There is an explanatory puzzle about the relationship between physics and the rest of science: if the lower-level facts from physics in some way underpin higher-level facts, why don’t the lower-level details matter more for the day-to-day practice of the special sciences? Is it just a pragmatic feature of the practice of science, or are the special science genuinely autonomous, and if so, in what sense? In this talk I will develop a new account of autonomy following Woodward (2018) which I dub ‘generalised autonomy’ since it unifies dynamical, causal and nomic autonomy. This account can solve the explanatory puzzle, and walk the tightrope of balancing the dependence on the lower level with the independence of the higher level. I'll argue that physics sometimes doesn't matter to sciences like ecology and biology in a genuine and objective sense - and prevents one objection to realism about these higher-level sciences.