Acquisition of empathy in child Japanese

ABSTRACT

This article investigates the acquisition of empathy verbs in child Japanese, focusing on verbs of giving/receiving: age-ru ‘give,’ kure-ru ‘give,’ and mora(w)-u ‘receive.’ These verbs are distinguished by which argument the speaker empathizes with when describing an event. For age-ru ‘give,’ the speaker empathizes with the subject (the giver); for kure-ru ‘give,’ the speaker empathizes with a non-subject (the recipient), and for mora(w)-u ‘receive,’ the speaker empathizes with the subject (the recipient). Using two diagnostics for empathy (alignment of first person with empathy loci; empathy loci being preferred antecedents in reflexive binding), 4- to 6-year-old children were tested. Our experiments show the following two findings: (i) children found kure-ru as most challenging, partially contradicting previous research; (ii) some children as young as age 4 have fully acquired the empathy-encoding properties of these verbs despite the speaker’s empathy being unobservable in the input. We discuss the challenges that kure-ru poses for children in light of the potential learnability problem that these empathy verbs pose.

View correction statement:

Correction

Acknowledgments

We would like to sincerely thank members of Language Acquisition Research Group at University of Hawaiʻi at Mānoa, members of Tokyo Psycholinguistics Lab, and audience at BUCLD 44 and GALANA 9 for valuable feedback. We are also grateful to all of the children and adults who participated in our study.

Data availability statement

The data that support the findings of this study are openly available in “figshare” at https://figshare.com/s/1a72b4cff5f747bf5f7c.

Supplemental Material

Supplemental data for this article can be accessed on the publisher’s website.

Disclosure statement

No potential competing interest was reported by the authors.

Notes

¹ The term point of view is used in many different ways in the literature, including how we define it here in this article. However, this usage is different from how it is traditionally used in the Japanese literature on empathy. In that body of work, best exemplified by the work of Kuno, point of view is often interchangeable with empathy. We make this distinction between point of view and empathy for the sake of exposition.

² It is worth noting that while Kuno (1987) judges sentences like (3) acceptable, there is disagreement on whether sentences with kure-ru ‘give’ are acceptable with a third-person empathy locus (as in 3). However, this controversy does not affect our experiments, and thus we will not discuss this issue any further.

³ Kuno (1987) defines age-ru ‘give’ as a semi-honorific verb, and used yar-u ‘give’ instead as an informal form that is parallel to kure-ru ‘give’ and mora(w)-u ‘receive.’ However, yar-u ‘give’ sounds rude in colloquial speech, and it is inappropriate for children, thus we consistently use age-ru ‘give’ (a far more child-appropriate verb) in this article.

⁴ Kuno (1987) argues that the empathy associated with age-ru as a main verb is not as strong as age-ru as a supporting verb (henceforth, a benefactive verb), but we will not discuss this difference any further because our experiments only used age-ru as a supporting verb.

⁵ Umeda et al. (2017) tested the L2 acquisition of two types of long-distance reflexive, zibun (empathic zibun and logophoric zibun) by Chinese L2ers of Japanese. Our second experiment adopted Umeda et al.’s (2017) sentence types of the empathic zibun (atode ‘after’-clauses and kae-ru ‘go.home’ as a matrix verb). See Umeda et al. (2017) for detailed discussions.

⁶ The data shown here are combined results of those verbs as simple verbs and benefactive verbs. See Horiguchi (1979) for separate results of simple verbs and benefactive verbs.

⁷ More specifically, overall, most of the errors were (incorrect) production of age-ru when the stimulus sentence was kure-ru, but younger children (4-year-olds) sometimes erroneously produced kure-ru when the stimulus sentence was age-ru and when the verb was used as a benefactive form (but not a simple form).

⁸ Ishiguro (1985) tested both canonical and scrambled word order sentences with age-ru, kure-ru, and mora(w)-u, and the results are far too complex for a discussion here. Moreover, those complexities are not relevant to the current article.

⁹ Okabe’s (2005, 2011) work is far deeper than what we can cover in this article. Okabe (2005) showed that 4- to 6-year-olds’ performance on mora(w)-u ‘receive’ improved when a source argument (i.e., a giver argument in (7)) was kara ‘from’-marked, as opposed to being ni- ‘DAT’-marked. Furthermore, Okabe (2011) showed that 4- to 6-year-olds performed well on mora(w)-u even with a ni-marked source argument when it is an indirect benefactive sentence (i.e., the subject argument of mora(w)-u is not an actual recipient of the accusative object but a recipient of a benefit), but not a direct benefactive sentence. None of this detracts from our point, which is that comprehension studies appear to show (contrary to the results of production studies reviewed earlier) that the locus of difficulty for children is mora(w)-u.

¹⁰ We found transcriptions that seem to be duplicated in the MiiPro corpus because there were some pairs of utterances that appeared in exactly the same contexts but were assigned different line numbers in the corpus. We excluded those seemingly duplicated utterances by hand.

¹¹ Japanese has a verbal noun, choodai, which requires the speaker to be a non-subject recipient argument like kure-ru. Choodai is frequently used in imperative forms, such as omotya choodai! ‘Give me the toy!. We do not report the input frequency of choodai above because choodai and kure-ru are different verbs, and we cannot assume that children who know kure-ru also know choodai or vice versa. For interested readers, Table 2a shows the input frequency of kure-ru and choodai:

¹² Here is more detailed explanation of why we did not include mora(w)-u in the second round of Experiment 1. In the second round, we added another task (after the TVJT, and so the additional task did not impact the results of the TVJT), which is not reported in this article. This second task was the Dimensional Change Card Sort (DCCS) task (Zelazo 2006). This task measures how quickly the children can switch their perspective from one dimension to another, and we examined whether children’s ability to switch perspective in the DCCS task correlates with their performance with age-ru and kure-ru. This additional experimentation meant we needed to streamline the TVJT in some way, and we did so by excluding items with mora(w)-u (because our crucial interest was the difference between age-ru and kure-ru). The results of the correlation between DCCS and TVJT are reported in Ohba & Deen (2020).

¹³ We ran a simple logistic regression model with Accuracy (correct vs. incorrect) as a response variable and with Verb Type (age-ru vs. kure-ru) and Experimental Session (with mora(w)-u or without mora(w)-u) as predictor variables.

¹⁴ One of the reviewers asked why the character uttering the test sentences was not varied between the one speaking first and second because this might have influenced the results. As shown in the following, the experiment is a complex one, and adding this as an additional factor would have resulted in unnecessary complexity. We think that the nonvariability of the speakers has little or no effect and does not explain our main finding (that young children performed well with age-ru and mora(w)-u but not with kure-ru, discussed in more detail later), since the same method was used for test sentences involving all three verbs.

¹⁵ Seven children were excluded because they answered “true” to at least one false practice item. We also excluded two children who answered “false” to a true practice item. Ten children were also excluded because they did not answer correctly to at least one comprehension question given before a test sentence. When the child could not satisfy at least aforementioned one criterion, we did not test the child any further and finished the session. Thus the large number of exclusions is partly due to these strict exclusion criteria. Moreover, the remaining 41 children performed perfectly with the practice items (100%, 123/123) and almost perfectly with the filler items (96.7%, 238/246).

¹⁶ We used different stories for each test sentence in the actual experiment. See Supplementary Appendix A.

¹⁷ One of the reviewers suggested a possibility that the children ignored the benefactive verbs at the end of the sentences and only paid attention to the main verb, ka(w)-u ‘buy.’ This suggestion can explain why some of the children incorrectly interpreted kure-ru with the first person subject, since without kure-ru, it is most natural to assign the first person to the null subject with the verb of buying. However, it cannot explain why even the 4-year-old children correctly comprehended mora(w)-u ‘receive.’ That is, without mora(w)-u, the subject is the agent of the buying event, while with mora(w)-u, the theta-role of the subject switches to the recipient of the buying event. Therefore, if the children did not pay attention to the benefactive verbs at all, they should not be able to answer correctly with the sentences containing mora(w)-u. We therefore think that children did not ignore the benefactive verbs attached to the main verb.

¹⁸ Okabe’s (2008) Experiment IV reported a “slight” preference toward local antecedents by 4- to 6-year-olds. However, Orita et al. (2021) point out that Okabe’s (2008) TVJT did not satisfy the condition of plausible deniability. Adding plausible deniability, Orita et al. show that children chose the LD antecedent for zibun 20.8% of the time with LD-true stories, while they chose the local antecedent for zibun 66.7% of the time with local-true stories. Putting this all together, we conclude that Japanese children (like children acquiring other languages) do indeed have a locality preference.

¹⁹ But there are also studies that showed that children preferred LD antecedents in specific conditions (Hyams & Sigurjónsdóttir1990 and Sigurjónsdóttir & Hyams 1992 on Icelandic sig ‘self’ and Joo 2014 and Joo & Deen 2019 on Korean caki ‘self’).

²⁰ Due to the pandemic, five out of 26 children were tested online using Zoom. We did not find any notable difference from children tested in person; thus we provide aggregated results.

²¹ Some previous acquisition studies of zibun ‘self’ in complex sentences used logophoric verbs as matrix verbs such as i(w)-u ‘say’ (Okabe 2008, Experiment I) and omo(w)-u ‘think’ (Otsu 1997; Orita et al. 2021). However, the logophoric verbs introduce logophoric zibun, which selects the agent of indirect discourse, thoughts, etc. (i.e., the matrix subject) as its antecedent (Oshima 2004, 2007, etc.). Because logophoric verbs loosen the preference for local antecedents (at least in the adult grammar), following precedent in the literature (Umeda et al. 2017), we used clause types that are not logophoric (using the connector atoni, ‘after’ and the matric verb kae-ru ‘go.home’).

²² It has been argued that the verbs such as ara(w)-u ‘wash’ and huk-u ‘wipe’ do not allow a local reading of zibun, and to have a local reading, we need to use body-part nouns such as (zibun-no)-karada ‘(self-GEN)-body’ instead of zibun (Noguchi 2013, 2014; Oshima 1979; etc.). However, Kishida (2011) proposes that the local reading of zibun with the verbs that normally do not allow a local identity with a subject becomes available when contrastive stress is added, inducing a meaning such as John washed himself but not anybody else (Kishida 2011:253). As we will see, our test stories always contained two characters, and one of them washes/wipes himself over the other animal or vice versa. This constitutes a contrastive meaning, and we assume that local interpretations of zibun should be naturally available. And as we will see in the results section, the adults in fact strongly preferred local interpretations with ara(w)-u ‘wash’ and huk-u ‘wipe’ in the no-empathy-verb condition. Hence, the main verbs we used do not pose a problem.

²³ We used different animals and events for each test sentence, but in this article, we will use the same context for each test condition for ease of explanation. See Supplementary Appendix B for the complete set of items.

²⁴ The fact that children correctly assign first person reference to the recipient subject with mora(w)-u rules out the possibility that children’s preference is “agent” empathy. Rather, the children seem to prefer to assign the speaker’s empathy to subject arguments regardless of theta roles.

²⁵ Alternatively, children may lack understanding that zibun allows a long-distance antecedent (Orita et al. 2021).

²⁶ This subject empathy default does not need to be abandoned in the course of development because subjects are most likely to be empathy loci in general. Kuno (1987:211) argues that “it is easier for the speaker to empathize with the reference of the subject than with the referents of other NPs in the sentence,” which suggests that subject empathy is a default even for adults.

²⁷ The distribution of second-person subjects is relevant here because second-person arguments are high in empathy (second only to first-person arguments) due to them being a discourse interlocutor, and thus second-person subjects may help align children’s preferences toward subject empathy.

²⁸ The following sentence is an utterance by Asato’s mother containing both age-ru and kure-ru. When the mother is a giver of a benefit, she uses age-ru ‘give’; when the mother is a recipient of a benefit, she uses kure-ru ‘give.’

(i) Ja:   Kakka   domburi   kaite-ageru   kara

Now  mother  bowl    draw-give   because

sono  naka-ni  raamen    kaite-kureru?

 this  inside-DAT ramen   draw-give

 ‘Now, a mother (i.e., I) am going to draw a bowl (for you), so can you draw ramen (for me)?’