Three forms of filter technique for measuring random and directional locomotion of leucocytes have been compared: (1) the conventional one filter technique of Boyden (lower surface count method); (2) the two filter system with a lower cell-impermeable filter designed to count the cells at the underside of the upper filter as well as those on the lower filter (two filter count method); and (3) two filter systems counting only cells associated with the lower filter (lower filter count method). In some instances all three methods produce qualitatively similar results. In others totally different results are reproducibly obtained with identical cell preparations, media and attractants. Compared to the two filter count method, the lower surface count method and the lower filter count method are not sufficiently reliable. The discrepancies are partly due to errors in measuring the response. They are caused by variable cell adhesion to the filters resulting in a varying distribution of cells between the upper and lower filter and/or detachment of neutrophils from the upper filter. Some of the discrepancies are not due to errors in assessing the response, but to differences in gradient formation and drift of chemokinetic and chemotactic materials from one compartment to the other.