This section further investigates the difficulty level of each sub-property, that is, Definite/Generic, Indefinite, 3rd person, 1st/2nd person, Non-stative, Stative, Atelic, and Telic. As discussed in Chapter Two, the ka-NP is generally definite or generic.
An indefinite NP rarely occurs in the construction. Also, a ka-NP can only be absent when it is a third person singular pronoun. Omission of a first or second person singular pronoun is prohibited in Taiwanese grammar. As for the constraint on the
ka-verb, a nonstative verb is considered to be more compatible with the ka construction than a stative verb. Finally, a progressive marker is expected to co-occur with an atelic ka-verb but not a telic one.
Table 4.2 illustrates the results of our subjects’ performances on each sub-property.
Table 4.2 Subjects’ Correct Responses on the Sub-Properties of the ka Construction (Average Means)
Property Group
Referentiality Omission Definite/Generic Indefinite 3rd Person 1st/2nd person
M SD M SD M SD M SD
Experimental 0.78 0.26 0.42 0.17 0.40 0.18 0.54 0.35
Control 1.00 0.00 0.53 0.12 0.36 0.24 0.13 0.29
Experimental 0.74 0.28 0.30 0.20 0.59 0.23 0.40 0.21
Control 0.89 0.16 0.00 0.00 0.56 0.14 0.27 0.11
As shown in Table 4.2, the performances of the experimental group and the control group on the sub-properties of Referentiality, Dynamicity and Progressive were quite similar. Regarding Referentiality, both groups showed a significant preference (p<0.05) for Definite/Generic (Experimental: M=0.78; Control: M=1.00) than for Indefinite (Experimental: M=0.42; Control: M=0.53); with regard to Dynamicity, they performed better (p<0.05) on Non-stative (Experimental: M=0.74; Control: M=0.89) than on Stative (Experimental: M=0.30; Control: M=0.00); in the case of Progressive, they achieved apparently a higher score (p<0.05) on Atelic (Experimental: M=0.59;
Control: M=0.56) than on Telic (Experimental: M=0.40; Control: M=0.27). On the other hand, the two groups performed differently on the sub-properties of Omission.
The experimental group obtained a significantly better score (p<0.05) on 1st/2nd Person (M=0.54) than on 3rd Person (M=0.40) while the control group got a higher score (p<0.05) on 3rd Person (M=0.36) than on 1st/2nd Person (M=0.13).
Table 4.3 summarizes the statistic significances for the two groups.
Table 4.3 A Summary of the Statistic Significant Differences between the Experimental and Control Groups
Property Sub-Property Experimental Control
Referentiality Definite/Generic vs. Indefinite p=0.000* p=0.000*
Omission 3rd Person vs. 1st/2nd Person p=0.007* p=0.011*
Dynamicity Non-stative vs. Stative p=0.000* p=0.000*
Progressive Atelic vs. Telic p=0.000* p=0.000*
With the above results, a paired samples test was conducted with the age groups as an independent variable to see if the same tendencies can be found.
Table 4.4 Subjects’ Correct Responses on the Sub-Properties of the ka Construction (Average Means)
Property Group
Referentiality Omission Definite/Generic Indefinite 3rd Person 1st/2nd person
M SD M SD M SD M SD
Nonstative Stative Atelic Telic
M SD M SD M SD M SD
Group 1 0.58 0.28 0.38 0.18 0.48 0.17 0.39 0.20
Group 2 0.80 0.25 0.36 0.16 0.61 0.24 0.47 0.20
Group 3 0.84 0.24 0.16 0.18 0.69 0.23 0.33 0.20
Group 4 0.89 0.16 0.00 0.00 0.56 0.14 0.27 0.11
Table 4.4 shows that in the case of Referentiality, Indefinite was consistently found more difficult than Definite/Generic for each age group (p<0.05) to use in ka sentences. However, the one-way ANOVA result showed that significant differences were found among the performances of each age group. Regarding Definite/Generic, Groups 1 (M=0.58) and 2 (M=0.81) performed significantly worse (p<0.05) than Group 4 (M=1.00) while Group 3 (M=0.95) showed no significant difference (p>0.05). Furthermore, Group 1 scored significantly lower (p<0.05) than Groups 2 and 3 whereas Group 2 had a similar ability with Group 3. The result also suggested that in responding to Definite/Generic, there was a great improvement from ages 4 to
5 while from ages 5 to 6 they performed better gradually. As for Indefinite, the only significant difference (p<0.05) was found between Group 1 (M=0.33) and Group 4 (M=0.53), where Group 4 showed a higher preference for Indefinite than Group 1, which seems to be a counter intuitive result since our adults were supposed to know an indefinite ka-NP is impossible in the ka construction. Later in this section, we will argue that the result is not that surprising if the specific nature of the ka-NP is taken into consideration.
With regard to Omission, Group 1 scored better (p<0.05) on 1st/2nd Person (M=0.59) than on 3rd Person (M=0.41) while Group 4 showed a higher preference (p<0.05) for 3rd Person (M=0.36) than for 1st/2nd Person (M=0.13). Also, no significant difference was found between Groups 2 and 3 in the two sub-properties (p>0.05). However, concerning 3rd Person, the results of between group comparisons showed that no significant difference was found among the four groups (p>0.05). As for 1st/2nd Person, Groups 1 (M=0.59) and 2 (M=0.63) were found to show a significantly higher preference (p<0.05) for 1st/2nd Person than Group 4 (M=0.13) while Group 3 (M=0.41) showed a similarly low preference for 1st/2nd Person as Group 4 (p>0.05).
Regarding Dynamicity, each age group displayed a higher preference (p<0.05) for Non-stative than Stative. Comparing the performances on Nonstative of each group, it was found that Group 1 (M=0.58) scored significantly lower (p<0.05) than Groups 3 (M=0.84) and 4 (M=0.89) whereas no significant differences were found among the other groups (p>0.05). Regarding Stative, the results showed Groups 1 (M=0.38) and 2 (M=0.36) performed significantly higher (p<0.05) than Group 4 (M=0.00); moreover, Group 3 (M=0.16) also displayed a significant reluctance (p<0.05) to use Stative than Groups 1 and 2.
With respect to Progressive, each group showed a preference for Atelic than
Telic. However, significant differences were only found in Groups 3 and 4 (p<0.05).
The between group comparison was also conducted. No significant difference (p>0.05) existed among the four groups on Atelic. As for Telic, the only significant difference was found in Groups 2 (M=0.47) and 4 (M=0.27), where Group 2 showed a higher preference for Telic than Group 4, implying that there was a decrease in Group 2 on Telic. Tse, Tang, Shi and Li (1991) conducted a longitudinal study on children’
acquisition of the Mandarin ba construction. The spontaneous speech data collected in their study was further analyzed by Cheung (1992). The error ratio was shown in Table 4.5 (Cheung 1992, pp. 125-126):
Table 4.5 Spontaneous Speech Samples: Error Ratio
Correct Error Fragment Total
2 yr 62 (67%) 17 (18%) 13 (14%) 92
3 yr 65 (68%) 22 (23%) 9 (9%) 96
4 yr 185 (76%) 25 (10%) 34 (14%) 244
5 yr 127 (69%) 22 (12%) 34 (19%) 183
As indicated in Fahn (1993), the correct use of the ba construction increased as the children grew up except for the five-year-olds. At the age of five, there was a decrease from 76% to 69%. No explanation was provided in Cheung (1992).
Observing the present study, a decrease also found in our five years old’s performances on Telic.
In sum, regarding Referentiality, Indefinite was consistently more difficult than Definite/Generic for each age group (p<0.05). With regard to Omission, it was found that 1st/2nd Person was significantly easier (p<0.05) than 3rd Person for Group 1 while the opposite result was found in Group 4 (p<0.05); also, the two sub-properties were of equal difficulty (p>0.05) for Groups 2 and 3. As for Dynamicity, each group performed better (p<0.05) on Nonstative than on Stative. Regarding Progressive,
Groups 1 and 2 showed no significant preference for either Telic or Atelic while Groups 3 and 4 attained higher scores (p<0.05) on Atelic than on Telic. Table 4.6 summarizes the statistic significances for each age group.
Table 4.6 A Summary of the Statistic Significances for Each Age Group
Property Sub-Property Group 1 Group 2 Group 3
Referentiality Definite/Generic vs. Indefinite p=0.002* p=0.000* p=0.000*
Omission 3rd Person vs. 1st/2nd Person p=0.041* p>0.05 p>0.05 Dynamicity Nonstative vs. Stative p=0.043* p=0.000* p=0.000*
Progressive Atelic vs. Telic p>0.05 p>0.05 p=0.000*
The present results supported our previous discussion that Group 3 had reached an adult-like grammar since their responses to the sub-properties of each main property patterned like the control group, except for Omission. This also confirms our claim that our subjects performed better gradually from ages 4 to 6. And it was found that age five was a transitional point for the acquisition of the ka construction. Our results showed that the five-year-olds performed similarly with the six-year-olds on most sub-properties; comparing to the adults, the five-year-olds obtained similar scores as the adults concerning most of the sub-properties.
The above results confirmed the previous theoretical analysis that Definite/Generic should be an important constraint on the ka construction since each age group showed a positive tendency in judging and producing sentences with a definite or generic ka-NP. However, the high scores for Indefinite in some cases showed that our subjects did not see eye to eye on the general view that the ka-NP cannot be indefinite. The contradictory result might be attributed to the probability that the indefinite NP in the present study was regarded as “specific,” that is, it refers to the referent that is known to the speaker but not the hearer (Li and Thompson 1981,
Radden and Dirven 2007). According to some studies, an indefinite ka-NP is sometimes possible if it refers to something that the speaker has in mind but is not accessible to the hearer (Hung 1995, Chen and Tsao 1995, and Tsao 2003)1.
In the case of Omission, in contradiction to our expectation, not only Groups 1 to 3 did not perform better on 3rd Person than on 1st/2nd Person but Group 1 even showed a significant preference for 1st/2nd Person, indicating that our subjects had not yet acquired the contrast between 3rd and 1st/2nd Person. Besides, the results can be attributed to the existence of strong task effects, that is, the pictorial cued production may not be sufficient for eliciting the ka sentences. However, some inference could still be gained from the results here. That is, Group 1 still did not have the distinction between 3rd and 1st/2nd Person since they totally performed incorrectly on this constraint. Furthermore, Groups 2 and 3 might just partially master the contrast since they showed no preference for either 3rd or 1st/2nd Person, implying that Omission might be mastered relatively late.
With respect to Dynamicity, each group showed a significant preference for Nonstative, confirming the analysis that stative verbs are incompatible with the ka construcitn and that the contrast between Nonstative and Stative had been acquired.
The result, hence, supported the previous theoretical analysis that the function of the ka construction is to emphasize the end result of the object that undergoes the action or to prepose the object to make the action become the semantic focus (Hung 1995;
Tsao 2003). In other words, the action or the result of the action is the key to the construction. However, the degree of action that stative verbs have is relatively low.
Hence, these verbs rarely occur in the ka construction. Also, Nonstative was scored
1 An error ratio of Indefinite was then calculated. It was found that our subjects did judge 83% of the test items with an indefinite ka-NP as grammatical; however, less than 1% sentences produced by them contained an indefinite NP, showing that the indefinite NP in the present study was considered as
“specific.”
slightly lower than Definite/Generic. An item analysis was further conducted to find out the reason. It was found that our subjects performed especially worse on producing the target sentence Banga ka muemue ting ‘The mosquito stung the little sister’. The slightly lower score of the subjects, thus, supported our previous claim that the subjects avoided using a ka expression in responding to our question ‘What did X do to Y’. They simply found that the pragmatic context was inappropriate for a ka sentence with a bare verb to occur since the ka sentence with a bare verb preposes the verbal object to make the verb become the focus.
As for Progressive, it was expected that our subjects would accept Atelic and reject Telic. Contrary to the theoretical expectation, the two younger groups showed no preference for either Atelic or Telic (though their scores of Atelic were slightly higher than those of Telic), however, Groups 3 and 4 did show a strong preference for Atelic than for Telic, implying that not until age six, can the contrast between Atelic and Telic be acquired. Moreover, though our subjects generally preferred Atelic than Telic, their scores on Atelic were not as high as we expected. This phenomenon, again, strongly supported the claim that the discourse function of the ka construction with a bare verb is to make the verb the semantic focus. The progressive marker, however, can only show up in the ka construction with a bare verb. Hence, it was not that our subjects refused to use a progressive marker in the ka construction with an Atelic verb, but that they perceived the discourse context in the present study was not suitable to use the ka expression.
In light of the above results, a refinement of Pinker’s (1989) Theory of Conservativism was made. According to Pinker (1989), three levels of semantic representations:
i) The specific representation of each individual words
ii) The narrow-range (NR) subclasses, which are the set of elements that share particular elements of meanings
iii) The broad-range (BR) classes, which are larger sets of elements that share more general elements of meanings
It is argued that after children master the contrast between the subclass that is permitted in a certain construction and that is prohibited in the construction, they apply the contrast to the construction. According to the present results, we may hypothesize that in acquiring the contrast which is necessary for using a certain construction, children will go through three stages. That is, we further developed Pinker’s (1989) second level of semantic representation into three developmental phases. We may argue that before children fully master the contrast between elements allowed or prohibited in the ka contrast, they have to go through three developmental stages: Stage 1: No Contrast Stage 2: Partial ContrastStage 3: Full Contrast. Viewing these three stages from a statistic perspective, we may claim that Stage 1 should equal to the situation that the scores of an element prohibited in the construction were significantly scored higher (p<0.05) than those of an element admitted. In an analogy, Stage 2 should be equivalent to the situation that the performances of prohibited elements display no statistic difference from those of admitted elements (p>0.05). Stage 3, thus, should be an equivalent to the situation that the score of an admitted element is significantly higher (p<0.05) than that of an prohibited one.
What follows displays which stage our groups were at in acquiring the four constraints according to the aforementioned statistic results:
(4) Dynamicity
As shown in (4), at the age of 4, our subjects had proceeded to the full contrast stage in distinguishing Non-stative from Stative since there was a significant difference between Non-stative and Stative in each group.
(5) Referentiality
As illustrated in (5), in mastering the distinction among Definite, Generic and Indefinite, our subjects were all at the stage of full contrast since all of them performed significant better on Definite/Generic than on Indefinite.
(6) Progressive
As can be seen in (6), in differentiating Telic from Atelic, the 4-year-olds and 5-year-olds were still at the stage of partial contrast as they showed no significant preference for either Telic or Atelic, and the 6-year-olds had moved to the full contrast stage as they displayed a significant preference for Atelic.
(7) Omission
Finally, for the distinction of 1st, 2nd, and 3rd Person, our 4-year-olds were still at the no contrast stage since they significantly performed better on 1st/2nd Person than on 3rd
Person while our 5-year-olds and 6-year-olds had moved to the partial contrast stage as they showed no significant preference for either of the sub-properties.
4.3 Task Effects
To avoid extrapolations caused by a single task, both the comprehension task (i.e., the GJ task) and the production task (i.e., the PP task) were designed in the present study. The GJ task examined the children’s linguistic competence of the Taiwanese ka construction according to the subjects’ instinct judgments on the provided ka sentences. The PP task, on the other hand, investigated the subjects’
performance of the ka construction with the pictorial scenarios and the experimenter’s elicitation. Table 4.7 illustrates the subjects’ mean scores of the two tasks.
Table 4.7 Subjects’ Correct Responses to the Two Tasks (Average Means) Task
Group
PP GJ M SD M SD
Experimental 0.48 0.25 0.59 0.12
Control 0.67 0.09 0.77 0.07
It was found that the experimental group performed significantly better (p=0.000) on the comprehension task (M=0.59) than on the production task (M=0.48). The same finding was obtained in the adult group. For the control group, the PP task (M=0.67) was significantly (p=0.001) more challenging than the GJ task (M=0.77).
The differences between the two tasks for each group are shown in Figure 4.2.
Figure 4.2 Each Group’s Correct Responses to the Two Tasks (Average Means) As shown in Figure 4.2, the PP task was consistently more difficult than the GJ task for all age groups. Also, except for Groups 2 and 3, where no significant preference (p>0.05) was detected, the other two groups displayed a significantly better performances (p<0.05) on the GJ task than on the PP task. Group 1 got a score of merely 0.28 in the PP task but obtained a higher score of 0.56 in the GJ task (p=0.000). The scores of Groups 2 and 3 increased to 0.52 and 0.64 in the PP task, and to 0.55 and 0.67 in the GJ task, respectively. However, no significant difference was found (p>0.05). Generally speaking, task effects were obtained in each group’s perforamnces.
To further examine whether the same task effects can be found on each property of the ka construction, the subjects’ mean scores of the properties in the two tasks are shown in Table 4.8.
Table 4.8 Subjects’ Correct Responses on the Properties of the ka Construction in the Experimental 0.59 0.38 0.54 0.15 0.05 0.19 0.60 0.22
Control 0.91 0.13 0.53 0.12 0.03 0.13 0.78 0.22 Experimental 0.66 0.31 0.63 0.23 0.41 0.32 0.61 0.21
Control 0.89 0.16 1.00 0.00 0.53 0.18 0.75 0.13
As illustrated in Table 4.8, both experimental and control groups obtained a higher score on the GJ task than on the PP task in general except for Referentiality, where opposite results were found. The results of a paired samples test further showed that the experimental group and the control group performed on Omission and Progressive in the two tasks significantly differently (p<0.05).
Regarding Referentiality, though our children’s responses were higher in the PP task (M=0.59) than in the GJ task (M=0.54), the difference did not reach a statistic significance (p>0.05). As for Omission, the experimental group showed a great discrepancy in their performances on the two tasks. Their score of the GJ task was 0.60 while that of the PP task was merely 0.05 (p=0.000). With regard to Dynamicity, the score of the PP task (M=0.66) was similar to that of the GJ task (M=0.63), which showed no statistic significance (p>0.05). Lastly, the children’s comprehension of Progressive (M=0.61) was better than their production of
Progressive (M=0.41) (p=0.000). For the control group, they performed significantly better (p>0.05) on the GJ task than on the PP task for all properties except for Referentiality, where they got a significantly higher score (p=0.000) on the PP task (M=0.91) than on the GJ task (M=0.53). Thus, task effects were found in the subjects’ performances on each property.
To see if the same effects existed in each group’s performances, a paired samples test was conducted with age group as an independent variable. As shown in Table 4.9, task effects were generally found in the performance on each property for each group except on Referentiality, where the subjects’ performances on the PP task exceeded those on the GJ task.
Table 4.9 Each Group’s Correct Responses on the Properties of the ka Construction in In response to Referentiality, Groups 1 and 2 performed equally on the PP (Group 1:
M=0.36; Group 2: M=0.61) and GJ (Group 1: M=0.53; Group 2: M=0.55) tasks (p>0.05), while the other two groups performed on the PP task (Group 3: M=0.81;
Group 4: M=0.91) significantly better than on the GJ task (Group 3: M=0.55; Group 4:
M=0.53) (p<0.05). The exception of Referentiality supported our claim that in judging the grammaticality of the items with an indefinite ka NP, our subjects considered those items as “specific.” Thus, they wrongly judged the items as grammatical. As for Omission, a surprisingly strong task effect was found (p<0.05).
Each group made at least 50% correct responses to the GJ task. However, their
performances on the PP task was merely a far cry. Notice that there was a slightly decrease in the performance of our five-year-olds on the GJ task, though not significant. As for Dynamicity, the subjects’ comprehension was in general prior to production. An exception was found again in Group 2, where production (M=0.73) was significantly (p<0.05) higher than comprehension (M=0.58), suggesting that our subjects may undergo an overgeneralization stage. Finally, Groups 1 and 4 performed better on the comprehension task (Group 1: M=0.61; Group 4: M=0.75) than on the production task (Group 1: M=0.20; Group 4: M=0.53) concerning Progressive (p<0.05). Also, Groups 2 and 3 performed slightly better on the GJ than
performances on the PP task was merely a far cry. Notice that there was a slightly decrease in the performance of our five-year-olds on the GJ task, though not significant. As for Dynamicity, the subjects’ comprehension was in general prior to production. An exception was found again in Group 2, where production (M=0.73) was significantly (p<0.05) higher than comprehension (M=0.58), suggesting that our subjects may undergo an overgeneralization stage. Finally, Groups 1 and 4 performed better on the comprehension task (Group 1: M=0.61; Group 4: M=0.75) than on the production task (Group 1: M=0.20; Group 4: M=0.53) concerning Progressive (p<0.05). Also, Groups 2 and 3 performed slightly better on the GJ than