Semantic reversal anomalies under the microscope: Task and modality influences on language‐associated event‐related potentials

• Published in: The European journal of neuroscience 2020-10, Vol.52 (7), p.3803-3827
• Main Authors: Kyriaki, Louise, Schlesewsky, Matthias, Bornkessel‐Schlesewsky, Ina
• Format: Article
• Language: English
• Subjects: Auditory stimuli; Electrophysiology; Event-related potentials; Language; N400; neurolinguistics; P600; Semantics; sentence processing; Statistical analysis; Statistics; Visual perception
• ISSN: 0953-816X; 1460-9568
• DOI: 10.1111/ejn.14862

Semantic reversal anomalies (SRAs)—sentences where an implausibility is created by reversing participant roles—have attracted much attention in the literature on the electrophysiology of language. In spite of being syntactically well formed but semantically implausible, these sentences unexpectedly elicited a monophasic P600 effect in English and Dutch rather than an N400 effect. Subsequent research revealed variability in the presence/absence of an N400 effect to SRAs depending on the language examined and the choice of verb type in English. However, most previous studies employed the same presentation modality (visual) and task (acceptability judgement). Here, we conducted two experiments and three statistical analyses to investigate the influence of stimulus modality, task demand and statistical choices on event‐related potential (ERP) response patterns to SRAs in English. We reproduced a previous study's procedure and analysis (N. Bourguignon et al. (2012) Brain and Language, 122, 179–189) and further introduced between‐subjects factors of task type and modality, using mixed‐effects modelling to analyse the data. We observed an N400 effect to typical English SRAs (agent subject verbs, e.g. “the fries will eat the boys”), which contrasts existing literature and was not predicted by existing theories that account for SRA processing. Task demand modulated the ERPs elicited by SRAs, while auditory presentation led to increased comprehension accuracy and a more broadly distributed ERP. Finally, the statistical methods used influenced the presence/absence of ERP effects. Our results suggest a sensitivity of language‐related ERP patterns to methodological parameters, and we conclude that future experiments should take this into careful consideration. Building on a reproduction of an existing study, we systematically examined the effect of stimulus modality, task demand and statistical analysis (ANOVA versus linear mixed‐effects modelling) on language‐associated event‐related potentials. We observed an N400 effect to typical English semantic reversal anomalies (e.g. “the fries will eat the boys”), contrasting with the existing literature.