Mandarin Chinese and Thai are Generalized Classifier Languages (Chierchia 1998), allowing definite bare nouns in argument positions and requiring numeral classifiers. Yet Mandarin Chinese and Thai have striking differences in the marking of modifiers. While Mandarin makes use of a single polyfunctional marker de to mark relative, possessive, and mensural modifiers, Thai makes use of three different morphemes thii/khawng/khanaat in the three different constructions, all historically derived from nouns. This paper adopts the analysis of de as a type-shifter (i.e., <et,e>, proposed by Huang (2006) to account for the distribution of simple and complex adjectives in modificational environments in Chinese, but argues that this analysis cannot be extended to thii/khawng/khanaat, which are shown to be predicate-forming operators. The authors hypothesize that nominalization performed by de is done covertly in Thai, and demonstrate how dialectal data reveals that de might be the fusion of two distinct functional markers (Zhu 1961, 1993, Huang 2006), one for predication/abstraction, just like the Thai modifier markers, and one for nominalization. The authors suggest that by unpacking de into two distinct functions, namely predication and nominalization, and by positing a silent nominalizer in Thai modification structure, some apparent differences between Chinese and Thai modifier markers in their ability to license NP-ellipsis (e.g. Simpson 2003, Saito et al 2008, and Cheng & Sybesma 2009) can be explained. Some empirical and theoretical challenges are presented at the end of the paper.