Individual Differences in Mentalizing Tendencies

Mentalizing is an umbrella term used to refer to a wide range of socio-cognitive processes (Frith & Frith, 2006). To illustrate this breadth, the ability to represent the visual perspective, intentions, knowledge, and emotions of others have, at times, all fallen under the broader category of mentalizing (Gilead & Ochsner, 2021). While these capacities arguably share a common denominator in that they all require the ability to represent “invisible” mental states in causal terms (e.g., seeing, feeling, intending, believing, and knowing, see Penn et al., 2008), they also differ in important ways. For instance, while one’s ability to represent the visual perspective of a co-present other may strongly relate to one’s more general visuospatial representational abilities (Cardillo et al., 2020), the ability to represent the knowledge of other individuals may relate more strongly to one’s abstract reasoning and problem-solving skills (Astington et al., 2002).

It is therefore apparent that a comprehensive account of mentalizing must grapple with the ways in which an individual’s general cognitive capacities relate to and support one’s mentalizing capacities. While elucidating these relationships is necessary for a basic understanding of the cognitive foundations of mentalizing, insights regarding these relationships may shed light on ways in which individuals come to mentalize in meaningfully different ways as well. Indeed, due to the subtle, dynamic, and demanding challenge associated with navigating one’s social environment, it is likely that individuals will vary considerably in the mentalizing orientations they adopt.

The notion that individuals mentalize about their social environment in meaningfully different ways raises the possibility that these differences might relate to broader dimensions of daily function as well. For instance, individuals who tend to ruminate about what others are thinking of them might also be likely to experience social anxiety (Brown & Stopa, 2007, though see Ballespí, Vives, Sharp, et al., 2019). Alternatively, while individuals who approach mentalizing inferences with an overestimated sense of direct access to the inner workings of other minds may experience social interactions with greater ease, they may ultimately fall short of capturing an adequately complex representation of the target of their mentalizing efforts, potentially leading to misunderstanding or discord. More generally, different mentalistic tendencies may play a role in a wide range of outcomes pertaining to general cognitive functioning, personal relationships, and general well-being.

Prior investigations of individual differences in mentalizing have explored several dimensions upon which individuals are argued to vary in meaningful ways. In the present study, we explore how variability in four such dimensions: 1) egocentrism/altercentrism; 2) accuracy/inaccuracy; 3) over-mentalizing/under-mentalizing; 4) certainty/uncertainty relate to individual differences in general fluid intelligence and a range of self-reported measures. We have focused on these dimensions due to a growing body of literature indicating that individual variability along these dimensions may significantly relate to instances of psychopathology (see De Meulemeester et al., 2020; Drayton et al., 2018; Kronbichler et al., 2019; Luyten et al., 2020). We will now provide a brief overview of the literature exploring these dimensions.

One characterization of the mentalizing process has been that humans are “reflexively mindblind” (Lin et al., 2010) and need to deliberately engage in inferences about other perspectives in order to overcome an automatic egocentric anchor (Epley et al., 2004). In contrast to this approach, others (e.g. Apperly, 2018; Brown-Schmidt & Hanna, 2011) have argued that, rather than acting as a fundamental bias or default anchor, egocentrism is but one of several constraints that are at play when navigating a social encounter. For example, Samson et al. (2010) found that, under certain conditions, individuals may automatically represent the visual perspective of another agent, even when unprompted—a phenomenon they termed altercentric interference. On this view, just as individuals may experience some difficulty in inhibiting their own perspective when making inferences regarding others (“egocentric interference”), individuals may also experience some difficulty inhibiting others’ perspective when considering their own (“altercentric interference”).

While initial exploration of egocentric and altercentric interference demonstrated their expression at an aggregated level, subsequent research has argued that individuals may vary in the extent to which they experience these forms of interference when mentalizing. Furthermore, these differences are argued to relate to broader individual differences in meaningful ways. For example, Wu & Keysar (2007) found that Mandarin-speaking university students who had recently moved to Chicago from mainland China were able to overcome egocentric interference more efficiently than a non-Asian Native-English speaking comparison group. The authors suggested that an interdependent cultural context might have fostered less egocentrism in the Mandarin-speaking group thus boosting their ability to overcome their egocentric interference.

In the case of altercentric interference, Nielsen et al. (2015) found that self-reported empathic concern and perspective-taking on the interpersonal reactivity index (IRI, Davis, 1983) were positively correlated with altercentric interference (though see Mattan et al., 2016). More recently, working with a sample of incarcerated men, Drayton et al. (2018) found that offenders with more psychopathic traits demonstrated less altercentric interference in comparison to offenders with less psychopathic traits. A potential implication of these studies is that individuals process their visual environment in meaningfully different ways which relate to broader patterns of interpersonal function in the world. In the case of psychopathy, this suggests that certain people, even at a basic level of observing a visual scene, process the environment with less concern for the perspective of others. Similarly, it is possible that individuals who are highly empathetic may process their visual environment in a way that is more altercentrically-biased than less empathetic individuals.

Notably, the methods used to assess egocentric and altercentric interference exert considerable cognitive demands. Specifically, interference scores index the ease with which participants successfully represent a given perspective (either of their own or of another person/avatar) in the presence of a competing and incongruent alternative. The ease with which one completes these tasks successfully may therefore reflect a more general ability to rapidly modify one’s behaviors in relation to a changing set of rules. Consequently, individual differences in egocentric and altercentric interference may simply reflect broader differences in abilities such as working memory span, task maintenance and processing speed. Important outstanding questions are therefore the extent to which egocentric and altercentric interference capture distinctively social phenomena and whether these forms of interference capture two distinct biases as opposed to a general singular processing difficulty. To illustrate, in the case of Wu & Keysar (Wu & Keysar, 2007, see also Wu et al., 2013) discussed above, it is possible that the Asian students simply had better general processing skills (working memory, task maintenance, and pattern detection) than the Native-English-speaking comparison group. A comprehensive understanding of the egocentric and altercentric interference phenomena must therefore address how these constructs relate to general cognitive abilities as well.

Perhaps the most central dimension used to investigate individual differences in mentalizing has been the construct of accuracy. In the case of the “Reading the Mind in the Eyes Test” (RMET; Baron-Cohen et al., 2001), individuals are scored on their ability to label a series of decontextualized facial emotional expressions. Similarly, in the case of the Movie for the Assessment of Social Cognition (MASC; Dziobek et al., 2006), participants observing a scripted interaction between actors are asked to select one of four interpretations of the actor’s behaviors, one of which was pre-determined to reflect inferred mental state “accuracy”. Using these methods, accuracy scores have been found to be significantly lower in certain clinical populations such as Autism Spectrum Disorder (ASD) but not in others (e.g., Narcissistic Personality Disorder, see Ritter et al., 2011).

While several studies have explored the relationship between general intelligence and RMET performance (e.g., Baker et al., 2014), less is known regarding the relationship between accuracy scores on the MASC and general cognitive abilities. Since accurate mental state inference involves flexible and dynamic integration of complex information, a greater understanding of the relationship between the inferred mental state accuracy construct and general cognitive abilities is warranted. Furthermore, the relationship between inferred mental state accuracy and broader personality dimensions is also not well understood. Since inferred mental state accuracy is presumably an important skill, understanding how variability in this dimension relates to broader aspects of the self is also of interest.

Another dimension used to explore individual differences in mentalizing revolves around the constructs of “over/hypermentalizing” and “under/hypomentalizing” (e.g., Sharp et al., 2011). In this framework, certain individuals are said to generate long and overly detailed accounts that have little or no relationship to observable (testable) reality (“over-mentalizing”). Conversely, other individuals tend to have difficulty constructing complex models of their own mind (or that of others) and thus exhibit “insufficient” mental state inference (“under-mentalizing”).

In extreme cases, specific mentalizing tendencies (e.g., over-mentalizing or under-mentalizing) are argued to characterize the socio-cognitive deficits associated with specific disorders (Crespi & Badcock, 2008). Furthermore, even within a diagnosis, meaningful differences in specific mentalizing tendencies may relate to significant real-life differential outcomes (see for example, Engelstad et al., 2019) On the other hand, there is currently mixed evidence regarding the notion that certain conditions are neatly associated with a particular type of mentalizing style (see Langdon & Brock, 2008). For example, in a recent meta-analytic review, McLaren et al. (2022) found that over-mentalizing was associated with a wide range of psychopathology symptoms, calling into question the notion that over-mentalizing is specifically tied to specific personality disorders such as borderline personality disorder.

An important outstanding question is therefore the extent to which under-mentalizing and over-mentalizing represent distinctly different patterns. An alternative account might suggest that a general difficulty may underlie both over-mentalizing and under-mentalizing response patterns on the MASC. Finally, to the extent that over-mentalizing/under-mentalizing represent distinct tendencies, less is known regarding how these tendencies relate to broader cognitive and personality dimensions.

Fonagy et al. (2016), have argued that individuals vary considerably in the extent to which their mentalizing processes are characterized by high (or low) degrees of certainty about mental states. Furthermore, these differences are argued to relate to various forms of psychopathology. In this framework, excessive certainty (assessed by strong agreement with statements such as “I can tell how someone is feeling by looking in their eyes”) reflects a failure to recognize the opacity of mental states and excessive uncertainty (assessed by strong agreement with statements such as “I don’t always know why I do what I do”) reflects difficulties with constructing adequately complex mental state inferences. In comparison to a non-clinical control sample, a clinical sample of outpatient individuals with Borderline Personality Disorder or Eating Disorder had higher excessive uncertainty about mental states scores on the RFQ (Fonagy et al., 2016). While high certainty scores were related to certain aspects of psychopathology as well, this pattern was far less consistent. These findings suggest that the certainty/uncertainty dimension (particularly the uncertainty pole) may play a role in the socio-cognitive challenges experienced by individuals with a range of clinical diagnoses.

One set of unexplored issues related to the certainty/uncertainty dimension is the extent to which variability along this dimension relates to personality dimensions such as agentic orientation (Baryla et al., 2019) and personal sense of power (Anderson et al., 2012). Specifically, it is not clear how certainty/uncertainty about mental states relates to general personality dimensions that might bear on the extent to which one feels certain about things such as mental states. Elucidating these relationships further may shed light on the documented relationships between excessive uncertainty about mental states and various instances of psychopathology.

Most of the dimensions discussed in this section (accuracy/inaccuracy, over/under, certainty/uncertainty) operate under the assumption that an objective and appropriate standard of mental state inference exists. Given the documentation of considerable cultural differences in the conceptualization of desirable forms of mentalizing (Duranti, 2015), we do not imply that these classifications reflect absolute judgements regarding optimal mentalizing function. Furthermore, specific mentalizing tendencies might be adaptive or desirable in one context but not the other. Notably, while a degree of accuracy is undoubtedly an important aspect of mentalizing, the operationalization of the accuracy construct is far from straightforward. Specifically, mentalizing involves a process of constructive interpretation both on the part of the individual engaging in mentalizing (“what am I feeling right now”) and the mentalizing target (“what is this person feeling right now”). The notion that mentalizing simply involves “accurately” inferring the content of another’s mind (or one’s own mind for that matter) thus potentially overlooks the constructive nature of mentalizing and subjective emotional experience more generally (see Burner, 1997).

Despite these limitations, we see value in exploring the extent to which the existing dimensions used to assess individual differences in mentalizing tendencies relate to one another and to broader aspects of the self. The preregistration for this study is available here: https://osf.io/srd45. In this study, we examine:

1. If mentalizing tendencies, such as egocentrism and altercentrism, are related across tasks.

2. The relationship between performance on fluid intelligence tasks and performance on mentalizing tasks.

3. If specific clusters of self-reported psychopathology symptoms are related to specific mentalizing tendencies.

4. If individuals with higher levels of self-reported traits associated with psychopathy have lower levels of altercentrism on mentalizing behavioral tasks.

5. How individual differences in sense of interpersonal power, agency and social interaction anxiety are associated with altercentrism in mentalizing behavioral tasks.

We finally explore the relationship between various self-report measures and performance on mentalizing tasks. We selected self-report measures based on our own suspicion that variability in these measures may be conceptually tied to the mentalizing dimensions explored in this project. For example, we suspected that a tendency to anthropomorphize might relate to a tendency to over-mentalize and that a tendency to experience empathic concern might relate to a tendency to experience altercentric interference.

In this project, we aim to address these questions by measuring performance on a comprehensive battery of tasks and questionnaires tapping mentalizing, fluid intelligence, emotion regulation, self-reported psychopathology symptoms and various personality dimensions. We thus adopt a comprehensive individual-differences approach towards understanding mentalizing tendencies. We hope this approach will shed light on the cognitive foundations underlying different mentalizing tendencies and provide a basis for future research aimed at alleviating mentalizing-related challenges faced by individuals in both clinical and non-clinical settings.

We tested 161 participants who were recruited through the online platform Prolific (www.prolific.co). Participants had to be at least 18 years of age, be fluent in English, based in the US, have a minimum Prolific approval rating of 90%, and not have participated in previous versions of the study (including pilot versions). 11 participants were excluded for poor performance or failing attention checks, resulting in a final sample of 150 participants (77 men, 73 women, mean age = 34.15 years, sd = 11.45 years). Participants were compensated $15 for completing the study. All participants gave informed consent, and ethical approval was obtained from the Institutional Review Board at the University of Oregon.

Tasks were implemented and hosted on Gorilla (Gorilla.sc, Anwyl-Irvine et al., 2020). Participants completed a battery of behavioral tasks and questionnaires. The study took approximately two hours to complete, and participants were given three and a half hours to complete the measures. Participants were required to use Chrome or Firefox as their browser and complete the study on a computer or tablet. The order of presentation was fixed across all participants as follows:

Movie for the Assessment of Social Cognition (MASC; Dziobek et al., 2006). The MASC is a real-time, video-based assessment of social cognition which measures accurate mentalizing and dysfunctions in mentalizing including no mentalizing, over-mentalizing, and under-mentalizing. Participants’ scores were computed by summing the amount of each answer type yielding four scores for each participant, corresponding to each mentalizing answer type (accurate, over-mentalizing, under-mentalizing, non-mentalizing).

6/9 Task(Quesque et al., 2018). Participants in this task are presented with an image of a person sitting at a desk containing 15 different symbols. At the center of these symbols was a digit that appears as a “6” from the vantage point of the participant but as a “9” to the person sitting at the desk. Participants were then asked to report the identity of that central number. “6” responses were coded as egocentric while “9” responses were coded as altercentric.

Dot Task(Samson et al., 2010). The dot task (which has been referred to as the “avatar” visual perspective taking task in previous work) requires participants to report their own visual perspective in a scenario where a co-present agent either shares their perspective or has a different visual perspective. Conversely, the task may also request that participants report the visual perspective of the agent rather than their own perspective. We adopted the stimuli used in Ferguson, Brunsdon & Bradford (2018). The difference (in RT and number of errors) between self-congruent trials (in which participants must report their own visual perspective which is congruent with that of the avatar) and self-incongruent trials (in which participants must report their own visual perspective which is incongruent with that of the avatar) are a participant’s altercentric interference scores. Conversely, the difference between other-congruent trials (in which participants must report the visual perspective of the avatar which is congruent with their own visual perspective) and other-incongruent trials (in which participants must report the visual perspective of the avatar which is incongruent with their own visual perspective) are a participant’s egocentric interference scores.

Ambiguous Reference Task(Keysar, 1994). In this task, participants read a short series of vignettes and are asked to interpret how a friend of the speaker might perceive them. The messages may seem sincere, but privileged background knowledge about the speaker’s intentions suggests a sarcastic interpretation. After reading the vignettes, participants rate the extent to which they believe the friend of the speaker would take the comment as sarcastic on a 6-point Likert scale from 1 (definitely not) to 6 (definitely yes). Higher scores indicate higher levels of egocentrism.

Interpersonal Reactivity Index (IRI; Davis, 1983). The IRI is a 28-item measurement tool for the multi-dimensional assessment of empathy. The IRI has 4 subscales (perspective-taking, fantasy, empathic concern and personal distress), each made up of seven different items. Responses are in the form of a 5-point Likert scale from 1 (Does not describe me very well) to 5 (Describes me very well).

Director Task (computerized version) (adapted from Dumontheil et al., 2012). In this task, the participant follows the instructions of a pre-programmed virtual “agent” to move around various objects in a vertical grid of squares. The agent is located on the other side of the grid and cannot see all of the objects, because some of the cells are occluded on their side. Crucially, the agent is supposed to be ignorant of the contents of those cells, and when, for example, they ask the participant to “move the small candle,” the smallest of three candles is visible only to the participant.

Individual Differences in Anthropomorphization Questionnaire (IDAQ; Waytz et al., 2010). The IDAQ is a 30-item self-report measure of anthropomorphism, the tendency to attribute human mind-like properties to an agent. Respondents rate the extent to which they believe a non-human entity (e.g., a machine) has human characteristics (e.g., intentions, emotions, free will) on a scale from 0 (not at all) to 10 (very much). Items on the IDAQ represent three groups of commonly anthropomorphized entities, including nonhuman animals, natural entities, and technological devices.

Social Interaction Anxiety Scale (SIAS; Safren et al., 1998). The SIAS is a 20-item questionnaire designed to measure distress in social situations and interactions.

Personality Assessment Inventory-Borderline (PAI-BOR; Stein et al., 2007). The PAI-BOR is a 24-item questionnaire containing four subscales related to the core features of BPD. The four subscales are affective instability, identity problems, negative relationships, and self-harm/impulsivity (not limited to self-harm behaviors).

Difficulties with Emotion Regulation Scale – Short Form (DERS-SF; Kaufman et al., 2016). The DERS-SF incorporates 6 subscales in the measurement of emotion regulation, including 1) awareness and understanding of emotional responses, 2) acceptance of emotions, 3) the ability to control impulsive behaviors when experiencing negative emotions, 4) the ability to employ situationally appropriate emotion regulation strategies to meet one’s goals, 5) the ability to engage in goal-directed behavior while experiencing negative emotions, and 6) the extent to which one is clear about which emotions one is experiencing. Responses are in the form of a 5-point Likert scale from 1 (“Almost never”) to 5 (“Almost always”).

Systemizing Quotient (SQ; Baron-Cohen et al., 2003). The SQ is a 60-item questionnaire designed to measure systemizing, or the drive to analyze or construct systems. The SQ contains 40 questions related to systemizing and 20 filler questions.

Empathy Quotient (EQ-60; Baron-Cohen & Wheelwright, 2004). The EQ-60 is a 60-item questionnaire consisting of 40 empathy-related questions and 20 filler questions. The questionnaire is designed to measure individual differences in empathy.

Triarchic Psychopathy Measure (TriPM; Blagov et al., 2016). The TriPM is a 58-item self-report questionnaire designed to measure psychopathy in terms of three constructs, each measured by a separate subscale. These constructs are boldness, meanness, and disinhibition.

Raven Advanced Progressive Matrices(Raven et al., 1989). The Raven’s Progressive Matrices consists of 36 sets of geometric patterns arranged in 3x3 cells. For each set, one of the patterns is missing and the task for the participant is to select among several alternatives, the one that correctly completes the overall series of patterns in that set. Participants in the present study were given the opportunity to complete up to 18 of the 36 sets in 10 minutes (as abridged by Unsworth & McMillan, 2014). A participant’s score is the total number of correct solutions.

Letter Sets (Ekstrom, French, Harman, & Dermen, 1976). In this task, participants saw five sets of four letters. Of these five sets, four were unified by a rule that consistently applied to their composition and ordering. Participants were required to indicate the set that violates the rule. Following two examples, participants had 5 minutes to complete 20 test items. A participant’s score was the total number of items solved correctly.

Number Series(Thurstone, 1962). In this task, participants saw a series of numbers organized according to some unstated rule. Participants were required to figure out this rule in order to determine what the next number in the series would be. Participants selected their answer out of five possible numbers that were presented. Following five practice items, participants had 5 minutes to complete 15 test items. While participants are typically granted 4.5 minutes to complete this task, we extended the allotted time to 5 minutes in our online version of the study.

Perspectives Questionnaire(Baryla et al., 2019). The Perspectives Questionnaire is a 20-item self-report measure of propensities to take agentic (a person who performs an action) or recipient (a person at whom the action is directed and who experiences its effects) perspectives in the context of social interaction. Respondents rate the extent to which certain statements (e.g. “I like to make decisions”) apply to them on a 7-point Likert scale from 1 (definitely not) to 7 (definitely yes).

Personal Sense of Power Scale(Anderson et al., 2012). The Sense of Power scale is an 8-item scale that measures an individual’s personal sense of power. Respondents rate the extent to which they agree with statements such as “my wishes do not carry much weight” on a 7-point Likert scale from 1 (strongly disagree) to 7 (strongly agree).

Need for Cognition Scale (NCS; Cacioppo et al., 1984). The NCS is an 18-item self-report questionnaire that measures an individual’s tendency to engage in and enjoy effortful cognitive endeavors. Reponses are in the form of a 9-point Likert-scale from -4 (very strong disagreement) to +4 (very strong agreement).

Reflective Functioning Questionnaire – 8 item (RFQ-8; Fonagy et al., 2016). The RFQ-8 is an 8-item brief screening measure of reflective functioning, the tendency to understand people in terms of intentional mental states, such as feelings, desires, wishes, goals and attitudes. Responses are in the form of a 7-point Likert scale from 1 (strongly disagree) to 7 (strongly agree). Critically, this measure addresses the extent to which people feel certain about the nature of mental state inferences (e.g. “I always know what I feel”) resulting in a certainty and uncertainty score for each participant.

The Ten-Item Personality Inventory (TIPI; Gosling et al., 2003). The TIPI is a 10 item self-report measure of the Big-Five personality dimensions. For each dimension, one item represents a positive pole, the other a negative pole. Participants rate how each trait applies to themselves using a 7-point Likert scale from 1 (strongly disagree) to 7 (strongly agree). We recoded the reversed score items (items 2, 4, 6, 8 & 10) and computed the average of the two items (the standard item and the reverse-scored item) that make up each personality dimension.

We operationalized altercentrism and egocentrism on the Dot Task in two different ways. First, we computed altercentric interference (RT) scores by subtracting RTs for self-congruent trials from self-incongruent trials. For egocentric interference (RT) scores, we subtracted RTs for other-congruent trials from other-incongruent trials. For the Dot task, we excluded all observations that were over 2.5 SDs from the mean RT of that participant. In addition to the RT scores, we also computed altercentric and egocentric error scores. To this end, for altercentric interference (error) scores, we subtracted the number of errors committed in self-congruent trials from self-incongruent trials. Similarly, for egocentric interference (error) scores we subtracted the number of errors committed in other-congruent trials from other-incongruent trials. This resulted in two altercentrism and egocentrism scores (RT and errors) for each participant. We computed a single director altercentrism score for each participant on the Director Task by dividing the number of errors committed by the total responses on trials in which a co-present avatar shared the same perspective as the participant. Conversely, for the director egocentrism score, we computed the same error ratio for trials in which the participant and director had different perspectives. We computed a single egocentrism score based on responses to the first three items of the Ambiguous Reference Task. Of the four items used, the first three were positively correlated with one another (see Supplementary Materials) while the fourth item was negatively correlated with the other three. Since we were primarily interested in extracting a single value reflecting the general performance pattern of participants on this task, we removed the fourth item from our analysis. We then used exploratory factor analysis to test a single factor model in which the three items loaded onto a single factor. We will refer to this score as ambiguous reference egocentrism throughout this paper (for brevity purposes, this score will be identified as keysar in all figures). For the 6/9 Task, selections of the central number as “6” were coded as egocentric while “9” responses were coded as altercentric. If a participant provided any other response besides “6” or “9”, they were removed from the analysis. In this study, we operationalized mentalizing accuracy, over-mentalizing, and under-mentalizing with the MASC, and we operationalized mentalizing certainty and uncertainty with the RFQ.

We used confirmatory factor analysis to test a single factor model in which the three fluid intelligence tasks (Raven’s matrices, number series, letter series) loaded onto a single factor. We will refer to this single factor as fluid intelligence or Gf throughout this paper.

We measured levels of psychopathology in participants through the BSI-18 scales for depression, anxiety, and somatization. Psychopathy was measured through TPM and its subscales meanness, boldness, disinhibition. Emotion regulation was measured through the DERS-SF and its subscales strategies, non-acceptance, impulse, goals, awareness, and clarity. Levels of Borderline traits were measured through the PAI-BOR and its subscales affective instability, identity problems, negative relationships, and self-harm.

Finally, we measured empathizing through the IRI subscale for empathy, as well as the Empathizing Quotient (EQ).

We adopted a model comparison approach to explore relationships between more than two variables. As this study is largely exploratory, we performed correlational analysis to examine the relationship between most measures. For all analyses, we will treat an effect size of r = .05 as “very small”, r = .10 as “small”, r = .20 as “medium” and r = .30 as “large” (Funder & Ozer, 2019). We set our alpha at .05 when examining the relationship between variables, but discuss the overall support for a given hypothesis based on the pattern of findings observed across analyses. We take this approach as we are interested in reporting both similarities and differences between measures of the same construct in this study.

The correlations between our egocentrism and altercentrism variables are illustrated in Figure 1 (95% confidence intervals for all correlations can be found in the supplementary materials). We found that altercentric and egocentric RT scores on the Dot Task were not significantly correlated with their corresponding error scores. In contrast, we found significant positive correlations between altercentric and egocentric variables (RT and error scores) on the Dot task. However, after controlling for fluid intelligence, the only significant relationship observed between Dot Task variables was between altercentric and egocentric interference (RT) scores (r = 0.38, p < 0.0001). Along those lines, egocentric errors and altercentric errors were highly correlated with each other on the Director task (r = 0.77, p < 0.0001) even when controlling for fluid intelligence.

While egocentric variables were largely unrelated across tasks, we found several significant relationships between altercentric variables across tasks. Specifically, other than a significant positive correlation between egocentric errors on the Dot task and egocentric errors on the Director task, none of the egocentric variables were significantly correlated with one another. Furthermore, this dot-egocentric-error/director egocentrism relationship was not significant when controlling for fluid intelligence. In contrast, the relationship between altercentric errors on the Dot Task and Director Task was significant even when controlling for fluid intelligence (r = 0.42, p < 0.0001).

In addition to the significant correlations observed between some of our altercentrism measures, several of the altercentrism/egocentrism relationships were significant even when controlling for fluid intelligence as well. Specifically, the relationships between egocentric errors on the Director Task and altercentric errors on the Dot Task (r = 0.33, p < 0.0001), and between ambiguous reference egocentrism and altercentric errors on the Dot Task (r = 0.25, p = 0.002), were significant even when controlling for fluid intelligence. The relationships between egocentric errors on the Director Task and altercentric interference (RT) on the Dot Task, altercentric errors on the Director Task and ambiguous reference egocentrism, and altercentric interference (RT) on the Dot Task and ambiguous reference egocentrism, were no longer significant when controlling for fluid intelligence.

To summarize, after controlling for fluid intelligence, we observed no significant correlations between our various egocentrism variables. Altercentrism variables, on the other hand, were significantly correlated with other altercentrism and egocentrism variables. While several of these relationships were not significant when controlling for fluid intelligence, several remained significant. Finally, we did not observe a significant relationship between any of the altercentric or egocentric variables with selection on the 6/9 task.

The correlations between our mentalizing variables are illustrated in Figure 2. Mentalizing accuracy, as indexed by the MASC, was negatively correlated with scores on most altercentrism/egocentrism variables (r range: -0.35 to -0.42, p < 0.0001). Over-mentalizing, as indexed by the MASC, was not significantly correlated with any of the altercentrism/egocentrism variables. In contrast, under-mentalizing and lack of mentalizing, as indexed by the MASC, were positively correlated with most of our egocentrism/altercentrism variables. These relationships remained significant when controlling for fluid intelligence for correlation coefficients which were equal to or over r = 0.25. We did not observe a significant relationship between performance on the MASC with selection on the 6/9 task. Certainty/uncertainty scores, as indexed by the RFQ, were not correlated with egocentrism/altercentrism variables, under/overmentalizing, or mentalizing accuracy.

The confirmatory factor analysis to examine if Raven’s matrices, number series, letter series loaded onto a single factor revealed a model that was acceptable, χ²(3)= 138.358, p < 0.0001, RMSEA = .0, SRMR = .0, NNFI = , CFI = 1, AIC = 2312.691. This factor was negatively correlated with all egocentric/altercentric scores (r range -0.24 to -0.53, p < 0.01) (see Figure 3). Fluid intelligence was positively correlated with MASC accuracy scores (r = 0.39, p < 0.0001) and negatively correlated with MASC no-mentalizing (r = -0.31, p < 0.001) and MASC under-mentalizing scores (r = -0.4, p < 0.0001). Fluid intelligence was not significantly correlated with MASC over-mentalizing or with certainty/uncertainty scores and did not predict selection on the 6/9 task.

The results of our examination of how self-reported psychopathology symptoms are related to measures of mentalizing tasks are demonstrated in Figure 4.

Egocentrism/altercentrism. Somatization, as measured by the BSI, was positively correlated with Dot altercentric errors (r = 0.19, p = 0.02) and not correlated with the other altercentric/egocentric variables. Anxiety and depression symptoms were not significantly correlated with any of the altercentrism/egocentrism scores. Individuals with higher levels of somatization symptoms were more likely to select egocentrically on the 6/9 task. Levels of anxiety and depression were not related to selection on the 6/9 task. PAI-BOR total score was not correlated with any of the altercentric/egocentrism variables. Identity problems, affective instability, and negative relationships, as assessed by the PAI-BOR, were not significantly correlated with altercentrism/egocentrism scores. Self-harm was positively correlated with egocentric scores on the Director Task (r = 0.16, p = 0.045) and not significantly correlated with the other variables. Identity problems significantly predicted selecting altercentrically on the 6/9 task. The other PAI-BOR scores did not significantly predict selection on the 6/9 task. While the goals subscale of the Difficulties with Emotion Regulation (DERS) measure significantly predicted responding altercentrically on the 6/9 task, and the impulsivity subscale was positively related with altercentric errors on the Dot task (r = 0.2, p = 0.012), most subscales with the DERS were not correlated with altercentric/egocentric scores.

Accuracy and Over-mentalizing/Under-mentalizing. Somatization was negatively correlated with MASC accuracy (r = -0.25, p = 0.002), positively correlated with MASC no-mentalizing (r = 0.17, p = 0.032) and MASC over-mentalizing (r = 0.22, p = 0.008) and not significantly correlated with MASC under-mentalizing. Anxiety and depression scores were not significantly correlated with any of the MASC scores. Affective instability, identity problems and negative relationships were not significantly correlated with MASC scores. Self-harm was negatively correlated with MASC accuracy (r = -0.21, p = 0.009) and positively correlated with MASC no-mentalizing (r = 0.27, p < 0.001), but was not significantly correlated with over-mentalizing or under-mentalizing. Certainty about mental states scores were negatively correlated with all PAI-BOR scores (r range: -0.39 to -0.52, p < 0.0001) and uncertainty about mental states was positively correlated with all PAI-BOR scores (r range: 0.36 to 0.55, p < 0.0001). Accuracy (on the MASC) was positively correlated with DERS goals (r = 0.20, p = 0.016), negatively correlated with DERS impulsivity (r = -0.18, p = 0.03), and not significantly correlated with the other egocentrism/altercentrism variables. No-mentalizing (on the MASC) is positively correlated with DERS impulsivity (r = 0.23, p = 0.004) and not significantly correlated with the other variables. MASC under-mentalizing was negatively correlated with DERS goals (r = -0.23, p < 0.005) and not with other variables. MASC over-mentalizing was not significantly correlated with any of the DERS scores.

Certainty/uncertainty. Certainty about mental states was negatively correlated with all BSI (r range: -0.24 to -0.34, p < 0.01) PAI-BOR (r range: -0.39 to -0.45, p < .0001) and DERS scores (r range: -0.35 to -.52, p < 0.0001). Uncertainty about mental states was positively correlated with all BSI (r range: 0.39 to 0.45, p < 0.0001), PAI-BOR (r range: 0.36 to 0.52, p < 0.0001) and DERS scores (r range: 0.19 to 0.60, p < 0.05).

Correlations between psychopathy and mentalizing scores can be found in Figure 5.

Altercentrism/Egocentrism. Meanness and disinhibition were positively correlated with the Director Task altercentrism and egocentrism (r range 0.25 to 0.33, p < 0.01) but not significantly correlated with the other altercentrism/egocentrism variables. Boldness was significantly correlated with director altercentrism (r = 0.18, p = 0.026) and none of the other variables. The total psychopathy score was positively correlated with egocentric errors on the Dot Task, egocentrism on the Ambiguous Reference Task and both director egocentrism and altercentrism (r range 0.17 to 0.38, p < 0.05). With the exception of the relationship between disinhibition and director altercentrism, these relationships were significant even when controlling for fluid intelligence. None of the psychopathy scores were significantly correlated with altercentrism (RT or errors) on the Dot Task.

Accuracy and Over-mentalizing/Under-mentalizing. All psychopathy scores were negatively correlated with accuracy (r range: -0.20 to -0.38, p < 0.05). Meanness and total psychopathy scores were positively correlated with no mentalizing (r range: 0.30 to 0.37, p < 0.001), over-mentalizing (r range: 0.16 to 0.18, p < .05) and under-mentalizing (r range: 0.17 to 0.26, p < 0.05). While Boldness was positively correlated with under-mentalizing (r = 0.24, p < 0.003) and not significantly correlated with no-mentalizing and over-mentalizing, disinhibition was positively correlated with no-mentalizing (r = 0.31, p < 0.0001) and over-mentalizing (r = 0.17, p = 0.04) but not significantly correlated with under-mentalizing.

Certainty/Uncertainty. Meanness was negatively correlated with certainty (r = -0.23, p < 0.01) and not significantly correlated with uncertainty. While Boldness was positively correlated with certainty about mental states (r = 0.32, p < 0.0001) and negatively correlated with uncertainty (r = -0.26, p = 0.001), disinhibition showed the opposite pattern and was negatively correlated with certainty (r = -0.42, p < 0.0001) and positively correlated with uncertainty (r = 0.44, p < 0.0001). Finally, meanness was negatively correlated with certainty (r = -0.23, p = 0.005) and not significantly correlated with uncertainty while total psychopathy scores were not significantly correlated with certainty/uncertainty scores.

The correlations between sense of power, agent/patient perspective, and social interaction anxiety and mentalizing scores can be found in Figure 6.

Egocentrism/Altercentrism. Sense of power was positively correlated with egocentrism on the ambiguous reference task, Dot altercentric errors and both altercentrism and egocentrism on the Director Task (r range: 0.18 to 0.34, p < 0.05). In contrast to this robust pattern, while agentic perspective scores were positively correlated with dot altercentric interference (RT) (r = 0.17, p = 0.037) and recipient perspective scores were positively correlated with director egocentrism (r = 0.2, p = 0.013). These perspectives were not significantly correlated with the other egocentrism/altercentrism variables. Social interaction anxiety was not significantly correlated with any of the egocentrism/altercentrism variables. Finally, sense of power, agentic/recipient perspective scores and social interaction anxiety do not significantly predict selection on the 6/9 Task.

Accuracy and Over-mentalizing/Under-mentalizing. Sense of power was positively correlated with no-mentalizing and under-mentalizing scores on the MASC (r range: 0.34 to 0.35, p < 0.0001) and negatively correlated with accuracy on the MASC (r = -0.36, p < 0.0001). Sense of power was not significantly correlated with over-mentalizing. Both agentic and recipient perspective scores were negatively correlated with accuracy on the MASC (r = -0.17, p = 0.036) and neither were significantly correlated with over-mentalizing. Agentic perspective scores were positively correlated with under-mentalizing on the MASC (r = 0.22, p = 0.006), and not significantly correlated with no-mentalizing or over-mentalizing. Recipient perspective scores were positively correlated with no-mentalizing and under-mentalizing (r range: 0.19 to 0.21, p < 0.05). Social interaction anxiety was negatively correlated with under-mentalizing (r = -0.17, p = 0.042) and not significantly related with the other MASC variables.

Certainty/Uncertainty. Certainty about mental states was significantly negatively correlated with social interaction anxiety (r = -0.46, p < 0.0001) and recipient perspective (r = -0.26, p = 0.001), and positively correlated with agentic perspective (r = 0.33, p < 0.0001). Certainty scores were not significantly correlated with sense of power. Uncertainty scores were positively correlated with social interaction anxiety (r = 0.37, p < 0.0001), recipient perspective (r = 0.37, p < 0.0001), and sense of power (r = 0.28, p < 0.001).

The correlations between self-reported empathy and systemizing and mentalizing scores can be found in Figure 7.

Egocentrism/Altercentrism. None of the four IRI subscales (perspective-taking, empathic concern, fantasy, and personal distress) were correlated with any of the egocentrism/altercentrism variables. In contrast, self-reported Empathizing Quotient Scores (EQ) were negatively correlated with director egocentrism (r = -0.23, p = 0.004) and ambiguous reference egocentrism (r = -0.19, p = 0.02).

Accuracy and Over-mentalizing/Under-mentalizing. None of the four IRI subscales were correlated with any of the MASC variables. In contrast, self-reported Empathizing Quotient Scores (EQ) were positively correlated with accuracy (r = 0.20, p = 0.012) and negatively correlated with no-mentalizing (r = -.17, p = .03) and over-mentalizing (r = -0.20, p = 0.01). Empathy was not correlated with under-mentalizing scores. In contrast, Systemizing (SQ) was positively correlated with under-mentalizing (r = 0.22, p < 0.005) and not significantly correlated with the other three MASC measures.

Certainty/Uncertainty. Certainty about mental states was positively correlated with self-reported perspective-taking (r = 0.30, p < 0.001) and empathic concern (r = 0.22, p < 0.006) and negatively correlated with personal distress on the IRI (r = -0.38, p < 0.0001). Certainty scores were not significantly correlated with fantasy. Uncertainty about mental states scores were positively correlated with personal distress (r = 0.35, p < 0.0001) and not significantly correlated with the other three IRI subscales. Both self-reported empathizing (EQ) and systemizing (SQ) were positively correlated with certainty about mental states (r range: 0.37 to 0.44, p < 0.0001) and not significantly correlated with uncertainty about mental states.

The correlations between anthropomorphizing, need for cognition, the Big-5 personality dimensions and mentalizing scores can be found in Figure 8.

Egocentrism/Altercentrism. Anthropomorphizing scores were positively correlated with most of our egocentrism/altercentrism variables with the exception of egocentric and altercentric (RT) interference scores on the Dot task (r range: 0.19 to 0.33, p < 0.05). A similar pattern was observed for the need for cognition though altercentric interference (RT) on the Dot task was positively correlated with need for cognition scores (r = 0.20, p = 0.015) and Dot task egocentric errors were not significantly correlated with need for cognition. Extroversion was positively correlated with egocentric errors on the Dot task (r = 0.17, p = 0.042) and not significantly correlated with any of the other egocentrism/altercentrism variables. Emotional stability was positively correlated with altercentric interference (RT) and egocentric errors on the dot task (r range: 0.16 to 0.17, p < 0.05). Conscientiousness, openness, and agreeableness were not significantly correlated with egocentrism/altercentrism variables.

Accuracy and Over-mentalizing/Under-mentalizing. Anthropomorphizing was negatively correlated with MASC accuracy (r = -0.42, p < 0.0001) and positively correlated with no-mentalizing and under-mentalizing scores (r range: 0.33 to 0.35, p < 0.0001). Anthropomorphizing was not significantly correlated with over-mentalizing. We observed the same pattern for the need for cognition. Extroversion was negatively correlated with accuracy (r = -0.21, p = 0.008) and positively correlated with under-mentalizing (r = 0.30, p < 0.001). Extroversion was not significantly correlated with no-mentalizing and over-mentalizing scores. Emotional stability scores performed in the same way as extroversion scores. Conscientiousness scores were not significantly correlated with any of the four MASC scores. Openness scores were positively correlated with under-mentalizing (r = 0.18, p < 0.05) and not significantly correlated with the other three MASC variables. Agreeableness was negatively correlated with over-mentalizing (r = -0.24, p < 0.01) and not significantly correlated with the other three MASC variables.

Certainty/Uncertainty. Anthropomorphizing was not significantly correlated with certainty or uncertainty about mental states scores. In contrast, need for cognition scores were negatively correlated with certainty (r = -0.20, p = 0.014) and positively correlated with uncertainty about mental states scores (r = 0.31, p < 0.001). Extroversion and openness were positively correlated with certainty about mental states (r range: 0.18 to 0.27, p < 0.05) and not significantly correlated with uncertainty. Emotional stability, conscientiousness and agreeableness were all positively correlated with certainty about mental states (r range: 0.31 to 0.45, p < 0.001) and negatively correlated with uncertainty (r range: -0.21 to -0.46, p < 0.01).

In this study, we assessed the extent to which four mentalizing dimensions (egocentrism/altercentrism, accuracy, over-mentalizing/under-mentalizing, and certainty/uncertainty) relate to broader variability in cognitive and social functioning. To this end, we administered a battery of behavioral and self-report measures tapping mentalizing, fluid intelligence, personality dimensions, sense of self, cognitive tendencies, and psychopathology symptoms.

First, we found that different measures of egocentrism were not correlated with one another and while some altercentric measures were correlated with each other, this pattern was mixed. For example, altercentric errors on the Dot task were correlated with both altercentrism and egocentrism scores on the Director task while altercentric interference (RT) and altercentric errors on the Dot task were not correlated with one another. One potential interpretation of these results is to conclude that, since our various tasks are measuring different facets of mentalizing, they need not correlate with one another. Specifically, while the Dot and Director tasks tap aspects of visual perspective taking, the Ambiguous Reference task taps conceptual, non-visual perspective taking (Surtees & Apperly, 2012, though see Cole & Millett, 2019). From our perspective, on the other hand, given the fact that these measures are all purportedly tapping the construct of egocentrism, we would expect some amount of cross-task correlation. In fact, our pattern of results raises questions regarding the extent to which existing tasks are successfully capturing divergent egocentric/altercentric dimensions in contrast to a more general difficulty associated with representing two alternative perspectives. This underscores the need to continue to develop assessment tools of these dimensions with a focus on obtaining adequate construct validity.

Second, we found that variability along our mentalizing dimensions was tightly related to fluid intelligence scores. Specifically, with the exception of over-mentalizing on the MASC, fluid intelligence was significantly negatively correlated with all mentalizing variables tapping mentalizing difficulties (egocentrism, altercentrism, under-mentalizing, no-mentalizing) and positively correlated with mentalizing accuracy as assessed by the MASC. In light of prior research pointing to a link between general intelligence and performance on the “Reading the Mind in the Eyes” task (RMET) (e.g., Baker et al., 2014), this pattern of results was expected. However, the magnitude of these effects was generally higher than observed in previous research. To illustrate, while Baker et al.’s (2014) meta-analysis of the relation between IQ and RMET scores revealed an effect of r=.24, 95% CI [.19, .29], we found that accuracy scores on the MASC were positively associated with fluid intelligence scores at r=.39, 95% CI [.25, .52], and negatively associated with director egocentrism at r = -.53, 95% CI [-.63, -.39].

The association between fluid intelligence and mentalization could be interpreted in two distinct ways. On the one hand, one might argue that the fluid intelligence/mentalizing link observed in these data simply reflects the more general relationship between general intelligence and mentalizing abilities. On the other hand, it is possible that existing measures, which exert considerable task-maintenance abilities, confound general cognitive abilities (fluid intelligence marking but one aspect of a broader set of abilities undergirding performance on mentalizing tasks) with mentalizing abilities or tendencies. Future research can continue to explore this question by extending the scope of general cognitive abilities studied in relation to mentalizing tasks on the one hand and by devising mentalizing tasks that vary in the extent to which they exert general task-maintenance demands on the other hand.

Despite these potential issues with construct validity and relation to general fluid intelligence, we found several noteworthy relationships between self-reported psychopathology symptoms and mentalizing dimensions. In particular, somatization (BSI-18) was positively correlated with altercentric errors (Dot task) and no-mentalizing (MASC), and was negatively correlated with accuracy (MASC) and over-mentalizing (MASC). In addition to somatization, self-harm (PAI-BOR) was positively correlated with egocentrism (Director task) and no-mentalizing (MASC) and negatively correlated with accuracy (MASC). In other words, somatization (BSI-18) and self-harm (PAI-BOR) scores were systematically related to variability along many of the mentalizing dimensions studied (see Ballespí, Vives, Alonso, et al., 2019; Laghi et al., 2016). In contrast, anxiety and depression scores (BSI-18), as well as identity problems, affective instability and negative relationships (PAI-BOR), were generally not significantly correlated with altercentrism/egocentrism or MASC variables. Interestingly, with the exception of an r = .22, 95% CI [.058, .4] association with somatization (BSI-18), over-mentalizing on the MASC was not significantly associated with any of the other measures tapping psychopathology. This pattern of results stands in contrast to previous findings pointing to an association of r = .25, 95% CI [.17, .31] between over-mentalizing and a wide range of psychopathology instances (McLaren et al., 2022). Future research can address this discrepancy.

In contrast to the altercentrism/egocentrism, accuracy and over-mentalizing/under-mentalizing dimensions, the associations between self-reported psychopathology and the certainty/uncertainty dimension was more widespread. In particular, all BSI-18 (somatization, anxiety, depression), PAI-BOR (self-harm, negative relationships, identity problems, affective instability) and DERS (goals, strategies, non-acceptance, impulse, awareness) scores were negatively correlated with certainty about mental states (RFQ-8). Conversely, these scores were all positively correlated with uncertainty about mental states (RFQ-8). Consistent with previous work (e.g., Müller et al., 2021), The effect sizes for these associations tended to be quite large.

The fact that uncertainty about mental states was more robustly associated with psychopathology than the other mentalizing dimensions we measured can be interpreted in several ways. On the one hand, given the fact that RFQ-8 (in contrast to tasks such as the Dot or Director task) was designed within a clinically oriented framework (see Fonagy et al., 2016), some of these correlations are to be expected. Furthermore, in contrast to the rest of the mentalizing dimensions which were behaviorally probed, certainty/uncertainty was the only mentalizing dimension which we measured by self-report. The relatively stronger associations observed between this dimension and psychopathology may therefore be due to the specific wording of key items on the RFQ (see Müller et al., 2021), or, at least in part, to the fact that behavioral and self-report measures tend to be weakly correlated (see Dang et al., 2020). On the other hand, the remarkably consistent patterns observed in our data underscores the extent to which the certainty/uncertainty dimension may be promising for future investigation of the relationship between mentalizing tendencies and psychopathology.

Due to prior findings indicating that psychopathic traits were related to mentalizing dimensions (e.g., less altercentric interference, Drayton et al., 2018) we were particularly interested in the psychopathy construct for the present study. We found that self-reported psychopathy (TPM) scores were significantly positively correlated with both altercentrism and egocentrism scores. Thus, while it is possible that individuals with psychopathy do in fact tend to experience less altercentric interference on tasks such as the Dot task (a claim in need of replication and extension) our data painted a more complex picture. In particular, while the positive correlation between egocentric interference and psychopathy may conceptually support the notion that individuals with psychopathic traits may indeed process their visual environment with less concern for co-present others, the fact that psychopathy was also positively correlated with altercentric interference points to the need for future research to elucidate the nature of the relationship between psychopathy and mentalizing.

In addition to the relationship between psychopathy and egocentrism/altercentrism, we also found significant correlations between psychopathy scores and the accuracy and over-mentalizing/under-mentalizing dimensions. Specifically, individuals with higher psychopathy scores were more likely to have lower mentalizing accuracy and higher levels of over-mentalizing, under-mentalizing and no-mentalizing. Finally, in contrast to our other measures of psychopathology (DERS, BSI-18, PAI-BOR)—which were consistently negatively correlated with certainty about mental states and positively correlated with uncertainty—boldness and disinhibition as measured by the TPM showed the opposite pattern. Specifically, these were positively correlated with certainty and negatively correlated with uncertainty scores. Teasing apart the role certainty about mental states plays in the context of psychopathic traits (boldness and disinhibition but not meanness) from the role of certainty in individuals who are low in psychopathic traits would therefore be an interesting question for future research.

Aside from self-reported psychopathology and psychopathy, we were also interested in exploring relationships between variability along mentalizing dimensions and more general personality and interpersonal individual differences. Along those lines, we found that subjective sense of power during social interactions was positively correlated with several egocentrism (as well as altercentrism) measures. Furthermore, sense of power was negatively correlated with accuracy (MASC) and positively correlated with no-mentalizing and under-mentalizing scores. Subjective sense of power scores were thus linked to our mentalizing dimensions in a similar manner to that of our psychopathy scores, suggesting that individuals who experience high levels of subjective sense of power in social interactions may be more likely to experience challenges associated with mentalizing.

In a potentially related set of findings, significant (though inconsistent in terms of directionality) relationships between self-reported empathy and egocentrism/altercentrism have been reported (Mattan et al., 2016; Nielsen et al., 2015). While self-reported empathic concern and perspective-taking on the IRI were not significantly related to altercentric interference in our data (replicating neither Nielsen et al., 2015 or Mattan et al., 2016), self-reported empathy scores on the Empathizing Quotient (EQ) were negatively correlated with director egocentrism and ambiguous reference egocentrism. This pattern suggests that there is a negative relationship between self-reported empathy and egocentrism on both visual and non-visual measures. Future research is needed to understand the nature of this association (see Erle & Topolinski, 2015). The nature of the inconsistency between the empathic concern subscale score on the IRI and the Empathy Quotient (EQ) score in relation to egocentrism is also in need of further understanding.

Along those lines, while IRI scores were not significantly correlated with the accuracy or over-mentalizing/under-mentalizing dimensions, empathizing quotient scores were positively correlated with accuracy (MASC) and negatively correlated with no-mentalizing and over-mentalizing. Furthermore, while empathizing quotient scores were not correlated with under-mentalizing scores, systemizing scores were positively correlated with under-mentalizing. In sum, self-reported empathy on the empathizing quotient, but not the inter-personal reactivity index, was related to variability along several of our mentalizing dimensions.

One central limitation of our study is that psychopathology was assessed via self-report measures and was conducted with a general non-clinical sample. The extent to which our findings relate to clinical populations is therefore not clear. However, the meaningful relationships found between variability along dimensions used in psychopathology research and mentalizing dimensions within a general population highlights the potential associated with merging clinical assessment tools with social cognition research.

In our data, replicating previous findings (e.g., Fonagy et al., 2016), robust correlations were observed between uncertainty about mental states and psychopathology. At first glance, the conclusion might be that uncertainty about mental states is associated with psychopathology. However, from a culture-centered perspective, this conclusion is questionable. Specifically, the extent to which individuals feel that the contents of other people’s minds are readily accessible is culturally variable. For example, Wice et al. (2020) demonstrated that American participants reported greater mental state access than did Japanese participants when reflecting upon their knowledge of others mental states (see also Duranti, 2015). Concretely, it is plausible that a cultural assumption of gratuitous mental state access might positively relate to the certainty individuals experience regarding mental states. The positive correlation between uncertainty about mental states and psychopathology may therefore reflect the absence of a “good fit” between one’s professed uncertainty and the cultural norms underlying social relationships promoting mental state access and (arguably) an elevated level of certainty about mental state inferences.

The more general point here is that certainty/uncertainty about mental states, which in our data was the mentalizing dimension that was most consistently tied to psychopathology, may be inexorably tied to broader cultural norms regarding perceptions of mental state access. Consequently, we might predict that the relationship between certainty/uncertainty about mental states and psychopathology would manifest in different ways (if at all) in cultural contexts which are more circumspect in regards to the notion that mental states are readily accessible or interpretable. Concretely, in contrast to the emerging picture from previous data as well as the present study, it is possible that uncertainty about mental states would not be predictive of psychopathology in cultural contexts which endorse notions of greater mental state opacity, restrict the scope of presumed mental state access, or conceptualize mental state access in different ways.

To illustrate the importance of considering mentalizing within cultural context, consider the MASC. In this task, participants observe Michael’s pushy and unsuccessful romantic pursuit of Sandra before and during a dinner party. While watching these interactions, unfold they are asked to make mentalizing inferences regarding the characters (e.g. “what is Sandra feeling”). We argue that “accurate” responses are made possible when the perceiver integrates their perceptions of the normative nature of the actor’s behavior in their mental state inference. Thus, it would be impossible to “accurately” assess what one actress (Sandra) is feeling if norm violations on the part of another actor (Michael) are not registered by the participant.

It is therefore evident that a thorough consideration of the implicit cultural norms underlying the development of mentalizing tasks as well as the extent to which participants integrate culturally-based normative reasoning with mentalizing inferences is critical in order to improve our understanding of what it means to “accurately” mentalize. This culture-centered approach may also shed light on the relationship between mentalizing tendencies and psychopathology in contexts which endorse different norms and encourage different mentalizing orientations. Indeed, a culture-centered approach would likely lead to the discovery of many more mentalizing dimensions which have yet to be empirically explored.

Contributed to conception and design: Netanel Weinstein, Lucy Whitmore and Kathryn Mills

Contributed to acquisition of data: Lucy Whitmore and Kathryn Mills

Contributed to analysis and interpretation of data: Netanel Weinstein and Lucy Whitmore

Drafted and/or revised the article: Netanel Weinstein and Kathryn Mills

Approved the submitted version for publication: Netanel Weinstein, Lucy Whitmore and Kathryn Mills

We would like to acknowledge Dare Baldwin, David Condon, Yoel Everett, Sara Hodges, Ulrich Mayr, Nash Unsworth and Sara Weston for providing helpful feedback regarding the study design.

This research was funded by the University of Oregon College of Arts and Sciences and a grant awarded to Netanel Weinstein from Gorilla.sc. Kathryn L Mills was supported by the Research Council of Norway (RCN) grant number 288083.

We have no competing interests. KLM is an associate editor at Collabra: Psychology. She was not involved in the review process of this article.

The work to produce this manuscript was consistent with the transparency and openness policies of Collabra: Psychology. As such, the analyses for this project were pre-registered and the data will be made available upon request by reviewers and will be made publicly available upon acceptance (https://osf.io/d7rgq/).