An earlier study showed that responses are remembered better when subjects produce them from cues, than when subjects read cue-response pairs. The decided memory advantage for generated targets relative to read ones is known as the generation effect. The present research is designed to study the generation effect for cues, following a depth-of-processing interpretation. The task described is to produce a word, opposite in meaning to the cue, given the cue and target with two interior adjacent letters missing. Thirty-two undergraduate students viewed a cue-target pair for five seconds, wrote the target word, and viewed the completed target for three seconds. Control subjects received the cue and the entire target, wrote the target, and reviewed the target for three seconds. All subjects received twelve items and provided data on free recall, cued recall, and recognition. Half were tested for recognition of targets, and half for recognition of cues. Results showed that the generation effect does apply to cues as well as responses, although the effect for cues is weaker, suggesting that depth of processing cannot be the only process contributing to the generation effect. (LMO)