Three-quarters of a million hadith: the sum that counts channels
The number that paralysed the sceptics
In the previous article we saw how the Muwatta’ of Imam Malik ﵀ was copied from Morocco to Khurasan as the first publicly circulating hadith compilation. In the generations after him everything accelerated: more teachers, more students, more books, and above all more numbers. From the third century of the Hijra figures have come down to us that take the breath away from any modern reader. Al-Bukhari ﵀ is said to have chosen his collection out of six hundred thousand transmissions. Ahmad ibn Hanbal ﵀ spoke of millions. And the total stock of a few thousand ahadith in the first generation had, by the middle of the third century, swollen to roughly three-quarters of a million.
For the contemporaries themselves these numbers were not a problem but a given, the familiar measure of their own discipline. The problem arose only twelve centuries later, when Western readers encountered the figures without knowing the science that had produced them, and applied to them the arithmetic that this article dissects. Few misunderstandings in the study of Islam have been so small in their origin and so large in their consequences: a confusion of the unit of measure, nothing more, and yet it still carries to this day a considerable share of the popular distrust of the hadith.
Around these numbers an entire sceptical edifice has been raised. The reasoning sounds, on first hearing, irrefutable. If in the days of al-Bukhari ﵀ six hundred thousand transmissions were in circulation, and his Sahih, after strict selection, contains fewer than three thousand unique ones, then roughly everything that circulated must have been false or suspect: fewer than one in two hundred passed the test. The British orientalist Alfred Guillaume drew precisely that conclusion, with precisely that ratio, and he was not alone in it; comparable readings are found, as the footnote enumerating the whole school makes clear, in Muir, Robson, Nicholson, Gibb and others besides. The misunderstanding was, in other words, no slip of a single author, but the shared reading of an entire generation of Western readers, who took the six hundred thousand for a count of texts because the unit of measure of the discipline was unknown to them. In this picture the early hadith world becomes an ocean of forgery in which a single little island of authenticity floats, and the rigour of the collectors is, ironically, forged into evidence against the tradition.
Mohammad Mustafa al-Azami ﵀ devoted a separate appendix of his Studies in Early Hadith Literature to these numbers, and his answer is not rhetoric but a lesson in arithmetic. Modern scholars, he writes, were unacquainted with the nature of this material, were bewildered by its extent, and so arrived at very strange conclusions. Anyone who does know the nature of the material sees that the dizzying numbers count something other than the unwary reader supposes. They do not count texts. They count paths.
Before we follow that arithmetic lesson, it is worth seeing how much hangs upon its outcome. The bewilderment over numbers is, after all, no marginal ornament of the sceptical school; it is one of its foundations. Whoever once believes that the early centuries produced six hundred thousand predominantly false transmissions has thereby the backdrop against which every further theory sounds plausible: a community that fabricates on such a scale will fabricate chains too, and a corpus fished out of such an ocean deserves distrust until proven otherwise. The reverse holds equally. If the great numbers turn out to measure something other than texts, then not only the one-in-two-hundred ratio evaporates but the backdrop as well; what remains is then no ocean of lies in which the canonical books would be little islands, but a documented network of roads around a far smaller and far more stable body of texts. The question of what the counter counted is therefore no philological trifle. It decides the entire picture one forms of the early hadith world.
Azami himself notes in this connection that al-Bukhari ﵀ was not the only one about whom such enormous numbers are transmitted; of other transmission scholars too, gigantic figures of memorisation and writing are recorded, and he chooses the case of al-Bukhari ﵀ merely as the best known through which to clarify the problem. What is demonstrated for this one figure therefore holds for the whole family of numbers from which the sceptical literature plucked its denominator.
What the counter counted
To understand the sum one must separate two questions that the sceptical reading silently merges: what did the early scholars call a hadith, and when did they count two of them?
The first question seems simple, but it was not. For some transmission scholars the word hadith embraced the sayings, deeds and tacit approvals of the Prophet ﷺ; in the definitions of other scholars it extended also to the sayings and legal decisions of the Companions ﵃ and the Successors. Whoever counts with the broad definition therefore counts not only the Prophetic word but the whole juridical-historical memory of three generations, and his totals lie correspondingly higher without a single transmission having been added to the Prophetic corpus. This is no theoretical possibility but a documented practice, one the reader of this map has met before: in the census of the writing Companions ﵃ we saw that Abdullah ibn Amr ﵁ collected, alongside the words of the Prophet ﷺ, also the legal decisions of Umar ﵁, and that the books of the early scholars were filled with the rulings of Companions ﵃ on inheritance law, marriage law and governance. All that material travelled in the same books, through the same chains, and fell for the broad counters under the same word. Whoever reads the six hundred thousand as six hundred thousand sayings of the Prophet ﷺ therefore reads a wrong content into it to begin with, before the actual counting error has even come up.
The second question is the decisive one, and Azami formulates the answer in a single sentence: every channel of transmission is counted as a separate hadith. When the same saying of the Prophet ﷺ reached a scholar by way of thirteen different chains, then that scholar said he possessed thirteen ahadith. Not out of carelessness, but because for his science each chain was a separate piece of evidence, with its own series of witnesses who each had to be weighed individually. The example Azami cites comes from the mouth of one of the greatest critics of the second century, Abd ar-Rahman ibn Mahdi ﵀: I have thirteen transmissions from al-Mughirah ﵁ on the wiping over the leather socks in the wudu. It concerns, strictly speaking, one action, one practice of the Prophet ﷺ, reported by one Companion ﵁; however often the action was repeated, as a report it is one. But because the report reached Ibn Mahdi ﵀ along thirteen routes, his accounting counted thirteen.
Once one sees this, one sees it everywhere in the literature. Ibn Khuzaymah ﵀ gives, in a single chapter of his Sahih, about thirty isnads for one hadith about a single action of Aisha ﵂, the cleaning of a garment. Muslim ﵀ adduces, for the one incident in which the young Ibn Abbas ﵁ placed himself to the left of the Prophet ﷺ in the night prayer and was drawn by him to the right, thirteen isnads, thirteen hadith therefore in the count, and moreover uses that multiplicity for precisely what it is meant for: one of the thirteen routes, that of Yazid ibn Abu Ziyad, says, mirror-fashion, that the boy stood on the right and was drawn to the left, and by laying the twelve other routes beside it the error of this one transmitter can be pinpointed exactly. And Azami adds an observation that only fully reveals the scale of the phenomenon: Muslim ﵀ documents the ramifications of his chains mainly up to about the year one hundred and thirty; had he carried the full genealogy through into his own generation, there would have been at least fifty of them.
The tahajjud example deserves a further moment of attention, because it shows that the multiplicity of channels was no dead weight but a working instrument. Twelve routes relate that the boy stood on the left and was drawn to the right; one route, running through Yazid ibn Abu Ziyad on the authority of Kuraib, relates the mirror image. Without the twelve other routes that single version would not have been recognisable as an error; there would simply have existed a transmission with a reversed detail, and no one could have determined which detail was the original. With the twelve routes beside it the diagnosis is exact: one transmitter in one chain has confused left and right, and the science can name his name. This is what the bookkeeping of channels served for. Each additional route was an additional witness in a permanently running cross-examination, and the more witnesses, the smaller the chance that the error of one of them went unnoticed. The numbers the sceptic read as a symptom of chaos are in reality the measure of the control.
By the third century this multiplication had become a consciously pursued ideal. There were transmission scholars who claimed to possess every hadith through as many as a hundred channels, and many wrote out every hadith from twenty or thirty channels. One must not read this as a collecting mania for its own sake. The more independent routes a report had, the better it was protected against the error of an individual link; the thirty chains around Aisha’s ﵂ garment are not a thirtyfold repetition, they are a thirtyfold control. But it does mean that the bookkeeping of this science exploded, and that any outsider who read its totals as counts of texts would be off by orders of magnitude.
Azami calls the first four centuries of the Hijra the golden age of the science of transmission, and notes in this connection that the number of transmitters grew enormously in that period. It is precisely in that growth that the explanation of the numbers lies. Every new generation of students that took over an existing corpus added no new texts but did add new routes, for each student became himself a link along which the material ran onward, and the accounting of the discipline counted each route. Whoever has read the previous article on tahammul al-'ilm can picture the mechanics of it: teaching was organised per shaykh, each student kept the ahadith of each teacher separately, and of the great teachers dozens of writing students are known by name. A teaching system arranged in this way cannot but make its own counts explode, even when not a single new text is added.
The scholars of the third century saw this more sharply than anyone, for they bore the burden of it. The books grew so voluminous, Azami notes, that they became difficult to handle, and Shu’bah ﵀ therefore gave the practical advice to note the well-known ahadith with a reference to the well-known scholars instead of writing everything out in full each time. That is the behaviour of archivists collapsing under their own administration, not of forgers padding out a corpus. A community that had wanted to enlarge its material artificially would have had no need of abbreviation systems; a community that honestly booked every route did.
The arithmetic of al-Zuhri’s ﵀ school
How fast such an explosion goes Azami demonstrates with a model the reader of this map will recognise at once, because it begins with al-Zuhri ﵀, the compiler from an earlier article. Of al-Zuhri ﵀ at least fifty students are known by name who assembled their own written collections of his material. Suppose now, Azami reckons, that each of those fifty wrote down on average only five hundred transmissions from him, a deliberately modest assumption; then the accounting already counts twenty-five thousand. Let each student in turn have two or three students of his own, and by the end of the second century there stand, from this one school, about seventy-five thousand registered transmission channels in the books; by the time of al-Bukhari ﵀ and his contemporaries they number in the hundreds of thousands. In a footnote he makes it still more concrete: through the mere fact that at least fifty students passed on the book of al-Zuhri ﵀, his material could within twenty-five years become multiplied thirty- to fortyfold.
Let us note what this model is and is not. It is emphatically an illustration, with assumptions that Azami himself designates as averages and suppositions, not a measurement. But it shows something that needs no measurement: that the growth from a few thousand ahadith in the first generation to about three-quarters of a million by the middle of the third century requires no wave of forgery, but is nothing other than the normal propagation of a limited body of texts through an expanding teaching system that books each route separately. The previous article on tahammul al-'ilm described that system from within: hundreds of teachers, each with writing students, each channel neatly administered per shaykh. The great numbers are the bookkeeping shadow of that school. Where the sceptic sees an ocean of fabrications, there stands in reality a cadastre of roads.
With this key one can then do what Azami himself demonstrates to the reader: reckon backwards. When transmission scholars of the third century used to write out every hadith from twenty or thirty channels, and some even claimed to possess every hadith through a hundred channels, then one may infer from this, he writes, what the real numbers were of the transmissions described as six hundred thousand. The division is simple enough to do in one’s head. Six hundred thousand channels, divided by a few dozen routes per report, brings the number of separate reports down to the order of tens of thousands, and whoever then considers that the word hadith in the broad count also embraced the material of Companions ﵃ and Successors sees the Prophetic portion within it shrink further still. The denominator of Guillaume’s ratio thus loses two orders of magnitude in two steps, and with the denominator the ratio disappears.
The same key unlocks the most famous numbers of the third century. The statement of Ahmad ibn Hanbal ﵀ that upwards of seven million transmissions were reliable, of which Abu Zur’ah ﵀ knew six million by heart, is as a count of texts an absurdity and as a count of channels a routine communication: it says that the web of documented, soundly assessed transmission routes ran into the millions in his generation. In the classical literature it is transmitted besides that Ahmad ﵀ is said to have chosen his Musnad from about three-quarters of a million transmissions; that statement does not stand in Azami’s corpus and we pass it on here solely as a classically transmitted figure, but whoever wishes to understand it understands it with the same key, for it uses the same bookkeeping. And the Musnad itself? About forty thousand transmissions, without repetitions an estimated thirty thousand, whereby Azami soberly adds that systematic research into those numbers was still lacking at the time.
Al-Bukhari ﵀ never said what Guillaume makes him say
With this the famous one-in-two-hundred ratio is placed in its proper light. The numerator and the denominator of that ratio measure, after all, different things. The six hundred thousand from which al-Bukhari ﵀ selected are channels: every route of every report, including the thirty routes around one garment and the thirteen around one night prayer, including moreover, in the broad definition, the material of Companions ﵃ and Successors. The 7,397 transmissions his Sahih contains with repetitions, and the 2,602 without repetitions, are the residue of a series of selections that had little to do with authenticity alone: selection by subject, by brevity, by the strictest class of evidence.
That this is no benevolent reinterpretation after the fact is clear from the words of al-Bukhari ﵀ himself, which Azami quotes: I have included in my book al-Jami’ only what was authentic, and I have left out more authentic transmissions than I have included, in order to avoid unnecessary length. The man who according to the sceptical reading declared that nearly everything outside his book was suspect declared in reality the opposite: that outside his book lay more authentic material than within it. He wrote no inventory of all that is genuine; he wrote, at the request of his master Ishaq ibn Rahwaih ﵀, a handbook. Even the title of the work carries that modesty within it, for in the full name, al-Jami’ al-Musnad as-Sahih al-Mukhtasar of the affairs of the Messenger of Allah ﷻ and his Sunnah and his days, al-Mukhtasar means nothing other than: the abridged extract. An epitome is not reproached for not being complete; that it does not wish to be complete stands on its title page.
Whoever wants a check-sum, moreover, need only open al-Bukhari’s ﵀ book itself, for the counting of channels is visible in it down to the very last page. With repetitions the Sahih contains 7,397 transmissions, without repetitions 2,602. Within one and the same book, by one and the same author, the manner of counting thus yields almost a tripling, simply because the same hadith is adduced in several chapters and along several chains, and each adducing counts. If the accounting of a single carefully edited handbook can already inflate its own contents by a factor of three, then it needs no argument what happens when one adds together the loose notes of thousands of scholars over three centuries, with twenty, thirty or a hundred routes per report. The ratio between 600,000 and 2,602 is then no scandal any longer; it is the same ratio as that between 7,397 and 2,602, extended over a whole civilisational domain.
Azami sums it up in the rule of thumb that carries this whole appendix: when the transmission scholars name enormous figures, they mean channels and sources of transmission, not real numbers of hadith; and when they name, on the contrary, small figures, such as the statement that al-Zuhri ﵀ had a thousand hadith and al-Qasim ﵀ two hundred, they most probably mean hadith as subject, not counted per isnad. Large counts paths, small counts content. Whoever confuses the two countings can construct any desired scandal; whoever keeps them apart sees a consistent bookkeeping.
How great the real number of separate Prophetic transmissions then is, Azami treats with the same honesty with which he marks his own uncertainties elsewhere: the exact number is unknown. There is a transmitted statement, ascribed to a series of masters of criticism, al-Thauri ﵀, Shu’bah ﵀, Yahya al-Qattan ﵀, Ibn Mahdi ﵀ and Ahmad ﵀ himself, that there would be only four thousand of them, and Azami calls that statement outright incomprehensible. There is the tendency of later authors to stay, on the authority of al-Hakim ﵀, below ten thousand, whereby Azami shows that al-Hakim’s ﵀ figure concerned only the very highest class of evidence, the transmissions that meet the conditions of al-Bukhari ﵀ and Muslim ﵀, and that al-Hakim ﵀ himself already objected to the broader reading with the question of how the total could be so low, when four thousand Companions ﵃ associated with the Prophet ﷺ for upwards of twenty years and transmitted from him. The margin between those estimates is real and Azami does not polish it away. But it plays itself out between thousands and tens of thousands of subjects, two orders of magnitude below the channel-counts with which the sceptical ratio fills its denominator.
It is worth pausing at this passage over the tone, for it says something about the reliability of the guide this map follows. An apologist who wanted to reassure his reader would have given a smooth answer here: he would have explained away the four thousand of the masters, elevated al-Hakim’s ﵀ ten thousand into dogma, or presented a total of his own with the assurance readers like to hear. Azami does none of the three. He calls the four thousand incomprehensible and leaves it incomprehensible; he explains what al-Hakim’s ﵀ figure does and does not measure; he gives with the Musnad the current estimates and immediately adds that real research is still lacking. Whoever writes thus would rather count accurately than favourably. It is the same scholarly conscientiousness that the reader of this map saw earlier when he called his own arrangement arbitrary and reported without disguise the memory errors of the dictating scholars, and it is the best guarantee that his main conclusion too, channels and not texts, is no wish but an observation.
Inauthentic does not mean: fabricated
One last confusion remains, and it is perhaps the most stubborn of all, because it is hidden in the word rejected itself. If the scholars did not include hundreds of thousands of channels in their strictest books, were all those channels then lies?
The science of hadith judged a transmission in two parts, and it always began with the chain. If the isnad is defective, then the hadith is called defective, without the content having to be weighed separately, for authentic in this science is only what is sound in both parts. A perfectly true saying with an unsound chain is therefore rejected, not because the content would be false, but because the evidence for it along this route does not hold. Azami illustrates it with a document the reader of this map already knows from the Cambridge article and the article on the nuskhah of Suhail ﵀: the nuskhah in the name of al-Zubair ibn Adi ﵀, which is held as a whole to be spurious because the transmitter Bishr ibn al-Husain, according to the critics, falsely ascribed it to his source, while nonetheless about a quarter of its content stands, along other and indeed sound chains, quite simply in al-Bukhari ﵀ and Muslim ﵀ and is there called authentic. The same words, rejected on the one route and accepted on the other: the distinction between judgement of route and judgement of content cannot be drawn more sharply.
Therefore, Azami concludes, the statement that two hundred thousand hadith were not authentic does not mean that two hundred thousand lies were in circulation; it means that the chains of those channels did not meet the requirements, while the content itself may be true or untrue. The great sifting of the third century was no bonfire of fabrications, but an evidentiary sorting of paths: this route is passable, that one not, and about the cargo we speak per route separately. That there also really existed fabricated material besides, this map has never concealed; the community catalogued its forgeries with name and particulars, as appeared earlier with the false copies in the name of Anas ﵁. But the arithmetic that would make a scandal of the sifting itself reckons with the wrong unit.
From the counting to the counters
Thus the sum ends where it had to begin: with the insight that the numbers of the third century are the proof of a science, not the indictment against it. What began as the strongest weapon of the sceptical school, the dizzying denominator from which the canonical books were supposedly fished, turns out on closer inspection to be its weakest piece: a unit of measure misread by readers who did not know the counting system. And it says something about the strength of the tradition that the correction did not have to come from outside. The rule of thumb that unravels the misunderstanding, large counts paths and small counts content, is no modern discovery; it lay already contained in the way Ibn Mahdi ﵀ spoke of his thirteen transmissions and the way Muslim ﵀ set his thirteen chains side by side. Azami did not invent it, he unearthed it. A community that registers every route of every report separately, until of one domestic detail thirty administered paths exist, is the opposite of a community in which fabrications have free play; it has made of the control itself a bookkeeping. But a bookkeeping of paths immediately raises the next question: who guarded the quality of all those paths? Who decided, by what standards, that the one chain was sound and the other not, and who kept the files on the thousands of men and women through whose hands the transmissions passed?
The answer to that question is no side issue, for an administration is only as good as its controllers. Thirty chains around one action prove something only when one knows that each of those thirty routes ran through people whose love of truth, memory and history of encounter had been separately investigated, and that a route with a weak link was also really rejected, however pious its content sounded. The examples of this article already showed that apparatus in the distance, in the unmasking of Yazid’s confusion and in the rejection of a complete nuskhah on account of one unsound transmitter; now it is time to enter the apparatus itself.
That is the subject of the next article: Jarh wa-Ta’dil, the science of the criticism of transmitters, the instrument with which the community subjected the reliability of every link in every chain to an independent judgement. There we shall see how the counters of the channels were at the same time their strictest assayers, how their biographical dictionaries grew into the largest database of persons of the pre-modern world, and why precisely the combination of both, the administration of this article and the criticism of the next, makes the later sceptical theories of a back-projected transmission untenable. Whoever once knows what was counted wants to know who did the counting; and whoever comes to know the counters understands why their numbers deserve the trust the sceptics wished to deprive them of.