Neurobiological correlates and attenuated positive social intention attribution during laughter perception associated with degree of autistic traits

Laughter plays an important role in group formation, signaling social belongingness by indicating a positive or negative social intention towards the receiver. In adults without autism, the intention of laughter can be correctly differentiated without further contextual information. In autism spectrum disorder (ASD), however, differences in the perception and interpretation of social cues represent a key characteristic of the disorder. Studies suggest that these differences are associated with hypoactivation and altered connectivity among key nodes of the social perception network. How laughter, as a multimodal nonverbal social cue, is perceived and processed neurobiologically in association with autistic traits has not been assessed previously. We investigated differences in social intention attribution, neurobiological activation, and connectivity during audiovisual laughter perception in association with the degree of autistic traits in adults [N = 31, Mage (SD) = 30.7 (10.0) years, nfemale = 14]. An attenuated tendency to attribute positive social intention to laughter was found with increasing autistic traits. Neurobiologically, autistic trait scores were associated with decreased activation in the right inferior frontal cortex during laughter perception and with attenuated connectivity between the bilateral fusiform face area with bilateral inferior and lateral frontal, superior temporal, mid-cingulate and inferior parietal cortices. Results support hypoactivity and hypoconnectivity during social cue processing with increasing ASD symptoms between socioemotional face processing nodes and higher-order multimodal processing regions related to emotion identification and attribution of social intention. Furthermore, results reflect the importance of specifically including signals of positive social intention in future studies in ASD. Supplementary Information The online version contains supplementary material available at 10.1007/s00702-023-02599-5.


Supplementary participant information
Notes: p-values are given for the corresponding independent sample t-test per variable between groups; AQ percent = percent of maximum attainable score on the Autism-Spectrum Quotient (Baron-Cohen et al., 2001). BDI = Beck Depression Inventory (Beck et al., 1996); MWT-B IQ = calculated IQ value from the Mehrfachwahl-Wortschatz-Intelligenztest (Lehrl, 2005).

Supplementary behavioral methods and results
Attributional biases were operationalized over all participants as the mean individual laughter intention attribution rating correlated with AQ score (Baron-Cohen et al., 2001). Furthermore, an individual bias score was calculated per participant as the regression coefficient representing the change (slope) in answer choice selection with increasing positivity of answer choices from 1 (strongly negative) to 4 (strongly positive). Assuming no answering bias towards or away from more positive or more negative answer choices, a participant's regression coefficient would not differ from zero. The association of answering tendencies with AQ was calculated by correlating participants' regression coefficients with AQ score. Given systematic answering tendencies, such that fewer answer choices with increasing positivity are made, or answer choices of increasing negativity are chosen more often, less positive / more negative slopes would be expected with increasing AQ.

Fig. S1 Bias of social intention attributions and AQ percent scores
Attribution biases operationalized per participant as the slope of change in answer choice frequency with increasing positivity of attributions. More negative regression coefficients indicate lower endorsement of increasingly positive intention attributions. Scatter plot depicts less positive / more negative regression coefficients (y-axis) with increasing AQ scores (x-axis), indicating that with higher autistic trait scores, the reduction in answer choice selection corresponding to more positive attributions became stronger (larger negative slope).
The relationship between individual bias scores and AQ percent scores was significant (Table 1: r = -.35, p = .028, Figure S1), and remained significant when controlling for depressive symptoms (BDI scores) via partial correlation (r = -.35, p = .033). Taking a group approach, comparing participants with and without ASD diagnosis, an independent sample t-test did not show significant group differences in the mean standardized regression coefficient (mean ASD = -0.30; mean without ASD = 0.11; t(29) = 1.01, p = .160). Finally, the correlation between regression coefficients and AQ percent score showed a marginally significant negative correlation of medium effect size both within the ASD diagnosis group (r = -.46, p = .092) as well as in the group without ASD (r = -.31, p = .087). Answer choice selection for the category "strongly positive" differed significantly between groups (mean ASD = 25.7% answer choices; mean without ASD = 16.6% answer choices; t(29) = 2.09, p = .023), while answer choice selection in the remaining answer categories were not significantly different (all t(29) < 1.17, all p > .312).

Localizer Tasks for Psychophysiological Interaction (PPI) Seed Determination
Localizer tasks were performed to determine the peak face sensitive region within the bilateral fusiform gyrus (seed for the fusiform face area, FFA) as well as the peak voice sensitive region within the bilateral superior temporal gyrus (seed for the temporal voice area, TVA). Both localizer tasks were based on established experimental paradigms for face and voice sensitive area determination (voice localizier: Belin et al., 2000; face localizer: Kanwisher et al., 1997, adaptation see Hoffmann et al., 2016;Schwarz et al., 2019) In the face localizer task, eight 30-second blocks of 45 static pictures were shown. Pictures were presented block-wise from each of the following four categories: human faces, everyday objects, landscapes and houses (Kanwisher et al., 1997). Each category was presented in two blocks. Blocks were separated by an interblock interval of 20 seconds, in which a fixation cross was presented in the middle of the screen. Participants completed a one-back task to assure attention. Thereby, a button was to be pressed with the right index finger when the presented picture was identical to the preceding picture. Four regressors of interest were defined, one each for faces, objects, landscapes and houses. The contrast image Faces > (Houses, Objects, Landscapes) was calculated to identify face sensitive regions. The resulting group-level bilateral clusters situated within the fusiform gyrus were extracted as masks for the FFA seeds.
The voice localizer task implemented 24 blocks of sounds from the following three categories: 12 blocks of human vocal sounds, including speech, cries and laughter, 6 blocks of environmental sounds such as screeching tires or church bells, and 6 blocks of animal sounds such as mooing or galloping (Belin et al., 2000). In addition, 12 blocks of eight-second silence were presented. Blocks were randomized with the restriction that two blocks of silence were not presented sequentially. Participants conducted this task with eyes closed as a passive listening task. Three regressors were defined for human vocal sounds, environmental and animal sounds. The contrast Voices > (Environmental Sounds, Animal Sounds) was calculated to identify voice sensitive regions. The resulting group-level bilateral clusters situated on the superior temporal gyrus were extracted as masks for the TVA seeds.
Finally, as a central node of socioemotional saliency recognition, masks for the bilateral amygdalae were extracted from the automated anatomical labeling atlas (AAL) implemented in xjView SPM 8 toolbox (Tzourio-Mazoyer et al., 2002). The six final seed regions are shown in Supplementary Figure S2 and the localizerbased seeds are defined in Supplementary Table S2.

Supplementary neuroimaging methods and results (between-group analyses)
Methods: In addition to the fully dimensional approach according to the spectral characterization of ASD, a supplemental group comparison was also calculated. Hereby, an independent sample t-test was performed between the groups with (n = 10) and without (n = 21) a clinical ASD diagnosis. As in the main analysis, age and sex were controlled for as covariates of no interest. The between-groups comparison was performed for possible differences during neural activation (GLM-approach) as well as in connectivity (PPI-approach) during laughter processing.
Results: Group differences in GLM and PPI are shown in Figure S3 and reported in Table S3.
Participants with ASD showed reduced activation within the left inferior frontal gyrus (IFG) during laughter processing compared to participants without ASD ( Figure S3a). Reduced connectivity between the right FFA and bilateral IFG as well as bilateral supramarginal / angular gyrus and inferior parietal lobe was found in participants with compared to those without ASD (Figures S3b). The pattern of results did not change when including BDI as an additional covariate of no interest, and BDI alone did not explain any unique variance in the neural response.
Discussion: In the overall laughter processing activation, both the dimensional and group-based approaches showed subthreshold bilateral IFG activation differences dependent on AQ percent score or ASD diagnostic status. In the dimensional approach, the reduced activation with increasing AQ scores in the left IFG cluster remained significant after voxel and cluster-wise thresholding, while in the group comparison, the right IFG cluster remained significant after thresholding. Regarding connectivity, reduced right FFA to frontal and parietal cortical regions in ASD remained stable in the group comparison. Thus, the group analyses largely confirm the findings of the dimensional analyses. Overall, these data provide evidence of linear neural correlates of laughter processing spanning a continuum of autistic traits from the nonclinical to clinical domain, independently of depressive symptom severity.

Fig. S3 Group differences in neural activation and connectivity during laughter processing. (a) Reduced
BOLD response during laughter processing in ASD in left inferior frontal gyrus. (b) Reduced connectivity during laughter processing in ASD between the right FFA (seed) and bilateral inferior and middle frontal gyrus and bilateral inferior parietal lobe. Whole-brain statistical thresholding at p < .001, uncorrected at voxel level; FWE corrected for multiple comparisons at cluster level p < .05. Notes: All results represent reduced BOLD response and reduced connectivity in participants with ASD compared to participants without ASD. FFA = fusiform face area. GLM = General linear model, BOLD-based activation approach; PPI = Psychophysiological interaction-based connectivity approach.

Supplementary neuroimaging methods and results (AQ-independent analyses)
Methods: In addition to the AQ-dependent analyses, two supplementary, AQ-independent whole-brain analyses were calculated. The purpose of these analyses was to explore overall attribution and task-related BOLD associations independently of autistic trait level. To this end, the first-level contrasts created for the attribution effect (rating-dependent parametric analysis) and task effect (self vs. other condition) were used.
For the attribution effect, each laughter sequence regressor was parametrically weighted with the participant's demeaned behavioral attribution rating for the respective sequence, with positive weights indicating more positive intention attributions and negative weights indicating more negative attributions. Second level analyses (one-sample t-tests) were conducted on the group level for attribution and task-related contrasts.
Results: AQ-independent activation patterns yielded a significant result related to attribution ratings. More positive intent attributions were associated across all participants with an increase in activation in the ventral anterior cingulate cortex (vACC) during laughter processing ( Figure S4). No task-related AQ-independent BOLD responses were found.

Fig. S4
Autistic trait-independent attribution effect BOLD response in the ventral anterior cingulate cortex during laughter processing was associated with more positive attribution ratings across all participants. This effect was independent of autistic trait score. MNI peak coordinates: x = 0, y = 36, z = -6; tpeak = 6.38, pFWE < .001; cluster size k = 249 voxels. Whole-brain statistical thresholding at p < .001, uncorrected at voxel level; FWE corrected for multiple comparisons at cluster level p < .05.
Discussion: Among all participants, attributions of increasing positive social intention were associated with greater vACC activation. Within the limited literature on the neural correlates of social acceptance, associated neural processing has been localized among lateral frontoparietal regions (Bolling et al., 2013(Bolling et al., , 2012Ethofer et al., 2020). Medial structures including the medial prefrontal (mPFC) and (dorsal) anterior and posterior cingulate (dACC/PCC) cortices, in contrast, are more commonly associated with social pain, activated in paradigms of social exclusion or rejection (Eisenberger and Lieberman, 2003;Vijayakumar et al., 2017;Williams and Jarvis, 2006) as well as during perceived socially excluding laughter (Ethofer et al., 2020). The ventral ACC has been implicated less often in such paradigms, yet has been (negatively) associated with attenuated social pain following social support (Onoda et al., 2009), and with both exclusion as well as fairplay, relative to control, conditions (Bolling et al., 2011). In the discussion surrounding the conceptualisation of the (d)ACC as part of a neural sociometer (Eisenberger et al., 2011), ever more studies report ACC responding to social evaluation and (particularly unexpected) social involvement in both an inclusive (positive) as well as an exclusive (negative) direction (Cheng et al., 2020;Dalgleish et al., 2017;Perini et al., 2018). The current results indicate that this broader, valence independent, processing of perceived self-referential judgements and social saliency may extend ventrally within the cingulate cortex, in particular when active attributional processes regarding the social intention of the social counterpart are involved.