A study was conducted to determine if covert reader-generation of interspersed prequestions would affect recall of science-oriented prose. Sixty college freshmen in a basic skills reading course were divided into three groups: Group I received 5 hours of training and practive in the construction of self-generated questions, including recognition of topic sentences, paragraph organization, and differentiation of various question types; Group II was given, in 1 hour, a condensed version of the same training; and Group III, the control group, received alternative training. Subjects read science passages written at the 9th and 16th grade levels and completed multiple choice and competition criterion tests immediately and 1 week later. Results showed that (1) there was no overall effect for treatment; (2) treatment groups scored higher on the difficult passage and lower on the easier passage than did the control group, with Group I the highest and lowest respectively; (3) a strong trend was indicated for the treatment groups to score higher on delayed and lower on immediate recall measures; (4) there were no significant differences between training groups; and (5) no between-group differences resulted from the two types of criterion posttests. (Author/FL)