The tilt-shift effect is a photographic technique in which blur is exaggerated by slanting the image plane, which causes the scene to look small. This makes sense because discriminable blur only occurs when objects in a real scene are relatively close. We used the effect to study how focus cues—blur and accommodation—affect perceived size. A computer-generated image was rendered in sharp focus and blurred in various ways. Four conditions were used, with the validity of focus cues increasing in the following order. 1) The image was shown on a single display surface. 2) A linear blur gradient was applied to the image on a single display surface and oriented to roughly match or lie orthogonal to the depth gradient in the scene (Vishwanath, 2007). 3) The image was shown on one surface but each pixel was blurred to yield the same retinal-image blur as the original scene; this required measuring the aberrations of each observer's eye and appropriately modeling blur due to defocus and aberrations. 4) The sharp image was presented on three planes at different focal distances; blur and accommodative signals from the stimulus were therefore consistent with the signals that would be created by the original scene. The blur and/or accommodation signals were manipulated to be consistent with near (small) scenes or far (large) scenes. We presented two stimuli on each trial and observers indicated the one in which a central unfamiliar object appeared smaller. The results showed that blur and accommodation had a large effect on perceived size in condition 4. Blur had successively less effect in the other conditions as the validity of the cue decreased. Thus, perceived miniaturization depends on how closely the pattern of blur is matched to the depth values within an image, and on whether dynamic blur and accommodation are appropriate.