Making Sense of Wason
Parallel mentalistic/mechanistic cognition resolves the controversy.
Posted May 05, 2012 | Reviewed by Ekua Hagan
One of the longest-running debates in modern psychology centers on the correct interpretation of the Wason Selection Task. Peter Wason originally designed the test to see if people applied logic that would disprove a hypothesis by falsifying it as well as by confirming it. It goes like this:
Imagine that you have the job of checking that cards have been sorted correctly. You have to make sure that, if a person has a "D" on one side of a card, they have a "3" on the other. The four below have to be checked to see if they conform. Which card or cards do you definitely have to turn over in order to see if it violates the rule?
Less than one in four people get this right. Most people either choose D and 3, or D alone. The correct answer is D and 7. The reason is that the rule says simply that a card with a D on one side must have a 3 on the other, not that a card with a 3 must have a D on the other side. So whether the 3 has a D on its back or not is irrelevant. However, it is important to know what is on the other side of the 7, because, if that were a D, the rule would be violated, and the correction would have to be made.
Now consider a second situation:
You are serving at a bar and have to enforce the rule that if a person is drinking beer, they must be over 20 years of age. The four cards below have information about people sitting at a table. One side of the card tells you what a person is drinking and the other side tells their age. Which card or cards must you turn over to see if the rule is being broken?
The correct answer is the cards with beer and 16. However, about 75 percent of people get this one right. Why is this so when only one in four get the previous version right? How is it that people are seemingly more intelligent with drinks and ages of drinkers than with letters and numbers?
The evolutionary psychologists, Cosmides and Tooby,* claim that both tests are logically exactly the same, but that in the second one, a “cheat-detection” mental module kicks in which helps people to get it right most of the time. They argue that the problem with enforcing the law about the drinking age is a “social contract” issue that mobilizes the cheat-detection module of the mind to get the answer right in the majority of cases. However, if cheat-detection is not involved, as it isn't with the card-checking problem, most people get the answer wrong.
Evolutionary psychologists often give the impression that this conclusion is widely accepted in the Wason-testing community. But in fact, this is not the case. On the contrary, an alternative explanation is that in the second example it is immediately apparent that a person who is 16 years old may be violating the rule. However, in the first case, it is not obvious until you think about it that the rule which says that a card with D on one side must have a 3 on the other also means that a card with any number other than 3 must be checked to see if the letter on its other side violates the rule. In spite of a superficial similarity, these are simply different tasks. Indeed, some believe that reasoning is not involved in the selection task at all, which is much more a question of judging the relevance of the information presented, as the following example of the test illustrates:
The City Council has asked for volunteers to take care of visiting school children. Volunteers have to fill in a card. Rossi and Bianchi, two clerks of the City Council, are about to sort the cards. There are two scenarios:
Scenario A. Bianchi believes that only women will volunteer, but Rossi asserts that there are male volunteers. Bianchi bets Rossi that all the male volunteers are married, and Rossi accepts the bet. Cards filled in by the volunteers show sex on one side and marital status on the other. Because sex and marital status are on different sides of the card, it is impossible to find out if Bianchi has won the bet without turning over one or more cards.
Your task is to decide which card or cards it is absolutely necessary for Rossi to turn over in order to see if Bianchi has won the bet that if a volunteer is male, then he is married. The four cards are:
Scenario B. Bianchi bets that men with dark hair will volunteer to take the children and Rossi accepts the bet. Cards filled in by the volunteers show sex on one side and hair color on the other. In front of the clerks on the table are four cards.
Your task is to decide which card or cards it is absolutely necessary for Rossi to turn over in order to see if Bianchi has won the bet that if a volunteer is male, he has dark hair. The four cards are:
Thirty-six undergraduates at the University of Padua took part in the experiment and were randomly assigned scenarios A or B. The correct answers are cards 1 and 4 in both scenarios. However, only 16 percent got the right answer in scenario B, while 65 percent got it right in scenario A. Sperber, Cara, and Girotto interpret this finding to prove that the relevance of information to the selection task is more important than any other factor. For example, in this case, marital status is often relevant to childcare, whereas hair color never is.
However, these findings also cast doubt on Cosmides and Tooby’s claim that a specific “cheat-detection module” is involved because there is no question of deception in either case, and both versions are logically and semantically the same.
Nevertheless, the cheat-detection versus social-relevance interpretations can be reconciled as stressing different aspects of mentalistic cognition. Detecting cheating is an obvious aspect of mentalism understood as interpersonal, psychological interaction, as indeed is putting things in their correct social context and judging their relevance accordingly. The original, purely abstract version of the Wason Selection Task, by contrast, epitomizes mechanistic cognition. Indeed, the latter is something you can imagine a computer being programmed to do very readily. But serving drinks at a bar or assigning school children in the other two examples could not be entrusted to machines because both rely on good social skills, common-sense, and contextual understanding of people’s behavior: in a word, mentalism. And because people are people and not machines, the mentalistic versions of the Wason Task are easy, and the original, mechanistic version, hard.
An obvious implication is that autistics — or at least high functioning ones — ought to do worse on the cheat-detection Wason Task than normal thanks to their symptomatic deficits in mentalism. A paper published in 2009** claiming to be the first to do so compared 20 autistics with 20 non-autistic controls matched for age, sex, and IQ on a Wason test involving intentional versus unintentional cheating. The autistics did indeed do worse than the controls on both, but showed exactly the same 30 percent difference between the intentional versus the unintentional cheating scenarios: controls 80/50 percent correct on intentional cheating/unintentional cheating; autistics 60/30 percent.
The researchers offer their own explanations, but my prediction is that if tested on the non-cheating, social relevance version of the Wason Task above, the autistics would do worse still. This is because autistics are good at rule-learning and rule-enforcement — often naively and very literally so, thanks to their mentalistic deficits and tendency to think mechanistically. This may explain their relative success with scenarios involving the detection of cheating, which is always a question of breaking a rule or acting unfairly in some way. Clearly, more research is needed, but it would help if researchers were aware that the Wason Task findings might in fact hinge on the difference between mechanistic and mentalistic modes of cognition.
* Cosmides, L. & Tooby, J. Cognitive Adaptations for Social Exchange. in The Adapted Mind: Evolutionary Psychology and the Generation of Culture (eds. Barkow, J.H., Cosmides, L. & Tooby, J.) pp. 163-228 (Oxford University Press, New York, 1992).
** Rutherford, M.D. & Ray, D. Cheater Detection is Preserved in Autism Spectrum Disorders. Journal of Social, Evolutionary and Cultural Psychology 3, 105-117 (2009).