The relationship between scene and eye movements

Individual differences make it difficult to recognize similarities between individuals in eye movement patterns. However, if consistencies can be found, eye movements could be used for a variety of purposes. In this study, the consistency of eye fixation patterns is explored using statistical evaluation as well as neural networks. Eye tracking data is used as input units into neural networks to test whether they can learn to associate task patterns with that task data. A network is successfully used to determine whether a participant was searching text on the screen or counting groups of arrows from eye fixation information. Another network is trained to determine whether the participant was counting a group of 1-3 arrows or 4-6 arrows, as would be expected according to the tendency to "subitize," or recognize small numbers of objects without counting each of them. The implications of the findings are discussed.