A fundamental concern when using visual presentations to study cerebral asymmetry is to ensure that stimuli are presented with the same degree of retinal eccentricity from a central fixation point in either visual field. However, a widely used procedure intended to control fixation location merely instructs participants to fixate appropriately without any other means of ensuring that central fixations actually occur. We assessed the validity of assuming that instructions alone ensure central fixation by using the traditional RVF advantage for words and either (a) only instruction to fixate centrally, or (b) eye-tracking device that ensured central fixation on every trial. Experiments 1 and 2 found that when only instructions were given, the vast majority of fixations were not central, and more occurred to the right of centre than to the left. Moreover, the prevalence of non-central fixations was otherwise disguised by the finding that both fixation procedures produced similar RVF advantages in overt performance. The impact of typical non-central fixations on performance was revealed by systematically manipulating fixation location in Experiment 3, where deviations in fixation of only 0.25 degrees from centre had a reliable impact on visual field effects. Implications of these findings for studies of cerebral asymmetry are discussed.