Chapter 2 Literature review
2.4 Syllable deletion in Mandarin, Cantonese and Taiwan Southern Min
國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
On the whole, most of the cross-linguistic studies showed a strong relationship between stress and syllable in children’s phonological acquisition. However, in Mandarin, the salient status of stress is lower. Duanmu (2000) suggested that it is difficult to detect stress in Mandarin since the most important phonetic cue for stress is F0, but F0 is already for tone use in order to differentiate word meanings, and thus cannot be used freely for stress. Moreover, Zhu and Dodd (2000) showed that weak stressed syllables have a very short duration and a much reduced pitch range. The pitch of a weak syllable is primarily determined by the preceding tone. Compared with the saliency of tones, consonants and vowels, the saliency of weak stress is far less important because of the low value in differentiating word meaning (Zhu & Dodd, 2000). Consequently, because this study attempts to investigate the syllable acquisition in Mandarin and types of syllable deletion, the following section will be laid out the relative phonological elements in Mandarin.
2.4 Syllable deletion in Mandarin, Cantonese and Taiwan Southern Min
As previously stated, studies have indicated that young children’s types of syllable omission will be easily influenced by the factor of stress because stress is salient in stress languages. However, children’s syllable deletion in tonal languages may not be related to stress. According to Zhu and Dodd (2000), they worked on the phonological acquisition of Beijing Mandarin, and they found that children acquiring
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Beijing Mandarin acquired syllable elements in the following order: tone was acquired first; and followed by syllable-final consonants and vowels; syllable-initial consonants were mastered last. Results showed that tone errors were rare, even in the youngest group of children. Tone has the highest saliency in Mandarin: it is compulsory for every syllable structure; change of tone will change the lexical meaning (Zhu & Dodd, 2000; Yip, 2002). It is suggested that the saliency of the components in a language system determines the order of acquisition. Consequently, based on the comparison of phonological processes used by children acquiring different languages, phonological patterns showed both universal tendencies and language-specific constraints on acquisition.
In terms of Cantonese acquisition, Tse (1991) investigated the process of the acquisition of the phonological system based on a longitudinal study of one Cantonese-speaking child. According to the analysis of child’s speech data, the most used structure in the order of frequency was CV, and was then followed by CVG and CVC. However, CVG syllable remained a relatively unstable structure with omission of the final glide. Furthermore, while analyzing the structural simplification, Tse (1991) observed that the child was likely to delete the consonant cluster [kw] to be either [k] or [w]. In addition to initial consonant, final consonant deletion was seen as a fairly active process, with 17.6% of the final consonant deletion. The most
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
frequently affected type was postnuclear glides with 24.6%, followed by nasal finals with 10.6% and stop finals with 1.4%. Hence, Tse (1991) found the evidence for universal trends in the sequence of phonological acquisition, and concluded that these phonological patterns, simplification strategies and salient phonological distributions were accounted for Cantonese children. In order to further attest phonological patterns, So and Dodd (1995) enrolled 268 Cantonese-speaking children to analyze their speech errors in a picture-naming test. Based on the comparison of phonological structures of different languages, the younger Cantonese-speaking children made more errors in the syllable-initial position than in the syllable-final position. Results revealed that only a relatively small proportion (10.7%) of the total number errors were made on final syllable positions, including fronting, backing and final consonant deletion. However, for syllable-initial consonants, So and Dodd (1995) indicated that the acquisition patterns of children acquiring Cantonese exhibited a interaction of linguistic universal tendency with language-specific effects.
On the other hand, Tsay (2006) found that the patterns of syllable deletion occurred differently in terms of the word-position and the number of the syllables.
Tsay (2006) explored the relationship between prosodic model and the syllable omission patterns produced by children acquiring Taiwan Southern Min, aged from 1;6 to 3;0. The children used syllable omission as simplification strategies to shorten
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
long utterances. There were two patterns of the children’s syllable deletion: (1) word-initial syllables were omitted more frequently than word-final syllables and (2) omission occurred in multisyllabic words more frequently than in monosyllabic word.
Thus, on the basis of different factors and elements in syllable omissions, we may deduce that stress is irrelevant with syllable deletion in tonal language, Mandarin.
To sum up, previous studies have focused on a shared developmental order of syllables and examined the universal syllable types at children’s early phonological stage in terms of different theoretical approaches. Most of cross-language studies have highlighted the strong relationship between stress and syllable deletion; however, stress is not the most salient in Mandarin. Based on the observation of syllable patterns in Mandarin, the patterns may reveal both universal tendencies and language-specific constraints on syllable acquisition and syllable deletion.
Consequently, this thesis aims at types and patterns of syllable acquisition and syllable deletion of the four young children acquiring Taiwan Mandarin in a longitudinal observation method.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
38
Chapter 3
Methodology
There are two parts in this section. The first part consists of the data collection, and the second part involves the data analysis. The data have been collected by the researcher and the research team. There are in total 84 hours; 39 hours have been collected and transcribed by the researcher and 45 hours have been collected and transcribed by the research team in Phonetics and Psycholinguistics Lab at National Chengchi University. The study has been sponsored by MOST research projects,
“Consonant acquisition in Taiwan Mandarin (MOST 100-2410-H-004-187-)”,
“Consonant acquisition in Taiwan Mandarin: Evidence from longitudinal and experimental studies (MOST 101-2410-H-004-182-)” and “Consonant acquisition in Taiwan Mandarin: Evidence from observational and experimental studies (MOST 102-2410-H-004-107-).”
Section 3.1 presents the way in which the participants have been recruited and their background information. The procedure and the recording equipment used during the data collection will also be introduced in this section. For data analysis in section 3.2, I will present the methods of data transcription, the criteria for choosing target words, the formulas used in obtaining syllable type frequency and how the syllable deletion patterns have been organized.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
3.1 Data collection
This section presents how the researcher and the research team collected the speech data from the children and their families’ background information. The families participating in the research projects were recruited from an advertisement posted on a non-profit parent forum – Babyhome (http://www.babyhome.com.tw/).
On the forum, the research team posted the recruiting page, which listed the background of the research information and the age of recruiting children. Parents who would like to join in the research were supposed to fill a registration form out.
Sixteen children were enrolled under the projects.
According to children’s speech data and family backgrounds, some of the children in the MOST research projects lived with their grandparents who spoke Taiwanese Southern Min, so the children would produce Taiwanese Southern Min words during the observation. Besides, because some parents used to speak English to the children, the children might also produce English in the data. In order to rule out the influence of the languages (i.e. Taiwan Southern Min and English), this study excludes these children and focuses on the children acquiring Taiwan Mandarin. This study therefore collected four children and their caretakers. There were in a total number of 16515 tokens produced by four children and 4586 tokens produced by four caretakers, among which 8132 tokens were transcribed by the assistants in the
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
research team and 12692 tokens were transcribed by the researcher.
3.1.1 Participants
The background of the four children, two girls and two boys, were selected and enrolled in this study. They were from middle class family in Taipei, Taiwan. The four children lived with their parents and their mothers took care of them in the day time.
They were the only child in their families during the observation. Their mothers used Taiwan Mandarin to communicate with them, so the children’s first languages were Taiwan Mandarin. The participants’ background information is presented in the following Table 3.1 and Table 3.2.
Table 3.1 Participants’ age range and recording duration
Participant Gender Age Range Recording Duration
Child #1 Female 1;0 – 2;5 18 months
Child #2 Female 0;10 – 2;3 18 months
Child #3 Male 1;1 – 2;6 18 months
Child #4 Male 0;11 – 2;4 18 months
The children’s ages were 0;10 to 1;1 at the beginning of the observation. The data in the boys’ and girls’ babbling stages were excluded in this study. The age when they were in non-reduplicated babbling stages and they had produced first meaningful words was selected. At the end of the observation, the their ages were 2;3 to 2;6.
Besides, they were reported to have no speech, hearing or visual impairments.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Table 3.2 Participants’ gender and relation with children
Participant Gender Relation Language
Caretaker #1 Female Mother of child #1 Taiwan Mandarin Caretaker #2 Female Mother of child #2 Taiwan Mandarin Caretaker #3 Female Mother of child #3 Taiwan Mandarin Caretaker #4 Female Mother of child #4 Taiwan Mandarin Four adult participants were four children’s caretakers respectively. They were mothers and their native languages were Taiwan Mandarin. Each mother was the primary caretaker of the child during the whole longitudinal observation.
3.1.2 Procedures
The data collection of four children started from January 2012 to March 2014.
There were eight research assistants in the research team. Two assistants were sent to
a child’s house to record the spontaneous speech of the natural daily activities
between the child and the mother every other week. On average, a 60-minute
recording was made from every observation. Sometimes the recording time might be
shorter if the children were tired or cried and it might be longer if the children were
reading or playing with their mothers. In order to create a natural context, the
activities during recording were various, including share-book reading, object-naming,
or playing with toys, and the mothers were encouraged to play with their children. For
most of the time, the children played with the mothers and the mothers were the
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
people the children were more familiar with. The choice of home environments as
observation settings was based on the studies of Lewedag, Oller and Lynch (1994)
and Chen and Kent (2005). Children were reported to vocalize more frequently and produced more well-formed syllables in home environments than in laboratory settings (Lewedag et al., 1994).
In order to collect the more natural speech data from children’s vocalization, no
systematic planning of elicitation was done during the recording, except for natural
elicitations in daily life. For instance, when the children were playing toys with their
mothers, sometimes the mothers would ask them to name the toys. Therefore, the
speech data of this study were ensure to be elicited from spontaneous conversations,
rather than elicited and planned sounds. The target words for analysis were selected in
these spontaneous speeches. Although the number and the type of children’s phonetic
errors might be greater than controlled experimental tasks, the data in a more natural
context reflected the processes of phonological acquisition. Thus, the researcher
selected these spontaneous speech data as analysis in this study.
During the observation, one of the assistants held the video recorder and the
other assistant held the sound recorder. The one who held the video recorder would
make sure to film the child’s face, mouth and the objects they played with. The
assistant who held the sound recorder had to get closer to the child in order to achieve
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
good quality of speech sounds. At the end of every recording, the assistants would
note down every recording date and time, making it easier to file the recording data.
Before the research project, the participants have had signed the human subject
consent forms. At the end of the research project, the families would receive an album
of video recordings as a souvenir. Whole funds were supported by the MOST research
projects (MOST 100-2410-H-004-187-, MOST 101-2410-H-004-182-, and MOST 102-2410-H-004-107-).
3.1.3 Recording equipment
To achieve flexible recording, without distracting subjects, the wireless equipment of video-recording and sound-recording were used. Sony HDR-CX380 Handycam digital video camera recorder, Sony ICD-UX543F digital voice recorder and Shure SV100 multi-purpose microphone were all used during the recording. The microphone was used to connect the voice recorder. During the observation, the video-recorder and sound-recorder were powered with batteries instead of using cables, in order to provide more flexibility in choosing observation locations in the home environments. The size of the equipment was small, so assistants could carry the equipment easily while interacting with children.
In addition, the video helped us record children’s gestures, lip movements and
items they played with. It also provided us some clues to decode the referential
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
meaning of children’s utterance. The sound-recorder offered us high quality sound
files. The multi-purpose microphone with the unidirectional pattern helped reduce
ambient noise and provided clear sound quality for spoken words. All recording equipment was used for providing high recording quality for perceptual transcription.
3.2 Data Analysis
The four children in the observation period were from one-word stage and two-word stage to the stage when the utterances were longer than two syllable words.
Moreover, it is noted that Mandarin has a “dual-vocabulary” pattern, where most of the monosyllabic words have a disyllabic form (Duanmu, 2007). Children’s syllable production may vary from monosyllables to longer utterances. Accordingly, this study will carefully analyze children’s monosyllabic words, disyllabic words and multisyllabic words separately in different sections. The following sections consist of how the data have been transcribed, coded and analyzed.
3.2.1 Transcription and coding
The data from the recording files were transcribed by the researcher and the assistants of the research team. If there were disagreements, the tokens would be discussed and checked by the third assistant of research team. All the assistants were native speaker of Taiwan Mandarin and had good training of phonetic transcription of speech data.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
In order to determine the participants’ production in transcription and assess the coding reliability, intra-reliability and inter-reliability were concerned (Vihman &
McCune, 1994; Chen & Kent, 2005). To estimate intra-transcriber reliability (93.1%
for consonants and 97.7% for vowels), the researcher checked the consistency of the transcriptions by re-transcribing a random one-hour sample of the children’s recordings. In addition, to check inter-transcriber reliability (92.3% for consonants and 95.6% for vowels), two other experienced transcribers of the research team were asked to transcribe the same recording data. The intra-transcriber and inter-transcriber reliability of the transcription reached a percentage higher than 90% under the study.
With regard to phonetic transcription, this study focuses on syllable types and do not deal with specific segment acquisition. Thus, broad transcription was used by the IPA system.
The child’s utterances of words and phrases were transcribed into five parts:
actual produced words in IPA transcription, syllable type, tone, number of occurrences and possible meaning. The transcribed examples are shown below in Table 3.3.
‧
Table 3.3 The sample of coding IPA
Transcription
Syllable Type
Tone Occurrence Possible Meaning
Table 3.3 represents the way how the researcher coded the child’s utterances. The first column showed the child’s actual produced words transcribed with IPA symbols. The second column was to show the syllable type of each meaningful word. There were twelve syllable types, coded as V, CV, VG, GV, VN, CVG, CVN, CGV, GVG, GVN, CGVG, and CGVN. The third column was to mark the tone of words. The tones were coded as [55], [35], [21] and [51] in Mandarin, representing level tone, rising tone, falling rising tone, and falling tone. The neutral tone was coded without any tone number in the transcription. The forth column was the number of occurrence of each word. For example, in the row of the first example, the child produced “flower” as
[wa] for 3 times. Finally, the fifth column marked the possible meaning of each word,
which could be inferred by contexts, gestures or adult’s repetition. The meaningless tokens were excluded in the study.
3.2.2 Criteria for target word
Since this study attempts at investigating the children’s syllable patterns and
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
types of syllable deletion, the criteria for target word and syllables should be concerned. The following were the criteria for choosing the appropriate target words of the children’s production. Based on the previous analyses (e.g., Vihman & McCune, 1994; Stoel-Gammon, 2011; Sosa and Stoel-Gammon, 2012; H. Y. Wang, 2014), there were three major criteria adopting in this study. Firstly, the participants’ sound
quality must be clear. The mumbling speech or overlapping speech of adults would be
excluded. Moreover, since background noise from playing with toys and rustling
noise from contacting with recorders would result in fuzzy sounds, these noises would
be excluded. Secondly, the children’s meaning of each word must be clear and plain.
To be judged a meaningful word, the children’s speech forms should be
systematically related with the context. Furthermore, words would be chosen when a
Mandarin gloss could be identified, and the meaning of words could be inferred by
examination from contexts or repetition of adults’ speech. For instance, if the child
pointed at an animal toy and uttered [ja21 ja35], we suggested that the intended
meaning was “a goat.”
Thirdly, the deleted forms that were selected in the study could not be
completely different from target words. Because this study focuses on the syllable
deletion, the syllable of words that reduce from complex syllable types to simpler
syllable types, such as from GVN-GVN in [jɑŋ21 jɑŋ35] goat to GV-GV [ja21 ja35],
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
would be included. Nevertheless, if the child produced CV-CGVN in [mi51 ljɑŋ51]
moon, CV [mi51] would not be calculated because [mi51] was totally different from
GV [ɥɛ51]. Moreover, the target lexicons should be nouns, verbs, adjectives and adverbs in children’s production. If they produced a whole sentence, the lexicon
within the production would be excluded in order to avoid syntactic problems.
Fourthly, the imitated production would be included in examining child
acquisition (e.g., Ferguson & Farwell, 1975; Vihman & McCune, 1994; Sosa &
Stoel-Gammon, 2012; Macrae, 2013). Ferguson and Farwell (1975) proposed that a large number of young children’ productions were imitated words and children could imitate words spoken by adults with “a considerable separation in time.” In this way, it is difficult to exclude children’s imitation production from the analyses. Vihman and McCune (1994) pointed out that the assessment of the reliability of children’s imitation word could use the three criteria, including imitated tokens, phonetic identity and no inappropriate uses. For the analysis of imitated tokens, if children produced the imitation with apparent understanding and clear meaning, then it should be credited. For the analysis of phonetic identity, all the instances of the words should exhibit the same phonological shape, conforming to phonetic match. For the appropriate word-use, it means that if children’s word forms were not used in conflicting contexts and could infer the same word, the forms would be credited.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Furthermore, Sosa and Stoel-Gammon (2012) noted that the child’s production
Furthermore, Sosa and Stoel-Gammon (2012) noted that the child’s production