Multiword units lead to errors of commission in children’s spontaneous production: “What corpus data can tell us?”

NB: This is is a deliberately BAD blog post that I wrote for a workshop on writing a GOOD blog post. You can read the GOOD version of this post at https://benambridge.wordpress.com/2022/04/25/do-fish-have-barcodes-and-other-baffling-questions/

Researchers at the ESRC International Centre for Language and Communicative Development (LuCiD) based at the University of Manchester recently published a study investigating non-inversion errors in English-speaking children’s production of wh-questions. In English, the subject (e.g., I) and the auxiliary (e.g., can) swap positions when forming questions:

I can draw a house –> What can I draw?

This is known as subject-auxiliary inversion. But often typically-developing children learning English as a first language fail to perform subject-auxiliary inversion (e.g., *What I can draw?). This is known as a subject-auxiliary inversion error, also called uninversion errors or non-inversion errors.

The aim of this research was to test different theoretical accounts of why children make these errors. The study took the form of a corpus-based study of longitudinal audio recordings of children and their caregivers, with the children aged from 1;8.22 to 2;0.25 at the beginning of the study, half boys and half girls. First, the research found a correlation between the frequency of multiword bigrams from uninverted structures in the linguistic input (e.g., what+I+can+draw) and the probability of children making the corresponding error when attempting the relevant question (e.g., What I can draw?). Second, the research found a negative correlation between the frequency of multiword bigrams from inverted structures (i.e., questions) in the linguistic input (e.g., What+can+I+draw) and children’s error rates.

These results are important because they challenge movement-based accounts of children’s question production, and suggest instead that children learn to ask questions by generalizing across the input. This is evidence for constructivist or usage-based accounts of language acquisition. They also constitute evidence for interventions for children with Developmental Language Disorder (DLD) for whom questions constitute a particular challenge. 

The research was conducted by Stewart McCauley (now at the University of Iowa) along with Colin Bannard, Anna Theakston, Michelle Davis, Thea Cameron-Faulkner and Ben Ambridge, all at the University of Manchester and the ESRC International Centre for Language and Communicative Development (LuCiD). It was published in the journal Developmental Science.

Reference:

McCauley, S. M., Bannard, C., Theakston, A., Davis, M., Cameron‐Faulkner, T., & Ambridge, B. (2021). Multiword units lead to errors of commission in children’s spontaneous production:“What corpus data can tell us?*”. Developmental Science, 24(6), e13125.