The desirable properties of the selected sample can be upset if the survey suffers from non-response, which pretty much all surveys do. Non-response will not necessarily bias survey estimates, but it will do so if the non-respondents are systematically different from the respondents. The reaction of researchers to the increased difficulty of obtaining high response rates has taken two forms. One has been to search for better ways of persuading people to take part in surveys. The other has been to search for better ways of making estimates in the situation where considerable non-response has affected the data. This chapter explores different perspectives on the search for better ways of persuading people to take part in surveys. To keep non-response under control to the extent that continued use of the classic randomization-based inferential paradigm can continue to be employed, researchers will need to be ever more inventive and alert both to opportunities and to challenges.