This study focuses upon voice data from mass media to disclose the tone features of Taiwan Mandarin and to identify its acoustic and perceptual varieties. The study consists of perceptual judgment and acoustic analyses of spontaneous speech. Each candidate represents one unique kind of Taiwan Mandarin. Candidate Chen's speech represents the Mandarin variety which is heavily influenced by Taiwanese. Candidate Soong's speech represents the Mandarin variety of "Mainlanders", whose parents were born and raised in one of the provinces in Mainland China. Candidate Lien's speech represents the variety that is employed by bilingual speakers of Mandarin and Taiwanese. Based on perceptual judgments, the prosodic features of each candidate are structured. The acoustic analyses interpret some but not all of the perceptual judgments. Subtle tone variants were detected by perceptual judgments, rather than acoustic measurements.