This study investigated adult age differences in the cohesion of narrative retellings in both the reference and conjunction discourse systems and explored the role of information-processing factors in accounting for any such differences. Twenty Ss in each of the age groups 18-25, 26-55, and 60-87 either read a story or its parallel cartoon version, then retold it twice. Stories were coded for recall, clarity of referencing, and types of propositional connectives. We also obtained Daneman's (1980) measure of sentence memory span. The oldest group scored significantly lower on the memory span measure, recalled less story information, and made more referential errors in retellings. There were no age differences in complexity of conjunction usage. Working memory span scores partly accounted for the differences observed in referential quality both within and between age groups. Results are generally consistent with an information-processing account of story telling and aging.Research on the development of narrative skills has focused to date on the earlier portion of the life span. It has demonstrated that younger children frequently have difficulties managing the systems of cohesion that serve to tie the propositions of a story together and make it a text (e.g., Piaget, 1926;Pratt & MacKenzie-Keating, 1985). However, despite some interest in the role of narrative in the lives of older adults (e.g., Mergler, Faust, & Goldstein, 1985), so far there has been little detailed study of narrative cohesion during the later portions of the life span.Halliday and Hasan (1976) described a series of cohesion systems in English, of which the most thoroughly studied is reference. A number of studies of children have shown that preschoolers show some rudimentary mastery of aspects of the reference system, such as pronouns or the article system. However, they have considerably more difficulty than adult narrators in making clear the systematic links between earlier and later text elements (Menig-Peterson & McCabe, 1978;Zehler & Brewer, 1982). In particular, younger children show difficulties introducing and marking new referents appropriately (e.g., Warden, 1976) and frequently produce ambiguous pronouns or terms that could refer to more than one previous referent, thus con-This research was supported by grants from Wilfrid Laurier University and from Mount Saint Vincent University. Portions of these findings were reported at the Canadian Psychological Association Meetings, Vancouver, British Columbia, June 1987. We thank Kathleen Kitching and Vince Walsh for their help with stimulus materials and data analysis as well as the participants of this research for their good humor and interest.