When a speaker of English and a speaker of Chinese think about the same object, their brains are representing a shared concept. However, we don’t know how similarity in the concepts evoked by words is manifested in the brains of speakers of different languages. We have previously shown that neural similarity relations are strongly conserved across subjects, allowing across-subject decoding. Here we extend that result to translating word-elicited activations across groups of speakers of Chinese and English. Specifically, by matching the neural similarity relations elicited by a set of seven Chinese words, presented to Chinese speakers, with the neural similarities elicited by the equivalent English words presented to English speakers, we are able to translate between the English and Chinese words with 100% accuracy, based only on the patterns of functional activity that they elicit. This demonstration provides evidence for the conservation of semantic relations between concepts across different languages.