![]()
1
A brief history of the Wason selection task
Jonathan St B T Evans
My talk at Ken Manktelow’s Festschrift conference was titled ‘Whatever happened to the selection task?’. However, some recent study of Web of Science has convinced me that my reports of its demise were somewhat exaggerated. While certainly past its prime, research on the task continues to feature in a number of recent publications, as we shall see. What the task has contributed overall to the psychology of reasoning in its 50 odd years of life is, however, quite remarkable. I offer here a brief, selective, affectionate and doubtless biased account of its history.
The selection task was first presented to an unsuspecting academic world in a book chapter. Wason (1966) reported an experiment in which the subjects, as he called them, were to decide whether a rule was true or false. The rule was ‘If a card has a vowel on one side, then it has an even number on the other side’. Subjects were told that the rule applied only to four particular cards which had a letter written on one side and a number on the other. The visible sides of the cards were
E K 4 7
Wason and Johnson-Laird (1972, p. 173) later described the results as follows: ‘The vast majority of subjects say either “E” and “4” or “only E”. Both answers are wrong. The correct answer is “E” and “7”. Any odd number on the other side of the E falsifies the rule in exactly the same way as would any vowel on the other side of the 7.’ Wason (1966, p. 146) comments that ‘In spite of instructions to the contrary, they [the subjects] cannot inhibit a tendency to see whether the statement is “true”.’ He later adds (p. 147) that ‘this apparent bias towards verification’ is analogous to that observed on his 2–4–6 task, discussed earlier in his chapter. In fact, there is a striking parallel between Wason’s (1966) first account of the task and that of his (1960) original paper on the almost equally famous ‘2 4 6’ problem (see Evans, 2016). In both cases a single condition is reported, without controls and comparisons, and the reader invited to wonder at the difficulty of a task which (in Wason’s opinion) should be easy. Wason was wont to refer to both these problems as ‘deceptively simple’ and expressed wonder that anyone would call them ‘deceptively difficult’. I still have not worked that one out.
The impact of Peter Wason
Wason was discharged from the army as an officer at the end of World War 2 and took up the opportunity offered to read for a university degree. However, this was in English. On graduation, he decided he wanted to be a psychologist, for reasons he never explained to me, and went to University College London (UCL) to take a second degree in psychology. He never left UCL but after obtaining his PhD moved to the linguistics department because, as he told me when supervising my own PhD, ‘I refused to teach’ (!). He worked most of his career as a Reader in Psycholinguistics despite concentrating the bulk of his research on the psychology of reasoning. And indeed, he did not teach, or do any administrative work so far as I can tell. His position, in effect, was that of full-time researcher.
Given this happy situation and his later fame, one might suppose him to have a prolific publication output and to feature in the world’s elite journals. In fact, neither is the case. Web of Science (WoS) records just 31 journal articles published by Wason between 1953 and 1984. (There were a couple of books and a few book chapters in addition.) The bulk of these were published in British journals and not one APA publication is to be found on his list. In a modern British university, such an output would be considered modest at best, even for someone who did his share of teaching and administration. When we look at his impact, however, the picture is quite different. Because of the age of the papers it is difficult to get a fully accurate picture of his citations. For example, Web of Science shows a total of 1,576 which does not include his most cited paper, Wason (1960), or his original book chapter, (Wason, 1960), although elsewhere in the WoS database these are given at 571 and 254, respectively (the last, being a book chapter, as well as old, is almost certainly underestimated). Even on these figures, however, we can see a total citation count between 2,000 and 2,500.
What is most striking, however, is the distribution of citations to Peter Wason (Figure 1.1). Twenty years on from his last published journal article, citations of Peter Wason’s work are still increasing.1 The citations to his first published account of the selection task in a journal article (Wason, 1968) are shown in Figure 1.2, and they also have increased in recent years – all this despite my perception that actual studies of the task are on the decline. Identifying how many papers actually report work on the task is very difficult without reading all the citing papers. However, to get an indication, I searched for papers that give both ‘Wason’ and ‘selection task’ in their topic, with the results shown in Figure 1.3. A scan of the abstracts of the more recent ones suggests that most of these do in fact report new empirical work on the task. So the number of such studies is indeed decreasing, but not so much as I had thought.
Figure 1.1 Distribution of 1,567 citations to Peter Wason – Web of Science, December 2014. Excludes Wason (1960) – 571 citations – and Wason (1966) – 254 citations
Figure 1.2 Distribution of 432 citations to Wason (1968) – Web of Science, December 2014
Much more interesting than this quantitative impact, however, is the influence the selection task had on the development of reasoning theory. Table 1.1 presents a highly selected list of important papers based on the selection task, together with citations rates. Each of these develops an important new theoretical perspective on the psychology of reasoning and all involve either the report of new experiments using the selection task or the extensive re-analysis of selection task data in order to support the theoretical claims made. Note that there are a number of other important papers on the selection task, some of which will be discussed later in this chapter. Table 1.1 does not include papers which simply discovered new empirical phenomena, however important, or whose main thrust was the understanding of the selection task itself. Of these kinds of papers, well-cited (100+) examples include Evans and Lynch (1973, 148 citations) – early report of ‘matching bias’; Griggs and Cox (1982, 259 citations) – key experimental paper elucidating the thematic facilitation effect; Fiddick, Cosmides and Tooby (2000, 129 citations) and Gigerenzer and Hug (1992, 148 citations) both developing the evolutionary theory of reasoning.
Figure 1.3 Distribution of 289 papers featuring ‘Wason’ and ‘selection task’ in a topic search – Web of Science, December 2014
Table 1.1 Seminal theoretical papers featuring the Wason selection task, with citations in Web of Science (December 2014) | Authors | Subject | WoS |
Wason and Evans (1975) | Dual-process theory of reasoning | 125 |
Cheng and Holyoak (1985) | Pragmatic reasoning schemas | 624 |
Cosmides (1989) | Evolutionary theory of reasoning | 814 |
Manktelow and Over (1991) | Decision theoretic treatment of reasoning | 151 |
Oaksford and Chater (1994) | Rational analysis; Bayesian theory of reasoning | 362 |
Sperber, Cara and Girotto (1995) | Relevance theory of reasoning | 182 |
I will discuss three phases of work on the selection task: the early work (1966–1979), the heyday (1980–1999) and the modern work of 2000 onwards.
Early work on the selection task (1966–1979)
Peter Wason was very enthused by the selection task following his 1966 book chapter and discussed it with me often during my PhD years (1969–1971) under his supervision and in the decade that followed in which we kept close contact. This was one of his most productive periods with a relative flurry of papers on the selection task published by him between 1968 and 1976, some of them in collaboration with Phil Johnson-Laird and, later, myself. It was during this period also that his influential book Psychology of Reasoning was published (Wason & Johnson-Laird, 1972), including several chapters on the selection task and helping to spread its fame. Even so, as Figure 1.2 shows, citations were no more respectable prior to the post-1980 heyday.
When I first joined Wason as a PhD student, he was working on the abstract version of the selection task, similar to the letters-numbers rule given earlier. He had decided by then that his subjects were incorrigibly irrational victims of ‘verification bias’ (later known as confirmation bias). He had shown their failures to be insensitive to a number of experimental variations such as binary materials or the use of both values on one side of the card with masks and so on. He was fascinated by their ‘irreversible’ thinking, quoting verbal protocols and trying to induce better performance by various therapies (Wason, 1969, is a particularly fascinating read). Two developments were, however, radically to change this picture (Wason hated split infinitives). The first was the discovery of the so-called thematic facilitation effect, in which the difficulty of the task apparently evaporated when realistic rather than abstract materials were used. The two early studies on this were run while I was working with him at UCL (Johnson-Laird, Legrenzi & Legrenzi, 1972; Wason & Shapiro, 1971). The second, which I discuss later, was the discovery that matching rather than verification bias explained the abstract selection patterns (Evans & Lynch, 1973).
Wason appeared disappointed when Diana Shapiro reported the first experiment to show that the difficulty of the task could be removed using realistic materials – but not for long. He soon became fascinated by how the task could be made easy in such a simple way, and the thematic facilitation effect featured strongly in the book he and Johnson-Laird were writing at the time (Wason & Johnson-Laird, 1972). The simple idea that realistic materials improve reasoning became a powerful meme (Dawkins, 1976) that escaped from the small reasoning field into psychology more generally. It was the thing that everyone who had never studied reasoning knew about reasoning. And like all memes, it proved very hard to recall when later research showed that this idea was oversimplified and inaccurate. (I originally wrote ‘simplistic’ before I remembered that Wason hated this word.2)
On starting my lecturing career, my (equal) first PhD student was Ken Manktelow. As an undergraduate, Ken had little interest in cognitive psychology and none at all in reasoning. But he had top marks in his class, and when I offered him a PhD studentship which had dropped into my lap, he thought about it for a while and then – with some apparent reluctance – accepted. (The rest is history.) We decided to combine two hot topics on the selection task – matching bias and the thematic facilitation effect. Showing matching bias requires the introduction of negatives into the conditional statements. Suppose the standard abstract rule is
If there is an A on one side of the card, then there is a 3 on the other side of the card.
We know that people tend to choose A and 3, and Wason had originally seen this as a verification bias. He thought people were trying to confirm the rule by looking for an A and 3 combination rather than a falsifying A and not-3 combination. But suppose the statement was this:
If there is an A on one side of the card, then there is NOT a 3 on the other side of the card.
And suppose that for both statements the visible sides of the card show
A D 3 7
If people have a verification bias, they should select A and 3 for the affirmative rule and A and 7 for the negative rule (looking for a confirming combination of A and not-3). What they actually tend to do is to choose A and 3 on both problems, the so-called matching bias effect (Evans, 1998; Evans & Lynch, 1973). This was a very difficult finding for Peter Wason and he was somewhat shocked when I first showed it to him. But he almost immediately accepted that the matching bias account must be right, and we subsequently collaborated on work to reconcile this with previous findings (Evans & Wason, 1976; Wason & Evans, 1975).
Now back to Ken Manktelow’s PhD. We thought that since realistic materials remove the difficulty of the selection task, they might remove matching bias as well. So Ken ran a series of experiments comparing abstract and realistic materials and including negations in the conditional sentences. Most of his materials used food and drink problems, with statements such as
- If I eat haddock, then I drink gin.
While somewhat arbitrary, these materials are definitely thematic, and because people have no preconceptions about these relations, they work fine with negations added such as
- If I eat haddock, then I do not drink gin.
Participants were told that each card represented what was eaten and drunk at a particular meal and shown cards with visible words such as
Letter-number rules were used as abstract controls. The logic of the design was impeccable. According to the thematic facilitation meme, logical performance should have been better with thematic materials. The only problem was that these materials did not facilitate. At all. Not even a little bit. In fact, they behaved exactly the same as the abstract materials, including full matching bias effects (Manktelow & Evans, 1979).
At the time Ken was desperately disappointed with these negative results. He eventually r...
