Abstract
Measures of scanpath similarity across different conditions or participants are essential in many research domains. Traditional string-edit methods compare fixations within a scanpaths according to their numerical and more recently temporal position within the paths. These procedures are reasonable when the to-be-compared scanpaths are executed in response to a relatively stable environment such as comparing scanpaths during picture viewing. In sequential sensorimotor tasks, participants actively change their environment while extracting task-relevant visual information with the eyes and these dynamical changes differ across trials. Some sub-actions are for instance elongated and accompanied by more fixations while others are shortened and accompanied by fewer fixations than during prior execution. Therefore, traditional methods for determining scan path similarity are not adequate in this case. The functional units of sequential sensorimotor tasks are the sub-actions in which the task is structured. Based on these sub-actions, a functional matching procedure for determining scanpath similarity had been developed (Foerster et al., 2011). This procedure evaluates the similarity of scanpaths according to the linear distances between fixation locations that have been performed during the same sub-action of a sensorimotor task. Distance values are evaluated by testing them against random distance values calculated from the same scanpaths. The method reveals whether participants look at similar locations when they are engaged in the same sub-action compared to different sub-actions. Extending our earlier work, the functional matching procedure was applied to fictitious scanpaths as well as to real scanpaths that had been recorded during a high-speed stacking task. The strength of the method could be demonstrated by comparing it with traditional string-edit methods.
Meeting abstract presented at VSS 2012