Chapter 1 Introduction
1.5 The framework of the thesis
國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
6
What is the most frequent tone produced by subjects? What is the ordering of frequency in the four tones, including the neutral tone? What is the ordering of accuracy rate in all the Mandarin tones? Do easier produced tones present higher accuracy rate? Do higher frequency tones also have higher accuracy rate? Do tones that appeared earlier have a higher frequency and accuracy rate in the occurrences? Would there be universal ordering of stabilization among different Asian tone acquisition studies?
(3) Regarding the issue of substitution pattern in tonal errors:
What kind of strategy would children use to replace the un-mastered tones? Would they replace the un-mastered tone with more frequent tones or more accurate tones? Which tone is more unstable and is replaced by other tones more frequently? Which tone is more likely to replace the unstable tones? Are there obvious individual differences among children's substitution strategies?
1.5 The framework of the thesis
In chapter 1, I provided a brief introduction to indicate what kind of issues are going to be discussed in the whole thesis, and what is the motivation for dealing with these issues; the four main research questions were also addressed. In Chapter 2, firstly I will mention the language universals in first-language acquisition in 2.1, and the Mandarin tonal system representations will be introduced in 2.2. Section 2.3 and 2.4 include studies specifically on tone acquisition. The comparison among tone acquisition studies of Asian
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
7
languages is in 2.3; the foundation of tone acquisition studies in Mandarin is in 2.4. The last section 2.5 will introduce the tone acquisition theories. Chapter 3 contains the methodology which can be viewed in two separate parts. Section 3.1 is the data collection method describing how I obtained the speech tokens, and section 3.2 is the data analysis method explaining how the data were arranged. Chapter 4 will display the results and analysis in tables and graphs. Section 4.1 shows the overall results in all data. Section 4.2 applies different measure to analyze the monosyllabic and disyllabic tokens separately. In section 4.3, the substitution patterns in tonal errors are offered, and in section 4.4, the ages of emergence and stabilization in all subjects are summarized. The discussion and explanation are provided in chapter 5. Section 5.1 summarizes the findings in chapter 4.
Section 5.2 and 5.3 are the comparison among different studies in Mandarin and different cross-linguistic studies. Section 5.4 is the concluding remarks and the suggestion for further research.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
8
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
9
Chapter 2
Literature review
This chapter consists of three sections. Firstly, I will review studies with universals in children’s phonological acquisition in section 2.1. Secondly, the various representations of Mandarin tonal system will be introduced in 2.2. Section 2.3 consists of a theory of tonal markedness. Later on, the important studies related to children’s tone acquisition in Asian tone languages will be presented in 2.4. Lastly, the most related references focusing on Mandarin tonal acquisition will be shown in 2.5.
2.1 Language universals in first language acquisition
Many researchers proposed that there were chronological stages in children’s early vocalization, which means that children across the world acquired languages by following similar steps. (Lenneberg 1967; Kaplan & Kaplan, 1971). The four stages of children’s early vocalization divided by Kaplan & Kaplan (1971) are presented below:
Stage 1: Crying (0;0-0;6)-The cries are identified by parents as angry or physical pain.
Stage 2: Pseudocry and Noncry vocalization (0;4-0;5)- Begin to use the articulatory organs, and the utterances become different from crying.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
10
Stage 3: Babbling and Intonated vocalization (0;5-0;8)- The vocalization becomes speech-like. Vowel-like and consonant-like sounds are combined into reduplicated syllable structure. Infants begin to imitate adults’ intonation pattern.
Stage 4: Patterned Speech (0;9-0;12)- A large number of sounds appear, and child’s first word begins to appear.
Although there were some overlapping among each stages and not all researchers agreed the division of the stages, the chronology of these development still gained consensus.
The reason why the chronology is proved universally may be attributed to the biological maturation of brain and motor control. Neurologists and psychologists proposed that babbling occurs automatically when the relevant structures in the brain have reached a critical level of maturation (Preyer, 1882). Before the maturation of the brain and motor control, Oller (1974) found that children would simplify the speech sound they perceived and construct a simplified sound system to ease the burden on their short-term memories and motor coordination.
2.1.1 Syllables
It was observed that children’s early syllable structures in the babbling stage were simplified and would go through a regular progression: V, CV, and CVCV (Kaplan &
Kaplan, 1971; Kies 1995).
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
11
1. V (The vocalic sounds are like aaa, eee, and uuu.)
2. CV (Children combine consonants and vowels into open syllables, such as ta and ma.)
3. CVCV (There are reduplications repeating frequently. The consonants and vowels are in ABAB form, such as papa, nene, and tata.)
The syllable structures began from simply vowels without consonants. Vowels are easier to produce than consonants for there is no need to exercise the articulators to impede the air flow in the vocal tract. To pronounce a vowel, the only articulator involved is the tongue, but consonants with different manners and places need more articulators working together such as lips, jaw, and velum. The progress from V to CV surely implied the consistent growth of children’s muscular control.
Lightfoot (1982) also believed that language development were biologically determined, and children all over the world acquired languages with uniformity. There were several examples extracted from Demuth (2010) for how children in different languages used similar rules to truncate adults’ target forms.
Table 2.1 The examples of children’s simplified form in English and French Adults’ form Children’s form Meaning Age
English [spәˈk ɛ ti] [ˈk ɛ ti] ‘spaghetti’ 1;2 French [paˈtat] [pәˈtæ:] ‘potato’ 1;4
In Table 2.1, an English-speaking child imitated adult’s [spәˈ kɛ ti] as [ˈ kɛ ti], and a French-speaking child produced [paˈtat] as [pәˈtæ:] . Both children used the same
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
12
phonological rule to simplify adults’ trisyllabic words into disyllables, which were the CVCV structure in the study of Kaplan and Kaplan (1971) mentioned above. Children’s immature muscular control unables them to produce consonant clusters like [sp] or close syllable like [tat]. Thus, children naturally simplified or truncated consonants when producing syllables more complicated than open syllables.
2.1.2 Suprasegmentals
Regarding the stages Kaplan & Kaplan (1971) presented above, they mentioned that in the third stage, infants began to imitate adults’ intonation pattern. This finding suggested that children’s suprasegmental acquisition, such as stress, intonation, pitch, and tone, started very early in the developmental process. Many studies proved that these suprasegmental characteristics were acquired very early and even earlier than segments (Lenneberg, 1967; Kaplan, 1970; Demuth, 1996). Crystal (1970) also cited that children at ages as young as 0;7 to 0;10 could exercise suprasegmental characters much stably and readily than segmental characters.
2.1.2.1 Intonation
For English-speaking children, it was found that infants could distinguish and produce falling and rising intonation contours at 0;8 (Kaplan 1970). When infants were in their late babbling stage, they could well imitate adults’ intonation patterns despite with immature segments (Kies 1995). Kies provided an example to prove that an
‧
English-speaking child could uttered two different intonation patterns at the age of 0;8.
Table 2.2 The two intonation patterns produced by an English-speaking child
(1) (Declaratives)
(2) (Interrogatives)
In these two sentences, although most of the syllables were simplified to V or CV forms, he had already developed the rising and falling in declarative and interrogative intonation patterns which refer to different meanings in English.
2.1.2.2 Stress
What’s more, linguists also noticed that English-speaking children used stress to distinguish meanings as early as they uttered their first words (Kies 1995). Engel (1973) reported that one child used [‘mama] to call his mother but [ma’ma] to his father at 0;10.
The only difference of these two utterances laid on the stress position that ‘mother’ was stressed on the first syllable and ‘father’ was stressed on the second syllable. These examples could clarify that children do acquire prosodic features in their early phonological development.
2.1.3 Reduplication of motherese
Motherese, also named child-directed speech (CDS), refers to the speech form used o i tu
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
14
by adults in talking to young children. Adults would naturally shift the way of normal speech to a high pitched, slow, reduplicated, and simplified way when talking to children (Carroll, 2008). There was a study pointing out that infants paid more attention to repetitive form than regular conversation, and the use of child-directed speech facilitated children to in language learning (Matychuk, 2005). The use of reduplication in motherese facilitated children's language development in several aspects: Phonologically, children could perceive the same syllable twice because the consonants and vowels in reduplications are identical in the first and second syllables. Morphologically, the syllables are simplified to CVCV forms, and the consonants and vowels are replaced to earlier acquired phonemes. Therefore, children better repeated syllables that were reduplicated.
Reduplication was the most common form applied in children’s early production (Grunwell, 1982). As mentioned above in the study of Kaplan and Kaplan (1971), the CVCV structure were repeated frequently by children, and the consonants and vowels were in ABAB form, such as papa, mama, and tata. In the reduplication of motherese in Mandarin, Yang (2012) found that in Taiwan, most care-takers would produce reduplications in a particular tone combination which is a low-falling followed by a rising tone, such as [njow21-njow35] ‘a cow’ and [y21-y35] ‘a fish.’ Children also uttered most reduplications in [21-35] combination. It was obvious that children’s preference of tone
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
15
combination was identical to that of adults’ reduplication in motherese. Whether children’s form was affected by adults’ input needs more specific methods to examine, but the examination would not be conducted in this current study.
2.2 Introduction to Mandarin tonal system
Mandarin tones has been classified and transcribed into several different systems. In this section, I will introduce tonal representations both in Standard Mandarin and Taiwan Mandarin in 2.2.1 and 2.2.2. A summary and the adopted tonal representation will be provided in 2.2.3.
2.2.1 The tonal representation systems of Standard Mandarin
The Standard Mandarin (henceforth SC) is the official language in China. The standard pronunciation of SC is based on Beijing dialect, but it absorbed many other Chinese dialects in different regions (Duanmu 2000). Despite the variance, the tonal representation systems of SC reviewed here are the descriptive view of Mandarin.
2.2.1.1 Chao (1968)
Chao (1968), also known as a musician, established the five-point scale to transcribe the pitch of tones in Standard Mandarin (henceforth SC). He proposed that the tone classification in Mandarin could be realized as pitch differences, and the pitch register was shown in a range numbered 1 to 5. The bigger the number was, the higher the pitch register would be.
‧
Table 2.3 Chao’s (1968) tonal representation system of SC Tone # Chinese Name Tonal feature Pitch value
1 Yinping High Level 55
2 Yangping High Rising 35
3 Shangsheng Low Falling-Rising 214
4 Qusheng High Falling 51
In Table 2.3, the four tones named yinping, yangping, shangsheng, qusheng were described as High Level [55], High Rising [35], Low Falling-Rising [214], and High Falling [51]. The five-point scale which was invented by Chao (1930) was widely used and cited in most research on tone languages for its convenience to depict a tone clearly and specifically. The tonal acquisition studies reviewed in the following sections also adopted Chao’s five-point scale for describing pitch values.
2.2.1.2 Yip (2001)
Yip (2001) provided the link between the phonetic targets and their corresponding representations in Mandarin tonal system. Her tonal representation was shown below.
Table 2.4 Yip’s (2001) tonal representation system of SC Tone # Pitch value Tonal feature
1 55 H
2 35 MH
3 21 L
4 53 HM
In this tonal system, tones were considered to have a head and a contour in the following.
The capital H and L in Tone 1 and Tone 3 represented level tones, and MH and HM in Tone 2 and Tone 4 were contour tones. With regard to the pitch values, the representation of Tone 1 and Tone 2 agreed with those in Chao (1968), but Tone 3 was presented as [21]
‧
without the final rising and Tone 4 was presented as [53] with partial falling from high to mid.
2.2.1.2 Lin (2007)
The system Lin (2007) adopted in her studies concerning SC was presented both with tone features and pitch values. She mentioned that most analyses of SC used only high and low distinction, but it was not specific enough. There were three distinctions of register in Lin’s system, which were high, mid, and low, and they would be presented in capital letters (H, M, and L).
Table 2.5 Lin’s (2007) tonal feature and pitch value Pitch value Tonal feature
4 or 5 H (high)
3 M (mid)
1 or 2 L (low)
In Table 2.5, it defined the corresponding tonal features of the pitch values respectively.
The highest two numbers, 4 and 5, were categorized in H; number 3 which referred to a middle pitch sound fitted in M; the lowest two digits, 1 and 2, had the low pitch feature, so were sorted in L. The tonal features could be used to transcribe into the digital system.
Table 2.6 Lin’s (2007) tonal representation system of SC
Tone # Tonal feature Pitch value
1 HH 55
2 MH 35
3 LH (in phrase final syllable) 214 LL (before another tone) 21 4 HL (in phrase final syllable) 51 HM (before another tone) 53
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
18
The four tones in Table 2.6 were transcribed into pitch register features. For Tone 1, the high level [55] pitch value mapped with HH in this system; Tone 2 which with [35] pitch value was presented by MH. There were respectively two representations in Tone 3 and Tone 4 for a more detail distinction regarding different syllable positions. The basic forms of these two tones were shown in LH for Tone 3 and HL for Tone 4. Yet, it was believed that final part of Tone 3 and Tone 4 would be reduced when followed by another tone.
When these two tones were not in syllable-final position, the rising part of Tone 3 would be missing and presented as LL; the falling part would become [53] and be presented as HM.
The neutral tones which usually appear in grammatical particles or the second syllable of reduplicated kinship terms were precisely depicted in Lin’s (2007) book.
Phonologically, a neutral tone was unstressed and was a low tone underlyingly. However, the pitch values were found to be different when following different stressed tones.
Table 2.7 Pitch values of the neutral tone
Tone # Tone feature Example
T1 + T0 55 + 2 [ma55 ma2] ‘mother’
T2 + T0 35 + 3 [lai35 lə3] ‘came’
T3 + T0 21 + 4 [tɕ je21 tɕ je4] ‘older sister’
T4 + T0 53 + 1 [khan53 lə1] ‘saw’
The Table 2.7 showed the pitch values after each tone. The tone values remained low when following T1 and T4, but when following T2 and T3, the values would rise to [3]
and [4]. The examples in the right column showed where the neutral tone could appear
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
19
and which pitch value would be surfaced respectively.
2.2.2 The tonal representation systems of Taiwan Mandarin (TM)
Many linguists observed that the Mandarin in Taiwan (henceforth TM) was distinct from SC which spoken in Mainland China (Kubler, 1985; Fon, 1997; Fon & Chiang 1999;
Duanmu, 2000; Lin, 2007). The tone change in TM may result from the influence of Taiwanese, the dialect widely spoken in Taiwan. The specific tone values of TM have been reanalyzed acoustically and the studies and their tonal representations would be presented below.
2.2.2.1 Shih (1988)
An acoustic study conducted by Shih (1988) pointed out that the pronunciation of Tone 3 was particularly different from people between Mainland China and Taiwan. In TM, Tone 3 was often produced without the rising part no matter in word-final position or not. The same description was also stated by Kubler (1985) that in Taiwan Mandarin, Tone 3 tended to be pronounced as [21] instead of [214]. The representation Shih adopted was presented below.
Table 2.8 Shih’s (1988) tonal representation system of TM Tone # Tonal feature
1 HH
2 MH
3 LL
4 HL
In Table 2.8, Shih used HH to represent the high-level tone, and MH to represent the
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
20
rising tone. Tone 4 was presented in HL which was considered a complete falling. Due to the dialectical variance mentioned by Kubler (1985) and Shih (1988), Tone 3 which often produced without the rising part was showed to be LL in Shih’s Taiwan Mandarin tonal system.
2.2.2.2 Fon (1997)
Fon (1997) measured the pitch height, duration, and slope steepness of the four tones by the Computerized Speech Lab, KAY Elemetrics. Analyzing the tones produced by the subject who was a 22-year-old female college student in National Taiwan University, Fon (1997) has characterized the tonal representation system for TM in Chao’s five-point scale.
Table 2.9 Fon’s (1997) tonal representation system of TM Tone # Pitch value
1 44
2 323
3 312
4 42
The pitch values of the four tones presented in Table 2.9 were totally different from the system invented by Chao (1968) which has been shown in Table 2.3. Fon (1997) argued that the system she proposed has acoustic fact, and the subject enrolled was born in Taiwan. Thus, the result of the tonal representation would be so distinct from that in Chao’s (1968) study. In fact, the two main distinctions between SM and TM lied in the pitch range and the contour changes of Tone 2 and Tone 3. The pitch range narrowed to a
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
21
four-point scale. The height in Tone 1 lowered from [55] to [44], and the pitch range for Tone 4 also narrowed from [51] to [42]. On the other hand, Fon presented Tone 2 as a dipping tone whose contour was [323], and the contour was similar to the pitch value in Tone 3 presented as [312].
A further perceptual study conducted by Fon and Chiang (2004) illustrated that Tone 2 and Tone 3 had crucial differences between duration, steepness, and height. The duration in Tone 2 was longer than Tone 3, and the slope of Tone 2 was not steeper than Tone 3. The higher pitch of Tone 2 in the ending point was a cue for distinguishing Tone 2 and Tone 3.
2.2.3 Summary
While there were several different tonal representation provided by phonologists and phoneticians in Standard Mandarin and Taiwan Mandarin, how to determine which tonal system is the most suitable one? Duanmu (2000) suggested that it was allowed to slightly modify the transcription of the tonal representation if only if the modified form does not cause meaning contrast in Mandarin. For example, the [21] and [11] in Mandarin do not distinguish meanings, so it does not matter whether we call Tone 3 as a low falling or a low level.
In this thesis, I will mainly adopt the five-point scale invented by Chao (1968) because it is the basic system adopted by most of the phonetic studies. To accommodate
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
22
the change of tone in Taiwan Mandarin which proposed by Shih (1988) and Kubler (1985), Tone 3 would be modified to [21] for the dialect variation in TM.
Table 2.10 The tonal representation adopted in this thesis Tone Pitch value Tonal feature
Tone 1 55 high-level
Tone 2 35 rising
Tone 3 21 low-level
Tone 4 51 falling
Neutral neut neutral
Table 2.10 showed the pitch value and tonal features which will be adopted in this study.
The transcription used for the four tones respectively would be [55], [35], [21], and [51].
The corresponding tonal features of the tones would be high-level, rising, low-level, and falling. The neutral tone would be transcribed in the abbreviation ‘neut.’
2.3 Theory of markedness of tone
Yip (2001) presented a theory of tonal markedness to distinguished marked features and unmarked feature in tone. The tonal markedness theory was derived from three types of data. The first type of data was from Hashimoto’s (1987) survey on tone sandhi in 83 dialects of Chinese. Second, she provided data from Cheng’s (1973) quantitative study of Chinese tones in which the tonal inventories of 73 dialects were studied. The third type of evidence was from the acquisition studies conducted by Clumeck (1980) and Li and Thompson (1978). The evidences could be generalized into three rules, and Yip stated that the markedness rules was used “to minimize articulatory effort.”
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
23
Minimize articulatory effort
a. contour tones are more marked than level tones b. rising tones are more marked than falling tones c. high tones are more marked than low tones
Yip summarized Hashimoto’s (1987) study that contour tones were more likely to be leveled in tone sandhi, and similar results were also found in children’s acquisition data (Li & Thompson,1978; Clumeck, 1980). In numerous tonal inventories which presented by Cheng (1973), it was more likely to have falling tones in the tonal system than rising tones. The third constraint focused on comparing level tones. She found that high level tones were more marked because it needed more strength and could not “minimize articulatory effort.” From this point of view and the evidences from numerous dialects, she concluded that low level tones were more unmarked. The three tonal markedness rules could be examined by the acquisition data from cross-linguistic studies to determine the universal features of tones.
2.4 Tone acquisition studies in East Asia
In this section, I will review several tone acquisition studies cross-linguistically, including Thai, Cantonese, and Taiwanese. Linguists are interested in whether there are language universals in first language acquisition. The tone acquisition studies reviewed here will be compared to the results of the current study in the discussion section. In this
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
24
section, a Thai study focusing on phonetic and phonological development in early speech will be presented in 2.4.1. Two longitudinally conducted researches on Cantonese tone
section, a Thai study focusing on phonetic and phonological development in early speech will be presented in 2.4.1. Two longitudinally conducted researches on Cantonese tone