Autism spectrum disorder (ASD) is a range of neurodevelopmental conditions characterized by impairments in social interaction as well as restricted/repetitive behavior and interests [1]. Although challenges with language is no longer a core symptom of ASD, it is still more extensive than assumed in autistic individuals. In addition, there is marked variability in language proficiency of ASD patients, ranging from silent to verbally fluent. Compared to language production, language comprehension difficulties may be more severe and may serve as an early sign of ASD [2,3]. Semantic processing, which means understanding and categorizing the meaning of stimulus, is considered to play a vital role in language comprehension [4]. Some studies suggested that semantics might provide a foundational language skill for social interactions and social skills in children with autism [5]. Event-related potential (ERP) with millisecond resolution is an efficient and sensitive technique for characterizing subtle differences in semantic processing [6]. As reflecting the neural mechanism of language comprehension, the N400 and the P600 are the most important ERP components in language progressing [4,7].

The N400 component is one of the most powerful tools used for semantic processing and integration research. It was first noted by Kutas and Hillyard as a language measure [8]. Although it is a negative peaking over the central-posterior electrode sites, it need not be negative in absolute terms [9]. The N400 effect refers to a difference ERP created via subtraction of a congruent condition from an incongruent one [10] and reflects the activation of related semantic networks [11]. The N400 response is often elicited by two paradigms. The semantic-anomaly paradigm generates a narrative with contextually congruent or incongruent sentence-final words. The amplitude of the N400 is greater for the incongruent condition because integration is difficult in incongruent context [12]. The semantic-priming paradigm provides a related or unrelated word before a target word and the priming word is considered to be the context for the target word to be integrated [13]. The decrease in the amplitude of N400 reflects the response to the target word, which is semantically related to the previous words [14]. Both of the two paradigms reflect the process of semantic integration of the critical word with the working context. Another theory supported that the N400 response reflects facilitated activation of the long-term memory as N400 is sensitive to word frequency [15]. Generally speaking, high-frequency and early-acquired words result in shorter N400 latency and a smaller N400 amplitude relative to less frequent words as they are easier to access from memory [16]. In addition, other factors that may alter the N400 amplitude include language proficiency and modalities [9]. Now, the function of N400 has been expanded to reflect meaning processing more broadly. Nonlinguistic context such as picture sequences was found to elicit N400 effect. However, the N400 effects elicited by nonlinguistic context is more frontally-distributed in scalp topography [17].

So far, there are considerable variabilities in N400 response among individuals with autism. A few studies have concluded that autistic people exhibit decreased or disappeared N400 effect, suggesting destruction of the neural response in language processing [5,18]. Meanwhile, in some studies, the N400 effect is typical in autistic people, with more significant negativity elicited by the incongruent stimulation [19–21]. By comparing the N400 component between autistic and non-autistic group, some studies found that autistic people showed abnormal N400 response [22]. Previous studies have confirmed that N400 could also be elicited by paradigms represented with the nonverbal stimuli such as visual narratives and environmentally sound. Some studies suggested that autistic people experience deficits in the semantic processing of language, but the semantic processing of non-verbal stimuli is intact [23,24]. While some literature suggested that autistic person showed a potential benefit of using verbal material but a medium difficulty with visual long-term memory [25].

P600, a positive deflection in centro-parietal is assumed to be a family member of the late positive components (LPC). P600 is considered as another important language-related ERPs and was first characterized in the context of syntactic processing [26]. Recent studies suggested that P600 might reflect more general conflict monitoring mechanisms rather than purely syntactic processes [27]. Previous studies have not yet determined whether autistic individuals present abnormal P600 under semantic or syntactic violation conditions.

The inconsistent results of language related ERPs in ASD may be derived from the heterogeneity generated by individual differences, modalities, and stimulus materials. The individual heterogeneity is assumed to originate from age (children/adolescents or adults), diagnosis (autism or Asperger syndrome), and cognitive ability. For the heterogeneity in modalities, the N400 effect was elicited either from unimodal tasks (visual or auditory) or from cross-modal paradigms given the high multisensory nature of language cues. For the heterogeneity in materials, the N400 effect was obtained from lexical associative priming or sentence-level congruity.

To date, evidence assessing the language-related ERPs in ASD is lacking, especially when compared the language-related ERPs among different experimental stimulus and paradigms. This meta-analysis therefore updates the current knowledge on semantic processing disturbance in patients with autism. The aims of this study were (1) to examine the neural characteristics of semantic processing in autistic people; (2) to explore the factors affecting semantic processing ERP feature in patients with ASD.

This review was registered at the Prospero International Prospective Register of Systematic Reviews (PROSPERO no. CRD 42021265852) and followed guidance of conducting and reporting systematic reviews from the Cochrane handbook and the PRISMA checklist assessed by the editorial team of PRISMA.

A total of 157 primary literature related to search protocol were discovered: 42 from PubMed, 81 from Web of Science and 34 from Embase. After selecting titles and abstracts, the full texts of 30 were considered potentially relevant. In the end, 14 articles were included in the meta-analysis after excluding 16 articles for the following reason: data no available, no case-control study, and no semantically related condition. The flowchart of study selection and inclusion is shown in Fig. 1.

The characteristics of the included studies are listed in Table 1. All of the studies were conducted between 2005 and 2022. In total, this meta-analysis pooled results from 254 autistic people and 262 neurotypical controls, in which half of the subjects were children and adolescents and half were adults. Most of the studies included in the meta-analysis enrolled autistic people without intellectual disability. Two studies recruited subjects with Asperger syndrome. In one study, autistic children with minimally verbal were recruited. There are two kinds of stimuli in the studies including verbal materials and visual materials. According to the “stimulus characteristics”, articles were divided into two categories (linguistic and non-linguistic stimuli).

This is the first meta-analysis examining semantic processing in autistic individuals from the ERPs perspective and providing a comprehensive overview of current knowledge about language-related ERPs in individuals of ASD. The meta-analysis pooled the results of 18 studies and showed that the autistic people tend to display a reduced overall N400 amplitude and N400 effect compared to nonautistic people. Besides, the reduced N400 amplitude is limited to linguistic materials and no statistically significant difference was observed under non-linguistic stimuli, indicating a language-specific deficit of semantic processing in autistic individuals.

N400 is a well-studied ERP component of semantic processing [10]. Our meta-analysis revealed an overall reduced N400 amplitude in the autistic group, which suggests an abnormal semantic processing in autistic individuals. According to the assumption of weak central coherence theory, autistic individuals only focus on fine-grained detail and have difficulty integrating information within the linguistic context [43]. Studies of functional magnetic resonance imaging (fMRI) and intracranial recordings have revealed that the posterior temporal cortex, anterior temporal cortex, and inferior frontal cortex play a role in the semantic processing of linguistic [9]. The evidence from fMRI studies also supported that there is inefficient activation of the brain areas associated with semantic processing in autistic individuals [44].

The N400 effect is the difference wave between congruent and incongruent conditions. Previous studies have shown that the N400 effect is sensitive to individual’s comprehensive skills [45]. In our meta-analysis, a reduced N400 effect was found in autistic individuals, which suggests a possible developmental disability of semantic processing in autistic individuals. Although no significant difference between two groups was found in incongruous or congruous conditions, autistic individuals showed a subtle reduced N400 amplitude in the incongruous conditions and enhanced negativity in the congruous condition. This trend could be the result of a reduced N400 effect in the autistic group. Attenuated N400 under incongruent condition indicates a decreased activation of the incongruent context in autistic individuals, and the enhanced N400 in congruent condition reflects the difficulty in using context to generate expectancies in autistic individuals [46].

In addition, there is a contentious debate on whether the impairment of semantic integration in autistic individuals is a language-specific or a global impairment. The synthesized result of our meta-analysis suggested a language-specific deficit, which can also be supported by the fMRI analysis indicating that individuals of ASD rely more on visual processing in the occipital cortex than verbal processing in the temporal cortex [44,47]. Some researchers speculated that pictorial semantic processing is superior to verbal semantic processing in ASD [23]. As the fundamental role of perception processing in language development, perception deficits are closely related to language ability. The ERP study of speech-specific perception deficits in ASD also supported the result of our meta-analysis [48].

Furthermore, the results of meta-regression revealed that the effect size of N400 amplitude might be affected by the verbal IQ, indicating the difference of N400 amplitude between two groups is negatively related to language capacity. A similar finding could also be observed in subgroup analysis. For people diagnosed with Asperger syndrome, a part of family of ASD that preserved cognitive and verbal facilities, the difference of N400 amplitude was comparable. The result is consistent with the literature reported previously that N400 had a solid relationship to vocabulary learning in children and sentence comprehension in adults [49]. Furthermore, gender may be another potential moderator affecting the N400 amplitude. The result may be explained by the fact that better linguistic abilities are more common in females than in males [50]. The group differences of N400 response were observed in semantic anomalies and semantic priming paradigms, which suggests the possible difficulty in integrating semantic context and facilitating lexical access in autistic individuals [9]. However, the inverted-out-of-category words task showed no significant difference between the two groups, which means that the lexical frequency effect is less sensitive in distinguishing autistic from the nonautistic group.

The P600 is a positive deflection of compositional semantic integration processes [51]. The P600 effect can be observed in responses to syntactic violations, semantic incongruency, and pragmatic anomalies. Some studies reported that P600 was not specifically related to the language process but was more likely to reflect a general monitoring process and modulated with instructions [52]. According to our meta-analysis from limited literature, contrary to Pijnacker et al. [53], the comparison group showed significantly greater negativity than autistic individuals under linguistic stimuli. One possible reason is that the effect of P600 overlapped by the sustained anterior negativity. Previous findings have reported that the sustained anterior negativity could be explained by the increased demands on working memory load and the decreased monitoring of violations. Therefore, the presence of negativity masked the expected P600 effect of plausibility [54]. In addition, language capacity did not affect the P600 component according to the result of meta-regression. These results indicated that P600 might not be a stable index of semantic processing in ASD, even though a significant difference was detected between the two groups. Further research is needed to explore the potential role of P600 in ASD. In the contrary, the attenuated amplitude of N400 might be a reliable component that related to the poor semantic ability in ASD. More work is needed to address whether N400 has the potential to be a marker of language impairment in ASD.

Our meta-analysis contains some limitations. First, only a few studies recruited autistic participants with minimally verbal ability [55]. The characteristic of N400 waveform in individuals of ASD with intellectual disability should be further explored. Second, the number of studies was relatively limited. Some researches were excluded from this study because no N400 response was elicited. Third, given that the amplitude response of N400 was most susceptible to manipulation and the latency was generally stable [10], only a limited number of studies reported the N400 latency. Therefore, our meta-analysis only incorporated the N400 amplitude index and did not include the latency indicators. Last but not least, as much of the literature reported previously, the ASD group shows atypical lateralization patterns of ERPs to speech stimuli [56]. The relationship between cerebral lateralization and language-related ERPs in autistic individuals should be considered in future studies.

This meta-analysis indicates an overall pattern of a reduced N400 amplitude and difference wave in autistic people under linguistic stimuli. The N400 amplitude is modulated by verbal intelligence and gender. In conclusion, our findings suggest a language-specific semantic processing deficiency in autistic people from some aspects of neurocognitive functions. The decrease of N400 amplitude might be a promising indication of poor language capacity in ASD.