Electrocorticographic evidence of a common neurocognitive sequence for mentalizing about the self and others

Tan, Kevin M.; Daitch, Amy L.; Pinheiro-Chagas, Pedro; Fox, Kieran C. R.; Parvizi, Josef; Lieberman, Matthew D.

doi:10.1038/s41467-022-29510-2

Download PDF

Article
Open access
Published: 08 April 2022

Electrocorticographic evidence of a common neurocognitive sequence for mentalizing about the self and others

Nature Communications volume 13, Article number: 1919 (2022) Cite this article

10k Accesses
10 Citations
222 Altmetric
Metrics details

Subjects

Abstract

Neuroimaging studies of mentalizing (i.e., theory of mind) consistently implicate the default mode network (DMN). Nevertheless, the social cognitive functions of individual DMN regions remain unclear, perhaps due to limited spatiotemporal resolution in neuroimaging. Here we use electrocorticography (ECoG) to directly record neuronal population activity while 16 human participants judge the psychological traits of themselves and others. Self- and other-mentalizing recruit near-identical cortical sites in a common spatiotemporal sequence. Activations begin in the visual cortex, followed by temporoparietal DMN regions, then finally in medial prefrontal regions. Moreover, regions with later activations exhibit stronger functional specificity for mentalizing, stronger associations with behavioral responses, and stronger self/other differentiation. Specifically, other-mentalizing evokes slower and longer activations than self-mentalizing across successive DMN regions, implying lengthier processing at higher levels of representation. Our results suggest a common neurocognitive pathway for self- and other-mentalizing that follows a complex spatiotemporal gradient of functional specialization across DMN and beyond.

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

Article Open access 09 May 2019

Single-neuronal predictions of others’ beliefs in humans

Article 27 January 2021

The brain’s default network: updated anatomy, physiology and evolving insights

Article 06 September 2019

Introduction

Humans are social by nature: our central nervous systems have evolved many mechanisms to support our rich and complex social worlds¹. Although high levels of sociality are seen throughout the animal kingdom^2,3, humans are exceptional in their capacity for mentalizing: the ability to consider the mental states and traits of others and oneself^4,5. The field of social neuroscience seeks to understand how mentalizing and other social functions are implemented at the level of brain and biology⁶. In humans, social neuroscience primarily relies on functional magnetic resonance imaging (fMRI), a neuroimaging modality with high spatial resolution but low temporal resolution⁷. Hundreds of fMRI studies have shown that mentalizing recruits default mode network (DMN) regions – including temporoparietal junction (TPJ), posteromedial cortex (PMC), and medial prefrontal cortex (mPFC) – with remarkable consistency across countless mentalizing paradigms instantiated in various sensory modalities^{4,8,9,10,11,12}. Nevertheless, the specific social cognitive functions of individual DMN regions remain unclear. When seen through fMRI, DMN regions appear to respond concurrently, yet electrophysiological studies demonstrate that critical neurocognitive dynamics occur at millisecond timescales throughout DMN¹³. Thus, the limited temporal resolution of fMRI may preclude more precise neurocognitive accounts of mentalizing and its component processes.

Several studies have investigated the fast spatiotemporal dynamics of mentalizing processing using source-space electroencephalography (EEG) and magnetoencephalography (MEG), neuroimaging modalities with millisecond temporal resolution but coarse spatial resolution¹⁴. These studies reveal a general spatiotemporal sequence of cortical recruitment, starting in visual cortex, followed by mirror neuron system regions (MNS; e.g., intraparietal sulcus and premotor cortex), then lastly in DMN regions^{15,16,17,18,19,20}. These findings exemplify the emerging consensus that visual representations are used by MNS to identify observable actions (e.g., grasping for food), which are then used by DMN to infer unobservable mental states (e.g., hunger)^{8,21,22,23,24}. Taken together, EEG/MEG studies of mentalizing suggest that visual cortex, MNS, and DMN act as a hierarchical neurocognitive pathway that transforms low-level visual inputs into high-level mentalistic inferences. However, despite broad agreement at the network level, these studies report inconsistent recruitment across individual DMN regions. These inconsistencies may reflect limitations in EEG/MEG source localization, particularly in deeper regions such as mPFC and PMC²⁵, which were not sampled in many of these studies. As such, the sequence of mentalizing processing across individual DMN regions remains unclear.

We sought a more spatiotemporally precise and mechanistic account of mentalizing by exploring neuronal population activity across individual DMN regions and beyond. Leveraging the benefits of human intracranial electrophysiology²⁶, we recorded high-frequency broadband activity (HFB; 70‒180 Hz), which reflects the rapid aggregate spiking of neuronal populations²⁷. In contrast, fMRI measures slow metabolic changes, although fMRI and HFB correspond in the anatomy and direction of measured effects (see Parvizi & Kastner, 2019²⁶). We show that self- and other-mentalizing share a complex spatiotemporal gradient of functional specialization at millisecond, millimeter, and cross-regional scales. Our findings demonstrate that high spatiotemporal resolution methods can provide critical insights on the neurocognitive mechanisms of human social cognition.

Results

Data and design

We recruited sixteen human participants who had electrocorticography (ECoG) electrodes surgically implanted onto the cortical surface for epilepsy monitoring and treatment (Supplementary Table 2). Recordings were obtained from all 2125 electrode sites in our cohort (Fig. 1b). Our behavioral task (Fig. 1a) consisted of true/false text prompts for three conditions of interest: self-mentalizing (e.g., “I am honest”), other-mentalizing (e.g., “My neighbor is honest”), and a non-social ‘cognitive’ task involving simple arithmetic (e.g., “9 + 86 = 95”; Supplementary Table 4). Sites and trials underwent exclusion criteria for epileptic activity, noise, poor behavioral performance (trials only), and statistical outliers (see Methods).

We began by parcellating the brain into seven regions-of-interests (ROIs; Fig. 1c) using each participant’s native-space cortical surface (Supplementary Fig. 2). We included six DMN ROIs that are strongly implicated in mentalizing¹²: temporoparietal junction (TPJ), anterior temporal lobe (ATL), posteromedial cortex (PMC), anteromedial prefrontal cortex (amPFC), dorsomedial prefrontal cortex (dmPFC), and ventromedial prefrontal cortex (vmPFC). Visual cortex was included as a control ROI. Of 2125 electrode sites, 555 were included in our ROIs.

We examined HFB activity from each site in two ways: single-trial and trial-averaged analyzes (Fig. 1d–f). Single-trial analysis compared trialwise HFB responses relative to the pre-stimulus baseline preceding each trial (p_FDR < 0.05, corrected across timepoints, trials, and sites). Single-trial analysis captured five key metrics of HFB activations: onset, peak, and offset latencies, duration, and peak power (defined in Fig. 1e). Trial-averaged analysis used linear mixed-effects models (LMEMs) to estimate mean timecourses of task-evoked HFB power relative to pre-stimulus baseline (Fig. 1f). Trial-averaged analysis identified sites with significant activations or deactivations for each task condition (p_FDR < 0.05, corrected across timepoints and sites). See Fig. 2 for within-site analyses of exemplar ROI sites. All statistical tests herein were two-tailed.

Functional specificity for mentalizing strengthens from visual cortex to mPFC

To explore the fast spatiotemporal dynamics of mentalizing processing, we first examined its functional-anatomic foundations. We began by identifying sites with significant HFB responses to either self- or other-mentalizing, regardless of functional specificity or selectivity (Fig. 3a, b). Using trial-averaged results, sites were considered ‘mentalizing-active’ (cyan) or ‘mentalizing-deactive’ (orange) if they produced higher or lower HFB power, respectively, relative to pre-stimulus baseline (p_FDR < 0.05). Sites were considered ‘mentalizing-nonresponsive’ if they produced nonsignificant HFB responses to mentalizing. We found mentalizing-active sites in nearly all parts of cortex, while mentalizing-deactive sites were rarer and generally located outside DMN (Fig. 3a). Overall, most cortical sites were mentalizing-nonresponsive (55% whole-brain; Fig. 3b).

**Fig. 3: A neurocognitive sequence for mentalizing.**

Next, we examined the functional specificity of mentalizing-active sites (Fig. 3c, d). Using trial-averaged results, we identified which mentalizing-active sites also produced significant HFB responses to arithmetic (cognitive task) relative to pre-stimulus baseline (p_FDR < 0.05). Additionally, we directly compared single-trial peak power across mentalizing and arithmetic using robust regression (p_FDR < 0.05, corrected across sites). Sites were considered ‘mentalizing-specific’ (light and dark turquoise) if they were (1) mentalizing-active but not arithmetic-active, and (2) produced significantly higher peak power for mentalizing over arithmetic. Sites were considered ‘non-specific’ (lime, pink, and red) if they coactivated for mentalizing and arithmetic, regardless of peak power differences.

Overall, most mentalizing-active sites were non-specific (68% whole-brain), while the remaining mentalizing-specific sites were unevenly distributed across cortex (Fig. 3c, d). Within our ROIs, the lowest mentalizing-specificity was found in visual cortex (1%). Intermediate mentalizing-specificity was found in TPJ (38%), ATL (32%), and PMC (58%). Very high mentalizing-specificity was found in amPFC (87%), dmPFC (94%), and vmPFC (100%). Taken together, these results show a gradient of mentalizing-specificity from visual cortex to mPFC.

Mentalizing activations propagate from visual cortex to mPFC

Next, we explored the timing of mentalizing-evoked activations across ROIs. To this end, we analyzed single-trial HFB latency metrics from mentalizing-active ROI sites (Fig. 3e). Pairwise ROI comparisons (Fig. 3f) measured trial-by-trial ROI differences while controlling for behavioral response times (RT_Behav) and participant-related heterogeneity using LMEMs (nested within Trial within Participant). Critically, pairwise comparisons only included participants with mentalizing-active sites in both ROIs.

We found that visual cortex produced the earliest activation onsets (M = 101 ± 1 ms) of any ROI (Fig. 3f; p_FDR < 0.05, corrected across unique ROI pairs). Afterwards, mid-latency onsets were produced by TPJ (M = 303 ± 7 ms), ATL (M = 316 ± 7 ms), and PMC (M = 322 ± 5 ms), with nonsignificant differences between them. Later onsets were produced by amPFC (M = 465 ± 10 ms) and dmPFC (M = 466 ± 7 ms), with nonsignificant differences between them, followed lastly by vmPFC (M = 537 ± 15 ms). Critically, propagation of activations across ROIs had robust trial-by-trial consistency (Supplementary Fig. 3). Peak latencies showed a similar pattern of cross-ROI differences as onsets, although intra-mPFC differences were nonsignificant (Fig. 3f). Offset latencies showed the least cross-ROI differentiation. Earlier offsets were produced by visual cortex, TPJ, and ATL, with nonsignificant differences between them. Later offsets were produced by amPFC, dmPFC, and vmPFC, with nonsignificant differences between them. We also performed post-hoc analysis of offset latencies relative to RT_Behav (Fig. 4c), revealing that mPFC activations more closely preceded RT_Behav than other ROIs combined (b₂₀₂₀₄ = 138 ± 31 ms, p = 4.28e−6). Nevertheless, despite these robust cross-ROI latency differences, ROIs predominantly activated at overlapping times (Figs. 2, 3e, 5c and 6; single-participant results in Supplementary Fig. 2).

**Fig. 4: Spatiotemporal gradients of functional and behavioral relevance for mentalizing.**

In sum, mentalizing evoked largely concurrent activations across ROIs, as might be expected from neuroimaging literature. Nonetheless, fine-scale cross-ROI differences in onset, peak, and offset latencies depict an overarching spatiotemporal sequence of activation from visual cortex to mPFC. This sequence extended across individual sites throughout cortex (Fig. 4a).

A spatiotemporal gradient of mentalizing-specificity from visual cortex to mPFC

Thus far, we have found spatial gradients in the timing (Fig. 3e, f) and functional specificity (Fig. 3c, d) of neuronal population responses to mentalizing. To examine the correspondence between these gradients, we used logistic mixed-effects classification to predict mentalizing-specificity from mean onset latencies across mentalizing-active ROI sites, nested within Participant (Fig. 4b). We found that the probability of mentalizing-specificity over non-specificity rose by 1.4% per one-millisecond increase in onsets (odds ratio = 1.014, b₂₅₅ = 0.014 ± 0.002, p = 9.21e−11). This robust effect was largely attributable to between-ROI differences, as within-ROI effects were weaker and significant only in TPJ (p = 0.047), ATL (p = 0.039), and PMC (p = 0.005). In sum, ROI sites with later activations were more likely to be mentalizing-specific.

These results portray a neurocognitive sequence^28,29,30,31 that reflects both the timing and specificity of mentalizing processing across ROIs. Within this sequence, visual cortex appears to be a pre-mentalizing stage, as it produced the earliest activations and negligible mentalizing-specificity. Next, temporoparietal DMN regions (TPJ, ATL, and PMC) appear to be lower-level mentalizing stages, as they produced mid-latency activations and intermediate mentalizing-specificity. Finally, mPFC regions (amPFC, dmPFC, and vmPFC) appear to be higher-level mentalizing stages, as they featured the latest activations and overwhelming mentalizing-specificity.

Behavioral responses are best predicted by TPJ and dmPFC activity

To explore the relationships between neuronal and behavioral responses, we determined which ROIs predicted RT_Behav and Choice_Behav (choosing ‘true’ or ‘false’ for task prompts) by analyzing single-trial HFB metrics from mentalizing-active ROI sites. We used LMEMs to simultaneously estimate neuronal associations with RT_Behav and Choice_Behav while controlling for stimulus visual dissimilarity (see Supplementary Methods), nested within Site and Participant. We found that onset latencies had significant (p < 0.05) positive associations with RT_Behav in visual cortex, TPJ, PMC, amPFC, and dmPFC (Table 1). Moreover, we found significant positive associations between RT_Behav and peak latencies (all DMN ROIs; Fig. 4d), offset latencies (all ROIs; Table 1), and activation duration (all ROIs). In contrast, activation magnitude (peak power) had significant RT_Behav associations in only two ROIs: TPJ and dmPFC (Fig. 4e). Intriguingly, dmPFC was the only ROI that significantly predicted Choice_Behav (Table 1). Taken together, behavioral responses were best predicted by TPJ and dmPFC, which had more numerous significant neurobehavioral associations than other ROIs.

Table 1 Aggregate ROI analyses of mentalizing.

Full size table

Next, we examined whether behavioral responses were better predicted by activity in successive ROI sites. Specifically, we compared mean onset latencies (Fig. 4a) and within-site RT_Behav random effect sizes (b from analyses in previous paragraph) using LMEMs nested within Participant. We found that sites with later onsets had stronger RT_Behav effects for peak latency (b₂₅₅ = 0.972 ± 0.094, p = 1.51e-23) but not peak power (b₂₅₅ = 0.019 ± 0.074, p = 0.796). In sum, RT_Behav was better predicted by successive ROIs in terms of activation latency (Fig. 4d) but not activation magnitude (Fig. 4e).

Self- and other-mentalizing share a common neuroanatomical basis

To fractionate mentalizing’s neurocognitive sequence across mentalizing type, we first explored the anatomical interrelations between self- and other-mentalizing (Fig. 5a, b). We therefore identified sites that produced significant trial-averaged activations for each mentalizing type relative to baseline (p_FDR < 0.05, corrected across timepoints and sites). We also directly compared self/other differences in single-trial peak power using robust regression (p_FDR < 0.05, corrected across sites). Sites were considered ‘self-only’ or ‘other-only’ if they (1) produced significant trial-averaged activations for only one mentalizing type, and (2) produced significantly greater peak power for that mentalizing type over another. Sites that activated for both mentalizing types were labeled by peak power differences: ‘self-greater’ (Self > Other), ‘other-greater’ (Other > Self), or ‘non-selective’ (Self = Other).

**Fig. 5: A common neurocognitive sequence for self- and other-mentalizing.**

We found that mentalizing-active sites were overwhelmingly coactive for both mentalizing types (non-selective + self-greater + other-greater = 97% whole-brain; Fig. 5a,b). Moreover, non-selective sites formed the largest category in all ROIs and the whole-brain (range: 55-86%). We also compared amounts of ‘self-selective’ sites (self-only + self-greater) versus ‘other-selective’ sites (other-only + other-greater) using McNemar χ² (Yates-corrected; df = 1). This revealed that other-selective sites significantly (p < 0.05) outnumbered self-selective sites in visual cortex (3% self/23% other; χ² = 9.39), PMC (6% self/34% other; χ² = 8.45), and the whole-brain (5% self/13% other; χ² = 24.40). Nonsignificant self/other differences were found in TPJ (5% self/14% other; χ² = 0.57), ATL (3% self/11% other; χ² = 0.37), amPFC (9% self/13% other; χ² = 0), dmPFC (6% self/8% other; χ² = 0), and vmPFC (0% self/45% other; χ² = 3.20).

In sum, self- and other-mentalizing recruited near-identical cortical sites in a largely non-selective manner, though selective sites were mostly other-selective in visual cortex and PMC. Unexpectedly, other-selective and self-selective sites did not significantly outnumber each other in TPJ, amPFC, and dmPFC, which contradicts previous fMRI work from our group³² and elsewhere^33,34,35.

Other-mentalizing evokes slower and lengthier activations across successive ROIs

Given the highly overlapping neuroanatomy of self- and other-mentalizing, we explored whether self/other differences might be better characterized by fast spatiotemporal functional dynamics. We therefore examined single-trial HFB metrics from mentalizing-active ROI sites. Mentalizing type was analyzed by LMEMs that controlled for RT_Behav and visual dissimilarity, nested within Site and Participant.

We found that self- and other-mentalizing evoked a common spatiotemporal sequence of HFB activations across ROIs (Fig. 5c). Within this sequence, onset latencies showed nonsignificant self/other differences (Table 1). However, other-mentalizing evoked significantly (p < 0.05) later peaks and offsets than self-mentalizing in all DMN ROIs except TPJ (Fig. 5c). Concordantly, activation duration was significantly longer for other- versus self-mentalizing in all DMN ROIs except TPJ (Fig. 5d). In contrast, self/other differences in activation magnitude were significant only in PMC, which produced greater peak power for other-mentalizing (Fig. 5e). Crucially, although HFB metrics often reflected RT_Behav and visual dissimilarity, significant self/other differences were ultimately dissociable from these covariates (Supplementary Table 1).

Next, we examined whether self/other differences became stronger in successive ROI sites. We therefore compared mean onset latencies (Fig. 4a) with single-site random effect sizes (b from analyses in previous paragraph) for peak latency (Fig. 5c) and peak power (e.g., self/other selectivity; Fig. 5e) using LMEMs nested within Participant. We found that later onsets predicted stronger self/other differences in peak latencies (b₂₅₅ = 0.404 ± 0.131, p = 0.002). However, onsets did not significantly predict peak power (b₂₅₅ = −0.303 ± 0.0744, p = 0.796).

In sum, we found that self- and other-mentalizing recruited near-identical sites in a common spatiotemporal sequence (Fig. 5a–c). Within this sequence, other-mentalizing evoked slower (Fig. 5c) and lengthier (Fig. 5d) activations in all DMN ROIs except TPJ. Intriguingly, successive ROI sites had greater self/other differentiation in the timing, rather than selectivity, of activations. Taken together, self/other functional differences were primarily characterized by the timing of activations throughout successive DMN regions.

Summary of evoked neuronal activity

To summarize task-evoked neuronal population activity (Fig. 6a), we identified sites with significant HFB responses during specific time windows relative to pre-stimulus baseline (p_FDR < 0.05, corrected across sites and time windows). From 0-250 ms, activations were largely localized to visual cortex with negligible task condition differences. From 250–500 ms, activations also encompassed temporoparietal and lateral frontal regions, where mentalizing and arithmetic (cognitive task) began to diverge. Specifically, in temporoparietal DMN regions, self- and other-mentalizing primarily evoked activations, while arithmetic evoked interdigitated activations and deactivations. From 500–750 ms, mentalizing activations began to encompass mPFC, especially for self-mentalizing. In contrast, arithmetic evoked mPFC deactivations, which continued for all successive time windows. From 750–1000 ms, both mentalizing types evoked similar mPFC activations. From 1000–2000 ms, other-mentalizing evoked more sustained activations than self-mentalizing, particularly in mPFC.

Controlling for stimulus visual dissimilarity (VD)

To distinguish mentalizing-related neuronal effects (mentalizing type, RT_Behav, and Choice_Behav) from stimulus VD (e.g., prompt length), we used computer vision^36,37 (see Supplementary Methods). After controlling for VD, mentalizing-related effects were still significant in all DMN ROIs (Figs. 4d, e and 5c–e and Table 1). However, in visual cortex, marked self/other differences in peak power became nonsignificant (Fig. 5e). In sum, mentalizing-related effects in DMN ROIs were not explained by prompt length and other visual features.

Behavioral results

To confirm that participants performed mentalizing for task prompts, we analyzed Choice_Behav for self/other biases towards positive or negative affective traits (Supplementary Fig. 4c) using logistic mixed-effects classification (nested within Trait within Participant). Overall, we found a greater probability of ‘true’ choices for positive versus negative traits (odds ratio = 2.53, b₁₄₇₁ = 0.931 ± 0.069, p = 1.04e−39). This bias was stronger during self-mentalizing versus other-mentalizing (odds ratio = 1.38, b₁₄₇₁ = 0.326 ± 0.064, p = 4.56e−7). These results indicate that participants engaged the mentalistic content in task prompts with a self-positivity bias, which is a canonical feature of mentalizing^38,39,40. Moreover, high accuracy in arithmetic trials (median = 92.5%) suggests effortful attention to task prompts (Supplementary Fig. 4b).

To determine whether RT_Behav varied by Mentalizing Type, Choice_Behav, and stimulus visual features (VD1 and VD2; see Supplementary Results), we used a LMEM nested within Participant (Supplementary Fig. 4a). We found faster RT_Behav for self-mentalizing (M = 2406 ± 49 ms) versus other-mentalizing (M = 2797 ± 52 ms; t₁₆₄₆ = 2.85, p = 0.004). Additionally, RT_Behav was faster for ‘true’ choices (M = 2482 ± 59 ms) versus ‘false’ choices (M = 2807 ± 45 ms; t₁₆₄₆ = 2.16, p = 0.031). However, the Mentalizing Type x Choice_Behav interaction was nonsignificant (t₁₆₄₆ = 1.47, p = 0.140). Unsurprisingly, longer prompt lengths (VD1) elicited slower RT_Behav (t₁₆₄₆ = 4.85, p = 1.36e-6). Meanwhile, remaining visual features (VD2) evoked nonsignificant RT_Behav effects (t₁₆₄₆ = −0.25, p = 0.803). In sum, mentalizing type, Choice_Behav, and VD1 had significant and dissociable effects on RT_Behav.

Discussion

Using electrocorticography (ECoG), we probed the neurocognitive substrates of mentalizing at the level of neuronal populations. We found that mentalizing about the self and others recruited near-identical cortical sites (Fig. 5a, b) in a common spatiotemporal sequence (Figs. 5c and 6). Within our ROIs, activations began in visual cortex, followed by temporoparietal DMN regions (TPJ, ATL, and PMC), and lastly in mPFC regions (amPFC, dmPFC, and vmPFC; Fig. 3e, f). Critically, regions with later activations exhibited greater functional specialization for mentalizing as measured by three metrics: functional specificity for mentalizing versus arithmetic (Figs. 3c, d and 4b), self/other differentiation in activation timing (Fig. 5c, d), and temporal associations with behavioral responses (Fig. 4d and Table 1). Taken together, these results reveal a common neurocognitive sequence^28,29,30,31 for self- and other-mentalizing, beginning in visual cortex (low specialization), ascending through temporoparietal DMN areas (intermediate specialization), then reaching its apex in mPFC regions (high specialization).

Our results are consistent with gradient-based models of brain function, which posit that concrete sensorimotor processing in unimodal regions (e.g., visual cortex) gradually yields to increasingly abstract and inferential processing in ‘high-level’ transmodal regions like mPFC^41,42. We found that the strength of self/other differences in activation timing increased along a gradient from visual cortex to mPFC. Specifically, other-mentalizing evoked slower (Fig. 5c) and lengthier (Fig. 5d) activations than self-mentalizing throughout successive DMN ROIs. These self/other functional differences corresponded with self/other differences in RT_Behav (Supplementary Fig. 4), although the two were often dissociable (Table 1). Thus, perhaps because we know ourselves better than others, other-mentalizing may require lengthier processing at more abstract and inferential levels of representation, ultimately resulting in slower behavioral responses.

What might our results imply about extant fMRI findings? Hundreds of fMRI studies consistently suggest that: (1) TPJ and dmPFC are most crucial for mentalizing^{6,8,11,12,43,44,45,46}, and (2) dmPFC is selective for thinking about others over oneself ^32,33,34,35. However, when examined with ECoG, we found that both pieces of received wisdom are not what they seem. Below, we discuss both issues before moving onto our peculiar vmPFC results, and then conclude with systems-level discussion.

Unsurprisingly, we found that DMN regions such as TPJ and dmPFC contained higher proportions of ‘mentalizing-specific’ sites (i.e., mentalizing-active but not cognitive-active) relative to the whole-brain average (Fig. 3c, d). The spatial distribution of mentalizing-specific sites roughly resembles the ‘mentalizing network’ reported in countless fMRI studies^{4,6,8,9,10,11,47}. However, our DMN ROIs were by no means functionally homogenous. Relative to other ROIs, TPJ and dmPFC activity best predicted RT_Behav (Fig. 4d, e and Table 1), supporting the notion that both regions are most crucial for mentalizing performance^{6,12,43,45,46,48,49}.

We also found numerous functional distinctions between TPJ and dmPFC, which is surprising given their remarkably similar functional profiles in fMRI literature^{4,6,8,9,10,11,47,50}. Specifically, we found that TPJ produced earlier activations (Fig. 3e, f) that were notably coactive for mentalizing and arithmetic (cognitive task; Fig. 3c, d). Indeed, the onsets of TPJ activations were the earliest of any DMN ROI. In contrast, dmPFC produced significantly later activations (Fig. 3e, f) that were overwhelmingly mentalizing-specific (Fig. 3c, d), indicating that dmPFC sits at a higher level of mentalizing’s neurocognitive sequence than TPJ. Moreover, aggregate ROI analyses revealed no significant self/other differences in TPJ (Table 1), while dmPFC featured robust self/other timing differences (Fig. 5c, d), suggesting that dmPFC is more sensitive to variation in mentalistic content. In terms of behavior, although TPJ and dmPFC best predicted RT_Behav, dmPFC was the only ROI that predicted Choice_Behav (choosing ‘true’ or ‘false’ for task prompts; Table 1), implying that dmPFC instantiates more aspects of mentalistic behavior than TPJ. Strikingly, unlike TPJ, dmPFC activation offsets closely preceded behavioral responses (within 124 ± 7 ms; Fig. 4c), suggesting that dmPFC is more deeply involved in the final stages of mentalistic reasoning. Taken together, while TPJ and dmPFC are both clearly crucial for mentalizing performance, dmPFC appears more specialized for mentalizing itself.

Given the marked functional differentiation between TPJ and dmPFC, what specific neurocognitive roles might they play in mentalizing? In social neuroscience, TPJ is often considered a functionally-specific locus for explicit belief reasoning^45,46,51,52. Yet here, TPJ was less functionally specialized relative to dmPFC (Figs. 3c–f, 4 and 5c, d). To explain this discrepancy, we suggest that TPJ provides crucial antecedents for explicit belief reasoning in dmPFC. Given TPJ’s central role in automatic evaluations of thematic semantics^{53,54,55,56,57,58,59,60,61}, we propose that TPJ automatically represents integrative psycho-semantic models of exemplar contexts for a given inference. In simpler terms, TPJ may help us ‘see’ the psycho-semantic gestalt of a given situation⁶². Accordingly, tasks that ‘show’ concrete mentalistic content (e.g., social animations) often recruit TPJ but not dmPFC, while tasks that require mentalistic logical inferences (e.g., false belief) often recruit dmPFC in addition to TPJ^10,11,12. Thus, when mentalistic content feels ‘seeable’ from perceptual processing, TPJ could generate mentalistic inferences without explicit belief reasoning. Indeed, work on implicit and spontaneous mentalizing show that TPJ can encode an actor’s beliefs without any explicit reasoning^{16,63,64,65,66,67,68,69,70,71}. Taken together, TPJ may implicitly set the psycho-semantic stage for explicit belief reasoning that occurs later in dmPFC when necessary (e.g., our trait judgment task; Fig. 1a).

The dmPFC may be well-suited for explicit belief reasoning^72,73,74,75. We found substantial concurrent activation between dmPFC and all other ROIs (Figs. 2, 3e and 6), suggesting that dmPFC could work iteratively with lower-level regions to refine what is ‘seen’, thus providing dmPFC with increasingly-useful inputs from which to draw better inferences^76,77. Moreover, studies on strategic reasoning show that dmPFC can arbitrate between multiple mental models⁷⁸ and prospective choices⁷⁹ by simultaneously evaluating multiple possibilities⁸⁰ through ‘fuzzy’ propositional reasoning^77,81. As such, dmPFC may arbitrate between multiple exemplar contexts to help extract the most relevant and enduring semantic features for a given psychological inference. Taken together, dmPFC may integrate and refine representations throughout mentalizing’s neurocognitive pathway to strategically reason about minds.

As for mentalizing about the self or others, fMRI studies routinely suggest that dmPFC is ‘other-selective'^32,33,34,35. What underlying neuronal population dynamics could result in stronger hemodynamic responses for one mentalizing type over another? The standard assumption would be that the magnitude (i.e., intensity) of neuronal activations differs across mentalizing type. This might be seen in aggregate ROI activity, or perhaps across individual ROI sites. We tested both possibilities by examining self/other differences in activation magnitude (HFB peak power). Unexpectedly, aggregate ROI analysis revealed that dmPFC produced nonsignificant self/other differences in peak power (Fig. 5e and Table 1). Similarly, single-site analysis showed that most dmPFC sites had nonsignificant self/other differences in peak power (86%), while the remaining ‘self-greater’ and ‘other-greater’ sites did not significantly outnumber each other (Fig. 5a, b). Strikingly, dmPFC did not contain sites that only activated for one mentalizing type. In sum, both mentalizing types recruited identical dmPFC sites (100% overlap) at equivalent intensities, which appears inconsistent with numerous fMRI reports of ‘other-selective’ dmPFC responses.

We instead found robust self/other differences in the timing of dmPFC activations. Specifically, other-mentalizing evoked slower and lengthier activations compared to self-mentalizing (Fig. 5c, d). In other words, dmPFC activity remained significantly above baseline for longer during other-mentalizing (see Figs. 2e and 6a). This suggests a different account of why dmPFC produces stronger hemodynamic responses for mentalizing about others over oneself. The typical story is that dmPFC is highly specialized for thinking about other people’s minds^32,33,34,35. Alternatively, dmPFC could be sensitive to the inherently greater difficulty of other-mentalizing, which may necessitate additional processing cycles before completion. This additional processing may involve ‘anchoring-and-adjustment’; the use of self-representations as an anchor from which to adjust other-mentalizing inferences – a function strongly linked with dmPFC and its role in strategic reasoning^75,82,83. Self-mentalizing may be simplified by the rich compendium of accessible information we have about ourselves, thus resulting in brief but equally intense processing. Given that standard fMRI analysis does not distinguish activation intensity from activation duration, it appears that the latter has been mistaken for the former – though we cannot exclude the possibility of confounds related to differences between ECoG and fMRI²⁶.

Our peculiar vmPFC results may shed light on schematic contributions to mentalizing. Although there is growing consensus that vmPFC is important for schema processing^{32,84,85,86,87,88}, our results appear contrary to this at first blush. We found that vmPFC produced the latest activations across ROIs (Fig. 3e, f), which contradicts numerous reports of early (<200 ms) schema-based ‘gist’ predictions in vmPFC^{89,90,91,92,93,94,95}. Early vmPFC activation has even been observed during mentalizing¹⁷. Of note, many studies have observed ‘double waves’ of early and late vmPFC activity: early activity may reflect vision-based gist construal, while late activity may reflect elaborative situational construal^{93,96,97,98,99,100}. There is some evidence of a ‘double wave’ in the present study, as three posterior vmPFC sites produced very early activations (<200 ms; Figs. 2k and 4a) that met outlier exclusion criteria for ROI analyses. These early vmPFC sites produced equivalent activations for all task conditions (Figs. 2k, 3c and 5a), aligning with characterizations of rapid magnocellular gist processing, which likely cannot discern our alphanumeric stimuli^91,101. In contrast, late-onset vmPFC sites produced longer activations for other- versus self-mentalizing (Figs. 2f and 5c, d), perhaps reflecting other-mentalizing’s greater reliance on schematic feedback, especially in our trait judgment task^{84,102,103,104}. Nevertheless, self- and other-mentalizing recruited near-identical vmPFC sites (91% overlap; Fig. 5a, b), suggesting common schematic underpinnings^32,84,105. Taken together, we propose that vmPFC provides schematic contributions to mentalizing in two ways: rapid predictive processing from coarse visual afferents, followed by slow situational processing involving schematic feedback and integration across DMN.

At the systems level, we found complex and hierarchical processing dynamics across mentalizing’s putative neurocognitive pathway. We observed consistent trial-by-trial propagation of activation onsets across ROIs (Supplementary Fig. 3), portraying an initial ‘feedforward sweep’ of coarser processing^28,106,107 along the pathway. Indeed, onset latencies were insensitive to self/other differences (Table 1). Onsets were followed by considerable concurrent activations across all ROIs (Figs. 2, 3e, 5c, 6 and S2), suggesting that distinct pathway regions largely work together (i.e., recurrent processing^28,108,109) within an overarching processing sequence. Recurrent processing may be crucial for self/other differentiation, as self/other differences in aggregate ROI activity did not reach significance until concurrent activation was achieved across all ROIs (e.g., activation peaks and offsets; Fig. 5). Taken together, mentalizing may be supported by a brief initial ‘feedforward sweep’ of coarser processing along the pathway, followed by substantial recurrent processing that may integrate and refine representations across pathway regions. These dynamics could obscure cross-regional functional distinctions in fMRI studies. However, further research involving connectivity analyses are needed for more conclusive claims.

Temporoparietal DMN regions (tpDMN; TPJ, ATL, and PMC) may help integrate representations throughout mentalizing’s neurocognitive pathway. Classification analysis revealed two distinct functional types (Fig. 4b) that were anatomically interdigitated in tpDMN ROIs: sites with earlier non-specific activations, and sites with later mentalizing-specific activations (Fig. 3c). Intriguingly, non-specific tpDMN sites often coactivated with lower-level regions like visual cortex (e.g., Fig. 2b–d), suggesting attunement to lower-level afferents. Meanwhile, mentalizing-specific tpDMN sites often coactivated with higher-level regions like mPFC (e.g., Fig. 2e, f, h, i), suggesting attunement to higher-level feedback. Critically, we also found lengthy concurrent activations across site types and ROIs (Figs. 2 and 6), which could recurrently integrate low- and high-level representations^73,76,77,110. Indeed, we found that visual and mentalistic representations were simultaneously encoded in all tpDMN ROIs (Supplementary Table 1). Taken together, we propose that interactions between distinct interdigitated neuronal populations in tpDMN^111,112,113 may help integrate and refine distant representations across mentalizing’s neurocognitive pathway.

This study is not without confounds and limitations. Some of these limitations are inherent to ECoG: the use of participants with epilepsy, inconsistent brain coverage across participants, and sampling bias for cortical gyri²⁶. Although these limitations were mitigated to the best of our ability (see Methods), they cannot be completely ameliorated. Thus, our ECoG findings could be corroborated by examining healthy participants with recent advances in source-space EEG/MEG, such as ultra-high density EEG¹¹⁴, optically-pumped MEG¹¹⁵, and laminar source localization¹¹⁶. Another important confound was the sparse right-hemisphere coverage of our cohort, which may limit the interpretability of our ATL and TPJ results¹¹⁷. Nonetheless, our few right-hemisphere sites appear functionally analogous to their left-hemisphere homologs (Supplementary Figs. 6–9). A task-related limitation was the short pre-stimulus baseline (Fig. 1a), which sometimes contained residual activity from prior trials, likely resulting in artifactual ‘deactivations’ in somatomotor sites (Fig. 6a). Another task-related confound was greater prompt length for other-mentalizing (e.g., “My neighbor is…”) versus self-mentalizing (e.g., “I am…”; Supplementary Table 4), which we controlled for using computer vision^36,37 (see Supplementary Methods). Relatedly, mentalizing-specificity (Fig. 3c, d) could arise from differences between sentences and arithmetic equations. However, mentalizing-specific sites were not concentrated in reading-related regions^{118,119,120,121}, but rather in the ‘mentalizing network’ reported by countless fMRI studies¹².

Distributed hierarchical processing is a central organizing principle of neurocognitive systems^28,31,42,109. Characterizing such hierarchies has enabled incisive neuromechanistic accounts of many psychological functions^30,122. Here we provide a comprehensive electrophysiological exploration of the human social brain, revealing that mentalizing is characterized by complex and hierarchical neurocognitive dynamics at millisecond, millimeter, and cross-regional scales. While many questions remain, our findings contribute to a solid foundation upon which more conclusive neurocognitive accounts of mentalizing can be built.

Methods

All research activities herein were conducted in accordance with protocol approved by the Stanford Institutional Review Board for human experimentation. All computational procedures and analyses herein were implemented in MATLAB¹²³ unless otherwise specified.

Participants

We employed a cohort of sixteen human participants while they underwent neurosurgical treatment for drug-resistant epilepsy (demographics in Supplementary Table 2). Each participant provided written informed consent in accordance with the Stanford Institutional Review Board. Participants were not compensated as per Stanford Institutional Review Board guidelines for inpatient clinical research.

As part of their presurgical evaluation, participants were implanted with ECoG at Stanford University Medical Center. The anatomical placement of electrode sites was determined according to each participant’s clinical needs. Participants were included in this study’s cohort if they had electrode coverage in key DMN regions: mPFC, PMC, TPJ, and ATL. Each participant was monitored in hospital for six to ten days prior to surgery, during which the study was conducted.

Behavioral task

ECoG data was recorded while participants performed an event-related behavioral task with six conditions (trial types; Fig. 1a and Supplementary Table 4). Five of these conditions required true/false responses to written prompts, while one condition consisted of cued rest. Two conditions featured mentalizing prompts, either about oneself (e.g., “I am honest”) or others (e.g., “My neighbor is honest”). Participants were instructed to select a single neighbor (current or past) as the target for other-mentalizing. Cognitive task trials consisted of basic arithmetic (e.g., “9 + 86 = 95”). Two conditions featured memory-related prompts: episodic (e.g., “I ate candy yesterday”) and self-semantic (e.g., “I eat a lot of candy”). The cued rest condition required no response and displayed a fixation crosshair for 5-10 seconds. The memory and rest conditions were not relevant to the current analyses have been reported elsewhere^112,113,124. Stimuli were presented in random order and were self-paced, advancing to the next trial after the participants’ response, or up to 15 seconds if no response. The inter-trial interval (ITI) occurred -200-0 ms before each trial. The experiment was broken into two separate runs (mean run duration=12.50 ± 1.64 min). Participants were allowed a short break between behavioral runs. On average, each run featured 25 trials of each sentence condition, 40 cognitive trials, and 36 rest trials. Each non-rest trial contained unique prompts; prompts were not repeated within participant. Responses were made via a handheld keypad using either the ‘1’ (true) or ‘2’ (false) key. Participants were instructed to perform the task as accurately and as quickly as possible. All stimuli were presented in white font on a black background (1200 × 800 pixels) using Psychophysics Toolbox 3¹²⁵.

For all analyses, trials were excluded if meeting any of the following criteria: high-frequency epileptic oscillations, no behavioral responses, irrelevant button presses, or RT_Behav under 400 ms.

Electrocorticography data acquisition

ECoG recordings were obtained via 2125 subdural electrodes (Fig. 1b). Electrodes (platinum plates with diameter of 1.2–2.3 mm) were implanted subdurally onto the cortical surface in grids or strips with center-to-center interelectrode spacing of 4-10 mm (Adtech Medical Instruments). Electrodes were connected to a multichannel recording system (Nihon Kohden; Tucker Davis Technologies) with sampling rate of 1000 Hz or above. Anatomical data was acquired using a GE 3-Tesla SIGNA Magnetic Resonance Imaging (MRI) scanner at Stanford University. A T1-weighted anterior-posterior commissure-aligned pulse sequence was used. T1 data was resampled to 1 mm isotropic voxels, then segmented to distinguish gray and white matter using FreeSurfer¹²⁶. To facilitate electrode localization, postimplant computerized tomography (CT) scans were coregistered to the preoperative MRI anatomical brain volume¹²⁷. For each participant, electrodes sites were localized in BioImage Suite¹²⁸ and displayed on the participants’ own reconstructed 3D cortical surface using the iELVis toolbox¹²⁹. Electrode positions were corrected for postimplantation brain shift, allowing for the accurate anatomical localization of electrodes sites¹³⁰.

Defining regions of interest (ROIs) and brain networks

Each participant’s native-space cortical surface reconstruction (e.g., Supplementary Fig. 2) was used to classify electrode sites into a priori ROIs that are strongly implicated in mentalizing, with visual cortex included as a control ROI (Fig. 1c). Standardized brain-based parcellation was avoided due to known transformation inconsistencies in ECoG¹²⁹. ROIs were defined through FreeSurfer cortical parcellation combined with visual inspection of anatomical landmarks. The ROI for ‘visual cortex’ consisted of occipital cortex, lingual gyrus, posterior fusiform gyrus, and posterior inferotemporal cortex. The ‘ATL’ ROI consisted of a bilateral anterior subregion of temporal cortex with precentral sulcus as the posterior bound, comprising the temporal poles and adjacent sections of entorhinal cortex and superior, middle, and inferior temporal sulci/gyri. The ‘TPJ’ ROI was a bilateral posterior subregion of inferior parietal lobule with lateral sulcus as the anterior bound, comprising angular gyrus and adjacent sections of supramarginal gyrus and superior temporal sulcus/gyrus. The ‘PMC’ ROI consisted of precuneus, posterior cingulate, and retrosplenial cortex. The ‘amPFC’ ROI was an mPFC subregion bounded between the ventral and dorsal reaches of corpus callosum. The ‘dmPFC’ ROI was a mPFC subregion ventrally bounded by the amPFC ROI and posteriorly bounded by the callosal rostrum. The ‘vmPFC’ ROI was an mPFC subregion dorsally bounded by the amPFC ROI and posteriorly bounded by the callosal rostrum, including the medial orbitofrontal surface.

ECoG preprocessing

Preprocessing was performed on a single-site/single-participant basis using custom routines (see Code Availability). First, data were notch filtered for power-line noise (57–63 Hz) and harmonics (117–123 Hz, 177–183 Hz). Electrode sites were discarded from further analyses if they were marked as pathological or ‘noisy’ by postclinical evaluation. The data was then rereferenced by subtracting the mean signal of the remaining electrodes from each electrode’s signal. The rereferenced data underwent time-frequency decomposition into 4–200 Hz spectra in 1–10 Hz bands using 5-cycle Morlet wavelet transforms. The power of the signal in each frequency band was z-transformed across time; this helps correct the 1/frequency decay of neurophysiological signals and improves interpretability. Data was then epoched into trials that were time-locked to stimulus presentation, ranging from 200 ms pre-stimulus to 5000 ms post-stimulus. For each trial and frequency band, baseline correction was performed by subtracting the mean power across the pre-stimulus baseline period (−200–0 ms) from all timepoints within a trial. To reconstruct the high-frequency broadband (HFB) signal, the primary signal of interest, z-transformed power of frequency bands within 70–180 Hz were averaged to produce a single HFB timecourse per electrode. Lastly, HFB timecourses from each electrode were low-pass filtered with a gaussian window (width = 50 ms) for further analysis. Trials were rejected from further analyses if epileptic high-frequency oscillations were observed. Within epochs, timepoints were discarded after presentation of the next trial’s stimulus (i.e., timepoints after RT_Behav + 200 ms ITI).

Statistics

Statistical analyses described below were implemented in MATLAB Statistics and Machine Learning Toolbox¹²³. Data was visualized using the GRAMM toolbox¹³¹. All statistical tests were two-tailed. All multiple comparisons corrections maintained the False Discovery Rate (FDR) under 0.05 through the Benjamini-Yekutieli procedure for data with any dependence structure¹³², with p-values adjusted accordingly (p_FDR).

We primarily used mixed models due to their design flexibility and robustness to sampling heterogeneity, multicollinearity, and statistical confounds¹³³. Specifically, we employed linear mixed-effects modeling (LMEM) for continuous outcomes and logistic mixed-effects classification (LMEC) for binary outcomes. To minimize estimation bias, mixed models used restricted maximum likelihood-based estimators¹³³. To account for full data dependence structure with reduced bias and assumptions, mixed models used unconstrained variance-covariance matrices with log-Cholesky parametrization¹³⁴. To account for heterogenous variances across mixed model terms, Satterthwaite approximation for degrees of freedom (effective DF) was used¹³⁵. To rectify violations of assumptions and overparameterization in mixed models, we evaluated objective function Hessian matrices and ensured positive definiteness¹³⁶. Additionally, non-mixed models were used when specified below.

Within-site analyses

Within-site analyses (Fig. 1d–f) were performed to provide the bases for the primary multi-site analyses. The dependent variable for within-site analyses was z-scored HFB power across timepoints and trials (Fig. 1d). Trials were excluded if containing high-frequency epileptic oscillations, no behavioral responses, irrelevant button presses, or RT_Behav under 400 ms. Timepoint-by-timepoint observations met outlier exclusion criteria if HFB power exceeded three median absolute deviations from other observations of the same timepoint and task condition.

Trial-averaged analysis (within-site)

To identify sites with significant evoked HFB responses for each task condition (p_FDR < 0.05, corrected across timepoints and sites), we used LMEMs (Fig. 1f). The intercept (null distribution) consisted of timepoints within the pre-stimulus baseline (−200–0 ms). Each peri-stimulus timepoint (0–5000 ms) was represented as a separate dummy variable. The intercept was nested within trial to account for trial-specific variance. This LMEM specification estimates mean timecourses of z-scored HFB power (β) for each task condition. To dampen spikes and other noise, timepoints were not considered significant unless p_FDR < 0.05 was maintained for 50 ms consecutively. For each task condition, sites were considered ‘active’ or ‘deactive’ if evoked HFB power was significantly higher or lower, respectively, than pre-stimulus baseline; if sites produced both, the polarity of the greatest deflection was used. Sites with nonsignificant differences from baseline were considered ‘nonresponsive’.

Single-trial analysis (within-site)

To identify timepoints with significant evoked HFB activations for each trial (Fig. 1e), timepoints between stimulus onset and the forthcoming trial (RT_Behav + 200 ms) were run through a sliding window test (width = ±10 ms). Observations (z-scored HFB power) in each sliding window were tested against observations from the pre-stimulus baseline via two-sample Welch’s t-tests, which accounted for unequal variances and sample sizes. This analysis identified timepoints within individual trials that featured significant stimulus-evoked responses (p_FDR < 0.05, corrected across timepoints, trials, and sites) relative to the pre-stimulus baseline preceding each trial. To dampen spikes and other noise, timepoints were not considered significant unless p_FDR < 0.05 was maintained for 50 ms consecutively.

Single-trial analysis provided five key metrics of HFB activity (Fig. 1e). Onset latency is the earliest timepoint with significant activation (green squares). Peak latency and peak power are the timepoint and magnitude, respectively, of the strongest activation (white squares). Offset latency is the latest timepoint with significant activation (red squares). Duration is the total number of timepoints with significant activations (brown areas).

Multi-site analyses

Multi-site analyses used results from within-site analyses as response measures. Outliers were identified with respect to the two-dimensional distance between stimulus onset and RT_Behav using bisquare robust regression¹⁰⁶. For each site and single-trial HFB metric, observations were considered outliers and discarded if residuals were greater than three median absolute deviations (MAD). For ROI sites, this method was applied a second time using observations from all mentalizing-active sites within each ROI. Sites were excluded from ROI analyses if over 50% of its observations exceeded three MAD. Of all ROI sites, only three sites in vmPFC were excluded.

Functional specificity and selectivity (multi-site)

Functional specificity (Fig. 3c, d) of each site was determined by comparing mentalizing (collapsed across self/other) and arithmetic (cognitive task) results from trial-averaged analysis, along with direct comparisons of single-trial peak power (including trials with nonsignificant activations) for mentalizing versus arithmetic using bisquare robust regression. Specifically, sites were considered ‘mentalizing-specific’ if they produced (1) significant trial-averaged activations for mentalizing but not arithmetic (p_FDR < 0.05, corrected across sites and timepoints), and (2) produced greater peak power for mentalizing over arithmetic (p_FDR < 0.05, corrected across sites). Sites with significant trial-averaged coactivations for mentalizing and arithmetic were considered ‘non-specific’, regardless of peak power differences. All mentalizing-active sites with nonsignificant mentalizing/arithmetic peak power differences were labeled as ‘non-specific’ (Mentalizing active = Cognitive active).

Self/other selectivity (Fig. 5a, b) of each site was determined by trial-averaged results and direct comparisons of single-trial peak power (including trials with nonsignificant activations) for self- versus other-mentalizing using bisquare robust regression. Sites were considered ‘self-only’ or ‘other-only’ (not considering cognitive task) if they only produced (1) significant trial-averaged activations for only one mentalizing type (p_FDR < 0.05, corrected across sites and timepoints) and (2) produced greater peak power for that mentalizing type over the other (p_FDR < 0.05, corrected across sites). Sites that activated for both mentalizing types were labeled by self/other differences in peak power: ‘self-greater’ (Self > Other), ‘other-greater’ (Other > Self), and ‘non-selective’ (Self = Other). Sites that activated for only one mentalizing type but had nonsignificant self/other differences in peak power were considered ‘non-selective’ (Self = Other). In sum, self-only and self-greater sites were considered ‘self-selective’, while other-only and other-greater sites were considered ‘other-selective’.

Aggregate ROI analyses (multi-site)

To reveal the spatiotemporal dynamics of mentalizing processing, we analyzed single-trial HFB metrics from mentalizing-active ROI sites. Each ROI and HFB metric was analyzed using separate LMEMs. All LMEMs were nested within Participant, which accounted for within- and between-participant heterogeneity through unconstrained variance/covariance matrices across within-participant random effects. LMEMs that included Site or Trial as nesting factor likewise accounted for site- and trial-related heterogeneity, respectively. Stimulus visual dissimilarity was controlled for by including two visual dimensions (VDs) as random effects specified below; VDs were derived from a popular computer vision model based on the ventral visual stream^36,37 (see Supplementary Methods). In addition to outlier exclusion criteria in the multi-site analyses section above, trials with RT_Behav over 5000 ms were excluded given our 5000 ms epoch lengths.

LMEMs for self/other differences (Fig. 5c–e and Table 1) represented mentalizing type as a fixed and random effect. We also included random effects for RT_Behav and VDs, which controlled for RT_Behav and stimulus visual dissimilarity. All random effects were nested within Site and Participant to account for site- and participant-related heterogeneity.

LMEMs for neurobehavioral associations (Fig. 4d, e and Table 1) simultaneously estimated RT_Behav and Choice_Behav (controlling for one another) as fixed and random effects, while VDs were specified as random effects. All random effects were nested within Site and Participant to account for site- and participant-related heterogeneity.

Pairwise ROI comparisons (Fig. 3f) were performed using LMEMs with ROI as a fixed and random effect nested within Trial within Participant, which estimated within-trial ROI differences and accounted for trial- and participant-related heterogeneity. Additionally, a random effect for RT_Behav was nested within Trial within Participant to control for RT_Behav. Critically, each pairwise comparison was restricted to participants with mentalizing-active sites in both ROIs.

Whole-brain HFB responses within time-windows (multi-site)

To provide a broad overview of the neuronal spatiotemporal dynamics evoked by each task condition, we performed whole-brain analysis of HFB responses within specific time-windows (Fig. 6a). For each site and task condition, separate LMEMs were used to analyze trial-by-trial HFB power (z-scored). The intercept (null distribution) consisted of observations within the pre-stimulus baseline (−200–0 ms). Dummy variables consisted of observations within each time window. All model terms were specified as fixed and random effects nested within Trial, which accounted for trial-related heterogeneity. This specification estimates HFB responses evoked by each task condition within time windows (p_FDR < 0.05, corrected across sites and time windows).

Grand-average ROI timecourses (multi-site)

To summarize task-evoked neuronal dynamics within ROIs, we performed grand-averaged analysis (Fig. 6b–h) of trial-by-trial HFB timecourses (z-scored) from ROI sites that were identified as active or deactive for a given task condition by trial-averaged analysis (see Figs. 1f and 2). Separate LMEMs were used for each ROI and task condition. The intercept (null distribution) consisted of timepoints within the pre-stimulus baseline (−200–0 ms). Peri-stimulus timepoints (0–3000 ms) were represented as dummy variables. Model terms were nested within Site and Participant to account for site- and participant-related heterogeneity. This specification estimates grand-average HFB timecourses evoked by each task condition within ROIs (p_FDR < 0.05, corrected across ROIs and timepoints).

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability

Anonymized preprocessed ECoG data can be shared upon reasonable request to Kevin M. Tan (kevmtan@ucla.edu), subject to a data-sharing agreement between the requestor(s), study authors, and Stanford University. The data-sharing agreement will be tailored to the aims of the requestor(s). Source data are provided with this paper.

Code availability

Custom code is available here: https://github.com/pinheirochagas/lbcn_preproc. Additional information, annotation, and organization of code used here can be provided upon request to Kevin M. Tan (kevmtan@ucla.edu).

References

Lieberman, M. D. Social: Why Our Brains Are Wired to Connect. (Crown, 2013).
Leonhardt, S. D., Menzel, F., Nehring, V. & Schmitt, T. Ecology and Evolution Of Communication In Social Insects. Cell 164, 1277–1287 (2016).
Article CAS PubMed Google Scholar
Fox, K. C. R., Muthukrishna, M. & Shultz, S. The social and cultural roots of whale and dolphin brains. Nat. Ecol. Evol. 1, 1699–1705 (2017).
Article PubMed Google Scholar
Frith, C. D. & Frith, U. The neural basis of mentalizing. Neuron 50, 531–534 (2006).
Article CAS PubMed Google Scholar
Schaafsma, S. M., Pfaff, D. W., Spunt, R. P. & Adolphs, R. Deconstructing and reconstructing theory of mind. Trends Cogn. Sci. 19, 65–72 (2015).
Article PubMed Google Scholar
Lieberman, M. D. Social Cognitive Neuroscience. in Handbook of Social Psychology (American Cancer Society, 2010). https://doi.org/10.1002/9780470561119.socpsy001005.
Friston, K. J. Functional and effective connectivity: A review. Brain Connectivity 1, 13–36 (2011).
Article PubMed Google Scholar
Van Overwalle, F. & Baetens, K. Understanding others’ actions and goals by mirror and mentalizing systems: A meta-analysis. NeuroImage 48, 564–584 (2009).
Article PubMed Google Scholar
Mars, R. B. et al. On the relationship between the “default mode network” and the “social brain”. Front. Hum. Neurosci. 6, 189 (2012).
Article PubMed PubMed Central Google Scholar
Amft, M. et al. Definition and characterization of an extended social-affective default network. Brain Struct. Funct. 220, 1031–1049 (2015).
Article PubMed Google Scholar
Molenberghs, P., Johnson, H., Henry, J. D. & Mattingley, J. B. Understanding the minds of others: A neuroimaging meta-analysis. Neurosci. Biobehav. Rev. 65, 276–291 (2016).
Article PubMed Google Scholar
Schurz, M. et al. Toward a hierarchical model of social cognition: A neuroimaging meta-analysis and integrative review of empathy and theory of mind. Psychol. Bull. 147, 293 (2021).
Article PubMed Google Scholar
Fox, K. C. R., Foster, B. L., Kucyi, A., Daitch, A. L. & Parvizi, J. Intracranial electrophysiology of the human default network. Trends Cogn. Sci. 22, 307–324 (2018).
Article PubMed PubMed Central Google Scholar
Schoffelen, J.-M. & Gross, J. Source connectivity analysis with MEG and EEG. Hum. Brain Mapp. 30, 1857–1865 (2009).
Article PubMed PubMed Central Google Scholar
Vistoli, D., Brunet-Gouet, E., Baup-Bobin, E., Hardy-Bayle, M.-C. & Passerieux, C. Anatomical and temporal architecture of theory of mind: A MEG insight into the early stages. NeuroImage 54, 1406–1414 (2011).
Article PubMed Google Scholar
Van Duynslaeger, M., Van Overwalle, F. & Verstraeten, E. Electrophysiological time course and brain areas of spontaneous and intentional trait inferences. Soc. Cogn. Affect Neurosci. 2, 174–188 (2007).
Article PubMed PubMed Central Google Scholar
Mossad, S. I. et al. Thinking about the thoughts of others; temporal and spatial neural activation during false belief reasoning. NeuroImage 134, 320–327 (2016).
Article PubMed Google Scholar
Wang, H., Callaghan, E., Gooding-Williams, G., McAllister, C. & Kessler, K. Rhythm makes the world go round: An MEG-TMS study on the role of right TPJ theta oscillations in embodied perspective taking. Cortex 75, 68–81 (2016).
Article PubMed Google Scholar
Yuk, V. et al. Do you know what I’m thinking? Temporal and spatial brain activity during a theory-of-mind task in children with autism. Dev. Cogn. Neurosci. 34, 139–147 (2018).
Article PubMed PubMed Central Google Scholar
Zhang, L., Gan, J. Q., Zheng, W. & Wang, H. Spatiotemporal phase synchronization in adaptive reconfiguration from action observation network to mentalizing network for understanding other’s action intention. Brain Topogr. 31, 447–467 (2018).
Article PubMed Google Scholar
Spunt, R. P. & Lieberman, M. D. Dissociating modality-specific and supramodal neural systems for action understanding. J. Neurosci. 32, 3575–3583 (2012).
Article CAS PubMed PubMed Central Google Scholar
Barrett, L. F. & Satpute, A. B. Large-scale brain networks in affective and social neuroscience: Towards an integrative functional architecture of the brain. Curr. Opin. Neurobiol. 23, 361–372 (2013).
Article CAS PubMed PubMed Central Google Scholar
Catmur, C. Understanding intentions from actions: Direct perception, inference, and the roles of mirror and mentalizing systems. Conscious. Cogn. 36, 426–433 (2015).
Article PubMed Google Scholar
Tan, K. M., Burklund, L. J., Craske, M. G. & Lieberman, M. D. Posttraumatic stress disorder and the social brain: Affect-related disruption of the default and mirror networks. Depression Anxiety 36, 1058–1071 (2019).
Article PubMed Google Scholar
Akalin Acar, Z. & Makeig, S. Effects of forward model errors on EEG source localization. Brain Topogr. 26, 378–396 (2013).
Article PubMed PubMed Central Google Scholar
Parvizi, J. & Kastner, S. Promises and limitations of human intracranial electroencephalography. Nat. Neurosci. 21, 474 (2018).
Article CAS PubMed PubMed Central Google Scholar
Mukamel, R. et al. Coupling between neuronal firing, field potentials, and fMRI in human auditory cortex. Science 309, 951–954 (2005).
Article ADS CAS PubMed Google Scholar
Felleman, D. J. & Van Essen, D. C. Distributed hierarchical processing in the primate. Cereb. Cortex. Cereb. Cortex 1, 1–47 (1991).
Article CAS PubMed Google Scholar
Mormann, F. et al. Latency and selectivity of single neurons indicate hierarchical processing in the human medial temporal lobe. J. Neurosci. 28, 8865–8872 (2008).
Article CAS PubMed PubMed Central Google Scholar
Diaconescu, A. O. et al. A computational hierarchy in human cortex. arXiv:1709.02323 [q-bio] (2017).
Hilgetag, C. C. & Goulas, A. ‘Hierarchy’ in the organization of brain networks. Philos. Trans. R. Soc. B: Biol. Sci. 375, 20190319 (2020).
Article Google Scholar
Lieberman, M. D., Straccia, M. A., Meyer, M. L., Du, M. & Tan, K. M. Social, self, (Situational), and affective processes in Medial Prefrontal Cortex (MPFC): Causal, multivariate, and reverse inference evidence. Neurosci. Biobehav. Rev. https://doi.org/10.1016/j.neubiorev.2018.12.021 (2019).
Denny, B. T., Kober, H., Wager, T. D. & Ochsner, K. N. A meta-analysis of functional neuroimaging studies of self and other judgments reveals a spatial gradient for mentalizing in medial prefrontal cortex. J. Cogn. Neurosci. 24, 1742–1752 (2012).
Article PubMed PubMed Central Google Scholar
Bzdok, D. et al. Segregation of the human medial prefrontal cortex in social cognition. Front. Hum. Neurosci. 7, 232 (2013).
Article PubMed PubMed Central Google Scholar
de la Vega, A., Chang, L. J., Banich, M. T., Wager, T. D. & Yarkoni, T. Large-scale meta-analysis of human medial frontal cortex reveals tripartite functional organization. J. Neurosci. 36, 6553–6562 (2016).
Article PubMed PubMed Central Google Scholar
Jarrett, K., Kavukcuoglu, K., Ranzato, M. & LeCun, Y. What is the best multi-stage architecture for object recognition? in 2009 IEEE 12th International Conference on Computer Vision 2146–2153 https://doi.org/10.1109/ICCV.2009.5459469 (2009).
Leeds, D. D., Seibert, D. A., Pyles, J. A. & Tarr, M. J. Comparing visual representations across human fMRI and computational vision. J. Vis. 13, 25–25 (2013).
Article PubMed PubMed Central Google Scholar
Pahl, S. & Eiser, J. R. Valence, comparison focus and self-positivity biases. Exp. Psychol. 52, 303–310 (2005).
Article PubMed Google Scholar
Igou, E. R. ‘How long will I suffer?’ versus ‘How long will you suffer?’ A self-other effect in affective forecasting. J. Personal. Soc. Psychol. 95, 899–917 (2008).
Article Google Scholar
Silani, G., Lamm, C., Ruff, C. C. & Singer, T. Right supramarginal gyrus is crucial to overcome emotional egocentricity bias in social judgments. J. Neurosci. 33, 15466–15476 (2013).
Article CAS PubMed PubMed Central Google Scholar
Margulies, D. S. et al. Situating the default-mode network along a principal gradient of macroscale cortical organization. PNAS 113, 12574–12579 (2016).
Article CAS PubMed PubMed Central Google Scholar
Huntenburg, J. M., Bazin, P.-L. & Margulies, D. S. Large-scale gradients in human cortical organization. Trends Cogn. Sci. 22, 21–31 (2018).
Article PubMed Google Scholar
Apperly, I. A., Samson, D., Chiavarino, C. & Humphreys, G. W. Frontal and temporo-parietal lobe contributions to theory of mind: Neuropsychological evidence from a false-belief task with reduced language and executive demands. J. Cogn. Neurosci. 16, 1773–1784 (2004).
Article PubMed Google Scholar
Samson, D., Apperly, I. A., Chiavarino, C. & Humphreys, G. W. Left temporoparietal junction is necessary for representing someone else’s belief. Nat. Neurosci. 7, 499–500 (2004).
Article CAS PubMed Google Scholar
Saxe, R. & Powell, L. J. It’s the thought that counts: Specific brain regions for one component of theory of mind. Psychol. Sci. 17, 692–699 (2006).
Article PubMed Google Scholar
Saxe, R. The right temporo-parietal junction: a specific brain region for thinking about thoughts. Handbook of Theory of Mind 1–35 (2010).
Schurz, M., Radua, J., Aichhorn, M., Richlan, F. & Perner, J. Fractionating theory of mind: A meta-analysis of functional brain imaging studies. Neurosci. Biobehav. Rev. 42, 9–34 (2014).
Article PubMed Google Scholar
Cook, J. L. Task-relevance dependent gradients in medial prefrontal and temporoparietal cortices suggest solutions to paradoxes concerning self/other control. Neurosci. Biobehav. Rev. 42, 298–302 (2014).
Article PubMed Google Scholar
Spunt, R. P., Meyer, M. L. & Lieberman, M. D. The Default Mode Of Human Brain Function Primes The Intentional Stance. J. Cogn. Neurosci. 27, 1116–1124 (2015).
Article PubMed Google Scholar
Andrews-Hanna, J. R., Smallwood, J. & Spreng, R. N. The default network and self-generated thought: component processes, dynamic control, and clinical relevance. Ann. N. Y Acad. Sci. 1316, 29–52 (2014).
Article ADS PubMed PubMed Central Google Scholar
Saxe, R. & Kanwisher, N. People thinking about thinking people: The role of the temporo-parietal junction in “theory of mind”. NeuroImage 19, 1835–1842 (2003).
Article CAS PubMed Google Scholar
Saxe, R. & Wexler, A. Making sense of another mind: The role of the right temporo-parietal junction. Neuropsychologia 43, 1391–1399 (2005).
Article PubMed Google Scholar
Sass, K., Krach, S., Sachs, O. & Kircher, T. Lion–tiger–stripes: Neural correlates of indirect semantic priming across processing modalities. NeuroImage 45, 224–236 (2009).
Article PubMed Google Scholar
Binder, J. R. & Desai, R. H. The neurobiology of semantic memory. Trends Cogn. Sci. 15, 527–536 (2011).
Article PubMed PubMed Central Google Scholar
Geng, J. J. & Vossel, S. Re-evaluating the role of TPJ in attentional control: Contextual updating? Neurosci. Biobehav. Rev. 37, 2608–2620 (2013).
Article PubMed PubMed Central Google Scholar
Davey, J. et al. Automatic and controlled semantic retrieval: TMS reveals distinct contributions of posterior middle temporal gyrus and angular gyrus. J. Neurosci. 35, 15230–15239 (2015).
Article CAS PubMed PubMed Central Google Scholar
Lewis, G. A., Poeppel, D. & Murphy, G. L. The neural bases of taxonomic and thematic conceptual relations: An MEG study. Neuropsychologia 68, 176–189 (2015).
Article PubMed PubMed Central Google Scholar
Lambon Ralph, M. A., Jefferies, E., Patterson, K. & Rogers, T. T. The neural and computational bases of semantic cognition. Nat. Rev. Neurosci. 18, 42–55 (2017).
Article Google Scholar
Murphy, C. et al. Fractionating the anterior temporal lobe: MVPA reveals differential responses to input and conceptual modality. NeuroImage 147, 19–31 (2017).
Article PubMed Google Scholar
Teige, C. et al. Dynamic semantic cognition: Characterising coherent and controlled conceptual retrieval through time using magnetoencephalography and chronometric transcranial magnetic stimulation. Cortex 103, 329–349 (2018).
Article PubMed PubMed Central Google Scholar
Patel, G. H., Sestieri, C. & Corbetta, M. The evolution of the temporoparietal junction and posterior superior temporal sulcus. Cortex 118, 38–50 (2019).
Article PubMed PubMed Central Google Scholar
Lieberman, M. Seeing minds, matter, and meaning: The CEEing model of pre-reflective subjective construal. Psychol. Rev.129, 1–43 (In press).
Ma, N., Vandekerckhove, M., Overwalle, F. V., Seurinck, R. & Fias, W. Spontaneous and intentional trait inferences recruit a common mentalizing network to a different degree: Spontaneous inferences activate only its core areas. Soc. Neurosci. 6, 123–138 (2011).
Article PubMed Google Scholar
Kestemont, J., Vandekerckhove, M., Ma, N., Van Hoeck, N. & Van Overwalle, F. Situation and person attributions under spontaneous and intentional instructions: an fMRI study. Soc. Cogn. Affect Neurosci. 8, 481–493 (2013).
Article PubMed Google Scholar
Kovács, Á. M., Kühn, S., Gergely, G., Csibra, G. & Brass, M. Are all beliefs equal? Implicit belief attributions recruiting core brain regions of theory of mind. PLOS ONE 9, e106558 (2014).
Article ADS PubMed PubMed Central Google Scholar
Hyde, D. C., Betancourt, M. A. & Simon, C. E. Human temporal-parietal junction spontaneously tracks others’ beliefs: A functional near-infrared spectroscopy study. Hum. Brain Mapp. 36, 4831–4846 (2015).
Article PubMed PubMed Central Google Scholar
Schuwerk, T., Vuori, M. & Sodian, B. Implicit and explicit Theory of Mind reasoning in autism spectrum disorders: The impact of experience. Autism 19, 459–468 (2015).
Article PubMed Google Scholar
Naughtin, C. K. et al. Do implicit and explicit belief processing share neural substrates? Hum. Brain Mapp. 38, 4760–4772 (2017).
Article PubMed PubMed Central Google Scholar
Schneider, D., Slaughter, V. P. & Dux, P. E. Current evidence for automatic Theory of Mind processing in adults. Cognition 162, 27–31 (2017).
Article PubMed Google Scholar
Bardi, L., Six, P. & Brass, M. Repetitive TMS of the temporo-parietal junction disrupts participant’s expectations in a spontaneous Theory of Mind task. Soc. Cogn. Affect Neurosci. 12, 1775–1782 (2017).
Article PubMed PubMed Central Google Scholar
Boccadoro, S. et al. Defining the neural correlates of spontaneous theory of mind (ToM): An fMRI multi-study investigation. NeuroImage 203, 116193 (2019).
Article PubMed Google Scholar
Mitchell, J. P. Mentalizing and Marr: An information processing approach to the study of social cognition. Brain Res. 1079, 66–75 (2006).
Article CAS PubMed Google Scholar
Krol, M. & Krol, M. Is recursive belief inference the engine of mentalizing? J. Neurosci. 30, 15711–15712 (2010).
Article CAS PubMed PubMed Central Google Scholar
Tamir, D. I., Thornton, M. A., Contreras, J. M. & Mitchell, J. P. Neural evidence that three dimensions organize mental state representation: Rationality, social impact, and valence. PNAS 113, 194–199 (2016).
Article ADS CAS PubMed Google Scholar
Tamir, D. I. & Thornton, M. A. Modeling the predictive social mind. Trends Cogn. Sci. 22, 201–212 (2018).
Article PubMed PubMed Central Google Scholar
Cunningham, W. A. & Zelazo, P. D. Attitudes and evaluations: a social cognitive neuroscience perspective. Trends Cogn. Sci. 11, 97–104 (2007).
Article PubMed Google Scholar
Mitchell, C. J., Houwer, J. D. & Lovibond, P. F. The propositional nature of human associative learning. Behav. Brain Sci. 32, 183–198 (2009).
Article PubMed Google Scholar
Lee, S. W., Shimojo, S. & O’Doherty, J. P. Neural computations underlying arbitration between model-based and model-free learning. Neuron 81, 687–699 (2014).
Article CAS PubMed PubMed Central Google Scholar
Doll, B. B., Duncan, K. D., Simon, D. A., Shohamy, D. & Daw, N. D. Model-based choices involve prospective neural activity. Nat. Neurosci. 18, 767–772 (2015).
Article CAS PubMed PubMed Central Google Scholar
Barron, H. C., Dolan, R. J. & Behrens, T. E. J. Online evaluation of novel choices by simultaneous representation of multiple memories. Nat. Neurosci. 16, 1492–1498 (2013).
Article CAS PubMed PubMed Central Google Scholar
Mitchell, J. P. Social psychology as a natural kind. Trends Cogn. Sci. 13, 246–251 (2009).
Article PubMed PubMed Central Google Scholar
Tamir, D. I. & Mitchell, J. P. Neural correlates of anchoring-and-adjustment during mentalizing. Proc. Natl Acad. Sci. 107, 10827–10832 (2010).
Article ADS CAS PubMed PubMed Central Google Scholar
Tamir, D. I. & Mitchell, J. P. Anchoring and adjustment during social inferences. J. Exp. Psychol.: Gen. 142, 151–162 (2013).
Article Google Scholar
Lieberman, M. D., Jarcho, J. M. & Satpute, A. B. Evidence-based and intuition-based self-knowledge: An fMRI study. J. Personal. Soc. Psychol. 87, 421–435 (2004).
Article Google Scholar
Aminoff, E. M. Putting Scenes in Context. (The MIT Press, 2014).
Spalding, K. N., Jones, S. H., Duff, M. C., Tranel, D. & Warren, D. E. Investigating the neural correlates of schemas: Ventromedial prefrontal cortex is necessary for normal schematic influence on memory. J. Neurosci. 35, 15746–15751 (2015).
Article CAS PubMed PubMed Central Google Scholar
Bein, O., Reggev, N. & Tompary, A. Working with schemas, predicting with schemas. J. Neurosci. 38, 1608–1610 (2018).
Article CAS PubMed PubMed Central Google Scholar
Giuliano, A. E., Bonasia, K., Ghosh, V. E., Moscovitch, M. & Gilboa, A. Differential influence of ventromedial prefrontal cortex lesions on neural representations of schema and semantic category knowledge. J. Cogn. Neurosci. 1–28 https://doi.org/10.1162/jocn_a_01746 (2021).
Cardinal, R. N., Parkinson, J. A., Hall, J. & Everitt, B. J. Emotion and motivation: The role of the amygdala, ventral striatum, and prefrontal cortex. Neurosci. Biobehav. Rev. 26, 321–352 (2002).
Article PubMed Google Scholar
Bar, M. et al. Top-down facilitation of visual recognition. Proc. Natl Acad. Sci. USA 103, 449–454 (2006).
Article ADS CAS PubMed PubMed Central Google Scholar
Kveraga, K., Boshyan, J. & Bar, M. Magnocellular projections as the trigger of top-down facilitation in recognition. J. Neurosci. 27, 13232–13240 (2007).
Article CAS PubMed PubMed Central Google Scholar
Olofsson, J. K., Nordin, S., Sequeira, H. & Polich, J. Affective picture processing: An integrative review of ERP findings. Biol. Psychol. 77, 247–265 (2008).
Article PubMed Google Scholar
Barrett, L. F. & Bar, M. See it with feeling: affective predictions during object perception. Philos. Trans. R. Soc. B: Biol. Sci. 364, 1325–1334 (2009).
Article CAS Google Scholar
Kveraga, K. et al. Early onset of neural synchronization in the contextual associations network. PNAS 108, 3389–3394 (2011).
Article ADS CAS PubMed PubMed Central Google Scholar
Guillory, S. A. & Bujarski, K. A. Exploring emotions using invasive methods: Review of 60 years of human intracranial electrophysiology. Soc. Cogn. Affect Neurosci. 9, 1880–1889 (2014).
Article PubMed PubMed Central Google Scholar
Rudrauf, D. et al. Rapid interactions between the ventral visual stream and emotion-related structures rely on a two-pathway architecture. J. Neurosci. 28, 2793–2803 (2008).
Article CAS PubMed PubMed Central Google Scholar
Grabenhorst, F. & Rolls, E. T. Value, pleasure and choice in the ventral prefrontal cortex. Trends Cogn. Sci. 15, 56–67 (2011).
Article PubMed Google Scholar
Wyczesany, M., Grzybowski, S. J. & Kaiser, J. Emotional reactivity to visual content as revealed by ERP component clustering. J. Psychophysiol. 29, 135–146 (2015).
Article Google Scholar
Panichello, M. F., Kveraga, K., Chaumon, M., Bar, M. & Barrett, L. F. Internal valence modulates the speed of object recognition. Sci. Rep. 7, 361 (2017).
Article ADS PubMed PubMed Central Google Scholar
Tan, K. M. & Tarr, M. J. Top-down or bottom-up? The spatiotemporal dynamics of affective object perception differ across valence strength. Poster presented for the Social and Affective Neuroscience Society. https://doi.org/10.13140/RG.2.2.31730.91840 (2016).
Chaumon, M., Kveraga, K., Barrett, L. F. & Bar, M. Visual predictions in the orbitofrontal cortex rely on associative content. Cereb. Cortex 24, 2899–2907 (2014).
Article PubMed Google Scholar
Shenhav, A., Barrett, L. F. & Bar, M. Affective value and associative processing share a cortical substrate. Cogn., Affect., Behav. Neurosci. 13, 46–59 (2013).
Article Google Scholar
FeldmanHall, O. & Shenhav, A. Resolving uncertainty in a social world. Nat. Hum. Behav. 3, 426–435 (2019).
Article PubMed PubMed Central Google Scholar
Raykov, P. P., Keidel, J. L., Oakhill, J. & Bird, C. M. The brain regions supporting schema-related processing of people’s identities. Cogn. Neuropsychol. 37, 8–24 (2020).
Article PubMed Google Scholar
Jenkins, A. C., Macrae, C. N. & Mitchell, J. P. Repetition suppression of ventromedial prefrontal activity during judgments of self and others. PNAS 105, 4507–4512 (2008).
Article ADS CAS PubMed PubMed Central Google Scholar
DiCarlo, J. J. & Maunsell, J. H. R. Using neuronal latency to determine sensory–motor processing pathways in reaction time tasks. J. Neurophysiol. 93, 2974–2986 (2005).
Article PubMed Google Scholar
Alilović, J., Timmermans, B., Reteig, L. C., van Gaal, S. & Slagter, H. A. No evidence that predictions and attention modulate the first feedforward sweep of cortical information processing. Cereb. Cortex 29, 2261–2278 (2019).
Article PubMed PubMed Central Google Scholar
Lamme, V. A. & Roelfsema, P. R. The distinct modes of vision offered by feedforward and recurrent processing. Trends Neurosci. 23, 571–579 (2000).
Article CAS PubMed Google Scholar
Hebb, D. O. The Organization of Behavior: A Neuropsychological Theory (Wiley, 1949).
Fischmeister, F. P., Martins, M. J. D., Beisteiner, R. & Fitch, W. T. Self-similarity and recursion as default modes in human cognition. Cortex 97, 183–201 (2017).
Article PubMed Google Scholar
Parvizi, J., Van Hoesen, G. W., Buckwalter, J. & Damasio, A. Neural connections of the posteromedial cortex in the macaque. PNAS 103, 1563–1568 (2006).
Article ADS CAS PubMed PubMed Central Google Scholar
Foster, B. L., Rangarajan, V., Shirer, W. R. & Parvizi, J. Intrinsic and Task-dependent Coupling Of Neuronal Population Activity In Human Parietal Cortex. Neuron 86, 578–590 (2015).
Article CAS PubMed PubMed Central Google Scholar
Daitch, A. L. & Parvizi, J. Spatial and temporal heterogeneity of neural responses in human posteromedial cortex. PNAS 115, 4785–4790 (2018).
Article CAS PubMed PubMed Central Google Scholar
Robinson, A. K. et al. Very high density EEG elucidates spatiotemporal aspects of early visual processing. Sci. Rep. 7, 16248 (2017).
Article ADS PubMed PubMed Central Google Scholar
Boto, E. et al. On the Potential of a new generation of magnetometers for MEG: A beamformer simulation study. PLOS ONE 11, e0157655 (2016).
Article PubMed PubMed Central Google Scholar
Bonaiuto, J. J. et al. Non-invasive laminar inference with MEG: Comparison of methods and source inversion algorithms. NeuroImage 167, 372–383 (2018).
Article PubMed Google Scholar
Igelstrom, K. M., Webb, T. W., Kelly, Y. T. & Graziano, M. S. A. Topographical organization of attentional, social, and memory processes in the human temporoparietal cortex. eNeuro 3, 6–16 (2016).
Article Google Scholar
Llorens, A., Trébuchon, A., Liegeois-Chauvel, C. & Alario, F.-X. Intra-cranial recordings of brain activity during language production. Front. Psychol. 2, 375 (2011).
Article PubMed PubMed Central Google Scholar
Wu, H. C. et al. Gamma-oscillations modulated by picture naming and word reading: Intracranial recording in epileptic patients. Clin. Neurophysiol. 122, 1929–1942 (2011).
Article PubMed PubMed Central Google Scholar
Miller, K. J., Abel, T. J., Hebb, A. O. & Ojemann, J. G. Rapid online language mapping with electrocorticography. J. Neurosurg. Pediatr. 7, 482–490 (2011).
Article PubMed PubMed Central Google Scholar
Long, L. et al. Feed-forward, feed-back, and distributed feature representation during visual word recognition revealed by human intracranial neurophysiology. https://doi.org/10.21203/rs.3.rs-95141/v1 (2021).
Demirtaş, M. et al. Hierarchical Heterogeneity across Human Cortex Shapes Large-Scale Neural Dynamics. Neuron 101, 1181–1194 (2019).
Article PubMed PubMed Central Google Scholar
MATLAB version R2021a. The MathWorks Inc. (2021),
Raccah, O., Daitch, A. L., Kucyi, A. & Parvizi, J. Direct cortical recordings suggest temporal order of task-evoked responses in human dorsal attention and default networks. J. Neurosci. 38, 10305–10313 (2018).
Article CAS PubMed PubMed Central Google Scholar
Kleiner, M. et al. What’s new in Psychtoolbox-3. Perception 36, 1 (2007).
Google Scholar
Fischl, B. FreeSurfer. Neuroimage 62, 774–781 (2012).
Article PubMed Google Scholar
Hermes, D., Miller, K. J., Noordmans, H. J., Vansteensel, M. J. & Ramsey, N. F. Automated electrocorticographic electrode localization on individually rendered brain surfaces. J. Neurosci. Methods 185, 293–298 (2010).
Article PubMed Google Scholar
Papademetris, X. et al. BioImage suite: An integrated medical image analysis suite: An update. Insight J. 2006, 209 (2006).
PubMed PubMed Central Google Scholar
Groppe, D. M. et al. iELVis: An open source MATLAB toolbox for localizing and visualizing human intracranial electrode data. J. Neurosci. methods 281, 40–48 (2017).
Article CAS PubMed Google Scholar
Dykstra, A. R. et al. Individualized localization and cortical surface-based registration of intracranial electrodes. NeuroImage 59, 3563–3570 (2012).
Article PubMed Google Scholar
Morel, P. Gramm: grammar of graphics plotting in Matlab. J. Open Source Softw. 3, 568 (2018).
Article ADS Google Scholar
Yekutieli, D. & Benjamini, Y. Resampling-based false discovery rate controlling multiple test procedures for correlated test statistics. J. Stat. Plan. Inference 82, 171–196 (1999).
Article MathSciNet MATH Google Scholar
Hox, J. J., Moerbeek, M. & Schoot, R. van de. Multilevel Analysis: Techniques and Applications, Third Edition. (Routledge, 2017).
Barr, D. J., Levy, R., Scheepers, C. & Tily, H. J. Random effects structure for confirmatory hypothesis testing: Keep it maximal. J. Mem. Lang. 68, 255–278 (2013).
Article Google Scholar
Gaylor, D. W. & Hopper, F. N. Estimating the degrees of freedom for linear combinations of mean squares by Satterthwaite’s Formula. Technometrics 11, 691–706 (1969).
Article Google Scholar
Schielzeth, H. et al. Robustness of linear mixed-effects models to violations of distributional assumptions. Methods Ecol. Evolution 11, 1141–1152 (2020).
Article Google Scholar

Download references

Acknowledgements

We are grateful for public financial support from the United States and Canada. This work was supported by National Institute of Neurological Disorders and Stroke Grant R01NS078396 (J.P.), National Institute of Mental Health Grant 1R01MH109954-01 (J.P.), National Science Foundation Grant BCS1358907 (J.P.), Department of Defense Grant 13RSA281 (M.D.L.), National Institute of Child Health and Human Development Postdoctoral Fellowship F32HD087028 (A.L.D.), Stanford University School of Medicine Medical Scholars Research Fellowship (K.C.R.F), Natural Sciences and Engineering Research Council of Canada Postdoctoral Fellowship (K.C.R.F.), and National Science Foundation Graduate Research Fellowship DGE-1650604 (K.M.T.). We thank the Laboratory of Behavioral and Cognitive Neuroscience for generously providing access and support for these data. We thank members of the Social Cognitive Neuroscience Lab for their continued support. We also thank Carolyn Parkinson for her continued support.

Author information

Authors and Affiliations

Social Cognitive Neuroscience Laboratory, Department of Psychology, University of California, Los Angeles, CA, USA
Kevin M. Tan & Matthew D. Lieberman
Laboratory of Behavioral and Cognitive Neuroscience, Department of Neurology and Neurological Sciences, Stanford University, Stanford, CA, USA
Amy L. Daitch, Pedro Pinheiro-Chagas, Kieran C. R. Fox & Josef Parvizi
School of Medicine, Stanford University, Stanford, CA, USA
Kieran C. R. Fox & Josef Parvizi

Authors

Kevin M. Tan
View author publications
You can also search for this author in PubMed Google Scholar
Amy L. Daitch
View author publications
You can also search for this author in PubMed Google Scholar
Pedro Pinheiro-Chagas
View author publications
You can also search for this author in PubMed Google Scholar
Kieran C. R. Fox
View author publications
You can also search for this author in PubMed Google Scholar
Josef Parvizi
View author publications
You can also search for this author in PubMed Google Scholar
Matthew D. Lieberman
View author publications
You can also search for this author in PubMed Google Scholar

Contributions

J.P. and A.L.D. contributed to experimental design and data acquisition. J.P., A.L.D, P.P., and K.M.T. developed preprocessing and analysis tools. K.M.T. and M.D.L. conceptualized the analysis plan. K.M.T. performed formal data analysis. All authors contributed to writing of the manuscript.

Corresponding author

Correspondence to Kevin M. Tan.

Ethics declarations

Competing interests

All authors declare no conflicts of interest.

Peer review

Peer review information

Nature Communications thanks Ziv Williams and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Peer reviewer reports are available.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Supplementary Information

Peer Review File

Reporting Summary

Source data

Source Data

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and permissions

About this article

Cite this article

Tan, K.M., Daitch, A.L., Pinheiro-Chagas, P. et al. Electrocorticographic evidence of a common neurocognitive sequence for mentalizing about the self and others. Nat Commun 13, 1919 (2022). https://doi.org/10.1038/s41467-022-29510-2

Download citation

Received: 19 February 2021
Accepted: 11 March 2022
Published: 08 April 2022
DOI: https://doi.org/10.1038/s41467-022-29510-2

This article is cited by

Reactivity of the ventromedial prefrontal cortex, but not the amygdala, to negative emotion faces predicts greed personality trait
- Kun Deng
- Weipeng Jin
- Qiang Wang
Behavioral and Brain Functions (2023)
Direct intracranial recordings in the human angular gyrus during arithmetic processing
- Pedro Pinheiro-Chagas
- Fengyixuan Chen
- Josef Parvizi
Brain Structure and Function (2023)

Comments

By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.