Disambiguation of two-tone images reveals semantic contributions to object recognition in the EEG

Electrophysiological responses to visual objects carry information about stimulus identity and semantic category, but it remains difficult to know whether such information represents semantic knowledge or merely regularities in physical image features. Here, we presented two-tone images while recording EEG to dissociate the learned semantic concept from physical stimulus properties in the electrophysiological signal. Seventeen healthy participants completed a semantic disambiguation experiment with three phases: ambiguous two-tone images before disambiguation, their corresponding grayscale photographs, and the same two-tone images after disambiguation. We quantified representational change using Mahalanobis distances between EEG response patterns. After disambiguation, EEG responses to learned two-tone images became more similar to the response of their photo counterparts, indicating semantic enrichment of stimulus-specific representations. This effect emerged approximately 100 ms after stimulus onset and persisted across an extended time window. However, we found no evidence for an increase in semantic category-level responses after disambiguation. Furthermore, responses to spontaneously recognized two-tone images did not already show stronger alignment with the responses to their corresponding photo images before explicit visual disambiguation was performed. Together, these findings show that EEG can capture learned semantic modulations of responses to visually degraded stimuli, while also indicating that low-level visual properties dominate the representational structure. Two-tone images can therefore serve as a useful paradigm for probing semantic learning in the EEG, and may guide future research in children and patient populations whose ability to behaviorally confirm learning is limited.