Parts of speech in early Serbian child language

Abstract

The study was aimed at the preliminary quantitative exploration of distribution of parts of speech in Serbian children’s early language production and comparisons with the distribution in language input. The Serbian Corpus of Early Child Language (SCECL) was used in retrieving production of children and adults longitudinally recorded in interaction sequences at 16 age levels (18-48 months). Automatic lemmatization and manual check and corrections were made on six parts of speech, mainly content words: nouns, verbs, adjectives, adverbs, pronouns, and prepositions. In order to make a better differentiation between language structure and language use, information on the distribution of parts of speech from a large corpus of standard Serbian written language was also included in the analysis. The findings reveal a high degree of similarity between the parts of speech distributions in child language and conversational language input. In comparison to the distribution in written Serbian language, it reveals that formation of structure in child language is not influenced by language structure itself but by functional aspects of its use in natural face-toface interaction. Child language deviates from language input at early stages of development, and incremental approximation towards the model distribution is recorded. The findings are in accordance with distributional accounts in the theory of language acquisition.