R2.1. The double-sided argumentative function of reasoning

Why does reasoning exist at all, given that it is a relatively high-cost mental activity with a relatively high failure rate? To answer this question, we proposed to step back from the study of individual cognitive processes and to look at the evolution of human communication. Humans are immersed in a flow of socially transmitted information and are highly dependent on it. For communication to have evolved, it had to be advantageous to both communicators and receivers (who are, of course, the same individuals but acting in two different capacities). What makes communication advantageous to receivers is that it provides them with rich information that they could not, or not easily, have obtained on their own. For this, the information they receive has to be genuine information; that is, close enough to truth. What makes communication advantageous to communicators is that it allows them to achieve some desirable effect in the receivers. For this, the information they emit has to be conducive to this effect, whether it is true or false.

Dessalles, who has himself developed a perspective in some respects comparable to ours, understands us to claim that “the biological function of reasoning is to achieve shared knowledge optimization” (para. 2) and that this is done not at the individual but at the group level. We do argue that the main function of reasoning is indeed social but by serving the social interests of individuals rather than the collective interests of the group.

To reap the benefits of communication while limiting the risk of being misled, receivers must exercise what we have called epistemic vigilance (Sperber et al. Reference Sperber, Clément, Heintz, Mascaro, Mercier, Origgi and Wilson2010). There is no fail-safe algorithm to sort genuine from spurious information; hence, we argue, various cost-effective heuristics that may contribute to approximating such a sorting are likely to have evolved. The main heuristic of epistemic vigilance consists in assessing the trustworthiness of communicators. Thus, we agree with Opfer & Sloutsky that “children detect trustworthiness long before they detect argument inconsistency” (para. 4) (e.g., see Mascaro & Sperber Reference Mascaro and Sperber2009). But if detecting the trustworthiness of communicators were the only heuristic used, then receivers would end up rejecting a good amount of genuine and relevant information when they lack sufficient ground to accept it on trust. For instance, few if any readers of our article would accept its conclusions just out of trust in its authors! To be more effective, epistemic vigilance must be exercised not only towards the source of information but also towards its content. Independently of its source, a message may have a greater or a lesser believability. This believability is assessed by considering its coherence with background knowledge. Coherence checking, we argue, is the second major heuristic used in filtering communicated information, and is at the basis of reasoning proper.

Coherence checking starts as a method for receivers to filter information; it ends up being exploited also by communicators who engage in coherence displays in order to have their messages accepted. Just as receivers would filter out some genuine information if they relied only on the trustworthiness of the source, communicators would fail to communicate some believable messages if they relied only on their own authority. Arguing consists in displaying coherence-based reasons for the acceptance of a given message. It is, in essence, an “honest display” strategy opened to evaluation and aimed at the audience's epistemic concerns. Of course, what is displayed may be an invalid argument made in the service of deception and in the hope that its invalidity won't be detected. Contrary to what Dessalles attributes to us, we do not believe that communicators argue in order “to correct or update others' beliefs” (para. 3) when it is not to their advantage. They argue for whatever it is advantageous to them to have their audience believe. Often enough, for instance, when communicating to coordinate action, communicator and audience have convergent interests in sharing true information, but this is far from being always the case.

In an evolutionary perspective, receivers' coherence checking creates selective pressure for communicators' coherence displays in the form of arguments, which in turn creates selective pressure for adequate evaluation of arguments on the part of receivers. At least in some cultural contexts, this results in a kind of arms race towards greater sophistication in the production and evaluation of arguments. Of course, argumentation can be misused and abused–for instance, by arguing above the head of one's audience (Sperber Reference Sperber2009) or by lacing arguments with appeals to emotion. Doing so, however, is more likely to serve the interests of the communicator than those of the audience. Contrary to what Opfer & Sloutsky maintain, “hot” persuasion is advantageous to communicators only to the extent that receivers yield to it, but it is not advantageous to receivers who care to be well informed. For this they had better reason, as Petty and Cacioppo (whom Opfer & Sloutsky oddly cite in support of their claim) have shown in numerous experiments that demonstrate precisely this point: When people are motivated to reason, they do a better job at accepting only sound arguments, which is quite generally to their advantage (e.g., see Petty et al. Reference Petty, Cacioppo and Goldman1981).

R2.2. Other functions of reasoning?

Several commentators, while agreeing that argumentation may be an important function of reasoning, suggest that it may serve other functions, as well: either social functions other than the production and evaluation of arguments (Baumeister, Masicampo, & DeWall [Baumeister et al.]; Dessalles; Frankish; Pietraszewski) or functions contributing to individual cognition (Evans; Frankish; Godfrey-Smith & Yegnashankaran). We recognize the possibility. Our claim is that argumentation is the main function of reasoning, and we tried to demonstrate this claim by showing the signature of this function in the way reasoning actually functions, and in particular in what has been seen as flaws and that we argue are features of reasoning. Any evolved mechanism can be put to a variety of uses: The mouth can be used to play a wind instrument, the skin can be used to display tattoos, and the sense of smell can be used to develop wine expertise. Whether these additional uses become functions in the biological sense, or even main functions (in which case Gould and Vrba's term exaptation might be properly applied) depends on the relative contribution these uses make to the fitness of organisms endowed with the mechanism.

Dessalles and Frankish suggest that argumentation could have evolved as a means to display one's intellectual skills. Indeed, argumentation can be put to such a use. However, the human drive to show off is so strong that just about everything in our physical and cognitive makeup has been recruited to that effect. People may indeed run, fight, or argue to impress others. But what makes these performances not only impressive but relevant is that running, fighting and arguing are useful for purposes other than impressing others. Moreover, the production of arguments typically involves a high degree of mere satisficing (Mercier, submitted a). People do not look for the best formulation of the best possible argument. Instead, they use the first minimally decent argument that comes to mind. If it works, then it wasn't worth expending more effort. If it doesn't, it is easy to try a rebuttal or another argument. This way of arguing is to be expected only if the goal is to convince, but it would be very surprising if reasoning had the function to display one's skills by producing impressive arguments. In other words, reasoning is more like a crow's than a peacock's tail: It may be a bit drab, but it serves its main function well. Its occasional use, for instance, in academic milieus, to display one's intellectual skills is unlikely to contribute to fitness to the point of having become a biological function, let alone the main function of reasoning.

Pietraszewski rightly points out that argumentation is used not just in the defense of factual claims but also of claims that are not properly speaking matters of fact but more matters of choice or of social alignment. He distinguishes two classes of cases, one where the goal of arguments is “to change representations of the payoff structure of pursuing certain future plans” (para. 8). When argumentation is used for such purpose, it recruits intuitions that bear on the domain at hand–for example, what is advantageous or what is well regarded–and puts forward reasons for a conclusion. Such cases may fall squarely under the characterization we have given of reasoning. Pietraszewski also draws attention to a second class of cases where “agreement and disagreement … become commodities in themselves as a way of signaling the coordination strength and challenging others. This class of argumentation psychology should be designed to conflate evaluations of the argument with the source and social context of the argument; who is arguing should be just as important as what they are saying when considering the ‘goodness’ of an argument” (para. 9). This is a welcome and important observation, but does it point to another function of reasoning? We would suggest rather that it highlights that communication typically involves a mix of means and goals. The main relevance of a communicative act may be in its explicit content, in its implicatures, or in the very fact that it took place at all (Sperber & Wilson (Reference Sperber and Wilson1995); it may have to do with transmission of factual information or, indeed, with signaling agreement and disagreement. This can be done in particular by using arguments not so much to convince but to polarize. The phenomenon of polarization that, in the target article, we discussed mostly in negative terms, should probably be seen as fulfilling a function along the lines suggested by Pietraszewski. This said, it is not clear that it bears so much on the function of reasoning rather than on a wider range of mechanisms of social cognition that exploit and interact with argumentation in a variety of communicative interactions. In particular, as Opfer & Sloutsky insist and as Sell Reference Sell2006, quoted by Pietraszewski) has shown with the example of anger, “hot” communication interferes with argumentation in important ways, reminding us that argumentation is only one device for persuasion among several.

Baumeister et al. draw attention to two major phenomena linked to reasoning: consciousness and culture. We noted that reasoning is a typically conscious activity but we did not elaborate for lack of sufficiently clear ideas regarding consciousness in general and its relationship to reasoning in particular. Baumeister et al. propose to extend the social hypothesis towards conscious thought in general: “Conscious thought enables people to talk to others and thereby enables small groups to resolve differences” (para. 6). Their arguments are indeed very congenial to the argumentative theory. Reasoning could then be seen as one part of a larger set of mental processes that are not typically thought of as being social by nature, even though they actually are. Baumeister and colleagues have made forays in this direction, and we hope that more research will follow, shedding new light on well-known but still puzzling results.

Regarding the role of reasoning in culture, we agree that indeed it is a source of cultural innovation (for example, in technology, law, and the sciences) and that it plays a role in cultural communication (with great cultural variations, we surmise). However, we are not convinced that this is the function or even a function of reasoning. Reasoning is advantageous to individuals who are better able to persuade others or to discriminate good from bad arguments. Most of the arguments humans produce and evaluate are about very local concerns: Who forgot to close the door? Should we walk or take the bus? Was John lying or mistaken? Reasoned answers to these local questions don't ever reach a cultural level of distribution. Reasoning in order to improve laws or to discover new techniques is very rare. It may well be beneficial to the group, but that is not enough to assume that reasoning evolved, through group selection, for such a benefit.

Godfrey-Smith & Yegnashankaran, drawing on Vygotsky, make the interesting suggestion that “reasoning is deliberative [i.e., individualistic] in function but dialogic in structure” (para. 3). We do not dispute that reasoning can be used for individual ratiocination. Even though solitary reasoning may not be the most effective way to enhance individual cognitive goals, we do not exclude that such enhancement may have sufficiently contributed to the relative fitness of reasoners to constitute a function of reasoning. What we would dispute is that this is at all likely to be the main function of reasoning, and the Godfrey-Smith & Yegnashankaran commentary is helpful to spell out the argument. If they are right and individualistic reasoning is dialogic in structure, why should this be so? Were a computer scientist to design a reasoning program (not specifically aimed at argumentation), would she opt for such a structure? In fact, many reasoning program have been developed in artificial intelligence (AI) and in the psychology of reasoning (e.g., Johnson-Laird Reference Johnson-Laird2006; Rips Reference Rips1994), and typically they generate arguments (in the formal sense of the term) without any dialogic back-and-forth between two agents. We suggest that this dialogic structure of individual reasoning is the signature of its primarily argumentative main function. We would argue moreover that the main contribution of reasoning to individual cognition is in helping people evaluate other people's arguments.

Both Evans and Frankish suggest an interesting way to reconcile our suggestion that reasoning may have evolved for argumentation and the more classical view that it serves first and foremost individual cognitive goals. This individualistic function, Evans argues, might be an exaptation from an initial argumentative function favored by the concomitant evolution of “language, metarepresentation, and large forebrains” (para. 6). Frankish proposes that reasoning, even if it primarily evolved for argumentation, might have been substantially “co-opted to play a role in individual cognition” (para. 5). As we said in the preceding paragraph, we do not exclude the possibility that reasoning may also be adaptive in helping individual ratiocination. Still, our argument remains that the many apparent failings of reasoning–our knowledge of which owes much to Evans himself–make better sense as the signature of a main argumentative function.

Evans insists on the role of reasoning in anticipating the future. While anticipating the future is a major aspect of human cognition, it is not clear to us what specific role reasoning plays in this process, by contrast with other abilities such as imagination and simulation. It is even less clear what features of reasoning, if any, are specifically tailored for this role. Frankish points out that reasoning can be used to strengthen our resolve by buttressing our decisions with supporting arguments. Indeed, reasoning can do exactly that. However, if weakness of resolve had been enough of a problem to favor the evolution of a mechanism to deal with it, presumably the natural selection solution would have been to strengthen our resolve directly rather than to adapt reasoning to find resolution-strengthening arguments. Moreover, this very tendency to use reasoning to bolster our beliefs and decisions is likely to have more negative than positive consequences, as highlighted in section 4.2 of the target article.

R3.1. Are we really good at argumentation?

If reasoning evolved for argumentation, humans should possess decent argumentative skills: They should be able to engage in an informal debate, constructing, evaluating, and rebutting arguments. Several commentators question that this is so, pointing to research that shows significant improvement in argumentative ability through learning and casting doubt on the data we presented to defend argumentative skills.

Harrell, Kuhn, and Wolfe all point to very interesting data showing improvement in argumentation skills with training–as well as the relative poverty of these skills before training. Most of the studies cited, however, bear on meta-argumentative skills. Kuhn et al. (Reference Kuhn, Goh, Iordanou and Shaenfield2008) report gains in “meta-level communications about the discourse” (p. 1310). Larson et al. (Reference Larson, Britt and Larson2004) study the understanding of long written arguments. Critical thinking skills are often evaluated through the ability to draw argument schemas or to recognize the different components of an argument and their relationship to one another. Such skills may very well be crucial for success in modern academic life or even be of relevance to participation in democratic life, and their study in important in its own right. However, they are not part of what one can expect basic argumentative skills to be. Language provides a useful analogy. One can be a fluent language user without being able to tell what a verb is or, a fortiori, to parse a sentence. Likewise, one can be a skilled arguer without being able to recognize argument forms or draw argument schemas. In both cases, these abilities can help–a mastery of syntax can make one a better poet, a mastery of argument schemes a better essayist–but they are not necessary.

One study seems to tap into a more fundamental argumentative skill–the ability to draw counterarguments (Goldstein et al. Reference Goldstein, Crowell and Kuhn2009). In this experiment, sixth and seventh graders had to choose between a counterargument and a novel argument in a fictitious discussion, and they tended to choose the latter instead of the former. Even though the context successfully mimicked a debate, it does not follow that the participants were highly motivated to defend a position that had been arbitrarily assigned to them. In such a context, it is not surprising that they should engage in satisficing and pick an argument that might just be enough. Moreover, although the counterargument option might have been formally superior, it is not clear that it would have been the most efficient one in a real discussion.

Harrell and Wolfe also dispute the data used as evidence of people's basic argumentative skills. Harrell suggests deflationary interpretations for several of the findings we cite, but, however ingenious, these interpretations are not compelling. To be more specific: Bailenson and Rips (Reference Bailenson and Rips1996) do observe that being the first speaker increases the burden of proof. But with contentious first sentences such as “abortions should not be illegal” (p. S7), this is hardly surprising–or nonnormative. In Rips (Reference Rips2002), the repetitions are nothing but innocuous, and so participants are right to discount arguments accordingly. Neuman et al. (Reference Neuman, Weinstock and Glasner2006) do not really observe worse performance among people who argue. Participants were merely avoiding pointing out a fallacy in a polite dialogue (cf. Rips (Reference Rips2002, n. 1), but they were still perfectly able to detect it in a rougher discussion. Thompson et al. (Reference Thompson, Evans and Handley2005) observed more belief bias when people were engaged in an argument … because people were asked about their opinion and not logical validity. Far from having “difficulty producing evidence for a claim” (Harrell, para. 7), the participants in the study by Sá et al. Reference Sá, Kelley, Ho and Stanovich2005) were able to muster nearly six arguments to defend their opinion on difficult topics (for example, Why do children fail in school?). If these arguments were not always the best, this is only to be expected in such circumstances (as explained in sect. 2.3 of the target article).

Uleman, Kressel, & Rim [Uleman et al.] point out that the strong and weak arguments used in most persuasion research are “pretested … to have these properties” (para. 10), so that it is hardly surprising that strong arguments should be more convincing. That researchers want to make sure their intuitions about argument strength are appropriate does not mean that they are without grounding, however. For instance, when trying to persuade students that a new exam would be beneficial, a strong argument might be that “average starting salaries are higher for graduates of schools with the exams” (Petty & Cacioppo (Reference Petty and Cacioppo1979, p. 1921) accompanied by relevant statistics. By contrast, a weak argument might read “by not administering the exams, a tradition dating back to the ancient Greeks was being violated.” It is not a stretch to find some objectivity in this judgment of strength: It is not a mere experimental artifact that the stronger arguments, in these experiments, do actually provide more support for their conclusion. Finally, Wolfe mentions a study in which “implausible reasons and warrants … yielded higher agreement than the same claims without support” (para. 2). Famously, Langer et al. Reference Langer, Blank and Chanowitz1978) had already demonstrated the effect of poor reasons more than 30 years ago. However, they also showed that the effect mostly disappeared when the stakes increased–as argued in the target article, whether people genuinely evaluate arguments depends on how much they care about (and disagree with) the conclusion.

So, without denying that spontaneous argumentation skills are imperfect and can be improved by teaching–and that this is linked to the variable importance given to argumentation in different cultures and institutions–we maintain that they display a remarkable superiority to the reasoning skill elicited in nonargumentative contexts.

R3.2. How efficient is group reasoning?

This question has elicited contrary opinions from the commentators. Khlentzos & Stevenson think that good performance in groups is obvious since “subjects share a common goal of finding the correct solution” (para. 11). However, the same participants can face the same problems with the same goal but individually fail, so having the correct answer as a goal can hardly be the whole story. Johnson, on the other hand, questions the generality of good group performance, and Sternberg claims that groups are actually very poor at reasoning or decision making, citing as a support the groupthink syndrome or group polarization. First, it should be stressed that the argumentative theory does not predict that groups will always make better decisions, but merely that reasoning should work better in the context of a genuine debate. Many other factors besides reasoning can impact the outcome of a discussion–strategic considerations, face saving, and so forth. And reasoning in group can also bring poor outcomes when there is no genuine deliberation. Actually, in section 2.3 of the target article, we offer an explanation based on the argumentative theory for group polarization. Without a refutation of this explanation, we don't see how this very phenomenon can be used as evidence against the theory. Finally, Opfer & Sloutsky mention one study that showed groups of children performing more poorly after a discussion (Levin & Druyan Reference Levin and Druyan1993). It is true that sometimes the best arguments will point in the wrong direction. When, in the early 20th century, geologists argued against Alfred Wegener's theory of continental drift, their mistaken conclusions stemmed not from poor reasoning but from the state of knowledge at the time. Moreover, the explanation offered by Opfer & Sloutsky–that the group member with the correct answer is simply more confident–has already been refuted in section 2.3 of the target article (for instance, how could that account for groups performing better than their best member?).

R3.3. The strength of the confirmation bias

When we look for arguments in a debate, we are mostly interested in arguments for our side or against the other side. This is why, we surmised, the confirmation bias is a feature of reasoning, at least in its role of argument producer. De Neys, Poletiek, Stupple & Ball, and Wolfe have reservations about the prevalence and robustness of this bias.

Poletiek questions the evidence from hypothesis testing problems cited in support of the idea that the confirmation bias is specific to reasoning (sect. 3.1). We agree with Poletiek that most of hypothesis testing is actually not directed by reasoning, and that confirmatory strategies are the result of heuristics that do not display a genuine confirmation bias. But this does not explain why people fail to adopt falsificatory strategies when they are asked to and adopt them spontaneously when they test someone else's hypothesis. It seems as though reasoning is unable to correct our own intuitions even though it can easily try to correct those of others.

Wolfe mentions a number of studies about the myside bias; that is, the tendency for participants to mostly–or only–give arguments that support their opinion. One of these studies show that, although “pervasive,” the myside bias could be oriented by the instructions (Wolfe & Britt (Reference Wolfe and Britt2008). Participants could be made to write essays against their opinion about a pretend requirement to “impose a 2-year math requirement for all students” (p. 8). But in this experiment, participants did not have to generate the arguments themselves. Instead, they were provided with a series of arguments for and against the position. As a result, they did not have to fight their confirmation bias and engage in the truly hard task of generating arguments against their own point of view. The resulting myside bias merely reflects a belief that it is better to provide arguments only for one's side rather than also for the other side. As Wolfe and Britt Reference Wolfe and Britt2008) observed, essayists often mention arguments for the other side–typically to rebut them and give even more weight to their own argument. But writing essays is as much of a basic argumentative skill as writing novels is a basic linguistic skill. It is therefore not surprising that untrained participants should exhibit a myside bias and that, as Wolfe points out, training can attenuate it.

The belief bias is one of the phenomena that, we surmised, show that people have a confirmation bias: They will take into account their beliefs about the conclusion when evaluating the logical validity of a syllogism. De Neys and Stupple & Ball question our interpretation of the belief bias data. They both point out that people–at least some people–try to engage in logical reasoning when faced with such problems. That they try is hardly surprising: Participants are merely responding to instructions that emphasize logical validity. We agree that in reasoning tasks people try to provide the correct, logically valid answer. What is more interesting is that most of them fail. Given that the tasks are not computationally hard, this indicates that reasoning is not geared towards pure logical validity, but that it takes into account other factors, such as believability.

R4.1. The algorithmic level

Our target article focuses on the ultimate level of explanation: What is the function of reasoning? A theory at that level has implications for the algorithmic implementation of reasoning, but this was not the subject matter of our article, and, to be candid, we are still in the process of working out these implications satisfactorily. At first blush, moreover, the argumentative approach does not imply a single, narrow characterization of this implementation. We therefore acknowledge this present limitation of our contribution that in particular Khlentzos & Stevenson have underscored, and we are grateful for the positive suggestions of several commentators.

Oaksford's Bayesian probabilistic approach (para. 4) allows for a fine-grained analysis of argument strength and of the changes in degree of beliefs they warrant. Like ours, this a computational theory, but it specifies a proximal function–how to change our beliefs in response to arguments–rather than an ultimate one. As a consequence, it has the potential of taking us one step further in the direction of an algorithmic theory.

Dessalles points out the sequential nature of reasoning, and asks how the argumentative theory can account for this feature. First, it is important to mention that even if the explicit part of reasoning is sequential–we make only one argument step at a time–other processes within reasoning (such as argument search) may well work in a parallel fashion. The sequential nature of explicit reasoning can be explained by its linguistic character (that might not be necessary for pure solitary reasoning if it exists, but that cannot be bypassed in argumentation), language itself being sequential, for several reasons unrelated to reasoning or argumentation (Pinker & Bloom Reference Pinker and Bloom1990; see also Carruthers Reference Carruthers1996).

Wiles mentions abduction as a plausible mechanism through which reasoning could find arguments. She points to other psychological mechanisms that also rely on abduction, such as spatial navigation, suggesting that reasoning might have been exapted from these mechanisms. However, given that reasoning deals with inputs and outputs that are very different from those of these other systems, the possibility of an evolutionary exaptation scenario remains very speculative. This does not mean that these other mechanisms have nothing to teach students of reasoning. To the extent that the problems solved by, for example, mechanisms of spatial navigation are similar to the problem facing reasoning–finding an acceptable argument within a large set of potentially relevant propositions–then, as suggested by Wiles's commentary, we can expect commonalities in the solutions used in both cases (for further elaboration on this point, see Mercier, submitted a).

Uleman et al. present interesting evidence and arguments on intuitive (or “spontaneous”) inferences. We agree that these inferences constitute most of cognition, and that they influence which arguments are used and how they are evaluated. We agree moreover that much more must be done on the relationship between intuitive inference and reasoning, even if much relevant work has already been done in the framework of dual-system approaches to reasoning.

Finally, Weber & Johnson offer a process-level specification of how reasoning works in decision making. According to their theory–query theory–participants facing a decision query their memory for relevant information or experiences. Two features of this theory are particularly relevant in the context of the argumentative theory. First, as a result of this process, “choice follows from the resulting balance of evidence,” so that this theory predicts reason-based choice. The second important point is that “the first query produces richer representations because of output interference” (para. 5). To the extent that further queries might represent perspective that are less congenial to the individual–for instance, trying to see the problem from someone else's perspective–a mechanism that favors the first query can create a form of confirmation bias. Given that query theory predicts reason-based choice and that it might also explain some forms of confirmation bias, it is quite congenial to the argumentative approach. However, rather than being a question of “intrapsychic and interpersonal argumentation” (para. 8), as Weber & Johnson suggest, the question can perhaps be more usefully framed as a difference in level of analysis. We believe that query theory could also help explain aspects of interpersonal argumentation–how we find arguments–whereas the argumentative theory also makes predictions regarding intrapsychic phenomena–such as reason-based choice. The difference is that query theory offers predictions based on the workings of reasoning, whereas the argumentative theory offers predictions based on the function of reasoning. Given that they are not at the same level, these explanations do not compete. Assume, for instance, that reason-based choice is entirely due to the processes delineated in query theory. We would still need to understand why this process is there in the first place, and why the outcomes of such a process are generally adaptive–questions that the argumentative theory aims at answering.

Another way to better understand the processes of reasoning is through modeling. Fox's Logic of Argument is a possible solution that is congenial to our proposal because it distances itself from formal logic to encompass the subtleties of argument structure. The fields of AI and informal logic are now teeming with models of arguments that can be conveniently simulated on computers, and this is certainly an interesting way to develop for any theory of reasoning or argumentation.

R4.2. Reasoning outside the lab

The target article focused its review on experiments carried out in the laboratory, mostly with Western college students. Narvaez rightly points out the limitations of such a narrow focus. In their review of cross-cultural psychology work, Henrich et al. (Reference Henrich, Heine and Norenzayan2010) have shown that in many domains WEIRD people–people from western educated industrialized rich democratic countries–behave in ways that are different from the rest of the world (para. 9). In the case of reasoning and argumentation, scholars have hypothesized that such skills are a mostly Western tradition, born in classical Greece and nurtured in the Enlightenment. It would indeed by a deadly blow to the theory if some cultures were unwilling to argue or unable to reason. Happily for us (and for these cultures), the available data do not point in that direction. While there certainly are differences in reasoning and argumentative style (e.g., see Norenzayan et al. Reference Norenzayan, Smith, Kim and Nisbett2002), there is no report of a culture that would be deprived of these skills. The two most frequently alleged cases are illiterate societies–which are supposed to be unable to reason–and Eastern cultures–which are supposed to be unwilling to argue. Yet members of illiterate populations can solve logical problems in the proper contexts, and the intellectual history of Eastern cultures is as fraught with debates as that of the West. While our theory could certainly be strengthened if the data from non-WEIRD cultures were richer, available data point to similar patterns the whole world over (Mercier, in press a).

Narvaez also regrets the absence of developmental data from the target article, accusing us of “implicit biological determinism” (para 5). Our thesis is evolutionary and pays attention to biological factors, but development itself is of obvious biological relevance. Again, language can be a useful analogy. Researchers who think that language is an adaptation do not deny that different languages are acquired in cognitive development. Likewise, the argumentative theory does not need argumentative skills to be just innate. The main reason so little space was devoted to an analysis of developmental evidence is that there is so much of it that it warranted a whole other article (Mercier, in press b). In this other article, it is shown that children can argue from very early on–long before they can do any abstract reasoning–that they are also prone to the confirmation bias, and that they reason better in groups–which is why collaborative learning has proven to be so successful in education.

Narvaez is joined by Wu in drawing attention to reasoning in the political sphere. The results discussed by Wu offer further support for the argumentative theory. In particular, he refers to the development of deliberative democracy as illustrating the power of group reasoning. Many arguments can be garnered in support of an increased participation of citizens in policy debates: more legitimate and fairer outcomes, increase in trust and positive public behavior, etc. But Wu points to a more recent trend that emphasizes the epistemic value of deliberative democracy: Beyond all their other advantages, debates can also yield epistemically superior outcomes (Cohen Reference Cohen1986; Estlund Reference Estlund2007; Landemore Reference Landemore2007). Empirical results in political science support the claim that groups can perform well, with deliberations yielding more informed and more coherent opinions and decisions (for review, see Mercier & Landemore, in press). Incidentally, the argumentative theory is also in a good position to help explain some of the blatant failures of political debates, whether it is polarization (Landemore & Mercier, submitted) or the dangers of widely publicized debates (Mercier, submitted b).

Johnson correctly argues that the theory should apply to scientists and philosophers–including the authors of this target article–as well as to laypersons. Indeed, it is not unlikely that, despite genuine efforts to remain objective, we have been guilty of the confirmation bias, thereby illustrating our argumentative theory by the very manner in which we were arguing for it. Argumentation and debates have always played a central role in philosophy, be it in classical Greece, Akbar's India, or the Warring States Period in China. The lone philosopher always runs the risk of using her great reasoning abilities to build a system of high internal coherence on shaky intuitive foundations. Even if scientists rely more on empirical evidence than do philosophers to correct their intuitions, their reasoning is still deeply argumentative. A scientific article or book is always a “long argument” in Darwin's famous words. Moreover–and contrary to popular representation of the lone genius–groups have always been the crucial place for scientific reasoning (e.g., see Dunbar Reference Dunbar, Steinberg and Davidson1995).

Johnson also points out the importance of arguments in the moral domain, suggesting that people might be especially vulnerable to arguments that exploit moral “oughts.” Some moral psychologists would disagree and suggest instead that people are rarely receptive to moral arguments, being more easily influenced by narratives or emotional appeals (Bloom Reference Bloom2010; Haidt & Bjorklund Reference Haidt, Bjorklund and Sinnott-Armstrong2007). The argumentative theory predicts an intermediary state of affairs: People should be somewhat receptive to moral arguments while evaluating them on the basis of their own moral intuitions (Mercier, in press c).