An overview of detection, classification and localization of signals in an uncertain ocean environment is presented from a Bayesian perspective. The Bayesian approach allows one to both model and incorporate, in an optimal way, the inherent uncertainty that often exists in the knowledge of ocean acoustic environmental parameters. The relationship between environmental parameter estimation and signal detection is illustrated, and it is shown how environmental parameter estimation can be viewed as an inherent part of the optimum Bayesian detector. Furthermore, by using the Bayesian approach, one can design optimum detection algorithms that are robust with respect to precise knowledge of environmental parameters. Using the receiver operating characteristic (ROC), one can illustrate the trade-off among detection performance, environmental uncertainty, and signal-to-noise ratio. Several examples are taken from an uncertain shallow water environment to illustrate these trade-offs. Finally, it is shown how depth classification can be viewed as an optimal detection problem. [Work supported by ONR.]