The distinction between polling and public relations - publicizing particular angles, setting expectations, and encouraging certain actions - is problematic. So, too, the assumptions that accurate predictions depend on pollsters asking unbiased questions, and that opinion polls are miniature referendums. The attempt to predict the outcome of the 1951 referendum on communism in Australia by Roy Morgan's Gallup Poll illustrates these points. Morgan, wanting the referendum to pass, framed the issue accordingly. But he also thought the Government's framing offered the best basis for predicting the result. Opponents reframed the issue and the referendum was defeated. This article, in exploring the relationship between polling and public relations, analyses Morgan's questions, his forecasts, and his explanations for predicting the wrong result.