SCANPATHS: DEFINITIONS, THEORY AND APPLICATIONS

The term “scanpath” was first used by Noton and Stark in their papers published in 1971 in Vision Research and Science, entitled “Scanpaths in saccadic eye movements while viewing and recognising patterns” and “Scanpaths in eye movements during pattern perception”, respectively. Noton and Stark claimed that when a particular visual pattern is viewed, a particular sequence of eye movements is executed, and furthermore that this sequence is very important in accessing the visual memory for this pattern. Although Noton and Stark’s scanpath theory in this strong form has been disputed strongly over the years, especially the second postulate above, considerable evidence has been accumulated in experiments with natural images that the human gaze selects informative details, i.e. that eye fixations tend to fall on or near important details of images. In addition, arising from this interest in scanpaths, a number of powerful techniques have been developed recently that explore the statistical properties of eye-movement patterns and relate these with important characteristics of the visual stimuli.

For the purpose of this archive we will consider a “scanpath” to be any eye-movement data collected by a gaze-tracking device, where information is recorded about the trajectories (paths) of the eyes when scanning the visual field and viewing and analysing any kind of visual information. Such data usually consists of gaze-direction, fixation position and duration, and saccade duration. Our definition of scanpath is broader than the one used by most researchers, including Noton and Stark’s original definition, but it allows us to include data collected from different experiments and in various applications in a single framework. Although data from reading, eye-typing and other experiments can also be considered as scanpaths if we use this definition, the data that is included in the Scanpaths.org archive will be primarily from experiments dealing with the visual analysis of natural or computer generated images and video.