Deductive reasoning and matching-bias inhibition ...

MATCHING-BIAS THINKING AND REASONING, 2002, 8INHIBITION (3), 205–224TRAINING

205

Deductive reasoning and matching-bias inhibition training: Evidence from a debiasing paradigm

Sylvain Moutier, Nathalie Angeard, and Olivier Houdé Universities of Caen and Paris 5, France

Using the matching bias example, the aim of the present studies was to show that adults’ reasoning biases are due to faulty executive inhibition programming. In the first study, the subjects were trained on Wason’s classical card selection task; half were given training in how to inhibit the perceptual matching bias (experimental group) and half in logic without the inhibition component (control group). On the pre- and post-tests, their performance was assessed on the Evans conditional rule falsification task (with a negation in the antecedent of the rule), a task that also involves matching bias. In addition, subjects were tested for perceptual field dependence/independence using the Embedded Figures Test. The results brought out a specific inhibition training effect, as well as a clear-cut relationship in the experimental group between receptiveness to training and perceptual field independence. In the second study, the training paradigm was the same except that on the pre- and post-tests, the negation was in the consequent of the conditional rule (in this case, the perceptual matching response corresponds to the logical response). The subjects succeeded on the pre-test, and the matching-bias inhibition training had a negative effect on post-test performance. This specific negative priming effect confirms the inhibitory impact of our experimental training and outlines the dissociation of inhibition and logical components.

The concept of inhibition has a long and diverse history (Macmillan, 1996; Smith, 1992), and since the early 1990s, it has benefited from a new thrust of energy in cognitive psychology (see Dagenbach & Carr, 1994). Our studies on number, categorisation, and reasoning attempt to show that from early childhood to adulthood, cognitive development cannot be reduced to the coordination– activation of structural units (as in Jean Piaget’s structuralist theory and in the neo-structuralist models), but that development also often involves inhibiting a competing structure or scheme (Houdé, 2000; Houdé & Guichart, 2001; Houdé, et al., 2000, 2001; see also Bjorklund & Harnishfeger, 1990; Dempster, 1992; Requests for reprints should be sent to Olivier Houdé, Groupe d’Imagerie Neurofonctionnell e (UMR 6095), Université Paris 5, Sorbonne, 46 Rue Saint-Jacques, 75005 Paris, France. Fax: 33 (0) 1 40 46 29 93. E-mail: [email protected] r © 2002 Psychology Press Ltd http://www.tandf.co.uk/journals/pp/13546783.htm l DOI:10.1080/13546780244000033

206

MOUTIER, ANGEARD, HOUDÉ

Dempster & Brainerd, 1995; Diamond, 1991). In adult cognition, this idea is particularly applicable to the study of deductive reasoning biases (Houdé, 1997) such as the perceptual matching bias (for a review, see Evans, 1998). The matching bias was first demonstrated by Evans (1972; see also Evans, 1989; Evans, Newstead, & Byrne, 1993; Evans & Over, 1997) using a conditional rule falsification task. If both the antecedent and the consequent of a rule are true (TT), then the conditional is true; if the antecedent is true and the consequent is false (TF), then the conditional is not true; and if the antecedent is false, the conditional is true regardless of the truth value of the consequent (FT or FF). One of the rules Evans used is “If there is a blue square on the left, then there is not a green diamond on the right”. In this example, the blue-square/greendiamond case (TF)—the only case that is inconsistent with the rule stated—is the logically correct answer to the falsification task. Of Evans’s 24 participants, 23 found this answer. However, when the negation is in the rule’s antecedent rather than in its consequent (“If not-p then q” instead of “If p then not-q” as in the previous example), performance changes considerably. So with the rule “If there is not a red square on the left, then there is a yellow circle on the right”, only seven participants found the correct answer (TF) by putting, for example, a blue square (not a red square) on the left of a green diamond (not a yellow circle). Everyone else erroneously put a red square on the left of a yellow circle, thereby making the conditional rule true (FT). According to Evans, this clear performance difference is indicative of a strong deductive reasoning bias. The bias in question is the perceptual matching bias, which consists of preferring the instances (here, the geometric figures) mentioned in the rule and neglecting the logically relevant instances whenever they do not match the antecedent or consequent. Evans also used the matching bias concept to account for participants’ performance on Wason’s card selection task (Wason, 1968). The materials for this classical reasoning task are cards with a letter on one side and a number on the other. Four cards are lined up on a table, with two letters and two numbers facing up (e.g., A, D, 3, 7). The rule “If there is an A on one side of a card, then there is a 3 on the other” is stated, and the subject is asked to indicate which card(s) must be turned over to verify the rule. The most frequent responses are A and A-3 (and less often A-3-7). None of these solutions is logically correct: the right answer is to turn over A and 7. Turning over the A checks to see whether there is a true antecedent and a true consequent (TT), and turning over the 7 proves that one does not have a true antecedent with a false consequent (TF). Cards D and 3 are irrelevant, because for the D the antecedent is false (FT or FF), and for the 3 the consequent is true (TT or FT): they therefore cannot be incompatible with the rule. Despite the simplicity of this task, most participants fail (answering either that turning over A suffices, or that A and 3 have to be turned over). These results, which have given rise to a large number of interpretations, are explained in an

MATCHING-BIAS INHIBITION TRAINING

207

original way in Evans’s analysis. The author proposed the following modification: replace Wason’s rule with the rule “If A then not-3”. Now the logically correct answer is to turn over A and 3. Card 3 in this case is used to make sure the antecedent is not true when the consequent is false (TF). This time, the answer “A and 7” is wrong. Most participants succeed on this new version of the task. The only possible explanation, according to Evans, is that participants fall prey to the matching bias. No matter what the rule is (“If p then q” or “If p then not-q ” ), participants consider only the instances that match the antecedent and the consequent, even when those instances are not logically relevant (as in Wason’s classical task). Evans (1998) proposed a critical overview of the various theoretical interpretations of matching bias, including heuristic-analytic theory, mental models theory, the theory of optimal data selection, relevance theory, and the specific processing-negations account. What stands out from this analysis is that Oaksford and Stenning’s (1992) processing-negations account is consistent with most of the findings, as is Evans’s own (1989) heuristic-analytic account. In the processing-negations account, negation entails difficulty (particularly when abstract content is used) in constructing contrast classes corresponding to the normally large set of possible states of affairs that the negative assertion permits. For example, in the conditional rule falsification task described earlier, the difficulty lies in considering the set of “not-a-red-square” figures (or in considering “7” as a case of “not-3” in Wason’s classical task). According to heuristic-analytic theory, the difficulty stems from selective attention to, or representation of, problem features. Heuristic processes are described as preconscious or preattentive, and are essentially pragmatic in nature; the relevance is psychological and not necessarily logical. In contrast, analytic processes correspond to deductive competence. [See also Evans and Over’s (1997) distinction between two forms of rationality: Rationality 1 is a form of everyday reasoning that people use to reach their goals without having to conform to some normative theory; Rationality 2 corresponds to deductive competence.] The specific account of the matching bias in this case relies on heuristics like the “not-heuristic”. According to Evans (1989), the linguistic function of the word “not” is to direct attention to the proposition that it negates, and he insists on the fact that in everyday language, people rarely use negation to supply new information. Thus, in the conditional rule falsification task (where “not” is introduced in the antecedent: “If there is not a red square on the left, …” ), the not-heuristic would induce a high activation level for the matching strategy. In fact, as Evans (1998) stressed, these two accounts (processing-negations and heuristic-analytic) make complementary contributions to explaining matching-bias responses, whether it be the conditional rule falsification task or Wason’s card selection task. Of interest to us here from the standpoint of the role of inhibition in deductive reasoning is the emphasis placed by Evans on selective

208


attention processes. Clearly, if the linguistic function of “not” (not-heuristic) is to direct attention to the proposition that it negates (hence, the matching bias), then inhibition is precisely the cognitive mechanism that allows one to redirect attention towards the logically relevant items (i.e., towards the contrast set in Oaksford and Stenning’s processing-negations account). In this perspective, Houdé and Moutier (1996; see also Houdé, 2000) suggested that observation of reasoning biases does not necessarily lead to the conclusion that the essence of human reasoning is heuristic and inconsistent with the canons of logic, and that deductive competence is doomed to be overridden (when abstract content is used) by reasoning biases. A “presumption of rationality ” (expression taken from the philosophy of mind, Engel, 1993) is indeed justified, in so far as the biases underlying reasoning errors, such as the matching bias, may not be rooted in faulty deductive logic but in faulty cognitive inhibition in working memory (Baddeley, 1986, 1992; Pascual-Leone, 1988, 2000). Evans (1998) himself stated that “in the heuristic-analytic theory, biases occur even when participants have logical capability, through the analytic system [Rationality 2, Evans & Over (1997)], to make correct deductions” (p. 54). Houdé and Moutier (1996) experimentally trained a population of adolescents and young adults to inhibit the matching bias on Wason’s classical card selection task (see Figure 1). A control procedure was also used without the inhibition component, where “by subtraction”, only the logic component remained (i.e., only the explanations of the correct and incorrect answers). The specific feature of the inhibition training was that it introduced warning elements, i.e., executive instructions. So the experimental training began with, “In this problem, the source of the error lies in a habit we all have of concentrating on cards with the letter or number mentioned in the rule [the experimenter points to cards A and 3, and to the place where they are mentioned in the rule] and not paying attention to the other cards. This can have a very misleading effect on us: we think this makes things easier when in fact we’re falling into a trap!” Then, for each matching response (A, A-3), printed on a card, the subject was led to explain why it is misleading and then to put it under a transparent hatched area (which depicted inhibition) in the response box (see board a in Figure 1B). [In line with Perner (1998), inhibition and set-shifting depend on a metarepresentation of the habitual act as maladaptive.] The training in the control condition simply explained the logic underlying the task (which was also part of the experimental training condition) without giving any warnings (Figure 1Bb). On the pre- and post-tests (Figure 1A), the performance of the experimental and control group participants was assessed on the Evans conditional rule falsification task (with rules like “If there is not a red square on the left, then there is a yellow circle on the right ” ). So, rather than being limited to practice or instruction about this pretraining task, our experimental training was aimed to help subjects transfer a


209

self-regulatory—or self-feeling—mechanism, i.e., matching-bias inhibition, from the training task (Wason’s task) to the post-training task (Evans task). The results indicated an inhibition training effect but no effect of logic training without the inhibition component, which leads us to confirm that the matching bias does not seem to be due to the lack of deductive logic (in agreement with Evans, 1998, cited earlier), but to a specific attentional or executive inhibition failure. In addition to this psychological study, recent neuroimaging data (Houdé et al., 2000, 2001) has provided evidence of a striking shift from the posterior part of the brain (a perceptual network), to a prefrontal network, right at the time when this type of inhibitory control is setting in (and no such neuroanatomical shifting after logic training without the inhibition component). However, in Houdé and Moutier’s (1996) results, not all of the experimental group participants were receptive to the matching-bias inhibition training (some were still biased after training). The interpretation proposed by Houdé and Moutier was that unreceptive participants may be perceptual field-dependent (FD) (in the traditional sense of the term defined in differential psychology by H.A. Witkin), while the other participants may be field-independent (FI). Remember that field-dependence is a measure of an individual’s ability to disregard the perceptual context and implement an analytic process. Thus, because of their cognitive style, FD participants would need more efficient inhibition (and thus, more in-depth training) than FI participants in order to ward off the matching strategy. In the framework of Pascual-Leone’s (1989) model (see also Pascual-Leone, 1988, 2000), which describes the relationships between a field operator (F) and attentional activation and inhibition operators in working memory, one can assume that reasoning is initially guided by the application of the perceptual schemes with the greatest activation strength (in the Evans conditional rule falsification task, a red square on the left of a yellow circle, i.e., the matching scheme), that is, schemes whose triggering conditions (see processing-negations and heuristic-analytic accounts described earlier) correspond to the salient aspects of the field (the geometric figures mentioned in the rule). Yet according to Pascual-Leone, F can either facilitate reasoning, or on the contrary, hinder it. In the matching-bias case, F would impede deductive reasoning in all participants (whether FD or FI) on the pre-test, and in FD participants, could reduce the potential benefits of the experimental inhibition training. These participants may therefore need more training than FI ones. Using the same paradigm as in Houdé and Moutier’s (1996) experiment, but with an additional field-dependence task on the pre-test (a French version of the Embedded Figures Test or EFT: Oltman, Raskin, & Witkin, 1985; Witkin, Oltman, & Karp, 1971), the present study (Study 1) tested the hypothesised relationship between receptiveness to matching-bias inhibition training and perceptual field dependence/independence.

210


Figure 1. (A) Pre- and post-test materials. Evans conditional rule falsification task (1972, 1989): Example of matching bias (red square, yellow circle) and logical response (blue square, green diamond) for the rule “If there is not a red square on the left, then there is a yellow circle on the right”. (B) Materials used for the training with the Wason (1968) card selection task: (a) Experimental training in matching-bias inhibition (board a depicts the executive processes involved in correctly performing the task, i.e., the inhibition process, shown as a hatched area); (b) Materials used for training without the inhibition component (only the logical explanation of the wrong answers A and A-3, and the right answer A-7).

210


211

To solve the EFT, participants had to find simple geometric figures embedded in a complex one, by outlining them with a pencil. We already know (Huteau, 1987) that this test can be used to classify individuals (FD or FI) on the basis of their ability to perceive an element separate from its perceptual context (just like the elements in the contrast sets, according to processing-negations theory) and to adopt an analytic attitude in problem solving (just like the analytic processes, according to heuristic-analytic theory). This psychometric technique tests “flexibility of closure”, i.e., participants’ ability to break down one perceptual form (the complex figure) and consider another (Huteau & Lautrey, 1999). This is precisely what is required to inhibit the perceptual matching bias in deductive reasoning tasks, where closure corresponds to the attentional focus on the elements mentioned in the conditional rule (geometric figures, letters, or numbers). We expected to observe a post-test effect of inhibition training in the experimental group that would replicate Houdé and Moutier’s (1996) results (see also Houdé et al., 2000, 2001). We also expected experimental group participants who were not receptive to the training (ones who still exhibited the matching bias after inhibition training) to be FD, and training-receptive participants to be FI.

STUDY 1 Method Participants

The experimental sample was made up of 36 Parisian adults whose mean age was 24;6. All were at least high-school graduates, and there was an equal number of men and women. The criteria for assigning participants to the experimental conditions are given in the procedure.

Procedure

Testing was individual and took place in two sessions held 1 week apart. The first session consisted of a pre-test, which included the perceptual fielddependence test or Embedded Figures Test (EFT), and the Evans conditional rule falsification task. After the pre-test, the sample was divided into an experimental group and a control group, each containing 18 participants matched on their Evans and EFT test scores. The second session was the experimental training in matching-bias inhibition or control training (logic training without the inhibition component), using Wason’s card selection task, and an immediate post-test in which participants were tested again on the Evans task (with a delayed post-test taken a week later).

212


Tests (pre- and post-) and training

Field-dependence test or Embedded Figures Test. The Embedded Figures Test was part of the pre-test. Participants had to find simple figures embedded in a complex one by outlining them with a pencil, without being able to view the simple and complex figures at the same time. This was achieved by having the simple figures (A to H) drawn on the back of the test booklet. The subject could turn over the booklet to look at them as many times as desired during the exercise. The EFT has three parts. The first part includes seven practice items and is used to make sure the subject understands the exercise. The other two parts each consist of nine more difficult items that make up the test proper. Test scores thus range from 0 to 18. For each part, a time limit of 5 minutes is set. A detailed description of the testing instructions is given on pages 20–22 of the ECPA manual (Oltman et al., 1985). Evans conditional rule falsification task. This task (described in the introduction) was used on the pre- and post-tests. The participants were given 12 coloured geometric figures (circles, squares, and diamonds in blue, yellow, red, and green). They had to falsify the conditional rule “If there is not a red square on the left, then there is a yellow circle on the right” (this rule triggers a matching bias in most participants, who answer “red square, yellow circle”). The rule was written on a board and the instructions were to “make the rule false”. Participants responded by putting two coloured figures, one for the antecedent and the other for the consequent, in a double box drawn on the board. Experimental training in matching-bias inhibition and control training without an inhibition component. The training procedure began with Wason’s card selection task described in the introduction. Faced with cards A, D, 3, and 7, the subject had to state which card or cards had to be turned over to verify the rule “If there is an A on one side of a card, then there is a 3 on the other”. The four cards were lined up on a board with the rule written across the top. After Wason’s task, the experimental group underwent the matching-bias inhibition training and the control group underwent the no-inhibition-component training (exact replication of the procedure used in Houdé & Moutier, 1996; see also Houdé et al., 2000, 2001). The training lasted approximately 30 minutes. The following is an excerpt from the matching-bias inhibition training (see Figure 1Ba). As stated in the introduction, the specificity of this training condition lies in its warning elements (shown in italics). “… Thus, the goal here is (1) to not fall into the trap of the two cards A and 3 mentioned in the rule, and (2) to consider all of the cards, A, D, 3, 7, one by one by imagining the number or the letter it might have on the other side to see whether these cards can make the rule false … To help you understand, let’s consider the


213

different answers and eliminate the wrong ones—the ones that make you fall into the trap—to find the right answer.” The participant is then shown the second part of the experimental materials (board a in Figure 1B). It is another board placed right next to the first, on which the response repertoire is depicted by a box, the inhibition process is depicted by a transparent hatched area on top of the box, and the activation process is depicted by an unhatched circle cut out from the middle of the hatched area. The answers (A, A-3, and A-7) are represented on three small cards which can be slid into the response repertoire. The answer cards are of different colours, depending on whether the answer is wrong (A, A-3) or right (A7). The experimenter says to the subject: “In the box you see here, we’re going to put the different answers written on these cards, while clearly separating the wrong answers which make you fall into the trap—we’ll put them under the hatching—from the right answer. Let’s start with answer A [...].” The training ends when the subject is capable of producing correct explanations (without experimenter prompting) for wrong answers A and A-3.

The control condition (Figure Bb) used exactly the same procedure except for the inhibition warning elements shown in italics. (Detailed instructions for these two training conditions are available upon request.)

Results

The analyses presented here pertain to (1) failures on Wason’s task at the beginning of the training phase, (2) training effects on the Evans task, and (3) relationships between the training effects and perceptual field dependence/ independence on the Embedded Figures Test.

Wason task failures

The response patterns on Wason’s task indicated massive failure; this is in line with the findings in the literature. There was not a single correct answer (A-7) in the entire sample. The most frequent wrong answer, observed in 18 participants out of 36, was A-3, which corresponds to the matching bias. The answers A and 3, which can be called partial matching, were given by 10 participants (8 and 2, respectively). The remaining 8 participants gave a variety of wrong answers including 7, A-D-7, A-D, and D-7. Thus, a matching strategy (A-3, A, or 3) appears to be the typical bias triggered by Wason’s selection task: 28 participants (78% of the sample) relied on total or partial matching.

Inhibition training effects on the Evans task

The pre-test results indicated as a whole that participants’ logical performance in both the experimental and control groups was very low (remember that, by construction, the scores of the two groups were comparable: see Method). The participants’ scores divided them into two sub-populations

214


(see Figure 2a), a biased one (n = 15) and a logical one (n = 3) (binomial test significant at .004 for these two groups, with N = 18 and k = 3; see Siegel & Castellan, 1988). The clearly predominant group exhibited biased behaviour, which in the present case corresponded to the matching bias (total or partial: matching with the figure mentioned in the antecedent and/or consequent: red square and/or yellow circle). The question raised here is: What were the respective proportions of the biased and logical sub-populations within the experimental group, as compared to the control group, on the post-test? The results showed that the proportions changed considerably between the pre- and post-tests (Figure 2a). After inhibition training, the experimental group could no longer be called globally biased (nonsignificant binomial test at .407 for the success/failure contrast with N = 18 and k = 8). On the other hand, the control group, which had undergone noinhibition-component training (where, by subtraction, only the logic component remained), still had a clearly predominant matching-bias sub-population (binomial test significant at .001 with N = 18 and k = 2). An intra-individual analysis of changes from biased to logical behaviour between the pre-test and the post-test indicated an inhibition training effect (existence of biased logical patterns) in the experimental group only: MacNemar’s 2 = 5.14, p < .05 with df = 1 (no effect in the control group). Note that the results in Figure 2a are the delayed post-test scores, which means that the inhibition training effect was robust, persisting even after a week. The immediate post-test analyses yielded comparable results (binomial test nonsignificant for the experimental group and significant at .001 for the control group; MacNemar’s 2 = 4.17, p < .05 with df = 1).

Inhibition training effects and perceptual field dependence

As Figure 2a shows, not all experimental group participants were receptive to the inhibition training (existence of individual patterns of the biased biased type), although the training did significantly change the respective post-test proportions of biased and logical participants on the Evans task (a finding already obtained in Houdé & Moutier, 1996). From a differential standpoint, and in reference to the Embedded Figures Test scores, the question was whether receptiveness to inhibition training was linked to perceptual field dependence. Remember that the EFT score ranges between 0 and 18. Based on the scaling chart by education and sex (taken from the French adaptation of this test; see Oltman et al. 1985, p. 19), field-dependent (FD) participants are ones located in the two lower quartiles, i.e., score between 0 and 15 for men and between 0 and 13 for women, and field-independent (FI) participants are ones located in the two upper quartiles, i.e., score between 16 and 18 for men and between 14 and 18 for women. Figure 2b shows the pre- and post-tests scores of the experimental group


a

b

Pre-test

n 16 14 12 10 8 6 4 2 0

n

Logical Biased

Control

215

Pre-test (experimental group)

9 8 7 6 5 4 3 2 1 0

Experimental

FD FI

Logical

Biased

Post-test (experimental group)

Post-test n

n 10

16 14 12 10 8 6 4 2 0

8 Logical Biased

FD

6

FI

4 2 0

Control

Experimental

Logical

Biased

Figure 2. (a) Effect of experimental inhibition training: Evans task scores on the pre-test (logical or biased response) and on the post-test for the experimental group (training in matching-bias inhibition) and for the control group (logic training without the inhibition component) (n = 36). (b) Inhibition training effect and perceptual field dependence /independence on the EFT: Evans task scores (logical or biased response) on the pre- and post-tests, as a function of field dependenc e (FD) or independence (FI) in the experimental group (n = 18). (Note that these results are the delayed post-test scores. The immediate post-test analyses yielded comparable results.)

on the Evans task (biased or logical responses), as a function of whether the participants were FD or FI on the EFT. The Evans task scores on the post-test were significantly linked to field dependence [delayed post-test: Fischer test significant at .025 with A + B = 10 and C + D = 8 (totals in right margin), A = 7 and C = 1; see Siegel, 1956; immediate post-test: Fischer test significant at .01]. An analysis of intraindividual performance changes between the pre- and post-tests indicated that all experimental group participants who had progressed by the immediate post-test after inhibition training (biased logical patterns) were FI, while those who had not progressed with training (biased biased patterns) were mostly FD (Fischer test significant at .025 with A + B = 8, C + D = 7, A = 5, and C = 0).

216


Discussion

In line with what is traditionally stressed in experimental research on deductive reasoning, our data pointed out a massive amount of “irrational” behaviour in adults [where irrational means a deviation with respect to the normative model of classical deductive logic, what Evans & Over (1997) call Rationality 2]. The participants ’ irrationality was attested on the pre- and post-tests by a matching bias on the Evans conditional rule falsification task (total or partial matching), and by their massive failure on Wason’s card selection task performed at the beginning of the training phase (100% failure, 78% of which was due to the matching bias). All of these results provide empirical arguments that could lead the psychologist or philosopher to reject any presumption of rationality. This conclusion will not be drawn here though. The first essential point brought out by our results is the effect of experimental training in matching-bias inhibition (contrasted to the lack of a logic training effect without the inhibition component), suggesting that participants can be rational (achieve deductive competence) while exhibiting inefficient inhibition which prevents the expression of that rationality. Houdé and Moutier’s (1996) results were replicated here (see also Houdé et al., 2000, 2001). The interesting thing about Evans’s (1972, 1989) data, presented in the introduction, is not so much that most participants turn out to be “irrational” (that is, they fall prey to the matching bias), but that those participants are so while other participants in the same sample exhibit rational behavior that reflects a logical analysis of the task. If, as Evans claims, a matching bias is systematically triggered by the task (which makes it potentially present in everyone), there is every reason to believe that the difference between the former and latter participants lies in their ability to inhibit the bias, not in the lack of a relevant logical scheme (in agreement with Evans, 1998). This was confirmed here by the observed effectiveness of the inhibition training. The second essential point in our results is related to the previously observed (Houdé & Moutier, 1996) large interindividual variability in participants’ receptiveness to inhibition training. This type of training did significantly change the respective proportions of participants who answered in a biased or logical way on the Evans task, but it was far from being beneficial for everyone. The question raised at the onset of this study concerned whether or not receptiveness to inhibition training is linked to perceptual field dependence/independence (in reference to Pascual-Leone’s model, 1989). The results for the Embedded Figures Test lead us to answer “yes ” to this question: participants who still produced a biased answer after inhibition training (individual patterns of the biased biased type) mainly had a FD cognitive style, whereas the opposite was true for participants whose response pattern was biased logical. It appears as though FD participants need greater inhibitory efficiency (and thus, more indepth inhibition training) than FI participants, in order to resist the matching bias.


217

It should nevertheless be mentioned that the EFT, like any other psychometric technique, is only an indicator, and as such, its validity might be questioned (see Linn & Kyllonen, 1981). In any case, although our result needs replication, it was very clear-cut, and thus opens up a new research pathway, one that combines the cognitive psychology of inhibition, the psychology of deductive reasoning, and differential psychology. To support our interpretation in terms of inhibition (rather than logic), it would also be interesting to apply the same training paradigm, but to use conditional rules on the pre- and post-tests that have negation in the consequent, not in the antecedent. As seen earlier at the beginning of the introduction, it has been shown in this case that people succeed naturally in the Evans rule falsification task because the perceptual matching response coincides with the logical response (and hence would lead here to pre-test success). Consequently, for this type of rule, training in matching-strategy inhibition should lower posttest performance. If it does, then this will show that the specific effect of the training was to discourage matching rather than encourage logical thinking. This was tested experimentally in the second study.

STUDY 2

Based on our data, Study 1’s discussion stated that difficulty in considering the contrast sets (Oaksford & Stenning, 1992) or redirecting attention to the logically relevant elements (Evans 1989, 1998) is linked to the ability to inhibit the matching strategy triggered by perceptual cues. Hence (1) the observed effect of matching-bias inhibition training, and (2) the link between training receptiveness (biased logical pattern) and perceptual field independence. In the present control study, we wished to test the role of cognitive inhibition in a priming adaptation of our paradigm, through a two-part, prime–probe selective attention or executive task. We exactly replicated the Houdé and Moutier paradigm (1996) except that on the pre- and post-tests, conditional rules with a negation in the consequent were used, as in “If there is a blue square on the left, then there is not a green diamond on the right”. In this case, as Evans (1972, 1989) stressed, the matching response corresponds to the logical response (hence the high success rate). For this new version of the paradigm, a prediction based on the inhibition account is that experimental group performance will be worse after matching-bias inhibition training than control group performance will be after logic training without the inhibition component. Such a prediction corresponds to a negative priming effect (see Bruce & Tipper, 1998; Houdé, 2001; Neill, Valdes, & Terry, 1995; Tipper, 1985). In part A (the prime), a subject responds to an “attended” strategy (or stimulus), S1, ignoring another strategy S2. In Part B (the probe), the same subject must unexpectedly respond to the just-ignored strategy S2, or to a new strategy, S3. If the processing of S2 was inhibited in order to further process S1, subsequent processing of S2 must

218


overcome any persistence of that inhibition. Consequently, S2 should be more difficult to process than S3. In many experiments that consider the prediction of attention as inhibition, subjects are slower and/or less accurate in responding to strategies related to recently ignored strategies (see Neill et al., 1995; Tipper, 1985; Tipper, Bourque, Anderson, & Brehaut, 1989; Neill et al., 1995). The opposite prediction, where inhibition is an epiphenomenon, is the lack of a priming effect or a positive priming effect if S2—a misleading but possible strategy in A—has not yet dissipated by the time the subject answers in B. Initially, the negative priming paradigm was defined in terms of stimuli, with S1 as the target and S2 as the distractor. The present set-up is a variation applied to reasoning strategies (see also Houdé & Guichart, 2001, for an application in the field of numerical cognition). During matching-bias inhibition training, the processing of the matching strategy (S2) was inhibited in order to redirect attention to the logically relevant elements (S1). Consequently, if on the pre- and post-tests the matching response (S2) corresponds to the logical response, we expect to observe worse experimental group performance on the post-test (compared to a high success rate on the pre-test). So, in the present paradigm where the matching strategy corresponds to the logical answer on the Evans task, the inhibition training is designed to create a cognitive conflict on the post-test.

Method Subjects

The experimental sample was made up of 30 Parisian adult subjects whose mean age was 20;7. All were at least high-school graduates, and there was an equal number of men and women. The criteria for assigning subjects to the experimental conditions are given in the procedure.

Procedure

Testing was individual and was run in two sessions held 1 week apart. The first session consisted of a pre-test on the Evans conditional rule falsification task with a negation in the consequent of the rule. After the pre-test, the sample was divided into an experimental group and a control group, each containing 15 subjects matched on their Evans scores. The second session consisted of either experimental training in matching-bias inhibition or control training, both using Wason’s card selection task, and an immediate post-test in which subjects were tested again on the Evans task. A delayed post-test was taken a week later.

Tests (pre- and post-) and training

Evans conditional rule falsification task. The Evans task (described in the introduction) was used on the pre- and post-tests. The subjects were given 12


219

coloured geometric figures (circles, squares, and diamonds in blue, yellow, red, and green). They had to falsify four conditional rules (presented in random order) with a negation in the consequent: e.g., “If there is a blue square on the left, then there is not a green diamond on the right”. Four rules were used in this study because negative priming cannot be measured in an all-or-nothing situation. These rules trigger a matching bias response that corresponds here to the logical response (“blue square, green diamond” in the previous example). The rule was written on a board and the instructions were “to make the rule false”. Subjects responded by putting two coloured figures, one for the antecedent and the other for the consequent, in a double box drawn on an adjacent board. Experimental training in matching-bias inhibition and control training without an inhibition component. We used exactly the same training procedures (matching-bias inhibition and control) as in Study 1.

Results

The analyses presented here pertain to (1) failures on Wason’s task at the beginning of the inhibition training, and (2) training effects on the Evans task for the negative priming paradigm.

Wason task failures

As in Study 1, the response patterns on Wason’s task indicated massive failure. Once again, there was not a single correct answer (A-7) in the entire sample. The most frequent wrong answer, observed in 20 subjects out of 30, was A-3, the matching-bias response. Partial matching answers A or 3 were given by 10 subjects (7 and 3, respectively). Thus, a matching strategy (A-3, A, or 3) appears to be the typical bias triggered by Wason’s selection task: 30 subjects (100% of the sample) relied on total or partial matching.

Inhibition training effects on the Evans task

Our prediction in this modified paradigm was a negative priming effect in the matching-bias inhibition training condition (experimental group), i.e., worse logical performance on the post-test, which is the opposite prediction to Study 1 (see Figure 2). This is exactly what the observed success patterns showed (see Figure 3). In the experimental and control groups alike, all subjects had a success pattern (four out of four rules) on the pre-test, replicating Evans’s results. On the post-test, however, the experimental group subjects had failure patterns (failure on one to four rules out of four) and could no longer be called “logical” (nonsignificant binomial test at .304 for the success/failure contrast with N = 15 and k = 6). An intra-individual analysis of changes from logical (or matching in

220

MOUTIER, ANGEARD, HOUDÉ Pre-test n 16 14 12 10

+

8

-

6 4 2 0 Control

Experimental

Post-test n 16 14 12 10

+

8

-

6 4 2 0 Control

Experimental

Figure 3. Additional study with the same paradigm as Houdé and Moutier (1996) (see Figure 1), but using conditional rules with negation in the consequent on the pre- and post-tests (e.g., “If there is a blue square on the left, then there is not a green diamond on the right”). N = 30, 15 in the experimental group (training in matching-bias inhibition) and 15 in the control group (logic training without the inhibition component). Effect of experimental inhibition training: Evans task scores on the pre-test and on the post-test for the experimental group and for the control group ( + denotes patterns with success on all four rules; – denotes patterns with failure on one to four rules).

this case) to illogical behaviour between the pre-test and the post-test indicated an inhibition training effect (existence of six logical illogical patterns) in the experimental group only; no such patterns were observed in the control group, where consistent success was still achieved (significant binomial test at .02 with N = 6 and k = 0). This negative priming effect after inhibition training was no longer observed on the delayed post-test, which is consistent with the temporary nature of the priming phenomenon.

Discussion

For this study, which used a new version of the Houdé and Moutier (1996) paradigm with negative consequent conditional rules, our prediction based on the inhibition account was that after matching-bias inhibition training, experimental group performance would be worse than control group performance (logic training without the inhibition component). This is exactly


221

what the observed success patterns suggested. In the experimental and control groups alike, all subjects had an overall pattern of success on the pre-test, replicating Evans’s result. On the other hand, on the post-test, the experimental group had failure patterns (failure on one to four rules out of four); no such patterns were observed in the control group, where consistent success was still achieved. This additional study based on a negative priming paradigm thus clearly confirms that our experimental training (in Study 1 and in Study 2) discouraged perceptual matching rather than encouraging logical thinking. This result is consistent with Study 1, where a clear-cut relationship was obtained in the experimental group between receptiveness to inhibition training (biased logical pattern) and perceptual field independence. So, the present data lead us to the conclusion that difficulty in making correct deductions is linked to the ability to inhibit the matching strategy triggered by perceptual cues.

CONCLUSION

The essential point brought out by the results of Study 1 is related to the previously observed effect (Houdé & Moutier, 1996) of experimental training in matching-bias inhibition, where interindividual variability was linked to fielddependent/independent cognitive style. In relation to the two theoretical interpretations pointed out by Evans (1998 ), i.e., Oaksford and Stenning’s (1992) processing-negation account and Evans’s (1989) heuristic-analytic account, the present data suggest that difficulty in considering contrast sets (processing the negations) or redirecting attention towards logically relevant elements—as the not-heuristic first attracts attention to the negated proposition —is linked to the greater potentiality of field-independent subjects to inhibit the matching strategy triggered by perceptual cues (geometric figures, letters, or numbers mentioned in the rule). The essential point brought out by the results of Study 2 using a new version of our paradigm was the negative priming effect, which points out the dissociation of the inhibition and logical components. Given that such a negative priming effect provides further empirical evidence for inhibitory mechanisms in deductive reasoning, this point leads us to confirm that matching bias does not seem to be due to a lack of logic, but to a specific attentional or executive inhibition failure. This interpretation is in line with the selective attention account already proposed by Evans (1989, 1998) and allows us to explain why bias may occur even when subjects have logical capabilities. The lack of a logic training effect can thus be accounted for by the fact that what is missing is not deductive competence per se, but inhibitory control of reasoning (Houdé, 1997). On this point, our analysis is comparable to that of authors in the field of developmental psychology, like Diamond (1991), who stressed that “cognitive development can be conceived of, not only as the progressive acquisition of

222


knowledge, but also as the enhanced inhibition of reactions that get in the way of demonstrating knowledge that is already present ” (p. 67) (see also Dempster & Brainerd, 1995; Houdé, 2000). Similar effects of experimental training in matching-bias inhibition have also been observed in children (Moutier, 2000). The behavioural data reported here, in conjunction with our recent brainimaging data (Houdé et al., 2000, 2001), argue in favour of the capacity of the human mind/brain—when trained under certain conditions and as a function of cognitive style (field dependence/independence)—to overcome deductive reasoning bias. This is what Evans (1989) called “debiasing” (p. 113 et seq.), although he remained quite sceptical about the possibility of triggering this correction mechanism: “On the whole, there is very little evidence that deductive reasoning biases can be removed by verbal instructions relating to the underlying logical principles” (pp. 116-117). Our own results confirm this contention, that training based solely on logical explanations (our control groups) is in fact ineffective. On the other hand, we now know that executive training aimed directly at bias inhibition, which Evans had not envisaged, can radically change reasoning performance and lead to neural reconfigurations. The results obtained here thus add a new piece of evidence to the debiasing case. Manuscript received 18 September 1998 Revised manuscript received 7 September 2001

REFERENCES

Baddeley, A.D. (1986). Working memory. Oxford: Oxford University Press. Baddeley, A.D. (1992). Working memory. Science, 255, 566–569. Bjorklund, D.F., & Harnishfeger, K.K. (1990). The resources construct in cognitive development : Diverse sources of evidence and a theory of inefficient inhibition. Developmental Review, 7, 93–130. Bruce, M., & Tipper, S.P. (1998). Attention and inhibition. In H. Pashler (Ed.) Attention (pp. 141–173). Hove, UK: Psychology Press. Dagenbach, D., & Carr, T.H. (Eds.) (1994). Inhibitory processes in attention, memory, and language. New York: Academic Press. Dempster, F.N. (1992). The rise and fall of the inhibitory mechanism: Toward a unified theory of cognitive development and aging. Developmental Review, 12, 45–75. Dempster, F.N., & Brainerd, C.J. (Eds.) (1995). Interference and inhibition in cognition. New York: Academic Press. Diamond, A. (1991). Neuropsychological insights into the meaning of object concept development. In S. Carey & R. Gelman (Eds.), The epigenesis of mind: Essays on biology and cognition (pp.67–110). Hillsdale, NJ: Lawrence Erlbaum Associates Inc. Engel, P. (1993). Philosophie de la pensée logico-mathématique. In O. Houdé & D. Miéville (Eds.), Pensée logico-mathématique [Logico-mathematical thinking] (pp. 205–228). Paris: Presses Universitaires Français. Evans, J.St.B.T. (1972). Reasoning with negatives. British Journal of Psychology, 63, 213–219. Evans, J.St.B.T. (1989). Biases in human reasoning. Hove, UK: Lawrence Erlbaum Associates Ltd. Evans, J.St.B.T. (1998). Matching bias in conditional reasoning. Do we understand it after 25 years? Thinking & Reasoning, 4, 45–82.


223

Evans, J.St.B.T., Newstead, S.E., & Byrne R.M.J. (1993). Human reasoning: The psychology of deduction. Hove, UK: Lawrence Erlbaum Associates Ltd. Evans, J.St.B.T., & Over, D.E. (1997). Rationality in reasoning: The problem of deductive competence. Current Psychology of Cognition, 16, 3–38. Houdé, O. (1997). The problem of deductive competence and the inhibitory control of cognition. Current Psychology of Cognition, 16, 108–113. Houdé, O. (2000). Inhibition and cognitive development: Object, number, categorization, and reasoning. Cognitive Development, 15, 63–73. Houdé, O. (2001). Interference and inhibition (Psychology of). In J.N. Smelser & P.B. Baltes (Eds.), International encyclopedia of the social and behavioral sciences. Oxford: Pergamon, Elsevier Science. Houdé, O., & Guichart, E. (2001). Negative priming effect after inhibition of number/length interference in a Piaget-like task. Developmental Science, 4, 119–123. Houdé, O., & Moutier, S. (1996). Deductive reasoning and experimental inhibition training: The case of the matching bias. Current Psychology of Cognition, 15, 409-434. Houdé, O., Zago, L., Crivello, F., Moutier, S., Pineau, A., Mazoyer, B., Tzourio-Mazoyer, N. (2001). Access to deductive logic depends on a right ventromedial prefrontal area devoted to emotion and feeling: Evidence from a training paradigm. NeuroImage, 14, 1486–1492. Houdé, O., Zago, L., Mellet, E., Moutier, S., Pineau, A., Mazoyer, B., & Tzourio-Mazoyer, N. (2000). Shifting from the perceptual brain to the logical brain: The neural impact of cognitive inhibition training. Journal of Cognitive Neuroscienc e, 12, 721–728. Huteau, M. (1987). Style cognitif et personnalité: La dépendance/ind épendance à l’égard du champ [Cognitive style and personality: Field dependence /independence]. Lille: Presses Universitaire Lille. Huteau, M., & Lautrey, J. (1999). Evaluer l’intelligence: Psychométrie cognitive [Evaluating intelligence: Cognitive psychometrics]. Paris: Presses Universitaire Français. Linn, M.-C., & Kyllonen, P. (1981). The field dependence /independence construct. Journal of Educational Psychology, 73, 261–273. Macmillan, M. (1996). The concept of inhibition in some nineteenth century theories of thinking. Brain and Cognition, 30, 4–19. Moutier, S. (2000). Deductive reasoning and experimental matching-bias inhibition training in school children. Current Psychology of Cognition, 19, 429–452. Neill, W.T., Valdes, L.A., & Terry, K.M. (1995). Selective attention and the inhibitory control of cognition. In F.N. Dempster, C.J. Brainerd (Eds.), Interference and inhibition in cognition (pp. 207–261). New York: Academic Press. Oaksford, M., & Stenning, K. (1992). Reasoning with conditional containing negated constituents. Journal of Experimental Psychology: Learning, Memory and Cognition, 18, 835–854. Oltman, P.K., Raskin, E., & Witkin, H.A. (1985). Manuel du test des figures encastrées [Manual for embedded figures test]. Paris: Editions du Centre de Psychologie Appliquée. Pascual-Leone, J. (1988). Organismic processes for neo-Piagetian theories. In A. Demetriou (Ed.), The neo-Piagetian theories of cognitive developmen t (pp. 25–64). Amsterdam: North-Holland. Pascual-Leone, J. (1989). An organismic process model of Witkin’s field-dependence independence. In T. Globerson & T. Zelniker (Eds.), Cognitive style and cognitive development (pp. 36–70). Nordwood, NJ: Ablex. Pascual-Leone, J. (2000). Reflections on working memory: Are the two models complementary ? Journal of Experimental Child Psychology, 77, 138–154. Perner, J. (1998). The meta-intentional nature of executive functions and theory of mind. In P. Carruthers & J. Boucher (Eds.), Language and thought (pp. 270–283). Cambridge: Cambridge University Press.

224


Siegel, S. (1956). Nonparametric statistics for the behavioral sciences. New York: McGraw-Hill. Siegel, S., & Castellan, N.J. (1988). Nonparametric statistics for the behavioral sciences (Second Edition). New York: McGraw-Hill. Smith, R. (1992). Inhibition. History and meaning in the sciences of mind and brain. London: Free Association Books. Tipper, S.P. (1985). The negative priming effect: Inhibitory priming by ignored objects. Quarterly Journal of Experimental Psychology, 37, 571–590. Tipper, S.P., Bourque, T., Anderson, S., & Brehaut, J. (1989). Mechanisms of attention: A developmental study. Journal of Experimental Child Psychology, 48, 353–378. Wason, P. (1968). Reasoning about a rule. Quarterly Journal of Experimental Psychology, 20, 273–281. Witkin, H.A., Oltman, P.K., & Karp, S.A. (1971). Manual for embedded figures test, children’s embedded figures test, and group embedded figures test. Palo Alto, CA: Consulting Psychologists Press.