Visual field testing in conventional perimetry is critically dependent on a subject's ability to hold his/her gaze on a fixation target. Although a variety of methods has been designed to ascertain proper fixation in commercial perimeters (e.g. stimuli in the anatomical blind spot; subject observation), there are few reports concerning fixation stability during actual testing. We have recently started using the Scanning Laser Ophthalmoscope (SLO) for routine perimetry in patients with central scotomas, and performed a retrospective video tape analysis of their fixation. The position of a retinal landmark was followed throughout the test, and two measures of fixation stability were studied: short term, i.e. stability during a one-second target presentation; and long term, i.e. throughout the test. Preliminary data from 4 subjects indicate that long-term fixation varies over angles of 3 to 5 degrees, while shifts up to 2 degrees within a single target presentation occurred in all subjects; in one subject shifts up to 4 degrees were observed. Subjects' ability to maintain stable fixation varied greatly: the percentage of trials during which fixation shifted by less than 15' ranged from 0% to 52% (mean 26%); the percentage in which fixation shifted by more than a degree ranged from 29% to 63% (mean 41%). The distribution of momentary fixation loci also varied: some subjects show a broad two- dimensional distribution of these loci, whereas in others these loci form a small number of clusters. We conclude that the use of conventional perimetry in patients with central scotomas entails serious risks of inaccuracy. We are currently collecting additional SLO perimetry data in subjects with central scotomas, as well as in subjects with normal vision or intact central fields.