The oddity performance of five preschool children at risk for mental retardation was facilitated by increasing the number of nonodd elements in a visual array. A combination intrasubject reversal and multiple baseline across subjects design indicated the internal validity of interventions designed to enhance the perceptual salience and consequent stimulus control of the odd stimulus. Results demonstrate that transfer and maintenance of oddity learning can be obtained even with individuals for whom correct oddity responding is uncommon. The typically poor performance of young and developmentally delayed children as compared to nondelayed children on tasks such as the oddity task may be attributable to a lower sensitivity to relational information.