Appearance
3.11 — Writing, Counting and the Invention of Memory
The first writing was not poetry, prayer or law. It was a receipt.
The earliest tablets from Uruk in southern Mesopotamia, around 3300 BCE, record quantities of barley, beer, sheep, cloth and labour, with the name of an official. The oldest readable human documents are inventory and payroll.
That is not a disappointing origin. It tells you exactly what writing was for: a state that could not remember everything it was owed. Chapter 3.9 described the surplus and the officials who managed it. This chapter is about the technology that made managing it possible, and about what else that technology turned out to do.
From tokens to tablets, one step at a time
The sequence in Mesopotamia is unusually well preserved, and it is a good example of a technology nobody invented.
Step one: tokens, from around 8000 BCE. Small clay shapes — cones, spheres, discs, cylinders — each standing for a quantity of a commodity. A cone might be a small measure of grain, a sphere a large measure. These appear across the Near East for five thousand years before writing, and they are found in the same settlements as the first farming.
Step two: sealed containers, around 3500 BCE. To send a consignment with a shipment, the tokens were sealed inside a hollow clay ball so nobody could alter the count in transit. But then you cannot see the contents without breaking the seal. So the shapes of the enclosed tokens were pressed into the outside of the ball before it dried.
Step three: the realisation. If the marks on the outside record the contents, the tokens inside are redundant. Flatten the ball and you have a tablet with impressed signs, which is writing.
Step four: from pictures to sounds. Early signs were pictorial — a stylised head of grain for barley, an ox head for cattle. Pictures work for objects and fail for names, verbs and abstractions. The solution, found independently in several writing systems, is the rebus principle: use a sign for its sound rather than its meaning. In English this would be drawing an eye to write the word "I", or a bee and a leaf to write "belief". Once a sign can stand for a sound, any word in the language can be written, and writing stops being accounting and becomes language.
The script that resulted is cuneiform — from the Latin for wedge-shaped, because the reed stylus pressed into clay makes wedges rather than lines. It was used for over three thousand years, for at least a dozen unrelated languages including Sumerian, Akkadian, Hittite and Old Persian, in the same way the Latin alphabet is used for hundreds of languages today.
Independent inventions
Writing was invented from scratch at least three times, probably four.
Mesopotamia, around 3300 BCE, as above.
China, by around 1200 BCE with the oracle bone inscriptions of the Shang dynasty, and probably earlier in forms that did not survive. Chinese writing has been in continuous use ever since, which makes it the longest continuously used writing system in the world, and a literate Chinese reader today can recognise many oracle-bone characters.
Mesoamerica, by around 500–300 BCE, with the Maya script fully developed by the early centuries CE. Complete independence is certain here — there was no contact of any kind.
Egypt, around 3200 BCE, appearing rather suddenly in nearly complete form. Whether it was independent or stimulated by knowledge that Mesopotamians were doing something is argued; the scripts themselves share nothing, so any influence was the idea rather than the system.
Stimulus diffusion — knowing that writing is possible, then inventing your own — accounts for many other cases, and the clearest modern example is the Cherokee syllabary devised by Sequoyah in the 1820s, who could not read English but understood that marks could carry speech, and built a complete system for Cherokee alone.
The Indus script, and why we are honest about it
The Indus civilisation (Chapter 4.3) produced thousands of short inscriptions on seals and pottery from around 2600 to 1900 BCE. It has not been deciphered, and the reasons are specific rather than mysterious.
The inscriptions are extremely short — typically five signs, rarely more than fourteen. There is no bilingual text of the kind that unlocked Egyptian hieroglyphs through the Rosetta Stone. And the underlying language is unknown, with Dravidian, Indo-Aryan and other candidates all argued, largely along lines that track modern political preferences.
There is even a serious argument that it is not writing at all but a system of non-linguistic symbols — trade marks, clan emblems, or religious signs — and while most specialists disagree, the argument is not frivolous.
Anyone claiming a decipherment should be asked three questions: does it produce sensible readings across a large body of inscriptions rather than a handful; does it work on texts the proposer did not choose; and does it predict anything checkable. No proposed decipherment has passed. This book will say so plainly rather than adopting one, and Chapter 4.3 explains how much this costs us — an entire literate civilisation of five million people whose own words we cannot read.
The alphabet, and why it mattered
Cuneiform had many hundreds of signs. Egyptian hieroglyphs had over seven hundred. Chinese has tens of thousands of characters, of which a few thousand are needed for literacy. Learning any of these took years, which meant scribes were a professional class and literacy was a privilege.
The alphabet changed the arithmetic. Around 1800 BCE, in the Sinai peninsula, workers — probably Semitic-speaking labourers in Egyptian turquoise mines — took a small number of Egyptian signs and used them for the first sound of the Semitic word for the object depicted. A house is bayt, so the house sign now means "b". An ox is aleph, so the ox head means "a".
The result needed about twenty-two to thirty signs to write anything. That is learnable in weeks rather than years.
The Phoenicians spread it, because they were the Mediterranean's traders and it suited merchants who needed to keep their own records. From Phoenician came Greek — which added vowels, since Phoenician wrote consonants only and Greek needs vowels to be readable — and from Greek came Etruscan, then Latin, and from Latin the letters on this page. Aramaic, from the same root, gave rise to Hebrew, Arabic, and the Brahmi script.
Brahmi is where the Indian side runs, appearing clearly in Ashoka's inscriptions of the third century BCE (Chapter 6.4). Its own origin is argued — most scholars derive it from Aramaic, some argue for indigenous development. From Brahmi descend Devanagari, Bengali, Tamil, Telugu, Kannada, Malayalam, Sinhala, Tibetan, Thai, Burmese, Khmer and Javanese. A very large fraction of the world's scripts descend from one Indian ancestor.
And the Indian grammatical tradition deserves its own line. Panini's Ashtadhyayi, from roughly the fifth or fourth century BCE, describes Sanskrit in about four thousand compressed rules, using a formal metalanguage with ordered rule application and something very close to what a modern computer scientist would call a generative grammar. It is the most sophisticated linguistic analysis produced anywhere before the twentieth century, and it was constructed to preserve exact pronunciation of oral texts — which is the point of the next section.
Memory without writing
Oral cultures are not simply cultures waiting for writing. They developed techniques for exact transmission that are genuinely astonishing, and India is the world's best example.
The Vedas were transmitted orally for well over a thousand years before being written, and with a fidelity that is verifiable. The method was deliberate redundancy: in addition to the plain recitation, students learned the text in pada form (word by word, isolated), krama (each word paired with the next), jata (pairs recited forwards, backwards, forwards), and ghana (a still more elaborate permutation). Each layer is a checksum on the others. A word cannot drift without breaking several patterns at once.
The result is that the Rigveda has come down with its pronunciation and accent preserved to a degree that comparative linguistics can verify against related Iranian texts. This is one of the most impressive feats of information preservation in human history, and it was done entirely in heads.
Other techniques appear everywhere. Metre and rhyme constrain what words can go where, so an error breaks the rhythm audibly. Formulaic phrases — "swift-footed Achilles", "the wine-dark sea" — give a reciter building blocks. Homer's epics were composed and transmitted this way, as Milman Parry demonstrated in the 1930s by studying living oral poets in Yugoslavia who could perform epics of comparable length. Aboriginal Australian songlines encode routes across hundreds of kilometres in sung sequences, and some have been argued to preserve information about coastlines that were drowned by post-glacial sea level rise thousands of years ago.
So what did writing add? Not accuracy — the Vedic method is more accurate than most manuscript traditions. What writing added was scale, durability without living carriers, and the ability to check. An oral tradition requires an unbroken chain of trained people; break the chain and it is gone. A text survives neglect. And crucially, a written claim can be compared with another written claim by someone who was not present at either, which is the precondition for law, science and history as this volume understands them.
Counting and number
Counting is older than writing. Tally marks on bone go back tens of thousands of years — the Ishango bone from central Africa, around 20,000 years old, carries grouped notches whose interpretation is argued but whose deliberate grouping is not.
Number bases came from bodies. Base 10 from ten fingers, base 20 from fingers and toes — which survives in French quatre-vingts for eighty and in Maya numerals. Base 60 in Mesopotamia is thought to come from counting the three joints of each of four fingers with the thumb, giving twelve, then using the other hand's five fingers to count twelves. Sixty has an unusual number of divisors, which makes fractions easy, and it survives in your clock and in the 360 degrees of a circle.
Positional notation is the deep invention, and it is worth being precise about what it means. In Roman numerals, the symbol X always means ten. In positional notation, the same digit means different amounts depending on where it sits: in 333, the three digits mean three hundreds, three tens and three units.
And positional notation requires a symbol for nothing, because you must be able to distinguish 305 from 35. This is where zero comes in, and it is an Indian achievement. Zero as a placeholder existed in Mesopotamia and among the Maya. Zero as a number in its own right — something you can add, subtract and multiply with, with stated rules — appears in India. Brahmagupta, in 628 CE, gives the arithmetic of zero and of negative numbers explicitly, including that a number multiplied by zero is zero and that debts and fortunes are negative and positive quantities.
The Bakhshali manuscript, containing a dot used as zero, has been radiocarbon dated with parts as early as the third or fourth century CE, though the manuscript is composite and the dating has been questioned. The system travelled to the Islamic world, where al-Khwarizmi wrote on it in the ninth century, and to Europe through Latin translations, which is why the digits are called Arabic numerals in English and Hindse — Indian — in Arabic. Volume II Part 11 covers the mathematics.
Where this shows up in your life
Every institution in this book runs on records. Property depends on a register. Money is a ledger. Law is a text. Science is published claims that others can check. Debt is a written promise. Chapter 9.1 shows that money and writing appeared together in the same offices in the same cities, and that is not a coincidence — they are two applications of one technology.
Literacy is still the sharpest dividing line in development. India's literacy rate at independence in 1947 was around 12 percent. It is now over 77 percent, and the gap between states — Kerala above 95 percent, several northern states well below the average — predicts income, health and child mortality better than almost any other single number. Chapter 9.14 uses this hard.
And the alphabet's arithmetic still applies. A writing system learnable in weeks rather than years determines who can participate. The same argument recurs with printing in Chapter 8.2, with mass schooling in Chapter 9.5, and with the internet.
What the next page covers
Stone can be sharpened but not shaped; wood can be shaped but not made hard. Metal can be both, and the arrival of metal reorganised warfare, trade, agriculture and social hierarchy in that order. Chapter 3.12 closes this Part with copper, bronze and iron — how each was smelted and why each was harder to make than the last, why bronze created the first genuinely international trade system, why iron was the democratic metal, and how the material a society could make decided who it could defeat.