Brewer, Kenneth2015-12-131369-1473http://hdl.handle.net/1885/83544This paper is based on the author's recollections of work done for and with E.K. Foreman, and later developments of that work. It describes most particularly the author's early intimations of the potential for using prediction models in survey sampling, the reasons why the use of such models, especially for inferential purposes, was so strongly resisted in the second half of the 20th century, the manner in which randomization and prediction inferences can be combined, and some advantages flowing from that combination.Keywords: Design-based; Inference; Model-based; Prediction; Randomization; Survey sampling200510.1111/j.1467-842X.2005.00404.x2015-12-12