Traffic flow data from 21 roundabout 'weaving' sections have been analysed to check the validity of Wardrop's formula for the design of conventional roundabouts, and to search for improved methods of capacity prediction. On average, capacities were eight percent below the 'practical capacity' predicted by Wardrop's formula. The proportion of weaving traffic made no significant contribution to accuracy; with this factor removed a revised version of the formula was developed, and has already been published in the Department of the Environment Technical Memorandum H2/75. Further work centred round the fitting of a linear relationship between traffic entering the roundabout from a saturated approach and that already in the circulating carrtiageway. Results showed a moderate (but non-significant) improvement over the original and revised versions of Wardrop's formula in terms of between-sites variation, and a considerably improved representation of the within-sites variation. Further checks will be made against more data now being obtained