Abstract
The ability to discriminate a tilted target among vertical distractors is reduced by increasing the set size, but attention to the target eliminates this effect. Here we test and model orientation search for the target under different stimulus and attention conditions. We measured performance for identifying the tilt of a small peripheral Gabor patch tilted either clockwise or counterclockwise from vertical among a varying number of distractors (1 to 7). We varied the nature of the stimulus either by adding orientation noise to both target and distractors, displayed at high contrast, or by displaying the stimuli at a near-threshold contrast. Attention was modulated by the use of a peripheral cue indicating the location of the target. The cue was simultaneous with the stimuli and lasted for the same duration (100 msec). The results show that cueing the target increased sensitivity by a factor of two, causing a leftward shift of the psychometric functions without affecting their shape; the functions in the cued and uncued conditions were parallel. This trend is constant across all the different condition. We used these results to test the ‘signal enhancement’ and ‘noise reduction’ hypotheses of attention through a model of search that differs from previous proposals. Our model includes early noise in the templates, which we have shown to be the main limiting factor in similar tasks. Moreover, threshold estimates reveal that the cue not only flattens the set size functions, but that there is also a 50% decrease in thresholds from set size 2 to set size 8, for both no noise and high noise. This result suggests that once the uncertainty about the target location is eliminated through an effective cue, observers use the noise samples to get a better knowledge of the noise distribution. The presence of this effect in the no noise condition suggests further that the observer does not know in advance what the distribution of internal noise is.