Traditional theories of attention rely on the idea that when we search for a target in a visual display the brain boosts the activity of neurons optimally tuned for the target features. In this issue of Neuron, Navalpakkam and Itti take a computational approach to show that this strategy is actually very inefficient when the target is surrounded by distractors with similar features. Instead, the optimal strategy is to boost the activity of neurons that best discriminate between target and distractors, while essentially ignoring the neurons that respond best to the target.