Automimicry is the existence of palatable individuals in an unpalatabl
e, warningly coloured (i.e. aposematic) prey species. Because automimi
cs are visually indistinguishable from their models, they present a sp
ecial problem for the evolutionary stability of aposematism. Tradition
ally this problem has remained either unrecognized, or has been consid
ered solved by the actions of naive predators that attack warningly co
loured prey, eating them if palatable, rejecting them if unpalatable.
Although naive predators can stabilize aposematism against invasion by
automimics, the conditions under which they can do so are restricted.
It is argued here that an adjustment to the traditional view of how a
posematic signals operate can offer a more general solution to the pro
blem of automimicry. The ''go-slow signalling'' solution proposes that
some predators learn not to avoid aposematic prey, but to sample them
cautiously in order to determine their true palatability more accurat
ely. Because unpalatable prey can benefit by advertising conspicuously
to go-slow predators, but cannot benefit by advertising to naive pred
ators (unless predators are innately cautious of conspicuously coloure
d prey), go-slow signalling can stabilize aposematism under a wider ra
nge of conditions than can traditional naive predators. The assumption
s, predictions, and empirical implications of the go-slow signalling h
ypothesis are discussed.