Reversals in perceived direction of motion of a grating when its spati
al frequency exceeds half that of the sampling mosaic provide a potent
ial tool for estimating sampling frequency in peripheral retina. We us
ed two-alternative forced-choice tasks to measure performance of three
observers detecting or discriminating direction of motion of high con
trast horizontal or vertical sinusoidal luminance gratings presented e
ither 20 or 40 deg from the fovea along the horizontal meridian. A fov
eal target at a comfortable viewing distance aided fixation and accomm
odation. A Maxwellian view optometer with 3 mm artificial pupil was us
ed to correct the refraction of the peripheral grating, which was pres
ented in a circular patch, 1.8 deg in diameter, in a surround of simil
ar colour and mean luminance (47.5 cd . m(-2)). The refractive correct
ion at each eccentricity was measured by recording the aerial image of
a point after a double pass through the eye. The highest frequency wh
ich can reliably be detected (7-14 c/deg at 20 deg, 5.5-7.5 c/deg at 4
0 deg) depends critically on refraction. Refraction differs by up to 5
D from the fovea to periphery, and by up to 6 D from horizontal to ve
rtical. Direction discrimination performance shows no consistent rever
sals, and depends less on refraction. It falls to chance at frequencie
s as low as one-third of the highest that can be detected. Gratings wh
ich can be detected but whose direction of motion cannot be discrimina
ted appear as irregular speckle patterns whose direction of motion var
ies from trial to trial. The absence of motion reversals may reflect i
rregularity of sampling, and suggests that reversals are not a simple
tool for studying sampling in peripheral vision.