Reliable, noninvasive biomarkers that reveal the internal state of a subject are an invaluable tool for neurological diagnoses. Small fixational eye movements, called microsaccades, are a candidate biomarker thought to reflect a subject's focus of attention [Z. M. Hafed, J. J. Clark,
VisionRes.
42
, 2533–2545 (2002); R. Engbert, R. Kliegl,
VisionRes.
43
, 1035–1045 (2003)]. The linkage between the direction of microsaccades and attention has mainly been demonstrated using explicit and unambiguous attentional cues. However, the natural world is seldom predictable and rarely provides unambiguous information. Thus, a useful biomarker must be robust to such changes in environmental statistics. To determine how well microsaccades reveal visual-spatial attention across behavioral contexts, we analyzed these fixational eye movements in monkeys performing a conventional change detection task. The task included two stimulus locations and variable cue validities across blocks of trials. Subjects were adept at the task, showing precise and graded modulations of visual attention for subtle target changes and performing better and faster when the cue was more reliable [J. P. Mayo, J. H. R. Maunsell,
J. Neurosci.
36
, 5353 (2016)]. However, over tens of thousands of microsaccades, we found no difference in microsaccade direction between cued locations when cue variability was high nor between hit and miss trials. Instead, microsaccades were made toward the midpoint of the two target locations, not toward individual targets. Our results suggest that the direction of microsaccades should be interpreted with caution and may not be a reliable measure of covert spatial attention in more complex viewing conditions.