I believe that science must rest on an a priori foundation (and is not empirical) and I wanted to use this forum to get a sense for how people might respond to these ideas. Here are the reasons I give for why science must ultimately have an a priori foundation: Mathematics is a priori. I don't believe this can seriously be doubted. It is universal and non-analytic (per Kurt Godel) and cannot be empircal (per the problem of induction). The problem of induction basically shows that science either has an a priori foundation, or cannot constitute anything like knowledge. Many scientific discoveries (including many of the most important ones in history) take place by means of thinking. "Thought experiments" are a major means of making scientific discoveries. Thus Newton and Einstein made their major discoveries just "by thinking about it for a long time" (rather than by experiments or observations). Many scientific claims clearly go far beyond what could be justified by experiment. To use a simple example, Newton's 3rd law is a universal statement about every action. But neither Newton nor anyone else since has ever tested every action in the history of the universe. To draw universal laws from a handful of examples would be a gross fallacy if science were empirical. Simple mechanics is clearly intuited a priori. Levers, wheels, and gears all work in ways that can be clearly intuited. Ironically, what happens to pool balls in Hume's classic example is also intuited a priori. I believe much (but not all) of classical physics is also based on a priori intuition. I believe that experiments and the scientific method interact with a priori intuitions in the following way. From a known class of events (by events I mean a physical phenomenon that can be repeatedly tested), scientists use a sort of repeated-hypothetical-deduction process to come up with a "hypothesis" (a general statement) that fits with the set of known events, eliminate those that lead to contradictions, and when they find one that does not contradict any known events, design experiments to test whether other events entailed by the hypothesis actually obtain. Thoughts?