A perception experiment with Japanese listeners is conducted to investigate the nature of place shift phenomenon that was previously found with French and English listeners. Hallé et al. [1] showed that unattested consonant sequences /tl, dl/ are perceptually repaired to form grammatically acceptable consonant clusters /kl, gl/ in the listeners’ native language.

In this study, a similar experiment with Japanese listeners, whose mother tongue lacks the onset clusters altogether, is conducted. The result explicitly shows that the place shift phenomenon ought not to be interpreted in relation to the top-down phonotactic feedback. Rather, I will argue that both labial and velar shift reflect an autonomous, signal-driven process. As such, language specificity in speech perception must reside in the listeners’ cue weighting, rather than encoded linguistic knowledge.