Retroactive interference (RI) in prose learning was investigated in an experiment where passages were constructed on the basis of a predetermined logical structure. This structure made it possible to operationally define similarity and assess the effects of RI for inferential information as well as that stated directly in the original passage. Subjects, 40 college students, read the same passage, and half of them were immediately administered a retention test. Subjects then read either a passage highly similar to the first or a completely different passage. Three retention tests were then given to all subjects. A 2x2 design was used to test the effect of similarity and the influence of an immediate test of original learning upon later retention. Overall retention was significantly higher when the original learning test was included, but similarity appeared to affect RI on some measures only in the absence of the test. (Author/MS)