The Modified Packard method of discourse analysis, based on eight functional aspects of discourse relationships--elaboration, reason, comparison/contrast, temporal, emphasis, coordination, subordination, and superordination--was applied to the open-ended recalls of randomly selected honors students trained in the Evelyn Wood method and to a control group. The analysis revealed that when time, but not amount of material was held constant, slower readers showed significantly greater gains on recall of operations, reason, coordination, temporal, and emphasis nodes, and averaged more nodes per cluster. Rapid readers recalled more clusters. When time and amount of material were held constant, rapid readers showed significantly greater gains on total operations: reason, comparison/contrast, and coordination nodes, and more clusters. An objective test designed to measure concept application, paraphrase, and verbatim recall showed no significant difference between the experimental and control groups. A discourse analysis of the test failed to reveal why it did not discriminate the comprehension differences between groups. (Author)