Validating the Use of Percent Grammatical Utterances for Assessing Mandarin-Speaking Children's Grammatical Skill: Evidence From 3-Year-Olds

J Speech Lang Hear Res. 2024 Apr 8;67(4):1173-1185. doi: 10.1044/2024_JSLHR-23-00557. Epub 2024 Mar 27.

Abstract

Purpose: The study examined the use of percent grammatical utterances (PGUs) for assessing grammatical skills in Mandarin-speaking 3-year-old children.

Method: Participants were 30 Mandarin-speaking 3-year-olds with typical development. Language samples were collected in two visits for each child using a picture description task. Children were asked to talk about 16 pictures in response to questions and prompts at each visit. Pictures for the language sample collection were identical across the visits. PGUs were computed, and the grammatical errors that children produced in the task were coded and tallied for error types at each visit. Test-retest reliability, split-half reliability, and concurrent criterion validity of PGUs were evaluated.

Results: The mean PGU level was approximately 78% at Visit 1 and 81% at Visit 2, both of which were significantly below the mastery level (i.e., 90%). The correlation coefficient for test-retest reliability of PGU was large (r = .70, p < .01); the correlation coefficient for split-half reliability was medium at Visit 1 (r = .47, p < .01) and large (r = .65, p < .01) at Visit 2. In addition, the correlation coefficient for concurrent criterion validity of PGU was medium for both visits (rs ≥ .35, ps ≤ .03). The ranking and proportion of each error type were similar between the visits.

Conclusion: The initial evidence from psychometric properties suggests that PGU computed from the picture description task is a reliable and valid measure for evaluating grammatical skills in Mandarin-speaking 3-year-old children.

Supplemental material: https://doi.org/10.23641/asha.25395499.

MeSH terms

  • Child
  • Child Language
  • Child, Preschool
  • Humans
  • Language Development Disorders*
  • Language Tests
  • Language*
  • Reproducibility of Results