This section addresses the first research purpose— to find out whether the
English learners’ performance varies with the three types of RVCs, i.e., Act-R, Sem-R and Sta-R RVCs. In the following, we will first present the results of the grammaticality judgment task and then examine whether the English learners’
performance differ according to the three types of RVCs. The results of the three RVC types will be reported first.
Table 4-1 illustrates the subjects’ correct responses to the Act-R, Sem-R and Sta-R RVCs obtained from the GJ task. The task consisted of 18 items divided into three groups according to the type of RVC; that is, 6 items for each type of RVC. Each of the subjects’ correct response would get one point, so the highest mean score for
each RVC type is 6 and the total score is 18.
Table 4-1. Correct responses to the three RVC types in the GJ task
Activity-Result Semelfactive-Result State-Result Type of RVC
Group
Mean SD Mean SD Mean SD
Mid 2.85 1.785 1.70 1.380 3.45 1.504
High 4.10 1.334 2.90 1.714 4.00 1.298
Control 5.55 0.686 5.40 0.821 5.85 0.489
As the mean scores in Table 4-1 indicate, the Chinese controls’ performance was better than that of the two experimental groups. The results of ANOVA revealed that the accuracy of judgment between the control and experimental groups was significantly different (Scheffe, p=.000* for the mid group and p=.000* for the high group), implying that the English learners’ knowledge of the property that RVCs cannot occur with durative linguistic labels was far from native-like. Table 4-1 also illustrates that the L2 learners’ performance differed according to types of RVCs. That is, their performance on Sta-R RVCs was the best, the mean score for the correct responses being 4.0 for the high group and 3.45, for the mid group. Next was the Act-R RVC; the mean scores of the advanced and intermediate learners were 4.10 and 2.85, respectively. The Sem-R RVC had the lowest mean score among the three types of RVCs, the mean scores being 2.90 and 1.70 for the high and mid groups,
respectively. A mixed design ANOVA analysis was conducted to check if the mean differences between any two of the three RVC types reached the level of significance.
The results revealed that the performance on the Sta-R RVC was significantly better than that on the Sem-R RVC, with a p value of .000 (p<.05*). In addition, a significant difference was observed between the learners’ performances on the Act-R and Sem-R RVCs (p=.000*). Though learners’ correct responses to the Act-R RVCs were not the same as those to the Sta-R RVCs (p=.272, p>.05), the difference did not reach the level of significance.
The above findings accorded with our prediction of the acquisition of Sta-R /Sem-R RVCs (and that of Sta-R /Act-R RVCs), but did not correspond to that of Act-R/ Sem-R RVCs. The acquisition of Action-Result RVCs28 is expected for the following reasons. It has been shown that in SLA, the transference of one linguistic label from the L1 to the L2 has the involvement of an obvious correspondence between the learner’s native language and the target language (e.g., Lado 1957; Ard and Homburg 1992; Odlin 1994). As mentioned in Chapter Two, Chinese allows some States to serve as the first constituents of RVCs, while English does not. Moreover, the cross-linguistic comparison indicates that Act-R and Sem-R RVCs are similar to English Accomplishments in that all of them present events consisting of both an
28 Activity-Result and Sem-R RVCs are Action-Result RVCs.
action and a result. However, it reveals that Action-Result RVCs, composed of only a result in their meaning, resemble English Achievements in grammatical properties.
From the discussion, State-Result RVCs display a relatively lower degree of L1-L2 differences than Action-Result RVCs. Thus, it was predicted that the L2 acquisition of State-Results RVCs, which involves less L1 interference, would be easier than that of Action-Result RVCs. Our result has indeed provided empirical evidence showing that the learners did better on State-Result RVC than on Action-Result RVCs and thus supported the analysis that the different degrees of L1 interference leads to unsuccessful L2 learning.
So far, the findings have suggested that the English learners did better on Sta-R and Act-R RVCs. The Sem-R RVCs, on the other hand, posed the greatest difficulty in acquisition to the learners. Since the results of Act-R RVCs and Sem-R RVCs did not conform to our prediction,29 the performances on the three RVC types were further examined to find out why most learners were able to detect the incompatibility between the meaning of the Act-R RVC and a durative grammatical structure, but unable to notice such a feature in the Sem-R RVCs. A closer look was paid to the three types of RVCs when they appear with the imperfective marker zhengzai, the inceptive verb kaishi and the terminative verb tingzhi. Table 4-2 presents the mean
29 We predicted that the performances on Act-R and Sem-R would be similar because the two types resemble each other in the following ways. First, they select both an action and a result as their components. Second, they present an event with a final endpoint.
accuracy scores obtained from the learners’ performances on the three durative markers. There were 18 test items in the GJ task. The 18 items were divided into three groups according to the three durative markers-- zhengzai, kaishi and tingzhi, 6 items for each marker. Hence, the highest mean score for each marker is 6 and the total
score is 18.
Table 4-2. The mean scores of correct responses to questions with the durative linguistic labels zhengzai, kaishi and tingzhi in the GJ task
Imperfective
The mean accuracy scores displayed in Table 4-2 showed that the subjects in the control group outperformed those in the two experimental groups. A significant difference was found between the control group and the two experimental groups (Scheffe p=.000** for the mid group and p=.000** for the high group). Moreover, Table 4-2 indicates the mean score in the inceptive verb kaishi were the highest, the scores being 3.20 and 4.70 for the mid and high groups, respectively, and .the average mean score being 3.95. The learners’ average mean accuracy score of the test items
with the imperfective marker zhengzai is 3.0 and that of the items with the terminative verb tingzhi is 2.55. A one-way ANOVA analysis showed that the learners’ correct responses to the questions with kaishi were statistically significantly higher than those with the other two durative linguistic labels (p=.001* for kaishi vs. zhengzai and p=.000** for kaishi vs. tingzhi), signaling that questions with both an RVC and the inceptive morpheme kaishi posed less difficulty to the L2 learners. Notice also that the average mean scores observed from Table 4-2 seem to indicate that the learners did better on ‘zhengzai’ than on ‘tingzhi’, but the result of ANOVA revealed that their performances on zhengzai items than on tingzhi items, but no significant difference was found statistically (p=.174, p>.05).
The findings presented above suggest that the learners did better on test questions with ‘kaishi’. It was also found that that the L2 learners did not perform like the controls on all the three durative markers, implying that they did not have complete understanding of this feature of RVC. Therefore, the interactions between the three RVC types and the three durative linguistic labels need to be scrutinized in detail in order to find out why the learners did not exactly comprehend that the RVC is semantically instantaneous. Presented in Figure 1 is the learners’ performance on test items with kaishi:
3
Figure 4-1. The learners’ performance on test questions with kaishi
(The numbers 1, 2 and 3 on the X axis stand for the Act-R, Sem-R and Sta-R RVCs, respectively.)
It is apparent in Figure 4-1 that the learners did worst on Sem-R RVCs, indicated with No. 2 on the X axis. An ANOVA analysis revealed that the learners’ performance on Act-R RVCs was significantly better than that on Sem-R RVCs (p=.001*). The analysis also revealed that the learners’ correct responses to the Sem-R and Sta-R RVCs were significantly different (p=.000**). It seems that the L2 learners were least familiar with the meaning of the Sem-R RVCs. More importantly, the unfamiliarity with the Sem-R RVCs may also account for why the learners’ performance on kaishi was not close to the Chinese controls’. A further analysis of the learners’ performance also showed that they did poorly on questions that contained the Sem-R RVCs and durative expressions like the terminative verb tingzhi ‘stop’, as illustrated in Figures
4-2 and 4-3, respectively.
Estimated Marginal MeansEstimated Marginal MeansEstimated Marginal MeansEstimated Marginal Means
mid
Figure 4-2. The learners’ performance on test questions with tingzhi
(The numbers 1, 2 and 3 on the X axis stand for the Act-R, Sem-R and
Estimated Marginal MeansEstimated Marginal MeansEstimated Marginal MeansEstimated Marginal Means
mid
Figure 4-3. The learners’ performance on test questions with zhengzai
(The numbers 1, 2 and 3 on the X axis stand for the Act-R, Sem-R and Sta-R RVCs, respectively.)
Figures 4-2 and 4-3 clearly show that the learners’ performance on Sem-R RVCs (indicated with 2 on the X axis) was the worst among the three RVC types, implying that the Sem-R RVCs were truly confusing to the L2 learners. Though in questions with tingzhi, the Act-/Sta-R RVC enjoyed higher mean scores than the Sem-R RVC, no statistical difference was found (p=.649 for Sem-R vs. Act-R RVC and p=.300 for the Sem-R vs. Sta-R RVC). On the contrary, we found that in test items with the imperfective aspect marker zhengzai, the performance on Act-/Sta-R RVCs was significantly different from that on Sem-R RVCs (p=.000** for Sem-R vs. Act-R RVC and p=.000** for Sem-R vs. Sta-R RVCs). Based on the findings, we can state that most learners were able to detect the incompatibility between the meaning of the Act-R RVC and a durative grammatical structure, but unable to notice such a feature in Sem-R RVCs. It was stated earlier that the learners’ poor performance on Sem-R RVCs is perhaps due to their unfamiliarity with the meaning of this type of RVC. A closer examination of all the test items designed for Sem-R and Act-R RVCs shows that the low mean score of the Sem-R RVCs can indeed be accounted for with the fact that the learners were less familiar with the elements in Sem-R RVC than with those in the Act-R RVC. To illustrate this point, consider tables 4-3 and 4-4:
Table 4-3. Accuracy rates of questions with Activity-Result RVCs
Table 4-4. Accuracy rates of questions with Semelfactive-Result RVCs
RVCs
As shown in Tables 4-3 to 4-4, the Sem-R RVC gua-shang ‘scratch-hurt’ has the lowest accuracy rate, the accuracy rate being 20%. On the other hand, the learners appreciated the Sem-R RVC da-si ‘hit-die’ and kan-duan ‘cut-break’ more, the accuracy rate of which is 50%. As for Act-R RVCs, learners performed best on the item he-guang ‘drink-empty’; the accuracy rate reached 82.5%. The lowest accuracy rate obtained from the Act-R RVCs is 30%, while that of Sem-R is 20%. A further exploration of these RVCs is conducted to find out whether the learners’ poor performance on the Sem-R RVCs results from their unfamiliarity with the component morphemes of the RVCs. The learners’ unfamiliarity with the component morphemes comprising the Sem-R RVCs may be closely related to the frequency of occurrence of these morphemes in their textbook. In order to substantiate this notion, we consulted
the learners’ textbooks30 and found that Activity verbs such as chi ‘eat’, he ‘drink’
and du ‘read’ occurred more frequently than Semelfactives verbs like kan ‘cut’, da
‘hit’31 and qiao ‘knock’. The frequency of input hence seems to play an influential role in the acquisition of RVCs by the English learners. Or learners’ unfamiliarity with the RVC did place extra burden in acquiring the grammatical property of RVC.
The Input Hypothesis, introduced by Krashen (1985), can further account for the learners’ poor performance on the Sem-R RVCs. According to Krashen (1985), for SLA to take place, learners need comprehensible input. That is, ‘acquisition’ takes place when the learner understands input that is a little beyond the current level of his/her competence. Thus, the Act-R RVCs in the tasks, the components of which the learners already know, enjoyed higher accuracy mean score than the Sem-R RVCs, which consisted of morphemes that were not familiar to the learners. The findings can also serve as cogent evidence that supports Chen’s (2004) conclusion that English learners of Chinese are more proficient in using the RVCs introduced as examples in their textbooks.