1. A number still belonged to something

Three people have a flock of sheep to divide.

They may not need to write down a single numeral. Drive the animals into three groups, move a few from one group to another, and keep going until everyone is satisfied. Perhaps nobody can say afterward exactly how many sheep they took. That afternoon, at least, nobody objects.

Now change the arrangement. Part of this year's flock is owed to an institution, and the amount depends on last year's shortfall. The animals are scattered among several villages. The person collecting them is somewhere else. Some must be converted into an allowance of grain at a rate set by a third party. Three roughly equal groups will no longer do. Somebody needs a record that another person, in another place and another month, can read in the same way.

That is the sort of work the earliest surviving tablets were doing.

To see what made their notation unfamiliar, imagine two records from Uruk around 3300 BCE. This is an illustrative comparison, not a claim about two identifiable museum objects. One records livestock; the other records a quantity of barley. We might expect to recognize the same number wherever it appears. The scribe could not simply make that assumption.

Late-fourth-millennium scribes used several numerical systems. Discrete objects, including people and animals, belonged to one; grain capacity to another; land area and other measurements had their own conventions. Certain products and rations required further distinctions. These were not merely different fonts for a single universal notation.

A numerical sign could therefore take its value from the kind of quantity being recorded. Copy it unchanged from one system into another and you might no longer have the same amount. Before reading the number, you had to establish what was being counted or measured.

This does not mean conversion was impossible. Relations between units could be learned and used. It means the notation did not carry everything a later reader would need inside itself. The commodity, the unit, and the activity still helped determine the reading.

A tablet could consequently lose some of its intelligibility when it left the setting that produced it. The original writer knew what everyone in the room knew. There was no need to state it all. An archaeologist confronting an unprovenanced object may be trying to recover precisely what the scribe had been able to leave unsaid.

It would be a mistake to call this simply primitive. The systems grew close to the things they served. Capacity units were related to containers, area units to ways of surveying land. Counting a shepherd's animals and receiving grain into a storehouse were different activities. They had accumulated different working conventions.

Specialization could also be convenient. A storekeeper who spent his life measuring grain did not always need notation suited to counting livestock. A local practice could leave out what it did not use. The difficulty became more visible when records from different activities had to be brought together.

Cities made that happen more often. Storehouses, temples, work crews, and continuous deliveries created obligations that could not all remain within separate little worlds. Who owed what had to be compared. Much later, an abstract calculating layer would help connect quantities expressed in different units. That development took more than a thousand years; it did not arrive with the first numerical tablet.

The oldest records already had a clear practical destination. So much received. So much still outstanding. So much issued to someone else.

They were not being written for us. They were being written so that somebody could check next month.

2. Most of the earliest writing was an account

The proto-cuneiform corpus is usually discussed in terms of roughly six thousand tablets and fragments, about five thousand from Uruk itself, with others from sites such as Jemdet Nasr. In the conventional chronology, these materials span approximately 3350–3000 BCE. Such totals describe a scholarly corpus, not a complete inventory of everything once written.

Hans Nissen's familiar classification put economic and administrative documents at about 85 percent and non-administrative material, principally lexical lists, at about 15 percent. The lists became more prominent in the later phases. In the earlier material, administrative recording is more overwhelmingly dominant still.

So, among the earliest written objects we actually possess, accounts are not a narrow majority. They constitute most of the evidence.

These were not account books in the modern sense. A record might combine a quantity sign, a commodity or personnel sign, and a mark for an institution or individual. The script did not yet freely reproduce all the possibilities of spoken language. Quantities and categories were often more legible than anything we would call a sentence. It was particularly good at saying how much of something had gone to someone.

Even a possible personal name reaches us through this environment. A pair of signs conventionally read as Kushim recurs on eighteen Uruk tablets associated with barley and brewing materials. One records grain handled over thirty-seven months. Whether Kushim names a person, an office, or an institution remains uncertain. A common reading makes it a person. If that reading is right, one of the earliest recoverable human names belongs to someone on a barley account.

But the percentage needs a brake attached to it. Eighty-five percent of the surviving corpus is not the same as 85 percent of everything people originally wanted writing for.

Clay preserved certain activities exceptionally well. A discarded tablet could dry out; a building fire could harden it. Other practices relied on wood, fiber, spoken instructions, temporary marks, or wet clay rubbed smooth and used again. They did not all have the same chance of entering an archive, much less a modern collection.

Administrative demand was central to the environment in which writing took shape. That is a strong historical claim. Accounting was the sole reason anyone invented writing. That is a different claim, and the evidence does not carry it.

Nissen, Peter Damerow, and Robert Englund connected early writing with the increasingly complex management of resources by urban institutions. Their account identifies a setting and a set of pressures. A neat origin story supplies a single intention. No inventor's statement from 3400 BCE survives to settle that difference.

The small clay objects usually called tokens pose a related problem. Cones, balls, disks, hollow clay envelopes, and impressions on their surfaces are real archaeological finds. Denise Schmandt-Besserat proposed a long developmental sequence: simple tokens stood for quantities of goods, more elaborate tokens represented a widening inventory, tokens were enclosed in clay envelopes, impressions recorded them outside, and eventually two-dimensional marks replaced the three-dimensional objects. Writing emerged from that process.

There is real material beneath the proposal. Some envelopes really did contain tokens. Yet resemblance in shape does not, by itself, establish continuity of use and meaning across thousands of years. Many objects lack a secure accounting context. Their shapes do not map onto the earliest written signs as consistently as a strong version of the theory requires. They may have served several purposes.

A more limited account is easier to defend: some tokens, sealings, and numerical impressions belong to the history of accounting technologies. A unified prehistoric token system, stable in meaning over millennia and capable of yielding the signs of writing one by one, is not an established consensus.

The difference matters because origins are unusually hospitable to tidy stories. A sequence of objects can look like a sequence of intentions. Those are not necessarily the same sequence.

3. A person becomes a workday

By the Third Dynasty of Ur, the documentary world is much larger. Known administrative texts from this period number on the order of 120,000.

One revealing form is the balanced account. Start with what should have been delivered or accomplished during a period of responsibility. Subtract authorized expenditures, losses, and completed work. Establish the remainder as a surplus or deficit. Add dates, the responsible person's name, and whatever the accounting procedure requires.

The decisive operation is conversion. Labor becomes workdays. Agricultural tasks, textile production, grain, and wool can be related through prescribed rates of production. At places such as Umma, groups were organized under foremen or supervisors. The administration wanted more than a head count. It wanted to know how much labor this group was supposed to deliver.

The person at the center of the account was therefore often the supervisor.

Once labor had been converted into an accounting unit, something else became harder to see. One worker was ill today. Another worked more slowly. Another had been given the heaviest part of the job. A total expressed in workdays did not necessarily deny these differences; it did not necessarily have a place for them. Subsequent arithmetic could operate on the figure without encountering the people again.

A group of records from years 37–40 of Shulgi's reign follows hundreds of female workers and their labor and grain obligations under Lu-Urub, son of Minar-ni. Whatever the uncertainties of reconstructing the institution, the supervisor was not merely someone taking attendance each morning. His name could function as a continuing point of administrative responsibility.

A deficit might carry forward into the next period. A surplus might be credited. The next balanced account could absorb the previous one. The name and the number could remain attached across successive years.

Supporting records lay behind the total: receipts, sealed documents, and smaller acknowledgments retained for checking. Ur III accounting even left evidence of containers used to keep these supporting tablets. A balanced account was not necessarily just one tablet. It could be one tablet resting on a basketful of others.

Here an attractive claim sometimes outruns the records. We are told that, when a worker died, an unfinished labor quota passed to the family. There are records of deaths and replacements. There are provisions for dependants and orphans. There are supervisors' deficits carried forward. But placing those facts next to one another does not produce a tablet stating that a named dead worker's outstanding workdays became a named relative's inherited debt.

What the evidence securely shows is the persistence of a deficit in an account. The further claim about the hereditary transfer of an individual labor obligation needs direct evidence of its own.

It is easy to miss the step because the extrapolation seems so plausible. A system capable of reducing living labor to a number might also be capable of passing the number to a son. But what an arrangement could do is not identical to what we know it did.

The supervisor's responsibility needs a similar qualification. An administrative account in his name does not automatically establish personal, unlimited liability in a later legal sense. Was Lu-Urub expected to cover every deficit from his private property? The documents do not give us enough to say that. They tell us whose son he was. They do not provide an inventory of his home.

Even without the stronger claims, the change is considerable. An administration in southern Mesopotamia could take hundreds of people, a year of their effort, and the grain they consumed, convert them into comparable quantities, and place the result beneath a responsible person's name.

The calculation became easier to carry. The lives it represented did not become equally easy to carry with it.

4. The layer where the calculation happens

Another development becomes visible toward the end of the third millennium, particularly in Ur III practice. Scribes increasingly converted concrete measurements into abstract sexagesimal place-value numbers, calculated there, and converted the answer back into the appropriate units.

Two developments should not be collapsed. Counting systems in which sixty mattered already existed at Uruk. A place-value notation in which the position of a sign could give it different powers of sixty became clearly established much later. It entered practical calculation in the late third millennium and reached a mature form in Old Babylonian schools.

Tables made that abstract layer usable.

A standard reciprocal table did not list every integer from one to fifty-nine. It concentrated on numbers whose reciprocals terminate in base sixty, often in some twenty or thirty entries. Other divisors required approximation, decomposition, or another procedure. Multiplication tables supplied products for a chosen number. Metrological tables connected customary units of length, area, capacity, and weight with the numerical system.

A central practical move was to replace division with multiplication by a reciprocal. Find or remember the reciprocal, then multiply. The tables were not merely helpful appendices to a technique that already existed independently of them. They were part of the technique.

For a student, knowledge and memory were closely joined. Learn the table and a class of divisions becomes possible. Fail to learn it and the written numbers sit there waiting. Repeated work on clay was one way to put the procedure into a person's hands and head.

The notation still left something outside the signs. Early place-value writing ordinarily lacked a fixed radix point and the later forms of an internal zero placeholder. The same written sequence might need to be read at scales differing by sixty, by 3,600, or by a fraction of those amounts. Units and context supplied the scale.

That absence should not instantly become a story about constant mistakes. Identifying a particular ancient error as the necessary result of a missing zero, rather than a misplaced sign or another copying problem, would require evidence. The limitation of the notation is visible; a particular failure cannot simply be inferred from it.

The figures handled relative numerical relations. A trained reader supplied the order of magnitude and the connection to grain, silver, or a daily ration. This was not an entirely self-sufficient writing system. It worked with a person beside it.

Nor did standardization mean that a king abolished all customary units overnight. Capacity, weight, length, and area continued to have distinct units. What became more unified was the layer at which calculation took place. Conversion tables were the passage between that layer and the things being measured.

The phrase Shulgi's reforms is convenient for the wider standardization of administration, taxation, labor, and measurement during his long reign. It should not make us imagine a surviving modern-style metrology act, with a royal decree defining every unit relation line by line. Much is reconstructed from the documentary practices that became consistent.

Shulgi's hymns offer something different. In one, the king presents himself as having studied scribal learning, Sumerian and Akkadian, calculation, addition, subtraction, and accounting. Other praise poetry likewise celebrates his learning. Yet the surviving copies are chiefly Old Babylonian school manuscripts, written roughly two or three centuries after his lifetime.

They may preserve earlier court traditions. They cannot function as an independently verified school record. Their secure subject is the image of learned kingship and the value attached to it.

Still, there is something worth noticing even at that level. A king could be praised for knowing how to keep accounts.

5. What survived in one house

At Nippur, a building known by its excavation label, House F, yielded roughly 1,425 tablets and fragments. More than nine-tenths were school material. Unlike so many mathematical tablets bought without a recorded provenance, these objects could be related to actual rooms and archaeological deposits.

The tablets themselves show something of the work. On one common type, a teacher's model occupies a portion of the front and a student's copy sits beside it. The reverse may preserve longer passages or earlier lessons. Small round tablets can place a short model beside its repetition. Student areas were often wiped and reused.

A cheap teaching surface was an expensive loss for the historian. A child could smooth away a mistake and start again. What the child first wrote, how it went wrong, and how many attempts preceded the surviving copy may all have vanished in that gesture.

The objects also differed in how a person could hold them. Yale's round tablet YBC 7289, about eight centimeters across, fits in a palm. British Museum tablet BM 13901 measures about 19.4 by 11.7 centimeters: a more substantial surface for the hands or a table. Wet clay could be corrected; dry clay was less cooperative. Crowded signs, erasures, and recutting preserve some of that negotiation with the material.

Scholars including Steve Tinney, Eleanor Robson, and Niek Veldhuis have reconstructed sequences of instruction from tablet types, combinations of texts, and directions of copying. Making wedges and individual signs led into syllables, personal names, and organized word lists. More complex lexical work, metrology, numerical tables, and literary copying followed or overlapped.

This is a reconstructed curriculum, not an ancient timetable recovered intact. A pupil might repeat, skip, or pursue several things at once. Nor can one house stand without qualification for every school in Babylonia. Robson has been particularly careful about that limit.

School literature brings us closer to a pupil's day, but in a different way. In the composition usually called Schooldays, a student goes to school with a tablet, recites, eats, prepares more work, and returns home to repeat the day's lesson to his father.

Next morning, the teacher finds something missing. The child is beaten.

More beatings follow. The official in charge of the work strikes him; the reason is lost in a damaged passage. Another objects to his talking. Someone else complains about the position of his head. A drawing supervisor asks why he has left his place. The gatekeeper asks why he has gone outside. Another complaint concerns an object whose name is missing. The Sumerian-language supervisor catches him speaking Akkadian. His own teacher dislikes his handwriting. The sequence accumulates punishments until the day seems to contain little else.

That evening, the father invites the teacher home, gives him food, beer, and gifts, and the teacher eventually praises the boy.

The scene almost makes checking it feel unnecessary. That is precisely when the distinction matters: we are reading school literature, not a disciplinary log. The composition belongs to the Old Babylonian world and may draw upon recognizable practices. Its comic or satirical accumulation does not prove that every school operated this way every day.

The tablets independently support a less dramatic picture. Teachers supplied models. Pupils copied them. Work was checked and corrected. Advanced pupils copied literary compositions, including texts about being pupils.

At Ur, another house, given the modern excavation address No. 7 Quiet Street, yielded more than two thousand documents and exercises associated with Ku-Ningal and his family. Calling it simply the home of a hereditary family of professional scribes smooths away important distinctions. Ku-Ningal and several sons had priestly roles. Teaching within the household is well supported; reducing the family's occupations to scribal employment is not. The seal of his son Ur-Nanna gives him religious and archival responsibilities.

There was strong family continuity without evidence of an entirely closed legal caste. Literacy, collections, contacts, and offices could accumulate in a household and give its children an obvious advantage. That does not mean every title was hereditary in exactly the same way.

Nor is there a single salary scale for the profession. Grain, wool, land income, rations, and the privileges of office might all contribute. The prestige of scribal learning cannot establish the prosperity of every minor copyist.

It is tempting to end with an encouraging story about gifted poor children finding a ladder into a different life. The surviving student records do not allow that story to become a general fact. They disproportionately reveal families already connected to administration and temples. The chances of an ordinary farming family's child entering advanced scribal education cannot be calculated from them.

We can see the lesson. We cannot see everyone who was missing from the room.

6. Problems that lead to no payment

BM 13901 contains four columns and about twenty-four problems involving sides and areas of squares and rectangles. Its acquisition did not preserve a secure archaeological findspot.

Read the problems as a sequence and three features stand out. They are arranged by difficulty. Operations return in altered forms. The chosen numbers are exceptionally accommodating.

That arrangement is already a kind of invention. A grain account does not naturally generate a second exercise slightly harder than the first. Somebody has to identify where a learner may stumble, then devise problems that expose and vary that step. The object in the author's mind is no longer only a field. It is a difficulty in another person's understanding.

Such problems commonly wear the clothes of administration: fields, bricks, canals, workers, silver, grain. Those were familiar objects in the scribal world. Their appearance on a tablet does not prove that the problem records a real transaction. Dimensions can be chosen to make a procedure work cleanly rather than to describe a particular plot of land.

YBC 7289 offers another example. On its face are a square, its diagonals, and numerical inscriptions, including a remarkably accurate value for the relation between a square's side and diagonal. Its reverse has been partly erased. A corresponding constant appears in other tables. The student may have been applying a known value, not deriving it on the spot.

The tablet demonstrates sophisticated handling of this numerical relation. It does not by itself demonstrate a Greek-style system of propositions and deduction. Attaching a later theorem's name can import an entire organization of knowledge along with a correct mathematical observation.

Plimpton 322 is more contested. The surviving clay is roughly thirteen centimeters wide and nine high, a few centimeters thick, with a damaged edge and a marked surface. Four columns and fifteen complete rows remain; material may be missing on the left.

George Plimpton acquired it from an antiquities dealer around 1922–23. The dealer associated it with Senkereh, ancient Larsa. Without a recorded excavation, that attribution cannot become a secure findspot. The tablet subsequently entered Columbia University's collection.

In 1945, Otto Neugebauer and Abraham Sachs identified numerical relations corresponding to the sides of right triangles. Those relations can be checked. The dispute concerns the purpose of the table.

Calling it the world's first trigonometric table gives it an immediately familiar identity. Robson's reassessment in 2001–02 challenged that identity. Old Babylonian evidence does not supply the later conceptual apparatus of angular trigonometric tables. Reciprocal pairs, school tables, problem construction, and the tablet's format have closer parallels within its own documentary setting. She read it as connected with a teacher's preparation of problems. A strong claim for exact sexagesimal trigonometry was advanced again in 2017.

The right-triangle relations are not in doubt. The historical label remains a matter of interpretation. A defensible account can explain reciprocal pairs and educational uses without pretending that the debate has a single uncontested verdict.

Something similar happens when historians translate verbal procedures into algebra. The tablets do not present symbolic equations in our notation. Akkadian instructions describe lengths, widths, adding, removing, holding together, and completing areas. Neugebauer and François Thureau-Dangin's major decipherments made the underlying mathematical relations strikingly visible in modern algebraic form.

That success carried a risk: a modern equation on the printed page could be mistaken for the object the ancient scribe had consciously manipulated. Jens Høyrup's work on the language instead emphasized operations on measured lines and areas—cutting, joining, and completing shapes. Different modern formalisms can reproduce the result. Understanding what the scribe understood requires attention to the words.

Return to the graded problems. They do not end in a settlement. The same scribal world was making two sorts of record: one would go to a storehouse or an office; another would go into the hands of a learner.

They shared material, vocabulary, and notation. Their destinations differed. One followed deliveries and harvests. The other could follow a sequence of difficulties. Once that second rhythm existed, it did not need a new administrative order to justify every next question. A bad harvest did not make it pointless for the next exercise to be harder than the last.

There is no reliable percentage for how much of the corpus was purely practical or purely intellectual. A reciprocal table could be both schoolwork and preparation for future accounting. A fictitious field problem could train a real professional skill. Counts of known mathematical cuneiform texts—often put at roughly two thousand, with a substantial share still awaiting full publication—do not sort their purposes for us.

An apparatus developed for keeping track of things had begun producing work that was not itself keeping track of any consignment at all.

In the same buildings. With the same hands.

7. On the other river

The imbalance in surviving evidence is striking. Surveys often count only about twenty-five specialized Egyptian mathematical documents, containing around a hundred problems, most concentrated in two papyri. Compare that with roughly two thousand mathematical cuneiform texts.

Preservation is an unavoidable part of the comparison. Clay can survive drying and sometimes benefit from fire. Papyrus, leather, and wood decay in damp deposits. Much of the Nile Valley and Delta—the very places where agricultural administration flourished—offered poor conditions for their survival. Twenty-five against two thousand is not a fair league table of mathematical activity. Among other things, it compares soils.

The best-known Egyptian document is the Rhind Mathematical Papyrus. Its two major sections in the British Museum are about 296 and 198.5 centimeters long, both roughly thirty-two centimeters high. Fragments are also held in Brooklyn. Even the two large sections together approach five meters; the original roll was longer. Imagine not a book to leaf through but a narrow sheet extending beyond the width of an ordinary living-room wall.

At its beginning, the scribe Ahmose gives his name and dates the copying to year 33 of King Apepi. He also says he worked from an older text associated with the reign of Nimaatre, generally identified as Amenemhat III.

That earlier copy no longer exists. Ahmose's statement, made roughly three centuries later, is our evidence for it. We cannot check whether he copied every line, selected material, updated numbers, or combined sources. The tradition he reports is valuable. It is not an invitation to invent the missing manuscript.

Copying was not necessarily a secondary act. It kept something in existence. It also gave the copyist an opportunity to determine what would travel onward. Ahmose left a name and a date, more than most people who held a pen in his world.

One Egyptian text explains why a child might be encouraged to join that profession. A father takes his son on a journey and disparages occupation after occupation. Carpenters exhaust themselves. Brickworkers are covered in dirt. Laundry workers labor near crocodiles. Fishermen, arrow-makers, messengers: each trade is made to look miserable. The conclusion returns insistently. Be a scribe.

The composition belongs to a Middle Kingdom tradition, but the complete copies and many school manuscripts that survive are later. Generations of pupils copied an argument for becoming the sort of person who could copy it. Whether read as satire or professional self-promotion, it is not an occupational income survey.

A later school document, Papyrus Lansing, makes the farmer's position particularly bleak. A scribe arrives to assess the crop, accompanied by armed attendants. Grain is demanded; the farmer has none. Beating and restraint follow, extending to his family. The repeated advice is, again, to become a scribe.

This is an instructional text, not the record of a particular collection visit. Yet the person arriving to demand grain belongs to the profession the text recommends. That reversal should not disappear when we admire literacy.

Returning to the Rhind mathematical roll, its modern history also contains gaps. Rhind bought the principal material at Luxor in 1858, reportedly from the area near the Ramesseum, without a documented excavation. The British Museum acquired the main sections in the mid-1860s. Edwin Smith bought further fragments, later recognized as joining the roll; his daughter donated his collection in 1906.

The other mathematical objects have their own unusual shapes and histories. The Moscow Mathematical Papyrus is roughly five meters long but only four to eight centimeters high: more a strip than a broad sheet. Its twenty-five problems include the volume of a truncated pyramid. Vladimir Golenishchev acquired it on the antiquities market in 1893, again without a secure archaeological findspot.

The Egyptian Mathematical Leather Roll, about twenty-five by forty-three centimeters, preserves repeated decompositions into unit fractions. It entered the British Museum with Rhind's material but remained rolled up for decades. Dry leather had become so brittle that opening it risked destroying it. Treatment finally allowed it to be unrolled in 1927.

By contrast, the mathematical material from Lahun belongs to excavations of a Middle Kingdom settlement. It can be placed within an archaeological context rather than only a dealer's account. Mathematics forms a small part of the papyri recovered there.

Egyptian methods ordinarily expressed a fractional quantity as a sum of distinct unit fractions, with some special conventions. The Rhind roll opens with a table decomposing a class of fractions into workable components. It supports division, doubling, rationing, and the distribution of quantities that cannot be handled merely by repeated halving. Problems about bread, beer, grain, and labor draw upon the same repertoire.

A smaller related table from Lahun helps establish that the Rhind table belongs to a tradition rather than appearing from nowhere.

These methods fitted certain operations naturally. Other calculations required longer expressions and trained choices among decompositions. Calling the system clumsy measures it against our notation. Corrections and auxiliary tables demonstrate a demand for skill; they do not tell us that the scribes themselves experienced their system as inefficient.

Egyptian school mathematics also produced problems that no real account was likely to require. Rhind problem 79 presents a cascade of sevens: houses, cats, mice, ears of grain, measures of grain. Each level multiplies the next. The question concerns their total.

No granary official needed to find exactly seven cats in each of seven houses. The familiar nouns carry an exercise in powers and summation. It has the extravagant pleasure of a riddle. Someone turned an escalating numerical pattern into work for a student.

Other problems, notably 56–60, concern the seked: a way of expressing a slope through horizontal displacement in palms and finger-widths for a vertical rise of one cubit. We can describe the relation with later mathematical language, but the text does not use our angles or trigonometric functions.

For Old Kingdom construction, Merer's logbook from Wadi el-Jarf supplies a different kind of first-hand evidence. It belongs late in Khufu's reign, to the year after the thirteenth cattle count, conventionally converted to approximately regnal year 26 or 27. Merer supervised a group of roughly forty boatmen within a larger crew. The log records journeys carrying limestone from Tura to the works at Giza.

Merer was an inspector, not a mathematician in the later professional sense. He had people and stone to deliver on a schedule. His records show the coordination that an enormous building project required. Quarry traces, crew marks, and workers' settlements provide further evidence of organized construction.

What does not survive is a Khufu-period worksheet proving that its designers followed a particular procedure from the Rhind roll, copied nearly a millennium later. We can reconstruct a pyramid's slope in seked units and notice convenient ratios. Inferring some equivalent method of slope control is reasonable. Identifying the exact written procedure remains an inference.

Capacity measures offer another example. The hekat could be divided repeatedly, and surviving problems use those fractional units. Later writers connected that sequence with a beautiful myth about a damaged eye. The beauty deserves a closer look.

For now, Ahmose's name remains at the head of the roll. He identifies himself through the act of copying. The modern label mathematician comes from our side of the encounter.

8. The words that went missing

The most familiar story about geometry's beginnings comes from Herodotus.

An Egyptian king, he says, divided land into equal plots and collected an annual rent. When the Nile took part of a plot away, surveyors measured the loss and the rent was adjusted. Herodotus then offered his own conclusion: he thought geometry had originated in this way and later reached Greece.

There are two distances to keep in view.

One is chronological. Herodotus wrote in the fifth century BCE. If the royal traditions behind his story chiefly reflect the Middle Kingdom, they concern a world roughly thirteen to fifteen centuries earlier. Even a looser reading leaves him much later than the mathematical tradition represented by the Rhind roll.

The other distance is inside the sentence. He marks the origin claim as his inference. Remove that qualification and a traveler's explanation becomes an ancient archival fact.

The unqualified version is easier to remember. It travels further.

There is abundant evidence that Egyptians surveyed land: area problems, land records, measuring practices, and taxation documents. What we do not possess is an Egyptian document announcing that geometry was invented because floods erased the boundaries.

The story acquired accessories. A rope divided by twelve knots, a three-four-five triangle, the recovery of a right angle: these details are not in Herodotus's account. Their familiar use in stories about Egyptian geometry belongs to much more recent histories.

As the picture becomes more detailed, the details can become younger.

The fractions of the Eye of Horus offer a further example. In the attractive modern account, Seth tears the eye apart. Each recovered piece corresponds to a successive fraction: one-half, one-quarter, one-eighth, down to one-sixty-fourth. The pieces nearly complete the eye; Thoth supplies what is missing. Myth, mathematics, damage, and restoration seem to fit perfectly.

Jim Ritter traced the modern construction of that correspondence. The relevant capacity-fraction signs in early Middle Kingdom material are hieratic signs. No surviving ancient text identifies the six marks as six parts of the eye. Strong visual correspondences were developed in modern Egyptology and became familiar through sign lists and histories of mathematics, especially in the early twentieth century.

The myth of the injured and restored eye is genuinely ancient. The explicit pairing of six parts with those six fractional signs is another matter. A compelling synthesis can be thousands of years younger than the objects it seems to explain.

These stories have an additional advantage over a sober account of the archive: they do not need an identifiable person. There are too few surviving names to give an inventor a biography, so the invention is attributed to Egypt, or to a river.

The nilometer invites a comparable shortcut. Egyptians recorded inundation levels from early periods. Columns, stairs, and wells could serve as measuring installations, with cubits marked on them. Too little water threatened irrigation; too much could damage crops. It is reasonable to connect knowledge of water levels with estimates of agricultural resources and the state's capacity to collect revenue.

But that broad connection is not a surviving Pharaonic tax table that automatically converts so many cubits of water into a particular tax rate. Detailed evidence connecting water measurement and taxation is richer for Ptolemaic, Roman, and especially Islamic periods. The famous nilometer at Roda dates to 861 CE. It cannot be borrowed unchanged for a much earlier world.

Returning from origin stories to surviving records brings their unevenness into focus. Estimates of known cuneiform objects range from several hundred thousand toward a million; by 2016, more than 300,000 had been catalogued. Only a small part is mathematical. Ur III administrative documents alone vastly outnumber the specialized mathematical corpus, while Egyptian mathematical texts survive in only a few dozen objects. The exact totals depend on the date and the counting rules. The asymmetries are unmistakable.

They produce several kinds of blindness.

The cheapest and most frequent exercises were often the easiest to erase. The Nile floodplain is much less audible than dry settlements and desert burials. Sophisticated written work is more visible than an ordinary person's mental arithmetic, knotted records, estimate by container, or spoken account of a debt. Repetition and memorization mattered in schools, but no calculation gives us the exact share of teaching that happened orally.

There is also the question of who held the stylus. It is not correct that all scribes were men. Evidence for female scribes reaches back into the Akkadian and Ur III periods and becomes clearer in Old Babylonian material. Scholars have identified school texts signed by women, ranging from elementary exercises to literary work. Women in particular Old Babylonian communities also managed property and documents; Amat-Mamu is a well-known example.

At the same time, many women and enslaved or dependent people appear in the accounts as people being counted. Textile workers, children, and laborers become rations and workdays under someone else's administrative responsibility.

Both observations belong in the picture. One cannot cancel the other.

An archive is produced by people with access to recording. Its silence about everyone else is a feature of that production, not a census of intellectual ability. The absence of mathematical tablets signed by ordinary farm workers cannot prove that farm workers were unable to calculate.

Who first taught children outside scribal households to count? How much arithmetic remained entirely unwritten? What complete design procedures did early builders use? How did oral accounting work before writing at Uruk? We do not know enough to reconstruct these worlds in full.

The surviving tablets are the ones someone kept, or the ones a fire happened to harden. The exercises smoothed away for tomorrow's lesson have nothing more to tell us.

The child rubbing the clay flat did not know that he was removing evidence.

The reckoning continues. What counts as mathematics is still not finally settled.