| An experimental study was conducted with humans performing a conjunctive visual search task introduced by Treisman and Gelade (1980). A detailed analysis of eye movements and fixations confirms the central elements identified by Motter and Belky (1998) with regard to the roles played by perception, selection, attention and object recognition in their experiments with trained monkeys, but further reveals and characterizes the critical role played by memory . This finding raises important questions about the widespread notion and use of animal models as a proxy for humans in basic visual search tasks.; A major effort and innovation of the present study consisted in developing a dynamical model and corresponding computer simulations that faithfully reproduce the main patterns and characteristics of the observations. The statistical analysis of the measured data indicates that the probability of saccading from object to object depends mainly on three elements, namely, the distance between objects, the angle-shift between consecutive saccades, and memory. The element of randomness in the visual search is captured by the model through a Monte Carlo selection of a particular eye movement from one location to another based on the probability of that particular eye movement, which is in turn determined by the values assumed by the three factors for the considered consecutive fixations and their connecting and preceding saccades and fixations. The angle-shift and the memory factors depend on two corresponding parameters adjusted to match the measured angle-shifts and return rates. The memory factor further depends on the size of the area of conspicuity (AC) for object recognition, which is adjusted so that computer simulations optimally reproduce the saccade amplitude distribution, the distribution of the number of saccades needed to capture the target, and the failure rate.; The computer simulations are consistent with the statistical analysis of the observations in all critical aspects, confirming most notably the prevalence of circular or spiraling trajectories, and explaining that prevalence in terms of the inhibition of return to previously inspected locations and the additional bias to saccade in the forward direction forcing together the tendency to saccade along the display boundary, once reached within the AC. |
