top of page

Summary of The Evolved Apprentice

Updated: Sep 3, 2022

Book by Kim Sterelny

summary by Haggeo


Sterenly distinguishes, following Peter Godfrey-Smith, philosophy of science vs philosophy of nature. Philosophy of science studies science. Philosophy of nature studies nature. Sterenly is after a philosophy of nature. This work is philosophy because he uses philosophical tools: synthesis, argument, and interdisciplinary integration. It is also philosophy because it asks about humans and our place in nature.

The book has empiricist themes as it emphasizes learning over pre-installed competencies. But it is not individualist or internist. It argues that “...human cognitive competence is a collective achievement and a collective legacy…” (xii). Moreover, cognitive achievement depends on us engineering the environment around us. This has strands with Andy Clark’s social Epistemology.

This book differs from various theories on evolutionary theories of mind. The book denies key innovation models and argues for positive feedback loops. The book denies nativism. The author focuses on the cognitive journey as much as the cognitive destination. Instead of focusing on cognitive mechanisms, Sterelny focuses on archaeology, paleoanthropology, and ethnography. The book emphasizes the importance of niche construction. We construct and are constrained by our social niche.

Ch 1, The Evolved Apprentice - Sterelny

Human species have developed remarkable characteristics. They have complex social lives, linguistic-cognitive abilities, expansive habitats, and so forth. According to the standard model, there are two main hypotheses used to explain this human evolutionary trajectory.

  • Social intelligence hypothesis: increasing complexities in social life required increasing sophistication in cognition.

Many theorists have filled in the details. Robin Dunbar argues that social complexity means we need to increase memory and conflict management. An influential approach is Humphrey’s Machiavellian model. The idea here is that every person had to learn how to gain profit from others, often via cooperation.

A second hypothesis is a hypothesis about how our psychology is composed:

  • Modularity: we come pre-wired to analyze information in certain ways.

The motivation for modularity is that humans are able to solve a vast amount of problems easily and efficiently. Just like some believe we come pre-wired to pick up a language, many think we come pre-wired to pick up norms and other information (mind-reading) to navigate our complex physical and social environment.

Sterelny rejects modularity. Sterelny accepts with the standard model that cooperation drove human evolution, but he denies that it was the only driver. He thinks that pre-wired modules are insufficient because the environments faced by humans are vastly diverse and varied, so no modules could help. According to modularity, human psychology is the same everywhere. Sterelny denies this. He holds that our minds develop differently depending on the environment and how we choose to interact with our environment. Sterelny agrees with the mainstream that we need to explain how it is that people do so well in solving novel problems in novel environments.

1.2 The Social Intelligence Hypothesis

According to the standard model, we need to cooperate with each other to survive. And this requires complex cognitive capabilities. The standard view faces the problem of free riders. How do we get free riders to contribute? Theorists here focus on the effects of punishment, error, coercion, rewards, norms, and so forth. What they don’t do is focus on the mechanisms that make cooperation rewarding.

Sterelny explains what makes cooperation difficult: 1. Our complex needs 2. Specialization. 3. Social hierarchies. 4. Longevity of life, memories, and relationships. 5. Increasing population sizes. 6. Sex and gender roles. 7. We need communication, but not all of it is honest. 8. Complex norms. 9. Partial transparency. Given all these factors, Sterelny notes that we don’t just need to study the motivation for cooperative behavior. We also need to explain what makes it profitable.

1.3 Cooperative Foraging

At some point, we became social foragers. This requires complex ecological information, cooperation, and technology. The amount needed will vary depending on time and place. Groups of people would need to coordinate and thus communicate on how to hunt large game. Even here, hunting plans didn’t always go as planned. Emergencies occur and decisions have to be made on the fly. This increases cognitive requirements.

1.4 cooperative foraging and knowledge accumulation

Sterelny holds that what is necessary for coordination is necessary for cooperation. Early humans needed trust, teamwork, mind reading, social information, ecological information, and talent. This is what makes joint action profitable. The information and abilities needed, as well as the technologies required, became so complex that it could not be invented every generation. Skills, tools, and other practices had to be handed down generation to generation. He cites as an illustration work by Morrison (1981) who documents aboriginal children learning to recognize tracks by learning how to reproduce them.

Sterelny then introduces us to the archaeologist concept of “behaviorally modern.” These are cultures that depend on high volume, high fidelity cultural learning. Humans have unique social environments that allow for behaviorally modern ways of life. Skeptics say that social learning has to be a recent development, and hence not a main driver of human evolution. Sterelny responds that adding cognitive capital might be a new phenomenon, but preserving it is not. “Around 400,000 years ago, humans were using material technology that could not have been reinvented anew each generation…” (15).

Life in a changing world

Our physical environment changed radically: climate change has always been a challenge we face. Humans moved from dry to cold to forests. We encountered different kinds of animals, from small to big, to dry and wet. Often, we even changed the environment, say by burning it. Our social environments have always changed. We develop norms for esteem and prestige. Population sizes changed too. Different levels of specialization were enabled by different population sizes. “While some principles of biology and naive physics are constant across the ecological challenges those environments pose, the constant features are extremely coarse-grained. Most of what these different peoples need to know will be specific to their circumstances” (19). This makes the modularity hypothesis suspect.

Chapter 2: Accumulating Cognitive Capital

2.1 A Lineage Explanation of Social Learning

One central factor in human evolution is social-cultural learning. Humans live in changing worlds–in worlds with few constants, and that change is multidimensional: “hominin social environments changed size, degree of inidivual specialization, life history, patterns of intergroup relations, forms of hiearchy and within group control, sexual division of labor, and the role of norms and ritual” (24). To explain this feature of humans, we need a “lineage explanation.” This is an explanation that explains how, gradually, we became agents with capacities for social learning. This is the task of this chapter.

The account “...begins without special adaptations for cross-generational cultural learning, but these evolve as a consequence of the establishment of cultural learning…” (26). The account begins with cognitive mechanisms for a-social learning that get adapted for social learning, thus expanding the scope of social learning. Moreover, the environment is often constructed to facilitate learning. So most social learning is hybrid, involving social mediation as well as individual resources (experience/trial and error).

2.2. Feedback Loops

Sterelny claims that there are two important feedback loops driving human evolution. The first is between expertise, social learning, and life history. Animals socially use information about the position of others to make decisions. By contrast, humans developed long childhoods, adolescence, and an extended end of life in which they share information. Children are expensive, so this requires ecological innovation. So, life span changes co-evolve with innovation.

The second feedback loop involves the interaction between social learning environments and individual cognitive adaptations. Parents can structure the environment so a child learns by trial and error. Sterelny says this is social learning even if the cognitive mechanisms are a-social. But this new dynamic can put pressure on social learning adaptations. Older generations become more tolerant of the young, and the young become more focused on older generations. Moreover, individual learning and social learning can be enhanced by similar mechanisms: memory, attention, inhibition, and monitoring one’s actions.

2.3 The Apprentice Learning Model

Sterelny proposes we understand information transmission via the apprentice model. This involves “...learning by doing in an environment seeded with informational resources” (35). The cooperation can be jointly beneficial. The expert can get work out of the learner. Sterelny claims four benefits of this model. First, as discussed above, the model does not require a special cognitive adaptation for social learning, but it can build on it. Second, apprentice learning enables high fidelity, high-bandwidth knowledge flow. Third, the model fits ethnographic data. And fourth, it illuminates the archaeological records (next chapter).

Apprentice learning does not require formal instruction or institutions. Experts can structure the environment without explicit teaching, say because early learners can hear their conversations. Learners can see intermediate steps of the process. Sometimes, instruction may have to be more guided or explicit.

  • Incremental construction of apprentice learning. Early hominins probably did not have a fully developed apprentice practice, but the model still applies to them because the young would learn in structured environments.

  • Ethnographic plausibility. Foraging societies sometimes provide their young with toys that match their size but relate to their activities (e.g., small hunting bows). They teach the young how to make tracks of the animals they follow, or they keep semi-wild pets so that the young become familiar with the animals. Children are taken on safe versions of adult hunting trips. Others report learning what types of mushrooms were poisonous by being presented examples. Still other ethnographic records reveal practices very much akin to formal apprenticeships.

Sterelny concludes that any theory of the human mind needs to explain two things. First, the world is constantly changing, yet we have done relatively well as a species. Second, we accumulate cognitive capital. How is all this achieved?

Chapter 3: Adapted individuals, Adapted Environments

3.1 Behavioral Modernity

Sterelny’s model of social learning involves older generations structuring the learning environment of the younger generation so that the younger generation can learn via observation, trial and error, and explicit instruction. For social learning to go well, both the right kind of mind and the right kind of environment is needed. This chapter uses this insight to resolve two paleoanthropological puzzles. One puzzle is the puzzle of why neanderthals went extinct. The other puzzle is the lag (100-150k years) between human species and humans as behaviorally modern species. We start with the latter.

Anatomically modern Humans appear around 200,000 years ago (46), but their behavior is radically different from modern humans. Their technology is simpler, their foraging is narrower, and their social organization is rudimentary. Researchers often refer to later humans, the ones with more complex tools, social structure, and practices as “behaviorally modern.” Sterelny acknowledges that the difference might be due to the lack of evidence. The tools they used might have disappeared, and we might be wrong about their social organization. Nevertheless , Sterelny holds that, “...the apparent differences are so large and long-lasting that researchers generally agree that the cultures of the First Sapiens and Moderns differ qualitatively” (p. 46).

There is another wrinkle. The development of technologies isn’t gradual, with more primitive ones coming first and more advanced ones coming later. Rather, the record is mixed. What first was thought to be a late invention is sometimes discovered in very early civilizations. So the situation is this. Comparing sapiens from 200,000 years ago to the ones from 40-60 thousand years ago, one finds a dramatic shift in the ways of life: technological, agricultural, cultural, and so forth. Nevertheless, the change is not gradual. Characteristically modern features appear rather early on. We can put the puzzle as follows. If we think that the early sapiens were cognitively different than the later ones, what changed? If we think they were the same, why did modern behavior take so long?

3.2 The Symbolic Species

A common answer to the puzzle of behavioral modernity is that humans use symbols, and symbols indicate a distinct cognitive sophistication. The existence of changing signals in our evolutionary history, such as beads and artwork, is evidence for a change in cognitive power. So, the thought continues, maybe what changes from the early period to the late period is a kind of cognitive sophistication.

According to this approach, symbol use is a criteria for modern humans. Why think that symbols point to cognitive sophistication? One answer is that symbols don’t usually resemble what they symbolize. The word rabbit, for example, does not look or sound like a rabbit. The other answer is that symbols refer to things temporally and spatially displaced. The word rabbit can refer to a rabbit even when none is around.

Sterelny agrees that symbols are important, for we don’t just form groups, we identify ourselves with groups. Symbols help us do this. But he notes that there is a difference between symbols that make a heavy cognitive demand, and symbols that help us become “self-conscious collectives.” For example, if someone wears face paint like that of their tribe, the color is present, so it is spatially and temporarily contiguous. If someone wears expensive beads, this is a symbol of high social status, but this convention isn’t arbitrary. Expensive beads carry an honest signal of success, just like driving a ferrari would. The symbols of group identity need not be cognitively demanding symbols. Sterelny concludes that what is distinctive about the change in symbols, and so what is explanatorily important, is not a change in cognition but a change in social organization, and he holds that symbols become both a cause and effect of social organization.

3.3 Public Symbols and Social Worlds

Sterelny begins with a discussion comparing ocher and beads as symbols. Ocher occurs early on in hominin history and beads occur later. Ocher can be useful as a preservative, repellent, or glue. It can also be used to affect signals, say as camouflage. By contrast, beads can encode more precise information. How they are arranged on a garment or how many one has can mean something, say that one has high social status. But this would only mean something in an arbitrary way; it would require a convention, and people outside the group won’t recognize the convention.

Sterelny then discusses the population structure hypothesis:

  • symbols of group membership were not for identifying with the group. Instead, they were a form of advertising. As group size increased, as social relations became more complex, and as trade increased, individuals needed a way to distinguish themselves in various ways: as successful, as trustworthy, and so forth.

Symbols help do this. Sterelny concludes that symbols have social roles, and these occur because of complex, thick social networks. Sterelny writes that “if these considerations are persuasive, the appearance of ocher, beads, and the like in the archaeological record is an effect of demographic and social change, not the first emergence of self-identifying groups” (54). Sterelny seems to be saying that the uses of these materials goes in line with demography and social change, not group identification. So, ocher can be used for camouflage. Beads can be used to show ones status in a group.

3.4 Preserving and Expanding Information

Theorists usually think that Australia was populated by individuals who were behaviorally modern. The conditions required to make it to Australia were too harsh to resemble anything like “a pregnant woman on a log.” However, they left no sign of behavioral modernity upon arrival. This is a puzzle for those who think that behavioral modernity rests on a kind of cognitive sophistication. Proponents might say that humans can be behaviorally modern without leaving evidence of it.

Sterelny proposes an alternative: behavioral modernity can be lost if one of its key supporting factors (e.g., demography) are lost. Behavioral modernity consists both in preservation of information and innovation on it, and behavioral modernity is a collective capacity. And what allows all this involves three factors: individual cognitive capacity, a structured learning environment, and population structure. In regard to cognitive capacity, Sterelny cites cultural learning capacities. Such capacities include theory of mind, communication, observation, and reflective understanding of one’s skills.

Next Sterelny turns to features of population that allow for “stable, high bandwidth, high fidelity flow of information.” On average, a larger group can better support specialization. In turn, specialists are most likely to innovate. They are in the best position to refine their craft. Also, a higher population means that cultural information is stored in more minds (redundancy). As a result, accidental deaths in some members are less likely to result in loss of cultural information. Because of redundancy, the problem of low information fidelity can be met. If the information is stored in multiple minds, an agent with poor details can acquire information from multiple sources.

On Sterelny's view, individuals entering Australia would have been stretched thin as they advanced into a large area, thus lowering demography. More generally, Sterelny explains the puzzle of behavioral modernity by appeal to group size. Behavioral modernity is a more recent phenomena since population size, structured learning environments, and social organization all co-evolve to produce it.

Neanderthal Extinction

Sterelny argues that his model can explain neanderthal extinction. As the climate was changing, the environments of large game were shrinking. This in turn forced neanderthals to dwindle in number. Smaller numbers are bad for large game hunting as higher risks come with hunting in smaller numbers. The change might have been too rapid for neanderthals to develop cultural adaptations to live in other environments. The change would have required them to quickly learn about new food resources, new tools and weapons, and to pay the learning costs. It seems they were stuck in a fitness trap. Following their known ways of life would eventually not workout for them. Trying new ways of life was a high risk.

Chapter 5 Costs and Commitments

5.1 Free Riders

The chapter begins with the idea that humans have evolved by extensive cooperation. We depend on each other for informational, ecological, and reproductive cooperation. Evolutionary biology tells us that cooperation is fragile. Cheats, free riders, defectors, bullies, and the untrustworthy threaten the stability of cooperation. This chapter begins to address the question of how cooperation can be stable despite the risk of cheats and the untrustworthy.

Sterelny rejects the idea that the problem is solved by being able to identify cheats. In small scale societies, defection may be easily identifiable. But the problem here will be in controlling for defection. By what mechanism is the temptation to defect controlled? One idea is that we engage in partner choice. The problem with this solution is that we don’t always get to choose who we will interact with. This is the case with bullies. So we need to explore how cooperation manges to be a stable feature of human life.

5.2 Control and Commitment

If an agent wants to take your resources, you should retaliate. However, if you are strong enough to retaliate, you may consider taking resources from others. This scenario is one where cooperation is unstable. One solution is for multiple agents to band together. They must decide to collectively retaliate against other agents who threaten to take their resources. And they must find a way to control free riders in the band. In short, they must make threats of consequences to outsiders and insiders.

The question is what makes these threats credible? Delivering a sanction is costly. There is always the temptation to make false threats. To keep threats of sanction reliable, we need a story about how we commit. This is a commitment device:

An agent optimizes his (her) future utility if (s)he acts at time T to ensure that at

T + n (s)he will carry through option X (if some triggering event occurs), even

though choosing X at T + n does not then optimize utility (104).

The next question is whether evolutionary processes can assemble commitment devices.

5.3 Commitment Mechanisms

Sterelny begins with Schelling’s distinction between external and internal solutions to commitment dilemmas. An external solution involves modifying the environment in order to make commitment valuable. A dramatic example is one where the general burns the bridge from which their army can retreat. Even in the face of losing, continued battle is the best option because it is the only option. Similar strategies involve public ceremonies. One barrier to this route is that some external strategies come too late in human evolution. For example, legal contracts exist in formalized institutions–ones where cooperation has already been well established.

Frank (1998) proposed a classic internal solution. Internal solutions depend on the agent's psychology. The idea here is that agents are such that when temptation comes, the committed agent puts a high price on loyalty, fairness, and sanctions. Thus, while the resource maximizing option may be to defect, agents subjectively evaluate commitment as the most valuable option. Frank proposes that our social and moral emotions are commitment devices.

Sterelny asks two challenging questions about Frank’s proposal. Defection sometimes means giving up on resources. And resources are motivating. So, how do emotions function to motivate commitment in a way that overcomes the motivations prompted by resources? Second, even if some are trustworthy as Frank says, how can we identify those who merely pretend to be trustworthy? Sterelny concedes that Frank’s account has insights, but he does not think it is the whole picture. Sterelny will detail how we have managed to construct environments where temptations to defect are reduced.

5.4 Signals, Investments, and Interventions

One strategy employed for identifying the trustworth is the costly signals model. Sterelny says, “costly signals are honest signals of sender characteristics, because only honest signalers can afford the handicap of sending them” (109). Sterelny rejects this model because he sees many situations where all members of a group are sending the signal. If everyone is able to send the signal, then the function of the signal cannot be to differentiate between the reliable and the unreliable. Nevertheless, he thinks costs play three important roles.

First, costs amplify the effects of arousal. Singing emotionally changes both the sender and the receiver, and this can in effect reinforce mutual bonds. Second, signaling costs are sometimes so high that one becomes “invested.” Examples of this include situations where one joins a gang or clan that leaves one marked. Third, honesty has by products that the dishonest must fake, and this imposes a unique cost on the dishonest. If I want to fake love for rock music, I have to go to concerts, learn songs, bands, and trivia. I have to fake enjoyment of songs I hear. These are costs I have to pay that the true rock lovers do not have to pay.

5.5 Hunting and Commitment

Sterelny does not think that the costly signal models are able to explain “all in” signals, such as why everyone hunts. He thinks that, instead, costly signals function in two ways. Internally, costly signals change the psychology of the agent. Externally, costly signals change the cost of defection. He starts with the internal signal. The main idea here is that humans enjoy collective activity. Moreover, high cost, collective activity creates psychological tendencies towards loyalty. Collaborative efforts such as hunting, combat, sports, and dance make people more trusting and trustworthy. Sterelny notes that this does not depend on some costs being higher for unreliable signalers.

5.6 Commitment Through Investment

Another way the problem of cooperation can be addressed is by committing by investment. The clearest example of this is individuals who become irreversibly marked to join an outcast group. Irreversibly marking members of a hated war tribe means that anyone who defects will be killed either by insiders or outsiders. Gang initiation rituals often involve commiting crimes that make members outcasts of society. The idea is that one can make one’s investment public so that defection no longer pays. Sterelny points out that this is no longer a handicap signal: it is not a signal that is costly only to the unreliable.

Sterelny explains that both internal and external solutions to cooperative dilemma co-evolve and are mutually reinforcing. External investment leads to group activity, which leads to internal pressures to be reliable, which in turn means that one is even more invested.

5.7 Primitive Trust

The challenge of this chapter is to explain the reliability of trust. Various forms of human cooperation depend on trust. Sterelny appeals to three human features. First, the ability to construct niches. By investing in relationships, we alter the cost-benefit structure of defection. Second, evolutionary mechanisms co-evolve. In this case, as cooperation increases in a myriad of ways, the benefits of defection are reduced. Third, cooperation is a sort of joint activity, and joining activity breeds trust.

Trustworthiness has often been seen as an internal feature of an agent’s psychology. For this reason, it has been natural to focus on signals of reliability and hence costly signals. Sterelny notes that his account calls such picture into question. Instead, it seems that trust and trustworthiness is built up by the three mechanisms just mentioned.


Sterelny K. (2014). The evolved apprentice : how evolution made humans unique. MIT


39 views0 comments
bottom of page