College London .4bstractmPC'sin_g ;1s data base the fatal pedestrian accidents occurring in London in 1970-71. the usefulness of police reports for estimating the speeds of road traffic accidents was investigated. Two approaches were used: correlating cstlmates by ditkrent people of the initial speeds and relating behick domagc to cstimakd impxt speed. The first approach established quite high correlations (around 0.6) hetwccn diflcrent ectimatcs. and in the second a positive association was found hctuecn speed and damage. Furthcrmorc. the arcrage dilkence between estimates made by indrpcndcnt witnesses und by the drtscrs of the \chlclcs involved was small (3 m.p.h.). In addition. a table is siven showing the distrlhutlon of cstlmntcd impact speeds for this sample of accidents. INTRODCICTIOS Speed is one of the most important hctors in road itccidcnts. both in influencing whcthcr or not an accident occurs and in dctcrmining the scvcrity of' injury sustained once the accirlcnt has hnppcncd. II' H'C arc studying the cticct of sonic deign chnngc to the interior of the car on drivers' injuries. LVC ~vould like to cumparc accidents occurring at the same speed. The s;~mc goes for evaluating the protection alforctcd by scat hulk. or crash hclmcts klr motorcyclists. II' wc arc intcrcstcd in whcthcr one model of car is more likely than iltlOthCr tu roll over in accidents. \vc would like to know whcthcr thcrc is any ditkrcnco in the speeds or the accitlcnts the tkvo moclcls arc involved in. sincc roll-over is more likely at higher speeds. and it would hc mislcding to damn moclcl A as being niorc prone to this if moticl I3 wcrc gcncrally intolvcd in Iowa--speed accidents. WC would like to know how much of the high xcidcnt rate of young drivers can bc xttributcd to high speed driving in unsuitable conditions. and part of the inlormation to rlatcrminc this would bc the spds of the accicicnts they arc involved in.