The basic structure of the “Wason selection task” is a rule and the instruction to validate the rule. In one variation of the task, students were asked to test the rule “if a card has D on one side, it has a 3 on the other”. They were then shown four cards, which had either a letter (D or F) or a number (3 or 7) on them, and were asked which cards they would turn over to validate the rule. The correct answer would be the D and 7 cards. If a D card had anything other than a 3 on the other side, the rule was disconfirmed; ditto if a 7 had a D on the other side.
Most students got it wrong – they said they would look at the F and 3 cards. Their error was in seeking to confirm the rule, rather than disconfirm it. But a rule is only a rule if it applies across the board; therefore, all you have to do is find one instance where it doesn’t apply and the rule is invalidated. Neither the F nor 3 cards had any bearing on the question because nothing on the other side could disconfirm the rule.
In the real world, we talk more of causes than rules, but the process of establishing a causal relation is similar to that of validating a rule: seek cases that disconfirm the proposition that x causes y. In other words, find cases of x without y and y without x (the equivalent of turning over the D and 7 cards in the Wason task). For example, to disconfirm the proposition that stress is the cause of poor job performance, one would look for cases where stressed workers perform just fine and bad workers are not stressed. If one finds such cases, one cannot say poor work performance is always the result of stress or that stress always impairs work performance.
Of course, causes rarely work like that in the real world, which is why scientists speak such in such convoluted terms. To say “x partly accounts for a portion of y given certain assumptions and conditions and only at high levels of x” lacks the emotional bunch of “x causes y” but that’s often how the world works.