We report six experiments, involving nearly 1000 subjects, that investigate people’s tendency to indicate “I knew it all along” with reference to facts that we are sure they would not have recalled without benefit of study. Subjects studied factual statements before doing a general knowledge test. “Knew-it” subjects were told to write only the answers they would have recalled without benefit of study, but they wrote about 20% too many answers; based on this finding, we would conclude that people mistake recently learned answers for previous knowledge. However, “learned-it” subjects, who were told to write only answers they learned during the experiment, tended to write the correct number; given this finding, we would conclude that people have an excellent ability to discriminate between the two sources of available answers. The two conclusions can’t both be right. Either people have the needed information in mind, but the knew-it test distracts them from it, or they don’t, but can adopt strategies for the learned-it test that give accurate performance. We conclude it’s a bit of both.