Ancestry Without Origin: Isnād-cum-Matn Analysis and the Question of Early Hadith

Ignaz Goldziher’s 1890 Muhammedanische Studien argued that much of what the canonical hadith collections present as prophetic reports is the deposit of later theological and legal controversies, retrojected into earlier authorities through the invention or adjustment of transmission chains. Joseph Schacht, sixty years later, translated the suspicion into a rule: the isnād grows backward, so a chain that reaches a first-century figure through a single narrator at any point in its descent is more likely constructed than remembered. The rule has the virtue of falsifiability, and it generates a specific prediction about what the data should look like if Schacht is right. It also produces a methodological impasse. The classical hadith sciences respond to individual narrator reliability through the rijāl apparatus, rating each link in the chain for probity and accuracy of memory, but the rijāl technique presupposes the very thing Schacht denies, that the chains in question record transmission rather than attribution. A positive rijāl verdict on every narrator in a fabricated chain tells us nothing about whether the report they appear to carry ever left the hand of the person who attached their names to it.

Harald Motzki’s work on ʿAbd al-Razzāq’s Muṣannaf, beginning in the 1980s, and extended through his later studies of specific legal and exegetical reports, was an attempt to move past this impasse without capitulating to Schacht’s default reading. If the isnād by itself could not carry the historical argument, and if Schacht was sometimes right that chains grew backward, it did not follow that the corpus as a whole was late, nor that every common link was the point at which a tradition was invented. The question was whether a method existed that could distinguish, within the available evidence, between genuine transmission and back-projection. Motzki’s proposal, which came to be called isnād-cum-matn analysis, begins from a procedural commitment: to analyze every available version of a report not chain by chain and not text by text, but chains and texts together, looking for patterns whose joint distribution is difficult to account for on the hypothesis of late fabrication.

Procedure

One collects every attested version of the report in question, canonical and non-canonical, Sunni and Imami where relevant, early compilations and late, including variant recensions of the same collection. The collection stage already imposes discipline, because it forces the analyst to treat as evidence not the canonical form of the hadith but the distribution of its forms across the tradition. One then draws out the transmission network version by version, noting where chains share narrators and where they diverge. What emerges is a directed graph, rooted at the originating authority (typically the Prophet or a Companion, or in Imami material one of the Imāms) and branching out through successive generations of rāwīs to the collectors in whose books the versions are preserved. The graph is in general not a tree. Chains converge and diverge at various nodes, and some nodes carry many more chains through them than others. The nodes at which multiple independent chains converge are what Schacht and his successors call common links.

Matn comparison distinguishes Motzki’s method from purely isnād-based analysis. For each version in the collection, one records the wording in full, marks every variant against every other, and maps these textual variants back onto the transmission graph. The question is whether the distribution of variants correlates with the structure of the graph, and if so, in what way. The inference that closes the procedure is constrained by these correlations. The shape of the graph and the distribution of textual variants across it together restrict what can be said about the tradition’s history, and in particular about whether a given common link represents a real teaching event, in which a historical figure transmitted material to a community of students who in turn transmitted it onward, or whether the common link represents a later attribution from which the chains below it fan out as a matter of citation rather than memory.

The common link

Schacht’s proposal was that the narrator at whom multiple chains converge is the person responsible for putting the report into circulation, and that, absent evidence to the contrary, this means the common link had either invented the report or was the first to attribute it to the authority at the top of the chain. If a hadith attributed to the Prophet comes to us through twenty chains that all pass through Ibn Shihāb al-Zuhrī in early second-century Medina, Schacht’s default reading is that Zuhrī, or someone using Zuhrī’s name, is the point of origin, and that the single strand above him, running up through one or two narrators to a Companion and then to the Prophet, is a reconstruction rather than a record.

The position has intuitive force. A forger wanting to give a report prophetic authority would need to attach it to a chain, and the cheapest chain to construct is a short one with few intermediary names. The inverted-tree shape (convergence from above, divergence below) is consistent with this picture. The convergence above the common link marks the point beyond which the forger did not bother or could not afford to invent; the divergence below marks the real history of the report’s circulation after its attachment.

Motzki’s revision is not that Schacht was wrong about every case but that his default reading runs ahead of the evidence in a specific way. The inverted-tree shape is also what one would expect if the common link were a real teacher whose material did not enter the written record until somewhat later, whose students transmitted it widely, and whose own teachers taught it to him or to a narrow circle whose other students did not themselves become major transmitters. Both hypotheses, invention and transmission, predict convergence-then-divergence. The question is whether further features of the data distinguish them.

What the pattern tests

Motzki identifies several features of the data whose presence is more costly for the fabrication hypothesis than for the transmission hypothesis, and whose joint presence makes the transmission reading difficult to avoid. The partial common link is the first and most important. A real teacher who taught a report to a substantial group of students will tend to produce not one common link but several generations of nested common links. Below Zuhrī one expects to find Maʿmar b. Rāshid, Ibn Jurayj, Mālik b. Anas, Shuʿayb b. Abī Ḥamza, each himself a convergence point for multiple chains, each heading a sub-cluster with its own downstream network. To simulate this pattern by fabrication would require seeding variants of the report into multiple independent student lineages, arranging for each lineage to develop its own descent pattern consistent with the independently attested biographical and bibliographical data about the narrators involved, and doing so without leaving traces of the seeding process in the chronology of citation. The cost of the fabrication scales multiplicatively with the number of partial common links that need to be simulated, because each of them must be equipped with its own sub-network; the cost of genuine transmission scales additively, because each partial common link is simply a student who taught his own students.

The second test, which depends on the first, is the covariance of matn variants with the branches of the transmission network. If the versions of a report that descend through Mālik share wording choices among themselves, the versions descending through Maʿmar share different wording choices, the Ibn Jurayj versions a third set, and these branch-specific variants remain stable within each branch across several generations of transmitters, the pattern is what teaching produces: each major student retained a slightly different recollection or redaction of the material he received and passed on his version to his own students. A fabricator working from a single text would tend to produce either uniform wording across all branches, if he copied the text faithfully into each manufactured chain, or unstructured noise, if he worked quickly. Clustered variation that tracks branch structure, with variants stable within branch and distinct across branches, is the signature of real descent.

The third feature concerns the shape of the network above the common link, and it separates the proper inverted tree from what Juynboll called the spider: a network in which a single strand runs up through thin chains to an early authority, with branching appearing only among the late collectors who attribute the report back through that single strand. The spider is the topological signature of citation rather than transmission. If a report surfaces for the first time in several late collections, each of which cites it through a chain that shares one or two narrators at its lower end but runs up through a single strand to the Prophet, the simplest reading is that the late collectors draw on a common late source and that the single strand above it reflects attribution rather than memory. The proper inverted tree has body at the middle of the chain, with multiple narrators attested in multiple collections, transmitting to multiple students, at several levels of the descent.

A fourth consideration, which Motzki handles with more caution, concerns the content of the matn rather than its distribution. Fabricated material tends to serve an identifiable polemical or legal purpose and to read as advocacy, with wording chosen to deliver the needed conclusion. Material that transmitted through generations of teaching often preserves wording that is awkward for later readers, underdetermined for later legal debates, or positively inconvenient for the transmitters who kept it in circulation. The presence of such wording in a report is not proof of authenticity, since a careful forger might insert awkwardness to produce the appearance of transmission, but in the aggregate it shifts the prior. Reports whose only features are their utility to later positions invite suspicion; reports with content that no one had reason to invent are easier to read as inheritance.

A phylogenetic aside

The formal structure of these tests bears an affinity to phylogenetic reconstruction in biology, where one infers the descent relationships of species or sequences from the distribution of shared derived characters across a set of taxa. The logic is the same in both cases. If two taxa share a derived feature that is unlikely to have arisen independently in each, the economical hypothesis is common descent from an ancestor in which the feature first appeared, and the pattern of such features across a set of taxa allows the descent tree to be reconstructed. In isnād-cum-matn analysis the taxa are the recorded versions of a report, the characters are the wording choices, and the descent tree is the transmission network. The covariance of variant wording with branch structure is a set of shared derived characters, and the nested common link pattern is the branching topology of the tree.

The analogy is not decorative, because phylogenetic methods have well-understood failure modes and those failure modes translate directly into conditions under which ICMA becomes unreliable. Phylogenetic reconstruction works when the characters being tracked evolve at rates slow enough to preserve signal over the depth of the tree and fast enough to distinguish branches, when horizontal transfer between lineages is minimal, and when taxon sampling is dense enough to resolve the nodes of interest. Hadith transmission meets these conditions unevenly. Horizontal transfer has a specific form in the isnād literature, the phenomenon of narrators receiving the same report through multiple teachers and blending versions, which the rijāl literature calls tadlīs in some of its presentations. When tadlīs is common for a given narrator, the clean branch-specific pattern that ICMA relies on degrades into a reticulation. Rate variation also matters. Wording that is remembered by formula (a proverb, a legal maxim, a prayer formula) evolves slowly and preserves little signal; wording that reflects the transmitter’s own phrasing evolves quickly and may drift too far for cross-branch comparison. The phylogenetic framing allows one to predict in advance which reports will be amenable to ICMA and which will not, rather than discovering case by case that the method does not apply.

The analogy also clarifies what the common link is, ontologically. A common link is not a causal origin of the report; it is the most recent common ancestor of the extant branches, which is a statement about the tree, not about history before the tree begins. This is why Motzki’s method cannot, even in principle, reach past the common link to adjudicate the material above it. The single strands above the common link are pendant edges, and a single pendant edge carries no phylogenetic information about the depth of its attachment to the authority named at the top. The method is blind there by construction.

The Imami reception

The method was developed on the Sunni hadith corpus, and its paradigmatic applications (Motzki on ʿAbd al-Razzāq, Schoeler on biographical traditions) operate in a literature where the transmission networks are dense enough for the required patterns to emerge. The Imami hadith corpus presents a different evidentiary situation. The major compilations, the Kutub al-Arbaʿa of Kulaynī, Ibn Bābawayh, and Ṭūsī, are late-third and fourth-century AH texts, and the chains that reach the Imāms frequently thin out to single or near-single strands in the middle generations, reflecting the historical circumstances of transmission under the later Umayyads and the ʿAbbāsids and the small size of the community that carried this material in the first two centuries. Where Sunni hadith on a given question might be attested through ten chains converging on a second-century common link with four partial common links below him, Imami hadith on a comparable question may be attested through three chains converging on a Companion of al-Ṣādiq with no detectable partial common link structure. The formal conditions under which ICMA produces strong inferences are, on average, less often met.

Seyfeddin Kara’s work on the traditions reporting ʿAlī’s muṣḥaf (Kara, In Search of ʿAlī Ibn Abī Ṭālib’s Codex) is one of the more disciplined recent attempts to apply the method, or something close to it, to Imami material, and it illustrates both the method’s reach and its limits in this corpus. Kara is able to show that certain reports about ʿAlī’s collection of the Qurʾan are in stable circulation by the early second century and cannot be dismissed as late Imami polemical construction, which is a real gain against the Schachtian default. He cannot show, and does not claim to show, that the reports go back to ʿAlī himself, because the chains above the second-century common links are pendant strands of the kind the method cannot resolve. The result holds within the scope the method licenses and does not exceed it.

There is also a tradition-internal reason why ICMA registers differently within Imami scholarship than within Sunni scholarship, which has to do with the status of khabar al-wāḥid in classical Imami uṣūl. Al-Sharīf al-Murtaḍā’s position, which defined the Imami mainstream until Ṭūsī relaxed it, was that reports not rising to the level of tawātur yield no certain knowledge and cannot serve as a basis for legal obligation independently of reason or consensus. Murtaḍā’s Muʿtazilī-rationalist epistemology treats the single-strand report with much more suspicion than the classical Sunni fuqahāʾ treat it, and this suspicion is not the same as Schachtian historical skepticism, but it operates in the same space. A finding that ICMA produces, that a report is in stable circulation at a given date with a plausible common link, is a historical finding about the report’s circulation, not an epistemic finding about its authority. For a Murtaḍawī Imami, circulation at a given date is not what licenses the report’s legal or theological use; only tawātur, or a connection to an Imām, or a conformity with reason does that. The gap between what ICMA can establish and what Imami uṣūl demands is, for this tradition, inherent rather than incidental.

The later Imami rijāl tradition, from Najāshī and Ṭūsī’s Fihrist and Rijāl through Ibn Dāwūd, ʿAllāma al-Ḥillī, Mamaqānī’s Tanqīḥ al-Maqāl, and Khūʾī’s Muʿjam Rijāl al-Ḥadīth, evaluates individual narrators and groups of narrators along lines parallel to the Sunni jarḥ wa taʿdīl literature but with its own categories (thiqa, ṣaḥīḥ, muwaththaq, ḥasan, the Kashshian grading of the aṣḥāb al-ijmāʿ) and its own canon of authorities. It does not pose, as a formal question, the question ICMA poses, which is how the shape of the whole network constrains what can be said about the tradition’s origin. This is not a deficiency of the Imami tradition, because ICMA poses a historical question that the Imami uṣūl tradition had other reasons not to ask, its own epistemological commitments having settled the status of single-strand reports on other grounds. But it means that ICMA, if it is to do work within Imami scholarship, must be introduced as what it is, a tool for a specific historical question about when and through whom a tradition was in circulation, not a replacement for the rijāl apparatus and not an answer to the questions rijāl addresses.

Scope

Rijāl, uṣūl, and ICMA address distinct questions. Rijāl evaluates whether each named narrator in the chain is personally reliable by the standards of jarḥ wa taʿdīl. The uṣūlī evaluation of khabar al-wāḥid determines the authority of a report established as being in circulation, whether it generates ʿilm, ẓann, or obligation. ICMA addresses the prior question that neither of these directly poses: whether the chain is a record of transmission at all, and if so, when the material first entered stable circulation and through whom. The three operate on the same object and their outputs bear on each other, but the questions are distinct and the answer to one does not substitute for another.

What ICMA produces in a successful case is a dated node in a transmission graph, associated with a reconstructed matn or a tight range of variants, and a degree of confidence that scales with the density of the observed pattern. Motzki’s standard treats three or more partial common links with branch-specific matn variation as sufficient for the transmission inference; Juynboll demanded more, was correspondingly skeptical of common links with thinner PCL structure, and the threshold remains a parameter of the method rather than a result of it. The dating the method yields is asymmetric. It fixes a terminus post quem for the common link’s activity, typically in the early-to-mid second century AH for the major Sunni reports Motzki studied, but cannot fix an upper bound on the age of the material the common link was teaching, because the pendant edges above the common link carry no internal evidence about their depth of attachment to the named authority at the top. For Imami material the common link, where one can be identified at all, tends to sit in the third generation of transmitters from al-Bāqir or al-Ṣādiq rather than in the immediate circle of the Imām, which raises the effective terminus post quem by a generation or more and restricts the methodologically available conclusion to the stability of the report within the circles of those later students rather than to its connection to the Imām himself. Combining ICMA’s output with the classical rijāl apparatus and with the uṣūlī evaluation of khabar al-wāḥid requires deliberate work, since each of the three operates at a different level of the question (individual narrator reliability, network-level dating, and the epistemic authority of the conclusion), and their outputs do not concatenate without explicit bridging assumptions.

Paimana