Is System 2 Reasoning Really Constrained by Working Memory Capacity?

A theory that has come to into prominence in recent years is the Dual-Process Theory of Reasoning. This theory proposes that humans have two different reasoning systems.

System 1 is considered to be old in evolutionary terms and shared with other animals: it comprises a set of autonomous subsystems that have specific purposes and include both innate knowledge and domain-specific knowledge acquired by domain-general learning mechanisms. System 2, on the other hand, is evolutionarily recent and distinctively human: it permits abstract reasoning and analytical thinking, but is constrained by working memory capacity (WMC) (Evans, 2008; De Ney, 2006; Evans, 2003). WMC is a limited-capacity system for temporary storage and manipulation of information for complex tasks such as comprehension, learning and reasoning (Goldstein, 2008). The strength of Dual-Process theory is that it ties together a wide-range of research on reasoning that once seemed disparate. Of particular interest to Dual-Process theorists is research based on the Wason Selection Task, a classical reasoning task developed by Peter Wason in 1966.

As part of the Wason Selection Task participants are shown four cards lying on a table showing two letters and two numbers (A, D, 3, 7). They are told that each card has a number on one side and a letter on the other. They are then given the following rule: "If there is an A on one side then there is a 3 on the other". The participant is then told that he/she must turn over whichever cards are necessary to determine whether the rule is true or false. Performance on this task is extremely low. Typically, less than 10% of individuals make the correct selections of the A card and the 7 card. The most common incorrect choices are the A card and the 3 card or the selection of the A card only (Capon, Handley & Dennis, 2003; Evans, 2003). Interestingly, however, a modification to the task introduced by Griggs and Cox (1982) which couches the rule and card information in concrete everyday terms (i.e. "If a person is drinking beer, then he or she must be over 18 years old") has consistently produced substantially improved performance. For example, in the Griggs and Cox study, 73% of the participants answered correctly.

It is the stark difference in performance between the original abstract version of the task and the newer thematic version introduced by Griggs and Cox that makes the Wason Task so interesting to Dual-Process theorists; this is because they believe that the thematic version taps into System 1 reasoning while the abstract version taps System 2 reasoning. Therefore, much of the Dual-Process theory is substantiated by research that provides insight into why people find the thematic version of the Wason Task easier than the abstract version (Evan, 2003). Some of the studies that Dual-Process theorists regularly cite follow below.

Cosmides (1989), an evolutionary psychologist, argued that specific characteristics of the mind, that are intrinsic to all humans, have evolved overtime because they have proven to be advantageous for survival. One such characteristic is the 'social exchange algorithm'. This algorithm is geared to the generic rule "if you take a benefit then you must pay a cost" and has sensitive procedures for detecting "cheaters" (those who violate the rule by taking the benefits without paying the cost). In regards to the Wason Tasks, Cosmides proposes that the thematic version of the task activates the social exchange algorithm and is therefore better understood by participants (Goldstein, 2008; De Ney, 2003). To test this hypothesis Cosmides constructed several Wason Selection Tasks that varied in their degree of social relevance. Cosmides found that the more socially relevant the task the more likely participants were to answer it correctly (Cosmides, 1989). These findings help to substantiate the Dual-Process idea that System 1 reasoning is evolutionarily old and includes sub-systems that are innate (Evans, 2008; De Neys, 2003; Evans, 2003).

Cheng and Holyoak (1985) offer a different explanation; they suggest that the thematic version of the task triggers Pragmatic Reasoning Schemas. These schemas are specific ways of thinking about cause and effect in the world and are learned as part of experiencing everyday life. One schema that people learn is the Permission Schema, which states that if a person satisfies condition A (such as being the legal age for drinking), then they get to carry out action B (being served beer). Performance on the thematic tasks is facilitated because the rules and card information fit a pre-existing set of production rules for permission schemas (Goldstein, 2008; De Ney, 2003; Evan, 2003). To test this hypothesis Cheng and Holyoak (1985) carried out an experiment in which participants were tested with a version of the Wason Selection Task that triggered a Permission Schema and a version that was arbitrary. They found that participants were more likely to answer the Permission Schema version correctly than the arbitrary version. Dual-Process theorists use Cheng and Holyoak's findings to substantiate the idea that System 1 reasoning includes domain-specific knowledge acquired by domain-general learning mechanisms (Evans, 2008; De Neys, 2003; Evans, 2003).

Research that seeks to shed light on what makes a person more likely to answer the abstract version of the Wason Task correctly has been carried out by Stanovich and West (1998). Their study produced strong evidence to suggest that higher cognitive ability, in particular analytical reasoning, is involved. In order to measure cognitive ability the participants' Scholastic Aptitude Test (SAT) scores were used; SAT scores are known to load highly on general intelligence. The study found significant differences in the SAT scores between those who responded correctly to the abstract tasks and those who responded correctly to the thematic tasks; those who correctly responded to the abstracts tasks had significantly higher SAT scores. Further to this, participants who answered the thematic tasks correctly were no likelier at all to be higher in cognitive ability than those answering them incorrectly. This finding helps to substantiate the dual-process idea that analytical thinking is a feature of System 2 reasoning and that System 1 reasoning does not require this particular mental ability (Evans, 2008; De Neys, 2003; Evans, 2003).

While the above research helps to validate the different characteristic of System 1 and 2 reasoning no study exists currently (as far as the author is aware) that looks at the relationship between the abstract version of the Wason Task (the task that taps Systems 2 reasoning) and WMC. So far, the only evidence that Dual-Process theorists have to defend their idea that System 2 reasoning is constrained by WMC has been taken