Although research subscales have been developed (Buckelew, DeGood, Schwartz, & Kerler, 1986) to assess the cognitive and somatic item patterning among chronic pain patients using the SCL-90-R, no information is available concerning the reliability or validity of those subscales. This study was designed to assess the internal consistency of these experimental subscales, evaluate the validity of these scales using standardized clinical scales measuring similar constructs, and to further examine the item response patterning with a new pain clinic sample and a rehabilitation center patient sample. The SCL-90-R was administered to 78 patients with chronic pain, 52 patients with spinal cord injury, and a control group of 145 college students. Estimates of internal consistency, represented by coefficient alpha, were .69 for the Somatic Depression subscales and .91 for the Cognitive Depression subscales. Data suggestive of construct validity included differential correlations of r = .78 between the experimental Cognitive Depression scale and the SCL-90-R Obsessive Compulsive scale, r = .66 between the experimental Somatic Depression scale and the SCL-90-R Somatization scale. Similar evidence of divergent validity was not clearly demonstrated with the remaining SCL-90-R subscales. Multivariate analysis of variance revealed that previously found item response patterning was replicated with a new chronic pain sample and with a spinal cord injured sample. Refinement of the research subscales and implications for clinical use among medical/rehabilitation samples is discussed.