Objective-Speed at the time of a bicycle crash is an important determinant of the amount of energy transmitted. Controlling for speed is thus important in the evaluation of outcomes and effectiveness of intervention strategies. This study was conducted to evaluate the accuracy of self reported speed in a population of recreational cyclists. Methods-Children's and adults' bicycle speeds were measured with a radar gun as they rode along a closed road at weekend recreational events. Cyclists were then stopped and asked to estimate their speed. Measured speed, cyclist's estimate oftheir speed, age, and sex were documented. Parents were also asked to estimate their child's speed. Results-One hundred and fifty two cyclists from 4 to 80 years of age participated. Seventy per cent were children 13 years of age or younger. The mean (SD) speed of this group was 8.9 (2.5) mph. Cyclists age 14 and older were traveling at a mean speed of9.7 (2.87) mph. Estimated speeds were significantly higher than measured speeds for this older group, but there was no significant difference between mean measured and estimated speeds for the younger riders. There was also no significant difference between measured and estimated speed for males and females. Parents estimated their child's speed quite accurately. Conclusions-Self reported speeds for children were in close agreement with measured speeds and, thus, are sufficiently accurate to be a useful measure of crash severity in evaluating heimet effectiveness. (Injury Prevention 1997; 3: 43-45)