Autistic traits and positive psychotic experiences modulate the association of psychopathic tendencies with theory of mind in opposite directions

Various clinical disorders, including psychopathy, and autism and schizophrenia spectrum disorders, have been linked with impairments in Theory of Mind (ToM). However, although these conditions can co-occur in the same individual, the effect of their inter-play on ToM abilities has not been investigated. Here we assessed ToM abilities in 55 healthy adults while performing a naturalistic ToM task, requiring participants to watch a short film and judge the actors’ mental states. The results reveal for the first time that autistic traits and positive psychotic experiences interact with psychopathic tendencies in opposite directions to predict ToM performance—the interaction of psychopathic tendencies with autism traits was associated with a decrement in performance, whereas the interaction of psychopathic tendencies and positive psychotic experiences was associated with improved performance. These effects were specific to cognitive rather than affective ToM. These results underscore the importance of the simultaneous assessment of these dimensions within clinical settings. Future research in these clinical populations may benefit by taking into account such individual differences.


Results
Statistical analysis. We tested for the effects of autism traits (Autism Quotient; AQ), positive psychotic experiences (Community Assessment of Psychic Experiences positive subscale; CAPEp), primary psychopathic tendencies (Levenson Self Report Psychopathy primary subscale; P-LSRP), and all two-way interactions on total ToM errors using ordinary least squares regression models. All regression models were calculated using z standardized predictor values. The rationale for the two-way interactions was motivated by evidence for the co-occurrence of autism and conduct disorders/psychopathic tendencies, on the one hand, and for the co-occurrence of psychosis/schizophrenia and psychopathic tendencies, on the other. While controversial, we also included the interaction of autism and psychosis based on evidence for the co-occurrence of ASCs and SSDs 13,15,16 . We did not include the three-way interaction as, to our knowledge, there is no evidence suggesting that psychopathy, autism and psychotic spectrum disorders simultaneously co-occur. Significant interactions were probed with the Johnson-Neyman method using MODPROBE for SPSS 41 . This method provides a 'high-resolution picture' of the interaction by estimating the value(s) of one predictor, at which the other predictor has a significant effect on the outcome measure. This is established by identifying the precise value(s) along the continuum of one predictor for which the regression slopes of the other predictor are estimated to be significantly different from zero. Table 1 summarizes the sample scores on the various questionnaires and theory of mind measures.
The Johnson-Neyman analysis, presented in Fig. 1a,b, probes the interaction of P-LSRP and AQ. The analysis showed that including the interaction improved the model by 5.65% (F = 5.27, p = 0.026). As shown in Fig. 1a, the relationship of P-LSRP with the number of ToM errors was positive when AQ scores were at and higher than −0.53 SD from the mean (at AQ = −0.53 SD; β (SE) = 1.17 (0.58), t = 2.01, p = 0.05, CI 95 = 0.00-2.34). Conversely, in Fig. 1b, there was a significant positive relationship of AQ with number of ToM errors when the P-LSRP scores were at and higher than 0.54 SD above the mean (at P-LSRP = .54 SD; β (SE) = 1.00 (0.50), t = 2.01, p = 0.05, CI 95 = 0.00-2.00) of P-LSRP. Figure 1c,d probes the interaction of P-LSRP and CAPEp. The analysis showed that including the interaction improved the model by 15.54% (F = 14.49, p < 0.001). As shown in Fig. 1c, P-LSRP was associated with an increased number of ToM errors when the CAPEp scores were at and lower than 0.39 SD from the mean (at CAPEp = 0.39 SD; β (SE) = 1.02 (0.51), t = 2.01, p = 0.05, CI 95 = 0.00-2.04). However, this effect was reversed when the CAPEp scores were at and higher than 2.40 SD from the mean (at CAPE = 2.40 SD; β (SE) = −2.42 (1.20), t = −2.01, p = .05, CI 95 = −4.84-0.00). In Fig. 1d, there was a significant positive relationship of CAPEp and number of ToM errors when the P-LSRP scores were at and lower than −0.96 SD from the mean (at P-LSRP = −0.96; β (SE) = 1.31 (0.65), t = 2.01, p = 0.05, CI 95 = 0.00-2.62). However, this relationship reversed such that there was a significant negative relationship of CAPEp with number of ToM errors when the P-LSRP

Measure
Mean ± SD Range   For all panels (a-d), regression lines represent the significant zone (p ≤ 0.05) of the focal predictor at the specified value of the other predictor. Panels 1a and 1b probe the interaction between the P-LSRP and AQ scores. Panel 1a shows that increasing P-LSRP scores significantly increases ToM errors when AQ scores are equal or more than −0.53 SD from the mean. Panel 1b shows that increasing AQ scores significantly increase ToM errors when P-LSRP scores are equal or more than 0.54 SD from the mean. Panels 1c and 1d probe the interaction between the P-LSRP and CAPEp scores. Panel 1c shows that increasing P-LSRP scores significantly increase ToM errors when CAPEp scores are equal or less than 0.39 SD from the mean, but reduce ToM errors when CAPEp scores are more than 2.40 SD from the mean. Panel 1d shows that increasing CAPEp scores significantly increase ToM errors when P-LSRP scores are equal or less than −0.96 SD from the mean, but reduce ToM Errors when P-LSRP scores are equal or more than 0.49 SD from the mean.
Probing the interaction of P-LSRP with CAPEp showed that including the interaction improved the model by 18.3% (F = 16.67, p < 0.001). We found a significant positive relationship of P-LSRP scores with the proportion of cognitive ToM errors when the CAPEp scores were at and lower than .21 SD from the mean (at CAPEp = 0.21 SD; β (SE) = 0.026 (0.013), t = 2.01, p = 0.05, CI 95 = 0.00-0.051). However, this effect was reversed such that there was a significant negative relationship of P-LSRP with the proportion of cognitive errors when CAPEp scores were at and higher than 1.79 SD from the mean (at CAPE = 1.79 SD; β (SE) = −0.051 (0.025), t = −2.01, p = 0.05, CI 95 = −.10-0.00). Conversely, there was a positive relationship of CAPEp and cognitive ToM errors when the P-LSRP scores were at and lower than −1.15 SD from the mean (at P-LSRP = −1.15; β (SE) = 0.038 (0.019), t = 2.01, p = 0.05, CI 95 = .00-0.075). However, this relationship reversed such that the association between CAPEp and proportion of cognitive ToM errors was negative when the P-LSRP scores were at and higher than .21 SD from the mean (at P-LSRP = 0.

Discussion
In the present study, we examined the interplay between psychopathic tendencies, autism traits, and positive psychotic experiences in predicting performance on the MASC, a naturalistic and realistic test of ToM abilities. We found that primary psychopathic traits were associated with ToM task performance, but that these relationships depended on levels of both autism traits and positive psychotic experiences. In particular, we found that the interaction of psychopathic tendencies and autism traits was associated with a decrement in ToM performance, whereas the interaction of psychopathic tendencies with positive psychotic experiences was associated with improved ToM performance.
Consistent with our hypothesis, there was a significant interaction of primary psychopathic tendencies with autism traits such that high scores on both measures were associated with a greater overall number of ToM errors. This finding is in contrast to those of an earlier study that failed to find an interaction of psychopathic and autistic traits in a developmental sample 32 . The results reported here are therefore the first to show support for the expected deleterious effects of this 'double-hit' of psychopathic tendencies and autism traits on ToM abilities. Conversely, the interaction of primary psychopathic tendencies with positive psychotic experiences showed that increasing scores on both scales was associated with significantly improved overall ToM performance. Notably, these results are consistent with findings suggesting that extreme levels of psychopathic tendencies among patients with schizophrenia was associated with better ToM abilities, compared with patients with lower psychopathic tendencies 12 . The results reported here are the first to show evidence for a similar effect in relation to the broader phenotype of these disorders in a non-clinical sample.
When considering the proportion of cognitive and affective ToM errors, it transpires that the interactions we observed for overall ToM errors were specific to cognitive ToM errors. As for the overall ToM errors, the effect of psychopathic tendencies on cognitive ToM errors depended on the relative expression of autism and psychosis traits such that, on the one hand, increasing psychopathic and autistic traits were associated with increased cognitive ToM errors, and on the other hand, increasing psychopathic and psychosis traits were associated with reduced cognitive ToM errors. Affective ToM errors were uniquely and positively associated with psychopathic tendencies.
In an attempt to shed light on the mechanism underlying this pattern of results, we make two main observations. First, autism traits and positive psychotic experiences interacted with psychopathic tendencies in opposite directions. This pattern of results is consistent with the diametric model 42 , which posits that ASCs and SSDs have opposing effects on ToM abilities, whereby ASCs are associated with reduced ToM and SSDs with hyper-ToM, a tendency to over-attribute/-generate mental states about others 43 . It has been shown that psychopathy and autism are associated with distinct neurocognitive underpinnings 33,35,44 , and that the features associated with these conditions may not overlap 32 . Thus, the combined worsening effect of autistic traits and psychopathic tendencies on ToM performance is conceivable in light of robust evidence associating autism, and its broader phenotype, with impaired cognitive ToM 20, 31, 34 .
The mechanism underlying better cognitive ToM abilities in individuals scoring high on both psychopathic tendencies and positive psychotic experiences is less clear. However, we speculate that if positive psychotic experiences are associated with a greater ability to generate hypotheses about the mental states of others 42,43 , then this might be advantageous for individuals with co-occurring high psychopathic tendencies. Imaging studies have shown that individuals with clinically elevated levels of psychopathy or ASPD show a greater reliance on cognitive processes, thought to represent a compensatory mechanism in the absence of affective processing 36,45,46 . Thus, when combined with a greater number of alternative mental state attributions, it is possible that this cognitive route may prove beneficial for individuals with co-occurring high psychopathic tendencies and psychosis traits. Consistent with this hypothesis, there is evidence to suggest that the co-occurrence of psychopathic tendencies in SSDs may be associated with an increased propensity for interpersonal charm, cunning, and instrumental aggression 12 .
Second, we observe that while psychopathic tendencies had a unique and negative effect on affective ToM errors, its effect on cognitive ToM errors was moderated by the relative expression of autistic traits and psychotic experiences. It has been argued that situations that involve affective ToM entail more self-reflection, and require the integration of affective and cognitive information, compared with situations involving cognitive ToM 6 . Psychopathic tendencies are associated with reduced affective resonance to others emotional expressions, and reduced activation of regions including the amygdala and anterior insula that are involved in emotion and empathy 47 . Hypoactivation of these regions has also been observed in relation to psychopathic tendencies during affective ToM task performance, and may reflect a failure to process emotionally salient cues, including facial expressions and body postures 48 . In contrast, autistic traits, as pointed above, are related to impairments in cognitive ToM, but not in affective resonance or emotional empathy, that is, the ability to feel what another is feeling 31,33,44 . Indeed, individuals with autism may resonate with the emotional experience of others to a normal, or even heightened, degree 49,50 . This distinction between psychopathic tendencies and autism traits likely accounts for the absence of a psychopathy-autism interaction in predicting affective ToM.
Taken together, our findings may have implications for understanding the aggressive behaviours associated with psychopathic tendencies. For example, better ToM abilities may be associated with a more cunning and deceitful interpersonal style in psychopathy, and a greater incidence of premeditated or goal directed aggression 12,51,52 . Thus, it may be predicted that the aggressive behaviours of those with elevated psychopathic tendencies and high levels of positive psychotic experiences would be more instrumental compared with those who score relatively more highly on one measure compared with the other. In contrast, adolescents with ASCs and elevated psychopathic traits do not appear to show elevated rates of conduct problems 14 . This finding is interesting and suggests that psychopathy related impairments that are not seen in autism, including in resonating with others emotions and understanding others affective states, may account for increased rates of aggression in relation to psychopathic, but not autistic traits 33,44 .
Our results need to be interpreted in light of several limitations. The sample recruited in this study was relatively small with the majority being female college students. While we observed no effect of participant sex on ToM task performance in the current study, the extent to which there may be sex differences in the relationships should be investigated further, given previous reports highlighting differential relationships among boys and girls in the relationship of psychopathic traits with empathic abilities 53 . Furthermore, despite recent findings on the relationship of psychopathic tendencies with ToM in patients with schizophrenia 12 , these results are yet to be mirrored in clinical patient samples with extreme levels of autism.
Our study is the first to show that autistic traits and positive psychotic experiences interact with psychopathic tendencies in opposite directions to predict ToM performance. This modulation is specific to cognitive rather than affective errors. We are aware of the debate regarding the relevance of effects associated with attenuated clinical expressions in the healthy population to clinical entities. However, these results underscore the importance of the simultaneous assessments of these dimensions within clinical settings, particularly given the increased recognition of the co-occurrence of these conditions, and the dimensional view of psychopathology. Perhaps the inconsistencies in the literature with regard to the relationship between psychopathic tendencies and ToM abilities reflect individual differences in the expression of autism and psychosis. Future research concerned with socio-cognitive abilities in these clinical populations may benefit by taking into account such individual differences.

Method
Participants. A sample of 55 healthy adults (16 males, 39 females), aged between 18 and 37 (M (SD) = 20.00 ± 2.59) were recruited from the student population at the University of Birmingham, UK, and participated in return for course credit. Participants reported no history of psychiatric illness, epilepsy, neurological disorders, brain injury or alcohol or substance abuse problems. Written informed consent was received from each participant after the nature of the study had been explained. All experiments were carried out in accordance with relevant guidelines and regulations and approved by the University of Birmingham Research Ethics Committee.

Materials
The Movie for the Assessment of Social Cognition (MASC). The MASC 38 is a 15-minute video film depicting several scenarios revolving around four friends making arrangements to meet for dinner, preparing and eating a meal, and playing a game. The scenarios are played out in different locations between two, three, or all four of the characters, both in person and over the phone. Across these different social situations the characters show misunderstandings, express irony and ambiguous body language, act flirtatiously, and make insulting comments. The task consists of 51 multiple-choice questions of which six were control questions. The critical 45 questions aimed to test the participants' abilities to infer the thoughts, intentions, feelings, and emotions of the characters. Each question was presented following a short clip from the movie. The total number of correct responses reflects an individual's general ToM performance. In addition, the MASC can distinguish between cognitive and affective ToM on the basis of questions that ask about knowledge, beliefs and intentions (28 questions), versus feelings and emotions (17 questions). An example cognitive question asks, "What does Michael think Cliff is laughing about?" An example affective question asks, "What is Sandra feeling?" Levenson Self Report Psychopathy scale (LSRP). The LSRP 54 assesses psychopathic traits in non-institutionalized populations. It contains 26 items measured on a four-point Likert scale. Sixteen items measure 'primary' traits, including selfishness and a lack of care for others, and the remaining items tap 'secondary' traits, which include proneness to boredom and impulsivity. Internal consistency in the present study was Cronbach's α = 0.88 for the primary subscale (P-LSRP) and 0.70 for the secondary subscale (S-LSRP). There is no recommended clinical cut-off for the LSRP.
The Community Assessment of Psychic Experiences (CAPE). The 42-item CAPE questionnaire 55 is a reliable measure for the assessment of the expression of psychosis dimensions in clinical and research settings.
Positive psychotic experiences were assessed using the CAPE's 20-item positive subscale (CAPEp). The internal consistency of this subscale in this study is good (Cronbach's α = 0.74). Although there is not a well-established cut-off score for the CAPEp, a score of 50 has been shown to have some validity 56 . The Autism Spectrum Quotient (AQ). The AQ 29 consists of 50 items that measure the presence of traits associated with the autistic spectrum within the general population. Each item is given a score of 0 or 1. The AQ's internal consistency in this study is good (Cronbach's α = 0.71). A score of 32 has been recommended as a clinical cut-off for the AQ 57 .
The National Adult Reading Test (NART). The NART 58 consists of 50 words of irregular pronunciations.
The total number of correct pronunciations is used to predict the full, verbal and performance Wechsler Adult Intelligence Scale (WAIS) IQ 59 . -Revised (CESD-R). The CESD-R 60 is a 20-item self-report scale that closely reflects the DSM-IV criteria for depression, and assesses the individual's level of depressive symptomatology experienced over the last two weeks. The internal consistency in this study is excellent (Cronbach's α = 0.95). Although a score equal to or above 16 has been recommended for identifying a person at risk for clinical depression, this value has been criticised for producing too many false positives 61 .

Centre for Epidemiologic Studies Depression Scale
Data availability. The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.