Visual selection is imperfect; whenever a complex array of objects is processed, representations of multiple objects are likely to be active simultaneously. A full account of attentional processing must explain how these representations affect one another and how they interact to produce a response. Evidence on these interactions comes from measures of distractor interference and from dilution of distractor effects by other nontargets. Based on these data, different principles have been proposed to help understand target-distractor interactions, including accounts based on perceptual load and on dilution among nontargets. We review evidence from a number of experiments, including some using Yantis and Jonides's (Journal of Experimental Psychology: Human Perception and Performance, 10, 601-621, 1984, Journal of Experimental Psychology: Human Perception and Performance, 16, 121-134, 1990) methods for preventing abrupt onsets, which can disrupt spatial attention. The results underscore spatial constraints on the allocation of attention to include targets and exclude distractors. Selection is most effective when a single region can be selected that includes all possible target locations and excludes possible distractor locations. This region can be expanded or contracted as needed for the task, as suggested by C. W. Eriksen and St. James's (Perception & Psychophysics, 40, 225-240, 1986) zoom lens model. This attentional zoom setting is probably affected by a number of factors, including the number of nontargets, the similarity among stimulus elements, the discriminability of the possible targets, and the discrimination difficulty of a concurrent task. A narrower attentional zoom setting that excludes a distractor will prevent interference from that distractor. Interference from a distractor will be diluted by nontargets, but only if they are within the attentional zoom region.