Vowel length contrasts in Japanese, e.g., chizu “map” vs. chiizu “cheese”, are cued primarily by ... more Vowel length contrasts in Japanese, e.g., chizu “map” vs. chiizu “cheese”, are cued primarily by vowel duration. However, since short and long vowel durations overlap considerably in ordinary speech, learning to perceive vowel length contrasts is complex. Meanwhile, infant-directed speech (IDS) is known to “exaggerate” certain properties of adult-directed speech (ADS). If so, then it is possible that vowel length contrasts might also be exaggerated in IDS. To investigate this, the present study analyzed vowel durations in the RIKEN Japanese Mother-Infant Conversation Corpus, which contains 11 hours of IDS by 22 mothers talking with their 18-to-24-month-old infants, and 3 hours of ADS by the same mothers. Results indicated that vowel length contrasts were generally not exaggerated in IDS, except at the end of prosodic phrases. Furthermore, several factors that systematically affected vowel duration in IDS were identified, including phrase-final lengthening and “non-lexical lengthening”, i.e., the lengthening of vowels for emphatic or other stylistic purposes. These results suggest that vowel duration in Japanese IDS could not only potentially facilitate learning of lexical distinctions, but also signal phrase boundaries, emphasis, or other communicative functions.
The Journal of the Acoustical Society of America, 2017
Non-native speakers often have difficulty accurately producing and perceiving the syllable struct... more Non-native speakers often have difficulty accurately producing and perceiving the syllable structure of a second language. For example, Japanese learners of English often insert epenthetic vowels when producing English words, e.g., stress produced as /sutoresu/. Similarly, when asked to count syllables in spoken English words, they frequently overestimate the number of syllables, suggesting that they tend to perceptually insert epenthetic vowels between adjacent consonants. These tendencies suggest the possibility that learners may have difficulty distinguishing between English words that contrast in syllable count, i.e., words that differ in the presence/absence of a vowel, e.g., blow-below, sport-support. Furthermore, if listeners perceptually insert epenthetic vowels, then they should misperceive blow as below more often than below as blow. To test these predictions, Japanese listeners participated in a 2AFC identification task, using 78 English minimal pairs contrasting in syllable count such as blow-...
10th International Conference on Speech Prosody 2020, 2020
It is well attested that vowels are longer before voiced than voiceless consonants in English. Vo... more It is well attested that vowels are longer before voiced than voiceless consonants in English. Vowel duration in productions by American English speakers (AE) is affected especially by the voicing of postvocalic consonants and the quality of the target vowel [1]. Similar voicing effects in Japanese infants and adults, and also in English produced by Japanese learners (JE) were found as well [2]. However, previous studies on voicing effects in JE overlooked an important factor: gemination in loanwords. Data from a production experiment using monosyllabic English words with and without gemination in their borrowed counterparts were analyzed. Participants were 15 university students in Tokyo from intermediate to advanced proficiency levels in English, and two AE speakers for control. Results showed that vowel duration for JE was conditioned not by post-vocalic voicing, but instead by geminatability of the word, which was independently judged by 12 Japanese listeners in a different experiment. Post-vocalic stop closure duration was longer for voiceless than voiced stops, and longer for words that were judged to have a geminate stop than words that did not, for JE but not for AE speakers. Thus, voicing effects attested in AE speech is not straightforwardly realized in JE because of the interference from Japanese prosody.
The present study investigated the extent to which lexical factors, e.g. lexical status and compe... more The present study investigated the extent to which lexical factors, e.g. lexical status and competition, affect phonetic realization of phonological structure, by examining the singleton-geminate contrast in Japanese. Stimuli were twoand three-mora minimal pairs contrasting in singleton vs. geminate stops, half of which were real words, e.g. /kako/, while the other half were similar-sounding nonwords, e.g. */nako/. Furthermore, half of the items had a lexical competitor contrasting in quantity, e.g. /kako//kakko/, while the other half did not, e.g. /tako/*/takko/. Thirty-two native Japanese speakers read the target items interspersed with filler items. Results revealed that the vowel preceding the singleton or geminate stop was shorter for real words than for nonwords. However, no significant effect of lexical competition was found. These results suggest that phonetic realization of words is affected by some lexical factors even in a non-stress-based language such as Japanese.
Phonetics and phonology both study the sounds of language, so it is not surprising that one is re... more Phonetics and phonology both study the sounds of language, so it is not surprising that one is relevant to the other. However, in reality, there seems to be a deep chasm that separates the two fields. For one thing, phonetics has sometimes been treated as "extra-grammatical" and outside of linguistics proper, while phonology is considered to be part of the grammar (Anderson (1981)). There are likely many reasons for this division, one of which is the mutual incompatibility between the continuous, gradient nature of phonetic phenomena and the discrete, categorical nature of phonological patterns. Another reason might pertain to the goals of each subdiscipline; the goals of phonetics are to understand how speech is produced and perceived, what its physical characteristics are, and how it relates to linguistic structure, while the goals of phonology are to make generalizations about language-general and language-specific characteristics of sound patterns, and how they relate ...
Native English listeners’ perception of Japanese phonemic length contrasts was investigated using... more Native English listeners’ perception of Japanese phonemic length contrasts was investigated using a minimalpair identification task with Japanese words that differed by the length of a vowel or consonant, e.g., “haken”–“hakken”. Identification accuracy was found to be higher for listeners with Japanese exposure than for unexposed listeners, and higher after five days of perceptual identification training than before. Even though listeners were trained to identify words containing vowel length contrasts only, accuracy improved for other length contrasts also. However, accuracy was relatively poor when the contrasts occurred in certain positions within the target word. In sum, while training improved English listeners’ perceptual ability, a significant gap remained before they reached native-like performance.
Studies on phonetic realization of phonological structure have revealed that certain lexical prop... more Studies on phonetic realization of phonological structure have revealed that certain lexical properties are reflected in fine details of speech production. Examples are vowel space expansion for words with dense neighbors, and VOT increase in voiceless stops for words with voicing minimal pairs. These lexical effects are primarily found in English, but there is much to explore in phonemically and prosodically different languages, such as Japanese. The present study investigated the duration of intervals around the burst of word initial velar stops in the Corpus of Spontaneous Japanese which has about 200 speakers’ annotations given by trained phoneticians. Minimal pair competitors and word familiarity data are drawn from an 80,000-word database. Results show significant effects of competitors on the duration of the following vowel and word familiarity on the closure duration, which suggest that lexical effects do exist in Japanese, but in a radically different manner from English.
A speech production experiment was conducted to investigate how well native Japanese speakers can... more A speech production experiment was conducted to investigate how well native Japanese speakers can produce a non-contrastive sound in American English, namely, intervocalic alveolar flaps, and how their production varies as a function of language experience. Native Japanese speakers who have prior experience living in the US produced English sentences containing potentially flappable segments. Results showed that Japanese speakers produced alveolar flaps to varying degrees. Acoustic analysis revealed that flaps produced by Japanese speakers show properties that have previously been associated with them, such as short duration and continuous voicing. Results also showed that Japanese speakers produced alveolar flaps within words, e.g. letter, and in certain types of phrases, e.g. get on. Finally, speakers who frequently produced alveolar flaps tended to be those who arrived in the US early in their life, who stayed in the US for a long time period, and who had high scores on TOEFL iBT.
In this paper, the results of the experiment for young normal-hearing listeners were presented, w... more In this paper, the results of the experiment for young normal-hearing listeners were presented, which is a part of the whole project, the perception experiment for young normal-hearing and elderly listeners. The perception experiment was conducted by using geminate and non-geminate words with /shu/ and /shi/ (referred to as /Su/ and /Si/), 'shukan,' 'shukkan,' 'shikan' and 'shikkan,' having devoiced and non-devoiced vowels to investigate if the lower limit frequency (F L) and its falling affect to discriminate /Su/ from /Si/ with devoiced vowels by young normal-hearing listeners. The synthesised fricative sound was used for C 1 , and the frication has three types of F L , 1,440 Hz, 2,760 Hz and 4,080 Hz. Moreover, there are two patterns of the frication with each F L : 1) F L of the frication was kept in the same frequency until the end of the frication in each F L , and 2) the frequencies fall from each F L to 480 Hz in the middle of the frication. The rate which they perceive as /Su/ (referred to as the response rate of /Su/) for stimuli with devoiced vowels in F L falling condition was higher than in F L non-falling condition. The response rates of /Su/ in the both devoiced/ non-devoiced condition were also higher in case that the F L was 1,440 Hz regardless of F L falling. Hence, these results implied that F L itself and F L falling affect to distinguish /Su/ from /Si/ with devoiced vowels by young normal-hearing listeners. In the results of geminate perception, the tendency was observed that the response rates of geminate for the stimuli with devoiced vowels were almost same as the response rates for the stimuli with non-devoiced vowels, suggesting it was possible for young normal-hearing listeners to discriminate geminate from non-geminate even when vowels were devoiced.
Alveolar flaps are non-contrastive allophonic variants of alveolar stops in American English. A l... more Alveolar flaps are non-contrastive allophonic variants of alveolar stops in American English. A lexical decision experiment was conducted with Japanese learners of English (JE) to investigate whether second-language (L2) learners are sensitive to such allophonic variations when recognizing words in L2. The stimuli consisted of 36 isolated bisyllabic English words containing word-medial /t/, half of which were flapfavored words, e.g. city, and the other half were [t]-favored words, e.g. faster. All stimuli were recorded with two surface forms: /t/ as a flap, e.g. city with a flap, or as [t], e.g. city with [t]. The stimuli were counterbalanced so that participants only heard one of the two surface forms of each word. The accuracy data indicated that flap-favored words pronounced with a flap, e.g. city with a flap, were recognized significantly less accurately than flap-favored words with [t], e.g. city with [t], and [t]-favored words with [t], e.g. faster with [t]. These results suggest that JE learners prefer canonical forms over frequent forms produced with context-dependent allophonic variations. These results are inconsistent with previous studies that found native speakers' preference for frequent forms, and highlight differences in the effect of allophonic variations on the perception of native-language and L2 speech.
A computer-based instruction system for language learners, dubbed ATR CALL, is described. This sy... more A computer-based instruction system for language learners, dubbed ATR CALL, is described. This system is a collection of speech perception, production and/or comprehension training tools, which have been mainly developed for second language learners. Each component focuses on the acoustic-phonetic, prosodic, lexical, or semantic decoding level of spoken language. Speech analysis tools for language learners are also provided. State-of-the-art technologies are used for each tool. The current system consists of a Japaneseteaching system as well as an English-teaching system.
The paper reports the results of an experiment in second-language speech that provides further ev... more The paper reports the results of an experiment in second-language speech that provides further evidence for the view that syllable onsets and codas are asymmetric, and for the view that English proficiency affects Japanese listeners’ speech perception performance. When two groups of Japanese listeners (Japanese college students and Japanese teachers of English) were asked to count syllables in spoken English words and nonwords, the performances of both groups declined as the number of consonants in the target item increased, but onsets led to a more drastic decline in performance than did codas. The results of Japanese teachers of English further revealed that upper-level English teachers performed the task significantly better than Japanese college students whereas lower-level English teachers did significantly worse. Among the three groups of participants, lower-level English teachers were the most strongly affected by phonological constraints on syllables in Japanese.
Vowel length contrasts in Japanese, e.g., chizu “map” vs. chiizu “cheese”, are cued primarily by ... more Vowel length contrasts in Japanese, e.g., chizu “map” vs. chiizu “cheese”, are cued primarily by vowel duration. However, since short and long vowel durations overlap considerably in ordinary speech, learning to perceive vowel length contrasts is complex. Meanwhile, infant-directed speech (IDS) is known to “exaggerate” certain properties of adult-directed speech (ADS). If so, then it is possible that vowel length contrasts might also be exaggerated in IDS. To investigate this, the present study analyzed vowel durations in the RIKEN Japanese Mother-Infant Conversation Corpus, which contains 11 hours of IDS by 22 mothers talking with their 18-to-24-month-old infants, and 3 hours of ADS by the same mothers. Results indicated that vowel length contrasts were generally not exaggerated in IDS, except at the end of prosodic phrases. Furthermore, several factors that systematically affected vowel duration in IDS were identified, including phrase-final lengthening and “non-lexical lengthening”, i.e., the lengthening of vowels for emphatic or other stylistic purposes. These results suggest that vowel duration in Japanese IDS could not only potentially facilitate learning of lexical distinctions, but also signal phrase boundaries, emphasis, or other communicative functions.
The Journal of the Acoustical Society of America, 2017
Non-native speakers often have difficulty accurately producing and perceiving the syllable struct... more Non-native speakers often have difficulty accurately producing and perceiving the syllable structure of a second language. For example, Japanese learners of English often insert epenthetic vowels when producing English words, e.g., stress produced as /sutoresu/. Similarly, when asked to count syllables in spoken English words, they frequently overestimate the number of syllables, suggesting that they tend to perceptually insert epenthetic vowels between adjacent consonants. These tendencies suggest the possibility that learners may have difficulty distinguishing between English words that contrast in syllable count, i.e., words that differ in the presence/absence of a vowel, e.g., blow-below, sport-support. Furthermore, if listeners perceptually insert epenthetic vowels, then they should misperceive blow as below more often than below as blow. To test these predictions, Japanese listeners participated in a 2AFC identification task, using 78 English minimal pairs contrasting in syllable count such as blow-...
10th International Conference on Speech Prosody 2020, 2020
It is well attested that vowels are longer before voiced than voiceless consonants in English. Vo... more It is well attested that vowels are longer before voiced than voiceless consonants in English. Vowel duration in productions by American English speakers (AE) is affected especially by the voicing of postvocalic consonants and the quality of the target vowel [1]. Similar voicing effects in Japanese infants and adults, and also in English produced by Japanese learners (JE) were found as well [2]. However, previous studies on voicing effects in JE overlooked an important factor: gemination in loanwords. Data from a production experiment using monosyllabic English words with and without gemination in their borrowed counterparts were analyzed. Participants were 15 university students in Tokyo from intermediate to advanced proficiency levels in English, and two AE speakers for control. Results showed that vowel duration for JE was conditioned not by post-vocalic voicing, but instead by geminatability of the word, which was independently judged by 12 Japanese listeners in a different experiment. Post-vocalic stop closure duration was longer for voiceless than voiced stops, and longer for words that were judged to have a geminate stop than words that did not, for JE but not for AE speakers. Thus, voicing effects attested in AE speech is not straightforwardly realized in JE because of the interference from Japanese prosody.
The present study investigated the extent to which lexical factors, e.g. lexical status and compe... more The present study investigated the extent to which lexical factors, e.g. lexical status and competition, affect phonetic realization of phonological structure, by examining the singleton-geminate contrast in Japanese. Stimuli were twoand three-mora minimal pairs contrasting in singleton vs. geminate stops, half of which were real words, e.g. /kako/, while the other half were similar-sounding nonwords, e.g. */nako/. Furthermore, half of the items had a lexical competitor contrasting in quantity, e.g. /kako//kakko/, while the other half did not, e.g. /tako/*/takko/. Thirty-two native Japanese speakers read the target items interspersed with filler items. Results revealed that the vowel preceding the singleton or geminate stop was shorter for real words than for nonwords. However, no significant effect of lexical competition was found. These results suggest that phonetic realization of words is affected by some lexical factors even in a non-stress-based language such as Japanese.
Phonetics and phonology both study the sounds of language, so it is not surprising that one is re... more Phonetics and phonology both study the sounds of language, so it is not surprising that one is relevant to the other. However, in reality, there seems to be a deep chasm that separates the two fields. For one thing, phonetics has sometimes been treated as "extra-grammatical" and outside of linguistics proper, while phonology is considered to be part of the grammar (Anderson (1981)). There are likely many reasons for this division, one of which is the mutual incompatibility between the continuous, gradient nature of phonetic phenomena and the discrete, categorical nature of phonological patterns. Another reason might pertain to the goals of each subdiscipline; the goals of phonetics are to understand how speech is produced and perceived, what its physical characteristics are, and how it relates to linguistic structure, while the goals of phonology are to make generalizations about language-general and language-specific characteristics of sound patterns, and how they relate ...
Native English listeners’ perception of Japanese phonemic length contrasts was investigated using... more Native English listeners’ perception of Japanese phonemic length contrasts was investigated using a minimalpair identification task with Japanese words that differed by the length of a vowel or consonant, e.g., “haken”–“hakken”. Identification accuracy was found to be higher for listeners with Japanese exposure than for unexposed listeners, and higher after five days of perceptual identification training than before. Even though listeners were trained to identify words containing vowel length contrasts only, accuracy improved for other length contrasts also. However, accuracy was relatively poor when the contrasts occurred in certain positions within the target word. In sum, while training improved English listeners’ perceptual ability, a significant gap remained before they reached native-like performance.
Studies on phonetic realization of phonological structure have revealed that certain lexical prop... more Studies on phonetic realization of phonological structure have revealed that certain lexical properties are reflected in fine details of speech production. Examples are vowel space expansion for words with dense neighbors, and VOT increase in voiceless stops for words with voicing minimal pairs. These lexical effects are primarily found in English, but there is much to explore in phonemically and prosodically different languages, such as Japanese. The present study investigated the duration of intervals around the burst of word initial velar stops in the Corpus of Spontaneous Japanese which has about 200 speakers’ annotations given by trained phoneticians. Minimal pair competitors and word familiarity data are drawn from an 80,000-word database. Results show significant effects of competitors on the duration of the following vowel and word familiarity on the closure duration, which suggest that lexical effects do exist in Japanese, but in a radically different manner from English.
A speech production experiment was conducted to investigate how well native Japanese speakers can... more A speech production experiment was conducted to investigate how well native Japanese speakers can produce a non-contrastive sound in American English, namely, intervocalic alveolar flaps, and how their production varies as a function of language experience. Native Japanese speakers who have prior experience living in the US produced English sentences containing potentially flappable segments. Results showed that Japanese speakers produced alveolar flaps to varying degrees. Acoustic analysis revealed that flaps produced by Japanese speakers show properties that have previously been associated with them, such as short duration and continuous voicing. Results also showed that Japanese speakers produced alveolar flaps within words, e.g. letter, and in certain types of phrases, e.g. get on. Finally, speakers who frequently produced alveolar flaps tended to be those who arrived in the US early in their life, who stayed in the US for a long time period, and who had high scores on TOEFL iBT.
In this paper, the results of the experiment for young normal-hearing listeners were presented, w... more In this paper, the results of the experiment for young normal-hearing listeners were presented, which is a part of the whole project, the perception experiment for young normal-hearing and elderly listeners. The perception experiment was conducted by using geminate and non-geminate words with /shu/ and /shi/ (referred to as /Su/ and /Si/), 'shukan,' 'shukkan,' 'shikan' and 'shikkan,' having devoiced and non-devoiced vowels to investigate if the lower limit frequency (F L) and its falling affect to discriminate /Su/ from /Si/ with devoiced vowels by young normal-hearing listeners. The synthesised fricative sound was used for C 1 , and the frication has three types of F L , 1,440 Hz, 2,760 Hz and 4,080 Hz. Moreover, there are two patterns of the frication with each F L : 1) F L of the frication was kept in the same frequency until the end of the frication in each F L , and 2) the frequencies fall from each F L to 480 Hz in the middle of the frication. The rate which they perceive as /Su/ (referred to as the response rate of /Su/) for stimuli with devoiced vowels in F L falling condition was higher than in F L non-falling condition. The response rates of /Su/ in the both devoiced/ non-devoiced condition were also higher in case that the F L was 1,440 Hz regardless of F L falling. Hence, these results implied that F L itself and F L falling affect to distinguish /Su/ from /Si/ with devoiced vowels by young normal-hearing listeners. In the results of geminate perception, the tendency was observed that the response rates of geminate for the stimuli with devoiced vowels were almost same as the response rates for the stimuli with non-devoiced vowels, suggesting it was possible for young normal-hearing listeners to discriminate geminate from non-geminate even when vowels were devoiced.
Alveolar flaps are non-contrastive allophonic variants of alveolar stops in American English. A l... more Alveolar flaps are non-contrastive allophonic variants of alveolar stops in American English. A lexical decision experiment was conducted with Japanese learners of English (JE) to investigate whether second-language (L2) learners are sensitive to such allophonic variations when recognizing words in L2. The stimuli consisted of 36 isolated bisyllabic English words containing word-medial /t/, half of which were flapfavored words, e.g. city, and the other half were [t]-favored words, e.g. faster. All stimuli were recorded with two surface forms: /t/ as a flap, e.g. city with a flap, or as [t], e.g. city with [t]. The stimuli were counterbalanced so that participants only heard one of the two surface forms of each word. The accuracy data indicated that flap-favored words pronounced with a flap, e.g. city with a flap, were recognized significantly less accurately than flap-favored words with [t], e.g. city with [t], and [t]-favored words with [t], e.g. faster with [t]. These results suggest that JE learners prefer canonical forms over frequent forms produced with context-dependent allophonic variations. These results are inconsistent with previous studies that found native speakers' preference for frequent forms, and highlight differences in the effect of allophonic variations on the perception of native-language and L2 speech.
A computer-based instruction system for language learners, dubbed ATR CALL, is described. This sy... more A computer-based instruction system for language learners, dubbed ATR CALL, is described. This system is a collection of speech perception, production and/or comprehension training tools, which have been mainly developed for second language learners. Each component focuses on the acoustic-phonetic, prosodic, lexical, or semantic decoding level of spoken language. Speech analysis tools for language learners are also provided. State-of-the-art technologies are used for each tool. The current system consists of a Japaneseteaching system as well as an English-teaching system.
The paper reports the results of an experiment in second-language speech that provides further ev... more The paper reports the results of an experiment in second-language speech that provides further evidence for the view that syllable onsets and codas are asymmetric, and for the view that English proficiency affects Japanese listeners’ speech perception performance. When two groups of Japanese listeners (Japanese college students and Japanese teachers of English) were asked to count syllables in spoken English words and nonwords, the performances of both groups declined as the number of consonants in the target item increased, but onsets led to a more drastic decline in performance than did codas. The results of Japanese teachers of English further revealed that upper-level English teachers performed the task significantly better than Japanese college students whereas lower-level English teachers did significantly worse. Among the three groups of participants, lower-level English teachers were the most strongly affected by phonological constraints on syllables in Japanese.
Uploads
Papers by Keiichi Tajima