Two experiments contrasted processing manipulations designed to focus readers' attention on sequential and relational aspects of passages with processing activities presumed to focus readers' attention on individual lexical items and the propositions in which they were embedded. Employing an intentional recall paradigm, both experiments found recall to be superior when the readers engaged in sequential and relational processing to recall when they performed individual item-specific lexical processing. These results disagree with those of previous work on individual item-specific lexical processing conducted with an incidental paradigm but agree closely with a broad range of studies that have focused on relational and sequential processing. The results are discussed in terms of differences between intentional and incidental learning paradigms.In recent years, considerable work has indicated the importance of understanding the relational and sequential aspects of texts in memory for prose. Typically, processing activities, such as listing key terms, answering lower level factual questions, and copying parts of the text-activities that focus readers' attention on superficial or highly specific elements of text-have not been effective in enhancing readers' memory for prose (e.g., Andre & Womack, 1978;Glover, Plake, Roberts, Zimmer, & Palmere, 1981). For example, Glover et al. (1981) contrasted readers' memory for text in several conditions including a key word condition, in which subjects listed "key words" found in paragraphs; a model statements condition, in which subjects identified statements in each paragraph as fitting a model; and a logical inference condition, in which readers focused on the relational aspects of paragraphs in order to draw logical inferences. The results indicated that the key words and model statements conditions obtained recall scores highly similar to those of a control group, while subjects in the logical inferences condition recalled significantly greater amounts of text. Similarly, Andre and Womack (1978) found that highly specific, lower level questions 297