5.2. Inferring social intentions

In social tasks, patients overall tended to perform less well than healthy controls, but this tendency did not reach significance. As expected, schizophrenic patients were also found to significantly favour the ‘tit-for-tat’ (TFT) mode of reciprocity, by preferentially attributing intentions congruent with that mode to the second agent (i.e. defect if previous defection, cooperate if previous cooperation) to the expense of others (i.e. always defecting, always cooperating). However, two main results made them diverge from healthy participants. First, while healthy participants showed an early preference toward TFT intentions in the baseline session (that is, prior to being biased), patients did not. As a result, patients were found to provide lesser responses toward TFT intentions than controls did in the bias session. Second, this lesser interest for TFT intentions was concomitantly associated with the number of correct responses improving to a greater extent than those of healthy participants as the amount of visual information increased. Finally, the more the negative symptoms were severe, the less patients were sensitive to the bias, i.e. the less they responded toward the preferred mode of reciprocity. Accordingly, the more the negative symptoms were severe, the more patients relied on the visual information to make their decision.

Crucially, this pattern of performance was the exact opposite of the one we observed in the non-social tasks. One explanation for this result might reside in the fact that social conditions – which required paying attention to both the first and second player’s intention –, were more demanding than non-social ones. However, if that was the case, poorer performances should have been observed in these conditions. Yet, schizophrenic patients performed at comparable levels across non-social and social experiments, indicating that the lesser effect of the bias on patients’ social decision is accounted for by the type of intention considered rather than by the difficulty of the task.

Two possible explanations may account for why the bias did not influence patients’ social inference to the same extent as it did for that of controls’. In healthy participants, situations identified as involving social interactions are generally prone to trigger domain-specific expectations concerning the way agents are likely to behave in such situations (Frith & Frith, 2006; Castelli, Happé, Frith & Frith, 2000). In the present study, healthy participants exhibited such expectations as early as the baseline session – with faster RTs in predicting intentions congruent with the TFT mode of reciprocity –, while patients did not. Moreover, we found that the more the negative symptoms were severe, the fewer patients responded toward this mode of reciprocity. It is therefore possible that social situations do not induce the same domain-specific expectations in negative patients which are normally observed in healthy participants. Impoverished social knowledge has been consistently argued to be an intrinsic feature of schizophrenia (Cutting & Murphy, 1990). Previous evidence also suggests that observable behavioural information derived from social interaction is abnormally used by negative patients in problem solving tasks (Bedell et al., 1998). The present results suggest, accordingly, that patients with severe negative symptoms do not properly use social domain-knowledge to predict what interacting agents are most likely to do. This is further echoed by the fact that patients compensated for this by overly relying on visual evidence.

Such over-reliance on visual, outer information, together with impoverished social expectations is consistent with the observation that patients with negative symptoms – such as anhedonia or alexithymia – tend to excessively focus on directly observable, external information, rather than inner experience (Taylor, 1994). However, fewer social expectations in schizophrenia do not necessarily imply that patients have none. Their responses in the bias session reveal that they are able to acquire and use knowledge about social situations. This suggests an alternative explanation for their pattern of performance: in a series of ToM tasks with stories depicting interacting agents, Sarfati and colleagues have shown that patients preferentially based their interpretation of intention on socially familiar experiences rather than on the context of the story (Sarfati et al., 1997). Drury and coworkers also found that schizophrenic patients are prone to be distracted by the relations between the story characters and their own personal experiences (Drury et al., 1998). One cannot therefore preclude that patients do have strong expectations of how agents are supposed to interact, but that these expectations are abnormally related to their own (inappropriate?) social experience. Selecting social intentions congruent with the context of the task would therefore require them to disengage from these prior expectations. This would be a costly process which in our task could be accounted for by patients’ increased reaction times toward TFT intentions.