Talmudic Logic — the reasoning system
Talmudic argument is not a style of reading. It is a rule system: named inference patterns, invoked explicitly, applied under constraints, and argued about when misapplied. That is precisely what makes it tractable to formalize — and over the last two decades a body of peer-reviewed mathematics has done exactly that, rendering individual Talmudic inference rules as explicit algorithms with defined inputs, defined outputs, and a defined right to say undecided.
This page is the short form of that system: the thirteen hermeneutic rules by which the Torah is expounded, the one rule that has received the most developed formal treatment, and the seven named moves a claim undergoes in the give-and-take of a sugya — each read as an operation on an argument graph.
The thirteen middot of Rabbi Ishmael
Each of the thirteen מידות (middot, "measures") is a named inference pattern with a canonical Talmudic instance. They are not habits of interpretation but rules — cited by name in the argument that uses them, and challengeable on the grounds that the rule was applied where it does not hold.
On dating: the thirteen are attributed to Rabbi Ishmael, a second-century-CE Tanna. They expand the earlier seven rules attributed to Hillel, roughly first century BCE. The codification is therefore late-Second-Temple through Tannaitic — not a single moment, but a rule-set that was itself revised and extended.
| # | Hebrew | Transliteration | Translation | Inference pattern |
|---|---|---|---|---|
| 1 | קל וחומר | Kal va-Chomer | Light and heavy (a fortiori) | Holds in the lesser case ⟹ holds at least as strongly in the greater |
| 2 | גזירה שוה | Gezerah Shavah | Equal decree (verbal analogy) | Shared term across two passages licenses transfer of a rule between them |
| 3 | בנין אב מכתוב אחד | Binyan Av mi-katuv echad | Prototype from one verse | One authoritative case becomes the paradigm for all relevantly similar cases |
| 4 | בנין אב משני כתובים | Binyan Av mi-shnei ketuvim | Prototype from two verses | Two cases sharing a feature and a consequence generalise to all cases with that feature |
| 5 | כלל ופרט | Klal u-Ferat | General then particular | The particular restricts the general: scope collapses to the stated instance |
| 6 | פרט וכלל | Perat u-Klal | Particular then general | The general expands the particular: scope extends to everything of that kind |
| 7 | כלל ופרט וכלל | Klal u-Ferat u-Klal | General–particular–general | The middle term defines the category; scope is what shares its defining features |
| 8 | כלל שהוא צריך לפרט | Klal she-hu tzarich li-Ferat | General requiring the particular | An underspecified general is unusable until a particular fixes its content |
| 9 | פרט שהוא צריך לכלל | Perat she-hu tzarich li-Klal | Particular requiring the general | An isolated particular is unusable until a general supplies its principle |
| 10 | דבר שהיה בכלל ויצא מן הכלל | Davar she-hayah ba-klal ve-yatza min ha-klal lelamed | Singled out to teach about the whole | An item named separately teaches a principle back onto its entire category |
| 11 | דבר שהיה בכלל ויצא לטעון טען אחר | Davar she-hayah ba-klal ve-yatza lit'on to'an acher | Singled out under a new rule | An item named separately is governed by its own rule and exits the general one |
| 12 | דבר הלמד מעניינו | Davar ha-lamed me-inyano | Learned from its context | A term's meaning is a function of its surroundings, not of the term alone |
| 13 | שני כתובים המכחישים זה את זה | Shnei ketuvim ha-machishim zeh et zeh | Two mutually contradicting verses | An apparent contradiction is resolved by a third passage supplying the distinction |
Rules 5–9 are not a list but a connected sub-system about scope — how a category's extension is fixed by the interaction of general and particular language. That cluster is what Studies in Talmudic Logic, Volume 2 formalizes, under the title The Textual Inference Rules Klal uPrat: How the Talmud Defines Sets.
Rule 2, gezerah shavah, carries the tradition's own best argument for formal inference: Hillel used it to re-establish that the Passover offering may be brought on the Sabbath — a ruling the sages had forgotten, and which he recovered from the shared language of two passages. Knowledge reconstructed from the structure of the text rather than from the chain of memory.
Kal va-chomer, formalized: matrix abduction
The first and most-used of the thirteen rules is also the one with the most developed formal treatment. In a 2009 paper in Studia Logica, Abraham, Gabbay and Schild introduced matrix abduction and used it to model the Talmudic a fortiori rule.
A reasoning situation is encoded as a matrix of cases against attributes, entries in
{0, 1}, with one or more blanks. An algorithm defined over superiority relations on
partial orders then decides whether each blank should be filled as 0, as
1, or left explicitly undecided.
ai,j = ? ⟹ 0 │ 1 │ undecided
Two things make this more than a curiosity. First, it is general: the authors present it as a general method for executing non-deductive inferences, and demonstrate it on a modern sentencing problem — sentences for traffic offences — as well as on the Talmudic case study from Kidushin 5a–5b. Second, it is comparative: the paper sets the Talmudic rule beside its counterparts in other legal traditions, Islamic Qiyas and the Hindu Kaimutika Nyaya, treating a fortiori reasoning as a cross-cultural formal object rather than a parochial one. Appendices extend the same machinery to argumentation networks, to voting paradoxes, and to paradoxes of judgement aggregation.
Why an AI reader should care
The feature worth dwelling on is undecided as a first-class output. The algorithm is permitted to decline. A method that must always produce a value will produce one whether or not the evidence supports it; this one is built to report when it cannot tell — and to make that report a legitimate result rather than a failure.
Analysis of the Talmudic Argumentum A Fortiori Inference Rule (Kal Vachomer) using Matrix Abduction. M. Abraham, D. M. Gabbay & U. Schild, Studia Logica 92(3): 281–364 (2009). Special issue New Ideas in Applied Logic, ed. D. M. Gabbay & J. Malinowski. DOI: 10.1007/s11225-009-9202-5
The general treatment sits in the Handbook of Philosophical Logic: Abraham, Gabbay and Schild, "Principles of Talmudic Logic," Volume 18 (2nd edn, Springer, 2018), pp. 133–375 — 243 of the volume's 387 pages, and its single longest item. Full bibliography on the Gabbay page.
The seven operations of a sugya
A sugya — a unit of Talmudic discussion — puts every claim through a small set of named moves. The left column is the traditional operation. The right is its reading as a computation over an argument graph.
| Operation | What it does | Computational reading |
|---|---|---|
| Kushiya (קושיא) objection |
Raise the strongest available attack on a claim: what evidence contradicts it, what assumption is hidden | A typed attack edge into the claim node. Mandatory, not optional — a claim that has faced no objection has no standing |
| Terutz (תירוץ) resolution |
Answer the objection: what evidence rescues the claim | A defence edge that can reinstate the attacked node, restoring it to the accepted set |
| Chiluk (חילוק) distinction |
Show that two apparently contradictory positions hold in different cases or under different assumptions | Matrix completion: find the column that distinguishes the cases, then re-index each claim to its own condition. Both survive, consistently |
| Machloket (מחלוקת) preserved disagreement |
Record both positions with their proponents and their support; never flatten to a winner | Both nodes retained with their evidence. A practical ruling is a separate node, so issuing one does not delete the minority |
| Nafka Mina (נפקא מינה) practical difference |
Determine what actually changes if this position is true | Defeasible consequence: trace the downstream conclusions that depend on the node |
| Safek (ספק) uncertainty |
Register doubt, and of what kind | Typed uncertainty on the node: factual, model, source, measurement, or adversarial. Distinguishes ignorance from risk instead of merging them into one number |
| Teiku (תיקו) unresolved |
Record that the question stands open; do not manufacture a resolution | A logged non-resolution, retained as a citable output. An open question is a result, not a failure |
Most AI systems treat knowledge as answers. The Talmud treats knowledge as structured disagreement. Joshua Lazoff, The Talmudic Epistemic Stack — epistemic case-study brief
Which yields an operating rule for any system built this way:
Do not optimize for answers. Do not optimize for persuasion. Do not optimize for consensus. Optimize for preserving and improving the structure of truth-seeking itself. Joshua Lazoff, The Talmudic Epistemic Stack — epistemic case-study brief
The seven operations are treated as an engineering methodology — typed moves, retained minority positions, logged non-resolutions — on the Sugya Protocol page.
Go deeper
TalmudicLogic.com →
The deep interactive site: each of the thirteen middot with its canonical sugya, worked examples, and the formal frameworks that model it. Start here if you want the rules in motion rather than in summary.
TruthKnowledgeBot.com →
The same discipline applied to a working system: claims that carry their sources, uncertainty that stays typed, and questions that are allowed to remain open.
Dov Gabbay & the mathematics
The logician behind the formalization — labelled deduction, fibring, reactive Kripke semantics, and the Talmudic Logic Project, with full citations.
The Sugya Protocol
Structured disagreement as an engineering methodology for trustworthy, updateable knowledge bases.
Sources
The middot table is condensed from a dataset that carries per-rule Talmudic citations and academic references — Strack & Stemberger (1996); Maccoby (1991); Walton, Reed & Macagno, Argumentation Schemes (Cambridge University Press, 2008); Daube (1949); Jackson (1975); Rosensweig (2003); Neusner (1987). The formalization of kal va-chomer is Abraham, Gabbay & Schild, Studia Logica 92(3): 281–364 (2009). The seven operations and their computational readings are from the author's own epistemic case-study brief, The Talmudic Epistemic Stack.
Scope note, stated deliberately: what this literature establishes is that particular Talmudic inference rules admit formal models and decision procedures. It is not a computability theorem about Talmudic reasoning in general, and nothing on this page should be read as one.
Corrections and citation challenges are welcome — joshua@digitaltwinpro.com.