The purpose of this research was to study the development of learners' memory for their own writing, using the cloze procedure as a sensitive measure of retention performance. In particular, the objectives were to (a) examine students' retention of self-generated texts at four levels (sixth, eighth and tenth grades, and college undergraduates), (b) compare learner's cloze performance on self-generated writing and on a peer's writing, and (c) examine developmentally the changes made in exact replacement cloze performance and the changes in meaning. Although it was hypothesized that developmental trends would occur in cloze accuracy, only very minor developmental trends appeared. Writers' accuracy on their own texts in a cloze format was 84% (exact replacements) while it only reached 60% accuracy on the writing of their peers. Thus, by standard cloze measures, writers are independent on their own and their peers' writing. Shifts in meaning showed a slight developmental trend with sixth graders deviating from original meaning more than undergraduates. The study implies that teachers should be aware that students remember their self-generated texts very well and that students should invest considerable time in revision processes.