Quantifying eye stability during a fixation task: a review of definitions and methods.

Several definitions, measurements, and implicit meanings of 'fixation stability' have been used in clinical vision research, leading to some confusion. One definition concerns eye movements observed within fixations (i.e., within periods separated by saccades) when observing a point target: drift, microsaccades and physiological tremor all lead to some degree of within-fixation instability. A second definition relates to eye position during multiple fixations (and saccades) when patients fixate a point target. Increased between-fixation variability, combined with within-fixation instability, is known to be associated with poorer visual function in people with retinal disease such as age-related macular degeneration. In this review article, methods of eye stability measurement and quantification are summarised. Two common measures are described in detail: the bivariate contour ellipse area (BCEA) and the within-isolines area. The first measure assumes normality of the underlying positions distribution whereas the second does not. Each of these measures can be applied to two fundamentally different kinds of eye position data collected during a period of target observation. In the first case, mean positions of eye fixations are used to obtain an estimate of between-fixation variability. In the second case, often used in clinical vision research, eye position samples recorded by the eyetracker are used to obtain an estimate that confounds within- and between-fixation variability. We show that these two methods can produce significantly different values of eye stability, especially when reported as BCEA values. Statistical techniques for describing eye stability when the distribution of eye positions is multimodal and not normally distributed are also reviewed.