The purpose of this study was to examine the effect of different kinds of cues on the recall of a text. Forty-four subjects read a short research paper. After a delay of three weeks, they answered questions on the paper. There were three types of cues: free cues, which encouraged unrestricted responses; bibliographic cues, which asked subjects to recall elements of bibliographic description; and structural cues, which were based on the text-linguistic structure of research reports. Recall protocols were analyzed to establish the degree to which protocol vocabulary matched index terms and propositions taken from the article. Bibliographic cues produced relatively few matching terms, and short protocols. Free cues produced many matching terms and propositions, but very long protocols containing large numbers of non-matching terms and intrusions. Structural cues produced as many matching terms and propositions as free cues, but significantly fewer non-matching terms and intrusions. Implications of these findings for known-item retrieval and the design of information system interfaces are discussed.