This study determined the effectiveness of a group training procedure in improving performance of intellectually average students on the Raven Learning Potential measure, and examined the differential effectiveness of training by race and social class. The sample of 379 second through fifth graders, half of whom were black, were pretested on Raven's Coloured Progressive Matrices, randomly assigned to trained or non-trained groups, and posttested on the Raven test. Multiple regression analysis on posttest scores indicated that trained students improved their nonverbal reasoning ability more than nontrained students and that training was equally effective with black and white children. Middle class children were found to benefit more from training than lower class children. (Author)