As discussed in Section 1.2, children may be sensitive to the referential direction in their pronominal acquisition. Previous literature has found that children would use direction to restrict anaphora (Tavalolian, 1977). And Chinese children did not reject backward coreference as much as the English children did (Lust et al., 1996; Wilcoxon, 1991) Thus, the difference between forward and backward coreference is also one of the major concerns in the present study (cf. Huang, 2011; Wei, 2001).
4.2.1 Overall Findings
The mean scores of the subjects’ responses to forward coreference and backward coreference are presented in Figure 4-2.
1 The mean scores of G1 and G2 for the multiple-interpretation patterns were below average (G1: M = 0.25, G2:
M = 0.40).
63
Figure 4-2 Subjects’ Responses to Different Referential Directions
As shown in Figure 4-2, one-way ANOVA indicates a significant difference among the six age groups in the two referential directions (forward: F (5, 90) = 31.773, p < .001;
backward: F (5, 90) = 48.321, p < .001). Table 4-2 further shows the within-group differences between the two referential directions.
Table 4-2 The p values of Subjects’ Responses to Different Referential Directions
G1 G2 G3 G4 G5 Control
F 23.430 9.679 9.587 5.484 0.147 .523
P .000* .007* .007* .033* .707 .481
A significantly better performance on forward coreference than backward coreference was found in G1 (F (1, 30) = 23.430, p < .001), G2 (F (1, 30) = 9.679, p < .01), G3 (F (1, 30)
= 9.587, p < .01), and G4 (F (1, 30) = 5.484, p < .05). Although there was no significant difference in G5 (F (1, 30) = 0.147, p > .05), they still performed better on forward than backward coreference (M = 0.81 > 0.80). The mean score of the control group for forward coreference was lower than that for backward coreference (M = 0.89 < M =0.90), which did not yield a significant difference (F (1, 30) = 0.005, p > .01).
The post hoc analysis shows that in response to forward coreference, G3 was significantly better than G1-G2 (p > .001), G5 performed significantly better than G1-G3 (p
64
< .05), and the control group did significantly better than all the experimental groups except
for G5 (p > .05). As for backward coreference, G4 significantly outperformed G1-G2 (p
< .001), G5 significantly outperformed G1-G4 (p < .05), and there was no significant
difference between G5 and the control group (p < .05). Based on this finding, we can conclude that G5 had acquired an adult-like grammar with regard to forward and backward coreference of the pronominal renjia.
Table 4-3 further illustrates the differences between forward and backward coreference in the single- and the multiple- interpretation patterns.
Table 4-3 The Comparisons of Subjects’ Responses to Different Referential Directions in Single- and Multiple-Interpretation Patterns
Type Group
Single Multiple
Forward Backward Forward Backward
M SD M SD M SD M SD
As can be seen in Table 4-3, in response to the singe-interpretation patterns, from G1 to G4 the correct responses to forward coreference increased regularly (0.50 > 0.58 > 0.68 >
0.77), whereas the regular increases on backward coreference were just obtained from G2 to G3, and G4 to G5 (G1 = 0.46, G2 = 0.46, G3 = 0.64, G4 = 0.69, G5 = 0.81). This is the main reason why the significant differences between the two directions in the single-interpretation patterns were only found in G2 (p < .01) and G4 (p < .05). In response to the multiple-interpretation patterns, G1-G2 scored below average on the two referential directions (forward: G1 = 0.43, G2 = 0.45; backward: G1 = 0.31, G2 = 0.35). The mean
65
scores for both coreference increased in G3, and G4 significantly performed better than G3 on backward coreference. Finally, concerning the different referential directions in different interpretation patterns, G5 did not show a significant difference with the control group, implying that they had acquired the adult grammar.
4.2.2 General Discussions
As pointed out by Kao (1993), there are no precedence constraints on the direction of referential pronouns in Mandarin Chinese, and backward coreference is considered acceptable in the adult grammar. Huang (1982) and Teng (1985) also provide evidence that backward coreference for pronouns is possible in Chinese.
It has been reported that English children tend to show a stronger preference for forward than for backward coreference (Lust, 2006). As stated in Section 1.2.3, children may be sensitive to the head-direction of their L1s in the acquisition of grammatical anaphora. Some researchers have found that children whose L1 is head-initial prefer forward over backward coreference (e.g., C.S Chomsky, 1969; Goodluck, 1981; Lust et al., 1986; Solan, 1983), and the others have reported that children whose L1 is head-final did a better job on backward coreference (e.g., Lust & Chien, 1984; Lust et al., 1982). In the present study, our Chinese children did not reject backward coreference as much as English children did. This finding is consistent with Huang’s (2011) result that the Chinese children were able to obtain both forward and backward coreference. However, our children aged 4-7 performed significantly better on forward than backward coreference of the pronominal renjia, which seems to contradict to the Principal Branching Direction2 (PBD) in Chinese (Lust & Chien, 1984: 51).
According to Lust & Chien (1984), Chinese is principally left-branching (PLB) similar to Japanese and different from English (Lust & Chien, 1984: 53). On the contrary, Chinese
2 ‘If the PBD of the language is ‘right’, children establish anaphora in a mainly forward direction. If the PBD is
‘left,’ they establish a backward direction of anaphora as unmarked (Lust & Chien, 1984: 51).’
66
shares the same word order (i.e., SVO) as English (e.g., Chao, 1968, 1973; Huang, 1982;
Smith, 1981). As claimed in Slobin & Bever (1982), word order plays an essential role for linguistic constraints on the surface structures of sentences in children’s early syntax.
Therefore, it is likely that our children’s higher mean scores for forward coreference reflected the superiority of word-order strategy used in the acquisition of pronominal coreference.
Moreover, Chinese is distinct from both English and Japanese in that it has been categorized as a topic-comment language (Huang, 1982; Li & Thompson, 1976, 1981). A discourse topic is “semantically relevant to the following comment but syntactically independent of the rest of the sentence (Li & Thompson, 1976: 446).” According to Lust & Chien (1984), the grammatical concept of the PBD is not subject to a discourse anaphora (e.g., a discourse-bound pronominal), which is generally relevant to a discourse topic and expected to work forward in direction. Therefore, having regarded renjia as a discourse-bound pronominal, the children of our study were found to reject more backward than forward coreference in the target sentences.
Another possible account for the better performance on forward than backward coreference in the children’s acquisition of this pronominal might be due to the processing complexity or attentional demand of forward and backward serial recall from cognitive perspective (Case, 1972; Case & Globerson, 1974). Case (1972) argues that the performance on forward serial recall is superior to backward serial recall because the former uses little attentional resources and only requires passive storage. The attentional resources mean the executive process span, or the so-called working memory. In contrast, backward serial recall involves an attentional demanding transformation, which means more attentional resources will be used (Case, 1972). Therefore, a better control of forward than backward serial recall might cause more correct responses to forward coreference in pronominal acquisition.
The following issue concerns with the comparisons of our subjects’ responses to different referential directions in the single- and the multiple-interpretation patterns.
67
Generally speaking, all the subjects performed better on forward than on backward coreference in the single-interpretation patterns. Moreover, with regard to forward coreference in the single-interpretation patterns, the performance of all the experimental groups made regular progress. To contrast, regarding the backward coreference in the single-interpretation patterns, the correct responses of the 6-year-olds increased sharply compared with the 4- and 5-year-olds.
Like the single-interpretation patterns, all the subjects did better on forward than backward coreference in the multiple-interpretation patterns. The 6-year-olds performed well on forward coreference in the multiple-interpretation patterns. It was not until age 7 (G4) did the children significantly performed better on backward coreference in the multiple-interpretation patterns. Therefore, we conclude that there is a pivotal phenomenon that ages 6-7 are transitional stages that the children have become sensitive to different referential directions in different interpretation patterns of the pronominal renjia.