The holes in hemoglobin-free ghosts generated by osmotic lysis of human erythrocytes have been characterized. The efflux from ghosts of water-soluble probes with a Stokes radius of 5,.2 to 61 hi was measured, The kinetics was usually first order, suggesting a homogeneous ghost population, but became complex because of sieving when hole and probe were approximately the same size. The area of exit per ghost, calculated from Fick's law of diffusion, varied widely but the number of holes per ghost, calculated from the sieving of pairs of probes, was always unity. The hole had a circular rather than elliptical shape and a path length of 60 A, approximately the thickness of the membrane. Ghosts were induced to partially seal so as to trap various probe molecules at diffusional equilibrium. The sustained retention of probes ruled out the possibility that the holes are impermanent, intermittent breaches in membrane continuity. The dispersion of hole size under a given set of conditions was fairly narrow. Centrifugation on density barriers composed of appropriate solutes separated ghosts populations into fractions with holes larger than the solute (which pelleted) and smaller (which floated). The fractional floatation of ghost populations on barriers of dextran, sucrose, mannitol, and CsCl was calibrated with estimates of their hole radii obtained from efflux kinetics to establish a rapid, simple, and precise technique for estimation of mean ghost hole size. The average difference between the hole radius measured by density barrier and equilibrium trapping methods was 4 & 8 S.D. A. Holes could be reduced at high ionic strength to 7 hi in radius and ilated at very low ionic strength to >lo4 b (1 pm), at which point they became visible in the dark-field microscope as a single round 1esion.The hemolytic hole is thus a continuously-tunable molecular filter, the area of which can be modulated over more than a million-fold in a defined fashion.