3. METHOD
3.4 Data Analysis
國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
participants still have any problems on vocabulary that wordlist does not provide, they were allowed to ask questions any time.
3.3.2 Formal Testing
All problematic items in the pilot study had been replaced. Before the start of the task, the directions to the test were illustrated and an example was displayed by the tester. The test began after the confirmation of participants’ understanding about the test. Besides, participants were told to ask any question any time if they had any difficulty during the experiment.
Both of the tasks were conducted in the classroom. Participants were reminded to do the experiment as carefully as they could in order to assure the experiment of variability and reliability of the experiment.
3.4 Data analysis
Here is some of the policy on how to calculate the collected raw data. The directions of scoring will be demonstrated in this section.
In GJ, three possible answers, active voice, passive voice, and both are marked by numbers 1, 0, and 2, respectively. The number, 3, represents the case that
participants do not choose any answers, which are regarded as discarded item. They would not be calculated in the study. In CET task, two possible answers, actives and passives, are marked by number 1 and 0. The number 3 was reserved for those
‧
discarded items. The examples of discarded items are listed as follows.
Table 8: The examples of discarded items from participants’ response
Test item Participants’ response
車禍發生了 There was a car accident happened/Car
accident happening
這位先生中毒了 He get the poison
這隻貓咪動了 The cat started to move
餐廳預約了
I have been reserved the restaurant/I reserve the restaurant/ Reserve to (without subjects)
車禍發生了 It just happened a car accident
那份文件扔掉了 Throw away the document
這台腳踏車修好了 This bike is finish to repair
The standard of whether or not to abandon data is based on the
argument-predicate structure. Even though the voice form of the main verb happen in the predicate is correctly used as in It just happened a car accident, the participant otherwise regards the verb happen as a transitive verb. It will be contradictory if the verb use happen is thought of as correctness on the usage of non-alternating
unaccusatives. Besides, the data with incorrectly or alternatively using verbs such as He get the poison, The cat started to move and This bike is finish to repair would also
be discarded since the main verbs are not what we concerned. Moreover, There-be sentence pattern such as There was a car accident happened would be discarded because this is not a canonical sentence word order in English.
Although the voice forms in the main verb is what we concerned, the argument of a sentence is also the factor in establishing the voice forms differences (Croft,
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
1995). L2 learners are likely to use passive voice when an object appears in the subject position (Yip, 1995). Thus, those imperative sentences such as Throw away the document and the agent insertion as in I have been reserved the restaurant and I
reserve the restaurant would be discarded. Undoubtedly, fragment sentences such as
Reserve to are viewed as discards. Those discards would not be analyzed in the study.
For the level of low-intermediate proficiency, a great deal of double verbs appears in L2 learners’ production such as The lost kid was find and The man is poison. In the case, the criteria of identifying active or passive voice to L2 learners
with respect to double verbs is based on the occurrence of passive tense — copular verb with past participle. In other words, if participants could use the passive voice in any other item such as A car accident was happened in his or her test sheet, they would be regarded to be able to use passives since they could use “be V-ed.”Therefore, their production in any other item in the sheet such as The lost kid was find and The man is poison would not be regarded as passive voice. In the data, many participants
could use Car accident was happened in the sentence 車禍發生了 “che huo fa sheng le” or Examine4 had been bribed, you can pass exam in the sentence 主考官被買通 了，你一定可以通過考試 “zhu kao guan bei mai tong le, ni yi ding ke yi tong kuo
kao shi.” Once these kinds of sentences were found in the data, double verbs in the
4 Obviously, the word examine here is a typo. Participant just substitutes examine for examiner. This
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
40
test sheet would be regarded as active voice; otherwise doubles would be thought of as passives.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
CHAPTER FOUR
RESULTS AND DISCUSSION
The chapter aims to present the results conducted in chapter three and discuss the findings of the research questions. There are three parts according to the research questions in the following sections: (a) the U-shaped curve in learning of
non-alternating unaccusatives, (b) the influences of alternating unaccusatives and passives in the U-shaped learning of non-alternating unaccusatives, and (c) animacy effect on learning of non-alternating unaccusatives.
4.1 The Results and Discussion of U-shaped Curve in Learning of
Non-alternating Unaccusatives
4.1.1 Results
The study of U-shaped curve of non-alternating unaccusatives is tested in CET.
The test involves four groups, in which low proficiency group had not learned passive voice while the other groups had. Unaccusative Trap Hypothesis suggests that
overpassivization on unaccusatives occurred in the second stage when L2 learners had learned passives while no overuse passivization happened in the first stage (Oshita, 1998, 2001). Research also supported that the use of unaccusatives would be subsumed after learners acquired passive voice (Zobl, 1989). Participants’
performances on the use of non-alternating unaccusatives are presented as follows.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
42
Table 9: The correct rate of non-alternating unaccustives
Groups Proficiency Correctness on nn-alternating unaccusatives (%)
A Low 87.91
B Low-intermediate 79.92
C Intermediate 86.80
D High-intermediate 88.19
Figure 3: The correct rate of non-alternating unaccusatves from Chinese to English
In Table 7, the correct rate for Group A is 87.91%, which is nearly as high as Group D, 88.19%. There is an abrupt descend from Group A to B, reaching 79.92%, followed by a sharp increase to 86.80% for Group C. The curve represented in the Figure 3 has a “steep cliff” in learning of non-alternating unaccusatives. Although it is more like a V-shaped curve, the trend of the curve is as expected.
The experiment has conducted on 4 groups, group differences for independent variable, participants’ scores for dependant variable. After the collected data were decoded, the data were submitted to the one-way ANOVA on SPSS software. The analysis showed no significant difference in the use of non-alternating unaccusatives
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
among groups, F(3, 113) =1.931, p >.05, p=.129. However, as we looked into the analysis between two groups in post hoc. The analysis revealed a statistically significant difference between Group A and Group B, and marginally significance between Group B and Group C. There exists a huge drop between Group A and Group B. The possible reasons for the gap of performances might be supposed to
overpassivization and the phenomenon of subsuming unaccusatives. Nevertheless, overall it appears to be no differences with respect to the understanding of language development of non-alternating unaccusatives. Additionally, Levene’s test for equality of variance between groups is not significant, that is, the numbers of sample
differences between groups will not influence the result. The result does not reject the first hypothesis: it will be a U-shaped curve in learning of non-alternating
unaccusatives. Moreover, the result evidences the fact that unaccusatives seem to be subsumed as learners acquired passive voice (Zobl, 1989) and matches the three stages of Unaccusative Trap Hypothesis in learning of unaccusatives (Oshita, 1998, 2001).
4.1.2 Discussion
The findings showed that L2 Chinese learners do overpassivize non-alternating unaccusatives in certain point of language development. The phenomenon occurred in the time when L2 learners acquired the passives, as the result displayed in Group B.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
44
After L2 learners acquire passives, they seem to be more willing to use passive voice and were reluctant to accept NP-V order. It seems that the use of unaccusatives is subsumed and substituted for passives once an object appeared in the subject position (Yip, 1995).
4.1.2.1 Salience in Language and Reluctance in NP-V Word Order
The phenomenon of overpassivization might result from two reasons: the salience of passives and reluctance in NP-V word order. In the viewpoint of salience, study showed that the frequency of L2 structure could be the reason for being salient (Bardovi-Harlig, 1987). In the study, the construction of preposition stranding (marked form) and the construction of preposition piped piping (unmarked form) were examined by L2 learners of English as a second language. The results showed that preposition stranding (marked form) is acquired before preposition piped piping (unmarked form), which is opposed to markedness hypothesis: unmarked forms are acquired before marked forms. The study suggested that a factor of facilitating language learning is salience. According to Bardovi-Harlig, salience is defined as the frequency in the target language. That is, the higher frequency the structures are, the more salient they are. In the way, language learning would be facilitated through greater input of high frequency. Likewise, we defined salience as the availability of input as Bardovi-Harlig. Even though unaccusatives are commonly distributed in the
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
world (unmarked form) and the structures of passives are not so common in the world (marked form), the structures of unaccusatives are less seen than that of passives in L2 learning English. In other words, passives have much more input than
unaccusatives in English. According to (Bardovi-Harlig, 1987), greater input leads to the learning of passives with ease and learning of unaccusatives with difficulty even though unaccusatives belong to unmarked and passives are marked. Another
viewpoint comes from “Novelty effects”: the structure in L2 is quite different from L1, and this effect would facilitate language learning (Kleinmann, 1977). “Novelty effects”
happened in different languages. In our study, the use of non-alternating unaccusatives with NP-V order with patient subject in English is quite similar to Chinese structure.
(31) Bi mai LE Pens buy LE ‘Pens are bought’
Since Chinese is a topic-comment structure, a number of NP-V word order with a patient/theme subject could be observed in Chinese. Therefore, unaccusatives with NP-V word order are not such “novelty” to L2 learners. However, the elements of passives in English, such as copular verb followed by a participle, are never seen in Chinese. Consequently, based on the assumption of Kleinmann (1977), passives would be more salient than non-alternating unaccusatives in L2 English.
‧
Additionally, the reluctance of accepting unaccusatives with NP-V order is evidenced (Kellerman, 1978; Oshita, 1998; Yip, 1995). Therefore, once the fact that passives are more salient than non-alternaing unaccusatives and L2 learners are reluctant to accept NP-V order with unaccusatives, it is plausible that L2 learners of English overuse passivization on unaccusatives.
4.1.2.2 Transitivization in Non-alternating Unaccusatives
The U-shaped curve in learning of non-alternating unaccusatives illustrated that non-alternating unaccusatives could be correctly used for low proficiency learners.
The phenomenon seems to reveal that there is a successful transfer from L1 NP-V word order to L2 unaccusatives. The determination of cross-language transfer appears to result from language distance (Kellerman, 1978). In his study, language distance involves language-neutral and language-specific items5. Language-neutral items are believed to be common across languages whereas language-specific ones are
considered more unique in his or her language, such as idioms, phonology structure of language, and slang expressions. Kellerman further demonstrated that
language-neutral items are more likely to be transferred than language-specific ones in second language acquisition. Therefore, the successful transfer of NP-V word order in the use of non-alternating unaccusatives from L1 Chinese into L2 English might be
5 In Kellerman’s study, language-neutral items also refer to the more core meanings of lexical items while language-specific ones are associated with more peripheral non-core meanings.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
because non-alternating unaccusatives belong to language–neutral items. However, as proficiency increases, the transferability is decreasing.
What is of interest in the study is that participants transitivize non-alternating unaccusatives. The example of tested items is as follows.
Table 10: The example of transitivizing unaccusatives in CET
爸爸抵達機場了 Father arrived airport
The phenomenon of regarding arrive as a transitive verb even happened in high-intermediate learners. There are two possible reasons for the phenomenon. First, perhaps it might be the successful transferability from L1 to L2 as mentioned before.
The semantic meaning of arrive is like a language-neutral item since the meaning of arrive is believed in many languages. According to Kellerman (1987), L2 learners
therefore might tend to map the L1 language habit (arrive 抵達 “di-da” as a
transitive verb in Chinese) into L2 language. Second, transitivizing unaccusatives might reveal that L2 learners underlyingly view unaccusatives as transitive verbs.
Previous study showed that the reason for overpassivization with respect to
unaccusatives is because learners thought of unaccusatives as transitive verbs (Yip, 1990, 1995). Yip (1990) cited from L1 literature from Bowerman (1983) and L2 adult learners from Ruthford (1987) to support transitivization hypothesis.
(32) *He disappeared himself. (L1) (33) *Do you want to see us disappear our heads? (L1)
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
48
(34) *The shortage of fuels occurred the need for economical engine. (L2) (35) *This construction will progress my country. (L2)
Likewise, the production Father arrived airport by L2 learners of English seems to support Yip’s hypothesis. Nevertheless, the argument of transitivization hypothesis might not be so strong unless the production Airport was arrived by Father is also found in L2 learners’ production since transitives could be always passivized.
4.2 The Influences of Alternating Unaccusatives and Passives with Respect to
Learning of Non-alternating Unaccusatives.
4.2.1 Results
Interactions to non-alternating unaccusatives would be illustrated below. Only three groups, low-intermediate, intermediate, high-intermediate, participated in the task since low proficiency learners had not learned passives; it would be not
convincing on validity to have them choose answers from active/passive voices in GJ task. Again, there are three kinds of verbs tested in the task, alternating unaccusatives, non-alternating unaccusaives, and passives. The correct rate of each verb is shown as follows.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Table 11: Correct rate of unacusatives and passives with different proficiency.
Correct Rate(%)
Proficiency Passive Verbs Non-alternating Unacc. Alternating Unacc.
Low-intermediate 68.56% 41.28% 7.57%
Intermediate 76.38% 67.01% 15.27%
High-intermediate 90.97% 81.25% 7.63%
From the table 9, the correct rate of passives begins at 68.56% while the correct rate of non-alternating unaccusatives and alternating unaccusatives starts at 41.28%
and 7.57%, respectively. Compared to passives, low-intermediate learners seem to have difficulty on alternating and non-alternating unaccusatives. As proficiency level increases, the growing scale of non-alternating unaccusatives seems to be more obvious than that of passives. It seems that the difficulty of the use of non-alternating unaccusatives was overcome soon.
The above table revealed that the correct rate on alternating unaccusatives in the intermediate group is about two times better than the groups of low-intermediate and high-intermediate. The difference of correct rate on alternating unaccusatives among groups of low-intermediate, intermediate, and high-intermediate seems huge. We put the data of the correct rate on alternating unaccusatives in the calculation in ANOVA on SPSS.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
50
Table 12: The analysis of correct rate of alternating unaccusatives in different groups
The result of the analysis showed that there is a significance in different groups on the correct rate of alternating unaccusatives, F (2, 84) = 3.145, p<.05. The result showed that as learners’ proficiency grows, the correct rate on alternating
unaccusatives would be different. However, the correct rate on alternating
unaccusatives would not be getting better or worse all the time as proficiency grows.
Instead, the performances on the correct rate on alternating unaccusatives would become better first and then get worse as the following figure shows.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Figure 4: Correct rate on passives, alternating unaccusatives, and non-alternating unaccusatives with different proficiency.
In Figure 4, as proficiency increase, the correctness of alternating unaccusatives goes up first and then goes down at once. It seems that the understanding of
alternating unaccusatives is nothing to do with proficiency. As for non-alternating unaccusatives and passives, correct rate would be getting higher as learner’s proficiency grows. The language development of alternating unaccusatives is quite different from that of passives and non-alternating unaccusatives.
Simply looking into correctness on unaccusatives and passives could not understand the whole picture of the verbs. The error rate with respect to different verbs could disclosure the learning difficulty and provide whole picture of learning among alternating and non-alternating unaccusatives, and passives.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
52
Table 13: Error rate on alternating and non-alternating unaccusatives and passives with different proficiency
Error Alter. Unacc. (%) Non-alter.
Unacc.(%)
Passives (%)
Error as Proficiency
Non-alter.
Unacc.
Passives Alter.
Unacc.
Passives Alter.
Unacc.
Non-alter.
Unacc.
Low-Intermediate 38.63% 53.03% 7.19% 51.51% 7.57% 23.10%
Intermediate 44.44% 40.00% 6.26% 26.38% 8.33% 14.93%
High-intermediate 60.41% 31.49% 4.86% 13.88% 4.16% 4.86%
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Figure 5: Error rate on alternating and non-alternating unaccusatives and passives with different proficiency
In Figure 5, Y-axis is the error rate and verb differences are represented in X-axis in two rows. On the bottom of Figure 5 are the errors that participants do not correctly use on the type of verbs. In X-axis from left to right, there are three kinds of errors of verbs: error of alternating unaccusatives, error of non-alternating unaccusatives,and error of passives. Over each kind of error of verbs on X-axis are two subsections referring to the verb that participants misanalyzed for. In the figure, three different kinds of curve represent different groups of proficiency: for low-intermediate group, for intermediate group, and for high-intermediate group.
As we could see in the error of non-alternating unaccusatives in Figure 4, participants tend to make the error of viewing non-alternating unaccusatives as
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
54
passives. In other others, participants would overextend passivization on non-alternating unaccusatives. According to the above figure, the error rates of overpassivization on non-alternating unaccusatives are 51.51%, 26.38%, and 13.88%
with the different English proficiency from low to high. On the other hand, the errors with respect to misanalyzing non-alternating unaccusatives as alternating ones are few in participants. No matter how proficient they are, the misanalysis remains in a low level, 7.19% for low-intermediate group, 6.26% for intermediate group, and 4.86%
for high-intermediate group.
According to the above data, they are calculated in one-way ANOVA on SPSS software. The results showed that the error rate of passivizing non-alternating
unaccusatives exist significant differences in different proficiency, F (2,84) = 25.523, p=.000. Since the data is calculated by three groups, we could go further to examine the differences between groups by the analysis of post hoc. The post hoc. data is displayed as follows.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Table 14: The analysis of post hoc. about overpassivization in non-alternating unaccusatives
The above table revealed that the error of overpassivization on non-alternating unaccusatives showed significant differences between groups, p=.000 or p=.029 Further, Levene’s test for equality of variance between groups is not significant.
Additionally, the influence of alternating unaccusatives on the learning of non-alternating unaccusatives is also examined on SPSS. Analysis revealed no
Additionally, the influence of alternating unaccusatives on the learning of non-alternating unaccusatives is also examined on SPSS. Analysis revealed no