A New Riddle of Induction
Suppose that tomorrow the sun does not rise, bread does not nourish, and swans are blue. Does that show that nature is not uniform, that the past is not projectable to the future, and that induction has broken down? Can we conclude that what we observe tomorrow does not resemble the past? Not unless we know the past—unless we know that the sun used to rise every day, that bread used to nourish, and that previous swans were white. But memory is fallible and vulnerable to skepticism. If we are wrong about the past in these respects, then when we suppose that the future diverges from the past, we are mistaken—actually the future does resemble the past (blue swans etc). So unless we have an answer to skepticism about the past we cannot infer from an apparent breakdown in the uniformity of nature that there is a real breakdown.  Given that we have no such answer, we cannot know that the future fails to resemble the past. If bread never actually nourished in the past, then its failure to nourish tomorrow is perfectly uniform and projectable from its past properties. So it is not just that we can’t establish that nature is uniform; we also can’t establish that it is not uniform. We can’t describe a situation in which we discover that the previous laws of nature have broken down, or were not laws after all, for it is always possible that we are wrong about how things were in the past. This makes the skeptical problem of induction ever harder. We can know that our predictions have been falsified, but it doesn’t follow that we can know that the future does not resemble the past, since we could be wrong about the past. Even a total failure in all our inductive predictions would not establish that the future diverges from the past. Nature might be completely uniform and yet appear to us not to be. We can’t know that nature will continue the same into the future and we can’t know that it has not continued the same.
 There are two sources of potential error about the past: first, we might just be wrong that bread ever nourished (we have false memories); second, we might have made an inductive error about bread in the past, inferring that all past bread nourishes from the limited sample of bread we have encountered (maybe the uneaten bread was poisonous). If we make the latter error, our observation tomorrow that some bread is poisonous actually follows the way bread was in the past, so there is no breakdown of uniformity.