Group games were used to teach basic number concepts to educable mentally retarded children. The number concepts were integral but incidental to the intentional teaching of game skills. Subjects were 21 boys and 19 girls from eight classes, free of gross defects (means were IQ=66.22, age=7.9, mental age=5.11), and matched in pairs for chronological age, mental age, IQ, and scores on pre-experimental tests of number concepts, game skills, preference for social play, and voluntary social participation in school. The experimental group of 20 participated 100 minutes a week in a 9-month game program; the control group, also 20, attended a traditional math program; and both were tested at the middle and end of this period. The experimental group scored higher on both mid-experimental (p<.003) and post-experimental (p<.0003) measures on the Number Knowledge Test (test reliability=98% on Form A and 93% on Form B), higher on the post-experimental measure (p<.002) on the spontaneous use of quantitative terms, and made fewer errors o n both the mid-experimental (p<.005) and post-experimental (p<.0005) tests on General Game Skills (test reliability=91%). Game and modeling procedures and attention directing variables (use of excitement and rewards) are suggested as teaching techniques for retarded children. , and a handbook for teaching game skills is included. (Author/SN)