A pretest/post-test procedure for measuring information gain from discourse was investigated together with several other aspects of discourse processing. The main purpose was to determine the effect of a pretest on discourse learning as measured by post-test performance. The study also investigated (1) serial position effects in learning from discourse, (2) learning of factual versus relational information, and (3) information chunking of discourse material. Four hundred fifth-graders were used as subjects. The results indicated that (1) the pretest was an essentially neutral event, neither facilitating or depressing post-test performance; (2) almost all learning or retention was on factual as opposed to relational information; (3) negative recency serial position effects were obtained as a function of the order in which the information was presented in the passage, but no serial position effect was obtained as a function of test item order; and (4) no evidence was obtained for information chunking on a supra-sentence level. Tables, figures, and references are included. (Author/AW)