correlation was observed be- medium ES and corresponds to a Cohen's d = 0.63. An r = 0.10 is considered a small ES
tween the two tests (r(30) = 0.58, p = 0.001), although no significant and corresponds to a Cohen's d = 0.20.
 A.M. Achim et al. / Psychiatry Research 196 (2012) 207–213 211

Table 3 (r = −0.01, p = 0.940) or social knowledge (r = 0.13, p = 0.486).

Results from the univariate analysis of covariance (ANCOVA). Note however that none of the significant correlations from these ex-
Source of variance F(1,57) p ploratory correlation analyses would have survived a Bonferroni cor-
rection for multiple comparisons; these results should thus be
Covariates
Non-social reasoning 4.48 0.039 interpreted with caution.
Social knowledge 12.0 0.001
Social cue recognition 0.20 NS 4. Discussion
Independent variable
Group 5.07 0.028 This study suggests that mentalizing is an aspect of social cogni-
tion that is particularly impaired in FEP patients, with a highly signif-
NS = not significant.
icant between group difference (Effect size r = 0.39, equivalent to a
Cohen's d = 0.84). The magnitude of the between group difference
conditions from the Combined stories test, namely the first-order in- observed in this study was in fact in the same range in terms of its ef-
ference questions (U = 434.0, Z = −1.07, p = 0.283, ESr = − 0.14) fect size than what has previously been reported in SZ patients with a
and memory/attention questions (U = 446.0, Z = 0.64, p = 0.523, longer duration of illness who, as was the case with the patients in
ESr = 0.08). this study, were outpatients or inpatients about to be discharged at
In a second and more crucial step because non-social reasoning, the time of testing (Bora et al., 2009). Moreover, the between group
social knowledge and social cue recognition are all suggested to con- difference remained significant even after controlling for perfor-
tribute to higher order mentalizing capacities, we used a univariate mance on other measures including social cue recognition, social
analysis of covariance (ANCOVA) to test whether these three mea- knowledge as well as non-social reasoning. This study thus confirms
sures had a significant impact on mentalizing performance, and that FEP patients do present specific mentalizing impairments similar
whether the between group difference on our mentalizing test was to those observed in patients with a longer duration of illness, at least
still observed after entering these three measures as covariates for for second-order mentalizing from situations of social interactions.
the between group analysis. As shown in Table 3, the ANCOVA The fact that FEP outpatients were recruited for this study (except
revealed a significant impact of non-social reasoning (F(1,57) = for a few inpatients waiting to be discharged) further suggests that
4.48, p = 0.039) and of social knowledge (F(1,57) = 12.04, p = these second-order mentalizing deficits represent a trait of the illness
0.001) on mentalizing performance (i.e. these variables are significant and that such deficits can be observed early after illness onset.
predictors of mentalizing performance), suggesting that both these Unlike some of the previously published SZ studies (Sprong et al.,
variables contribute to mentalizing. However, a significant between 2007; Bora et al., 2009), however, we did not observe a deficit on the
group effect on mentalizing performance was still observed with first-order inference items in our FEP patients. Contrary to the
this analysis (F(1,57) = 5.07, p = 0.028, ESr = 0.28), meaning that second-order mentalizing items, these first-order items do not re-
the mentalizing deficit observed in our patients could not be fully quire that participants understand interactions between the story
explained by performance on the other three measures. characters. Our study thus suggests that the mentalizing deficit in
Redoing this whole set of analyses after excluding the five control FEP patients is specific to mentalizing about social interactions. This
subjects that were identified to fulfill DSM-IV criteria for a specific observation is consistent with the previous FEP study by Inoue and
phobia or for substance abuse/dependence did not change the pattern collaborators (Inoue et al., 2006) that reported a significant impair-
of results, with the exception that the direct group comparison no ment only for their second-order mentalizing question, and not for
longer reached significance for the non-social reasoning measure their first-order mentalizing question. Taken together, these results
(p = 0.080, instead of 0.050). could reflect that first-order mentalizing abilities deteriorate with
the progression of SZ or that patients with impaired first-order abili-
3.3. Correlations between our social cognition tests and the clinical char- ties are overrepresented in samples of people with chronic schizo-
acteristics of the FEP patients phrenia (which could be the case if these patients have more
regular contacts with psychiatric services than those individuals
We also explored the pattern of correlations between perfor- that do not show such impairments). Longitudinal studies would
mance on our tests in the FEP group and social functioning as however be required to clarify this question.
assessed with the SOFAS. As displayed in Table 4, significant positive We also failed to observe a significant social cue recognition deficit
correlations were observed both with the mentalizing task (r = 0.45, in our FEP group, contrary to a few previous FEP studies. It should
p = 0.011) and with the social knowledge task (r = 0.37, p = 0.038), however be noted that the only two studies that used social cue rec-
whereas the correlation with social cue recognition task did not ognition tests that were methodologically similar to the one used
reach significance (r = 0.31, p = 0.095). As for the duration of illness, here, i.e. that presented a series of emotional faces and asked partici-
a significant negative relationship was observed with the social cue pants to identify the basic expressed emotion, reported effect sizes of
recognition test (r = −0.36, p = 0.049), but not with mentalizing low or medium magnitude in FEP (Edwards et al., 2001; Addington et
al., 2006a). Of those two previous studies, a significant between group
difference was only observed in the study that included larger groups
Table 4 of participants and thus had greater statistical power (Addington et
Correlation of social cognition performance with functioning and duration of illness in
patients with first episode psychosis (FEP).
al., 2006a). The significant relationship observed between social cue
recognition and duration of illness in our study further suggests that
SOFAS Duration of illness these deficits are more subtle in the early stages of psychosis. Other
Mentalizing (N = 31) r = 0.45 r = − 0.01 studies using a wider range of facial expressions (i.e. not restricted
p = 0.011⁎ NS to basic emotions; (Kucharska-Pietura et al., 2005)), or assessing
Social knowledge (N = 31) r = 0.37 r = 0.13
emotion recognition through emotional prosody (Edwards et al.,
p = 0.038⁎ NS
Social cue recognition (N = 31) r = 0.31 r = − 0.36 2001; Kucharska-Pietura et al., 2005) in FEP have observed deficits
p = 0.095 p = 0.049⁎ of greater magnitude in terms of their effect sizes. Taken together,
NS = not significant (p > 0.1).
these results suggest that while no significant social cue recognition
⁎ Significant at p b 0.05. But note that none of these exploratory correlations would deficits were detected in the current study, social cue recognition
have survived a Bonferroni correction for multiple comparisons. can be affected in FEP patients. Indeed, a deficit might have been
212 A.M. Achim et al. / Psychiatry Research 196 (2012) 207–213

observed in our patient group had we used a more sensitive social cue It could also have been interesting to widen the focus of the social
recognition measure and/or larger groups of participants. cue recognition and the social knowledge tests, which were restricted
To our knowledge, social knowledge about mental states has not in terms of the range of mental states being assessed, (i.e. focused pri-
previously been assessed in FEP and though we did not observe a sig- marily on emotions). Knowledge of other mental states such as inten-
nificant deficit, performance on that test was observed to significantly tions/desires, knowledge/beliefs and/or other more complex
contribute to mentalizing performance. This suggests that the menta- emotions were thus overlooked. General social knowledge questions
lizing deficits observed in our patients could be more pronounced in targeting these other types of mental states could certainly be
patients with poorer social knowledge, a relationship that should be added to the social knowledge task in the future. Other aspects of so-
taken into account when considering social cognition remediation cial knowledge that are not specifically related to mental states (e.g.
strategies, especially since some aspects of social knowledge have social schemas that relate more to standard action sequences)
been found to be related to improvement in functional outcome fol- might also have an influence on mentalizing and could also be inter-
lowing cognitive enhancement therapy in people with SZ (Eack et esting to include in future studies. For social cue recognition, though a
al., 2011). Taken together with these previous results, the pattern of person's knowledge or intentions is less directly assessed through a
results observed in this study argues for the importance of consider- contextual observation of social cues, the ability to identify eye gaze
ing social knowledge about mental states when assessing social cog- direction, certain body movements or certain aspects of prosody can
nition in clinical settings. contribute to real life capacities to infer mental states in others. Adding
Non-social reasoning performance, which was significantly im- some probes to assess these abilities in future studies could thus pro-
paired in our FEP patients, was also observed to play a role in explain- vide more complete information on the range of difficulties leading to
ing mentalizing performance, suggesting that the deficits observed on mentalizing impairments in FEP patients.
such social cognition measures are not fully independent from other Another limitation of this study is that we did not include a com-
non-social cognitive processes. Non-social reasoning deficits have plete set of neuropsychological tests, which could have contributed to
previously been reported in SZ (e.g. Young and Bentall, 1997) and highlight potential relationships with our social cognition measures.
the idea that non-social neurocognitive processes can contribute to As for patient sampling, recruiting patients from a first-episode
social cognition performance is not new. For example, it has been rec- psychosis clinic offers several advantages, including having a patient
ognized by Addington et al. (2006b), who showed that social cogni- sample that is more representative of the diversity of future out-
tion acts as a mediator between basic cognition and functioning in comes. Our decision to include patients with durations of illness of
FEP patients. Consistent with that previous study, we also observed up to 5 years (range 1 to 57 months) could however have limited
significant positive relationships between our social cognition mea- the interpretation of our results as being specific to first-episode sam-
sures and social functioning in our FEP group, attesting to the func- ples, though the main mentalizing measure showed no evidence of a
tional significance of identifying such deficits early in the SZ illness relationship with illness duration in our patients.
process.
4.2. Conclusion
4.1. Strengths and limitations of our study
Even after controlling for other social and non-social cognitive abili-
One of the strengths of this study is that we relied on a well- ties, mentalizing as assessed with our comprehensive task was signifi-
controlled, comprehensive mentalizing task for which we observed cantly affected in our FEP group. Although our social cognition
good psychometric properties. Mentalizing tasks used in the litera- measures showed positive correlations with FEP patients' social func-
ture often suffer from ceiling effects at least in the control group. tioning, mentalizing showed the strongest and most significant correla-
The inclusion of more items, along with the open questions strategy tion, attesting to its importance and the need to assess the presence of
and the scoring scheme on a scale from zero to two points per item, these deficits and offer targeted interventions (i.e. interventions focused
allowed additional variability in scores and made it possible to per- on the specific social cognition difficulties of each patient) early in the
form parametric analyses on our mentalizing data, which can be treatment to favor a positive outcome. This seems particularly relevant
more informative than relying on criterion-based measures. Because among FEP patients, as they are usually young adults who are at impor-
no standard measure of mentalizing is yet available, reporting such tant life junctures, such as choosing and establishing a career, founding
psychometric properties could also be an important step toward a family, etc., when the psychosis occurs. Preventing negative impacts of
identifying adequate measures for future studies and eventually for the illness and promoting a more positive functional outcome for these
clinical evaluations. patients through remediation of mentalizing difficulties could have a
One limitation of the study is that our mentalizing task relied major positive influence on these people's lives.
purely on verbal descriptions and written presentation of the charac-
ter's verbalizations and thus did not call for much social recognition
processing. This may have restricted our ability to observe a link Acknowledgments
with our independent social cue recognition measure. It would thus
be interesting to assess the relationship between mentalizing and so- This work was supported by a NARSAD Young Investigator Award
cial cue recognition using a mentalizing task that is more ecological, to PLJ, a Canadian Institutes of Health Research (CIHR) operating grant
i.e. that relies on a more complete set of social cues, including percep- (#MOP-77647) to MAR, and a CIHR postdoctoral fellowship to AMA.
tual cues. PLJ was supported by a salary grants from the Fonds de la Recherche
Another limitation that could eventually be circumscribed by en Santé du Québec (FRSQ) and the CIHR. RO was supported by stu-
using a more ecological task is the fact that social cue recognition dentships from CIHR and FRSQ. We would like to thank the clinical
and social knowledge were assessed with material that was indepen- and research team from Clinique Notre-Dame des Victoires for their
dent from the material presented in the context of the mentalizing interest and support with this study, and Marie-Audrey Lavoie who
task (i.e. the combined stories). The relationship between the two un- re-scored the mentalizing task for the inter-rater validation.
derlying social cognition processes of interest, social cue recognition
and social knowledge, could thus only be tested indirectly. A more di- References
rect assessment of this relationship would have required using the
Addington, J., Saeedi, H., Addington, D., 2006a. Facial affect recognition: a mediator be-
same material as in the mentalizing test for the evaluation of social tween cognitive and social functioning in psychosis? Schizophrenia Research 85,
cue recognition and social knowledge. 142–150.
 A.M. Achim et al. / Psychiatry Research 196 (2012) 207–213 213

Addington, J., Saeedi, H., Addington, D., 2006b. Influence of social perception and social Happe, F.G., 1994. An advanced test of theory of mind: understanding of story charac-
knowledge on cognitive and social functioning in early psychosis. British Journal of ters' thoughts and feelings by able autistic, mentally handicapped, and normal chil-
Psychiatry 189, 373–378. dren and adults. Journal of Autism and Developmental Disorders 24, 129–154.
American Psychiatric Association, 1994. Diagnostic and Statistical Manual of Mental Heinrichs, R.W., Zakzanis, K.K., 1998. Neurocognitive deficit in schizophrenia: a quan-
Disorders, 4th ed. American Psychiatric Association Press, Washignton D.C. titative review of the evidence. Neuropsychology 12, 426–445.
Baron-Cohen, S., 1989. The autistic child's theory of mind: a case of specific develop- Herold, R., Tenyi, T., Lenard, K., Trixler, M., 2002. Theory of mind deficit in people with
mental delay. Journal of Child Psychology and Psychiatry 30, 285–297. schizophrenia during remission. Psychological Medicine 32, 1125–1129.
Baron-Cohen, S., O'Riordan, M., Stone, V., Jones, R., Plaisted, K., 1999. Recognition of Inoue, Y., Yamada, K., Hirano, M., Shinohara, M., Tamaoki, T., Iguchi, H., Tonooka, Y.,
faux pas by normally developing children and children with Asperger syndrome Kanba, S., 2006. Impairment of theory of mind in patients in remission following
or high-functioning autism. J Autism Dev Disord 29, 407–418. first episode of schizophrenia. European Archives of Psychiatry and Clinical Neuro-
Bertrand, M.C., Sutton, H., Achim, A.M., Malla, A.K., Lepage, M., 2007. Social cognitive sciences 256, 326–328.
impairments in first episode psychosis. Schizophrenia Research 95, 124–133. Kay, S.R., Fiszbein, A., Opler, L.A., 1987. The positive and negative syndrome scale
Blair, R.J., Cipolotti, L., 2000. Impaired social response reversal. A case of ‘acquired (PANSS) for schizophrenia. Schizophrenia Bulletin 13, 261–276.
sociopathy’. Brain 123 (6), 1122–1141. Kee, K.S., Horan, W.R., Salovey, P., Kern, R.S., Sergi, M.J., Fiske, A.P., Lee, J.H., Subotnik,
Bless, H., Fiedler, K., Strack, F., 2004. Social Cognition: How Individuals Construct Social K.L., Nuechterlein, K., Sugar, C.A., Green, M.F., 2009. Emotional intelligence in
Reality. Taylor & Francis, New York. schizophrenia. Schizophrenia Research 107, 61–68.
Bora, E., Yucel, M., Pantelis, C., 2009. Theory of mind impairment in schizophrenia: Kendler, K.S., Neale, M.C., Walsh, D., 1995. Evaluating the spectrum concept of schizo-
meta-analysis. Schizophrenia Research 109, 1–9. phrenia in the Roscommon Family Study. American Journal of Psychiatry 152,
Brunet-Gouet, E., Achim, A.M., Vistoli, D., Jackson, P.L., Hardy-Baylé, M.-C., Passerieux, C., 749–754.
2011. The study of social cognition with neuroimaging as a means to explore future Koelkebeck, K., Pedersen, A., Suslow, T., Kueppers, K.A., Arolt, V., Ohrmann, P., 2010.
directions of deficit evaluation in schizophrenia? Psychiatry Research 190, 23–31. Theory of mind in first-episode schizophrenia patients: correlations with cognition
Corcoran, R., Mercer, G., Frith, C.D., 1995. Schizophrenia, symptomatology and social and personality traits. Schizophrenia Research 119, 115–123.
inference: investigating “theory of mind” in people with schizophrenia. Schizo- Kohler, C.G., Walker, J.B., Martin, E.A., Healey, K.M., Moberg, P.J., 2010. Facial emotion
phrenia Research 17, 5–13. perception in schizophrenia: a meta-analytic review. Schizophrenia Bulletin 36,
Couture, S.M., Penn, D.L., Roberts, D.L., 2006. The functional significance of social cogni- 1009–1019.
tion in schizophrenia: a review. Schizophrenia Research 32, S44–S63. Kucharska-Pietura, K., David, A.S., Masiak, M., Phillips, M.L., 2005. Perception of facial
Cutting, J., Murphy, D., 1990. Impaired ability of schizophrenics, relative to manics or and vocal affect by people with schizophrenia in early and late stages of illness.
depressives, to appreciate social knowledge about their culture. British Journal of British Journal of Psychiatry 187, 523–528.
Psychiatry 157, 355–358. Malla, A., Norman, R., Schmitz, N., Manchanda, R., Béchard-Evans, L., Takhar, J.,
Eack, S.M., Pogue-Geile, M.F., Greenwald, D.P., Hogarty, S.S., Keshavan, M.S., 2011. Haricharan, R., 2006. Predictors of rate and time to remission in first-episode psy-
Mechanisms of functional improvement in a 2-year trial of cognitive enhancement chosis: a two-year outcome study. Psychological Medicine 36, 649–658.
therapy for early schizophrenia. Psychological Medicine 41, 1253–1261. Miller, D.C., 1991. Hollingshead's two-factor index of social position. In: Handbook of
Edwards, J., Pattison, P.E., Jackson, H.J., Wales, R.J., 2001. Facial affect and affective pros- research design and social measurement, 5th ed. Sage Publications, Newbury
ody recognition in first-episode schizophrenia. Schizophrenia Research 48, Park, pp. 351–359.
235–253. Ringe, W.K., Saine, K.C., Lacritz, L.H., Hynan, L.S., Cullum, C.M., 2002. Dyadic short forms
Ekman, P., Friesen, W.V., 1976. Pictures of Facial Affect. Consulting Psychologists Press, of the Wechsler Adult Intelligence Scale III. Assessment 9, 254–260.
Palo Alto. Sarfati, Y., Brunet, E., Hardy-Bayle, M.C., 2003. Comic-Strip Task: Attribution of inten-
First, M.B., Spitzer, R.L., Gibbon, M., Williams, J.B.W., 1998. Structured clinical interview sions to Others. Service de Psychiatry Adulte, Hôpital de Versailles, Le Chesnay,
for DSM-IV patient edition (SCID-I/P V and SCID-I/NP Version 2.0). Biometric Re- France.
search Department, New York State Psychiatric Institute, New York, NY. Schimmelmann, B.G., Conus, P., Edwards, J., McGorry, P.D., Lambert, M., 2005. Diagnos-
Frith, C.D., Corcoran, R., 1996. Exploring ‘theory of mind’ in people with schizophrenia. tic stability 18 months after treatment initiation for first-episode psychosis. Journal
Psychological Medicine 26, 521–530. of Clinical Psychiatry 66, 1239–1246.
Green, M.F., 1996. What are the functional consequences of neurocognitive deficits in Sprong, M., Schothorst, P., Vos, E., Hox, J., van Engeland, H., 2007. Theory of mind in
schizophrenia? The American Journal of Psychiatry 153, 321–330. schizophrenia: meta-analysis. British Journal of Psychiatry 191, 5–13.
Green, M.F., Kern, R.S., Braff, D.L., Mintz, J., 2000. Neurocognitive deficits and functional Young, H.F., Bentall, R.P., 1997. Probabilistic reasoning in deluded, depressed and nor-
outcome in schizophrenia: are we measuring the “right stuff”? Schizophrenia Bul- mal subjects: effects of task difficulty and meaningful versus non-meaningful ma-
letin 26, 119–136. terial. Psychological Medicine 27, 455–465.
Green, M.F., Penn, D.L., Bentall, R., Carpenter, W.T., Gaebel, W., Gur, R.C., Kring, A.M.,
Park, S., Silverstein, S.M., Heinssen, R., 2008. Social cognition in schizophrenia: an
NIMH workshop on definitions, assessment, and research opportunities. Schizo-
phrenia Bulletin 34, 1211–1220.