Patrick Chu, PhD
Research on Interlanguage Speech Intelligibility Benefit (ISIB)
Bibliographies of Interlanguage Speech Intelligibility Benefit (ISIB)
How Cantonese listeners process Mandarin tones: Implications for the second-language phonological lexicon
Patrick Chun Kau Chu and Marcus Taft University of New South Wales, Sydney, Australia
There are major corresponding relationships between the tones in Cantonese and Mandarin words. For example, 93% of the words that are pronounced as tone 1 in Cantonese are pronounced as tone 1 in Mandarin (e.g., 光 ‘light’, pronounced gwong1 in Cantonese and guang1 in Mandarin). Pronunciation errors arise when Cantonese speakers overgeneralize these regularities to ‘irregular-tone’ words. For example, the character 詢 ‘inquire’
(pronounced soen1 in Cantonese and xun2 in Mandarin) may be mispronounced as xun1 in Mandarin by Cantonese speakers due to the inappropriate application of the major corresponding rules. A disyllabic word transcription task was carried out on Cantonese and Mandarin listeners to examine whether non-native listeners are better than native listeners in understanding Mandarin tone-mispronounced words which show characteristics of Cantonese accent. Listeners heard mispronounced (e.g., xun1wen4) or the correctly pronounced version (e.g., xun2wen4) of words (e.g., 詢問 ‘inquire’) spoken by a native Mandarin speaker and wrote down the words in Chinese characters. They were told that the speaker was from a
Cantonese background and were asked to guess the intended words spoken by the speaker. Results showed that Cantonese listeners understood these mispronounced words better than Mandarin listeners, supporting an interlanguage speech intelligibility benefit (c.f. Hayes-Harb et al., 2008). When Cantonese listeners’ phonological proficiency improves, their recognition accuracy for the correct pronunciation improves while that of the mispronounced word does not deteriorate. A dual-route second language spoken word recognition and production model is proposed to account for these findings. Word recognition is purely sublexical while production is both lexical and sublexical, and the strengths of the two routes depend on
language proficiency.
Rachel Hayes-Harb, Bruce Smith, Tessa Bent & Ann Bradlow. (2008) The interlanguage speech intelligibility benefit for native speakers of Mandarin: Production and perception of English word-final voicing contrasts, Journal of Phonetics, 36, 664-679.
Summary:
With low L2 proficiency, Cantonese/Mandarin bilinguals often incorrectly pronounce Mandarin words with corresponding Cantonese tones. However, ability to understand tone-mispronounced Mandarin words is unaffected by L2 proficiency. A dual-route L2 spoken word recognition and production model is proposed whereby recognition is purely sublexical while production is both lexical and sublexical.
(pronounced soen1 in Cantonese and xun2 in Mandarin) may be mispronounced as xun1 in Mandarin by Cantonese speakers due to the inappropriate application of the major corresponding rules. A disyllabic word transcription task was carried out on Cantonese and Mandarin listeners to examine whether non-native listeners are better than native listeners in understanding Mandarin tone-mispronounced words which show characteristics of Cantonese accent. Listeners heard mispronounced (e.g., xun1wen4) or the correctly pronounced version (e.g., xun2wen4) of words (e.g., 詢問 ‘inquire’) spoken by a native Mandarin speaker and wrote down the words in Chinese characters. They were told that the speaker was from a
Cantonese background and were asked to guess the intended words spoken by the speaker. Results showed that Cantonese listeners understood these mispronounced words better than Mandarin listeners, supporting an interlanguage speech intelligibility benefit (c.f. Hayes-Harb et al., 2008). When Cantonese listeners’ phonological proficiency improves, their recognition accuracy for the correct pronunciation improves while that of the mispronounced word does not deteriorate. A dual-route second language spoken word recognition and production model is proposed to account for these findings. Word recognition is purely sublexical while production is both lexical and sublexical, and the strengths of the two routes depend on
language proficiency.
Rachel Hayes-Harb, Bruce Smith, Tessa Bent & Ann Bradlow. (2008) The interlanguage speech intelligibility benefit for native speakers of Mandarin: Production and perception of English word-final voicing contrasts, Journal of Phonetics, 36, 664-679.
Summary:
With low L2 proficiency, Cantonese/Mandarin bilinguals often incorrectly pronounce Mandarin words with corresponding Cantonese tones. However, ability to understand tone-mispronounced Mandarin words is unaffected by L2 proficiency. A dual-route L2 spoken word recognition and production model is proposed whereby recognition is purely sublexical while production is both lexical and sublexical.
The mental representation of Second Language Phonological Lexicons: Implications from the Recognition of Mandarin Mispronounced Words by Cantonese and Mandarin speakers
Patrick Chun Kau Chu and Marcus Taft University of New South Wales, Sydney, Australia
Cantonese speakers tend to mispronounce monosyllabic words in Mandarin where both words are homophones in Cantonese but not in Mandarin (Zeng, 2009). For example, the words 聲 ‘sound’ and 星 ‘star’ are both pronounced sing1 in Cantonese but sheng1 and xing1 in Mandarin respectively. Hence, some Cantonese speakers may mispronounce the word 聲 ‘sound’ as xing1 in Mandarin. This kind of mispronunciation is due to the negative transfer at the lexical rather than the phonological level from the first language because there is an involvement of lexical entries from both the first and second languages. A disyllabic word transcription task was carried out on Cantonese and Mandarin listeners to examine whether non-native listeners are better than native listeners in understanding Mandarin mispronounced words which show characteristics of Cantonese accent. Listeners heard mispronounced version (e.g., xing1yin1) or the correctly pronounced version (e.g., sheng1yin1) of words (e.g., 聲音 ‘sound’) spoken by a native Mandarin speaker and then wrote down the words in Chinese characters. They were told that the speaker was from a Cantonese background and may have mispronunciations in her Mandarin productions and the participants were asked to guess the intended words spoken by the speaker. Results showed that Cantonese listeners understood these mispronunciations better than Mandarin listeners, supporting an interlanguage speech intelligibility benefit for listeners (Hayes-Harb et al., 2008). It was further shown that this benefit cannot be completely accounted for by the similarity of sounds between the Cantonese pronunciation (e.g., sing1) and the Mandarin mispronounced version of the words (e.g., xing1) in a subsequent sound similarity judgment task. When Cantonese listeners’ phonological proficiency improves, their recognition accuracy for the correct pronunciation improves while that of the mispronounced word does not deteriorate. A dual-route second language spoken word recognition and production model is
proposed to account for these findings. Word recognition is purely sublexical while production is both lexical and sublexical, and the strengths of the two routes depend on language proficiency.
References:
曾子凡（2009）《香港人學說普通話》。香港﹕三聯書店。
Rachel Hayes-Harb, Bruce Smith, Tessa Bent & Ann Bradlow. (2008) The interlanguage speech intelligibility benefit for native speakers of Mandarin: Production and perception of English word-final voicing contrasts, Journal of Phonetics, 36, 664-679.
proposed to account for these findings. Word recognition is purely sublexical while production is both lexical and sublexical, and the strengths of the two routes depend on language proficiency.
References:
曾子凡（2009）《香港人學說普通話》。香港﹕三聯書店。
Rachel Hayes-Harb, Bruce Smith, Tessa Bent & Ann Bradlow. (2008) The interlanguage speech intelligibility benefit for native speakers of Mandarin: Production and perception of English word-final voicing contrasts, Journal of Phonetics, 36, 664-679.
Interlanguage Speech Intelligibility Benefit and the Mental Representation of Second Language Speech Sounds
Patrick Chun Kau Chu and Marcus Taft University of New South Wales, Sydney, Australia
The interlanguage speech intelligibility benefit is where non-native listeners are better at understanding non-native speech than native listeners. Whether this really occurs is unclear from previous research. We hypothesize that the benefit can only be found for words that are mispronounced phonologically as other words by non-native speakers (e.g., ‘thin’ mispronounced as ‘fin’ by Cantonese speakers). If such a mispronunciation is due to the non-native speaker’s inability to perceive second language (L2) phonemic contrasts (e.g., /θ-/f/), then L2 listeners with a Cantonese background may treat minimal pairs having these phonemic contrasts as homophones (e.g., ‘thin’ – ‘fin’). Therefore, when hearing the mispronounced version of the word ‘thin’ (i.e., ‘fin’), the L2 listeners may have an advantage over native listeners in understanding the intended word.
To test this, an English monosyllabic word transcription task was carried out for Cantonese and native English listeners. Target words which were expected to be confused and mispronounced by Cantonese speakers were included in two separate experiments (e.g., ‘thin’ in Experiment 1 and ‘fin’ in Experiment 2). Control items were also included where Cantonese speakers were not expected to produce any phonological errors (e.g., ‘low’). The words were produced either by a native English speaker or a Cantonese speaker.
Results from Experiment 1 showed that Cantonese listeners recognized more Cantonese-accented words than English listeners for target but not for control words. In addition, Cantonese listeners understood the intended word better for target words that had a higher frequency (e.g., ‘thin’) than the mispronounced word (e.g., ‘fin’) relative to words that had a lower frequency (e.g., ‘buzz’) than the mispronounced word (e.g., ‘bus’). Such an impact of word frequency suggests that the two languages are represented in a single phonological system for L2 listeners.
Across the two experiments, we compared the percent of baseword responses (e.g., ‘thin’) where Cantonese and English listeners heard ‘thin’ in Experiment 1 and ‘fin’ in Experiment 2. There were more baseword responses in Experiment 1 than Experiment 2 for Cantonese listeners, and the difference was larger for the speech produced by the native English speaker than the Cantonese speaker. This demonstrates that Cantonese speakers can perceive and produce second language speech contrasts (though in a non-native manner), suggesting that, in contrast with Experiment 1, the two languages are represented in separate phonological systems for L2 listeners. Theoretical issues regarding the mental representation of second language speech sounds in native and non-native listeners and the seemingly contradictory findings in the two experiments will be discussed in terms of the Speech Learning Model.
To test this, an English monosyllabic word transcription task was carried out for Cantonese and native English listeners. Target words which were expected to be confused and mispronounced by Cantonese speakers were included in two separate experiments (e.g., ‘thin’ in Experiment 1 and ‘fin’ in Experiment 2). Control items were also included where Cantonese speakers were not expected to produce any phonological errors (e.g., ‘low’). The words were produced either by a native English speaker or a Cantonese speaker.
Results from Experiment 1 showed that Cantonese listeners recognized more Cantonese-accented words than English listeners for target but not for control words. In addition, Cantonese listeners understood the intended word better for target words that had a higher frequency (e.g., ‘thin’) than the mispronounced word (e.g., ‘fin’) relative to words that had a lower frequency (e.g., ‘buzz’) than the mispronounced word (e.g., ‘bus’). Such an impact of word frequency suggests that the two languages are represented in a single phonological system for L2 listeners.
Across the two experiments, we compared the percent of baseword responses (e.g., ‘thin’) where Cantonese and English listeners heard ‘thin’ in Experiment 1 and ‘fin’ in Experiment 2. There were more baseword responses in Experiment 1 than Experiment 2 for Cantonese listeners, and the difference was larger for the speech produced by the native English speaker than the Cantonese speaker. This demonstrates that Cantonese speakers can perceive and produce second language speech contrasts (though in a non-native manner), suggesting that, in contrast with Experiment 1, the two languages are represented in separate phonological systems for L2 listeners. Theoretical issues regarding the mental representation of second language speech sounds in native and non-native listeners and the seemingly contradictory findings in the two experiments will be discussed in terms of the Speech Learning Model.
The role of relative word frequency in interlanguage speech intelligibility benefit
Patrick Chun Kau Chu and Marcus Taft University of New South Wales, Sydney, Australia
Previous research on interlanguage speech intelligibility benefit do not have consensus on the issue whether non-native listeners are better at understanding non-native speech than native listeners. Our study showed that the relative frequency of the intended word and the mispronounced word is the mediating factor explaining the discrepant results reported in the literature. English and Cantonese listeners transcribed English words spoken by English and Cantonese speakers. Monosyllabic words which are expected to be mispronounced by Cantonese speakers as another word are included as target items (e.g. ‘thin’ pronounced as ‘fin’). Results showed that Cantonese listeners recognized more Cantonese-accented words
than English listeners, particularly when the intended word (e.g. ‘thin’) has a higher frequency than the mispronounced version (e.g. ‘fin’). However, Cantonese listeners were not better than English listeners in understanding Cantonese-accented baseline words in which there are no phonological mispronunciations by Cantonese speakers (e.g. ‘luck’). Theoretical issues regarding the mental representation of second language speech sounds in native and non-native listeners will be discussed.
than English listeners, particularly when the intended word (e.g. ‘thin’) has a higher frequency than the mispronounced version (e.g. ‘fin’). However, Cantonese listeners were not better than English listeners in understanding Cantonese-accented baseline words in which there are no phonological mispronunciations by Cantonese speakers (e.g. ‘luck’). Theoretical issues regarding the mental representation of second language speech sounds in native and non-native listeners will be discussed.
