For migratory species, the timing of arrival at breeding grounds is an important determinant of fitness. Too early arrival at the breeding ground is associated with various costs, and we focus on one understudied cost: that migrants can experience a higher risk of predation if arriving earlier than the bulk of the breeding population. We show, using both a semi-analytic and simulation model, that predation can select for later arrival. This is because of safety in numbers: predation risk becomes diluted if many other individuals, either con- or heterospecific, are already residing in the area. Predation risk dilution can also select for more synchronous arrival because deviating from the current population-wide norm to earlier or later dates leads to higher predation risk or to failures in territory acquisition, respectively. The fact that selection for high arrival synchrony can in some cases be more important than selection for a specific date (early or late) within the season is an example of an ‘evolutionary priority effect’: whichever strategy – in this case a particular arrival time – becomes established in a population can remain stable over long periods of time; there are many possible equilibria (multiple stable states) which the population can remain at. Mixed arrival strategies are also possible under some circumstances.