Looking for Arguments

1

SUBMITTED VERSION – NOT TO BE QUOTED

Looking for Arguments

Hugo Mercier

Philosophy, Politics and Economics Program

University of Pennsylvania

313 Cohen Hall

249 South 36th Street

Philadelphia, PA 19104

[email protected]

http://hugo.mercier.googlepages.com/

267-340-2285

Keywords: Argumentation, Reasoning, Dual-process theories, Relevance, Satisficing.

Looking for arguments

2

Abstract: How do people find arguments while engaged in a discussion? Following an

analogy with visual search, a mechanism that performs this task is described. It is a

metarepresentational device that examines representations in a mostly serial manner until it

finds a good enough argument supporting one’s position. It is argued that the mechanism

described in dual process theories as ‘system 2’, or analytic reasoning fulfills these

requirements. This provides support for the hypothesis that reasoning serves an argumentative

function.


3

How do people find arguments? When engaged in a discussion, in a debate, everyone is able

to find arguments defending her position, and do to so quickly (Shaw, 1996). Even very

young children can perform this feast—something their parents are not always happy about

(see Mercier, submitted, for review). Despite the omnipresence of argumentation in daily life,

we know very little about the psychological mechanisms underpinning this ability. A goal of

this paper is to describe and ground in empirical data a model describing the psychological

mechanisms that are used to find arguments. ‘Argument’ is to be understood here as a

synonym of ‘reason’ or ‘supporting statement’—what we say when we want to convince

someone—and not in its technical, logical meaning. These are the arguments we use when

arguing in daily life.

Several lines of work are relevant to the psychology of argumentation, but none of them has

asked directly the question ‘how do we find arguments?’ First, when people argue, they

reason. If psychologists agree that reasoning is involved in argumentation, standard theories

of reasoning would seem to have little to say on this topic: The word ‘argumentation’ does not

appear in the index of the three books describing the major theories of reasoningi. Moreover,

most tasks in this literature involve participants either evaluating the conclusion of an

argument or trying to determine if a logically valid conclusion follows from some premises.

In neither case do they have to actually find premises for a given conclusion. Despite these

apparent methodological shortcomings—regarding the question at hand—I will argue that the

ability psychologists of reasoning have been studying is the same ability we use to find

arguments, thus rendering their conclusions highly relevant.

A small but growing field of research investigates argumentative abilities more specifically.

With a few exceptions, these experiments usually deal with the understanding, not the

production, of arguments. They have shown that participants can adequately follow the

commitments of different speakers (Rips, 1998), attribute the burden of proof (Bailenson &

Rips, 1996) and that they react appropriately to fallacies of argumentation (Baum, Danovitch,

& Keil, 2007; Corner, Hahn, & Oakfsord, 2006; Hahn & Oaksford, 2007; Hahn, Oaksford, &

Bayindir, 2005; Neuman, 2003; Neuman, Weinstock, & Glasner, 2006; Oaksford & Hahn,

2004; Rips, 2002; Weinstock, Neuman, & Tabak, 2004). Psychologists interested in

argumentation and informal reasoning have recorded the production of arguments by

participants, but there have not developed processing models as those offered in standard


4

psychology of reasoning (e.g., Kuhn, 1991; Perkins, 1985). In social psychology dozens of

experiments have tested the effect of the argument strength on persuasion but, again, this only

speaks to evaluation and not production (see Petty & Wegener, 1998 for review). Linguistics

is another domain suffering from the same asymmetry between the study of understanding

and that of production. But what comparatively little work there is in language production

(e.g., Levelt, 1989; Levelt, Roelofs, & Meyer, 1999) does not speak to the much more specific

issue of argument production. Finding a good argument and knowing how best to express it

are two relatively distinct problems. As I explain later, standard communicative skills are not

sufficient to find good arguments.

Aside from the experimental literature, one can find many models of argumentation (see van

Eemeren, Grootendorst, & Henkemans, 1996, for an introduction). These models are often

concerned with creating typologies of arguments or arguments parts. For instance, Stephen

Toulmin devised the famous distinction between claim, datum, warrant, backing, rebuttal and

qualifier (Toulmin, 1958). These models may be very helpful heuristics in designing

psychological theories but they are not, in and of themselves, psychological theories either of

argument evaluation or argument production.

Given how little research has been dedicated to the topic, starting from scratch may seem to

be the only way to proceed. I will argue otherwise. In the next section the argumentative

theory of reasoning will be briefly introduced. According to this theory, arguing—finding and

evaluating arguments—is the very function of reasoning. If this theory is correct then the

capacity that has been studied and described by psychologists of reasoning is precisely the

capacity required to find arguments. In order to prove this point, the first step will be to

predict what kind of psychological mechanism would be best fitted to the task of finding

arguments. It will then be possible to compare this list of requirements to the actual attributes

of reasoning. I will try to show that there is a very good match between what is required of an

ability that has to find arguments and reasoning as described by current theories. This will

fulfill the two goals of the paper: providing a model of how we find arguments and, while

doing so, supporting the argumentative theory of reasoning.

The argumentative theory of reasoning


5

Many fields of psychology have converged on the idea that the mind can usefully be divided

into two broad categories of mechanisms. Thus one can find a distinction between automatic

and controlled attention (Posner & Snyder, 1975), implicit and explicit memory, learning or

attitudes (Berry & Dienes, 1993; Reber, 1993; Schacter, 1987; Wilson, Lindsey, & Schooler,

2000) and heuristic and systematic evaluation processes (Chaiken, Liberman, & Eagly, 1989;

Gawronski & Bodenhausen, 2006; Petty & Cacioppo, 1986). More recently (though see

Wason & Evans, 1975) these dual process theories have spread to the field of reasoning and

decision making (Evans & Over, 1996; Kahneman, 2003; Kahneman & Frederick, 2002;

Sloman, 1996; Stanovich, 2004). Despite some underlying commonalities in the way the

distinction is drawn, it is not entirely clear that the two categories of psychological processes

have the same extension in these different domains of psychology. Here we will focus on the

distinction used in the field of reasoning and decision making and well represented by Evans’

heuristic-analytic theory (Evans, 2006, 2007). According to this theory, cognitive mechanisms

can be divided into heuristic processes—fast, nearly costless and generally unconscious—and

analytic processes—slow, effortful and generally conscious. When first looking at a choice of

cookies your heuristic system will direct your attention towards the most delicious or the most

colorful looking treats—or simply those you usually buy. Then your analytic system may (or

may not) kick in and urge you to compare the calories you would get from different brands in

order to make a more rational (diet-wise) choice. There is growing evidence both in reasoning

and in decision making supporting the validity of such a distinction (see Evans, 2003, 2008;

Evans & Frankish, 2009, for reviews). A related way to frame this dichotomy is to draw a line

between intuitive inferences (or intuitions) and reflective inferences (or reasoning) (Mercier &

Sperber, 2009). In this model, reasoning is a very specific cognitive mechanism that finds and

evaluates reasons. In this article, ‘reasoning’ will only refer to this sense of reasoning: a

mechanism that bears on reasons. It is to the function of this ability that we now turn.

Most theories of reasoning—in psychology or in philosophy—assume that reasoning fulfils

an overall epistemic and/or practical function: it generates new beliefs, creates knowledge,

and drives us towards better decisions. Pointing out some flaws in this assumption Sperber

has suggested instead that reasoning has an argumentative function (Sperber, 2000, 2001).

Building on standard theories of the evolution of communication, he makes a good case that

reasoning may have evolved in order to convince others and only be convinced when it is

worth it. This theory—the argumentative theory of reasoning—has profound implications for


6

the way reasoning works, when it works well, and the results it should achieve. Drawing from

different literatures—psychology of reasoning and decision making, social psychology,

developmental psychology—it has been possible to find a great wealth of empirical support

for this view of reasoning (Mercier, submitted, in press; Mercier & Landemore, submitted;

Mercier & Sperber, 2009, in press; Sperber & Mercier, in press). For instance, people are

good at arguing but bad in abstract reasoning tasks, even though the later should be much

easier. Reasoning becomes much more efficient in argumentative contexts. The workings of

reasoning display a consistent and robust confirmation bias. And when we reason before

making a decision, reasoning drives us toward a decision for which we can argue but not

necessarily a good one. But for one—the confirmation bias, that will be mentioned later—

these predictions bear more on the contexts that lead to felicitous reasoning and on the effects

of reasoning than on the way reasoning actually works. Making the hypothesis that one of the

main functions of reasoning is to find arguments, it should be possible to draw predictions

regarding the workings of reasoning. The remainder of this article will be dedicated to making

and testing these predictions. I will try to show that they fit in very well with standard

descriptions of the workings of reasoning, descriptions that were arrived at with very different

theoretical assumptions. This fit can thereby be seen as further validation of the argumentative

theory of reasoning.

The task: finding arguments

When trying to find arguments several elements have to be taken into account. The most

important is the message (the conclusion) that we want to communicate. We are trying to find

arguments either because this message has been rejected or because we think it may be if we

try to deliver it on its own. The other important elements are the audience and the context.

Different arguments may be effective for a child or an adult, and different arguments expected

in a pub or a courtroom. Here is a very mundane example in which the need to find arguments

arises:

Simon and Margo want to go to the opera.

Simon says ‘let’s take the Tube’.

Margo answers ‘I’d rather walk’.


7

At this point Margo and Simon may try to convince each other to use their favorite mode of

transportation. The task they face is to find good arguments supporting their respective ideas.

The first problem they encounter is that the space within which they have to search is huge.

Many different beliefs could be used as potential arguments. Beliefs about the weather—is it

too cold outside, or just warm enough? Beliefs about the esthetic qualities of parts of

London—it’s nice to walk along Tottenham Court Road. Beliefs about the location of good

shops—if we walk we could buy food for dinner on the way. Beliefs about previous

interactions—last time I let you choose. Beliefs about public transportation in London—the

central line is out of order. Beliefs about a person—I know you recently broke your ankle,

you shouldn’t walk too much. The list could go on to encompass beliefs that are only dimly

related to the topic at hand, such as beliefs about astronomical phenomena—if we walk we

may have a glimpse of the lunar eclipse that will happen tonight.

The size of the search space would not be a problem if it were structured in such a manner

that the best arguments would simply be the most accessible elements. But this will only

rarely be the case: what makes a good argument is highly contextual and the whole cognitive

system is not geared towards finding good arguments. Beliefs about the weather could be

good arguments for Simon or for Margo depending on their tastes and on what the weather

actually is like. Beliefs about esthetic qualities can be either relevant as arguments for Margo

(if the scenery is beautiful), for Simon (if it’s downright ugly), or simply irrelevant (if it’s

mediocre). Beliefs about good shops could be relevant as arguments for Margo (if the shops

are on the way), for Simon (if they are close to the Tube station), or irrelevant (if they have

nothing to buy). This could go on for all potential arguments. For instance, the first things that

Margo will think of may be ‘if we walk we will arrive on time’, which is a good argument

only if the Tube doesn’t allow that; ‘it’s a bit cold outside’ which would be a

counterargument; ‘I usually walk there’ which is rather irrelevant as an argument, etc. The

order in which the beliefs are examined will only be loosely related to their value as

arguments. What this means is that the task of finding arguments comes down to a filtering of

beliefs until one that is deemed to be a relevant argument is foundii. In order to better

understand how doing such a thing is possible, we can use an analogy with another

mechanism that involves a similar filtering: visual search.


8

Visual search can work in very different manners, depending on the target and the context of

the search (see Li, 2002). It can proceed in a parallel, fast fashion if the target does not share

many attributes with the surrounding objects: if they belong to different and well defined

categories (Duncan & Humphreys, 1989). If you are looking for a tennis ball that was

misplaced in the cutlery drawer, there is no need to scan different objects: you are going to

focus immediately on the ball. On the other hand, if either the other objects are similar to the

target (a tennis ball among other yellow balls of a similar size—or the proverbial needle in a

haystack), or if the other objects do not belong to a well defined category (a tennis ball in the

closet you have been filling with junk for 20 years), then you cannot immediately focus on the

target. Instead, your visual system focuses on different objects, in turn, and decides if they are

the one you are looking for—it uses a serial search. Reasoning, the ability we use to find

arguments, works in a similar manner, but instead of searching through objects, it searches

through representations. Much in the same way as searching through objects implies creating

representations of these objects, searching through representations requires creating

representations of these representations. Thus reasoning is a metarepresentational mechanism:

a psychological device that creates and deals with representations of representations (Sperber,

2000). This is the first important prediction to be tested, in the next section. But the task

reasoning has to perform is in fact even more complicated that finding a needle in a haystack,

because it does not even know what the needle looks like.

When we are looking for an argument we rarely, if ever, know exactly what it is that we are

looking for: Often, there will be no way to tell in advance exactly what will be a good

argument for a given conclusion in a given context. There may be some broad indications,

explored later in the article, but rarely a precise picture as in the case of the tennis ball or the

needle. A better analogy than visual search for a known object is that of a search for an ‘ad-

hoc’ category (Barsalou, 1983). For instance, you just moved in a new place and you have to

drive a nail but you do not have a hammer. You can look for an object that could be used to

drive a nail instead of a hammer. ‘Objects that could be used to drive a nail instead of a

hammer’ is a novel, ad-hoc category that you developed on the spot. Compared to the

category ‘hammers’ this category is very poorly defined. The object should not be too small

or too big. It should not be too fragile. It should not be too expensive. It should be relatively

easy to handle. But apart from that, it can have many shapes, any color, etc. What you do to

find such an object is scan your environment and for each object that attracts your attention


9

see if it matches your list of requirements. The order in which objects are going to be scanned

is influenced by these requirements—very small or very large objects may not even be

considered for instance—but it will mostly be directed by factors largely irrelevant to the task

at hand. Attributes such as color or proximity will drive the order in which the search is

conducted. Likewise, the order in which reasoning considers potential arguments is driven by

considerations that are partly irrelevant to the task of finding a good argument, as illustrated

earlier by the thoughts that are the most accessible for Margo but that are not relevant as

arguments. The order in which beliefs are considered is explored in the fourth section that

deals with relevance. Another property of this type of search is its seriality. Only one object

(in visual search) or one representation (in the search for arguments) is considered at a given

time. That reasoning mostly functions in such a manner will be demonstrated in section five.

But before this, we turn to the most fundamental property a mechanism whose function is to

find arguments should have: that of being metarepresentational.

Reasoning as a metarepresentational device

The question of whether reasoning should be seen only as a metarepresentational device is

both semantic and substantial. On the semantic level, the answer will depend on the field in

which ‘reasoning’ is used. In some areas of philosophy and psychology reasoning has

acquired a very broad meaning that is, in fact, much closer to the meaning of ‘inference’ than

to the common sense meaning of ‘reasoning’ as conscious ratiocination. This ‘reasoning’

clearly does not only refer to metarepresentational mechanisms. I will argue, however, that

within most dual-process theories at least part of what is usually called analytic or system 2

reasoning (and what is simply called reasoning here) is indeed a specific type of

metarepresentational device. This is a substantial claim, that there exists a specific mental

mechanism exerting a specific function in a specific manner—whether one wants to call it,

and only it, reasoning or not.

If one grants that being an argument—good or bad—is a property of representations, as

opposed to other things in the world, then any mechanism that tries to find and evaluate

arguments must be metarepresentational: it must construct representations of representations

in order to evaluate them as argumentsiii. So the question is: are the psychological

mechanisms described by psychologists of reasoning metarepresentational?


10

Among the proponents of dual process theories, at least two argue directly, if in other words,

for a metarepresentational view of reasoning. Stanovich suggests that reasoning is (at least in

part) a device that allows us to create ‘representations of our own representations’ (Stanovich,

2004, p. 51). This, however, still leaves open the possibility that there may be other forms of

reasoning. Evans is clearer on the fundamental nature of this distinction. He is careful to

oppose his epistemic mental models—those underlying reasoning—to semantic mental

models—representations of the world (Evans, 2006, 2007). For Evans, epistemic mental

models are akin to epistemic attitudes (I am sure that X, I doubt that Y). Epistemic attitudes

are metarepresentational because they judge other representations: you can doubt ‘that there is

a chair’ (a representation) but you cannot doubt ‘a chair’ (a simple object).

Beyond these broad characterizations of reasoning, it may be more useful to look at specific

examples in order both to illustrate the difference between representational and

metarepresentational mechanisms, and to make a stronger point that reasoning is purely

metarepresentational. Non-metarepresentational processes can lead to inferences or behaviors

that are superficially similar to the ones performed by reasoning. Margo (or, for that matter,

her cat) knows that there is food in either one of two places. Upon visiting the first place and

finding it empty, she moves on to the second. Even thought this inference could be formalized

as a disjunctive syllogism, there is no need for reasoning proper: all that is needed is a

mechanism that stacks intentions and switches to the next one in line if the current one is not

fulfilled. On the other hand, Margo (but not her cat) is also able to solve a disjunctive

syllogism such as “there is food in location A or B; there is no food in A; therefore there is

food in B” through reasoning. In this case Margo explicitly uses the two premises as

arguments to draw the conclusion. She knows the reasons for drawing the conclusion, and it is

because she understands them that she draws the conclusion. Since the premises are

considered as arguments, or as reasons, they must have been metarepresented.

Even though, in this case, intuitions and reasoning can lead to similar outcomes, the two

processes are very different. In the case of reasoning, of the explicit consideration of reasons,

Margo can explain on the spot why she drew the conclusion, without having to try to retrace

his steps. When decisions or inferences are the result of other mechanisms, people are

generally only able to offer a rationalization in guise of explanation (Evans & Wason, 1976;


11

Nisbett & Wilson, 1977; Wilson & Dunn, 2004). If Margo was under time pressure or

stressed her ability to perform intuitive inferences would likely be unchallenged but her

ability to reason may very well be impaired (see e.g., Markman, Maddox, & Worthy, 2006).

The distinction between intuitive inferences and reasoning may be clearest when we consider

premises that can be used as arguments or not. Margo and Simon, still arguing about how to

go to the opera, are coming out of their flat. Simon steps out just before Margo and says ‘it’s

windy.’ Upon hearing this, Margo can have two reactions (that are not mutually exclusive).

She can draw an intuitive inference from this new piece of information, inference that will

make her less likely to choose to walk, much in the same way as if she had experienced the

wind herself. In this case, once she has accepted the communicated information, and if she

does draw the inference, then she will accept its output. But Margo can also understand

Simon’s utterance as an argument in their ongoing conversation: as a reason not to walk. In

this case, she will represent the same proposition and evaluate its validity as an argument for

not walking. She may very well believe Simon but still decide that it is a bad argument

because it is very unlikely that the wind will be strong enough to make a difference. She can

perfectly well accept the truth of the utterance without accepting it as an argument. Moreover,

the opposite can also happen. Margo accepts Simon’s utterance as an argument. Then she

steps out and decides that there is so little wind that it does not even qualify as windy. At this

point she refuses Simon’s utterance: she does not think it is windy. But she can still think that,

had it really been windy, it would have been a good reason not to walk. This shows two

things. First, that reasoning is a distinct mechanism. It is possible to separate its effects from

that of other inferences, even in cases where the premises (‘it’s windy’) and the conclusion

(‘walking is not a good idea’) are similar. Second, that reasoning is metarepresentational.

What matters in evaluating ‘it’s windy’ as an argument is not its relation to the world—

whether it is, in fact, windy or not windy. What matters is the relation between ‘it’s windy’

and the conclusion ‘we should not walk.’ And such relationships between representations

have to be evaluated by metarepresentational mechanisms, in the same way as relationships

between things in the world have to be evaluated by representational mechanisms.

Seriality


12

Even if one grants that reasoning is metarepresentational, it does not follow that it should

work as a mechanism that relies in great part on serial evaluation. In the case of visual search,

some searches rely on highly parallel processes. When searching for a tennis ball in the

cutlery drawer, the properties of many different elements are evaluated in parallel, allowing

you to focus directly on the tennis ball. One could imagine a metarepresentational process that

would function in a similar manner. Parallel search is made possible, however, by the fact that

we have a category for the target of the search (the tennis ball) and a category for the other

objects (the cutlery). In the case of arguments, we have neither: we generally do not know

what a good argument will be, and the representations that are not good arguments do not

belong to a single category (they may all be bad arguments for different reasons). In a latter

section some stable categories reasoning can rely upon are discussed, and it is only to the

extent that these categories are used that reasoning can work in a parallel manner. Otherwise

parallel processes are ill-suited to the task. Instead, reasoning should proceed mostly in a

serial fashion: considering one representation at a time until one that fits with some criteria of

being a good argument is found.

Several psychologists have claimed that reasoning can only focus on one item at a time.

Legrenzi and his colleagues have described several errors that may result from such a

phenomenon, known as ‘focusing’ (Legrenzi, Girotto, & Johnson-Laird, 1993). But it is

Evans who has made the most convincing claim for the seriality of reasoning, when defending

his singularity principle. According to the singularity principle, ‘when we think hypothetically

[i.e. when we reason] we only consider one possibility or mental model at a time’ (Evans,

2007, p.17). Evans gathers several pieces of evidence in support of his claim that I will

quickly summarize here. First he reviews work on hypothesis testing showing that people

consider only one hypothesis at any given time (Mynatt, Doherty, & Dragan, 1993, see also

Shaklee & Fischhoff, 1982). He also cites classical work on inductive reasoning that has come

to the same conclusion (in particular Bruner, Goodnow, & Austin, 1956, who speak of an

‘abhorrence of disjunctiveness’ (p.160, see also Levine, 1966). If people can only consider

one possibility at a time, they should have trouble with tasks that require the concurrent

representation of two possibilities. And indeed performance plummets when problems involve

exclusive disjunction (‘A or else B’), as illustrated by the famous THOG problem (Wason &

Brooks, 1979). More recently Johnson-Laird and his colleagues have created problems

involving ‘or else’ that can mislead more than 90% of participants (Johnson-Laird, Legrenzi,


13

Girotto, & Legrenzi, 2000). Studies of everyday decision making come to the same

conclusion: people only consider one option at a time when reasoning about a problem (see

e.g., Klein, 1998). While this evidence does not rule out the possibility of parallel processes

working in a manner that would be hard to assess experimentally, it shows that serial

processes play a considerable role in reasoning.

Relevance

If the type of search that has been advocated above is necessary, it is because the objects

among which the search is conducted do not present themselves in the right order. If it were

the case, there would be no need for a serial search mechanism: picking the first answer

would be enough. When we search for an unusual object, such as something to drive a nail

instead of a hammer, a good solution is not going to pop out of the visual scene. Instead,

solutions are going to present themselves in an order determined by some partly irrelevant

criteria. The same goes for arguments. In this case as well the order in which representations

are be evaluated is determined by factors that are partly irrelevant to the issue at hand. Which

representations are considered depends on their relative accessibility in the context of the

search. For instance, Simon’s beliefs about walking, about the Tube, and about Margo will be

more salient, more easily accessible. This is both a blessing—an insufficient one but a

blessing nonetheless—and a curse. This is a blessing because there is a very high probability

that good arguments will be found among such beliefs, much more than among, say, beliefs

about the foreign policy of Guatemalaiv. It is an insufficient blessing because most of these

beliefs are nevertheless bad arguments (or not arguments at all). And it can also turn into a—

very mild—curse in that it will make it very improbable that a good argument that lies beyond

these easily accessible beliefs will be considered.

Whatever its supposed advantages or drawbacks may be, what evidence is there that such

considerations of accessibility actually guide the search? Once again, we can turn to Evans

who has suggested that reasoning is guided by a relevance principle (Evans, 2006, 2007).

According to this principle, ‘mental models are generated by heuristic or pragmatic processes

that are designed to maximize relevance in a particular context, given the current goals of the

reasoner’ (Evans, 2007, p.18). What is important to understand here is that relevance is not

only determined by the problem that reasoning has to solve but by other factors such as prior


14

beliefs, goals, and context. There are two ways to demonstrate this. The first is by showing

that the order in which beliefs are considered is not solely determined by their value for the

problem at hand. The other is by showing that this order is not random but can be explained

by the considerations of other factors.

The first point can be easily deduced from the fact that people’s ‘logical performance in

abstract reasoning tasks is generally quite poor’ (Evans, 2002, p. 981). If participants faced

with reasoning tasks were able to focus on the most relevant arguments for the task at hand,

they would solve them all effortlessly. The tasks used—conditional inferences, syllogisms and

the like—are computationally trivial; one might say that the solution is right under the

participant’s eyes. For instance, in the famous Wason selection task (Wason, 1966), all the

possible answers are easily accessible. What are not easily accessible are good reasons for the

good answers. Moreover, participants who failed at the task can be convinced that there was a

better answer (i.e. an answer supported by better arguments, see Maciejovsky & Budescu,

2007; Moshman & Geil, 1998). So it is not as if participants were unable to understand that

there are better arguments, it is simply that they do not consider them spontaneously.

Given that the order in which representations are considered is not simply determined by their

relevance for the task at hand, by what is it determined? A host of well known cognitive

factors will play a role. They go by different names and may refer to different psychological

mechanisms: availability, anchoring, salience, accessibility, framing, relevance, etc. The

heuristic and biases program has investigated several of these phenomena and showed how

the ease with which a solution is brought to mind has an important effect on the final answer,

even in cases in which the influence is wholly unwarranted (Gilovich, Griffin, & Kahneman,

2002; Kahneman, Slovic, & Tversky, 1982). In an oft-cited experiment, participants were

given a random number from 1 to 100 (the anchor) and this influenced their later answer to

the question ‘how many countries are there in Africa?’ (Tversky & Kahneman, 1974).

Mussweiler and his colleagues have convincingly argued that this effect is mediated by an

increase in the accessibility of anchor consistent knowledge: participants will think first of

arguments that are consistent with the anchor (Mussweiler & Strack, 1999a, 1999b;

Mussweiler, Strack, & Pfeiffer, 2000). Likewise, the framing of problems affects choices

through ‘the generation of arguments for available choice alternatives’ (Milch, Weber,

Appelt, Handgraaf, & Krantz, 2009, p. 244, see Shafir, Simonson, & Tversky, 1993). The


15

most precise formulation of these diverse influences can be found in relevance theory, as

formulated by Sperber and Wilson (Sperber & Wilson, 1995). Several experiments have

demonstrated that considerations of relevance determine what conclusion is going to be

considered most spontaneously when faced with a reasoning problem (Sperber, Cara, &

Girotto, 1995; Van der Henst, 2006; Van der Henst, Sperber, & Politzer, 2002). It is plausible

to think that the theory could account as well for the order in which arguments are then

considered.

There is ample evidence that the order in which arguments are considered is not driven

exclusively by their relevance for the task at hand. However, even if the order in which

reasoning considers different arguments is not optimal, the evaluation procedure could still be

so demanding that it ends up finding the best arguments. Some of the evidence we have

reviewed—the poor performances in reasoning tasks for instance—argue against this

possibility and the next section will explain why a mechanism that has to find arguments

should not be expected to look for the best ones.

Satisficing

Since the notion of satisficing was introduced by Simon (1982), it has become nearly a truism

that psychological mechanisms satisfice, that they do not behave in line with standards

models of optimization. This is only to be expected if one takes costs into account. Given that

oftentimes more precise calculations would be much more costly while yielding little

improvement, it makes sense to use heuristics instead—in some sense, it becomes optimal to

use heuristics (see e.g., Gigerenzer, Todd, & ABC Research Group, 1999). Claiming that

reasoning satisfices would not be terribly interesting, as all cognitive mechanisms are bound

to satisfice to some extent. My claim here is stronger: that reasoning satisfices much more

than most psychological mechanism.

When looking for arguments, there are two reasons for wanting to find a decent enough

argument. The first is that there is always a risk of looking foolish by uttering something

idiotic. But this risk does not apply only to arguments, it applies to anything one might say.

Therefore, it is probably dealt with by other control mechanisms linked with communication

more generally and can safely be ignored here. The second reason to find decent arguments is,


16

obviously, so that we can convince our interlocutor. What is remarkable about this task is that

failing to find a convincing argument is as close as possible to a free error. First because it is

generally possible to try again: if our first argument fails to carry the day, we can suggest

another, and then another again, etc. There is no pressure to start by the best argument. This is

very different from most other situations; often a decision will entail a fork: if you think that

taking a given road will be faster, or that courting a given person is better, by shifting to

another solution you will generally incur a cost, sometimes a very important one. Not only is

failing to find the best argument first not a problem, but failing to convince at all is also often

not a big issue. Argumentation is typically used to gain something extra. Only very rarely will

it be a question of life or death, or even of important costs. Again, this is quite different from

many other psychological mechanisms for which failure is extremely costly, with a predator

recognition mechanism being a somewhat extreme example.

So, if reasoning is indeed a device designed to find arguments, it should be satisfied with

arguments that, far from being the best, are just good enough that they might have a chance to

work. In one sense, the whole of the psychology of reasoning bears testimony to the fact that

reasoning satisfices. If reasoning worked until it found the best possible argument, then it

would provide the correct answer for most reasoning problems, something that it clearly does

not do. Studying informal reasoning, Perkins comes to a similar conclusion: ‘many reasoners

could be characterized as "makes sense epistemologists." Such reasoners proceed to analyze a

situation only to the point where the analysis makes superficial sense.’ (Perkins, 1985, p.568).

Likewise, for Nisbett and Ross ‘the lay scientist seems to search only until a plausible

antecedent is discovered’ (Nisbett & Ross, 1980, p. 119, see also Rozenblit & Keil, 2002).

Analogical reasoning follows the same pattern: in standard tasks people are mostly guided by

superficial features (e.g., Gentner, Rattermann, & Forbus, 1993).

Such results, particularly those dealing with informal reasoning, may seem to undermine the

earlier claim that people are good at arguing, but the tension is only superficial. On the one

hand, being satisfied by relatively weak arguments when the context does not call for sounder

arguments is what one should expect once costs are taken into account. This is part of the

good overall organization of our cognitive systems that we only devote extra effort when

extra effects are expected (Sperber & Wilson, 1995). On the other hand, this ‘adaptive

laziness’ does not preclude people both from evaluating arguments accurately and from being


17

able to find better arguments when the need arises. A good illustration of this is the use made

of explanations (make sense causal theories) and evidence (data) by arguers. In a study of

informal reasoning, Kuhn asked participants to take a stance and defend their views on a

variety of topics (Kuhn, 1991). One of the ‘flaws’ that she observed was a lack of preference

for evidence over explanations: people were much more likely to be satisfied with a

superficial explanation than to resort to hard data. In the present framework, this should be

expected given that participants were not faced with a debating partner, but with an ‘inert’

experimentalist who would not contest their claims. People who had to engage in real debates

showed a marked improvement in the quality of their arguments—presumably because their

weakest arguments were refuted and they had to make the effort to find better ones (Kuhn,

Shaw, & Felton, 1997). Moreover, when people were offered easy access to evidence they

quickly realized that it would make for better arguments; they were also more likely to be

convinced by arguments relying on evidence than explanation (Brem & Rips, 2000). Children

are also able to provide better arguments, or warrants and backings for their arguments when

pressed by teachers or experimenters (Anderson, Chinn, Chang, Waggoner, & Yi, 1997;

Anderson et al., 1997; Anderson, Chinn, Waggoner, & Nguyen, 1998; Webb et al., 2008).

The picture that has been painted of reasoning so far fits perfectly with what is required of a

mechanism that has to find arguments. It is metarepresentational; it evaluates potential

arguments one by one, in an order that is driven by general considerations of relevance; and it

is satisfied by weak arguments if they are not contested. The next feature to be explored is

particularly interesting because, if it is expected of a mechanism that has to look for

arguments, it would be deeply maladaptive for other kinds of mechanism.

The confirmation Bias

When Margo wants to convince Simon to walk to the opera, she is only interested in

arguments that will support her point of view or that will undermine Simon’s. More generally,

a mechanism that looks for arguments should have a very strong confirmation bias: it should

only try to find arguments that will support the individual’s conclusion. There is a huge

literature in psychology—from reasoning to decision making through social psychology—

claiming that people suffer from such a bias (see Nickerson, 1998, for review). In many cases,

however, these results can be accounted for by the use of a ‘positive testing strategy,’ based


18

on the assumption that one’s original hypotheses are not too far off the mark (Klayman & Ha,

1987). Given that using such a strategy is the rational thing to do in most situations, it hardly

deserves to be called a ‘bias.’ More importantly, it is also misleading to call it a

‘confirmation’ or ‘confirmatory’ bias since its goal is not to confirm one’s initial ideas. Such a

strategy can sometimes lead to undue confirmation of our ideas, but only as a side-effect of its

overall search for accuracy. Many phenomena that were once explained as instances of a

confirmation bias are now explained as the results of other mechanisms that are not

confirmatory in nature (Evans, 1998; Klayman, 1995). But these mechanisms are intuitive. It

is only by turning to reasoning that one can find evidence of a genuine confirmation biasv.

Maybe the best example is that of the Wason selection task. This is a simple reasoning task

that involves a conditional rule and requires that participants understand how to falsify the

rule in order to succeed. At first, it was thought that participants had an intuitive tendency to

confirm the rule instead of falsifying it (Wason, 1966). Later research has shown, however,

that the answers are better explained by positing a two stages process. First, intuitions related

to utterance comprehension will guide the participants’ attention towards a given answer. In

the standard task, this answer will confirm the rule, but slight changes in the task will lead

these intuitions towards an answer that falsifies the rule (Evans & Lynch, 1973; Girotto,

Kemmelmeier, Sperber, & Van der Henst, 2001; Sperber et al., 1995). This means that these

intuitions do not aim either at confirming or at falsifying the rule. After this initial stage

reasoning kicks in and, in the vast majority of the cases, it only looks for justifications

supporting participants’ initial intuitions (Evans, 1996; Lucas & Ball, 2005; Roberts &

Newton, 2002). Here is a genuine confirmation bias. Reasoning is not trying to produce the

best answer; instead finding confirmatory arguments is its very goal. And the case of the

Wason selection task is not an exception: ‘[The] confirmation bias is perhaps the best known

and most widely accepted notion of inferential error to come out of the literature on human

reasoning’ (Evans, 1989, p. 41).

It is important to stress that the confirmation bias does not stem from an inability to grasp

falsification. Participants are perfectly able to use diverse and sophisticated strategies in order

to falsify something they disagree with. Coming back to the Wason selection task, when

participants are motivated to prove that the rule is false, they become much more apt at

finding the answer that does exactly that (Dawson, Gilovich, & Regan, 2002). Likewise, when


19

an hypothesis is given by someone else and we have reasons to doubt its validity, falsification

comes much more easily (Butera, Legrenzi, Mugny, & Pérez, 1992; Cowley & Byrne, 2005).

And when a syllogism has a conclusion that contradicts our beliefs we look for sophisticated

arguments to justify its rejection (Klauer, Musch, & Naumer, 2000). In all these cases,

falsifying strategies are only used to confirm our initial hunch that a proposition is false.

That reasoning displays a prevalent and robust confirmation bias would be hard to contest.

This is quite striking, especially since it can lead to rather poor outcomes when reasoning is

used in the wrong contexts (see Mercier & Sperber, in press). If reasoning was supposed to

help us improve our epistemic status or help us make better decisions, then the existence of

such a bias would be a deep mystery. On the other hand, it is an expected feature for a

mechanism designed to find arguments.

Categories of arguments

Rhetoricians have developed elaborate classifications of arguments—ad hominem argument,

argument by analogy, argument by examples, etc. Moreover, we know that categories are

extremely useful in facilitating other kinds of searches, such as visual search. Thus, one would

think that these categories could be used to find arguments. But it is not clear that naïve

people can rely on such categories while looking for arguments. In fact, these categories are

partly orthogonal to the problem at hand, which is to find a good argument, whatever kind of

argument it may be. When you want to find a good restaurant, and if you have ecumenical

tastes, then the categories ‘Japanese restaurant,’ ‘Italian restaurant’ or ‘French restaurant’ are

rather irrelevant. What you are looking for is a category ‘good restaurant.’ The analogy is

limited because in the case of restaurants, it would be possible to create a category ‘good

restaurants,’ but this is generally not possible in the case of argumentsvi. Depending on the

present topic, the context, the interlocutor, etc., any given representation can be a good or a

bad argument (or not an argument at all). Still, sometimes you will not have had the time to

form a category ‘good restaurants’—you are in a new city, say. In this case, the analogy with

the restaurants can provide us with a useful clue. For instance, there may be an area in this

city that is likely to have several such restaurants, and you might just decide to stroll around

these streets until you find a place that satisfies your requirements. This is possible because


20

cities are organized in a manner that will make your search easier. Are our minds organized in

the same helpful way?

By ranking possible arguments in a manner that is, if insufficient, at least extremely helpful,

relevance makes it at least possible to find arguments. This is the first manner in which the

organization of the mind helps us find arguments. But this ranking is agnostic regarding other

important properties that a representation has to have to be a good argument. For instance, a

proposition may have equally relevant consequences that have a positive and a negative

valence. But the former will be a good argument to support the proposition and the latter will

not. This pattern holds generally: positive consequences of one’s proposition, as well as

negative consequences of one’s interlocutor’s proposition, often make good arguments, while

the converses don’t. So if our minds are organized in such a way that negative and positive

consequences are somehow compartmentalized, then we could rely on this

compartmentalization for a more efficient search. Much like we can choose to only look for

restaurants in a given area, maybe we can choose to focus on positive or negative

consequences. At least that would save us the trouble of looking at representations that will

only very rarely turn out to be good arguments. It may be possible to direct the search in other

ways. Instead of consequences, one could focus on antecedents—representations from which

we can draw inferences leading to the conclusion—in which case restricting the search to the

antecedents of one’s own position will be a good idea. The converse would be to look for

representations that are incoherent with our interlocutor’s position. Or we may look for

representations that have a similar relationship with one another as one premise and the

conclusion we want to convey, creating an argument by analogy. So it is possible that the

search for arguments can be oriented in a given direction.

All of these suggestions make assumptions about the structure of the mind that are not trivial,

and there is a dearth of empirical work examining this question of categories of arguments.

Moreover, the theory presented here does not have to make precise predictions regarding what

strategies, or categories, can be used when looking for arguments, so I will not belabor the

point any further.

***


21

Starting from the description of a computational problem, it is possible to develop hypotheses

regarding the kind of mechanism that would be best able to solve it. The problem that has

been tackled here is that of finding arguments. In this case, I have argued that one should

expect a metarepresentational mechanism, so that it can evaluate representations as

arguments. This mechanism that works in a serial manner, examining each possible argument

in turn. The need to examine several arguments arises partly from the fact that representations

are ordered following general considerations of relevance, and not only their relevance as

arguments. Moreover, such a mechanism should satisfice and it should have a confirmation

bias. Evidence has been presented showing that reasoning, as investigated in the psychology

of reasoning and decision making, fits very well with these requirements. This fit is taken as

further evidence in favor of the argumentative theory of reasoning, according to which

reasoning evolved precisely to fulfill an argumentative function. The last two properties—an

extremely high degree of satisficing and a very strong confirmation bias—are particularly

striking in that they are not expected of a psychological mechanism that would have another

function.

Given that the question of finding arguments has not been thoroughly explored in the

psychological literature, the model suggested here has to be considered as a tentative first

step. Hopefully, it will lead to more elaborate suggestions and stimulate empirical research on

this topic.


22

Anderson, R. C., Chinn, C., Chang, J., Waggoner, M., & Yi, H. (1997). On the logical

integrity of childrenʼs arguments. Cognition and Instruction, 15(2), 135–167.

Anderson, R. C., Chinn, C., Waggoner, M., & Nguyen, K. (1998). Intellectually stimulating

story discussions. In J. Osborn & F. Lehr (Eds.), Literacy for All (pp. 170–196). New

York: Guilford.

Bailenson, J. N., & Rips, L. J. (1996). Informal reasoning and burden of proof. Applied

Cognitive Psychology, 10(7), 3-16.

Barsalou, L. W. (1983). Ad hoc categories. Memory and Cognition, 11(3), 211-227.

Baum, L. A., Danovitch, J. H., & Keil, F. C. (2007). Children's sensitivity to circular

explanations. Journal of Experimental Child Psychology, 100(2), 146-155.

Berry, D. C., & Dienes, Z. (1993). Implicit learning. Hove: Erlbaum.

Brem, S. K., & Rips, L. J. (2000). Explanation and evidence in informal argument. Cognitive

Science, 24, 573–604.

Bruner, J. S., Goodnow, J. J., & Austin, G. A. (1956). A Study of Thinking. New York: Wiley.

Butera, F., Legrenzi, P., Mugny, G., & Pérez, J. A. (1992). Influence sociale et raisonnement.

Bulletin de Psychologie, 45, 144–154.

Chaiken, S., Liberman, A., & Eagly, A. H. (1989). Heuristic and systematic processing within

and beyond the persuasion context. In J. S. Uleman & J. A. Bargh (Eds.), Unintended

thought (pp. 212-252). New York: Guilford Press.

Corner, A., Hahn, U., & Oakfsord, M. (2006). The slippery slope argument: probability,

utility and category reappraisal. Proceedings of the 28th Annual Meeting of the

Cognitive Science Society.

Cowley, M., & Byrne, R. M. J. (2005). When falsification is the only path to truth. Paper

presented at the Twenty-Seventh Annual Conference of the Cognitive Science Society,

Stresa, Italy.

Dancy, J. (2000). Practical reality. Oxford: Oxford University Press.

Dawson, E., Gilovich, T., & Regan, D. T. (2002). Motivated reasoning and performance on

the Wason selection task. Personality and Social Psychology Bulletin, 28(10), 1379.

Duncan, J., & Humphreys, G. (1989). Visual search and stimulus similarity. Psychological

Review, 96, 1–26.

Evans, J. S. B. T. (1989). Bias in Human Reasoning: Causes and Consequences. Hillsdale,

NJ: Lawrence Erlbaum.


23

Evans, J. S. B. T. (1996). Deciding before you think: Relevance and reasoning in the selection

task. British Journal of Psychology, 87, 223-240.

Evans, J. S. B. T. (1998). Matching bias in conditional reasoning: Do we understand it after

25 years? Thinking and Reasoning, 4, 45-82.

Evans, J. S. B. T. (2002). Logic and human reasoning: an assessment of the deduction

paradigm. Psychological bulletin, 128(6), 978-996.

Evans, J. S. B. T. (2003). In two minds: dual-process accounts of reasoning. Trends in

Cognitive Sciences, 7(10), 454-459.

Evans, J. S. B. T. (2006). The heuristic-analytic theory of reasoning: Extension and

evaluation. Psychonomic Bulletin and Review, 13(3), 378-395.

Evans, J. S. B. T. (2007). Hypothetical Thinking: Dual Processes in Reasoning and Judgment.

Hove: Psychology Press.

Evans, J. S. B. T. (2008). Dual-processing accounts of reasoning, judgment and social

cognition. Annual Review of Psychology, 59, 255-278.

Evans, J. S. B. T., & Frankish, K. (Eds.). (2009). In Two Minds. Oxford: Oxford University

Press.

Evans, J. S. B. T., & Lynch, J. S. (1973). Matching bias in the selection task. British Journal

of Psychology, 64(3), 391-397.

Evans, J. S. B. T., & Over, D. E. (1996). Rationality and Reasoning. Hove: Psychology Press.

Evans, J. S. B. T., & Wason, P. C. (1976). Rationalization in a reasoning task. British Journal

of Psychology, 67, 479-486.

Gawronski, B., & Bodenhausen, G. V. (2006). Associative and propositional processes in

evaluation: An integrative review of implicit and explicit attitude change.

Psychological Bulletin, 132(5), 692-731.

Gentner, D., Rattermann, M. J., & Forbus, K. D. (1993). The roles of similarity in transfer:

Separating retrievability from inferential soundness. Cognitive Psychology, 25(4),

524-575.

Gigerenzer, G., Todd, P. M., & ABC Research Group. (1999). Simple Heuristics That Make

Us Smart. Oxford: Oxford University Press.

Gilovich, T., Griffin, D., & Kahneman, D. (Eds.). (2002). Heuristics and Biases: The

Psychology of Intuitive Judgment. Cambridge, UK: Cambridge University Press.


24

Girotto, V., Kemmelmeier, M., Sperber, D., & Van der Henst, J.-B. (2001). Inept reasoners or

pragmatic virtuosos? Relevance and the deontic selection task. Cognition, 81(2), 69-

76.

Hahn, U., & Oaksford, M. (2007). The rationality of informal argumentation: A bayesian

approach to reasoning fallacies. Psychological Review, 114(3), 704-732.

Hahn, U., Oaksford, M., & Bayindir, H. (2005). How convinced should we be by negative

evidence? Proceedings of the 27th Annual Meeting of the Cognitive Science Society.

Johnson-Laird, P. N. (2006). How We Reason. Oxford: Oxford University Press.

Johnson-Laird, P. N., Legrenzi, P., Girotto, V., & Legrenzi, M. S. (2000). Illusions in

reasoning about consistency. Science, 288(5465), 531.

Kahneman, D. (2003). A perspective on judgment and choice: Mapping bounded rationality.

American Psychologist, 58(9), 697-720.

Kahneman, D., & Frederick, S. (2002). Representativeness revisited: Attribute substitution in

intuitive judgement. In T. Gilovich, D. Griffin & D. Kahneman (Eds.), Heuristics and

Biases: The Psychology of Intuitive Judgment (pp. 49-81). Cambridge, UK:

Cambridge University Press.

Kahneman, D., Slovic, P., & Tversky, A. (1982). Judgment Under Uncertainty: Heuristics

and Biases: Cambridge University Press.

Klauer, K. C., Musch, J., & Naumer, B. (2000). On belief bias in syllogistic reasoning.

Psychol Rev, 107(4), 852-884.

Klayman, J. (1995). Varieties of confirmation bias. In J. R. Busemeyer, R. Hastie & D. L.

Medin (Eds.), Decision Making from the Perspective of Cognitive Psychology (pp.

385-418). New York: Academic Press.

Klayman, J., & Ha, Y. (1987). Confirmation, disconfirmation, and information in hypothesis

testing. Psychological Review, 94, 211-228.

Klein, G. (1998). Sources of Power: How People Make Decisions. Cambridge, MA: MIT

Press.

Kuhn, D. (1991). The Skills of Arguments. Cambridge: Cambridge University Press.

Kuhn, D., Shaw, V. F., & Felton, M. (1997). Effects of dyadic interaction on argumentative

reasoning. Cognition and Instruction, 15, 287-315.

Legrenzi, P., Girotto, V., & Johnson-Laird, P. N. (1993). Focussing in reasoning and decision

making. Cognition, 49, 37-66.


25

Levelt, W. J. M. (1989). Speaking: From Intention to Articulation. Cambridge, MA: MIT

Press.

Levelt, W. J. M., Roelofs, A., & Meyer, A. S. (1999). A theory of lexical access in speech

production. Behavioral and Brain Sciences(22), 1-45.

Levine, M. (1966). Hypothesis behaviour by humans during discrimination learning. Journal

of Experimental Psychology, 71, 331-338.

Li, Z. (2002). A saliency map in primary visual cortex. Trends in Cognitive Sciences, 6(1), 9-

16.

Lucas, E. J., & Ball, L. J. (2005). Think-aloud protocols and the selection task: Evidence for

relevance effects and rationalisation processes. Thinking and Reasoning, 11, 35-66.

Maciejovsky, B., & Budescu, D. V. (2007). Collective induction without cooperation?

Learning and knowledge transfer in cooperative groups and competitive auctions.

Journal of personality and social psychology, 92(5), 854-870.

Markman, A. B., Maddox, W. T., & Worthy, D. A. (2006). Choking and excelling under

pressure. Psychological Science, 17, 944–948.

Mercier, H. (submitted). Reasoning and argumentation: not everything has to be learned.

Mercier, H. (in press). On the universality of argumentative reasoning. Journal of Cognition

and Culture.

Mercier, H., & Landemore, H. (submitted). Understanding the successes and failures of

deliberation.

Mercier, H., & Sperber, D. (2009). Intuitive and reflective inferences. In J. S. B. T. Evans &

K. Frankish (Eds.), In Two Minds. New York: Oxford University Press.

Mercier, H., & Sperber, D. (in press). Why do humans reason? Arguments for an

argumentative theory. Behavioral and Brain Sciences.

Milch, K. F., Weber, E. U., Appelt, K. C., Handgraaf, M. J. J., & Krantz, D. H. (2009). From

individual preference construction to group decisions: Framing effects and group

processes. Organizational Behavior and Human Decision Processes.

Moshman, D., & Geil, M. (1998). Collaborative reasoning: Evidence for collective rationality.

Thinking and Reasoning, 4(3), 231-248.

Mussweiler, T., & Strack, F. (1999a). Comparing is believing: A selective accessibility model

of judgmental anchoring. In W. Stroebe & M. Hewstone (Eds.), European Review of

Social Psychology (Vol. 10, pp. 135-168). Chichester: Wiley.


26

Mussweiler, T., & Strack, F. (1999b). Hypothesis-consistent testing and semantic priming in

the anchoring paradigm: A selective accessibility model. Journal of Experimental

Social Psychology, 35, 136-164.

Mussweiler, T., Strack, F., & Pfeiffer, T. (2000). Overcoming the inevitable anchoring effect:

Considering the opposite compensates for selective accessibility. Personality and

Social Psychology Bulletin, 26(9), 1142.

Mynatt, C. R., Doherty, M. E., & Dragan, W. (1993). Information relevance, working

memory and the consideration of alternatives. Quarterly Journal of Experimental

Psychology, 46A, 759-778.

Neuman, Y. (2003). Go ahead, prove that God does not exist! On high school students' ability

to deal with fallacious arguments. Learning and Instruction, 13(4), 367-380.

Neuman, Y., Weinstock, M. P., & Glasner, A. (2006). The effect of contextual factors on the

judgement of informal reasoning fallacies. The Quarterly Journal of Experimental

Psychology, 59(2), 411-425.

Nickerson, R. S. (1998). Confirmation bias: A ubiquitous phenomena in many guises. Review

of General Psychology, 2, 175-220.

Nisbett, R. E., & Ross, L. (1980). Human inference: Strategies and shortcomings of social

judgment. Englewood Cliffs, N.J: Prentice–Hall.

Nisbett, R. E., & Wilson, T. (1977). Telling more than we can know. Psychological Review,

84(1), 231-259.

Oaksford, M., & Hahn, U. (2004). A Bayesian approach to the argument from ignorance.

Canadian Journal of Experimental Psychology, 58(2), 75-85.

Perelman, C., & Olbrechts-Tyteca, L. (1969). The New Rhetoric: A Treatise on

Argumentation. Notre Dame, IN: University of Notre Dame Press.

Perkins, D. N. (1985). Postprimary education has little impact on informal reasoning. Journal

of Educational Psychology, 77, 562-571.

Petty, R. E., & Cacioppo, J. T. (1986). The elaboration likelihood model of persuasion. In L.

Berkowitz (Ed.), Advances in Experimental Social Psychology (Vol. 19, pp. 123-205).

Orlando, FL: Academic Press.

Petty, R. E., & Wegener, D. T. (1998). Attitude change: Multiple roles for persuasion

variables. In D. Gilbert, S. Fiske & G. Lindzey (Eds.), The Handbook of Social

Psychology (Vol. 1, pp. 323–390). Boston: McGraw-Hill.


27

Posner, M. I., & Snyder, C. R. R. (1975). Attention and cognitive control. In R. L. Solso

(Ed.), Information Processing and Cognition: The Loyola Symposium. Hillsdale, NJ:

Erlbaum.

Reber, A. S. (1993). Implicit Learning and Tacit Knowledge. New York: Oxford University

Press.

Rips, L. J. (1994). The Psychology of Proof: Deductive Reasoning in Human Thinking.

Cambridge, MA: MIT Press.

Rips, L. J. (1998). Reasoning and conversation. Psychological Review, 105, 411–441.

Rips, L. J. (2002). Circular reasoning. Cognitive Science, 26, 767–795.

Roberts, M. J., & Newton, E. J. (2002). Inspection times, the change task, and the rapid

response selection task. Quarterly Journal of Experimental Psychology, 54, 1031-

1048.

Rozenblit, L., & Keil, F. (2002). The misunderstood limits of folk science. Cognitive Science,

26, 521-562.

Schacter, D. L. (1987). Implicit Memory: History and Current Status. Journal of experimental

psychology. Learning, memory, and cognition, 13(3), 501-518.

Shafir, E., Simonson, I., & Tversky, A. (1993). Reason-based choice. Cognition, 49(1-2), 11-

36.

Shaklee, H., & Fischhoff, B. (1982). Strategies of information search in causal analysis.

Memory & Cognition, 10(6), 520-530.

Shaw, V. F. (1996). The cognitive processes in informal reasoning. Thinking & Reasoning,

2(1), 51-80.

Simon, H. A. (1982). Models of Bounded Rationality. Cambridge, MA: MIT Press.

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

Bulletin, 119(1), 3-22.

Sperber, D. (2000). Metarepresentations in an evolutionary perspective. In D. Sperber (Ed.),

Metarepresentations: A Multidisciplinary Perspective (pp. 117-137). Oxford: Oxford

University Press.

Sperber, D. (2001). An evolutionary perspective on testimony and argumentation.

Philosophical Topics, 29, 401-413.

Sperber, D., Cara, F., & Girotto, V. (1995). Relevance theory explains the selection task.

Cognition, 57, 31-95.


28

Sperber, D., & Mercier, H. (In press). Reasoning as a social competence. In J. Elster & H.

Landemore (Eds.), Collective Wisdom.

Sperber, D., & Wilson, D. (1995). Relevance: Communication and Cognition. Oxford:

Blackwell.

Stanovich, K. E. (2004). The Robot's Rebellion. Chicago: Chicago University Press.

Stein, N. L., & Miller, C. A. (1993). The development of meaning and reasoning skill in

argumentative contexts: Evaluating, explaining, and generating evidence. In R. Glaser

(Ed.), Advances in Instructional Psychology (Vol. 4, pp. 285-335). Hillsdale:

Lawrence Erlbaum Associates.

Toulmin, S. (1958). The Uses of Argument. Cambridge: Cambridge University Press.

Tversky, A., & Kahneman, D. (1974). Judgement under uncertainty: heuristics and biases.

Science, 185, 1124-1131.

Van der Henst, J.-B. (2006). Relevance effects in reasoning. Mind & Society, 5(2), 229-245.

Van der Henst, J.-B., Sperber, D., & Politzer, G. (2002). When is a conclusion worth

deriving? A relevance-based analysis of indeterminate relational problems. Thinking

and Reasoning, 8, 1-20.

van Eemeren, F. H., Grootendorst, R., & Henkemans, F. S. (1996). Fundamentals of

Argumentation Theory. Mahwah, NJ: Erlbaum.

Wason, P. C. (1966). Reasoning. In B. M. Foss (Ed.), New Horizons in Psychology: I (pp.

106–137). Harmandsworth, England: Penguin.

Wason, P. C., & Brooks, P. J. (1979). THOG: The anatomy of a problem. Psychological

Research, 41, 79-90.

Wason, P. C., & Evans, J. S. B. T. (1975). Dual processes in reasoning? Cognition, 3, 141-

154.

Webb, N. M., Franke, M. L., Ing, M., Chan, A., De, T., Freund, D., et al. (2008). The role of

teacher instructional practices in student collaboration. Contemporary Educational

Psychology, 33(3), 360–381.

Weinstock, M., Neuman, Y., & Tabak, I. (2004). Missing the point or missing the norms?

Epistemological norms as predictors of students’ ability to identify fallacious

arguments. Contemporary Educational Psychology, 29(1), 77-94.

Wilson, T. D., & Dunn, E. W. (2004). Self-knowledge: Its limits, value, and potential for

improvement.


29

Wilson, T. D., Lindsey, S., & Schooler, T. Y. (2000). A model of dual attitudes.

Psychological Review, 107(1), 101-126.

i The Psychology of Proof (Rips, 1994) for mental logic theory, How we Reason (Johnson-Laird, 2006) for mental models theory and Hypothetical Thinking (Evans, 2007) for (one) dual process theory. It should be said however that Rips has devoted a lot of work to argumentation aside from his theories of reasoning. ii A similar idea is expressed by Perelman and Olbrechts-Tyteca in their classic study of argumentation: ‘One datum in argumentation consists of the agreements available to the speaker as supports for his argument. But this element is so large and capable of being used in so many different ways, that the manner in which one makes use of it is of paramount importance. Accordingly, before examining the use of this datum in argumentation, it is essential that we say something of the part played by preliminary selection of the elements that are to serve as the starting point of the argument and by the adaptation of these elements to its purpose’ (Perelman & Olbrechts-Tyteca, 1969, p.115).

iii See Sperber (2000). This point does not garner unanimity (Dancy, 2000) but is consensual enough to warrant accepting it, at least provisionally, here.

iv Accessibility, or relevance, will be particularly effective at organizing the space of possible arguments after someone has already given an argument. In this case, relevance mechanisms may triangulate quite effectively on counter-arguments. For instance, if Margo says ‘walking will be fast’, one of the first things that may spring to Simon’s mind is ‘but taking the Tube will be even faster,’ precluding the need for an extensive search.

v This argument is developed in Mercier & Sperber (in press) so it will only be brushed here.

vi The exception would be arguments that are made very often in similar contexts and for similar audiences. In this case, it should be possible to have an idea of what a good argument may be. A teacher may think, for instance, that arguments by analogy are particularly efficient in explaining a certain class of concepts.

Looking for Arguments

Documents