TY - JOUR
T1 - Applying item response theory to the development of a screening adaptation of the goldman-fristoe test of articulation–second edition
AU - Brackenbury, Tim
AU - Zickar, Michael J.
AU - Munson, Benjamin
AU - Storkel, Holly L.
N1 - Publisher Copyright:
© 2017 American Speech-Language-Hearing Association.
PY - 2017/9
Y1 - 2017/9
N2 - Purpose: Item response theory (IRT) is a psychometric approach to measurement that uses latent trait abilities (e.g., speech sound production skills) to model performance on individual items that vary by difficulty and discrimination. An IRT analysis was applied to preschoolers' productions of the words on the Goldman-Fristoe Test of Articulation–Second Edition (GFTA-2) to identify candidates for a screening measure of speech sound production skills. Method: The phoneme accuracies from 154 preschoolers, with speech skills on the GFTA-2 ranging from the 1st to above the 90th percentile, were analyzed with a 2-parameter logistic model. Results: A total of 108 of the 232 phonemes from stimuli in the sounds-in-words subtest fit the IRT model. These phonemes, and subgroups of the most difficult of these phonemes, correlated significantly with the children’s overall percentile scores on the GFTA-2. Regression equations calculated for the 5 and 10 most difficult phonemes predicted overall percentile score at levels commensurate with other screening measures. Conclusions: These results suggest that speech production accuracy can be screened effectively with a small number of sounds. They motivate further research toward the development of a screening measure of children's speech sound production skills whose stimuli consist of a limited number of difficult phonemes.
AB - Purpose: Item response theory (IRT) is a psychometric approach to measurement that uses latent trait abilities (e.g., speech sound production skills) to model performance on individual items that vary by difficulty and discrimination. An IRT analysis was applied to preschoolers' productions of the words on the Goldman-Fristoe Test of Articulation–Second Edition (GFTA-2) to identify candidates for a screening measure of speech sound production skills. Method: The phoneme accuracies from 154 preschoolers, with speech skills on the GFTA-2 ranging from the 1st to above the 90th percentile, were analyzed with a 2-parameter logistic model. Results: A total of 108 of the 232 phonemes from stimuli in the sounds-in-words subtest fit the IRT model. These phonemes, and subgroups of the most difficult of these phonemes, correlated significantly with the children’s overall percentile scores on the GFTA-2. Regression equations calculated for the 5 and 10 most difficult phonemes predicted overall percentile score at levels commensurate with other screening measures. Conclusions: These results suggest that speech production accuracy can be screened effectively with a small number of sounds. They motivate further research toward the development of a screening measure of children's speech sound production skills whose stimuli consist of a limited number of difficult phonemes.
UR - http://www.scopus.com/inward/record.url?scp=85029760696&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85029760696&partnerID=8YFLogxK
U2 - 10.1044/2017_JSLHR-L-16-0392
DO - 10.1044/2017_JSLHR-L-16-0392
M3 - Article
C2 - 28859203
AN - SCOPUS:85029760696
SN - 1092-4388
VL - 60
SP - 2672
EP - 2679
JO - Journal of Speech, Language, and Hearing Research
JF - Journal of Speech, Language, and Hearing Research
IS - 9
ER -