RSP - Animal & Machine Intelligence

Nothing in biology makes sense except in the light of evolution.
Theodosius Dobzhansky

As Jean Piaget used to say, intelligence is what you use when you don't know what else to do.
William H Calvin

Reflections on the natural basis of intelligence have traditionally tended to revolve around the question of brain size. More brain, it has long been assumed, must result in more intelligence. Although superficially appealing, this assumption doesn't hold up under scrutiny—brain size doesn't correlate more than trivially with observed (or expected) intelligence (e.g. Seyfarth et al 2002; Dunbar 1998). [1] Sperm whales, for instance, have the largest brains in the animal kingdom, averaging around 9000 grams—roughly six times the size of human brains (typically 1500 grams, e.g. Nelson 1982, p262)—yet no one seriously claims that humans are any less intelligent than whales, never mind six times less. Likewise, the astonishing feats routinely performed by insects, birds, and other animals with modest brain mass are often discounted on the grounds that "real" intelligence involves language, technology, explicit reasoning and so on—in other words what humans do. Specialized or "modular" intelligence, such as the ability to accurately navigate across miles of desert or ocean, or the capacity for indoor horticulture, somehow aren't quite the real thing. In a paper critical of those who are reluctant to accept that some human intelligence might also be modular, Seyfarth et al (2002) remark:

For ethologists studying animals in their natural habitats, the notion of modular intelligence hardly comes as a surprise. Specialized, domain-specific performance almost seems the rule rather than the exception. Arctic terns migrate each year from one end of the earth to another, Cataglyphis ants navigate across featureless deserts, bees dance to signal the location of food, and some corvid species hide thousands of seeds in the fall, recovering them with unerring accuracy throughout the winter; yet despite these specialized skills we don't think of terns, ants, bees, or crows as generally more intelligent than other species. They are, instead, [regarded] more like nature's idiots savants. (Seyfarth et al, 2002)

Perhaps it is inevitable that we have been as homocentric in our thinking as we have. Maybe "speciesism," as Richard Dawkins suggests, is the next great prejudice we have to face. (Dawkins 2003, p21) There are signs however, that this bias is beginning to recede—at least within the cognitive sciences. The much quoted similarity between chimpanzee and human DNA, for example, although already something of a cliché, hopefully augurs well for a broader and more progressive view of natural intelligence that no longer unquestioningly locates humans as the exemplar par excellence. Having said that, the target of this paper is primate brains, so with that caveat I hope any inadvertent anthropomorphizing or homocentrism on my part will be forgiven.

In this paper I ask: why are there big brains in the world? How did they come to be, and what advantages or protections do they confer? In summarizing the central hypotheses that address these questions, I adumbrate a multi-factorial hypothesis of primate brain evolution that reflects the complex ecological dynamics present in the EEA (environment of evolutionary adaptedness). In particular I highlight the putative roles of sexual selection, fitness indicators, reproductive and dietary strategies, sudden environmental change, and phenotypic plasticity (the Baldwin Effect), as well as the demands of group dynamics and social interaction, as proposed by Dunbar (1993, 1998), Whiten & Byrne (1997), and others.

The lesson is clear: trying to deduce intelligence by measuring brain size is a mistaken approach. Brain size is strongly correlated with body size, but not with observed intelligence. If one were set a task of matching brains to bodies it would be a simple matter to determine which brains come from the biggest creatures (particularly if the brains are from animals of the same class, say mammals or reptiles), but it would not be at all obvious which brains came from the most intelligent animals. (See Figure 1).

Absolute brain weight then, is a poor predictor of intelligence. But as we can see, simple brain to body weight ratios are equally unreliable, not only because geometrical ratios favour small animals over bigger ones, but also because brain size varies widely across species, and body size varies within species. To address the first of these problems—the nonlinearity of brain to body size ratios—various equations have been devised that control for body size by relating the log of observed brain weight to the log of a measured reference quantity (typically body-mass, metabolism, or cortical volume) multiplied by an empirically derived constant (in Fig 1 this is the slope of the black lines describing the trend inside the polygons). The Encephalization Quotient (EQ) (Jerison 1973) is perhaps the best known of these measures—an equation that seeks to measure "extra" cortical neurons by revealing how much deviation there is in cortical mass from an expected figure in a given population. [2] Scores exactly on the slope lines within the polygons in Fig 1 have an EQ value of one. Scores above the lines reflects an EQ greater than one, while those below the lines indicate an EQ of less than one. This formula predicts that creatures with differing brain sizes but similar EQ should have similar intelligence (e.g., a lion and a house cat), while creatures with similar size brains but different EQ should display variation in intelligence (e.g., a chimp and a cow). Dunbar (1993) summarizes the different versions of this formula and concludes that the results they yield (at least in respect to his research on primates) are reasonably consistent with one another.

The bigger the body, the bigger the brain required to run it—which means it's not so much a question of big brains per se that interests us, but rather, why have some creatures evolved brains that are cortically richer than seems evolutionarily "necessary"?

Estimates of the total number of extant animal species vary widely. Two hundred and fifty years ago, in his Systema Naturae, Linnaeus catalogued over four thousand animal species and estimated that there could be as many as ten thousand. Recent estimates put the number anywhere between two and fifty million (e.g., Erwin 1988), but whatever the actual number most experts agree that since life began on Earth somewhere between three and four billion years ago, the vast majority (99% or more) of all species that ever existed are now extinct (Dennett 1995, p302). Life, in other words, has existed in an astonishing variety of forms, most of it without any sort of brain at all. Nevertheless, brains have become increasingly commonplace in the five hundred million years or so since they first appeared in vertebrates (e.g., in the early fish Haikouichthys—see Shu et al, 2003), though vanishingly few exhibit an EQ greater than one with respect to today's ecosystem (Macphail 1982, p243).

If bigger brains reliably resulted in greater reproductive success we ought to see many more of them (in the fossil record and alive today) than we do. There must be constraints that strictly limit the evolution of larger brains. What are these constraints? And what are the environmental and selective pressures that allow or favour such a costly piece of anatomy?

Brain size has increased in many lineages over evolutionary time. In the earliest stages of encephalization this may partly be due to what Stephen J Gould called the "Left Wall Effect" (Gould 1996, Fig 2; and Marino 1998, Fig 3). Organisms must be of a minimum size and complexity in order to be viable (the statistical limit of the left wall), and when all the smallest and simplest available designs and ecological niches are filled, the only way life can evolve is up in scale, or rightward in Gould's plan (Gould 1996; Marino 1998). This random drift effect is called a "passive" trend and could apply to brains as well as to other traits. Add selection pressures that eliminate the least able or fecund and the trend can become "driven," with swiftly increasing complexity. (Fig 3)

Figure 2 - The "Left Wall" & "Right Tail" of complexity in the Earth's biomass (after Gould 1996)

Gould (1996) conjectures that in the absence or relaxation of active selection pressures, there might be a slight tendency for complexity to drift back to simpler designs, and he cites numerous examples of lineages that have downsized, streamlined, and economized, so to speak. Typically however, trends are expected to follow a trajectory, the progress of which is ratcheted by selection. [3]

There are several simple selection pressures that can drive a trend, one of the most obvious being the classic arms race of the predator-prey relationship. When a predator catches her prey she will most likely have killed one of the less able members of the prey population. In so doing she nudges the average fitness of the prey population up a little, thus making tomorrow's lunch that much harder to catch. Over time, this means prey become ever more fleet of foot (or whatever talents they evolve to avoid being eaten), and predator lineages must also evolve in order to catch them, or else perish. Senses, skills, powers of prediction, and various anatomical features are often developed to extraordinary levels in this way—traits that can also result in an increase in brain size.

But this only goes so far it seems. Although lions and antelopes, eagles and fish, frogs and flies, have been engaged in this relentless struggle for millions of years, it isn't clear that any big-brained species can attribute their cerebral prowess to this mechanism alone. Such prolonged combat may have forced many of these antagonists to evolve bigger bodies (and therefore somewhat bigger brains by virtue of having bigger bodies), as well as a variety of other traits, but predator-prey arms races are insufficient by themselves, it seems, to account for either the existence of the biggest brains or the most intelligent creatures.

In a sense, these predator-prey arms races are strategies in a larger, Malthusian 'energy race'. Diet (energy) is obviously an important factor in brain development, maintenance and evolutionary encephalization, and dietary strategies can have profound evolutionary consequences on physiological evolution. Creatures who feed on scarce foodstuffs cannot afford to sit in a single tree all day as some foliovores do; they must be able to find their food, perhaps in neighbouring (and hostile) territories, which may require remembering where certain landmarks are, when and where certain trees are in fruit, as well as the ability to navigate. They might also need to employ more flexible or ingenious foraging tactics, all of which requires the right sort of neurological equipment. (Aiello and Wheeler 1995; Seyfarth & Cheney 2002)

A concentrated, high calorie, high protein diet may take less time to eat, less time to digest, and less digestive equipment than a field of grass or a tree full of leaves, but it probably also takes time and ingenuity to hunt or find. Brain evolution is thus driven (to some extent) by dietary and behavioural strategies, yet these strategies depend in turn on the right kind of brain—a classic feedback loop. (Dunbar 1998, Aiello & Wheeler 1995, and Appendix A).

The first serious constraint on brain size now becomes glaringly apparent. As Dunbar (1998) puts it:

Because the cost of maintaining a large brain is so great, it is unlikely that large brains will evolve merely because they can. Large brains will evolve only when the selection factor in their favour is sufficient to overcome the steep cost gradient.

And steep it is. Brains are hugely expensive to evolve and run: "Although the [human] brain represents only 2% of body weight, it receives 15% of the cardiac output, 20% of total body oxygen consumption, and 25% of total body glucose utilization" (Davis et al 2002; Miller 2000). Brains, then, take a large share of available resources. In a Malthusian world we would therefore expect them to be as frugal as can be got away with. Large brains must pay their way—by increasing inclusive fitness—or else perish. But whatever size of brain a creature has, the dietary strategy that feeds it must be able to provide enough nutrition, and in the Hobbesian world of natural selection this is sometimes best achieved by hunting or foraging in groups.

Having lots of eyes, ears, and noses on the lookout for competitors, predators and food (possibly prey) is an effective strategy that many species (and most primates) adopt. Being part of a group offers many advantages over a solitary life, including the possibility of the division of labour, with all the economic advantages that can bring. But group living—at least among birds and mammals—can have its problems too. Unlike the relatively ordered and rule driven world of insect colonies, the seemingly chaotic and much more complex world of warm-blooded group life requires considerably more real time information processing (Dunbar 1998). In particular, constellations of personal relationships need to be kept track of—both between self and others, and between others and others—especially when the "shadow of the future" [4] falls on them. It turns out that the more relationships one needs to keep track of, the more one needs a bigger brain to keep track of them.

"Primates are, above all, social animals." says Robin Dunbar (1993). The "Social Brain Hypothesis" he advocates (1993, 1998) follows from the deduction that because big brains are so costly, a substantial advantage must be gained in having one. He compared the ratio of neocortex to the rest of the brain (or brain part, such as the medulla) in a variety of primate species, and after controlling for body size, showed that neocortex size is strongly correlated with social group size—making brain size the best predictor of group size among primates. This, he says, yields "much the best [statistical] fit, accounting for 76% of the variance in mean group size among 36 genera of Prosimian and Anthropoid primates." (Dunbar 1993)

Further noting that on closer inspection the relative EQ (or CR-cortical ratio-in his nomenclature) of apes is greater than that of monkeys, which in turn is greater than that of prosimians, Dunbar says: "It is as if apes require more computing power to manage the same number of relationships that monkeys do, and monkeys in turn require more than prosimians do. This relation corresponds closely to the perceived scaling of social complexity." (Dunbar 1998) In other words it seems that the socio-political concerns of primates become subtler and finer grained as neocortex size increases. [5] Dunbar also suggests that male reproductive success in primates is directly related to neocortex size. Given that courtship and mating behaviours are extremely elaborate in many species, this is not entirely surprising. Individuals with defective brains would likely be readily identifiable by their strange behaviour, and afflicted males in particular could well be denied mating opportunities. Male brains, in other words, could act as fitness indicators-which in highly social species may translate into competence/status indicators of an individual within the group (Miller 2000, p134).

Geoffrey Miller and Peter Todd have gone further and suggested that sexual selection plays a central role in the evolution of brain encephalization. Pointing out that the pattern of evolution in hominid brains seems to have gone in fits and starts, Miller (2000) observes that traits formed through sexual selection follow the same pattern-they rapidly evolve to a local maximum then plateau out. Any noticeable trait, he suggests, can be operated on by mate choice, whether through sensory bias, as a fitness indicator, or even as pure ornament. He even speculates (convincingly) on the origin of human aesthetic sensibilities in connection with this process. He notes that males greatly outnumber females in their production of ornament, song, combat and so on (in humans this translates into music, literature, war, etc), and that it is hard to see any survival advantages in aesthetic behaviour: "Since there are no plausible survival benefits for music production, reproductive benefits seem worth a look." (Miller 2000b) Elsewhere he says:

human art . . . [is] a biological adaptation: it evolved through sexual selection to serve the same courtship functions as almost all other examples of organic beauty and complex behavioral signals observable in nature. Such ornamentation often evolves as a reliable, costly indicator of the signaller's good health, good brain, and good genes." (Miller 2001).

While small brained creatures (in which there is at least some female mate choice) typically select mates through rigid, well established signals, larger-brained animals, it is argued, can and frequently do use more flexible and complex behavioural cues on which to base their choices.

The statistician and geneticist Ronald Fisher launched the idea of "Runaway" sexual selection in 1930, in his book The Genetical Theory of Natural Selection, although he had floated the suggestion of arbitrary signs of male fitness in a paper some years earlier (Miller 2000, p54). Sexual selection had fallen out of favour soon after Darwin proposed the idea in 1871 (in The descent of man, and selection in relation to sex) and the idea had been all but consigned to oblivion after the Nobel Prize winning geneticist Thomas Hunt Morgan (who showed that chromosomes contain the genetic material) ridiculed the idea in 1904.

The peacock's tail is perhaps the canonical example of a "Runaway" sexually selected trait. Thanks to the neurology of the pea hen's brain, the more blue a male can flutter at her in spring, the more she is likely to allow him to copulate with her. Males with the most blue on offer will do best when it comes to fathering offspring, and some of the male offspring will tend to resemble their successful fathers. Females have exactly the same desire to leave descendents, half of which will be males, so they pick the "best" fathers for their offspring they can—those who flash the most blue—if they don't want their sons to die childless. Anything that increases the size of the display, the intensity of the colour, and so on will be strongly favoured.

But sooner or later the tail gets too big, the colours too bright (the song too loud, etc), and the trait becomes a handicap; predators can no longer be outrun (prey can see or hear you coming from miles away). This imposes a limit on the lengths a tail can get to. Sooner or later females will have to start choosing on the basis of some other trait. But in the meantime, the very fact that a male can make it through the winter carrying such a costly and risky burden, and keep it in perfect condition, means that he is either very lucky, or else he "has what it takes"—and his very existence acts as a good fitness indicator.

Darwin's "other" theory has at last been integrated into biological thinking, but there is still resistance to the idea that mate choice could influence brain evolution. The peacock's tail was long considered an oddity, one of the many puzzles of nature. Large brains are, according to Miller, Todd, Wills and others, no different—not an oddity, rather a fantastically elaborate mate-attracting and mate-choosing device. As they remind us, mating choices are important-get them wrong and your lineage comes to a screeching halt-so it is not surprising that females in particular are so picky when it comes to choosing a mate, as they almost always commit much more energy into offspring than males.

The so-called "Baldwin Effect" [6] is another factor that could be implicated in primate brain evolution. This is the idea that phenotypic plasticity can allow some immediate adaptation, which can then "guide" or favour genetic mutations that achieve the same result. For instance, the skin of some creatures, when exposed to sunlight, can adapt by producing a protective pigment. But a lineage that is exposed to sunlight for many generations may evolve to produce the pigment permanently. With respect to organisms with relatively substantial brains (and, vitally, substantial memory), this process might mean that a learned skill which confers a survival or reproductive advantage, and that can be taught to kin, can become ubiquitous in a population-with natural selection weeding out those individuals who fail to learn well enough. In connection with brain evolution there might even be a case for asking whether a "runaway" Baldwin effect could occur. In this scenario, a member of a population learns a trick that some members cannot grasp or master. Genes that underlie this better learning would then increase in frequency proportionally to the advantage of the particular skill learned. Language, for instance, could have developed in this way. Certainly, evolutionary brain growth appears to have been patchy and episodic, (Dunbar 1998) and in concert with environmental change (e.g. the frequent, high-magnitude, and sometimes extremely rapid climate changes that have occurred during the last 2.5 million years [Calvin 1998]), unusual brain evolution in some species, far from being miraculous, might sooner or later be inevitable.

In contrast to the Mesozoic (248-65 mya) and Paleozoic (543-248 mya) eras, which were characterized by the development of size and force, the Cenozoic (the last 65 million years) could be described as the age of intelligence. The ideas touched on here, put together, hint at the complex story of how some brains became so large-but why they did so particularly in the last 65 million years remains unclear.

Presupposing the veracity of biology's "central dogma"-that genotype leads to phenotype [7]-and given that genes, bodies, and sometimes even whole populations are subject to varying environmental pressures, the question of why some brains are big (highly encephalized) can be answered by tracing the trajectories of large-brained lineages through evolutionary space. What were the selective pressures that moulded such an unlikely piece of anatomy? The primate EEA (environment of evolutionary adaptedness) is still being pieced together, but already a strong case can be made for the roles of feedback, sexual selection, dietary adaptation, and social living, as well as geo-physical factors such as rapid climate change.

Where to put your resources - into growing big brains or big guts? Diagrams from: Aiello and Wheeler (1995). The "expected" figures are calculated using a standard scaling rule such as is commonly used (e.g., Kleiber's Law) to predict BMR (basal metabolic rate) and EQ. Aiello and Wheeler suggest that the answer to the paradox of how humans can afford, energetically, such an expensive organ without a corresponding increase in BMR is by eating a much richer diet that requires less energy and gut to digest. See: http://www.beyondveg.com/billings-t/comp-anat/comp-anat-4a.shtml

A positive feedback loop is proposed whereby a bigger brain leads to more efficient and profitable feeding strategies, further reducing energy consumption in physiology, digestion and foraging.

Aiello, L.C., & Wheeler, P., (1995), The expensive tissue hypothesis: the brain and the digestive system in human and primate evolution, Current Anthropology, vol. 36, p.199-221

Byrne R. and Whiten A., (eds.), (1988), Machiavellian Intelligence: Social Expertise and the Evolution of Intellect in Monkeys, Apes, and Humans, Oxford: Oxford University Press

Brosnan S.F. & de Waal F.B.M., (2003), "Monkeys Reject Unequal Pay", Nature, 425: 297-299

Clutton-Brock T.H. and Harvey P.H., (1980), Primates, brains and ecology, Journal of the Zoological Society of London, 190: 309-323

Cohen J.E., Jonsson T., Carpenter S.R., (2003), Ecological Community Description Using The Food Web, Species Abundance, And Body Size, Proceedings Of The National Academy Of Sciences Of The United States Of America, 100 (4): 1781-1786

Crick, F.H.C., (1958), On Protein Synthesis, Symposium of the society for experimental biology XII, New York: Academic Press, p153.

Davis, Kenneth L., Charney, Dennis, Coyle, Joseph T., & Nemeroff, Charles, (Eds.), (2002), Neuropsychopharmacology: The Fifth Generation of Progress, American College of Neuropsychopharmacology
http://www.acnp.org/g4/GN401000064/CH064.HTML

Dunbar, R.I.M., (1992), Neocortex size as a constraint on group size in primates, Journal of Human Evolution 20: 469-493

Erwin, Terry L., (1988), The Tropical Forest Canopy: The Heart of Biotic Diversity, in E.O.Wilson, ed., Biodiversity, National Academy Press, Washington, D.C., pp.123-129.

Gibson K.R., (2002), Evolution of Human Intelligence: The Roles of Brain Size and Mental Construction, Brain, Behavior and Evolution, 59:10-20

Gould, S.J., (1996), Life's Grandeur: The Spread of Excellence from Plato to Darwin, London, Random House

Humphrey, N., (2002), The Deformed Transformed, in Humphrey, N., (Ed), The Mind Made Flesh: Essays from the Frontiers of Psychology and Evolution, chapter 14, p 165-199, Oxford, OUP

Humphrey, N., (1998), Cave Art, Autism, and the Evolution of the Human Mind, Cambridge Archaeological Journal 8(2):165-191

Jerison, H.J., (1979), The evolution of diversity in brain size, in: Hahn, M.E., Jensen, C. and Dudek, B.C., (Eds), Development and Evolution of Brain Size. Behavioral Implications. Academic Press, New York, p. 29-57

Kull, K., (2000), Organisms can be proud to have been their own designers, Cybernetics and Human Knowing, vol 7(1), p. 45-55 http://www.zbi.ee/~kalevi/chkfin.htm

Marino, L., (1998), A Comparison of Encephalization between Odontocete Cetaceans and Anthropoid Primates, Brain, Behavior and Evolution, 51:230-238
http://www.emory.edu/LIVING_LINKS/Marino.html

Miller, G. F., (2001), Aesthetic fitness: How sexual selection shaped artistic virtuosity as a fitness indicator and aesthetic preferences as mate choice criteria, Bulletin of Psychology and the Arts 2(1), 20-25 http://www.unm.edu/~psych/faculty/aesthetic_fitness.htm

Miller, G.F., (2000), The Mating Mind: How sexual choice shaped the evolution of human nature. New York, Doubleday

Miller, G.F., (2000a), Mental traits as fitness indicators: Expanding evolutionary psychology's adaptationism. Annals of the New York Academy of Sciences, 907: 62-74

Miller, G. F., (2000b), Evolution of human music through sexual selection, in: N. L. Wallin, B. Merker, & S. Brown (Eds.), The origins of music, MIT Press, pp. 329-360 http://www.unm.edu/~psych/faculty/music.htm

Prusiner, S.B., (1997), Prion diseases and the BSE crisis, Science, 10; 278(5336):245-51.

Sloan Wilson, D., Near D., and Miller R., (1996), Machiavellianism: A Synthesis of the Evolutionary and Psychological Literatures Psychological Bulletin, Vol. 119, No. 2, 285-299

Turney, P., (1996), Myths and Legends of the Baldwin Effect, Proceedings Workshop on Evolutionary Computation and Machine Learning at the 13th International Conference on Machine Learning, pages 135-142

Whiten, A., & Byrne, R. W., (Eds.), (1997), Machiavellian Intelligence II: Extensions and Evaluations, Cambridge, Cambridge University Press

[1] Quite what we mean by the term "intelligence" is a much-disputed matter that I shall not address here. Let me simply offer Alfred Binet's definition: "the totality of mental processes involved in adapting to the environment." (The Columbia Encyclopedia, Sixth Edition, 2001)

[2] Based on the linear equation y = mx + b, where y = predicted brain weight, x = observed body weight, m = slope and b = the value of y where the slope intercepts the vertical axis. To correct for the nonlinearity of body weight to brain weight the logs of body and brain weight are used: log(y) = log(b) + m[log (x)]

[3] But as Gould emphasizes, the complexity, size, or intelligence of a lineage can go down as well as up. Although life has generally become more complex over time, that is not to say it will in any particular lineage or era. The trend, he thought, is weak.

[4] As Robert Axelrod (1984) put it. This is the idea that how one behaves in a particular social situation depends (in part at least) on whether one will encounter (and remember) one's interlocutor again in the future. As the likelihood of future interactions increases, the more the current interaction will be constrained by that prospect (and the future interaction determined by the current one). In social primates the shadow of the future must hang over most, if not all relationships within the group to some extent.

[5] An observation that students of culture have yet to fully capitalize on, and one that should be of interest to "Darwinian" humanities scholars.

[6] Named after J. Mark Baldwin, the American naturalist who first described it in 1896.

[7] The idea that information only flows one way-from DNA to RNA to proteins-first proposed by Francis Crick (1958). Recent work has led to some questioning of this position. (E.g., Prusiner 1997)