This article addresses the problem of small nonprobability samples in research in the education of the deaf and hard of hearing in the face of a current and increasing emphasis on "scientifically based research" as required by recent No Child Left Behind (NCLB) federal legislation. The authors examine the gains and losses in information generated using non-probability samples in our field, describe the conditions required for generalizable research results, and identify several factors in the field of research in the education of the deaf that limit the ability to generate probability samples. Finally, the authors consider possible solutions to the problem, including more creative recruitment strategies, alternate research designs, and alternate reviewing strategies.