In the surface structure of Chinese nominal modifiers (quantifiers, determiners, adjectives, measure phrase, relative clause, etc.) may occur either before or after a modified noun. In most of the transformational studies of Chinese syntax (e.g. Cheng 1966; Hashimoto 1966; Mei 1972; Tai 1973; Teng 1974), it has been assumed that such NP's have the underlying order of modifier + noun (M-N, henceforth) with the variant surface order of noun + modifier (N-M, henceforth) being derived. This paper, however, first argues that Chinese has the underlying order of N-M and that the variant M-N order is due to an optional transformation that reorders the N-M sequence. It is shown that the M-N hypothesis requires a set of complex constraints on the rule postposing modifiers in order to account for various surface patterns exhibited by NP's containing more than one modifier. It will be argued that in the N-M hypothesis proposed here various surface patterns can best be accounted for by constraining the N-M inversion rule to the following effect; namely, it can apply to any NP, either a higher one or a lower one, but only once within the domain of a possible complex NP. (Author)