In cognitive science there is a paradox: Researchers studying decision making have repeatedly shown that people employ simple and often less than optimal strategies when integrating information from multiple sources. However, researchers working in fields such as categorization, memory, and perception have had great success using optimal models to account for information integration. Is this conflict due to the use of different materials and procedures? We test the hypothesis that stimuli requiring more controlled information integration lead to suboptimal performance, while stimuli that lend themselves to more automatic processing produce more optimal integration. We test for one canonical example of sub-optimal information integration, the dilution effect, using stimuli more commonly found in perception experiments. Dilution was indeed reliable across several conditions. The largest effects occurred in stimuli manipulated so as to discourage automatic processing. We use the Multi-component Information Accumulation model to explain how stimulus presentation influenced cognitive processing.