The ' Extreme Female Brain ' : Increased Cognitive Empathy as a Dimension of Psychopathology

Baron-Cohen's 'extreme male brain' theory postulates that autism involves exaggerated male-typical psychology, with reduced empathizing (considered here as social-emotional interest, motivation and abilities) and increased systemizing (non-social, physical-world and rule-based interest, motivation and abilities), in association with its male-biased sex ratio. The concept of an 'extreme female brain', involving some combination of increased empathizing and reduced systemizing, and its possible role in psychiatric conditions, has been considerably less well investigated. Female-biased sex ratios have been described in two conditions, depression and borderline personality disorder (BPD), that also show evidence of increases in aspects of empathy in some studies. We evaluated the hypothesis that BPD and depression can be conceptualized in the context of the 'extreme female brain' by: (1) describing previous conceptualizations of the extreme female brain model, (2) reviewing evidence of female-biased sex ratios in BPD and depression, (3) conducting meta-analyses of performance on the Reading the Mind in the Eyes test (RMET) among individuals with BPD, clinical or subclinical depression, and other psychiatric conditions involving altered social cognition and mood (schizophrenia, bipolar disorder, eating disorders, and autism), in relation to disorder sex ratios, and (4) evaluating previous evidence of increased empathic performance in these, and related, psychiatric conditions, and (5) synthesizing these lines of evidence into models for causes and effects of an 'extreme female brain'. Our primary empirical results are that RMET performance is enhanced in AC C EP TE D M AN U SC R IP T ACCEPTED MANUSCRIPT subclinical depression, preserved in borderline personality disorder, and reduced in other disorders (by meta-analyses), and that across disorders, more malebiased patient sex ratios are strongly associated with worse RMET performance of patients relative to controls. Our findings, in conjunction with previous work, suggest that increased cognitive empathizing mediates risk and expression of some psychiatric conditions with evidence of female biases, especially nonclinical depression and borderline personality disorder, in association with increased attention to social stimuli, higher levels of social and emotional sensitivity, negative emotion biases, and over-developed mentalist thought. These results link evolved human sex differences with psychiatric vulnerabilities and symptoms, and lead to specific suggestions for future work.


Introduction
Evolved sex differences in psychological and biological traits play important roles in the development, causes, and manifestations of many psychiatric conditions (Rutter et al., 2003). The 'extreme male brain' (EMB) theory of autism postulates that autism spectrum conditions reflect extreme manifestations of 'male-typical' psychology due in part to high prenatal testosterone (Baron-Cohen, 2002;Baron-Cohen et al., 2005). This theory can help to account for several notable features of autism, including its strongly male-biased sex ratio and reduced performance in measures of empathy and theory of mind (Baron-Cohen et al., 2011). Given the usefulness of the EMB theory in generating testable hypotheses and A C C E P T E D M A N U S C R I P T

ACCEPTED MANUSCRIPT
mechanistic cognition (which are non-social), are each basically synonymous for our purposes, and these two systems are subserved by different networks of distributed regions of the brain, with activation patterns that tend to be inversely related (Jack et al., 2012;Jack, 2014). Both sets of constructs can be quantified using ether self-report or task-based metrics of the relevant interacts and abilities. Given that higher fetal testosterone mediates both decreased social and empathic interest and abilities, and increased scores on metrics of systemizing (review in Baron-Cohen et al., 2011), and that their neural bases are inversely associated, empathizing and systemizing might be expected to be negatively correlated with one another, as found in some studies (e. g., Nettle, 2007b;Grove et al., 2013). However, the psychological and neurological bases of the relation of empathizing with systemizing remain to be studied in detail.
A primary sex difference in cognition, Baron-Cohen (2002, 2009 postulates, is represented by the balance of empathizing with systemizing, such that on average, males demonstrate a stronger drive to systemize, and females, on average, tend toward empathizing (Baron-Cohen, 2002;Baron-Cohen et al., 2005; Chakrabarti and Baron-Cohen, 2006;Wakabayashi et al., 2007). Under the E-S model, autism represents an extreme expression of male-typical cognition involving a strongly skewed profile of enhanced systemizing and reduced empathizing (extreme Type S in Baron-Cohen et al., 2005). This pattern of exaggerated psychological 'maleness' is consistent with the strong male bias in autism spectrum disorder prevalence (especially among individuals with lesssevere autism), as well as with evidence linking autistic phenotypes with elevated

A C C E P T E D M A N U S C R I P T
ACCEPTED MANUSCRIPT exposure to prenatal androgens, hormones that play important roles in 'masculinizing' the developing brain (Baron-Cohen, 2002;Lutchmaya et al., 2002aLutchmaya et al., , 2002bManning et al., 2001;Baron-Cohen et al., 2011;Wu and Shah, 2011).
Together, the E-S and EMB theories predict that a primarily-female proportion of the population will exhibit an E-S profile opposite to that observed in autism, one that is skewed toward increased empathizing and reduced systemizing (extreme Type E in Baron-Cohen et al., 2005). Consistent with this prediction, Goldenfeld et al. (2005) reported that, based on distributions of scores from the Empathy Quotient and Systemizing Quotient drawn from individuals with and without high-functioning autism, a small and all-female proportion of the sample (7%, and none with autism) exhibited this 'extreme female' profile of high empathizing and low systemizing. Similarly, data from Baron-Cohen et al. (2014) showed that 'extreme type E' individuals (those in the lowest 2.5th percentile for SQ and highest 2.5th percentile for EQ) demonstrated a strong female bias (with 59 of 60 individuals being female, from a total population of 2562 females and 1344 males;  2 = 28.9, P < 0.0001); similar results are also described in Wheelwright et al. (2006). However, the question of whether or not an EFB, defined in this manner, manifests in aspects or diagnoses of psychiatric conditions has been addressed by only a small number of previous studies.

Previous Accounts of the 'Extreme Female Brain'
A C C E P T E D M A N U S C R I P T ACCEPTED MANUSCRIPT Baron-Cohen (2002, 2012 suggested that increased empathizing drive and abilities need not negatively impact social functioning and engender psychiatric illness, and that low systemizing would be unlikely to cause impairment in psychological functioning. However, in these articles he did not explicitly consider extreme high levels of empathizing or combinations of high empathizing with low systemizing in the context of psychopathologies. A suite of researchers (Abu-Akel, 1999;Abu-Akel and Bailey, 2000;Dammam, 2003;Frith, 2004;O'Connor et al., 2007;Crespi and Badcock, 2008;Zahn-Waxler et al., 2008;Sharp and Venta, 2012;Dinsdale and Crespi, 2013;Sharp et al., 2013) has suggested causal roles for high levels of empathy or hyper-developed mentalistic cognition in psychotic-affective spectrum disorders, which include a set of psychiatric conditions (mainly schizophrenia, major depressive disorder, bipolar disorder and borderline personality disorder)(defined just below) that overlap substantially in their phenotypes, genetic bases, and correlates (Lieb et al., 2004;Perugi et al., 2011;Blackwood et al., 2007;Kendler, 2005;Barnow et al., 2010;Brosnan et al., 2010;Glaser et al., 2010;Crespi, 2011;Moritz et al., 2011;Slotema et al., 2012). Such overlaps, and the high comorbidities between these conditions, highlight their indistinct boundaries and partially-overlapping genetic and environmental causes. In this context, borderline personality disorder (defined below) was indeed originally conceptualized as a condition at the interface of depression with schizophrenia, given its high comorbidity (coincidence in diagnosis) with depression (e. g., Luca et al., 2012), and its notable incidence (e. g., on the order of 25-50% of patients) of psychotic A C C E P T E D M A N U S C R I P T
Borderline personality disorder is a psychiatric condition that involves unstable and intense personal relationships, disturbance of self-identity, feelings of emptiness, high rejection sensitivity and expectations of abandonment, selfdamaging behavior, impulsivity, anger, mood instability, and psychotic symptoms, that is highly comorbid with depression and bipolar disorder (Gunderson, 1984(Gunderson, , 2009Lieb et al., 2004;Crowell et al., 2009;Fonagy and Luyten, 2009;Barnow et al., 2010;Glaser et al., 2010;Leichsenring et al., 2011). Damman (2003)  Zahn-Waxler and colleagues have described a model whereby excessive levels of emotional empathy and social sensitivity, in conjunction with dysfunctional family environments and high levels of internalizing emotions (e. g., M A N U S C R I P T

ACCEPTED MANUSCRIPT
guilt, shame, sadness and embarrassment), may increase liability to depression and anxiety, primarily among adolescent and young-adult females (Zahn-Waxler et al., 2006, 2008O'Connor et al., 2002O'Connor et al., , 2007Zahn-Waxler and Van Hulle, 2012 person's mental states that go so far beyond observable data that the average observer will struggle to see how they are justified'. Sharp and Venta (2012) summarize evidence that hyper-mentalizing is characteristic of individuals (especially females) with borderline personality disorder and borderline features, and that it may centrally involve difficulties in emotion regulation especially in social contexts. They note that despite such hyper-mentalizing, BPD has been associated in multiple studies, predominantly of females, with superior mentalizing abilities compared to controls for some tasks (reviewed in Dinsdale A C C E P T E D M A N U S C R I P T ACCEPTED MANUSCRIPT and Crespi, 2013;Mitchell et al., 2014). Such enhanced abilities may be related to high levels of attention and vigilance in social situations (Sieswerda et al., 2007;Domes et al., 2009;Frick et al., 2012), increased reliance on, and experience with, implicit, automatic and non-conscious social cognition (Fonagy and Luyten, 2009;Sharp and Venta, 2012;Sharp et al., 2013), and increased expression in females of psychological defense mechanisms in social contexts (del Giudice 2014), but such mechanisms have yet to be analyzed directly.
Hyper-mentalizing in disorders such as BPD can be consistent with increased accuracy in mentalization tasks, compared to controls, to the extent that errors are reduced on average but tend to involve over-interpretation of social stimuli (Sharp et al. 2011).
Taken together, these conceptualizations of the EFB, and sex differences in social psychopathology, all suggest that it may involve psychological dysfunctions typical of the psychotic-affective spectrum, and high aspects of empathizing or high empathizing relative to systemizing. In particular, the studies considered above suggest that BPD and depression represent among the strongest psychiatric candidates for consideration as 'extreme female brain' disorders, closely associated with one another, but they also represent the only two psychiatric conditions that have been linked, in replicated studies, with enhanced empathic expertise.
The considerations described above have motivated our primary hypothesis for conceptualization of the EFB as evaluated here: that it centrally involves empathic abilities that are enhanced, in the context of increases in sociallyfocused cognition and affect, but that such enhancements engender increased risk for depression and borderline personality disorder and their subclinical manifestations. By this hypothesis, females are thus differentially strongly affected by these two psychiatric conditions, mainly due to their increased social and empathic interests and abilities compared to males. Social and empathyrelated cognitive-affective foci are thus increased to extremes in BPD and

ACCEPTED MANUSCRIPT
depression, with enhanced performance in some social-empathic contexts (compared to neurotypical females) but social dysfunction overall, due to hypermentalizing and cognitive-affective biases. Moreover, by our hypothesis there should also be an association, among psychiatric conditions involving social cognition, between disorder sex ratios and empathic abilities, with more femalebiased disorders showing relatively increased, preserved, or less-reduced empathic skills in patients compared to controls.
Our hypothesis thus predicts: (1) female-biased sex ratios in depression and borderline personality disorder; (2) evidence of empathic enhancements in borderline personality disorder and depression (clinical or subclinical), relative to controls (or at least preserved performance compared to other disorders); and (3) a negative correlation between empathic skills (in patients relative to controls) and disorder sex ratio (percent males), such that disorders with more femalebiased sex ratios involve enhancement, or less reduction, in such abilities.

Sex Ratios in Borderline Personality Disorder and Depression
Baron-Cohen's extreme male brain theory for autism was inspired, in part, by the strong male biases found among individuals with autism spectrum disorders, which suggest that males are predisposed to autism as a consequence of how they differ, psychologically, from females ( Baron-Cohen et al., 2011). In parallel M A N U S C R I P T

ACCEPTED MANUSCRIPT
to this reasoning, an extreme female brain theory for psychiatric conditions should, as noted above, apply most directly to disorders that show female biases in their prevalence.

ACCEPTED MANUSCRIPT
The strongest correlates of heterogeneity appear to be variation between females and males in symptom profiles, ascertainment biases due to sex differences in help-seeking, and demographic differences among the populations analyzed (e. g., Oliver et al. relationships, whereas males scored higher than females for impulsiveness.
Given evidence for stronger female biases among clinical than non-clinical populations for BPD diagnoses and symptoms, the borderline traits reported more commonly by females across studies (including fear of abandonment, affective instability, intense and unstable relationships, paranoia, and chronic feelings of emptiness) may engender help-seeking at lower thresholds than those traits more commonly reported by males (i. e., impulsivity and anger), and thus lead to the stronger female biases observed in most clinical settings (Widiger and Trull, 1993;Skodol and Bender, 2003;Oliver et al., 2005). Symptom profiles are especially important for inferences regarding sex ratios because to meet DSM criteria for a BPD diagnosis, an individual must exhibit any combination of five (or more) symptoms from a nine-item list, thus creating a situation where individuals with substantially different sets of symptoms (with 256 combinations in total, and overlap in as few as one criterion) may receive the same diagnostic label of BPD (Gunderson, 2010). Unbiased sex ratios in some samples assessing fit to BPD diagnoses thus appear to reflect some combination of symptom profile differences between males and females, methodological issues, and differences between the sexes in thresholds for seeking psychological help (Skodol and Bender, 2003;Oliver et al., 2005;Sansone and Widerman 2014). With regard to the EFB hypothesis, it is the symptom profiles that are more characteristic of females that should be associated with increased empathizing; this hypothesis has yet to be directly addressed but it is consistent with the symptom profile differences described above, and with the female biases found in all of the populations of individuals with BPD who were analyzed for RMET performance, as described below.
The evidence for strong female biases in depression is extensive and highly consistent across studies, as described extensively in previous work. Thus, two recent meta-analyses (Luppa et al., 2012;Ferrari et al., 2013), and systematic or narrative reviews (e. g., Kessler, 2003;Kuehner, 2003)

Borderline Personality Disorder, Depression, and the Extreme Female Brain
The findings described above for female biases in depression and BPD suggest that sufficient evidence exists to merit focused investigation of the degree to which risk and expression of these psychiatric conditions may be modulated, in part, by high empathizing or some combination of high empathizing A C C E P T E D M A N U S C R I P T  , Hall, 1978;Hall and Matsumoto, 2004). It is important to note, however, that the RMET measures only one aspect of empathy (cognitive empathy, in the context of visually-based theory of mind abilities); 'emotional' empathy, which includes for example emotional resonance and empathic concern, is not addressed.
In the next section of this paper, we systematically evaluate the EFB model with regard to performance on the RMET, by analyzing RMET performance across a suite of psychiatric disorders, of which BPD and depression are predicted to best fit an EFB model. In accordance with Baron-Cohen's conceptualization of empathizing and systemizing as mediating psychopathologies that show strong sex biases, we predict that individuals with EFB-associated psychiatric disorders should exhibit enhanced performance on the RMET or less-impaired performance than observed in other psychiatric disorders (especially autism). Moreover, under the EFB and EMB models, disorders demonstrating enhancements, or less impairment, on the RMET should A C C E P T E D M A N U S C R I P T

ACCEPTED MANUSCRIPT
exhibit relative female biases in their prevalence.

Reading the Mind in the Eyes in BPD, Depression and other Disorders
To conduct meta-analyses of RMET performance among different psychiatric conditions, we accessed Web of Science to compile a list of all publications that   (Figure 3). Note that this effect is not due to sex differences between patient and control groups, which were almost always matched in each study.
These findings indicate that overall, more male-biased disorders tend to involve larger reductions in cognitive empathic abilities, as measured by the RMET, and that more female-biased disorders (here, BPD and non-clinical depression in particular) tend to involve relatively spared, or enhanced, empathic abilities by this test. Given that RMET performance is also higher among neurotypical females than neurotypical males (meta-analysis in Kirkland et al., 2013), these data are consistent with the general prediction of the EFB model that such empathic abilities should be highest among females with the relevant disorder, next highest among neurotypical females, and lower still among neurotypical males; in the same way, under the EMB model, males with autism outperform neurotypical males, who outperform neurotypical females, for some tests that show well-established higher performance among males ( Baron-Cohen et al., 2011). It is important to note, however, with regard to this analysis of RMET performance, that almost none of the studies presented their results separately for females and males (when both sexes were tested); more-robust tests of this EFB prediction will require analyses that take account of sex.
Moreover, more male-biased disorders, especially autism and schizophrenia, are

ACCEPTED MANUSCRIPT
generally considered to be relatively severe (compared to the other disorders analyzed here) as regards overall cognitive deficits, which may have influenced RMET performance (although major depression, which is strongly female biased, can also be regarded as comparably severe in this general regard). Finally, we note that although anorexia exhibits strong female biases, it also involves significantly reduced performance on the RMET (Figure 3); these findings suggest that this condition differs from borderline personality and subclinical depression in important ways related to theory of mind and empathic abilities, such that mentalizing may tend to be reduced rather than hyper-developed (e. g., (2014) also noted that individuals with BPD were as accurate or better than controls across a diversity of complex social judgement tasks, although there was notable heterogeneity in performance across tasks and sample populations.

Schulte
These findings indicate that, of all clinically-defined psychiatric and personality disorders, BPD represents the only condition to show evidence across many, M A N U S C R I P T

ACCEPTED MANUSCRIPT
diverse studies of empathic enhancements over control individuals, in a parallel way to which autism is the only disorder showing diverse evidence for enhancements over controls in visual-spatial skills and systemizing (e. g., Caron et al., 2006;Baron-Cohen et al., 2011;Mottron et al., 2013). Despite such extensive evidence of empathic enhancements in BPD, it is also important to note that a substantial proportion of studies demonstrate empathic deficits in this disorder, and that the causes of such variation in findings remain largely unclear.
Moreover, although BPD did not involve enhanced RMET performance by the meta-analysis conducted here, the number of studies involved was small, which, especially given the high symptom heterogeneity of BPD, suggests that this analysis was under-powered. Additional studies of BPD are thus required for more robust evaluation of this hypothesis in relation to alternatives.
In our analysis, evidence for enhanced empathic (RMET) skills in association with depression was limited to individuals with sub-clinical manifestations and individuals with a maternal history of depression. Similarly, Lane and DePaulo (1999) found that individuals with sub-clinical depressive symptoms (dysphoria) were better than non-dysphoric individuals at detecting social deception in two experimental paradigms, and Nettle and Liddle (2008) found that higher neuroticism predicted better performance on a test of cognitive empathy involving social story interpretation. Given that BPD, as a personality disorder, may be more similar to sub-clinical rather than major depression in the severity of its negative effects on general cognitive functioning, these findings suggest that empathic enhancements are found primarily among individuals with relatively

ACCEPTED MANUSCRIPT
less-severe manifestations of psychotic-affective conditions, where sex ratios also appear to be relatively female biased, as also discussed above (see also Bebbington, 1991). On the autism spectrum, sex ratios are more (male) biased towards the higher-functioning end of the spectrum, but the degree to which enhancements on tasks associated with visual-spatial skills and systemizing are differentially found among males (compared to females) with autism, or higherfunctioning males with autism, remains unknown (Caron et al., 2004(Caron et al., , 2006Manjaly et al., 2007;Baron-Cohen et al., 2011). Finally, the strong evidence of RMET enhancement in subclinical depression, but the non-significant evidence for BPD, suggests that subclinical depression, especially among females, may represent our best current model for EFB effects, at least in the domain of cognitive empathy and its correlates.
Despite findings that show enhancements in some aspects of empathy among individuals with subclinical depression, empathic skills, as indexed by the RMET, were clearly reduced among individuals with major depression. These results are concordant with findings from large sets of studies that show depressionassociated deficits in cognitive and empathic functioning, in conjunction with biased attention and enhanced attention to, and recognition of, negativelyvalenced emotional stimuli (e. g., Inoue et al., 2006;Leppänen, 2006;Cusi et al., 2011;Schreiter et al., 2013). In contrast to low mood and mild depression, major depression, like many other severe psychiatric conditions, is associated with highly impaired executive function, reduced motivation and attention, and working memory deficits (Burt et al., 1995;Fossati et al., 2002; Hasler et al., 2004;Galecki et al. 2015). Empathizing performance in depression (as well as relatively-severe cases of BPD) is expected to be confounded by the presence of these deficits (Lee et al., 2005); for example, Zobel et al. (2010) found that chronic depression predicted poor theory of mind performance on a cartoonpicture story task, but after controlling for logical memory and working memory, depressive status no longer predicted theory of mind ability. Studies that evaluate aspects of empathy along a continuum from very low to high levels of depression are suggested as especially useful for future work, by our results.

ACCEPTED MANUSCRIPT
Similar considerations regarding general deficits (e. g., Hay and Sachdev, 2011) may apply to the reductions in RMET performance found among individuals (mainly females) with eating disorders in our meta-analyses, although this hypothesis remains conjectural.
Considered together, these findings provide evidence that for the RMET and other measures of empathic abilities, higher performance by females compared to males tends to parallel higher or preserved performance by individuals with BPD or subclinical depression, compared to control individuals (with strong female biases among both such groups). These findings support a central prediction of the EFB model, at least with regard to its empathizing dimension.
Further evaluation of the model require additional tests, especially involving relatively realistic and socially-interactive empathy-related paradigms (Dinsdale and Crespi, 2013), individuals with subclinical depression or high neuroticism (e. g., Pasquier and Pedinielli, 2010) and individuals with varying degrees of severity and different symptom profiles for BPD.

Depression and Other Conditions
In autism, a combination of enhanced perceptual, visual-spatial abilities, and systemizing, with reduced empathizing, are considered to jointly contribute to dysfunction in social interactions. By contrast, this review, and the previous work For BPD, some early studies attributed enhanced empathy to dysfunctional parenting, whereupon children develop enhanced sensitivities to cues indicating their caregivers' mental states especially with regard to potential rejection or abuse (Krohn, 1974;Carter and Rinsley, 1977;Frank and Hoffman, 1986;Park et al., 1992;Linehan, 1993). By this hypothesis, the social difficulties characteristic  et al., 2009;Dyck et al., 2008;Fertuck et al., 2009;Fonagy and Luyten, 2009). In the context of RMET performance, mentalization would

ACCEPTED MANUSCRIPT
presumably be controlled, such that empathic enhancements could be expressed in the relative absence of social-emotional dysregulation. Figure  Under the hypotheses discussed here, preserved or enhanced cognitive empathy and empathic skills, in both BPD and subclinical depression, thus develop primarily from biological underpinnings and social circumstances that engender increased attention and sensitivity to interpersonal cues, with social dysfunction and more-severe disorder following from some combination of increased social attention and sensitivity with emotional dysregulation, negatively-biased interpretations of social stimuli due to maladaptive developmental schema, hyper-mentalization, and increasing perceived and actual intensity of social problems (Zahn-Waxler et al., 2006, 2008O'Connor et al., 2007;Fonagy and Luyten, 2009;Sharp et al., 2011Sharp et al., , 2013Zahn-Waxler and A C C E P T E D M A N U S C R I P T ACCEPTED MANUSCRIPT Van Hulle, 2012;Mitchell et al., 2014;Unoka et al., 2015). This model for BPD and depression is consistent with their high comorbidity (Lieb et al., 2004;Zimmerman and Mattia, 1999), and findings that depressive symptoms in BPD, or combined BPD and depression, are positively associated with mental state discrimination abilities in some studies, especially for negative emotions (Fertuck et al., 2009;Mitchell et al., 2014;Unoka et al., 2015). Both enhancements and dysfunction may also be related to higher intrinsic social and emotional sensitivities due to effects from differential sensitivity ( in social cognition more generally, appear to represent the strongest contrasts of BPD and depression with autism, and the clearest current evidence regarding the EFB hypothesis in comparison with the extreme male brain. These considerations also suggest that quantification of social-relational focus, attention, and motivation, rather than just empathizing per se, may be a more direct route to understanding the relationship between gender differences and psychopathologies linked with the EMB and EFB.

Conclusions, Limitations and Implications
We have provided here the first comprehensive theoretical and empirical framework for analysis and understanding of the EFB. Our main conclusion is that a notable body of evidence supports the hypothesis that the EFB model may be applicable to subclinical depression and BPD in particular, and psychoticaffective conditions more generally, just as the EMB model is applicable to autism spectrum conditions. The primary lines of evidence relevant to this inference include: (1)  The primary limitation of the analytic aspect of our study is its restriction to one test of cognitive empathy, the RMET, which is due to this being the only such test that has been performed across enough disorders and conditions for meaningful, synthetic tests to be performed. Moreover, numbers of studies are small for some disorders (including BPD, subclinical depression, and depression), which reduces statistical power and increases effects from possible heterogeneity among studies in methodology. Future studies should also consider the roles of increased affective empathy in female-biased psychopathology (Zahn-Waxler et al., 2006, 2008, especially in the context of how higher social sensitivity among females (in contrast to reduced social sensitivity in autism, and in males) may contribute to risks of depression and BPD.
Our analysis of theory and evidence regarding the EFB has several implications for clinical work, and for specific future studies that would provide further tests of predictions regarding EFB models and conceptualizations. First, Sharp et al. (2013) found, with regard to borderline personality traits in adolescents, that hyper-mentalizing, but not other measures of social-cognitive reasoning, exhibited malleability through mentalization-based therapies. This finding suggests that dysfunctionally-increased empathy-related psychological A C C E P T E D M A N U S C R I P T

ACCEPTED MANUSCRIPT
phenotypes may provide more-effective targets for therapy than social-emotional or social-cognitive 'deficits' that are usually considered to involve reductions of trait expression. Second, given that aspects of female gender 'protect' females from autistic impairments (Robinson et al., 2013), some aspects of male gender may also tend to 'protect' males from depression and BPD; determining the nature of such protective factors may provide clues for improving preventatives and therapies. McHenry et al. (2014) suggested, for example, that 'testosterone may have protective benefits against anxiety and depression'. In this context, our hypothesis also predicts that depression and BPD should be associated with low prenatal testosterone (or low prenatal testosterone relative to prenatal estrogen), in direct contrast the relatively-high prenatal testosterone associated to some A C C E P T E D M A N U S C R I P T

Supplementary Materials
Supplementary information for this article can be found online. M A N U S C R I P T

FIgure Legends
Fig. 1. By the empathizing-systemizing model extended to understanding extreme female brain phenotypes, dysfunctionally-high empathizing and low systemizing are expected to be associated with borderline personality disorder, depression, and other manifestations of psychotic-affective conditions. As such, these conditions can be considered to represent psychological 'opposites' to autism spectrum conditions, with regard to empathizing and systemizing.    Figure   4) and Zahn-Waxler and Hulle (2012, Figure 25.2), also with the addition of explicit hypothesized effects from empathizing and systemizing. The model of borderline personality disorder is based primarily on work by Fonagy and colleagues (e. g., Fonagy and Luyten, 2009;Sharp et al., 2011Sharp et al., , 2013, with the addition of explicit hypothesized effects from empathizing and systemizing as conceptualized by Baron-Cohen (2009). The potential roles of low systemizing under both models remain unclear.