We used a motion coherence paradigm to test the hypothesis that patients with retinitis pigmentosa (RP) have difficulty discriminating the direction of spatial displacements because of a random loss of motion-sensitive units owing to cone photoreceptor dropout. Minimum (Dmin) and maximum (Dmax) displacement thresholds of patients with typical RP or Usher syndrome were compared with those of age-similar, visually normal subjects. Two-frame random dot cinematograms were used, in which a group of target dots, which comprised 40-100% of the dot array in steps of 20%, were displaced in one of four directions, whereas the non-target dots were randomly repositioned between frames. Reducing the dot coherence in this way increased Dmin and reduced Dmax for both the RP patients and control subjects. Furthermore, the displacement thresholds of the RP patients were displaced laterally from normal along a log coherence axis, consistent with the hypothesis that the patients had a reduced effective (intrinsic) coherence. However, the displacement thresholds of control subjects, when measured at a reduced coherence, did not mimic those of RP patients at full coherence when both groups were tested with a range of dot contrasts and dot areas. These apparently discrepant findings can be reconciled if it is assumed that the patients' effective coherence varies with stimulus visibility.