Abstract: | This paper examines the pretreatment intercorrelation matrix of the subtests of the Porch Index of Communication Ability to derive factors reflecting components of language impairment. In addition to a general factor reflecting overall language dysfunction, five specific factors reflecting verbal competency (fluency), graphic-verbal expression, gestural-verbal expression, gestural-nonverbal comprehension, and graphic-geometric comprehension were found which accounted for 90.6% of the total variance. A cluster analysis of the five pretreatment factor scores yielded six groups which accounted for 79% of the variance for these factors. These groups were differentiated primarily in terms of impairment and showed a high rate of accuracy prediction for group membership (97.5%). These groups were extremely stable, as shown by the 75% accurate prediction rate derived from the post-treatment factor scores. The effectiveness of this model in generating homogeneous and highly discriminable groups based on objective measures of the patient's language functions indicates that classification of patients into empirical groups should be a major concern of any research design. Implications for assessing effectiveness of therapeutic intervention and the design of test battery were also discussed. |