This is a study of Mandarin third tone (T3) sandhi produced by learners learning Mandarin as a second language, third language, or heritage language. I investigate factors affecting learners’ Mandarin T3 sandhi performance measured by two constructs, comprehensibility and accentedness. T3 sandhi in Mandarin is a phonological property that learners of Mandarin need to acquire on top of the four lexical tones (i.e., Tone 1 high level, Tone 2 rising tone, Tone 3 dipping tone, and Tone 4 falling tone). T3 sandhi is a process that which lexical tones alternate categorically, changing from the underlying tone sequence of T3T3 to T2T3. This process is motivated by the underlying trochaic feet of Mandarin (Qu, 2013). 67 Chinese learners passed the screening test (i.e., reading monosyllabic words with satisfactory tone production) to ensure that they could produce basic lexical tones before their tone sandhi production was evaluated. The eligible learners’ reading of the experimental wordlist that consists of 40 disyllabic words (i.e., 20 sandhi words and 20 non-sandhi words) was judged by 20 native speakers of Chinese in terms of comprehensibility and accentedness on a scale of 1 to 9 (e.g., Munro & Derwing, 1995; Saito, Trofimovich & Isaacs, 2017). The primary findings of the study are 1) Comprehensibility and accentedness were confirmed to be two distinct constructs as learners were found to perform significantly differently on these two constructs (both p < .05). 2) Previously learned foot structure, either from first or second language (L1footness or L2footness), were found significantly predict L3 comprehensibility and accentedness. L1footness was correlated with better performance: higher comprehensibility and lower accentedness ratings. The finding of L2footness’ correlation with worse performance in comprehensibility and accentedness was confounding but caused by low exposure to the target language Mandarin. 3) Exposure to the target language, measured by total learning length, the number of Chinese courses taken and total time spent in China, was found significantly influence comprehensibility and accentedness. This shows the importance of teasing apart effects of exposure and language transfer in L3 acquisition studies, which echoes with Puig-Mayenco and Rothman (2020). 4) Heritage learners were not found to have any acquisitional advantages over non-heritage learners as there were no significant differences between heritage versus non-heritage learners. Furthermore, Cantonese learners were found to perform worse than L2 learners on T3 sandhi words (where T3 sandhi rules need to apply) but not on non-sandhi words due to their lack of foot structure in their heritage language Cantonese. This suggests the heterogeneous nature of the Chinese heritage learner population, and Cantonese heritage learners and Mandarin heritage learners should be distinguished at least for prosodic feature acquisition. / Graduate