Lexically conditioned phonetic variation: A test with the voicing contrast in Japanese

Many lexical factors have been shown to influence phonetic realization of words. For example, studies have shown that voice onset time (VOT) of word-initial stops is shorter in high-frequency words than in low-frequency words, and is longer in words that form a voicing minimal pair, e.g., cod-god, than in words that do not, e.g., cop-*gop. The present study begins to ask whether such lexically conditioned phonetic variations are language-general, by examining productions of words in Japanese. The stimuli were two-mora Japanese minimal pairs contrasting in word-initial /k/ vs. /g/, half of which were real words, e.g., /kara/, while the other half were similar-sounding nonwords, e.g., */kapa/. Furthermore, half of the items had a lexical competitor contrasting in voicing, e.g., /kara/-/gara/, while the other half did not, e.g., /kana/-*/gana/. The stimuli were split so that each participant read only one member of each minimal pair. Twenty-four native Japanese speakers read the target items interspersed with filler items. Results showed opposite trends from those previous reported. Specifically, VOT for /k/-initial words was longer for (high-frequency) real words than for (low-frequency) nonwords, and was shorter for words that had a lexical competitor than for words that did not. [Work supported by JSPS.]