The Reliability of Developmental Sentence Scoring as a Function of Sample Size

Abstract

Language samples taken from 50 preschool children were used to estimate the reliabilities of the total and component measures of the latest edition of the Developmental Sentence Scoring (DSS) system (Lee and Koenigsknecht, 1972; Lee, 1974). Each sample was divided into five, five-sentence response segments, and DSS total and component scores were tabulated for each segment. Estimates of reliability and standard error of measurement were obtained for the measures involved for varying numbers of response segments. A method is discussed whereby standard error of measurement data can be evaluated by comparison with normative data for the purposes of making decisions about sample size. Application of this method to the total score of DSS showed that the sample size needed for separating normal from abnormal language users was great enough that its use was impractical for this purpose. Because of the absence of normative data on the component measures, the method outlined could not be used for making decisions about sample sizes, which would be appropriate for component measures.