This article evaluates the effect of the choice of survey recruitment mode on the value of water quality in lakes, rivers, and streams. Four different modes are compared: bringing respondents to one central location after phone recruitment, mall intercepts in two states, national phone-mail survey, and an Internet survey with a national, probability-based panel. The modes differ in terms of the representativeness of the samples, non-response rates, sample selection effects, and consistency of responses. The article also shows that the estimated value of water quality can differ substantially depending on the survey mode. The national Internet panel has the most desirable properties with respect to performance on the four important survey dimensions of interest.