In four experiments, we examined selection processes in visual search using a probe detection task to measure the allocation of attention. Under preview search conditions, probes were harder to detect on old relative to new distractors (Experiment 1). This cannot be attributed solely to low-level sensory factors (Experiment 2). In addition, probe detection was sensitive to color-based grouping of old distractors and to color similarity between old distractors (Experiments 3 and 4). These effects were dissociated when the color of the old distractors changed but probe detection effects remained. Collectively, the data indicate both group-based suppression of distractors and the separate inhibition of distractor features in search.