We develop an equilibrium model in which agents learn about both the mean and volatility of dividends and differ in ambiguity aversion. Because their conservative portfolios give them higher marginal risk-bearing capacity, ambiguity-averse investors step in as buyers when volatility rises. In contrast to economies populated only by Bayesian agents, volatility learning under ambiguity aversion has first-order effects: uncertainty premia rise with perceived volatility and portfolio flows predict returns. Using the concavity of the option-implied volatility surface as a proxy for ambiguity and index futures trading data, we provide empirical support for the model's predictions.