Abstract
Purpose: In the Posner task, the observer has to detect a target that might appear in one of two locations. On each trial, a precue indicates the probability of the target appearing at that location. Accuracy in those trials where the target appears at the cued locations (valid cue trials) is typically higher than in the trials where the target appeared at the uncued location (invalid cue trials). The cueing effect is commonly interpreted as a shift of attentional resources to the cued location leading to enhanced processing. Alternatively, an ideal observer model (in which selective attention simply weights the target likelihood at each location by the cue validity) also predicts a cueing effect. The classification image technique allows the investigator to estimate how the observer weights the data to reach a decision. Here, we use the technique to study how information is combined at the cued and uncued locations. Methods: The task was to detect a contrast increment (4.7 % contrast)) in one of two Gaussian disks (sd = 12 min; ) located in a field of white noise (stdev = 3.5 cd/m2, mean luminance = 29.8 cd/m2). The pre-cue was presented for 50 ms. The stimuli were presented for 150 ms. The target was present on 50 % of the trials. On target present trials, the target appeared at the cued locations on 80 % of the trials. The noise fields from the false alarm trials were averaged to obtain classification images for the cued and uncued locations. Monte Carlo simulations were used to generate classification images for the ideal Bayesian observer. Results: The classification images for both the ideal and human observers at the cued location show a significantly larger magnitude than at the uncued location. Conclusions: Classification images suggest that human observer weight the information at the cued location more heavily than the uncued location. This weighting of the information is consistent with the ideal observer and predicts a cueing effect.
Supported by NIH 53455, NASA-NCC 1027 and NASA NAG 9-1157 to ME.