Synonyms, as an important part of English vocabulary, have always been a key but difficult point in EFL (English as a foreign language) teaching. In the traditional vocabulary teaching, teachers mainly distinguish synonyms via describing, translating and explaining with examples in the dictionary. However, most dictionaries can only provide separate word meanings and limited examples without explaining the different usage and meanings of the synonyms. Therefore, drawing upon Sinclair's model of extended lexical units (ELUs) and employing a corpus-assisted approach to the study of synonyms preserve and conserve, this study tries to analyse the similarities and differences of these synonyms from the perspectives of their frequency, colligation, collocation, semantic preference and semantic prosody. The findings show that corpus-assisted approach is more effective, objective and reliable in distinguishing synonyms and a number of implications can be concluded in English language teaching, learning, testing and research.