Effects of observation cost on sequential search behavior

Sixty Ss participated individually in a search problem attempting to find a single target object located in one of several distinguishable regions. The search consisted of sampling observations with replacement from the various regions. Ss were told the prior probability of finding the target in a given region, the conditional miss probability that if the target is in a given region it will not be detected on a single observation there, and the cost per observation for each region. The effect of observation costs was investigated. Procedures devised for comparing actual and optimal search behavior showed that relative to the optimal policy Ss erred in the direction of maximizing detection probabilities rather than in the direction of minimizing the cost per observation, and that this error increased in magnitude the larger the differences among observation costs. A Markov model was tested and rejected. Implications of the study to visual search problems are discussed briefly.