Let's go back and look at those three drawings, again. If you said for the first drawing, that person is extremely suspicious of other people, that would be a mistake. Because strangely drawn eyes do suggest suspicion. But in this series that you saw, strangely drawn eyes were less likely to be associated with suspicion than normal eyes. If you said, is worried about his masculinity, for the second picture. That's a mistake, because exaggerated muscles suggest worries about masculinity but they were actually less likely to be associated with worries about masculinity than normal muscles. Well, if you made one more mistakes, you are in good company.q most people make this kind of errors, that's because was susceptible to seeing illusory correlations. When we're prepared to see a particular association, we are particularly likely to see it. In this case, funny eyes suggest suspicion, so we're likely to remember the slide that showed this association. Emphasis on muscle suggests worries about masculinity so, that's the slide that we remember. Since we have theories about all kinds of things, we can't trust our memory about associations to tell us what the relationship is. If we have a theory that X goes with Y, we're likely to notice and to remember the cases where that's true and not remember or not notice the cases where it's not true. So even when the data don't actually support a particular plausible belief, we're quite likely to think it does. The drawings and symptom pairings that you just saw were actually invented by my research team, but the prism test is real. It was given by psychiatrists and psychologists for decades, even though it is literally worthless. You can't tell anything about anybody from the picture they draw except that it's a very bad test of intelligence. Smarter people do trended draw better drawings. But nothing else about them predicts anything. So how could they do that for decades, continue to make this error? Well, they never set up a two by two table. The slides that showed a plausible association got remembered and the other ones didn't. And that principle explains why psychologists and psychiatrists believe the Rorschach could tell them something about their client. That's the ink blot, you'll remember. Some of the alleged associations that were claimed about the Rorschach test were those that lay people would find intuitive. Some were arcane, they were suggested by psychoanalytic theory. But whatever the origin of the theory, if you expect an association, that's what you're likely to see. Then the erroneous belief about correlation gets sustained, because you keep on seeing confirmations. So if it's the draw a person test that you've been giving for years, you say just as I thought, this guy who drew the big muscles is worried about his masculinity. Nailed it. We'll be talking about this kind of problem in lesson six. We're subject to what's called a confirmation bias in all kinds of ways. We rest content with confirming data, and we don't think to look for data that might disconfirm our hypothesis. I've seen lots of drawings with big muscles by men who were worried about their masculinity. But, how about the drawings with big muscles by men who weren't worried about their masculinity and how about all the men who were worried about their masculinity who didn't draw big muscles? You have to keep in mind all four cells of the 2 by 2 table, and this fact escaped hundreds if not thousands of psychiatrists and psychologists. The plausibility principle is extremely important because we can't trust our casual observations about associations. If an association is plausible, we're likely to see it, whether it's there or not. If it's not plausible, we're likely to fail to see it, even if it exists and is quite strong. This is true even for animals. Dogs are almost impossible to teach that they can get fed if they yawn. That's not been the way dogs have been getting food these many eons. Dogs readily learn to go left in a maze if a light comes on at the left, and to go right, when it comes on at the right. But it's almost impossible for them to learn to go left, if it comes on at the top, and right if it comes on at the bottom. They're prepared to see directions that make sense, left and right. They're not prepared to see left and right matched with top and bottom. Cats readily learn that they can get food by pressing a lever or pulling a string. But it's only with very great difficulty that they learn that they can get food by licking or scratching themselves. Pigeons will actually die before they learn that not pecking at a light will get them food. Pigeons haven't made at this far by assuming that not pecking could ever got them fed. So we've just seen that there is or important. They determine what you're going to be able to see. And if you don't have any theories at all it's going to be hard to see any kind of association at all. In the next segment we'll look at the case of when you don't have any theory at all. How easy is it to detect correlations?