Further investigation of each sub property helps us understand more about its difficulty levels and how linguistic factors influence our subjects’ language acquisition.
Table 4-2 shows the results of the experimental group and the control group.
5 Please refer to the task effect in Section 4.3.
6 For example, our children usually answered question 21 in the PP Task with an active sentence Baba siong mama ‘Father misses Mother’ instead of the target passive sentence Baba bun mama siong ‘Father was missed by Mother.’ We will discuss the difficulty of each sub property in Section 4.2 later.
Table 4- 2: Subjects’ Correct Responses on the Sub-Properties of Hakka Passives in the Two Tasks (in average means)
TC SC N-Truncated Truncated Long-distance Short-distance
Property
Experimental 0.63 0.24 0.73 0.27 0.63 0.30 0.74 0.25 0.50 0.26 0.59 0.33
Control 0.97 0.13 1.00 0.00 1.00 0.00 1.00 0.00 0.83 0.24 1.00 0.00
VT AV Resultative Action Experiential Adversative N Adversative
Property
As can be seen in Table 4-2, with respect to TC, the experimental group performed better on non-truncated (M=0.67) than truncated (M=0.60) passives. Regarding SC, long-distance passives (M=0.59) were more challenging than short-distance (M=0.67) passives for our subjects. Non-gapped passives (M=0.50) were even more difficult than long-distance passives. As for AM, the subjects performed better on passives with inanimate agents (A-I: M= 0.73; I-I: M= 0.74) than those with animate agents (A-A, I-A:
M=0.63). With regard to VT, experiential verbs were the toughest one to be acquired for our children (M=0.56). Moreover, statistic significances were found by the Paired-simples T-test7 in response to these properties. However, there was no statistic difference found in response to AV or animacy of patient roles, i.e., A-I vs. I-I; A-A vs.
7 Please refer to Table (iii) in Appendix E.
I-A. With respect to the control group, the statistic difference was only found in response to RP, where non-gapped passives were scored the lowest. With the above results, let us see if the same tendencies can be found in each age group. Table 4-3 shows the performance of each age group on sub properties.
Table 4- 3: Correct Responses of Each Group on the Sub Properties of Hakka Passives in the Two Tasks (in average means)
TC SC N-Truncated Truncated Long-distance Short-distance
Property
Control 0.97 0.13 1.00 0.00 1.00 0.00 1.00 0.00 0.83 0.24 1.00 0.00
VT AV
As can be seen in Table 4-3, truncated passives were consistently more challenging than non-truncated passives for each age group. However, Groups 1 and 2, and Groups 2 and 3 performed similarly (p> 0.05)8, showing that our subjects performed better
8 Please refer to Table (iv) in Appendix E.
gradually from ages 4 to 6, but that they improved dramatically from ages 6 to 7. In response to SC, similarity can also be obtained between Groups 2 and 3 on both long- and short- distance passives. Neither was a significant improvement between Groups 1 and 2 on short-distance passives. Nonetheless, with respect to long-distance passives the youngest group (i.e., Group 1) scored the worst (M=0.18), contributing to a significant difference with results of Group 2. Concerning RP, our subjects performed better on long-distance than non-gapped passives, except Group 1. As for non-gapped passives with resumptive pronouns, no significant difference was found from Groups 1 to 3 (p>
0.05), scoring averagely low. Regarding AV, the performances of Groups 1 and 2, and Groups 2 and 3 on adversative passives displayed a gradual development, with no difference being found. On non-adversative passives, a significant improvement was shown from Groups 1 to 2. However, Groups 2 and 3, and Groups 3 and 4 performed similarly. As for AM, Group 3 (at age 6) performed as well as our adults on passives with inanimate agents, i.e., A-I and I-I passives, where no significant difference was found between Group 4 and the control group. With respect to passives with animate agents, our subjects attained full acquisition at age 7. As for VT, the experimental groups tended to perform worst on experiential verbs, except Group 3.
The results above displayed profound implications for theoretical premises. With regard to No Truncation, the results showed that our older children had no
problem producing full (or non-truncated) passives with specified objects. However, they were apt to omit the post-bun NPs (or objects) which were unspecified. In other words, the application of a general human pronoun in the post-bun position, as in Mama bun
ngin hak do ‘Mother was frightened (by someone),’ was proven to be challenging to the
children. This difficulty may result from two reasons. First, our subjects needed more time and efforts mastering unspecified common nouns than specified common nouns (cf.
Brown 1973). The other was the influence of Mandarin Chinese passives, which sets no
restriction on the presence of an object. According to Kuo (1995), the full form is much more extensively used in Chinese children’s speech, and is hypothesized as the prototypical form. And thus, our subjects might omit objects based on the pragmatic requirement. In other words, full forms are usually uttered with a generic agent in
‘normal’ Chinese passives (c.f. Wang 1944), while truncated forms are made especially when they intend to stress the result in specific contexts. However, Hakka passives lack this flexibility; therefore, more time would be needed for our children to acquire the obligatory presence of the post-bun NP in Hakka, both specified and unspecified.
As for Animacy/Inanimacy, we found that agent animacy had significant influence on children before age 7. This result showed that our children preferred inanimate to animate agents. The former is as in Mama bun coido got do le (‘Mother was cut by the knife’) (A-I) and Bozhi bun tai fung chui zeu le (‘Newspaper was blown away by the wind’) (I-I); while the latter is as in A-Bo bun fai ngin zok ziu le (‘A-Bo was caught away by the bad person’) (A-A) and Tong bun cengin siit tet le (‘Candy was eaten by the kid’) (I-A). This finding seemed to be different from the traditional belief that animate NPs draw more attention than inanimate NPs (cf. Givon 1984, Kuo 1995). However, when delving into the passives with animate agents, we noticed that our children performed worst on A-A passives, which are the so-called reversible passives. The poor performances of our subjects on reversible passives confirmed Baldie’s (1976) account that young children usually treat reversible passives as other patterns, e.g., active sentences. On the other hand, our children’s preference to the A-I passives lent a support to what Harris’s (1978) and Bock’s (1986) finding that children passivized more to A-I passives than I-A passives. Accordingly, English-speaking adults and children are considered to passivize more to test pictures with animate patients and inanimate agents than to those with inanimate patients and animate agents (cf. Bock 1986, Harris 1978, Tseng 1997). This indicates that the combination of an animate patient and an inanimate
agent is frequently used in a passive sentence. However, the same preference is not found from the adults. As for patient animacy, it was found insignificant as the role in the acquisition of Hakka passives, which is in an agreement with Tseng’s finding in the acquisition of Chinese passives (1997). Our study confirmed the literature that animacy did play a role in children’s language acquisition (cf. Bock 1986, Dewart 1979, Harris 1978, Kuo 1995, Lempert 1984, 1989, 1990, Tseng 1997).
Concerning Verbal Transitivity, the results substantiated previous findings by indicating that children at 4 and 5 performed better on passives with action and resultative verbs than experiential verbs. In their studies of English, Maratsos et al. (1987) found that action verbs were used more frequently than non-action verbs in children’s passive constructions. Sudhalter and Braine (1985) also found that action verbs preceded experiential verbs in passives acquisition. Maratsos et al. (1987) found that action verbs were much easier than mental verbs in passives. This might result from the main attention children pay to the semantic prototypical role signified by the main verb. Kuo (1995) confirmed that transitivity might matter in the acquisition of Chinese passives by observing children aged 2-6. In Tseng (1997), with an introduction of a new subtype unique to Mandarin, i.e., the resultative verb, she reported that her children preferred the action verbs passives and resultative verbs, but that disliked experiential verbs. The present study shows a similar situation in the acquisition of Hakka passives.
As for Adversity, the study showed that only the four-year-olds preferred adversative passives, but that the older kids and the control group expressed no special preferences.
Our youngest subjects’ choices may reflect the unmarked settings in the brain. However, when they were exposed to adults’ input, they began to accept non-adversative passives.
In fact, the prototype of adversative reading can be found in Mandarin Chinese: If the verb entails a positive meaning in the construction, the sentence will be considered awkward (cf. Tseng 1997). Also, in spoken Chinese, passive sentences are sustained as
the adverse expressions (Chao 1968, Hashimoto 1988, Li and Thompson 1981). However, the neutralization of adversative and non-adversative passives is commonly seen in Mandarin Chinese. It has been found that due to the ‘translatese’ (Chao 1968), i.e., the non-adverse bei sentence from translating foreign passives (Li and Thomson 1981:496), adversity becomes an unnecessary property of Chinese passives, especially in written language (Chao 1968). In particular, scientific and academic writing are mainly involved.
Hakka natives were probably under the same influence of the neutralization. The other possibility for the insignificant difference between adversative and non-adversative passives may lie in children’s interpretations. They may see non-adversative passives as
the ones with adversative implications and interpreted them as negative sentences.
Turning to the sentence patterns involved in properties of Syntactic Complexity and Resumptive Pronoun, we noted that the syntactic derivation of the non-gapped passive and the long-distance passive in Chinese have been put into question9 in linguistic literature. They are said to have same effects with what can be observed in relative clauses. However, theories on relativization strategies are somewhat controversial (cf.
Huang et al. 2008, Mei 1978, Sanders et. al 1972), which results in their undetermined status and thus complex in grammatical structure. When we put them into the child’s language acquisition, we found them more difficult than the simple patterns, i.e., the short-distance passives. The result accorded to the viewpoint that the more complex a structure is, the more time a child would need to get full acquisition (Borer and Wexler 1987). However, the difference between long-distance passives and short-distance passives disappears when children grew older than 5. On Resumptive Pronoun, on the other hand, the non-gapped passive was more difficult than the long-distance passive except for Group 1, suggesting that our children’s poor performance on the non-gapped passives afterwards were highly influenced by adults’ speech.
9 Please refer to Section 2.1.3.
After discussing the subjects’ responses on each sub property with the literature findings, let us see if the data collected in this section corresponds to the results concerning the semantics-syntax mismatch obtained from Section 4.1. As can be seen in Table 4-2, the properties showing significant differences were mainly those which exert more syntactic concepts, i.e., No Truncation, the Resumptive Pronoun, and Syntactic Complexity. On the other hand, comparisons within semantic-related properties such as Adversity, Verbal Transitivity, and Animacy were found less significant. However, one may notice that among semantic properties such as Animacy and Verbal Transitivity, unexpected differences were found, e.g., inanimate agents (A-I and I-I) were scored higher than animate agents (A-A and I-A). In fact, the significant difference was attributed to syntactic caterogies rather than semantic notions. In the former combinations, when the object positions were occupied by inanimate roles, whether the subject was animate or not was without any difference. In the latter where the object positions were occupied by animate roles, whether the subject was animate or inanimate did not matter, either. However, when comparing I-I with I-A, and A-I with A-A, significant differences were revealed. When the subject positions were occupied by inanimate roles as in I-I and I-A, whether the object positions were animate posed a significant difference. Similarly, the situation was found in A-I and A-A. Therefore, the difference in Animacy/Inanimacy was actually due to a subject-object asymmetry10. The present study indicates that the importance of the subject position weighs over that of the object position. The other significant difference found in semantic properties was the experiential verb. Except the reason that the experiential verb was not prototypical in passives, our subjects’ poor performances might also result from its infrequent use in Hakka passives. As what we
10 Subject and object positions exert asymmetrical influences on children’s language acquisition (cf. Cao 2003, Karin 1995, Suzuki 2000, etc.). The phenomenon can be found in the acquisition of Chinese relative clause (Cao 2003), Enlighs Wh-question (Karin 1995), Japanese Case-Marking Particles (Suzuki 2000), and so forth.
suggested before, our children treated the patterns with experiential verbs as cause-and-effect questions and thus failed to produce passive sentences.