Few studies of Taiwanese tones, which have a complex tonal structure, used computer-assisted acoustic analysis to measure time-varying dynamic pitch patterns in contour tones. After acquiring raw F0-over-time traces of lexical tones, we calculated F0-over-normalized time data and plotted F0 profiles along 11 equidistant time axis. Mean acoustic parameters, individual tone descriptions, and tone numbers are described. Our data show that seven lexical tones are grouped into seven tightly clustered, but distinct patterns. However they differ from previous studies in (1) Tone-5 is a mid falling-rising, rather than low-rising tone. (2) Short tones are level tones but not falling tones. (3) Short tone (entering tone) is not shorter in duration than long tones (non-entering tones). In conclusion, our study presents F0-over-normalized time patterns and duration of Taiwanese lexical tones and provides time normalization as a robust model to study tone categorization, gender tone contrast and pathologic tones.