Regressions re-visited

Abstract

We revisit the definition of regressions in eye tracking, having found existing definitions, formulated within a reading paradigm, unsuitable for visual display assessment. The new definition is tested using eye movement data recorded during a usability evaluation of two series of graph designs. The new definition gave a stronger result in the usability assessment than was obtained using the previous measure, and gave a result consistent with other eye movement usability metrics derived for the two series of graph designs. The newly defined measure is easily computed and is sensitive enough to reveal a correlation between improved performance and a decrease in regressions.