by Dan Sperber
Scientists, it is true, do achieve some serious and, over time, increasing degree of convergence. This may be due in part to the many carefully developed methods that play a major role in conducting and evaluating scientific research. There are no instructions, however, for making discoveries or for achieving breakthroughs. Judging from scientists’ own accounts, major advances result from hunches that are then developed and fine-tuned in the process of searching, through backward inference, for confirming arguments and evidence while fending off and undermining counterarguments from competitors.
Arguably, science is the area of human endeavor where rationality and good reasoning are most valued (as we will discuss in detail in Chapter 18). Reasoning, however, does not cause scientists to spontaneously converge on the best theories. It causes them, rather, to elaborate and vigorously defend mutually incompatible competing theories. It also helps them—with a higher, even if still imperfect, degree of convergence this time—to evaluate competing theories and reach some degree of tentative agreement on the current winners of the competition. Winning in this competition may well be, to a large extent, a matter of epistemic luck—having invested in a better initial hunch for which there are stronger evidence and better arguments to be found—more than of better reasoning in producing new ideas. What scientific practice suggests is, on the one hand, that various specific methods can be relevant to specific reasoning tasks and, on the other hand, that there is no such thing as a general method for reasoning.
The study of reasoning has been dominated by a normative goal and a descriptive expectation: the goal of discovering and making explicit a general method that could produce good reasoning and the expectation that actual human reasoning would be guided by an approximation of such a method. Both the goal and the expectation have, so far, been disappointed. This failure is not an accident. There is a principled explanation for it. The procedures of intuitive inference, we have argued, are unconscious, opportunistic, and diverse. The idea that intuition might consist in following a general method makes little sense.
Our higher-order intuitions about reasons for some intuitive conclusion take into account many of the properties of the lower-order intuitions they are about, properties that may vary greatly from one domain to another. This is not bad, let alone fallacious, reasoning. In putting our reasons in argumentative form to convince others, sometimes we appeal to their own nuanced higher-order intuitions, sometimes we choose to be schematic, and often we do a bit of both. The schematism we employ, rooted in ordinary language, has inspired one of the great intellectual achievements of human history: the development of the science of logic. Logic, however, tells us neither how we reason nor how we should reason. There is no general method that we could and should follow when reasoning, either on our own or in dialogue with others.
10
Reason: What Is It For?
Look at the piece of furniture that appears in Figure 16 on the following page. Hardly a well-designed chair, is it? The seat is so low that you’d practically have to squat on it. The back is too high; the upper bar is a horizontal slab on which you could not lean comfortably. How come? Actually, this is not a chair. It is a church kneeler. The height is right for kneeling. The “back” is in fact the front, and the upper slab is there to rest not one’s back but one’s praying hands. Once this artifact’s real function is recognized, what looked like flaws turn out to be well-designed features.
Like the object in Figure 16, reason seems to have an obvious function: to help individuals achieve greater knowledge and make better decisions on their own. After all, if using reason doesn’t help one reach better beliefs and choices, what is it good for? However, like a kneeler used as a chair, reasoning serves this function very poorly.
We’ve already examined quite a few failures of reasoning, and we will look at many more in Part IV, but the upshot of experimental research on how we reason is that we do so in a way that is biased and lazy and that often fails to solve even simple but unfamiliar problems. It would be easy to stop here and simply conclude, as many psychologists have done, that human reason is just poorly designed.
Another conclusion is possible. The true function of reasoning may have been misunderstood. Reasoning might be the kneeler-mistaken-for-a-chair of modern psychology. In Chapters 7, 8, and 9, we have argued that reasons are commonly used in the pursuit of social interaction goals, in particular to justify oneself and to convince others. Here, we adopt an evolutionary approach to argue that these social uses of reason are not side effects or minor functions of a cognitive mechanism having as its main function the enhancement of individual cognition and that, on the contrary, the main function of reason is social. Why resort to an evolutionary approach? Because this is the only approach that explains the fact that complex inheritable traits of living things tend to produce beneficial effects. Outside of an evolutionary perspective, it is quite unclear why human reason, or anything else for that matter, should have any function at all.
Figure 16. A poorly designed chair?
Functions before and after Darwin
That human reason has a function is an idea commonly taken for granted and one that we want not to challenge but to develop on a new basis. To begin with, let us be clear about the very notion of a function: discussions about the evolution of reason are often marred by a superficial understanding of the notion.
What is a function? How can one decide what is the function of an artifact such as a chair, a natural quality such as a flower’s scent, a body part such as a wing, or a mental capacity such as human reason? The two questions—what is a function, and how do we identify a particular function—are closely related. For artifacts, a rough answer is generally easy to produce. The function of an artifact is what it is made for. Who decides what an artifact is for? Whoever designed it. A chair is made for sitting, a kneeler for kneeling, and so on.
Attributing a function to a biological trait is much less straightforward. No one made wings for flying or the heart for pumping blood. And yet, who would doubt that wings are for flying? Who would contest William Harvey’s discovery of the function of the heart?
Before Darwin, attributing a function to some trait of an animal or a plant involved little more than answering the question: What is it good for? Find a good effect of a biological trait and just assume that what it is good for is what it is for. What could be more useful than flying? So, the function of wings is to enable flying. Such a commonsense notion of function is not useless; it has played an important role in science before Darwin, such as in Harvey’s discovery of the function of the heart, and it still plays at least a heuristic role in modern science. What question such an ahistorical notion of function raises and fails to solve is: Why, in the first place, should we expect biological traits to have useful effects? We don’t expect anything of the sort in the case of other natural objects such as stones, planets, and subatomic particles. These don’t have functions. For a long time, the only answer was that God, the maker of all things, has made wings for flying, hearts for pumping blood, and so forth—an answer that replaced a series of interesting problems with an unfathomable mystery.
It was Darwin’s theory of natural selection that provided the basis for a coherent and useful notion of a function based on a genuine explanation of why organs and other traits of living things have the functions they have.
Any inheritable trait of an organism has many effects. Most of these effects are without consequences for reproductive success. The scent of a flower, for instance, might make some local animals dizzy without this harming or benefiting the plant in any way. Some effects of a trait may happen to enhance reproductive success: the scent of a flower may attract insects and recruit them as pollinators. Still other effects of a trait may actually compromise reproductive success: a scent might attract florivores and entice them to eat the flower, thereby diminishing the amount of pollen available to pollinators.
When overall the effects of a heritable trait are more benef
icial than harmful to the reproductive success of its carriers, then this trait is likely to be selected and to propagate over generations. Selected traits have at least one of their effects that has contributed to the reproductive success of their carriers and thereby to their own propagation. The scent of wildflowers, for instance, has been selected because by attracting pollinators, it contributed to the reproductive success of the plants. When biologists talk of the function of a selected trait, they refer to such a beneficial effect.
Selected traits may have more than one function, and these functions may change in the course of evolution. Penguins, for instance, are descendants of birds that had used their wings to fly, but the function of penguins’ wings is to help them swim. At some point in their evolution, these wings must have served both functions, as they do in some seabirds, such as in razorbills.
Traits that owe their propagation to their function(s) are adaptations. Not all features of an organism are adaptations. Some features result from physical or chemical constraints on biological development. The fact that all organisms contain carbon, for instance, is a basic chemical property of life itself (as it is found on earth, at least). Not all features of an adaptation are adaptive, either. Many are side effects. The regular contractions of the heart are an adaptation. They have the function of causing blood flow throughout the body. The noise produced by these contractions, on the other hand, isn’t adaptive. It is a side effect.
The Function of Reason from Aristotle to Twentieth-Century Psychologists
Just like ideas about the function of wings, ideas about the function of reason emerged well before Darwin and the development of a scientific notion of function. Even after Darwin, much evolutionary thinking on reason shares with pre-Darwinian views a sense that the function of reason, just as that of wings, is obvious enough. Reason is a means for individuals to acquire superior knowledge and to make better decisions. Reason, by performing this intellectual function, elevates humans above all other animals.
Darwin himself expressed this same general idea: “Of all the faculties of the human mind,” he wrote in The Descent of Man, “it will, I presume, be admitted that Reason stands at the summit.”1 He put, however, an evolutionary twist on this classical view. Through his intellectual faculties, he argued, man “has great power of adapting his habits to new conditions of life. He invents weapons, tools, and various stratagems to procure food and to defend himself. When he migrates into a colder climate he uses clothes, builds sheds, and makes fires; and by the aid of fire cooks food otherwise indigestible.” Because these intellectual faculties are variable and tend to be inherited, “they would have been perfected or advanced through natural selection …. In the rudest state of society, the individuals who were the most sagacious … would rear the greatest number of offspring.”2
The functional effects of reason are roughly the same for Darwin as they were for Aristotle. What is new with Darwin, however, is the use of these effects to explain why reason should have evolved. And because Darwin is making a more precise claim about the function of reason, his claim is more open to a scientific challenge. If reason evolved to help individuals think on their own, then it should really provide for truly better thinking in terms of both cognitive benefits and mental costs, and if it does not, as much of modern psychology of reasoning suggests is the case, then we are indeed faced with a serious problem.
Psychologists who discovered what looked to them like major flaws of human reason must have, you would think, understood and discussed the challenge this presented for a post-Darwinian understanding of the function of reason. Actually, they did not. Until the 1990s, evolution, natural selection, and biological function were hardly ever mentioned in the psychological literature on reasoning. Psychologists just took for granted that the function of reason—with “function” understood as a commonsense rather than biological notion—was to enhance individual cognition. They then concluded that reason was not performing its function as well as had been assumed.
If anything, psychologists saw flaws in reasoning as evidence against an evolutionary approach. Evolved mechanisms should perform their function well enough to have been selected. Reasoning, with all its flaws, is not properly geared to the pursuit of knowledge and good decision. Hence, many concluded, the case of reasoning shows the irrelevance of the evolutionary approach to human psychology.3 Of course, psychologists might have rethought the function of reason instead of uncritically accepting common wisdom in the matter, but they did not.
While the foundations of an evolutionary approach to psychology were laid by Darwin himself and some relevant work had been done over the years, especially in the study of animal behavior, a general and systematic discussion of the evolution of human cognitive mechanisms and of reasoning in particular started only when psychologist Leda Cosmides and anthropologist John Tooby outlined an ambitious program of “evolutionary psychology” in the 1980s.4
“Has Natural Selection Shaped How Humans Reason?”
From the start, human reasoning has been a major topic of research in evolutionary psychology and a main focus of polemics for scholars opposed to this approach. The polemic erupted when, in 1989, Leda Cosmides published what became the most controversial article in the history of psychology of reasoning. It was entitled “The Logic of Social Exchange: Has Natural Selection Shaped How Humans Reason?”5 (hence the title of this section).
There were in Cosmides’s article two major theoretical claims. From an evolutionary point of view, she argued, one should expect reasoning mechanisms to have evolved as responses to specific problems that had been recurrent in the ancestral environment. Specialized adaptations do a much better job of addressing the problems they evolved to handle than would any general problem-solving ability. In fact, it is not even clear what an ability to solve problems in general might consist of (if not, precisely, of a complex articulation of a great many more specialized mechanisms). A general reasoning ability, if such a thing could evolve at all, might, at best, help in dealing with the residue of problems not properly handled by specialized mechanisms. On tasks of no evolutionary significance (such as those typically given in reasoning experiments), Cosmides argued, one should expect reasoning to be much less effective than on problems for which dedicated mechanisms have evolved.
Cosmides’s second and central claim was that among the many cognitive adaptations most likely to have evolved, there had to be an inferential mechanism aimed at solving a major problem raised by cooperation. Cooperation is an interaction between two or more individuals where each incurs a cost and receives a benefit. Each member of a party of big game hunters, for instance, spends time and takes risks to help the group catch a prey. Provided that the benefits are greater than the costs and are shared fairly, cooperation is advantageous to all cooperators. However, in many cases, each cooperator stands to gain even more by sharing in the benefits of cooperation without paying the full cost, in other terms by cheating or free riding. Hunters who avoid taking risks but take the same share of the meat as others are at an advantage. Of course, if none of the hunters take risks, the prey will escape. Widespread free riding is likely to result in failure of cooperation. For cooperation to be profitable and hence to endure, cheaters must be identified and either controlled or excluded.
Cosmides and Tooby argued that the identification of cheaters should be seen as a major problem in human evolution, favoring the emergence of a specialized module capable of computing the rights and duties of cooperators as a basis for detecting cheaters. To test this hypothesis, they relied massively on the Wason selection task, which we talked about in Chapter 2 and which we don’t see as adequate for the job.6 Our goal here, however, is not to evaluate Cosmides’s arguments for the existence of an evolved ability to detect cheaters (which we see as strong) or her experimental evidence (which we see as weak). It is to use her approach as a source of inspiration and of contrast to propose another, no less radical and even more ambitious evolutionary hypothesis about t
he function of human reason in general, and reasoning in particular.
Like Cosmides and Tooby, we expect an evolved mind to consist principally in an articulation of modular mechanisms. Modules are specialized—each helps solve some specific type of problem or take advantage of some specific type of opportunity in a way that contributes to biological fitness. Reasoning as classically understood, on the other hand, is allegedly in the business of addressing any kind of problem and finding ways of taking advantage of any kind of opportunity. Classical reasoning is not a specialist but a generalist, not a narrow module but a broad faculty.
If the job of reasoning is much too broad to be that of a module, there seem to be only two possible conclusions: either the central and most distinctive component of the human mind is not modular, or else reasoning as classically understood doesn’t even exist. Cosmides and Tooby favor the second conclusion: contrary to a long-held dogma of philosophy and psychology, there is no such thing as reasoning in general.7 Other researchers working on the evolution of the mind, such as the psychologist Cecilia Heyes and the philosopher Kim Sterelny, favor the first possible conclusion: the existence of domain-general reasoning is evidence that the human mind is much less modular than Cosmides and Tooby and others have argued.8
Are these the only two possibilities? No. As we have shown (in Chapter 6), a metarepresentational module processes a very special type of object, representations, and attends to properties specific to these representations. Such a module may nevertheless indirectly provide information about the very topics of the representations it processes. Mindreading, for instance, informs us about not only the thoughts of others but also all the things and events these thoughts are about. Reasoning, we have argued, is produced by a metarepresentational module, the specific domain of which is the relationship between reasons and the conclusions they support. These reasons and conclusions, however, can themselves be about any topics. As a result, inferences about reasons-conclusions relationships indirectly yield conclusions in all domains, indirectly providing a kind of virtual domain-generality.
