Global and Local Imperialism in Statistical Mechanics

Abstract Statistical mechanics applied on a universal scale (the Mentaculus) attaches probabilities to all possible events in the universe, far outstripping its local thermodynamic evidence base. This ‘imperialistic’ tendency is often criticised as requiring untenable dynamical choreography in phase space. However, I will argue that imperialistic issues arise even in local statistical mechanics; if we tolerate them here, we should tolerate them in the Mentaculus. In an attempt to sidestep these issues, some have suggested swapping the Statistical Postulate for a set of initial distributions claimed to yield the same thermodynamic predictions but indeterminacy elsewhere. I show that the result is still largely imperialistic, but for this very reason might be preferable to the standard Mentaculus as providing the same goods from weaker assumptions.