An influential definition of human intelligence is “a very general mental capability that, among other things, involves the ability to reason, plan, solve problems, think abstractly, comprehend complex ideas, learn quickly and learn from experience. It is not merely book learning, a narrow academic skill, or test-taking smarts. Rather, it reflects a broader and deeper capability for comprehending our surroundings – ‘catching on,’ ‘making sense’ of things, or ‘figuring out’ what to do” (Gottfredson 1997a, p. 13).
One of the striking facts about our universe is that it contains intelligence. Somehow, planet Earth went from a barren, scorching planet into a flowering display of life. Eventually, Homo sapiens (“wise man”) was the first species on Earth, and possibly in the universe, that started to talk, generate rich cultures, send conspecifics and electronics to space, and invent abstract algebra and calculus.
Why did intelligence evolve in the first place? This question seems unreasonable at first sight because it is hard to imagine anything more useful than intelligence. Nevertheless, intelligence comes packed with disadvantages. A key hint about the cost of intelligence is the fact that it is relatively rare in nature. In fact, many organisms do just fine without a brain at all. Moreover, intelligence has evolved independently only a few times (convergent evolution) in some mammals, cephalopods (e.g., octopi), and birds (Emery and Clayton 2004). Further, although just 2% of the total body weight of humans, the brain consumes 20% of the calories consumed (Aiello and Wheeler 1995), and Mother Nature does not spend calories unnecessarily. Nevertheless, the brain expanded threefold in just under 2 million years, suggesting that selective pressures forcefully aligned to produce intelligence in the Homo line.
This entry will briefly discuss (1) the science and psychometrics of human intelligence, (2) how intelligence works from an information processing point of view, and finally (3) models and evidence on why human intelligence evolved.
The Psychometrics of Intelligence
Key parts of human intelligence are the abilities to learn, reason, and solve problems (Plomin and von Stumm 2018). These abilities to learn and reason also differ dramatically between people and can be reliably measured with tests such as the Wechsler intelligence scale (Wechsler 2008) and Ravens progressive matrices (Raven and Court 1998). The latter test consists of a series of multiple choice questions based on matrices. The subject is shown a 3 × 3 matrix containing symbols where the last one is omitted, the subject and then chooses a symbol from multiple possible responses based on identification of the underlying rule generating the matrix. In modern measurements of intelligence, a person’s intelligence is estimated by taking the subject’s raw score on several mental abilities and comparing it to the average performance of a representative group in the subject’s age range. This is called deviation intelligence quotient (IQ) (Urbina 2011). The normalized IQ distribution has an average of 100 and a standard deviation of 15. Intelligence tests are important clinically, because they are used to diagnose intellectual disability in children and adults (a diagnostic criteria being IQ below 70) and to assess milder cognitive deficits caused by brain injury and neurodegenerative disease.
A classic empirical finding, first made by Spearman (1904), is that test scores on several specific and different intellectual abilities covary and hence form the statistical abstraction called general intelligence (g). In other words, people who score well on one particular cognitive ability tend to score well on others. Based on these findings, general intelligence is understood psychometrically as a latent trait, meaning that it is not measured directly, but inferred from several measured cognitive abilities such as verbal and spatial abilities. However, every cognitive domain also has some degree of specificity. For instance, it is possible for an individual to score above average in logical ability but below average in verbal ability. In fact, a large controversy in the history of intelligence theory and research concerns how to understand both the general and the specific aspect of intelligence. Many theoreticians agree that g is important for all mental abilities (Jensen 1998), while others claim that humans have multiple and separate intelligences, such as interpersonal and musical ability (e.g., Gardner 2011). However, the latter theory seems to be a synthesis of general and specific intelligence, as well as noncognitive abilities and personality traits (Visser et al. 2006).
Today intelligence has emerged as a generalist concept that influences many real-life outcomes. Most obviously, it predicts educational attainment (Jensen 1998), but intelligence is also a strong predictor in a wide range of nonacademic areas. In fact, intelligence predicts general health and longevity (Hart et al. 2003), and it is the single best predictor of success in job training and performance, with predictive power increasing with job complexity (Schmidt and Hunter 2004). Overall, intelligence is the best single predictor for both favorable socioeconomic outcomes, such as good education, occupation, and income, and negatively for unfavorable outcomes such as adult poverty, incarceration, and chronic welfare use (Gottfredson 1997b; Gottfredson and Deary 2004).
How Intelligence Works
How can a vast collection of unintelligent neurons produce intelligent behavior? Intelligence cannot, of course, arise out of lumps of brain matter alone. Instead, intelligence is produced by the computations going on within the neural tissue. That is, human minds can do intelligent things because nervous systems represent, process, and transform information (Clark 2014). One key property of a computational system is the ability to change its operation based on information (Cosmides and Tooby 2002). Hence, computational processes can be seen as independent of the physical machinery in which it occurs. For example, the plastic cover on a thermostat is not a computational system, but the thermocouple that responds to a particular room temperature and turns on the switch that activates the heat-generating parts of a furnace is a computational system. Similarly, the visual system in the human brain that detects the presence of a predator and activates muscles that either makes a person run away if the predator has seen it, or freeze if it has not, is a computational system (Cosmides and Tooby 2002).
How does the computing in neural circuitry actually work? One approach in cognitive science, called connectionism, uses the insights gained from artificial neural networks to shed light on how intelligence can be produced (Elman et al. 1998). The design of computer-programmed artificial neural nets works by many interconnected nodes that are modified by a learning rule that, over time, strengthen particular connections to produce the desired outcome (Striedter 2005). Hence, in the connectionist perspective, the knowledge base residing within neural structures are composed of sets of simple processors linked by a daunting mass of wiring and connections (Clark 2014).
However, another important approach, called the physical symbol system, is that computation works by more abstract symbol manipulation (Newell and Simon 1976). This is akin to how a digital computer works, namely, by manipulating symbols (the famous zeros and ones), while the CPU do operations on them.
Although you can get far with simple connectionist models for modeling intelligent human behavior, it also runs into problems. Generally, they can fail to capture the flexibility and power of everyday reasoning. For example, when representing complex relations between bits of knowledge or generalizing abstract relationships, hence, brain circuitry that produces intelligent behavior might best be construed as a complex combination of both connectionism and symbol manipulation (Marcus 2003).
Within classical evolutionary psychology, it is often thought that the mind consists of several special-purpose modules that solve particular problems that were relevant in past evolutionary environments (Barrett and Kurzban 2006). This has been dubbed the massive modularity hypothesis. In this view, the mind contains a vast array of tools that solve particular problems, rather than content-independent, abstract reasoning devices.
There are good reasons to believe that complex brains are modular. For example, within robotics research, installation of special-purpose tricks is often the only effective ways of generating adaptive behavior in real time (Clark 2014). It is likely, therefore, that the mind consists of many specialized subsystems running in parallel, rather than a few general problem solvers.
However, how can the modularity view explain intelligence, that clearly involves usage of content-independent reasoning to solve novel problems? This is in itself a difficult problem because it is hard to conceive how the alternative, namely, content-free and general-purpose minds, could generate intelligence. This would be akin to a smart phone solving complex tasks (reading maps, taking pictures, and so on) without any special-purpose apps installed on it.
A useful distinction is to consider two types of intelligence: (1) dedicated intelligence and (2) improvisational intelligence. Dedicated intelligence is the degree to which a neural program, or module, is able to efficiently solve a particular set of computational problems. Improvisational intelligence, on the other hand, reflects the ability to exploit transient or novel conditions to achieve adaptive outcomes (Cosmides and Tooby 2002).
How is improvisational intelligence achieved? An answer might be that human g relates more closely to evolution of interactions between evolutionarily recent neural circuitry, rather than the evolution of evermore specialized modules alone (Jung and Haier 2007). Although brain networks are highly modular, there is evidence that individual differences in human general intelligence is not associated with the number or size of modules but in the connectivity between and within modules, particularly in frontal and parietal brain regions (Hilger et al. 2017). Nevertheless, concrete knowledge about how this connectivity produces intelligence might depend more on progress in artificial neural nets and general artificial intelligence than from the direct study of the messy biochemical reality of neuroscience.
The Evolution of Intelligence
Around 2 million years ago, Homo erectus (also known as Homo ergaster) appeared in Africa. This was the ancestor of ancient Homo sapiens and Homo neanderthalensis (Klein 2000). The size of the brain of Homo erectus was twice as large as any living nonhuman apes and had roughly 75% the brain capacity of modern humans (Gabora and Russon 2011; Ruff et al. 1997). Homo erectus has left evidence of human-like characteristics; they made stone axes, engaged in long-distance hunting, and had the ability to adjust lifestyle based on season (Gabora and Russon 2011).
Two major brain expansions occurred during the evolution of the homo line. The first was around 2 million years ago (Klein 2008), and the second one was between 600,000 and 150,000 years ago, marking the appearance of anatomically modern Homo sapiens (Gabora and Russon 2011; Ruff et al. 1997).
The fully grown human brain weighs around 1500 g, which is more than three times heavier than for every other living primate (Striedter 2005). Nonetheless, brain weight correlates with the overall size of an organism, so information about brain weight alone is not a useful proxy for the relative intelligence of a species. Therefore, it is more informative to look at ratios between brain and body weight. However, this ratio can be misleading too because smaller animals generally have higher brain to body ratio than larger ones (pocket mice have a higher brain-body ratio than humans do, for instance). Given this nonlinearity, the most useful measure of relative interspecies brain size is the encephalization quotient (EQ). EQ is the ratio between predicted and actual brain mass of a particular species given information from several reference species (Striedter 2005). Examples of EQ scores for various species are the African elephants with an EQ of 1.3, the mouse with 0.5, chimpanzees 2.2–2.5, dolphins 5.3, and humans, with the highest score of all species, 7.4 (Roth and Dicke 2005).
The cradle of humans was in equatorial Africa (Klein 2000), and it was here that humans first occupied and exploited what has been dubbed the cognitive niche (DeVore and Tooby 1987). A niche often refers to what an animal does in its local ecology: It can be said to describe the profession, biologically speaking, of an organism (Odum 1959). Rather than exploiting resources by purely physical and chemical means (e.g., running faster, stronger toxins, and so on), humans are able to extract resources from the environment by using cognitive models of causal relationships in the world. This allows humans to invent tools and weapons, extract poison and drugs from other living organisms, and engage in coordinated action with other people; all embedded in culture (Pinker 2010).
The Cost of Evolving Intelligence
As mentioned in the introduction, evolving large amounts of brain tissue is calorically expensive. Furthermore, greater intelligence requires larger brains and therefore a larger cranium in which to contain it, and giving birth to increasingly big-brained children was risky due to the close correspondence in size between the maternal pelvis and the neonatal cranium (Rosenberg and Trevathan 2002). This evolutionary problem leads to humans being born at an earlier stage in development, giving the brain time to grow postnatally instead. This gives rise, in part, to the particularly long childhoods among humans, years that are spent in apprenticeship within the local culture, using ever-increasing intelligence to acquire skills for survival and social know-how (Bjorklund and Hernandez Blasi 2016; Pinker 2010).
All the extreme traits that are characteristic of humans such as language, sociality, long childhoods, and long investments from fathers and grandparents, and usage of different food sources, have likely all contributed to the development of Homo intelligence (Street et al. 2017). Language, for instance, allows for effective communication of know-how, which in turn allows for more sophisticated problem-solving and social cooperation.
However, evolution of the human brain and human cognitive ability was probably not just simply due to cooperative hunting, meat eating, or tool usage because the fossil records show that these behaviors preceded the main brain expansions in humans (Klein 2000).
Another set of unusual human traits that might have coevolved with increased intelligence are the unusual sexual characteristics of humans. The human species have a reduced sexual dimorphism: Men are still larger than females but less so compared to other apes, suggesting selection for monogamy, biparental care, and general mating cooperation. This might be a crucial step given the extended childhood of humans. Babies require very costly parental investment in order to survive to reproductive age themselves, and human males do provide care and protection toward their offspring to a greater degree than other male apes (Flinn et al, 2005; Gabora and Russon 2011; Pinker 2010).
What selective pressures drove the brain expansions, despite all the costs, within the homo line and in anatomically modern Homo sapiens?
Models of the Evolution of Human Intelligence
There are several models on why intelligence evolved in humans. The following section will briefly discuss two important models on how it got started: ecological dominance and the social brain model. Then, lastly, we discuss the possible role for sexual selection in explaining the extraordinary intelligence of humans.
The ecological dominance model on the evolution of human intelligence stresses that the hominin line was able to increasingly master the hostile forces of nature, such as procuring food, keeping warm, survive, and finding mates (e.g., Flinn et al. 2005). Several parts of the human mind are useful for ecological dominance. For example, more than any other animal, humans have “scenario-building” abilities, including foresight, planning, and “autonoetic” functions – the ability to do mental time travel. All of these functions are related to the most recently evolved parts of the brain, namely, the prefrontal cortex.
In fact, humans are the only species that can mold its environment and even entirely remove other aspects of nature, including other species, from its environment (Alexander 1990), suggesting that intelligence evolved at least partly as a means to overcome classic ecological challenges.
Ecological dominance is likely to be an important factor in the evolution of human intelligence. However, a problem with the model is that it does not explain why humans evolved the exceptional cognitive abilities that they have. Many other species hunt, live on savanna habitats, have extended lifetimes, and have generally gone through similar selection pressures in their ecology without ever coming close to evolving the relative brain size and intelligence of Homo sapiens (Flinn et al. 2005).
The social brain model postulates that human intelligence evolved due to the social chess humans have played throughout evolution – namely, in dealing with other intelligent hominin creatures. After humans dominated and mastered their ecology, these social selective pressures might have been even more consequential. Predicting future moves and countermoves of competitors and cooperators is complex, especially when amplified by deception and shifting coalitions. Climbing social hierarchies is crucial for mating success, a complicated game that strongly rewards intellect (Dunbar 1998; Flinn et al. 2005). Hence, a crucial part of the social brain model is that people who are more sophisticated socially can manipulate other individuals to gain control over resources and other people’s behavior. Moreover, humans have also evolved the capacity for prosociality and cooperation. That is, individuals display moral and altruistic sophistication toward both kin and nonkin, making them more attractive as social partners (Nesse 2007).
There is a correlation in group size of a species and brain size (Dunbar 1998). Many highly developed mental abilities in humans are also directly related to social functioning. For example, theory of mind (ToM), the ability to ascribe and understand the intentions and mental states of others, is well developed in humans and have severe negative consequences (autism) when dysfunctioning (Baron-Cohen 1997; Dennett 1989; Nesse 2007).
Sexual Selection: Is Human Intelligence a Fitness Indicator?
In sexual courtship many animals advertise their genetic quality with their bodies and minds (Miller 2000, 2011). The male bowerbird, for example, is famous for its exquisite architectural abilities (Day et al. 2005). Male bowerbirds arrange their bowers out of twigs and ornament them with petals, glossy beetles, and even colorful human-made objects such as pens and dollar bills. Arguably, the male applies its skills as an architect, carpenter, and interior designer sending potential mates a signal saying “I am able to build this particularly complicated bower because my genome contains below-average numbers of mutations and hence my brain was able to develop properly, therefore you should mate with me, because my high-fitness-brain is likely to be passed on to your children.”
Higher phenotypic and genetic variation.
Positive correlations with other fitness-relevant aspects of an animal, such as parasite resistance, longevity, and body symmetry.
All of these attributes fit human intelligence well (see Prokosch et al. 2005).
Hence, a strength of the sexual selection perspective on intelligence is that it can account for facts about human intelligence not accounted for by other models. For example, that intelligence is highly heritable, meaning that about half of the observed variation in intelligence across people are attributable to genetic differences (Plomin and von Stumm 2018). Another example is that intelligence in men is related to sperm quality (Arden et al. 2009), indicating that intelligence is associated with other conceptually unrelated but fitness-relevant traits. Furthermore, and more straightforwardly, intelligence is sexually attractive and one of the most desired traits for long-term sexual partners for both sexes (Buss 1989; Geher and Miller 2008).
Sexually selected traits in the animal kingdom often portray large sex differences, typically with males competing by showing off conspicuous displays and females choosing the best fit males (although this is sometimes flipped around, depending on which sex invests more in the offspring). Among humans, there are no sex differences in general intelligence (Jensen 1998). However, the evolution of fitness indicators does not require that one sex displays a sexually selected trait and that the other prefers it. Human mate choices are often best described by mutual mate choice (MMC) – both sexes are choosy about their mates, and both men and women select for intelligence (Stewart-Williams and Thomas 2013). However, the average fitness payoff for showing off high intelligence might still be higher among human males (Miller 2011). If the latter scenario is the case, it would result in high-intelligence men having more high-quality offspring and low-intelligence men being kept out of the mating game entirely, causing more variance in intelligence and general fitness among men. Indeed, this is the observed pattern among humans: There is greater variance in intelligence among human males (Deary et al. 2003; Stewart-Williams and Thomas 2013), with the consequence that there are slightly more men with extremely low and extremely high intelligence.
The fitness indicator model is a very unlikely explanation for why intelligence started to evolve in several Homo species (and in other mammals) in the first place. This is because, in general, traits that evolve into fitness indicators are selected (e.g., by female choice) for their hard-to-fake complexity and hence high mutational target size, but the trait is typically arbitrarily chosen (Keller 2008; Miller 2011). That these evolutionary events arose independently across species is therefore unlikely.
Moreover, some argue that the coevolution between language, sociality, and cognition alone can account for why the human brain expanded so quickly and hence regard sexual selection as explanatory superfluous (Pinker 2010). Future empirical research and analysis of genomic data will help shed light on how important or unnecessary the fitness indicator model is in explaining the evolution of human intelligence.
Intelligence, the ability to learn, reason, and solve problems, can predict many fitness-relevant life outcomes such as educational achievement, income, and health. There is still much to learn about the nuts and bolts of intelligence. However, connectionism and other approaches in cognitive science have shed some light on how neural circuitry can process information.
There are several models on why intelligence evolved in humans. Some highlight the mastery of the hostile forces of nature, such as finding food and keeping warm (ecological dominance), as a tool for social deception, competition, and cooperation (social brain), and that intelligence evolved for the purpose of displaying fitness in mate competition (fitness indicator).
The models on the evolution of intelligence are not always competing perspectives but rather different pieces of the puzzle that belong together (e.g., Flinn et al. 2005). It is nevertheless scientifically useful to differentiate them to derive clear testable predictions, in order to find out where they compete and where they align (Dunbar 1998). For example, one study found that human-sized brain and bodies were predicted by a model where individuals faced selection pressures that was 60% ecological, 30% socially cooperative, and 10% between-group competitive (González-Forero and Gardner 2018) (the fitness indicator model was not included), suggesting that several models are important but to different degrees.
The fitness indicator model can potentially explain why intelligence evolved so quickly and is sexually attractive and highly heritable and why it correlates with other fitness-relevant traits. However, simpler models are probably better accounts for how the evolution of intelligence in the Homo line got started, and it is still debatable if sexual selection is superfluous or essential in explaining the existence of our big and costly brains.
- Alexander, R. D. (1990). How did humans evolve? Reflections on the uniquely unique species (Special Publication No. 1). Ann Arbor: Museum of Zoology.Google Scholar
- Andersson, M. (1994). Sexual selection. Princeton: Princeton University Press.Google Scholar
- Baron-Cohen, S. (1997). Mindblindness: An essay on autism and theory of mind. Cambridge, MA: A bradford book.Google Scholar
- Bjorklund, D. F., & Hernandez Blasi, C. (2016). Evolutionary developmental psychology. In D. M. Buss, C. H. Blasi, & B. J. Ellis (Eds.), The handbook of evolutionary psychology (Vol. 2, pp. 904–924). New Jersey: John Wiley & Sons.Google Scholar
- Clark, A. (2014). Mindware: An introduction to the philosophy of cognitive science (2nd ed.). New York: Oxford University Press.Google Scholar
- Cosmides, L., & Tooby, J. (2002). In R. J. Sternberg & J. C. Kaufman (Eds.), Unraveling the enigma of human intelligence: Evolutionary psychology and the multimodular mind (pp. 145–198). Mahwah: Lawrence Erlbaum Associates Publishers.Google Scholar
- Dennett, D. C. (1989). The intentional stance. Cambridge, MA: MIT press.Google Scholar
- DeVore, I., & Tooby, J. (1987). The reconstruction of hominid behavioral evolution through strategic modeling. In W. G. Kinzey (Ed.), The evolution of human behavior: Primate models (pp. 183–237). Albany: SUNY Press.Google Scholar
- Elman, J. L., Bates, E. A., & Johnson, M. H. (1998). Rethinking innateness: A connectionist perspective on development (Vol. 10). Cambridge, MA: MIT press.Google Scholar
- Gardner, H. (2011). Frames of mind: The theory of multiple intelligences. Hachette: Basic Books.Google Scholar
- Geher, G., & Miller, G. (2008). Mating intelligence: Sex, relationships, and the ,ind’s reproductive system. New York: Routledge.Google Scholar
- Jensen, A. R. (1998). The g factor: The science of mental ability. Westport: Praeger.Google Scholar
- Keller, M. C. (2008). The role of mutations in human mating. In G. Geher & G. Miller (Eds.), Mating intelligence: Sex, relationships, and the Mind's reproductive system (pp. 173–193). New York: Routledge.Google Scholar
- Marcus, G. F. (2003). The algebraic mind: Integrating connectionism and cognitive science. Cambridge, MA: MIT Press.Google Scholar
- Miller, G. (2000). Sexual selection for indicators of intelligence. In G. Bock, R, J. A. Goode, & K. Webb (Eds.), Novartis Foundation Symposium (Vol. 233). West Sessex, UK: Wiley.Google Scholar
- Miller, G. (2011). The mating mind: How sexual choice shaped the evolution of human nature.Google Scholar
- Odum, E. P. (1959). Fundamentals of ecology (2nd ed.). Philadelphia: WB Saunders.Google Scholar
- Prokosch, M. D., Yeo, R. A., & Miller, G. (2005). Intelligence tests with higher g-loadings show higher correlations with body symmetry: Evidence for a general fitness factor mediated by developmental stability. Intelligence, 33, 203–213. https://doi.org/10.1016/j.intell.2004.07.007.CrossRefGoogle Scholar
- Raven, J. C., & Court, J. H. (1998). Manual for Raven’s progressive matrices and vocabulary scales. Oxford: Oxford Psychologists Press.Google Scholar
- Striedter, G. (2005). Principles of brain evolution. Sunderland: Sinauer.Google Scholar
- Wechsler, D. (2008). Wechsler adult intelligence scale–fourth edition (WAIS–IV). San Antonio: The Psychological Corporation.Google Scholar