In this work package, we used corpus analyses and eye-tracking to understand how children learn to favour contextual information over lexical information in their interpretation of simple and complex sentences, a development critical to children’s ability to understand longer conversations and text.

Early in development, children are highly sensitive to both frequent multiword chunks (e.g. he’s catching the X) and to aspects of the discourse context such as what is ‘new’ information for their conversation partner. But they don’t always coordinate the two. For example, they don’t necessarily understand that they should use a lexical noun phrase (e.g., the boy) to refer to a new referent and pronouns (e.g., he) to refer to a given referent. And sometimes these sources of information compete (e.g., conversation partners use a lexical noun phrase to refer to a given referent), so children must learn to prioritise the most reliable information in any given context.

Our corpus analyses of subject and object relative clauses suggest that in child-directed and child speech, the lexical and semantic features that are linked to specific noun phrases within these clauses largely align with findings from adult-directed speech. For example, noun phrases referring to patients tend to be inanimate, and noun phrases referring to agents tend to be expressed by pronouns. For discourse-based features, however, both our corpus analyses and our eye-tracking data suggest that these features take time to develop as children do not systematically expect subject noun phrases to refer to discourse-old entities, for example.

Project Team 

Ben Ambridge, Silke Brandt (lead), Jacky Chan, Anna Theakston 

Duration 

3 years, starting 1 May 2020 

Project Number

2.2 

Selected publications 

“Putting lexical and semantic cues into discourse context: A multifactorial corpus study of relative clauses in child-directed and child speech” – currently under review at Language Development Research

Related