SENSORY OVER-RESPONSIVITY AND ATYPICAL NEURAL RESPONSES TO SOCIALLY RELEVANT STIMULI IN AUTISM

Abstract

Although aversive responses to sensory stimuli are common in Autism Spectrum Disorder (ASD), it remains unknown whether the social relevance of aversive sensory inputs affects their processing. We used functional magnetic resonance imaging (fMRI) to investigate neural responses to mildly aversive nonsocial and social sensory stimuli as well as how sensory over-responsivity (SOR) severity relates to these responses. Participants included 21 ASD and 25 Typically-Developing (TD) youth, aged 8.6–18.0 years. Results showed that TD youth exhibited significant neural discrimination of socially relevant versus irrelevant aversive sensory stimuli, particularly in the amygdala and orbitofrontal cortex (OFC), regions that are crucial for sensory and social processing. In contrast, ASD youth showed reduced neural discrimination of social vs. nonsocial stimuli in the amygdala and OFC, as well as overall greater neural responses to nonsocial compared to social stimuli. Moreover, higher SOR in ASD was associated with heightened responses in sensory-motor regions to socially-relevant stimuli. These findings further our understanding of the relationship between sensory and social processing in ASD, suggesting limited attention to the social relevance compared to aversiveness level of sensory input in ASD vs. TD youth, particularly in ASD youth with higher SOR.

LAY SUMMARY

In this study, we used functional magnetic resonance imaging (fMRI) to examine brain responses to aversive sensory stimuli with and without social relevance in youth with Autism Spectrum Disorder (ASD) and typically-developing (TD) youth. TD youth showed greater neural responses to social compared to nonsocial sensory inputs, whereas the opposite was true in ASD youth. Results suggest that compared to TD youth, ASD youth (particularly those with more severe sensory over-responsivity symptoms) attend more to the level of aversiveness rather than to the social relevance of sensory input.

INTRODUCTION

Individuals with Autism Spectrum Disorder (ASD)¹ often exhibit altered sensory processing. Although the prevalence of sensory processing atypicalities in ASD is quite high, ranging from 69% (Baranek et al., 2006) to 95% (Leekam et al., 2007; Tomchek & Dunn, 2007), ASD research has historically focused more on social cognition (see Leekam, 2016 for review). However, the interest in the interactions between sensory and social difficulties in ASD has been growing (Thye et al., 2018). A few recent fMRI studies have suggested that allocation of attention to socially meaningful information requires actively filtering out extraneous information (e.g., Hernandez et al., 2020) and that sensory distractions affect brain activity during social information processing in ASD children (Green et al., 2018; Patterson et al., 2021). However, much is still not understood about how the sensory and social symptoms in ASD affect or relate to each other. Here, we investigate how social relevance affects brain processing of aversive sensory inputs in ASD.

Sensory over-responsivity (SOR) refers to an exaggerated and aversive response to sensory stimuli. SOR is associated with autistic traits including social communication impairments, suggesting a link between sensory and social information processing in ASD (Schwarzlose et al., 2023; Tavassoli et al., 2014). For instance, busy social environments that include loud, unpredictable, and often human-generated sounds (e.g., restaurants, schools, airports, shopping malls) are commonly recognized as a significant challenge for autistic individuals with SOR (Ben-Sasson et al., 2009). Several fMRI studies have demonstrated that SOR is related to heightened sensory-limbic activation, reduced neural habituation to sensory stimuli, and over-attribution of salience to extraneous information, particularly when multiple modalities of sensory input are presented simultaneously (Green et al., 2015, 2016, 2019). The processing of salient social information is a crucial aspect of human cognition and behavior, yet neurobiological studies of SOR such as the ones described above have mainly focused on nonsocial stimuli. Therefore, the effect of the social relevance of aversive stimuli on attention and over-responsivity is not well understood.

The relationship between sensory processing atypicalities and social functioning has been examined at several hierarchical levels (sensory, perceptual, or attentional; see Thye et al., 2018 for review), and current evidence suggests that aberrant attentional processes could reduce the salience of social information. First, children with ASD have enhanced pitch processing of non-speech sounds (Yu et al., 2015) and preferentially attend to nonsocial objects (Kikuchi et al., 2009). Second, individuals with ASD have diminished capacity to selectively attend to one sound source in the presence of multiple competing sources (Teder-Sälejärvi et al., 2005). Third, autistic individuals show altered connectivity and activation patterns within the salience network, a suite of brain regions involved in selective attention and determining the relative importance of incoming information, (Dichter et al., 2012; Uddin et al., 2013). These atypical patterns of connectivity in the salience network have been linked to reduced engagement with social information and greater SOR in autism (Green et al., 2016; Odriozola et al., 2016). The hub of the salience network, the anterior insula, has also been found to be hypoactive in response to social cognition tasks and hyperactive in response to basic sensory information (Di Martino et al., 2009; Green et al., 2015). Taken together, these findings suggest that social stimuli may be less salient than nonsocial stimuli for youth with ASD due to an overallocation of attention to extraneous sensory input.

Two key brain regions implicated in the processing of both sensory and social information are the amygdala and the orbitofrontal cortex (OFC) (Bachevalier & Loveland, 2006). The amygdala and OFC are involved with detection and adaptive responses to emotional and social significance of stimuli (Easton & Emery, 2005). The amygdala is well-known for its involvement in the formation of social judgements (Adolphs, 2010). The OFC, located in the frontal lobes of the brain, is implicated in the integration of sensory information with emotional and reward-related signals and in guiding social behavior based on the anticipated value of social cues (Baron-Cohen et al., 2000; Dichter et al., 2012). Several functional imaging studies have shown that the amygdala and OFC are highly connected but have reciprocal functions: that is, the amygdala identifies the salience of stimuli, and the OFC adjusts behavioral responses when the salience of stimuli have changed (see Bachevalier & Meunier, 2010 for review). Thus, altered activity of the amygdala and OFC is thought to contribute to ASD-related impairments in processing of salient social information (Tam et al., 2017; Thye et al., 2018). Nevertheless, the role that these brain regions play in atypical differentiation of social and nonsocial properties of sensory stimuli in autism remains unclear.

Here, we compared neural responses to nonsocial aversive sensory stimuli versus similarly aversive but socially relevant (human generated) sounds to determine whether ASD and TD youth show different neural responses depending on the social relevance of the stimuli. Given that real-life sensory environments usually include multiple modalities of sensory information, we were particularly interested in examining differential responses to social versus nonsocial sounds when they were also paired with tactile information, as in our prior studies (e.g., Green et al., 2015; 2019). We expected that ASD youth would have increased neural activity, particularly in the amygdala and OFC, to both social and nonsocial aversive stimuli, compared to TD youth, but that ASD youth would show reduced neural discrimination to these two types of stimuli compared to TD. We focused on the amygdala and the OFC given their importance in sensory, social, and attentional processes. We further hypothesized that within ASD youth, higher SOR would relate to activation in sensory and attention processing regions regardless of social relevance. Identifying differences in neural responses to sensory versus social relevance can inform our understanding of how different types of information are selectively prioritized in ASD and provide valuable insights into the impact of SOR on social functioning.

METHODS

Participants

Participants were 21 children and adolescents with ASD and 25 TD controls (mean age, 14.49 years; [SD] = 2.64; range, 8.6–18.0 years). The initial sample included 29 ASD and 27 TD youth. Six ASD and 1 TD participants were excluded due to motion, with all included participants having a mean absolute motion of <1mm and mean relative motion (using framewise displacement metrics) of <.03mm during the fMRI task. One TD and 2 ASD participants were excluded due to scanner artifacts or errors. All ASD participants had a prior clinical diagnosis of autism spectrum disorder, which was confirmed using the Autism Diagnostic Interview–Revised (ADI-R; Lord et al., 1994) and the Autism Diagnostic Observation Schedule–Generic (ADOS-G; Bishop & Norbury, 2002). Ten ASD participants were taking psychoactive medications (selective serotonin reuptake inhibitors, N=1; psychostimulants, N=3; multiple medications, N=6). The ASD group had significantly greater age, mean absolute motion, and number of outlier motion volumes (i.e., volumes scrubbed) but lower Full-Scale IQ (FSIQ) and Verbal IQ compared to the TD group (Table 1). Therefore, these variables were covaried in all analyses (see fMRI Data Analysis section). Diagnostic groups did not differ significantly in sex, race, ethnicity, or Performance IQ. All study procedures were approved by the University of California, Los Angeles, Institutional Review Board.

Table 1.

Descriptive Statistics

	ASD (mean ± SD)	TD (mean ± SD)	t or χ2
N	21	25
Age (years)	15.73 ± 2.41	13.61 ± 2.29	p<0.001
Sex
males	17	17	p=0.33
females	4	8
Ethnicity
Hispanic or Latino/a	7	7	p=0.621
Not Hispanic or Latino/a	13	18
Unknown/Not reported	1	0
Race
American Indian/Alaska Native	0	0	p=0.461
Asian	2	6
Black or African American	0	0
White	13	12
More than One Race	5	4
Unknown or Not Reported	0	0
WASI full-scale IQ	108.57 ± 14.66	117.56 ± 10.82	p=0.03
WASI verbal IQ	104.38 ± 15.65	117 ± 11.47	p<0.001
WASI performance IQ	110.76 ± 15.36	114.04 ± 10.69	p=0.42
SOR total score (parent-rated)	77.48 ± 26.24	47.32 ± 3.35	p<0.001
Scanner Motion
Mean absolute motion	0.48 ± 0.23	0.35 ± 0.20	p=0.06
Mean relative motion	0.13 ± 0.74	0.12 ± 0.05	p=0.31
Volumes scrubbed	30.76 ± 20.86	19.28 ± 14.96	p=0.04

¹Fisher’s exact test was used to assess independence of variables.

(SD: standard deviation; ASD: Youth with Autism Spectrum Disorder; TD: Typically-Developing Youth; WASI: Wechsler Abbreviated Scale Intelligence; SOR: sensory over-responsivity. Higher SOR scores indicate more severe SOR.)

Behavioral Measures

SOR total scores were calculated by summing the number of SOR items for which parents indicated atypical responses on the Sensory Processing 3-Dimensional (SP3-D) Inventory (Miller et al., 2017; see Table 1). The SP3-D Inventory is a questionnaire in which respondents indicate which of a list of visual, tactile, and auditory stimuli their child is over-responsive to, under-responsive to, or seeks out, although only SOR scores were used for the current study.

MRI Data Acquisition

MRI data were acquired on a Siemens Prisma 3-Tesla MRI scanner. A high-resolution structural T2-weighted echo-planar multi-band imaging volume (spin echo, TR = 5000 ms, TE=60 ms, 208 mm FOV, 72 slices, 2 mm thick) was acquired coplanar with the functional scans to ensure identical distortion characteristics. fMRI scans were collected using an EPI multi-band acquisition and covering the whole cerebral volume (TR = 720 ms, TE = 37 ms, flip angle = 52, 208 mm FOV, 72 slices, voxel size = 2×2×2 mm). Auditory stimuli were presented to the participant using magnet-compatible headphones with active noise cancellation (Optoacoustic OptoActive II ANC) under computer control. Participants wore earplugs underneath to reduce interference of the operating scanner noise. Visual stimuli were presented through MR-compatible goggles (Resonance Technology Inc. VisuaStimDigital).

fMRI Sensory Paradigm

Participants focused on a center fixation cross for 12.5 seconds before, between, and after sensory-evoked tasks. Participants were exposed to the following 15-second stimulus conditions in a counterbalanced block design: three blocks of nonsocial auditory, three blocks of social auditory, four blocks of joint nonsocial (tactile plus nonsocial auditory), four blocks of joint social (tactile plus social auditory), and four blocks of tactile only. Social and nonsocial audio stimuli were matched for aversiveness level based on pilot testing and consisted of sounds of children screaming during play and pink noise (weighted towards lower frequencies), respectively. The OptoAcoustic volume knob was adjusted to a 9 o’clock orientation, ensuring that audio stimuli were presented at a consistent volume across participants. The tactile stimulus was a scratchy bath glove wrapped around the head of a spatula which was brushed onto the participants’ inner left forearm at a rate of one stroke per second. The timing was standardized with a countdown timer viewed by the experimenter. Total scan length was 8 minutes and 28 seconds. Immediately after the sensory paradigm, participants were asked to report on how much each stimulus bothered them, on a scale from 1 to 10 with cartoon faces to anchor the scale. Participants were shown the rating scale prior to the scan to ensure understanding.

Sensory Ratings

To test for group by condition differences in stimulus ratings, a repeated-measures ANOVA was performed with sensory ratings as the dependent variable, group (ASD vs. TD) as a between-subjects factor and Social Relevance (social vs. non-social) and Stimulus Type (auditory vs. joint) as within-subject factors. FSIQ, age, mean absolute motion, and sex were tested as covariates and included in the final models if there were significant effects at p<.1.

fMRI Data Analysis

Analyses were performed using FSL Version 6.0.6 (www.fmrib.ox.ac.uk/fsl). Preprocessing included motion correction to the mean image, spatial smoothing (Gaussian Kernel full width at half maximum=5mm), and high-pass temporal filtering (t>0.01 Hz). Functional data were linearly registered to a common stereotaxic space by first registering to the in-plane T2 image (6 degrees of freedom) then to the MNI152 T1 2mm brain (12 degrees of freedom). FSL’s fMRI Expert Analysis Tool (FEAT), version 6.0, was used for statistical analyses. Fixed-effects models were run separately for each subject, then combined in a higher-level mixed-effects model to investigate within- and between-group differences. Single-subject models for all analyses included 12 motion parameters as covariates. Volumes that were significant outliers for motion were identified using the fsl_motion_outliers tool and regressed out in the single-subject models. Age, mean absolute motion, and FSIQ were included as covariates of no interest in subsequent within- and between-group analyses. Correlations between age and neural responses in the two conditions of focus (i.e., Joint Nonsocial and Joint Social) are presented in Tables S1-2 and Figure S1. While number of outlier volumes also differed between groups, it was significantly correlated with mean absolute motion (r=.69; p<.001), so it was not included to avoid autocorrelation of regressors.

Each experimental condition (auditory social, auditory nonsocial, joint social, and joint nonsocial) was modeled with respect to fixation and social vs. nonsocial conditions were directly compared. Higher-level group analyses were carried out using FSL’s Local Analysis of Mixed Effects State (FLAME 1&2) (Beckmann et al., 2003; M. Woolrich, 2008; M. W. Woolrich et al., 2004). Within- and between- group statistical images for each condition (vs. fixation) were thresholded at Z>2.3 and corrected for multiple comparisons (Gaussian-random field theory based) at p<.05. Between-group comparisons were masked (posthoc) by within-group contrasts at a liberal threshold (Z>1.7; p<.05) to clarify which group drove the between-group differences. Between-condition comparisons were masked (posthoc) by within-condition contrasts at a liberal threshold (Z>1.7; p<.05) to specify greater activation and not less deactivation by the condition of interest (i.e., Joint Nonsocial condition for the Joint Nonsocial>Joint Social contrast). The focus of whole-brain analyses was on the joint conditions which best represent real-world conditions where multiple sensory modalities are present and have been found to best differentiate ASD and TD groups (Green et al., 2015).

OFC and Amygdala Neural Activation

Given a priori interest in the amygdala and OFC as regions of key relevance for sensory reactivity and regulation, we took a region-of-interest (ROI) approach to examine how activation in these regions differed across conditions and diagnostic groups. For each participant, parameter estimates were extracted from the four conditions using right and left OFC and amygdala masks. Masks were selected from the Harvard-Oxford probabilistic atlas (Desikan et al., 2006). To test for group by condition differences in amygdala and OFC activation, we ran two repeated-measure ANOVAs with amygdala and OFC activation as the dependent variables, respectively. Each ANOVA included Group (ASD vs. TD) as a between-subjects factor and Social Relevance (nonsocial vs. social), Stimulus Type (auditory vs. joint), and Laterality (left vs. right) as within-subject factors. Associations between neural activation and SOR severity was also evaluated. FSIQ, age, mean absolute motion, and sex were tested as covariates and included in the final models if significant at p<.1.

Correlation with SOR scores

To evaluate the correlation of SOR with neural activity during each condition, regression analyses were performed with the demeaned SOR total score as the independent variable. These analyses were performed only within the ASD group, due to the TD group having extremely low variability of SOR severity. The comparisons were thresholded at Z>2.3, p<.05. Parameter estimates were extracted from significant clusters for each participant and plotted to verify that the correlations were not driven by outliers. The effect of SOR on group by condition interactions in the amygdala and OFC was examined by entering SOR as a covariate into the repeated-measures ANOVAs previously performed for ROI analyses of neural responses.

RESULTS

Behavioral Results

We found a significant effect of Stimulus Type and a trend-level effect of Group on sensory rating (Table 2), with ASD youth rating sensory stimuli as more aversive than TD youth. Post-hoc analysis showed that both groups reported joint auditory and tactile stimuli as significantly more bothersome than the auditory condition alone; (ASD: (F(1,42)=5.18, p=.03, η_p²=.11), TD youth: (F(1,42)=4.40, p=.04, η_p²=.10)). There were no significant differences in sensory ratings of nonsocial and social stimuli, indicating that participants found both of these stimuli to be similarly aversive.

Table 2.

ANOVA—Sensory Ratings

Analysis	Mean Square	F
Main Effect of Group	71.63	2.84†
Main Effect of Stimulus Type	71.17	9.58*
Group × Stimulus Type	0.53	0.07
Main Effect Social Relevance	7.35	1.75
Group × Social Relevance	0.26	0.06
Group × Social Relevance × Stimulus Type	0.01	0.00

Results of repeated-measures ANOVA predicting sensory aversiveness ratings with one between-subjects factor of Group (ASD vs. TD) and two within-subjects factors of Social Relevance (nonsocial vs. social) and Stimulus Type (auditory vs. joint). (ASD: Youth with Autism Spectrum Disorder; TD: Typically-Developing Youth)

^†p<0.10.

^*p<0.05.

Whole Brain fMRI Results

First, the brain activity elicited by each stimulus condition (Joint Nonsocial and Joint Social) was examined in each diagnostic group separately (Tables 3–4; Figure 1). Second, between-group analyses were conducted to compare activity between ASD and TD groups. Third, between- and within-group analyses using contrasts [Joint Nonsocial>Joint Social] and [Joint Social>Joint Nonsocial] were performed to investigate condition-by-group effects (Table 5; Figure 2). Results for the auditory and tactile conditions are presented in the Supplement (Tables S3-6; Figures S2-4).

Table 3.

Montreal Neurological institute (MNI) coordinates for the Joint Nonsocial condition as compared to fixation

Joint Nonsocial > Fixation
	L/R	Location of peak	Additional regions covered	Voxels	Z-max	x	y	z
ASD	R	Primary auditory cortex	Supramarginal gyrus	4951	6.03	42	−28	20
	L	Cerebellar lobule VI	Cerebellar lobule V	3697	7.71	−26	−44	−30
	L	Primary auditory cortex	Supramarginal gyrus	3689	5.74	−46	−38	16
	R	Postcentral gyrus	Precentral gyrus	1936	4.62	26	−36	56
	R	Cerebellar lobule VI		1747	4.53	30	−46	−38
	L	Brain stem		583	4.99	−10	−34	−38
TD	R	Insular cortex	Primary auditory cortex	2541	5.06	42	−4	−4
	L	Primary auditory cortex	Supramarginal gyrus	1061	4.89	−48	−24	14
	R	Post central gyrus	Primary auditory cortex, Precentral gyrus, Superior parietal lobule	729	4.16	24	−38	72
	L	Cerebellar lobule V	Cerebellar lobule VI	594	5.94	−22	−44	−24
	L	Cerebellar lobule VIIIa	Cerebellar lobule VIIIb	532	4.26	−22	−62	−52
ASD>TD1	L	Lingual gyrus		536	3.65	0	−70	4
	L	Supramarginal gyrus		293	3.68	−68	−42	14
	L	Angular gyrus	Lateral occipital cortex	127	2.99	−40	−52	36
	L	Superior temporal gyrus	Middle temporal gyrus	107	3.58	−68	−32	2
	R	Lingual gyrus	Occipital pole	81	3.39	4	−90	−8
TD>ASD		No significant findings

Note. ‘Regions’ listed in bold are peaks. For large clusters, ‘Additional regions covered’ by the cluster beyond the peak are described. X, y, and z refer to the left-right, anterior-posterior, and inferior-superior dimensions, respectively; Z-max refers to the Z-score at those coordinates (local maxima or submaxima). Voxels indicate cluster size. Within- and between-group analyses were cluster corrected for multiple comparisons, z>2.3, p<.05. Analyses were covaried for age, mean absolute motion, and full-scale IQ.

¹Masked by activation in the ASD group from the [Joint Nonsocial>Fixation] contrast at z>1.7, p<.05 to identify ASD>TD group differences only in clusters that showed positive activation in the ASD group.

(ASD: Youth with Autism Spectrum Disorder; TD: Typically-Developing Youth)

Table 4.

Montreal Neurological institute (MNI) coordinates for Joint Social condition as compared to fixation

Joint Social > Fixation
	L/R	Location of peak	Additional regions covered	Voxels	Z-max	x	y	z
ASD	R	Primary auditory cortex	Superior temporal gyrus, Supramarginal gyrus	4569	7.2	44	−26	8
	L	Primary auditory cortex	Superior temporal gyrus	4055	7.67	−32	−34	14
	R	Postcentral gyrus	Precentral gyrus, Superior parietal lobule	1923	5.06	22	−44	68
	L	Cerebellum lobule VIIIa	Cerebellum lobules VIIb, VIIIb	780	5.52	−20	−64	−50
	L	Cerebellum lobule VI	Cerebellum lobule V	486	5.2	−30	−44	−28
TD	L	Cerebellum lobule V	Cerebellum lobule VI, I-IV	9824	7.54	−18	−46	−24
	R	Primary auditory cortex				36	−22	6
	R	Thalamus				18	−20	12
	L	Primary auditory cortex	Orbital frontal cortex, Frontal pole, Frontal medial cortex, Superior frontal gyrus, Superior temporal gyrus, Supramarginal gyrus	8683	7.06	−50	−24	10
	L	Insula			3.99	−40	−4	−8
	R	Precentral gyrus	Postcentral gyrus, Superior parietal lobule	890	4.29	26	−20	66
	L	Cerebellar Crus II	Cerebellum lobule VIIIb	878	3.85	−12	−72	−36
	R	Cerebellar Crus I		675	3.68	26	−94	−32
ASD>TD		No significant findings
TD>ASD1	L	Frontal pole	Orbital frontal cortex, Superior frontal gyrus, Inferior frontal gyrus	2682	4.52	−42	40	−18
	R	Frontal medial cortex		22	2.95	10	46	−20

Note. ‘Regions’ listed in bold are peaks; those listed in italics are subpeaks within the same cluster as the coordinates above them. For large clusters, ‘Additional regions covered’ by the cluster beyond the peak are described. X, y, and z refer to the left-right, anterior-posterior, and inferior-superior dimensions, respectively; Z-max refers to the Z-score at those coordinates (local maxima or submaxima). Voxels indicate cluster size. Within- and between-group analyses were cluster corrected for multiple comparisons, z>2.3, p<.05. Analyses were covaried for age, mean absolute motion, and full-scale IQ.

¹Masked by activation in the TD group from the [Joint Social>Fixation] contrast at z>1.7, p<.05 to identify TD>ASD group differences only in clusters that showed positive activation in the TD group.

(ASD: Youth with Autism Spectrum Disorder; TD: Typically-Developing Youth)

Figure 1.

Within- and between-group results for the Joint Nonsocial and Joint Social conditions with simultaneous tactile stimulation to the left arm. Group comparisons were masked with within-group results at z>1.7.

Table 5.

Montreal Neurological institute (MNI) coordinates for a) Joint Nonsocial condition as compared to Joint Social condition and b) Joint Social condition compared to Joint Nonsocial condition

a) Joint Nonsocial > Joint Social
	L/R	Location of peak	Additional regions covered	Voxels	Z-max	x	y	z
ASD1	L	Cerebellar Crus I	Cerebellar Crus II	1054	4.55	−52	−62	−28
	R	Cerebellar Crus I	Cerebellar Crus II	459	4.1	38	−66	−32
	L	Supramarginal gyrus	Lateral occipital cortex	445	4.26	−48	−64	48
	R	Lingual gyrus	Intracalcarine cortex, Occipital pole, Precuneus cortex	352	4.37	2	−76	0
	L	Frontal Pole	Orbital frontal cortex	191	3.64	−50	38	−6
	L	Insular cortex	Putamen	110	4.22	−32	12	−12
	L	Middle temporal gyrus	Inferior temporal gyrus	109	4.33	−58	−34	−20
	R	Brainstem		95	3.39	6	−38	−48
TD		No significant findings
ASD>TD2	L	Supramarginal gyrus	Angular Gyrus	337	3.75	−38	−50	32
	L	Cerebellar Left VIIIa	Cerebellar Left VIIb, Crus II	323	3.48	−32	−44	−46
	L	Lingual gyrus	Precuneus cortex	263	3.67	0	−60	8
	L	Insular cortex	Planum polare, Amygdala	248	4.27	−36	−2	−20
	L	Orbital frontal cortex	Frontal pole, Inferior temporal gyrus	239	3.78	−50	28	−8
	R	Planum polare	Hippocampus, Amygdala	201	4.28	42	−12	−12
	L	Middle temporal gyrus		175	3.95	−68	−38	−12
	L	Cerebellar Crus I		134	3.33	−50	−66	−32
	R	Occipital pole		120	3.71	12	−94	−16
	R	Brain stem	Cerebellar lobule IX	101	4.03	8	−44	−44
	R	Cingulate gyrus		39	3.66	6	−40	6
	L	Hippocampus		31	4.33	−24	−16	−18
	L	Inferior temporal gyrus		27	3.51	−48	−40	−14
	L	Middle temporal gyrus		14	3.01	−62	−44	0
TD>ASD		No significant findings
b) Joint Social > Joint Nonsocial
	L/R	Location of peak	Additional regions covered	Voxels	Z-max	x	y	z
ASD3	L	Anterior superior temporal gyrus	Primary auditory cortex	1508	5.76	−58	−6	0
	R	Posterior superior temporal gyrus	Primary auditory cortex, Planum temporale	1492	7.23	54	−22	4
TD4	R	Superior temporal gyrus	Primary auditory cortex	2886	5.48	68	−16	4
	L	Posterior superior temporal gyrus	Primary auditory cortex	2767	5.74	−52	−24	8
	L	Inferior temporal gyrus	Insular cortex	580	4.34	−34	6	−28
	L	Frontal medial cortex	Paracingulate gyrus	555	4.51	−6	38	−22
	L	Frontal pole		484	4.6	−14	64	14
	L	Hippocampus	Amygdala	54	3.64	−24	−16	−18
ASD>TD		No significant findings
TD>ASD5	L	Orbital frontal cortex		827	4.98	−42	44	−20
	L	Frontal Pole		691	4.49	−10	64	10
	L	Middle temporal gyrus		347	3.97	−68	−40	−8
	L	Cingulate gyrus		256	4.01	−8	34	−6
	R	Superior frontal gyrus		232	4.27	22	58	22
	L	Hippocampus	Amygdala	67	4.69	−26	−16	−16
	R	Lingual gyrus		42	3.71	12	−38	−4
	L	Inferior frontal gyrus		27	3.19	−40	22	14
	R	Orbital frontal cortex		21	3.26	24	8	−18
	R	Brain stem		16	3.14	2	−18	−40
	L	Subcallosal cortex	Frontal medial cortex	15	3.33	−4	26	−24

Note. ‘Regions’ listed in bold are peaks. For large clusters, ‘Additional regions covered’ by the cluster beyond the peak are described. x, y, and z refer to the left-right, anterior-posterior, and inferior-superior dimensions, respectively; Z-max refers to the Z-score at those coordinates (local maxima or submaxima). Voxels indicate cluster size. Within and between-group analyses were cluster corrected for multiple comparisons, z>2.3, p<.05. Analyses were covaried for age, mean absolute motion, and full-scale IQ.

¹Masked by activation in the ASD group from the [Joint Nonsocial>Fixation] contrast at z>1.7, p<.05 to identify significant differences in the Nonsocial compared to Social conditions only within clusters that were activated in the ASD Joint Nonsocial condition.

²Masked by activation in the ASD group from the [Joint Nonsocial>Fixation] and [Joint Nonsocial>Joint Social] contrasts at z>1.7, p<.05 to identify ASD>TD group by condition differences only in clusters that showed positive activation in the ASD group during Joint Nonsocial compared to Joint Social conditions.

³Masked by activation in the ASD group from the [Joint Social>Fixation] contrast at z>1.7, p<.05 to identify significant differences in the Social compared to Nonsocial conditions only within clusters that were activated in the ASD Joint Social condition.

⁴Masked by activation in the TD group from the [Joint Social>Fixation] contrast at z>1.7, p<.05 to identify significant differences in the Social compared to Nonsocial conditions only within clusters that were activated in the TD Joint Social condition.

⁵Masked by activation in the TD group from the [Joint Social>Fixation] and [Joint Social>Joint Nonsocial] contrasts at z>1.7, p<.05 to identify TD>ASD group by condition differences only in clusters that showed positive activation in the TD group during Joint Social compared to Joint Nonsocial conditions.

(ASD: Youth with Autism Spectrum Disorder; TD: Typically-Developing Youth)

Figure 2.

Within-group and between-group results for the neural activation during the Joint Nonsocial compared to the Joint Social condition.

Nonsocial condition.

During joint white noise and tactile stimulation, both groups exhibited greater in bilateral primary auditory cortices and associated regions, right primary motor and sensory cortices, and left cerebellum compared to fixation. In addition, the ASD group further activated left brain-stem and right cerebellum while the TD group further activated right insula and superior parietal lobule. Between-group analyses indicated that the ASD group showed significantly greater activation than the TD group within the visual cortex.

Social condition.

When sounds of children screaming on a playground were presented with tactile stimulation, both groups showed greater activation in bilateral primary auditory cortices and associated regions, right primary motor and sensory cortices, right superior parietal lobule, and left cerebellum compared to fixation. The TD group had additional activation in right cerebellum. A between-group comparison showed that the TD group had significantly greater activation in multiple frontal regions within the left hemisphere (i.e., medial and orbital frontal cortices as well as superior and inferior frontal gyri) compared to the ASD group.

Nonsocial vs. Social comparisons.

[Joint Nonsocial>Joint Social] contrasts were masked by regions that were significantly activated in the within-group Joint Nonsocial condition analysis at Z>1.7. Similarly, [Joint Social>Joint Nonsocial] were masked by regions of significant activation in the Joint Social condition at Z>1.7.

The ASD group showed greater activation to the Joint Nonsocial compared to Joint Social condition in bilateral visual and temporal cortices, left frontal pole, OFC, insula, putamen, bilateral cerebellum, and right brainstem. In the TD group, no brain regions activated more to the Joint Nonsocial compared to the Joint Social condition.

In the [Joint Social>Joint Nonsocial] contrast, both ASD and TD groups had greater activation in bilateral auditory cortices. Additionally, the TD group showed greater activation in left frontal pole, OFC, paracingulate gyrus, insular cortex, hippocampus, and amygdala.

Condition-by-group interactions.

The ASD group had greater activation for the Nonsocial versus Social stimuli, compared to the TD group, in bilateral visual cortex, OFC, right cingulate gyrus, bilateral hippocampus and amygdala, left insular cortex, and right brain-stem. The TD group showed greater activation for the Social versus Nonsocial stimuli, compared to the ASD group, in several frontal regions including bilateral frontal pole and OFC, left inferior frontal and cingulate gyri, subcallosal cortex, thalamus, and hippocampus, as well as right visual cortex and brain stem. Taken together, youth with ASD exhibited greater neural responses during nonsocial stimuli compared to social stimuli while the opposite pattern was observed for the TD group.

ROI Analysis Findings:

Repeated-measures ANOVAs with Group (ASD vs. TD) as a between-subjects factor and Social Relevance (nonsocial vs. social), Stimulus Type (auditory vs. joint), and Laterality (left vs. right) as within-subject factors were conducted for neural activation in two a priori regions of interests (amygdala and OFC). ANOVA statistics are reported in Table 6 and interaction effects are depicted in Figure 3.

Table 6.

ANOVA—Amygdala and OFC activity during sensory tasks

	Amygdala		OFC
Covariate	FSIQ		Age
Analysis	Mean Square	F	Mean Square	F
Main Effect of Group	0.01	0.27	0.03	0.57
Main Effect of FSIQ	0.21	4.87*	0.01	0.19
Main Effect of Age	0.15	3.52†	0.04	0.73
Main Effect Social Relevance	0.09	3.59	0.95	3.3†
Age × Social Relevance	0.01	0.25	0.12	4.29*
Group × Social Relevance	0.17	6.75**	0.16	5.5*
Main Effect of Social Type	0.08	1.20	0.01	0.06
Group × Social Type	0.01	0.17	0.04	0.51
Main Effect of Laterality	0.02	3.76†	0.00	0.04
Group × Laterality	0.01	2.26†	0.01	1.02
Group × Laterality × Social Relevance	0.00	0.78	0.03	4.30*
Main Effect of SOR	0.00	0.00	0.00	0.00
SOR × Social Relevance	0.00	0.04	0.01	0.20

Results of repeated-measures ANOVA of amygdala and orbitofrontal cortex (OFC) activation with one between-subjects factor (Group, ASD vs. TD) and three within-subjects factors of Social Relevance (nonsocial vs. social), Stimulus Type (auditory vs. joint), and Laterality (left vs. right).

^†p<0.10.

^*p<0.05.

^**p<0.01.

Figure 3.

Results from group by condition repeated-measures ANOVA predicting responses in a) amygdala and b) orbitofrontal cortex (OFC) for the ASD and TD group during the Nonsocial and Social conditions. ANOVA results indicated a group by condition interaction showing that the TD group had higher amygdala and OFC responses during the Social condition compared to the Nonsocial condition, whereas the ASD group did not show significant differences between the conditions. Individual points indicate each participant’s mean activation averaged across left and right hemispheres and auditory and joint conditions. Boxplots denote 25^th percentiles, medians, and 75^th percentiles of each distribution; vertical extending lines show the range of values; rhombuses denote the mean values.

(ASD: Youth with Autism Spectrum Disorder; TD: Typically-Developing Youth).

Amygdala.

We found a significant interaction between group and stimulus relevance on amygdala activation (F(1,43)=6.75, p=.01, η_p²=.14). Pairwise comparisons showed that while the TD group had significantly greater activation for social compared vs. nonsocial conditions (F(1,43)=6.16, p=.02, η_p²=.13), the ASD group showed no significant differences in amygdala activation between the two conditions (F(1,43)=1.84, p=.18, η_p²=.04). FSIQ was included in the model as a covariate due to its main effect on amygdala activation with greater FSIQ related to lower amygdala activation (F(1,43)=4.87, p=.03, η_p²=.10).

OFC.

There was a significant interaction between group and social relevance on OFC activation (F(1,43)=5.54, p=.02, η_p²=.11). In addition, there was a significant three-way interaction between group, social relevance, and laterality of OFC activity (F(1,43)=4.30, p=.04, η_p²=.10), indicating that the group by social relevance interaction was mainly driven by the left OFC. Pairwise comparisons indicated that, similarly to the amygdala results, TD OFC activation during nonsocial conditions was significantly lower than during social conditions (F(1,43)=7.95, p=.007, η_p²=.16), but the ASD group did not show significant differences in OFC activation in nonsocial vs. social conditions (F(1,43)=.59, p=.45, η_p²=.01). Age was included in the model as a covariate due to its significant interaction with social relevance in OFC activation (F(1,43)=4.29, p=.04, η_p²=.10). Specifically, age was positively correlated with OFC activation during the social condition (F(1,43)=4.06, p=.05, η_p²=.09). There were no other significant main effects or interactions.

Correlation With SOR Scores:

For the ASD group, regions where activation was significantly correlated with SOR severity in joint conditions are reported in Table 7 and depicted in Figure 4. During the joint nonsocial condition, SOR was negatively associated with activation in left insular cortex and putamen. During the joint social condition, SOR was associated positively with left cerebellum, bilateral lateral occipital cortex and superior parietal lobule, and left precuneus, as well as bilateral postcentral and left precentral gyri. Results from the auditory-only conditions are presented in Table S7 and Figure S5.

Table 7.

Montreal Neurological institute (MNI) coordinates for brain regions where neural activation patterns in ASD youth were significantly correlated with SOR total score during Joint conditions.

Joint Nonsocial
	L/R	Location of peak	Additional regions covered	Voxels	Z-max	x	y	z
SOR+		No significant findings
SOR-	L	Insular cortex	Putamen	588	3.75	−38	−8	8
Joint Social
	L/R	Location of peak	Additional regions covered	Voxels	Z-max	x	y	z
SOR+	L	Superior parietal lobule	Lateral occipital cortex, Postcentral gyrus, Precuneus cortex, Supramarginal gyrus	1718	4.34	−20	−72	60
	R	Superior parietal lobule	Lateral occipital cortex, Postcentral gyrus, Precentral Gyrus, Supramarginal gyrus	1098	4.39	32	−54	68
	R	Middle temporal gyrus	Inferior temporal gyrus, Temporal occipital fusiform cortex	531	4.17	62	−32	−28
	L	Middle frontal gyrus	Precentral gyrus	379	3.73	−40	2	40
	L	Cerebellar lobule IX	Cerebellar lobule VIIIb	355	3.9	−12	−52	−54
SOR-		No significant findings

Note. ‘Regions’ listed in bold are peaks. For large clusters, ‘Additional regions covered’ by the cluster beyond the peak are described. x, y, and z refer to the left-right, anterior-posterior, and inferior-superior dimensions, respectively; Z-max refers to the Z-score at those coordinates (local maxima or submaxima). Voxels indicate cluster size. SOR correlational analyses were cluster corrected for multiple comparisons at a threshold of z>2.3, p<.05.

(ASD: Autism Spectrum Disorder; SOR+: coordinates where neural activation was positively correlated with sensory over-responsivity; SOR-: coordinates where neural activation was negatively correlated with sensory over-responsivity)

Figure 4.

Areas of signal change that were correlated with sensory over-responsivity (SOR) scores in ASD youth. Scatter plots show parameter estimates (PE) which were extracted from significant clusters where brain activation was significantly correlated with SOR and plotted to demonstrate that they were not driven by outliers.

Multivariate analyses showed no significant main effects of SOR or interactions between social relevance of stimuli and SOR scores for amygdala activation (F(1,18)=.04, p=.84, η_p²=.02) or OFC activation (F(1,18)=0.20, p=.66, η_p²=.01).

DISCUSSION

Sensory processing challenges and difficulty interpreting social information are common characteristics of ASD. Yet, how stimulus aversiveness interacts with social relevance to affect information processing in ASD is not well understood. In this study, we used fMRI to investigate how responses to aversive sensory stimuli differ in ASD and TD youth based on the social relevance of the stimuli. Overall, we found that TD youth showed greater activation to social compared to nonsocial aversive stimuli, suggesting that TD youth may be more sensitive to the social relevance of the stimuli compared to ASD youth who may instead be more responsive to the overall aversive qualities of the sensory experience.

Our results confirm previous findings that youth with ASD exhibit unique neural processing patterns during sensory stimulation (Green et al., 2013, 2015, 2019) and extend these findings to show additional distinctions in the context of socially relevant stimuli. At the whole brain level, ASD youth had greater brain activation in response to nonsocial stimuli (white noise paired with a brush on the arm) compared to TD youth, consistent with our earlier findings (Green et al., 2015; 2019). These heightened neural responses to nonsocial stimuli in ASD youth were notably characterized by hyperactivation of the insula and visual cortex. The insula, a hub of the salience network, plays an important role in detection of salience and emotional significance (Menon & Uddin, 2010; Seeley, 2019) and has previously been found to be hyperactive in ASD during sensory processing (Odriozola et al., 2016; Patterson et al., 2021). Activation of visual cortex during a variety of auditory, perceptual, and cognitive tasks in youth with ASD may reflect difficulty in disengaging visual processing and shifting attention to other sensory modalities (Green et al., 2019; Keehn et al., 2013, 2021; Samson et al., 2012). Our findings are consistent with impaired downregulation of visual cortex in ASD during non-visual stimulation, suggesting reduced distinction of sensory networks, which could lead to less efficient processing of sensory stimuli (Joanne et al., 2017). Additionally, the TD group had greater neural responses compared to the ASD group during socially relevant aversive stimuli (children screaming in play paired with the same brush on the arm), particularly in frontal regions. This finding is in line with prior findings wherein TD youth generally prioritize social over nonsocial sensory information and show greater neural activation in frontal brain regions important to social information processing when sensory information is social in nature (Green et al., 2018).

Whole-brain interaction analyses indicated that the ASD group showed greater responses than the TD group to nonsocial compared to social stimuli whereas the TD group showed the opposite pattern. These results support existing literature that underscores enhanced attention to nonsocial information in ASD (Elison et al., 2012; Gong et al., 2021; Stavropoulos et al., 2018) and extend it by demonstrating that mildly aversive nonsocial information takes precedence over comparably aversive social information in terms of neural response and resource allocation. Specifically, compared to TD youth, ASD youth had greater engagement of regions involved in sensory, emotional, and cognitive processing during nonsocial compared to social stimuli. Increased activity in the visual cortex and hippocampus in ASD youth suggests a reliance on visual and memory-related processes to interpret sensory input devoid of a social context. Additionally, greater activity in the insular cortex and amygdala, which play an important role in salience detection, attention, and interpretation of sensory inputs (Menon & Uddin, 2010; Šimić et al., 2021; Zhang et al., 2022), supports our hypothesis that youth with ASD may attribute greater significance to nonsocial rather than social aversive stimuli. In contrast, the TD group showed greater neural activation in response to socially relevant compared to nonsocial stimuli in regions related to emotional and cognitive processing. These regions, including the amygdala and OFC, are involved in emotional and reward processing and social cognition (Adolphs, 2010) while the frontal pole is implicated in cognitive control (Koechlin, 2011), suggesting that TD assign more significance to socially relevant sensory information. This prioritized engagement with socially relevant information in TD youth may be important to interpreting and responding effectively to complex social cues.

Region-of-interest analyses, focused on the amygdala and OFC, two regions that are highly important to both SOR and social information processing, were consistent with whole-brain analyses in showing modulation of amygdala and OFC activity in the TD group, with heightened responses to socially relevant stimuli. Notably, there was a significant interaction effect indicating that, in these two brain regions, the TD group discriminated social relevance more than the ASD group, which did not show significant between-condition differences in activation of these regions. These ROI findings support our hypothesis that TD individuals efficiently recruit the amygdala and OFC to assess and discriminate social relevance, whereas the salient aspects of incoming sensory information may differ for individuals with ASD. That is, the aversiveness of stimuli may be prioritized for processing in ASD over and above the social significance of the information. This may affect the development of ability to engage in social interactions in youth with ASD. The association between older age and greater OFC, and not amygdala, activation during social conditions highlights that the engagement of prefrontal regions important to sensory regulation and discrimination of social relevance may increase with age (Cakar et al., 2023a). As such, these findings emphasize the importance of developmental considerations and the need for longitudinal studies to capture change over time in neural responses to sensory stimuli.

Correlational analyses between SOR severity and neural activation provided insight into how heterogeneity within ASD youth relates to processing nonsocial versus social aversive stimuli. SOR severity was negatively related to activation in regions related to sensory interpretation and integration during the joint nonsocial condition, suggesting that individuals with more severe SOR may show reduced higher-level processing of aversive sensory information. It could also suggest a more adaptive response in individuals with lower SOR, who may be more able to integrate multiple types of sensory information for accurate interpretation (e.g., Green et al., 2018). In contrast, the positive association between SOR severity and activation in sensory, motor, and higher-order cognitive regions during the social condition likely reflects the intensified primary sensory processing response commonly exhibited in youth with greater SOR severity (Green et al., 2015, 2019), and could suggest more attention being paid to the primary sensory aspects of the stimuli rather than the social relevance, which is consistent with the reduced neural discrimination seen in the ASD group.

Importantly, both ASD and TD groups rated the social and nonsocial stimuli as equally aversive, indicating that there were no differences in ratings of condition within either group, although the ASD group overall rated both stimuli slightly higher than the TD group. Thus, the group by condition interactions in neural activation may be driven by attentional differences rather than differential experiences of aversiveness between the two conditions. This result also suggests that social stimuli may not be inherently more aversive for individuals with ASD when the aversiveness of basic sensory qualities is held constant (see Clements et al., 2018), though more research is needed with a broader range of social stimuli to build on the current findings. Additionally, both groups rated joint stimuli (auditory plus tactile) as more aversive than auditory alone, consistent with prior findings that neural over-activation in ASD versus TD is more pronounced during joint than during single modality stimulation (Green et al., 2015, 2019). Limited group differences to auditory stimuli alone may be partly due to the loud MRI environment which could dampen the contrast between the auditory condition and fixation. Furthermore, joint conditions better represent everyday sensory experiences which usually involve multiple sensory modalities.

Limitations

While the present study provides valuable insights into the neural correlates of processing socially relevant stimuli in ASD, it does have some limitations. First, the sample size was relatively small, which could limit the generalizability of findings to broader populations, and results should be replicated in a larger sample. Notably, the ASD and TD samples significantly differed on IQ and age. Although both were included as covariates in all analyses, future studies should strive to match participants on age and IQ. This would be particularly important given that across both groups, youth with higher IQ had lower amygdala responses. Further, we found that age was correlated with stronger responses to social stimuli, particularly in the medial prefrontal cortex in the TD group. While IQ has not been shown to relate to SOR (Wood et al., 2021), lower IQ is related to higher amygdala response to potential threat (Choe et al., 2015). Age may be related to SOR, and particularly to prefrontal engagement during aversive sensory stimulation (Cakar et al., 2023). Further investigations of the relationship between IQ, age and amygdala and prefrontal responses are warranted. Second, this study involved passive sensory exposure, and future studies should examine whether findings extend to active social information processing tasks to further increase ecological validity. Third, while the stimuli were matched on aversiveness, they were not matched for basic sound properties. However, stimuli were consistent across groups and rated as comparably aversive, so basic sound properties cannot account for the observed group differences. Fourth, the use of parent-report measures to assess SOR and self-report of aversiveness of stimuli may be subject to biases or variations in individual perception. Further studies could incorporate objective measures or information from multiple informants to provide a more comprehensive assessment of sensory experiences. Despite these limitations, the study provides important insights into the relationships between sensory and social processing in individuals with ASD and sets the stage for future investigations to deepen our understanding of these phenomena and inform targeted interventions.

Summary and Conclusions

In sum, youth with ASD exhibited greater neural activation to aversive stimuli without social relevance compared to socially relevant but similarly aversive stimuli, particularly in regions implicated in emotional arousal and visual information processing. In contrast, TD youth showed greater neural responses to stimuli with social relevance, particularly in frontal regions important for social information processing and emotion regulation. The TD group had significant differences in amygdala and OFC activity in response to nonsocial and social sounds, whereas the ASD group showed similar activation in these regions regardless of the social relevance of the sounds. Finally, we found that SOR severity was correlated with neural responses in ASD youth, such that those with lower SOR showed more activation in regions important to higher-level sensory integration and interpretation during the nonsocial stimuli whereas those with higher SOR had greater activation of sensory-motor regions during socially relevant stimuli. Taken together, while TD youth may upregulate regions important to social attention and social information processing in a context-dependent manner, ASD youth, and particularly those with higher SOR, may respond to the general aversiveness of the conditions. These findings highlight the complex interplay between sensory and social information processing and provide insight into the neural mechanisms underlying disruptions in social information processing that is common among individuals with ASD and heightened SOR (e.g., Green et al., 2018; Schwarzlose et al., 2023).