Temporal reliability of length-complexity index

Abstract

The purpose of this study was to determine the test-retest reliability of a new measure of children's language, the Length-Complexity Index (LCI). There were seventeen subjects who participated in this study, seven males and ten females. The subjects were five-year-old children attending Kindergarten in the Sullivan, Illinois Public Schools. Each of the subjects had normal hearing, average or above average intelligence as measured on the Peabody Picture Vocabulary Test (Dunn, 1965), no obvious neuromuscular disorders, and American English family language background. None were twins, or had a history of stuttering. All subjects came from middle or lower class socioeconomic level families. Language samples were elicited from these children on three separate occasions within a ten-day period by two examiners. The stimulus material constructed to elicit the verbalizations consisted of three sets of pictures judged to be of interest to five-year-old children. The intraclass correlation coefficient for MLR was r i = .65 for the individual child's responses on subsequent retests of single 50-response language samples. This indicates the considerable variability of MLR as a measure of a child's daily verbal output. The intraclass correlation coefficient for the LCI was r i = .80 for the individual child's responses on subsequent retests of single 50-response language samples. The results indicate that within the limits of this study as a language measure the LCI is not as variable as the MLR; that it tends to measure children's language output more reliably over time. This is interpreted to mean that a composite linguistic analysis of length and complexity will yield a more consistent picture of an individual child's verbal output than will a measure of length alone.