The purpose of this study was to test associative skills by means of a paired-associate task with varied mediational instructions. Subjects were 84 first grade ghetto children, 25% of whom were from Spanish-speaking homes; virtually all the other subjects were black. The 3 x 2 x 4 factorial design included: three mediation sets (imagery, verbalization-sentences, and verbalization-naming), two ability levels (mediational ability and production ability), and four trials. A control group was tested for spontaneous production. Results indicate that performance was better (a) under the sentential and imagery sets than under the naming set, (b) when mediational rather than production ability was required, and (c) on later trials, although rate of learning varied by condition. An interaction of mode, elaboration, and ability was found. Results suggest the same patterns of mediational ability obtain for first graders from ghetto and white middle class backgrounds. The interaction of mediational deficiency with mediation set requires the revision of the stage model of mediational deficiency. (Author/NH)