80 emotionally disturbed and 80 neurologically impaired boys in a public school setting were matched for WISC Full Scale IQ and chronological age. An analysis was made of the Similarities subtest responses utilizing both the error-analysis method of Spence and the cognitive style procedure of Sigel for purposes of comparison. The Wilcoxon matched-pairs signed-ranks technique was used in analyzing the data. The results were as follows: (1) the error analysis approach provided the greater potential for differential diagnosis between the two groups under consideration; (2) use of the cognitive style scores permitted improved definition of the dynamic implications of the error-analysis categories, inadequate abstracting and narrative-descriptive responses; (3) diagnostic power was not increased when all responses as opposed to errors alone were considered.