The third research question addresses the contextual effects on the two constructions.
Section 4.3.1 presents the overall findings of comparisons between contextual types in each group, and Section 4.3.2 is a discussion about the findings.
4.3.1 Overall Findings
An overall comparison between contextual types is presented in Figure 4-5, where the
subjects scored the highest on sentences in supporting context, the second on sentences in isolation, and the least on sentences in biasing context (M=0.76 > 0.66 > 0.59).
Figure 4-5. Subjects’ Overall Interpretations of the Contextual Types
With the utilization of the Kruskal-Wallis rank sum test, an overall between-type difference
within the four groups was found (χ2(2)=32.90, p<.001). The two-sample Wilcoxon tests were used to examine pairwise comparisons; hence, as can be seen in Table 4-9, the comparisons between sentences in isolation and in supporting context, and between sentences in supporting and in biasing context were found significant (p<.02).
Isolation Supporting Context Biasing Context
Table 4-9. The Overall between-type Differences
p-value
1.808E-07 0.3489 1.591E-06Note: The adjustment of p-values for three paired comparisons was .024
Impacts of contextual types on donkey sentences and bare conditionals were taken into investigation respectively. Table 4-10 provides the between-type comparisons for donkey sentences.
Table 4-10. Each Group’s Interpretations of DS in the Three Contextual Types
Context
As seen in Table 4-10, the mean scores of KS, Grade 2 and Grade 4 were higher in both supporting and biasing contexts comparing to the scores on sentences in isolation. KS and Grade 2 did not show significant differences among the three contextual types (KS: χ2(2)=1.12,
p=.57; Grade 2: χ
2(2)=3.72, p=.16), while Grade 4 was the only group that showed a significant difference (χ2(2)=10.46 p<.01). Since the mean score of Grade 4 greatly increased in supporting4 Three pairwise comparisons – in isolation vs. in supporting context, in isolation vs. in biasing context, and in supporting context vs. in biasing context – were needed; therefore, the adjusted p-value was .05/3 = .0167, and rounding it to the nearest hundredth was .02. Hence, if the result was a p-value of .04, it was counted as insignificant.
context, pairwise between-type comparisons for Grade 4 indicated that a significance was found in the isolation-supporting context comparison (p<.02). The adults, on the other hand, received a higher mean score in supporting context only, but a lower score in biasing context;
however, the between-type differences were not significant (χ2(2)=3.62, p=.16). Between-group comparisons in each contextual type are shown in Table 4-11.
Table 4-11. The between-group Differences for DS in the Three Contextual Types Context Isolation
Note: The adjustment of p-values for six paired comparisons was .01.
Significant differences were found in the between-group comparisons of sentences in isolation (χ2(3)=57.71, p<.001), in supporting context (χ2(3)=55.76, p<.001), and in biasing
context (χ2(3)=31.05, p<.001). The two-sample Wilcoxon tests reported that the three child groups all interpreted sentences in isolation significantly differently from the adults (p<.01).
KS showed significant differences from the two older groups respectively (p<.01), while the Grade 2-Grade 4 comparison did not show any significance (p=.60). In supporting context,
Grade 2 and Grade 4 did not show a significant difference (p=.06), and only Grade 4 showed no significant difference from the adults (p=.18), while Grade 2 and KS did (p<.01). In biasing context, significant differences were found in the comparisons between KS and the older groups (p<.01), but no difference was found in the comparisons between Grade 2 and Grade 4 (p=.58), between Grade 2 and the adults (p=.36) and between Grade 4 and the adults (p=.72).
After reporting contextual effects on donkey sentences, let’s look into the effects on bare conditionals as shown in Table 4-12, which reports the between-type comparisons.
Table 4-12. Each Group’s Interpretations of BC in the Three Contextual Types
Context
Except KS, who received the highest mean score on sentences in biasing context, the older child groups and the adults all obtained the highest mean scores on bare conditionals in supporting context, followed by sentences in isolation and those in biasing context. The differences of mean scores were found insignificant in KS (χ2(2)=2.88, p=.24), but significant in Grade 2, Grade 4 and the adults (Grade 2: χ2(2)=30.75 p<.001; Grade 4: χ2(2)=26.50 p<.001;
Adults: χ2(2)=78.45 p<.001). Accordingly, the two-sample Wilcoxon tests were used to explore their differences, as can be seen in Table 4-13.
Table 4-13. The between-type Differences for BC in Each Group
Adults
7.147E-05 4.954E-13 4.8E-14Note: The adjustment of p-values for three paired comparisons was .02.
All the three groups showed significant differences in the comparison between isolation and biasing contexts and between supporting and biasing contexts (p<.02). The adults also showed a significant between-type difference in the comparison between isolation and supporting context (p<.02).
The pairwise between-group comparisons in sentences of the three contextual types is shown in Table 4-14.
Table 4-14. The between-group Differences for BC in the Three Contextual Types Context Isolation
Note: The adjustment of p-values for six paired comparisons was .01.
First, it was found that comparisons of mean scores in isolation were insignificant (χ2(3)=8.01,
p=.05). However, a significant discrepancy was found in sentences in supporting context
(χ2(3)=14.49, p<.01). The two-sample Wilcoxon tests showed that comparing with the adults, KS and Grade 2 each showed significant differences (p<.01), but Grade 4 did not (p=.06), and that the three experimental groups did not show significant difference between one another (KS-Grade 2: p=.85; KS-Grade 4: p=.50; Grade 2-Grade 4: p=.45). For sentences in biasing context, a significance difference was found in the overall between-group comparison (χ2(3)=
56.16, p<.001). Pairwise between-group comparisons indicated that only Grade 2 and Grade 4 did not display a significant difference (p=.50), but comparisons between the other groups did (p<.01).
4.3.2 Discussion
Previous studies have proposed that contexts are influential to interpretations of sentences with syntax-semantic interface (Prévost & Paradis 2004, Foppolo 2009, Ireri et al. 2012). In the present study, contextual effects on both donkey sentences and bare conditionals were found.
Although concepts of quantification are abstract, providing context can offer children some clues and time to process target sentences. Nevertheless, the two constructions exhibit different degrees of explicitness in quantification, that is, wh-words in bare conditionals are less explicit than quantifier types in donkey sentences. Hence, different degrees of explicitness render
different degrees of contextual effects, where there were more significant effects on donkey sentences than on bare conditionals.
In donkey sentences, it was found that sentences in context helped the older child groups receive higher mean scores than sentences in isolation did. Regardless of supporting or biasing context, the older children interpreted donkey sentences more accurately. It can be reasoned that memory spans of children are limited; however, with contexts lengthening the span of time to figure out the situations of tasks, children can then interpret better with processing devices working more properly to interpret quantificational sentences. In this case, context suffices for interpreting the vague concepts (DeVault & Stone 2004); hence, children can answer donkey sentences better.
On the contrary, wh-words in bare conditionals are not as explicit as quantifiers in donkey sentences. Bare conditionals themselves are implicit and ambiguous that could not simply be resolved by the provision of context. As found in the mean scores of Grade 2, Grade 4 and the adults, their interpretations of bare conditionals were quite flexible, and easily influenced by context. It is worth noting that the adults’ mean scores in both supporting and biasing contexts were highly affected by context. Since the expected readings for all bare conditionals were universal readings, the impact of context rendered a high mean score in supporting context but a very low score in biasing context. This difference can also be found in the older child group though its mean scores did not display as a huge difference as the adults. It can be reasoned
that Chinese is more a discourse-oriented language (Huang 1984, Ni 1987) and that readings of bare conditionals are more in a pragmatic matter (Dekker 2001, Herburger 2015) where their readings fluctuate on the basis of contexts; hence, when a factor determining a reading of a sentence is implicit and unpredictable, context can easily influence its reading.
In sum, context is certainly a crucial factor that helped the subjects obtain readings more easily and straightforwardly, no matter it was positive or negative contexts. Yet the explicitness of the operator that assigns a reading to a sentence matters. If the operator is explicit enough, like quantifiers, context functions as a synergy, enhancing mean scores and making sentences to be interpreted more easily; however, if the operator, like wh-words, is implicit, context will then dominate readings and thus children will easily be influenced by contextual clues.
For the developmental pattern for the contextual effects, the results show that our children over seven years old were keen to contextual clues; hence, they interpreted sentences with the influence of context. The child subjects under seven years old were not aware of contextual clues (Yatsushiro 2008) since they were still of the stages of learning and producing discourse (Pearson & de Villiers 2005). Preschoolers are still at a sentence level where decontextualized speech is often uttered especially when syntactically complex utterances are expressed (Tannen 1982); accordingly, children above seven years old can start appreciating context to infer readings of sentences.