Autism Symptom Severity Predicts Accuracy of Facial Emotion Recognition Under Conditions of Low Theta-Power Early in Visual Processing

Background: Individuals with autism spectrum disorder (ASD) demonstrate deficits in facial emotion recognition (FER; Harms et al., 2010), which requires holistic visual processing (Duchaine et al., 2006). The weak central coherence model (WCC) predicts that individuals with ASD have deficits in holistic processing (Happé and Frith, 2006). While, the level of impairment on FER tasks generally relates to ASD symptom severity (SS) (Spencer & O’Brian, 2006; Humphreys et al., 2007), some individuals with high ASD SS can have intact performance on FER task, possibly through the engagement of compensatory neural processes that don’t involve holistic processing (Harms et al, 2010). However, different neural processes that reflect cognitive compensation during FER task have yet to be identified in individuals with ASD. One possible neural correlate of holistic processing that might shed light on this question is early posterior theta activity. Early Posterior theta in visual processing task can be measured during the P100 event-related potential (ERP) and is associated with cognitive integration during visual working memory tasks (Sauseng et al., 2009), and abnormalities in theta are associated with deficits in object recognition and memory in individuals with ASD (Chan et al., 2011). The extent to which ASD SS relates to FER, as a function of early posterior theta, has yet to be examined. This interaction can clarify the possible compensatory neural processes utilized by individuals with ASD, and is important for understanding the heterogeneity in FER.

Objectives: We examine whether ASD SS predicted performance on a FER task and if this relation varied as a function of early theta-power during the P100 ERP window.

Methods: Fifty-three youth (M_age=11.60, SD_age=2.96; 38 male) with IQ≥70 (M_IQ=103.49, SD_IQ=15.40) and ADOS-2-confirmed ASD diagnosis and SS. Electroencephalography (EEG) data were recorded during a FER task (DANVA-2; Nowicki, 2004). Natural log transformed theta power at the P100 ERP was calculated using wavelet analysis for central frequencies 4-7.5Hz using 20 logarithmic steps at 4 cycles. Moderated linear regression analyses were used to test the relation of ASD SS and accuracy for identifying adult and child emotional faces at varying levels of the theta frequency during the P100.

Results: No correlations between ADOS SS and FER performance, theta-power and FER performance, or theta-power and SS (p’s> 0.26) were found. However, a significant interaction was found between SS and P100 theta-power in predicting FER performance (B= 0.463 p = 0.026), such that at lower levels of theta-power and greater SS predicted better FER performance (Figure 1).

Conclusions: Our results suggest that greater SS predicts better FER performance, but only among those exhibiting lower occipital theta-power during the P100 ERP window. Heightened posterior theta is associated with holistic processing during FER tasks (Sauseng et al., 2009). These results suggest that, when such holistic processing is not engaged, individuals with greater symptoms of ASD may utilize alternate or compensatory strategies to decode faces, while those with fewer symptoms may rely on underdeveloped holistic processing abilities. Our results may help explain heterogeneity in relations between ASD SS and FER performance.