People represent their own mental states more distinctly than those of others

One can never know the internal workings of another person—one can only infer others' mental states based on external cues. In contrast, each person has direct access to the contents of their own mind. Here, we test the hypothesis that this privileged access shapes the way people represent internal mental experiences, such that they represent their own mental states more distinctly than the states of others. Across four studies, participants considered their own and others' mental states; analyses measured the distinctiveness of mental state representations. Two fMRI studies used representational similarity analyses to demonstrate that the social brain manifests more distinct activity patterns when thinking about one's own states vs. others'. Two behavioral studies complement these findings, and demonstrate that people differentiate between states less as social distance increases. Together, these results suggest that we represent our own mind with greater granularity than the minds of others.

simulate the drawing of new samples from the population. Sample sizes from 10 to 50 were simulated. Sample sizes of 30 and 35 respectively were estimated to offer 95% power detecting the effects in the earlier study, with the difference in the sample size between the studies due to the differing number and identities of mental states selected. The sample size for Study 3 was based on the sample size for Study 2, with additional participants collected to allow for the potential of attrition from the online experiment. For Study 4, we calculated an expected standardized effect size for the difference between close and far targets (d = .20) using data from Study 3. A parametric power analysis indicated the need for 192 participants to achieve 95% power. We increased our collection target to 350 to allow for the possibility that the effect size estimate from Study 3 was positively biased due to the procedure we used to select states for Study 4.
The far target's biography was comprised of filler material combined with key phrases related to their self-reported politics, religion, hobbies, and college major. It was crafted to be as dissimilar as possible to the participant, as well as being naturally less familiar and less likeable than the self. In Study 1, the distant target's gender was automatically set to match the participant. Names for the distant target were randomly selected amongst the three most common male and female names among college-aged individuals in the United States, so long as that did STATES MORE DISTINCT IN SELF VS. OTHERS 2 not match the participant's own name. In Study 2, the far target's gender was matched to the gender of the close target.
In both imaging studies, participants reported their relationship with each target (e.g. how much they liked them and how close, familiar, and similar they felt to that target), how warm, competent, and extraverted each target was. In Study 2 participants rated how much time, per week, they typically spent with the close target (in person, on the phone, texting, and on social media). In Study 2, participants also rated each of the 25 mental states in the imaging study on three dimensions: valence, social impact, and rationality. In both studies, we also solicited openended feedback about the experiment and its purpose.
Participants also provided their demographics and completed several individual difference measures. In Study 1, these measures consisted of the Narcissistic Personality Inventory 2 , the Reading the Mind in the Eyes task 3 , the MOS social support survey 4 , the UCLA Loneliness Scale Version 3 5 , a two-item extraversion measure, a measure of social network size, and the Autism-Spectrum Quotient 6 . In Study 2, participants completed the measures loneliness, extraversion, social network size, and Autism, as well as the Social Interaction Anxiety Scale 7 , and a moral judgment questionnaire 8 . These individual difference measures were collected for combination with other data sets in cross-study analyses, and thus we do not report analyses of these data in the present investigation. The design of Study 2 grouped targets into blocks during the imaging task. This was done to reduce task demands on the participants, so that they would not have to switch between targets on each trial as in Study 1. However, as a result, the design matrix featured collinearity between mental states within the same target person. This collinearity produced artifactual pattern correlations between some pairs of mental states. Monte Carlo simulation using null data revealed that the artificially induced pattern correlations were approximately equal to the partial correlations between the respective GLM regressors, controlling for the rest of the design matrix.
Although the artifactual pattern similarity was necessarily orthogonal to the effect of social distance which we were investigating, we nonetheless wished to mitigate this source of noise.
Thus, for all Study 2 analyses involving dissimilarity matrices, we first regressed out a matrix of the partial correlations between regressors of interest. All representational similarity analyses in Study 2 were conducted using the residuals from this regression in place of the raw dissimilarity matrices.
We used multidimensional scaling (MDS) to visualize the similarity between mental states. This technique finds a 2-D configuration, or map, of objects (in this case, mental states) that reproduces a set of measured distances (or dissimilarities) or dissimilarities between said objects. The specific algorithm we used was a nonmetric MDS implemented in the smacof package in R. The MDS mapping allowed us to visualize changes in the size of the mental state representational space as a function of target person in Studies 2 and 3, both of which featured all three targets. For the Study 2 analysis, we computed dissimilarity matrices using state-specific activity patterns from within the voxels which were significant in the Study 1 searchlight analysis (Figure 1). These matrices were averaged across participants prior to MDS. For Study 3, we averaged (and reverse-coded) ratings of pairwise similarity across participants to produce a STATES MORE DISTINCT IN SELF VS. OTHERS 5 single dissimilarity matrix ( Figure 3B). The averaged dissimilarity matrices from each study were divided into three partsone for each target personand 2-dimensional MDS was applied separately to each target dissimilarity matrix. Procrustes analysis was used to rotate the resulting configurations into a similar orientation to facilitate comparison of targets within each study.
These configurations were then plotted togetherappropriately scaled to indicate the average dissimilarity in each conditionto illustrate the changing similarity of mental state patterns as a function of social distance. The average dissimilarities between states for each target were plotted as circles in the same space.
In previous research 9 , we identified three psychological dimensionsrationality, social impact, and valencewhich explained a substantial amount of the neural activity underlying mental state representation. Here we tested whether these dimensions changed in importance as a function of whose mental states participants considered. To do so, we correlated neural pattern similarity between state-specific patterns in Studies 1 and 2, with predictions based on how close the corresponding states were on each of these psychological dimensions, (as well as a fourth dimension from that previous study, termed human mind). Neural similarity was examined within the same feature-selected regions specified above in the primary representational similarity analyses. The analysis was carried out separately for each target person within each participant. Resulting correlations between neural similarity and dimensional proximity were then Fisherized and subjected to paired t-tests (within participant, across target) to determine whether any dimension changed in importance as a function of target person.
In Study 1, we found one significant difference between the self and far target: valence was significantly better predictor of the similarity between one's own states than the similarity between the far target's states (Δr = .07, d = .62, p < .002). This result remained significant STATES MORE DISTINCT IN SELF VS. OTHERS 6 controlling for multiple comparisons across the four dimensions tested. However, this effect did not replicate in Study 2. Indeed, in Study 2, none of the 12 tests (4 dimensions X 3 target comparisons) detected a significant difference in dimensional importance between targets, even prior to controlling for multiple comparisons. Together, the results of these dimensional representational similarity analyses suggest that the shape of the mental state space remains generally consistent across targets, despite the smaller overall size of the space for close and far targets, relative to the self. In other words, the different in the overall granularity of mental state representation between self and other cannot be attributed to a specific reduction in dimensionality along any of the psychological dimensions considered.