The importance of constructive comprehension processes in learning from tests

The goal of these experiments was to introduce and test the constructive retrieval hypothesis, according to which retrieval practice will be most effective when it encourages constructive elaborations of text content. Experiment 1 provided baseline performance data for the materials included in Experiments 2 and 3. In Experiment 2, instilling inference-based test expectancies before an initial retrieval attempt led to more constructive retrieval practice and better final test performance than instilling detail-based expectancies. In Experiment 3, instructions to construct explanations during initial retrieval attempts led to more constructive retrieval practice than free recall, and better final test performance than free recall or rereading instructions. These experiments support a constructive retrieval account of testing effects, and demonstrate that it is not retrieval practice alone, but rather the kind of constructive processing invoked during retrieval attempts that can improve both retention and comprehension when learning from text.