Summary.Forty‐eight subjects (16–17 years old), randomly assigned to two recall conditions (immediate and delayed), were each given two prose passages to recall, one after reading, the other after listening. The passages were judged to be difficult in content and unfamiliar. An incidental recall procedure was adopted. Following Meyer's (1975) analytic procedures, recall protocols were scored for the presence of idea units, content units, relationship units and number of units at each level of the content structure. Number of errors were scored independently of Meyer's scheme. As hypothesised for a difficult recall task involving connected prose, results demonstrated the superiority of recall after reading, both in terms of total units recalled and in number of errors. This suggests the value of more systematic research into the conditions which favour one mode of presentation over the ot