Abstract  This study provides supporting evidence for the claim in Sohn (Studies in Phonetics, Phonology and Morphology 12:307–324, 2006) that the nature of the geminate output in Korean sonorant assimilation is crucially dependent on the stem-final sonorant. Given the generally accepted claim of positional faithfulness to the onset (Lombardi, Natural Language and Linguistic Theory 13:39–74, 1995; Beckman, Positional faithfulness, University of Massachusetts, Amherst, 1998; Casali, Resolving hiatus, University of California, Los Angeles, 1998), however, the coda saliency effect observed in Korean sonorant assimilation is intriguing. Drawing on a theory of Licensing by Cue and the P-map hypothesis (Steriade, Phonetics in phonology: The case of laryngeal neutralization, UCLA, 1997; Steriade, The phonology of perceptibility effects: The P-map and its consequences for constraint organization, UCLA, 2001), this study claims that coda saliency is a language-specific phonological reflex of the contrast asymmetry in the lexical representation and in acoustic cues. The asymmetry in functional load depending on the position is reflected in the perception of similarity, namely whether the difference between two alveolar sonorants is less distinctive or robust in word-initial vs. word-final position. Based on similarity ranking by reference to the perceptibility difference, this study argues that the marginally contrastive sonorant in the onset is more likely to be the target of neutralization than is the sonorant in the coda, whose contrastiveness is well-grounded in Korean phonology.