The purpose of this study is to present a thorough acoustic investigation of VOT for stops in connected speech in Taiwan Mandarin (TM) and Japanese. Aspirated and unaspirated stops in TM and voiced and voiceless stops in Japanese not only in isolated syllables but also in word-initial/-medial (Japanese) or phrase-initial/-medial (TM) positions were involved. The study found that TM aspirated stops definitely belong to the long-lag category, and unaspirated stops belong to the short-lag category. Conversely, VOT for Japanese voiceless stops in the initial position falls somewhere between short-lag and long-lag; while, VOT in the medial position becomes extremely short, almost as short as voiced stops with positive VOT. Moreover, the occurrence of prevoicing for voiced stops is random. These results show that Japanese stops are acoustically quite different from TM stops in terms of VOT, which further implies that Japanese stops might be difficult for TM speakers to distinguish correctly.