A central bias in spatial selection has been proposed to explain the decreasing search efficiency with increasing target eccentricity that results when distractors can occur closer to fixation than the target (J. M. Wolfe, P. O'Neill, & S. C. Bennett, 1998). The authors found evidence for such a bias using an odd-man-out variant of conjunction search. However, the bias was absent for the same displays when the identity of the odd-man-out target was known in advance. The authors propose that (a) top-down knowledge of a target feature supports grouping on this feature and (b) grouping links a peripheral target to central distractors expressing the same feature, increasing the attentional weighting afforded to the target and, consequently, facilitating its detection. The effects are independent of bottom-up priming effects occurring across trials. Thus, feature-based grouping can be driven top-down and can overrule the central bias in spatial selection.