skip to primary navigation skip to content

Prof Horace Barlow

staff photo

Tel: +44 (0)1223 333813, E-mail:

I'm currently interested in testing a hypothesis about the role of motion blur in helping us to determine the direction of motion of moving images. Motion blur results from the failure of the photoreceptors to follow the rapid fluctuations of intensity that occur when an image moves, and it causes a severe loss of acuity in the direction of motion, even when the velocity is only a few degrees/sec. This is shown at top left by the 2-D spatial frequency filter calculated for motion at 2 deg/sec. Our hypothesis is that the static patterns shown below (called Glass patterns after their discoverer) mimic the patterns of attenuation of high spatial frequencies that occur from the patterns of motion caused by optic flow, which are the image movements that occur when an observer moves through a fixed environment. A paper (with Bruno Olshausen) describing the hypothesis and some supporting observations is available at

I have previously worked on other problems in vision and have titles and links to some other papers available on

fig03 image

Figure: Glass figures made by defining random positions for an array of dots, and then duplicating each dot either with a constant linear displacement, a constant outward displacement from a fixed centre, or a constant angular rotation about a fixed point. Most observers report seeing streaks, or apparent "flow", in addition to the actual dots. It is thought that this results from activation of a mechanism used by the eye to derive the directions of motion of patches of the image from the direction of the blur that results from motion.

spatial image

Above: This figure shows the extra attenuation at all orientations and spatial frequencies caused by 2 deg/sec motion of an image downward and to the right. It was calculated from measurements of Koenderinck and van Doorn (1979) made using counterphase flickering gratings. The colours crossed by the solid line show the attenuations at different frequencies for spatial frequency components modulated in the direction of motion. The red segments show the lack of attenuation at all spatial frequencies for modulation directions orthogonal to the motion, but the segments are narrow and should yield accurate information about the direction of the orthogonal. The dotted line is at 84° to the direction of motion, 6° from the orthogonal, and there is already considerable attenuation of high spatial frequencies.