+ All Categories
Home > Documents > Assessing Anaphoric Relations via the Phased Choice Methodology

Assessing Anaphoric Relations via the Phased Choice Methodology

Date post: 17-Nov-2023
Category:
Upload: ncu
View: 0 times
Download: 0 times
Share this document with a friend
34
Darcy Sperlich* Assessing anaphoric relations via the phased choice methodology DOI 10.1515/iral-2015-0017 Abstract: There has been a continued interest in discovering how learners of another language acquire a target languages anaphoric system. While our understanding of the issues has improved, the development of methodologies to assess the interpretation of anaphora in a learners interlanguage has not. This paper is concerned with introducing a methodology named Phased Choiceto assess anaphoric reference, which more effectively tackles the issues facing anaphoric testing. The data was gathered from an experiment that focused on English and Korean learners of Chinese acquiring the Chinese reflexive ziji self, results showing that the Korean learners are at an advantage to their English counterparts due to the similar pragmatic strategies Korean and Chinese share. Moreover, the methodology in comparison to the literature shows its strength, dealing effectively with problems past tests faced. Keywords: anaphora, reflexive, binding, Chinese, ziji 1 Introduction Applied linguistic research strives to find better mechanisms to assess the objects of its investigations, engendering better informed theory by increasingly robust data. In second language acquisition (SLA), the object of study is inter- language (IL), which has been used to assess a variety of theories, including Binding theory (Chomsky 1995). While there is a strong tradition of investigating anaphora in SLA, the methods used to assess anaphoric interpretation have not been significantly advanced. Past SLA researchers developed a wide range of methodologies in response to the various inadequacies of the anaphoric tests they faced. Demirci (1997) noted four different test types in the literature; Picture Identification Task (PIT, e.g. Finer 1991); multiple choice (e.g. Thomas 1989); Truth-value Judgement Task (TVJT, e.g. Chen 1996); Acceptability Interpretation Task (named by Demirci 1997 *Corresponding author: Darcy Sperlich, Department of Applied Foreign Languages, National Kaohsiung University of Applied Sciences, 415 Chien Kung Road, Sanmin District, Kaohsiung 80778, Taiwan, E-mail: [email protected] IRAL 2015; 53(4): 355388 Authenticated | [email protected] author's copy Download Date | 11/6/15 4:16 AM
Transcript

Darcy Sperlich*

Assessing anaphoric relations via thephased choice methodology

DOI 10.1515/iral-2015-0017

Abstract: There has been a continued interest in discovering how learners ofanother language acquire a target language’s anaphoric system. While ourunderstanding of the issues has improved, the development of methodologiesto assess the interpretation of anaphora in a learner’s interlanguage has not.This paper is concerned with introducing a methodology named ‘Phased Choice’to assess anaphoric reference, which more effectively tackles the issues facinganaphoric testing. The data was gathered from an experiment that focused onEnglish and Korean learners of Chinese acquiring the Chinese reflexive ziji ‘self’,results showing that the Korean learners are at an advantage to their Englishcounterparts due to the similar pragmatic strategies Korean and Chinese share.Moreover, the methodology in comparison to the literature shows its strength,dealing effectively with problems past tests faced.

Keywords: anaphora, reflexive, binding, Chinese, ziji

1 Introduction

Applied linguistic research strives to find better mechanisms to assess theobjects of its investigations, engendering better informed theory by increasinglyrobust data. In second language acquisition (SLA), the object of study is inter-language (IL), which has been used to assess a variety of theories, includingBinding theory (Chomsky 1995). While there is a strong tradition of investigatinganaphora in SLA, the methods used to assess anaphoric interpretation have notbeen significantly advanced.

Past SLA researchers developed a wide range of methodologies in responseto the various inadequacies of the anaphoric tests they faced. Demirci (1997)noted four different test types in the literature; Picture Identification Task (PIT,e.g. Finer 1991); multiple choice (e.g. Thomas 1989); Truth-value Judgement Task(TVJT, e.g. Chen 1996); Acceptability Interpretation Task (named by Demirci 1997

*Corresponding author: Darcy Sperlich, Department of Applied Foreign Languages, NationalKaohsiung University of Applied Sciences, 415 Chien Kung Road, Sanmin District, Kaohsiung80778, Taiwan, E-mail: [email protected]

IRAL 2015; 53(4): 355–388

Authenticated | [email protected] author's copyDownload Date | 11/6/15 4:16 AM

but based upon Berent and Samar 1990), which has also been called a ‘sentenceinterpretation test’ by Lakshmanan and Teranishi (1994) and ‘acceptabilityjudgement test’ by Sorace (1996).

The key question is whether these tests tap what is fully allowed by alearner’s IL, or simply a preferred interpretation. ‘Preferred interpretation’means rather than tapping what is allowed in a learner’s IL grammar, onlytheir preferred choice is captured (as Thomas 1991a observed; see alsoAkiyama 2002; Demirci 1997; White et al. 1997). Presenting an English learnerwith John thinks that crazy Tom will shoot himself, the learner must be given aclear way to assess each antecedent of himself, to avoid one selection affectingthe other. E.g. a learner may select Tom first of all, and then not consider Johndue to Tom already being selected – note that this is not pertinent to John beingungrammatical, as it may in fact be allowed in the learner’s IL due to L1 transfer.This is particularly pertinent to the Chinese reflexive ziji ‘self’, which can havemultiple antecedents. The problem created is that a learner may only select oneantecedent as perhaps it is more economical to do so, rather than fully con-sidering each possible antecedent (making the sentence ambiguous) – thus notgiving a full picture of the IL. Therefore, it is tantamount that the testingmethodology allows an unbiased assessment of each possible antecedent tofully capture what is allowed in a learner’s IL.

This study investigates how English and Korean learners of Chinese acquirethe Chinese monomorphemic reflexive ziji by assessing their understanding in aseries of neutral and biased sentences. The main focus however will be onassessing the performance of the new ‘Phased Choice’ methodology againstother methodologies assessing anaphoric interpretation. The guiding hypothesesof this study are:1. English learners of Chinese will suffer from negative transfer (of their syntactic

binding rules) in their Chinese, and as a result will underperform in comparisonto the Korean learners of Chinese (behaving more like English native speakers).

2. Korean learners of Chinese will benefit from positive transfer (of theirpragmatic binding rules) in their Chinese, and outperform the Englishlearners of Chinese (behaving more like Chinese native speakers).

2 Theoretical background

The foundation of this research rests upon the assumptions that reflexiveregulation falls under two major areas, syntax and pragmatics. Huang (2000)theorised that languages such as Chinese and Korean are more pragmatic in

356 Darcy Sperlich

Authenticated | [email protected] author's copyDownload Date | 11/6/15 4:16 AM

nature, and languages such as English and German are more syntactic. Thisfollows Givón (1979), who suggested such a pragmatic-syntactic continuumof languages, with Huang applying this to anaphora.

Importantly, the route taken here involves both syntactic and pragmatictheories, and influences the views taken on English, Chinese and Korean reflex-ives. Firstly reflexives can be classed as either polymorphemic (a.k.a. complexreflexive, e.g. himself) and monomorphemic (a.k.a. simplex reflexive, e.g. ziji).Chinese and Korean are alike as both share simplex reflexives, whereas Englishonly has a complex reflexive. The Chinese reflexive studied here is ziji, which ischaracterised by being able to be long-distantly bound (e.g. Huang 2000), as isthe other Korean reflexive studied, casin (e.g. Kim et al. 2009). English isdifferent from the two languages as it only has a polymorphemic reflexivepronoun+self, and is syntactically regulated (Reuland 2011), unless it is placedwithin a logophoric environment (see Kuno 1987; Baker 1995).

There have been many attempts to find a syntactic solution to ziji (e.g.Huang and Tang 1991), which have been refuted both on empirical and theoryinternal grounds (Huang 1994, 2000). Huang (2000) concluded that ziji is prag-matic in nature, accounted by his revised neo-Gricean theory of anaphora. Evensyntactians have abandoned a pure syntactic approach and instead opt for adual syntactic-pragmatic ziji solution (e.g. Huang and Liu 2001).1

Moreover this is supported syntactically as since the advent of Minimalism,Binding theory has been reformulated into interactions between grammaticalprimitives rather than stipulated theoretical modules. Moreover, room has beengiven to pragmatics as in Reuland’s (2011) syntactic theory of anaphora, wheresyntax does not play a part in the regulation of anaphora, discourse becomes thepervading factor. Hence overall, a pragmatic approach to ziji would be sup-ported by both syntacticians and pragmatists alike, which in turn further colla-borates Givón’s (1979) syntactic-pragmatic continuum.

The theoretical viewpoint here is that anaphora can be explained by acombination of syntax and pragmatics. However, previous SLA studies into zijihave been constructed from a mainly syntactic viewpoint suggesting that ziji issyntactically constrained (e.g. Yuan 1998), which is at odds with current theore-tical approaches. Therefore, this forces a rethink of so-called syntactic transfer ofziji, as previous SLA studies on ziji focus on syntax alone without considerationof the pragmatics of ziji.

1 Binding theory has since been reformulated into general grammatical principles underMinimalism, e.g. Heinat (2008), Hicks (2009), Reuland (2011), and Rooryck and Wyngaerd (2011).

Assessing anaphoric relations 357

Authenticated | [email protected] author's copyDownload Date | 11/6/15 4:16 AM

3 Methodological background

Past anaphoric tests used are neither uniform in their stimuli type nor question-ing methodology. For example, a researcher provides a short story stimulus,followed by a truth-value question, contrasting with another researcher whouses multiple choice questions – these two methodologies can lead to differentresults. Accordingly, the different types of stimuli will be discussed first, fol-lowed by the questioning methodologies, focusing on their difficulties in testingreflexive binding.

3.1 Stimuli and question types

The first stimulus discussed is using a picture, where participants are usuallyshown picture(s) of two people doing something to each other. This type ofstimulus is useful as it avoids using a learner’s L2 to understand the story (Finer1991) but does not prevent the participant from creating the linguistic constructin their L1 or L2 via their ‘mentalese’ (Pinker 1994). The second type of stimulusis a short paragraph, usually of two to three people where something happens tosomeone. It is usual that it does not contain any reflexives as this may confuseparticipants, or introduce bias when responding to a question (which has areflexive) about the stimulus. The final stimulus type is one to two sentencescontaining the reflexive under scrutiny.

Turning to the question types, the first discussed is ‘forced choice’, wherethe participant choose between antecedent A or B (e.g. Cook 1990). The secondtype is ‘multiple choice’, which assesses all possible antecedents, as in (1):

(1) Who does ziji refer to?1) Zhangsan2) Lisi3) Both Zhangsan and Lisi

The third question type is termed ‘staggered choice’, meaning that each ante-cedent is assessed individually (e.g. Yip and Tang 1998) in (2) below:

(2) Can ‘himself’ refer to John? Yes NoCan ‘himself’ refer to Tom? Yes No

The final type is labelled ‘negative staggered choice’ (e.g. Lakshmanan andTeranishi 1994), a variation of the previous type, as seen in (3):

358 Darcy Sperlich

Authenticated | [email protected] author's copyDownload Date | 11/6/15 4:16 AM

(3) Himself cannot be John. Agree DisagreeHimself cannot be Tom. Agree Disagree

The combinations of these stimuli and questions lead to the anaphoric tests. ThePIT is a combination of picture and forced choice (Finer and Broselow 1986) orstaggered choice (White et al. 1997). Lakshmanan and Teranishi (1994) criticisedPITs as L2 learners may use non-grammatical strategies and select interpreta-tions based upon preference (also noted by Hua 1994). This was further sup-ported by White et al. (1997), who found that their own PIT results were inferiorto their Truth-value Judgement Test, noting pictures do not provide enoughcontext for divergent interpretations. Moreover, one cannot control what parti-cipants do; if the instructions say to look at the picture first and then thesentence, participants may do the opposite. The implication is that a preferredreading has already been developed, and when the picture shown does notmatch this, the picture is rejected – a problem of preference.2

The second type is combining story or sentence stimuli with multiple,staggered or negative staggered choice, providing a wide range of choices tothe participant. These tests have been called ‘acceptability interpretation task’,‘sentence interpretation’, ‘acceptability judgement test’ and ‘acceptability judge-ment task’, referred here as ‘Interpretive Judgement Test’ (IJT), for they test theinterpretation of a sentence and ask for judgement.

Focusing on tests with multiple choice questions, despite giving thesechoices, the participant may assess all antecedents at once and then choosethe most preferred one. Lakshmanan and Teranishi (1994) criticised Hirakawa’s(1990) use of multiple choice, as it remains unknown which antecedent is notacceptable, thereby not fully reflecting what is allowed in the learner’s ILgrammar and only preference.

The final pairing is with staggered choice questions, a strong question typewhich reduces preference by questioning each antecedent separately. Adopterswere Demirci (1997) and Yip and Tang (1998), first appearing in Berent andSamar (1990). In the words of Yip and Tang (1998: 175), “the task thus ‘walked’the respondents through a variety of options, hence reducing the chance of arespondent’s omitting (or misreading in an unintended way) the consideration ofa possible interpretation”. The other alternative is negative staggered choice(Lakshmanan and Teranishi 1994), which assesses each possible antecedent byasking what cannot be the reflexive’s antecedent. When Lakshmanan andTeranishi (1994) compared their results to previous studies that found preference

2 This problem with the PIT could be avoided by presenting the stimuli and questions about itone after the other, and not together immediately (see Runner and Goldwater 2011).

Assessing anaphoric relations 359

Authenticated | [email protected] author's copyDownload Date | 11/6/15 4:16 AM

for long-distance (LD) binding for zibun (a Japanese LD reflexive), they insteadshowed that preference disappeared, replaced by binding by either antecedent.Demirci (1997) however commented that cognitively it is difficult to process anegative question about a possible antecedent, echoed by Blackwell (2000).

Another test type using the forced choice method assessing truth-value givesthe Truth-value Judgement Test (TVJT, developed by Crain and Mckee 1986). Theadvantage of judging the truth-value is that one can infer what the antecedent ofthe reflexive is without asking the participant outright (White, 1995), engagingimplicit knowledge. However, Demirci (1997) noted various studies using this (e.g.White et al. 1997) that did not overcome preference problems (as MacLaughlin1995 also found). White et al. (1997) themselves showed that the TVJT is flawed toa certain degree, especially using pictures, as preference is likely to be elicited.

Furthermore, it is crucial that the stimulus itself is not ambiguous. Observethis problem in Akiyama (2002: 35) in (4), which gives context for LD bias:

(4) John and Taro were best friends. John fell in love with Taro’s girlfriend.Taro’s girlfriend became interested in John too, and she broke up withTaro. Since then Taro has never tried to see John.

John thought that Taro hated himself. Yes/No

The underlying assumption is that if the participant thinks this statement isfalse, he/she will select ‘No’ (himself does not refer to Taro). If a person selects‘Yes’, it is assumed that the antecedent of himself is John, which entails that it isfalse that himself is anteceded by Taro. However, it is quite plausible that aperson may think ‘Taro hates himself for letting this happen, so himself refers toTaro, so the answer is ‘yes’’, which is different from ‘John is a bad person. Nowonder Taro hates him, I’ll select ‘yes’.’ These two thoughts lead the researcherto mark the ‘yes’ response as selecting the LD antecedent, when in fact the firstthought decides it to be the local antecedent.

A second subtype of the TVJT is using a sentence as the stimulus. Again, ifnot carefully designed this presents a problem in assessing Chinese ziji ‘self’, asseen in (5) taken from Chen (1995: 44):

(5) Wang Xianshen shou Li Xianshen zhidao ziji de taitai hen piaoliang.‘Mr. Wang said that Mr. Li knew that self’s wife was very beautiful.’

Gengju Wang Xianshen de guan dian, Wang Xianshen de taitai henpiaoliang ma?‘According to Mr. Wang, was Mr. Wang’s wife very beautiful?’

360 Darcy Sperlich

Authenticated | [email protected] author's copyDownload Date | 11/6/15 4:16 AM

In this experiment, the participant is given the options of YES/NO/NOT CLEAR. If oneselects YES, this entails LD binding, with NO indicating local binding. The problem isthat no option has been given for possible ambiguous reference as ziji can take bothantecedents. Furthermore, it would be erroneous to assume that selecting YES entailsthe falsity of NO, due to the nature of ziji. Selecting NOT CLEAR itself is unclearbecause it could mean YES/YES, YES/NO, NO/YES, NO/NO, or ‘I don’t know’.

To counteract these problems, Thomas (1991b) suggested training participantsto detect ambiguity. However, training awareness brings unwanted bias into thetest as the participantsmight actively search for ambiguity evenwhen there is none,betrays the test’s purpose, and brings unwanted explicit knowledge to the forefront.

In sum, test designs must deal with Carroll et al.’s (1981: 370) observation,“speaker/hearers are subjectively preoccupied; they are not ordinarily aware ofunaccceptabilities, presuppositions, and ambiguities in their own utterances, orin the speech of others,” for if they do not, data collected will not be valid.

4 Introducing the phased choice methodology

This section presents the ‘Phased Choice’ methodology introduced by this article.To begin, there are two questioning methodologies that work well but operatedifferently, the staggered choice and the truth-value judgement. In (6) the IJTpresents the antecedence question directly, but the TVJT in (7) questions indirectly:

(6) John thinks that Tom doesn’t like himself.Himself = John Y/NHimself = Tom Y/N

(7) John thinks that Tom doesn’t like himself.Tom doesn’t like John. Y/NTom doesn’t like himself (Tom). Y/N

A major point is that the physical order of the questions can affect how they areanswered – selection bias is introduced when both antecedent questions arepresented side by side, and in a fixed questioning order.

In light of this, the proposed new step in this continual testing evolution iscalled the ‘Phased Choice’ questioning methodology.3 Firstly, stepping back and

3 Kim (1993: 199–200) suggested this fledging idea in passing, “In order to determine whetherL2 learners’ responses represent their preference for one interpretation over the other(s) or theirunderlying grammars, a new methodology must be used. That is, we need to elicit L2 learners’judgement of different interpretations of the same sentences by presenting the learner with onecontext and one sentence at a time.”

Assessing anaphoric relations 361

Authenticated | [email protected] author's copyDownload Date | 11/6/15 4:16 AM

considering how these tests were implemented, many were of the traditionalpencil and paper type. The implication is that it was difficult to produce aunique script for every participant, while providing no mechanism to stop theparticipants from returning to change their answers, among others. The solutionto these problems is to turn to computers.

Consider (8) below, the participant is forced to only consider the question athand (without comparing it to another question alongside it):

(8) John thinks that Tom doesn’t like himself.Himself = John Y/N

Subsequently, the same stimulus appears later on inquiring about the otherantecedent, as in (9):

(9) John thinks that Tom doesn’t like himself.Himself = Tom Y/N

The major advantage of this phasing method is that it greatly increases theprobability that the participant will fully appraise all antecedents (avoidingpreference), giving a fuller picture of his/her IL grammar. Moreover, as thispresents a higher likelihood of the reflexive being bound by more than oneantecedent, this avoids a problem noted by MacLaughlin (1995: 4), “linguis-tically-naive subjects… are notoriously bad at reporting multipleinterpretations”.

Finally, it is important that these judgement questions be randomised.That is, for the sentence John said that Tom doesn’t like himself, the order ofquestions might first be Tom then John, or John then Tom (interrupted bydistracters or other test questions). This is advantageous as the participantswill be unable to predict the questioning order. In sum, by adapting the PhasedChoice questioning methodology will bias apparent in other studies be effec-tively reduced.4

4 A reviewer has pointed out the similarity between Phased Choice and the Latin squaredesign. The key points of departure for the Phased Choice is that it allows for a full assessmentof all choices for all participants, whereas the Latin square requires subgroups to be formedamong participants – logistically this is difficult to fulfill as large numbers of SLA participants(usually of differing levels) are usually not the norm (as compared to experiments using nativespeakers). Secondly, Phased Choice allows for comparison of the items in the stimuli tested,whereas the Latin square divides it. Comparing how a participant’s judgements (in this articlefor example) on their anaphoric choices is a very useful mechanism to have as it provides an indepth view on the individual’s considerations in a complicated syntax-pragmatic interfacingenvironment, which would not be available if judgements were spilt.

362 Darcy Sperlich

Authenticated | [email protected] author's copyDownload Date | 11/6/15 4:16 AM

5 Methodology

This section introduces the full methodology used by this study, whichbuilds upon the past literature and the adoption of the Phased Choicemethodology.

5.1 Measuring implicit and explicit knowledge

It is difficult to know whether a participant will use implicit or explicit knowl-edge, thus causing a conundrum for an experimenter who wishes for implicitknowledge use only. Implicit knowledge is unconscious, whereby the partici-pant is unable to vocalise judgement knowledge. Conversely, explicit knowledgeis conscious, a learner is able to verbalise a grammatical rule/relevant explana-tion to support his/her judgement.

Implicit knowledge elicitation is important as Ellis (2009: 13) stated that“default L2 production relies on implicit knowledge”, and this study utilisedEllis (2009a) for guidance in drawing upon implicit knowledge. The first relatesto the learner’s attention – explicit knowledge may be used if the task is focusedupon form and its explanation, however the task here encouraged implicitjudgements by focusing on meaning by judging antecedence. The chances ofexplicit knowledge use here was low, as Jiang (2009: 470) observed, “it is quiterare that learners are explicitly told or taught about the locality constraintson the binding of English reflexives”, noted earlier by Akiyama (2002).Additionally, several Chinese for Speakers of Other Languages lecturers werequeried whether they included explicit teachings of reflexives, and the answerwas ‘no’.

The second issue concerns awareness of the task. The participant was givenminimal information about the test and distracter items were used to keep theparticipant in the dark about the test’s goals.

The third factor regards instructions, the participant was encouraged to use‘feel’ when answering to facilitate implicit knowledge use. However, whileGordon and Hendrick (1997) found that giving implicit instructions had noobservable effect (also found by Cowart, 1997), there is no evidence in doingso causes harm.

The final factor is timing. Loewen (2009) suggested that a time limit helpsprevent L2 learners accessing explicit knowledge, for if they have time to musethey could access explicit knowledge (Ellis 2004). The approach taken here isthat the participants were encouraged to work as quickly as possible, aiming for

Assessing anaphoric relations 363

Authenticated | [email protected] author's copyDownload Date | 11/6/15 4:16 AM

their first impression.5 Time pressure was given by ‘suggesting’ a maximum timefor the whole test, which encouraged a quicker pace (45 minutes for the learnersand 20 minutes for the native speakers, as guided by my pilot study).6

5.2 Test sentence construct

This section introduces the sentences used. Tested were biclausal and triclausalsentences, which were neutral or biased to certain referents. The reflexive occursafter all antecedents, termed forward reflexivisation. The embedded clauseswere either finite or nonfinite (but only finite in triclausal sentences). Thisdichotomy applies only to English; even though the label is used to describeChinese and Korean sentences (for ease of comparison to English), this does notimply that the distinction is made in these languages.7

In total there were 11 different sentence types tested with 2 representativetokens each, for a total of 22 target items assessed. Distracter items wereincluded, an equal amount to the tested sentences (22), totalling 44 sentences.The target types were named as, N-F/NF (Biclausal-No Pragmatic and SemanticBias-Finite/nonfinite), P-F/NF, (Biclausal-Pragmatic Bias-Matrix Subject-Finite/nonfinite) PSB-F/NF, (Biclausal-Pragmatic and Semantic Bias-Matrix Subject-Finite/nonfinite) T-N (Triclausal No Pragmatic and Semantic Bias-Finite), P-LD1(Triclasual-Pragmatic Bias-Matrix Subject-Finite) PS-LD1 (Triclasual-Pragmaticand Semantic Bias-Matrix Subject-Finite), P-LD2 (Triclasual-Pragmatic Bias-

5 Davies and Kaplan (1998) made the point that while a timed test may elicit more ‘feel’answers, it also may increase the amount of guessing due to the short time given to makejudgements. Moreover, Purapura (2004) was not in favour of timed tests, as he argued that test-takers could be anxious which may affect their judgements.6 The pilot study was conducted to assess the effectiveness of the test sentences used, and toset a time limit. This information was used to modify the final version of the test sentences andtiming limit.7 If Chinese exhibits finiteness has been debated in the literature, with linguists (e.g. Huang1982; Tang 2000) striving to find the finite/nonfinite distinction in Chinese. Nevertheless, Hu etal. (2001) provided a strong argument concluding that one cannot truly distinguish between thetwo clause types. Similarly regarding Korean, Lee (2009) found problems with previous ana-lyses on finiteness in Korean, arguing for a role of Mood and Modality in its licensing. Hence,the position taken here for Korean is as the position taken on Chinese. Hence, the finite/nonfinite distinction, made in English does not translate across to Chinese or Korean. Areviewer questions the point of making the distinction apparent here given the positiontaken, the reasoning is because as there are other scholars who might take the oppositecamp’s view in allowing for the distinction, this allows an easier assessment of the experimentfrom their point of view.

364 Darcy Sperlich

Authenticated | [email protected] author's copyDownload Date | 11/6/15 4:16 AM

First Embedded Subject-Finite), and PS-LD2 (Triclasual-Pragmatic and SemanticBias-First Embedded Subject-Finite). Below are examples of two sentence types,the Appendix containing all test sentences.

Biclausal, Pragmatic and Semantic Bias to the Matrix Subject, Finite andnonfinite (PS-F/NF). In this sentence type, there are pragmatic and semanticfactors which encourage LD binding. In (10) below it can be reasonably assumedthat a villain would poison the hero so that the villain would not be caught:8

(10) The poisoned hero doesn’t know that the devious villain poisoned himself.

Triclausal, No Semantic or Pragmatic Bias, Finite (T-N). This sentence is‘neutral’, i.e. without any pragmatic or semantic influence on the binding ofhimself in (11):

(11) Tom heard that Dick thinks that Harry had hurt himself.

5.3 Psycholinguistic measures

Using a participant’s judgement as a basis of theorising makes several assump-tions on how the judgement was made. The following discusses using simpletechniques which provide greater explanatory power on what underlies judge-ment data. These are reaction timing, confidence scales, and applying a methodfrom cognitive psychology into the SLA of anaphora for the first time, knowledgesource attribution.

The first technique is reaction timing. This was recorded from when thestimulus first appeared, to when the participant made a Y/N judgement. Thisoffers further perspective on the thoughts leading to the judgement, e.g. if he/she took three seconds longer on judging the first antecedent compared to thesecond one, this could indicate that sentence bias is playing a role.

The second is the use of confidence scales (cf. Gass 1994). After a Y/Njudgement, confidence was indicated through a Likert scale with verbal

8 For those that would simply dismiss this sentence (in English) as ungrammatical, considerthis: the reason for judging it ungrammatical is that himself is in an argument position, whichshould be locally bound. Indeed, this sentence does not seem right due to the pragmaticcomponent pushing himself to be LD bound, in contrast to syntactic local binding. But if himselfis subject to strict local binding here, nothing should be considered odd and be simply locallybound without further consideration, which is clearly not the case. Therefore, anaphoricresolution includes pragmatic effects, as the interplay between syntax and pragmatics showsin the above sentence (cf. Huang 2000).

Assessing anaphoric relations 365

Authenticated | [email protected] author's copyDownload Date | 11/6/15 4:16 AM

categories below in (12),9 which alerts the investigator to the quality ofjudgement:10

(12) 4 – Certain, you have no doubts that you are right3 – Quite confident2 – More or less guessing1 – Guess (i.e. you could have flipped a coin, 50/50 chance of getting it

right)

The final technique, new to SLA, is concerned with eliciting the belief ofwhat the judgement knowledge is based on.11 The scheme was termed ‘sourceattribution categories’, as used by Dienes and Scott (2005) in (13):

(13) G = Guess – you guessed the answer, just like flipping a coin.I = Intuition – you don’t know why you are right, it’s a ‘feeling’ that you

have.R = Rule – you have learnt the rule before and you are knowingly applying

it to the sentence, and you can say what the rule is.

Using guess and intuition implies an unconscious judgement, while rule sup-poses a conscious judgement.12 They also included ‘memory’, but this would notbe useful as participants were not being asked to memorise strings.

This system is better than relying on verbal reports, as Rebuschat (2008)believed that they are not completely accurate as typically participants areunable to verbalise their awareness, leading to the misconception that implicitknowledge was used. Moreover, Ellis (2004) found that one’s explicit knowledge

9 This scale is based on Dienes (2008). In cognitive psychology, Dienes (2008) compared sixdifferent types of confidence scales, comparing the performance of binary (high/low), binary(guess/sure), numerical 50–100%, numerical 50–100% with ratings explained, numerical cate-gories (e.g. 50–59) and verbal categories (e.g. sure, very sure). He expected the verbal categoriesto be the most sensitive, but there were no difference between the scales. Therefore, the verbalcategories used in this study are sufficient.10 Yule et al. (1985) found that a high accuracy rating may not be necessarily related to highconfidence, as learners may misperceive the state of their own linguistic knowledge.11 The technique is relatively new, see Loewen (2009) for his use of Rule and Intuition to assistmeasurement of explicit and implicit knowledge in SLA. Moreover, Sorace (1988) suggestedusing confidence scales as a measure to distinguish between a participant’s perception of whathe/she knows, compared to what he/she really knows.12 Rebuschat (2008: 103) stated, “the combined use of confidence rating and source attribu-tions appears to be a promising method for assessing awareness”, echoed by Rebuschat andWilliams (2012) recommending its use for SLA.

366 Darcy Sperlich

Authenticated | [email protected] author's copyDownload Date | 11/6/15 4:16 AM

may not be accurately measured by what one can verbalise – the existence ofthe former does not depend on the latter.13

5.4 Randomisation scheme

Three different types of randomisations were used insuring that each participantreceived a unique test script. The first type was intersentential, where a questionabout a possible antecedent randomly appeared. The second type was intrasen-tential, where the stimulus randomly appeared. This was constrained followingCowart’s (1997) suggestion, using block randomisation. The test had two scriptswithin, helping against the possibility of similar test items occurring closelytogether (each block had its own distracter items).14

5.5 Test selection

This study made use of two separate tests, the IJT and the TVJT, using thePhased Choice questioning methodology, along with sentence stimuli. The IJTwas completed first, followed by the TVJT, with on average a few days betweenthem to avoid a learning effect. The tests were conducted on a computer (via theInternet) in a computer lab/private room. The test program itself generated arandom login number, and the participant is given a password to access thetest.15 The participant then selected his/her language (Chinese, Korean orEnglish). Afterwards, he/she entered his/her biographical information. Thisincluded sex, age, country of origin, and previous linguistic study. The learnergroups had extra questions regarding their Chinese. Following this, instructionswere given about the test and an example to try. The test followed.

The test sentences were presented randomly to the participants, e.g. as (14):

(14) John knows that Tom thinks highly of himself.Himself = Tom YES NO

13 Ellis (2005) noted that verbalisations vary in the degree of sophistication, e.g. ‘The poly-morphemic anaphor himself can only be anteceded by a local subject or object, subject tospecial syntactic structures and discourse conditions’, compared to, ‘himself can’t refer to thatperson, but this person is OK’; both are explicit regardless.14 Note that the first sentence of the first script is a distracter item to ease the participants intothe test.15 The test program was a custom made one using Adobe Flash.

Assessing anaphoric relations 367

Authenticated | [email protected] author's copyDownload Date | 11/6/15 4:16 AM

There were three possibilities of what could follow, 1) a distracter item, 2) a testsentence, or 3), the same sentence but it would ask about the second antecedentJohn. Note with 1) and 2) that only one antecedent question would randomlyappear.16 The keyboard was used to enter the judgement (Y for Yes and N forNo). Reaction timing from when the stimulus appeared to when the participantanswered Y/N was collected. Following this, there were two scales to selectfrom, the confidence and source of knowledge scales (accessible after a Y/Njudgement was made). Confidence (15) was selected via the keyboard numberpad, and then knowledge source (16) was chosen using the keys G, I and R:

(15) 4 3 2 1

(16) G I R

The participant then progressed to the next screen after all selections weremade.

Thus the entire test was completed in this manner. To assist participants, atthe bottom of every screen showed the definitions for the confidence and knowl-edge source scales. Moreover, a vocabulary list was attached to each questionfor the learners to counteract any possible unknown vocabulary, as Birdsong(1989: 86) indicated, “learners in judgment elicitation experiments are oftenconfronted with unfamiliar sentence types and lexical items.”

5.6 Participants

This study encompassed eight different groups totalling 100 participants overnative control and learner experimental groups. Participants completed the testsvoluntarily, and in return they entered a draw to win one of three cash prizes.There were four learner groups, split into English and Korean speakers, whowere international students studying at a university in China (grouped accordingto proficiency from their classes). The English group had intermediate (ELCi) andadvanced (ELCa) participants, the intermediate group containing two males(from the USA) and one female (from New Zealand), their ages ranging between17–21. The advanced group had two females (20 and 33), both from the USA. The

16 What are the chances that the exact same stimuli appears with the other antecedent question?Calculating at the beginning of the test, the probability is a low 2.2% (for a biclausal sentenceoccurring). This changes as the test progresses, but will remain at a low level. An alternative is tocreate an algorithm which disallows the above scenario (pseudo-randomisation).

368 Darcy Sperlich

Authenticated | [email protected] author's copyDownload Date | 11/6/15 4:16 AM

Korean learner of Chinese group was split into an intermediate (KLCi) and anadvanced group (KLCa). The intermediate group had two females, aged 20 and21. The advanced group had five males and one female, their ages ranging from22–28. Both groups were South Korean.

The next group were the near-native Chinese speakers (NNCS), havingacquired Chinese to a near-native degree according to their self-report and myknowledge of their language proficiency.17 Three females and four males tookpart, ages ranging from 27–61. Four came from New Zealand, one from Australia,one from the UK and another from China (who did not grow up in China). Theyspanned from 10–43 years speaking Chinese. This group was also the mostlinguistically aware, with 2 PhDs, 1 BA and one person spending 20 years inthe field of linguistics. This is because the community of NNCS (in New Zealand)is very small, therefore my recruitment efforts were concentrated at the tertiarylevel.

There were three native speaker control groups, Chinese native speakers(CNS, speaking northern Mandarin), having 6 male and 19 female participants,whose ages ranged from 17–60; Korean native speakers (KNS, from South Korea)made up of 8 male and 11 female participants, whose ages ranged from 18–55;English native speakers (ENS, from New Zealand) having 16 males and 20females, whose ages ranged from 25–49.

6 Results

The focus is to review the data so to compare the Phased Choice methodology toother methodologies. Firstly an overview of binding patterns per group is pre-sented, with more focus on the native speaker groups. The learner groups on theother hand, while providing interesting patterns of binding that will be com-mented on, would benefit from further verification. Nevertheless, the data aresimilar to that of the literature (e.g., Yuan 1998, Ying 1999, Dugarova 2008), butprovide a more refined view. Due to space constraints, only Y/N answers will becovered – forming the basis of comparing methodologies. This leaves timing,confidence levels and timing unexplored due to given the amount of data there

17 A reviewer noted that self-report is not the best measure to use for this group. I also elicitedtheir language use of Chinese and background in Chinese. Moreover I supplemented thisinformation with my own knowledge of their Chinese ability from their professional involve-ment in Chinese.

Assessing anaphoric relations 369

Authenticated | [email protected] author's copyDownload Date | 11/6/15 4:16 AM

is to review, going beyond the scope of this article. However, further on I presentsome specific examples as to give an idea of this data’s use.

These results are the combination of the IJT and TVJT, seen in Tables 1 and 3,significance in Tables 2 and 4. It should be noted that there was very littledifference between the two surveys in terms of the answers given, timing, con-fidence levels and knowledge source. The highlighted cells in Tables 1 and 3represent the highest proportion of answers between Y/N for each antecedent.

Beginning with the biclausal neutral sentences (Table 1), the ELC groups havea preference for local binding, as do ENS. The KLC groups on the other hand,combining both the N and NF neutral sentences have a rough 50/50 distributionbetween both antecedents. The LD binding in the NF sentence is most likely due toinadvertent bias to the LD subject which was present in one of the tokens, as alsoevidenced by CNS in their LD binding in the NF type (but a rough 50/50 in the Fsentence). Moreover, this is evident in NNCS who LD bind strongly, but still haveheavy L1 influence as seen through their local binding choices. The interesting caselies within CNS, as there is a 62% selection of the local antecedent, and 57% of theLD antecedent. Comparing the YES vs. YES answers, 48% of YES choices go to theLD antecedent, whilst 52% of YES answers are for the local antecedent – aBinomial test shows that the differences are not significant (p=0.767). This isinteresting as ziji has shown a preference to the LD subject (Dugarova 2008) orthe local subject (Hua 1994). In an overview of the experimental literature (Hua1994; Yip and Tang 1998; Yuan 1998; Ying 1999; Dugarova 2008), ziji shows arough 50/50 distribution in neutral sentences, which is clearly seen in the datahere. KNS on the other hand show the LD antecedent is preferred.

Focusing now on the biased sentences, there is little difference between theP and PS types; the addition of semantic bias had little observable effect.Looking at ELCi, the bias generally confused them, as there is no clear patternof local or LD binding. The only consistent pattern is that there is a majority NOanswer in these sentences, showing the difficulty pragmatic bias causes them.ELCa, although with residue local binding, respond well to the bias (includingNNCS). The KLC groups, KNS and CNS predictably follow the bias. ENS providean interesting case because they respond to bias, but in PS-F they do not.18

Looking back at the tokens for these sentence types, himself is occurring inprepositional phrases, not in an argument position; it has been well documentedthat himself can be LD bound in such contexts (e.g. Kuno 1987).19 Hence, in PS-F,

18 Note in Table 2 the groups have no significant differences between them due to theirfollowing of bias.19 Note that these are English-specific effects and do not apply to Chinese and Korean, giventhe LD nature of their reflexives.

370 Darcy Sperlich

Authenticated | [email protected] author's copyDownload Date | 11/6/15 4:16 AM

the amount of LD binding appears minimal in comparison to the other types.Excluding the prepositional phrase tokens and concentrating on the tokens withargument reflexives, binding towards the matrix subject is 45% in P-F (token 2)and 28% in PS-F (both tokens), not an insignificant result.

Turning to the triclausal sentences (Table 3), their patterns closely follow thebiclausal ones. In the neutral sentences, the ELC groups and NNCS exhibit localbinding, however the KLC groups have no real preference. KNS show no real

Table 2: Statistical differences between ENS, CNS and KNS per biclausal sentence.

ReferentSentence

LD LC

N-F χ(, N=)=., p < . χ(, N=)=., p < .N-NF χ(, N=)=., p < . χ(, N=)=., p < .P-F χ(, N=)=., p < . χ(, N=)=., p=.P-NF χ(, N=)=., p < . χ(, N=)=., p=.PS-F χ(, N=)=., p < . χ(, N=)=., p < .PS-NF χ(, N=)=., p=. χ(, N=)=., p=.

20 ‘※=prediction reversed’ means the opposite holds true. For example, in the LD binding forN-F sentence note that ‘N-F’ is bolded, meaning that it is predicted that groups will bind here.But ENS has the symbol ※, which means that this prediction does not hold for them.

Table 1: Biclausal % proportion Y results for all groups.20

GroupSentence

ELCi ELCa KLCi KLCa NNCS ENS CNS KNS

LD N-F ※ ※

N-NF ※ ※

P-F ※ ※

P-NF ※ ※

PS-F ※ ※

PS-NF ※ ※

LC N-F

N-NF

P-F ※ ※

P-NF ※ ※

PS-F ※ ※

PS-NF ※ ※

Note: LD= Long-distant antecedent LC= local antecedent ※=prediction reversedBold=binding expected Shaded cell=unpredicted binding.

Assessing anaphoric relations 371

Authenticated | [email protected] author's copyDownload Date | 11/6/15 4:16 AM

binding preference, but CNS are inclined to the matrix subject. Moving to thebiased sentences, only ELCi do not follow the bias. Finally, ENS respond to bias,when in fact they should not (but notice the parallel amount of local binding).Unlike the biclausal sentences, all the triclausal sentences contain himselfoccurring in an argument position.

Table 3: Triclausal % proportion Y results for all groups.

GroupSentence

ELCi ELCa KLCi KLCa NNCS ENS CNS KNS

LD N ※ ※

P-LD ※ ※

P-LD

PS-LD ※ ※

PS-LD

LD N ※ ※

P-LD

P-LD ※ ※

PS-LD

PS-LD ※ ※

LC N

P-LD ※ ※

P-LD ※ ※

PS-LD ※ ※

PS-LD ※ ※

Note: LD= Long-distant antecedent LC= local antecedent ※=prediction reversedBold=binding expected Shaded cell=unpredicted binding.

Table 4: Statistical differences between ENS, CNS and KNS per triclausal sentence.

ReferentSentence

LD LD LC

T-N χ(, N=)=.,p < .

χ(, N=)=.,p < .

χ(, N=)=.,p < .

P-LD χ(, N=)=.,p < .

χ(, N=)=.,p=.

χ(, N=)=.,p < .

P-LD χ(, N=)=.,p < .

χ(, N=)=.,p < .

χ(, N=)=.,p < .

PS-LD χ(, N=)=.,p < .

χ(, N=)=.,p=.

χ(, N=)=.,p < .

PS-LD χ(, N=)=.,p < .

χ(, N=)=.,p < .

χ(, N=)=.,p < .

372 Darcy Sperlich

Authenticated | [email protected] author's copyDownload Date | 11/6/15 4:16 AM

Overall, several patterns emerge. Firstly ELCa are at a clear advantage overELCi; they have largely obtained the rules surrounding ziji, allowing LD binding,whereas ELCi either are stuck in local binding, or have difficulty with LDbinding. It is evident however that ELCa in neutral sentences revert to localbinding, but are performing close to par with the KLC groups (as do NNCS). TheKLC groups are at an advantage (having similar binding rules to Chinese) overtheir ELC counterparts by following bias, and oscillate somewhat between CNSand KNS in the neutral sentences (while the ELC groups bind as ENS does). ENSprove to be an enigma, as they are not immune from pragmatic/semantic bias toan antecedent binding himself in an argument position. Strong differences existbetween CNS and KNS in neutral sentences, which may be due to differencesbetween ziji and casin (the Korean reflexive tested here), but both groups arevery similar in following bias.

6.1. Indeterminacy

The final analysis determines how consistent the participants were with theirresponses. This was investigated by checking if participants’ Y/N answers wereconsistent over both tokens (see Hua 1994 for a similar approach). An alternativeroute involves criterion levels, e.g. 3/4 selections for the LD antecedent showsthe acquisition of a binding rule; but is avoided here. Using criterion levels tojudge if a participant has ‘acquired’ a rule is subjective; how did the researcherarrive at the conclusion that 75% ‘correct’ selections shows the rule’s acquisi-tion, whilst another researcher used an 85% level? What happens if a participantobtains a slightly lower/higher percentage? Lakshmanan and Selinker (2001:402) also offered criticism, “a problem in relation to the use of criterion levels,(especially high criterion levels) is that one cannot conclude that the failure onthe part of learners to reach the criterion level for acquisition is because of lackof knowledge,” (see also Hamilton 1996: 440). The point is that there has beenno research that has discovered a rate for rule acquisition.

A second issue for criterion levels is that the researcher has assumed whatthe ‘correct’ answer to the question is. If the answer is simple (e.g. ‘what is thepast tense of walk, -ed or -en?’), then this is justified. However in this study, thequestions posed are complex due to the nature of reflexives in each language.For example, in Chinese neutral sentences containing two antecedents followedby ziji, either may be the antecedent. A pragmatically biased to the matrixsubject sentence with two antecedents followed by himself in English drivesbinding by the matrix subject, but English native speakers may not choose it due

Assessing anaphoric relations 373

Authenticated | [email protected] author's copyDownload Date | 11/6/15 4:16 AM

to syntactic influence. Therefore, the preference is to give a descriptive view thana prescriptive one in light of how syntax and pragmatics interface.

Moving to the analysis, Table 5 below shows how consistent participants ineach group are in the same tokens over different tests (e.g. IJT Token 1 versusTVJT Token 1), and different tokens in the same tests (e.g. IJT Token 1&2, repeatfor TVJT). It is important to read this table in conjunction with the result Tables 1and 3.

Focusing on ELCi’s responses, they have mixed levels of consistency over allsentence types, but recalling from Tables 1 and 3 above they either had difficultyfinding an antecedent or bound locally. Turning to ELCa, consistency is muchmore apparent given that they often converge on one antecedent. ConsideringKLCi/a, they are more consistent in biased sentences than neutral ones. TheNNCS group on the other hand still suffer from L1 influence, and it can be seenthat there is no clear division between biased and neutral sentence consistency.What was said of KLC can be said of CNS and KNS as well. ENS producereasonably consistent results, the highest being in the neutral sentences.Notice that this drops in the biased sentences, when two different bindingstrategies present themselves.21 If CNS/KNS are compared to ENS in the neutralsentences it is very clear that ENS are binding syntactically with high consis-tency – CNS/KNS on the other hand have noticeably lower levels of consistencyin the neutral sentences which reflects their pragmatic binding, or noticeably alack of rule driven local binding seen in ENS.22

Looking at the overall results (Table 6), percentages from these two methodsdiffer slightly from one another, inconsistent Y/N judgements ranging roughlybetween 15% to 30%. What is the cause of this indeterminacy? A possibleanswer is the construct of anaphoric regulation – the rules are not set andexceptions are bound to occur, considering the various linguistic interfacesinvolved. Furthermore, these are not grammaticality judgements, rather it isabout possible antecedents, which creates more room to manoeuvre. Hence,the limited variability of these judgements is not unusual considering the abovefactors, and has been found in the literature as well (e.g. Hua 1994).

Finally, as mentioned previously some data are shared in the areas ofconfidence, timing and knowledge source as to see how this correlates withthe anaphoric judgements. The example picked is comparing how ELCaanswered a biased sentence compared to KLCa. Recall in the biclausal P-F

21 Hua (1994) reported the percentage of inconsistency in ENS in neutral finite and nonfinitesentences at 4%, while CNS at 28% and 17%, respectively.22 I would like to thank the reviewer whose comment made me re-look at the data, resulting inthis observation.

374 Darcy Sperlich

Authenticated | [email protected] author's copyDownload Date | 11/6/15 4:16 AM

Table5:

Con

sisten

cyof

Y/Njudg

emen

tspe

rsenten

ce.

Sen

tenc

eGroup

N-F

N-NF

P-F

P-NF

PS-F

PS-NF

T-N

P-LD

P-LD

PS

-LD

PS-LD

ELCi

DTo

k,STe

st

.

.

.

.

.

.

STo

k,DTe

st.

.

.

.

.

.

ELCa

DTo

k,STe

st.

.

.

.

.

.

.

STo

k,DTe

st

.

KLC

iDTo

k,STe

st

.

.

.

.

.

STo

k,DTe

st

.

.

.

.

KLC

aDTo

k,STe

st

.

.

.

.

.

.

.

.

.

STo

k,DTe

st

.

.

.

NNCS

DTo

k,STe

st.

.

.

.

.

.

.

.

.

.

.

STo

k,DTe

st.

.

.

.

.

.

.

.

.

.

.

ENS

DTo

k,STe

st.

.

.

.

.

.

.

.

.

STo

k,DTe

st.

.

.

.

.

.

.

.

.

.

.

CNS

DTo

k,STe

st

.

.

.

.

STo

k,DTe

st.

.

.

.

.

.

.

.

.

.

.

KNS

DTo

k,STe

st.

.

.

.

.

.

.

.

STo

k,DTe

st.

.

.

.

.

.

.

.

Note:

DTo

k,STest=Different

Toke

n,Sam

eTest

STo

k,DTest=Sam

eTo

ken,

Different

Test.

Assessing anaphoric relations 375

Authenticated | [email protected] author's copyDownload Date | 11/6/15 4:16 AM

sentence both groups have overwhelmingly responded to the anaphoric bias,which suggests both groups are on par. However, the psycholinguistic datapaints another picture, as ELCa are overall slower (25.7 seconds versus 13.7seconds), have lower confidence (2.67 versus 3.45) and show a higher degreeof guessing (17% versus 5%). Hence processing wise, the two groups are cer-tainly not on par with one another even though on the surface they might be.23

Therefore, combining these measurements with the methodology presented hereprovide a highly valid data set.

6.2 Phased choice versus other methodologies

Tables 7–10 below make comparisons between data elicited in this study andother studies on the acquisition of reflexives. However, there are several pointsto note. Firstly, there is currently no comparable data for Korean learners ofChinese, this being the first study of its kind to include them, hence theirexclusion here. Secondly, the studies included here are not uniform in theirmethodology, therefore these cannot be considered as uniform results. Finally,to facilitate comparison, the results of the biased biclausal sentences (P and PS)have been combined for comparison to what are biased sentences reported inthe literature.24

In Table 7, in the neutral sentences ELC prefer local binding, which thecurrent study is in line with. However, the current study shows the LD bindingalternative being minimally selected, compared to other studies which haverelatively high numbers of LD selections (local/LD being added to the LD

Table 6: Overall consistency of Y/N judgements.

GroupType

ELCi ELCa KLCi KLCa NNCS ENS CNS KNS

DTok, STest .% .% .% .% .% .% .% .%STok, DTest .% .% .% .% .% .% .% .%

Note: DTok, STest=Different Token, Same Test STok, DTest=Same Token, Different Test.

23 There is limited data on triclausal sentences, hence reference to them will be made infootnotes.24 As this goes beyond the scope of the current article, this data shall be focus of futurepublication.

376 Darcy Sperlich

Authenticated | [email protected] author's copyDownload Date | 11/6/15 4:16 AM

category). This shows the Phased Choice methodology did not capture this trendfound in the literature, due to how the questions were presented. Turning to thebiased sentences, this study’s ELC INT are much less likely to LD bind althoughthere is strong bias, compared to the other studies. The ADV group on the otherhand respond strongly to bias, which is in line with what has been reported.However, the Phased Choice methodology is eliciting a stronger LD bindingresponse, with lower local binding responses. In sum, the data here is compa-tible with what has been reported in the literature, showing a finer grainedpicture of binding choices.

In Table 8, focusing first on the neutral sentences, this study’s B-N-F elicits arough 50/50 distribution, but in the NF sentence type it is weighed to an LDinterpretation.25 Compared to the rest of the literature, it is difficult to find apattern as the binding oscillates between local and LD antecedents or local/LD.Combining these results reveals a rough 50/50 distribution over the two ante-

Table 7: English learners of Chinese binding patterns (%).

Study Yuan () Ying () Dugarova () Current Study

Methodology → Multiple

choice

sentence IJT

Negative

staggered choice

sentence IJT

Multiple

choice

sentence IJT

Phased Choice

sentence

IJT & TVJT

Sentence Type Level INT ADV INT ADV BEG INT ADV INT ADV

B-N-F Local

-b

-a

b

LD ‒Local/LD

B-N-NF Local

LD

Local/LD

B-BLD-F Local

LD

Local/LD

B-BLD-NF Local

LD

Local/LD

Note: a – Low percentages under 3% are recorded as ‘-’; b – Ying referred to the sentence typeas neutral only.

25 One token out of two for this particular sentence type had inadvertent semantic bias to the LDNP.

Assessing anaphoric relations 377

Authenticated | [email protected] author's copyDownload Date | 11/6/15 4:16 AM

cedents, captured by the Phased Choice methodology. Moving onto the biasedsentences, throughout the LD antecedent is the preferred binder. However, thereis still local and local/LD binding. In contrast, the current study records a lowerlevel of local binding, and a very strong LD binding preference, due to themethodology used here (Table 9).

KNS in the neutral sentences are divided between the two antecedents, apartfrom Lee (2008). In fact, this study aligns with Lee’s LD binding. Lee useddiscourse text stimuli and multiple choice, while this study achieved similarstrong results by using single sentences and a new questioning methodology.

In the final Table 10,26 the studies’ neutral sentences are in line with oneanother, showing dominant local binding. The key differences lie with the

Table 8: Chinese native speakers binding patterns (%).

Study

Sentence Type

Hua()

Yip andTang

()a

Yuan()

Ying () Dugarova()

CurrentStudy

Methodology → Multiplechoice

sentenceIJT

Staggeredchoice

sentenceIJT

Multiplechoice

sentenceIJT

Negativestaggered

choicesentence

IJT

Multiplechoice

sentenceIJT

PhasedChoice

sentenceIJT & TVJT

B-N-F Local -

d

LD

Local/LD b

B-N-NF Local

LD

Local/LD c

B-BLD-F Local

LD

Local/LD

B-BLD-NF Local

LD

Local/LD

Note: a – Their CNS are Cantonese speakers; b – They also included a category of ‘LD or local/LD’, which is 92%; c – They also included a category of ‘LD or local/LD’, which is 84%; d – Yingtested neutral sentences only.

26 In triclausal, neutral nonfinite sentences, Hirakawa (1990) found ENS locally bound 98%,and Kim (1993) found ENS locally bound 100%. In triclausal, neutral finite sentences, Hirakawa(1990) found ENS locally bound 98%. These results are consistent with this study’s findings.

378 Darcy Sperlich

Authenticated | [email protected] author's copyDownload Date | 11/6/15 4:16 AM

biased sentences, noting the strong LD binding obtained by this study and Lee(2008), and to a lesser extent Demirci (2000). That ENS are susceptible topragmatic influence appears to be an emerging trend in the experimentalliterature.

Overall, the data collected by the current study is in line with what has beenfound previously, reviewed above. Moreover, the data collected reveals nuanceddifferences which would have been unattainable using other methodologies.28

Thus, it can be seen that the Phased Choice methodology delivers robust datawhich is valid, as compared to other methods.

Table 9: Korean native speakers binding patterns (%).27

Study

Sentence Type

Kim andMontrul()

Lee () Kim andYoon ()

Kim et al.()

CurrentStudy

Methodology → Forcedchoice

picture TVJT

Multiplechoice

discoursetext IJT

Forcedchoice

picture TVJT

Forcedchoice

picture TVJT

PhasedChoice

sentenceIJT & TVJT

B-N-F Local

-a

-a

-a

LD

Local/LD

B-N-NF Local

LD

Local/LD

B-BLD-F Local

LD

Local/LD

B-BLD-NF Local

LD

Local/LD

Note: a – They presented neutral sentences without the F/NF label.

27 In triclausal, neutral nonfinite sentences, Kim (1993) found KNS locally bound (NP3) 23%,LD (NP2) 50%, LD (NP1) 8%, and NP2/3 15%. This contrasts to this study’s own results in havinga similar distribution across all three referents.28 When Lakshmanan and Teranishi (1994) compared their methodology to previous studies,they found LD binding preference for zibun (a Japanese LD reflexive) disappeared, showing a50/50 distribution.

Assessing anaphoric relations 379

Authenticated | [email protected] author's copyDownload Date | 11/6/15 4:16 AM

Table10:En

glishna

tive

spea

kers

bind

ingpa

tterns

(%).

Study

Sen

tenc

eTy

peTh

omas

()

Kim

()

Hua

()

Wells

()

Yipan

dTa

ng()

Dem

irci

()

Lee()

Curren

tStudy

Metho

dology

→Multiple

choice

senten

ceIJT

Multiple

choice

senten

ceIJT

Multiple

choice

senten

ceIJT

Forced

choice

storyTV

JT

Stagg

ered

choice

senten

ceIJT

Stagg

ered

choice

senten

ceIJT

Multiple

choice

discou

rse

text

IJT

Phas

edCh

oice

senten

ceIJT&

TVJT

B-N-F

Local

LD–

––

Local/LD

––

––

B-N-NF

Local

LD–

Local/LD

––

B-BLD

-FLo

cal

LD

Local/LD

––

B-BLD

-NF

Local

LD

Local/LD

380 Darcy Sperlich

Authenticated | [email protected] author's copyDownload Date | 11/6/15 4:16 AM

7 Discussion

Firstly addressing the hypotheses posed at the beginning of the article, it can beclearly seen by comparing the native speaker groups’ binding patterns to thelearners’ that a), the ELC groups suffer from negative transfer of their syntacticstrategies (but are better at the advanced level, as are NNCS), and b) the KLCgroups benefit from positive transfer of pragmatic strategies, which in turn arevery similar to Chinese.

The results of this study demonstrate the success of the current methodologyin testing possible antecedents of a reflexive. Previous tests used are susceptibleto introducing bias into the results, therefore providing a clouded view of aspeaker’s L1/2. The Phased Choice questioning methodology effectively reducesthis bias to near nil, as exemplified by the 50/50 binding of ziji by two ante-cedents in a neutral biclausal sentence by CNS. In the literature, results havebeen partial in either direction, but observed as a whole the distribution closelymirrors to what has been found here in a single study.

Further results exemplify the strength of the data obtained, for example thefollowing of pragmatic bias by ENS is unprecedented, with many studies show-ing local binding is preferred. As mentioned, the results drew close parallels toone other study for ENS, Lee (2008) who used a strong discourse focus within atext, whilst here just single sentences were used. That aside, it should beemphasised that this study’s primary results largely follow previous studies,but with much finer grained distinctions.

On another note, this methodology can be flexibly used with different types ofanaphoric judgement tests. As it was seen, both an IJT and TVJT were used withPhased Choice, successfully providing results that were quite consistent with eachother. By assessing each possible antecedent one at a time, the antecedent isassessed upon its own merits while not being influenced by another antecedentquestion that occurs alongside it. To facilitate this, clear instructions were given tothe participants to evaluate each possible antecedent separately, a strong rando-misation scheme was employed in order to avoid any potential response bias andprovided a unique script for every participant, while half the test containeddistracter items, all of which led to a valid and reliable output.

8 Conclusion

Various SLA studies on reflexives have had problems in discovering the fullextent of binding in native speakers and learners alike due to the methodology

Assessing anaphoric relations 381

Authenticated | [email protected] author's copyDownload Date | 11/6/15 4:16 AM

used. This has resulted in forming conclusions based upon data which might notreflect a speaker’s true competence. Testing anaphoric reference is never asstraight forward as one would like it to be; all experiments have their owncomplexities, the current being no exception. Nevertheless, using the PhasedChoice methodology reduces a test’s exposure to various biases, and used inconjunction with the IJT or TVJT is an advance over previous types.

Acknowledgements: I would like to thank both reviewers for their constructivecomments which have improved this paper from what it was. Any remainingerrors are my own. This paper represents re-worked sections from Sperlich (2013).

References

Akiyama, Yasuhiro. 2002. Japanese adult learners’ development of the locality condition onEnglish reflexives. Studies in Second Language Acquisition 24(1). 27–54.

Baker, C. L. 1995. Contrast, discourse prominence, and intensification, with special reference tolocally free reflexives in British English. Language 71(1). 63–101.

Berent, Gerlad P. & Vincent J. Samar. 1990. The psychological reality of the subset principle:Evidence from the governing categories of prelingually deaf adults. Language 66(4).717–741.

Blackwell, Sarah E. 2000. Anaphora interpretations in Spanish utterances and the neo-Griceanpragmatic theory. Journal of Pragmatics 32(4). 389–424.

Birdsong, David. 1989. Meta linguistic performance and interlinguistic competence. Berlin:Springer-Verlag.

Carroll, John M., Thomas G. Bever & Chava R. Pollack. 1981. The non-uniqueness of linguisticintuitions. Language 57(2). 368–383.

Chen, Dongdong. 1995. Chinese reflexive ziji in second language acquisition. Penn WorkingPapers in Linguistics 2(2). 37–52.

Chen, Dongdong. 1996. The L2 acquisition of Chinese reflexive ziji. McGill Working Papers inLinguistics 12. 1–15.

Chomsky, Noam. 1995. The Minimalist Program. Cambridge, MA: MIT Press.Cook, Vivian J. 1990. Timed comprehension of binding in advanced L2 learners of English.

Language Learning 40(4). 557–599.Cowart, Wayne. 1997. Experimental syntax: Applying objective methods to sentence judgements.

Thousand Oaks, CA: Sage Publications.Crain, Stephen& Cecile Mckee. 1986. Acquisition of structural restrictions on anaphora. In Stephen

Herman, Jae-Woong Chee & Joyce McDonough (eds.), Proceedings of the 16th annual meetingof the North East Linguistics Society, 94–110. Amherst, MA: GLSA Publications.

Davies, William D. & Tamar I. Kaplan. 1998. Native speaker vs. L2 learner grammaticalityjudgements. Applied Linguistics 19(2). 183–203.

Demirci, Mahide. 1997. The role of pragmatics in the acquisition of reflexive binding in L2. EastLansing, MI: Michigan State University Add PhD dissertation.

382 Darcy Sperlich

Authenticated | [email protected] author's copyDownload Date | 11/6/15 4:16 AM

Demirci, Mahide. 2000. The role of pragmatics in reflexive interpretation by Turkish learners ofEnglish. Second Language Research, 16(4). 325–353.

Dienes, Zoltan. 2008. Subjective measures of unconscious knowledge. Progress in BrainResearch 168. 49–64.

Dienes, Zoltan & Ryan Scott. 2005. Measuring unconscious knowledge: Distinguishingstructural knowledge and judgment knowledge. Psychological Research 69(5–6).338–351.

Dugarova, Esuna. 2008. Interpretation of the Chinese reflexive ‘ziji’ by English speakers. InBoping Yuan (ed.), Theoretical and empirical approach to applied Chinese languagestudies, 7–16. London: Cypress Book Company.

Ellis, Rod. 2004. The definition and measurement of L2 explicit knowledge. Language Learning54(2). 227–275.

Ellis, Rod. 2005. Measuring implicit and explicit knowledge of a second language. Studies inSecond Language Acquisition 27(2). 141–172.

Ellis, Rod. 2009. Implicit and explicit learning, knowledge and instruction. In Rod Ellis, ShawnLoewen, Catherine Elder, Rosemary Erlam, Jenefer Philp & Hayo Reinders (eds.), Implicitand explicit knowledge in second language learning, testing and teaching, 3–25. Bristol:Multilingual Matters.

Ellis, Rod. 2009a. Measuring implicit and explicit knowledge of a second language. In Rod Ellis,Shawn Loewen, Catherine Elder, Rosemary Erlam, Jenefer Philp & Hayo Reinders (eds.),Implicit and explicit knowledge in second language learning, testing and teaching, 31–64.Bristol: Multilingual Matters.

Finer, Daniel L. & Ellen I. Broselow. 1986. Second language acquisition and reflexive binding. InStephan Herman, Jae-Woong Chee & Joyce McDonough (eds.), Proceedings of the 16thannual meeting of the North East Linguistics Society, 154–183. Amherst, MA: GLSAPublications.

Finer, Daniel L. 1991. Binding parameters in second language acquisition. In Lynn Eubank (ed.),Point counterpoint: Universal Grammar in the second language, 351–374. Amsterdam:John Benjamins.

Gass, Susan M. 1994. The reliability of second-language grammatically judgements. In Elaine E.Tarone, Susan M. Gass & Andrew D. Cohen (eds.), Research methodology in second-language acquisition, 303–322. Hillsdale, NJ: Lawrence Erlbaum.

Givón, T. 1979. On understanding grammar. New York: Academic Press.Gordon, Peter C. & Randall Hendrick. 1997. Intuitive knowledge of linguistic co-reference.

Cognition 62(3). 325–370.Hamilton, Robert. 1996. Against underdetermined reflexive binding. Second Language Research

12(4). 420–466.Heinat, Fredrik. 2008. Probes, pronouns and binding in the Minimalist Program. Saarbrücken:

VDM Verlag.Hicks, Glyn. 2009. The derivation of anaphoric relations. Amsterdam: John Benjamins.Hirakawa, Makiko. 1990. A study of the L2 acquisition of English reflexives. Second Language

Research 6(1). 60–85.Hu, Jianhua, Haihua Pan & Liejiong Xu. 2001. Is there a finite vs. nonfinite distinction in

Chinese? Linguistics 39(6). 1117–1148.Hua, Te-Fang. 1994. The interpretation of L2 reflexive pronouns by adult Chinese learners of

English. University of Hawai’i Working Papers in ESL 13(1). 53–87.

Assessing anaphoric relations 383

Authenticated | [email protected] author's copyDownload Date | 11/6/15 4:16 AM

Huang, C.-T. James. 1982. Logical relations in Chinese and the theory of grammar. New York:Garland Publishing.

Huang, C.-T. J. & C.-C. J. Tang. 1991. The local nature of the long-distance reflexive in Chinese. InJ. Koster & E. Reuland (eds.), Long distance anaphora, 263–281. Cambridge: CambridgeUniversity Press.

Huang, C.-T. J. & C.-S. L. Liu. 2001. Logophoricity, attitudes and ziji at the interface. In P. Cole,G. Hermon & C.-T. J. Huang (eds.), Syntax and semantics 33: Long-distance reflexives,141–195. New York: Academic Press.

Huang, Yan. 1994 The syntax and pragmatics of anaphora: A study with special reference toChinese. Cambridge: Cambridge University Press.

Huang, Yan. 2000. Anaphora: A cross-linguistic study. Oxford: Oxford University Press.Jiang, Lin. 2009. A referential/quantified asymmetry in the second language acquisition of

English reflexives by Chinese-speaking learners. Second Language Research 25(4).469–491.

Kim, Hye-Ryun. 1993. First and second language acquisition of binding: An investigation ofKorean-speaking children and Korean adult learners of English. Gainesville, FL: Universityof Florida dissertation.

Kim, Ji-Hye & Silvina Montrul. 2004. First language influence on second language attrition:Interpretation of the Korean binding system. Harvard Studies in Korean Linguistics 10.86–98.

Kim, Ji-Hye & Silvina Montrul & James Yoon. 2009. Binding interpretations of anaphors byKorean heritage speakers. Language Acquisition 16(1). 3–35.

Kim, Ji-Hye & James Hye Suk Yoon. 2008. An experimental syntactic study of binding of multipleanaphors in Korean. Journal of Cognitive Science 9(1). 1–30.

Kuno, Susumu. 1987. Functional syntax: Anaphora, discourse and empathy. Chicago, IL:University of Chicago Press.

Lakshmanan, Usha & Larry Selinker. 2001. Analysing interlanguage: How do we know whatlearners know? Second Language Research 17(4). 393–420.

Lakshmanan, Usha & Keiko Teranishi. 1994. Preferences versus grammatically judgements:Some methodological issues concerning the governing parameter in second languageacquisition. In Elaine E. Tarone, Susan M. Gass & Andrew D. Cohen (eds.), Researchmethodology in second-language acquisition, 185–206. Hillsdale, NJ: LawrenceErlbaum.

Lee, Kum-Young. 2008. The role of pragmatics in reflexive interpretation by Korean learnersof English. In Melissa Bowles (ed.), Selected proceedings of the 2007 Second LanguageResearch Forum, 97–112. Somerville, MA: Cascadilla Proceedings Project.

Lee, Kum-Young. 2009. Finite control in Korean. Iowa City, IA: University of Iowa PhD dissertation.Loewen, Shawn. 2009. Grammaticality judgment tests and the measurement of implicit and

explicit L2 knowledge. In Rod Ellis, Shawn Loewen, Catherine Elder, Rosemary Erlam,Jenefer Philp & Hayo Reinders (eds.), Implicit and explicit knowledge in second languagelearning, testing and teaching, 65–93. Bristol: Multilingual Matters.

MacLaughlin, Dawn. 1995. Universal grammar and L2 acquisition of reflexive binding: Somelearners acquire a non-L1/non-target system. Paper presented at the Second LanguageResearch Forum, Cornell University.

Pinker, Steven. 1994. The language instinct. London: Allen Lane The Penguin Press.Purapura, James E. 2004. Assessing grammar. Cambridge: Cambridge University Press.

384 Darcy Sperlich

Authenticated | [email protected] author's copyDownload Date | 11/6/15 4:16 AM

Rebuschat, Patrick. 2008. Implicit learning of natural language syntax. Cambridge: Universityof Cambridge PhD Dissertation.

Rebuschat, Patrick & John N. Williams. 2012. Implicit and explicit knowledge in second lan-guage acquisition. Applied Psycholinguistics 33(4). 829–856.

Reuland, Eric. 2011. Anaphora and language design. Cambridge, MA: MIT Press.Rooryck, Johan & Guido V. Wyngaerd. 2011. Dissolving Binding Theory. Oxford: Oxford

University Press.Runner, Jeffrey T. & Micah B. Goldwater. 2011. On the representational nature of representa-

tional noun phrases. In Jeffrey T. Runner (ed.), Syntax and semantics 37: Experiments at theinterfaces, 219–250. Bingley: Emerald.

Sorace, Antonella. 1988. Linguistic intuitions in language development: The problem of inter-determinacy. In James Pankhurst, Mike S. Smith & Paul V. Buren (eds.), Learnability andsecond languages: A book of readings, 167–190. Dordrecht: Foris.

Sorace, Antonella. 1996. The use of acceptability judgements in second language acquisitionresearch. In William C. Ritchie & Tej K. Bhatia (eds.), Handbook of second languageacquisition, 375–409. San Diego, CA: Academic Press.

Sperlich, Darcy. The acquisition of long-distance reflexives in Chinese as an interlanguage: Anexperimental study. Auckland: The University of Auckland PhD Dissertation.

Tang, C.-C. Jane. 2000. Finite and nonfinite clauses in Chinese. Yuyan ji Yuyanxue 1(1). 191–214.Thomas, Margaret. 1989. The interpretation of English reflexive pronouns by non-native

speakers. Studies in Second Language Acquisition 11(3). 281–303.Thomas, Margaret. 1991a. Do second language learners have “rouge” grammars of anaphora?

In Lynn Eubank (ed.), Point counterpoint: Universal Grammar in the second language,221–239. Amsterdam: John Benjamins.

Thomas, Margaret. 1991b. Universal Grammar and the interpretation of reflexives in a secondlanguage. Language 67(2). 211–239.

Wells, Terri L. 1998. L2 acquisition of English binding domains. In Maria-Luise Beck (ed.),Morphology and its interfaces in second language knowledge, 227–256. Amsterdam: JohnBenjamins.

White, Lydia. 1995. Input, triggers, and second language acquisition: Can binding be taught?In Fred R. Eckman, Diane Highland, Peter W. Lee, Jean Milcham & Rita R. Weber (eds.), Secondlanguage acquisition: Theory and pedagogy, 63–78. Mahwah, NJ: Lawrence Erlbaum.

White, Lydia, Joyce Bruhn-Garavito, Takako Kawasaki, Joe Pater & Philippe Prévost. 1997.The researcher gave the subject a test about himself: Problems of ambiguity andpreference in the investigation of reflexive binding. Language Learning 47(1).145–172.

Ying, H. G. 1999. Access to UG and language transfer: A study of L2 learners’ interpretation ofreconstruction in Chinese. Second Language Research 15(1). 41–72.

Yip, Virginia & Gladys Tang. 1998. Acquisition of English reflexive binding by Cantoneselearners: testing the positive transfer hypothesis. In Maria-Luise Beck (ed.), Morphologyand its interfaces in second language knowledge, 165–193. Amsterdam: John Benjamins.

Yuan, Boping. 1998. Interpretation of binding and orientation of the Chinese reflexiveziji by English and Japanese speakers. Second Language Research 14(4). 324–340.

Yule, George, Jerry L. Yanz & Atsuko Tsuda. 1985. Investigating aspects of the language learner’sconfidence: An application of the theory of signal detection. Language Learning 35(3).473–488.

Assessing anaphoric relations 385

Authenticated | [email protected] author's copyDownload Date | 11/6/15 4:16 AM

Appendix

Target Sentences for English, Chinese and Korean.

N-F/NF

Tom knows that Dick thinks highly of himself.老王认为老陈觉得自己很了不起。

철이는 민수가 자신을 높이 평가한다는 것을 알고 있다.

Tom thinks that Dick saw a photo of himself in the newspaper.老王认为老陈看见了自己在报纸上的照片。

민수는 철이가 신문에서 자신의 사진을 봤다고 생각한다.

Tom wants Dick to give himself more time.老王要老陈给自己多一点时间。

민수는 철이가 자신에게 더 많은 시간을 줄 것을 원한다.

Tom wants Dick to post himself a letter.老王要老陈寄一封信给自己。

민수는 철이가 자신에게 편지를 부치기를 원한다.

P-F/NF

Confucius believed that the new disciple had not yet studied the writings abouthimself.孔子相信新的弟子还没有读过关于自己的文章。

공자는 새로운 제자가 자신에 대한 글을 아직 공부하지 않았다고 믿었다.

The teacher believes that the student will give himself an apple in every class.老师相信学生每一堂课都会给自己一颗苹果。

선생님은 학생이 매 수업중 자신에게 사과를 줄 것을 믿는다.

The Emperor commissioned the artist to paint a portrait of himself.皇帝派画家画一幅自己的画像。

황제는 화가에게 자신의 초상화를 그릴 것을 위임했다.

The Emperor commissioned the architect to design a new royal capital forhimself.皇帝派设计师设计一座新的皇都给自己。

황제는 건축가에게 자신을 위한 새로운 황도를 설계할 것을 위임했다.

386 Darcy Sperlich

Authenticated | [email protected] author's copyDownload Date | 11/6/15 4:16 AM

PS-F/NF

The poisoned hero doesn’t know that the devious villain poisoned himself.中毒的英雄不知道是坏人毒死自己的。

중독된 영웅은 교활한 악한이 자신에게 독을 쓴 것을 알지 못한다.

The injured policeman said that the drunken motorist assaulted himself.受伤的警察说酒醉的驾驶攻击自己。

부상 당한 경찰은 음주 운전자가 자신을 폭행했다고 말했다.

The old soldier wants to tell the young man the terrible war stories about himself.老兵想要跟年轻人说自己可怕的打仗的故事。

늙은 군인은 젊은 남자에게 자신에 대한 끔찍한 전쟁 이야기를 말하기를

원한다.

The arrogant movie star wants to show the weak director the rewritten partsabout himself.自大的电影明星给软弱的导演看自己角色重改的部分。

거만한 배우는 나약한 감독에게 자신에 대해 재작성된 부분을 보여주기를

원한다.

T-F

Tom heard that Dick thinks that Harry had hurt himself.老王听说老陈认为老李伤害了自己。

철이는 민수가 동수는 자신을 해쳤다고 생각한다고 들었다.

Tom thought that Dick said that Harry hates himself.老王以为老陈说老李讨厌自己。

철이는 민수가 동수는 자신을 미워한다고 말했다고 생각했다.

P-LD1

The visiting president knows that the school teacher said that the student hadmet himself once before.来参观的总统知道学校老师说学生有见过自己。

방문한 대통령은 선생님이 학생은 자신을 전해 한번 만났다고 말했다는 것을

안다.

The school boy hopes that the teacher will guarantee that the school bully won’ttease himself in class tomorrow.小学生希望老师会保证校园流氓明天上课不会欺负自己。

남학생은 선생님이 학교에서 따돌림이 내일 교실에서 자신을 괴롭히지 않을

것을 보장해 줄 것을 바란다.

Assessing anaphoric relations 387

Authenticated | [email protected] author's copyDownload Date | 11/6/15 4:16 AM

P-LD2

The doctor found out that the patient doesn’t know that the chemist prescribedhimself with the wrong medication.医生发现病人不知道药剂师开错药给自己。

의사는 환자가 약사는 자신에게 잘못된 약을 처방했다는 것을 알지 못 한다는

것을 간파했다.

The cook heard that the customer complained that the waiter served himselfwith the wrong main.厨师听见客人抱怨服务生上错菜给自己。

조리사는 고객이 웨이터는 자신에게 잘못된 식사를 제공했다고 불평하는 것을

들었다.

PS-LD1

The famous scientist thinks that a colleague knows that the journalist willinterview himself tomorrow.有名的科学家认为同事知道明天会有记者来采访自己。

유명한 과학자는 동료가 신문기자는 내일 자신을 면담 한다는 것을 안다고

생각한다.

The imprisoned president suspected that the colonel knew that the generalwould overthrow himself.被囚禁的总统怀疑中校知道将军会推翻自己。

수감된 대통령은 중령이 장군은 자신을 전복 시킨다는 것을 알고 있다고

의심했다.

PS-LD2

The student heard that the lazy teacher knows that the angry principal will firehimself tomorrow.学生听说懒惰的老师知道生气的校长明天会解聘自己。

학생은 게으른 선생님이 화난 교장은 내일 자신을 해고 시킨다는 것을 안다고

들었다.

The happy teacher believes that the bad student knows that the principal willexpel himself tomorrow.快乐的老师相信坏学生知道校长明天会开除自己。

행복한 선생님은 불량 학생이 교장은 내일 자신을 쫓아 낸다는 것을 안다고

믿는다.

388 Darcy Sperlich

Authenticated | [email protected] author's copyDownload Date | 11/6/15 4:16 AM


Recommended