Let there be M hypotheses H 1 ,... , HM, and let Y be a random variable, taking values in a set y, with different probability distribution under each hypothesis. A quantizer -': Y-+ {1,..., D} is applied to form a quantized random variable 7(Y). We characterize the extreme points of the set of possible probability distributions of -(Y), as 7y ranges over all quantizers. We then establish optimality properties of likelihood-ratio quantizers for a very broad class of quantization problems, including problems involving the maximization of an Ali-Silvey distance measure. Some new results are also obtained for a Neyman-Pearson decentralized detection problem.