Sex Differences in Automatic Emotion Regulation in Adolescents with ASD

Abstract

Autism may be underdiagnosed in females because their social difficulties are often less noticeable. This study explored sex differences in automatic facial emotion processing in 45 adolescents with autism spectrum disorder (22 female, 23 male), age 10–16 years, performing active target detection task and Go/NoGo tasks where faces with positive and negative emotional expressions served as irrelevant distractors. The combined sample demonstrated more accurate performance on the target detection (response initiation) than the Go/NoGo task (response inhibition), replicating findings previously reported in typical participants. Females exhibited greater difficulty than males with response initiation in the target detection task block, especially in the context of angry faces, while males found withholding a response in the Go/NoGo block with happy faces more challenging. Electrophysiological data revealed no sex differences or emotion discrimination effects during the early perceptual processing of faces indexed by the occipitotemporal N170. Autistic males demonstrated increased frontal N2 and parietal P3 amplitudes compared to females, suggesting increased neural resource allocation to automatic emotion regulation processes. The associations between standardized behavioral measures (autism severity, theory of mind skills) and brain responses also varied by sex: more adaptive social functioning was related to the speed of perceptual processing (N170 latency) in females and the extent of deliberate attention allocation (P3 amplitudes) in males. Together, these findings suggest that males and females with autism may rely on different strategies for social functioning and highlight the importance of considering sex differences in autism.

Lay Summary:

Females with autism may exhibit less noticeable social difficulties than males. This study demonstrates that autistic females are more successful than males at inhibiting behavioral responses in emotional contexts, while males are more likely to initiate a response. At the neural level, social functioning in females is related to the speed of automatic perceptual processing of facial cues, and in males, to the extent of active attention allocation to the stimuli. These findings highlight the importance of considering sex differences in autism diagnosis and treatment selection.

Autism Spectrum Disorder (ASD) is characterized by impairments in social communication and restricted, repetitive behaviors and interests (APA, 2013). While generally considered to be more common in males, with a ratio of 4:1 (Maenner et al., 2020), recent data suggest that ASD may be underdiagnosed in females (Kim et al., 2011; Loomes et al., 2017) because of their effective camouflaging, such as deliberately performing the expected social behaviors in order to make their difficulties less noticeable (Dean et al., 2017; Dworzynski et al., 2012; Lai et al., 2015; Corbett et al., 2020a).

Previous studies examining social functioning in ASD focused on the ability to process socially relevant information, such as faces and facial expressions (e.g., see Jemel et al., 2006, Harms et al., 2010 for reviews). The latter provide additional cues needed to understand what other people might be feeling or thinking, contributing to theory of mind (TOM) skills and allowing to adaptively modify behavior (Mier et al., 2010; Trevisan & Birmingham, 2016). Although emerging data suggest that autistic females may show higher social motivation (Sedgewick et al., 2016) and greater visual attention to faces compared to males (Harrop et al., 2019), little is known about sex differences in emotion processing in ASD.

Emotion processing includes both detection of emotional expressions in others (i.e., emotion perception) as well as voluntary or automatic regulation of one’s own feelings and actions (Whittle et al., 2011). Previous research on sex differences in emotion perception in neurotypical participants noted that across the lifespan, females might be slightly faster and more accurate in facial expression recognition than males (Rahman et al., 2004; Hall et al., 2010), although the effect sizes for such sex differences were often small (Hall & Matsumoto, 2004; McClure, 2000). With regard to specific emotions, some studies on sex differences reported females to be better than males at recognizing positive expressions (Lambrecht et al., 2014; Bonebright et al., 1996), while others identified female advantage in detecting negative emotions (Thompson & Voyer, 2014). The neural mechanisms underlying emotion perception may also vary between sexes, with greater levels of limbic activation supporting faster emotion identification in females, and increased frontal and parietal cortical activation corresponding to slower and more analytical emotion processing in males (Whittle et al., 2011).

In participants with ASD, most findings on emotion perception come from studies with predominately male samples and suggest difficulties with facial emotion recognition (Ghanouni & Zwicker, 2018; Lozier et al., 2014), especially for anger (Ashwin et al., 2006; Leung et al., 2015), although specific findings have been inconsistent across tasks (Harms et al., 2010). A meta-analysis concluded that recognition of happiness might be relatively typical in ASD, while negative emotions (e.g., anger, fear) are recognized less accurately regardless of age or IQ (Uljarevic & Hamilton, 2013). Limited examination of sex differences to date yielded inconsistent results, with some reporting that female adults with ASD recognized six basic emotions significantly better than males (Sucksmith et al., 2013), and others finding no sex differences (Ketelaars et al., 2016).

The connection between efficient emotion recognition and social difficulties in ASD may not be direct because performance on emotion perception tasks did not always correlate with caregiver reports of social functioning in ASD (Griffiths, 2019), and emotion recognition training did not result in improvement of social skills in individuals with ASD (Zhang et al., 2021). Thus, closer examination of other forms of emotion processing, such as automatic emotion regulation following incidental exposure to affective stimuli, may be more informative when considering social functioning and associated sex differences in ASD.

Emotional faces are salient stimuli (Batty & Taylor, 2003; Vuilleumier & Schwartz, 2001), and when they are unexpected or task-irrelevant, performance on the main task will depend on efficiency of automatic emotion regulation aimed at inhibiting the distraction (e.g., Mauss et al., 2007; Blair et al., 2007; Taylor et al., 2018). Effective emotion regulation is also related to better social adjustment (Hopp et al., 2011). A review of studies in neurotypical individuals concluded that males compared to females may be more efficient in automatic emotion regulation because of their greater reliance on the fronto-parietal than limbic networks for emotion processing (Whittle et al., 2011). Sex-based differences have also been reported in the types of emotion regulation strategies, with females relying on more verbal strategies (e.g., social support, rumination) and males utilizing avoidance and suppression (Cai et al., 2018; John & Gross, 2004; Zimmermann & Iwanski, 2014; Tamres et al., 2002). According to parent- or self-reports, individuals with ASD generally exhibit more emotion regulation difficulties compared to typical peers (Cai et al., 2018; Mazefsky et al. 2014), which may contribute to social difficulties (Beck et al., 2020; Hartmann et al., 2019). While greater than typical use of suppression to regulate negative emotions has been reported in a predominately male sample of children and adolescents with ASD (Samson et al., 2015), sex differences in active or automatic emotion regulation in autistic individuals have not been systematically characterized (Cai et al., 2018). Furthermore, a systematic review of emotion regulation processes in ASD suggested caution in interpreting the findings because most studies used self-report measures, which could be skewed by communication difficulties or inaccurate reporting of inner experiences (Weiss et al., 2014).

Measures of brain activity are frequently used to examine underlying neural processes that may not be readily observable in overt responses. In particular, event-related potentials (ERP) offer millisecond-level temporal resolution that allows to characterize distinct stages of information processing, from stimulus detection to evaluation and response selection. In the context of social information processing, prior ERP studies noted that perceptual encoding of faces is reflected by the occipito-temporal N170 response (Bentin et al., 1996; Rossion, 2014). In ASD, the N170 may be slightly delayed compared to typical peers (Kang et al., 2018), but does not appear to be directly associated with social behavior (Key & Corbett, 2020). The evidence is mixed regarding the N170 sensitivity to facial expressions (Batty & Taylor, 2003; Eimer & Holmes, 2002): in adults, it may distinguish between emotional and neutral faces (albeit the effects sizes are small), but not between different types of emotion (Hinojosa et al., 2015). Previously, typically developing school-age children did not demonstrate significant N170 modulation in response to emotional faces (Batty & Taylor, 2006). Conversely, a frontal N2 response, observed 200ms after stimulus onset might be sensitive to processing of emotional information and index regulatory efforts to inhibit actions in the presence of negative emotion (Lamm & Lewis, 2010). In particular, when used in a Go/NoGo paradigm with faces requiring speeded responses to a frequent stimulus (Go trials) and response withholding to a different stimulus presented infrequently (NoGo trials), Zhang and Lu (2012) interpreted the N2 response to Go trials to reflect automatic attention to emotional stimuli while the N2 response to NoGo trials was considered to index response conflict monitoring. The following frontal P3 response, occurring within 300–500 ms, on Go trials reflected general attention to motivationally relevant stimuli. The NoGo P3 response served as a marker of motor response inhibition associated with automatic emotion regulation processes (Zhang & Lu, 2012). Of note, a parietal rather than frontal maximum could be observed for the inhibition-related ERP responses in children compared to adults (Ciesielski et al., 2004). A parietal P3 response reflecting attention and memory processes, including response selection, overlaps in space and time with the late positive potential (LPP) in tasks requiring active stimulus evaluation (Polich, 2007). The LPP is modulated by emotional content (Schupp et al., 2000); thus, automatic emotion processing and regulation could also affect the parietal P3 (Dennis & Hajcak, 2009).

Examination of the effects of stimulus valence on automatic emotion regulation in neurotypical adults revealed reduced behavioral accuracy on trials requiring response inhibition in the positive compared to negative emotional context (Albert et al., 2010; Hare et al., 2005; Taylor et al., 2018), consistent with the idea that happy faces encourage approach behaviors and thus necessitate greater inhibitory control, while negative faces would trigger behavioral inhibition that reduces the demand for attentional resources needed to successfully withhold a response (e.g., Urbain et al., 2017). NoGo trials presented in the context of happy vs. negative faces elicited larger frontocentral P3 amplitudes in neurotypical adults (Albert et al., 2010), while a magnetoencephalography study (Taylor et al., 2018) noted increased activation of frontal and temporal regions involved in face and emotion processing in response to NoGo trials presented with angry vs. happy faces within 225–425ms after stimulus onset, the time window that overlaps with the N2-P3 responses. Preadolescent neurotypical children also demonstrated increased frontal activation in response to NoGo trials presented with angry vs. happy faces (Urbain et al, 2017). However, they were less successful in inhibiting button press responses on the angry vs. happy face trials, suggesting difficulty with attentional regulation of behavioral responses in negative contexts, possibly due to immaturity of the inhibitory networks in the brain (see also Vidal et al., 2012).

The goal of this exploratory study was to investigate whether automatic emotion regulation processes differ between females and males with ASD and contribute to their social functioning. We selected the emotional faces paradigm previously used in studies of neurotypical adults (Taylor et al., 2018) and children (Urbain et al., 2017) because it included a response inhibition block (Go/NoGo task) and a control condition that did not require withholding of a prepotent motor response to the same stimuli (active target detection task). Assuming that sex differences in automatic emotion regulation would interfere with the behavioral response inhibition to a greater extent than with the perceptual encoding of the stimuli, we predicted that differences between males and females with ASD would be most pronounced in the NoGo N2 and P3 responses, especially on angry face trials. The direction of such between-group differences could not be specified a priori due to insufficient prior data. Sex differences in the N170 response associated with face perception were examined for completeness.

Method

Participants

Forty-five adolescents with ASD between the ages of 10 and 16 years representing four consecutive cohorts of participants in a randomized clinical trial of a social skills treatment (SENSE Theatre®; www.clinicaltrials.gov ID# NCT02276534) contributed ERP data for this study. The sample included all available females (n=22) and 23 males matched on age to the females. All participants were recruited from the university clinic, support groups and schools. The diagnosis of ASD was made in accordance with the Diagnostic and Statistical Manual-5 (APA, 2013) based on: (1) a previous diagnosis by a psychologist, psychiatrist, or pediatrician with autism expertise; (2) current clinical judgment (B.A.C.); and (3) the Autism Diagnostic Observation Schedule (ADOS; Lord et al., 2000), administered by research-reliable personnel. Social Communication Questionnaire (SCQ; Rutter et al., 2003) further corroborated the diagnosis (scores of ≥15). Parent reports indicated that many participants scored in the clinically significant range on anxiety measures: Child Behavior Checklist (CBCL; Achenbach & Dumenci, 2001) anxiety subscale score > 63: 18 females/14 males; Multidimensional Anxiety Scale for Children-2 (MASC-2; March, 2013) total score > 60: 15 females/14 males.

The demographic and clinical information for the study sample is presented in Table 1. The two groups were not significantly different on standardized behavioral measures of IQ, autism severity, or social functioning. However, caregivers of the autistic females compared to males reported more (p=.048) affective/depressive symptom concerns on the CBCL.

Table 1.

Demographic and diagnostic characteristics of the study participants.

		Female	Male	Total	p-value
		n=22	n=23	n=45
Age	M	12.70	13.00	12.85	0.608
	SD	1.75	2.06	1.90
ADOS severity	M	6.57	7.17	6.89	0.245
	SD	1.72	1.67	1.70
SCQ total	M	16.41	18.91	17.69	0.250
	SD	7.72	6.65	7.22
WASI Verbal	M	108.38	102.70	105.41	0.280
	SD	14.25	19.50	17.24
WASI Performance	M	94.29	100.87	97.73	0.239
	SD	18.31	18.24	18.36
WASI Composite	M	100.05	101.87	100.98	0.735
	SD	17.55	18.33	17.77
CBCL - affective	M	68.33	62.52	65.30	0.048
	SD	9.30	9.62	9.81
CBCL - anxiety	M	68.57	64.70	66.55	0.074
	SD	6.18	7.71	7.21
MASC total (parent)	M	66.43	63.70	65.00	0.426
	SD	12.77	9.70	11.22
NEPSY Theory of Mind: Verbal	M	18.59	16.83	17.69	0.122
	SD	2.89	4.42	3.81
NEPSY Theory of Mind: Contextual	M	4.91	4.87	4.89	0.903
	SD	0.87	1.25	1.07
NEPSY Theory of Mind: Total	M	23.50	21.70	22.58	0.149
	SD	3.07	4.92	4.18
SRS T-score	M	79.14	74.65	97.50	0.740
	SD	8.04	8.22	8.35

All participants had normal or corrected-to-normal vision, no color blindness or medical history of seizures, traumatic head injury, or other serious medical conditions affecting the central nervous system (confirmed by parent report). Parents/guardians of the participants provided written informed consent, and participants provided assent. The Institutional Review Board of Vanderbilt University Medical Center approved the study procedures. The ERP and neuropsychological measures reported in this manuscript were collected during the baseline study visits.

Procedure

ERP measures

Emotion processing tasks: Based on the procedures of Taylor et al. (2018), participants completed two blocks of 128 trials designed to elicit automatic emotion regulation: (1) behavioral response inhibition (Go/NoGo task; 75% Go trials requiring a button press, 25% NoGo trials requiring withholding of the motor response) and (2) active target detection (75% standard trials not requiring a response; 25% target trials requiring a response). Within each block, the participants viewed color photographs of prototypical emotional (50% angry, 50% happy) expressions from 38 unfamiliar young adults (50% female) drawn from the Radboud Faces Database (Langner et al. 2010). Each photograph was presented inside a blue or purple frame (1.7 cm wide) on a white background in the center of a 19” compute monitor. From the viewing distance of 90 cm, the entire stimulus subtended visual angles of 10.5° (h) × 8.6°(w). The participants were instructed to attend to the color of the frame and respond via a button press (target detection block) or withhold a response (Go/NoGo block) when they saw the target color. They were encouraged to perform as accurately and quickly as possible. The order of the blocks and the target color were counterbalanced across participants.

Each trial began with a black fixation point (plus sign; Courier New font, size 48) that remained on the screen for the duration of the intertrial interval (varied randomly between 900–1100ms) followed by the stimulus presentation (500 ms) and another fixation point until a button press response was made (up to 1000ms). On average, target detection block lasted ~6 minutes, while the Go/NoGo block was ~3 minutes long. The difference in block duration was due to the varied number of active button press responses. A researcher was present in the testing room to monitor participants’ attention to the stimuli. During periods of inattention or excessive motor activity, stimulus presentation was suspended until the participant was ready to continue with the task.

ERP Acquisition

A 128-channel Geodesic Sensor net (EGI, Inc., Eugene, OR) was used to record the ERPs. Data were sampled at 250Hz with impedance levels at or below 50 kOhm. All electrodes were referred to vertex and then re-referenced during data analysis to an average reference (Picton et al., 2000). Data were recorded using NetStation 5.4 software and NetAmps 400 amplifier. E-prime (v.2.0, PST, Inc., Pittsburgh, PA) controlled stimulus presentation and recorded behavioral performance (accuracy and reaction time).

ERP Data Analysis

Continuous EEG recordings were filtered using a 0.1–30Hz bandpass filter and segmented on stimulus onset to include a 100-ms prestimulus baseline and a 900-ms poststimulus interval. Ocular and movement artifacts were removed using independent components analysis procedure implemented in EEGLab (Delorme et al., 2007). Data for electrodes with poor signal quality were reconstructed using spherical spline interpolation (Junghofer et al., 2000). If more than 20% of the electrodes within a trial required interpolation, the entire trial was discarded. Data retention rates were comparable across emotions within each trial type and across the subject groups (Go/NoGo task: Happy Go M = 32.76 +/− 8.00; Happy NoGo M = 11.09 +/− 2.53; Angry Go M = 34.00 +/− 7.05; Angry NoGo M = 10.58 +/− 2.47; Target detection task: Happy Target M = 10.60 +/− 2.61; Happy Standard M = 30.47 +/− 8.36; Angry Target M = 10.44 +/− 2.61; Angry Standard M = 24.51 +/− 9.98). There were no significant sex-related differences in data retention rates for all but one condition: females retained 1.5 trials more than males in the happy target condition (p = .04).

Following artifact removal, individual ERPs were averaged separately for trials that did or did not require a response within target detection and Go/NoGo blocks and for each emotional expression. The resulting data were re-referenced to an average reference and baseline-corrected by subtracting the average microvolt value across the 100-ms prestimulus interval from the poststimulus segment. Next, mean amplitudes and peak latencies were derived for the occipito-temporal N170 (150–230ms), fronto-central N2 (200–300ms), and parietal P3 (300–600ms) responses (Table 2). These scalp locations¹ and time intervals were selected a priori based on prior studies examining these ERP components in children with and without autism (Smith et al., 2004; Key & Corbett, 2014; Zhang & Lu, 2012) and confirmed by visual inspection of the grand-averaged waveforms. Mean amplitude measures are more robust than maximum peak amplitude measures to the detrimental effects of high-frequency noise or variability in the number of artifact-free trials (Luck, 2014; Thomas et al., 2004). Averaging amplitude and latency measures across the electrodes within each cluster (Figure 1) further maximized signal-to-noise ratio. Following the analysis approach of Taylor et al. (2018), the dependent measures were grouped based on task, emotion, and the presence of the motor response (e.g., NoGo and standard trials). Furthermore, based on the findings of Zhang & Lu (2012) suggesting that Go trials may provide additional information about attention to the emotional stimuli, we opted to include them in the analysis instead of calculating NoGo-Go difference scores. Thus, amplitude and latency measures for each ERP peak of interest were entered into separate repeated measures analyses of variance (ANOVA) with Sex (2: female, male) as the between-subject factor and Task (2: Go/NoGo, target detection) × Trial Type (2: respond, not respond) × Emotion (2: happy, angry) × Electrode (2: left/right for N170; anterior/posterior for N2, P3) as the within-subject factors. Significant interactions were further explored using one-way ANOVAs and/or pairwise comparisons with Bonferroni correction.

Table 2.

Mean amplitude and peak latency for the N170, N2, and P3 responses for each emotion and trial type.

Measure	Emotion	Trial Type N170	Female				Male				Total
			left OT		right OT		left OT		right OT		left OT		right OT
			M	SD	M	SD	M	SD	M	SD	M	SD	M	SD
amplitude	angry	Go	−0.04	2.70	−0.35	1.99	0.40	2.58	0.00	4.82	0.19	2.62	−0.17	3.68
		NoGo	0.31	3.48	−0.04	3.66	2.40	5.68	1.02	4.75	1.38	4.80	0.50	4.24
	happy	Go	0.29	3.03	−0.46	2.07	−0.23	5.14	0.55	4.97	0.03	4.20	0.06	3.83
		NoGo	−0.24	2.84	−0.09	2.95	1.31	3.80	0.28	3.51	0.56	3.42	0.10	3.22
	angry	Standard	0.43	3.15	0.64	6.94	0.95	4.48	−0.32	4.23	0.70	3.85	0.15	5.67
		Target	−0.49	3.70	−0.11	3.51	0.68	4.60	−0.87	4.94	0.11	4.18	−0.50	4.27
	happy	Standard	0.11	2.91	1.01	3.28	1.11	2.68	0.06	4.42	0.62	2.81	0.52	3.89
		Target	0.76	3.43	−0.01	3.23	0.27	4.79	−1.07	5.23	0.51	4.14	−0.55	4.36
latency	angry	Go	185.82	24.06	179.27	22.17	193.57	29.83	192.70	32.65	189.78	27.14	186.13	28.52
		NoGo	191.27	22.75	184.36	22.83	194.09	28.97	194.61	30.45	192.71	25.86	189.60	27.19
	happy	Go	185.82	24.59	178.55	24.64	197.39	31.53	186.61	29.68	191.73	28.63	182.67	27.33
		NoGo	198.36	26.78	177.09	20.48	187.30	29.14	188.35	31.70	192.71	28.25	182.84	27.11
	angry	Standard	183.09	23.35	179.45	18.41	182.43	27.55	190.96	30.92	182.76	25.30	185.33	25.95
		Target	184.91	23.74	184.73	24.30	191.83	32.69	191.48	31.87	188.44	28.56	188.18	28.31
	happy	Standard	182.36	24.05	181.82	25.92	187.13	32.74	195.83	30.76	184.80	28.60	188.98	29.05
		Target	186.00	26.50	192.91	24.25	184.17	27.06	190.78	30.07	185.07	26.50	191.82	27.09

		N2	FCz		FCz		FCz
			M	SD	M	SD	M	SD
amplitude	angry	Go	−1.76	2.39	−0.69	2.70	−1.21	2.58
		NoGo	−1.83	3.07	−2.69	3.56	−2.27	3.32
	happy	Go	−1.71	2.21	−0.75	3.72	−1.22	3.08
		NoGo	−2.10	2.88	−1.73	3.35	−1.91	3.10
	angry	Standard	−1.53	3.13	−1.24	2.35	−1.38	2.73
		Target	−1.29	2.25	−0.42	2.76	−0.84	2.53
	happy	Standard	−2.11	2.62	−1.32	2.97	−1.71	2.80
		Target	−2.21	3.04	−1.62	3.05	−1.91	3.03
latency	angry	Go	262.73	32.74	257.57	37.44	260.09	34.92
		NoGo	261.82	28.30	255.65	27.26	258.67	27.63
	happy	Go	261.09	35.50	263.65	33.05	262.40	33.90
		NoGo	258.18	31.90	252.87	22.14	255.47	27.16
	angry	Standard	266.73	31.19	262.96	37.43	264.80	34.18
		Target	258.00	33.55	257.91	32.02	257.96	32.40
	happy	Standard	273.82	26.78	263.13	37.62	268.36	32.85
		Target	260.36	27.57	262.43	34.43	261.42	30.93

		P3	FCz		Pz		FCz		Pz		FCz		Pz
			M	SD	M	SD	M	SD	M	SD	M	SD	M	SD
amplitude	angry	Go	−1.34	1.86	1.59	1.96	−0.14	2.54	2.53	2.24	−0.73	2.29	2.07	2.14
		NoGo	−0.27	2.91	3.04	2.82	−1.30	3.63	4.33	4.65	−0.80	3.30	3.70	3.88
	happy	Go	−1.38	1.62	2.32	1.88	−0.43	2.99	3.76	3.10	−0.90	2.44	3.06	2.65
		NoGo	−1.44	3.00	2.30	2.36	−0.19	4.24	3.95	4.54	−0.80	3.70	3.14	3.70
	angry	Standard	−0.69	2.67	2.07	2.37	−1.18	3.35	1.91	2.36	−0.94	3.01	1.99	2.34
		Target	−1.92	2.62	4.34	3.36	−0.80	3.43	5.71	4.10	−1.35	3.08	5.04	3.78
	happy	Standard	−1.45	1.86	1.22	1.83	−0.88	2.84	2.75	2.60	−1.16	2.40	2.00	2.36
		Target	−2.46	2.44	4.58	3.80	−1.44	3.63	6.63	3.32	−1.94	3.11	5.63	3.67
latency	angry	Go	475.82	114.14	382.55	94.06	474.09	113.43	342.78	66.59	474.93	112.48	362.22	82.72
		NoGo	471.27	99.13	436.36	115.08	442.96	96.67	417.91	90.36	456.80	97.81	426.93	102.42
	happy	Go	461.27	107.20	384.36	94.50	453.22	111.88	380.35	76.50	457.16	108.44	382.31	84.81
		NoGo	440.73	103.24	424.55	96.66	465.22	100.89	393.74	88.32	453.24	101.63	408.80	92.75
	angry	Standard	449.09	95.42	431.45	129.20	470.09	109.97	386.96	79.34	459.82	102.50	408.71	107.80
		Target	428.18	105.56	423.45	86.34	477.57	112.92	409.74	72.75	453.42	110.98	416.44	79.07
	happy	Standard	461.09	93.58	412.00	114.78	463.83	90.87	385.04	89.37	462.49	91.16	398.22	102.31
		Target	428.55	106.77	426.55	91.88	474.96	116.97	411.48	89.92	452.27	113.28	418.84	90.17

Figure 1.

128-channel electrode layout and the selected clusters used in the analysis.

Behavioral Performance Analysis

Button press response accuracy and reaction time for each emotion and trial type were analyzed using separate repeated measures ANOVAs with Sex (2: female, male) as the between-subject factor and Task (2: Go/NoGo, target detection) × Emotion (2: happy, angry) × Trial Type (2: respond, not respond) as the within-subject factors. To ensure accurate representation of response speed, reaction time data were curated to exclude values below 100 ms.

Measures of social functioning

NEPSY Theory of Mind (TOM) subtests (Korkman, Kirk, & Kemp, 2007) were administered to formally assess social perception as a skill likely to be affected by automatic emotion recognition abilities (e.g., Coricelli, 2005; Henry et al., 2006). The verbal portion (TOM-V) measured ability to understand false belief and figurative language, recognize mental states and imitation. The contextual portion (TOM-C) assessed the ability to relate emotion to social context. It required the participants to identify a picture that best represents the feelings of a character in six different scenarios. Both TOM-V and TOM-C subtest scores as well as the total score were used in the analyses.

Social Responsiveness Scale (SRS; Constantino & Gruber, 2005) is a 65-item questionnaire completed by caregivers to measure social functioning (e.g., communication, cognition). Individual differences in emotion regulation abilities can contribute to performance in the social domains measured by the SRS (Hartmann et al., 2019). The Total T-score was used in the analyses. T-scores between 60 and 75 are clinically significant, with scores above 76 indicating more severe ASD symptoms.

Exploratory Brain-Behavior Association Analysis

Exploratory analyses examined correlations between the ERP responses, behavioral Go/NoGo and target detection task performance, ASD symptoms (ADOS), and social functioning (SRS, NEPSY TOM). To provide the least conservative evaluation of possible brain-behavior associations, no correction for multiple significance testing was applied for the correlational analyses.

Results

Behavioral performance

Behavioral performance data were not available for one female participant due to a technical error. All remaining participants performed at better than chance level. Summary accuracy and reaction time data are presented in Table 3. The repeated measures ANOVA using accuracy as the dependent measure identified a main effect of Task, F(1,41) = 29.55, p < .001, η_p² = .42, indicating that the participants were more accurate in the target detection than inhibition task. There was also a Trial Type × Sex interaction, F(1,41) = 15.67, p < .001, η_p² = .28. Follow-up analyses revealed that females were more accurate than males on trials requiring response withholding (NoGo trials in the inhibition block, standard trials in the target detection block), F(1,41) = 6.11, p = .02.

Table 3.

Mean accuracy and reaction times for behavioral performance on each trial type ad emotion context.

Measure	Emotion	Trial	Female		Male		Total
		Type	M	SD	M	SD	M	SD
Response Inhibition
Accuracy	Happy	Go	0.85	0.17	0.94	0.12	0.90	0.15
		NoGo	0.81	0.15	0.73	0.16	0.77	0.16
	Angry	Go	0.86	0.17	0.94	0.10	0.91	0.15
		NoGo	0.85	0.15	0.72	0.20	0.78	0.19
Reaction Time	Happy	Go	361.73	51.81	363.39	62.79	362.59	57.15
		NoGo	351.47	122.11	313.59	109.03	331.63	115.61
	Angry	Go	363.32	52.31	365.43	64.07	364.42	58.09
		NoGo	301.23	52.79	312.87	104.62	307.83	85.30
Target detection
Accuracy	Happy	Target	0.81	0.25	0.95	0.10	0.89	0.20
		Standard	0.98	0.02	0.91	0.15	0.95	0.11
	Angry	Target	0.84	0.24	0.91	0.16	0.88	0.20
		Standard	0.98	0.02	0.93	0.14	0.95	0.11
Reaction Time	Happy	Target	409.36	58.67	434.40	93.19	422.76	79.16
		Standard	326.02	77.44	429.73	156.73	390.84	140.25
	Angry	Target	403.07	58.29	436.98	100.97	421.21	84.68
		Standard	426.80	345.38	471.35	283.44	453.84	303.78

The interaction of Task × Emotion × Trial Type × Sex, F(1,41) = 4.01, p = .05, η_p² = .09, indicated females were more accurate than males in withholding responses on NoGo trials presented with the angry faces, F(1,42) = 5.53, p = .02. Conversely, males were more accurate than females initiating the response to the target trials presented with happy faces, F(1,42) = 5.67, p = .02. Within-group analyses noted that in both emotional conditions, females had difficulty with response initiation in the target detection block (lower accuracy for target than standard stimuli; happy: t(19) = 2.99, p < .01, d = .67; angry: t(19) = 2.67, p = .02, d = .60), while males found withholding a response challenging in the inhibition block (lower accuracy for NoGo than Go trials; happy: t(22) = 6.38, p < .001, d = 1.33; angry: t(22) = 5.79, p < .001, d = 1.21).

Analysis of the reaction times identified a main effect of Task, F(1,17) = 12.66, p < .01, η_p² = .44, indicating that the participants responded slower in the target detection than inhibition task. No other main effects or interactions reached significance.

ERP results

N170: For the amplitude, there was a main effect of Trial Type, F(1,43) = 6.42, p = .02, η_p² = .13, due to larger N170 response on trials requiring a behavioral response (Figure 2). No other effects reached significance.

Figure 2.

Left and right occipito-temporal N170 responses in males and females with ASD during the (A) inhibition (Go/NoGo) and (B) target detection (active oddball) tasks.

Analysis of the N170 latency identified interactions of Task × Electrode, F(1,43) = 9.86, p < .01, η_p² = .19, Task × Electrode × Emotion, F(1,43) = 4.85, p = .03, η_p² = .10, Trial Type × Electrode × Sex, F(1,43) = 6.48, p = .02, η_p² = .13, and Task × Trial Type × Emotion × Sex, F(1,43) = 4.05, p = .05, η_p² = .09. Follow-up analyses noted no between- or within-sex differences in the timing of the N170 response on trials requiring withholding or initiating a button press as well as across trials with different emotions. In the combined sample, there were no significant between-emotion differences in the N170 latency for either the target detection or the inhibition tasks. Between-task differences were observed at the left temporo-occipital cluster, with faster latencies in the target detection than inhibition tasks for both angry and happy emotions, t(44) = 2.03, p = .05, d = .30 and t(44) = 2.41, p = .02, d = .36, respectively. At the right temporo-occipital sites, task-related differences were observed for the happy faces only, with shorter latencies in the inhibition than target detection task, t(44) = 2.57, p = .01, d = .38.

N2: The fronto-central N2 response was examined in the context of the inhibition task only to minimize the overall number of significance tests (Luck & Gaspelin, 2017) because there were no specific hypotheses for the N2 effects in the target detection task. The amplitude was characterized by the main effect of Trial Type, F(1,43) = 7.54, p = .01, η_p² = .15, and a Trial Type × Sex interaction, F(1,43) = 4.04, p = .05, η_p² = .09. Follow-up analyses revealed larger N2 amplitudes for the NoGo than Go trials in males, t(22) = 3.17, p < .01, d = .66 (Figure 3), while females exhibited no significant N2 amplitude differences (p = .58). The two groups were not significantly different in their responses to the individual conditions. Analysis of the N2 latency measures did not identify any significant effects.

Figure 3.

Fronto-central N2-P3 responses during the inhibition task in males and females with ASD.

P3: For the amplitude, there was a main effect of Electrode F(1,43) = 91.97, p < .001, η_p² = .68, and interactions of Task × Trial Type, F(1,43) = 5.10, p = .03, η_p² = .11, Task × Trial Type × Electrode, F(1,43) = 18.64, p < .001, η_p² = .30, Emotion × Electrode, F(1,43) = 4.48, p = .04, η_p² = .09, and Emotion × Trial Type × Electrode, F(1,43) = 3.99, p = .05, η_p² = .09. Follow-up analyses identified no significant differences between tasks or trial types at the fronto-central locations (p = .09–.98; Figure 3). At the parietal cluster, larger amplitudes were observed for NoGo than Go trials, t(44) = 2.09, p = .04, d = .31, and for target than standard trials, t(44) = 7.28, p < .001, d = 1.09 (Figure 4). Further analyses noted larger P3 amplitudes for the Target than NoGo trials, t(44) = 3.96, p < .001, d = .59.

Figure 4.

Parietal P3 responses in males and females with ASD during the (A) inhibition (Go/NoGo) and (B) target detection (active oddball) tasks.

Sex-related differences for the P3 amplitude were present as the interaction of Trial Type × Emotion × Sex, F(1,43) = 4.20, p = .05, η_p² = .09. Follow-up analyses noted that males elicited larger amplitudes than females on happy and angry face trials requiring a response, F(1,43) = 8.42, p < .01and F(1,43) = 6.48, p = .02, respectively. For response-withholding trials, sex differences were observed for the happy face condition only, F(1,43) = 4.16, p = .05, with males having larger amplitudes than females.

Analyses of the P3 latency identified the main effects of Task, F(1,43) = 70.86, p < .001, η_p² = .62, Electrode, F(1,43) = 49.68, p < .001, η_p² = .54, and Trial Type, F(1,43) = 55.02, p < .001, η_p² = .56. These effects were further qualified with the following interactions: Task × Electrode, F(1,43) = 60.17, p < .001, η_p² = .58, Task × Trial Type, F(1,43) = 46.59, p < .001, η_p² = .52, Trial Type × Electrode, F(1,43) = 82.16, p < .001, η_p² = .66, and Task × Trial Type × Electrode, F(1,43) = 14.44, p < .001, η_p² = .25. Follow-up analyses identified latency differences only for the inhibition task, with faster parietal P3 responses for the Go than NoGo trials, t(44) = 3.34, p < .01, d = .50.

Associations with behavioral measures

Analysis of the associations between the diagnostic measures, neuropsychological assessments, and behavioral task performance revealed no significant correlations with the ADOS scores, but showed that higher accuracy on the frequent trials (Go, Standard) in both happy and angry face conditions was associated with higher NEPSY TOM total scores, r = .41–.55, p < .05. For the infrequent trials, greater inhibition (NoGo) accuracy for happy faces and response initiation (Target) accuracy for angry faces were also related to higher NEPSY TOM total scores, r = .35, p = .02 and r = .46, p < .01, respectively. There were no associations with the reaction time measures (all p > .05).

Significant correlations between the behavioral task performance and the ERPs sensitive to the experimental condition differences were observed only for the P3 response. In the target detection task, the predicted association between greater parietal P3 amplitudes on target trials and higher behavioral accuracy was observed for both happy and angry face conditions, r = .39, p = .01 and r = .35, p = .02, respectively. Greater accuracy on target trials with happy faces was also associated with shorter parietal P3 latency, r = −.42, p < .01. In the inhibition task, greater accuracy in NoGo trials in happy and angry face conditions was associated with shorter frontal P3 latency, r = −.34, p = .02 and r = −.31, p = .04, respectively, while greater Go accuracy correlated with faster parietal P3 responses, r = −.32, p = .04 and r = −.30, p = .05, respectively.

Analysis of the associations between the ERP responses and standardized diagnostic or neuropsychological measures revealed distinct patterns of brain-behavior correlations between males and females with ASD. ADOS severity scores correlated with delayed NoGo latency of the frontal N2 response to happy faces, r = −.42, p = .05, and reduced parietal P3 amplitude to angry targets, r = −.50, p = .02, in males. In females, higher ADOS severity scores were associated with increased parietal P3 amplitudes to Go and NoGo trials with happy faces, r = .44, p = .04, and r = .50, p = .02, respectively, as well as with delayed right occipito-temporal N170 latencies for happy and angry Go trials, r = .51, p = .02, and r = .44, p = .04, and happy NoGo trials, r = .53, p = .01.

Fewer social difficulties (lower SRS scores) were associated in males with larger parietal P3 amplitudes for Go and standard trials regardless of face emotion, r = −.44– −.58, p < .01–.04, as well as with the larger NoGo response to angry faces, r = −.42, p = .05. None of the SRS associations were significant in females.

Higher scores on TOM-Contextual measure in males were related to increased parietal P3 amplitudes on NoGo and target trials in both emotion conditions, r = .52–.58, p < .01. Smaller left N170 amplitudes to happy Go trials, r = .70, p < .001, and larger responses to happy target trials, r = −.45, p = .03, were also associated with higher scores. In females, better TOM-Contextual performance correlated with shorter right N170 latencies for Go and standard angry trials, r = −.50, p = .02, and r = −.44, p = .04, respectively as well as faster left N170 responses to NoGo angry and target happy trials, r = −.46, p = .03, and r = −.56, p < .01, respectively.

Discussion

Spontaneous emotion processing and the associated sex-based differences in ASD are understudied. The current study explored automatic emotion regulation processes in male and female adolescents with ASD performing target detection and Go/NoGo tasks where facial expressions served as irrelevant distractors. Analysis of behavioral responses for the combined sample revealed more accurate performance on the target detection (response initiation) than the Go/NoGo task (response inhibition) and faster response times during the Go/NoGo than the target detection task, replicating the findings in typical adults previously reported by Taylor et al. (2018). Sex differences were present in both emotional conditions in the form of females having greater difficulty than males with response initiation in the target detection block, while males found withholding a response challenging in the Go/NoGo block. Electrophysiological data indicated no sex differences in the early perceptual processing of faces indexed by the N170. Conversely, males and females differed in the pattern of the frontal N2 and parietal P3 responses during both positive and negative emotion conditions. The brain-behavior associations also varied between the sexes, suggesting differences in the specific approaches to processing social cues.

Behavioral results indicated that more effective response inhibition in females than males with ASD was particularly pronounced on the NoGo trials presented with angry faces, resembling the stimulus effects reported by Taylor et al. (2018). Slower responses to angry than happy faces, which could support more effective inhibition, have been previously reported in typical populations (Hare et al., 2008). Sensitivity to anger expressions increases during puberty (Thomas et al., 2007; Lawrence et al., 2015), and recent reports suggest that females with ASD may enter puberty earlier than males (Corbett et al., 2020b). Therefore, females with ASD may evidence more effective automatic detection of and response regulation in the presence of angry faces compared to same-age male peers. Conversely, males were more accurate than females initiating behavioral responses on target trials with happy faces. The finding appears consistent with prior reports demonstrating that typical adolescent boys were more likely to identify faces as happy compared to girls (Lee et al., 2013), and motor responses to happy faces were less likely to be inhibited (Hare et al., 2008; Urbain et al., 2017). Overall, the behavioral performance of the adolescents with ASD followed the patterns observed in neurotypical populations. Furthermore, autistic females exhibited more effective response inhibition compared to males.

Analysis of electrophysiological data demonstrated general sensitivity of the occipitotemporal N170 response to top-down attentional modulation, evidenced by larger amplitudes on trials requiring response initiation compared to withholding. Previously, increased N170 amplitudes were reported for target than standard trials in neurotypical female adults (Choi et al., 2015). Similarly, faster N170 latencies were observed in the target detection than Go/NoGo block regardless of emotional content, suggesting accelerated stimulus detection when a motor response had to be actively prepared for execution. However, there were no significant sex- or emotion-related differences for the N170 amplitude or latency, consistent with the findings from typical populations (Batty & Taylor, 2006; Choi et al., 2015; but see Nowparast Rostami et al., 2020 for evidence of a larger and faster N170 in typical adult females than males). Nevertheless, sex differences were observed for the associations between the N170 metrics and clinical measures: in females but not males, greater ADOS severity scores correlated with slower right occipito-temporal N170 latencies for happy and angry Go and happy NoGo trials. Conversely, faster right N170 latencies for the frequent (Go, standard) angry trials and left N170 latencies to angry NoGo and happy target trials were related to more typical scores on NEPSY TOM-Contextual subscale, which involved processing of emotional cues. For males with ASD, reduced left N170 amplitudes to happy Go trials and larger responses to happy target trials correlated with higher TOM-Contextual scores. In combination, these results suggest that speed of early perceptual face processing (reflected by the N170 latency), without regard for specific emotional content, may play a role in social adaptive functioning in females with ASD. On the other hand, in autistic males, the extent of neural resource allocation to perceptual processing of positive facial expressions (marked by the N170 amplitude) appears to be the primary contributing factor. This pattern of results aligns with the previous findings in neurotypical populations suggesting that females rely on faster, predominately bottom-up perceptual face processing mechanisms, while males may engage more top-down processes (Whittle et al., 2011)

Analysis of the fronto-central N2 response indexing response conflict detection in the Go/NoGo task noted larger amplitudes for trials requiring response inhibition vs. initiation, consistent with the results previously reported for such paradigms in neurotypical children and in participants with ASD (Høyland et al., 2017). However, this finding in the current study was observed only in males with ASD. Paired with their reduced behavioral accuracy on the inhibition trials compared to female participants, the increased N2 amplitudes could reflect more effortful activation of conflict detection and automatic emotion regulation mechanisms in the presence of positive or negative emotional distractors (Lewis et al., 2006). These results suggest that autistic males may be less efficient in automatic emotion regulation compared to females, while an opposite pattern of sex differences has been previously suggested by the studies of neurotypical participants (Whittle et al., 2011).

Contrary to the findings by Taylor et al. (2018), adolescents with ASD as a group did not demonstrate increased processing of task-irrelevant negative emotions in the 200–300ms window, as the N2 amplitudes were not significantly different for trials with angry or happy faces. Inability to differentiate between happy and angry faces is not a likely explanation of these results as behavioral performance varied based on the emotional context. Existing data also note that children with ASD without cognitive disability are not significantly different from typical controls in their ability to identify facial expressions of basic emotions (Fink et al., 2014; Tracy et al., 2010). Furthermore, in cases where the group with ASD was less accurate than the controls, their performance for prototypical (i.e., high intensity) expressions was greater than chance (Griffiths, 2019). On the other hand, similar lack of emotion-related differences in the N2 response was previously reported in children with clinical anxiety, while typical controls elicited larger N2 responses to angry than happy faces (Hum et al., 2013). Children with ASD often exhibit symptoms of anxiety (White & Roberson-Nay, 2009; Bauminger & Kasari, 2000; van Steensel & Heeman, 2017), which could explain the current N2 findings. In the current sample, many participants scored in the clinically significant range on parent reported anxiety measures.

There were no significant sex, task, trial type, or emotion-related effects for the fronto-central P3 response. Reduced or absent anterior NoGo P3 responses have been reported in neurotypical children and interpreted to suggest that the associated inhibitory processes may not become apparent until older ages because motor response inhibition abilities develop slower than response conflict detection processes indexed by the N2 (Jonkman et al., 2003; Johnstone et al., 2005; Smith et al., 2004). Indeed, prior work indicated that the NoGo N2 response could be present by 7 years of age, while the frontal P3 NoGo response reached maturity in young adulthood (Jonkman, 2006). Task difficulty could also explain the observed pattern of the fronto-central N2 and P3 findings, as the amplitude of the NoGo P3 decreases with increasing task difficulty without affecting the NoGo N2 response (Gajewski & Falkenstein, 2013). Furthermore, a shift from frontal toward parietal maxima for the ERP responses indexing inhibition processes has been previously reported in children compared to adults (Ciesielski et al., 2004).

The parietal P3 response revealed typical amplitude modulation by stimulus probability with larger, more positive responses for rare (target, NoGo) than frequent (standard, Go) trials. Furthermore, the parietal P3 response was sensitive to sex differences. Males compared to females with ASD elicited larger P3 amplitudes on happy- and angry-face trials requiring response initiation. These results are consistent with previous investigation of sex differences in typical emotion processing demonstrating greater activation of frontal and limbic circuits in males compared to females when presented with affectively charged images (Whittle et al., 2011). In addition, autistic males generated larger parietal P3 response on happy-face trials requiring response withholding. Similar to the frontal N2 results, the combination of reduced behavioral accuracy and increased P3 amplitude could suggest that males with ASD had to allocate greater attentional resources than females to inhibit a motor response. Previously, Jonkman et al. (2003) interpreted the combination of reduced behavioral accuracy and larger parietal P3 responses as the evidence of a developmental lag in sustained attention processes.

Examination of brain-behavior associations for the parietal P3 response noted that in autistic females, increased amplitudes on Go and NoGo trials with happy faces, indicative of possible difficulty with response selection in the context of positive emotions, correlated with higher ADOS severity scores. Similar associations between increased ASD symptoms and higher P3 amplitudes have been reported in adults with ASD during an active emotion identification task (Keifer et al., 2019). In males, larger parietal P3 amplitude on target trials presented with angry faces was related to reduced ADOS severity while larger P3 amplitudes on angry NoGo trials were associated with fewer social difficulties (lower SRS scores). Anger has been reported as a particularly difficult emotional expression for participants with ASD to recognize (Ashwin et al., 2006; Humphreys et al., 2007; Garman et al., 2016), possibly because it is often encountered in the context of social rules (Leung et al., 2015). The correlation between the P3 amplitude and the SRS scores was also observed for the frequent (Go, standard) trials regardless of face emotion, suggesting that sustained attention to social stimuli and the associated response initiation vs. inhibition choices may be needed for more successful adaptive social functioning in males with ASD.

While providing novel information on automatic emotion regulation and sex differences in ASD, this study has a number of limitations. In the behavioral task, the instructions emphasized accuracy as well as speed rather than speed alone. This manipulation could have reduced the magnitude of response conflict detection; nevertheless, the behavioral results replicated the findings by Taylor et al. (2018) who prioritized the speed of response. Given the reported evidence of difficulties with executive functioning in ASD (Lewis et al., 2006), explicitly encouraging participants to pay attention maximized the likelihood of them actually performing the Go/NoGo and target detection tasks instead of indiscriminately pushing the response button. Also, while we collected behavioral performance data for the experimental tasks, we did not analyze ERP trials based on accuracy. This choice was motivated by the relatively high overall performance and the need to have a sufficient number of usable EEG segments for analysis without making the task excessively long. It is possible that the occasional inclusion of the error trials obscured the N2 and P3 responses. However, our data demonstrated the expected N2 response on the infrequent trials requiring response inhibition and the P3 response to target vs. standard trials. Therefore, the relative influence of the error trials on the overall results was likely minimal. Finally, the study design (based on Taylor et al., 2018, Urbain et al., 2017) did not include neutral faces, which limited our ability to further investigate possible sex differences in automatic emotion perception and regulation in adolescents with ASD.

In conclusion, our results demonstrate that adolescents with ASD engage in automatic emotion regulation, but the specific face processing strategies vary between the sexes. Specifically, females with ASD are better at response inhibition, particularly for angry faces. Conversely, males with ASD are more successful with response initiation, especially for happy faces. Neural results noted no sex differences in the N170 response, but larger N2 and P3 amplitudes in males with ASD, interpreted to reflect more effortful automatic emotion regulation compared to females. The associations between standardized measures (autism severity, social functioning, theory of mind skills) and brain responses also varied by sex: better adaptive functioning was related mainly to the speed of perceptual processing (N170 latency) in females and the extent of attention allocation (P3 amplitudes) in males.

Together, these findings provide additional support for the argument that emotion processing abilities are relevant for broader social functioning in ASD and highlight the importance of biological sex is an neurobehavioral index that warrants intentional focus in research with translational clinical relevance. In clinical practice, males and females with ASD are often conceptualized as a homogenous diagnostic group. Yet, the psychophysiological findings underscore sex-based differences in attentional resource allocation, speed of perceptual processing as well as brain-behavior associations with symptom severity and adaptive functioning. Thus, a unisex approach to autism is outdated. Our results fit into an emerging literature showing unique neural, behavioral and emotional response patterns between females and males with ASD that emphasize the need for additional research to improve diagnostic practice and inform evidence-based interventions.