Source URL: View this document on the ASHA website
Background: This article was one of the first to investigate nonword repetition as dynamic assessment. It also highlighted its importance as a less biased measure of language impairment for individuals from culturally and linguistically diverse backgrounds.
Hypothesis: The authors hypothesized that nonword repetition might better predict enrollment in language intervention than a knowledge-dependent measure of spoken language skill on which performance could be influenced by factors such as language background, experiential history, and vocabulary knowledge. Further, they predicted that the nonword repetition task would yield more informative likelihood ratios than a knowledge dependent language test.
Methodology and Participants: Study 1 included a group of 20 language impaired (LI) and 20 language normal (LN) English speaking children between the ages of 6;0 and 9;9 years from an urban area. Children were identified as LI if they were receiving SLP intervention at the time of the study. The authors of the study did not use standardized language tests to identify LI children in order to avoid biases inherent in those tests. Each child was asked to repeat, one by one, a set of 16 phonemically balanced nonwords between 1 and 4 syllables, totaling 96 different phonemes. Percentage of phonemes correct (PPC) was calculated for each syllable set (i.e., all 3 syllable nonwords) and as a total for all nonwords presented. Study 2 investigated the nonword repetition task’s effectiveness in distinguishing LI versus LN children. It included the 40 children from the previous study plus 45 other LI (also identified due to receiving SLP intervention) and LN children.
Conclusion: Study 1 revealed statistically significant differences between the LI and LN in the 3 syllable set, 4 syllable set, and total PPC. Analysis of Study 2 revealed that identifying a child as LI based on a total percent of phonemes correct (TOTPPC) of 70% or less produced a positive likelihood ratio of 25.15. Conversely, identifying a child as LN with a TOTPPC of 80% or higher produced a negative likelihood ratio of .03. These numbers exceed the minimum diagnostic accuracy required to demonstrate reasonable sensitivity and specificity. Acceptable positive likelihood ratios should be at least 10.00 and acceptable negative likelihood ratios should be no more than .10.
Relevance to the Field: This article provided evidence that children with language impairments perform significantly worse than their non-impaired peers on measures of nonword repetition, and that such tasks accurately distinguish between children with and without language impairments. Nonword repetition tasks require skills necessary for language learning, including auditory processing, working memory, and organization of articulatory output. Unlike currently used standardized language tests, nonword repetition tasks are unbiased by socioeconomic status and parent education level as they do not test previous linguistic knowledge and skills. It would thus be appropriate to administer nonword repetition tasks during language assessments to assist in accurately identifying children as LI or LN.