teorema

Vol. XXXVIII/1, 2019, pp. 00-00

ISNN 0210-1602
[BIBLID 0210-1602 (2019) 38:1; pp. 00-00

Replies to Critics

Hugo Mercier and Dan Sperber

We are very thankful to our colleagues who have provided such

thoughtful and constructive discussions of our book, The Enigma of Rea-
son. Since these commentaries each raise a different set of issues, we re-
spond to them one by one.

KAROLINA PROCHOWNIK raised three questions, each a great invitation

for further elaboration. Could reason serve some other social function,
she asks, besides argumentation and justification? She suggests two ways
in which this could be the case.
Prochownik starts by suggesting that “reason could help people to
decide with whom to make alliances and facilitate the recognition of
group members. In particular, the evaluation of reasons provided by
others to justify their views or actions, in addition to testing their quality
and the level of commitment of the reasoner, could serve to filter indi-
viduals who really share the audience`s views from those who don’t.”
We agree that reasons can be used to signal one’s ideological or, more
generally, one’s coalitional allegiances. However, we would say that the
cognitive mechanisms that influence what contents we express as a func-
tion of how these contents inform others of our allegiances are distinct
from reason per se. The same mechanisms would be at play whether we
use reasons, plain statements of opinions, reactions of approval or dis-
approval to opinions expressed by others, or any other communicative
act (such as wearing a religious or political sign) to serve this end. As a
result, the use of reasons to communicate allegiances might not have di-
rectly influenced the evolution of our faculty of reason, and thus would
not be part of its function in the evolutionary sense of the term, even if it
can, on occasions, be a goal or even the main goal, of putting forward
reasons.

1
2 Hugo Mercier and Dan Sperber

In the same vein, Prochownik mentions that “when alliances or

groups are already formed, public production and sharing of reasons can
help to shape group identity (as built upon shared views, behaviors and
reasons to hold or perform them), decide common goals, and motivate
joint action.” Again, we agree with her that such goals can be served by
producing reasons. As Prochownik points out, the fact that reason can
be productively used to define group identity even by closely knit groups
who “share views on moral, political, religious or any other matters”
seems to go against our suggestion that such groups are prone to polari-
zation. Actually, there is no contradiction. The exchange of reasons in a
like-minded group can bear, for instance, on practical issues that aren’t
agreed on. In such a case, it should not lead to polarization (for an
extended discussion, see Mercier, submitted). One of our favorite exam-
ples — continuing on the theme of the abolition of slavery broached in
The Enigma of Reason — is that of a group of twelve Quakers who would
prove formidably influential in defending the abolition of the slave trade
in the U.K. As described by Hochschild, these “twelve were of one
mind”, and, as a result, they could easily have become polarized to the
point, say, of defending the immediate abolition of slavery and emanci-
pation of all slaves — a goal unfortunately rather unrealistic at the time.
Instead, their meetings led them to focus on more realistic goals, such as
the end of the slave trade in the U.K. One possible explanation for the
efficiency of argumentation even in conditions that seem propitious for
polarization, is that, while the twelve Quakers agreed on a common ulti-
mate goal, they disagreed on the best way to reach it. As a result, argu-
mentation played a useful role in helping them work out the best means
to reach the agreed upon end.
Exchange of arguments might also have reminded these twelve
Quakers of how right their cause was — even if they all agreed on that
— and bolstered their morale. This effect of the exchange of arguments
might be closer to what Prochownik has in mind. We would say that, as
in the case of reasons used to mark our allegiances, the use of reasons to
bolster group morale would not be specific to reasons and would thus not
have borne directly on the evolution of the faculty of reason, even if bol-
stering morale can certainly be the proximal function of some reasons.
Prochownik’s second suggestion is that “the evaluation of reasons
provided by a speaker [might be] sensitive to the specific features of the
social context the reasons are communicated in (e.g., who is reasoning,
whom is this person to us, what is the record of our past interactions,
what kind of interactions we aim for with that person in the future).”

teorema XXXVIII/1, 2019, pp. 00-00

 Précis of The Enigma of Reason 3

Our predictions on that point would be as follows. The status of the

speaker — are they someone we respect, etc. — should influence how
much effort we’re willing to put in understanding their reasons. Howev-
er, for a given amount of effort attributed to processing some reasons,
the output of the evaluation process should be similar regardless of the
source of the reason, a point classically made by pointing out that a
mathematical proof put forward by a scoundrel is just as convincing to a
competent audience as it would be if put forward by a paragon of hones-
ty. We have attempted to test this hypothesis by providing people with
equally strong arguments coming from sources that were either highly
trusted or highly distrusted. This manipulation did not seem to affect
participants’ evaluation of the arguments [Trouche, Shao, & Mercier
(2019)]. We would be tempted to reconcile this theorizing, and these pre-
liminary results, with Prochownik’s observation that “people may be
more likely to accept or reject other people`s reasons when they have
other social reasons to do so” by pointing out that reason’s verdict on
whether a given reason is good or not is not the only factor that affects
our decision to overtly agree or disagree with this reason. An employee
might opine to their boss’s asinine reason. A political party member
might vocally reject arguments challenging their party’s platform, even if
they are internally bothered by their cogency.
Prochownik’s comment also raises an interesting question not
about reason itself but about theories of reason. She asks “why the intel-
lectualist theories of reason have been so culturally successful and re-
sistant to being refuted for around 2000 years?” We’re obviously hardly
the first to stress the importance of the social dimension of reason [see,
e.g. Billig (1996); Gibbard (1990); Haidt (2001); Kuhn (1991); Perelman
& Olbrechts-Tyteca (1958); Piaget (1928)], but it seems that more indi-
vidualistic approaches dominate Western psychology and philosophy. If
we assume that something along the lines of our interactionist theory is
indeed the right take on reason, then one of its predictions should be
that “wrong theories … gradually disappear in the process of critical
evaluation and discussion by the scientific community,” and thus that
some version of the interactionist theory should become dominant. Well,
we are hopeful that this might come to pass!
We don’t have a worked out explanation for the success of intellec-
tualist theories (again, assuming they are essentially mistaken). Still one
relevant empirical factor is that humans are good at argumentation but
bad at thinking about argumentation in the abstract. People consistently
underestimate the efficiency of the exchange of arguments taking place

teorema XXXVIII/1, 2019, pp. 00-00

4 Hugo Mercier and Dan Sperber

in group discussions [Mercier, Trouche, Yama, Heintz, & Girotto

(2015)]. This is an instance of a more general difficulty humans have
with thinking abstractly about the reliability of socially distributed infor-
mation. They are very good at aggregating information from different
social sources — taking into account how many people support an opin-
ion, how competent and benevolent these people are, etc. — but they
are terrible at thinking about the aggregation of information from differ-
ent social sources [Mercier (in press)]. For instance, participants behave
nearly optimally when they have to update their answer to a problem as a
function of how many other participants have given a different answer
[Morgan, Rendell, Ehn, Hoppitt, & Laland (2012); for review, see Mercier
& Morin (submitted)]. However, when presented with an abstract voting
situation in which we expect the Condorcet Jury Theorem to apply, partic-
ipants don’t believe that the majority of a group is more likely to be right
than its average member ([Mercier, Dockendorff, & Schwartzberg,
(submitted)]. We shouldn’t particularly expect participants to have an ab-
stract understanding of the epistemic benefits of information aggregation
mechanisms. However, the fact that they appear to consistently underes-
timate the efficacy of majority rule, averaging, and argumentation might
help explain why theories that stress the importance and the value of the
social exchange of ideas have proven less convincing than they arguably
should have.

LAURA MACCHI and MARIA BAGASSI [L&M] comment our book from a
perspective inspired by the pioneering work of Giuseppe Mosconi
(1990), a perspective to the development of which they have contributed
in a series of original papers. There are interesting commonalities be-
tween their perspective and ours. In particular, we share the view that
the role of logic in thinking and reasoning has been overestimated and
that the role of language with its conceptual richness and pragmatic re-
sources has been underestimated. Manifestly there are also important dif-
ferences. We focus our discussion on a couple of clear disagreements,
one fairly general, the other about the interpretation of interesting exper-
imental work with the horse-trading task [Maier & Solem (1952)].
M&B object to our claim that in argumentation, logic is mainly
used for rhetorical purposes, to streamline and highlight the structure of
non-demonstrative arguments, rather than to try and give a genuine logi-
cal demonstration of their conclusions. Their objections to our claim,
however, are not about the rhetoric of argumentation but about com-
plexity of thought. To the extent that we understand it properly, this ob-

teorema XXXVIII/1, 2019, pp. 00-00

 Précis of The Enigma of Reason 5

jection is grounded in a genuine disagreement that M&B do not articu-

late: they see talk of thinking, reasoning, and argumentation as address-
ing the same range of phenomena from overlapping perspective. We
argue at length in our book that there is a distinct, typically human men-
tal mechanism that performs its function in communicative interactions
rather than in individual cognition. This mechanism produces reasons
for social consumption in the form of justifications and arguments.
When we talk of reason, we refer just to this mechanism. We don’t see
this mechanism as responsible for human inferences in general. Hence,
the interesting remarks M&B make about the role of language and con-
text in individual cognition are no objections to our claim.
An original claim we make in our book is that people are able to be
much more objective when evaluating arguments presented to them by
others than in producing arguments to convince others. One of the pre-
dictions that follows from that claim is that, in an exchange of arguments
among people with different solutions to some problem but a common
interest is solving it, while most individual members of the group are likely
to produce poor arguments, these arguments are likely to be recognized as
poor by the other members and to be eliminated so that, eventually, the
group is likely to converge on the right solution. There are dozens of ex-
periments confirming this prediction. When cooperative groups try to
solve demonstrative problems, a single individual with the correct answer
is nearly always able to convince the others; in other terms ‘truth wins’ is
by far the most common outcome [Moshman & Geil (1998); Trouche,
Sander, & Mercier (2014); for review, see Laughlin (2011)].
M&B, however, offer apparent counter-evidence to our prediction.
Using the horse-trading problem, Macchi and Bagassi (2015) showed
that participants who had given a wrong answer to a simple accounting
problem didn’t change their mind when they were presented with the
correct solution unless this correct solution was explained in a way that
addressed the source of their error. This is interesting in itself but we
don’t see how it warrants the conclusion that “usually, in fact, the best
idea does not prevail, but only the most shared one.”
There are two main reasons why the interesting results M&B dis-
cuss are no counter-evidence to our prediction:
(1) Our prediction is about group interaction. This experiment does
not involve a group discussion but just an interaction between
the participant and the experimenter. M&B do report that an
earlier study with group discussion [Mosconi, Bagassi, & Serafini

teorema XXXVIII/1, 2019, pp. 00-00

6 Hugo Mercier and Dan Sperber

(1988), which alas we could not find] produced similar findings,

but then Maier and Solem (1952) report that participants who
had given the wrong answer to the same problem changed for
the correct one when it was presented in their group. All this is
hard to interpret without having greater information on the pro-
cedures that were followed.
(2) Most interesting in M&B’s study is their demonstration that the
mistake of some of the participants was based on taking into ac-
count an irrelevant link between two transactions (the protago-
nist of the story twice buys and then sells a horse at a profit, his
total profit being the sum of the two; irrelevantly, it is the same
horse that is bought and sold by the protagonist, and it is bought
the second time at a price higher than the price at which it has
been sold, a fact that participants interpret as a loss that they
then wrongly subtract from the overall profit). Participants who
made this mistake were not convinced when presented with the
correct computation. Note that this correct computation was in-
cluded in their own computation, to which they wrongly added
another computational step. So, it is not surprising that they also
needed a non-misleading presentation of the fact that had misled
them in order to change their mind. Interesting indeed, but not
something on which our approach to reasoning commits us to
any prediction.
Although SALVADOR MASCARENHAS is in broad agreement with our
thesis, he suggests that, in the intellectualist theory, we might have con-
structed a bit of a straw man. More specifically, he reminds us that the
tradition of the psychology of reasoning, and of judgment and decision
making, does not only bear on reason (or System 2, as it might be called),
but on inferences more generally — for instance, the first-order infer-
ences that might draw participants towards the 10c answer to the bat and
ball, or towards the p and q cards in the standard Wason selection task.
As a result, when these psychologists claim that human reason mostly
serves individual, meliorative functions, they would mostly be correct, as
most of these first-order inferences do serve such functions.
We acknowledge, with Mascarenhas, that much important work on
inferential mechanisms other than reason has, since the 1950s, been done
under the heading of ‘psychology of reasoning.’ This sub-disciplinary
framing however may not have been that helpful: the study of specific
types of inference was seen as primarily relevant as a source of evidence

teorema XXXVIII/1, 2019, pp. 00-00

 Précis of The Enigma of Reason 7

for one or another view of reasoning (‘mental logic’ or ‘mental models’ in

particular) and other questions about the mechanisms involved — their
specific procedures, their function, their evolved underpinnings, for in-
stance — got little or no attention.
Moreover, we also believe that distinguishing between reason and
other intuitive processes, has been one of the valuable insights stemming
from the field (even if we do not share the System 1/System 2 distinction).
In this context, our target in The Enigma of Reason was the view that some
psychologists have formulated of the function of reason in a more restrict-
ed sense — what they would call System 2. When Kahneman, Stanovich,
or Evans, in some of their writings, suggest that System 2’s main function
is to correct the mistakes of our first-order inferences, and to help us reach
better decisions and better beliefs on our own more generally, their claims
clearly do not bear on first-order inferences across the board. As a result,
even if our book focuses on reason by contrast with first-order infer-
ences, we do not believe that it represented the field unfairly.
Finally, we want to stress that our thesis regarding the social func-
tions of reason only applies to reason as we define it, and not to other
inferential mechanisms. Throughout the book, we insist that reason is
only one sub-type of inference, or more precisely a sub-type of metarep-
resentational inferences, which are themselves a subtype of inference.

CATHAL O’MADAGAIN suggests that we underestimate the importance

of cultural transmission in the development of human reason. He points
out that some of reason’s most striking achievements — such as advanced
mathematics — are the outcome of a cultural ratchet effect: a human pe-
culiarity that allows us to add on to the cultural edifices built by our ances-
tors, potentially reaching new heights at each step. O’Madagain also points
out the importance of learning in more mundane forms of reasoning, as
when children learn to discount, say, ad populum arguments thanks to a
parent or teacher pointing out their weakness.
On the whole, these points are well taken. No one could rediscover
modern mathematics on their own — it is indeed built on the shoulder of
giants. The contrast between the travails that were necessary for advanced
mathematics to emerge, and the relative ease with which (some) students
can incorporate past discoveries is striking. But what is learnt, exactly,
when one learns new mathematical principles? In some cases at least, it’s
quite plausible that little or no reasoning is involved. Students might apply
the rules for long division with very little understanding of why they work.

teorema XXXVIII/1, 2019, pp. 00-00

8 Hugo Mercier and Dan Sperber

When it comes to reasons proper, as in a mathematical proof, is it

right to say that one has learnt to reason in a different way? Clearly, one
has learnt to use specific reasons in a new manner, since this peculiar ar-
rangement of premises to support a given conclusion would likely never
have occurred without prompting. Moreover, vaguely echoing Plato, we
would argue that the first-order inferences necessary to understand the
soundness of the proof must have been present before the proof was
encountered. For the proof to be genuinely understood, students must
be able to grasp each step. What is new isn’t each inferential step, but the
fact that these specific inferences are called in, in this specific order, to
bear on this specific conclusion. In some cases at least, people are then
able to recreate the series of inferences themselves if they have to con-
vince someone else in turn [Claidière, Trouche, & Mercier (2017)].
Still, it is true that we can develop shortcuts, such that a link between
a premise and a conclusion that was not intuitive at first—it required going
through several intermediate intuitive steps — can become more intuitive
with use. It then becomes easier to embed these reasons into increasingly
complex chains of reasoning. This is a fascinating phenomenon, and we
join O’Madagain in wishing for more research in this direction.
The learning of more mundane arguments is also fascinating — and
equally understudied. Again, it is important to try to pinpoint what exact-
ly is being learnt when, say, a child encounters a new argument.
O’Madagain takes the example of the ad populum, and of its counter-
arguments (e.g. if all your friends jumped from a bridge, would you join
them?). First, we disagree that the ad populum, or other so-called fallacies
of argumentation, are by nature fallacious. Following the majority is a
sound heuristic in a wide range of cases [Condorcet (1785); Hastie &
Kameda (2005)]. As a result, attempting to teach children that ad populum
arguments are fallacious as a rule might do more harm than good [even if
the most likely effect is to do nothing, see Mercier, Boudry, Paglieri, &
Trouche (2017)]. Moreover, it seems that participants are well able to
discriminate between weaker and stronger forms of these so-called falla-
cies, even without explicit teaching [Hahn & Oaksford (2007)]. In the
case of the ad populum, participants (including children) might be able to
rely on well calibrated intuitions about when they should follow majority
opinions [Morgan, Laland, & Harris (2015); Morgan et al. (2012); for
review, see Mercier & Morin (submitted)]. As a result, we suggest that
the evaluation of arguments, as a rule, does not require explicit teaching.
The case of argument production is more interesting. Even if chil-
dren are able to evaluate arguments without explicit learning, they need a

teorema XXXVIII/1, 2019, pp. 00-00

 Précis of The Enigma of Reason 9

lot of feedback when it comes to argument production. As we suggest in

The Enigma of Reason, the importance of this feedback has likely influ-
enced how reason works, making us ‘lazy’ when it comes to producing
arguments: if easily available arguments persuade their intended audi-
ence, there is no point bothering to find better ones; if they do not, then
we can rely on the audience to point that out and, more often than not, to
provide a counter-argument that will help us improve our own arguments.
Children, like adults, rely on this ‘negative feedback’ to learn which
arguments do not work and, eventually, stop giving them altogether (as
when a child learns that “because I want it” isn’t a good argument for
grabbing something of off their sibling). But the feedback that
O’Madagain is more interested in is positive: how children might appro-
priate for themselves the arguments used by others. Indeed, when a child
(or an adult for that matter) encounters a new argument, they are able
not only to evaluate it and decide whether they should change their
mind, but also, to some extent, to keep the form of argument in mind
and use it in turn [see, e.g. Anderson et al. (2001)]. The inferences chil-
dren (or, again, adults) have to draw in order to decide the circumstances
under which the new argument would be relevant have not, to the best
of our knowledge, been studied. Clearly, the learner brings a lot to the
table. They do not merely copy the initial argument, as they would if they
used it only to defend the specific conclusion the argument supported
when they first encountered it. Learners do not either generalize the ar-
gument to any potential conclusion (otherwise most arguments would be
patently absurd). A process that might look like pure learning — i.e. in-
corporating into one’s argumentative arsenal the arguments one encoun-
ters — requires a lot of pre-existing intuitions.
Even if we wholeheartedly agree with O’Madagain’s injunction that
more attention should be paid to the interactions through which children
may learn to reason better, our stance would be that he puts too much
weight on learning. In this respect, we should keep in mind that the prac-
tice of adults exchanging reasons with children is essentially a peculiarity
of middle- and upper-classes in rich societies. In other cultures, adults do
not talk with children all that much, when they do, they mostly rely on
imperatives, and they certainly do not feel compelled to justify their re-
quests [see, e.g. Gauvain, Munroe, & Beebe (2013); Maratsos (2007); Pye
(1986)]. Yet this doesn’t seem to stop children growing in traditional cul-
tures from developing the skills to evaluate (at least some) arguments
[Castelain, Bernard, Van der Henst, & Mercier (2016)].

teorema XXXVIII/1, 2019, pp. 00-00

10 Hugo Mercier and Dan Sperber

These latter results, suggesting that children in traditional cultures

differentiate (some) strong from weak arguments, might be partly con-
tested. As O’Madagain points out, experiments showing that young chil-
dren, including preschoolers and even 2-year-olds [Castelain, Bernard, &
Mercier (2018); Koenig (2012); Mercier, Bernard, & Clément (2014);
Mercier, Sudo, Castelain, Bernard, & Matsui (2018)] are able to discrimi-
nate between strong and weak arguments suffer from a potential con-
found. Even though the children are presented with arguments, it is
possible that they do not understand the statements qua arguments, pro-
cessing instead the premise as if it were a piece of information independ-
ent from the conclusion it was offered to defend. This is true, and we are
developing ways of testing whether young children can understand ar-
guments qua arguments. However, as a rule, in communication at least,
production tends to follow evaluation, and children this age do produce
arguments [for review, see Mercier (2011), (2016)]. We do not believe
therefore that our interpretation in terms of genuine argument evaluation
is particularly farfetched.
More generally, while we have argued at length that reason is an
evolved adaptation, we keep an open mind about the ways and the degree
to which it may be modified by cultural inputs. We conclude our book by
stating that “much more must be done to find out to what extent and in
which ways [reason] can be harnessed, enriched, and codified differently in
various cultural traditions” (p. 334). We welcome O’Magadain’s relevant
suggestions in this respect.

IRA NOVECK in his friendly, thoughtful, and demanding review makes

three reasonable requests: (i) that we defend our deflationary view of log-
ical inferences, (ii) that we develop clear means of falsifying our theory, (iii)
that we offer guidelines for how to deal with the justifications participants
offer in experiments.
First, let us reassure Noveck that we do not throw the baby of
spontaneous deduction with the water of mistaken accounts of the place
of deduction in inference generally and reasoning in particular. Not only
do we agree with Noveck that spontaneous deduction has been well estab-
lished and occurs quite commonly, we would even argue that its prevalence
is underestimated. Spontaneous deductions studied in the psychology of rea-
soning are almost exclusively deductions licensed by the presence of a so-
called logical term, for instance a connective such as “or” in (1) where (a)
entail (b):

teorema XXXVIII/1, 2019, pp. 00-00

Précis of The Enigma of Reason 11

(1) (a) Ira is in Paris or in London; he is not in London

(b) Ira is in Paris
We would argue that most items in the lexicon of any ordinary language
license some deductions. For instance, in (2), (a) entails (b) in virtue of
the semantics of so-called non-logical terms such as the verb “kill” in (4).
When such entailments are potentially relevant, they are, we suggest,
spontaneously computed:
(2) (a) John killed the dog
(b) The dog is dead
A deduction is a logical derivation from premises to a conclusion that
necessarily follows from the premises. It is an abstract relationship, just
as is, say, a multiplication. Such abstract relationships can be mentally
represented. When we speak of spontaneous deductions, we speak of the
process of mentally representing a logical deduction. Mentally represent-
ed deductions are not in and of themselves inferences. An inference, as
we use the term, is a psychological process that results in the formation
(or change in strength) of beliefs or decisions. Inferences may but need
not include deductive steps. Making a mental deduction, spontaneously
or deliberately, may contribute to the formation of a new belief (or to the
confirmation or rejection of a belief already held) but it needs not do so.
Even when a deduction plays a role in an inference, the conclusions of
the inference may be quite different from those of the deduction. Sup-
pose you are told, “Ira is in Paris or in London; he is not in London.”
You may spontaneously deduce, “Ira is in Paris,” but if you happen to be
confident that Ira is not in Paris, what you will infer from this deduction is
that what you were told is false. Or suppose that Peter tells you, “John
killed the dog,” and that (a) you already knew that the dog was dead; (b)
you didn’t know that Peter knew that the dog was dead; and (c) you don’t
believe that Peter is in a position to know how the dog died. In that case,
the spontaneous deduction from Peter’s utterance would indirectly inform
you that Peter knew that the dog was dead, but it wouldn’t modify your
beliefs about what happened to the dog. Generally speaking, the study of
mental deductions and that of inference (and in particular or reasoning)
should be sharply distinguished. The study of deduction has some rele-
vance to that of inference, but its main relevance may be elsewhere.
Because deductive relationships are an essential aspect of linguistic
meaning, deduction — in the sense of the mental representation of these

teorema XXXVIII/1, 2019, pp. 00-00

12 Hugo Mercier and Dan Sperber

relationships — often occurs spontaneously and plays, we would argue,

an important role in verbal comprehension. We take, however, a strongly
pragmatic approach to the role of linguistic meaning in comprehension:
linguistic meaning is a piece of evidence from which the audience can in-
fer the speaker’s intended meaning; it is not an encoding of that meaning
[Wilson & Sperber (2012)]. The main relevance of spontaneous deduc-
tion, we suggest, is to the semantics and pragmatics of natural language.
In particular, as we argue in the book, a deductive format is often used to
highlight the structure of a probabilistic argument. This rhetorical role of
deduction in the pragmatics of argumentation is interesting in its own
right, but it is, of course, less central and less grand than the role at-
tributed to deduction in logistic approaches to reasoning.
Noveck also mentions interesting recent work [Cesana-Arlotti et al.
(2018)] suggesting that pre-verbal infants can detect violation of disjunc-
tion elimination. More generally, isn’t logical deduction something that
humans do not only with natural language but also in non-linguistic think-
ing (or with ‘language of thought’)? Don’t other animals also use logic?
According to Jerry Fodor, they must: “Darwinian selection guarantees that
organisms either know the elements of logic or become posthumous”
[Fodor (1981), p. 121)]. We argue that inferences that psychologists or phi-
losophers can schematize as deductive need not be so, for at least two rea-
sons: they may be probabilistic rather than deductive, and they may be
performed by specialized mechanisms that exploit empirical regularities
without representing them as premises in anything resembling a deduction.
So, to answer Noveck’s first request in a nutshell: while deductive
relationships clearly play a major role in linguistic semantics, the exact
role or roles they play in inference in general or in reasoning in particular
has been exaggerated and obscured by logicist dogma and is in need of
an open-minded reconsideration.
Noveck’s second request is that we develop clear means of falsify-
ing the interactionist theory. Our theory rests on some evolutionary,
functional hypotheses, namely, that reason evolved by serving justificato-
ry and argumentative functions. Although there are several ways of test-
ing evolutionary hypotheses — for instance looking at variations in
fitness, or at genetic data — in the case at hand we have relied on the
match between structure and function. Within a broadly adaptationist
framework, we expect biological mechanisms to have a structure that fits
their function, that serves it well [Williams (1966)]. Our argument thus
rests on two sets of claims: (i) claims regarding what the structure of rea-
son is, and (ii) claims regarding the fit between this structure and the

teorema XXXVIII/1, 2019, pp. 00-00

Précis of The Enigma of Reason 13

purported functions of reason. Our evolutionary hypotheses could thus

be falsified by showing that either of these claims are mistaken.
Regarding the structure of reason, the most relevant traits, for our
purposes, were summarized in Table 2, p. 235 of The Enigma of Reason
(reproduced as Table 1 of the précis). One way of falsifying our hypothe-
ses is thus to show that human reason does not, in fact, possess these
traits. For instance, the claim that people are able to evaluate reasons that
challenge their point of view objectively is not consensual (to say the
least). Evidence that people are deeply biased against such reasons would
argue against our hypotheses.
Even if one were to accept our characterization of the traits of rea-
son, one could still attempt if not to falsify, at least to undermine our hy-
potheses by showing that these traits are better explained as unavoidable
by-products of other traits, or that they fit better with another purported
function. For example, some psychologists have suggested that the my-
side bias (or confirmation bias) is a natural outcome of cognitive pro-
cessing across the board. If that were true, then the existence of this bias
could not argue in favor of our hypotheses anymore. We believe this is
not true (see Chapter 11), but if it were shown that the myside bias is
widespread in our cognitive system, instead of being restricted to reason
as we claim, our hypotheses would be weakened. Likewise, people have
suggested other functional explanations for the myside bias and; if their
explanations were better supported than ours, our hypotheses would
again be weakened.
Without falsifying the overall framework, it is possible to falsify
more specific claims related to our theory. Noveck mentions the issue of
small group discussions: when does these discussions improve perfor-
mance on reasoning and decision making problems? Here, our theory
makes broad predictions: on the whole, when groups of people who dis-
agree on a given point, but share some common incentives, exchange ar-
guments together, the best arguments should carry the day and
performance should improve. However, there will be exceptions. For ex-
ample, sometimes the participants who have the correct answer might
have reached it partly by chance, or without being able to formulate rea-
sons defending their correct answer in a way that would convince other
participants who came to a wrong solution in a roundabout way (as
might be the case for the horse seller problem presented by Macchi &
Bagassi). Such special instances do not invalidate our theory, since the
theory doesn’t claim that people are always able to turn their correct in-
tuitions into reasons apt at convincing the intended audience. Still, the

teorema XXXVIII/1, 2019, pp. 00-00

14 Hugo Mercier and Dan Sperber

theory requires some ancillary considerations to explain why argumenta-

tion fails to work in these situations (i.e. because people don’t seem to be
able to articulate why the correct answer is correct) and, if such situa-
tions were more common than situations in which argumentation works
well, our theory would be in trouble.
Another example is that of ‘reason-based choice.’ Relying on a long
tradition in judgment and decision making, we claim that many devia-
tions from normative behavior in decision making tasks reflect the oper-
ation of reason, as it drives participants towards the most easily
justifiable answer — whether it is otherwise the best answer or not (see
Chapter 14). In each specific case, this explanation can be tested, for in-
stance by looking at whether (i) making participants accountable ampli-
fies the deviation from normative behavior (it should), (ii) cognitive load
reduces the deviation from normative behavior (it should), (iii) non-
human animals make the same mistakes (they shouldn’t), or (iv) the justi-
fications participants produce are in line with the answers they give, and
appear somewhat convincing to others (they should).
This last point brings us to Noveck’s final request regarding the
usefulness of justifications in psychological research. First, a bit of ter-
minology: we would differentiate explanations from justifications. Justifi-
cations are based on reasons, whereas explanations may use reasons, but
they need not. For instance, if a participant explains their answer in a
psychological experiment by saying ‘I’m too tired to think’, this is an ex-
planation, but not a justification (not a reason). We’ll take it that when
Noveck talks about explanations, what he has in mind is justifications —
which would indeed form the majority of participants’ answers, whether
they are asked to explain or to justify their answers. (Given that one of
us has used ‘explain’ in prompts, even though he was looking for justifi-
cations, we can see how this might not have been clear!)
With this in mind, the answer to the question “Does it follow that a
justification does not count as reasoning when the task simply asks a par-
ticipant to provide one, as they are in the first phase of the Trouche et al.
(2015) experiment?” is a qualified no. We spend some time arguing that
the same reasons can be used either or both as justifications or as argu-
ments, with no clear demarcation between these two uses. So, justifica-
tions are indeed reasons, produced by our faculty of reason. Participants
who give a reason to justify their answer to the experimenter could as
well use it to convince another participant to agree with them. The fact
that these reasons likely played no causal role in arriving at the answers is
normal: we claim that this is generally true of the reasons we produce.

teorema XXXVIII/1, 2019, pp. 00-00

Précis of The Enigma of Reason 15

By contrast, the question of whether the cues described by Noveck

(“Please note that the sum of the two percentages must be 100%. For
example, Anne is either a mathematics teacher or a French literature
teacher” to help participants solve a probabilistic problem) are reasons is
more problematic. First, there is no intrinsic quality in a statement that
makes it a reason or not, it all depends on how it is processed — so that
the same statement can be processed as an implicit instruction to follow
in producing an answer by a participant and as a reason to be invoked in
justifying one’s answer by another participant (in the same way that an
imperative utterance can be processed as an order by someone and as an
advice by someone else, say –– this ambiguity raises difficulties in inter-
preting some experimental results, see the discussion of O’Madagain’s
final point above). Second, in the case at hand, there are no strong ele-
ments helping participants process the statement as a reason: no explicit
conclusion, no connectives, etc. As a result, it’s quite plausible that par-
ticipants would interpret the cue as a piece of information, from which
they are free to draw whatever conclusion they want. If that is the case, a
prediction would be that if the conclusion were spelled out, and the link
between the cue (which becomes a premise) and the conclusion clarified,
more people should be able understand the relevance of the cue, and
thus to reach the correct answer.
Unfortunately, testing whether or not a given statement is pro-
cessed as a reason is not trivial. As mentioned above, we’re working on
experimental paradigms that would allow such testing in children. Other
methods, aimed primarily at adults, using reaction times, eye tracking, or
even neuroscientific tools might be used as well, but they still have to be
developed (another work in progress!).

Institut Jean Nicod, Département d’études cognitives,

ENS, EHESS, PSL University, CNRS,
Paris France
E-mail: hugo.mercier@gmail.com

Department of cognitive science and Department of philosophy,

Central European University, Budapest, Hungary
Institut Jean Nicod, Département d’études cognitives,
ENS, EHESS, PSL University, CNRS,
Paris France
E-mail: dan.sperber@gmail.com

teorema XXXVIII/1, 2019, pp. 00-00

16 Hugo Mercier and Dan Sperber

ACKNOWLEDGMENTS
HUGO MERCIER’S work is supported by the Agence Nationale de la Re-
cherche, EUR FrontCog ANR-17-EURE-0017. DAN SPERBER’S work is sup-
ported by the European Research Council under the European Union’s Seventh
Framework Programme (FP7/2007-2013)/ERC grant agreement n° [609819],
SOMICS.

REFERENCES

ANDERSON, R. C., NGUYEN-JAHIEL, K., MCNURLEN, B., ARCHODIDOU, A.,

KIM, S., REZNITSKAYA, A., & GILBERT, L. (2001), “The Snowball Phe-
nomenon: Spread of Ways of Talking and Ways of Thinking Across
Groups of Children; Cognition and Instruction, 19(1), pp. 1-46.
BILLIG, M. (1996), Arguing and Thinking: A Rhetorical Approach to Social Psychology;
Cambridge: Cambridge University Press.
CASTELAIN, T., BERNARD, S., & MERCIER, H. (2018), “Evidence that Two-Year-
Old Children are Sensitive to Information Presented in Arguments”; Infancy,
23(1), 124–135.
CASTELAIN, T., BERNARD, S., VAN DER HENST, J.-B., & MERCIER, H. (2016),
“The Influence of Power and Reason on Young Maya Children’s En-
dorsement of Testimony; Developmental Science, 19(6), pp. 957-966.
CESANA-ARLOTTI, N., MARTÍN, A., TÉGLÁS, E., VOROBYOVA, L., CETNARSKI,
R., & BONATTI, L. L. (2018), “Precursors of Logical Reasoning in Preverb-
al Human Infants”; Science, 359(6381), pp. 1263-1266.
CLAIDIÈRE, N., TROUCHE, E., & MERCIER, H. (2017), “Argumentation and the
Diffusion of Counter-Intuitive Belief; Journal of Experimental Psychology: Gen-
eral, 146(7), pp. 1052-1066.
CONDORCET. (1785). Essai sur l’application de l’analyse à la probabilité des décisions
rendues à la pluralité des voix. Paris: L’imprimerie royale.
FODOR, J.A., (1981), Representations: Philosophical Essays on the Foundations of Cogni-
tive Science; Cambridge: MIT Press.
GAUVAIN, M., MUNROE, R. L., & BEEBE, H. (2013), “Children’s Questions in
Cross-Cultural Perspective a Four-Culture Study”; Journal of Cross-Cultural
Psychology, 44(7), pp. 1148-1165.
GIBBARD, A. (1990), Wise choices, Apt feelings; Cambridge: Cambridge University
Press.
HAHN, U., & OAKSFORD, M. (2007), “The Rationality of Informal Argumenta-
tion: A Bayesian Approach to Reasoning Fallacies”; Psychological Review,
114(3), pp. 704-732.
HAIDT, J. (2001), “The Emotional Dog and Its Rational Tail: A Social Intuition-
ist Approach to Moral Judgment”; Psychological Review, 108(4), pp. 814-834.
HASTIE, R., & KAMEDA, T. (2005), “The Robust Beauty of Majority Rules in
Group decisions”; Psychological Review, 112(2), pp. 494-508.

teorema XXXVIII/1, 2019, pp. 00-00

Précis of The Enigma of Reason 17

HOCHSCHILD, A. (2006), Bury the Chains: Prophets and Rebels in the Fight to Free an
Empire’s Slaves. Boston: Houghton Mifflin Company.
KOENIG, M. A. (2012), “Beyond Semantic Accuracy: Preschoolers Evaluate a
Speaker’s Reasons”; Child Development, 83(3), pp. 1051-1063.
KUHN, D. (1991), The Skills of Arguments; Cambridge: Cambridge University
Press.
LAUGHLIN, P. R. (2011), Group Problem Solving; Princeton: Princeton University
Press.
MACCHI, L., & BAGASSI, M. (2015), “When Analytic Thought is Challenged by a
Misunderstanding; Thinking & Reasoning, 21(1), pp. 147-164.
MAIER, N. R., & SOLEM, A. R. (1952), “The Contribution of a Discussion Lead-
er to the Quality of Group Thinking: the Efective Use of Minority Opin-
ions”; Human Relations, 5(3), pp. 277-288.
MARATSOS, M. P. (2007), “Commentary”; Monographs of the Society for Research in
Child Development, 72, pp. 121-126.
MERCIER, H. (in press), “A paradox of Information Aggregation: We Do It Well
but Think About It Poorly, and Why this is a Problem for Institutions”; In
N. Ballantyne & D. Dunning (Eds.), Epistemology and Psychology. New York:
Oxford University Press.
–– (submitted), Not Born Yesterday: The Science of Who We Trust and What We Be-
lieve; New York, Princeton University Press.
–– (2011) “Reasoning Serves Argumentation in Children”; Cognitive Development,
26(3), pp. 177-191.
–– (2012), “Looking for Arguments”; Argumentation, 26(3), pp. 305-324.
–– (2016), “The Argumentative Theory: Predictions and Empirical Evidence”;
Trends in Cognitive Sciences, 20(9), pp. 689-700.
MERCIER, H., BERNARD, S., & CLÉMENT, F. (2014), “Early Sensitivity to Argu-
ments: How Preschoolers Weight Circular Arguments”; Journal of Experi-
mental Child Psychology, 125, pp. 102-109.
MERCIER, H., BOUDRY, M., PAGLIERI, F., & TROUCHE, E. (2017), “Natural-Born
Arguers: Teaching How to Make the Best of our Reasoning Abilities”; Edu-
cational Psychologist, 52(1), pp. 1-16.
MERCIER, H., DOCKENDORFF, M., & SCHWARTZBERG, M. (submitted), “Demo-
cratic Legitimacy and Attitudes About Information-Aggregation Proce-
dures”.
MERCIER, H., & MORIN, O. (submitted), “Majority Rules: How Good Are We at
Aggregating Convergent Opinions”?
MERCIER, H., SUDO, M., CASTELAIN, T., BERNARD, S., & MATSUI, T. (2018),
“Japanese Preschoolers’ Evaluation of Circular and Non-Circular Argu-
ments”; European Journal of Developmental Psychology, 15(5), pp. 493-505.
MERCIER, H., TROUCHE, E., YAMA, H., HEINTZ, C., & GIROTTO, V. (2015),
“Experts and Laymen Grossly Underestimate the Benefits of Argumenta-
tion for Reasoning”; Thinking & Reasoning, 21(3), pp. 341-355.

teorema XXXVIII/1, 2019, pp. 00-00

18 Hugo Mercier and Dan Sperber

MORGAN, T. J. H., LALAND, K. N., & HARRIS, P. L. (2015), “The Development

of Adaptive Conformity in Young Children: Effects of Uncertainty and
Consensus”; Developmental Science, 18(4), pp. 511-524.
MORGAN, T. J. H., RENDELL, L. E., EHN, M., HOPPITT, W., & LALAND, K. N.
(2012), “The Evolutionary Basis of Human Social Learning”; Proceedings of
the Royal Society of London B: Biological Sciences, 279(1729), pp. 653-662.
MOSCONI, G., BAGASSI, M., & SERAFINI, M. G. (1988), “Solutori e benrispondenti.
II. Il problema della compravendita del cavallo: Discussione e ricerca di
gruppo”; Giornale Italiano Di Psicologia, XV, 4, pp. 671-694.
MOSHMAN, D., & GEIL, M. (1998), “Collaborative Reasoning: Evidence for Col-
lectiveRrationality”; Thinking and Reasoning, 4(3), pp. 231-248.
PERELMAN, C., & OLBRECHTS-TYTECA, L. (1958), The New Rhetoric: A Treatise on
Argumentation. Notre Dame, IN: University of Notre Dame Press.
PIAGET, J. (1928), Judgment and Reasoning in the Child; London: Routledge and
Kegan Paul.
PYE, C. (1986), “Quiché Mayan Speech to Children”; Journal of Child Language,
13(1), pp. 85-100.
SPERBER, D. ET AL., (2010), “Epistemic Vigilance”; Mind and Language, 25(4), pp.
359-393.
TROUCHE, E., SANDER, E., & MERCIER, H. (2014), “Arguments, More than
Confidence, Explain the Good Performance of Reasoning Groups”; Jour-
nal of Experimental Psychology: General, 143(5), pp. 1958-971.
TROUCHE, E., SHAO, J., & MERCIER, H. (2019), “How is Argument Evaluation
Biased?”; Argumentation, (in press).
WILLIAMS, G. C. (1966), Adaptation and Natural Selection; Princeton: Princeton
University Press.
WILSON, D. & SPERBER, D (2012), Meaning and Relevance; Cambridge: Cambridge
University Press.

teorema XXXVIII/1, 2019, pp. 00-00