Echols, C. 1 , Souza, A. 2 , Barbosa, P. 3 & Cardoso-Martins, C. 3
1 University of Texas
2 Concordia University, Montréal
3 Universidade Federal de Minas Gerais, Brazil
Building on research documenting the potential value, for language acquisition, of information in the input, we analyze distributional information in Brazilian Portuguese (BP), and compare it to English. BP is of potential interest because adult speakers often omit pronouns (which is grammatical) and other grammatical elements from their speech, raising questions about the effects on distributional information of grammatical and ungrammatical omissions from speech directed to children. Maternal speech samples collected from 18 Brazilian and 18 United States mother-child dyads during semi-naturalistic interactions when children were 9, 13 and 18 months old were transcribed and coded for various grammatical properties. Initial analyses from this sample showed that 61% of the BP mothers' productions included omissions (with only 13% being grammatical) contrasted with 16% incomplete sentences for the English mothers, verifying important differences between languages in the completeness of utterances directed to infants. To evaluate distributional information, we built on frequent frame (Mintz, 2003) and Construction Grammar (Langacker, 1992) approaches. We identified, for all possible 3-word sequences, a set of frequent frames-sequences following either a word-X-word or word-word-X format that met a frequency criterion. We conducted collostructional analyses to measure the strength of association between the frequent frames and particular grammatical categories, focusing here on noun and verb categories. In both languages, the frames predicting nouns are virtually non-overlapping from those predicting verbs, indicating that both languages contain sufficient information to distinguish basic grammatical categories. The results of the collostructional analyses suggest that although the frequent frames provide statistical information that distinguishes noun from verbs in each language, relative strengths of association vary across language, grammatical category and frame type (word_X_word, word_word_X). Results will be discussed in relation to differences across languages in the distributional contexts of linguistic information and will be compared to vocabulary acquisition data from the children.