Declining response rates to telephone surveys have prompted survey organizations to be creative and open-minded concerning implementing new survey procedures and considering alternative data collection strategies. One strategy that may be helpful to slow or reverse the trend in declining response rates is to offer sample members a variety of ways to respond to the questionnaire-essentially allowing the respondent to choose the mode with which he/she is most comfortable. For example, adding a mail option, a Web option, or both to a telephone survey creates an alternative method for obtaining responses from willing participants, thereby allowing more resources to be devoted to reluctant sample members and late responders. This paper describes some of the practical issues encountered in designing and implementing a multimode survey and presents data on the impact of using multiple modes on data quality. The National Survey of Recent College Graduates (NSRCG) is a biennial survey of science and engineering bachelor's and master's degree recipients sponsored by the National Science Foundation. Conducted as a CATI survey since 1974, in 2003 the NSRCG was converted to a multimode survey using mail, Web, and CATI. In this paper, we discuss some of the challenges faced in converting the NSRCG to a multimode survey, such as revising question wording and response formats to suit each mode, managing multiple instrument versions, and maintaining a common database. We then examine measures of data quality by mode, including item nonresponse and response error (as measured by rates of editing). We conclude with lessons learned and suggestions for researchers implementing a multimode design.