A KeyGraph-like algorithm, which incorporates the concept of structural importance with association rules mining, for analyzing closed-endedquestionnairedataispresentedin thispaper.Theproposed algorithm transforms the questionnaire data into a directed graph, and then applies association rules mining and clustering procedures, whose parameters are determined by gradient sensitivity analysis, as well as correlation analysis in turn to the graph. As a result, both statisti cally significant and other cryptic events are successfully unveiled. A questionnaire survey data from an instructional design application has been analyzed by the proposed algorithm. Comparing to the results of statistical methods, which elicited almost no information, the proposed algorithm successfully identified three cryptic events and provided five different strategiesfordesigning instructional activities.Thepreliminary experimental resultsindicated that the algorithm works outfor analyzing closed-ended questionnaire survey data.