Merge as a basic mechanism of language: Evidence from language acquisition

One of the most basic skills of a proficient language user is the ability to combine smaller units (e.g., words) into larger units (e.g., phrases). Therefore, a mechanism that combines such "syntactic" objects is a good candidate for a basic component of the human computational system. Chomsky (1995) defines a mechanism called Merge that concatenates two and only two syntactic objects (e.g., morphemes, words) in each application. This paper presents evidence from children's earliest productions that Merge operates on three levels in child grammar. First, Merge operates on the word-level, fusing two words into single lexical items. Merge also operates on the sub-word level providing a straightforward account of the so-called "mixed" productions of bilingual children. In addition, Merge combines words into phrases. The fact that this mechanism only concatenates two units in any single application yields an elegant account of the observed initial period in language development in which only two-word combinations are attested (Bowerman, 1990; Brown, 1973). Language acquisition data thus reveal a simple concatenation operation like Merge to be one of the most basic mechanisms of language production.