The Interpretation of Japanese Pronouns by L1

The interpretation of Japanese pronouns by L1 English and L1 Spanish speakers 1. Introduction This study investigates L2 acquisition of the interpretive properties of the Japanese pronoun kare ‘he’ in subject positions by L1 English and L1 Spanish speakers. The Overt Pronoun Constraint (OPC) (Montalbetti, 1984) states that overt pronouns cannot take bound variable interpretations in null subject languages, including Spanish and Japanese. This constraint is not operative in non-null subject languages, such as English. Most previous studies on the OPC were conducted on L2ers whose L1 is a non-null subject language (e.g., Kanno, 1997; Marsden, 1998; Pérez-Leroux & Glass, 1999; Gürel, 2002; Rothman & Iverson, 2007a, 2007b; Rothman, 2009). To my knowledge, no previous study has examined acquisition of the OPC by L2ers whose L1 is a null subject language. This novel study fills this gap by examining Japanese and Spanish L2‒L1 combination. The present study also contributes to the existing literature by clarifying coreferential interpretations of the subject kare, on which Kanno (1997) and Marsden (1998) diverged. This paper is structured as follows: Section 2 presents interpretive differences of pronouns in Spanish, Japanese, and English while Section 3 reviews previous L2 studies on the OPC in Japanese, then Section 4 suggests research questions and hypotheses followed by Section 5 which explains the experiment and findings and finally Section 6 discusses implications of the findings, and is followed by a conclusion. 2. Background 2.1 Coreferential and bound variable pronouns in English Pronouns are expressions that do not have descriptive content encoding a concept. The semantic content of pronouns is limited to basic features, including person, number, and gender (Panagiotidis, 1 2002; Büring, 2011). The use of pronouns includes three types, (i) deictic pronouns, (ii) coreferential pronouns, and (iii) bound variable pronouns, as in (1)-(3) (Evans, 1980; Büring, 2011). 2 (1) He’s up early. (2) Johni loves hisi mother. (deictic) (coreferential) (3) a. Every mani loves hisi mother. (bound variable) Every man (λx (x loves x’s mother)) b. Which boyi brought hisi bear? Which boy (λx (x bought x’s bear))? When pronouns are deictically used, as in (1), they refer to a salient object in the perceptual environment in which the conversation takes place. Therefore, deictic pronouns usually have their antecedents in the discourse rather than the sentence. By contrast, in the remaining two uses, pronouns have sentential antecedents. In (2), the pronoun his is coreferential 3 with the coindexed referential antecedent, John. Thus, coreferential pronouns are interpreted as coreferring to particular individuals; hence, they have fixed values. By contrast, bound variable pronouns do not have fixed values, as shown in (3). In (3a), the interpretation of his varies, depending on the choice of a man. In other words, the bound variable pronoun his is interpreted as a variable x bound by a λ operator. Similarly, in (3b), the interpretation of his varies, depending on the choice of a boy. Thus, when pronouns have quantificational antecedents, as in (3a), or wh-phrase antecedents, as in (3b), they are interpreted as variables bound by the antecedents. This interpretation is called a bound variable reading. In the following, I focus on the two uses of pronouns—those bound by quantified antecedents (i.e. bound variable pronouns) and those taking referential antecedents (i.e. coreferential pronouns) because null subject languages (e.g. Spanish and Japanese) and non-null subject languages (e.g. English) differ from each other in these two uses, as shown in 2.2. 2.2 Pronouns in null subject languages Unlike English, Spanish overt pronouns do not necessarily take a bound variable interpretation. In (4), for example, the overt pronoun él ‘he’ can not take nadie ‘nobody’ as its antecedent, while pro can. Thus, the distribution of null and overt pronouns is not in free variation in Spanish. When the antecedents are not quantified, this asymmetry between null and overt pronouns disappears, as in (5). In (5), both null and overt pronouns can have coreferential readings. Thus, the distribution of overt pronouns is more restricted than null pronouns in Spanish. From these observations, Montalbetti (1984) proposed the Overt Pronoun Constraint (OPC) in (6). (4) Quantified antecedent context 2 E-type pronouns, which are neither coreferential nor bound variables, are also considered as the fourth type in literature (e.g. Evans, 1980: Heim & Kratzer, 1998) but they are not investigated here. 3 Following Heim & Kratzer (1998), I use the term coreferential when two expressions refer to the same individual. 2 a. Nadiei sabe que él*i/j/proi vendra. Nobody know:3S that he/pro come:3S.Fut ‘Nobodyi knows that he*i/j /proi will come.’ b. Nadiei sabe que el profesor lo vigila a él *i/j/proi/j Nobody know:3S that the teacher HIM-CL watch-over:3S him/pro ‘Nobodyi knows that the teacher watches over him*i/j/proi/j.’ (Alonso-Ovalle & D’Introno, 2001) (5) Referential antecedent context a. Juani cree que éli/j/proi/j es inteligente. John believe:3S that he/pro is:3S intelligent ‘Johni believes that hei/j/proi/j is intelligent.’ b. Juani sabe que el profesor lo (Montalbetti, 1984) vigila a él i/j/proi/j. John know:3S that the teacher HIM-CL watch-over:3S him/pro ‘Johni knows that the teacher watches over himi/j/proi/j.’ (6) Overt pronouns may not take a bound variable interpretation when an overt/null alternation occurs. (In contrast, null pronouns may have a bound variable and a coreferential interpretation.) Montalbetti suggests that the OPC is also operative in Japanese. Japanese allows null pronouns (pro) and overt pronouns (kare ‘he’ and kanozyo ‘she’)4 and the distribution of overt pronouns is also restricted than null pronouns. (7) shows that Japanese overt pronouns in complement clauses cannot take quantified or wh-word antecedents, whereas null pronouns can, just like Spanish. The asymmetry between null and overt pronouns disappears when the antecedents are referential, as in (8). (7) Quantified antecedent context a. Darei-ga [kare*i/j-ga /proi/j kuruma-o katta Who-Nom he-Nom /pro car-Acc to] i-tta-no? bought that say-Pst-Q ‘Whoi said that he*i/j/proi/j bought a car?’ b. Daremoi-ga [Mary-ga kare*i/j-o / proi/j sitteiru to] i-tta. Everyone-Nom Mary-Nom he-Acc /pro know that say-Pst ‘Everyonei said that Mary knew him*i/j/proi/j.’ (8) Referential antecedent context a. Taroi-wa [Hanako-ga karei/j-o/proi/j kadaihyooka siteiru] to omo-tta. Taro-Top Hanako-Nom he-Acc/pro overestimate doing 4 that think-Pst Kare and kanozyo are treated as N-pronouns here, following Noguchi (1997) and Déchaine & Wiltschko (2002), although a variety of proposals regarding the status of kare and kanozyo have been developed, such as demonstratives (Hoji, 1991) and epithets (Yashima, in press). 3 ‘Taroi thought that Hanako overestimated himi/j/proi/j.’ (Mihara & Hiraiwa, 2006) b. Taroi-wa [Mary-ga karei/j-o/proi/j sitteiru to] i-tta. Everyone-Nom Mary-Nom he-Acc /pro know that say-Pst ‘Taroi said that Mary knew himi/j/proi/j.’ Note that the distribution and interpretation of Japanese pronouns do not always parallel those of Spanish pronouns. As seen in (6), the Spanish OPC is a conditional constraint, and the restriction on overt pronouns is only observed in syntactic positions where a null/overt alternation potentially occurs. In other words, Spanish overt pronouns can take a bound variable reading in Focus, PPs, and possessives, where null pronouns are not allowed (Montalbetti, 1984; Alonso-Ovalle & D’Introno, 2001). (9a) gives an example of a Spanish overt pronoun in Focus. Unlike Spanish overt pronouns, Japanese overt pronouns consistently cannot have a bound variable interpretation even when no null/overt alternation occurs. As shown in (9b), the Japanese overt pronoun kare in Focus alternates with zibun ‘self’, not pro.5 Nevertheless, kare still cannot have a bound variable interpretation. Thus, Spanish and Japanese overt pronouns in Focus, PPs, and possessives have different interpretations. Apart from these positions, however, Spanish and Japanese overt pronouns have the same interpretation. The interpretive differences between English and Spanish/Japanese which have been discussed so far are summarized below in Table 1. (9) Focus a. Spanish Ningún estudiantei piensa No student que (sólo) ÉLi/j pasó el examen. believe:3S that (only) he pass:3S.Pst the exam ‘No studenti believes that only hei/j passed the exam.’ b. Japanese Daremoi-ga [kare*i/j-dake-ga /zibuni-dake-ga siken-ni pasu suru to] omotteiru Everyone-Nom he-only-Nom/self-only-Nom exam-in pass do that think ‘Everyonei thinks that only he*i/j /selfi will pass the exam.’ 5 When he-only-Nom is replaced with pro in (9b), the meaning of Focus is lost. The Focus particle dake ‘only’ needs to attach to an overt form. 4 Table 1. Interpretation of pronouns in English, Spanish and Japanese language antecedents Pronouns Examples Bound Interpretation Corefential interpretation English Spanish/Japanese Ref Qua Referential Quantified Overt Overt Null Overt Null he él/kare pro él/kare pro - Yes - - No(/Yes6) Yes Yes - Yes Yes - - 3. Previous studies Kanno (1997) investigated whether or not the OPC is operative in the grammars of L1 English speakers of L2 Japanese. She compared English speakers with low intermediate levels of proficiency to native Japanese speakers in a written interpretation task. The participants read bi-clausal sentences which contained quantificational matrix subjects (dareka ‘someone’ and dare ‘who’) and embedded (null or overt) pronominal subjects, as in (10). Then they chose appropriate antecedents for the pronoun, from three options; (a) same as the matrix subject (i.e. bound-only interpretation), (b) another person (i.e. disjoint-only interpretation), or (c) both (a) and (b) (i.e. both bound and disjoint interpretation). She found that the L2ers observed the OPC, just like the native Japanese speakers. (10) Dareka-ga [ kare-ga/pro Suuzan-o Someone-Nom sitteiru to ] i-tteimasi-ta yo. he-Nom/pro Susan-Acc know that say-Prg-Pst EMPH ‘Somene was saying that he/pro knows Susan.’ Q. Dare-ga Suuzan-o sitteiru n deshoo ka? Who-Nom Susan-Acc know that suppose Q ‘Who do you suppose knows Susan?’ A. (a) Same as someone (b) another person Marsden (1998) replicated Kanno, using the same sentence structure and methodology. She also used the same L1−L2 combination but tested L2ers with broader proficiency levels than Kanno. She confirmed the finding in Kanno, suggesting that L1 English speakers of L2 Japanese observed the OPC at early stages. Marsden however found that native Japanese speakers (n=11) unexpectedly chose coreferential interpretations for overt pronouns only 11.5% of the time. 7 This low acceptance rate of coreferential kare is not compatible with the view in Montalbetti (1984), who assumes a simpler view 6 Spanish overt pronouns exceptionally can be bound by quantified antecedents in syntactic positions where null pronouns do not occur. 7 Similar to Marsden, Yamada (2005) also reports that Japanese monolinguals (n=6) accepted the coreferential interpretation of kare only 9.4% of the time. In contrast, in Kanno, native Japanese speakers (n=20) accepted it 47% and this was significantly higher than the acceptance of the bound variable kare (2%). 5 according to which null subject languages and non-null subject languages differ only in terms of the bound variable interpretations of overt pronouns. Moreover, the low acceptance rate of the coreferential kare in Marsden indicates that kare may have mimic the Turkish overt pronoun o, which allows neither coreferential nor bound variable interpretations in sentences like (7) and (8) (Gürel, 2002). The low acceptance rate of the coreferential kare in Marsden is potentially attributable to the task effect. In her experiments, sentences like (10) were presented without context to the participants, and they chose appropriate antecedents from three options: (i) the matrix subject, (ii) another person, and (iii) both (i) and (ii). This task may reflect the participants’ preferences and overlook less preferable options. In other words, not choosing some option does not mean that it is ungrammatical. In order to see the task effect, the experiment in the present study employs two tasks. Another potential factor which causes the low acceptance of coreferential kare is verb meaning. Kuno (1972) and Kanzaki (1994) suggest that complement clauses following some verbs, such as say, think, and believe, tend to directly express the speaker’s feelings; therefore, self or its null form rather than kare is preferred. To test how far the verb meaning is involved in interpreting kare, kare in both reported speech and nonreported speech are tested in the present study. 4. Research question and hypotheses (11) Research questions Q1: Is the OPC truly operative in Japanese as Montalbetti (1984) suggests? Q2: Is the OPC acquirable by L1 English and L1 Spanish speakers in the same way? The first question arises from the contradictory results on interpreting kare with referential antecedents, as we have seen in Section 3. In Kanno (1997), the native Japanese speakers accepted the coreferential kare more frequently than the bound variable kare, as Montalbetti suggests, whereas, they did not in Marsden (1998). The second research question addresses the effect of L1s on acquisition of the OPC. The Full Transfer/Full Access Hypothesis (FT/FA) (Schwartz & Sprouse 1994; 1996) proposes that (a) the initial state of the L2 grammar is the end state of the L1 grammar and that (b) the L1 property is acquirable by means of UG. Considering the differences between Japanese, English and Spanish, the following predictions are made. (12) Predictions P1: Assuming that the OPC is operative in Japanese, as Montalbetti suggests, native Japanese speakers will accept the coreferential kare significantly more often than the bound variable kare. P2: Adopting the FT/FA model, L1 Spanish speakers will initially outperform L1 English speakers in interpreting kare. At later stages, both L1 Spanish and L1 English speakers will have correct interpretations of kare. 6 5. Experiment 5.1 Participants 60 L2 Japanese speakers whose L1 was either English (n=30, 15 advanced, 15 intermediate) or Spanish (n=30, 14 advanced, 16 intermediate) were participated in the experiment. The L2ers’ proficiency was measured by a Japanese language proficiency test adapted from Umeda (2008). The L2ers’ use of Japanese null pronouns was confirmed in a pre-test in which they translated a dialog from English or Spanish into Japanese. Table 2 presents the results on the Japanese proficiency test and the pre-test. 15 native Japanese speakers also participated in the experiment as the control group. Table 2. L2ers’ proficiency and results on the pre-test8 group Proficiency test (%) mean range EA (n=15) EI (n=15) SA (n=14) SI (n=16) Drop of pronouns (%) object total subject range 80 71-91 70 45 62 25-100 52 37-66 61 43 56 25-92 78 69-97 93 70 85 25-100 50 40-63 87 75 83 67-100 5.2 Methodology 5.2.1 The CJT Two tasks were employed. The first task was a coreference judgment task (CJT) adapted from Kanno (1997) and Marsden (1998). In this task, the participants were presented with Japanese sentences, followed by questions, as in (13). They were asked to choose potential antecedents for the pronouns, kare and pro. (13) Type1 (Quantified antecedents) Minnai -ga kinoo kare*i/j-ga/proi konpyuutaa-o tukatta to everyone-Nom yesterday he-Nom/pro computer-Acc used itteimasita that was saying ‘Everyonei was saying that he*i/j/proi used a computer yesterday.’ Q. Darega konpyuutaa-o tukatta nodeshooka? ‘Who used a computer?’ A. (a) Minna to onaji ‘Same as everyone’ (b) Betsu no hito ‘Another person’ (c) Wakaranai ‘I don’t know’ 8 In the table, EA, EI, SA, and SI represent English Advanced, English Intermediate, Spanish Advanced, and Spanish Intermediate, respectively. 7 In answering the questions, the participants were instructed to choose all potential antecedents, as in Kanno (1997) and Marsden (1998). In (13), the native Japanese participants were expected to choose only (b) (i.e. the disjoint-only interpretation) as the antecedent for kare. They were expected to choose both (a) and (b) as the antecedents for pro (i.e. the coreferential and disjoint interpretations). Participants were instructed to choose (c) ‘I don’t know’ when they could not understand the sentence because of lack of vocabulary or being unfamiliar with the sentence structure. The test sentences consisted of 3 types and each type consisted of 8 sentences (4 sentences for overt/null pronouns). The example in (13) represents Type 1, which contained a quantified antecedent (dareka ‘someone’ or minna ‘everyone’) as the matrix subject, a speech verb (itteimasita ‘was saying’) as the matrix verb, and a pronoun as the embedded subject. All test sentences of Type 1 were identical with those in Marsden (1998) except for the proper names used and omission of a sentence final particle. This was done in order to be able to make a direct comparison of the results. Example (14) represents Type 2, which contained a referential antecedent as the matrix subject. The remaining elements in the sentences in Type 2 were same as those in Type 1. Assuming that overt pronouns can take referential antecedents in Japanese, as Montalbetti suggests, the native Japanese participants were expected to choose both (a) and (b) in (14). (14) Type 2 (Referential antecedents) Hayasii-san-wa atode karei/j-ga/proi denwa-o kakeru to itteimasita Hayasi-Mr-Top later he-Nom/pro telephone-Acc dial that was saying ‘Mr. Hayashii was saying that hei/j/proi would call later.’ Q. Darega denwa surunode shooka? ‘Who would call?’ A. (a) Hayashi-san ‘Mr. Hayasi’ (b) Hayashi-san toha betuno hito ‘Someone other than Mr. Hayashi’ (c) Wakaranai ‘I don’t know’ The example in (15) represents Type 3. This type was included to test whether native Japanese speakers interpret pronouns with referential antecedents in non-reported speech differently from reported speech (as tested in Type 2). Type 2 and Type 3 contained the same referential antecedents but different verbs. In Type 2, the verbs were ‘was saying’ followed by a complement clause to ‘that’ as in (14). In Type 3, other verbs (hiteisimasita ‘denied’ as in (15), kigatukimasita ‘realized’, and wasureteimasita ‘forgot’) were used. (15) Type 3 (Non-reported speech) Tanakai-san-wa karei/j-ga/proi shatyoo-ni naru to-iu uwasa-o hiteisimasita Tanaka-Mr-Top he-Nom/pro president-Dat become that-saying rumor-Acc denied. ‘Mr. Tanakai denied the rumor that hei/j/proi becomes a president.’ Q. Uwasaniyoruto, darega shatyoo ni naru nodeshooka? ‘According to the rumor, who will become a president?’ 8 ‘Mr. Tanaka’ A. (a) Tanaka-san (b) Tanaka-san toha betuno hito ‘Someone other than Mr. Tanaka’ (c) Wakaranai ‘I don’t know’ 5.2.2 The TVJ The second task was a truth value judgment task (TVJ) in which the participants judged whether the given Japanese sentence matched the context illustrated in a picture by choosing ‘true,’ ‘false,’ or ‘I don’t know.’ The test sentences in the TVJ were exactly same as the CJT. An example of the sentence and a bound context is given in (16). This stimulus examined whether the participants allowed kare and pro to be bound by a quantifier, minna ‘everyone’. In (16), the picture illustrates a situation in which the three men are individually talking about their use of computers. If the participants choose ‘true’, they assume a bound interpretation. Alternatively, if they choose ‘false,’ they assume a disjoint interpretation. For the test sentence in (16), native Japanese speakers were expected to choose ‘false’ when it included kare as the embedded subject. They were expected to choose ‘true’ when it included pro as the embedded subject. The remaining two conditions were tested in the same way as Type 1. (16) Type1-Disjoint context (Quantified antecedents) Minnai-ga kinoo kare*i/j-ga/proi konpyuutaa-o tukatta to itteimasita Everyone-Nom yesterday he-Nom/pro computer-Acc used that was saying ‘Everyone was saying that he*i/j/proi used a computer yesterday.’ Figure 1. The bound context tadasii ‘ture’/ matigai ‘false’/wakaranai ‘I don’t know’ 5.3 The CJT results Figures 2 and 3 show group means of the acceptance rates (%) of the bound variable interpretation, namely, the total of the bound only interpretation and the bound/disjoint interpretation, of kare and pro. The acceptance rates indicate the proportion of times the participants chose the interpretations of each pronoun. An ANOVA was run on the mean scores (out of 4) of the interpretations, not the mean percentages. It was predicted that the controls would accept the bound variable interpretation of pro but not kare. This prediction was verified. Figure 2 shows that the controls chose the bound variable interpretation of kare only 24% of the time, whereas they chose the bound variable interpretation of 9 pro 98% of the time. These results show the OPC is operative in the Japanese language. Moreover, as shown in Figure 3, the controls chose the coreferential interpretation of kare 39% of the time, which was significantly higher than the bound variable interpretation of kare, 24% of the time (t(14)=2.81, p<.05). This suggests that the prohibition of binding kare is not simply an across the board prohibition but applies only to coreferential kare, as Montalbetti (1984) suggests. Regarding the L2ers, it was predicted that the L1 English speakers would initially accept the bound variable kare more frequently than the controls. In contrast, the L1 Spanish speakers would not accept the bound variable kare, just like the controls; as a result, the L1 Spanish group would outperform the L1 English group. This prediction was partially supported in this task. The L1 Spanish group did not outperform the L1 English group with respect to the acceptance rate of the bound variable kare. In Figure 2, both the intermediate English group (i.e. the EI group) and the intermediate Spanish group (i.e. the SI group) chose the bound variable kare to the same extent, 40% of the time. A two-way ANOVA showed no significant main effect of L1 (F(1,56)=0.135, p=0.715>.05) but a significant main effect of proficiency (F(1,56)=5.20, p<.05) on the choice of the bound variable kare by the L2ers. However, the Spanish speakers outperformed the English speakers with respect to making a distinction between quantified and referential antecedents for kare. As shown in Figure 3, the SI group chose the bound variable interpretation of kare significantly less frequently than the coreferential interpretation of kare (40% vs. 65%, t(15)=2.9, p<.05). In contrast, the EI group did not make a distinction between the antecedents for kare, accepting both interpretations of kare to the same extent (bound variable kare 40% vs. coreferential kare 43%, no significant difference: t(14)=.61, p=.55). This suggests that the OPC was not fully operative in the EI group’s grammar. In contrast to the EI group, the advanced English group (i.e. the EA group) made a distinction between the antecedents in interpreting kare. They chose the bound variable kare significantly less frequently than coreferential kare (20% vs 43%, t(14)=2.86, p<.05), just like the controls. Figure 2. Acceptance of the bound variable kare and pro in the CJT (%) 98 98 95 88 100 94 80 Q-kare 60 40 40 24 20 control EA 40 16 20 0 EI SA 10 SI Q-pro Figure 3. Acceptance of the bound variable and coreferential and kare in the CJT (%) 80 65 60 43 39 40 24 40 43 20 36 40 Rerefential antecedents 16 20 Quantified antecedents 0 control EA EI SA SI Figure 4 presents the acceptance rates of the coreferential kare in reported and non-reported speech. All L2 group accepted kare in reported and non-reported speech to the same extent (e.g. the controls 39% vs 54%, t(13)=1.71, p=.11). Figure 4. Acceptance of kare in reported and non-reported speech in the CJT (%) 100 80 60 65 54 39 43 control EA 40 37 43 34 36 reported speech 55 non-reported speech 29 20 0 EI SA SI 5.4 The TVJ results In the TVJ, dareka ‘someone’ (n=2) and minna ‘everyone’ (n=2) were used as the quantified antecedents, just like the CJT. However, the results on ‘someone’ were excluded from the analysis because the pictures for ‘someone’ failed to provide appropriate context. 9 Consequently, only the results on ‘someone’ (n=2) are reported in this section. 9 The reason for using ‘someone’ was to make a direct comparison with Kanno (1997) and Marsden (1998). However, the controls unexpectedly accepted the bound variable kare when the antecedent was ‘someone’ as often as 58% of the time. This acceptance rate was significantly higher than when the antecedent was ‘everyone’, when it was only accepted 17% of the time (t(12)=3.82, p<.01). Although both someone and everyone are quantificational and bind pronouns, the former is indefinite while the latter is not. It has been pointed out that indefinites ambiguously permit either coreferential or existential quantifier interpretations (Fodor & Sag, 1982). 11 The results on the TVJ were overall similar to the CJT. As shown in Figures 5 and 6, the controls chose the bound variable interpretation of kare 17% of the time, which was significantly less frequently than the bound variable interpretation of pro (100%, t(14)=8.92, p<.001) and coreferential interpretation of kare (64%, t(14)=4.16, p<.05). These results suggest that the OPC is operative in Japanese, as Montalbetti suggests. Figure 7 shows that the coreferential interpretation of kare in reported and non-reported speech did not significantly differ (64% vs. 68%, t(14)=.70, p=.50), just like the results on the CJT. Regarding the L2ers, the EI group accepted the bound variable kare significantly more often than the controls (t(25)=2.09, p<.05), whereas the remaining L2 groups, including the SI group, did not differ from the controls (SI: t(29)=1.0 p=.325). Moreover, the EI group did not make a distinction between the bound variable kare and the coreferential kare, as shown in Figure 6 (50% vs. 47%, t(14)=0.52, p=.61). In contrast, the SI group almost made a distinction (31% vs. 55%, t(15)=2.08, p=.055). These results suggest that the English group with lower proficiency did not have knowledge of the OPC due to L1 transfer although they acquire it as their proficiency improved. Figure 5. Acceptance of the bound variable interpretation of kare and pro in the TVJ (%) 100 100 97 90 100 88 80 50 60 40 23 17 20 Q-kare 21 31 Q-pro 0 control EA EI SA SI Figure 6. Acceptance of the bound variable and coreferential and kare in the TVJ (%) 80 64 65 40 20 55 5047 60 17 55 31 23 21 0 control EA EI SA 12 SI Qunatified antecedents Referential antecedents Figure 7. Acceptance of kare in reported and non-reported speech in the TVJ (%) 100 80 6468 6565 4750 60 5558 67 55 40 Reported speech non-reported speech 20 0 control EA EI SA SI 6. Discussion In Section 4, the following predictions were made. P1: Assuming that the OPC is operative in Japanese, as Montalbetti suggests, native Japanese speakers will accept the coreferential kare significantly more often than the bound variable kare. P2: Adopting the FT/FA model, the L1 Spanish speakers will initially outperform the L1 English speakers in interpreting kare. At later stages, both L1 Spanish and L1 English speakers will have correct interpretations of kare. This section will discuss whether or not these predictions were supported. 6.1. P1 was supported This study examined potential variation among native Japanese speakers in interpreting coreferential kare. The motivation for this comes from the contradictory results reported in previous studies. In Kanno (1997), the controls (native Japanese students at the University of Hawaii) chose the coreferential interpretation (i.e. ‘coreferential-only’ or ‘both coreferential and disjoint’ interpretation) of kare 47% of the time. In contrast, in Marsden (1998), the controls (native Japanese speakers in the UK) chose that same interpretation only 11.5% of the time. The results in Marsden (1998) suggest that kare must have a disjoint interpretation irrespective of antecedents, just like the Turkish o. In other words, this does not seem to be an OPC effect but something more general. In the present study, the controls accepted the coreferential kare 39% of the time in the CJT. This result was more in line with the findings in Kanno (1997) rather than in Marsden (1998). In the TVJ, the controls in the present study gave a ‘true’ response to the coreferential interpretation of kare 64% of the time. This result from the TVJ can be seen as evidence that kare indeed allows a coreferential interpretation. It also shows that the Japanese kare differs from the Turkish o, which allows neither coreferential nor bound interpretations. Thus, the present study fails to replicate the results in Marsden (1998). 13 In order to clarify why coreferential kare was not accepted 100% of the time even in Kanno (1997), this study also tested coreferential kare in non-reported speech. If the verb meaning or the sentence structure is a crucial reason why coreferential kare in reported speech is not fully accepted, as Kuno (1972) and Kanzaki (1994) suggest, the controls should accept coreferential kare in non-reported speech more than reported speech. However, this prediction was not confirmed by the results. Although the controls accepted a coreferential interpretation of kare in non-reported speech more frequently than reported speech, the difference was not statistically significant (i.e. t (13) =1.58, p=.14 in the CJT, t (14) =0.70, p=.50 in the TVJ). In other words, the OPC is operative in the same way, irrespective of the verb meaning. 6.2. P2 was supported The FT/FA predicts that L1 Spanish speakers should correctly reject kare as a bound variable, like the controls, from early stages. They should transfer their L1, Spanish, in which the OPC holds, on the L2. In contrast, L1 English speakers should show a development of their knowledge of the OPC. The L1 English speakers with lower proficiency would wrongly accept the bound variable interpretation of kare, transferring from their L1, where overt pronouns can take a bound variable interpretation. It follows that the SI group should outperform the EI group in interpreting kare. This prediction was verified. The EI group accepted the bound variable interpretation of kare significantly more often than the controls, whereas the SI group was target-like in the TVJ. Moreover, the EI group was not sensitive to the referential/quantified asymmetry in interpreting overt pronouns; while in contrast, the SI group was sensitive to the asymmetry in the two tasks. These results suggest that the EI group did not fully acquire the knowledge of the OPC, while the SI group did. Thus, the SI group performed better than the EI group, supporting the FT/FA. The results also show the development of the knowledge of the OPC in L1 English speakers’ grammar, confirming the FT/FA. The FT/FA predicts that once L1 English speakers acquire the fact that Japanese allows null pronouns, the OPC takes effect and they should successfully reject kare as a bound variable. In the experiment, the EA group consistently had the same knowledge of pronouns as the controls, suggesting that they had successfully acquired the OPC. Thus, the applicability of the FT/FA model was confirmed in the domain of anaphoric use of pronouns by L1 English speakers. This result is also compatible with findings reported in previous studies (Kanno, 1997; Marsden, 1998) although these studies found that the OPC was operative in L2 grammar at earlier stages than the present study. In these studies, the L2 grammars observed the OPC before attaining intermediate proficiency levels. The implication of these results is that UG is operative in L2 acquisition. The interpretive differences of null and overt pronouns are not taught in language courses. In addition, it is unlikely that L2ers obtain negative evidence regarding OPC effects in naturalistic L2 input. Moreover, Kanno (1997) confirmed that OPC effects do not hold in English by investigating native English speakers’ interpretations of English pronouns. In order to solve this under determination problem, it is reasonable to assume that OPC effects are acquirable by means of UG. 14 One of the unexpected results in this study is that the SI group and the EI group did not show a robust difference in their performance. They did not differ from each other when the group means of the acceptance rates of the bound variable kare were compared in the CJT, as shown in Figure 2. Robust differences were found when the quantified/referential antecedent asymmetry was considered. The FT/FA suggests that the initial state of L2 grammar is the end state of L1 grammar, in addition, the OPC holds in Spanish but not in English. Therefore, more robust differences between the SI and EI groups were expected to be found. The two L2 groups performed similarly presumably because most of the EI group were residents of Japan and already had ample naturalistic input which indicates null subjects even though their proficiency levels were not high but intermediate. If intermediate L2ers who have little naturalistic input are compared, we would expect to see stronger L1 transfer effects. Another factor that could account for our findings is transfer from English to Japanese. Most of the L1 Spanish participants in this study spoke English as L2 (or L3) though they were not Spanish-English bilinguals. Therefore, if they used their knowledge of English in acquiring Japanese, it is not surprising that the SI group’s performance is similar to the EI group’s performance. 7. Conclusion This study investigates the interpretation of the Japanese pronoun (kare ‘he’) by adult L1 English and L1 Spanish speakers of L2 Japanese to clarify two points; (i) whether the OPC is truly operative in Japanese, as Montalbetti (1984) suggests, and (ii) whether L1 Spanish speakers initially have advantages over L1 English speakers in acquiring the OPC in Japanese. The first point arises from the contradictory results on interpreting kare with referential antecedents. In Kanno (1997), the native Japanese speakers accepted the coreferential kare more frequently than the bound variable kare, consistent with Montalbetti but not with other previous L2 studies. The second point addresses the effect of L1s on acquisition of the OPC. The FT/FA model predicts that L1 English speakers of L2 Japanese would initially allow a bound variable interpretation of kare, transferring from their L1s. In contrast, L1 Spanish speakers of L2 Japanese would correctly disallow a bound variable interpretation of kare from the beginning. It was found that native Japanese speakers accepted the bound variable interpretation of kare less frequently than the coreferential interpretation of kare or the bound variable interpretation of pro. This suggests that the OPC is operative in Japanese, as Montalbetti suggests. It was also found that L1 Spanish speakers initially have advantages over L1 English speakers in acquiring the OPC. The intermediate L1 English group accepted the bound variable kare more often than the native Japanese speakers while the intermediate L1 Spanish group did not in the TVJ. Moreover, the intermediate L1 English group was not sensitive to the referential/quantified antecedent asymmetry in interpreting kare while the intermediate L1 Spanish group showed sensitivity in the CJT and TVJ. These differences are attributable to their L1s, English, which does not observe the OPC, and Spanish, which does, just like Japanese. In contrast to the intermediate L1 English group, the advanced L1 English group showed evidence of a target-like grammar, suggesting the OPC in their grammars. Given that the OPC is 15 underdetermined in input, these results suggest that Universal Grammar (UG) is operative in L2 acquisition. 16