Normal Sexual Dimorphism in Theory of Mind Circuitry is Reversed in Schizophrenia

Abstract

The ability to mentalize, or theory of mind (ToM), is sexually dimorphic in humans and impaired in schizophrenia. This sex-stratified study probed cognitive (indexed by intelligence) and affective (indexed by olfactory tasks) contributions to ToM performance in 37 individuals with schizophrenia and 31 healthy controls. The schizophrenia group showed impairments in mental state identification and inferring intentions compared to controls. Higher intelligence was correlated with mental state identification and inferring intentions in healthy females, whereas better smell identification was associated with mental state identification in healthy males. Conversely, higher intelligence was associated with mental state identification and inferring intentions in schizophrenia males, while better smell identification was correlated with mental state identification in schizophrenia females. These findings suggest that for ToM circuitry, the cognitive influences in healthy females and affective influences in healthy males are reversed in schizophrenia and may be displaced to lower circuitries by disease pathology. Symptom associations with emotion and cognition are also dimorphic, plausibly due to similar pathology superimposed on normal sex-specific circuitries. Males appear to rely on limbic processing for ToM, and disruption to this circuitry may contribute to development of negative symptoms. These findings highlight the importance of utilizing sex-stratified designs in schizophrenia research.

Introduction

Observations of sex differences in schizophrenia were documented in early conceptualizations of the disease (Kraepelin, 1919/1971) and over the past few decades renewed interest has resulted in strong empirical support for sex differences. Compared to males, females with schizophrenia typically have a later illness onset (Hӓfner, Maurer, Lӧffler, & Riecher-Rӧssler, 1993; Szymanski et al., 1995), better premorbid functioning (McGlashan & Bardenstein, 1990), and less severe illness course (Goldstein & Link, 1988) with better social functioning post-onset (Morgan, Castle, & Jablensky, 2008). Most studies of sex differences in symptomatology find that males exhibit greater severity of negative symptoms (Andreasen et al., 1990; Chang et al., 2011; Gur, Petty, Turetsky, & Gur, 1996; Ring et al., 1991, Schultz et al., 1997) whereas females display more mood symptoms (Goldstein & Link, 1988; Morgan et al., 2008; Szymanski et al., 1995; Walsh-Messinger et al., 2018). Most, but not all studies, support greater cognitive impairment in males (Goldstein, 1998; Han et al 2012; Seidman et al., 1997; Sota & Heinrichs, 2003; reviewed in Mendrek & Mancini-Marie, 2016). Neuroanatomical differences have also been identified, including decreased amygdala volume (Gur et al., 2000; Niu et al., 2004) and increased orbitofrontal cortex to amygdala ratio in males (Gur et al., 2004).

Sexual dimorphism refers to the differences between males and females that may be an advantage to their reproductive behavior, including sex-differences in size, cognition and behavior. In schizophrenia, the sexual dimorphisms observed in healthy individuals are often reversed (Gur, Gunning-Dixon, Turetsky, Bilker, & Gur, 2002; Gur et al., 2004; Malaspina et al., 2012a; Strauss et al., 2015). For example, Strauss et al. (2015) found that heathy females were more accurate than healthy males at identifying affective body expression, but this sexual dimorphism was reversed in schizophrenia despite both sexes demonstrating impairment compared to the controls. It has been proposed that the neuroprotective effects of estrogen may partially explain some of these sex differences (e.g. Garcia-Segura, Azcoitia, & DonCarlos, 2001). However, it is also possible that an excess of testosterone in utero could contribute to the increased social and emotional impairment observed in schizophrenia males, as similar findings are reported in autism (Auyeung et al., 2009), which occurs more frequently in males (Chakrabarti & Fombonne, 2005) and for which social impairment is a predominant symptom.

An important component of social engagement is the ability to accurately identify other people’s emotions and interpret their intentions, known as mentalizing or “theory of mind” (ToM; Premack & Woodruff, 1978). Deficits in ToM are well documented in schizophrenia (Bora, Yucel, & Pantelis, 2009; Brune, 2005b) and likely contribute to social incapacity, a prominent and disabling feature of the disease. There is also evidence of ToM impairment in individuals identified as having ultra-high risk for schizophrenia and in unaffected first-degree relatives (Bora & Pantelis, 2013), suggesting that the ToM deficits observed in schizophrenia are not just artifacts of having a disease.

Strong associations are consistently reported between ToM performance and cognitive function, specifically generalized intelligence, attention, memory, and executive function (Brune, 2005b; Corcoran, Mercer, & Frith, 1995; Jha, 2012), which suggests that cognition plays an important role in mentalizing. However, cognitive function does not fully account for ToM capacity in schizophrenia (Bozikas et al., 2011; Brune, 2005a; Doody, Gotz, Johnstone, Frith, & Owens, 1998; Pickup & Frith, 2001), supporting dysfunction in other neural processes, including the processing of affective information. Additionally, some schizophrenia studies find that ToM deficits are associated with disorganization and paranoia (Langdon, Siegert, McClure, & Harrington, 2005; Pickup & Frith, 2001; Sarfati & Hardy-Bayle, 1999). Thus, a schizophrenia-related impairment that produces psychosis may also underlie ToM deficits. Affective processes, via limbic circuitry, may also be involved in this pathology, consistent with imaging studies of mentalizing (Martin, Dzafic, Robinson, Reutens, & Mowry, 2016; Sugranyes, Kyriakopoulos, Corrigall, Taylor, & Frangou, 2011).

The use of odor cues to identify the intentions and optimal interactions with conspecifics is characteristic of many species (Stockhorst & Pietrowsky), including primates, and is analogous to human mentalizing which employs visual (Ekman & Friesan, 1975) or auditory stimuli (Bachorowski, 1999) along with contextual cues. Chemodetection and processing is a more anciently evolved social signaling system through which threatened states, stress, and sexual receptivity can be sensed (Malaspina, Corcoran, & Goudsmit, 2006). Olfactory processing particularly entails the function of the limbic system, initially called the rhinencephalon, or smell brain, and olfaction is hypothesized to have driven the evolution and elaboration of the limbic system as the emotional brain circuitry (Insel & Fernald, 2004). Thus, olfactory paradigms present useful indicators of emotional functioning. While humans have less conscious awareness of olfactory processes, studies indicate that olfactory information can influence human emotions. For example, one study found that male fear chemosignals generated fearful facial expressions and increased sensory vigilance in females, while male disgust chemosignals produced disgusted facial expressions and behaviors in females (de Groot, Smeets, Kaldewaij, Duijndam, & Semin, 2012). Other work shows that chemosignals from an aggressive person can induce an anxiety reaction (Mutic, Parma, Brünner, & Freiherr, 2016). Better olfactory and social function are furthermore associated in older women, including number of friends, relatives, and social engagement, with no significant associations in older men (Boesveldt, Yee, McClintock, & Lundström, 2017).

Smell identification deficits are well documented in individuals with schizophrenia, and show strong and separate relationships to intelligence and social impairments (Malaspina & Coleman, 2003; Malaspina et al., 2012b), including those defined as negative symptoms (Cieslak et al., 2015; Ishizuka et al., 2010; Kamath, Moberg, Kohler, Gur, & Turetsky, 2013). Odor threshold and smell identification are distinguished as lower versus higher order olfactory processes (Martzke, Kopala, & Good, 1997), although both are under variable control by higher cortical structures (Tanabe, Iino, & Takagi, 1975).

Despite the knowledge on sexual dimorphisms in social processes, very few studies examining ToM have specifically investigated sex differences, particularly in adults. Baron-Cohen, Jolliffe, Mortimore, & Robertson (1997) found that healthy adult females outperform their male counterparts in mental state decoding. This finding was replicated in middle school aged children (Ibanez et al., 2013), although another study found only a slight female advantage for inferring the intentions of others in younger, but not older children (Charman, Ruffman, & Clements, 2002).

Schizophrenia studies accounting for sex effects have found group differences in ToM between patients and controls (Bliksted et al., 2014, de Achával et al., 2010, Montag et al., 2011). If there are sex differences, as we expect, these group differences point to the robustness of the deficit, which may be more pronounced in male patients who are more represented in these studies. Investigations of group by sex interaction effects and, optimally, sex specific comparisons between patients and controls are needed to fully understand sex effects for ToM. Indeed, the two previous studies that utilized this approach to examine sex differences with respect to ToM in schizophrenia show sex effects. One study compared individuals with the deficit syndrome (a form of schizophrenia marked by primary severe and chronic social and emotional impairment) to individuals with non-deficit syndrome schizophrenia and found that deficit syndrome females were significantly impaired in ToM compared to non-deficit syndrome females and healthy female participants, while males with and without deficit syndrome schizophrenia were similarly impaired compared to healthy males (Csukly, Polgar, Tombor, Benkovits, & Rethelyi, 2014). Another study found that compared to males, females with schizophrenia had better overall mentalizing ability, as well as superior mentalizing abilities for both self and other mental states (Abu-Akel & Bo, 2013). However, the latter study did not include a control group which limits our understanding of these sex differences in relation to normal sex differences in social function, which could provide important information on the pathology of schizophrenia.

More than a century ago, Bleuler (1950/1908) conceptualized that the split between affective and cognitive processes were central to schizophrenia pathology, yet to our knowledge, sex differences in the cognitive and affective contributions to ToM performance in schizophrenia have not been examined. To address this gap, the present study tested ToM in sex-stratified samples of individuals with schizophrenia and healthy comparison participants to probe differential cognitive (indexed by intelligence) and affective (indexed by olfactory tasks) contributions to ToM performance.

Methods

Participants

Participants with DSM-IV schizophrenia or schizoaffective disorder diagnoses (N = 37) were recruited from treatment settings at a Northeastern urban state psychiatric institute to participate in an NIMH funded study examining social function in schizophrenia. All participants in the schizophrenia group were clinically stable and maintained on stable doses of medication for at least one month prior to participating. Healthy control participants (N = 31) were recruited from postings at the medical centers and from internet advertisements. Schizophrenia and healthy control participants were excluded if they were pregnant or on oral contraceptives, had active substance dependence, upper respiratory infections, allergies or histories of epilepsy, rhinoplasty or a major head injury requiring medical treatment. Participants were asked not to smoke cigarettes, use cosmetics, or perfume on the day of testing and to refrain from eating or drinking for at least two hours before testing. Institutional Review Boards at both Columbia University and New York University approved the study and all participants provided written informed consent.

Measures

Diagnosis and symptoms

Lifetime and current DSM-IV psychiatric diagnoses were determined using the Diagnostic Interview for Genetic Studies (Nurnberger et al., 1994) and current symptom severity was assessed with the Positive and Negative Syndrome Scale (PANSS; Kay, Fiszbein, & Opler, 1987). PANSS items were rated on a scale of one (absent) through seven (extreme), and items were summed to yield positive (psychotic symptoms), negative (social and emotional deficits), and general psychopathology (including affective and anxiety features) scales. Research assessments were performed by master’s or doctoral level clinicians who were trained to maintain inter-rater reliability of 0.95 for DSM-IV diagnoses.

Theory of Mind (ToM)

The ability to identify another person’s mental state was assessed with the Reading the Mind in the Eyes (RME; Baron-Cohen, Wheelwright, Hill, Raste, & Plumb, 2001) test. Participants were shown 36 pairs of eyes and for each were asked to: 1) identify the gender of the person in the picture as a non-affective control task, and 2) identify the mental state of the person in the picture by choosing one of four possible complex mental states (e.g. playful, confused, arrogant) to measure the reflexive process of emotion decoding. Separate scores were computed for gender identification and the primary measure of mental state identification.

The Strange Stories Task (SST; Happe, 1994) was used to assess ability to infer intentions of others, a reflective process that entails reasoning. For the SST, participants read twelve brief vignettes that featured a main character speaking sarcastically or ironically. After reading each vignette, participants were asked to: 1) report the meaning or comprehension of what the character said, which requires inference of the character’s intentions, and 2) provide a justification for each response. Scores were summed to yield separate scores for the primary measure of inferring intentions, and for response justification. Both scores were also summed together to yield a total SST score.

Olfaction

Odor detection threshold, typically considered to be a lower order olfactory process, was assessed with the Smell Threshold Test (SST; Doty, 2000), which determines odor threshold based on detection of phenyl ethyl alcohol concentrations. The test employs a single staircase, forced-choice procedure in which the odorant is applied to each nostril separately in squeeze bottles, while the opposite nostril is occluded with foam tape. The test begins with a −6.00 log concentration of phenyl ethyl alcohol, after which concentrations presented are increased in full log steps until a concentration is reached for which the participant offers five consecutive correct odor detections. The staircase is then reversed and moved down or up in half log increments with two pairs of trials at each concentration. The mean of the last four staircase reversal points determines odor threshold, which is expressed as the absolute value, such that higher numbers correspond to better odor detection sensitivity.

Smell identification, a higher order olfactory process, was measured with the University of Pennsylvania Smell Identification Test (SIT; Doty, Shaman, Kimmelman, & Dann, 1984). The test includes 40 odor-impregnated strips and requires participants to scratch each strip to release an odor and to choose one of four possible odor names for each strip. Correct responses for each item are totaled to determine the SIT accuracy score, which can range from 0 to 40.

Cognition

The Wechsler Adult Intelligence Scale–Third Edition (WAIS-III; Wechsler, 1997)) was used to measure intellectual function. The WAIS-III yields a Full Scale Intelligence Quotient (FSIQ), as well as separate Verbal (comprehension and expression of language) and Performance (visuospatial skills and visual processing speed) IQs. As a global measure of intelligence, only FSIQ was used in all primary analyses.

Data Analysis

Data were entered and verified using the SIR Database Management Software (SIR 2002, SIR Pty Ltd) and data analyses were conducted with IBM SPSS (Statistics 23). Descriptive statistics of all measures were examined to identify key features of the data (e.g. non-normal distribution, outliers, skewness) that might influence inferential methods and all variables were determined to be normally distributed. Analysis of variance (ANOVA) was used to examine group, sex, and interaction effects across all demographic variables. Multivariate analysis of variance (MANOVA) was employed to examine these effects across the cognitive, olfaction, and ToM measures. Associations between the ToM, olfaction, FSIQ, and symptom measures were examined using nonparametric Spearman correlation, which is less affected by outliers and therefore the more conservative correlation to use in small samples (de Winter, Gosling, & Potter, 2016). For all analyses, a threshold of p < 0.05 was considered significant.

Results

Characteristics of the 37 schizophrenia participants (male N = 22; female N = 15 females) and 31 healthy controls (male N = 14 males; female N = 17) and comparisons across all measures are shown in Table 1. Groups were similar in sex composition (Χ² = 1.38, p = .239), age, and intelligence, without any sex or group-by-sex interactions on the latter two measures. The schizophrenia group was less educated than the control group (p = .009) and males were less educated than females (p = .019), although there was no significant diagnosis by sex interaction. As expected, the schizophrenia participants exhibited significantly more PANSS symptoms than controls (all p’s < .001). One control had subthreshold positive symptoms, and others demonstrated only minimal negative and general psychopathology symptoms.

Table 1.

Comparison of demographic data, intelligence, and symptoms in schizophrenia participants versus healthy control participants, by sex.

	Healthy Participants		Schizophrenia Participants		Statistics
	Males	Females	Males	Females	Diagnosis		Gender		Diag./Gen.
	Mean (SD)	Mean (SD)	Mean (SD)	Mean (SD)	F	p	F	p	F	p
	N = 14	N = 17	N = 22	N = 15
Age	29.6 (8.9)	36.4 (14.5)	31.4 (8.5)	30.5 (7.5)	0.69	.409	1.40	.241	2.35	.130
Education (category)	4.4 (1.0)	4.9 (0.8)	3.2 (1.7)	4.3 (1.4)	7.20	.009*	5.78	.019*	0.63	.432
Onset of Illness	---	---	21.3 (6.6)	25.7 (7.9)					t = 1.90a	.066
	N = 14	N = 12	N = 17	N = 13
	Multivariate Wilks’ Lambda for Verbal and Performance IQ				0.82	.445	0.02	.978	1.66	.201
WAIS-III Verbal IQ	108.6 (12.6)	106.2 (12.5)	101.8 (18.9)	105.9 (13.2)	0.76	.388	0.04	.835	0.66	.421
WAIS-III Performance IQ	96.8 (11.5)	100.1 (15.6)	99.9 (22.1)	97.7 (10.9)	0.01	.930	0.01	.905	0.40	.532
WAIS-III FSIQ	103.7 (11.4)	103.8 (12.8)	101.1 (20.8)	102.5 (11.4)	0.22	.645	0.03	.854	0.03	.865
	N = 13	N = 14	N = 21	N = 13
			Multivariate Wilks’ Lambda		12.22	.000**	0.83	.486	0.39	.760
Positive Symptomsb	7.0 (0.0)	7.0 (0.0)	12.2 (5.4)	10.2 (5.7)	15.07	.000**	0.90	.347	0.90	.347
Negative Symptomsb	7.9 (2.5)	7.3 (0.7)	13.6 (5.3)	12.1 (5.5)	22.72	.000**	0.95	.334	0.15	.697
General Psychopathologyb	16.1 (0.3)	17.2 (1.7)	24.6 (5.5)	23.1 (7.6)	32.41	.000**	0.03	.873	1.12	.294

Test Statistic: Analysis of variance (ANOVA), except where indicated.

^*p <.050,

^**p<.010

^aStudent’s T-test

^bAssessed with the Positive and Negative Syndrome Scale

WAIS-III = Wechsler Adult Intelligence Scale, Third Edition

Olfaction

There were no group, sex, or interaction effects for either of the olfaction tasks (see Table 2). No significant differences emerged between scores for the right and left nostril; therefore, mean odor threshold was used in all subsequent analyses. Correlations between olfaction tasks are displayed in Table 4. More sensitive odor threshold predicted better smell identification accuracy in healthy males (p = .018); the olfaction measures were not correlated in healthy females (p = .602), schizophrenia males (p = .891) or schizophrenia females (p = .380).

Table 2.

Comparison of olfaction performance in schizophrenia participants and healthy control comparison participants, by sex.

	Healthy Participants		Schizophrenia Participants		Statistics
	Males	Females	Males	Females	Diagnosis		Gender		Diag./Gen.
	Mean (SD)	Mean (SD)	Mean (SD)	Mean (SD)	F	p	F	p	F	p
	N = 13	N = 15	N = 20	N = 14
Smell Identification Test	31.5 (3.8)	33.1 (5.0)	30.6 (4.1)	32.4 (4.5)	0.53	.472	2.32	.133	0.01	.928
	N = 14	N = 13	N = 17	N = 13
	Multivariate Wilks’ Lambda for right and left Odor Threshold				0.19	.828	1.66	.201	1.50	.233
Right Odor Threshold	−4.77 (1.5)	−4.84 (1.8)	−4.26 (1.2)	−5.95 (2.5)	0.38	.538	3.34	.073	2.86	.097
Left Odor Threshold	−4.86 (1.7)	−4.80 (1.1)	−4.34 (1.9)	−5.77 (3.1)	0.16	.690	1.51	.224	1.81	.185
Mean Odor Threshold	−4.82 (1.5)	−4.82 (1.1)	−4.40 (1.4)	−5.86 (2.6)	0.46	.502	2.54	.117	2.54	.117

Test Statistic: Analysis of variance (ANOVA), except where indicated

^*p <.050,

^**p<.010

Table 4.

Spearman correlations of theory of mind, olfaction, symptom and intelligence measures, by diagnosis and sex. Marginal and significant R to Z comparisons are presented.

	Healthy Participants			Schizophrenia Participants			Correlation Comparisons (R to Z)
	All	Males	Females	All	Males	Females
	rho	rho	rho	rho	rho	rho
RME Mental State
WAIS-III FSIQ	.441*	.042	.765**	.579**	.736**	.310	Controls: Males vs. Females p=.035*; Males: Controls vs. Cases p=.030*
Smell Identification	.442*	.587*	.391	.295	−.034	.551*	Cases: Males vs. Females p=.090; Males: Controls vs. Cases p=.076
Mean Odor Threshold	.014	−.126	.019	.079	.442	−.540	Cases: Males vs. Females p = .010*
Positive Symptomsa	---	---	---	−.239	−.192	.006	---
Negative Symptomsa	−.172	−.075	−.204	−.239	−.029	−.323	---
General Psychopathologya	−.493**	−.194	−.665**	−.142	.201	−.381	---
SST Inferring Intentions
WAIS-III FSIQ	.452*	.397	.700*	.601**	.686**	.445	---
Smell Identification	.161	.259	.028	.200	.208	−.061	Females: Controls vs. Cases p=.039*
Mean Odor Threshold	.338	.384	.367	.180	.220	.199	---
Positive Symptomsa	---	---	---	−.138	−.187	.097	---
Negative Symptomsa	−.210	−.232	−.113	−.204	−.141	−.095	---
General Psychopathologya	−.219	.117	−.445	−.048	.005	−.017	---
Mean Odor Threshold
WAIS-III FSIQ	.064	−.289	.154	.365	.530*	.265	Males: Controls vs. Cases p=.042*
Smell Identification	.371	.640*	.160	−.062	−.037	−.279	Males: Controls vs. Cases p=.059
Positive Symptomsa	---	---	---	−.324	−.103	−.702*	Cases: Males vs. Females p=.087
Negative Symptomsa	−.068	−.063	.000	−.045	−.155	.229	---
General Psychopathologya	.220	.309	.194	.129	.256	−.055	---
Smell Identification
WAIS-III FSIQ	.113	−.158	.246	.149	.121	.198	---
Positive Symptomsa	---	---	---	−.133	−.068	.138	---
Negative Symptomsa	−.200	−.203	−.229	−.630**	−.724**	−.263	All: Controls vs. Cases p=.059; Males: Controls vs. Cases p=.088
General Psychopathologya	.034	.263	−.159	−.155	−.038	.142	---

^*p <.050,

^**p<.010

^aAssessed with the Positive and Negative Syndrome Scale

WAIS-III = Wechsler Adult Intelligence Scale, Third Edition; RME = Reading the Mind in the Eyes; SST = Strange Stories Task

Theory of Mind

Schizophrenia participants and healthy controls performed similarly on the RME gender identification task; however, schizophrenia participants were less accurate than controls on the RME mental state identification (p = .041), SST inferring intentions (p = .013) and SST justification (p = .039) measures (see Table 3). There were no significant effects for sex or sex-by-diagnosis interactions for the ToM performance.

Table 3.

Comparison of theory of mind performance in schizophrenia participants and healthy control comparison participants, by sex.

	Healthy Participants		Schizophrenia Participants		Statistics
	Males	Females	Males	Females	Diagnosis		Gender		Diag./Gen.
	Mean (SD)	Mean (SD)	Mean (SD)	Mean (SD)	F	p	F	p	F	p
	N = 14	N = 17	N = 22	N = 15
			Multivariate Wilks’ Lambda		2.36	.103	0.96	.390	1.58	.213
RME Mental State	25.9 (3.8)	25.6 (5.0)	21.0 (6.4)	24.9 (5.8)	4.36	.041*	1.81	.183	2.42	.125
RME Gender	34.8 (1.0)	34.3 (1.4)	34.3 (1.6)	34.4 (2.2)	0.26	.609	0.21	.647	0.61	.438
	N = 14	N = 13	N = 22	N = 15
Multivariate Wilks’ Lambda for Justification and Inferring Intentions					3.41	.040*	1.09	.344	0.73	.488
SST Inferring Intentions	11.4 (0.8)	11.2 (0.7)	10.0 (2.3)	10.6 (1.6)	6.53	.013**	0.30	.589	1.05	.310
SST Justification	21.5 (2.5)	21.7 (2.2)	19.1 (3.6)	21.0 (2.3)	4.44	.039*	2.01	.162	1.33	.254
Total Score (sum of above)	32.9 (2.8)	32.9 (2.5)	29.1 (5.5)	31.6 (3.6)	6.21	.015*	1.46	.231	1.48	.229

Test Statistic: Analysis of variance (ANOVA), except where indicated

^*p <.050,

^**p<.010

RME = Reading the Mind in the Eyes; SST = Strange Stories Task

Associations between Theory of Mind, Olfaction, Intelligence, and Symptoms

Presented in Table 4 are the Spearman correlations for ToM performance with all other measures. In both schizophrenia and healthy control groups RME mental state identification and FSIQ were correlated; however, this was differentially explained by sex, as higher intelligence and more accurate mental state identification was correlated only in female controls (p = .004) and schizophrenia males (p = .001). In contrast, better mental state identification was associated with better smell identification accuracy in male controls (p = .035) and schizophrenia females (p = .041). There was also a significant correlation between SST inferring intentions and FSIQ in female controls (p = .036) and schizophrenia males (p = .002), again showing a reversal of sex effects in the schizophrenia group.

Finally, we examined associations between ToM, olfaction, and schizophrenia-related symptoms, measured by the PANSS. Performance on both ToM tasks was not significantly associated with any of the PANSS symptom scales in the schizophrenia group or in male controls; however, in female controls less accurate mental state identification was related to more general psychopathology symptoms (p = .004). Higher negative symptom severity was associated with lower smell identification scores in schizophrenia males (p < .001), and this male-specific association accounted for the relation between smell identification and negative symptoms in the sex-combined schizophrenia group (p < .001). In females with schizophrenia, greater positive symptom severity was associated with less sensitive odor detection (p < .05). Positive symptoms were not analyzed in the control group, as only one female control had measurable positive symptoms, although she notably had the most sensitive odor detection.

Post hoc sex-specific multivariate regression analyses were conducted to detect separate contributions of FSIQ and smell identification to RME mental state identification, controlling for age. These results supported the univariate analyses, showing sex-specific associations that differed between the healthy controls and the schizophrenia participants. In female controls, FSIQ was the only significant predictor of mental state identification (t = 3.92, p = .004), with no significant variability attributable to smell identification ability (t = 1.10, p=.305). Conversely, in male controls, only smell identification predicted mental state identification (t = 2.54, p = .035), with no significant effect of FSIQ (t = 1.07, p = .317). In contrast, FSIQ was not a significant predictor of mental state identification accuracy in females with schizophrenia, consistent with cognitive dysfunction, whereas in males with schizophrenia (t = 3.14, p = .009) and female controls there was a significant effect for FSIQ, suggesting limbic dysfunction and compensatory (but inadequate) cognitive contributions. Controlling for education did not change the outcomes of these analyses.

Discussion

This study uniquely probed the sex-specific underpinnings of mentalizing in individuals with schizophrenia and healthy controls. The ability to the identify mental states and infer the intentions of others was assessed using standard ToM tasks, with cognitive and affective influences determined by associations with intelligence and olfactory function, respectively. The results support the centrality of mentalizing deficits in schizophrenia, as significant ToM deficits were demonstrated in individuals with schizophrenia compared to controls, despite similar intelligence and olfactory ability in both groups. Supporting the study hypotheses, sex differences in ToM were identified in individuals with schizophrenia and controls, but were converse between the groups. The associations between the ToM measures and intelligence suggest significant cognitive influences for emotion decoding and making social inferences in healthy females, while in healthy males, the association between mental state identification accuracy and smell identification suggest stronger limbic contributions to emotion decoding. Individuals with schizophrenia showed an apparent reversal of these normal sex differences, due to a group and sex dependent downward displacement of mentalizing to lower order circuitries. In addition, presumably because of impairments in the preferred limbic circuitry, there was a compensatory shift to some cognitive contribution for emotion decoding and inferring intentions in schizophrenia males, but this was not adequate for intact functioning.

The finding of ToM impairments in the schizophrenia group, despite equal intelligence across the groups, supports the theory that mentalizing deficits in schizophrenia do not simply emerge from generalized cognitive deficits. Although the present study did not find expected sex differences in ToM performance, other studies have reported a female advantage for ToM (Baron-Cohen et al., 2015) and sex differences have also been found in individuals with schizophrenia (Abu-Akel & Bo, 2013; Csukly et al., 2014). Better mentalizing by females could be a consequence of better task engagement or more vigorous recruitment of the neural regions involved in mentalizing. By contrast, the present results suggest that this female advantage may be explained by sexually dimorphic mentalizing circuitries that entail greater cognitive contributions in females. The similar ToM performance for males and females across both groups in this study did not obscure meaningful sex differences in the neural underpinnings of mentalizing that were revealed by differing associations with cognitive and affective tasks. These results are supported by a previous finding of sex differences in neural activation patterns during ToM tasks in which females showed greater prefrontal changes than males (Frank, Baron-Cohen, & Ganzel, 2015). The limbic contributions in males are consistent with a pivotal role for the amygdala (Fine, Lumsden, & Blair, 2001; Shaw et al., 2004; Stone, Baron-Cohen, Calder, Keane, & Young, 2003), but studies have not separately examined males and females.

While symptoms were not significantly associated with mentalizing in males or females with schizophrenia, less accurate metal state identification was related to general psychopathology symptoms in female healthy controls, suggesting that subthreshold psychopathology symptoms may impair emotion decoding in healthy females. To our knowledge, no other study has examined associations between schizophrenia related symptoms and ToM performance in healthy controls. Psychotic symptoms were not associated with any of the ToM or olfaction tasks at the group level; however, more severe positive symptoms were related to lower odor detection sensitivity in schizophrenia females. In contrast, the single female control with the most sensitive odor detection also had the greatest psychotic symptoms, suggesting a utility for these tests to identify persons at risk for psychosis in larger samples. The association of psychotic symptoms with ToM scores supports frontostriatal input to mentalizing ability, which may be disrupted by the schizophrenia disease pathology. Dopaminergic influences on mentalizing are consistent with findings in Huntington’s disease (Snowden et al., 2003) and Parkinson’s disease (Mengelberg & Siegert, 2003; Mimura, Oeda, & Kawamura, 2006; Peron, Grandjean, Drapier, & Verin, 2014; Saltzman, Strauss, Hunter, & Archibald, 2000). The association of negative symptoms with smell identification deficits for males with schizophrenia is well reported (Brewer, Edwards, Anderson, Robinson, & Pantelis, 1996; Malaspina et al., 2012a; Moberg et al., 2006) and consistent with schizophrenia pathology that impacts crucial social limbic circuitry in males, although the reasons for the association have been enigmatic. The present findings suggest that a general vulnerability for negative symptoms among males may pertain to their lesser cognitive involvement for interpreting social signals; instead relying more on the limbic circuitry that may be more directly involved in the schizophrenia pathology. The robust sex differences in controls may interact with schizophrenia related pathology to produce different profiles of sexual dimorphisms in schizophrenia.

The strong relation between ToM performance and intelligence in healthy females is consistent with greater cognitive influences on mentalizing, and in line with greater cognitive processing by females for interpreting the emotions and intentions of others. In schizophrenia females, disease related damage to fronto-limbic circuitry may shift mentalizing to the limbic circuitry, which appears to play a stronger role in healthy males. Among healthy females, those with less accurate mental state decoding showed more anxiety and affective symptoms, presumably attributable to subtle cognitive dysfunction. In males with schizophrenia, disrupted fronto-limbic circuitry, as evidenced by smell identification deficits, produced more negative symptoms.

This study is cross sectional and therefore cannot distinguish cause and effect. In healthy participants, mentalizing deficits may produce symptoms by disrupting social communication and emotional bonds essential for healthy functioning. Sex stratification in the present study yielded smaller sample sizes and increased the number of tests performed, which are limitations to this preliminary report. However, sex differences are evident and failing to account for these dimorphisms provides incomplete information about schizophrenia. The present findings are consistent with the large body of literature on sex differences in smell identification, revealing associations between impaired smell identification and negative symptoms in males with schizophrenia. These findings are also consistent with our previous report of sex differences in associations between olfactory processing and cognition, in which better smell identification predicted higher intelligence, memory, and attention in females with schizophrenia, with opposite effects in males (Malaspina et al., 2012b). Likewise, we previously found that schizophrenia males self-report greater negative affect in response to pleasant odors compared to healthy males, with no such difference between females with schizophrenia and female controls (Walsh-Messinger et al., under review). In the latter study, negative symptoms were associated with more negative affect in response to pleasant odors and more positive affect in response to unpleasant odors in both males and females with schizophrenia. This is consistent with disturbed affective processing of odors in schizophrenia, suggesting impaired fronto-limbic circuitry. This pathology may particularly impair the neural regions employed by males for social processing.

In the present study, FSIQ was used to infer the contributions of cognitive functioning to ToM, while lower and higher order olfactory measures were used to infer more limbic centered processes. These contributions converge on the orbitofrontal frontal cortex (OFC), where both visual and verbal social information and olfactory information interact for higher order interpretations. Smell identification is processed in the amygdala and frontal prepiriform cortex, or olfactory cortex, which is located on the OFC (Eslinger, Damasio, & Van Hoesen, 1982) and responsible for monitoring social behavior (Beer, John, Scabini, & Knight, 2006). The key role of the OFC for smell identification was first demonstrated in patients with OFC lesions who exhibited specific deficits in smell identification with intact odor detection, showing the relative independence of these higher and lower order olfactory tasks (Potter & Butters, 1980).

In addition to being the first study to consider sex differences in the neural underpinnings of mentalizing in schizophrenia, this study has other strengths. The individuals with schizophrenia and healthy controls who participated in this study were rigorously diagnosed with structured interviews and had standardized ratings to assess the presence and severity of schizophrenia related symptomatology. All participants underwent full intelligence testing and assessment of odor detection threshold and smell identification. The absence of the expected mean differences in smell identification, odor threshold, or intelligence may relate to the ascertainment methods and demands of the study, the rigorousness of which may have excluded the interest of more impaired individuals with schizophrenia. The recruitment of healthy controls from the internet, rather than from university or medical center sites, may have also rendered the schizophrenia and control groups more similar than would be expected. However, the findings remained robust against any such ascertainment bias and the schizophrenia participants still demonstrated significant ToM deficits. These results demonstrate that sex stratified research is essential for studying social processes in general, and especially for sexually dimorphic conditions such as schizophrenia.