Chapter 4 Results and analysis
4.2 Subgroup analyses in monosyllabic and disyllabic tokens
4.2.5 Reanalysis in overall data
國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
73
from the mass application of [21-35] tone combination in motherese. Other than [21] and [35], the accuracy rates between different syllable positions in high-level tones [55] and falling tones [51] were similar. The neutral tone, as mentioned before, was limited to the phonological constraints that it is illegal to appear in the utterance-initial position, so its accuracy rate in disyllable 1 was inapplicable.
To sum up, the results between section 4.2.2 and 4.2.3 demonstrated that the frequencies and accuracy rates dropped a lot in the first syllable position of low-level tones [21] and the second syllable position of rising tones [35] after excluding the tone combination [21-35]. It indicated that the occurrences and stabilization of [21] and [35]
separately in other syllable positions were not as high as those in [21-35] combination.
Therefore, we could conclude that [21] and [35] were acquired as a phonetic whole in [21-35] combination but were not acquired individually in monosyllables or other tone combinations in disyllables.
4.2.5 Reanalysis in overall data
After excluding the [21-35] combination in disyllabic tokens, the results of frequencies and stabilization in [21] and [35] became more reasonable. In the precious section, the results were presented separately in different syllable positions. In order to compare the ranking of frequency and accuracy rate among all tones, the syllables in
‧
monosyllabic and disyllabic tokens should be combined together. The modified overall data here contained all monosyllable and disyllabic tokens except for [21-35]
combination.
With regard to the modified data in frequency analysis, the results were different from the original overall data in Figure 4.1.
Table 4.14 Modified number of tokens and frequencies of tones in all syllabic
tokens[55] [35] [21] [51] Neut Total
Number of tokens 960 441 381 565 163 2510 Frequencies of tones 38.2% 17.6 % 15.2% 22.5% 6.5%Figure 4.10 Modified frequencies in overall data
Now that we excluded the tone [21-35] combination from the overall data, we could finally rank the frequencies among all Mandarin tones. Based on the significantly different results presented in Figure 4.3 (χ2=27.9, p<.001), firstly, the high-level tone [55]
was the most frequently used tone, accounting for 38.2% of the total 2510 syllables (3536-513x2=2510). Secondly, [51] were used less frequently, with the percentage of 22.5%. The third and fourth places were [35] and [21], which obtained 17.6% and 15.2%
38.2%
‧
respectively. The least used tone was neutral tone, accounting for 6.5%. The frequency ranking according to Figure 4.10 would be [55]> [51]> [35]> [21]> N.
The same modified data were applied to the calculation in accuracy rate. The accuracy rates below were arranged without separating different syllable positions after excluding [21-35] combination.
Table 4.15 Modified number of tokens and accuracy rates of tones in all syllabic
tokens[55] [35] [21] [51] Neut
Number of correct tokens
/ number of targeted tones 464/504 193/250 212/307 298/364 91/111 Accuracy rates of tones 92.1% 77.2 % 69.1% 81.9% 73.3%
Figure 4.11 Modified accuracy rates in overall data
In the tonal accuracy rates in Figure 4.11, the differences of the five percentages was significantly different (χ2=18.157, p<.001). The most frequent tone [55] also ranked as the most accurate tone, and it reached as high as 92.1% of correctness. The falling tone [51]
showed lower accuracy rate than [55], which accounted for 81.9% of accuracy. The rising
92.1%
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
76
tone [35] and neutral tone were rather unstable, and their accuracy rates reached 77.2%
and 73.3%. The low-level tone [21] had the lowest accuracy rate which was 69.1%. As a consequence, the Figure 4.11 showed that children produced [55] more stable than [51], [35] and neutral tone were less stable than [51], and [21] was the most immature tone among all tones. The ranking of the accuracy rates of the tones would be [55]> [51]>
[35]> Neut> [21].
All in all, the ranking of the frequencies before excluding [21-35] was [55]> [35]>
[21]> [51]> Neut, and became [55]> [51]> [35]> [21]> Neut after modification. After the adjustment, the ranking of tone emergence coincides with that in tone frequency. It revealed that the tone that appeared earlier was used more frequently by children. As for the ranking of accuracy rate, it was originally [35]> [55]> [21]> [51]> Neut, but it changed to [55]> [51]> [35]> Neut> [21] after excluding [21-35] combination. The unreasonable high accuracy of [35] in the first result switched to the third place at the reanalysis, and [21] became the least stable tone. The ranking of accuracy rate in reanalysis was similar to that in tone emergence and tone frequency except for the sequence of [21] and the neutral tone. However, the neutral tone was not compatible in frequency ranking, because there is positional constraint for it to appear only in utterance-final. Thus, we could hardly compare the development of neutral tone with other lexical tones in frequency analysis.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
77
4.3 Substitution pattern in tonal errors
When a tone has not fully acquired yet, children would make consistent tone errors.
If one tone is often replaced with another tone, the low accuracy rate would suggest that the tone has not reached stabilization (Dodd 1995). Tones that were different from the target tone would be counted as a tonal error. The tonal error pattern was also studied by many linguists, because it could reveal whether children would use a more unmarked tone to replace a more marked tone.
The substitution pattern in the following figure showed the target tones and the realized tones in a matrix. For instance, if a child wrongly produced [njow21-njow35] ‘a cow’ as [njo55-njo35], the target tone in the first syllable would be [21]. [55] would be
the selected tone that realized in the tonal error.
Table 4.16 The matrix of substitution patterns of tonal errors in disyllabic words
Target toneRealized tone
[55] [35] [21] [51] Neutral Total
[55]
32 32 49 7120
37.6%[35]
24 63 7 094
29.5%[21]
24 11 14 453
16.6%[51]
13 14 8 944
13.8%Neutral
0 5 2 18
2.5%Total 61 62 105 71 20 319
19.1% 19.4% 32.9% 22.3% 6.3%
There were totally 319 tonal errors in the overall data. Table 4.16 showed the matrix in
‧
specific numbers of tonal errors that the upper row represented the target tones and the left column represented the tones that realized in substituting the error tones. The percentages on the right-hand side indicated which tone was more likely to replace the inaccurate tones. The percentages in the lower row represented which tone was more frequent to make errors. The most frequent tone which children chose in replacing an error was [55] in 37.6%, and the most frequent tone in making tonal errors was [21] in 32.9%. The percentages on the right column were put into the bar graph in Figure 4.12.
Figure 4.12 Percentage of the realized tone in tonal errors
In Figure 4.12, [55] was the first choice to replace other tones in errors which accounted for 37.6%. The [35] was also frequently used in replacing immature tones, accounting for 29.6%. The low-level tone [21] was realized less frequently in tonal errors, accounting for 16.6%. The falling tone [51] was a highly frequent and mature tone, but it did not replace tonal errors frequently, accounting for 13.8%. The neutral tone was the last choice children choose in substituting for tonal errors (2.5%). To sum up, when a tonal error occurred, children preferred to replace the immature tone with [55] the most, with [35]
37.6%
‧
the second, with [21] and [51] the third, and with neutral tones last. The ranking of the realized tones in substitution would be [55]> [35]> [21]> [51]> Neut.
4.4 The age of tone emergence and stabilization
The table in the following summarized the age when tones were emerged and stabilized in every subject. The table also recorded the age when children stabilized a tone in 66.7% and 90% of correctness respectively. The 66.7% of the stabilization represented the fact that two third of the tone were produced correctly, and 90% of the stabilization represented higher degree of correctness.
Table 4.17 Age of emergence and stabilization of each tone
Tone #1 #2 #3 #4 #5 #6
Note: the capital letters NA which stood for “not applicable” indicates that the tone has not emerged or stabilized yet during the observation.
The age of emergence had been presented in the beginning of this chapter, the age of
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
80
stabilization in two different degrees were added in this table. There were several things which could be presented in Table 4.17. We could compare the age of stabilization of the same tone in different subjects or compare the same subject in different tones. The individual differences could be seen in the first kind of comparison, and the chronological ordering could be seen in the second kind of comparison. From the comparison, we could figure out several questions, including which tone was emerged first in each subject, which subject stabilized all tones first, at what age did he/she stabilized all the tones, how long did subjects take to stabilize a tone after the tone was emerged, and roughly at what age have all subjects stabilized all tones.
Overall, the earlier appeared tone was [55] that emerged at 0;11 produced by subject
#5. Subject #3 and #4 were the earliest children to have stabilized all tones in the degree of 90% stabilization at the age of 1;4, and subject #1 has reached the same goal at 1;6.
Subject #2 and #5 have not stabilized neutral tone and [35] yet respectively. Subject #6 has not stabilized any tone yet, and there was only [55] emerged during the observation age from 0;10 to 1;5.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
81
Chapter 5
Discussion
5.1 Summary of the findings
In this study, the acquisition of Mandarin tones in six children aged 0;10 to 1;6 is observed. To describe children’s tonal development specifically, we age-tracked the age of emergence, calculated the frequencies and accuracy rates, and presented the substitution pattern in tonal errors in the overall data. Moreover, we classified the total of 2062 tokens into monosyllabic and disyllabic tokens, and have noticed that there were significant differences in frequencies or accuracy rates among different syllable positions due to a mass usage of the tone combination [21-35] in reduplications. After all tokens of the combination [21-35] were excluded, we reanalyzed the data again and obtained the results below:
Table 5.1 Tone acquisition orderings in different measures (1) Tone emergence [55] > [51] > [35],[21] > N (2) Tone frequency [55] > [51] >[35] > [21] > N (3) Accuracy rate [55] > [51] > [35] > N > [21].
(4) Substitution pattern [55]> [35]> [21]> [51]> N
(1) Tone emergence: the high-level tone [55] tended to emerged the first, followed by the falling tone [51]. Rising tones [35] and low-level tones [21] were appeared later.
Neutral tones appeared last or together with the last appeared citation tone.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
82
(2) Tone frequency: high-level tones [55] had the highest frequency of occurrences, followed by falling tones [51]. The frequencies of [35] and [21] were originally high, but after excluding the outlier combination [21-35], their frequencies became lower than [51]. The occurrence of neutral tones is limited to utterance-final position, so its frequency revealed to be the last.
(3) Accuracy rate: results showed that high-level tones [55] were the most stabilized tone.
The rates on falling tones [51] and neutral tones were similar and lower than [55].
Rising tones [35] were acquired less stable than [51], and the most unstable tones was the low-level tone [21].
(4) Substitution pattern: high-level tones [55] realized the most in tonal errors. Rising tones [35] seemed to replace immature tones more often than [21]. The tone that least occurred in tonal errors were falling tones [51]. Neutral tones would scarcely replace other tones.
(5) The age of tone emergence of stabilization in each subjects were summarized in Table 4.17 and it presented the individual differences and the chronological ordering in each tone.
Coincidentally, the high-level tone [55] emerged the first, and it also ranked as the most frequent and stable tone. Falling tones [51] were consistently ranked in the second place within tone emergence, frequency, and accuracy rate. Rising tones [35] and
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
83
low-level tones [21] appeared late, and were also less frequent and stabilized later than [55] and [51]. For the time being, the four lexical tones, [55], [51], [35] and [21], ranked in the same order among tone emergence, tone frequency, and accuracy rate. From the similarities mentioned above, it seemed that the tone children acquired earlier was also the tone children preferred to use. The fact that the more frequently used tone obtained higher accuracy showed that children would avoid producing immature tones in the early developmental stages.
However, there were inconsistent rankings of the neutral tone between tone accuracy rate and other measures that it occurred last but stabilized earlier than [21]. For the tone emergence, the neutral tone could be seen to emerge at the same time with the last appeared lexical tone. Although the neutral tone seemed to appear last in most subjects, it still showed up in a short time after the lexical tones or even at the same time with the last appeared lexical tone. For tone frequency, due to the phonological constraint that neutral tone could only appear in the utterance-final position, it was unfair to rank neutral tones with other lexical tones in frequency ranking. Yet, the neutral tone could still be ranked in accuracy rate, and its ranking was higher than [21]. In short, the neutral tone was emerged and stabilized roughly at the same time with the last appeared and last acquired lexical tone.
The other inconsistency was found in the ranking of substitution pattern. The
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
84
substitution pattern had different order from tone emergence, tone frequency, and accuracy rate that the second highest tone [51] in frequency turned out to be the last appeared tone in replacing tonal errors. It means that children were less preferred to replace a tonal error to [51]. Nevertheless, from Table 4.16, we found that [51] was the most preferred tone in replacing tonal errors in neutral tone. The reason could explain why [51] was the less preferred tone in tonal errors. Because the neutral tone has the least occurrences in lexical use that it only occurred in the particular position and in specific types of lexicon, the chance for [51] to replace others in tonal errors were also few.
5.2 Comparison with tonal acquisition studies in Mandarin
5.2.1 Age of acquisition
Several studies have pointed out the age children acquired the tonal system. Chao (1951) stated the age of distinguishing [35] and [21] was at 28 months old, and Jeng (1979) showed that by the age of 1;7, his two subjects were able to produce all tones accurately. Zhu (2002) presented more specifically that [55] was first acquired at the age between 1;2 to 1;5, and [21] was acquired last at the age between 1;4 to 1;9. By the age of 1;9, all lexical tones were acquired by all of the four subjects. First of all, the age that Chao (1951) presented was too old that in the following literature and the results in this study, children were able to distinguish the four contrastive tone at as early as 1;2 in Zhu’s study and 1;1 in this study. Secondly, the age of acquiring the four tones that Jeng
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
85
(1979) pointed out was 1;7, and that in Zhu’s study was 1;9. In the current study, five of my subjects have reached 66.7% of stabilization by the age of 1;6. My subjects seemed to acquire the four lexical tones a little bit earlier than those in Jeng and Zhu’s studies, but the gaps between these ages were small. In substance, though children had individual differences, the earliest age that could master the four contrastive tone was at about 1;6.
5.2.2 Order of tonal acquisition
With regard to the ordering of tonal acquisition, Li and Thompson (1977) reported that [55] was the earliest acquired tone, and [51] was the second. The dipping tone [214]
and the rising tone [35] were acquired last. Zhu (2002) described the same ordering that the high-level tones [55] and falling tones [51] stabilized earlier than rising tones [35], and falling-rising tones [214] were the last to stabilized. The order of stabilization in this current study was similar to these two previous studies that the first stabilized tone was [55], followed by [51], and the rising tone [35] and low-level tone [21] were stabilized last. It was surprising that the sequences of tonal acquisition were similar from children who lived in Taiwan thirty-five years ago and children who acquired Mandarin in Beijing.
All of these studies showed that high-level [55] was the most unmarked tone, because children tended to acquire this tone first consistently. The more marked tones were [35]
and [21] for the reason that they were acquired late by children. Ohala (1978) proposed the articulatory effort theory that the rising tones needed more energy and were more
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
86
difficult to produce. This theory could be verified from all the results in the three studies.
5.2.3 The tonal combination of reduplication in Taiwan Mandarin
In chapter 4.2, the issue of the tonal combination [21-35] in reduplication in motherese was presented. However, there was almost no study noticing the particular use of tonal combination [21-35] of disyllabic reduplications in the Mandarin tonal acquisition in Taiwan. In many cross-linguistic studies, it was reported that reduplication was the major form in children’s early production (Grunwell, 1982). The combination [21-35] is widely used in motherese by care-takers who spoke Mandarin in Taiwan. The tonal acquisition would definitely be influenced by the tonal combination in reduplications.
There was also no research proposed that the tones in reduplicative forms may not be acquired individually, but may be learned together as a prosodic whole. The evidence presented in this study showed that in [21-35] combination, the accuracy rates of [21] and [35] individually were low, but they become much higher in this mass used combination [21-35]. It may be caused by the influences in adults’ input that care-takers tended to transfer words into reduplicative forms that were mostly in [21-35] tone combination.
Researchers believe that children were sensitive to the repeated and regular form of speech sound and could imitate the same pattern more easily (McClelland et al., 1986).
The repeated tone pattern [21-35] might be learned more quickly in children’s mind than
‧
the individual [21] and [35] due to the more occurrences of the form [21-35]. But more examination should be conducted to actually analyze the frequency of input in these two tones; otherwise, the prediction would only be an assumption.
5.3 Cross-linguistic comparison
In chapter 2, several cross-linguistic studies in tonal acquisition were reviewed.
These studies provided the developmental sequences of tones in several languages or dialects. In the following, I would like to compare the results of the current study to the previous tonal studies, and see whether language acquisition is universal or language-specific. I will also apply the cross-linguistic evidences to examine the theory of markedness presented by Yip (2002).
The result of the current study showed that the tone acquisition ordering in Mandarin was [55]> [51]> [35]> [21]. To compare the orderings cross-linguistically, I gathered the tone acquisition orderings below.
Table 5.2 Tone acquisition orderings of cross-linguistic studies Cross-linguistic studies Tone acquisition orderings
Thai (Tuaycharoen 1977) [33],[11] > [224] > [45],[51]Cantonese (Tse 1978) [55],[11] > [33] > [13],[22],[25]
Cantonese (So & Dodd 1995) [55],[33] > [25]> [11],[13],[22]
Taiwanese (Tsay 2001) [55] > [53],[33] > [13] > [11]
Mandarin (Li & Thompson 1977) [55] > [51] > [214],[35]
Mandarin (Zhu 2002) [55] > [51] > [214],[35]
In Table 5.2, though the tone values in their inventories were different from each other, we could still compare and contrast the chronological orderings by viewing the tones in
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
88
features. The level tones could be characterized by high, mid, and low tones, and contour tones could be categorized into falling and rising tones.
These cross-linguistic data could be used to examine the three constraints in the tonal markedness theory presented by Yip (2002).
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
All studies including the results in this study agreed the first constraint, contour tones were more marked than level tones. The first acquired tones were all level tones and all the contour tones were acquired later. The high-level tones [55] in Taiwanese, Mandarin, and Cantonese found to be the first acquired tones, and the mid-level [33] and low-level tones [11] in Thai and Cantonese also developed earlier than contour tones. There were no studies showed that contour tones were acquired earlier than level tones, so it could be generalized that level tones were more unmarked universally.
If we considered the second constraint, we could find that the sequences were different between the study in Thai and those in other language. The falling tones [53] in Taiwanese and [51] and Mandarin were acquired earlier than the rising tones, [13] and [35], but the sequence was opposite in Thai that Thai-speaking children in Tuaycharoen’s study learned the rising tone [224] earlier than the falling tone [51]. Cantonese has no
‧
attributed to individual differences.The last markedness rule could only apply to level tones, so contour tones would not be included in this comparison. Comparing the acquisition ordering among high-level, mid-level and low-level tones, only Thai fulfilled this constraint that its low level tone acquired earlier than the high-level tone. This constraint could hardly find evidences from other cross-linguistic data except for Thai. The first acquired level tones in Thai were mid-level [33] and low-level [11] tones, and the high-level tone [45]3 in Thai was acquired the last. However, in Taiwanese and Mandarin, the high-level tones [55] were acquired the earliest. Especially in Taiwanese, the three level tones were acquired in the sequence of [55]> [33]> [11] which revealed the opposite ordering to the third constraint Yip proposed. The results from this current study also disapproved the third constraint that the high-level tone [55] was acquired earlier than the low-level tone [21] in Taiwan Mandarin. In Cantonese, the evidence from Tse’s (1978) study was not comparable because it showed the spontaneous acquisition for high-level tone [55] and low-level tone [11]. The other Cantonese study done by So and Dodd (1995) suggested that high-level
3 Duanmu (2000) pointed out the of five-point scale of tonal representation had problems. For example, the differences between [21] and [11] could not be detected by Mandarin native speakers. That is, if the slight falling or slight rising would not contrast meanings, they could be seen as level tones. In the Thai tonal system, there is a [45] tone. Because there is no high level tones similar to this tone, we could treat this tone as a high-level tone.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
90
tones were more unmarked than low-level tones that the high-level [55] and mid-level [33]
were acquired earlier than the mid-low-level [22] and the low-level [11].
were acquired earlier than the mid-low-level [22] and the low-level [11].