How to be reasonable about the meaning of ‘ought’

Consider the following two popular and plausible principles. First principle: facts about what you ought to do are tightly bound up with facts about what there’s reason to do. For example, if you ought to Φ, then there must be more reason for you to Φ than there is for you to not-Φ. Second principle: ought is ‘upwards monotonic’—if Φing entails Ѱing, then if S ought to Φ is true, so is S ought to Ѱ. I argue that these two principles are incompatible. If I’m right, then respecting the connections between reasons and oughts requires giving up monotonicity, and so requires giving up the classical semantics for deontic modals. We need something new. I sketch a non-monotonic meaning for ‘ought’ which builds the ought/reason connections directly into the semantics, while aiming to accommodate the data that made monotonicity look attractive in the first place.