This article was downloaded by: [University of Chicago Library]

On: 09 August 2013, At: 12:08

Publisher: Routledge
Informa Ltd Registered in England and Wales Registered Number: 1072954
Registered office: Mortimer House, 37-41 Mortimer Street, London W1T
3JH, UK

Inquiry: An Interdisciplinary
Journal of Philosophy
Publication details, including instructions for
authors and subscription information:
http://www.tandfonline.com/loi/sinq20

The “Rationality Wars” in

Psychology: Where They Are
and Where They Could Go
a
Thomas Sturm
a
Universitat Autònoma de Barcelona, Spain
Published online: 18 Jan 2012.

To cite this article: Thomas Sturm (2012) The “Rationality Wars” in Psychology:
Where They Are and Where They Could Go, Inquiry: An Interdisciplinary Journal of
Philosophy, 55:1, 66-81, DOI: 10.1080/0020174X.2012.643628

To link to this article: http://dx.doi.org/10.1080/0020174X.2012.643628

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all
the information (the “Content”) contained in the publications on our
platform. However, Taylor & Francis, our agents, and our licensors
make no representations or warranties whatsoever as to the accuracy,
completeness, or suitability for any purpose of the Content. Any opinions
and views expressed in this publication are the opinions and views of
the authors, and are not the views of or endorsed by Taylor & Francis.
The accuracy of the Content should not be relied upon and should be
independently verified with primary sources of information. Taylor and
Francis shall not be liable for any losses, actions, claims, proceedings,
demands, costs, expenses, damages, and other liabilities whatsoever
or howsoever caused arising directly or indirectly in connection with, in
relation to or arising out of the use of the Content.

This article may be used for research, teaching, and private study
purposes. Any substantial or systematic reproduction, redistribution,
 reselling, loan, sub-licensing, systematic supply, or distribution in any form
to anyone is expressly forbidden. Terms & Conditions of access and use can
be found at http://www.tandfonline.com/page/terms-and-conditions
Downloaded by [University of Chicago Library] at 12:08 09 August 2013
 Inquiry,
Vol. 55, No. 1, 66–81, February 2012

The “Rationality Wars” in

Psychology: Where They Are and
Where They Could Go
Downloaded by [University of Chicago Library] at 12:08 09 August 2013

THOMAS STURM
Universitat Autònoma de Barcelona, Spain

(Received 25 September 2011)

ABSTRACT Current psychology of human reasoning is divided into several different

approaches. For instance, there is a major dispute over the question whether human beings
are able to apply norms of the formal models of rationality such as rules of logic, or prob-
ability and decision theory, correctly. While researchers following the “heuristics and
biases” approach argue that we deviate systematically from these norms, and so are per-
haps deeply irrational, defenders of the “bounded rationality” approach think not only
that the evidence for this conclusion is problematic but also that we should not, at least not
very often, use formal norms in reasoning. I argue that while the evidence for heuristics
and biases is indeed questionable, the bounded rationality approach has its limits too.
Most especially, we should not infer that formal norms play no role in a comprehen-
sive theory of rationality. Instead, formal and bounded rules of reasoning might even be
connected in a more comprehensive theory of rationality.

Introduction
Current psychology of human reasoning constitutes a field where some of
the most interesting debates about rationality are fought. In one such dis-
pute, there are two main camps. On the one side, there is the “heuristics and
biases” approach (“HB approach” for short), embodied most prominently
in the work of Amos Tversky and Daniel Kahneman. This party claims, on
empirical grounds, that human beings often and systematically violate norms
of rationality that derive from formal logic, probability and decision the-
ory. Instead, human judgment and decision making uses “heuristics”, rules

Correspondence Address: Thomas Sturm, Department of Philosophy, Universitat Autònoma

de Barcelona, Edifici B, E-08193, Bellaterra (Barcelona), Spain. Email: tsturm@mpiwg-berlin.
mpg.de

0020-174X Print/1502-3923 Online/12/010066–16 © 2012 Taylor & Francis

http://dx.doi.org/10.1080/0020174X.2012.643628
 The “Rationality Wars” in Psychology 67

of thumb that work fairly well in some contexts but are not generally valid
and thus lead to biases. On the other hand, an influential criticism of the
HB approach—indeed, the only one which has prompted Kahneman and
Tversky (1996) to a published reply—has been developed by defenders of the-
ories of “bounded” rationality (“BR approach” for short). For this, the work
by Gerd Gigerenzer and his colleagues stands out. They deny that human
beings are systematically irrational in their judgment and decision-making.
Also, they argue that it is wrong to take rules of logic, probability theory and
statistics as unquestionable norms of rationality. Instead, we should construct
norms out of an empirical study of the contents and contexts in which humans
Downloaded by [University of Chicago Library] at 12:08 09 August 2013

reason.
This debate has become so heated (see e.g., Kahneman & Tversky, 1996
and Gigerenzer, 1996a) that Samuels, Stich, and Bishop (2002) have called it
the “rationality wars”. They claim that the debate is largely due to rhetori-
cal excesses on both sides. Once the core assumptions of both programs are
soberly compared, it becomes clear that the dispute isn’t really about much.
For instance, they claim that the defenders of the HB approach do not all
maintain that humans are completely and unavoidably irrational in their judg-
ing and deciding about logical, probabilistic or statistical tasks. According
to this reconstruction of the debate, the issue at stake is simply how rational
human beings are: How many standard rules of good reasoning do we actu-
ally apply? How far are we able to train ourselves to become sound reasoners?
However, as I shall argue, the debate is also about different, more fundamen-
tal questions: What is the basis of norms of rationality? Can we use empirical
investigations of human reasoning to answer this question? These questions
are hard to answer, yet also important.1 Contrary to what Samuels, Stich
and Bishop argue, I insist that the theories of rationality assumed in the two
research programs are substantively different, especially in their normative
aspects. But, in contrast to both parties in the psychological debate, I doubt
that these aspects are incompatible. A comprehensive theory of rationality
could and should be composed of substantively different elements. We cannot
build a solid home from bricks alone; we need cement, wood, glass, steel and
other materials as well. (I should mention that another paper, co-authored
with Gerd Gigerenzer [Gigerenzer & Sturm, 2011], shows that in important
respects I agree with the BR approach, especially in that many norms of rea-
soning can indeed be validated empirically and why that is so. In the present
paper, in contrast, I spell out where I disagree with the approach.)
In Section I, I illustrate the “heuristics and biases” approach more closely.
In Section II, I explain how defenders of “bounded” conceptions of ratio-
nality criticize the HB approach by objections that are largely convincing.
In Section III, however, I point out three limitations of the BR approach, lead-
ing to the idea of connecting it more closely with more traditional normative
conceptions of rationality.
 68 Thomas Sturm

I. Are we highly irrational? The “heuristics and biases” approach

How could one find out whether human beings judge and decide in accor-
dance with standard norms of rationality, such as those from formal logic,
or probability and decision theory? A massive amount of the psychological
literature from the last decades (see Lopes, 1991; Evans & Over, 1996) has
applied the following procedure: Pick a particular norm, to be used in a task
of more or less ordinary reasoning, and see how many experimental subjects
apply it correctly. If they do that to a very large extent, then—given that the
sample is representative—it may be inferred that human beings are rational in
judgment and decision making; if not, they are irrational. Here is a concrete
Downloaded by [University of Chicago Library] at 12:08 09 August 2013

example, to which countless others could be added: the conjunction rule and
the “Linda problem”.
The conjunction rule of probability theory states that an event A can never
be less probable than a conjunction of the (independent) events A and B: Prob
(A) ≥ Prob (A & B). A typical psychological test concerning whether humans
follow the conjunction rule goes as follows:

Linda is 31 years old, single, outspoken and very bright. She majored
in philosophy. As a student, she was deeply concerned with issues
of discrimination, social justice, and also participated in antinuclear
demonstrations.
Which statement is more probable?
(T) Linda is a bank teller.
(T&F) Linda is a bank teller and active in the feminist movement.

Results: About 85% of 142 subjects chose the answer “T&F” and thus violated
the conjunction rule (Tversky & Kahneman, 1983, p. 299).
The test for the conjunction rule has been varied in many ways. Different
cases of persons, events etc. were described and then connected to var-
ious questions. Also, even when subjects were being informed about the
correct answer, the tendency to produce incorrect answers decreased very lit-
tle. Kahneman & Tversky (1996) insist that the results are stable (see also
Kahneman’s views in Mellers, Hertwig & Kahneman, 2001). They explain
this “cognitive illusion” by saying that subjects judge this way because of
the representativeness of certain features of the description of Linda: If some-
one is concerned with issues about discrimination, social justice, and nuclear
weapons, many people think that these properties are likely to be representa-
tive for her being a feminist as well. This heuristic makes subjects ignore the
conjunction rule.
It is similar with many other reasoning tasks: Rather than using formal
rules of logic, probability theory, or statistics and rational choice theory,
humans seem to judge and decide on the basis of “heuristics” (rules of thumb
which often work) which lead to “biases” (systematically incorrect results).
In well-known studies with the Wason Selection Task, it has been argued that
 The “Rationality Wars” in Psychology 69

most of us do not apply the rules of material implication: When asked to

check instances of P→Q, subjects overlook the relevance of instances of ¬Q
as potential falsifiers. For instance, subjects are given four cards with letters
on one side (e.g. “E” and “C”) and numbers on the other side (“3” and “4”).
Two cards are presented with the letter side up, two with the number side up.
Then subjects are asked to turn over only those cards which are necessary for
figuring out if the following statement is true: “If there is an E on the one
side, then there is a 4 on the other side.” Most subjects pick out only the E-
cards or the E- and the 4-cards (vowels and odd numbers). Less than 10% of
the test subjects choose the right solution: the E- and the 3-cards. This is, in
Downloaded by [University of Chicago Library] at 12:08 09 August 2013

any case, the right solution if one follows standard propositional logic, which
views statements of the form “If P, then Q” as material implications: these
are wrong if and only if P is true and Q is wrong. But subjects overlook the
importance of the falsifying 3-card (=¬Q), and are said to be biased towards
looking for confirmation (Wason, 1966). Such mistakes, so the story goes, are
not limited to laypeople. Statistically tutored Harvard medical students and
staff members (!), when tested, were unable to correctly estimate the proba-
bility of breast cancer given certain diagnostic evidence; they committed base
rate neglect (Casscells, Schoenberger, & Grayboys, 1978). They seemed to use
an “availability heuristic”, using certain directly available items of evidence
instead of base rates. We are prone to commit the gambler’s fallacy (thinking
that the probability of an independent event is somehow affected by a series
of previous events). And so on and so on (e.g., Tversky & Kahneman, 1974;
Kahneman, Slovic & Tversky, 1982; Nisbett & Ross, 1980; Gilovich, Griffin
& Kahneman, 2002).
What are the broader implications of these results? Many researchers have
asserted outright that they have “bleak implications for human rationality”
(Nisbett & Borgia, 1975, p. 935), or that “the selection task reflects [a tendency
towards irrationality in argument] to the extent that subjects get it wrong
. . . It could be argued that irrationality rather than rationality is the norm”
(Wason, 1983, p. 59). Or, in a similarly pessimistic vein: “One might draw
rather cynical conclusions . . . Human reasoning is fundamentally flawed”
(Reisberg, 1997, p. 469; see also Gilovich, 1991; Sutherland, 1992; Piatelli-
Palmarini, 1994).2 Samuels, Stich and Bishop (2002) claim that Kahneman
and Tversky have not made such radical claims. However, it must be said
that they did draw very similar inferences. With respect to the error called
the “law of small numbers”—the tendency to draw conclusions from very
small samples rather than heeding the rule that only large samples will be
representative of the population from which they are drawn—they have writ-
ten: “The true believer in the law of small numbers commits his multitude
of sins against the logic of statistical inference in good faith . . . His intu-
itive expectations are governed by a consistent misperception of the world”
(Tversky & Kahneman, 1971, p. 31). In another study on decision theory, they
have questioned the “favored position” of the “assumption of rationality” in
 70 Thomas Sturm

economics, allegedly treated by economists as “a self-evident truth, a reason-

able idealization, a tautology, a null hypothesis”. It was, they have claimed,
an assumption impervious to empirical falsification:

the assumption of rationality is protected by a formidable set of defenses

in the form of bolstering assumptions that restrict the significance of any
observed violation of the model. In particular, it is commonly assumed
that substantial violations of the standard model are (i) restricted to
insignificant choice problems, (ii) quickly eliminated by learning, or
(iii) irrelevant to economics because of the corrective function of market
Downloaded by [University of Chicago Library] at 12:08 09 August 2013

forces. (Tversky & Kahneman, 1986, p. S273)

II. Maybe we are not so irrational after all: experimental artifacts and
normative prejudices
Let us now look at how defenders of the bounded rationality approach in
psychology attack the HB approach.3 Two of the most important objections
here are the following:
(1) The results concerning the systematic deficiencies of human reasoning
are often, if not always, due to experimental artifacts.
(2) Rationality is bounded: Psychologists should not take traditional for-
mal rules of logic, probability theory and rational decision theory as
normatively unproblematic, as is done in the heuristics and biases tra-
dition. Rather, norms of reasoning are not valid independently of the
contents and contexts in which they are applied; and to figure out the
fitness between formal norms and contents and contexts of their use is
a matter of empirical research.4
In what follows, I shall argue that (1) is convincing, but that (2)—while it
contains important insights—also throws out some babies with the bathwater.
That there are experimental artifacts in the studies by Kahneman, Tversky
and others can again be shown by reference to the “Linda problem”. The cru-
cial point here is that the language used in these tests is by no means innocent,
and that the understanding of core terms in the task questions can be influ-
enced by prior information. In the “Linda problem”, Kahneman and Tversky
presuppose that the terms “probable” or “more probable than” and “and”
are all that counts when we test reasoning abilities. Moreover, they assume
that these terms have to be understood such that “and” is the logical “AND”
(&), and that “probable” conforms to principles of mathematical probability
theory.
However, ordinary subjects do not understand them in these ways, espe-
cially not within the context of the Linda problem (Fiedler, 1988; Hertwig
& Gigerenzer, 1999). The description of Linda as a politically sensitive per-
son, active in the antinuclear weapons movement, and so on, pushes subjects
 The “Rationality Wars” in Psychology 71

in the direction of certain interpretations of the statements (T) and (T&F).

Given the forced alternative of the test, i.e., that condition that subjects must
pick either “T” or “T&F”, many of them (20 to 50%) seem to infer that,
for instance, “T” means to exclude “F” (i.e., T = T&¬F). Other subjects
(10 to 20%), again, seem to understand “T&F” to mean “T→F”.5 Isn’t this
irrational? If you mean by “rational” the standard formal rules of logic, prob-
ability theory, statistics and decision theory, then yes. But, Gigerenzer argues,
that is not the only notion of rationality we have.6 Another notion includes
our ability to draw semantic7 inferences. In our society, there is something like
a correlation between having certain political attitudes and making choices
Downloaded by [University of Chicago Library] at 12:08 09 August 2013

for jobs. Bank tellers are not, in their overwhelming majority, people who
care all day long about social justice, equality between the sexes, and so on.
However, given the description of Linda, it seems implausible to many sub-
jects that Linda is not F; hence they pick T&F. Others, again, may understand
“T&F” roughly as “Well, if Linda has become a bank teller, then she is still
concerned about discrimination, social justice, etc., and so it is more probable
than not that she is a feminist too”. They might want to avoid the—indeed
dubitable—judgment that Linda cannot be a feminist anymore once she has
become a bank teller. Not only the logical and mathematical meaning of the
terms “probable” or “and” matter in this task. The description of Linda is
relevant as well, and reasonably so.
One can thus charitably reinterpret the violations of the conjunction rule
such that subjects apply a different rational rule, namely a conversational
rule (Hertwig & Gigerenzer, 1999, with reference to Grice’s (1975) “maxim
of relevance”). It is, therefore, not necessarily adequate to label their infer-
ence fallacious or irrational. (To indicate another example: Moral or altruistic
goals can reasonably ground decision making, even when the choices are
at odds with formal decision theory. This has been shown by violations of
“Property Alpha” or the “independence of irrelevant alternatives”; see Sen,
1993; Gigerenzer, 1998.)
Moreover, the alleged fallacies are avoidable, contrary to Kahneman and
Tversky’s claim. We must only represent them in more transparent and unam-
biguous ways. For instance, representing the “Linda problem” in terms of a
frequentist (rather than a subjective) interpretation of probability improves
performances dramatically, even for statistically untutored subjects. Take the
following task:

(Same description of Linda as before.)

There are 100 people who fit the description above. How many of them
are (a) bank tellers, (b) bank tellers and active feminists?

In the answers to this format, the conjunction fallacy has been reduced
from about 85% to 20% and less (Fiedler, 1988; Hertwig & Gigerenzer,
1999).8 Similar impressive results were achieved by using frequency formats in
 72 Thomas Sturm

Bayesian tasks, to the effect that subjects no longer overlooked the relevance
of base rates (Gigerenzer & Hoffrage, 1995).
The same points can be made, mutatis mutandis, with regard to other
alleged “cognitive illusions”. First, we can reinterpret data which apparently
reveal fallacious reasoning such that it becomes clear that subjects are perhaps
applying a rule which differs from the rule tested, but which may nevertheless
be rational. Secondly, by using more transparent representations of reasoning
tasks, even logically or statistically untutored subjects achieve correct answers
at a much higher rate.
As should be pointed out, both considerations indicate only possibilities:
Downloaded by [University of Chicago Library] at 12:08 09 August 2013

It is not shown in such studies that subjects actually use rules of logic, proba-
bility theory, statistics, or alternatives such as conversational rules, and so on.
Likewise, concerning the cases of avoiding “cognitive illusions” through more
transparent representation, Gigerenzer et al. have not argued that the mind
is actually a frequentist. To figure out what is actually going on is a rather
intricate matter. In any case, it is quite possible that we are relatively rational
after all, being able to learn rules of logic, probability theory etc. Since ought
implies can, and can appears to be possible for us humans here, there seems
to be no argument against the normative validity of those standard rules.

III. How far is rationality bounded?

However, Gigerenzer and with him many others9 do not think such modera-
tion is enough. According to them, there is a higher price to be paid for what
has been given so far. We must give up what Gigerenzer describes as “utopian”
dreams about rationality: it is extremely unlikely that our minds were made
to conform to truly universal norms of reasoning, and that the rules of logic,
probability theory, and decision theory are these norms. Instead, human
rationality is “bounded”.
What exactly does this mean? While there are various ways in which the
conception of bounded rationality is spelled out, it should be said first that
it does not mean that our reasoning abilities are limited and weak. This
would, after all, come down to the claims of the HB approach.10 In con-
trast to such views, Gigerenzer points to a metaphor used by Herbert Simon
(who coined the very terminology of “bounded rationality”): “Human ratio-
nal behaviour is shaped by a scissors whose two blades are the structure of
task environments and the computational capabilities of the actor” (Simon,
1990, p. 7). In other words, one cannot leave out either the reasoning abil-
ity or the environment and still hope to adequately explain specific reasoning
processes. Moreover, this is also normatively important. Instead of viewing
standard formal rules of logic etc. as normatively valid (and therefore as
appropriate tools for the psychological investigation of human reasoning),
empirical arguments—specifically, arguments concerning the “contents and
contexts” in which a certain rule works and those in which it does not—can
and should be used to assess the validity of rules of reasoning.11 So, the
 The “Rationality Wars” in Psychology 73

claim that rationality is bounded is sometimes meant descriptively and some-

times normatively; and it is sometimes meant as a statement about reasoning
processes or activities, and sometimes about rules of reasoning. The philo-
sophically most challenging version of the BR approach would obviously be
to claim that rules of reasoning are bounded, and that this is not merely a
descriptive but also a normative claim. This contention is also at the center of
the following considerations, in which I shall argue for certain limitations of
the BR approach.
I present three kinds of considerations. First, I point out that Gigerenzer’s
statements about the content-and-context dependency come in different vari-
Downloaded by [University of Chicago Library] at 12:08 09 August 2013

eties, some of which are acceptable, but not strong enough to support the BR
approach in its rejection of formal rules as norms of rationality. Second, I
consider some empirical studies of reasoning in order to show that arguments
undermining the idea of a content-and-context-independent validity of rules
are sometimes questionable. Third, I argue that fast and frugal heuristics and
formal rules of logic, probability theory etc. play different and interlocking
roles in a comprehensive theory of rationality.
As to the first point, consider the following general statements by
Gigerenzer:
(1) “I argued that psychological principles are indispensable for defining and
evaluating what sound judgment is. Axioms and rules from probabil-
ity theory and logic are, by themselves, indeterminate.” (1998, p. 464;
emphasis added)
(2) “The point I wish to defend . . . is that formal axioms and rules can-
not be imposed as universal yardsticks of rationality independent of
social objectives, norms and values; they can, however, be entailed
by certain social objectives, norms and values. Thus, I am not argu-
ing against axioms and rules, only against their a priori imposition as
context-independent yardsticks of rationality.” (1996b, p. 320; emphasis
added)
(3) “My thesis is that traditional axioms and rules are incomplete as
behavioral norms in the sense that their normative validity depends on
the social context of behavior, such as social objectives, values, and
motivations.” (Gigerenzer 1996b, p. 319; emphasis added)
These passages are not identical in meaning. Statement (1) is the weakest
and can be accepted: Rules of logic or probability theory do not themselves
contain rules or criteria for their correct application. In fact, no rule can, since
that would lead into an infinite regress, as already no one less than Kant has
pointed out (1781/1787, pp. A132–34/B171–74). Likewise, it is not true that
because rules of logic or probability theory can be proven within formal cal-
culi, they are therefore to be viewed as norms of good reasoning, because it
is open how formal rules have to be mapped onto concrete reasoning tasks
(see Goldman, 2008; Grice, 2001). Whether a rule should be applied to a cer-
tain task depends not only on the rule itself; we also need criteria for deciding
 74 Thomas Sturm

which rule out of a number of possible candidates should be applied to a

given task. This may well depend on the subject’s cognitive access to rules, his
understanding of the task, his resources, the significance the task possesses (or
should possess) for him, and so on. For all these issues, psychological knowl-
edge is relevant. Being blind about this, and trying to mechanically apply
certain formal norms to certain reasoning tasks, is a real weakness of the
heuristics and biases approach. Still, let us not confuse, in general, questions
of validity with questions of application.12 In other words, if the BR approach
would come down merely to claim (1), it would be restricted to determining
conditions or criteria that tell us which rule is to be applied to a given problem.
Downloaded by [University of Chicago Library] at 12:08 09 August 2013

Claim (2) states that the justification of a logical or probabilistic rule may be
given by reference to the fact that it is entailed by certain values or objectives.
Consider also how formal rules can be entailed by convincing instances of
reasoning, as in the frequentist versions of the Linda problem or in Bayesian
tasks. However, if viewing rules as being entailed by certain goals or success-
ful instances of reasoning is meant to show that these rules are justified, this is
a strategy that has nothing to do with the idea of bounded rationality or with
validating norms by means of relating them to contents and contexts of rea-
soning, or to the environment of reasoners. Rather, it is a strategy that takes
convincing instances of reasoning, showing what general rules are embodied
in these instances, and then perhaps reflecting on whether these rules work in
other instances as well, revising them if necessary, and so on. This is the well-
known reflective equilibrium strategy (see e.g. Cohen, 1981). This is a serious
option, but it does not entail that the validity of the norms as universal or
formal rules must somehow be given up in favor of a bounded conception of
reasoning norms.
Thus, only statement (3) expresses the view that the normative validity
of rules depends on certain contents and contexts of reasoning—a much
stronger claim than (1) or (2), one which, as I argue next, is not supported by
the way Gigerenzer and others have attacked questionable uses of formal rules
in the HB approach’s studies on the Linda problem or the Wason Selection
Task.
Consider first that there is an inconsistency in Gigerenzer’s views con-
cerning the Linda problem. He claims both that (a) answering T&F is not
“a violation of the major view of probability, the frequentist conception”
(1991, p. 92) and that (b) a representation of the test in frequency terms
“makes the ‘conjunction fallacy’ largely disappear” (ibid., p. 96) or causes
“more correct answers” (1996a, p. 594, cf. p. 595). It seems to me that (b) is
correct, for the reasons given by Gigerenzer himself. However, then answer-
ing that “T&F” is more probable than “T” is a violation of the conjunction
rule, which is inconsistent with (a). Representing tasks in frequency formats
improves performance; but saying this only makes sense if the two different
representations of the task—one in terms of subjective probability, one in fre-
quentist terms—concern the same rule. Gigerenzer must admit something like
 The “Rationality Wars” in Psychology 75

that as well. Even though performance can be drastically improved by using

frequency formats, he points out that still as many as 20% of the subjects
do not follow the conjunction rule even under the improved experimental
design. Because the other 80% of responses are regarded as improvements
or as correct answers, the 20% must be on the side of those who still err.
One might object to this that we should not assume the conjunction rule
to be the one and only rule that matters in the Linda problem. The conversa-
tional “maxim of relevance” might be a reasonable alternative. But then what
is at stake here is, once again, not the validity of the conjunction rule, or its
formal, content-independent validity, but whether it is the rule to be applied
Downloaded by [University of Chicago Library] at 12:08 09 August 2013

in this task. Whether reasoners should use the rule in any given reasoning task
may—and probably does—depend on the context and content of the task in
question, but that does not show that the validity of the rule does too.
Another problem may be illustrated with regard to the treatment of the
Wason Selection Task by Leda Cosmides and John Tooby, accepted by
Gigerenzer and others as an important argument in favor of the concep-
tion of bounded rationality (Gigerenzer, 1991; Gigerenzer, 1998; Gigerenzer
& Hug, 1992). Cosmides and Tooby’s crucial point is that while subjects
appear to reason badly when checking material implications with descrip-
tive or indicative contents they improve when it comes to implications with
deontic contents such as “cheater detection” (Cosmides, 1989; Cosmides &
Tooby, 1996). Take a conditional such as “If someone is under 18 years old,
(s)he is required to drink coke (a non-alcoholic beverage)”, where the conse-
quent is a deontic statement. Here, subjects performed much better than in
tasks where P and Q in P→Q were replaced by entirely indicative contents.
(Subjects did somewhat better when the indicative conditionals were concrete
rather than abstract, but they did still much better with deontic conditionals.)
In the deontic conditionals, the large majority of subjects easily detected the
relevance of instances of ¬Q—whisky drinkers, say—as potential violators of
the norm.
What explains this accuracy in cheater detection? Put briefly, Cosmides and
Tooby’s account goes as follows. At some point of in the days of hunting and
gathering, social cooperation and exchange became advantageous and even a
distinctive step of the evolution of homo sapiens. Only few species have devel-
oped the ability to cooperate in ways that are expressive of genuine sociability.
Cosmides and Tooby support this claim with reference to the work of Trivers
(1971) and others (e.g., Axelrod & Hamilton, 1981; Axelrod, 1984; Maynard
Smith, 1982) concerning evolutionary explanations of reciprocal altruism and
cooperation. Such behavior can only constitute evolutionary stable strategies
if the organisms possess the ability to spot those who cheat. Beings with-
out such a capacity would be open to exploitation, and be selected out. This
selection pressure, according to Cosmides and Tooby, has led to the evolution
of a cheater detection module. It is successful in the area of such condition-
als, with statements with these contents or purposes. There was no selection
 76 Thomas Sturm

pressure to understand all conditional statements properly, let alone to under-

stand them along the lines of the material implication. This is why we are such
fools in the original Wason Selection Task. According to these considerations,
good reasoning depends upon the content and context of the task; in this case,
upon specific goals we pursue due to our evolutionary adaptation.
However, this argument is problematic. It is an open matter whether sub-
jects succeed so often because they have a specific cheater-detection ability.
Fodor (2000) has argued persuasively that deontic conditionals are not suf-
ficiently similar to indicative ones to make the case. In the latter case, what
is asserted, and what has to be checked, is P→Q as a whole. By contrast,
Downloaded by [University of Chicago Library] at 12:08 09 August 2013

in the former case the requirement is contained merely in the consequent of

the conditional, Q (the drinking part). The antecedent, P, merely specifies to
whom the requirement applies (those under 18 years). This is a matter of fact
about which there can be no sensible requirement. Subjects must, but also
easily can, spot non-coke drinkers as potential violators. For this, they merely
need to apply the Law of Non-contradiction, ¬(Q &¬Q), which is much eas-
ier than to use the Law of Contraposition, ((P→Q) & ¬Q) → ¬P needed in
the case of indicative conditionals. Hence, the ability to reason successfully in
cheater detection isn’t necessarily due to the content (the goal), but might be
explained by reference to a formal feature of the reasoning.
This consideration leaves open whether or not the apparently poor rea-
soning with indicative conditionals is irrational or not. Like in the Linda
problem, perhaps subjects employ a different rule which may be reasonable as
well. It has been proposed that, in the Wason Selection Task, subjects might
understand the test sentence such that it does not refer to the four cards only,
but as an empirical generalization: “For all x, if Fx→Gx”. For instance, they
might consider checking whether all swans are white, and do so by looking at
swans and white things, leaving out thereby all non-white things. What do you
think needs more time, checking all the white things or all non-white things?
(Oaksford & Chater, 1994; Botterill & Carruthers, 1999, p. 125) If that is so,
then subjects might be viewed as in some sense rational. The difference to
the Linda case is that here, the material conditional might still be the rule
to be applied when it comes to the ultimate justification of the statement (as
opposed to the pursuit of a hopefully less costly way for the selection of data).
But even though subjects are, in this case, not perfectly rational, that by itself
does not undermine the universal validity of the material conditional.
One might object at this point that the material conditional is a rather ques-
tionable thing. It is well-known that the rules for material implication run into
paradox. A statement such as “If Rome is the capital of Mongolia, then the
FC Barcelona will win the Champions League seven times in a row” is true
simply because Rome isn’t the capital of Mongolia. But the validity of the
rules of material implication remains a matter for logicians, not for psychol-
ogists. The interesting question—and here Gigerenzer’s opposition to a blind
use of formal rules within psychology is quite in place—is why psychologists
 The “Rationality Wars” in Psychology 77

like Wason thought that a rule that is a cause for serious logical troubles
should be used for purposes of testing the rationality of human subjects.
An important caveat needs to be added here. What I have said so far applies
only to examples like those mentioned. Gigerenzer and his colleagues, how-
ever, have developed a whole battery of rules of bounded rationality, so-called
“fast and frugal heuristics” such as the “recognition heuristic”, the “fluency
heuristic”, or “take the best”. This is the most important and novel part of the
BR approach. These heuristics are adequate for problems characterized by
inevitable uncertainty, very little information, or very little available solution
time (e.g., Gigerenzer, Todd & the ABC Research Group, 1999). It is typical
Downloaded by [University of Chicago Library] at 12:08 09 August 2013

that these problems are often more realistic reasoning and decision problems
than the textbook examples I have concentrated on so far. Logic or probabil-
ity theory may not be very helpful in such cases. For instance, by using the
recognition heuristic—“If one of two objects is recognized and the other is
not, then infer that the recognized object has the higher value with respect to
the criterion”—we are able to draw highly accurate inferences about, say, the
sizes of cities or winners in sports tournaments. Surprisingly, if we recognize
only names of objects, we do better than if we have more information about
them (Gigerenzer & Goldstein, 1996). Rules for such problems often depend
on empirical knowledge, and are normatively valid only relative to certain
domains or contexts. In the case of the recognition heuristic, this rule works
if and insofar as recognition of city names correlates with the criterion. When
there is no such correlation, judgments are wrong more often, and in such
circumstances the heuristic should not be used.
I pointed out that the BR approach sometimes looks (and is characterized)
as if it were about adequate reasoning processes, about determining which
rules to apply to a given task or problem rather than about determining and
justifying the rules themselves which we ought to follow in reasoning. Given
the existence and normative validity of fast and frugal heuristics, it is clear
that the program isn’t merely about such issues. It also develops for large
numbers of problems and tasks, rules to be followed. But the fact that there
are such norms of rationality does not show that there are no others.
This point brings me to my third and final consideration about limita-
tions of the BR approach. It is perhaps tempting to infer from the foregoing
arguments that not only is the psychology of reasoning deeply fragmented,
but so is very concept of rationality. Our best instances of good reasoning
may bear no more than a family resemblance to one another, some being
guided by bounded norms, others by norms that are formal or strictly uni-
versal. However, this is probably incorrect, and not simply because we might
be tempted to ask back: “Well, why is it that we call all of these different
rules ‘norms of rationality’? They must have something in common after all!”
Rather, I think that the two kinds of norms do, at least partly, play function-
ally different roles, and that they partly interlock. For instance, formal rules
such as Bayes’ theorem or principles of optimization (such as maximizing
 78 Thomas Sturm

expected utility) continue to play an important role even within attempts

to construct fast and frugal heuristics, namely as a normative standard.
A heuristic of this kind can only be normatively recommendable if it com-
petes successfully with such formal rules—if it leads to correct or convincing
results at least as often as they do, or perhaps even outperforms them. Also,
one cannot even formulate certain fast and frugal heuristics without using
some basic concepts and tools of formal logic. If reasoners are to consciously
use the recognition heuristic, they has to have a minimal grasp of the con-
ditional, “if–then” structure of the heuristic. Other rules, such as “take the
best” require the ability to master disjunction, and so on. In this way, some
Downloaded by [University of Chicago Library] at 12:08 09 August 2013

formal rules are built into heuristics. They are the cement that holds the build-
ing blocks of heuristics together. However, the BR approach also contains a
fundamentally correct and important insight: because reasoning often has to
proceed on the basis of very little information and large amounts of uncer-
tainty, it makes little sense to expect logic or probability theory alone to be
sufficient in a comprehensive normative theory of rationality. Heuristics can
and should complement formal rules here.
Thus, the picture of rationality emerging from this might help overcoming
the deep—and partly heated—debates in psychology discussed in this essay,
at least in part. It does not do so by showing that both research programs have
overplayed their cards, or that the differences are merely rhetorical while both
programs share the same core assumptions about how far human beings are in
fact rational (as argued by Samuels, Stich & Bishop 2002). My point concerns
the different issue of the normative assumptions embodied in the programs.
Without neglecting their basic differences, we can try to make them coherent
by assigning different, interlocking functions to them. Certainly this will have
to be discussed in closer detail. Could a more lasting peace emerge from this? I
don’t know. But I hope the idea seems worth more research and collaboration
between philosophy and psychology.13

Notes
1. For instance, they touch on debates about the possibilities and problems of naturalis-
tic approaches to epistemology (e.g., Kitcher, 1992; Stein 1996, pp. 14–18). At the same
time, epistemologists have so far taken little if any notice of the psychological debate
about human rationality (exceptions being Bishop, 2008; Bishop & Trout, 2005; Goldman,
2008). This needs to change. Within the space of this essay, I cannot discuss the conse-
quences of the “rationality wars” for naturalizing epistemology. But I hope to lay some
groundwork for doing so.
2. The pessimism about human rationality that comes out of the HB approach has also
worried naturalistic philosophers (e.g. Kornblith, 1992), though not all (e.g., not an earlier
Stich, 1980; also, Bishop & Trout, 2005, try to exploit the HB approach positively for their
own normative naturalism). Again, I must leave this interesting issue for another occasion.
3. There have been a number of critical discussions of the HB approach by philosophers as
well, esp. Cohen (1981), Stein (1996), see also Botterill & Carruthers (1999), Cherniak
(1986), Schumacher (2002). It would lead too far to discuss them here, however.
 The “Rationality Wars” in Psychology 79

4. There are yet further objections from this camp against the HB approach. For instance,
Gigerenzer (1996b) argues that the language of biases and heuristics is not really
explanatory: talk of a “representative heuristic” or an “availability heuristic” are mere
redescriptions of the phenomena. In other words, the HB approach does not really deliver
explanations. Samuel, Stich and Bishop’s (2002) attempt to argue that the differences
between the “heuristic and biases” and the “bounded rationality” approaches are merely
superficial overlooks this important point.
5. As Gigerenzer remarks: “Recent studies using paraphrasing and protocols suggest that
participants draw a variety of semantic inferences to make sense of the Linda problem:
Some 10 to 20% seem to infer that and should be read as a conditional, and some 20 to
50% seem to infer that the alternative ‘Linda is a bank teller’ implies that she is not active
in the feminist movement. . . . These semantic inferences can lead to choosing T&F rather
Downloaded by [University of Chicago Library] at 12:08 09 August 2013

than T” (1996a, p. 593). I have corrected the last sentence in the quotation. It origi-
nally reads: “These semantic inferences can lead to choosing T rather than T&F.” This
is confused, as Gigerenzer has agreed in personal communication.
6. Gigerenzer often points out that there are serious debates within logic, statistics, and so on
over which formal system is correct, which notion of probability adequate, etc. I cannot
enter this side of his “reject the norm” argument here. See Vranas, 2000, 2001; Gigerenzer,
2001.
7. I ignore here whether it is best to call these inferences “semantic”.
8. Elimination of polysemy helps as well, but not as much as the use of frequentist language,
as Gigerenzer has pointed out to me.
9. For instance, Cosmides and Tooby in their approach of an evolutionary psychology of
rationality hold similar views. It should be mentioned that Gigerenzer’s BR approach is
not committed to an evolutionary account of human rationality. For instance, it does not
accept the strong modularity thesis that evolutionary psychologists maintain (Cosmides,
1989; Cosmides & Tooby, 1996). That the two views need to be distinguished is overlooked
by Samuels, Stich and Bishop (2002).
10. Indeed, defenders of the HB approach have understood the concept of bounded rational-
ity in this way (e.g., Gilovich, 1991; Kahneman, 2003).
11. E.g., “ . . . on Kahneman and Tversky’s (1996) view of sound reasoning, the content of
the Linda problem is irrelevant. All that counts are the terms probable and and, which
the conjunction rule interprets in terms of mathematical probability and logical AND,
respectively. In contrast, I believe that sound reasoning begins by investigating the content
of a problem to infer what terms such as probable mean” (Gigerenzer, 1996a, p. 593).
12. I say “let us not generally confuse” them because there are cases of rules where the ques-
tion of justification and application are not easily distinguishable. This holds especially for
the “fast and frugal heuristics” of the BR approach. See Gigerenzer and Sturm (2011).
13. I thank Uljana Feest and Gerd Gigerenzer for valuable comments and criticisms. Cynthia
Klohr made helpful suggestions for wording the text. Completion of this essay was
supported by the Spanish Ministry for Science and Innovation, Reference number FFI
2008-01559/FISO.

References
Bishop, M.A. & Trout, J.D. (2005) Epistemology and the Psychology of Human Judgment
(New York: Oxford University Press).
Bishop, M. (2008) “Reflections on a normative psychology”, in: A. Beckermann & Sven Walter
(Eds.), Philosophie: Grundlagen und Anwendungen, pp. 249–62 (Paderborn: Mentis).
Botterill, G. & Carruthers, P. (1999) “Reasoning and irrationality”, in: G. Botterill &
P. Carruthers, The Philosophy of Psychology, pp. 105–30. (Cambridge: Cambridge
University Press).
 80 Thomas Sturm

Casscells, W., Schoenberger, A. & Graboys, T.B. (1978) “Interpretation by physicians of clinical
laboratory results”, New England Journal of Medicine, 299, pp. 999–1001.
Cherniak, C. (1986) Minimal Rationality (Cambridge, MA: MIT Press).
Cohen, L.J. (1981) “Can human irrationality be experimentally demonstrated?”, Behavioral and
Brain Sciences, 4, pp. 317–31 (comments and responses, pp. 331–59).
Cosmides, L. (1989) “The logic of social exchange: Has natural selection shaped how humans
reason?”, Cognition, 31, pp. 187–276.
Cosmides, L. & Tooby, J. (1996) “Are humans good intuitive statisticians after all?”, Cognition,
58, pp. 1–73.
Evans, J. St.B. & Over, D.E. (1996) Rationality and Reasoning (Hove: Psychology Press).
Fiedler, K. (1988) “The dependence of the conjunction fallacy on subtle linguistic factors”,
Psychological Research, 50, pp. 123–29.
Downloaded by [University of Chicago Library] at 12:08 09 August 2013

Fodor, J. (2000) “Why we are so good at catching cheaters”, Cognition, 75, pp. 29–32.
Gigerenzer, G. (1991) “How to make cognitive illusions disappear: Beyond heuristics and biases”,
European Review of Social Psychology, 2, pp. 83–115.
Gigerenzer, G. (1993): “The bounded rationality of probabilistic mental models”, in: K.I.
Manktelow & D.E. Over (Eds.), Rationality: Psychological and Philosophical Perspectives,
pp. 284–313 (London: Routledge).
Gigerenzer, G. (1996a) “On narrow norms and vague heuristics: A rebuttal to Kahneman and
Tversky”, Psychological Review, 103, pp. 592–96.
Gigerenzer, G. (1996b) “Rationality: why social context matters”, in: P. B. Baltes & Ursula M.
Staudinger (Eds.), Interactive Minds, pp. 319–46 (New York: Cambridge University Press).
Gigerenzer, G. (1998) “Psychological challenges for normative models”, in: D. M. Gabbay &
P. Smets (Eds.), Handbook of Defeasible Reasoning and Uncertainty Management Systems,
vol. 1., pp. 441–67 (Dordrecht: Kluwer).
Gigerenzer, G. (2000) Adaptive Thinking (New York: Oxford University Press).
Gigerenzer, G. (2001) “Content-blind norms, no norms, or good norms? A reply to Vranas”,
Cognition, 81, pp. 93–103.
Gigerenzer, G. (2004) “The irrationality paradox”, Behavioral and Brain Sciences, 27, pp. 336–38.
Gigerenzer, G. & Goldstein, D. G. (1996) “Reasoning the fast and frugal way: Models of bounded
rationality”, Psychological Review, 103, pp. 650–69.
Gigerenzer, G. & Hoffrage, U. (1995) “How to improve Bayesian reasoning without instruction:
Frequency formats”, Psychological Review, 102, pp. 684–704.
Gigerenzer, G. & Hug, K. (1992) “Domain-specific reasoning: Social contracts, cheating, and
perspective change”, Cognition, 43, pp. 127–71.
Gigerenzer, G. & Sturm, T. (2011) “How (far) can rationality be naturalized?” Synthese. DOI:
10.1007/s11229-011-0030-6.
Gigerenzer, G., Todd, P.M. & the ABC Research Group (1999) Simple Heuristics That Make Us
Smart (New York: Oxford University Press).
Gilovich, T. (1991) How We know What Isn’t So: The Fallibility of Human Reason in Everyday
Life (New York: Free Press).
Gilovich, T., Griffin, D., & Kahneman, D. (2002) Heuristics and Biases: The Psychology of
Intuitive Judgment (New York: Cambridge University Press).
Goldman, A. (1986) Epistemology and Cognition (Cambridge, MA: Harvard University Press).
Goldman, A. (2008). Human rationality: Epistemological and psychological perspectives”, in:
A. Beckermann & S. Walter (Eds.), Philosophie: Grundlagen und Anwendungen, pp. 230–47
(Paderborn: Mentis).
Grice, P. (1975) “Logic and conversation”, in: P. Cole & J. L. Morgan (Eds.), Syntax and
Semantics 3: Speech Acts, pp. 41–58 (New York: Academic).
Grice, P. (2001) Aspects of Reason (Oxford: Clarendon Press).
Hertwig, R. & Gigerenzer, G. (1999) “The ‘conjunction fallacy’ revisited: How intelligent
inferences look like reasoning errors”, Journal of Behavioral Decision Making, 12, pp.
275–305.
 The “Rationality Wars” in Psychology 81

Kahneman, D., Slovic, P. & Tversky, A. (Eds.) (1982) Judgment under Uncertainty: Heuristics and
Biases (New York: Cambridge University Press).
Kahneman, D. & Tversky, A. (1996) “On the reality of cognitive illusions”, Psychological Review,
103, 582–91.
Kahneman, D. (2003) “Maps of bounded rationality: Psychology for behavioral economics”,
American Economic Review, 93, pp. 1449–75.
Kant, I. [1781/1787] (1998): Kritik der reinen Vernunft, Jens Timmermann (Ed.) (Hamburg:
Meiner).
Kitcher, P. (1992) “The naturalists return”, Philosophical Review, 101, pp. 53–114.
Lopes, L.L. (1991) “The rhetoric of irrationality”, Theory & Psychology, 1, pp. 65–82.
Lopes, L.L. (1992) “Three misleading assumptions in the customary rhetoric of the bias
literature”, Theory & Psychology, 2, pp. 231–36.
Downloaded by [University of Chicago Library] at 12:08 09 August 2013

Mellers, B. Hertwig, R. & Kahneman, D. (2001) “Do frequency representations eliminate

conjunction effects? An exercise in adversarial collaboration”, Psychological Science, 12,
pp. 269–75.
Nisbett, R.B. & Borgida, E. (1975) “Attribution and the psychology of prediction”, Journal of
Personal and Social Psychology, 32, pp. 932–43.
Nisbett, R.B. & Ross, L. (1980) Human Inference: Strategies and Shortcomings (Englewood Cliffs,
NJ: Prentice Hall).
Oaksford, M. & Chater, N. (1994) “A rational analysis of the selection task as optimal data
selection”, Psychological Review, 101, pp. 608–31.
Piattelli-Palmarini, M. (1994) Inevitable Illusions: How Mistakes of Reason Rule Our Minds
(New York: Wiley).
Reisberg, D. (1997) Cognition: Exploring the Science of the Mind (New York: W. W. Norton).
Samuels, R., Stich, S. & Bishop, M. (2002) “Ending the rationality wars: How to make dis-
putes about human rationality disappear”, in: R. Elio (Ed.), Common Sense, Reasoning and
Rationality, pp. 236–68 (New York: Oxford University Press).
Schumacher, R. (2002) “Wie irrational können Personen sein?”, Zeitschrift für philosophische
Forschung, 56, pp. 22–47.
Sen, A. 1993: “Internal consistency of choice”, Econometria, 61, pp. 495–521.
Simon, H. A. (1990) “Invariants of human behavior”, Annual Review of Psychology, 41, pp. 1–19.
Stein, E. (1996) Without Good Reason: The Rationality Debate in Philosophy and Cognitive
Science (Oxford: Clarendon).
Stich, E. (1980) “Could man be an irrational animal? Some notes on the epistemology of
rationality”, Synthese, 64, pp. 115–35.
Sutherland, S. (1992) Irrationality (London: Pinker & Martin).
Tversky, A. & Kahneman, D. (1971) “Belief in the law of small numbers,” Psychological Bulletin,
2, pp. 105–10. (Reprinted in Kahneman, Slovic & Tversky, 1982, pp. 23–31.)
Tversky, A. & Kahneman, D. 1974: “Judgment under uncertainty: Heuristics and biases”,
Science, 185, pp. 1124–31.
Tversky, A. & Kahneman, D. (1983) “Extensional versus intuitive reasoning: Conjunction fallacy
in probability judgment”, Psychological Review, 90, pp. 293–315.
Tversky, A. & Kahneman, D. (1986) “Rational choice and the framing of decisions”, The Journal
of Business, 59, pp. S251–S278.
Vranas, P.B.M. (2000) “Gigerenzer’s normative critique of Kahneman and Tversky”, Cognition,
76, pp. 179–93.
Vranas, P.B.M. (2001) “Single-case probabilities and content-neutral norms: A reply to
Gigerenzer”, Cognition, 81, pp. 105–11.
Wason, P.C. (1966) “Reasoning about a rule”, Quarterly Journal of Experimental Psychology, 20,
pp. 273–81.
Wason, P.C. (1983) “Realism and rationality in the selection task”, in: J. St. B. T. Evans (Ed.),
Thinking and Reasoning, pp. 44–75 (London: Routledge & Kegan Paul).