Uninversion errors in Wh- Questions in Children with DLD

The current study looks at Uninversion errors in children with Developmental Language Disorder (DLD). Specifically, this study will extended McCauley et al's 2021 study which considered the role of multiword units in early wh-question production by using distributional statistics from child-directed speech to predict children’s spontaneous uninversion errors in typically developing children by considering this in children with DLD.

Our hypotheses are that uninversion errors (e.g. What you are doing?) will be more likely to occur in children’s question productions (e.g. when the target is: What are you doing?) when:
1. The frequency of n-grams in the correct target form have a lower frequency in the input, and more specifically
1a. The target Wh- word and auxiliary combination (e.g. what are) has a lower frequency in the input.
1b. The target form of the second word pair (e.g. are you) has a lower frequency in the input
1c. The target form of the third word pair (e.g. you doing) has a lower frequency in the input.
1d. The target form of the first word triple (e.g. what are you) has a lower frequency in the input.
1e. The target form of the second word triple (e.g. are you doing) has a lower frequency in the input.

2. The frequency of n-grams in the uninverted form have a higher frequency in the input, and more specifically
2a. The uninverted form of the first word pair (e.g. what you) has a higher frequency in the input.
2b. The uninverted form of the second word pair (e.g. you are) has a higher frequency in the input.
2c. The uninverted form of the third word pair (e.g. are doing) has a higher frequency in the input.
2d. The uninverted form of the first word triple (e.g. what you are) has a lower frequency in the input.
2e. The uninverted form of the second word triple (e.g. you are doing) has a higher frequency in the input.

To do this, transcripts from the Conti-Ramsden 3 corpus of 4 children with DLD will be prepared and we will test our hypothesis using Bayesian mixed-effects logistic regression modelling using brms.

A detailed discussion of the study is provided in the document attached.