The Conquest Took Twenty Years; the Conversion Took Four Centuries

In the year 632, in the strip of settled land running from the Nile to the Tigris, almost nobody was Arab, almost nobody spoke Arabic as a first language, and almost nobody was Muslim — Islam itself was, at that point, a matter of months old. Twenty years later, an Arab state ruled every mile of that strip. Three centuries after that, most of the people living in it still were not. The gap between those two sentences is the whole subject of this essay: not whether the conquest happened, which it plainly did, and not whether it was cruel, which by the standards of any seventh-century war it sometimes was, but the fact that three different things — who governs, who prays, and what language a household uses at the dinner table — moved at three wildly different speeds, and that the story we usually tell collapses all three into one dramatic year.

They are kept apart here on purpose. Conquest is a state-change: a garrison arrives, a governor is appointed, a tax register is rewritten, and the transition is complete within a campaigning season or two. Conversion is a diffusion process: it spreads household by household, pulled by incentives that change slowly and unevenly across a huge territory, with no single date on which it is "done." Language change is a third process again, with its own clock, sometimes faster than conversion and sometimes much slower. Treating all three as one event — "the Arab conquest," full stop — is the single most common distortion in how this period gets told, in both the triumphalist and the grievance versions of the story. The sources support neither. What they support is a slow, uneven, mostly non-violent diffusion running for centuries under a fast military and political conquest that took about two decades.

The twenty years

Start with the part that really was fast, because it is not in dispute and does not need softening. Muhammad died in 632. The Ridda wars that followed — the "wars of apostasy" in which the new Medinan state reasserted control over Arabian tribes that considered their allegiance to have ended with the Prophet's death — took about a year and produced, among other things, a battle-hardened Arab military coalition with nowhere left to fight inside Arabia. It turned outward. At the Yarmouk river in 636, an Arab army defeated a much larger Byzantine force and effectively ended Byzantine rule in Syria. In the same rough window, at Qadisiyya, Arab forces broke Sasanian Persian resistance in Iraq, and by 637 they had taken Ctesiphon, the Sasanian capital on the Tigris. Jerusalem surrendered to the caliph Umar in 638, by the later Islamic tradition on terms that let the city's Christian population keep its churches and its bishop. Egypt fell to Amr ibn al-As's forces between 639 and 642. By roughly 700, Arab armies had fought their way across North Africa to the Atlantic; in 711 Tariq ibn Ziyad crossed into Spain; by 751, at the Talas river in Central Asia, an Arab-led army was fighting the Chinese.¹

Hugh Kennedy's account of these campaigns treats them as a genuine military-historical puzzle: a poor, thinly populated peninsula on the edge of two exhausted empires produced armies that overran both a superpower and a half-superpower in less than a generation.² Part of the answer is exhaustion — Byzantium and Persia had just fought each other to a standstill in a war that ended only in 628 — and part of it is that Arab rule, in its first decades, was a lighter hand than what it replaced: garrisons in new cantonment cities, a head tax on non-Muslims, and a fairly hands-off relationship with existing landowners and tax farmers, who often had more to fear from Byzantine or Sasanian collectors than from the new Arab governor.³ Robert Hoyland's survey of near-contemporary evidence — Christian chronicles, papyri, coins, inscriptions — from communities with no reason to flatter the conquerors largely confirms the speed, and confirms that resistance, where it happened, was usually about taxation and garrisons rather than religion.⁴ The Egyptian chronicler John of Nikiu, writing within a couple of generations of the conquest of his own country, describes it as a disaster — burned churches, extracted tribute, a collapsing Byzantine administration his own account blames as much as it blames the invaders — but he is describing a war and a change of ruler, not a change of religion; his Egypt, in his own telling, remains a Christian country under new management.⁵

By any test a political scientist would apply — who commands the garrisons, who appoints the governor, whose name is on the coinage, who collects the tax — the conquest was finished, region by region, within the lifetimes of people who had lived under the previous regime. That is the twenty-year story: fast, decisive, geographically enormous, essentially complete by the early 640s in the heartland and by 711 at the outer edge in the west. Nothing that follows is a quarrel with that timeline.

The four centuries

What did not change on that timeline was who the population prayed to. This is harder to measure than a battle date, because nobody in the seventh, eighth, or ninth century was taking a religious census. The best attempt anyone has made is Richard Bulliet's, and it is worth explaining both what it found and why it is contested, because the finding only means what it means once you understand the method.

Bulliet worked from Muslim biographical dictionaries — vast reference works recording the names, teachers, and lineages of thousands of individuals across generations. His insight was that a family's naming pattern tends to shift a generation or two after conversion: a first Muslim generation still carrying a grandfather's pre-Islamic name, then increasingly Arabic-Islamic names as the family's Muslim identity settles in. By counting that generational depth across thousands of individuals and modeling the curve, he produced rough conversion timelines for several regions.⁶ On his reconstruction, Iran crosses the halfway point — half the population functionally Muslim — around the middle of the ninth century, more than two hundred years after the Arab conquest of Iran was militarily complete. Iraq, Syria, and Egypt cross the same point later still, in roughly the tenth century, and Spain lags on a broadly similar curve. In every region, conversion remains a minority phenomenon a century or more after conquest, and does not become a majority phenomenon until three or four centuries after.

This method is genuinely contested, not just conventionally hedged. It depends on biographical dictionaries compiled by and about the literate, urban, often religiously prominent classes — exactly the population most likely to convert early and visibly. It assumes naming conventions shifted at a fairly uniform lag after actual conversion, which is a modeling choice, not an observed fact. And it says little about rural populations, the overwhelming majority of any medieval society and largely invisible in urban biographical literature. Bulliet himself was explicit that he was building a model from imperfect data, not reading a census; historians have used his curves for decades because nothing better exists, not because the objections have been answered.⁷ Take the specific graph shapes as an estimate with a wide margin; take the general finding — that conversion crossed the halfway mark centuries, not decades, after conquest, at different speeds in different provinces — as about as well supported as anything in this field gets.

Conquest is a state-change; conversion is a diffusion process, and people habitually compress the second into the first when they tell the story.

What sustained non-Muslim majorities for so long was not merely inertia. It was, in significant part, policy working exactly as the early Islamic state needed it to. The jizya, a tax on non-Muslim subjects, and the kharaj, a land tax that also fell disproportionately on non-Muslims, were major sources of state revenue in the seventh and eighth centuries. A wave of mass conversion would have been a fiscal problem, not a windfall, for an early Umayyad treasury that depended on those receipts — no need to invoke cynicism, a state does not typically go out of its way to convert the people who fund it. Early converts, moreover, did not join as equals. Non-Arab converts to Islam under the Umayyads generally entered Muslim society as mawali, clients attached to an Arab tribal patron, and in practice this often meant second-tier status, a grievance that shows up repeatedly in the political unrest of the late Umayyad period and that the Abbasid revolution of 750 drew on for support.⁸ Converting did not, for a long stretch, mean joining a privileged class on equal footing with Arab Muslims — one of the more important, and more frequently flattened, facts about this period.

Christians and Jews under this arrangement were dhimmis, a legal category usually rendered "protected people." Both halves of that phrase were real and neither cancels the other. Protected: their lives, property, and freedom to worship were guaranteed by treaty or by the caliph's law, communities kept their own courts for internal affairs, and — as Sidney Griffith has documented at length — Christian intellectual life did not merely survive under early Islamic rule but in places flourished, most strikingly in the translation movement of the eighth and ninth centuries, in which Christian scholars of the Church of the East, men like Hunayn ibn Ishaq, rendered Greek philosophy and science into Syriac and then Arabic and helped build the intellectual infrastructure of what gets called the Abbasid golden age.⁹ Subordinate: dhimmi status also meant a tax non-Muslims alone paid, and real legal disabilities that varied by time and place — on testimony, on building or repairing places of worship, on holding the highest offices — a permanent second-class position codified in law even where daily relations were easy. Say both; a reader looking for either total tolerance or total oppression will find the sources arguing back.

Egypt kept the mosaics, Iran kept the language

The essay's promised figure belongs here, because it is a small, concrete instance of the larger point: conquest did not erase the conquered culture's craft traditions, it absorbed them. The Umayyad Mosque in Damascus, built in the early eighth century on the site of a church that had itself replaced a Roman temple, is covered in gold-glass mosaics depicting a paradisal landscape of trees, rivers, and buildings — no human or animal figures, respecting the aniconism that had become a marker of Islamic religious art, but executed in a technique and an aesthetic vocabulary directly continuous with Byzantine Christian mosaic work. Tradition and later Islamic sources report that Byzantine craftsmen, or craftsmen trained in that tradition, worked on the mosque; the specifics of who held the tesserae are less certain than the visual evidence, which is unambiguous — Byzantine craft in service of a new religious program, produced within living memory of the conquest, by people who had learned their trade making exactly this kind of image for churches. Two religions, one workshop culture, in the same decade.¹⁰

Language tells a parallel but not identical story, with its own clock. For the first several decades after the conquest, administration in the new provinces continued in whatever language it had used before: Greek in the former Byzantine territories, Pahlavi (Middle Persian) in the former Sasanian ones. That changed under the caliph Abd al-Malik, who around 696 to 700 ordered the administration Arabized — Arabic became the language of the tax registers and the chancery — and reformed the coinage toward a purely epigraphic Arabic design. That is a real, datable event. But an administrative order is not what people spoke at home, and Coptic in Egypt, and Aramaic and its literary form Syriac across Syria and Iraq, persisted as living, spoken languages for a long stretch after Abd al-Malik's reform — Coptic as an everyday spoken language survived, by most reckonings, well into the medieval centuries, with the exact point it stopped being anyone's first language genuinely disputed, and it continues today as the liturgical language of the Coptic Church.¹¹

Here is the asymmetry that is the sharpest tool this essay has to offer. Egypt, Syria, and Iraq eventually did become majority Arabic-speaking — not instantly, but over the medieval centuries the older languages receded into liturgy and pockets, and Arabic became the vernacular. Iran did not. It converted to Islam on a broadly similar multi-century timeline to its Arab-conquered neighbors — Bulliet's own data puts Iran's fifty-percent crossing point earliest among the regions he studied, in fact — but it did not Arabize. Persian survived and then re-emerged in the tenth and eleventh centuries as New Persian, written in a modified Arabic script but grammatically a direct continuation of Middle Persian, and self-consciously reasserted as a literary language in Ferdowsi's Shahnameh, completed around 1010, an epic concerned with pre-Islamic Persian kingship, composed in Persian rather than Arabic as a deliberate act.¹² Two regions, comparably distant from Arab control at the point of conquest, converted to the conquerors' religion on similar timelines; one adopted the conquerors' language as its own, the other did not. That divergence is not explained by conquest at all, since the conquest looked much the same from both vantage points. It has to be explained by something in the diffusion process itself — most likely the depth and prestige of Persian's own prior literary and administrative tradition, though this too is argued rather than settled.

This is one of several places where Fred Donner's account of the earliest period is worth naming, because it complicates the sequence just laid out. Donner has argued that the movement Muhammad founded should be understood, in its first decades, not as "Islam" in the later, doctrinally fixed sense but as an ecumenical "Believers' movement" that included pious Jews and Christians alongside the new Arabian monotheists, with the sharper boundary between Islam and other Abrahamic religions only hardening over the following century.¹³ If Donner is right, "conversion to Islam" in the first Islamic century may be a slightly anachronistic category to project backward — the line a later ninth-century biographical dictionary draws between Muslim and non-Muslim may not have been nearly so bright a line in 650. This is a minority position, not a settled one; it is worth raising because it suggests the three-speed model itself may understate how blurry the middle category was in its earliest decades.

Why it moved slowly

Put the pieces together and a mechanism, not a story of the sword, is what actually explains the four-century timeline. States that need tax revenue from a subject population do not maximize for that population's rapid conversion to the tax-favored category; the incentives point the other way, at least until the fiscal system is redesigned to stop depending so heavily on non-Muslim taxpayers, which happened only gradually. Converts who did switch, for the first century or more, generally entered a lower social tier as mawali rather than an equal one, which blunted the incentive to convert for status reasons in exactly the period when status-seeking might otherwise have driven the fastest conversion. Meanwhile whole professional classes — Coptic scribes running Egyptian tax administration for generations after the conquest, Nestorian physicians and translators at the Abbasid court, Jewish geonic academies in Iraq continuing rabbinic scholarship under caliphal rule rather than in spite of it — had every reason to keep doing exactly what they had been doing, in the language and religion they already had, because the new rulers needed their skills and, for a long stretch, were not especially interested in disturbing the arrangement.⁹

What did eventually push conversion past fifty percent, region by region, over those centuries was not a single mechanism but the accumulation of small, mundane ones: intermarriage, which under Islamic law generally required the non-Muslim partner or the children to be raised Muslim; career advancement, as administration and army increasingly rewarded Muslims specifically once fiscal policy shifted; the social pull of belonging to the dominant community in an increasingly Muslim-majority city; and simple demographic drift, as each generation of a mixed household tilted further toward the majority-forming religion. None of this required forced conversion at swordpoint, and the sources — Christian chronicles as much as Muslim biographical dictionaries — describe a slow social gravity operating across generations, occasionally accelerated by episodes of real persecution in particular times and places, but never as the general mechanism across four centuries and a dozen provinces.⁴ Albert Hourani's synthesis of the whole Arab-Islamic period makes essentially this point: political dominance was established fast, and a much slower social and cultural transformation, uneven from province to province, unfolded underneath it for centuries.¹⁴

Christians and Jews were, at once and for centuries, protected and subordinate — both true, and neither cancels the other.

This region's later history — the sharpest of the several declines that eventually reduced thousand-year-old Middle Eastern Christian communities to a remnant — belongs to a different, later story, taken up separately in The Middle East Became Christian Slowly, Then All at Once. Worth flagging only to say: whatever did that did not happen in the seventh or eighth century, and reading it backward onto the conquest generation gets the chronology wrong by thirteen hundred years of continuous, if subordinate, communal life in between.

Compressing the story

Two comparisons from elsewhere in this project sharpen the pattern. The Arab sieges of Constantinople in the seventh and early eighth centuries failed; the more famous battle that supposedly stopped Arab expansion into Europe, at Tours in 732, is given more explanatory weight than the numbers support — see Constantinople Held, Tours Didn't Need To. And centuries later, a different Muslim conquest of Anatolia produced the opposite linguistic outcome from Egypt's: Anatolia became Turkish-speaking as well as Muslim, a case examined directly in How Anatolia Became Turkish. Set the essays side by side and a pattern in the pattern appears — every one of these transitions separates into a fast political event and a slow social one, and the ratio between the two speeds differs every time, for reasons that have to be argued case by case rather than assumed from the fact of conquest alone.

That is really the whole argument. It is tempting, and flattering to a certain kind of narrative, to tell this history as a single dramatic transformation — armies out of Arabia, and within a lifetime the whole Middle East is Arab and Muslim. It is equally tempting, in the opposite direction, to tell it as forced religious conversion imposed by the sword across the same short span. Both versions are wrong for the same reason: they assume that because the state changed hands in twenty years, the population's language and religion changed hands just as fast. They did not. The garrison arrived in a year; the tax register was rewritten in a generation; the language of the chancery changed by decree in a single reign. The religious and linguistic character of the population underneath all of that took three or four centuries to catch up, moving at a pace set not by armies but by tax incentives, legal status, intermarriage, and the ordinary math of who converts a household at a time. People tell it as one event because one event is a better story. The tax registers, the biographical dictionaries, and the mosaics tell it as three.