The optical disector is among the most efficient cell counting methods, but its accuracy depends on an undistorted particle distribution in the z-axis of tissue sections. Because the optical disector samples particle densities exclusively in the center of sections, it is essential for unbiased estimates of particle numbers that differential shrinkage or compression (and resulting differences in particle densities along the z-axis) are known and corrected. Here we examined, quantified, and compared differential shrinkage and compression of vibratome-, celloidin- and cryosections. Vibratome sections showed a significant z-axis distortion, while celloidin- and cryosections were minimally distorted. Results were directly compared with previous data obtained from paraffin and methacrylate sections. We conclude that z-axis distortion varies significantly between embedding and sectioning methods, and that vibratome-, methacrylate- and paraffin sections can result in grossly biased estimates. We describe a simple method for assessing differential z-axis shrinkage or compression, as well as simple strategies to minimize the bias of the optical disector. Minimal bias can be achieved by either adjusting the placement and extent of counting boxes and guard spaces for sampling, or by applying a correction factor in cases when guard spaces are deemed essential for particle recognition.