Crossbucks, flashers and gates are used to warn motorists about the danger of trains at grade crossings. The technological and institutional setting in which these devices evolved is described first.

The safety of a grade-crossing should be estimated by mixing information about causal factors (train and traffic flows, type of device, geometry etc.) with the accident history of the site. We suggest a coherent procedure for doing so. This procedure is of general interest and applies not only to grade-crossings but also to intersections, road segments, drivers and vehicles.

Currently used estimates of the safety effect of warning devices are incorrect. We provide support for this claim. The aforementioned coherent procedure for estimating safety enables us to furnish corrected estimates.