Sources of variability in measurements of intraocular pressure (IOP) by applanation tonometry were examined with a repeated 3 X 3 Latin square design that allowed independent extraction of variation attributable to nine patients, three observers and the order in which the observers performed the measurements. The observers, who were experienced, did not differ significantly in the general level of the IOP readings they obtained or in the variability of the readings. The effect of order was significant; there was an approximately linear decrease in IOP with successive measurements, though the difference between the first and second readings was the largest. The intrinsic variability of the IOP readings gave approximate 90% tolerance intervals of +/- 4.5 mm Hg for a single reading and +/- 3 mm Hg for the average of two readings. The mean decrement between the first and second IOP readings, 1.0 mm Hg, implies the need for a consistent protocol for successive sessions of applanation tonometry.