Bertrand Russell's 1903 masterpiece The Principles of Mathematics places great emphasis on the need to separate propositions from psychological items such as thoughts. In 1919 (and until the end of his career) Russell explicitly retracts this view, however, and defines propositions as "psychological occurrences". These psychological occurrences are held by Russell to be mental images. In this paper, I seek to explain this radical change of heart. I argue that Russell's re-psychologising of the proposition in 1919 can only be understood against the background of his struggle with the problem of the unity of the proposition in earlier work. Once this is recognized, and the solution to the problem offered by the 1919 theory is appreciated, new light is also shed on Russell's naturalism. I go on to compare Russell's psychological "picture theory" with the vehemently anti-psychological picture theory of Wittgenstein's Tractatus Logico-Philosophicus and suggest that, once the background of the dispute is brought into clearer focus, Russell's position can be seen to have many advantages over its more celebrated rival. © Springer 2006.