Patrick Chu - Research
1. Interlanguage Speech Intelligibility Benefit (ISIB)
When the English utterance (i.e., 'Please give me the thin book on the right') is mispronounced by a Cantonese speaker as 'Peace give me the fin book on the white', Cantonese listeners may understand the intended meaning better than native English speaker. In Mandarin, upon hearing a mispronounced nonword 'xing1yin1' (i.e., 'star' + 'sound'), Cantonese listeners may understand the intended meaning (i.e., 'sound' 聲音 sheng1yin1) better than Mandarin listeners. These are cases where non-native listeners may have an advantage over native listeners in understanding accented speech (i.e., Interlanguage Speech Intelligibility Benefit for the Listener (ISIB-Listener), Hayes-Harb, Smith, Bent, & Bradlow, 2008). Beginning Cantonese learners of Mandarin may even understand the intended meaning of a word (e.g., 'search' 搜查) better when its tone is mispronounced (i.e., sou3cha2) than when it is correctly pronounced (e.g., sou1cha2), indicating that non-native listeners find accented speech more intelligible than native speech (i.e., Interlanguage Speech Intelligibility Benefit for the Speaker (ISIB-Speaker), Hayes-Harb et al., 2008). My research aims to find out the underlying mechanisms responsible for these effects.
2. Second Language (L2) Word Production and Recognition Model
There are many cognates in Cantonese and Mandarin because most of the time the same Chinese character has a similar meaning and pronunciation in the two languages. Moreover, there are sublexical (i.e., onsets, rimes and tones) pronunciation correspondences between Cantonese and Mandarin (Zhang & Gao, 2000). For example, most of the words that are pronounced tone 2 in Cantonese are pronounced tone 3 in Mandarin. In my study, it is shown that Cantonese speakers’ phonological knowledge of Mandarin words is influenced by these regularities in a series of Mandarin pinyin transcription tasks and a character-sound matching task. For example, Cantonese speakers may mistakenly think that the word 搜 'search' (pronounced sau2 in Cantonese and sou1 in Mandarin) is pronounced with tone 3 instead of tone 1 in Mandarin (referred to as a regularity effect). However, there are some irregular onset words (e.g., 靠 'rely', pronounced kaau3 in Cantonese and kao4 in Mandarin) where Cantonese speakers seldom give a mispronunciation using the dominant correspondence (e.g., Cantonese onset /k/ to Mandarin onset /j/) due to the fact that they are pronounced the same in Cantonese and Mandarin (i.e., a congruency effect). A new three-route L2 Mandarin word production and recognition model has been proposed to account for these findings. Cantonese speakers gradually shift from a sublexical to lexical/concept route in their L2 Mandarin word production with increasing L2 phonological proficiency while both the concept/lexical and sublexical routes are used by beginning and proficient learners in their L2 Mandarin word recognition.
My ongoing and prospective research related to this model focuses on the following topics. Please feel free to contact me if you are interested in carrying out some collaborative research.
A. Extending the L2 word production model to Japanese
In order to strengthen the explanatory power of the model, it must be able to apply to other languages as well. Japanese serves as a good testing case. In Japanese, there are many onyomi kanji (音讀漢字) where the pronunciations of those characters have origins in ancient Chinese. Arising from this, Japanese moras sometimes have a systematic pronunciation correspondence with Cantonese codas (Lee, 1992). For example, the Cantonese codas /t/ and /k/ correspond to the Japanese moras /tsu/ and /ku/ respectively (e.g., 突 'sudden', pronounced dat6 in Cantonese and to.tsu in Japanese) while the Cantonese codas /m/ and /n/ both correspond to the Japanese mora /n/ (e.g., 減 'minus', pronounced gaam2 in Cantonese and ge.n in Japanese). Beginning Cantonese learners of Japanese are sensitive to these pronunciation correspondences as indicated by the above-chance accuracy rate in a forced choice pronunciation matching task. This demonstrates that the use of the sublexical route in L2 word production can also be applied to Japanese.
B. Testing the L2 Mandarin word recognition model using online and electrophysiological methods
The current L2 Mandarin word production and recognition model has been mainly developed on the basis of off-line tasks (i.e., word transcription, Mandarin pinyin transcription, and character-sound matching). The involvement of the use of the sublexical route in L2 Mandarin word recognition can be further examined using on-line methods such as lexical decision within the cross-modal priming paradigm (e.g., Broersma & Cutler, 2008, 2011), eye-tracking within the visual-world paradigm (e.g., Huettig & McQueen, 2007), and electrophysiological methods using event-related potential (ERP) techniques to investigate the time-course of the activation of different word candidates.
C. Computational modeling of the L2 Mandarin word production model
The proposed L2 Mandarin word production model has the potential to be implemented computationally using a large corpus of monosyllabic and multisyllabic Mandarin production data collected from many Cantonese speakers with different levels of L2 Mandarin phonological proficiency. The computational model can then be used to simulate the behavioral findings reported in our studies. It can also be used to examine the relative contribution of regularity and congruency to L2 Mandarin word production.
3. Accent Perception and Adaptation
A. Perceptual adaptation to accented speech
Previous studies have shown that native listeners can adapt to phonetically-deviant speech (e.g., words pronounced with a slightly deviant /s/) after only a brief exposure (e.g., Norris, McQueen, & Cutler, 2003). However, it is not clear whether such adaptation can be applied to phonological-deviant speech when a phoneme is mispronounced as another phoneme by non-native speakers (e.g., /th/ mispronounced as /f/). The phonological-deviant speech can produce real words (e.g., 'thin' mispronounced as 'fin') or nonwords (e.g., 'faith' mispronounced as /feif/). By exposing native listeners to phonological-deviant speech which are nonwords, it is expected that native listeners would expand their phonemic category (e.g., /th/) to accommodate the deviant phonological form (e.g., /f/) to facilitate their understanding of accented speech. However, our study shows that there was no increase in recognition accuracy in understanding phonological-deviant speech where a word is mispronounced as another word (e.g., 'thin' mispronounced as 'fin') after exposure to mispronounced nonwords (e.g., /feif/). My future investigations will examine whether this is due to insufficient exposure to the phonological-deviant mispronounced nonwords or the lack of feedback during the exposure phase, or simply due to the limit of accent adaptation for this kind of phonological-deviant accented speech.
B. Word recovery mechanisms from mispronounced words
When a mispronunciation is heard in English and the mispronounced word becomes a nonword, listeners tend to change the vowel instead of the consonant in recovering the intended meaning (van Ooijen, 1996). For example, upon hearing the mispronunciation 'kebra', English listeners tend to interpret it as 'kobra' rather than 'zebra'. Even though similar results were found in other Indo-European languages such as Dutch and Spanish (Cutler, Sebastian-Galles, Soler-Vilageliu, & van Ooijen, 2000), it still remains unclear whether a preference for vowels over consonants is a universal phenomenon. Moreover, there are languages like Mandarin which differ from Indo-European languages in several aspects. First, rimes are more prominent units than vowels in Mandarin, relative to Indo-European languages, because there are only two possible codas (i.e., /n/ and /ng/) and there is a simpler syllabic structure. Second, Mandarin has a tone component and, third, more syllables in Mandarin have their own meanings than happens in Indo-European languages. Upon hearing a mispronounced Mandarin disyllabic word that is a nonword (e.g., the potential Cantonese pronunciation of the word 'search' 搜查 sou1cha2 as the nonword sou3cha2), will native Mandarin listeners prefer to change the vowel (or rime) rather than the consonant in attempting to recover the meaning of the word? Or will they prefer to change the tone rather than the consonant or vowel/rime? Since the first syllable in the mispronounced word will activate some word candidates already, will they prefer to change the sublexical units in the second syllable rather than the first syllable? How does word frequency and the number of words sharing the syllable interact with the word recovery mechanisms? The answers to these questions may impose further constraints on current word recognition models.
C. Individual differences in accent adaptation ability
Almost every person who learns to speak a second language after the age of four or five will have a certain degree of accent. Sometimes accents may cause a problem for speakers of different language backgrounds who want to use English to communicate with each other. Although the ultimate goal of language learning is to ask non-native speakers to improve their pronunciation accuracy, it cannot be achieved in a short period of time and the improvement may vary from one person to another. In order to facilitate the communication between non-native and native speakers, we may approach this issue from the perspective of listeners. It is well known that some people are good at comprehending accented utterances while others are not. Having knowledge about the native language of the non-native speaker may enhance the perception of their accented speech. However, it is impossible for people to master the native languages of all non-native speakers. Instead, I would like to investigate individual differences in the ability to understand accented speech and whether accent adaption is related to cognitive factors involved in problem solving skills (e.g., IQ, attention span, short term memory, creativity, phonological awareness). Training could be provided in those areas which may indirectly improve skills on the perception of accented speech.
4. Using L2 knowledge to investigate the mental representation of L1 speech sounds
It is well documented that the mental representation of L2 speech sounds is influenced by the L1 phonological system (e.g., Flege, 1995). For example, beginning Cantonese learners of English have difficulty differentiating the English phoneme /th/ from /f/ due to the lack of a separate phonemic representation /th/ in their L1. Based on the assumption of language transfer, the speakers' L2 knowledge can be used to infer the mental representation of L1 speech sounds when there is no consensus on their categorization. For example, there are disputes over the categorization of Cantonese as a six- or nine tone system. As mentioned in our L2 word production model, Cantonese speakers are more accurate in producing a correct Mandarin pronunciation when the word follows the dominant Cantonese/Mandarin correspondence than when it does not. However, the pronunciation correspondence between Cantonese and Mandarin tones is different under the six- and nine-tone system. I have used a Mandarin pinyin transcription task with ANCOVA and regression analysis, to show that the correspondences using the six-tone system account better for the accuracy data than those of the nine-tone system, thus providing empirical support for the former .
5. Using Vietnamese to investigate the commonalities and differences between various word production and lexical processing models
The different processing mechanisms between languages (e.g., English, Chinese, Japanese, Korean) are thought to be partly influenced by the different writing systems in additional to their differences in phonological and morphological structures. While English is a phonemic language (written in Roman alphabetic letters) with a stressed based system and complex syllable structure (with consonant clusters) that is rich in inflectional morphology, Chinese (written in logographic Chinese characters) is a morphosyllabic language with a tonal system and simple syllable structure (without consonant clusters) that is rich in compounding morphology. In Chinese, a character represents a syllable and the syllable usually represents a morpheme. Korean is an alphabetic language written in hangul (The Korean alphabet) in a square-shaped format. Korean is not a tonal language and more than 70% of the words have Chinese origins which are also morphosyllabic in nature (e.g., 독립 獨立 ‘independence’). Japanese, on the other hands, is a mora-based language written in kana and kanji with a pitch-accent system. Due to these inherent differences, different phonological planning units have been proposed for different languages (i.e., phonemes for English, syllables for Chinese and moras for Japanese) and different morphological processing models have been proposed for English and Chinese.
However, as Chinese and English differs in various aspects (e.g., alphabetic letters vs. Chinese characters, stress vs. tones, complex vs. simple syllable structure, rich in inflectional vs. compounding morphonology), it is difficult to conclude whether the difference in processing mechanisms are caused by one or more of these differences. In addition, most of the new models proposed in Chinese has not be examined in other languages. Vietnamese serves as a perfect target language to reexamine these issues. Vietnamese is a phonemic language (written in alphabetic letters) with a tonal system (with tone marks shown on the alphabets) and simple syllable structure (without consonant clusters) that is rich in compounding morphology. In addition, more than 70% of the Vietnamese words have Chinese origins which are also morphosyllabic in nature (e.g., độc lập 獨立 ‘independence’). Therefore, Chinese-Vietnamese and Vietnamese-Korean are almost minimal pairs in terms of the language properties, where the former only differs in the scripts used (logographic Chinese characters vs. Roman alphabets) while the latter only differs in the use of tonal vs. tonal system. Therefore, Vietnamese can be used as the target language to reexamine the following issues to investigate the universality of the models proposed in Chinese:
Psycholinguistic issues that has been explored in English can also be reexamined in Vietnamese to investigate their universality across languages:
In addition to behavioral methods, eye-tracking and electrophysiological methods can also be employed to examine these issues.
6. The cognitive processing of linguistic and musical structures: A comparative approach
Music shares many similarities with languages. In language, there is a limited number of phonemes, but the phonemes can be combined to form thousands of words and infinite number of sentences. In music, there is a limited number of music notes (e.g., do re me fa so la ti) but they can be combined to for an infinite number of melodies. Western music tends to use all the seven notes (i.e., heptatonic scale) in their melodies while most Chinese music tends to use only five notes (i.e., do re me so la) per scale (i.e., pentatonic scale) in their melodies. This is similar to the case in language where different languages use different combinations of phonemes in forming words. There are syntactic rules in languages, so we have a sense of which sentences are grammatical and ungrammatical. There is also syntax in music such that we have a sense of what melodies are acceptable (e.g., melodies ending with do and la) and what melodies are strange (e.g., melodies ending with re and fa). In terms of pronunciation, some non-native speakers may pronounce some phonemes with a slightly different phonetic realization, but native speakers of those languages may still assimilate these deviant sounds into their native phonemic categories, though with some processing cost in comprehension depending on the degree of the foreign accent. For non-proficient musicians, when they hum a melody, their pitch may deviate from the standard pitch and the length of the musical note may not be the standard length (e.g., one beat, two beats) as in the melodies but proficient musicians may still be able to recognize the melodies from these deviant melodies. Apart from that, people can recognize the same melody even though it is hummed by different people or played by different musical instruments. This is similar to the case that people can understand the same sentence spoken by different speakers irrespective of their gender and voice qualities. When a few notes of the melodies are heard (e.g., so la so fa), one or more melodies may be activated in our mind (e.g., London bridge is falling down), which is similar to language processing where some word candidates (e.g., bee, beef, beetle) are activated by a certain phoneme sequences (e.g., /bi/). Therefore, psycholinguistic experiments that have been conducted in languages in the past (e.g., speech perception, accent perception, priming, spoken word recognition, grammatical judgment) can be reexamined using musical materials to investigate whether there are common underlying mechanisms in the processing of language and speech. In addition, eye-tracking studies can be carried out to investigate the similarities and differences between sentence processing and musical score reading.
When the English utterance (i.e., 'Please give me the thin book on the right') is mispronounced by a Cantonese speaker as 'Peace give me the fin book on the white', Cantonese listeners may understand the intended meaning better than native English speaker. In Mandarin, upon hearing a mispronounced nonword 'xing1yin1' (i.e., 'star' + 'sound'), Cantonese listeners may understand the intended meaning (i.e., 'sound' 聲音 sheng1yin1) better than Mandarin listeners. These are cases where non-native listeners may have an advantage over native listeners in understanding accented speech (i.e., Interlanguage Speech Intelligibility Benefit for the Listener (ISIB-Listener), Hayes-Harb, Smith, Bent, & Bradlow, 2008). Beginning Cantonese learners of Mandarin may even understand the intended meaning of a word (e.g., 'search' 搜查) better when its tone is mispronounced (i.e., sou3cha2) than when it is correctly pronounced (e.g., sou1cha2), indicating that non-native listeners find accented speech more intelligible than native speech (i.e., Interlanguage Speech Intelligibility Benefit for the Speaker (ISIB-Speaker), Hayes-Harb et al., 2008). My research aims to find out the underlying mechanisms responsible for these effects.
2. Second Language (L2) Word Production and Recognition Model
There are many cognates in Cantonese and Mandarin because most of the time the same Chinese character has a similar meaning and pronunciation in the two languages. Moreover, there are sublexical (i.e., onsets, rimes and tones) pronunciation correspondences between Cantonese and Mandarin (Zhang & Gao, 2000). For example, most of the words that are pronounced tone 2 in Cantonese are pronounced tone 3 in Mandarin. In my study, it is shown that Cantonese speakers’ phonological knowledge of Mandarin words is influenced by these regularities in a series of Mandarin pinyin transcription tasks and a character-sound matching task. For example, Cantonese speakers may mistakenly think that the word 搜 'search' (pronounced sau2 in Cantonese and sou1 in Mandarin) is pronounced with tone 3 instead of tone 1 in Mandarin (referred to as a regularity effect). However, there are some irregular onset words (e.g., 靠 'rely', pronounced kaau3 in Cantonese and kao4 in Mandarin) where Cantonese speakers seldom give a mispronunciation using the dominant correspondence (e.g., Cantonese onset /k/ to Mandarin onset /j/) due to the fact that they are pronounced the same in Cantonese and Mandarin (i.e., a congruency effect). A new three-route L2 Mandarin word production and recognition model has been proposed to account for these findings. Cantonese speakers gradually shift from a sublexical to lexical/concept route in their L2 Mandarin word production with increasing L2 phonological proficiency while both the concept/lexical and sublexical routes are used by beginning and proficient learners in their L2 Mandarin word recognition.
My ongoing and prospective research related to this model focuses on the following topics. Please feel free to contact me if you are interested in carrying out some collaborative research.
A. Extending the L2 word production model to Japanese
In order to strengthen the explanatory power of the model, it must be able to apply to other languages as well. Japanese serves as a good testing case. In Japanese, there are many onyomi kanji (音讀漢字) where the pronunciations of those characters have origins in ancient Chinese. Arising from this, Japanese moras sometimes have a systematic pronunciation correspondence with Cantonese codas (Lee, 1992). For example, the Cantonese codas /t/ and /k/ correspond to the Japanese moras /tsu/ and /ku/ respectively (e.g., 突 'sudden', pronounced dat6 in Cantonese and to.tsu in Japanese) while the Cantonese codas /m/ and /n/ both correspond to the Japanese mora /n/ (e.g., 減 'minus', pronounced gaam2 in Cantonese and ge.n in Japanese). Beginning Cantonese learners of Japanese are sensitive to these pronunciation correspondences as indicated by the above-chance accuracy rate in a forced choice pronunciation matching task. This demonstrates that the use of the sublexical route in L2 word production can also be applied to Japanese.
B. Testing the L2 Mandarin word recognition model using online and electrophysiological methods
The current L2 Mandarin word production and recognition model has been mainly developed on the basis of off-line tasks (i.e., word transcription, Mandarin pinyin transcription, and character-sound matching). The involvement of the use of the sublexical route in L2 Mandarin word recognition can be further examined using on-line methods such as lexical decision within the cross-modal priming paradigm (e.g., Broersma & Cutler, 2008, 2011), eye-tracking within the visual-world paradigm (e.g., Huettig & McQueen, 2007), and electrophysiological methods using event-related potential (ERP) techniques to investigate the time-course of the activation of different word candidates.
C. Computational modeling of the L2 Mandarin word production model
The proposed L2 Mandarin word production model has the potential to be implemented computationally using a large corpus of monosyllabic and multisyllabic Mandarin production data collected from many Cantonese speakers with different levels of L2 Mandarin phonological proficiency. The computational model can then be used to simulate the behavioral findings reported in our studies. It can also be used to examine the relative contribution of regularity and congruency to L2 Mandarin word production.
3. Accent Perception and Adaptation
A. Perceptual adaptation to accented speech
Previous studies have shown that native listeners can adapt to phonetically-deviant speech (e.g., words pronounced with a slightly deviant /s/) after only a brief exposure (e.g., Norris, McQueen, & Cutler, 2003). However, it is not clear whether such adaptation can be applied to phonological-deviant speech when a phoneme is mispronounced as another phoneme by non-native speakers (e.g., /th/ mispronounced as /f/). The phonological-deviant speech can produce real words (e.g., 'thin' mispronounced as 'fin') or nonwords (e.g., 'faith' mispronounced as /feif/). By exposing native listeners to phonological-deviant speech which are nonwords, it is expected that native listeners would expand their phonemic category (e.g., /th/) to accommodate the deviant phonological form (e.g., /f/) to facilitate their understanding of accented speech. However, our study shows that there was no increase in recognition accuracy in understanding phonological-deviant speech where a word is mispronounced as another word (e.g., 'thin' mispronounced as 'fin') after exposure to mispronounced nonwords (e.g., /feif/). My future investigations will examine whether this is due to insufficient exposure to the phonological-deviant mispronounced nonwords or the lack of feedback during the exposure phase, or simply due to the limit of accent adaptation for this kind of phonological-deviant accented speech.
B. Word recovery mechanisms from mispronounced words
When a mispronunciation is heard in English and the mispronounced word becomes a nonword, listeners tend to change the vowel instead of the consonant in recovering the intended meaning (van Ooijen, 1996). For example, upon hearing the mispronunciation 'kebra', English listeners tend to interpret it as 'kobra' rather than 'zebra'. Even though similar results were found in other Indo-European languages such as Dutch and Spanish (Cutler, Sebastian-Galles, Soler-Vilageliu, & van Ooijen, 2000), it still remains unclear whether a preference for vowels over consonants is a universal phenomenon. Moreover, there are languages like Mandarin which differ from Indo-European languages in several aspects. First, rimes are more prominent units than vowels in Mandarin, relative to Indo-European languages, because there are only two possible codas (i.e., /n/ and /ng/) and there is a simpler syllabic structure. Second, Mandarin has a tone component and, third, more syllables in Mandarin have their own meanings than happens in Indo-European languages. Upon hearing a mispronounced Mandarin disyllabic word that is a nonword (e.g., the potential Cantonese pronunciation of the word 'search' 搜查 sou1cha2 as the nonword sou3cha2), will native Mandarin listeners prefer to change the vowel (or rime) rather than the consonant in attempting to recover the meaning of the word? Or will they prefer to change the tone rather than the consonant or vowel/rime? Since the first syllable in the mispronounced word will activate some word candidates already, will they prefer to change the sublexical units in the second syllable rather than the first syllable? How does word frequency and the number of words sharing the syllable interact with the word recovery mechanisms? The answers to these questions may impose further constraints on current word recognition models.
C. Individual differences in accent adaptation ability
Almost every person who learns to speak a second language after the age of four or five will have a certain degree of accent. Sometimes accents may cause a problem for speakers of different language backgrounds who want to use English to communicate with each other. Although the ultimate goal of language learning is to ask non-native speakers to improve their pronunciation accuracy, it cannot be achieved in a short period of time and the improvement may vary from one person to another. In order to facilitate the communication between non-native and native speakers, we may approach this issue from the perspective of listeners. It is well known that some people are good at comprehending accented utterances while others are not. Having knowledge about the native language of the non-native speaker may enhance the perception of their accented speech. However, it is impossible for people to master the native languages of all non-native speakers. Instead, I would like to investigate individual differences in the ability to understand accented speech and whether accent adaption is related to cognitive factors involved in problem solving skills (e.g., IQ, attention span, short term memory, creativity, phonological awareness). Training could be provided in those areas which may indirectly improve skills on the perception of accented speech.
4. Using L2 knowledge to investigate the mental representation of L1 speech sounds
It is well documented that the mental representation of L2 speech sounds is influenced by the L1 phonological system (e.g., Flege, 1995). For example, beginning Cantonese learners of English have difficulty differentiating the English phoneme /th/ from /f/ due to the lack of a separate phonemic representation /th/ in their L1. Based on the assumption of language transfer, the speakers' L2 knowledge can be used to infer the mental representation of L1 speech sounds when there is no consensus on their categorization. For example, there are disputes over the categorization of Cantonese as a six- or nine tone system. As mentioned in our L2 word production model, Cantonese speakers are more accurate in producing a correct Mandarin pronunciation when the word follows the dominant Cantonese/Mandarin correspondence than when it does not. However, the pronunciation correspondence between Cantonese and Mandarin tones is different under the six- and nine-tone system. I have used a Mandarin pinyin transcription task with ANCOVA and regression analysis, to show that the correspondences using the six-tone system account better for the accuracy data than those of the nine-tone system, thus providing empirical support for the former .
5. Using Vietnamese to investigate the commonalities and differences between various word production and lexical processing models
The different processing mechanisms between languages (e.g., English, Chinese, Japanese, Korean) are thought to be partly influenced by the different writing systems in additional to their differences in phonological and morphological structures. While English is a phonemic language (written in Roman alphabetic letters) with a stressed based system and complex syllable structure (with consonant clusters) that is rich in inflectional morphology, Chinese (written in logographic Chinese characters) is a morphosyllabic language with a tonal system and simple syllable structure (without consonant clusters) that is rich in compounding morphology. In Chinese, a character represents a syllable and the syllable usually represents a morpheme. Korean is an alphabetic language written in hangul (The Korean alphabet) in a square-shaped format. Korean is not a tonal language and more than 70% of the words have Chinese origins which are also morphosyllabic in nature (e.g., 독립 獨立 ‘independence’). Japanese, on the other hands, is a mora-based language written in kana and kanji with a pitch-accent system. Due to these inherent differences, different phonological planning units have been proposed for different languages (i.e., phonemes for English, syllables for Chinese and moras for Japanese) and different morphological processing models have been proposed for English and Chinese.
However, as Chinese and English differs in various aspects (e.g., alphabetic letters vs. Chinese characters, stress vs. tones, complex vs. simple syllable structure, rich in inflectional vs. compounding morphonology), it is difficult to conclude whether the difference in processing mechanisms are caused by one or more of these differences. In addition, most of the new models proposed in Chinese has not be examined in other languages. Vietnamese serves as a perfect target language to reexamine these issues. Vietnamese is a phonemic language (written in alphabetic letters) with a tonal system (with tone marks shown on the alphabets) and simple syllable structure (without consonant clusters) that is rich in compounding morphology. In addition, more than 70% of the Vietnamese words have Chinese origins which are also morphosyllabic in nature (e.g., độc lập 獨立 ‘independence’). Therefore, Chinese-Vietnamese and Vietnamese-Korean are almost minimal pairs in terms of the language properties, where the former only differs in the scripts used (logographic Chinese characters vs. Roman alphabets) while the latter only differs in the use of tonal vs. tonal system. Therefore, Vietnamese can be used as the target language to reexamine the following issues to investigate the universality of the models proposed in Chinese:
- Phonological units in spoken word production (e.g., Chen, Chen, & Dell., 2002; O’Seaghdha, Chen, & Chen, 2010; Wong & Chen, 2008, 2009; Wong, Huang, & Chen, 2012)
- Morpheme ambiguity resolution (e.g., Tsang & Chen, 2010, 2013, 2014; Tsang, Wong, Huang, & Chen, 2014)
- Morphological processing models (e.g., Chen & Chen 2006; Zhou & Marslen-Wilson, 1994, 1995, 2009; Zhou, Marslen-Wilson, Taft, & Shu, 1999)
- The role of tone in language processing (e.g., Cutler & Chen, 1997; Lee, 2007; Sereno & Lee, 2015, Taft & Chen, 1992; Malins & Joanisse, 2010; Ye & Connine, 1999)
- Time course of monosyllabic spoken word recognition (e.g., Zhao, Guo, Zhou, & Shu, 2011)
Psycholinguistic issues that has been explored in English can also be reexamined in Vietnamese to investigate their universality across languages:
- Word recovery mechanisms from mispronounced words (e.g., Cutler, Sebastian-Galles, Soler-Vilageliu, & van Ooijen, 2000, Van Ooijen, 1996)
- Neighborhood density effect (e.g., Luce and Pisoni, 1998; Vitevitch 2002)
In addition to behavioral methods, eye-tracking and electrophysiological methods can also be employed to examine these issues.
6. The cognitive processing of linguistic and musical structures: A comparative approach
Music shares many similarities with languages. In language, there is a limited number of phonemes, but the phonemes can be combined to form thousands of words and infinite number of sentences. In music, there is a limited number of music notes (e.g., do re me fa so la ti) but they can be combined to for an infinite number of melodies. Western music tends to use all the seven notes (i.e., heptatonic scale) in their melodies while most Chinese music tends to use only five notes (i.e., do re me so la) per scale (i.e., pentatonic scale) in their melodies. This is similar to the case in language where different languages use different combinations of phonemes in forming words. There are syntactic rules in languages, so we have a sense of which sentences are grammatical and ungrammatical. There is also syntax in music such that we have a sense of what melodies are acceptable (e.g., melodies ending with do and la) and what melodies are strange (e.g., melodies ending with re and fa). In terms of pronunciation, some non-native speakers may pronounce some phonemes with a slightly different phonetic realization, but native speakers of those languages may still assimilate these deviant sounds into their native phonemic categories, though with some processing cost in comprehension depending on the degree of the foreign accent. For non-proficient musicians, when they hum a melody, their pitch may deviate from the standard pitch and the length of the musical note may not be the standard length (e.g., one beat, two beats) as in the melodies but proficient musicians may still be able to recognize the melodies from these deviant melodies. Apart from that, people can recognize the same melody even though it is hummed by different people or played by different musical instruments. This is similar to the case that people can understand the same sentence spoken by different speakers irrespective of their gender and voice qualities. When a few notes of the melodies are heard (e.g., so la so fa), one or more melodies may be activated in our mind (e.g., London bridge is falling down), which is similar to language processing where some word candidates (e.g., bee, beef, beetle) are activated by a certain phoneme sequences (e.g., /bi/). Therefore, psycholinguistic experiments that have been conducted in languages in the past (e.g., speech perception, accent perception, priming, spoken word recognition, grammatical judgment) can be reexamined using musical materials to investigate whether there are common underlying mechanisms in the processing of language and speech. In addition, eye-tracking studies can be carried out to investigate the similarities and differences between sentence processing and musical score reading.