In two experiments a total of 662 high school students read a prose passage, took a verbatim or paraphrase quiz, and a week later completed a verbatim or paraphrase delayed test. Taking a quiz significantly enhanced performance on the delayed test. Performance was consistently much higher on the verbatim than on the paraphrase forms of quizzes and tests. Fitting the data rather well was a theory which assumes that a verbatim question is best at evoking retrieval of phonologically coded information in short term memory whereas a paraphrase question is best at instigating transfer of the information into long-term semantic memory. (Author)