Abstract: many studies have related changes in accident frequency to the presence of street lighting, and a few have examined its variation over a range of lighting quality, as measured by illuminance. this investigation attempts to find which of several measures of lighting (describing quantity – as represented by luminance or illuminance – uniformity and glare) most clearly explain variations in accident frequency. about 100 lit sites, almost all in built-up areas, were measured for lighting quality in dry-road conditions. the lighting variables measured were related to the dark:day ratios of accident frequency for the same sites. the strongest relationship found was that for average road surface luminance: in the range 0.5-2.0 candelas/m2, it is estimated that an increase of 1 cd/m2 is associated with a 35 per cent lower accident ratio. other measures of luminance and illuminance were also found to be related to accident ratio (and to each other), but not as clearly as was average road luminance, which is therefore the preferred explanatory variable. analysis of pedestrian and non-pedestrian accidents separately did not reveal a relationship between the former and lighting quality. in contrast, non-pedestrian accidents showed similar relationships to those for all accidents, with the addition of a possible relationship with overall uniformity of luminance.(a)