The aim of this study of air traffic on North Atlantic tracks is to examine whether an increase in the reported sighting rate of aircraft, identified as straying from an adjacent track, can be used for estimating a change in lateral collision risk, i.e. in the risk which is due to imperfect track-keeping. It is found that the ratio of the frequency of such sightings to the collision frequency is not large enough to detect even a quite severe change in the collision frequency.