Abstract: Aiming at the prevalent mismatches between parts of speech and grammatical functions in Chinese, the author points out that the confusion brought about by word classification and the resulting disputes grow out of the lexicalist view that ascribes the formation of words and parts of speech (or categories) to the lexicon.The study shows that the relevant Chinese facts can be given a unified account under the framework of Distributed Morphology(DM)that is based on the non-lexicalist view, according to which words are generated in the syntactic computation by morphemes through formal devices such as “Merge”，so that roots are categorized by virtue of being in a local structural relationship with one of the category-defining functional heads.Therefore, parts of speech or categories are not regulations of the lexicon but are derivatives of the syntactic structure.Chinese facts provide further cross-linguistic evidence for the validity of DM.Under the basic assumptions of this theory，Chinese and English facts are subordinated to the same analytic mechanism and discrepancies between the two languages are attributed to the different properties of the roots and the local morphophonological operations at PF.