Problems of obtaining a representative sample for a national survey are discussed, and the experiences of Educational Testing Service are used as examples of some methods used to cope with these problems. It is suggested that researchers: (1) capitalize on some existing program; (2) establish in the minds of the possible examinees and the possible participating schools or sampling units that they have a stake in the enterprise because of what they do; (3) sample as many sources as possible, consistent with the cost and requirements of the rest of the job; and (4) make provisions for collecting easily supplied data from everybody who was first contacted, follow up with those who don't cooperate, find out why, and modify procedures to include the nonrespondents. These procedures are suggested to be crucial in educational research, involving national surveys. (Author/CTM)