1. Introduction

Interpretations of reflexives by L2 learners have attracted much attention in previous L2 research. Among reflexive expressions used in natural languages, previous studies have mainly focused on two types of reflexives: a bi-morphemic reflexive, such as the English himself and the Japanese kare-zisin, (himself), which must be bound by an antecedent within its local domain, and a mono-morphemic reflexive, such as Japanese zibun (self) and Chinese ziji (self), which can be bound by an antecedent outside of their local domain.1 For example, the English reflexive himself in (1) cannot refer to the matrix subject, Taro, as it is outside of the local domain. The same restriction does not apply to zibun in (2) or ziji in (3), for which both the matrix and local subjects are possible antecedents. Reflexive expressions such as zibun and ziji, which can be bound by a non-local antecedent, are called long-distance (LD) reflexives. In addition, most LD reflexives are subject-oriented, with co-reference restricted to a clausal subject.

    1. (1)
    1. Taro1 thinks [Ken2 loves himself*1/2].
    1. (2)
    1. Taro1-ga
    2. Taro-NOM
    1. [Ken2-ga
    2. Ken-NOM
    1. zibun1/2-o
    2. self-ACC
    1. aisiteiru-to]
    2. love-that
    1. omotteiru.
    2. thinks.
    1. ‘Taro1 thinks Ken2 loves him/himself1/2.’
    1. (3)
    1. Zhangsan1
    2. Zhangsan
    1. zhidao
    2. know
    1. [Lisi2
    2. Lisi
    1. xihuan
    2. like
    1. ziji1/2].
    2. self.
    1. ‘Zhangsan1 knows Lisi2 likes him/himself1/2.’

Extensive research has studied the L2 acquisition of bi-morphemic reflexives by learners whose L1 has a mono-morphemic LD reflexive (e.g. Finer & Broslow, 1986; Hirakawa, 1990) and L2 acquisition of LD reflexives by learners whose L1 lacks them (e.g. Thomas, 1993; Yuan, 1998; Shirahata, 2006). These studies focused primarily on whether or not L2 leaners are able to correctly figure out the appropriate binding domains for a reflexive in the target language. However, few studies have focused on L2 acquisition of LD reflexives by learners whose L1 also has an LD reflexive (e.g. Yuan, 1998; Yoshimura et al., 2012; Yoshimura et al., 2013). This is largely due to a broadly held assumption of L1 transfer. A learner’s L1 grammar is assumed to transfer onto the initial state of a nascent L2 grammar (e.g. Schwartz & Sprouse, 1996), and as a result, if a learner’s L1 has an LD reflexive, they would be expected to already know how LD reflexives are interpreted.

Although Japanese zibun and Chinese ziji are both LD reflexives, there are in fact differences between them in interpretation (e.g. Kuno, 1972; Huang et al., 1984; Oshima, 2004; Nishigauchi, 2015). The difference between the two LD reflexives comes from the type of LD antecedent that each reflexive can be bound by. In Japanese, there are two types of LD antecedents, empathic and logophoric, and both can bind zibun in a subordinate clause (e.g. Kuno and Kaburaki, 1977; Kuno, 1978; Oshima, 2004; Kishida, 2011). In Chinese, while LD binding of ziji by a logophoric LD antecedent is possible as well, LD binding by an empathic antecedent is marginal at best (Huang, 1984; Huang et al., 2009). In this study, we investigate whether or not Chinese-speaking learners of Japanese can acquire the two types of LD antecedents in Japanese and interpret zibun in a target-like way. The acquisition of the two types of LD antecedents was tested using a condition known as the blocking effect. The blocking effect refers to a phenomenon where LD binding of an LD reflexive is “blocked” in the presence of a first- or second-person pronoun in the local domain. In Japanese, an empathic LD antecedent is subject to the blocking effect but a logophoric LD antecedent is not. In Chinese, by contrast, the blocking effect does apply to a logophoric LD antecedent (i.e. the only LD antecedent permitted in Chinese). We conducted an experiment to test for the presence/absence of the blocking effect in L2 Japanese interlanguage grammars (ILGs) to examine whether Chinese-speaking (CS) learners of Japanese are able to acquire the interpretation of an LD reflexive different from that of the LD reflexive in their L1.

This paper is organized as follows: In section 2, the interpretations of zibun and ziji as well as their sensitivity to the blocking effect will be summarized. Section 3 summarizes relevant previous studies, and section 4 describes the task administered to investigate CS learners’ acquisition of Japanese zibun along with the results from the study. Section 5 discusses the findings from the study and conclusions will be given in section 6.

2. Mono-morphemic Reflexives in Japanese and Chinese

As mentioned above, although both zibun and ziji are LD reflexives, their interpretations are not identical (e.g., Kuno, 1972; Huang, et al., 1984; Oshima, 2004; Nishigauchi, 2015). In Japanese, it has been claimed that there are three possible types of antecedent for zibun; genuine, empathic, and logophoric (Oshima, 2004; Kishida, 2011). According to Kishida (2011), the genuine binding of zibun is when zibun is bound by a co-argument within a local domain. It is claimed that a genuine reflexive interpretation is constrained by the syntactic condition of locality. Empathic and logophoric antecedents are both LD antecedents that are not constrained syntactically; rather, their interpretations are induced from pragmatic information. An empathic LD antecedent is the participant in a sentence that the speaker empathizes with or most identifies with (Kuno and Kaburaki, 1977; Kuno, 1987).2 In (4), the sentence is uttered from the matrix subject’s point of view, and thus Takasi is assumed to be its empathic locus. Such a participant in a sentence may be the antecedent of zibun, despite not being in the local domain containing zibun.

(4) Takasi1-wa [sensei-ga zibun1-o yobi-ni kita toki] hon-o yondeita.  
  Takasi-TOP teacher-NOM self-ACC call came when book-ACC was.reading  
  ‘Takasi1 was reading a book when the teacher came to call him1.’
  (adapted from Nishigauchi, 2015)

Another type of LD antecedent is logophoric. A logophoric antecedent is a participant whose “speech, thoughts, feelings, or general state of consciousness are reported” in a sentence (Clements, 1975). An example of a logophoric antecedent is given in (5). The matrix subject is a logophoric participant and can be an LD antecedent of zibun.

(5) Taro1-ga [Ken-ga zibun1-o nagutta-to] itta.  
  Taro-NOM Ken-NOM self-ACC hit-that said.  
  ‘Taro1 said Ken hit him1.’
  (Adapted from Kuno (1978), p. 212)

The two different types of LD antecedents permitted for zibun lead to a number of differences (Oshima, 2004; Kishida, 2011). For example, an empathic antecedent of zibun cannot be a non-subject, but a logophoric non-local antecedent can be. In addition, while zibun bound by an empathic antecedent can be replaced by a pronoun, zibun bound by a logophoric individual cannot. The blocking effect is another difference between the two types of LD antecedent. When there is a first-person pronoun within the local domain, zibun cannot be bound by an empathic LD antecedent, yet zibun can be bound by a logophoric LD antecedent even in the presence of a first-person pronoun in the local domain (Kuno, 1978; Oshima, 2004).

    1. (6)
    1. *Taro1-wa
    2. Taro-TOP
    1. [boku-ga
    2. I-NOM
    1. zibun1-ni
    2. self-DAT
    1. kasita]
    2. lent
    1. okane-o
    2. money-ACC
    1. nakusite-simatta
    2. lost
    1. rasii.
    2. seems
    1. ‘It seems that Taro1 lost the money I lent to him1.’
(7) Taro1-wa [boku-ga zibun1-o butta]-koto-o mada urandeiru.  
  Taro-TOP I-NOM self-ACC hit-fact-ACC still resent.  
  ‘Taro1 still resents that I hit him1.’
  (Kuno, 1978, pp. 212–213)

In (6), zibun cannot be co-referential with the matrix subject, Taro, but in (7) where the matrix subject is a logophoric individual, zibun can be co-referential with it. Thus, in Japanese, the blocking effect is induced with an empathic LD antecedent but not with a logophoric one.

Turning to Chinese, Huang et al. (2009) claim that there are two distinct types of ziji. One is a locally bound genuine reflexive and the other is bound non-locally by a logophoric individual, as shown in (3) in the introduction. However, according to Huang et al., the empathic type of LD antecedent is only marginal as a co-referent of ziji for some native speakers of Chinese.

(8) ??[Zhangsan lai kan ziji1 de shihou], Lisi1 zheng zai kan shu.  
     Zhangsan come see self DE moment Lisi now at read book.  
     ‘Lisi1 was reading when Zhangsan came to visit him1.’
  (Huang et al., 2009, p. 346)

Chinese ziji also differs from Japanese zibun with respect to the blocking effect. Unlike in Japanese, a logophoric LD antecedent is subject to the blocking effect in Chinese (Huang, 1984), as shown in (9).

(9) Zhangsan1 juede [wo2 dui ziji*1/2 mei xinxin].  
  Zhangsan think I to self not confident.  
  ‘Zhangsan1 thinks I2 have no confidence in him*1/2.’  
  (adapted from Nishigauchi, 2015)

Table 1 summarizes the properties associated with the interpretations of Japanese zibun and Chinese ziji. In Japanese, there are three types of antecedents for zibun: genuine, empathic and logophoric; the latter two are LD antecedents. In Chinese, on the other hand, there are two types of antecedents for ziji, genuine for local binding and logophoric for LD binding.

Table 1

Japanese and Chinese mono-morphemic reflexives.

Local LD

Genuine Empathic Logophoric

Japanese zibun Yes Yes [blocking] Yes
Chinese ziji Yes ? Yes [blocking]

For Chinese ziji, it has been argued that the blocking effect is induced by a perspective conflict.3 Huang et al. (1984) argue that a first- or second-person pronoun enters into a perspective conflict with a logophoric individual. In (9), for example, there are two logophoric individuals, the sentence-internal speaker, Zhangsan, and the speaker of the sentence represented by the first-person pronoun. Thus, there are two logophoric individuals in the sentence, which creates a conflict, leading to the logophoric LD antecedent Zhangsan being unable to bind ziji. In Japanese, on the other hand, a logophoric individual in a sentence alongside the speaker of the sentence expressed by a first-person pronoun does not lead to a perspective conflict, while the empathic locus alongside the speaker (the first-person pronoun) does.

It is not clear to us at this point why it is empathy, not logophoricity, that induces the blocking effect in Japanese, but in Chinese, it is logophoricity that does so. Nevertheless, we might speculate that the difference could be caused by the use of empathy in Japanese. Empathy can be encoded linguistically and it is manifested in lexical items like yaru/kureru (give). The verb yaru (to give) has been analyzed as a giver-oriented verb but kureru (to give) as a receiver-oriented verb. In (10a), the verb yaru is acceptable if the empathy locus is on the subject Taro, but the verb kureru is also acceptable if the empathy locus is on the indirect object Hanako (because, for example, Hanako is the speaker’s daughter). However, in (10b) the receiver-oriented verb kureru must be used because the speaker is him- or herself, the receiver of the book and the empathy locus has to be on him or her.

    1. (10)
    1. a.
    1. Taro-ga
    2. Taro-NOM
    1. Hanako-ni
    2. Hanako-DAT
    1. hon-o
    2. book-ACC
    1. yatta/kureta.
    2. give/give
    1. ‘Taro gave Hanako a book.’
    1.  
    1. b.
    1. Taro-ga
    2. Taro-NOM
    1. boku-ni
    2. I-DAT
    1. hon-o
    2. book-ACC
    1. *yatta/kureta.
    2. give/give
    1. ‘Taro gave me a book.’

Sentences in (11) and (12) show how linguistic empathy associated with the verbs yaru/kureru interacts with the interpretation of zibun. In order for the matrix subject, Taro, to be the LD antecedent of zibun, it has to be the empathy locus. In (11), since the giver-oriented verb yaru is used, the empathy locus is on the embedded subject, Hanako, and zibun cannot refer to the matrix subject. In (12), on the other hand, the receiver-oriented verb kureru is used. In this case, the empathy locus is not on the giver, Hanako, so it can be on the matrix subject, Taro. Co-reference between Taro and zibun is, therefore, possible for (12).

    1. (11)
    1. *Taro1-wa
    2. Taro-TOP
    1. [Hanako-ga
    2. Hanako-NOM
    1. zibun1-ni
    2. self-DAT
    1. yatta]
    2. gave
    1. hon-o
    2. book-ACC
    1. yonda.
    2. read.
    1. ‘Taro read the book Hanako gave to him.’
(12) Taro1-wa [Hanako-ga zibun1-ni kureta] hon-o yonda.  
  Taro-TOP Hanako-NOM self-DAT gave book-ACC read.  
  ‘Taro read the book Hanako gave to him.’
  (adapted from Oshima, 2007)

We speculate that because Japanese is a language in which empathy plays an important role, it is empathy, rather than logophoricity, that enters into point-of-view conflict with a first-person pronoun, inducing the blocking effect.4

3. Previous Studies

A number of studies have investigated L2 acquisition of mono-morphemic reflexives, including zibun and ziji. The main focus of these studies were to investigate whether or not L2 learners, whose L1 lacks a mono-morphemic reflexive, acquire its non-local binding and its subject-orientation (Thomas, 1993; Yuan, 1998; Kano & Nakayama, 2004; Shiharata, 2006; Yoshimura et al., 2012). There are nevertheless a few studies that have included learners whose L1 does have a mono-morphemic reflexive. Yuan (1998), for example, tested English- and Japanese-speaking learners of Chinese at the intermediate and advanced levels using a multiple-choice comprehension test. Learners were asked to choose an LD subject, a local subject, or both as the possible antecedent(s) of ziji in test sentences. The results showed L1 influence for the interpretations of LD reflexive ziji with Japanese speakers outperforming English speakers. They accepted LD binding of ziji and showed no statistical difference from the native speaker group, while their English-speaking counterparts were statistically different from both the NS group and the Japanese-speaker group, showing that they were less able to access LD interpretations of ziji.

Yoshimura et al. (2012) examined the interpretation of Japanese zibun by English and Chinese speakers. They assume that locally bound zibun is syntactically constrained and subject to Binding Principle A (Chomsky, 1981), whereas LD bound zibun is logophoric, induced by pragmatic information (Kuno, 1972). The participants were tested on a truth-value judgment task. Yoshimura et al. used bi-clausal, logophoric sentences with a finite embedded clause. The reflexive zibun appeared in a Genitive position in the embedded clause, as shown in (13).

    1. (13)
    1. Taro-wa
    2. Taro-TOP
    1. Yasuo-ga
    2. Yasuo-NOM
    1. zibun-no
    2. self-GEN
    1. puramoderu-o
    2. plastic.model-ACC
    1. gomibako-ni
    2. trash.can-in
    1. suteta
    2. throw
    1. to
    2. that
    1. iimashita.
    2. said
    1. ‘Taro said that Yasuo threw his plastic model away in the trash can.’

The results showed that both Chinese speaker (CS) and English speaker (ES) groups allowed local binding of zibun (CS: 92.2%, ES: 94.9%), but the CS leaners allowed more LD binding than the ES learners (CS: 78.4%, ES: 56.4%). They argue that, like Yuan’s (1998) study, the knowledge of an LD reflexive in a learner’s L1 facilitates the acquisition of an LD reflexive in L2. However, they point out that the acceptance rate for LD interpretations (“true” responses) of zibun by the CS group was lower than for local interpretations. They argue that the CS learners had more difficulty with LD binding of zibun, i.e. they allowed LD binding of zibun less frequently than local binding of zibun, because it requires integration of pragmatic information into sentence processing in order to determine the referent of the reflexive.5

The studies conducted by Yuan and by Yoshimura et al. have found that L2 learners whose L1 does not have an LD reflexive tend to prefer the reflexive taking an antecedent in its local domain. In contrast, L2 learners whose L1 has an LD reflexive are better at accessing LD interpretations. The results from Yoshimura et al. further suggest that L2 learners with an LD reflexive background nevertheless initially prefer local binding over LD binding. It was assumed in these studies, however, that the mono-morphemic reflexives zibun and ziji have the same binding interpretations. In contrast, in our study below, we recognize that the interpretations of zibun and ziji are not identical and that CS learners must acquire the Japanese distinction between two types of LD antecedents, empathic and logophoric. In the following section, we present our study examining the acquisition of zibun by CS learners of Japanese.

4. Study

4.1. Research questions

As discussed earlier, Japanese has empathic and logophoric LD antecedents for reflexives, while Chinese has only the latter. Furthermore, in Japanese, only the empathic LD antecedents are subject to the blocking effect whereas in Chinese, its logophoric LD antecedents are. Assuming L1 transfer as the L2 initial state, CS learners are faced with learning that: i) zibun can be bound by empathic, in addition to logophoric, LD antecedents, and ii) the blocking effect holds for empathic but not for logophoric LD antecedents. If CS learners of Japanese fail to acquire these two properties associated with zibun, they should fail to accept empathic LD antecedents and show the presence of a blocking effect with logophoric LD antecedents. This study thus examines whether they in fact allow empathic LD antecedents for zibun and whether they recognize the blocking effect applying to empathic LD antecedents but not to LD logophoric antecedents.

4.2. Participants

Twenty-eight Mandarin-speaking learners of Japanese and 36 native speakers of Japanese participated in the study. All participants in the CS group were undergraduate or graduate students at a university in Japan at the time of testing. They were advanced-level learners of Japanese, having passed the highest level of the Japanese-Language Proficiency Test (N1) and/or scoring over 80% on the Minimal Japanese Test (MJT) (Maki, Dunton, & Obringer, 2003).6 Participants in the native-speaker (NS) group were undergraduate students living in Tokyo and its surrounding areas. Table 2 summarizes participant ages, length of Japanese study and length of residence in Japan.

Table 2

Participants.

Age Length of study Length of residence

NS (n = 36) Mean 21.4
Range (20–25)
CS (n = 28) Mean 25.5 5;09 2;08
Range (21–33) 2;00–10;11 0;07–8;04

4.3. Tasks and procedures

The CS learners were asked to take two tests: first the MJT and then a truth-value judgment test (TVJT). The NS group took only the TVJT. The TVJT was presented as a paper-and-pencil test with no time limit for participants to complete the task.

In the TVJT, there were two types of LD antecedents, empathic and logophoric, and each type was further divided into two; one sentence type included both a third person LD subject and third person local (LOC) subject, shown in (14A) and (15A), and another type contained a third person LD subject but a first person LOC subject, as shown in (14B) and (15B). The latter two sentence types were included to test L2 learners’ knowledge of the blocking effect.

    1. (14)
    1. Type 1: Empathic condition
    1.  
    1. A:
    1. Taro1-ga
    2. Taro-NOM
    1. [Hanako2-ga
    2. Hanako-NOM
    1. kyoositu-de
    2. classroom-at
    1. zibun1/2-o
    2. self-ACC
    1. hometa
    2. praised
    1. atode]
    2. after
    1. kaetta.
    2. went.home
    1. ‘Taro went home after Hanako praised him/herself in a classroom.’
    1.  
    1. B:
    1. Tosiko1-ga
    2. Tosiko-NOM
    1. [watasi2-ga
    2. I-NOM
    1. taikukan-de
    2. gym-at
    1. zibun*1/2-o
    2. self-ACC
    1. semeta
    2. criticized
    1. atode]
    2. after
    1. nakidasita.
    2. cried
    1. ‘Tosiko burst into tears after I criticized her in a gym.’
    1. (15)
    1. Type 2: Logophoric condition
    1.  
    1. A:
    1. Mituko1-ga
    2. Mituko-NOM
    1. [Ryota2-ga
    2. Ryota-NOM
    1. kittin-de
    2. kitchen-at
    1. zibun1/2-o
    2. self-ACC
    1. semeta-to]
    2. criticized-COMP
    1. itta.
    2. said
    1. ‘Mituko said Ryota criticized her/himself in the kitchen.’
    1.  
    1. B:
    1. Tutomu1-ga
    2. Tutomu-NOM
    1. [watasi2-ga
    2. I-NOM
    1. kafe-de
    2. café-at
    1. zibun1/2-o
    2. self-ACC
    1. syookaisita-to]
    2. introduced-COMP
    1. itta.
    2. said
    1. ‘Tutomu said I introduced him/myself (to someone) at a café.’

Both LOC and LD binding are allowed in all sentence types but Type 1B, in which LD binding of zibun is prohibited by the blocking effect of an intervening first-person pronoun in the LOC domain.

Each sentence type is presented with either a LOC context, where binding of zibun to a LOC subject is required for the sentence to be judged True, or an LD context where binding of zibun to an LD subject is required. Thus, there are eight conditions, summarized in Table 3, with predicted interpretations for zibun by Japanese NSs. Examples of context-sentence pairs from each of the eight conditions are provided in Appendix A.7

Table 3

Test conditions and predictions of True or False.

Sentence type A B

1 Empathic LOC LD LOC LD
True True True False
2 Logophoric LOC LD LOC LD
True True True True

Contexts and test sentences appeared in Japanese script with Chinese characters (i.e. kanji). Furthermore, kanji characters were annotated with syllabic transcriptions (hiragana) above to show their Japanese pronunciation. Test instructions appeared on the first page of the TVJT, directing participants to circle True if a test sentence matches the preceding context or circle False if it does not.

In total, 64 test items were created, eight items per condition. Since the same test sentences are used once in a LOC context and once in an LD context, two sets of TVJT test items were created using a 2 × 2 Latin square design to preclude any single participant from reading the same test sentence twice. Consequently, each participant read four test items per condition. In addition, 64 filler items were also constructed using pronouns kare (he) or kanozyo (she) in sentence structures similar to, but different from, the test items. Since there was an imbalance between true test items and false test items (see Table 3), the truth-values for the filler items were manipulated so that there would be an equal number of True and False answers across the complete set of stimuli. Each version of the TVJT consisted of 64 items, 32 test items and 32 fillers.

If CS learners adopt the interpretation of Chinese ziji for interpreting Japanese zibun, they should diverge from Japanese NSs in their LD interpretations of Types 1A and 2B, (Table 3). Predictions for the interpretation of zibun by CS learners based on L1 transfer are summarized in Table 4. They are predicted to reject LD binding for Type 1A, in which the LD antecedent is empathic, and to reject binding of zibun with a logophoric LD antecedent for Type 2B.

Table 4

Predictions based on transfer of Chinese ziji interpretation.

Sentence type A B

LOC LD LOC LD

1 Empathic True False True False
2 Logophoric True True True False

4.4. Results

4.4.1. Group results

Results from the Type 1 empathic condition are summarized in Table 5, showing the mean number of True judgments for each condition. Since each participant answered four tokens per condition, the maximum number is four. In the empathic condition, we expect three out of the four conditions to elicit True responses, while LD interpretations for Type 1B ought to be impossible due to the blocking effect. As can be seen in the table, native speakers did not choose True for Type 1A as frequently as expected for either LD or LOC interpretations of zibun, responding True at only about 60%. As for Type 1B, the NS group accepted LOC interpretations of zibun but rejected LD interpretations as expected, confirming the presence of the blocking effect. The CS group overall presents similar response patterns for Types 1A and 1B. In the LD interpretation of Type 1B, they also demonstrated sensitivity to the blocking effect. For the other three conditions, True response rates were lower than they were for the NS group but higher than the NS group for the LD condition of Type 1B.

Table 5

Type 1 empathic condition: The choices of True (out of 4 tokens).

1A 1B

Context LOC LD LOC LD

Target TRUE TRUE TRUE FALSE

NS (n = 36) Mean 2.72 (68.1%) 2.33 (58.3%) 3.22 (80.6%) 0.47 (11.8%)
SD 1.09 1.17 0.98 0.77
CS (n = 28) Mean 2.61 (65.2%) 1.89 (47.3%) 2.96 (74.1%) 0.75 (18.8%)
SD 1.40 1.47 0.83 0.93

One-way repeated measures ANOVAs conducted separately for Types 1A and 1B8 revealed a significant main effect for context (LOC vs. LD) (1A: F (1, 62) = 4.86, p = 0.031, η2 = 0.07, 1B: F (1, 62) = 268.99, p < 0.001, η2 = 0.81) but not for group (1A: F (1, 62) = 1.92, p = 0.171, 1B: F (1, 62) = 0.004, p = 0.952) and no significant interaction between context and group (1A: F (1, 62) = 0.42, p = 0.518, 1B: F (1, 62) = 3.13, p = 0.082). Thus, the results indicate that both groups distinguished LD and LOC contexts, but the CS group was not different from the NS group for Type 1 conditions.

Table 6 shows the mean numbers of True judgments for the Type 2 logophoric condition, for which both LOC and LD interpretations of zibun are predicted to be possible. As with Type 1, however, some native speakers responded False where True was expected. This is particularly evident for the LOC interpretations of Type 2A items and for both Type 2B conditions. CS learners, on the other hand, showed a sharp distinction between LOC and LD conditions with Type 2B items, rejecting the LD interpretation of zibun.

Table 6

Type 2 logophoric condition: The choices of True (out of 4 tokens).

2A 2B

Context LOC LD LOC LD

Target TRUE TRUE TRUE TRUE

NS (n = 36) Mean 1.86 (46.5%) 3.44 (86.1%) 2.30 (57.6%) 2.52 (66.1%)
SD 1.22 1.02 1.32 1.38
CS (n = 28) Mean 2.39 (59.8%) 2.53 (63.2%) 2.64 (66.1%) 0.92 (23.2%)
SD 1.25 1.34 1.12 1.05

For Type 2A, ANOVAs showed a significant main effect for context (LOC vs. LD) (F (1, 62) = 14.68, p < 0.001, η2 = 0.19) and a significant interaction between context and group (F (1, 62) = 10.22, p = 0.002, η2 = 0.14). However, there was no significant main effect for group (F (1, 62) = 0.84, p = 0.360). For Type 2B, there were significant main effects for context (F (1, 62) = 9.79, p = 0.003, η2 = 0.13) and group (F (1, 62) = 9.49, p = 0.003, η2 = 0.13), as well as a significant interaction between context and group (F (1, 62) = 6.50, p < 0.001, η2 = 0.21).

To compare the results from the NS and CS groups for Type 2 logophoric conditions, a t-test for independent samples was conducted. Results are summarized in Table 7.

Table 7

Type 2 logophoric condition: Statistical results from group comparisons.

A B

LOC LD LOC LD

t-value df = 62 1.70 3.06 1.07 5.07
p-value .093 .003 .286 <.001
Cohen’s d .280 .550 .177 .883

The differences between the NS and CS groups were significant for the LD conditions of both Types 2A and 2B. For Type 2A, the NS group accepted the LD interpretation of zibun, while the L2 group did so less frequently, leading to a statistically significant difference between the groups. For Type 2B, in which there is a first-person pronoun in the local domain of zibun, NS participants allowed LD binding of zibun, but the CS group rejected it, a response similar to the LD interpretation of zibun in the Type 1B condition.

Results from paired sample t-tests examining participants’ distinctions between LOC and LD contexts for Type 2 are shown in Table 8. The NS group distinguished between Type 2A conditions but not between Type 2B conditions. The distinction made for Type 2A follows from the preference NSs showed for LD interpretations. The CS group, on the other hand, distinguished between LOC and LD contexts for Type 2B but not for Type 2A. For Type 2B, the results in Table 8 suggest that the CS group rejected LD interpretations, as they did for Type 1B. In both types of sentences, a first-person pronoun appeared in the local domain of zibun, and participants showed preference for the local antecedent, as expected for Type 1B but not for Type 2B.

Table 8

Type 2 paired sample t-test results: LOC vs. LD contexts.

Type t-value p-value Cohen’s d

NS A 5.43 <.001 1.41
B 0.64 .521 0.16
CS A 0.41 .684 0.10
B 5.48 <.001 1.58

As shown above, responses from the CS group for the Type 1 empathic conditions were similar to those from the NS group, with no statistical difference between groups. Significant differences were found for the Type 2 logophoric conditions. First, the CS group did not seem to allow the LD interpretation for Type 2B sentences, which suggests that the interpretation is blocked by the presence of the first-person pronoun for the Chinese group, though not for the NS group. The second difference found was for the Type 2A conditions. The NS group showed preference for the LD interpretation of zibun, but the Chinese group did not.

4.4.2. Individual results

In order to see whether there are any individual differences, responses were examined per participant. Table 9 summarizes the number of NS participants who made each truth value judgment (True or False) for each condition. Participants included in the True row judged at least 3 out of 4 tokens True, and those in the False row at least 3 out of 4 tokens False. Those who chose True and False each twice appear in the Neither row. As Table 9 shows, there appears to be substantial individual variability in the responses from native speakers, except for Types 1B (LOC, LD) and 2A (LD). In the group results, mean percentages from the NS group were lower than expected. The individual results show that for a sentence where True judgments are predicted, many Japanese native speakers nevertheless chose False, especially for Types 1A (LD), 2A (LOC), and 2B (LOC, LD).

Table 9

Individual NS truth value judgments (n = 36).

Type Type 1 Type 2

Condition A B A B

Context LOC LD LOC LD LOC LD LOC LD

Predicted T T T F T T T T

TRUE 22/36
61.1%
17/36
47.2%
28/36
77.7%
1/36
2.7%
11/36
30.5%
30/36
83.3%
16/36
44.4%
24/36
66.7%
FALSE 6/36
16.7%
8/36
22.2%
3/36
8.3%
32/36
88.9%
18/36
50.0%
4/36
11.1%
11/36
30.6%
8/36
22.2%
Neither 8/36
22.2%
11/36
30.6%
5/36
13.9%
3/36
8.3%
7/36
19.4%
2/36
5.6%
9/36
25.0%
4/36
11.1%

The individual results from the CS group are shown in Table 10. As with the NS group, CS learners exhibit some variability in their responses, again with the exception of Types 1B (LOC, LD) and 2B (LD). Different patterns between the NS and CS groups emerge for Types 2A (LD) and 2B (LD), similar to our findings from the group results.

Table 10

Individual CS truth value judgments (n = 28).

Type Type 1 Type 2

Condition A B A B

Context LOC LD LOC LD LOC LD LOC LD

Predicted T T T F T T T T

TRUE 15/28
53.6%
11/28
39.3%
20/28
71.4%
1/28
3.6%
14/28
50.0%
15/28
53.6%
15/28
53.6%
2/28
7.1%
FALSE 6/28
21.4%
12/28
42.9%
1/28
3.6%
24/28
85.7%
6/28
21.4%
6/28
21.4%
4/28
14.3%
21/28
75.0%
Neither 7/28
25.0%
5/28
17.9%
7/28
25.0%
3/28
10.7%
8/28
28.6%
7/28
25.0%
9/28
32.1%
5/28
17.9%

5. Discussion

The aim of this study was to examine whether CS learners of Japanese are able to acquire target-like interpretations of the Japanese reflexive zibun. Recognizing their L1 knowledge of an LD reflexive, Chinese ziji, the focus was on whether they can acquire the following properties of zibun:

(16) a. Zibun can be bound by an empathic LD antecedent
  b. The blocking effect is present for empathic LD antecedents, not for logophoric LD antecedents.

The acquisition of (16a) can be assessed from results on the LD condition of Type 1A with an empathic LD subject. The CS group interpreted zibun with an empathic LD antecedent at a rate of 47.3% for Type 1A, while the NS group allowed an empathic LD antecedent at 58.3%.9 The results from both groups showed no statistical difference. Therefore, although the rate of LD interpretations was low, the results suggest that the CS learners allow binding between zibun and an empathic LD antecedent, showing that the acquisition of the property indicated in (16a) is possible.

As for (16b), results showed that the CS group did not allow LD binding of zibun in Types 1B and 2B conditions. This suggests that the blocking effect is present not only with logophoric LD antecedents but also with empathic LD antecedents, showing that the property indicated in (16b) has not been acquired.

The LD binding of zibun by an empathic antecedent can be acquired based on input such as (17), in which zibun is unambiguously bound by an empathic LD antecedent, Taro, as the meaning of the verb, kasu (lend) makes it pragmatically implausible to lend oneself money. This type of positive evidence potentially leads CS learners of Japanese to acquire that zibun allows non-logophoric LD antecedents.

(17) Taro1-wa [Hanako-ga zibun1-ni kasita] okane-o mudanisite-simatta.  
  Taro-TOP Hanako-NOM self-DAT lent money-ACC waste.up-ended.up  
  ‘Taro wasted all the money that Hanako lent to him.’
  (Kishida, 2011, p. 26)

Although CS learners appear to allow both empathic and logophoric LD antecedents for zibun, the application of the blocking effect seemed to remain non-target-like. As discussed in section 2, a perspective conflict causes the blocking effect in Chinese. The results from this study suggest that a perspective conflict between a first-person pronoun and an LD antecedent still appears to be operative in Chinese-Japanese ILGs. For both sentences in (18), the first-person pronoun watasi (I), appears in the embedded clause. This first-person pronoun enters into a perspective conflict with the matrix subject, Tosiko in (18a) and Tutomu in (18b), and neither of them can bind zibun as its antecedent. The results, therefore, suggest that the CS learners have failed to acquire that it is empathy, rather than perspective, that induces the blocking effect in Japanese.

    1. (18)
    1. a.
    1. Empathic LD antecedent
    1. Tosiko-ga
    2. Tosiko-NOM
    1. [watasi-ga
    2. I-NOM
    1. taikukan-de
    2. gym-at
    1. zibun-o
    2. self-ACC
    1. semeta
    2. criticized
    1. atode]
    2. after
    1. nakidasita.
    2. cried
    1. ‘Tosiko burst into tears after I criticized her in a supermarket.’
    1.  
    1. b.
    1. Logophoric LD antecedent
    1. Tutomu-ga
    2. Tutomu-NOM
    1. [watasi-ga
    2. I-NOM
    1. kafe-de
    2. café-at
    1. zibun-o
    2. self-ACC
    1. syookaisita-to]
    2. criticized-COMP
    1. itta.
    2. said
    1. ‘Tutomu said I introduced him/myself at a café.’

In order to acquire this property, learners must be exposed to sentences such as (18b) in contexts where the intended antecedent of zibun is the logophoric LD subject. Although positive evidence of this type is in principle available, it is likely to be extremely rare and more complex compared to examples like (17), as it has to occur with a first- or second-person pronoun in the embedded clause while zibun refers to the preceding logophoric LD antecedent. This may prevent CS learners from positing a blocking effect in Japanese logophoric sentences.

In addition, notions such as logophoricity and empathy are related to pragmatics. As argued by proponents of the Interface Hypothesis (IH) (e.g., Tsimpli & Sorace, 2006; Sorace & Filiaci, 2006), properties associated with external interfaces, such as the syntax-discourse interface, are problematic for L2 learners, even at the near-native level. Tsimpli & Sorace (2006) argue that external interfaces are more difficult because learners have to integrate pragmatic and contextual information into the language system, whereas internal interfaces only deal with formal properties of language. If the IH is on the right track, CS learners may have difficulties distinguishing the pragmatic notion of empathy from logophoricity, as used in the interpretation of zibun, or they may have difficulty acquiring a linguistic notion of empathy. In this study, however, whether the CS learners have mastered linguistic empathy in Japanese was not independently tested.10 Therefore, further investigation is necessary to determine whether or not the cause of CS learners’ inability to distinguish the two types of Japanese LD antecedents might lie with their problems with linguistic empathy.

A difference between the NS and CS groups was also found for the Type 2A condition. In this sentence type zibun has either a LOC or an LD subject as its antecedent. However, the NS group seemed to accept the LD antecedent, i.e. a logophoric individual, as the antecedent of zibun more easily, accepting the LD interpretation at 86%, but only accepting the LOC interpretation at 46%. The high acceptance rate of LD interpretations was expected, but the low acceptance rate of LOC interpretations was not, as the LOC interpretations should also be acceptable. We believe these results are an indication that Japanese speakers have a clear preference for an LD antecedent for zibun in logophoric sentences. This is likely because a logophoric individual in the sentence forces a pragmatic preference for zibun to refer back to that individual. The CS group, however, did not show such preference, accepting LOC antecedents at 57% and LD antecedents at 63%, with no statistical difference between them. These results, again, are in line with the IH and also with the results from Yoshimura et al. (2012) as L2 learners are less able than the NSs to integrate pragmatic information, i.e. logophoricity, into a sentence interpretation.

Lastly, unexpected variability in responses among NS participants was found for sentence types 1A and 2B. Both LOC and LD antecedents should be acceptable for both types, but acceptance rates were only about 50–60% for both interpretations. NSs did not show preference for one interpretation over another for these types of sentences; rather, they show that as a group, neither LOC nor LD interpretations appear to be fully available. One possible explanation for such variability might be attributable to problems with the test instrument. It could be that our TVJT, including contexts paired with test sentences and the instructions given to participants, was somehow inadequate. However, we find that unlikely since we were able to obtain target-like judgments from the NS group for Type 1B sentences, which tested the presence of the blocking effect. If the test materials themselves were flawed, we would also expect variable results from Type 1B sentences. Instead, we have concluded that response variability was most likely obtained due to ambiguity of the test sentences. Our original rationale for use of a TVJT was in the first place to examine separately the two interpretations for each ambiguous test sentence. However, the results might nevertheless reflect preferences for one interpretation over another. For example, for Type 1A in the individual results, 17 out of 36 (47.2%) NS participants accepted the empathic LD antecedent, but eight out of the 36 (22.2%) rejected it. This shows that there were some NS participants who consistently accepted LD antecedents, while others consistently rejected them. To be sure, a TVJT is indeed an appropriate task for testing ambiguous sentences, but in this experiment, contexts may not have been adequately biased to override individual participant preferences towards one interpretation of the ambiguous test sentences.

6. Conclusions

The present study investigated whether Chinese learners of Japanese acquire the distinction between two types of LD antecedents, empathic and logophoric, for the Japanese reflexive zibun. Although Chinese speakers permitted empathic LD antecedents for zibun, they were unable to distinguish the two types of LD antecedents with respect to the blocking effect. While in Japanese the blocking effect is present only for empathic LD antecedents, the results showed that the blocking effect is also present for logophoric LD antecedents among the CS learners. These results suggest that CS learners apply the blocking effect under the same conditions as in their L1, thus being unable to overcome the effects of L1 transfer. We proposed two possible explanations for their non-target-like interpretation of zibun. First, the positive evidence necessary to demonstrate that the blocking effect does not apply with logophoricity may be insufficient in Japanese input generally. Another possible reason we suggested was, following a claim put forward as part of the Interface Hypothesis, that Chinese speakers are unable to acquire the two types of LD antecedents because of their difficulty distinguishing two types of pragmatic conditions, empathy and logophoricity. Further investigation will be necessary to determine how linguistic empathy is interpreted and used by CS learners of Japanese.

Additional File

The Additional file for this article can be found as follows:

Appendix A

Test materials. DOI: https://doi.org/10.22599/jesla.14.s1