Although it has long been known that analyses that treat stimuli as a fixed effect do not permit generalization from the sample of stimuli to the population of stimuli, surprisingly little attention has been paid to this issue outside of the field of psycholinguistics. The purposes of the article are (a) to present a non-technical explanation of why it is critical to provide a statistical basis for generalizing to both the population subjects and the population of stimuli and (b) to provide instructions for doing analyses that allows this generalization using four common statistical analysis programs (JMP, R, SAS, and SPSS).