Researchers have proposed that children use several cues to successfully exploit sentence frames when when learning novel verb meanings (Cameron-Faulkner et al., 2003; Dittmar et al., 2008; Fisher, 1994, 2002; Gertner & Fisher, 2013; Noble et al., 2016). These cues include preferences for(A) interpreting the first noun as causal agent, (B) each noun to bear a distinct thematic role, and (C) more frequent constructions. Cues B and C have been cited to explain why English- and German-learning children easily map the transitive verb in “the duck is gorping the bunny” onto a causative action but have more difficulty mapping the verb in the conjoined-NP sentence“the duck and the bunny are gorping”onto a noncausative action. That is, the two NP arguments in transitive sentencesbear two distinct theta roles (Agent and Patient) whereas the NPs in the conjoined sentence share the Agent theta role; moreover, conjoined-NP sentences are more sparse in the input than transitive sentences.

The studies reported here investigate whether these cues are available to Japanese children. Given more flexible word-order—canonical overt SOV order is comparatively rare in the input—and a high preponderance of argument omission, English-type cues are much less available in Japanese. Matsuo et al. (2012) found the same transitive-intransitive asymmetry in Japanese2-year-olds. Study 1 examines the‘first-noun-as-causal-agent’ and ‘distinct theta-role cues’ (A & B); Study 2 examines the role of constructional frequency (C).

Study 1 presented 27 2-year-old Japanese children with the same videos as Matsuo et al. (2012) used; children heard one of the following word orders:

(1)a. Agent-ga Patient-o neketteiru. (Agent-first, transitive (-GA, -O))

b. Patient-o Agent-ga neketteiru. (Patient-first, transitive (-O, -GA))

Japanese childrenemployingcue A should correctly match (1a) with a causative scene and should (incorrectly) match (1b) with a scene in which the first NP is agentive (i.e., the noncausative scene here). Indeed, the children hearing (1a)shifted their looking significantly towards the causative action (p< .05); however, the childrenhearing (1b) showed a different looking preference; between-condition effects trended towards significance (p=.07; Fig.1).

In Study 2, 39 2-year-olds saw videos in which novel causative (2 characters) and non-causative (1 character) actions were presented simultaneously, along with a novel verb. The novel verbs were presented in 2 between-subject conditions,as in (2a,b).

2a. Agent-garuchitteiru. (1-NP condition)

Agent-nomgorping

‘Agent and Patient are gorping.’

2b. Agent-ga Patient-o ruchitteiru. (2-NP condition)

Agent-nomPatient-accgorping

‘Agent is gorping Patient.’

Children using cue C should perform better in the 1-NP condition than the 2-NPs condition, 1-NP sentences having greater constructional frequency (Matsuo et al., 2012). Childrenusing cue B should find both (2a) and (2b) equally straightforward because each contains the same number of thematic relations as NPs. However, neither prediction was borne out: children hearing (2b) looked significantly longer to the causative action (p=.02), whereas thosehearing (2a) showed no significant preferences (Fig.2).

The results of these studies confirm that Japanese childrenmake use of argument number and case markers when interpreting novel verbs, as they consistently preferred causative actions when hearing canonical SOV sentences (Gleitman, 1990). Use of cue A was supported; however, neither construction frequency nor a preference for NPs with distinct thematic roles was observed to play a role.

Figures 1 and 2: Mean percent looking time to the causative actionsduring the control and test trials (error bars indicate standard error)