Learning effects were assessed for the block design (BD) task, on the basis of variation in 2 stimulus parameters: perceptual cohesiveness (PC) and set size uncertainty (U). Thirty-one nonclinical undergraduate students (19 female) each completed 3 designs for each of 4 varied sets of the stimulus parameters (high-PC/high-U, high-PC/low-U, low-PC/high-U, and low-PC/low-U), ordered randomly within a larger set of designs with mixed stimulus characteristics. Regression analyses revealed significant, although modest, learning effects in all conditions. Negative-logarithmic learning slopes (growth factors) were greatest for high-U/high-PC designs and smallest for low-U/low-PC designs. Comparison of these slopes with known Wechsler Adult Intelligence Scale (3rd ed.; D. Wechsler, 1997; and 4th ed.; D. Wechsler, 2008) BD subtest gain scores demonstrated that presenting novel test items matched on stimulus parameters in multiple administrations reduced learning effects compared with the repeated use of the same test items. The results suggest that repeated administration of novel test items of the BD subtest, matched for PC and U, would result in more accurate assessments of changes in examinees' abilities over time than would the use of the same items. Difficulties inherent in implementing this method are also discussed. (Contains 3 footnotes, 3 tables, and 4 figures.)