As has been seen in several prior studies of language and gamma [13, 36–38], the language task in the current study elicited strong gamma event-related desynchronization (ERD; see Figures 1 and 2). ERD of gamma activity was seen in both controls and pASD, but greater ERD was seen in the control group. While previous studies with simple tone/click stimuli found evoked gamma-band power to be reduced in those with ASD [26, 27] and in pASD [27, 30] compared to control groups, the current study found pASD to show increased evoked power overall compared to controls. Gamma-band abnormalities in response to language stimuli have not previously been investigated in first-degree relatives of individuals with ASD, but this finding is consistent with previous research finding adults with ASD to show greater increases in evoked gamma activity in response to a sentence context evaluation task compared to controls .
A likely explanation for the disparity between the current results and previous results showing reduced evoked gamma in pASD compared to controls is that the complexity of the stimuli and the task demands differed: while previous studies of gamma-band abnormalities in pASD used pure tone and amplitude-modulated tone stimuli [27, 30], the current study used spoken language stimuli in the context of a word recognition task. Rather than listening passively to a simple auditory stimulus, higher order cognitive processes including language and sustained attention were necessary for word recognition. While this has not been studied in pASD previously, these findings are consistent with those of Braeutigam et al. , who reported increased evoked gamma in both control adults and adults with ASD during a task involving evaluation of sentence context, but found the ASD group to show greater increases in evoked gamma that were more widespread across the time window than controls.
Both the early transient gamma-band response (tGBR; 30–150 ms post-stimulus) and the later gamma-band response (200–800 ms post-stimulus) were assessed, with similar results found in both time windows. During the earlier tGBR, there were significant group differences in supramarginal gyrus (SMG) and lateral occipital cortex (LOC), and marginally significant group differences in auditory cortex and fusiform gyrus (FFG), in which pASD showed increased evoked gamma activity and controls showed decreased evoked gamma. The same pattern was seen in the later evoked gamma-band responses in both SMG and LOC. While both groups showed increased evoked power in the later gamma-band response in auditory cortex, the pASD group still showed a greater increase than controls. Additionally, in FFG and LOC in both the early and late gamma response, both groups showed decreases in total power. However, controls exhibited greater reduction than the pASD group. In both the early and late time windows, there were no hemispheric differences in evoked power in controls, but there were differences between evoked power in left and right hemispheres in pASD; pASD showed increased evoked power in the right hemisphere in FFG, but evoked power was reduced in the left hemisphere. There were no significant group differences in phase-locking factor (PLF), but in both auditory cortex and SMG, PLF was greater in the left hemisphere than the right. This is of note as the opposite has been seen in non-language auditory tasks  and suggests that PLF in this study may be specifically related to language processing.
The cognitive processes involved in the word recognition task could explain the reduction in total gamma power seen in both groups, and the increase in evoked gamma seen in pASD relative to controls. While increases in evoked gamma activity are often seen when attention is increased [5, 72–75], some studies have found reduced gamma activity following repetition or priming during cognitive tasks [76, 77]. Given that the current results are in response to stimuli repeated multiple times throughout the task, it may be that total gamma activity was reduced due to word repetition. Some hypothesize that reduced gamma activity following repeated presentations of stimuli reflects the need for less neural representation as a result of a more “sharpened” neural network [76, 77]. This could explain why pASD showed increases in evoked gamma activity, while controls showed decreased evoked gamma, and why pASD also showed less reduction in total gamma power compared to controls. The reduction in total and evoked gamma activity in controls could be a result of their better formation of target word representations. Conversely, the increased evoked gamma and attenuated reduction in total gamma seen in pASD may reflect that they were less able to form a neural representation of target words.
The current study also investigated group differences in neural activity in the beta-band (13–30 Hz), as beta activity may also be involved in language processing [12, 37, 44, 45] and increased beta activity has previously been observed in children with ASD compared to healthy controls . Indeed, while both groups demonstrated increased early evoked beta activity, the current study found a significantly greater increase in pASD compared to controls in SMG, with a similar trend in both LOC and FFG. While there were no group differences in early total beta activity, there was a significant main effect of hemisphere in FFG, with greater increases in total activity seen in left compared to right hemisphere, across groups. Similarly to gamma activity, an overall decrease in total beta activity was observed, but there were no significant group differences in total beta. Interestingly, both groups demonstrated early increases in beta activity, followed by a later decrease. This could indicate cognitive differences between earlier and later stages of language processing; previous studies have suggested increased beta activity to be associated with top-down cognitive processing and decreased beta to be associated with bottom-up cognitive processing . The group differences seen in beta activity also suggest disparate cognitive function during the language task, as with gamma differences. However, given the dearth of research on beta activity in ASD, further investigation is warranted.
As language deficits are sometimes seen as part of the BAP [35, 40, 41] and the current sample of pASD did indeed show worse performance than controls on measures of phonology and receptive language, it is reasonable that this task would involve greater cognitive effort for pASD compared to controls. This hypothesis is also consistent with the pattern of group differences in neural activity being similar across regions of interest in the brain. If gamma and/or beta activity reflect task-related activation of language networks, these brain regions would be expected to be synchronously active or inhibited . Additionally, the use of word stimuli in the continuous word recognition task likely explains why a later gamma-band response was seen in addition to the early tGBR seen in previous studies using tone/click-train stimuli. In their study comparing gamma-band activation between words and pseudowords using MEG, Pulvermuller et al.  saw activation even after 700 ms post-stimulus and suggested the use of word stimuli likely accounted for the prolonged effect seen compared to tone or click stimuli.
An alternative hypothesis is that the later reduction in gamma-band activity could be due to sustained attention. Brookes et al.  found reductions in evoked gamma activity during visual working memory tasks and found this decrease to be larger during more demanding phases of the tasks. They suggested that this reduction reflects gamma activity being suppressed during working memory maintenance to prevent attention shifting to non task-related distractions. Given that attention dysfunction is common in ASD [18, 23] and may be part of the BAP , the lack of evoked gamma reduction and attenuation of total gamma reduction in pASD relative to the control group could reflect attentional dysfunction.
There were a number of interesting relationships between gamma activity and scores on language measures in the current study. For figurative language, higher scores were related to reduced evoked and increased total gamma power in the control group. This relationship was not significant for the pASD group. Receptive language scores showed a negative relationship with late total gamma in the control group, with higher scores associated with reduced evoked gamma. This fits well in the context of the hypothesis that reduced gamma may reflect a sharpened neural representation of the target words. Given that receptive language skills are important in a spoken word recognition task, this suggests that those with better receptive language performance may have shown a greater reduction in total gamma as a result of being better equipped to form a neural representation of the words. However, as with figurative language, this relationship was not significant in pASD. Similarly, higher scores on expressive language, letter fluency, and category switching accuracy in the control group were largely associated with decreased total and evoked gamma activity, with few significant relationships observed in the pASD group. That the majority of the significant relationships between language performance and gamma activity were seen in the control group, but not in pASD may suggest that language performance and gamma activity are more clearly linked in controls than in pASD overall. It should be noted, however, that these analyses were explorative in nature and should be interpreted with caution. Future studies should further explore the relationship between language performance and gamma activity by directly measuring gamma during these tasks.
There were also some relationships of note between beta-band activity and the language measures. While both pASD and controls showed an association between higher nonword repetition scores and total beta activity, this relationship was stronger for the control group. Similarly, higher figurative language scores were associated with increased total beta activity in the control group, but not in the pASD group. Higher expressive vocabularly scores in the control group were associated with reduced evoked activity in controls, but increased evoked beta in pASD. In sum, similarly to gamma-band activity, the relationship between beta-band activity and language performance appears to be different between controls and pASD and merits further investigation.
As the continuous word recognition task has been used previously to assess lateralization of language function , we also investigated lateralization of activity. Consistent with previous findings, we found early evoked gamma activity (0–200 ms post-stimulus) to be largely bilateral (with the exception of left lateralization in DLPFC), with later evoked activity (200–800 ms post-stimulus) showing greater left lateralization. Results were similar for total gamma activity, although greater lateralization of function was seen earlier. Evoked and total beta activity were more lateralized than gamma activity, but this differed among brain areas; DLPFC activity was consistently left lateralized, auditory cortex was consistently right lateralized, FFG and SMG were left lateralized in the early time window and right lateralized in the late time window, and LOC was left lateralized in the late time window and right lateralized in the early time window. This difference in lateralization supports that gamma and beta activity may play different roles in language processing [43, 45]. Interestingly, in areas for which there were group differences (LOC and DLPFC for evoked gamma; auditory cortex and DLPFC for evoked beta; SMG and DLPFC for total gamma; SMG for total beta), pASD showed greater left lateralization compared to controls. Previous studies have suggested that individuals with ASD show atypical language lateralization, with lateralization reduced [81, 82] or more rightward  compared to controls. Since greater left lateralization was seen in pASD, it may suggest a compensatory mechanism not seen in individuals with ASD.
A potential limitation of the current study is that is not possible to parse out attention effects from language effects. As such, it is difficult to tell if these results differ from those in studies using simple tone stimuli as a result of the language stimulus itself and the cognitive processes required to distinguish repeated from novel words, or as a result of participants actively attending to stimuli during the task. Future studies will adjust for these variables separately to assess the impact of each on neural activity by varying the loading on language and attention processes independently. Furthermore, standardized measures of attention can be taken in the future to address the potential relationship between neural activity and attention dysfunction. It has been suggested that attention relies on gamma synchronization to form a working memory representation of an object , so it will be important to assess the relationship between them .