A sentence completion task was used to investigate age differences in childrens' ability or inclination to invoke premises and inferences during prose processing, tasks which (according to the constructive hypothesis) adults typically perform. A pilot study confirmed that the sentence completion task was preferable to the recognition paradigm, because the latter does not differentiate between subjects who remember premises, inferences, or integrated representations of an array. Participants, 112 subjects each from grades five and eight, were tested for premise and inference memory. The study design varied the following between-subject factors: (1) information (premise, inference)--to independently assess retention; (2) grade (fifth, eighth)--to verify age differences; (3) conditions--to vary degree of contextual support for inference making; and (4) test lists (eight lists)--for counterbalancing. On inference, as expected, eleventh graders scored higher than fifth graders. This pattern of performance was not altered with the addition of context. No age differences were evident for the premise task; both age groups were highly proficient. Contrary to the constructive hypothesis, the benefits of supplying context were not limited to the younger group. Furthermore, neither age group performed above chance expectations on inferences in the standard condition. (Author/CP)