A corpus-based analysis of Chinese

relative clauses produced by Japanese

and Thai learners
Yike Yang (The Hong Kong Polytechnic University)
[email protected]

PACLIC 2020 - The 34th Pacific Asia Conference on Language, Information and Computation, Hanoi
Outline of the presentation

▪ Introduction ▪ Discussion
▪ Relative constructions ▪ The productivity of RCs
▪ Subject-object asymmetry ▪ The SU RC preference
▪ Acquisition of Chinese RCs ▪ The Subject-Object Hierarchy
▪ No cross-linguistic influence
▪ The current study
▪ Research questions ▪ Conclusions
▪ Research methods
▪ Acknowledgements
▪ Results
▪ Selected references
▪ An overview
▪ Subject-object asymmetry
▪ The Subject-Object Hierarchy
2
Introduction: Relative constructions

▪ Relative construction: a nominal (head) + a subordinate

clause (relative clause, RC)
▪ The applei [that you ate ti] was green. (O’Grady, 2011: 19)
▪ Typology of RCs:
▪ Pre-nominal: RC-Head
▪ Post-nominal: Head-RC
Basic word order Order of RC and head Number Example
Verb-object RC-head 5 Mandarin
Verb-object Head-RC 416 Thai
Object-verb RC-head 132 Japanese
Object-verb Head-RC 113 Persian
Languages that do not fall into the four types 213 Kutenai
Table 1: Typology of word order and order of RC and
head (adapted from Dryer, 2013) 3
 Introduction: Relative constructions

▪Chinese: ▪Japanese: ▪Thai:

SU: [ti mai shu de] nanhaii SU: [ti watashi ni hon o kure-ta] hitoi SU: phéti [thîi ti mii khâa mahǎasǎan]

buy book REL boy me DAT book ACC give-PAST person diamond REL have value tremendous

‘the boy who bought a book’ ‘the person who gave me a book’ ‘the diamond that has tremendous value’

DO: [ta mai ti de] shui DO: [watashi ga kinoo ti at-ta] hitoi DO: dèki [thîi chăn líaŋ ti maa]

he buy REL book I NOM yesterday meet-PAST person child REL I bring up come

‘the book that he bought’ ‘the person I met yesterday’ ‘the child that I brought up’

(Yabuki-Soh, 2007: 228) (Sornhiran, 1978: 177)

SU RCs have a longer gap-filler dependency in Chinese and Japanese while DO RCs have a
longer gap-filler dependency in Thai. → Crosslinguistic influence?
4
Introduction: Subject-object asymmetry

▪ The Noun Phrase Accessibility Hierarchy (NPAH)

▪ A universal hierarchy of ease of relativisation: Subject > Direct
Object > Indirect Object > Oblique > Genitive > Object of
comparative (Keenan & Comrie, 1977)
▪ Consistent results from post-nominal Indo-European
languages in both L1 (Diessel & Tomasello, 2001) and L2
(Izumi, 2003) contexts.
▪ Mixed results from pre-nominal languages: some favour the
SU type (Lee, 1992), whereas others support the DO type
(Chen & Shirai, 2015).

5
Introduction: Subject-object asymmetry

▪ The Subject-Object Hierarchy (SOH):

▪ OS > OO/SS > SO (Hamilton, 1994)
Type of RC Illustration No. of discontinuity
[TP ti xuyao gongzuo] de nanreni bangzhu-le ta. 1
SS need job REL man help-PST her
‘The man who needed a job helped her.’
[TP mama [VP yang ti]] de tuzii chi-le huluobo. 2
SO mom feed REL rabbit eat-PST carrot
‘The rabbit that mom feeds ate the carrot.’
wo kanjian [DP [TP ti mai shu] de nanhaii]. 2
OS I see buy book REL boy
‘I saw the boy who bought a book.’
ta mai-le [DP [TP nver [VP xiangyao ti]] de huai]. 3
OO he buy-PST daughter want REL flower
‘He bought the flower that his daughter wanted to have.’
Table 2: Illustration of discontinuity in Chinese RC according to SOH
6
Introduction: Subject-object asymmetry
▪ The Subject-Object Hierarchy (SOH):
▪ OS > OO/SS > SO (Hamilton, 1994)

Languages Updated SOH

Chinese SS (1) > SO/OS (2) > OO (3)
Japanese SS (1) > SO/OS (2) > OO (3)
Thai OS (1) > OO/SS (2) > SO (3)

Table 3: An updated SOH

7
Introduction: Acquisition of Chinese RCs

▪ Conflicting results concerning subject-object asymmetry:

▪ A SU RC preference from L1 child comprehension (Hu et al., 2016),
L1 adult comprehension (Lin & Bever, 2006), L2 comprehension (Li
et al., 2016) and L2 production (Xu, 2014).

▪ A DO RC preference from L1 child production (Chen & Shirai,

2015), L1 adult comprehension (Chen et al., 2008) and L2
comprehension (Packard, 2008).

▪ A consistent preference for SU RCs found in corpus-based

studies of native speakers (e.g. 73.8% vs 26.2% in Pu (2007)
and 60.8% vs 39.2% in Wu et al. (2011)).
8
Introduction: Acquisition of Chinese RCs

▪ Mixed results for the hierarchy generated from the SOH:

▪ Cheng (1995): SS > SO/OS > OO.
▪ Wu et al. (2011) and Lee (1992) : SS > OS > SO > OO.
▪ Pu (2007): SS > OS > OO > SO.
▪ Chang (1984) : SS/SO > OO > OS.

9
The current study: Research questions

▪ Is there subject-object asymmetry in the production of RCs

by Japanese and Thai learners with different levels of
proficiency?

▪ Is the updated version of the SOH consistent with the

written data of Japanese and Thai learners?

10
The current study: Research methods

▪ A corpus-based approach was adopted because of the

correspondence between ease of processing and frequency
of occurrence (Hawkins, 2004; Wu et al., 2011).
▪ The corpus: The HSK Dynamic Composition Corpus
▪ Version 2.0: http://hsk.blcu.edu.cn/
▪ Developed by the Beijing Language and Culture University.
▪ More than 11,500 essays written by L2 learners of various
language backgrounds.
▪ A variety of essay topics.

11
The current study: Research methods

▪ Data extraction:
▪ Learners at the upper intermediate (henceforth, intermediate) and
advanced levels.
▪ Exhausted all the essays of advanced Japanese and Thai learners
and coded the typical SU and DO RCs.
▪ Matched the number of typical RCs produced by intermediate and
advanced learners.
▪ Considered the topic of the selected essays when choosing the
essays.
▪ In total, 80 essays were collected.

12
The current study: Research methods

▪ Data processing and analysis :

▪ The role of the head in the main clause, the role of the gap in the
RC, the type of the verb in the RC and the animacy of the head
noun were manually coded.
▪ For frequency of occurrence of RCs, we adopted the chi-square
test and the binomial test.
▪ The Fisher’s exact test was used when the sample size was
insufficient for a chi-square test.

13
Results: An overview Error-free
RCs

Group No. of essays No. of sentences No. of typical RCs

JP_IN 28 430 48 (11.16%)
JP_AD 19 270 47 (17.41%)
Production
JP_Sum 47 700 did not differ 95 (13.57%)
TH_IN 21 266 by language 26 (9.77%)
background
TH_AD 11 174 25 (14.37%)
TH_Sum 33 440 51 (11.59%)

Table 4: An overview of the corpus data

Note: JP_IN = intermediate Japanese learner group; JP_AD = advanced
Japanese learner group; JP_Sum = all Japanese learners as a group; TH_IN =
intermediate Thai learner group; TH_AD = advanced Japanese learner
group; TH_Sum = all Thai learners as a group.
14
Results: Subject-object asymmetry

▪ A clear subject preference was found in both Japanese and

Thai learners at both proficiency levels.
100%

80%

60%

40%

20%
No group
0%
JP_IN JP_AD TH_IN TH_AD
differences!
DO 15 12 6 7
SU 33 35 20 18
SU DO
15
 Results: The Subject-Object Hierarchy
35
32
30
Group Hierarchy
25 JP_IN OS > SS > OO > SO
21
20 18 JP_AD OS > SS > SO > OO
15 14 13 JP_Sum OS > SS > SO > OO
12 11 12
9 10 10
10
7 7
TH_IN SS > OO/OS > SO
6 6 6 5
5 4 3 3 2 4 3 4 TH_AD OS > SS/OO > SO
0 TH_Sum OS > SS > OO > SO
JP_IN JP_AD JP_Sum TH_IN TH_AD TH_Sum
OO OS SO SS
None followed the
predictions from the SOH
16
Results: The Subject-Object Hierarchy

▪ No significant association between language background

and type of RC (p = .893).
▪ No association between proficiency level and the RC type (p
= .937 for the Japanese groups; p = .365 for the Thai
groups).
▪ In general, the hierarchies of the language and proficiency
groups in our data did not distinguish from each other.

17
Discussion: The productivity of RCs

▪ The percentage of RCs in our data was higher than those

reported in native data.
▪ The productivity of RCs may be explained by the modality of
the data.
▪ Also, the learners might have been striving for higher scores
by writing sentences with diverse and complex structures.

18
Discussion: The SU RC preference

▪ Although SU RCs have a longer filler-gap dependency, SU

RCs occur much more frequently than DO RCs in L1 and L2
Chinese speakers.
▪ Explanation from ‘markedness’ (Givón,1993: 178):
▪ A null subject is much more likely to occur than a null object in
Chinese (Pu, 1997; Xu, 1986), due to the subject usually being the
topic of a sentence (Tsao, 1990).
▪ If the null subject is unmarked, it is reasonable to assume that SU
RC is also unmarked.

19
Discussion: The Subject-Object Hierarchy

▪ Our results did not exactly follow the predicted hierarchies.

▪ There is a tendency of development from the intermediate
groups to the advanced groups that fits approximately with
the predicted hierarchies.
▪ However, this pattern is surprising, because the data of
advanced learners showed more similarities to the learners’
native languages than the data of intermediate learners.
▪ Further studies are needed to provide satisfactory
explanations to this phenomenon.

20
Discussion: No cross-linguistic influence

▪ No statistical difference was found between the Japanese

and Thai learners.
▪ Chinese RCs are typologically scarce, and they diverge from
Japanese and Thai RCs.
▪ Consequently, the learners may treat RCs in Chinese as a
new structure that is not equivalent to the RCs in their L1s
(Comrie, 2007).

21
Conclusions

▪ Both Japanese and Thai learners produced native-like

Chinese RCs at the intermediate level.
▪ Our findings partially supported the hypotheses of the NPAH
and the SOH.
▪ No obvious cross-linguistic influence was found in our data.
▪ More data are needed to the examine the applicability of
the NPAH and SOH hypotheses in L2 Chinese and in
general.

22
Acknowledgements

▪ The audience at the Pacific Second Language

Research Forum 2016 in Chuo University, Tokyo,
Japan.
▪ We also acknowledge the useful comments from the
PACLIC reviewers.

23
Selected references
Chen, Jidong, and Yasuhiro Shirai, ‘The Second Language’, Language Learning, 44 (1994), Chinese Discourse’, Discourse Processes, 43
Acquisition of Relative Clauses in Spontaneous 123–57 (2007), 25–53
Child Speech in Mandarin Chinese’, Journal of Keenan, Edward L., and Bernard Comrie, ‘Noun Sornhiran, Pasinee, ‘A Transformational Study of
Child Language, 42 (2015), 394–422 Phrase Accessibility and Universal Grammar’, Relative Clauses in Thai’ (University of Texas at
Comrie, Bernard, ‘The Acquisition of Relative Linguistic Inquiry, 8 (1977), 63–99 Austin, 1978)
Clauses in Relation to Language Typology’, Lin, Chien-Jer Charles, ‘Subject Prominence and Tsao, Feng-Fu, Sentence and Clause Structure in
Studies in Second Language Acquisition, 29 Processing Dependencies in Prenominal Relative Chinese: A Functional Perspective (Taipei: Student
(2007), 301–9 Clauses: The Comprehension of Possessive Book, 1990)
Diessel, Holger, and Michael Tomasello, ‘The Relative Clauses and Adjunct Relative Clauses in Wu, Fuyun, Elsi Kaiser, and Elaine Andersen,
Development of Relative Clauses in Spontaneous Mandarin Chinese’, Language, 94 (2018), 758–97 ‘Subject Preference, Head Animacy and Lexical
Child Speech’, Cognitive Linguistics, 11 (2001), O’Grady, William, ‘Relative Clause: Processing Cues: A Corpus Study of Relative Clauses in
131–51 and Acquisition’, in The Acquisition of Relative Chinese’, in Processing and Producing Head-Final
Dryer, Matthew S, ‘Relationship between the Clause: Processing, Typology and Function, ed. by Structures, ed. by Hiroko Yamashita, Yuki Hirose,
Order of Object and Verb and the Order of Evan Kidd (Amsterdam: John Benjamins and Jerome Packard (Dordrecht: Springer
Relative Clause and Noun’, in World Atlas of Publishing Company, 2011), pp. 13–38 Netherlands, 2011), pp. 173–93
Language Structures Online, ed. by Matthew S. Packard, Jerome L., ‘Relative Clause Processing in Xu, Yi, ‘Evidence of the Accessibility Hierarchy in
Dryer and Martin Haspelmath (Leipzig: Max L2 Speakers of Mandarin and English’, Journal of Relative Clauses in Chinese as a Second
Planck Institute for Evolutionary Anthropology, the Chinese Language Teachers Association, 43 Language’, Language and Linguistics, 15 (2014),
2013) (2008), 107–46 435–64
Givón, Talmy, English Grammar: A Function-BasedPu, Ming-Ming, ‘Zero Anaphora and Grammatical Yabuki-Soh, Noriko, ‘Teaching Relative Clauses in
Introduction (Amsterdam/Philadelphia: John Relations in Mandarin’, in Grammatical Relations: Japanese: Exploring Alternative Types of
Benjamins Publishing Company, 1993) A Functionalist Perspective, ed. by Talmy Givón Instruction and the Projection Effect’, Studies in
Hamilton, Robert L., ‘Is Implicational (Amsterdam/Philadelphia: John Benjamins Second Language Acquisition, 29 (2007), 219–52
Generalization Unidirectional and Maximal? Publishing Company, 1997), pp. 283–322
Evidence from Relativization Instruction in a ———, ‘The Distribution of Relative Clauses in

24
 THANKS!

Contact: [email protected]

25