During the past 20 years, survey methodology has undergone a paradigm shift. The old paradigm was based on a statistical model that focused on the effects of survey errors on the estimates derived from survey data. The new paradigm is based on a social scientific model that focuses on the causes of survey errors. Several developments have helped bring about this shift—the application of methods and concepts from cognitive psychology to the reduction of survey measurement error, the development of new computerized methods of data collection, and the increase in concern about measurement and nonresponse as sources of error in survey estimates. The new paradigm has little to say about the topics, such as sampling error, which were central to the old one; similarly, the old paradigm had little to say about how to reduce or prevent errors, a major concern for the new one. Thus, the two paradigms do not clash so much as complement each other.