Background configuration and rod threshold.

1. This paper investigates the variation in rod threshold when a small test flash is seen against backgrounds of different sizes. Over a substantial range of luminances above absolute threshold, the test flash is less easily seen against small backgrounds than large. This confirms earlier results.2. If an annular surround is added to a small circular background, threshold is reduced when background and annulus are equiluminous (uniform field), but rises rapidly as the annulus is made brighter or dimmer than the background. This cannot be explained by the threshold-elevating effects of light scattered on to the background from the surround, for threshold rises with annulus luminance faster than it does on uniform fields of equal luminance.3. If the surround is not a complete annulus but a windmill-shaped cross, threshold is higher than on a uniform field, no matter what the windmill luminance. Thus it is not the addition of light per se to the surround which reduces threshold.4. This conclusion is reinforced by the results of another experiment. The test flash is seen on a large uniform field. When superimposed on this field, a thin ring, light or dark, which causes only a small change in mean luminance, produces an appreciable rise in threshold.5. The addition of an equiluminous red surround to a small red background so as to create a uniform field causes a marked drop in test flash threshold, but a scotopically equal blue surround, that creates a uniform field for rods, does not alter the threshold. Since the test flash is seen only by rods it follows that signals from cones can alter rod threshold.6. Known or probable behaviour of retinal mechanisms cannot account for our results. All the operations which elevate threshold above its level on a large uniform field produce contours in the vicinity of the test flash. This we take as evidence that signals from the test stimulus are suppressed or reduced by other signals present only when the background is locally non-uniform.