This article develops a choice model for environmental public goods, which allows for consumers to learn about their preferences through consumption experiences. We develop a theoretical model of Bayesian updating, perform comparative statics over the model, and show how the theoretical model can be consistently incorporated into a reduced form econometric model. Our main findings are that in a random utility model (RUM) discrete choice model, a subject�s scale should increase and the variability of scale should decrease with experience if subjects are Bayesians. We then estimate the model using field data regarding preferences for one particular public good, water quality. We find strong evidence that additional experience increases scale, thereby making consumer preferences more predictable from the econometrician�s perspective. We find supportive but less convincing evidence that experience decreases the variability of scale across subjects