The content of shared reading for pre-schoolers: comparing the distribution of complex grammar in books vs conservations

The content of shared reading for pre-schoolers: comparing the distribution of complex grammar in books vs conservations

First Author: Yaling Hsiao -- University of Oxford
Additional authors/chairs: 
Kate Nation
Keywords: Syntax, Text Complexity, children, Reading, Grammar and syntax
Abstract / Summary: 

Purpose – Analyses on developmental corpora show that speech and texts provide different distributional statistics of lexical items with higher lexical diversity found in texts. Few studies have examined structural frequencies in speech versus texts, especially in pre-literate children. The current study addresses this gap by comparing children’s exposure to complex grammar via listening to conversations vs shared reading.

Method – We extracted complex structures in corpora of speech (Manchester Corpus – CHILDES, 3.8 million words) and texts (ReadOxford, 320,000 words) directed at children age 0-5. We focused on relative clauses, specifically subject relatives (“the horse that galloped”), object relatives (“the goals that world leaders set), reduced object relatives (“the fun I had”), passive relatives (“things that can’t be touched”), reduced passive relatives (“the shadows made by the sun”). The parser software Benepar generated constituency parsing trees and the structure extraction software Tregex was used to extract the relative clauses.

Results –Speech was significantly lower than text in normalized frequencies of all relative clause structures except passive relatives.

Conclusions – Listening to written language via shared reading provides young children with exposure to more complex syntax than speech. This indicates a role for print exposure in language acquisition, even before the onset of literacy. However, structural complexity doesn’t single-handedly differentiate the language used in speech and texts. Speech may provide discourse contexts conducive to the emergence of certain structures, e.g. reduced passive clauses.