Models of the evolution of aposematic coloration assume that the first brightly coloured morphs to appear in a cryptic population of chemically defended insects would have attracted the attention of predators, putting them at a disadvantage by comparison with their less conspicuous relatives. However, if birds show a generalised avoidance response (neophobia) towards novel prey, this could allow conspicuous morphs to survive long enough to enable aposematic selection to take over. The present experiment attempted to evaluate this possibility by pre-training individually identifiable wild blackbirds (n = 19) and robins (n= 12) with pastry baits of a particular colour and then giving them a choice between familiar-coloured and novel-coloured baits. The number of trials taken for each bird to contact the novel colour of bait for the first time, and to eat it on a regular basis, was recorded. Birds were extremely variable in their responses: some accepted novel-coloured baits immediately while others took more than a hundred exposures before eating them regularly. Green baits were avoided for longer than red or yellow. The results suggest (a) that a new aposematically coloured morph would not necessarily suffer a higher predation rate than its cryptic ancestors, and (b) that colours that are normally considered to be aposematic on the basis of laboratory experiments may not be especially aversive to wild birds.