Three experiments tested a signal-detection theory (SDT) model of visual search (e.g. as described in J. Palmer, C. T. Ames, & D. T. Lindsey, 1993). In Experiment 1, participants searched for a 0° line among distractors at (a) 30°; (b) 1/3 at 30°, 2/3 at 50°; ( c) 1/3 at 30°, 50°, and 70°; (d) 1/3 at 30°, 2/3 at 70°. The SDT model predicts improved performance in the more heterogeneous conditions, as some distractors are more discriminable from the target. In contrast, in Experiment 1, human performance degraded in the more heterogeneous conditions ( c) and (d). In Experiment 2, sparser displays improved the performance of the SDT model. In Experiment 3, search for θ° among homogeneous θ + 20° distractors was compared with search for θ° among θ ± 20° distractors. Performance in the latter condition was often worse, relative to performance in the homogeneous condition, than predicted by the SDT model; however, this depended greatly on the identity of the target.