The effect of presentation context on the organization and recall of strongly related and weakly related words was examined in EMR adolescents and nonretarded fifth-grade children of the same mental age (131 months). In blocked presentation formats, subjects sorted words into experimenter-defined groups of four, in which the order of the groupings either changed from trial to trial (i.e., blocked-random), or was consistent across trials (i.e., blocked-consistent). In sort-prompt conditions, subjects were free to structure their own relationships among items during presentation, following instructions to form meaning-based groups (i.e., minimum-sort-prompt), or given explicit training in sorting categorization and sorting stability (i.e., maximum-sort-prompt). Both subject groups found the minimum-sort-prompt method to be as effective as the maximum-sort-prompt method for improving measures of recall and clustering relative to the blocked-random method. Nevertheless, nonretarded children showed higher levels of organization for strongly related items, and better recall and organization for weakly related items. Subsequent analyses suggested that the lower memory performance of the EMR individuals receiving weakly related items was the result of both inconsistent and poorly structured sorting schemes across trials.