In this note I wish to enter however belatedly into the discussion of error distributions in navigation (cf. Anderson and Anderson and Ellis). Not being a navigator my arguments will be purely statistical, but I hope that they will throw some light on the problem, which seems to be an important one.

Hampton and Mills have commented that large errors occur more frequently than is predicted by gaussian behaviour; and Anderson remarks that in practice the ‘skirts’ of the empirical distribution are commonly hitched high over the estimated gaussian tails. The conclusion drawn is that navigational errors follow a distribution closer to the two-sided negative exponential than to the normal though, instead of peaking to a sharp point at the centre, the empirical distributions seem to have rounded heads.