1.   In this earlier study a group of subjects gave unconstrained accident estimates for various manoeuvres at a set of traffic lights.

2.   To avoid some of the problems associated with such a broad range of responses it was decided to limit the potential variance in accident estimates for the following study.

3.   Accident estimates and risk ratings are strongly related for individual subjects.

4.   For accident estimates this is as would be expected since subjects were specifically estimating the total number.

5.   These ratings were significantly lower than the accident estimates and interacted differently with junctions, nonetheless, risk ratings and accident estimates were highly intercorrelated.

6.   In all cases the correlation of accident estimate with the other variables was higher than their correlation with risk ratings.

7.   However, the number of subjects is very small for such an analysis and self reported accident estimates themselves are not a very reliable measure.

8.   The interaction between risk rating and accident estimate across the ten junctions suggests that subjects were able to sensibly dissociate the two judgment tasks in the laboratory setting.