1. Responses of cat retinal ganglion cells to a drifting sinusoidal grating stimulus were measured as a function of the grating orientation and spatial frequency.2. The response at fixed frequency and contrast varied with orientation in the manner of a cosine function. A new measure was introduced to quantify this orientation bias in the response domain on an absolute scale of 0-100%. Under experimental conditions designed to maximize the effect, the mean bias for 250 cells was 16% and the range was 0-46%. In 70% of cells there was significant bias.3. Orientation bias varied with spatial frequency and was maximal near the high-frequency limit. The majority of biassed cells preferred the same orientation at high and low frequencies but in some cells a reversal occurred: the orientation which gave maximum response at high frequencies gave minimum response at low frequencies. The greatest variation of cut-off frequency with orientation was (2/3) octave.4. Orientation bias was due to neural, not optical, factors. Nevertheless, the phenomenon could often be imitated by deliberately introduced optical astigmatism of up to 4 dioptres for brisk-sustained units and over 10 dioptres for brisk-transient units.5. The grating orientation preferred by cells varied systematically with position in the visual field. The central tendency was for the grating which yielded maximum response to lie parallel to the line joining the cell to the area centralis. This generalization failed for units within 2 degrees of the centre of the area centralis.6. Analysis of orientation bias indicates a functional asymmetry of receptive fields such that the centre mechanism, and sometimes also the surround mechanism, is elongated along the line joining cell to area centralis.