Extralist cues can serve as effective memory aids. However, several factors can reduce this effectiveness by producing retrieval inhibition. One factor involves degree of control in accessing the domain of information specified by the test cue. The results of Experiment 1 indicate that random switching between rhyme and meaning-related cues disrupts this control. The results of Experiments 2-4, however, indicate good control when all test cues are related to their targets within the same domain. Other factors producing retrieval inhibition include the size of the search set defined by the test cue, and the strength of the cue both in relation to its target and to its category name. The results of Experiment 5 indicate that recall is reduced for rhyme and taxonomic-instance cues when the search set is large and when cue-to-target strength is weak. When multiple cues are involved, it is also reduced for cues that are only weakly related to the category name. Implications for the differential effectiveness of single-instance and category-name cues and for current models of retrieval inhibition are discussed.Extralist cues can be highly effective obtained for several types of cues, including memory aids for recovering recently encoded rhymes, word endings, associates, and taxinformation (e.g., Bahrick, 1970; Tulving onomic category labels (Nelson & McEvoy, & Pearlstone, 1966). This effectiveness is 1979a, 1979b). influenced by preexperimentally measured Nelson (1981) has attempted to develop strength from the cue to the target (e.g., a model that would explain these results as Bahrick, 197Q) and by the size of the search well as more complicated findings involving set defined by the presented cue (Nelson manipulations of the nature of study context & McEvoy, 1979b). The stronger the cue-cues and their influence on target set size to-target relationship and the smaller the effects. The principal assumptions of this defined set, the greater the likelihood of sue-model can be divided into those concerning cessful recall. These variables do not appear structure, processing, and retrieval. The to interact, and similar findings have been structural assumptions specify what information can be encoded and how this infor-This research was supported by National Institute of mation is organized. Given the presentation Mental Health Grant MH, 16360 to the first author. o f a word, such information would include The research was presented, in part at the First Colvigual phonemic and meaning characterisorado Conference on Learning and Memory, Boulder, .