The perception of the morae with devocalized vowels in Japanese language

It is difficult for the Japanese learners to perceive and produce the morae with devocalized vowels. To make it easier for the Japanese learners to acquire them, we investigate how Japanese native speakers perceive them. Usually the first vowel of the Japanese word ‘susugi’(rinse) disappears, and so the pronunciation of ‘susugi’ is not [susugi] but [ssugi]. We made two psychophysical experiments. In the first one we shortened the part of [ss] of [ssugides] and [korewassugides] (This is rinse.) in 6 stages. [korewassugides] with shortened [ss] is easy to perceive as [korewasugides], but [ssugides] with shortened [ss] is not easy to perceive of [sugides]. In the case of [korewassugidesu], [korewa] helps the listener in measuring the length of morae of this sentence in order to perceive the shortened [ss] as a mora or two. The reason why [ssugides] with shortened [ss] is difficult to perceive as [sugides] is not the pitch of the first and the second mora but is the pitch of the second and the third mora. In the second experiment we changed the pitches of [sugides] and researched how Japanese perceive these sounds. 2. JAPANESE PHONETIC In ths Japanese language, the devocalized vowel tend to happen in the following cases. 1.When the closed vowel ([i],[u]) is between voiceless consonants ( [s], [k] ). For example, /gakuse:/ (student) >>[gakse:]. 2. When the closed vowel is following the voiceless consonant and there is a rest on its back word. For example, /desu/ >>[des]. The phonetic units of English are syllables, and the units of Japanese are syllables and morae. 'Japanese' and 'English' are divided into syllables as 'ja·pa·nese' and 'eng·lish'. The same meaning in Japanese 'ni·hon·go' (Japanese) and 'e:•go' (English). The Japanese words are also divided into morae as 'ni•ho•n•go' and 'e•e•go'. Japanese native speakers divided the word 'nihongo' into syllables as 'ni•hon•go' and also into morae as 'ni•ho•n•go'. In the former, each syllable is the s ound cluster to pronounce with ease, and in the latter each mora is the unit of time. Japanese natives perceive that each mora has the same length. The sentences have intonations and the words have accent in the sporken languages. The functions of accent is to divided the chain of sounds into some parts that compose the sentence, and to distinguish the meanings of words composed with the same speech s ounds. Japanese has pitch accent. The Japanese sound, '_ni ̄ho_n ̄bu_n-ga-ku' (the pitch of the morae 'ni', 'n' and 'n-ga-ku' which follow "_", is low, and the pitch of the morae 'ho' and 'bu' which follow " ̄", is high) means 'Japan, literature' and '_ni ̄ho-n-bu ̄n-ga-ku' means 'Japanese literature'. We have a Japanese minimal pair: '_a ̄me' (candy) vs. ' ̄a_me' (rain). Only the difference of the accent distinguishes these two words. 2.THE EXPERIMENT WITH THE SOUND CONTROLLED BY THE DURATION 2.1. Procedure [sugides] (Japanese ceder), [ssugides] (rinse), [korewasugides] (this is ceder) pronounced by an adult male standard Japanese speaker were used for the experiment as the sound stimuli. The original sounds were recorded into a digital audio tape. In TEST1, the part of [ss] in [ssugides] was shortened by computer synthesizer in 6 stages by 10%. In TEST2, similarly, the part of [ss] of [korewassugides] has been shortened to 6 stages by 10%. Six kinds of the manipulated sound stimuli and two kinds of the original sounds were ordered at random comprising one set. 10 sets of these stimuli were recorded into a digital audio tape. Stage Short[%] Duration[ms] Stage Short[% Duration[ms] 1* 0 308 1* 0 308 2 10 278 2 10 278 3 20 248 3 20 248 4 30 218 4 30 218 5 40 188 5 40 188 6 50 158 6 50 158 7 60 128 7 60 128 8* 0 137 8* 0 137 Table1:The manipulated stage of the sound stimuli. ‘Duration’ indicate the duration of [ss]. On the left table, stage1 is the original sound of [ssgides] and stage8 is the original sound of [sugides]. On the right table, stage1 is the original sound of [korewa ssugides] and stage8 is the original sound of [korewa sugides]. The subjects were divided into two groups; 20 native Japanese speakers of mu-accent (‘mu’ is non in English.) dialects, Japanese speakers that don’t distinguish meaning by pitch accent (gr oup JK), 20 native Japanese speakers who speak accent dialects Japanese (gr oup JT). The subjects were asked to listen to the tape using headphone and to mark on the answer 5th International Conference on Spoken Language Processing (ICSLP 98) Sydney, Australia November 30 -December 4, 1998 ISCA Archive http://www.isca-speech.org/archive