Discriminant validity of lateral tactile and motor skills measures in learning disabled children

The discriminant validity of left and right hand sensorimotor measures from the Halstead-Reitan battery was tested using confirmatory factor analysis. A group of learning disabled children was divided into analysis (N=488) and cross validation (N=400) samples and the following measures were taken for both the left and the right hands: Tactile Perception, Finger Agnosia, Fingertip Number Writing, Tactile Form Recognition, Finger Tapping, Grip Strength, Grooved Pegboard, and Mazes. These 16 tests and 5 WISC subtests (Information Similarities, Vocabulary, Block Design, Object Assembly) were included in a 6 factor 2 indicator model (Tactile Skills, Astereognosis, Simple Motor, Complex Motor, Verbal, Visual Spatial Constructive) with correlated errors between left and right hand measures from the same test, and between each Grooved Pegboard measure and Object Assembly. Parameters in the model were estimated three times in each sample: 1) with all non-zero factor loadings free, 2) with factor loadings constrained to be equal for left and right hand measures from the same sensorimotor test using a) the correlation matrix and b) the covariance matrix as input. The unconstrained model fit the data well in both samples as indexed by the normalized residuals, modification indices. Goodness of Fit and Adjusted Goodness of Fit Indices, fit ratio, squared multiple correlations and the root mean squared residual. In the analysis sample, constrained and unconstrained models did not differ significantly when analysis was based on the correlation matrix (X [raised 2] [lowered (8)]=4.66, p>.250), but did differ when analysis was based on the covariance matrix (X [raised 2] [lowered (8)]=28.42, p<.001). Lack of model fit resulted from constraints on three variables. Grip Strength, Finger Tapping, and Tactile Perception. Virtually identical results were obtained in cross validation. Constrained and unconstrained models were not significantly different when analysis was based on the correlation matrix (X [raised 2] [lowered (8)]=4.15, p>.250), but differed under analysis of the covariance matrix (X [raised 2] [lowered (8)]=29.79, P<.001). Lack of fit was again attributable to constraints on Finger Tapping and Grip Strength. The results indicated that left and right hand tests were tau-equivalent for all but two sensorimotor measures. Those same two measures (Grip Strength and Finger Tapping) were found in both samples to be congeneric and not tau-equivalent. The six factor model adequately accounted for most relationships in the data and provides a potential starting point for future studies investigating sensorimotor functions in learning disabled children. Combining scores from the left and right hand versions of the same test was recommended in future factor analytic studies which are not expressly concerned with the discriminant validity of lateral sensorimotor measures. Use of composite scores would increase test reliabilities and eliminate problems of test specific relationships caused by inclusion of doublet scores. Results of the present investigation suggest that the use of composite scores would not significantly alter the obtained factor structure.