The Princeton Guide to Evolution

Cognition: Phylogeny, Adaptation, and By-Products

Marc D. Hauser

OUTLINE

1. What are we measuring?

2. The space of possibilities

3. Novel possibilities and unanticipated outcomes

4. Evolving limitless options

The mind consists of feelings, decisions, plans, and memories generated by the brain. To study how minds evolve, a comparative approach is necessary, one that seeks evidence of phylogenetic similarities and differences, together with evidence of adaptive function. This chapter describes a set of challenges associated with exploring mental evolution, together with a framework for exploring a corner of this problem, focused on the patterns and processes that led to the evolution of human minds. This is a story of phylogeny, adaptation, and by-products. The chapter examines the hypothesis that despite some similarities between human and nonhuman animal minds, there are far greater discontinuities. The uniqueness of human cognitive capacity is due to a suite of changes in brain function that generate the signature of both human universals and cultural variation in language, music, mathematics, technology, and morality.

GLOSSARY

Abstract Thoughts. The ability to represent or think in ways that are detached from the primary sensory and perceptual inputs. This capacity allows individuals to think about things that are beyond their direct experiences with the world, including concepts such as infinity and objects that are possible but do not exist, such as hippos with green fur and pink antlers.

Cognitive Decomposition Approach. Dissecting cognitive capacities into their component parts to study the evolutionary history of the components. Students pursuing this approach decompose language, music, mathematics, technology, and morality into a suite of capacities and explore the phylogenies of these components, as well as the evolutionary processes that may have favored them.

Cognitive Promiscuity. The capacity to combine thoughts and feelings from different domains of understanding to create novel solutions. When cognitive promiscuity is in play, capacities that evolved to solve a highly specialized problem are also used to solve problems in other, disparate situations.

Combinatorial Operations. Mixing or combining discrete elements to create novel combinations, and thus new meanings or functions.

Recursive Operations. Computable functions that generate a potentially infinite set of hierarchically organized expressions.

Symbolic Expression. The capacity to represent and express a thought or emotion with a discrete symbol, including spoken, signed, or written words, as well as nonlinguistic representations (e.g., McDonald’s golden arches, scales of justice). This capacity allows individuals to reduce memory load through compact storage of discrete and easily retrieved symbols while facilitating lossless information transfer.

The anthropologist Martin Muller has an arresting photo of his hands, each palm up, each holding a part of a recently killed chimpanzee from Uganda. In one hand is the chimpanzee’s brain; in the other, one of its testicles. Both are virtually the same size, about the dimensions of a tangerine. Both have distinctive functions; both are targets of selection, with ancient evolutionary histories. All these comparative points are true of humans as well, except one: a human brain would completely cover and outsize a human hand, whereas a human testicle would sit easily within the palm. For any healthy human adult, Muller would be holding a melon in one hand and a diminutive grape in the other.

If we look beyond brain size and male reproductive organs, as well as many other features of comparative anatomy, the differences between humans and chimpanzees are even more striking. Chimpanzees have shown minuscule changes in the 6–7 million years of their evolution. Not one chimpanzee has ever ventured out of Africa to another continent. Not one chimpanzee has moved out of its own country, say, from Tanzania to Kenya. Not one chimpanzee has moved out of its habitat of origin in the tropical forests to life in the mountains or on the beach or in the desert. It appears that chimpanzees use the same communication today as they did millions of years ago. They also have the same mating system and social organization. And although they have made a few clever technological innovations, their material culture is largely the same today as it was in the past.

If you believe these comments belittle chimpanzees, you are wrong. What chimpanzees can do is impressive. Moreover, they have persisted for 6–7 million years, so they are doing something right. Finally, the observation of cognitive and behavioral stasis is true of virtually every other animal—virtually all but one: Homo sapiens.

In the 6–7 million years of our evolution, humans have marched out of Africa, inhabited every continent and virtually every habitat offered on earth, and ventured to novel environments on other planets. Our communication today preserves elements of our past but goes far beyond it, quantitatively and qualitatively. Our unique faculties of language and mathematics allow for a limitless variety of expressions, including massive compression of ideas into words and patterns of 1s and 0s. Our material culture today would be incomprehensible to our human ancestors. We have gone from chipped stone hand axes to air-powered hammers, cell phones, airplanes, and scud missiles. All these developments have fueled our nomadic travels and capacity to inhabit a bewildering diversity of environments. We also have transitioned from universal polygyny to a diversity of mating systems that includes polygyny, lifelong monogamy, and serial monogamy, and within each of these systems, opportunities to engage heterosexually, homosexually, or bisexually.

This comparative summary raises several profound evolutionary questions: What accounts for the differences we observe between modern humans and chimpanzees? What changes in the brain allowed us, but not our closest living relatives, to change our material cultures, mating systems, living environments, and systems of communication and thought? What evolutionary processes led to the differences between humans and chimpanzees?

These questions keep scholars of cognitive evolution up at night. Our goal is to provide a few answers to these questions, perhaps allowing for greater rest; then again, perhaps they will stimulate even more restlessness.

A few caveats before starting. There are two reasons to focus primarily on primates, and especially the contrast between humans and chimpanzees. First, there is no other comparative contrast that yields such a striking cognitive gap, especially among such closely related species. To put it starkly, though the molecular evidence shows that humans are more closely related to chimpanzees than chimpanzees are to gorillas, the cognitive evidence shows few differences between chimpanzees and gorillas but massive differences between both apes and humans. This is a delicious problem. It shows both the difficulty of moving from genetic to phenotypic differences, and the challenges of creating phylogenies based on different types of evidence. Second, because of the richness of our understanding of human cognition, including its neural underpinnings, we are well equipped to ask deep questions about comparative differences and similarities. Thus, for example, when we ask whether other animals imitate, experience empathy, perceive musical patterns, or deceive like humans, we can rely on a wealth of empirical evidence from cognitive science, neurobiology, and most recently, molecular neuroscience. Homo sapiens has joined the ranks of C. elegans and Drosophila as a model species.

To make sure there is no confusion, none of these initial comments are meant to diminish the elegant comparative evidence for other taxonomic assemblages, including experimental investigations of spatial navigation in insects, fish, birds, and other mammals, and the proximal mechanisms and selective pressures on caching in birds and rodents. In a short chapter, choices are necessary. I hope mine provide a reasonable introduction to some of the most pressing challenges and exciting developments while acknowledging the particular theoretical biases I hold.

1. WHAT ARE WE MEASURING?

Unlike measuring testicles or the brain, asking questions about cognitive evolution poses a measurement problem. What can we measure so that we may both make phylogenetic comparisons and devise experimental methods to test for selective pressures and the adaptations they generate? One approach, common to this day, has been to seek comparative evidence of behaviors that are unambiguous indicators of human intelligence, including language, mathematics, culture, music, and cooperation. But this approach is flawed. None of these traits are clearly isolated phenotypes like testicles and brains, readily quantifiable for entry into a phylogenetic analysis or experimental test of adaptation. All these traits depend on a variegated set of mechanisms, some shared in common with other animals, and some uniquely human. Each mechanism, in turn, generates a set of signature behaviors that can be quantitatively measured. To productively explore questions of cognitive evolution, we must decompose these complex phenotypes into their component parts. This is called the cognitive decomposition approach.

The good news is that more recent work in cognitive evolution has recognized that phylogenetic and functional analyses require more precise and narrowly specified traits. For example, a number of scholars, such as Derek Bickerton, Terry Deacon, and Tecumseh Fitch, have recognized the need to decompose language into separate mechanisms, including computations for structuring words into sentences, mapping concepts to words, and articulating words with sounds or visual signs. Those interested in the evolution of music, such as Ray Jackendoff, Fred Lehrdahl, and Ani Patel, have appreciated the need to decompose this system into auditory perception, planning, memory, pattern analysis, and combinatorial computations, to name a few. Those interested in the evolution of mathematics, such as Susan Carey, Stanislas Dehaene, Randy Gallistel, and Elizabeth Spelke, have looked at different mechanisms of quantification, including those that rely on discrete and explicit symbols such as the integers, as well as those that do not; and the ways in which quantification relies on more general mechanisms of categorization, memory, and attention; and at how nonexplicit symbol systems for quantification evolve and develop into explicit ones. Students interested in cultural evolution, such as Robert Boyd and Michael Tomasello, have recognized the importance of looking at different mechanisms of transmission, including teaching, imitation, and observational learning, together with the importance of innovation, conformity, and social organization. From such decomposition, students of cognitive evolution have made great strides over the past 15 to 20 years.

There are two important points to note about the cognitive decomposition approach. First, it starts with questions concerning the nature of a particular mechanism and then follows with questions related to adaptive function, including the socioecological conditions that favored its original expression. Consider, for example, the capacity for teaching, an ability that is observed in every human culture, in a wide variety of contexts, and is dependent on different cognitive processes. When humans teach, it is recognized that some individuals are ignorant and may want to learn. To teach requires demonstrating, breaking problems down into simpler components, recruiting the learner’s attention, monitoring their progress, revising the pedagogical approach, and so forth. In other animals, teaching appears highly limited in taxonomic scope, and among those animals exhibiting some form of teaching, it appears limited to a single context. Thus, meerkats engage in a form of functional teaching in which adults help prepare pups to develop the skills to kill scorpion prey. It is a form of teaching in that the adults recognize a deficiency in young individuals and then break down the mature form of the skill into components to facilitate learning. Unlike humans involved in teaching, however, meerkats engage in this kind of pedagogy in only one context: predation on scorpions. At present, we lack a coherent account of why even this form of teaching does not occur in other functionally significant contexts among meerkats and why other species with similar social and ecological pressures lack teaching altogether.

Second, the decompositional approach seeks to understand whether the mechanism in question evolved to solve a suite of general problems or a highly specific one while recognizing that a specialized mechanism can be co-opted for more general purposes. Three examples highlight the significance of this point:

• The temporal lobe of the primate cortex evolved for object recognition, but in humans it has been co-opted for the added task of recognizing written word forms.

• A region of the parietal lobe operates in nonhuman and human primates for approximate number estimation, but in humans it is also recruited for precise number computation with explicit symbols.

• A region at the juncture between the temporal and occipital lobes is involved in face recognition in both monkeys and humans—a highly specialized function—but in humans it is also used for other within-category discriminations (e.g., cars) that are unique to our species.

These three cases illustrate why, for any given cognitive function in an organism, it is important to ask for what it evolved, for what it is presently used, for what it could be used, and whether each of these functions is general or specific to a particular problem. The comparative evidence available suggests that most cognitive functions observed in nonhuman animals have evolved for highly specialized problems and are restricted to use in this context. In many cases, we have inherited these highly adaptive capacities, but owing to evolutionary changes in the brain, coupled with the particular environments we inhabit, these specialized functions have been liberated, allowing us to tackle a much broader range of problems. In fact, it appears that many of our most revered cognitive capacities are by-products of a few specific changes in neural function, capacities that are nonetheless highly adaptive today.

2. THE SPACE OF POSSIBILITIES

The fact that traits can evolve for one function and subsequently be used for another raises a challenge for students of cognitive evolution, a problem that finds parallels within the field of functional and theoretical morphology. In particular, as students of evolution, we typically study what is observable or what has been observed in the past if we have access to a fossil record. What is less often studied, except perhaps by students of artificial life and evolution, is what is potentially observable. What we observe reveals information only about the options that species have explored over their history. It does not reveal what could have occurred had they confronted different ecological or social pressures. This is a problem of potential, of possible phenotypic outcomes. In the case of anatomy, say, the coiling patterns of ammonite mollusks, it is a question of the possible shapes they might take under different conditions. In the case of behavior, say, courtship displays in birds, it is a question of the possible movements of the body and vocal tract that might evolve, either to accommodate shifts in perception of the choosy sex or to accommodate changes in the physical environment. In the case of cognition, say, pattern recognition, it is a question of possible ways of computing and classifying patterns that might change as animals develop different systems of communication or social relationships. The idea that for every organism there exists a space of possible anatomical, behavioral, and cognitive forms that have yet to be realized changes how evolutionarily oriented research programs are carried out. It is no longer sufficient to study what animals do or have done. It is important to study what they might do when confronted with different situations. In the case of cognition, it becomes a matter of challenging nonhuman and human animals with novel problems. The process of domestication provides an elegant illustration of this problem, and of the unanticipated consequences that can emerge when the selective regime changes in a radical way.

3. NOVEL POSSIBILITIES AND UNANTICIPATED OUTCOMES

Artificial selection (or experimental evolution; see chapter III.6) provides one way to explore the space of possible outcomes. Artificial selection provides a way of tapping into human creativity, allowing us to use our imagination to select for what nature may never imagine. Could we create a square tomato or a hen that lays square eggs, both optimized for packing? Could we select for hairless pets, reducing the mess at home and allowing those with allergies to enjoy the company of other animals? Some of these questions have already been asked and explored, generating a list of successes and failures. The successes show the power of our imagination and selection. The failures reveal either the poverty of our imagination or the hidden constraints that limit the power of selection. Both successes and failures inform our understanding of evolution. Recent work on mammalian domestication provides an elegant illustration, especially with respect to the role of selection in both behavioral and cognitive evolution.

Molecular evidence shows that the domestic dog first differentiated itself from its ancestor, the wolf, about 100,000 years ago. Dogs begin to appear in cave art, and the archaeological record more generally, between 14,000 and 16,000 years ago. Though there is little explicit information about the process of domesticating dogs, most scholars believe that as human populations adopted a more sedentary lifestyle, which included the introduction of agriculture, wolves started scavenging for food. Over time, this process initiated a cascade of morphological, physiological, behavioral, and cognitive changes that led to the first dog breeds. This first wave of change was followed by a second in which humans played an increasingly more directed role in creating new breeds, selecting for differences in size, coat coloration, temperament, and behavioral skills. Some of these differences were selected for pure aesthetics (e.g., shorter snouts, pint-sized bodies, floppy ears), others for functional differences linked to human lifestyle (e.g., herding dogs for those with livestock).

Behavioral research in the last 10 years suggests that the process of domestication in dogs resulted in fundamental changes in cognitive ability—differences that are striking when contrasted with those of their close relatives, the wolves, and their more distant relatives, the nonhuman primates, including chimpanzees. There are two fundamental questions here: What critical differences in cognitive ability are specifically due to domestication? And did these capacities evolve as a result of selection or are they by-products of other cognitive changes? At least part of the answer to these questions comes from an elegant series of studies on dogs and wolves, focusing on their capacity to understand the communicative gestures of humans, especially in relationship to nonhuman primates. Another part of the answer comes from a more targeted project involving the domestication of the silver fox.

When humans point, or turn their heads in a particular direction, those watching immediately understand that the space of interest has shifted. People point and turn their heads to indicate a change in focus, and often, to cause others to change their focus accordingly. Early in development, without any training, human infants recognize that pointing is goal directed, designed to capture another’s attention. Infants point to indicate an object or event of interest, and to obtain help in grabbing something that is out of reach. If an observant infant knows that a desirable toy has been hidden in one of two boxes but doesn’t know which one, it will readily find the toy by following an adult’s pointing gesture. Strikingly, chimpanzees presented with the same situation largely fail to find the hidden object; many individual chimpanzees will even fail after repeated presentations. This is a robust and striking failure. In contrast, dogs presented with the same hide-and-point task readily find the hidden object, and often do so on the first presentation.

The comparative evidence on human infants, dogs, and chimpanzees led to the hypothesis, championed by Brian Hare, that perhaps domestication was responsible for the convergence in cognitive ability between humans and dogs. If domestication is responsible for the dog’s cognitive prowess, then wolves should behave like chimpanzees in this task, and fail to understand human pointing, even after repeated exposure. This assumption was borne out in the first series of studies, providing stronger evidence for the role of domestication in dog cognition.

One problem with this comparative evidence is that not only are dogs domesticated whereas chimpanzees and wolves are not but dogs are reared from birth to adulthood in a human environment by humans. Perhaps dogs succeed on the pointing task because of their experience in a human environment, one rich in human pointing. That is, dogs learn about the goal of pointing through either passive exposure and associative mechanisms or through training. To tease apart these factors, wolf pups were raised by humans and then tested on the same task. These individuals readily followed the pointing gesture to the target box. This suggests that the human environment makes a direct contribution to the cognitive abilities of dogs. Further support for this claim comes from the fact that stray dogs, with either no human input or minimal experience, perform like wild wolves. Thus even domesticated dogs, lacking significant exposure to a human environment—and thus functionally feral—have poor comprehension of human communicative gestures.

Where domestication appears to play a unique role in dog cognition is in the relative sophistication of their capacity to read human gestures, a point supported by the work of Hare as well as Adam Miklosi. When dogs reared by humans are compared with wolves reared by humans, dogs perform far better when more subtle forms of pointing are presented, including foot pointing, momentary pointing at a distance, and reading the direction of eye gaze. What these results suggest is that a combination of selective pressure and early experience has shaped the cognitive ability of dogs. This is a lovely illustration of the interplay between evolutionary and developmental processes, or evo-devo. What this work is unable to demonstrate is whether the cognitive capacity of dogs (at least with respect to understanding communicative gestures) was selected for in its origins or is the accidental by-product of selection for some other capacity. To address this problem, we turn to work on silver foxes.

The Russian biologist Dmitri Belyaev decided to gain a deeper understanding of the process of domestication by systematically starting the process and controlling it over time. He started with a large population of silver foxes and divided them into several groups. Once they were settled, a human experimenter approached a group and offered some food. Those individuals who stayed near were selected and bred. This process was iterated over 30 generations. The outcome: a tame fox, completely unafraid of humans. This is precisely what Belyaev was hoping for. What he didn’t expect, and certainly didn’t select for, was floppy ears, a piebald coat, a curly tail, smaller brain, higher levels of serotonin (a neurochemical that regulates self-control), and much greater social cleverness. When tested in the hide-and-point task, domesticated silver foxes outcompeted their wild relatives. They performed at the level of domesticated dogs.

The fox studies led to the conclusion that selection for tameness resulted in a suite of by-products, including social cleverness. Although this is a reasonable conclusion, there is an alternative: those foxes paying attention to the experimenter’s eyes, and recognizing the lack of threat, were most likely to stay close. These were socially clever foxes, and they were selected. Though Belyaev thought he was selecting for tameness, he was actually selecting for those individuals most likely to correctly interpret human gestures, including especially the direction of their eye gaze. Belyaev actually selected for cognitive capacity. To rule out this alternative, it would have been necessary for him to approach the foxes with covered faces.

Together, these studies of dogs, wolves, and foxes show both the power of selection and developmental processes to modify phenotypes, and the fact that even targeted selection can result in unforeseen consequences. What has evolved doesn’t completely inform what is ultimately possible. Hidden within every organism is a suite of unrealized possibilities, waiting for novel opportunities, either naturally occurring or artificially imposed.

4. EVOLVING LIMITLESS OPTIONS

I stipulated in the introduction that the discontinuity between human and nonhuman minds is massive. I also stipulated that the cognitive chasm separating these two taxonomic groups is much larger than any other comparative contrast. For example, though many have claimed that the monkeys and apes far outperform birds in most cognitive tasks, recent work by Nicola Clayton and Nathan Emery on corvids shows that the gap is actually small, and in some domains such as creative tool use, these birds have the upper hand. Some scholars, such as Marc Bekoff and Frans de Waal, fundamentally disagree with the thesis that there is significant discontinuity in the cognitive achievements of humans and other animals, seeing far greater continuity. Thus, for example, Klaus Zuberbuhler has suggested that monkeys have vocalizations with a semantic richness that is akin to human words, and in some cases they combine these vocalizations in ways that approximate human sentences and their syntactic structure. These claims are controversial. Most linguists conclude that the origins of language, especially its semantics and syntax, cannot be traced to any of the communicative systems observed in other animals. None of these systems have the generative capacity of human language, including the young child’s ability to spontaneously acquire a massive vocabulary and to combine and organize these words into novel expressions—that is, to tap the generative power of our syntax to create a limitless number of meaningful sentences. The same claim holds for music, mathematics, technology, and morality. Some components of these systems show continuity with other animals, but the most powerful components show no parallel at all. Why?

To answer this question, we need to consider how the human brain generates variation in behavior, and in particular, a flexible and creative set of responses to novel situations. Humans uniquely evolved four distinctive mechanisms that solve this problem: generative computation, cognitive promiscuity, abstract thought, and symbolic expression. Generative computations consist of recursive and combinatorial operations. Recursive operations are computable functions that generate a potentially infinite set of hierarchically organized expressions. Combinatorial operations entail mixing discrete elements to create novel combinations, and thus new meanings or functions. Cognitive promiscuity allows wildly different ideas to couple, blending thoughts and feelings from different domains of understanding. Abstract thoughts are detached from the primary sensory and perceptual inputs, allowing us to think about things that are beyond our direct experiences with the world. Symbolic expressions are discrete representations of thoughts, ideas, or emotions, such as spoken, signed, or written words, as well as graphic icons.

The generative computations provide the engines of variation. Cognitive promiscuity enables us to go beyond the often specialized and myopic mechanisms that evolved to solve problems in one context or domain, such as foraging, tool use, cooperation, or communication. Abstract thoughts allow us imagine possibilities that extend far beyond what we feel, see, smell, hear, and touch. Symbols allow us to reduce memory load through compact storage of discrete and easily retrieved symbols while facilitating lossless information transfer. The following makes this less abstract.

A wide variety of animals use tools to solve particular environmental problems. In this sense, tool use is most definitely not unique to humans. This is, however, where the convergence ends. Unlike animal tool use, including that by chimpanzees and New Caledonian crows, human tool use makes use of the four variation-generating mechanisms just noted. For almost any tool that one can think of today, materials are combined to create objects with multiple functions—think pencil or Swiss army knife. The design for almost any tool that one can think of derives from cognitive promiscuity—pencils are designed to write, thereby physically transforming blank pages into marked-up ones while conveying our thoughts and feelings. Here physics, language, and social psychology are combined. In the case of many tools, their inventors went beyond direct experience to create an object with a particular function—no one ever experienced a writing device until someone imagined that marking up stuff, as in the earliest cave art, was important, both for telling stories and conveying information linked to survival. For many tools, the object not only serves one or more functions but is a symbol itself—consider the scales of justice. Animal tools serve one function, never consist of more than one material, are created from direct experience, and do not function as symbols. Animal tools lack the signature of generativity, promiscuity, abstraction, and symbolism.

Though we do not know when, how, or why these four uniquely human ingredients of cognition evolved, what we can say is that they enabled almost everything that we think of as characteristically human: language, mathematics, technology, music, religion, and morality. Each of these systems of knowledge is capable of expressing a virtually limitless range of possibilities. Each of these systems is highly generative, capable of changing as a function of new environmental challenges. In this sense, though each system may have originally evolved as a by-product of combining distinctive and highly adaptive components, selection may act on such by-products to produce highly adaptive outcomes. For example, religion consists of cognitive processes that evolved for nonreligious functions, such as attributing intentions to hidden powers or inanimate entities, and holding beliefs that unify in-group solidarity and inspire out-group hatred. These cognitive processes, and others, played a significant role in the evolution of religion and, especially, the capacity of religious organizations to solve the problem of large-scale cooperation among unrelated strangers.

CONCLUDING REMARKS

This chapter has largely focused on a small branch of the phylogenetic bush, specifically, the comparison between human and nonhuman primates, and especially chimpanzees. This focus was intentional. No other comparative pairing among living organisms presents such small genetic distances accompanied by massive phenotypic differences. Chimpanzees have hair all over their body and are knuckle walkers. Humans are virtually bald and walk on two feet. Chimpanzees have never left the forests of Africa. Humans long ago left the forests of Africa to inhabit mountains, deserts, ice caps, and oceans within and outside Africa, and to explore other planets. Chimpanzees make tools and communicate through sounds and gestures, but each of these expressions is far too primitive to count as a precursor for human technology and language.

Though it has commonly been stated that human intellectual prowess stems from the evolution of language, or the capacity to create culture, or to invent new technologies, these explanations fail on two counts. First, they treat each of these causal factors as monolithic. Language, culture, and technology each rely on a suite of mechanisms, some unique to humans and others shared. To understand the evolution of language, culture, or technology—or any of the other distinctively human expressions—requires decomposition, studying each mechanism’s unique evolutionary history. Second, decomposition shows that language, culture, and technology each rely on the uniquely human ingredients of generative computation, promiscuity, abstraction, and symbolic expression. These capacities may each have evolved to solve a particular problem, but today they serve multiple functions. These capacities represent the engine of creativity, a device that delivers a limitless number of solutions to novel problems. These capacities allowed our species uniquely to go where no species has gone before.