Working Memory Capacity and Thinking Disposition as Predictors of the use of Heuristic or Analytic Processing in Syllogistic Reasoning

Working Memory Capacity and Thinking Disposition as Predictors of the use of Heuristic or

Analytic Processing in Syllogistic Reasoning

AbstractThe notion that human beings have the capacity for two different modes of thought, an effortless, intuitive, heuristic

mode and an effortful, logical, analytic mode, is a profound one which goes to the very nature of our consciousness.

This study uses the belief bias effect in syllogistic reasoning to determine the relationship between two individual

differences; working memory capacity and need for cognition, and the propensity to use one or the other system of

thought. A correlational design was used. An opportunity sample of 115 participants took part in the study which

consisted of a pencil and paper syllogistic reasoning task and a short form of the need for cognition scale (Cacioppo,

Petty, & Feng Kao, 1984), and an online operation span (OSPAN) task (Krantz, 2008). It was found that both need for

cognition and working memory correlated positively with the ability to avoid belief bias errors. When confined to a

model the amount that each predicted the ability to avoid belief bias was additive, the model explaining 30% of the

variance in belief inhibition. Findings are explained in terms of dual-processing theory; those with greater need for

cognition are more inclined to use the analytic processing as they take greater enjoyment from it, those with a larger

working memory capacity are as likely to use analytic or heuristic processing as those with a smaller working memory

capacity but have more effective analytic and heuristic systems. As an aside to this, participants degree of certainty to

items on the syllogistic reasoning task were analysed as an indication of participants ability to detect conflict between

logic and belief, although a significant result was found, possible methodological issues limited the interpretation of this

part of the study.

General Introduction

The idea that humans have the capacity for two qualitatively different modes of thinking is a profound one. It is perhaps

the one attribute that sets us apart from other animals. Although monikers and details vary, the concept of duality in

human thought process is universal. The first formal theory was posited by William James in 1890 (James, 1890) but

dual-processing theory in its current form was first put forward by Newell and Simon (1972) who named the two

systems “heuristic” and “universal” and used them in regard to problem solving. They stated that people will use the

heuristic mechanism, finding the solution to a problem from the application of past experience, whenever this is

possible. If the problem faced is novel, the heuristic system fails as there is no previous experience to relate it to, then

the universal system is used whereby innovative solutions are generated.

Stanovich and West (1982) built upon this theory relating it to human judgement and reasoning rather than problem

solving. They named the two modes of thought System 1 and System 2 processing and these terms have fairly stuck in

the literature. System 1 processing, related to Newell and Simon’s heuristics, is characterised by fast, intuitive

judgements requiring of little cognitive effort. System 2, related to the universal system of Newell and Simon, is

characterised by slow, effortful, serial processing which requires a greater degree of cognitive effort (De Neys, 2006b), it

is seen as more logical and analytic than system 1. In his review of dual-process theories, Evans (2007a) called the two

systems ‘heuristic’ and ‘analytic’ for systems 1 and 2 respectively; in the interests of simplicity this report will do the

same.

Although the dual process theory is successful in describing two systems of thought that appear to be different in

nature, it has been criticized for not explaining how the mechanisms underlying these systems actually work (e.g.

Osman, 2004; Evans 2007a, 2007b). To this end, Kahnemann (2002) in his Nobel Prize lecture placed heuristic and

analytic judgement along a continuum of accessibility of thought, with automatic perception on one end and hard

deliberation on the other. This is in effect a view that the two processes are not different qualitatively but differ

quantitatively, only in the amount of conscious effort that is required. More recent research however, using functional

magnetic resonance imaging (fMRI) has found evidence that the two systems use distinct pathways in the brain,

indicating a qualitative rather than simply quantitative difference (Goel V. , 2007; Reverberi, Shallice, D'Agostini, Skrap,

& Bonatti, 2009; Rodrigues-Moreno & Hirsch, 2009). Goel (2003) gave participants categorical syllogisms designed to

either elicit belief-based response or analytical response whilst blood flow to different brain regions was visualised using

fMRI. He found that that belief-based syllogisms elicited activation of the left-frontal temporal region in the areas

associated with long term memory, whilst the analytical-based syllogisms elicited activation of the bilateral parietal

system, areas associated with visuo-spatial reasoning. Goel’s findings are comfortably explained by dual-process

theory. When a syllogism accords with prior belief the brain uses the heuristic pathway to long term memory to

determine a response. When a conflict is detected between conflict and belief this pathway is blocked and the brain

uses the analytic process, which Goel surmises can be found in the Visuo-Spatial system. These forays into the

anatomical and physiological basis of dual-processes may soon elucidate its underlying mechanisms.

Whatever its nature, the concept of duality of thought pervades the literature and has been applied to a variety of

subjects such as problem solving (Gillard, Van Dooren, Schaeken, & Verschaffel, 2009), decision making (Dijksterhuis,

Bos, Nordgren, & Von Baaren, 2006) and social judgement (Petty & Brinol, 2006; Park, Levine, Westerman, & Foregger,

2007) among many others. The cognitive dual process theory has even been integrated into the psychodynamic

concepts of the conscious and unconscious (Epstein, 1994). The present research is interested in the dual-process

paradigm in relation to deductive reasoning, syllogistic reasoning in particular. An established approach to studying the

dual-process theory is through exploiting the belief bias effect. Belief bias is the tendency to agree with a conclusion

which concurs with one’s own beliefs even when that conclusion conflicts with deductive logic (Markovits & Nantel,

1989).

In the arena of dual-processing there is a history of using this effect as an indicator of (inappropriate) use of heuristic

processing e.g. (Evans, 1983; Denesraj & Epstein, 1994; Stupple & Ball, 2007; De Neys & Van Gelder, 2009). Belief-

Inhibition occurs when an individual suspends their belief and agrees with a conclusion that is logically valid even

although it conflicts with their previous beliefs; this is held as an indication of use of analytic processing.

The method that is most often used to measure belief bias is syllogistic reasoning (e.g. Evans, Handley, & Harper, 2001;

Roberts & Sykes, 2003). There are 4 types of syllogisms that are used; Believable-Valid and Unbelievable-Invalid

(whereby logical necessity and believability of the conclusion concur, these are non-conflict syllogisms), Believable-

Invalid and Unbelievable-Valid (whereby logical necessity and believability of the conclusion conflicts, these are conflict

or belief bias syllogisms). An example of an unbelievable-valid (conflict) syllogism is “No addictive things are

inexpensive, Some cigarettes are inexpensive, Therefore, some cigarettes are not addictive” (taken from Evans,

Barston, & Pollard, 1983).

The subject matter of this project is the individual differences in dual-processing using the syllogistic reasoning

paradigm described above, individual differences in question being need for cognition and working memory capacity.

Previous findings have shown that both need for cognition and working memory capacity predict the differential use of

heuristics or analytics (e.g. West, Toplak, & Stanovich, 2008; Bruine de Bruin, Fischhoff, & Parker, 2007; Kokis,

Macpherson, Toplak, West, & Stanovich, 2002) this research is original in (among other facets described further on) its

interest in whether these findings indicate a common or different origin. As an aside this study will also look at

participants’ awareness of the conflict between belief and logic in the conflict syllogisms.

Alternate Hypotheses:

1. Participants with a greater need for cognition will make fewer belief bias errors on a syllogistic reasoning task.

2. Participants with a larger working memory capacity will make fewer belief bias errors on a syllogistic reasoning

task.

3. Need for cognition and Working Memory Capacity will co correlate in their prediction of belief-inhibition.

4. Participants will be less sure of their answers to syllogisms where logic conflicts with belief than to syllogisms

where logic concurs with belief.

The same group of participants are used for the entire study so that comparisons can be made between all the

elements being studied. The report however is divided into sections for each hypothesis being examined.

1. Need for Cognition and Belief Bias

Thinking dispositions as a causative factor of thinking biases have received a good deal of attention in the literature

(e.g. West et al., 2008; Cavazos & Campbell, 2008; Kokis, Macpherson, Toplak, West, & Stanovich, 2002; Stanovich &

West, 1998), the motivation for which has often been to uncover individual differences in the ability to think critically.

The ability to inhibit one’s existing beliefs in order to evaluate the validity of an argument is one aspect of critical

thinking. Some of the personality dimensions that have been studied in relation to belief bias are; openness to

experience (Stanovich & West, 1997), Dogmatism (Sa, West, & Stanovich, 1999), epistemic self-regulation (Toplak &

Stanovich, 2002) and need for cognition (West, Toplak, & Stanovich, 2008; Kokis, Macpherson, Toplak, West, &

Stanovich, 2002) among others.

The notion of studying personality dimensions in relation to belief bias was first developed by Stanovich and West

(1998); prior to this, research into individual differences in dual-processing had focused on measures of cognitive ability

and general intelligence (e.g. Gilinsky & Judd, 1994). West and Stanovich studied a variety of related thinking

dispositions, for example actively open-minded thinking and dogmatism, in relation to numerous measures of critical

thinking, including belief bias in syllogistic reasoning. They found that the tendency to make belief bias errors was

correlated with thinking disposition. People high in dispositions such as open-mindedness were less likely to make belief

bias errors; whereas people high in dispositions such as dogmatism were more likely to make belief bias errors.

The study of need for cognition in belief bias began later in the literature with Macpherson and Stanovich (2007). Need

for cognition is a personality dimension which describes the extent to which a person enjoys, seeks out and engages in

effortful cognitive endeavours. The term was coined by Cohen, Stotland, & Wolfe (1955) but its current meaning was

concieved by Cacioppo & Petty (1982). It has been found to be a personality dimension distinct from others (Cacioppo,

Petty, Feinstein, & Jarvis, 1996; Sadowski & Cogburn, 1997). Macpherson and Stanovich (2007) found that need for

cognition was a predictor (significant albeit weak) of Belief Bias in syllogistic reasoning. In their experiment they

presented participants with 8 conflict syllogisms to test the belief bias effect and 8 non-conflict syllogisms; however

they only analysed the results from the conflict syllogisms ignoring the results from the non-conflict syllogisms. Results

for the non-conflict syllogisms could be assumed to represent participants’ general ability on syllogistic reasoning. The

effect they discovered therefore could be explained by general cognitive ability (which they also found to be predicted

by need for cognition) or by general ability on syllogistic reasoning. Although evidence from studies such as Stanovich,

West & Toplak (2008) has found need for cognition to predict the belief bias effect even when cognitive ability was

partialed out, it is still not clear whether this variability in belief bias could be explained by ability at syllogistic

reasoning in particular.

An interesting facet of the research done by Macpherson and Stanovich is the finding that although cognitive ability and

need for cognition predicted belief bias in syllogistic reasoning, these same individual differences did not predict

another critical thinking bias; myside bias. Myside bias is the tendency to bias thinking towards one’s own opinions; it is

different from belief bias which is the tendency to bias judgements towards one’s factual knowledge about the world.

Macpherson and Stanovich were somewhat at a loss to explain this phenomenon, but it can be explained by the dual

process theory: Whilst belief bias is an indicator of the use of one mode of processing (heuristics) over another, myside

bias is not thought to be related to heuristics or analytics (Stanovich & West, 2008). Thus this finding indicates that

people varying in need for cognition actually vary in their tendency to use one or the other system of processing, rather

than just varying in their ability to avoid bias. It is important to note however that the measures of belief bias and

myside bias used differed in their level of formality, the former using formal syllogistic reasoning, and the latter using

less formal argument generation tasks.

The subject of this part of the study is the relationship between the thinking disposition of need for cognition and the

belief bias effect. The aim is to find out if people differing in need for cognition vary in their tendency to use different

mechanisms of thought (heuristics and analytics as described above).

The present study differs from previous research in this area in that performance on belief bias syllogisms is controlled

for by performance on non-conflict syllogisms. In this way the hope is to separate the belief bias effect from cognitive

ability so that any relationship between correct answers and need for cognition can be said to come from propensity to

use one or the other system of processing rather than simply from level of ability at syllogistic reasoning.

1.2. Methods

1.2.1. Design

A correlational design was used; with the following variables: need for cognition operationalised with a questionnaire,

Belief-Inhibition Score on a syllogistic reasoning task (percentage of correct answers to syllogisms where logic conflicts

with belief), and the control variable Non-Conflict Score on a syllogistic reasoning task (percentage of correct answers to

syllogisms where logic and belief concur). Both the syllogistic reasoning task and the need for cognition scale were

administered with paper and pencil. The syllogistic reasoning task was administered before the Need for cognition scale

so as to control for ‘cueing’ effects of the need for cognition scale, i.e. Participants may have been cued to use their

analytic system of processing by the nature of the statements in the need for cognition scale. The entire test takes

about 20-30 minutes to complete.

1.2.2. Participants

117 participants took part in the study, results from 2 of whom were removed as they misunderstood the scales and

filled them in incorrectly, leaving 115 participants of which 66 (57%) female and 49 (43%) male. The participants were

all adults the youngest being 18 and the oldest being 67 (M = 36, SD = 13). Participants were drawn from an

opportunity sample of family, friends, colleagues and acquaintances of the researcher. They were from a variety of

backgrounds, >80% of participants fell into the following 5 categories; Students, Usability Consultants, Engineers, Local

Government workers and Sales people. At the time of taking part participants were currently residing in either Greater

London (57%), Somerset (31%) or Scotland (12%), Participants were well above average for the population in terms of

educational level (Highest level: GCSE 2%; A Level; 14%; Bachelors; 52%; Masters 28%; PhD 4%). None of the

participants were paid for taking part.

1.2.3. Materials

Syllogistic Reasoning Task: The syllogistic reasoning task was created by the student researcher. It was deemed

appropriate to create syllogisms for this study as there is no ‘gold standard’ syllogistic reasoning task available.

Previous studies involving syllogistic reasoning have also used their own syllogisms or have modified previous ones (e.g.

De Neys & Van Gelder, 2009).

The task consisted of 20 syllogisms in total. 12 syllogisms were designed to conflict between logic and belief; of which 6

Valid-Unbelievable (V-U) and 6 Invalid-Believable (I-B). The other 8 were designed to accord between logic and belief; of

which 4 Valid-Believable (V-B) and 4 Invalid-Unbelievable (I-U)1.

Syllogisms were laid out as such:

Syllogisms were designed to be reasonably challenging due to the expectation of educated participants (professionals

and university students), although care was taken to use language that all participants were likely to be able to

understand. Syllogisms from Markovits & Nantel (1989) have been used by many researchers (eg. De Neys & Van

Gelder, 2009) but these were found to be too simple for the current study, on preliminary testing using simpler

researcher designed syllogisms similar to those in Markovits & Nantel (1989), the 5 participants tested answered

correctly for all of the syllogisms, this would have prevented any useful results being obtained by the study.

Instructions included an example syllogism and explained that a syllogism is ‘a type of logic puzzle’. Participants were

instructed to answer whether the conclusion followed logically from the premises on each item. They were also asked to

indicate how certain they were of their answer, either ‘certain’ or ‘uncertain’ (See part 4 for explanation)

1 Valid syllogisms were of the types AII-1, EIO-2, OAO-3, EAO-4, IAI-4 (A = Universal affirmative, E = Universal negative, I = particular affirmative, O = Particular negative), numbers refer to the figural layout.Invalid syllogisms are invalid because of one of the following; illicit process of the minor term, illicit process of the major term, fallacy of the undistributed middle, affirmative conclusion from negative premises.None of the syllogisms commit the existential fallacy.

The syllogistic reasoning task was designed to be divided into two separate scales for analysis; A Belief-Inhibition Scale

made up of the 12 conflict syllogisms and a Non-Conflict Scale made up of the 8 non-conflict syllogisms. In the results

these are shown as percentage correct scores so that the two can be compared. See appendix 1 for the full paper and

pencil task given to participants.

Reliability of Scales: Cronbach’s alpha for the Belief-Inhibition Scale from the current sample was .820 showing high

internal consistency, each of the items add consistency (removing any one reduces Cronbach’s α). See appendix 2.1 for

SPSS readout. Cronbach’s alpha for the Non-Conflict Scale from the current sample was .804 showing high internal

consistency, each of the items adds consistency (removing any one reduces Cronbach’s α). See appendix 2.2 for SPSS

readout.

Participants scored higher on the non-conflict syllogisms than on the Belief-Inhibition (Conflict) syllogisms, the mean

difference in score was .229 the effect size was d = .48. On a paired samples t-test this difference was found to be

significant (t = 9.209, df = 114, p < .0005). This can be confirmed by noting that zero does not fall within the

confidence interval (.18, .28). Because the only difference between the two sets of syllogisms was whether belief

conflicted or concurred with logic we can conclude that this effect occurs as a result of belief bias. See Appendix 2.3 for

SPSS readout.

Need for Cognition Questionnaire: The 18 item Short Form of the Need for Cognition Scale (SF-NFC) (Cacioppo, Petty, &

Feng Kao, 1984) was used. Participants were instructed to indicate their level of agreement with each statement on a 5

point Likert scale (‘strongly agree’, ‘agree’, ‘don’t know’, ‘disagree’, ‘strongly disagree’) ‘don’t know’ was used as the

middle term rather than ‘neither agree nor disagree’ to discourage central tendency bias without forcing choice. The

scale includes both positive and negative statements. Positive statements such as: “I prefer life to be filled with puzzles

I must solve” and negative statements such as “Learning new ways to think doesn't excite me very much”. A score from

0 to 4 was given for each item (4 being high need for cognition, 0 being low need for cognition) and the 18 scores were

summed to give the need for cognition score.

1.2.4. Procedure

Participants were asked to complete the syllogistic reasoning test first and then the Need for Cognition Scale,

demographic information was also collected. The tests were administered on paper Most participants were instructed by

the researcher face to face, others were instructed over the phone and via email.

1.2.5. Ethics

Informed consent was obtained from all participants before they started the tests. They were informed of their right to

end the test at any point. Participants’ results were kept anonymous with individual results being identified by numbers,

which could also be used by anyone wishing to remove their results after the fact. Participants were given a debrief

sheet at the end of the study with contact details and further explanation of the aims and background to the research.

See appendix 1.2 and 1.3 for participant consent form and debrief sheet

1.3. Results

The results for Need for Cognition were: N = 115, M = 44.63, sd. = 9.70, results for percentage of correct conflict

syllogisms (Belief-Inhibition Score) were: M = 62.81, sd. = 23.31, results for the percentage of correct non-conflict

syllogisms (Non-conflict Score) were: M = 84.75, SD = 13.26. These results are summarised in table 1 (see appendix 3

for SPSS readout of all descriptive statistics).

Table 1: Descriptive Statistics for Need for Cognition, Belief-Inhibition and Non-Conflict Scores

Number Mean Std. Deviation

Need for Cognition

115 44.63 9.70

Belief-Inhibition Score

115 62.81 23.31

Non-Conflict Score 115 84.75 13.26

A sequential linear regression analysis was performed with need for cognition as the dependant variable. Non-conflict

Score was added first to the model then belief-inhibition score was added second. A partial correlation was also done

with non-conflict score as the controlling variable.

It was found that need for cognition is a significant predictor of belief-inhibition (F = 24.535, p < .0005). This was a moderate positive correlation (r = .422, N = 115) and explains 17% of the variance (Adjusted R2 = .171). The partial correlation between belief-inhibition and need for cognition was the same when non-conflict score was controlled.

Table 2: Correlations among Correct Conflict and Non-conflict Syllogisms and Need for Cognition

Belief-Inhibition Score

Non-Conflict Score Need for Cognition Score

Belief-Inhibition Score

- .01 .03

Non-Conflict Score .01 - .42*Need for Cognition Score

.42* .03 -

* Significant at the .0005% level ¹Numbers above the diagonal are zero order 2Numbers below the diagonal are partial correlations

In summary it was found that belief-inhibition shows a moderate positive correlation with need for cognition. The partial

correlation when non-conflict score was controlled for was the same as when it was not controlled for and remained

significant, so we can assume that the correlation is due to belief bias effect and not ability at syllogistic reasoning. Non-

conflict score did not correlate significantly with either belief-inhibition score or need for cognition. See appendix 4 for

SPSS readout.

1.4. Discussion

The results for this part of the study showed that people who score more highly in the need for cognition are better at

inhibiting their prior beliefs when these conflict with logic, added to this is the fact that this correlation remained the

same when non-conflict score was controlled for. These results accord with the findings of Macpherson and Stanovich

(2007) and strengthen the dual process interpretation that can be taken from them, as the correlation between belief-

inhibition and need for cognition was this time found to be entirely independent of general ability on syllogistic

reasoning.

Collectively these results allow for the interpretation that people scoring highly in need for cognition are more inclined

to use an analytic system of processing than those low in need for cognition and are thus more able to avoid belief bias.

However in saying this it must be noted that the effect size was only moderate, need for cognition explaining 17% of the

variance in belief-inhibition score thus there are more important factors at play here. One of these factors may be

working memory, which will be covered by the next section of this report.

Interestingly it was also found that the correlation between conflict and non-conflict scores was almost zero (and

insignificant), presumably this is an indication that there is a strong belief bias effect occurring on the conflict syllogisms

which is entirely independent of general syllogistic ability. However, previous studies have found that belief-inhibition is

positively correlated with general cognitive ability (e.g. West and Stanovich, 1998; Stanovich, West & Toplak, 2008); it

seems likely that this correlation would increase with general ability at syllogistic reasoning due to the greater task

similarity, rather than decrease as the results have found. This finding will be discussed in greater detail in the general

discussion.

Feedback from participants from the need for cognition part of the study was fairly few and far between, participants

were far more interested in their results to the syllogisms component (which will be discussed in the general

discussion). One feedback however was that several participants showed signs of attempting to guess the results that

the experimenter was looking for, one participant stated “I’m probably going to be the one who likes thinking but is

really bad at it!” (see appendix 5) this appears to be fairly explicit participants bias. The fact that the participants were

in general more educated than the general population would likely have amplified this participant bias effect. It was

considered on the design of the experiment that participants may perform differently on the syllogisms if given the

need for cognition scale to complete first. 18 questions about the degree to which one enjoys cognitive effort may cue

one quite blatantly to use the analytic system of processing. This is the reason that participants were given the

syllogisms to complete first and the need for cognition scale second. However it seems that this may have biased the

results also. Participants anticipating their degree of success or failure at the syllogisms may have modified their

answers to the need for cognition scale accordingly. For example if a participant felt that they did badly on the

syllogisms they may have answered the need for cognition scale in a way suggesting they do not enjoy cognitive effort.

Allowing them to feel more in control of their result i.e. ‘it was because I don’t like thinking, not because I am bad at it’

thus protecting their own self-perceptions. It is difficult to know whether this affected the results or not. This could

possibly have been controlled for by employing each part of the test at different times, but this would have been

extremely time-consuming for both researcher and participant.

In conclusion, this part of the study found that individuals varying in need for cognition also varied in their ability to

avoid the belief bias effect. This can be explained by dual processing theory. Those who take greater pleasure in

cognitive effort are more likely to use the analytic system of processing over the heuristic system and therefore are

more able to avoid belief bias. The next section looks at the relationship between working memory capacity and belief-

inhibition.

2. Working Memory Capacity and Belief Bias

The concept of working memory has a strong theoretical relationship with the concept of analytic processing; working

memory can be seen as the mental workspace that the analytic system of processing requires in order to perform its

effortful, reflective thought. The heuristic system should require little, if any, conscious mental workspace, as it has no

requirement to mentally manipulate information. It follows therefore that having a large working memory capacity

would improve one’s ability to use the analytic system of processing whilst having little effect on one’s ability to use the

heuristic system. People with a larger working memory capacity would therefore be less likely to make belief bias

errors.

Kyllonen and Christal (1990) make the argument that logical reasoning (analytic processing) and working memory are

basically the same thing, in their study they found that measures of one strongly predicted the other. However this is

not supported by findings from fMRI studies that have mapped working memory to different locations in the brain than

those mapped for the analytic processing used in syllogistic reasoning (working memory; Courtney, Petit, Maisog,

Ungerleider, & Haxby, 1998; Landro, et al., 2001); analytic processing: Goel, 2003; 2007).

Syllogistic reasoning ability, however, has been strongly linked to working memory capacity, for example Copeland and

Radvansky (2004) and Capon, Handley and Dennis (2003), both of which found that people with larger working memory

capacities performed considerably better on syllogistic reasoning, although these studies do not relate to the belief bias

effect or dual-processing as they did not compare conflict and non-conflict syllogisms.

The first study to relate working memory directly with belief bias was undertaken by Guayle & Ball (2000). They found

that there was an interaction between the tendency to make belief bias errors and spatial working memory span in

people with lower than average working memory capacities. However this effect was not present in people with an

above average working memory capacity. Although thier study was generally more concerned with task aspects of

belief bias (the amount of load different task manipulations put on working memory) rather than with individual

differences in working memory capacity.

De Neys (2006a) designed an experiment to determine the relationship between belief bias on syllogistic reasoning and

working memory capacity. He was interested in whether better performance on syllogistic reasoning by people with

larger working memory spans, was due to a qualitative difference; people with larger working memory spans are more

likely to use analytics over heuristics, or due to a quantitative difference; the tendency to use analytics is the same in

people with larger working memory spans but their analytic system is better at getting the answer right. This was done

by ‘loading’ the working memory span with an unrelated task while the participant completed the syllogisms. He found

that on non-conflict syllogisms extra cognitive load did not affect the number of errors made, but that on the conflict

syllogisms the number of errors increased significantly with extra cognitive load. Under cognitive load participants were

forced to use their heuristic mode of processing as there was no ‘room’ in working memory for analytic processing to

take place; this did not affect answers to the non-conflict items as either mode of processing should come to the same

conclusion but on the conflict items use of heuristics caused belief bias. This supports not only the dual process theory

itself but the idea that working memory capacity and analytic processing are highly related.

An interesting aside to this study was a later finding by De Neys and Schaeken (2007) that when the term ‘some’ was

used in syllogism (rather than ‘all’ or ‘none’) people were more likely to interpret it logically (rather than pragmatically

as in common language) under greater cognitive load, showing that increased cognitive load does not always inhibit

logical thinking. How this fits into the dual-process theory is unclear.

A similar relationship between working memory capacity and belief bias to that which was found by De Neys (2006) was

found to be present in children (Handley, Capon, Beveridge, Dennis, & Evans, 2004). However in contrast to this are

findings from Morsanyi & Handley (2008) that children with greater cognitive capacity (including a measure of working

memory) were more likely to respond heuristically to problems than were children with lower cognitive capacity.

This part of the current study aims to look at the relationship between working memory capacity and the belief bias for

three reasons; Firstly to determine whether the previous research can be replicated which would add validity to the

findings of this study. Secondly, to compare these results to participants’ general ability at syllogistic reasoning, this is

an original part of the study. Thirdly, to look at how the correlation (if one is found) between working memory capacity

and belief bias relates to the correlation between need for cognition and belief bias to determine whether they have a

common or separate cause.

2.2. Methods

2.2.1. Design

A correlational design was used; with the variables: Working Memory Capacity Score on an online operational-span

(OSPAN) test (percentage of maximum possible score), Belief-Inhibition Score on a syllogistic reasoning task

(percentage of correct answers to syllogisms where logic conflicts with belief), and the control variable Non-Conflict

Score on a syllogistic reasoning task (percentage of correct answers to syllogisms where logic and belief concur). The

syllogistic reasoning task was administered with paper and pencil, the OSPAN task was administered on a variety of PC’s

and Macs belonging to either the participant, their workplace or in one of the University of Westminster computer

rooms. The entire test takes about 45 minutes to complete.

2.2.2. Participants

83 of the 115 participants described in part 1 also completed this part of the study. They were 39 (47%) male and 44

(53%) female. The oldest participant was 67 and the youngest was 18 (m = 37, sd = 14). Educational level and

background was fairly representative of the total cohort participating in part 1.

2.2.3. Materials

The Short Form of the Need For Cognition Scale was used as described in part 1.

Working Memory Capacity Test: An online version of the Operation Span (OSPAN) test (Turner & Engle, 1989) was used,

obtained from the Hanover College website (Krantz, 2008) which is freely available through The Higher Education

Academy’s psychology website (The Higher Education Academy, 2009). Each item of the test was composed of a

random word presented in conjunction with a maths problem, participants were instructed to simultaneously remember

the word and indicate whether the answer given for the maths problem was correct or incorrect. After all items had

been presented in this way participants were given a list of words and asked to click on the words which they

remembered form the test. The test was composed of 9 items, participants were given 15 seconds to look at each item

and 30 seconds for recall at the end. The test was scored by taking the mean of the maths score and word recall part of

the study for each participant, which were both given as percentages.

2.2.4. Procedure

As described in part 1 with the difference that participants also completed an online OSPAN test after the other parts of

the test. Some participants completed this part in the same session as the other parts of the test and others completed

it at a different time. Due to the complexity of setting up the online OSPAN part of the study almost all participants were

tested face to face with the experimenter in front of a computer screen. Less than 15% completed the OSPAN test away

from the researcher; these participants were given one to one instructions given over the phone directly before they

took the test and were debriefed and the results collected over the phone directly after they took the test. This is also

the reason that not all of the participants took part in this test.

2.3. Results

Table 4 shows the descriptive statistics for participants who completed all parts of the study.

Table 4: Descriptive Statistics for Belief-Inhibition and WMC for Participants Completing all Parts of the Study

Number Mean Std. Deviation

Belief-Inhibition Score

83 .64 .23

Non-Conflict Score 83 .86 .14

Working Memory Capacity

83 .57 .23

Sequential linear regression was done with working memory as the dependant variable, non-conflict score was added

first to the model then belief-inhibition score was added second. A partial correlation was also done controlling for non-

conflict score. The results of these analyses can be found in table 5 below

Table 5: Correlations among Belief-Inhibition and Non-conflict Syllogisms and Working Memory Capacity

Belief-Inhibition Score

Non-Conflict Score Working Memory Capacity

Belief-Inhibition Score

- .07 .39**

Non-Conflict Score .07 - .35*Working Memory Capacity

.37** .33* -

*Significant at the 0.001% level ** Significant at the .0005% level ¹Numbers above the diagonal are zero order 2Numbers below the diagonal are partial correlations

A moderate positive correlation was found between Belief-Inhibition and Working Memory Capacity (r = .39) which was

highly significant (p = < .0005, one tailed), a moderate positive correlation remained after non-conflict score was

controlled for (r = .37, p = > .0005) adjusted r2 = .14, so 14% of the variation in Belief-Inhibition can be explained by

Working Memory Capacity. A significant, moderate positive correlation was also found between working memory

capacity and non-conflict score (r = .35, p = > .001, partial r = .33, p = > .001). The adjusted r2 = .12 so 12% of the

variance in general ability at syllogistic reasoning can be explained by working memory.

So to summarise, both belief-inhibition and general ability at syllogistic reasoning correlated moderately with working

memory capacity. As before, it was found in this smaller sample that general ability at syllogistic reasoning did not

correlate significantly with ability to inhibit belief in syllogistic reasoning.

2.4. Discussion

It was found in this part of the study that people with greater working memory capacity made fewer errors on the belief

bias syllogisms, and made fewer errors on the non-conflict syllogisms; working memory accounting for 14% of variation

in belief-inhibition score and 12% of variation in non-conflict score. This agrees with the findings of Guayle & Ball (2000)

and De Neys (2006a) discussed in the introduction.

There are three possible interpretations of these results, firstly that people will make a choice (consciously or

unconsciously) for each syllogism to use one or the other system of processing; possessing a greater working memory

capacity leads people to tend more toward analytic processing and thus make less belief bias mistakes. This theory is

supported by the fact that score on conflict syllogisms correlated more highly to working memory capacity than did

score on non-conflict syllogisms.

The second interpretation is that people always use the analytic system when reasoning about syllogisms; people with a

greater working memory capacity do so better as they have more resources at their disposal, belief bias having nothing

to do with it. This conclusion is supported by the findings that people with greater working memory spans also

performed better on non-conflict syllogisms than those with smaller working memory spans. If it was the case that

people were relying only the heuristic systems for non-conflict syllogisms then working memory span should not be

related.

The final interpretation of these results is that people with greater working memory capacity have the same tendency to

use either heuristic or analytic processing but they differ from people with smaller working memory capacity in that

both their heuristic and analytic systems are better. This third interpretation goes against what has been theorised

before, that working memory and analytic processing are very similar, if not the same thing (Kyllonen and Christal,

1990) and that heuristic processing does not require use of the working memory. However some theorists believe that

the working memory controls the output from both analytic and heuristic processing, and another brand of this theory

that analytic processing controls the output of heuristic processing (Evans, 2007). Proponents of this theory maintain

that analytic processing makes greater demands on working memory than heuristics, which explains the closer

correlation of working memory capacity with conflict syllogisms than non-conflict syllogisms. This idea that people with

high working memory capacity have better heuristic processes as well as analytic is supported by a finding of De Neys,

Schaeken, & d'Ydewwalle (2005) that people with greater working memory capacity were better able to choose

between heuristics, inhibiting those that were inappropriate, and use one that was more normatively correct.

For future research it would be interesting to see a study which looked deeper into the way heuristic processing uses

working memory and how this fits into the dual process theory in general. It may be wise to do this without the use of

syllogisms as there is the possibility that such a formal reasoning task automatically cues activation of the analytic

processing (which will be dealt with in more detail in the general discussion). A possible design for this future study

would be to ask participants to perform a task which one could be certain required heuristic processing (would be

performed less well by analytic processing) whilst variably ‘loading’ the working memory with an unrelated task. The

loading task used could be the same as that used by De Neys, (2006) and Schaeken (2007). In this way it could be

measured to what extent heuristic processing requires use of the working memory.

It is unfortunate that not all of the participants that took part in the written tests could take part in the online OSPAN

part. The reason for this was that it required both the participant and researcher to be together sitting in front of a

computer2 which was not possible in all cases. The OSPAN test was fairly complicated to set up so participants were not

able to do it themselves at home.

To conclude, this part of the study found that people with larger working memory capacity made fewer errors on both

the conflict and non-conflict syllogisms. There are several possible explanations for this. The explanation that best fits

the results is that people varying in working memory capacity have the same tendency to use either analytic or

heuristic processing but that they differ in the effectiveness of both their heuristic and analytic systems. The next

section looks at how this relates to the need for cognition element studied in section 1.

3. Working Memory Capacity, Need for Cognition and Belief Bias

The previous two sections of this study have found that both need for cognition and working memory capacity are

predictors of belief-inhibition. What is missing is how the two relate to one another in the way that they predict belief-

inhibition.

Previous studies have looked at the associations between need for cognition and belief bias in conjunction with

cognitive ability, using measures such as SAT scores (Stanovich & West, 2008; West, Toplak, & Stanovich, 2008;

Stanovich & West, 1998). Other studies have also looked at the relationship between belief bias and working memory

2 A few participants (N < 15%) who completed the OSPAN test did so over the phone, these participants received detailed instruction on how to set up the test over the phone as they were doing it and their results were collected over the phone directly afterwards.

capacity (Guayle & Ball, 2000; De Neys, 2006). However no study has so far looked at the interaction between working

memory capacity and need for cognition in thier relationship with the belief bias effect. This means that it is still unclear

if the two relationships have the same underlying cause (such as cognitive ability) or if they are different in origin.

This part of the current research will look at the relationship between these three things in the same group of

participants.

3.2. Methods

3.2.1. Design

A correlational design was used; with the variables: Working Memory Capacity Score on an online operational-span

(OSPAN) test, (percentage of maximum possible score), Need for Cognition Score and Belief-Inhibition Score on a

syllogistic reasoning task (percentage of correct answers to syllogisms where logic conflicts with belief). The need for

cognition scale and syllogistic reasoning task were administered with paper and pencil, the OSPAN task was

administered on a variety of PC’s and Macs. The entire test takes about 45 minutes to complete. Participants and

procedure are the same as that used in parts 1 and 2 above. Materials relevant to this part of the study were the Need

for cognition scale, syllogistic reasoning task described in 1.1.3. and the Online OSPAN test described in 2.1.3.

3.3. Results

The results of the WMC test were; N = 83, M = .565, SD = .225, and of the Need for Cognition test (for participants who

also completed the WMC test) were; N = 83, M = 44.43, SD = 10.00. This is outlined in the table 3.

Table 3: Descriptive Statistics for 83 Participants who completed all parts of the test

Number Mean Std. Deviation

Need for Cognition 83 44.43 10.00

Belief-Inhibition Score

83 .64 .23

Working Memory Capacity

83 .57 .23

Pearson’s test of correlation found that there was no significant correlation between need for cognition and working

memory capacity. So participants need for cognition is not related to their working memory capacity. A simultaneous

linear regression analysis was performed with belief-inhibition as the dependant variable and need for cognition and

working memory capacity entered together. A significant model emerged which accounted for 30% (Adjusted r 2 = .30)

of the variance in belief-inhibition (r = .561, F = 18.376, p = > .0005). Please see appendix 7 for SPSS readouts. A

second simultaneous linear regression was done with participants highest level of education added to the model;

although this was found to be a significant predictor it added very little to the predictive value of the model and so was

dropped from the results.

Table 7: Correlations among Belief-Inhibition and Non-conflict Syllogisms and Working Memory Capacity

Belief-Inhibition Score

Need for Cognition Working Memory Capacity

Belief-Inhibition Score

- .46** .39*

Need for Cognition .44** - .17Working Memory Capacity

.36* -.01 -

*Significant at the 0.001% level ** Significant at the .0005% level ¹Numbers above the diagonal are zero order 2Numbers below the diagonal are partial correlations

To summarise, it was found that participants’ working memory capacity and their need for cognition both predicted the

extent to which they could avoid belief bias in syllogistic reasoning. Furthermore, these factors predicted belief-

inhibition independently of one another and had an additive effect on the predictive value of the model.

3.4. Discussion

The results show that even though both working memory capacity and need for cognition correlate with belief-inhibition

they do so independently of each other. When these correlations are consigned together to a model, the extent to which

each predicts belief bias is additive. This is strong evidence that the two factors have separate underlying causes. The

implications of this are discussed in the general discussion.

4. Awareness of Conflict in Belief-Bias

When logic conflicts with prior belief, in order to avoid belief bias an individual must inhibit the heuristic system and

operationalise the analytic system. For this to transpire the individual must first be aware that such a conflict exists.

Much speculation exists over whether people consciously experience this conflict and are able to choose between the

two systems or whether the decision to use one or the other system happens automatically.

Kahneman and Fredrick (2005) aver that in general people rely on their less demanding heuristic system to make

judgements, in certain rare cases they will detect a conflict between belief and logic and call up the more demanding

analytic processing but much more often will be unaware of the conflict and have very little insight into how their

judgement came about or that it is not normatively correct.

Other researchers take a different view (e.g. Denesraj & Epstien, 1994; Sloman, 1996), that when conflict between logic

and intuition takes place both heuristic and analytic systems are simultaneously activated and the subject is aware of

the conflict between the two. The person then must make a conscious decision as to which route to take, when they

take the wrong route for that situation, choosing heuristics when it would be more appropriate to use analytics then

they make belief bias type errors. This view is interesting in that it implies a method of controlling belief bias where the

other view does not. If people are consciously aware that a conflict is taking place and that there are two possible

‘routes to follow’ it may be possible to train or instruct the person to avoid these errors. However, as in the previous

view, if this conflict is not consciously detected then training would likely not be effective.

This part of the current research is original in that previous attempts at determining peoples’ awareness of belief-logic

conflict have been explicit in asking people to explain their thought processes (De Neys & Glumicic 2008), or casual in

taking note of comments made by participants whilst reasoning (Denesraj & Epstien, 1994; Epstein & Pacini, 1999) in all

cases qualitative data. In this study the participants are simply asked to indicate how certain they are of their answer to

each syllogism. If participants are aware that there is a conflict occurring it follows that they would be less certain of

their answers to the conflict syllogisms than to the non-conflict syllogisms. This will add a fresh perspective from

quantitative data.

4.2. Methods

4.2.1. Design

A within-groups design was used; with the variables: Degree of certainty on non-conflict syllogisms (percentage times

participant was certain of their answer to a non-conflict syllogism) and Degree of Certainty on Conflict syllogisms

(percentage of times participant was certain of their answer to a Conflict syllogism). The task was administered with

paper and pencil and takes about 20 minutes to complete. Participants and procedure are those describes in section 1

above.

4.2.2. Materials

The syllogistic reasoning task as described in part 1.1.3. The part of the syllogistic reasoning task which corresponds to

this research question is participants’ self reports of certainty in their answer. A certainty score is given as a percentage

of number of times participant answered ‘certain’ to the conflict and non-conflict syllogisms.

4.3. Results

Overall results for participants’ degree of certainty are reported in table 5 below.

Table 5: Descriptive Statistics for Degree of Certainty

Number

Mean Std. Deviation

Degree of Certainty on Conflict Syllogisms

115 .73 .19

Degree of Certainty on Non-Conflict Syllogisms

115 .80 .17

Participants were more certain of their answers to the non-conflict syllogism that they were of their answers to the

conflict syllogisms, the mean difference was -.07. The effect size was small (d = .10) but was found to be significant

using a paired t-test (t = 4.659, df = 114, p = <.00025, 1-tailed). No significant relationship was found between

participants’ level of certainty and the number of correct answers they gave for either category of item.

Taking a cue from the previous results, it was also decided to test whether working memory capacity or need for

cognition predicted ability to detect belief-logic conflict. No significant correlations were found.

4.4. Discussion

From participants self reported degree of certainty we can infer that there was some degree of awareness that a conflict

was occurring between belief and logic on the belief-inhibition syllogisms, as participants were less certain of their

results to these items. However since the effect size is so small (d = .10), and the mean percentage of certainty for

belief-inhibition syllogisms was so high (73%) this is not a strong enough effect to draw any firm conclusions.

The level of certainty was not significantly related to either working memory capacity or need for cognition which

suggests that differences in performance on the syllogisms did not relate to ability to consciously detect the conflict.

However a criticism of this part of the study is that participants self reports of certainty do not take into account

whether they actually came up with the correct solution to the syllogisms they were certain or uncertain of and in fact

participants score and level of certainty. Because of this there is a limit to the interpretations that can be made from

this part of the study.

General Discussion

The current research has found that the ability to inhibit ones belief when it conflicts with logic is weakly positively

correlated with working memory capacity, and moderately positively correlated with need for cognition. However

despite this it has been found that working memory capacity is not significantly correlated with need for cognition.

When working memory capacity and need for cognition are consigned together to a model, the extent to which each

predicts belief-inhibition is additive. There is no overlap of the prediction that each one adds. This is strong evidence

that the two factors have separate underlying causes.

These results when viewed through a dual-process lens signify that those with a larger working memory capacity and

those with a greater need for cognition are both more likely to use analytic processing but that they do so for different

reasons.

One interpretation of these results is that high need for cognition people are more likely to use analytic processing as

they take greater enjoyment from thought, even though they may find it just as effortful as low need for cognition

people. People high in working memory capacity are more likely to use analytic processing however as they are more

able to do so, it requires less effort for them to consciously manipulate the variables than it does for people with lower

working memory capacity so they have less need to rely on heuristic processing.

However this interpretation does not explain the fact that both belief-inhibition and general ability at syllogistic

reasoning correlated with working memory independently of each other. Had they correlated together then it could

have been concluded that larger working memory capacity allowed people to be better at syllogistic reasoning as well

as causing them to be more inclined to use analytic processing. The fact that they both correlate independently of each

other suggests that something else is going on.

This can be explained however. It is possible that people with larger working memory spans are better at both analytic

and heuristic processing and thus make fewer mistakes on whichever one they use; however whether they choose to

use one or the other for any given task depends on something other than WMC. People with a greater need for cognition

do not have better analytic systems than those who do not, however they are more likely to choose to use it when it is

required (and perhaps more likely to realise that it is required). Thus it could be predicted that the individual who would

be best at inhibiting their prior beliefs when faced with conflicting logic would be those with high working memory

capacity and a greater need for cognition as they have a combination of motivation and ability.

In relation to this interpretation an interesting thing to note about results to the conflict and non-conflict syllogisms is

the almost complete absence of correlation between conflict and non-conflict score. This can be taken as evidence that

the syllogisms were well designed to study the belief bias effect without it being related to cognitive ability or general

ability at syllogistic reasoning, however several previous studies have found that ability to avoid the belief bias effect is

somewhat correlated with general cognitive ability (e.g. West and Stanovich, 1998; Stanovich, West & Toplak, 2008).

One would expect this correlation to increase with the greater task similarity, rather than the decrease that has been

found in the current research.

It is perfectly possible that this finding indicates a methodological flaw in the design of the syllogisms. However there is

an interpretation of this finding that does accord with previous research. That is that each participant chose either the

analytic or heuristic system of processing and stuck to it for the entirety of the syllogistic reasoning task. Participants

using analytic processing on the non-conflict items may, due to the difficulty of the syllogisms have made more

mistakes than those using heuristics as they were looking out for ‘trick’ questions. The general assumption in dual-

processing is that analytic processing is better and more likely to come up with an answer that is normatively correct

than heuristic processing simply by virtue of the fact that it takes more effort (e.g. Alter et al., 2007; Stevenson, 1997;

Evans in Holyoak and Morrison, 2005). However this may not be the case, several studies have found heuristic

processing to be ‘better’ that analytic processing, not only because it demands less effort, but also because in many

instances it is more liable to come up with an answer that is normatively correct (Witteman, van den Bercken, Claes, &

Godoy, 2009; De Vries, Holland, & Witteman, 2008; Engel in Engel and Wolf, 2008; Wilson & Schooler, 1991; also see

Wilson, 2002). One such study by Pretz (2008) found that more experienced people did better using the analytic system

whilst less experienced people were better to rely on heuristics. This allows interpretation of the results of the current

study; people who have a larger working memory capacity perform better using analytic processing whilst people who

have a smaller working memory capacity are better to stick to heuristic processing.

One very common feedback from participants (in fact most participants reported this) was that the syllogistic reasoning

task was “exam” like, many participants reported strong negative feelings towards this part of the study and many were

anxious that they would perform badly (see appendix 7 for some of the more notable participant feedback). This being

despite all efforts to reassure that they themselves were not being tested only the theories involved. It was made

absolutely clear to participants that they were under no obligation to participate at the beginning of the study and this

was reiterated when participants reported negative feelings towards the syllogistic reasoning component. This may

have been more of an ethical consideration but for the fact that the same participants who reported performance

anxiety also requested to be sent their results (which necessitated that their results not be anonymous to the

researcher). This feedback does however lead to a criticism that can be made about the syllogistic reasoning paradigm

in general; that such a formal ‘exam like’ task may subconsciously cue analytic processing. It has been found in

previous studies that participants are very susceptible to subtle cues in their choice of processing mode, directing

participants to furrow their brow, for example, has been found to propel participants towards using analytic processing

in syllogistic reasoning (Alter, Oppenheimer, Epley, & Eyre, 2007). This is compounded by the fact that in general

everyday life people are not exposed to such formal reasoning ‘puzzles’; exposure to a novel stimulus may also cue

analytic processing.

This criticism of the syllogistic paradigm immediately suggests designs for future study. Many determinants of

differential use of one or the other system of processing have been studied, but they can be coded into two main

categories; individual differences (that this research has focused upon) such as cognitive ability (Stanovich & West,

2008; Gilinsky & Judd, 1994), thinking disposition and personality (e.g. Stanovich & West, 1997), age (De Neys & Van

Gelder, 2009) etc.; and task differences such as figural effects (Stupple & Ball, 2007), pragmatic or logical use of the

term ‘some’ (DeNeys & Schaeken, 2007), the type of instruction given to participants (Macpherson & Stanovich, 2007;

Evans, Handley, & Harper, 2001) etc. What has not been studied3 however are possible situational effects, such as the

formality of the setting; participants perhaps could be cued to use the analytic system over the heuristic system in an

exam like environment or a formal interview environment and this could be compared with processing system use in a

less formal environment, such as at home or in a public house. This would have implications for the use of aptitude tests

in (the formal setting of) job interviews which are becoming more commonplace.

Another interesting area of study in the dual process arena would be the influence of temporary affect at the time of

testing on the system of processing used. A person in a better mood may be more likely to use the less cognitively

demanding heuristic process in order not to ‘spoil’ thier good mood. The same effect might be found for a person who

was stressed as they may be less inclined to put themselves under more cognitive load by using the analytic system

and therefore opt for the heuristic system.

3 Accorded by a literature search on

The results of this study have possible applications for education. If children can be taught that thinking and ‘puzzling

things out’ is enjoyable rather than a chore it may help them in later life to overcome thinking biases and improve their

critical thinking skills even if their ability level is not high.

Conclusion

The current research found that both need for cognition and working memory correlated positively with the ability to

avoid belief bias errors and that they do so independently of one another. This is explained in terms of dual-processing

theory that those with greater need for cognition are more inclined to use the analytic processing as they take greater

enjoyment from it, wheras those with a larger working memory capacity are as likely to use either analytic or heuristic

processing as those with a smaller working memory capacity but have more effective analytic and heuristic systems

and are thus more likely to come up with an answer which is more normatively correct. These findings infer that the

heuristic process requires use of the working memory capacity although to a lesser extent than the heuristic process.

Indications for future research are a more in depth look into the ways in which the heuristic system makes use of

working memory and the impact of situational and temporary affect variables on the differential use of one or the other

processing system. Applications of the findings are suggested to education where children could be taught to avoid

thinking biases by being encouraged to see thinking as enjoyable rather than taxing.

Works CitedAlter, A. L., Oppenheimer, D. M., Epley, N., & Eyre, R. N. (2007). Overcoming intuition: Metacognitive difficulty activates analytic reasoning. Journal of Experimental Psychology: General , 569-576.

Bruine de Bruin, W., Fischhoff, B., & Parker, A. M. (2007). Individual differences in adult descision making competence. Journal of Persoanlity and Social Psychology , 938-956.

Cacioppo, J. T., & Petty, R. E. (1982). The need for cognition. Journal of Personality and Social Psychology , 116-131.

Cacioppo, J. T., Petty, R. E., & Feng Kao, C. (1984). The efficient assessment of need for cognition. The Journal of Personality Assessment , 306-307.

Cacioppo, J. T., Petty, R. E., Feinstein, J. A., & Jarvis, W. B. (1996). Dispositional differences in cognitive motivation: The life and times of individuals varying in need for cognition. Psychological Bulletin , 197-253.

Capon, A., Handley, S., & Dennis, I. (2003). Working memory and reasoning: An individual differences perspective. Thinking and Reasoning , 203-244.

Cavazos, J. T., & Campbell, N. J. (2008). Cognitive style revisited: The structure X cognition interaction. Personality and Individual Differences , 498-502.

Clifford, J. S., Boufal, M. M., & Kurtz, J. E. (2004). Personality traits and critical thinking: Skills in college students empirical tests of a two-factor theory. Assessment , 169-176.

Cohen, A. R., Stotland, A., & Wolfe, D. M. (1955). An experimental investigation of the need for cognition. Journal of Abnormal Psychology , 291-294.

Copeland, D. E., & Radvansky, G. A. (2004). Working memory and syllogistic reasoning. The Quarterly Journal of Experimental Psychology , 1437-1457.

Courtney, S. M., Petit, L., Maisog, J. M., Ungerleider, L. G., & Haxby, J. V. (1998). An area specialized for spatial working memory in human frontal cortex. Science , 1347-1351.

De Neys, W. (2006b). Automatic-heuristic and executive-analytic processing during reasoning: Chronometric and dual-task considerations. The Quarterly Journal of Experimental Psychology , 1070-1100.

De Neys, W. (2006a). Dual Processsing in Reasoning: Two systems but one reasoner. Psychological Science , 428-433.

De Neys, W., & Glumicic, T. (2008). Conflict monitoring in dual process theories of thinking. Cognition , 1248-1299.

De Neys, W., & Schaeken, W. (2007). When people are more logical under more cognitive load: Dual task impact on scalar implicature. Experimental Psychology , 128-133.

De Neys, W., & Van Gelder, E. (2009). Logic and belief across the lifespan: the rise and fall of belief inhibition during syllogistic reasoning. Developmental Science , 123-130.

De Neys, W., Schaeken, W., & d'Ydewwalle, G. (2005). Working memory and everyday conditional reasoning: Retrieval and inhibition of stored counterexamples. Thinking and reasoning , 449-381.

De Vries, M., Holland, R., & Witteman, C. L. (2008). Fitting decisions: Diffuse affect and intuitive versus deliberative decision strategies. Cognition and Emotion , 931-943.

DeCaro, M. S., Thomas, R. D., & Beilock, S. L. (2008). Individual differences in category learning: Sometimes less working memory capacity is better than more. Cognition , 284-294.

Denesraj, V., & Epstein, S. (1994). Conflict between intuitive and rational processing - when people behave against thier better judgement. Journal of Personality and Social Psychology , 819-829.

DeWall, C. N., Baumeister, R. F., & Masicampo, E. J. (2008). Evidence that logical reasoning depends on conscious processing. Consciousness and Cognition , 628-645.

Dijksterhuis, A., Bos, M. W., Nordgren, L. F., & Von Baaren, R. B. (2006). On making the right choice: The deliberation without attention effect. Science , 1005-1007.

Engel, C. (2008). Institutions for intuitive man. In C. Engel, & W. Singer, Better than consciousness? Decision making, the human mind, and applications for institutions (pp. 391-411). Cambridge, MA: MIT Press.

Engle, R. (2000, October 4). What is working Memory Capacity? Retrieved March 25, 2009, from Georgia Institute of Technology: www.allgpsy.uzh.ch/graduate/mat/Seminar00/E1.pdf

Epstein, S. (1994). Integration of the cognitive and the psychodynamic unconsciousness. American Psychologist , 709-724.

Epstein, S., & Pacini, R. (1999). Some basic issues regarding dual-process theories from the perspective of cognitive-experiential self-theory. In S. Chaiken, & Y. Trope, Dual-Process theories in social psychology (pp. 462-482). New York: Guilford Press.

Evans, J. S. (2005). Deductive Reasoning. In K. J. Holyoak, & R. G. Morrisson, The Cambridge Handbook of Thinking and Reasoning (pp. 169-184). New York: Cambridge University Press.

Evans, J. S. (2007a). On the resolution of conflict in dual process theories of reasoning. Thinking and Reasoning , 321-339.

Evans, J. S. (2007b). The heuristi-analytic theory of reasoning: Extension and evaluation. Psychonomic Bulletin and Review , 378-395.

Evans, J. S., Barston, J. L., & Pollard, P. (1983). On the conflict between logic and belief in syllogistic reasoning. Memory and Cognition , 295-306.

Evans, J. S., Handley, S. J., & Bacon, A. M. (2009). Reasoning under time pressure: A study of causal conditional inference. Experimental Psychology , 77-83.

Evans, J. S., Handley, S. J., & Harper, C. N. (2001). Necessity, possibility and belief: A study of syllogistic reasoning. The quarterly journal of Experimental Psychology A: Human Experimental Psychology , 935-958.

Gilinsky, A. S., & Judd, B. B. (1994). Working memory and bias in reasoning across the lifespan. Psychology and Ageing , 356-371.

Gillard, E., Van Dooren, W., Schaeken, W., & Verschaffel, L. (2009). Proportional reasoning as a heuristic-based process: Time constraint and dual task considerations. Experimental Psychology , 92-99.

Goel, V. (2007). Anatomy of deductive reasoning. TRENDS in Cognitive Sciences , 435-441.

Goel, V. (2003). Evidence for dual neural pathways for syllogistic reasoning. Psychologica , 301-309.

Guayle, J. D., & Ball, L. J. (2000). Working memory, metacognitive uncertainty, and belief bias in syllogistic reasoning. The Quarterly Journal of Experimental Psychology A: Human Experimental Psychology , 1202-1223.

Handley, S. J., Capon, A., Beveridge, M., Dennis, I., & Evans, J. S. (2004). Working memory, inhibitory control and the development of childrens reasoning. Thinking and Reasoning , 175-195.

James, W. (1890). Necessary truths and the effects of experience. In W. James, The Principles of Psychology (pp. 617-689). New York: Henry Holt and Company.

Kahneman, D., & Frederick, S. (2005). A model of heuristic judgement. In K. J. Holyoak, & R. G. Morrison, The Cambridge handbook of thinking and reasoning (pp. 267-293). Cambridge: Cambridge University Press.

Kahnemann, D. (2002, December 8). Kahnemann Lecture. Retrieved March 14, 2009, from Nobel Prize: http://nobelprize.org/nobel_prizes/economics/laureates/2002/kahnemann-lecture.pdf

Kokis, J. V., Macpherson, R., Toplak, M. E., West, R. F., & Stanovich, K. E. (2002). Heuristic and analytic processsing: Age trends and associations with cognitive ability and cognitive styles. Journal of Experimental Child Psychology , 26-52.

Krantz, J. H. (2008, November 8). Instructions for the Working Memory Experiment. Retrieved March 8, 2009, from Cognition Laboratory Experiements: http://psych.hanover.edu/JavaTest/CLE/Cognition/Cognition/workingMemory_instructions.html

Kyllonen, P. C., & Christal, R. E. (1990). Reasoning ability is (little more than) working memory capacity. Intelligence , 389-433.

Landro, N. I., Rund, B. R., Lund, A., Sundet, K., Mjellem, N., Asbjornsen, A., et al. (2001). Honig's model of working memory and brain activation: An fMRI study. Neuroreport , 4047-4054.

Macpherson, R., & Stanovich, K. E. (2007). Cognitive ability, thinking dispositions, and instructional set as predictors of critical thinking. Learning and Individual Differences , 115-127.

Markovits, H., & Nantel, G. (1989). The belief bias effect in the production and evaluation of logical conclusions. Memory and Cognition , 11-17.

Morsanyi, K., & Handley, S. J. (2008). How smart do you have to be to get it wrong? The role of cognitive capacity in the development of heuristic-based judgement. Journal of Experimental Child Psychology , 18-36.

Newell, A., & Simon, H. A. (1972). Human Problem Solving. Oxford, UK: Prentice-Hall.

Njus, D., & Johnson, D. R. (2008). Need for cognition as a predictor of psychosocial identity development. The Journal of Psychology , 645-655.

Osman, M. (2004). An evaluation of dual-process theories of reasoning. Psychonomic Bulletin and Review , 988-1010.

Park, H. S., Levine, T. R., Westerman, C. Y., & Foregger, S. (2007). The effects of argument quality and involvment type on attitude formation and attitude change: A test of dual process and social judgement predictions. Human Communication Research , 81-102.

Petty, R. E., & Brinol, P. (2006). Understanding social judgement: Multiple systems and processes. Psychological Inquiry , 217-223.

Petty, R. E., DeMarree, K. G., Brinol, P., Horcajo, J., & Strathman, A. J. (2008). Need for cognition can magnify or attenuate priming effects in social judgement. Personality and Social Psychology Bulletin , 900-912.

Pretz, J. E. (2008). Intuition versus analysis: Strategy and experience in complex everyday problem solving. Memory and Cognition , 554-556.

Reverberi, C., Shallice, T., D'Agostini, S., Skrap, M., & Bonatti, L. L. (2009). Cortical bases of elementary deductive reasoning: Inference, memory, and metadeduction. Neuropsychologia , 1107-1116.

Roberts, M. J., & Sykes, E. D. (2003). Belief bias and relational reasoning. The Quarterly Journal of Experimental Psychology A: Human Experimental Psychology , 131-154.

Rodrigues-Moreno, D., & Hirsch, J. (2009). The dynamics of deductive reasoning: An fMRI investigation. Neuropsychologia , 949-961.

Sa, W. C., West, R. F., & Stanovich, K. E. (1999). The domain specificity and generality of belief bias: Searching for a generalizable critical thinking skill. Journal of Educational Psychology , 497-510.

Sadowski, C. J., & Cogburn, H. E. (1997). Need for cognition in the big-five factor structure. The Journal of Psychology , 307-312.

Simon, H. A., & Newell, A. (1971). Human Problem Solving: The state of the theory in 1970. American Psychologist , 145-159.

Sloman, S. A. (1996). The empirical case for two systems of reasoning. Psychological Bulletin , 3-22.

Stanovich, K. E., & West, R. F. (1998). Individual differences in rational thought. Journal of Experimental Psychology: General , 161-188.

Stanovich, K. E., & West, R. F. (2008). On the relative independance of thinking biases and cognitive ability. Journal of Personality and Social Psychology , 672-695.

Stanovich, K. E., & West, R. F. (1997). Reasoning independently of prior belief and individual differences in actively open minded thinking. Journal of Educational Psychology , 342-357.

Stevenson, R. J. (1997). Deductive reasoning and the distinction between implicit and explicit processes. Current Psychology of Cognition , 222-229.

Stupple, E. J., & Ball, L. J. (2007). Figural effects in a syllogistic evaluation paradigm: An inspection time analysis. Experimental Psychology , 120-127.

The Higher Education Academy. (2009, March 10). psychologypracticals.com :: your one stop shop for practical psychology learning and teaching. Retrieved March 10, 2009, from Psychology Network: http://www.psychologypracticals.com/

Toplak, M. E., & Stanovich, K. E. (2002). The domain specificity and generality of disjunctive reasoning: Searching for a generalizable critical thinking skill. Journal of Educational Psychology , 197-209.

Turner, M. L., & Engle, R. W. (1989). Is working memory capacity task dependent? Journal of Memory and Language , 127-154.

West, R. F., Toplak, M. E., & Stanovich, K. E. (2008). Heuristics and biases as measures of critical thinking: Associations with cognitive ability and thinking dispositions. Journal of Educational Psychology , 930-941.

Wilson, T. D. (2002). Strangers to ourselves; Discovering the adaptive unconscious. Cambridge, MA: Harvard University Press.

Wilson, T. D., & Schooler, J. W. (1991). Thinking too much: Introspection can reduce the quality of preferences and decisions. Journal of Personality and Socail Psychology , 181-192.

Witteman, C., van den Bercken, J., Claes, L., & Godoy, A. (2009). Assassing rational and intuitive thinking styles. European Journal of Psychological Assessment , 1015-5759.

Appendices:

I have removed the majority of the appendices on Scribd as formatting differences were making the tables completely illegible. The research materials I used/made are available separately on my Scribd profile, if you would like to see the raw data or the SPSS readouts please get in touch on [email protected]. I have left the descriptive statistics and the summary of participant feedback below.

Descriptive Statistics for all Results

Descriptive Statistics

N Minimum Maximum Mean Std. Deviation

Gender 115 0 1 .43 .497

Age 115 18 67 35.77 12.779

Highest Educational Level 115 0 4 2.19 .794

Need for Cognition Score 115 16 70 44.63 9.702

Working Memory Capacity

Score83 .10 1.00 .5651 .22547

Overall Correct Syllogisms 115 .40 1.00 .7081 .14943

Non-Conflict Score 115 .50 1.00 .8475 .13263

Belief-Inhibition Score 115 .08 1.00 .6181 .23311

Degree of Certainty for

Belief-Inhibition Items115 .17 1.00 .7258 .19038

Degree of Certainty for Non-

Conflict Items115 .25 1.00 .7983 .17356

Valid N (listwise) 83

Participant Feedback

Feedback was obtained by casual observations of participants as they completed the study and noting down any relevant comments they made, and by asking at the end if they had any comments on the research. It was found that in general participants did not have a great deal of feedback to give, mostly they just wanted to know the correct answers to the syllogisms, which were provided on request.

Examples of comments made by participants:

About the syllogistic reasoning component

“Aaah (sigh) this is really hard!”

“I’ll have to think about this”

“[laughing] I’m just answering these randomly!”

“I’m sorry I guessed most of them”

“I’m really bad at these” – a very common statement

“I feel like I’m in an exam”

About the need for cognition component

“I don’t really know the answer to this bit”

“I’m probably going to be the one who likes thinking but is really bad at it!”

About the working memory capacity component

“I think I did really badly”

“Oh no I didn’t remember anything!”

Examples of questions asked by participants (and answers given by researcher):

“Do you want me to answer if this is true or not?” – about the conclusions to the syllogisms (Researcher: “Please answer whether or not the conclusions follow logically from the premises”)