While there is a consensus that speakers plan their utterances before they start producing them, the scope of the initial planning unit remains controversial. In subject-initial utterances, is the planning unit the whole subject phrase or a smaller functional phrase within the subject phrase? Allum and Wheeldon (2007) reported that speakers show faster onset latencies in producing utterances like The flower above the house is red, where the subject consists of two functional phrases (the flower and above the house) than in producing The flower and the house are red, where there is a single, longer functional phrase (The flower and the house), both in head-initial languages like English and head-final languages like Japanese. This has been taken to suggest that the functional phrase is a preferred unit of planning, rather than the whole subject. Experiment 1 in the present study replicates Allum and Wheeldon's study with speakers of another head-final language (Mandarin Chinese) and finds similar results. Experiments 2 and 3 investigate whether syntactic processing or visual grouping could potentially explain the pattern of responses, and find that they cannot. Together, these results provide further empirical support for the claim that the functional phrase is a primary unit of grammatical planning for speech production.