3.1 Study I
3.1.3 Procedure
A consent form (see Appendix F) was firstly given to the subjects. After they
agreed to participant in the experiment, they were invited to a quiet meeting room with
an answer sheet given (See Appendix G). All the items were presented on the computer
screen in text with pictures and audio materials. The subjects were asked to listen to the
context and read test items on the screen, and then to write down their answers on the
sheet. In this study, the SM task was conducted first, followed by the AJ task. Each task
2 The test items in the AJ task were presented in written form with traditional Chinese characters without pinyin or English translations.
took the subjects about 10 to 12 minutes to complete.
As for the scoring, a correct word order was given 1 point while an incorrect order
was received no point in the SM task. The average score of different types of test items
was calculated and the incorrect responses were analyzed. In the AJ task, a four-point
scale was employed to evaluate the L2 learners’ acceptability, as follows:
1) Absolutely unacceptable: 1 point 2) Probably unacceptable: 2 points 3) Partly acceptable: 3 points 4) Absolutely acceptable: 4 points
The average of each type of test sentences was counted and compared. A summary of
the procedures of Study I is illustrated below:
Figure 3-1 The Summary of the Procedures of Study I 3.2 Study II
Study II mainly explores L2 learners’ production and acceptability of gapped and
gapless ORCs.
Consent form
SM AJ
Data analysis
3.2.1 Subjects
The subjects participated in Study II were the same as those in Study I. They were
asked to do Study II after finishing Study I.
3.2.2 Methods and Materials
The methods adopted in Study II were the same as those used in Study I. Study II
also contained two tasks: a sentence-making (SM) task and an acceptability judgement
(AJ) task. These tasks were conducted in the same ways as those in Study I but the
materials were different, as shown in the following sections.
3.2.2.1 The Sentence-Making (SM) Task
In the SM task, a total of eight ORCs were included and these sentences were
classified as gapped RCs and gapless RCs. A summary of the test items for the SM task
of Study II is presented in Table 3-6.
Table 3-6 A Summary of the Test Items for the SM Task of Study II
Major Types Example Number Questions
‘The running speed of the thief was faster than that of the policeman.’
4 Qs 2,4,6,8
In the SM task of Study II, each test item was separated into four parts: a DCL
phrase, a clause, a head noun, a VP. The task was conducted in the same way as Study
I. An example of the SM task of Study II is shown in Table 3-7. In the example, even
though the DCL phrase nage ‘that’ can be placed before the clause or placed before the
head noun, the best answer is that the DCL phrase is place before the head noun since
it can avoid the ambiguity, as shown in Table 3-7.
Table 3-7 An Example of the SM Task3 of Study II
Sentence Answer
na ge / shi meiguoren / Wangwu baifang de/ yeye ① ② ③ ④
③①④②
The whole task is presented in Appendix C.
3.2.2.2 The Acceptability Judgement (AJ) Task
A total of eight ORCs were included in the AJ task, dividing into 4 subtypes:
DCL-first gapped ORCs, DCL-second gapped ORCs, DCL-DCL-first gapless ORCs, DCL-second
gapless ORCs. Each type consisted of two sentences. A summary of the test items for
the AJ task is illustrated in Table 3-8. A complete version of this task can be found in
Appendix D.
3 The test sentences in the SM task were presented in written form with traditional Chinese characters without pinyin or English translations.
Table 3-8 A Summary of the Test Items for the AJ Task of Study II
Gapless ORCs Nage yundongyuan youyong de chengji yuelaiyuehao le.
The procedures of Study II were the same as those of Study I. In the SM task, the
data was analyzed in two ways. First, a correct word order was given 1 point while an
incorrect order received no points. The average score of different types of test items
was calculated in the same way as in Study I. Second, correct answers produced by the
subjects were grouped into DCL-first and DCL-second structures. Then the percentages
of gapped and gapless ORCs were calculated. As for the AJ task, the scoring was the
same as that in the SM task.
3.3 Study III
Study III investigates L2 learners’ interpretation on DCL-first and DCL-second
ORCs and ambiguity resolution of DCL phrases.
3.3.1 Subjects
The subjects recruited in Study III were the same as those who had participated in
Studies I and II. They took part in this study after finishing the previous studies.
3.3.2 Methods and Materials
In this study, only an interpretation (IT) task was conducted to investigate whether
context can take effect on L2 learners’ interpretation of ORCs with different DCL
positions. The task consisted of three parts with three types of context: context-free (i.e.
sentence in isolation), biasing and supporting. Each part had eight items but these items
were randomized. A summary of the test items in the two parts is illustrated in Table
3-9.
Table 3-9 A Summary of the Test Items for the IT Task of Study III
Major Types Number Questions
Part I Part II Part III
DCL-first 12 1,2,3,5 1,2,3,4 2,3,4,6
DCL-second 12 4,6,7,8 5,6,7,8 1,5,7,8
Total 24
In this task, two questions were included: one question for the interpretation of
RCs and the other for the ambiguity resolution in RCs. However, Question 1 was
included in every type but Question 2 was asked only in the case of DCL-first structures.
The subjects were asked to read a sentence and then choose an answer from three
options provided for each question based on their interpretation of the sentence.
Part I contained isolated ORCs without context. Table 3-10 is an example of a test
item in Part I.
Table 3-10 An Example of a Test Item of Part I4 in Study III
Zhe wei xiaopengyou xihuan de laoshi piqi hen hao.
‘The teacher that the child likes has good temper.’
1. Qingwen shi shui piqi hao?
‘Who has good tempter?’
( ) A. Laoshi ( ) B. Xiaopengyou ( ) C. Laoshi huo Xiaopengyou ‘teacher’ ‘little.child’ ‘teacher or little.child’
2. Qingwen “zhe wei” zhi de shi shui?
‘Who does “this” modify?’
( ) A. Laoshi ( ) B. Xiaopengyou ( ) C. Laoshi huo Xiaopengyou ‘teacher’ ‘little.child’ ‘teacher or little.child’
As can be seen above, Option A was the answer to Question 1 since laoshi ‘teacher’
was the head noun, and Option C was the preferable answer to Question 2 because the
DCL phrase could modify either the head noun laoshi ‘teacher’ or the adjacent noun
xiaopengyou ‘little child’.
Part II required the subjects to firstly read a biasing context and then read a test
sentence. The context biasing the answer to Question 1 and the association with the
4 The test sentences in the Part I were presented in written form with traditional Chinese characters without pinyin or English translations.
head noun for Question 2. On a basis of their interpretation, they needed to choose an
answer from three options provided. Table 3-11 is an example of a test item in Part II.
Table 3-11 An Example of a Test Item of Part II5 in Study III
You yi ge xiaopengyou zai youzhiyuan feichang ting laoshi de hua.
‘A child listens to teachers’ words at the kindergarten.’
bei qita xiaopengyou qifu le, ye buhui gen tamen chaojia.
‘Even if he is teased by other children, he will not argue with them.’
Zhe wei xiaopengyou xihuan de laoshi piqi hen hao.
‘The teacher that the child likes has good temper.’
1. Qingwen shi shui piqi hao?
In the example, the previous context discussed how xiaopengyou ‘little child’ is
kind and likely to have a good temper. But xiaopengyou is not the head noun of the test
sentence. Thus, the context is biasing and it may result in the wrong interpretation to
answer Question 1 by the subjects. Even though the context shows supportive
information for xiaopengyou ‘little child’, Option B is the answer to Question 1 in that
laoshi ‘teacher’ is still the head noun of the RC. To answer Question 2, the highlight of
5 The test sentences in the Part II were presented in written form with traditional Chinese characters without pinyin or English translations.
xiaopengyou in the context is likely to bias the association with the head noun and the
subjects may choose the adjacent noun xiaopengyou instead of the head noun. Option
B is the preferable answer to Question 2 since the DCL phrase can modify either the
head noun laoshi or the adjacent noun xiaopengyou in a DCL-ORC structure.
In Part III, the subjects were asked to read a supporting context and then a test
sentence. The context supported the answer to Question1 and supported the head noun
association for Question 2. Finally, they needed to choose an answer from the three
options for each question. Table 3-12 illustrates an example of a test item in Part III.
Table 3-12 An Example of a Test Item of Part III6 in Study III
Laoshi yidingyao you naixin, you xiaopengyou fancuo ye bukeyi shengqi.
‘A teacher must have patience. Although children make mistakes, he or she cannot be angry.’
Zhe wei xiaopengyou xihuan de laoshi piqi hen hao.
‘The teacher that the child likes has good temper.’
1. Qingwen shi shui piqi hao?
6 The test sentences in the Part III were presented in written form with traditional Chinese characters without pinyin or English translations.
From the example shown above, the sentences in context show that laoshi ‘teacher’
must be patient and have a good temper. In the test item, laoshi is the head noun of the
test sentence. Therefore, this context can facilitate the subjects’ interpretation for
Question 1. Option C is the answer to Question 1 because it is the head noun of the RC.
Additionally, for Question 2, since the context described more information about the
head noun laoshi, the context tended to support the subjects’ DCL phrase association
with the head noun laoshi instead of xiaopengyou ‘little child’. Option A is the
preferable to Question 2 in that either laoshi or xiaopengyou can be modified by the
DCL phrase.
For the complete version of the items of Part I, Part II and Part III in Study III,
please refer to Appendix E.
3.3.3 Procedure
Part I, Part II and Part III were conducted in order. The answer to Question 1 was
given 1 point and the accuracies of different RC types were calculated and analyzed.
As for Question 2, the frequencies of different options in every item were counted and
analyzed by Chi-square.
3.4 Summary of Chapter Three
In this chapter, we have presented the backgrounds of the subjects, the tasks and
the procedures of the three studies. Study I and Study II contained two tasks (the SM
task and the AJ task) while Study III conducted the IT task with three types of context
(context-free, biasing and supporting). The procedures of the current research are
summarized as follows:
Figure 3-2 The Summary of the Overall Procedures of the Current Study The next chapter reports and discusses the results obtained from Study I.
Consent form
SM AJ
Study I
SM AJ
Study II
Biasing Supporting Study III
Data analysis
Context-free
CHAPTER FOUR STUDY I
As mentioned in Chapter Two, RC types may have effect on learners’ production
and interpretation. Xu (2009, 2014) and Li (2013) found that SRCs were easier than
ORCs for L2 learners, while Wu & Sheng (2014) concluded that ORCs were easier to
produce. Owing to the fact that RC type advantages are still controversial and that task
effect has not been investigated on the L2 learners’ performance, in this study we
explored L2 learners’ performance on different RC types (SRCs and ORCs) and the
factors affecting their performance such as task types (production or acceptability), gap
condition, L1 influence, etc. The following sections display and discuss the findings
obtained from Study I.
4.1 SRCs vs. ORCs
To address the first research question, we explored the L2 learners’ performance
on SRCs and ORCs. As presented in Chapter Three, we designed two tasks (i.e., the
sentence-making (SM) and acceptability judgement (AJ)) for investigation. First of all,
the subjects’ overall performance on the SM task is illustrated in Table 4-1.
Table 4-1 Subjects’ Mean Scores on SRCs and ORCs in the SM Task
Note: ‘*’ means the p-value is smaller than .05 and the sign ‘a’ in the table means that the correlation cannot be computed since the standard error of the difference is 0.
It was found that the L2 learners’ mean score on ORCs (M= .84) was higher than that
on SRCs (M=.66), scoring that ORCs were easier than SRCs for them to produce. The
within-groups comparison showed that only the L2 group presented a significant
difference (p< .05) in producing these two types of RC sentences, whereas Chinese
native speakers (NS) did not show this tendency. This finding shows a presence of RC
type effect in the L2 learners’ production.
Table 4-2 P-values for the within-type Differences between the Subjects’ Performance on the SM Task
Types p-value
SRCs .00**
ORCs .00**
Note: ‘**’ means the p-value is smaller than .01.
In addition, observing the results within the two types of RCs by the subjects as
Table 4-2 shows, we found that the L2 group significantly performed differently from
the NS group, where their mean scores were lower than those of the NS group on both
SRCs and ORCs in the SM task. These differences in SRCs and ORCs were significant
(SRCs: p< .05; ORCs: p< .05), demonstrating the L2 learners had not acquired native
competence to produce both SRCs and ORCs well.
Likewise, a similar ORC preference was also found in the AJ task, as seen in
Table 4-3.
Table 4-3 Subjects’ Mean Scores on SRCs and ORCs in the AJ Task Types
Note: The mean range is from 1 to 4, and ‘**’ means the p-value is smaller than .01.
As illustrated in Table 4-3, the mean scores of the L2 group on SRCs and ORCs were
2.67 and 3.23, respectively. Thus, it is evident that ORCs were more acceptable than
SRCs for the L2 learners, indicating that ORCs were easier than SRCs. A significant
difference (p< .05) was also found in their acceptability of these two different RC types,
as shown in Table 4-4:
Table 4-4 P-values for the within-type Differences between the Subjects’ Performance on the AJ Task
Types p-value
SRCs .00**
ORCs .01*
Note: ‘*’ means the p-value is smaller than .05 and ‘**’ means the p-value is smaller than .01.
In comparison of the L2 group, the Chinese native speakers did not reveal any
significant difference between SRCs and ORCs. Within these two RC types, the L2
group performed significantly differently from the NS group (SRCs: p< .05; ORCs:
p< .05), also indicating that the L2 learners’ acceptability of both SRCs and ORCs was
not native-like.
In the present study, the L2 group showed an ORC preference in production, which
is similar to the findings of Wu & Sheng (2014), that their subjects’ accuracy rate of
producing ORCs was higher than that of producing SRCs. As discussed in Chapter Two,
the CWO predicts ORCs are easier than SRCs since ORCs exhibit similar canonical
order in Chinese. Thus, this prediction is supported by our subjects’ ORC preference.
Moreover, according to Liu (2016) and Xu (2009, 2014), SVO is the canonical word
order in Chinese, which is claimed to be unmarked by Huang (1982). Following Ellis’s
claim (1985) that L2 learners tend to firstly learn unmarked or less marked rules of
target languages and then the marked ones, we may find that the structures have more
canonical orders are able to be learned earlier for developing a core grammar. In this
case, it is not surprising that our L2 learners perform on the canonical order (ORCs)
better than the non-canonical one (SRCs), as Xu (2009, 2014) also claimed Chinese
ORCs should be more canonical than SRCs. Thus, the L2 learners’ ORC production
preference in the current study can be explained by the CWO. The findings of the AJ
task also revealed the same ORC preference, as predicted by the CWO. The findings of
the SM task and the AJ task patterned alike, indicating that there was no task effect in
our study.
In summary, to address the first research question, this section provides results
and discussions on the L2 learners’ performance on SRCs and ORCs. The overall
results of the L2 learners’ production and acceptability revealed that there were
significant differences between their performance on SRCs and ORCs. The acquisition
sequence can be: ORCs > SRCs.
4.2 Gap Influence
In order to address the second question in the current study, we further explored
the subjects’ performance on structures with or without gaps, as presented in Table 4-5.
Table 4-5 Subjects’ Overall Performance on SRCs and ORCs with Gapped and Gapless
Note: The mean score is the accuracy of sentences. ‘*’ means the p-value is smaller than .05 and the sign
‘a’ in the table means that the correlation cannot be computed since the standard error of the difference is 0.
Table 4-5 firstly shows that in the RCs with gaps, SRCs scored lower than ORCs
(M=.65 vs. M=.73). But the difference between these two RC types with gaps was not
significant (p> .05). However, in the RCs without gaps, the mean score on ORCs was
significantly higher than that of SRCs (p < .05). That is to say, the L2 learners produced
ORCs better than SRCs when the structure is gapless. Thus, it can be seen that the
gapless condition did influence the L2 learners’ production of RC types.
However, regarding the NS group, there were no significant differences between
SRCs and ORCs in both gapped and gapless structures since they performed exactly
the same, showing that gaps did not influence the native speakers’ production.
Likewise, we examined whether gapping in RCs would affect the L2 learners’
acceptability on SRCs and ORCs, as can be shown in Table 4-6:
Table 4-6 Subjects’ Mean Scores on SRCs and ORCs with Gapped and Gapless
As Table 4-6 shows in the case of gapped RCs, SRCs and ORCs scored 2.95 and 3.13,
respectively, and there was no significant difference (p> .05) between these two RC
types. However, ORCs (M=3.32) significantly (p< .05) scored higher than SRCs
(M=2.38) in the case of gapless RCs. These findings showed that the L2 learners’
acceptability of SRCs and ORCs varied when gaps were absent in RCs, indicating that
the gapless condition did influence the subjects’ acceptability as it did in the SM task.
As for the NS group, the differences between SRCs and ORCs were not significant
in neither gapped or nor gapless RCs (gapped: p> .05; gapless: p> .05).
The SDT, as mentioned in Chapter Two, predicts that SRCs are easier than ORCs
to process since ORCs exhibit more complexity with more XP categories between the
antecedent and the gap. However, the gaps in SRCs are located at a higher syntactic
position which can be reached earlier; hence it is easier to figure out the filler-gap
relationship in SRCs. This predicts that the L2 learners should produce and accept SRCs
more than ORCs. However, according to the CWO, ORCs exhibit Chinese canonical
orders; thus, ORCs are predicted to be produced and accepted more than SRCs in the
gapped condition. The findings we obtained showed that neither SRCs nor ORCs were
gapped condition. The findings we obtained showed that neither SRCs nor ORCs were