Language and symbolic thought

Foundations

The problem with this lesson

Everything else in this chapter leaves something behind. Bones fossilise. Stone tools survive almost indefinitely. Fire leaves ash and heated sediment.

Language leaves nothing. It is air, and it is gone. There is no fossil of a sentence.

So this lesson is about detective work: what to look for when the thing itself cannot be found, and how to avoid convincing yourself with evidence that does not carry the weight.

What we are actually looking for

Not speech. Speech is a delivery mechanism, and other animals make sounds. What matters is symbolic reference: something standing for something else by agreement rather than by resemblance.

There is nothing about the sound “water” that resembles water. It works only because a group has agreed that it does. That agreement is the whole trick, and it has two consequences that make symbols archaeologically detectable.

First, symbols are social by necessity. A private symbol is useless. To find one in the ground is to find evidence of a shared convention, which means it is evidence about a group rather than an individual.

Second, human language has displacement: it refers to things not present. Last summer. The next valley. A person who died. A thing that has never existed. Vervet monkeys have distinct alarm calls for eagles, leopards and snakes, which is genuinely impressive, and none of them can say “there was a leopard here yesterday”. Almost no animal communication does displacement. All human language does, constantly.

The oldest things that look like symbols

Pigment. Ochre appears at African sites from several hundred thousand years ago, sometimes ground, sometimes in quantity. It is tempting evidence and it is weak on its own, because ochre is also useful: it works as a component of adhesives for hafting stone points, as a hide preservative, and as sun protection. Presence of pigment is not proof of meaning.

Deliberate marks. This is stronger. At Blombos Cave in South Africa, pieces of ochre around 77,000 years old carry engraved cross-hatched patterns, made with a point, in a deliberate sequence. The same site has a pigment-processing workshop dated to about 100,000 years, with abalone shells used as containers and a recipe involving ochre, bone, charcoal and fat. Somebody was making pigment to a method, not just picking up a red rock.

Ornament. This is the strongest of the three, and it is the least visually impressive. Perforated shell beads appear at sites across Africa and the Levant from around a hundred thousand years ago, often from shells with no food value, sometimes carried a long way inland, sometimes worn enough to be polished.

A bead does nothing. It cannot cut, carry or contain. Its entire function is to be looked at and understood, which requires somebody else who knows what it means. Ornament is a message with no content except its own meaningfulness, and it is therefore about as clean a signature of symbolic convention as the ground can give you.

The oldest art is not where you were told

For most of the twentieth century, the story ran through Europe: the painted caves of France and Spain, Chauvet and Lascaux and Altamira, taken as the moment the modern mind switched on.

That story has collapsed, and how it collapsed is worth knowing.

In 2014 a team applied uranium-series dating to the thin mineral crusts that grow over cave paintings in Sulawesi, Indonesia. The paintings had been known for decades and assumed to be a few thousand years old, because everyone knew that Pleistocene art was European. The crusts came back tens of thousands of years old, meaning the paintings beneath them were older still.

It kept going. Figurative painting in Borneo at least 40,000 years old. A hunting scene on Sulawesi at least 43,900 years old, showing small human-like figures with what appear to be spears facing animals. A painting of a Sulawesi warty pig at least 45,500 years old.

The oldest securely dated figurative art in the world is in Indonesia.

Europe looked like the origin because Europe is where the archaeologists were, where the funding was, and where the dating was done first. The bias was not in the past. It was in us.

India belongs in this correction. The rock shelters at Bhimbetka in Madhya Pradesh, a UNESCO World Heritage site since 2003, hold hundreds of painted panels spanning many thousands of years, in continuous relationship with the villages nearby. The oldest paintings there are not securely dated, which is a genuine gap rather than an absence of interest, and it is the sort of gap that new dating work has repeatedly turned over.

Was anyone else doing this?

If symbolic behaviour is what makes us us, then either nobody else had it, or the story needs adjusting. The evidence increasingly says adjust.

Neanderthals used pigment, made perforated and modified objects, collected unusual items, and constructed deliberate structures deep inside caves. Some claimed cave markings in Spain have been dated to before modern humans are thought to have arrived, and those dates are contested. This is an actively moving area, and the honest summary is that the gap between Neanderthals and early modern humans in symbolic behaviour is smaller than it used to look, and not zero.

The genetics, and a cautionary tale

In 2001 a gene was identified in a family with a severe inherited speech disorder. FOXP2 became famous instantly as “the language gene”, and a 2002 paper suggested the human version had been the target of recent strong selection.

Almost every part of the popular version turned out to be too simple. FOXP2 is a regulator that switches other genes on and off, and it is found across vertebrates: mice have it, and songbirds use it in learning song. Its disruption impairs the fine motor control of the mouth and face along with language, so it is not a pure language module. Neanderthals turned out to carry the same protein-coding variant we do. And the evidence for a recent selective sweep was substantially revised in later analyses using larger and more diverse samples.

FOXP2 is real and it matters. It is not a switch, there is no single language gene, and the episode is a good example of how a clean story outruns the data.

And the anatomy says the same

If language required a special vocal tract, we could date it from bones. It does not appear to.

Neanderthal hyoid bones, which anchor the tongue muscles, are essentially identical to ours. Their inner ear bones suggest sensitivity across the frequency range that matters for speech. Their vocal anatomy was not what stopped them.

Which means the question moves away from hardware. Whatever separated modern human communication from what came before, it was not the throat, and lesson 6.2 already showed it was not brain size either.

Against the revolution

The traditional framing is a “human revolution” around 50,000 years ago, when art, ornament, complex tools and burial appear together and the modern mind switches on.

The evidence has not been kind to it. Pigment use is hundreds of thousands of years old. Engraving is at least 77,000. Beads are at least 100,000. Figurative art is 45,000 in Indonesia and later in Europe. Bone tools, long-distance material transport and complex hafting all appear at different times in different places, sometimes appearing and then vanishing from the record for tens of thousands of years.

That last pattern is the interesting one. Innovations show up, and then disappear, and then show up again somewhere else. That does not look like a species acquiring a new capacity. It looks like a species that had the capacity all along and could not always hold on to what it made.

Why that changed is the next lesson.

Deeper

Why ornament is the strongest signal

The archaeological argument for beads as symbolic rests on several features together: the shells are frequently non-dietary species, they are perforated in consistent positions, they show use-wear from stringing and contact with other beads, and they occur far from their source, sometimes hundreds of kilometres inland. Each feature individually admits alternative explanations; together they indicate deliberate selection, modification, wear on the body and transport, which is difficult to account for without shared meaning.

How the Indonesian dates were obtained

Uranium-series dating of coralloid speleothems, the small mineral growths that form over cave surfaces, gives a minimum age for anything beneath them. The method depends on the deposit behaving as a closed system for uranium and thorium, which is checked by dating multiple layers within a single deposit and looking for consistent stratigraphic ordering. Criticisms have focused on open-system behaviour and on whether the dated crust is directly over the pigment. Replication across multiple panels, sites and islands, with internally consistent age ordering, is what moved the results from contested to accepted.

The problem with behavioural modernity

The trait list used to define modern behaviour, typically including blade technology, standardised bone tools, ornament, art, long-distance exchange and burial, was assembled largely from the European Upper Palaeolithic and then applied elsewhere as a test. Applied to Africa, it produces a picture in which traits appear piecemeal across a hundred thousand years and sometimes disappear. Applied to living foragers, several groups would fail it. Recent syntheses treat the list as a description of one regional trajectory rather than a diagnosis of cognition.

What Neanderthal FOXP2 does and does not show

Sequencing established that Neanderthals shared the two human-specific amino-acid substitutions relative to chimpanzees. Regulatory differences outside the coding region have since been reported, so identical protein sequence does not establish identical expression. The broader lesson is that single-gene explanations for complex cognitive traits have consistently failed, and that FOXP2’s real value has been as an entry point into the network of genes involved in speech and motor learning.

Frontier

Dating the Indian rock art

Bhimbetka and comparable Indian sites have no secure absolute dates for their earliest panels, largely because the pigments and settings differ from those in Indonesia and Europe where uranium-series dating on overlying crusts has worked. Developing applicable methods for Indian contexts would fill one of the largest remaining gaps in the global picture of early image-making.

Whether language is old or recent

Two positions coexist. One holds that fully modern language is recent, within the last hundred thousand years, and accounts for the apparent behavioural changes. The other holds that language capacity is far older, present in the common ancestor of modern humans and Neanderthals, and that what changed later was demographic and social rather than cognitive. The second position has gained ground, and neither is settled.

Transcript and shot list

Target runtime about 8 minutes. Built as a bias correction followed by an admission of limits: move the evidence out of Europe, then confront the fact that the thing itself does not fossilise. Shot types: ANIM is rendered from code with Manim, AI is a generated shot from Higgsfield with the shared style reference, REAL is agency imagery with credit, CAM is the narrator.

[00:00] S01 · ANIM · A bone fossilising, a stone tool surviving, ash preserved in sediment; then a spoken word drawn as a waveform that dissipates and leaves nothing.

Everything else in this chapter leaves something behind. Bones fossilise. Stone tools survive almost indefinitely. Fire leaves ash and heated sediment. Language leaves nothing at all. It is air, and it is gone. There is no fossil of a sentence. So this lesson is detective work: what to look for when the thing itself cannot be found, and how not to convince yourself with evidence that will not hold the weight.

[00:35] S02 · ANIM · The word “water” beside water, with no resemblance; then a group agreeing on the convention; then displacement: last summer, the next valley, a person who died.

And what we are looking for is not speech. Speech is a delivery mechanism, and plenty of animals make sounds. What matters is symbolic reference: something standing for something else by agreement rather than resemblance. There is nothing about the sound “water” that resembles water. It works because a group agreed it does. Which has two consequences. Symbols are social by necessity, because a private symbol is useless, so finding one in the ground is evidence about a group. And human language has displacement: it refers to things that are not here. Vervet monkeys have different alarm calls for eagles, leopards and snakes, which is genuinely impressive, and not one of them can say there was a leopard here yesterday.

[01:25] S03 · ANIM · Ochre appearing at African sites; then three practical uses shown: adhesive for hafting, hide preservation, sun protection; the evidence marked as weak on its own.

So, the candidates. Pigment first. Ochre turns up at African sites from several hundred thousand years ago, sometimes ground, sometimes in quantity. Tempting, and weak on its own, because ochre is also useful. It works in adhesives for fixing stone points to shafts, as a hide preservative, and as sun protection. Pigment is not proof of meaning.

[01:55] S04 · REAL · The engraved ochre block from Blombos Cave (Christopher Henshilwood or the Iziko Museums; confirm credit line). S05 · ANIM · The cross-hatched pattern traced stroke by stroke; then the 100,000-year-old workshop: abalone shell containers, ochre, bone, charcoal, fat.

Deliberate marks are stronger. At Blombos Cave in South Africa, pieces of ochre about seventy-seven thousand years old carry engraved cross-hatching, made with a point, in a deliberate order. And the same site has a pigment workshop dated to about a hundred thousand years, with abalone shells used as containers and a recipe involving ochre, bone, charcoal and fat. Somebody was making pigment to a method, not picking up a red rock.

[02:35] S06 · ANIM · Perforated shell beads: non-food species, consistent perforation, stringing wear, transported hundreds of kilometres inland; then the argument: a bead cannot cut, carry or contain.

But the strongest evidence is the least impressive to look at. Perforated shell beads, at sites across Africa and the Levant from at least a hundred thousand years ago. Often shells with no food value. Often carried hundreds of kilometres inland. Often worn smooth from rubbing against each other on a string. And a bead does nothing. It cannot cut, carry or contain. Its entire function is to be seen and understood, which requires somebody else who knows what it means. Ornament is a message whose only content is that it means something. That is about as clean a signature of shared convention as the ground will ever give you.

[03:20] S07 · REAL · The Sulawesi hunting scene panel (Griffith University / Ratno Sardi; confirm credit line). S08 · ANIM · A map: the old story running through France and Spain; then dates appearing over Sulawesi at 43,900 and 45,500, and Borneo at 40,000.

Now the correction. For most of the twentieth century this story ran through Europe. Chauvet, Lascaux, Altamira, taken as the moment the modern mind switched on. That has collapsed. In 2014 a team dated the thin mineral crusts growing over cave paintings in Sulawesi, in Indonesia. The paintings had been known for decades and assumed to be a few thousand years old, because everyone knew Pleistocene art was European. The crusts came back tens of thousands of years old, which means the paintings under them are older still. And it kept going. Borneo, at least forty thousand. A hunting scene on Sulawesi, at least forty-three thousand nine hundred. A painting of a warty pig, at least forty-five thousand five hundred. The oldest securely dated figurative art in the world is in Indonesia.

[04:15] S09 · ANIM · A map of twentieth-century archaeological funding and dating effort concentrated in western Europe; the caption: the bias was not in the past. Then Bhimbetka marked in Madhya Pradesh, with an open question on its earliest dates.

Europe looked like the origin because Europe is where the archaeologists were, where the money was, and where the dating happened first. The bias was not in the past. It was in us. And India belongs in this correction. The rock shelters at Bhimbetka in Madhya Pradesh, a World Heritage site since 2003, hold hundreds of painted panels spanning thousands of years, right beside villages that are still there. The oldest paintings are not securely dated. That is a real gap rather than a lack of interest, and it is exactly the kind of gap that new dating work has repeatedly overturned.

[05:00] S10 · ANIM · 2001: FOXP2 identified in a family with inherited speech disorder; the headline “the language gene”; then corrections appearing one by one: it is a regulator, mice have it, songbirds use it, it affects orofacial motor control, Neanderthals carry the same coding variant, the selective sweep evidence revised.

Then the genetics, which is a cautionary tale. In 2001 a gene was identified in a family with a severe inherited speech disorder. FOXP2 became the language gene overnight, and a paper the following year suggested the human version had been strongly selected recently. Almost every part of the popular version was too simple. FOXP2 is a regulator that switches other genes on and off. Mice have it. Songbirds use it to learn song. Its disruption impairs fine motor control of the mouth and face along with language, so it is not a language module. Neanderthals carry the same protein-coding variant we do. And the evidence for a recent sweep was substantially revised once larger and more diverse samples were analysed. FOXP2 is real and it matters. It is not a switch, and there is no single language gene.

[06:00] S11 · ANIM · A Neanderthal hyoid bone beside a modern one, essentially identical; inner ear sensitivity across speech frequencies; then a callback to brain volume from lesson 6.2.

And the anatomy says the same thing. If language needed a special vocal tract we could date it from bones. Neanderthal hyoid bones, which anchor the tongue muscles, are essentially identical to ours. Their inner ear bones suggest sensitivity across the frequencies that matter for speech. Their vocal anatomy was not the obstacle. Which, combined with the brain sizes from the last lesson, means the question moves away from hardware entirely.

[06:35] S12 · ANIM · The traditional “human revolution” at 50,000 years; then the actual dates scattered across hundreds of thousands of years: pigment, engraving, beads, figurative art, bone tools; several traditions appearing, vanishing, reappearing elsewhere.

Which brings up the traditional framing: a human revolution around fifty thousand years ago, when art, ornament, complex tools and burial appear together and the modern mind switches on. The evidence has not been kind to it. Pigment use is hundreds of thousands of years old. Engraving is at least seventy-seven thousand. Beads at least a hundred thousand. Figurative art forty-five thousand in Indonesia. Bone tools, long-distance transport and complex hafting all appear at different times in different places. And here is the interesting part: some of them appear, then disappear from the record for tens of thousands of years, then turn up again somewhere else.

[07:20] S13 · AI into ANIM · A hand stencil on rock, held; then the same innovation appearing, fading and reappearing on a timeline; the calendar returns.

That does not look like a species acquiring a new capacity. It looks like a species that had the capacity all along and could not reliably hold on to what it made. Which is a demographic problem, not a cognitive one. Next lesson: what changed, and why it had nothing to do with anyone getting smarter.

Shot list

One AI shot, ten code-rendered shots, two research images. Estimated Higgsfield use: 2 to 4 generations including retries.