The distinction between similarity-based and rule-based strategies has instigated a large body of research in categorization and judgment. Within both domains, the task characteristics guiding strategy shifts are increasingly well documented. Across domains, past research has observed shifts from rule-based strategies in judgment to similarity-based strategies in categorization, but limited these comparisons to 1 prototypical environment, a linear task structure, and a restricted set of strategies. To systematically compare the 2 domains, we considered several instantiations of rule-based and similarity-based strategies and examined strategy choice across different types of judgment and categorization tasks. Between participants, we varied task characteristics from a 1-dimensional linear to a multidimensional linear and to 2 multidimensional nonlinear tasks. Irrespective of domain, strategies considered, or model comparison technique used, we find that more participants relied on similarity-based strategies when the functional relationship between the cues and the criterion was nonlinear. Shifts from rule-based strategies in judgment to similarity-based strategies in categorization, however, were rare and most pronounced in 1-dimensional environments. These results support the hypothesis that the cognitive strategies people select to solve a judgment or categorization task depend less on the domain but more on the complexity of the task.