Indian numerals reach Baghdad — and become the digits of the world

How counting worked before Indian numerals

The written number systems used by major literate cultures before the spread of the Indian decimal place-value system shared an awkward feature: they were unsuited to calculation. They were good for recording quantities. They were bad for operating on them.

The Roman numeral system, in use across the Mediterranean and most of Europe through the late medieval period, used distinct symbols for one (I), five (V), ten (X), fifty (L), one hundred (C), five hundred (D), and one thousand (M). Other quantities were represented by composition: VII for seven, MCMXLV for 1945. The system has no positional value — the position of a symbol does not affect its quantitative meaning, only its relationship to its neighbors. There is no symbol for zero because the system has no place that needs filling. The system can express any positive integer up to a few thousand without unwieldy strings, and educated Romans could read it fluently. What it could not do is support practical multiplication and division of large numbers. Anyone who has tried to multiply MCCXLVI by CCCLXXXVII without converting to Indian numerals first knows why; the cumulative non-positional structure resists the algorithms of long-form arithmetic that schoolchildren are taught today.¹

Serious Roman calculation was therefore not done in writing. It was done with a counting board (abacus) — a flat surface marked with horizontal lines or with grooves, on which pebbles or counters were moved to represent quantities and operations. The abacus was a positional device even when the written system was not. A practiced Roman accountant could perform substantial calculations on the abacus and write down only the final result in Roman numerals. The wealth-management of the Roman empire, the construction of the great public works, the censuses of imperial taxation — all of these were conducted with abacus arithmetic and notarial records.²

The Greek number system was, if anything, worse. Greek used the alphabet's letters as numerals: alpha for 1, beta for 2, gamma for 3, and so on through theta for 9, then iota for 10, kappa for 20, etc., with additional archaic letters (digamma, koppa, sampi) brought in to fill the gaps. Like Roman numerals, Greek alphabetic numerals had no place value and no zero. Calculation again required external aids. The mathematical Greeks who developed Euclidean geometry and the early Hellenistic astronomers worked through their proofs in geometric form precisely because the arithmetic available to them was not adequate to support algebraic manipulation. Diophantus of Alexandria's Arithmetica (third century CE) is sometimes called the first algebra, but its symbolism is rudimentary and its operations are presented in prose rather than in calculation steps.

The Mesopotamian sexagesimal (base-60) system, used in cuneiform mathematical and astronomical texts, did have place value. It is the system the modern world inherited for dividing the hour into sixty minutes and the minute into sixty seconds, the circle into 360 degrees, the day into twenty-four hours. The Babylonian astronomers of the late first millennium BCE and the early first millennium CE achieved astronomical precision unmatched in the Greek tradition because their place-value system supported the long calculations the work required. But the system had no clean symbol for zero in early periods (the Babylonians used a positional gap; later they introduced a placeholder), and it never spread beyond cuneiform-using bureaucratic cultures.³

Chinese numerals had a hybrid character — the everyday written numerals were non-positional, like Roman, but the rod numerals used by mathematicians and astronomers (a system in which counting rods of different orientations represented digits in different positions) were essentially positional. By the Han period, Chinese mathematicians had a working place-value tradition for calculation purposes, though it was carried in physical rod manipulation rather than in written notation. The Chinese tradition produced sophisticated mathematics — the Nine Chapters on the Mathematical Art of the early imperial period contains algorithms for solving systems of linear equations — but the written notation did not generalize to a universal place-value system.⁴

Indian mathematicians were operating in a different environment. By the late Gupta and early post-Gupta period (fifth through seventh centuries CE), they were using a written decimal place-value system with nine digit-glyphs (1 through 9) and — critically — a written symbol for zero that functioned as a placeholder. The earliest unambiguous dated written zero is on a temple inscription at Gwalior in central India in 876 CE, but the system itself is documented in mathematical works at least two centuries earlier and probably had been in scholarly use for several centuries before that.⁵

What Brahmagupta did

The Indian mathematician Brahmagupta, working at Bhillamala (modern Bhinmal in Rajasthan) in 628 CE, produced the Brāhmasphuṭasiddhānta ("Correctly Established Doctrine of Brahma"), a comprehensive treatise on mathematics and astronomy. The work was not the first Indian text to use the place-value system — Aryabhata's Aryabhatiya of 499 CE had used a related notation, and earlier Sulbasutras and grammatical literature had presupposed positional arithmetic. What Brahmagupta did was systematic. He gave explicit arithmetic rules for operations on positive numbers, on negative numbers, and on zero. He treated zero not just as a placeholder but as a number — a quantity that could be added to, subtracted from, multiplied with, and (with the famous exception that division by zero produces an undefined result) divided into other numbers.⁶

Brahmagupta's rules included: the sum of two positives is positive; the sum of two negatives is negative; the sum of a positive and a negative is their difference; the product of two like signs is positive, of two unlike signs is negative; zero subtracted from any number gives that number; any number divided by zero produces (Brahmagupta's word) kha-cheda — "divided by void" — which he treated as an unbounded quantity rather than as a clean undefined result. The treatment of division by zero would not be fully settled in mathematical thought for another thousand years, but the framework Brahmagupta had laid out was the framework all subsequent work would build on.

Brahmagupta also gave systematic procedures for solving linear and quadratic equations, computed the volumes and areas of geometric figures with substantial precision, and presented a sophisticated astronomical model that calculated planetary positions, eclipses, and the lengths of the year and the lunar month. The astronomical model was, in some respects, ahead of contemporary Greek astronomy in its handling of planetary motion. The treatise also extends earlier Indian work on Diophantine equations, on algebraic identities, and on what would later be called combinatorics.

The Indian mathematical tradition that produced the Brāhmasphuṭasiddhānta was not a single-author tradition. It built on earlier work — Aryabhata's, Bhāskara I's (a near-contemporary of Brahmagupta), the Jain mathematical school's. It would be extended by Bhāskara II in the twelfth century, whose Līlāvatī and Bījagaṇita are read in India today as foundational mathematical texts. The Brāhmasphuṭasiddhānta is the work that traveled west, but it traveled as the visible tip of a much larger and older tradition.

The transmission to Baghdad

The path from Indian scholarly mathematics to the Abbasid court at Baghdad ran through diplomatic and scholarly contacts that the older histories sometimes underemphasize. The Indian subcontinent and the Iranian world had been in sustained scholarly contact for centuries before any Arab caliph existed; Sasanian Persia, the empire that fell to the Arab conquests in the mid-seventh century, had hosted Indian astronomers and translated Indian astronomical works into Pahlavi (Middle Persian) under Khusraw Anushirwan (r. 531–579) at his court at Ctesiphon. Some of those Pahlavi versions of Indian astronomical works survived the Sasanian collapse and were translated into Arabic in the eighth and ninth centuries.

The most precisely dated tradition is the embassy of c. 770 CE. The bibliographer Ibn al-Nadīm, writing his catalog al-Fihrist in 988, records that an Indian scholarly embassy arrived at the court of caliph al-Manṣūr (r. 754–775) bringing Sanskrit astronomical texts. Al-Manṣūr ordered the texts translated into Arabic. The principal translator, named in Arabic sources as Yaʿqūb ibn Ṭāriq and Muḥammad ibn Ibrāhīm al-Fazārī, produced the Sindhind — an Arabic version that combined Brahmagupta's Brāhmasphuṭasiddhānta with related Indian astronomical works.⁷ Through the Sindhind, the Indian decimal numerals, the zero, and the rules for operations on them entered Arabic-language scholarship.

The House of Wisdom

The institutional setting in which the transmission was completed was the Bayt al-Ḥikma — the "House of Wisdom" — in Baghdad. The institution had existed in some form since al-Manṣūr's reign as a library and translation bureau, but it was greatly expanded under al-Maʾmūn (r. 813–833), who made it the center of one of the largest sustained translation enterprises the world had yet seen. Greek philosophical, mathematical, medical, and scientific works were being rendered into Arabic from Greek originals or from earlier Syriac translations; Sanskrit astronomical and mathematical works were being rendered from Sanskrit originals or earlier Pahlavi versions; Persian historical and literary works were being absorbed into the new Arabic learned tradition. The institutional commitment was deliberate: al-Maʾmūn and his successors were funding scholarship at unprecedented scale partly to compete with the Byzantine intellectual establishment and partly because they understood mastery of the inherited learning of the previous Mediterranean and Indic civilizations as a marker of caliphal legitimacy.

Muḥammad ibn Mūsā al-Khwārizmī (c. 780 – c. 850) was the figure at the House of Wisdom whose work made the Indian numerals canonical for the Arabic-reading world. Al-Khwārizmī's surname suggests origin in Khwārizm, a Central Asian region around the Aral Sea (his name in Latin sources became Algoritmi). He worked at the court of al-Maʾmūn and produced two foundational mathematical works.

The first, Kitāb al-Jabr wa-al-Muqābala ("The Book of Restoration and Balancing"), c. 825 CE, presented a systematic theory of equations — solving linear and quadratic equations through standard procedures, classifying the equation types into a small number of canonical forms, demonstrating each procedure with worked numerical examples and geometric proofs. The work gave English the word algebra (from al-jabr, the operation of moving a subtracted term to the other side of an equation). It was translated into Latin in the twelfth century — first by Robert of Chester at Toledo in 1145, then by Gerard of Cremona soon after — and entered the Latin curriculum at the new universities founding in the same period.⁸

The second, Kitāb al-Jamʿ wa-al-Tafrīq bi-Ḥisāb al-Hind ("Book on Addition and Subtraction by the Method of the Indians"), gave a clear Arabic exposition of the Indian decimal place-value arithmetic. The Arabic original is lost; the work survives only in twelfth-century Latin translations under the title Algoritmi de numero Indorum — "Al-Khwārizmī on the numbers of the Indians." From the Latin opening phrase the European Middle Ages took the word algorism, later algorithm, for any systematic calculation procedure. Every modern use of the word algorithm — in computer science, in advertising, in critique of platform capitalism — descends from a Latin transliteration of one Khwarizmian Arabic mathematician's name.⁹

The House of Wisdom's enterprise produced, within a century of al-Khwārizmī's work, a flourishing Arabic mathematical tradition. Al-Battānī (c. 858–929) refined trigonometric astronomy. Abū al-Wafāʾ al-Būzjānī (940–998) developed the trigonometric identities that Western trigonometry would use for the next millennium. Al-Bīrūnī (973–1048) wrote a comprehensive comparative study of Indian astronomy (Taḥqīq mā li-l-Hind) that remains a foundational document in the historiography of cross-cultural mathematical contact. Omar Khayyām (1048–1131) extended Khwarizmian algebra to the systematic solution of cubic equations. The cumulative achievement of Arabic mathematics across the ninth through twelfth centuries was, in absolute terms, the most productive period of mathematical work the Mediterranean world had seen since Hellenistic Alexandria — and it rested on the integration of Greek, Indian, and Persian sources that the House of Wisdom had carried out.

The arrival in Latin Europe

Latin Christian Europe was, in the ninth and tenth centuries, mathematically in worse shape than the Roman world had been a thousand years earlier. The classical Greek mathematical tradition had been substantially lost in the West; what remained was the late-Roman Boethius's textbooks, an abridged Euclid in Latin translation, and the practical arithmetic associated with the late Roman agrimensores (land surveyors). Place-value arithmetic was unknown. The abacus and Roman numerals were the operational tools of monastic accountants and royal exchequers.

The transmission of the Indian-Arabic numerals to Latin Europe ran through three channels. The first and most consequential was the Toledo translation school, working at and after the Christian conquest of Toledo in 1085. Toledo had been a center of Andalusi learning under Muslim rule; after the conquest, the city retained its multilingual scholarly community (Arabic-speaking Christians, Mozarabs; Arabic-speaking Jews; Arabic-speaking Muslims who remained for several generations) and became the principal site at which Arabic mathematical, philosophical, and scientific texts were rendered into Latin. The translators included Gerard of Cremona, who spent forty years at Toledo and produced Latin versions of more than seventy Arabic works including al-Khwārizmī's Kitāb al-Jabr, Ptolemy's Almagest (which the Arabs had preserved), and major Aristotelian and medical texts. The Latin scholarly culture of the early universities — Bologna founded c. 1088, Paris c. 1150, Oxford c. 1167 — was substantially built on the Toledo translations.¹⁰

The second channel was Sicily, which had been under Arab rule from the ninth century until the Norman conquest of 1091. Sicilian translators worked from Greek and Arabic sources for the Norman and later Hohenstaufen courts at Palermo. The third was the commercial channel — Italian merchants conducting business with North African and Levantine ports learned Arab arithmetic operationally and brought it home. Leonardo of Pisa, known as Fibonacci (c. 1170 – c. 1240), traveled with his merchant father to Bugia (modern Béjaïa, Algeria), studied Arab arithmetic with Muslim teachers, and on return to Italy wrote the Liber Abaci of 1202 — an introduction to Hindu-Arabic numerals and their use in commercial calculation that became, slowly, the standard text on the new arithmetic for European merchants. The system displaced Roman numerals in Italian commercial bookkeeping over the thirteenth and fourteenth centuries, in academic mathematics over the fourteenth and fifteenth, and in everyday literate life across Europe by the seventeenth.¹¹

What it replaced — and what it changed

The Hindu-Arabic numerals replaced, in their European reception, the Roman numeral system that had dominated medieval European literacy for over a millennium. They also replaced — slowly, in some sectors instantly, in others over centuries — the abacus as the principal calculation tool. With written place-value notation, the calculation could be done on paper instead of with counters; the result could be checked, the procedure could be taught, the methodology could be written into textbooks and examined.

What changed because of the new arithmetic is large. Italian double-entry bookkeeping, which emerged in the thirteenth and fourteenth centuries, is essentially impossible without place-value arithmetic — the underlying operation of credits and debits requires fluent addition and subtraction of long lists of figures of varying magnitudes, which abacus arithmetic could do but written Roman-numeral arithmetic could not. The growth of European commercial cities and the financial instruments that accompanied them — bills of exchange, partnerships, joint-stock arrangements — depended on accountants who could keep coherent books in the new arithmetic. The early modern European commercial revolution rested on the new numerals.¹²

The scientific revolution of the sixteenth and seventeenth centuries was even more dependent. Galileo, Kepler, Newton, and the systematizers of European mathematical natural philosophy worked in a notation that allowed algebraic manipulation, calculation of long decimal numbers, and the systematic representation of equations. The shift from Galen's qualitative Aristotelian medicine to the quantitative experimental tradition of seventeenth-century natural philosophy ran in part on the new arithmetic. Newton's Principia of 1687 is unimaginable in Roman numerals — not as a graphical impossibility but as an intellectual one; the symbolic apparatus that supports the Principia's reasoning depends on calculation procedures that the older notation could not sustain.

What the system replaced in the Indian and Islamic worlds is a subtler question. India retained its decimal notation; the system was already present and the reception in those territories was not a replacement but a continuity. The Islamic world had used multiple older systems before adopting the Indian numerals — including the abjad alphabetical numeral system, in which Arabic letters served as numbers — and the new system displaced these in scientific writing while abjad continued in use for some traditional applications. The shift from older to newer notation in the Islamic world was largely complete by the late Abbasid period.

What the cost was

This is the cleanest single transmission documented in the Hidden Threads atlas. The intellectual core — Brahmagupta's notation traveling to al-Khwārizmī's Baghdad to Fibonacci's Italy to the modern world — carried no extraction, no plague, no conquest at the moment of any transmission step. The Indian astronomers who came to al-Manṣūr's court did so as honored emissaries; the Arabic translators who rendered their texts did so as scholars under caliphal patronage; the Latin translators who later carried the system into Europe did so through commercial and academic exchange. The system itself benefited from the journey — Brahmagupta's rules were extended by al-Khwārizmī and his successors; the European adopters added accounting innovations the system supported.

But the contexts at the receiving ends were not all clean.

The House of Wisdom in Baghdad was destroyed in February 1258 when Hulagu Khan's Mongol army sacked the city. The destruction of Baghdad killed an estimated 90,000 to 200,000 inhabitants (medieval Arabic chroniclers give figures up to a million; modern estimates are lower) and burned the libraries that had housed the cumulative manuscript record of the previous four centuries of Arabic scholarship. The Tigris is supposed to have run black with ink from the destroyed manuscripts (a colorful detail of unverifiable accuracy). Whatever the precise figure, the institutional infrastructure that had carried the Indian numerals into Arabic disappeared in a week of sustained siege violence. The Mongol destruction of Baghdad is a permanent casualty of cultural transmission to compare against the gift of the numerals: an institution that had absorbed Greek, Indian, and Persian learning, synthesized them, and produced original work for four centuries, was wiped out in February 1258.¹³

The Latin European reception ran through territories the Christian powers were taking from Muslims and from Jews. The Toledo translations occurred at a city the Christian Castilian forces had taken in 1085; the Sicilian translations occurred at a court that had inherited rule from a recently displaced Arab population. The Iberian Reconquista that ended at Granada in 1492 was followed within months by the expulsion of the Iberian Jewish population (perhaps 200,000 displaced) and the forced conversion of the Iberian Muslim population that remained, with a final expulsion of the Moriscos in 1609 (perhaps 300,000 displaced). The Iberian Reconquista is documented separately in this atlas as the entanglement-era transmission "Aristotle returns to Europe through Toledo" (forthcoming), where the cost framing is treated in detail. The point here is that the transmission of the numerals from al-Andalus to Christian Europe rode atop the same political process that destroyed al-Andalus.

What the world inherited is the cleanest mathematical tool ever devised. What the world also inherited, on the parallel arcs, is the destroyed library at Baghdad and the destroyed cultural ecology of al-Andalus. The numerals on the page of this article — 1, 2, 3, 4, 5, 6, 7, 8, 9, 0 — are an Indic gift, an Arabic refinement, and a Latin transmission. The contexts of all three steps were good and bad in the way real history is good and bad. The system itself is unambiguously a gift. What was on the boats and across the borders alongside it was not all gift.