Human performance in visual search for multiple targets.

摘要

This study extended the models and data for locating a single target to search tasks for locating multiple targets. Multiple-target search tasks can be categorized into two types. In exhaustive visual search tasks, observers have to find all given targets, and in non exhaustive visual search tasks, observers must choose when to stop searching. The first topic of this study was to derive human performance models for exhaustive multiple-target search and then to validate the models. Random and systematic models for an exhaustive search task were derived which were expected to describe upper and lower bounds of human search performance. The proposed random search model was described by a hypo-exponential distribution, while the proposed systematic search model was a piece-wise curvilinear function. Human search performance data from a visual search experiment fitted between the two search performance models. Interestingly, based on an analysis of interview data and search performance data, participants' search behavior changed during these exhaustive multiple-target search tasks. The models showed that they searched the search fields (1) at faster speed, (2) with shorter dwell time in a fixation, and (3) with less revisits in a fixation in the initial period of the search task than in the late period. Such a phenomenon was more evident in a multiple-target search task than in a single-target search task. It was also more evident in a multiple-target search task for different-type targets than in a multiple-target search task for targets of the same type. The second topic of this study was to derive an optimal stopping time model for a non-exhaustive search task containing multiple targets. Three usage strategies of the optimal stopping time were compared: a self-stopping strategy, an externally forced stopping strategy and a hybrid stopping strategy. The self-stopping strategy was the most effective among the three strategies under almost all task conditions of different time pressure and different pre-information on the number of targets (known and unknown number of targets). Such effectiveness of the self-stopping strategy could be caused by human observers' situation awareness ability and their use of decision cues.