Scientists unveil a 1.8 million-year-old ‘Dmanisi’ skull discovered in the Dmanisi caves in modern-day Georgia. Photo by Valerie Kuypers/AFP/Getty Images


Did Homo erectus speak?

Early hominins who sailed across oceans left indirect evidence that they might have been the first to use language

by Daniel Everett + BIO

Scientists unveil a 1.8 million-year-old ‘Dmanisi’ skull discovered in the Dmanisi caves in modern-day Georgia. Photo by Valerie Kuypers/AFP/Getty Images

What is the greatest human technological innovation? Fire? The wheel? Penicillin? Clothes? Google? None of these come close. As you read this, you are using the winning technology. The greatest tool in the world is language. Without it there would be no culture, no literature, no science, no history, no commercial enterprise or industry. The genus Homo rules the Earth because it possesses language. But how and when did we build this kingdom of speech? And who is ‘we’? After all, Homo sapiens is just one of several species of humans that have walked the Earth. Does ‘we’ refer to our genus, Homo, or to our species, sapiens?

To discover the answers to these questions, we need to travel back in time at least 1.9 million years ago to the birth of Homo erectus, as they emerged from the ancient process of primate evolution. Erectus had nearly double the brain size of any previous hominin, walked habitually upright, were superb hunters, travelled the world, and sailed to ocean islands. And somewhere along the way they got language. Yes, erectus. Not Neanderthals. Not sapiens. And if erectus invented language, this means that Neanderthals, born more than a million years later, entered a world already linguistic.

Likewise, our species would have emerged into a world that already had language. In spite of the fact that many paleoanthropologists view erectus as little more than a skinny gorilla, of few accomplishments, far too stupid to have language, and lacking a vocal apparatus capable of intelligible speech, the evidence seems overwhelming that they had language. Erectus needed language. They were capable of language. And, though often denied in evolutionary studies, the ‘leap’ to language was little more than a long series of baby steps, requiring no mutations, nor any complex grammar. In fact, the language of erectus would have been every bit as much a ‘real language’ as any modern language.

Erectus was an imposing creature. Males stood between 173 cm and 180 cm. Their immediate ancestors, the Australopithecine males, were only about 137 cm tall (their immediate ancestors might have been Homo habilis, but only if we accept that habilis were not Australopithecines, or that they were a separate species from Homo erectus, neither of which is clear). The brains of these early humans averaged around 950 cubic centimetres in volume, double the size of the Australopithecines, though smaller than those of male Neanderthals (1,450 ccs) and sapiens (1,250-1,300 ccs), but still within the range of modern sapiens females. The vocal apparatus of erectus might not have been much more advanced than that of a modern gorilla or it might have been more similar to ours. But whether their speech sounded different than ours or not, it was nevertheless adequate for language.

Evidence that erectus had language comes from their settlements, their art, their symbols, their sailing ability and their tools. Erectus settlements are found throughout most of the old world. And, most importantly for the idea that erectus had language, open oceans were not barriers to their travel.

Erectus settlements show evidence of culture – values, knowledge structures and social structure. This evidence is important because all these elements enhance each other. Evidence from the erectus settlement studied at Gesher Benot Ya’aqov in Israel, for example, suggests not only that erectus controlled fire but that their settlements were planned. One area was used for plant-food processing, another for animal-material processing, and yet another for communal life. Erectus, incredibly, also made sea craft. Sea travel is the only way to explain the island settlements of Wallacea (Indonesia), Crete and, in the Arabian Sea, Socotra. None of these were accessible to erectus except by crossing open ocean, then and now. These island cultural sites demonstrate that erectus was capable of constructing seaworthy crafts capable of carrying 20 people or more. According to most archaeologists, 20 individuals would have been the minimum required to found the settlements discovered.

Because the stone tools of erectus were simple and slow to evolve, some have rushed to conclude that they lacked intelligence for language. But stone-cutting implements are simply not the whole story. The evidence for erectus island settlements means that they built water-transport craft. Erectus seem to have had art as well, as exemplified in the 250,000-year-old Venus of Berekhat Ram.

Further, archaeologists have discovered 400,000-year-old wooden thrusting and throwing spears in lower Saxony (called the ‘Schöningen spears’), which suggest a robust hunting culture. Thrusting spears, for example, require at least one member of a group to get close enough to the prey, such as mastodons, to pierce them with the weapon. Hunting culture entails cooperation and planning with others.

One might object of course that these spears and other tools belong not to erectus but to some other pre-Neanderthal, pre-sapiens species of Homo. This objection has little force. In spite of a proliferation of names for Homo species that preceded Neanderthals (rudolfensis, ergaster, heidelbergensis, naledi, floresiensis and so on), the evidence is unclear for all but three species – erectus, neanderthalensis and sapiens. The burden of proof still rests upon those trying to distinguish the other proposed species of Homo. Finer distinctions among early Homo species are questionable, not only due to the absence of convincing evidence for such divisions but also by other positive evidence.

One source is the discovery of the Dmanisi skulls, 1.8 million-year-old skulls found only a few years ago in the Dmanisi cave in the modern-day country of Georgia – all from the same species. These reveal roughly equivalent amounts of variation in shape to that found in purportedly distinct species. The other source of evidence comes from new research that Homo erectus in China seem to have evolved directly into Homo sapiens, a position underscored by more recent discoveries. So why are so many different species suggested? Partially because such proposals represent valid attempts to account for the diversity in form of different fossils and sites. But the proliferation of names is also spurred on by the attraction of being the first to name a new species.

The truth is likely much more mundane – Homo changed gradually, with no clear end or beginning points between species, other than the three I support, if even there. However, even if later scholarship establishes that there were indeed more intermediate species and that in fact one of these, such as Homo heidelbergensis, first discovered language, the basic thesis of this essay is unchanged. Language was invented hundreds of thousands of years before Neanderthals and sapiens. But for now, the most reliable scholarship supports the idea that the only clear human species prior to Neanderthals was erectus.

‘Evidence from Flores suggests that our early ancestors were successful seafarers’

Moreover, sailing demonstrates a level of cognitive development rivalling even that of modern humans. The erectus accomplishment of paddling together across one of the strongest ocean currents in the world, such as the Throughflow that then and now surrounds the island of Flores in Indonesia, required not only cooperation, but also corrections, instructions and commands. Few detailed instructions or corrections can be given without language.

Since the voyaging capabilities of erectus are so important to assessing their linguistic ability, this is something we need to be clear about. The only alternative to the idea that erectus built boats or rafts is that erectus individuals were accidentally transported to other islands via logs or naturally occurring vegetable rafts or some such. But this doesn’t account for the evidence. In Stone Age Sailors (2014), the anthropologist Alan Simmons puts it like this:

our ancestors have often been painted as unintelligent brutes … however … this simply is not the case. Evidence … suggests that at least Homo erectus and perhaps pre-erectus hominins were early seafarers … based on this evidence, it seems that our early ancestors were … successful seafarers … Biological studies suggest that considerable numbers of ‘founder populations’ are required for viable colonisation.

As Simmons concludes: ‘Given these variables, purposeful seafaring, involving intentionally constructed craft capable of carrying relatively large payloads (that is, people plus provisions) over considerable distances, is a more plausible model, in light of the increasing global evidence of early humans on many islands.’

To build and operate boats, erectus needed to talk about what material to collect, where to collect it, how to put the material together and so on – just what we ourselves would need to talk about in order to build a raft. In addition to the assembly of a raft, the planning for the trip as a whole, the reasoning for the undertaking, would have all required language.

We can therefore conclude that erectus required language. But how difficult would it have been for them to invent language, even with their massive Homo brains? Well, this depends on what is meant by language. There are two fundamental components to language that all linguists agree upon – grammar and symbols. Although some linguists take grammar to be the most important component of human language, others take symbols to be more important. As seen below, though, once symbols appear in language, grammar comes along nearly for free. To understand the nature of the erectus invention of language, it is first important to recognise the distinction between communication and language:

  • Communication is the transfer of information.
  • Language is the transfer of information by symbols.

In my book How Language Began (2017), I make the case that erectus symbols began with their tools. In addition to the quartzite hand-axe ‘Excalibur’ used in a burial rite some 350,000 years ago in modern-day Atapuerca in Spain, all erectus tools, like all sapiens tools, became symbols of labour, community and culture. The creation, care, transport and skilled use of tools all demonstrate that these tools meant something more than simply the task they were designed to perform. Just as a shovel represents not only the task of digging, but also evokes memories of killing snakes, preparing a camp site and so on, the tools of erectus had many functions and would have elicited memories of cultural values and activities when they were not present.

In other words, erectus tools represented culturally agreed upon meanings that referred to displaced – not immediately present – activities and meanings, the hallmark of symbols. Thus, as erectus invented symbols, activities, tools, sailing, settlement patterns and so on, they were simultaneously inventing culture and transforming mere communication into language. This means that language is no more a part of ‘human nature’ than any other invention. Rather, to the degree that there is anything like human nature, it is seen in the cognitive power and flexibility that supports our ability to innovate. Language is waiting to be invented by any creature with a sufficiently powerful brain, human, non-human or even alien.

From this vantage point, the stakes are high for theories of the origins of language. The main question that arises is whether humans possess special cognitive abilities absent from the brains of all other creatures or whether, more simply, humans have language because they are smarter than other creatures (whether through higher densities of neurons, or other advantages of brain organisation).

All animals depend on indexes to navigate their way through the world, to feed themselves and escape danger

The answer to me is simple – find an animal that can communicate via symbols and you have a linguistic animal. And this is the claim for Homo erectus – it was the first animal to communicate via symbols. Complex grammar is neither necessary nor sufficient for language. DNA replication follows a grammar-like procedure yet no one would say that genes have language. And as I have been claiming for years, some modern languages lack complex syntax. Symbols, not grammar, are thus the sine qua non of language. They alone guarantee communication that is displaced, that is shared by an entire community of speakers, that can be transmitted between speakers and between generations, and that can represent either abstract or concrete ideas or things.

Symbols can do these things because of three properties: arbitrariness, intentionality, and conventionalisation. Arbitrariness means that a word such as ‘apple’ has no necessary connection to the fruit referred to by that name. Another symbol could just as easily have been selected to represent the same fruit, as in the Portuguese word maçã. By ‘intentional’ we mean that a form is purposely directed at the meaning it represents. And by ‘conventionalised’ one means that a symbol’s form is a society’s agreed-upon way of referring to a particular quality, event or thing. Apple means ‘apple’ because we say it does.

Fascinatingly, the archaeological record supports exactly the progression of steps needed to get creatures from information-transfer without symbols to language (information-transfer via symbols). It is a sequence that Charles Sanders Peirce, the inventor of semiotics (the theory of symbols) would have predicted. I therefore call this sequence the ‘Peircean progression’. The steps in order of appearance in the archaeological record would have been (i) indexes (used by all animals); (ii) icons (first observed with Australopithecus africanus); and (iii) symbols (the tools, art and cultural creations of erectus). This progress is represented in the following diagram:

All animals depend on indexes to navigate their way through the world, to feed themselves, to escape danger, to find shelter, and to attend to other biological needs. For example, a lion hunts its prey by the index of smell or perhaps an index of sight (such as a moving tree branch, caused by the passing of another animal). A fox escapes fire by evading the index of smoke, caused by fire. Indexes are thus as old as life.

Next come icons. The first icon in the archaeological record demonstrates an intentional recognition that one object resembles another. This is the 3 million-year-old Makapansgat manuport, from South Africa. This small 7 cm x 8.3 cm pebble looks like a human face and had been in the possession of a small group of Australopithecus africanus. We know that Australopithecus carried this stone to their cave because it is unlike the other minerals found there.

A million years after this first icon, the evidence for symbols in the cultural accomplishments of Homo erectus appears.

In the Peircean progression, the different tools of erectus would have served initially as indexes of the tasks they represented. That is, there would have been a physical link between a spear with blood on the tip and the mastodon it killed. But as culture developed, these tools became more conventionalised more than 1 million years ago, as the archaeology shows, and thus would have gradually taken on the role of symbols. A tool would have represented for erectus a culturally significant task, a representation present even when the tool was not being used. Again, this is perhaps best exemplified by the red quartzite biface found in a possible grave in Atapuerca. As culture developed, tools became symbols, something we otherwise find only among our fellow humans.

Organisation and planning of erectus settlements and travel also demonstrate the requisite cognitive ability for grammar and convention. The aggregate archaeological evidence of erectus accomplishments therefore supports the idea that erectus had achieved language more than 800,000 years ago, assuming that sailing followed language, and more likely 1.5 million years ago (supported by extensive travel and tool standardisation). How big a step would the invention of language have been for erectus?

Symbols would have also emerged gradually from agreed-upon form-meaning correspondences (spoken words have sound forms and tools have visual, olfactory and tactile forms) interpreted by cultural context. These might have been simple exclamations with gestures and intonation, later conventionalised as symbols. For example, suppose that someone shouted ‘Shamalamadingdong!’ upon a close encounter with a saber-toothed cat. Gestures and intonation could then have broken this down into smaller parts through reverse engineering as many modern language games such as ‘Pig Latin’ do. We also see speakers analyse their languages through ‘back formations’, such as the breaking down of the word ‘alcoholic’ into two (faux) parts, ‘alco-’ and ‘holic’ and then producing a new word such as ‘chocoholic’. Speakers analyse their words and sentences with frequency and ease, however their analyses diverge from linguistic theory.

Symbols also likely arose from the conventionalisation of natural sound patterns, such as crying, conventionalised to create ritual wailing in some Ge languages of Brazil.

Once a symbol, such as a word, is invented, a phenomenon known as ‘duality-of-patterning’ has been established (a term introduced in the 1960s by the American linguist Charles Hockett). This is the association of a meaningless item, such as a string of sounds, with a meaning. Take the word ‘cat’. Although the three sounds ‘c’ [k], ‘a’ [æ] and ‘t’ [t] are individually meaningless, combined into the word ‘cat’ they take on a conventionalised meaning. The sequence of meaningless sounds in a word can be recognised qua sequence. Once this is done, we can substitute sounds, such as ‘p’ for ‘c’ to get ‘pat’, or ‘d’ for ‘t’ to get ‘cad’, or ‘augh’ [ᴐ] for [æ] to get ‘caught,’ and so on. This relies upon an understanding or discovery of ‘slots’ (the positions for ‘c’, ‘a’ and ‘t’ in the word and syllable) and ‘fillers’, the individual sounds of the language. From this simple slot-filler principle, in conjunction with meaning and culture, we can build grammars of varying levels of complexity.

Once you have a set of symbols and a linear order agreed upon by a culture, you have a language

Strings of words, phrases and sentences all have slots, the positions for the units that compose them, just as words are fillers for the slots of a sentence. In ‘The goat saw the monkey’ the fillers are nouns (such as ‘goat’ and ‘monkey’), articles (‘the’), and verbs (‘saw’), and the slots are the positions before the verb, the verb and the position after the verb. So just as we can derive ‘pat’ from ‘cat’, in a sense, we can also derive ‘The monkey saw the goat’ or ‘The man saw the boy’ from the slot-filler arrangement of ‘The goat saw the monkey’. Once you have a set of symbols and a linear order agreed upon by a culture, you have a language.

That is really all there is to it, though of course most languages become more complex over time. Likewise, the bow and arrow has many degrees of complexity across societies, yet otherwise seems to be universal (because it is an optimal solution to secure protein that moves faster than you do). In the Peircean progression that I am proposing, these distinctions are expressed by the different G(rammars), G1-G3. G1 is a linear grammar (symbols merely placed in a linear order). G2 has some hierarchy (as in the Reed-Kellogg diagrams of English sentences that some of us learned to produce in elementary school). G3 has both hierarchy and recursion (so one can not only say the hierarchically structured ‘Bill said that May came in the room’ – where the sentence ‘that May came in the room’ is inserted into the larger sentence ‘Bill said …’ – but also recursive sentences such as ‘Bill said that May said that Peter said that John said that …’ where the sentences have in principle no upper bound).

All of the embellishments of grammar such as hierarchical structures, recursion, relative clauses and other complex constructions are secondary, based on a slot-filler arrangement of and composition of symbols, in conjunction with cultural conventions and general principles of efficient computation. Such principles were first discussed in the paper ‘The Architecture of Complexity’ (1962) by the Nobel Prize-winning economist Herbert Simon. However, syntactic complexity is not required to have a language with the same expressive power as any other language. In fact, there are modern languages whose grammars appear to be little more than symbols in an agreed-upon order, lacking evidence for hierarchy or other hallmarks of complex syntax.

Thus, once cultures and symbols appear, grammar is on the way. If this is correct, then language did not begin as singing, as intonation or as gestures. Rather, symbols depended upon gestures, speech and intonation symbiotically and simultaneously – a case also made by David McNeill in How Language Began (2012).

Erectus had relative shortcomings of course, beyond possibly lacking the range of sounds of modern humans. It also lacked the modern form of the important FOXP2 gene that sapiens have. Do the shortcomings of vocal apparatus and primitive genes pose a problem for the idea that erectus had language? Not at all. For example, the evolution of speech was triggered by language – as we developed languages, the modes of expressing them improved over time. Yes, sapiens speech is likely better than erectus speech. But this doesn’t mean that erectus lacked speech. Any mammal could have speech with the sounds they are capable of producing today. They just need the right kind of brain. The sapiens version of FOXP2 helps us to articulate sounds more easily and to think more quickly and efficiently than erectus. But it is not a ‘language gene’. And though erectus might have had, as it were, the ‘Model T’ version of this gene while we possess the ‘Tesla version’, their ‘primitive’ FOXP2 would not have deprived them of language. FOXP2 and other genes adapted partially due to evolutionary pressure from language and culture.

How many sounds does an entity need to keep their speech distinct, after all? Computers use only two ‘sounds’ – 0 and 1. Anything that can be said in any language of the world, no matter how many speech sounds there are in a given language, can be translated into 0s and 1s on a computer – otherwise you couldn’t type your novel on your laptop. The larynx is a red herring. In fact, there are modern languages such as the Papuan language Rotokas that have fewer than a dozen sounds. Erectus was physically capable of at least as many sounds as a gorilla or my laptop’s binary language. In fact, even with the same vocal apparatus, erectus likely could have made many more sounds than gorillas because of its more advanced brain. Chimps don’t talk because they don’t have the brains to support symbols, not because they lack the right vocal apparatus.

Modern English has sentences as simple as ‘You drink. You drive. You go to jail.’ Yet in spite of such grammatical simplicity, we understand these examples just fine. In fact, one can construct similar sentences in any language that will be intelligible to all native speakers of the language. Interpretation requires cultural context, not complex grammar – but this facilitates it, explaining why so many languages have complex grammars, as I explain in Language: The Cultural Tool (2012).

Were Homo erectus incapable of modern language because their tools were so primitive? Bollocks

Such phrases demonstrate that humans can interpret and use a language even when it lacks any obvious grammar other than the ordering of words. The language of Homo erectus might have been no more complicated than these examples, or more so. The English examples underscore the principle that language is underdetermined apart from cultural context. Syntax alone is insufficient.

But what about the many modern paleoanthropologists, linguists and others who do not believe that erectus was capable of modern language because their tools were so primitive? Bollocks. This attribution of inability to erectus is based on a number of errors in reasoning: (i) it focuses almost exclusively on stone tools for erectus, ignoring evidence for bone and wooden tools; (ii) it errs in either assuming that settlements on multiple islands were the result of land bridges or accidents due to wind; (iii) it appears not to consider the significance of erectus village organisation; (iv) no study of erectus speech appears to recognise that speech came later than language and that the human vocal apparatus needs to be able to produce only a small number of sounds to have speech (but see the recent research on macaques led by the evolutionary biologist W Tecumseh Fitch at the University of Vienna); (v) it fails to understand that tools become symbols; (vi) it tends to overestimate the difficulty of having language and fails to realise how slot-filler grammars follow from symbols based on duality-of-patterning.

The conclusion that erectus invented language through their higher intelligence and cultural development is strong, as evidenced by the archaeological record. But if language is merely a technology based on symbols and grammar, other creatures could have also discovered it. If they didn’t, it would be because they lack culture. There are some claims that other animals have language as it is defined here – information-transfer via symbols. It is well-known, after all, that many animals can learn symbols. Some examples are horses, great apes and dogs. What is unclear is whether nonhumans invent symbols in the wild. They would need culture to do so. No strong evidence for this exists.

The available evidence then strongly suggests that erectus invented language more than a million years ago. In so doing, Homo erectus changed the world more than any creature since, including their grandchild, Homo sapiens.