Conventional stereoscopic displays present conflicting stimuli to vergence and accommodation, causing fatigue, discomfort, and poor stereo depth perception. One promising solution is 'depth filtering', in which continuous variations in focal distance are simulated by distributing image intensity across multiple focal planes. The required image-plane spacing is a critical parameter, because there are constraints on the total number that can be used. Depth-filtered images have been shown to support continuous and reasonably accurate accommodation responses with 1.1 dioptre (D) imageplane spacings. However, retinal contrast is increasingly attenuated with increasing image-plane separation. Thus, while such stimuli may eliminate the vergence-accommodation conflict, they may also unacceptably degrade stereoscopic depth perception. Here we measured stereoacuity, and the time needed for stereoscopic fusion, for real targets and depthfiltered approximations to the same stimuli (image-plane spacings of 0.6, 0.9 and 1.2 D). Stereo fusion time was reasonably consistent across conditions. Stereoacuity for depth-filtered stimuli was only slightly poorer than for real targets with 0.6 D image-plane separation, but deteriorated rapidly thereafter. Our results suggest that stereoscopic depth perception, not accommodation and vergence responses, is the limiting factor in determining acceptable image-plane spacing for depth-filtered images. We suggest that image-plane spacing should ideally not exceed ~0.6 D.