THE EPISTEMOLOGY OF EXPERIMENT*