We manipulated categorical typicality and the presence of conflicting information as participants categorized multifeatured artificial animals. In Experiment 1, rule‐irrelevant features were correlated with particular categories during training. In the test phase, participants applied a one‐dimensional rule to stimuli with rule‐irrelevant features that were category‐congruent, category‐incongruent, or novel. Category‐incongruent and novel features delayed RT and P3 latency, but had no effect on the N2. Experiment 2 used a two‐dimensional rule to create conflict between rule‐relevant features. Conflict resulted in prolonged RTs and larger amplitudes of a prefrontal positive component, but had no impact on the N2. Stimuli with novel features did elicit a larger N2 than those with frequent features. These results suggest limitations on the generality of the N2's sensitivity to conflicting information while confirming its sensitivity to attended visual novelty.