Proceedings of the Annual Meeting of the Cognitive Science Society

Speakers of Chinese and English share decodable neuralsemantic representations, which can be elicited by words ineach language. We explore various, common models ofsemantic representation and their correspondences to eachother and to these neural representations. Despite very strongcross-language similarity in the neural data, we find that twoversions of a corpus-based semantic model do not show thesame strong correlation between languages. Behavior-basedmodels better approximate cross-language similarity, butthese models also fail to explain the similarities observed inthe neural data. Although none of the examined modelsexplain cross-language neural similarity, we explore how theymight provide additional information over and above cross-language neural similarity. We find that native speakers’ratings of noun-noun similarity and one of the corpus modelsdo further correlate with neural data after accounting forcross-language similarities.