[1] 12w Are patients with schizophrenia rational maximizers? Evidence from an ultimatum game study

[1] 201w Schizophrenia is associated with impaired social cognition and community functioning. Social decision-making strategies of healthy controls and patients with schizophrenia were compared by using the ultimatum game (UG). In this game two players have to split a sum of money. The proposer offers a portion to the responder, who decides to either accept or reject the offer. Rejection results in no income to either of the parties. Unfair proposals are frequently rejected by nonclinical individuals, a phenomenon described as altruistic punishment. Patients and controls participated in a series of UG interactions as responders in a computerized test setting. We also tested the effect of the proposer's facial expression on decision-making. Our results indicate that patients with schizophrenia accepted unfair offers at a significantly higher rate than did healthy controls. In contrast, at fair proposals, the acceptance rate was lower in patients compared with controls. At higher offers, the proposer's facial expression (positive/negative) significantly influenced the acceptance rate (positive facial expression increased the likelihood of acceptance) in the control group. This effect was not observed in the patient group. These results suggest that schizophrenia patients are impaired in socioeconomic interactions requiring emotion recognition and decision-making, which may result in unstable behavioral strategies.

[1] 250w Impaired social cognition is a characteristic feature of schizophrenia, which is associated with the level of community functioning (Brune, 2005;Couture et al., 2006). The three primary domains of inquiry are emotion perception, theory of mind, and attributional style, though others have been identified as key areas of investigation as well (e.g., social perception and social knowledge). Patients with severe negative symptoms are more impaired on facial emotion recognition tasks and display poorer social skills (Penn et al., 1997). However, the majority of traditional psychological tests do not take into consideration the importance of interpersonal interactions: for example mentalizing in psychiatric disorders is usually measured by the observation and interpretation of pictures, jokes or short stories, thus assessing mainly or solely explicit processes that are available for verbalization. As shown by current literature, behavior and interpersonal decision-making is determined not only by explicit, conscious processes but automatic implicit processes as well. As argued by Frith and Frith (2008), implicit, automatic processes in social cognition often facilitate prosocial and altruistic behavior and are often opposed to conscious strategies. Socioeconomic games, which are well-established in behavioral economy and widely used to examine decision-making strategies in non-clinical individuals using healthy population samples (Henrich et al., 2010), can provide a useful method in the investigation of social decision-making also in psychiatric diseases such as schizophrenia. The strength of these games compared to classic paradigms is that they allow observing subjects directly engaged in interpersonal decision-making, thus real-time interactions with implicit processes become available to investigation.

[2] 149w One example is the ultimatum game (UG), which represents a negotiation scene, where two players split a sum of money. One of the players acts as a proposer, who suggests how the amount should be divided between the proposer and the responder. The other player, the responder, can only decide whether to accept the offer or to reject it. If the responder accepts the offer, both players can keep the amount they agreed on. If the responder rejects the offer, neither player gains anything. Since the first description of the ultimatum game (Güth et al., 1982), it has become one of the most popular game theory paradigms that has been studied extensively from economical, mathematical, psychological, sociological, and neurological perspectives. Its popularity is based on its simplicity, the involvement of both "rational and emotional" decision-making, and its potential to explain such abstract phenomena as "inequity aversion", or "altruistic punishment".

[3] 147w According to Fehr and Schmidt (1999), people tend to exploit their bargaining power in competitive markets but not in bilateral bargaining situations; and people tend to exploit free-riding opportunities in voluntary cooperation games. Yet, when they are given the opportunity to punish free riders, stable cooperation is maintained, even though punishment is costly for those who punish ("altruistic punishment"). Inequity aversion (IA) is a term used in sociology and economy, and describes a certain human behavior: the preference for fairness and resistance to incidental inequalities. Fehr and Schmidt (1999) postulated that, as opposed to classic socioeconomic theories, people tend to make decisions to minimize inequity in outcomes. They argued that even disadvantageous IA appears in humans as the "willingness to sacrifice potential gain to block another individual from receiving a superior reward". According to them, the apparently self-destructive response creates the possibility of cooperation and bilateral bargaining.

[4] 240w The classical theory of economic decision-making assumes that while playing the UG, people should decide rationally, minimizing losses and maximizing gains. Thus, the proposer should offer the smallest possible amount to maximize the gain and the responder should accept every offer regardless of the amount because even a small amount is better than not gaining anything. However, according to studies conducted in healthy volunteers in different cultural environments, this assumption is systematically violated: people tend to propose "fair" offers (such as 60:40), and offers considered unequal and "unfair" are usually rejected. The limit from which amount an offer is considered unfair is surprisingly stable in the different studies (70:30 and lower) (Fehr and Fischbacher, 2003;Chuah et al., 2007;Hack and Lammers, 2008) and seems to have genetic components (Wallace et al., 2007). Frith and Frith (2008) suggest that altruistic punishment observed in the ultimatum game represents an example for implicit prosocial behavior vs. rational, selfish decisions, while in a recent EEG-study Hewig et al. (2011) found evidence that deviation from rational choice in the ultimatum game is guided by subjective emotional markers. The decisions of the responder in the UG are influenced by the features of the proposer, such as previous information about the proposer (Burns, 2006) and his/her physical attractiveness (Solnick and Schweitzer, 1999). Social decisions are also modulated by emotions expressed by the partners; positive emotions expressed by a smiling face enhance cooperation among players (Scharlemann et al., 2001).

[5] 144w The degree to which schizophrenia patients show impaired functioning on decision-making tasks still remains unclear. Studies using a gambling paradigm (Premkumar et al., 2008) and a pennymatching task resulted in ambiguous conclusions (Kim et al., 2007). Only one study used UG in patients with schizophrenia, examining the patients both in the roles of the proposer and responder. Using a twostage UG paradigm, these authors found that patients with schizophrenia did not fully exploit their strategic power as proposers: adjustment of their 2nd offer was different from that of the control group. While healthy controls, attempting to maximize their payoff, lowered their 2nd offer after a positive response to their 1st offer, schizophrenic patients did not do so. Authors interpreted this difference as a relative failure in strategic thinking. However, they performed similarly to controls in the role of the responder (Agay et al., 2008).

[6] 66w We think that the '2 round' UG paradigm gives only a limited possibility to explore the potential differences in decision-making strategies between the study groups. For this reason in the present study we applied an iterative ultimatum game paradigm and assumed that subjects with schizophrenia and healthy controls would act differently as responders especially at lower offers, when the "inequity aversion" or "altruistic punishment" takes effect.

[7] 114w The purpose of this study was twofold. First, we studied the acceptance rates in the UG as a function of offers and study group (schizophrenia vs. controls). We assumed that in schizophrenia altruistic punishment is decreased; our hypothesis was that the acceptance rates will be higher at lower, unfair offers (if 10-20% of the whole amount is offered) in the schizophrenia group than in the control group. Furthermore, as an extension of the first hypothesis, we assumed that the acceptance characteristics of patients with milder positive and negative symptoms have a similar acceptance pattern to that of the controls, while subjects with severe positive or negative symptoms tend to display a weaker altruistic punishment.

[8] 98w Second, we investigated how emotional facial expressions of proposers influence the decision of responders with schizophrenia during an UG. According to the results of Scharlemann et al. (2001) social decisions are modulated by the emotional facial expressions of the partner, by positive emotions having a cooperation enhancing effect. As a deficit in facial emotion processing is present in schizophrenia, we assumed that in schizophrenia perceived emotions has a decreased influence on social judgments. The second hypothesis of the study was that positive emotional expression facilitates cooperation in healthy controls but fails to do so in patients with schizophrenia.

[1] 388w Acceptance (yes or no) as response variable was investigated by repeated measurement logistic regression (with GENMOD), exploring the main effects of study group, offer, emotional facial expression, and the interaction of these factors. There were significant main effects of offer (acceptance is more frequent for higher offers than lower offers; χ² = 50.8, n =87, d.f.=4, p b 0.0001) and emotion (acceptance is more frequent for positive emotions than for negative emotions; χ² = 7.5, n = 87, d.f.= 1, p b 0.006), but not of the study group (χ²=1.4, n =87, d.f.= 1, p = 0.23). There was a two way interaction between offer and study group (χ²=18.8, n = 87, d.f. = 4, p = 0.0009), indicating that the association between offer and probability of acceptance is different in schizophrenia patients and controls. The pattern of the results indicated that schizophrenia was associated with an increased likelihood of acceptance at lower offers and a decreased likelihood of acceptance at higher offers as compared with the controls (Fig. 1, Table 2). Furthermore, the three-way interaction among offer, emotion, and study group also reached the level of statistical significance (χ² = 10.9, n = 87, d.f. = 4, p = 0.027). To understand the interaction, post-hoc analyses were conducted comparing the probability of acceptance at each amount of offer as a function of proposers' emotional facial expression (positive vs. negative) in each study group. In the control group, we found significant differences between positive and negative emotions at higher offers indicating that the probability of acceptance was higher if the proposers' emotional facial expression was positive (OR 10 [95%CL]= 1[0.6-1.6], χ²=0, p =1; OR 20 [95%CL]= 1.2[0.7-2.1], χ² = 0.6, p = 0.44; OR 30 [95%CL]= 1.3[0.9-1.9], χ²=1.5, p =0.22; OR 40 [95%CL]= 3.6[1.2-10.6], 2 =5.4, p = 0.02; OR 50 [95%CL] = 3.9[1.1-14.8], χ² = 4.2, p = 0.04). In the schizophrenia group, no significant differences were found between emotional states (OR 10 [95%CL] = 1.2[0.9-1.5], χ²=2, p =0.15; OR 20 [95%CL]= 1.2[0.9-1.7], χ² = 2.3, p =0.13; OR 30 [95%CL]= 1.3[0.9-1.7], χ²=2.2, p =0.13; OR 40 [95%CL]= 1.1[0.8-1.5], χ²=0.6, p =0.45; OR 50 [95%CL]= 1.1[0.7-1.6], χ² = 0.1, p =0.77) (Fig. 2). The study groups differed significantly in years of education and age (Table 1). Including these variables in the main analysis of the UG as

[2] 80w Schizophrenia Group covariates (years of education: χ² = 0.04, d.f. = 1, p = 00.84; age: χ² = 6.1, d.f. = 1, p = 0.01) did not affect the results mentioned before, the main effects (offer and emotion) and the interactions (offer by study group and offer by emotion by study group) remained significant. However we have to add that age also has a significant effect on acceptance rates, namely older subjects tend to accept offers at a higher rate.

[3] 56w Comparing the total pay off in the study groups by GLM analysis, we found that subjects in the schizophrenia group earned more money (M = 857.8, S.D. = 281.4) numerically than the control group (M = 805.9, S.D. = 182.7), but the difference did not reach significance (F = 0.81, d.f. = 1, p = 0.37).

[4] 174w In a different analysis we investigated the effects of PANSS scores on the acceptance rates in the UG. The main effects of offer, PANSS scores, and the interaction of these factors served as independent, while the acceptance rate (yes or no) served as dependent variables in the repeated logistic regression model (in GENMOD). In the case of total positive symptom score (χ 2 = 15.3, n = 58, d.f. = 4, p = 0.004) the main effect of money, and the money by 'positive PANSS score' interaction were significant (χ 2 = 11.7, n = 58, d.f. = 4, p = 0.02). The results were very similar in case of total negative symptom score, where the main effect of money (χ 2 = 17.2, n = 58, d.f. = 4, p = 0.002) and the money by 'negative PANSS score' interaction were significant (χ = 11.1, n = 58, d.f. = 4, p = 0.025) (Fig. 3.). PANSS total score and the PANSS general symptom score had no significant effect on the acceptance rates.

[5] 169w Furthermore we analyzed the association between decisionmaking and certain PANSS items, which might influence decisions like delusions (P1: χ 2 = 5.2, d.f. = 1, p = 0.02), conceptual disorganization (P3: χ 2 = 0.8, d.f. = 1, p = 0.38), suspiciousness/persecution (P6: χ 2 = 9.8, d.f. = 1, p = 0.002), poor rapport (N3: χ 2 = 0.1, d.f. = 1, p = 0.74) and disturbance in volition (G13: χ 2 = 0.01, d.f. = 1, p = 0.92). In order to characterize the associations further, we computed the effect size in terms of odds ratios (OR). In particular, the odds ratio in this computation reflected the likelihood of increase of the acceptance rate for a unit increase on the given PANSS item (odds ratios were computed only for those items, where association reached significance). Results indicated that a unit increase in P1 (delusions) and P6 (suspiciousness/persecution) was associated with an increased likelihood of acceptance (P1: OR = 1.30 [95% CL: 1.04-1.62]; P6: OR = 1.63[95%CL: 1.21-2.21]).

[6] 117w The study groups differed significantly on WCST, RAVLT and WAIS-R Digit Span Backward (Table 1). Including the results of these neuropsychological tests in the main analysis of the UG as covariates (WCST, number of perseverative errors: χ 2 = 0.4, d.f. = 1, p =0.53; WCST, categories completed: χ 2 =0.5, d.f.=1, p = 0.49; RAVLT total score: χ 2 = 0.01, d.f. = 1, p = 0.94; Digit Span Forward: χ 2 =2.2, d.f.=1, p = 0.13; Digit Span Backward: χ 2 =0.3, d.f.=1, p = 0.59) did not affect the results mentioned before, the main effects (offer and emotion) and the interactions (offer by study group and offer by emotion by study group) remained significant.

[7] 217w As mentioned in the 'Methods' section, the Hexagon test was used to measure emotion recognition. Using the General Linear Model (GLM) analysis, we found that the schizophrenia group (M = 92.8, S.D.= 17.3) recognized the basic emotions at a significantly lower rate (F =21.4, n = 87, d.f. = 1, p b 0.0001) than the control group (M = 111.3, S.D. = 6.6). Similarly, the recognition rate for surprise, anger, disgust, fear and sadness were significantly lower in the schizophrenia group, but no significant difference was found in the recognition of happiness (Table 3). Including the results of the Emotion Hexagon test in the main analysis of UG as covariates (Hexagon, all the basic emotions: χ 2 =1.2, d.f. = 1, p = 0.27; Hexagon, Happiness: χ 2 =0.3, d.f.=1, p =0.59; Hexagon, Surprise: χ 2 = 0, d.f. = 1, p = 0.95; Hexagon, Anger: χ 2 = 1.3, d.f. = 1, p = 0.26; Hexagon, Disgust: χ 2 = 0.8, d.f. = 1, p = 0.37; Hexagon, Fear: χ 2 =1.4, d.f.=1, p = 0.24; Hexagon, Sadness: χ 2 =0.4, d.f.=1, p = 0.55) did not affect the results mentioned before, the main effects (offer and emotion) and the interactions (offer by study group and offer by emotion by study group) remained significant Table 4.

[1] 259w The most important finding of the study is that there were significant differences in acceptance rates between schizophrenia patients and healthy controls depending on the fairness of the proposal. Namely, at lower offers acceptance rate was higher in schizophrenia, as reflected by ORs of 4.55 at the proposal of 10 HUF, and 3.44 at the proposal of 20 HUF. Contrary to unfair proposals, at higher, fair proposals, the average acceptance rates were lower in the schizophrenic group (OR = 0.34 for the proposal of 40 HUF) (Fig. 1). This intriguing pattern of results indicate that schizophrenia patients behaved differently compared to healthy controls both at unfair and fair proposals, but in opposite directions. However, the difference was more pronounced at the lower tail as reflected by the aforementioned odds ratios. According to the higher acceptance rate of unfair proposals one might see schizophrenic patients as rational maximizers, without the emotional need to punish unfairness. However this interpretation is not supported by the observed lower acceptance rates at fair proposals, the purely "rational" range of the UG. Based on this pattern of results and on the significantly higher ratio of inconsistent decisions (rejecting higher offers after accepting lower offers) in the schizophrenia group, we can suggest that rather than being rational maximizers, schizophrenia patients seem to be "inconsistent maximizers", following a paradox strategy. Furthermore we found that this inconsistency did not depend on other factors like PANSS scores or demographic variables (age or education), only a marginally significant association was found between the suspiciousness subscale (P6) of PANSS and inconsistency.

[1] 181w Participants were 58 patients with schizophrenia and 29 healthy control volunteers (Table 1). The diagnosis was based on the DSM-IV criteria (American Psychiatry Association, 1994), as confirmed by the Hungarian version of the MINI 5.0 diagnostic interview (Balázs et al., 1998) and by information from the personal files of the patients (namely, medical files were used for screening and DSM IV and MINI 5.0 for confirmation of the diagnosis). The controls were also screened with the MINI interview. All patients were in a stabilized clinical state (there were no signs or symptoms of acute psychosis and the positive symptoms score was under 35) and were able to cooperate with the study protocol. Clinical symptoms were evaluated with the Positive and Negative Syndrome Scale (PANSS) (Kay et al., 1987). Detailed demographic data can be found in Table 1. Exclusion criteria for patients and controls included age older than 65 years and younger than 18 years, comorbid neurological disease, head trauma, mental retardation, and substance abuse. All patients gave written informed consent. The study was approved by the Semmelweis University Institutional Ethics Board.

[2] 199w Participants acted as responders in a series of 40 trials of the ultimatum game. Before completing the task, the participants were told that they will play with another person online whose picture will be depicted on the screen. Written instructions preceding the task were the following: "Dear Participant! You will play a game with four people who will log in the task via our computer network. The rules of the game are the following: one of the distant players gets 100 Hungarian Forints (HUFs). The player fellow divides this sum between you and himself/herself as he/she prefers: the fellow player can keep a part of it, and offer another part to you. Your task is to decide whether to accept this offer. If you reject the offer, neither you nor the other player can keep the money; in this case you both get zero HUF. If you accept the offer, then both you and the other player can keep the amount offered. The collected sum will be paid to you after the end of the task. Do you understand the task?" The task started after the examiner made sure that the participant had read and understood the instructions thoroughly.

[3] 145w In each trial, the participant first saw a picture of a person showing anger or happiness on his/her face and the person's name with the following sentence: "Peter offers you a deal." Next, the participant saw the offer for 5 s. The offer is a take-it-orleave-it split of 100 HUF (approximately 50 US cents), for example, "Peter keeps 80 HUF and gives you 20 HUF". Then, the participant saw the question "Accept or Reject?" on the screen. The participant had unlimited time to consider the offer and push the "Accept" or "Reject" button to respond. Finally, the outcome based on the response was presented (e.g., "You both get zero HUF" if the offer was rejected or "Peter gets 80 HUF, you get 20 HUF" if the offer was accepted). The presentation time for the result was 5 s and the intertrial interval was 1 s.

[4] 93w Participants received 40 offers from four different proposers males and two females) in a randomized order. Each offer (10-50 HUF) was offered twice by each proposer with negative (Anger) and with positive (Happiness) facial expression as well. The facial expressions were presented by healthy volunteers and both angry and happy faces were validated before the study by students. Before beginning, the participants were instructed about the contingencies of an "Accept" or "Reject" response. They were told that the offers were real and that both players would be paid according to the participant's decisions.

[5] 133w As the use of deception was necessary because of the nature of the study, we followed the recommendations on deception in scientific experiments (Wendler and Miller, 2004). Before consenting, participants were explained that during the study they will be asked to complete different test batteries and play with different games, during which they might play with a computer or with a person (telling more about being deceived might have biased the participant's attitude and thus the results). At the end of the experiment, the participants underwent debriefing to explain that during the ultimatum game they were playing with a computer instead of a person. None of the subjects reported any subsequent negative feeling about the deception. All subjects were given the same amount of money (500 HUF) at the end of the investigation.

[6] 166w Perception of facial expressions of emotions was assessed using The Emotion Hexagon Task, an experiment containing blended facial expressions posed by model "JJ" from the pictures of facial affect series by Friesen and Ekman (1976) and Sprengelmeyer et al. (1996). The test consists of blended continua ranging between the following six expression pairs: happiness-surprise, surprise-fear, fear-sadness, sadness-disgust, disgust-anger, and anger-happiness. Each continuum consists of five morphed images blended in the same proportions. For example, the images in the happy-surprised continuum contain the following percentages of the happy and surprised expressions, 90% happy-10% surprise, and then 70-30%, 50-50%, 30-70%, and 10-90% of the same two expressions. One stimulus set consists of 30 images (six continua × five morphed faces); the task consists of six stimulus sets (first set is for learning and results are not analyzed), so the whole test consists of 180 images. The task is a multiple choice test, where the responder has to select the most appropriate emotion label from the six basic emotions.

[7] 173w A battery of classical neuropsychological tests, covering a range of neurocognitive functions was used for neurocognitive assessment: Rey Auditory Verbal Learning Test (RAVLT), Wisconsin Card Sorting Test (WCST) and WAIS-R Digit Span Forward and Backward tests. The RAVLT is a 15-item word list for testing verbal learning and verbal memory (Strauss et al., 2006). The WCST is a commonly used instrument to measure frontal executive function, such as concept formation, set-shifting and flexibility. Subjects are required to sort cards according to different rule dimensions (color, form, and number), but the actual sorting principle is not explicit, participants have to deduce it from the verbal feedback of the examiner. In this study we used the manual, 128-card version, administration and scoring of the WCST was conducted according to the test manual (Heaton et al., 1993). WAIS-R Digit Span Forward and Backward tests require subjects to remember and rearrange brief lists of numbers. As part of the WAIS-R test battery, they are used for the assessment of short term memory span (Strauss et al., 2006).

[8] 139w The association between acceptance (yes or no) in the UG and dependent variables (amount of offer [10,20,30,40, and 50 HUF], proposer's facial expression [negative or positive], and study group [schizophrenic vs. control]) was investigated by Generalized Linear Model Analysis (GENMOD). In order to plot the response rates we collapsed the offers by calculating the mean of the binary response variable under given conditions, namely study group, offer and emotional facial expression of the proposer (e.g: control subjects, 40 HUF, positive emotion) (Figs. 123). The association between the dependent and independent variables in the model was tested by the likelihood ratio Chi-square statistic. The General Linear Model Analysis (GLM) was used to test the differences in emotion recognition tasks and in the neurocognitive tests. The statistical analysis was conducted by using the Statistical Analysis System (SAS for Windows, version 9.1.3).

[1] 246w In order to measure inconsistency two binary variables were introduced in the analysis of decision-making. We marked a decision (1st variable) 'inconsistent' if rejected a higher offer while any lower offers were accepted, and flagged a subject (2nd variable) if he or she has at least one inconsistent decision. The associations between study groups and decision inconsistency are analyzed by Logistic Regression (LOGISTIC), where decision inconsistency was the dependent binary variable and study group was the independent variable. The analysis revealed that study group has a significant main effect on decision inconsistency, namely the ratio of inconsistent decisions was significantly higher in the schizophrenia group (schizophrenia group: M = 0.11, S.D.= 0.31 vs. control group: M =0.04, S.D.=0.20; χ 2 =4.7, d.f.= 1, p = 0.03) and more subjects have at least one inconsistent decision in this group that in the control group (schizophrenia group: Mean = 0.38, S.D. = 0.49 vs. control group M = 0.14, S.D. = 0.35; χ 2 =6.0, d.f.=1, p = 0.01). It is important to note that including age and years of education as covariates did not affect these results and none of these covariates have a significant effect on inconsistent decisions (age: χ 2 = 0.5, d.f.= 1, p = 0.49; years of education: χ 2 =0.7, d.f.=1, p = 0.41) nor on the number of subjects with inconsistent decisions (age: χ 2 = 0.9, d.f.= 1, p = 0.34; years of education: χ 2 =1.1, d.f.=1, p =0.30).

[2] 154w The associations between symptom severity in the patient group and decision inconsistency were analyzed by repeated measurement logistic regression analysis (GENMOD) where inconsistency was the dependent variable and symptom severity as indexed by PANSS scores served as independent variables. We could not detect significant relationship between PANSS total score (χ 2 = 0.04, d.f. = 1, p = 0.84), 'positive PANSS score' (χ 2 = 0.00, d.f. = 1, p =0.99), 'negative PANSS score' (χ 2 =0.6, d.f.=1, p = 0.44) and decision inconsistency. Furthermore, we analyzed the association between inconsistency and specific PANSS items, which might influence decisions like delusions (P1: χ 2 =7.7, d.f.= 5, p = 0.17), conceptual disorganization (P3: χ 2 =6.5, d.f.=4, p = 0.16), suspiciousness/persecution (P6: χ 2 = 8.2, d.f. = 4, p = 0.08), poor rapport (N3: χ 2 =2.6, d.f.=4, p = 0.63) and disturbance in volition (G13: χ 2 =2.0, d.f.=3, p = 0.57).

[3] 162w The consequence of higher acceptance rates at unfair proposals among schizophrenia patients may indicate their decreased ability to respond to unfair proposals by rejection. In other words, patients lose their ability to perform altruistic punishment, which is the core concept of the UG bargaining. Altruistic punishment, the enforcement of social norms by altruistic sanctions, is considered as a distinguishing characteristic of humans that is necessary for cooperation in groups. The pivotal work of DeQuervain et al. using fMRI showed that the neural basis of altruistic punishment is the activation of the anterior dorsal striatum, a brain-region associated with anticipated rewards (de Quervain et al., 2004). Moreover, the activation of the anterior dorsal striatum was correlated with the willingness to engage in altruistic punishment. In light of the aforementioned results, it is conceivable that the lower levels of altruistic punishment in schizophrenia are explainable by decreased reward sensitivity or the inability to discount future gains connected to momentary losses (Gold et al., 2008).

[4] 253w In another functional neuroimaging study Sanfey et al. (2003) examined the emotional aspects and the neural correlates of UG decision-making. Responders of the UG displayed activation of the insula, the anterior cingulate cortex (AC) and the dorsolateral prefrontal cortex (DLPFC). Insula activation was consistently higher in case of unfair offers, suggesting negative affect. Hence, we can say that altruistic punishment is a bias in decision-making based on the negative emotions provoked by unfair offers during the game (Fowler et al., 2005;van 't Wout et al., 2006). This effect may be weaker in schizophrenia, reflecting the dysfunctional recruitment of the neuronal network of social decision-making (Baas et al., 2008). It is important to note that when the results of the neurocognitive tests were included in the analysis, the effects remained significant. Therefore, impaired performance on tests of executive functions, attention, and memory did not explain the observed dysfunction in UG. Furthermore we found that years of education and age as covariates did not affect the results mentioned earlier, however older subjects showed a tendency to accept offers at higher rates, and this tendency reached significance, while there was no association between inconsistency in decision-making and age. It should be noted that age effects on certain social cognition measures did not reveal the same pattern as observed in standardized neurocognitive tests: in some cases performance can develop with age. To our knowledge there is no literature on the association between age and the acceptance rates in UG, an issue deserving further investigation in further studies.

[5] 286w We also explored the relationship between clinical symptoms and UG performance. The results revealed that positive and negatives symptoms, as indexed by the PANSS, influenced the acceptance rates in the UG (Fig. 3). Acceptance characteristics of patients with milder positive and negative symptoms (below the median score) had a similar acceptance pattern to that of controls, while subjects with severe positive or negative symptoms (above the median score) tended to display a weaker altruistic punishment. Furthermore we analyzed the relationship between specific PANSS items and acceptance rates in the UG and found that two items, namely P1 (delusions) and P6 (suspiciousness/persecution) have a positive correlation with acceptance rates (other items have no significant correlation). Why delusional or suspicious subjects tend to accept offers at a higher rate? A possible explanation of these phenomena is that these subjects may be apprehensive of the consequences of rejection. However, further investigations are needed to explore this issue. Here we would like to get back to the earlier proposed paradox strategy. First, a subgroup of patients with higher scores on delusion and suspiciousness subscales of the PANSS can cause this trend by accepting all offers at a higher rate. Additionally it seems that several patients follow a paradox strategy by accepting some lower offers and rejecting higher offers and this strategy cannot be associated with any factors within the study group (no significant association were found between decision inconsistency and PANSS scores or PANSS item scores). It is likely that a further investigation with a larger patient sample will be able to detect relationship between inconsistent decision-making and psychotic symptoms as indexed by the PANSS (Here we refer to the marginally significant association between decision inconsistency and the suspiciousness subscale).

[6] 230w It has been demonstrated that the decisions of the accepter could be modulated by the emotional facial expressions of the proposer (Scharlemann et al., 2001). In the case of healthy subjects, we found that positive facial displays facilitate the cooperativeness at higher offers, but not at lower offers. A possible explanation of this finding is that the strong negative emotions (anger) invoked by unfair offers can make the accepter unaffected by any facial expression, thus cannot be modulated any longer. In this respect, the most important finding was that, the facial expression of the proposer did not influence the decision of schizophrenia patients, contrary to healthy subjects (Results Section 3.1 -Ultimatum Game, Fig. 2.). Since the poorer performance of the patients on facial expression recognition is well documented in the literature (Marwick and Hall, 2008;Carter et al., 2009) and in the case of negative emotions (but not in case of positive emotions) we also found that healthy controls performed better than patients with schizophrenia, we included the results of the emotion recognition task as covariates in the main analysis of the UG and found that the results (including the emotion by offer by study group threeway interaction) remained significant. This suggests that the finding indicating that the facial expression of the proposer had no effect on the performance of the patients cannot be explained by the decreased emotion recognition ability.

[7] 219w Before concluding the findings, we have to say a few words about the limitations of the present study. As we pointed out earlier, besides the disorder of schizophrenia itself, psychotic symptoms, as indexed by the PANSS, also affect the UG outcome. This finding raises the question if the difference between subjects with schizophrenia and controls is a state-like or trait-like deficit. Since the schizophrenia subjects in the present study were predominantly (40:18) inpatients showing more severe positive and negative symptoms than the outpatient subgroup, this issue cannot be resolved unambiguously based on the present results while the sample size of the outpatient subgroup is not large enough. Further investigations involving larger outpatient samples or the extension of the present investigation (including a larger sample of subjects without psychotic syndromes) are needed to disentangle this issue. Another limitation of the current study is that we did not scan the study population for any history of ADHD, which disorder has a relationship with decision-making. It has been reported recently that behaviors associated with ADHD have a correlation with maximizing tendency in decision-making (Schepman et al., in press). In conclusion, we presented empirical data based on experiments investigating the decision-making processes of patients with schizophrenia. The interpretation of the results is limited by the computerized setting where patients only acted as responders.

[8] 63w However, our results shed light on the differences between the social decision-making strategies of schizophrenia patients and healthy controls both at unfair and fair proposals. These findings might be implicated in psychosocial treatment and rehabilitation programs aiming to improve social cognition and community functions in schizophrenia patients. Further translational research should investigate the associations of clinical symptoms, cognitive dysfunction, social cognition and decision-making.