The Sonority Sequencing Principle (SSP) might explain a putatively universal preference for well-formed over ill-formed syllables with respect to onset sonority profile. In particular, a series of studies by Berent and colleagues (Berent et al., 2007) showed that blif is preferred over bnif, bnif over bdif, and bdif over lbif, by native listeners of various languages. Yet, the universal phonological nature of SSP effect is debated by authors proposing a phonetic explanation (Peperkamp, 2007). Zhao and Berent (2016) investigated Mandarin, which bans any cluster, with mixed results. We, therefore, reexamined the Mandarin case, using new materials and procedures. In Experiment 1, 16 Mandarin listeners completed a Pinyin transcription task. The materials were produced by a native Russian speaker and consisted of C1C2V(n) syllables (C1=/p, t, k/ followed by /l, r, n/ or /t, k, p/, respectively, or C1=/l/ followed by /p, t, k/; V(n)=/a, i, u, an, in/). Transcriptions showed phonological repairs, especially by vowel-epenthesis, of the offending clusters that followed SSP. In Experiment 2, 21 Mandarin listeners completed a discrimination test on no-epenthesis vs. epenthesis contrasts: kCa-kvCa and lka-lvka (C=/l, r, n, p/, v=/ə, u/). Discrimination performance followed SSP in that it was poorer for plateau than rising sonority profiles. Yet, the falling-profile lk-contrast was easier than the plateau-profile kp-contrast. Experiment 3 confirmed the pattern, using another 21 Mandarin listeners run on the kl-, kp-, and lk-contrasts with a two- or four-times shorter epenthetic /ə/. One likely explanation is suggested by the transcription data for /l/+stop syllables: /l/+stop is quite often repaired by prothesis (lka>elka/ulka) rather than anaptyxis (lka>leka), making lka-leka a rather easy contrast. On an alternative, phonetic explanation, Russian /l/ is realized close to [u] in /l/+stop, differently than in /l/+vowel, hence making lka-leka easy. These hypotheses will be tested in future work.