Speakers of Chinese and English share decodable neural semantic representations, which can be elicited by words in each language. We explore various, common models of semantic representation and their correspondences to each other and to these neural representations. Despite very strong cross-language similarity in the neural data, we find that two versions of a corpus-based semantic model do not show the same strong correlation between languages. Behavior-based models better approximate cross-language similarity, but these models also fail to explain the similarities observed in the neural data. Although none of the examined models explain cross-language neural similarity, we explore how they might 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 models do further correlate with neural data after accounting for cross-language similarities.