In Korean, the onset constraints based on the sonority preference have the following hierarchy in the initial position of word : *ONS/RHO >> *ONS/LAT >> *ONS/NAS, and it offers a possible explanation for the prohibition on word-initial lateral. However, *med[ONS/NAS dominates *med[ONS/LAT (and *med[ONS/RHO) in the middle of word. Lateralization and nasalization in Korean are the phonological variation to equalize the sonority between the coda and its following onset according to *med[ONS/NAS >> *med[ONS/LAT without violating SYLLCON and IDENT-IO(cor/ant). However, this hierarchy cannot explain why both nasalization and lateralization are possible in case of /n.l/. To find out the reason, the phonetic experiment was conducted with the subjects who speak Daegue dialect, one of Kyeongsang dialect that is considered a pitch language. According to the precedent studies, neutralization does not occur at the strong position. In case of /n.l/, when subjects pronounce the syllable with /n/ using higher pitch than the following syllable with /l/, the syllable with /n/ becomes a phonetically strong position. Accordingly, /n/ can not be pronounced as /l/ by lateralization and instead, /l/ of the following syllable is pronounced as /n/ by nasalization. If subjects pronounce the syllable with /n/ and the following syllable with /l/ with the same pitch, lateralization occurs. The pronunciation of /n.l/ can be explained by phonetically strong position, not by constraints hierarchy.

• 손일권(경북대학교)