Many years ago, C.B. Martin drew our attention to the possibility of ‘finkish’ dispositions: dispositions which, if put to the test would not be manifested, but rather would disappear. Thus if x if finkishly disposed to give response r to stimulus s, it is not so that if x were subjected to stimulus r, x would give response z; so finkish dispositions afford a counter-example to the simplest conditional analysis of dispositions. Martin went on to suggest that finkish dispositions required a theory of primitive causal powers; there, I think, he was mistaken. All that they require is an improved conditional analysis, and this improved analysis can be built upon whatever treatments of properties and of laws we may favour on other grounds.