This study provides experimental data on the acquisition of Japanese causatives. It has been observed that children begin to produce analytic causatives at the age of 2–3 years old. However, there have been no studies examining whether the child causatives are identical to adult causatives with respect to their meaning and structure.

Japanese has two types of causatives, the lexical and the productive causative, which share the same verb stem and case-marking pattern, as given below: Lexical causative: John-ga Mary-ni zibun-no hon-o mise-ta. John-

NOM

Mary-

DAT

self-

GEN

book-

ACC

show-

PAST

[special characters omitted] 'John showed self' book to Mary.' (self = John, *Mary) Productive causative: John-ga Mary-ni zibun-no hon-o mi-sase-ta. John-

NOM

Mary-

DAT

self-

GEN

book-

ACC

look.at-

CAUS-PAST

[special characters omitted]'John made Mary look at self' book.' (self = John, Mary) As indicated, the lexical causative is unambiguous as to the antecedent of the anaphor zibun, whereas the productive causative is ambiguous. Given that zibun must be bound by a subject, it is assumed that the lexical causative is mono-clausal, while the productive causative is bi-clausal. Also, the lexical causative exclusively denotes 'direct causation' while the productive causative expresses 'indirect causation.'

A series of experiments investigated whether children aged 4–6 know the semantic and syntactic differences between the two causatives. Although the first two experiments revealed that children as old as 6 disallowed the embedded subject of the productive causative to function as "subject", the third experiment with slightly changed scenarios revealed that the same children assigned a bi-clausal structure to the productive causative and a mono-clausal structure to the lexical causative.

Children's comprehension of other constructions with bi-clausal structure, such as complex sentences with finite embedded clauses and benefactives, was also examined and compared to that of the causatives in order to clarify the non-adult-like performance on the causatives. Based on the finding that children also had difficulty with the benefactives, I suggest that incremental processing and difficulty with on-line reanalysis account for the children's non-adult-like responses. I also discuss how children might acquire the semantic and structural properties of Japanese causatives under "poverty of the stimulus" situation.