- The Linguistic Society of Japan
- 言語研究 (ISSN:00243914)
- vol.1967, no.50, pp.85-99, 1967-01-20 (Released:2010-11-26)
The phonetic forms of the so-called soku-on (choked sound) and hatu-on (syllabic nasal) are accounted for in generative phonology by means of the introduction of three phonological rules, progressive assimilation, nasalization, and regressive assimilation, ordered in this way. It is shown that Chomsky's argument which denies the theoretical status of the phoneme in generative phonology is equally applicable to the soku-on and hatu-on; that is, there is no level of representation on which either the soku-on or hatu-on is represented by one and the same segment. However, if we assume that the soku-on and the hatu-on are, respectively, the segment “plus consonantal, minus nasal”, and the segment “plus consonantal, plus nasal”, the other features being unspecified, then the representation using the soku-on and hatu-on (say, of a word), is shown to be an inverse representation on the levels of each of the three phonological rules introduced above.(A representation is by definition inverse on the level of a rule R, if the application to it of R and the rules that follow R yields the correct phonetic representation.) The regressive assimilation rule is characterized among these three rules by the fact that on this level the inverse representation is minimal and nondistinct from the generative representation on the same level. A speculation on the perceptual mechanism is added which assigns special significance to this level. It is assumed that in perception during the analysisby-synthesis procedure the input phonetic signal is temporarily stored in the representation which contains the segments corresponding to the traditional notion of soku-on and hatuon, namely the minimal inverse representation on the level of regressive assimilation; it is then compared with the generative representation on the same level to see whether the two representations are distinct or not.