In an experiment using a conditional logic task, subjects were given a set of cards defining a particular conditional relationship. The subjects were required to determine the rule and predict the outcomes of the rule. Rule structure was varied by providing subjects with either a suggested rule, a partial rule, or no rule for each set of cards. Other factors that were manipulated included problem type (use of negations in syllogisms) and problem dimensionality (level of task difficulty based on number of choice items). It was found that when task difficulty was low, subjects with the most rule structure used a matching strategy and showed higher error rates than those with less rule structure. When task difficulty was high, all subjects showed the matching strategy. Qualitative differences were found in the errors made by the two groups, suggesting the need for further research on both qualitative and quantitative differences in conditional problem solving. Appended tables present statistical data from the experiment. (Author/MSE)