Language-specific phonotactic restrictions modulate the perception of L1 and L2 sound structures. However, sonority-driven onset markedness is also known to affect listeners’ perception of onset clusters; onsets of large sonority distances elicit more accurate perception than those of small sonority distances (e.g., bl versus bd versus lb , Berent et al. 2007). Although English admits only onset sequences of a large sonority rise, certain prohibited onset clusters can emerge due to word-initial schwa deletion (e.g., banana [bnǽnə], potato [ptéɪɾoʊ]). The study investigated whether native and non-native listeners were perceptually sensitive to the sonority-based onset markedness as well as to legal versus illegal English onset clusters resulting from word-initial schwa deletion. Native English, Korean, and Japanese listeners completed identity judgment tests with auditory nonce words. The nonce words contained ill-formed as well as well-formed onsets resulting from initial schwa deletion and ill-formed onsets were further divided into onsets of a sonority rise, flat, and fall. The results of accuracy tests indicated that all the listener groups differentiated between well-formed and ill-formed English onset clusters and response latency showed a similar trend. The results also revealed that the listeners showed an illusionary vowel effect as a function of the onset markedness irrespective of their L1s.