Serializing languages refer to those demonstrating serial verb constructions (SVCs), while nonserializing languages are those having secondary predicates. Taiwanese is a language that has both constructions and thus hardly can be clearly defined as one or the other type of language. This paper looks into the structures of these two constructions to see how different or similar these two constructions are. It is argued that they appear to have similar structures in that the nonhead phrase is either a (pseudo)complement or adjunct. However, they still differ in several aspects. To illustrate, either V1 or V2 can be the head verb in SVCs, while only V1 can be the head in secondary predication. The adjunct in SVCs is positioned preverbally, while that in secondary predication occurs postverbally. These two constructions also differ in several other ways, including the status of the covert NP, and the candidates of the verbs