To figure this out, we added another component to our next experiment. Imagine that you are taking part in an experiment similar to the previous one. You took the eightquestion quiz and answered four questions correctly (50 percent), but thanks to the answers at the bottom of the page, you claimed that you had solved six correctly (75 percent). Now, do you think that your actual ability is in the 50 percent range, or do you think it is in the 75 percent range? On the one hand, you may be aware that you used the answer key to inflate your score, and realize that your real ability is closer to the 50 percent mark. On the other hand, knowing that you were paid as if you really had solved six problems, you might be able to convince yourself that your ability to solve such questions is in reality closer to the 75 percent level.

This is where phase two of the experiment comes in. After finishing the math quiz, the experimenter asks you to predict how well you will do on the next test, in which you will be asked to answer a hundred questions of a similar nature. This time, its clear that there are not going to be any answers at the bottom of the page (and therefore no chance to consult the key). What do you predict your performance will be on the next quiz? Will it be based on your real ability in the first phase (50 percent), or will it be based on your exaggerated ability (75 percent)? Here is the logic: if you are aware that you used the answer key in the previous test to artificially inflate your score, you would predict that you would correctly solve the same proportion of questions as you solved unassisted in the first test (four out of eight, or around 50 percent). But lets say you started believing that you really did answer six questions correctly on your own and not because you looked at the answers.

Now you might predict that in this next test, too, you would correctly solve a much larger percentage of the questions (closer to 75 percent). In truth, of course, you can solve only about half of the questions correctly, but your self-deception may puff you up, crablike, and increase your confidence in your ability.

The results showed that participants experienced the latter sort of self-puffery. The predictions of how well they would perform on the second phase of the test showed that participants not only used the answer key in the first phase to exaggerate their score, but had very quickly convinced themselves that they truly earned that score. Basically, those who had a chance to check their answers in the first phase (and cheated) started believing that their exaggerated performance was a reflection of their true skill.