70 Mandarin-speaking advanced learners of English (level B2 and above) participated in a perceptual identification experiment eliciting their preferred Mandarin equivalent classifications of English fricatives and affricates (/s, ʃ, ʧ, ʤ, tr, dr, ʒ/) along with fitness rates. The degree of mapping between Mandarin and English consonants, ranging from poor to fair, and good, were compared against predictions by the Perceptual Learning Model, a theoretic model that predicts learning outcomes by phonetic distances. Overall, the perceived phonetic distances between Mandarin and English consonants predicted the learners’ correct identification of the L2 consonants except for a few number of individual sounds. The Findings suggest that phonetic similarity do predict most mappings as the learning models postulate, but other factors such as articulatory proximity and orthographic influences should be considered, too.