Angler surveys can yield inaccurate results because of the failure of some anglers to participate or return their questionnaires. If nonrespondents are different from respondents with respect to the variables in the survey, some segments of the angler population will be under- or overrepresented, and the inferences made from the respondents about the population of interest will be subject to nonresponse bias. Errors introduced by nonresponse can be greater than sampling error, making it impossible to assign useful confidence limits to survey results. A 10-page mail questionnaire was sent to a stratified random sample of 9,981 Texas fishing license holders. Information was collected on fishing experience, fishing participation, species preferences, attitudes, and orientation to fisheries management efforts. A response rate of 62% was obtained (exclusive of questionnaires that could not be delivered). Adjustments for nonresponse were made by means of response propensity stratification. Individual response probabilities were estimated by means of logistic regression with a binary variable indicating response as the dependent variable and age, sex, race, and purchase date of fishing license as independent variables. Response probabilities were inverted to obtain the nonresponse adjustment weights. Unadjusted survey variables were considered significantly different from adjusted variables if their 95% confidence interval did not contain the adjusted mean. Several variables were subject to significant nonresponse bias, which indicates the need to make adjustments when nonresponse is present.