At the Middleton Theater, the chance placement of objects sometimes changed routine practices. For example, when the manager decided to fix his home refrigerator, he brought in all the parts—sheet metal covering, fan, Freon pump, and so on—and stacked them in a precarious pile behind the concessions counter. Under a sign that said “Don’t move,” the pile stayed there through the summer, half blocking the butter machine, which resulted in lopsided popcorn buttering. Eventually, Jerry, the regional supervisor from Milwaukee, showed up and ordered the pile of junk to be removed. Had he not stepped in, the pile would eventually have become a fixture. If the Middleton Theater had multiplied as a franchise chain—instead of being razed in the 1990s—a pile of old refrigerator parts might be in some of the spin-off theaters and lopsided buttering in others. Material culture and behavior would have evolved together into new branches.
In the 1980s, a few anthropologists and archaeologists, including Mike, maintained that stone tools, pottery, and even language are subject to evolutionary processes in the same way that teeth, cells, and bones are. Not coincidently, this was about when Richard Dawkins coined the term extended phenotype to refer to inherited traits outside the body. Classic examples include a beaver dam, spider web, bird nest, and termite mound—all of which are “tools” that protect both the organisms and their genes. Those genes, or replicators, are the units that express an organism’s behaviors in succeeding generations.
Back then, the academic response ranged from skepticism to ridicule. Pots and arrowheads were tools used by the people who made them, but pots and arrowheads don’t breed, so the argument went. Furthermore, prehistoric change in technology, or culture generally, was intentional: humans had ideas, which are in no sense equivalent to genes, and they carried them out. End of story.
Over the next couple decades, however, cultural evolution grew as a field in anthropology and even branched off into many subspecialties that recently reached popular culture. Words such as meme are now widespread, for example, and the idea of technology as extended phenotype is no longer radical. By 2015, over 90 percent of a sample of a thousand Americans aged sixteen to fifty-five considered the Internet to be an extension of their brain, and almost half treated their smartphones as if they were part of their memory. Indeed, as people extend themselves through these devices, online connection becomes effectively obligatory. Since smart devices shape and increasingly constitute our personal environment, they ought to qualify as part of the human phenotype.
The Acheulean Hand Ax
As the popular conversation continues, we occasionally see images of the iPhone juxtaposed with the Acheulean hand ax—the Pleistocene age stone tool that was used by our ancestors from about 1.7 million to perhaps 100,000 years ago. The comparison is a profound comment on the significance of the iPhone, because the hand ax is considered a landmark in human evolution, having enabled Homo erectus to disperse out of Africa, and set up shop in Europe and Asia. Some paleoanthropologists argue that the Acheulean hand ax was hardwired into the hominid brain—or that it was at least partly under genetic control. The iPhone has superficial similarities to the hand ax, as both are handheld, multipurpose tools that shape the personal environment. Besides the obvious technological differences, though, there is also a big evolutionary difference: unlike smartphones, which seem to change overnight, the hand ax essentially never changed over hundreds of thousands of years. There were slight regional differences, but for all intents and purposes, an Acheulean hand ax was an Acheulean hand ax, no matter where you were in the world.
Imagine inheriting a technology that had remained unchanged for hundreds of thousands of years. Our Homo ancestors, who probably learned to make stone hand axes as children, would have had little concept of changing this tool. It’s all but impossible for us to imagine technological stasis for over hundreds of thousands of years. It’s like imagining the distances of interstellar space. How could so little change be possible? The change in Paleolithic tools was not even glacially slow; the glacially carved gorges and waterfalls of upstate New York, for example, are only about twelve thousand years old. Surely, accidental improvements every few generations would have changed the Acheulean hand ax faster than that, yet the archaeological record shows it did not. Why not?
Maybe our ancestors were too stupid to invent anything new. But this explanation is dubious, given that hominid brain size more than doubled during the Pleistocene, from, say, two million to five hundred thousand years ago, and yet stone tools barely changed during that time. If the tools had been constrained by brain power, we would expect changes in parallel with brain size, yet we don’t see them.
Alternatively, maybe hominids needed larger groups for technological change to occur. Yet this applies to more complex technologies, where it pays to learn from the expert in the group, or where the group can afford to support technological specialists. In the case of Pleistocene stone tools, probably every individual could knap a hand ax without necessarily learning from an expert—if there even was one. In fact, undergraduate students in archaeology class, given gardening gloves, a pile of flint nodules, and the chance to bang them together, learn the basics of flint knapping quickly, and can soon turn out a passable Acheulean hand ax.
Emulation versus Imitation
Given that students can make good progress in an afternoon, perhaps the unchanging hand ax also served as the physical model or blueprint that our ancestors used for making more hand axes. This would be emulation, which means copying just the outcome or goal, as opposed to imitation, which means copying the method of getting to the goal. The difference is essential to cultural evolution; imitation was critical for complex human technology and culture, whereas primatologists debate whether chimpanzees can truly imitate as opposed to just emulate.
The difference also taps into the mind of ancient Homo. We’d like to know when our ancestors started to imitate rather than merely emulate. While the archaeological evidence of the Middle Stone Age doesn’t directly show how they learned, it does reveal what they did, which can be reverse engineered from the distinctive patterns of debris they left behind. Find, for instance, a grapefruit-sized flint nodule. Knock off flakes around it to prepare a “core.” Rotate the core here, knock off flakes there, rotate, and repeat. Prepare a “platform” on the core. Knock off a flake from that platform. Rotate sixty degrees, knock off another flake, and so on.
To resolve the question of emulation versus imitation, however, we need evidence for how they learned to do it. Archaeologist Jayne Wilkins—who discovered the world’s earliest evidence that stone points were used for spears, some 500,000 years ago—reasons that imitators would leave behind similar scatters of stone flakes each time, whereas emulators, with their unique ways, would leave more varied patterns. At Kathu Pan, a site in South Africa dating to half a million years ago, early modern humans made flint blades—long, narrow stone tools, sharp enough to slice up cooked rabbit, use as projectile points, or shave animal hides. Here, Wilkins believes the variation in debris patterns favors emulation. If Acheulean hand axes were made through emulation, this could explain the slowness of change in the Middle Stone Age: the hand axe itself, as the blueprint for its own reproduction by emulation, was under natural selection as part of the extended phenotype of these hominins.
A half-million years later, the first big game hunters arrived in North America, following their ancestors’ migration over the Bering land bridge between Siberia and Alaska about 14,000 years ago, when the glacial sea level was about a hundred meters lower than it is now. To make their delicate and refined projectile points, called Clovis points, apprentices would have tried to master an expert’s knapping process by careful imitation. Emulation was not an option. Also, unlike the Acheulean hand ax, Clovis points changed in just the several hundred years they were in use, between about 13,300 and 12,500 years ago. Can we order those changes in time, just by looking at the artifacts themselves? We can, and therein we find insight for mapping technological evolution in general.
Evolutionary Trees
Evolution is often used as a word for “change,” but the true meaning is more specific. Evolution means there are different variants transmitted between generations, over which these variants are sorted as some are transmitted more frequently than others. Although first developed for tracking biological evolution, cladistics—tracking the history of related entities through their shared features—can be applied to anything undergoing evolution.
This includes technology, so let’s look at the phylogeny of those Clovis projectile points. First, we need to choose which technological features to focus on. As a simple example, we can track a single feature, called fluting, which is the removal of a long flake from the base of a projectile point. The trees in the figure show three snapshots in the evolution of a projectile point lineage. It begins with the unfluted ancestral state A, which continues its own lineage while also giving rise to ancestor B, the “derived” state, which is fluted. In this new lineage, ancestor B then gives rise to two more groups that are both fluted, which is the “shared derived” state of these groups because they share fluting only with their immediate common ancestor.
With us so far? Let’s go one more step. As shown in the third tree, fluting is now old hat for the two newest groups that have emerged, so for these two, fluting is their “shared ancestral state.” But if we were talking about these two new groups and one of the older fluted groups, then fluting becomes the derived state again because it’s what they all share with their common ancestor, B. For reconstructing historical relationships, shared derived traits are more useful than shared ancestral traits because they are inherited (derived) from the groups’ most recent common ancestor. Now that we have those basics covered, let’s see what evolutionary trees tell us about languages, technologies, and maybe even the future.
Languages and Folktales
Even more so than technology, human language is strongly inherited and evolves in a treelike pattern over time. Ancestral languages such as Latin branch into descendants such as the Romance languages, which share features derived from their common ancestor, as Spanish, Portuguese, Italian, Romanian, and French speakers would know. Robust language phylogenies—treelike representations of inferred evolutionary relationships—exist for major language groups that often reflect remarkably ancient human dispersals. The phylogeny of Austronesian languages reflects the pioneering voyages of ancient seafarers, from Madagascar all the way to Easter Island. The evolutionary tree of Indo-European languages—Hindi, Germanic, and Romance languages and more—parallels many of the human migrations in Europe over the past eight thousand years, from the spread of farming out of the Levant to more recent migrations of Anglo-Saxons and Vikings into Britain.
As these migrating peoples taught their languages to their children, they also told them stories—the same stories they had learned from their own parents. As we discussed in chapter 3, many folktales are incredibly old—their time depth being a measure of the faithfulness of transmission. The geographic extent of such stories is tied to the spread of groups across the landscape. In his phylogenetic study of Little Red Riding Hood, anthropologist Jamshid Tehrani collected fifty-eight contemporary versions of the folktale from around the world. To break the versions down into discrete features, Tehrani chose plot elements shared among some but not all versions of the tale. In some Asian versions, for example, the villain drinks oil or spring water to clear his throat after he fails to impersonate the child’s mother. In certain African versions, the wolf cuts his tongue to smooth out his voice. Identifying the essences of these plot elements, Tehrani gave them labels such as “excuse to escape,” “dialogue with the villain,” or “hand test” (the children ask the “grandmother” to show her hand through the door). Once he had painstakingly cataloged all these different narrative features worldwide—98 percent of the work, as he will tell you—Tehrani generated a phylogeny and estimated the age of the common ancestor. His estimate that it was at least two thousand years old appeared just as billboards for Red Riding Hood, the 2011 Warner Brothers movie, were advertising it as an eight-hundred-year-old tale. Like the pile of refrigerator parts at Middleton Theater, folktales have been around longer than we think and are surprisingly stable.
After that study, Tehrani and his colleague Sara Graça da Silva looked at seventy-six magic-based folktales, including Rumpelstiltskin and Beauty and the Beast. Their phylogenetic analysis revealed that one tale, about a blacksmith who makes a deal with the devil, was about six thousand years old, making it Proto-Indo European, the ancient ancestor of most European languages and Hindi. Yet rather than blacksmiths, Neolithic farmers, who had no metal tools, were believed to have brought Proto-Indo European to Europe. Remarkably, a phylogenetic study of ancient folktales has reopened a fascinating debate about ancient technologies.
Complex Technology
Language itself can also be a technology, as with computer programming languages. Tracking the evolution of computer languages since the 1950s, Sergi Valverde and Ricard Solé of the Santa Fe Institute suspected a branching pattern was involved. In the 1980s, for instance, the C++ programming language branched off with object-oriented programming. Another branch emerged in the early 1990s, when James Gosling invented the language that later became Java, which became the world’s most popular programming language, especially for web pages.
The figure, which greatly simplifies the analysis, boils the phylogeny down to four major computer languages: Basic, Pascal, Python, and Java. Based on their features, the tree shows that Python and Java are more similar to one another than either is to Basic or Pascal. All four evolved from Fortran, but Pascal, Python, and Java look more like their common ancestor, Algol-60, than they do Fortran. Fortran is what unites those three with Basic. In this tree, Python and Java, together with their common ancestor, C++, form a clade. Pascal, Python, and Java, and their common ancestor, Algol-60, form another, more inclusive clade, and so on.
Just as with stone tools or biological species, the phylogeny of programming languages shows abrupt bursts of diversification instigated by key inventions. Zoom in on this phylogenetic tree and you’ll find smaller changes at a finer scale, still branching off and nested within one another. If we remove the time scale, the general nested branching pattern could represent another technology, whether it is stone tools, metal weapons, or transistor radios. Researchers have explored the US patent database and found the same interdependence of innovations as with the evolution of stone tools, just accelerated a thousand times faster.
A cascade of inventions can often emerge from a novel combination of existing technologies. With a little creativity, this might help anticipate or even shape the future. Given that clades tell us what evolutionary space has been occupied, they might also show what nearby but still-unoccupied concept spaces are ready to be explored. For illustration, let’s say Elvis had looked at a cladogram of music in 1953. Phylogenetics wasn’t around then, but let’s pretend it was. He might have seen a clade for country and western and its closest relatives, and a neighboring clade for rhythm and blues and its closest relatives, with an inviting open space in between. This is not exactly how rock and roll began, but the music did, effectively, colonize open space between existing clades.
In addition to music, we could see phylogenetic trees and treelike structures being used to examine all kinds of opportunities. For example, César Hidalgo of the MIT Media Lab generates treelike diagrams to represent how closely related technological specializations of different countries are. The closer two products are in terms of branch links, the more likely a country could export the one product given that it exports the other. Unfilled space between two linked products might identify new opportunities.
In the next chapter, we’ll take cultural phylogenetics a step further and show how we can use one solid phylogenetic tree, especially a language tree, to infer evolutionary changes in other aspects of culture, such as family structure and political systems. We also briefly look at how this might be used to consider the future.
