In this paper reasons are given for accepting a paradigmatically ‘metaphysical’ thesis called Unrestricted Mereological Composition, or something close to it. These reasons are drawn from thoroughly empirical considerations that played a significant role in the history of physical sciences, particularly in the emergence of the Newtonian laws of motion. Roughly: this metaphysical thesis is supported by a principle of ‘inference to the best explanation’ of observations of the very same kind that paradigmatically furnish evidence supporting or undermining empirical theories (...) in physics. There are methodological morals in metaphilosophy to be drawn from this historical example – blurring the distinction that has been traditionally drawn between ‘metaphysical’ theses and ‘empirical’ hypotheses. (shrink)