When analyzing open-ended or categorical questions, many times responses are cross-classified by other categorical variables. The resulting contingency tables are then analyzed using the chi square test of independence. This procedure leads to multiple significance tests and provides no method to assess higher order interactions. Loglinear modeling, however, permits the user to assess interaction effects as well as the effects of variables. This study compared the results of using these two procedures. Data from three open-ended questions (about program strengths, weaknesses, and suggested improvements) in a survey of high school participants in an interactive video program were analyzed by multiple chi square tests of independence and loglinear modeling. Sample size ranged from 204 to 146 for the three questions. In two cases (strengths and suggested improvements), the results were similar. The question about weaknesses, however, required a three-way interaction. Since higher order interactions are not possible with the chi square test of independence, it is concluded that loglinear modeling provides a more effective way of analyzing multi-way contingency tables. (Contains three figures.) (Author/SLD)