1. Introduction

Schizophrenia is associated with poor social functioning, as shown by impairments that most patients exhibit in interpersonal communication and relationships(Corcoran, 2001, for review). To account for these deficits, Theory of Mind (ToM) function – the ability to explain and predict others’ behaviour on the basis of their mental states – has often been focused as a probable candidate (Frith, 2004; Harrington et al., 2005; Sprong et al., 2007). Though ToM dysfunction in schizophrenia has given rise to an extensive line of research these two last decades, both its nature and the extent to which it is present are still widely debated (McCabe et al., 2004; Zalla et al., 2006; Walter et al., 2009). The reason for this controversy may stem from the fact that, taken together, the existing data proves to be particularly contradictory. Most studies indeed failed to formalize coherent associations between symptom dimensions and patients’ defective abilities in intention attribution and mentalizing. Thus, while a number of authors report selective correlations between a mentalizing deficit and the positive dimension of the disease – such as formal thought and language disorder (Sarfati et al., 1999; Brunet et al., 2003), or reality distortion (Frith & Corcoran, 1996) –, some other studies found concurrent associations with either disorganization (Zalla et al., 2006) or psychomotor poverty syndroms (Mazza et al., 2001), or with negative behavioural signs (Pickup & Firth, 2001) such as flat affects, social withdrawal and alogia (Langdon et al., 2002).

It has been suggested that global inconsistency between patients’ symptoms and ToM dysfunction may be rooted in the great variety of the tasks proposed, both in terms of complexity and type of (verbal, iconographic) material being used (see Walter et al., 2009). Crucially, the heterogeneity of the data could also result from the lack of control of the notion being tested, namely, the other’s intention the patient has to infer. “Intention” is indeed an umbrella term that may embrace various sub-types each being assigned different scopes (i.e. proximal or distal goals) and targets (i.e. action directed at an object or at another agent) (Pacherie, 2000, 2008). The ability to appreciate other people’s intentions thus refers to separate processes which could be differentially impaired depending on the scope and/or the target of the intention represented. As such, one cannot preclude the possibility that the schizophrenic patients’ deficit in mentalizing may embrace one particular type of intention while sparing appreciation of the others. It then appears to be of an extreme importance to precisely assess at which (possible) level of intentional complexity this deficit might take place, in strictly controlling the type of intention being evaluated on the basis of a careful intentional typology.

To do so, we adapted from Chambon et al. (xxxx) a series of tasks that require taking into account four distinct types of intentions in a large sample of healthy controls and symptom-matched patients. Both groups were instructed to identify, from observation of videos of action sequences, intentions that varied depending on either their scope (motor vs. superordinate) or target (social vs. non-social). Thus, combining each of these two dimensions resulted in 4 distinct experiments, in which participants were successively required to infer

i) motor (proximal) intentions, i.e. intentions directed at simple motor goals, such as lifting an arm, pressing a button, reaching for an object, etc. (task1)

ii) superordinate (distal) intentions, i.e. intentions directed at somewhat complex or general goals, the achievement of which typically involves the completion of a number of subgoals or substeps (e.g. drinking a glass of wine: taking a glass, opening a bottle, pouring the wine into the glass, etc.) (task 2)

iii) social motor intentions, i.e. simple goals achieved in a context of social interaction with another agent (task 3)

iv) social superordinate intentions, i.e. general goals comprising a sequence of basic motor acts achieved in a context of social interaction (task 4)

So far, only few studies directly tested the patients’ ability to appreciate distinct types of intention within the same experimental settings, or using the same material across conditions. Among them, evidence revealed that schizophrenic disorganised patients show a specific impairment in evaluating distal non-social intentions underlying current action sequences, while their ability for inferring proximal intentions would be intact (Zalla et al., 2006). Incidentally, this impaired ability would arise together with executive disorders in action sequencing, that is, with difficulties to coordinate the different segments that compose the action in relation with the mental states of the protagonists (Zalla et al., 2004; Zalla et al., 2006). However, a recent study manipulating non-social and social intentions suggested that patients could be impaired, not in appreciating actions directed at inanimate objects, but specifically in inferring intentions achieved in a context of (present or forthcoming) social interaction (Walter et al., 2009). A pronounced inability to appreciate social intentions was further associated with an abnormal modulation of activity in regions typically activated in ToM tasks, i.e. paracingulate cortex and bilateral temporo-parietal junctions (Walter et al., 2009). Interestingly, this reduced activity for social intentions was partly the consequence of no signal drop in physical (non-intentional) conditions, revealing that patients’ mentalizing profile was also characterized by a significant tendency to (over-)attribute intentionality to inanimate objects (see also the ‘hyper-TOM’ assumption, Abu-Akel & Bailey, 2000).

In the present study, we hypothesized that this confusing array of findings may be disentangled by investigating patients’ mentalizing abilities at a more fine-grained level of meta-cognitive functioning; that is, not only by assessing patients’ raw performances in intention recognition, but in further exploring how individuals with schizophrenia deal with the information which usually contributes to that recognition. We indeed showed in a previous study that inferring other’s intentions requires taking into account two distinct (external and internal) sources of information: i) first, the sensory information available from the action scene and derived from the agent’s movement kinematics, and ii) second, the observer’s prior expectations about which intention is, given past experience, the most likely cause of what is observed. Intentional inference can be modelled as an adaptive interaction between these two sources of information (see Baker et al., 2006), with participants tending to rely progressively more on their priors as the reliability of sensory information decreases, and conversely (Chambon et al., xxxx ; see also Onishi et al., 2007). Convergingevidence indicates that difficulties in action understanding observed in schizophrenia might further reflect an abnormal over-reliance on either of these two types of information, depending on the patients’ prevalent symptoms. As an example, the mentalizing profile of productive patients may be characterized by a tendency to give excessive credit to endogenous, self-generated information – e.g. prior expectations of how people are supposed to behave under some circumstances – whereas negative patients would display a stimuli-induced mentalizing style that may be accounted for by an exaggerated tendency to focus on directly observable, external information, rather than inner experiences (Frith, 1992; Taylor, 1994).

In addition to assessing patients’ abilities of drawing inferences at each level of intentional complexity, we sought to test this hypothesis by varying the following within each of the 4 above-mentioned tasks

(1) the amount of visual information available from the action scenes. We proceeded by manipulating the degree of completeness of the action sequences, resulting in video clips of varying durations

(2) prior expectations that participants had about the possible intentions of the observed agent. To do so, we progressively increased the probability that one intention (termed biased or preferred intention) occurs over the sequence, at the expense of other intentions (non-biased or non-preferred intentions) with the same scope and target.

We then looked at whether, for each of the four intentions considered, patients’ inferences may be accounted for by an abnormal dependence on their priors and/or sensory evidence conveyed by the action scene, and whether over-relying on either of these sources of information did correlate with the severity of positive (SAPS; Andreasen, 1984), negative (SANS; Andreasen, 1983), or disorganisation symptoms of schizophrenia.