This paper studies the design of mechanisms that are robust to misspecification. We introduce a novel notion of robustness that connects a variety of disparate approaches and study its implications in a wide class of mechanism design problems. This notion is quantifiable, allowing us to formalize and answer comparative statics questions relating the nature and degree of misspecification to sharp predictions regarding features of feasible mechanisms. This notion also has a behavioral foundation which reflects the perception of ambiguity, thus allowing the degree of misspecification to emerge endogenously. In a number of standard settings, robustness to arbitrarily small amounts of misspecification generates a discontinuity in the set of feasible mechanisms and uniquely selects simple, ex post incentive compatible mechanisms such as second-price auctions. Robustness also sheds light on the value of private information and the prevalence of full or virtual surplus extraction.