The goal of this study was to examine the Taiwanese Tone Sandhi Group domain. Replicating Carlson et al.'s (2005) perception study, an experiment was conducted aiming to see whether Taiwanese listeners were able to predict the occurrence and strength of upcoming boundaries - Word, Tone Sandhi Group and Utterance. Stimuli were selected from a corpus of spontaneous Swedish speech and Taiwanese read speech, so as to vary in Boundary types (no break vs. weak break vs. strong break), Fragment size (long vs. short), and Filtering (normal vs. low-pass filtered). These fragments were presented to two Taiwanese listeners who were instructed to guess which a prosodic break would follow each fragment. Results revealed that Boundary type is a factor that influenced the judgments when stimuli with all boundary types, fragment sizes and filtering conditions were combined. Taiwanese listeners are able to differentiate strong break vs. weak break vs no break in Swedish and strong break vs. weak break in Taiwanese, but not weak break vs. no break in Taiwanese. In addition, Taiwanese listeners, like American English listeners in Carlson et al's study, were able to hear the boundaries in Swedish in normal speech. Acoustic and prosodic correlates from the Taiwanese stimuli were also examined. There was a significant but small correlation found between judgments and the pause duration. Final-word duration ratio and Energy seem to be the most reliable cues to differentiate all Boundary types. Moreover, voice quality analysis suggests that breathiness, rather than creakiness, was potentially used to distinguish Tone Sandhi Group boundaries from Utterance boundaries.