{"path":"research/what-a-rule-rests-on/philosophy-sources-as-read-2026-10-08.md","content":"> **Kept as it came back.** This is the report of a read-only helper (a language model) made on 2026-10-08 for [what-a-rule-rests-on.md](../what-a-rule-rests-on.md). It read philosophical sources on rules, cases and likeness. Claude did not open these sources itself. Its findings are its own reading, with the depth of reading stated per item. One closing line about the helper's own scratch files was removed. Nothing else was changed.\n\n# Rules, cases and likeness: what the philosophical sources say\n\nPhilosophy only. Public-domain works (pre-1930) are quoted exactly; every such quotation was string-checked against text I fetched in this run, except items tagged *snippet* (seen only in a search-result excerpt). For works still in copyright I paraphrase and give pinpoints instead of quoting, with one exception (entry 7b); the wording is at the URLs. \"Read\" gives depth. German renderings are mine. Prepared 2026-10-08.\n\n## 1. Hume, *Enquiry concerning Human Understanding* (1748)\n**Claim.** IX §82 opens by saying all reasoning about matters of fact rests on \"a species of Analogy\": perfect where causes are entirely alike, weaker as similarity falls. IV §31 says the same of all arguments from experience, then asks why one instance would not suffice, if reason drew the inference. He finds no such reasoning; V names custom.\n- IX §82: \"All our reasonings concerning matter of fact are founded on a species of Analogy, which leads us to expect from any cause the same events, which we have observed to result from similar causes.\"\n- IV §31: \"all arguments from experience are founded on the similarity which we discover among natural objects\"; \"Now where is that process of reasoning which, from one instance, draws a conclusion, so different from that which it infers from a hundred instances that are nowise different from that single one?\"\n- V: \"Custom, then, is the great guide of human life.\"\n- Extra, Locke *Essay* IV.xvi.12 (things beyond the senses): \"ANALOGY in these matters is the only help we have, and it is from that alone we draw all our grounds of probability.\"\n\n**URL.** https://www.gutenberg.org/cache/epub/9662/pg9662.txt ; Locke https://www.gutenberg.org/cache/epub/10616/pg10616.txt\n**Read.** Hume IX §§82–83, IV §§31–32, V custom passage; Locke IV.xvi.12. Not whole books.\n\n## 2. Mill, *A System of Logic* (1843)\n**Claim.** (a) II.iii: reasoning from particulars to particulars is primitive (child, dog, an old general's \"ungeneralized analogies\"); general propositions are records and formulas; the real premises are the observed instances. (b) III.iii §3 poses, without solving, why one chemical experiment yields a law while endless black crows do not. (c) III.xx: any resemblance argument short of complete induction is \"analogy\"; the formula cannot tell it from induction, the difference being whether invariable conjunction was established. Analogy serves only for properties \"merely not known to be connected\" with the target.\n- II.iii §3: \"He is not generalizing; he is inferring a particular from particulars.\"\n- II.iii §4: \"All inference is from particulars to particulars: General propositions are merely registers of such inferences already made, and short formulæ for making more\"\n- II.iii §8: \"But this is a security for good reasoning, not a condition of all reasoning\"\n- III.iii §3: \"Why is a single instance, in some cases, sufficient for a complete induction, while in others, myriads of concurring instances, without a single exception known or presumed, go such a very little way toward establishing a universal proposition?\"\n- III.xx §2: \"In the strictest induction, equally with the faintest analogy, we conclude because A resembles B in one or more properties, that it does so in a certain other property.\" Known laws put analogy \"out of court\".\n- III.xx §3: \"a mere guide-post, pointing out the direction in which more rigorous investigations should be prosecuted\"\n\n**URL.** https://www.gutenberg.org/cache/epub/27942/pg27942.txt\n**Read.** II.iii §§1, 3, 4, 8; III.iii §3; III.xx §§1–3 in full.\n\n## 3. Aristotle: Prior Analytics II.23–24, Rhetoric I.2, Posterior Analytics II.19\n**Claim.** Pr. An. II.23: all belief comes by syllogism or induction; induction strictly runs through \"an enumeration of all the cases\" and is \"clearer to us\". II.24: an *example* proves the major of the middle through a term that resembles the third (that fighting the Thebans is evil, argued from their own war on the Phocians); it goes part to part, applies the result to the minor term, and does not need all cases. Rhet. I.2 calls the example an induction. Post. An. II.19: the universal comes from perception, memory and experience; first premisses are known \"by induction\". SEP adds that Topics 156b10–17 separates argument from likeness from induction, since it does not secure the universal.\n- Pr. An. II.24: \"We have an 'example' when the major term is proved to belong to the middle by means of a term which resembles the third.\"\n- Pr. An. II.24: \"reasoning from part to part, when both particulars are subordinate to the same term, and one of them is known\"\n- Rhet. I.2: \"The example is an induction, the enthymeme is a syllogism\"; \"but of part to part, or like to like\"\n- Post. An. II.19: \"out of frequently repeated memories of the same thing develops experience; for a number of memories constitute a single experience\"; \"Thus it is clear that we must get to know the primary premisses by induction\"\n\n**URL.** https://classics.mit.edu/Aristotle/prior.mb.txt (Jenkinson); https://classics.mit.edu/Aristotle/posterior.mb.txt (Mure); https://classics.mit.edu/Aristotle/rhetoric.mb.txt (Roberts)\n**Read.** The cited chapters in full; Topics only via SEP (entry 10).\n\n## 4. Kant, Jäsche *Logic* §84\n**Claim.** Under \"inferences of the power of judgment\", §84 names two inferences from the particular to the empirical universal: induction (many things of a kind, so all) and analogy (many shared determinations, so the rest). Induction follows the principle of generalisation, analogy that of specification: it infers \"von partikularer Ähnlichkeit zweier Dinge auf totale\" (from partial likeness of two things to total). Neither is an inference of reason, only \"logical presumptions\". The printed text keeps them as two species. Student notes of Kant's Blomberg lectures (AA 24:287, Young's translation, as cited by McAndrew) fold analogy into induction, calling it an induction directed at the predicate rather than at members of a class.\n- \"Die erstere Schlußart heißt der Schluß durch Induktion; die andre der Schluß nach der Analogie.\" (The former is called inference by induction, the other inference by analogy.)\n- \"Induktion und Analogie sind daher keine Vernunftschlüsse, sondern nur logische Präsumtionen oder auch empirische Schlüsse\"\n- \"Eines in vielen, also in allen: Induktion; vieles in Einem (was auch in andern ist), also auch das übrige in demselben: Analogie.\" (One in many, hence in all: induction; many in one, hence the rest in it: analogy.)\n\n**URL.** https://www.textlog.de/kant-logik-induktion.html ; https://kantstudiesonline.net/uploads/files/McAndrew00214.pdf\n**Read.** German §84 in full; McAndrew on §84 and the lecture notes; IEP \"Kant: Logic\" summary. No English translation of §84 read.\n\n## 5. Peirce\n**(a) \"Deduction, Induction, and Hypothesis\" (1878).** Rule, case, result: deduction infers the result, induction the rule, hypothesis the case. Analogy gets no separate analysis; resemblance-inference is filed under hypothesis, which \"has been called an induction of characters\" but differs because characters cannot be counted like objects. *The analogy-as-compound claim is not in this paper* (see b, c).\n- \"So that induction is the inference of the rule from the case and result.\"\n- \"It is the inference of a case from a rule and result.\"\n- \"Or, where we find that in certain respects two objects have a strong resemblance, and infer that they resemble one another strongly in other respects.\"\n\nhttps://en.wikisource.org/wiki/Popular_Science_Monthly/Volume_13/August_1878/Illustrations_of_the_Logic_of_Science_VI . Read: whole article.\n\n**(b) \"Some Consequences of Four Incapacities\" (1868).** \"The argument from analogy, which a popular writer upon logic calls reasoning from particulars to particulars, derives its validity from its combining the characters of induction and hypothesis, being analyzable either into a deduction or an induction, or a deduction and a hypothesis.\" (The writer is unnamed; the phrase is Mill's, my inference.) https://www.peirce.org/writings/p27.html . Read: whole page fetched.\n\n**(c) Collected Papers 1.65, 1.69 (\"Lessons from the History of Science\").** Peirce lists deduction, induction and retroduction as the three basic kinds of reasoning and says analogy (Aristotle's *paradeigma*) combines the characters of induction and retroduction. §69 defines analogy as the inference that a small collection of objects agreeing in various respects may agree in a further one (earth and Mars). Paraphrase only: the edition is probably still in copyright. https://www.textlog.de/4224 . Read: §§65–70.\n\n**(d) \"The Doctrine of Necessity Examined\" (1892): the later ground.** Induction, hypothesis and analogy \"depend upon one principle\". The ground is sampling plus long-run self-correction, not a postulate of likeness.\n- \"All are essentially inferences from sampling.\"\n- On the claim that induction postulates \"that under like circumstances like events will happen\": \"But this is a blunder, or bevue, due to thinking exclusively of inductions where the concluded ratio is either 1 or 0.\"\n- \"our inference, which was only provisional, corrects itself at last\"\n\nhttps://en.wikisource.org/wiki/The_Doctrine_of_Necessity_Examined . Read: whole page.\n\n## 6. Keynes, *A Treatise on Probability* (1921), Part III\n**Claim.** Ch. XVIII splits Hume's eggs into two grounds, likeness and number of trials. Number matters only through variety: further instances help by enlarging the negative analogy. Ch. XIX defines positive analogy (shared) and negative analogy (differences), each \"known\" or \"total\". Ch. XXII justifies both methods by a postulate that nature has limited \"independent variety\". Ch. XXIII speaks of \"distinguishing ... analogy from pure induction\".\n- \"We argue from Analogy in so far as we depend upon the likeness of the eggs, and from Pure Induction when we trust the number of the experiments.\"\n- \"an increase in the number of experiments is only valuable in so far as, by increasing, or possibly increasing, the variety found amongst the non-essential characteristics of the instances, it strengthens the Negative Analogy\"\n- \"As the positive analogy measures the resemblances, so the negative analogy measures the differences between the two objects.\"\n- Ch. XXII §9: \"As a logical foundation for Analogy, therefore, we seem to need some such assumption as that the amount of variety in the universe is limited in such a way that there is no one object so complex that its qualities fall into an infinite number of independent groups\"\n\n*Naming:* the capitalised \"Principle of Limited Independent Variety\" is not in the text I searched; Keynes writes \"limitation of independent variety\".\n**URL.** https://www.gutenberg.org/files/32625/32625-pdf.pdf\n**Read.** Ch. XVIII §§3–4, XIX definitions, XX §2, XXII §§5–13, XXIII §7 (via text search). Not the whole Part.\n\n## 7. Goodman\n**(a) *Fact, Fiction, and Forecast*, ch. III, \"The New Riddle of Induction\" (1955; 1983 ed. pp. 63–64 per a 2024 *Synthese* article).** Rules of deductive inference are justified by their fit with the particular inferences we accept, and those inferences by their fit with the rules. Goodman calls the circle virtuous: neither side is privileged, each is revised in light of the other until they cohere, and that coherence is all the justification either needs. He says the same holds for induction, whose canons answer to the inductive practice we accept. This is the passage later called reflective equilibrium; Brewer (entry 17) links it to Rawls. On projectibility: evidence about one piece of copper bears on other copper, but one man in the room being a third son does not bear on the other men, though both are generalisations of the evidence. Syntax cannot mark the difference, so the task is to say in what respects hypotheses must be alike.\nhttps://fitelson.org/probability/goodman_1955.pdf . Read: key passages of ch. III as hosted (searched; 44k characters). The entrenchment chapter was not read.\n\n**(b) \"Seven Strictures on Similarity\" (1972).** First stricture: similarity is relative, variable and culture-dependent. Fifth: the claim that the future will be like the past is empty, since it always is in some way; the question is in which, so similarity gives induction no guidance. Goodman's last sentence there reverses the direction of dependence, and is my one quotation from a work in copyright: \"inductive practice may provide the basis for some canons of similarity\". Seventh: counting shared properties fails, because any two things share some.\nhttps://gwern.net/doc/philosophy/ontology/1972-goodman.pdf . Read: whole essay.\n\n## 8. Quine, \"Natural Kinds\" (1969)\n**Claim.** Quine asks what confirms an induction and answers the grue puzzle by appeal to similarity: we expect the next emerald to be green because two green emeralds are more alike. He treats the sense of similarity as basic to thought and language, holds that any rational expectation presupposes resemblance, and says the underlying standard is innate, shared with other animals, and needed for any learning. Its fit with nature he explains by natural selection, while noting failures (colour). Science supersedes it: by theory we regroup things into new kinds, moving from the subjective animal sense to a similarity fixed by scientific hypotheses and posits. Where such a theory exists, talk of similarity and disposition becomes in principle superfluous.\nhttps://gwern.net/doc/philosophy/ontology/1969-quine.pdf . Read: whole essay searched; the passages on the innate standard, induction and theoretical similarity read in context. Page numbers unverified.\n\n## 9. Hesse, *Models and Analogies in Science* (1966)\n**Claim.** Known similarities and differences between source and target are tabulated side by side. *Horizontal* relations are similarities across domains; *vertical* relations hold within a domain and are typically causal. Keynes's positive and negative analogy are joined by Hesse's *neutral* analogy: properties of the source not yet known to hold in the target (a paper quoting her p. 8 says these are what allow new predictions; *snippet*). Requirements: material (*pre-theoretic*) similarity, causal vertical relations, no essential difference (p. 87, via SEP). Per Bartha (2022), Hesse later accepted only two elementary inductive forms, next-case induction and analogical inference, and found that her Clustering Postulate supports the first but not the second.\nhttps://plato.stanford.edu/entries/reasoning-analogy/ (§§2.2, 3.3) ; https://philinq.it/index.php/philinq/article/download/114/61 ; https://philarchive.org/archive/HTTAPR . Read: SEP §§2–4; Bartha 2022 selectively. **Hesse's book was not read** (the only full-text copy found was blocked).\n\n## 10. Bartha: SEP \"Analogy and Analogical Reasoning\"; *By Parallel Reasoning* (2010)\n**Claim.** The SEP gives the schema (S resembles T in known respects; S has Q; so T plausibly has Q*), reports that no acceptable rule for valid analogical inference has been formulated (§2.4), and surveys justifications: deductive (Aristotle's example; Davies and Russell), inductive (single case; sampling, which founders on counting and bias), a priori (Keynes, Hesse, Bartha) and pragmatic (precedent). On the deductive reading with a generalisation, the source becomes irrelevant (§4.1). Bartha's articulation model replaces overall similarity with two tests: a *prior association* in the source between the known similarities and the projected one, and *potential for generalization* of that relation to the target. Similarity is judged relative to a model of the source (abstract, *snippet*). Its argument is limited to establishing prima facie plausibility, not degree of probability (§4.3). The NDPR reviewer reads Bartha as holding that analogy cannot do much more than make a conclusion plausible.\nhttps://plato.stanford.edu/entries/reasoning-analogy/ ; https://philpapers.org/rec/BARTAM-19 ; https://ndpr.nd.edu/reviews/by-parallel-reasoning-the-construction-and-evaluation-of-analogical-arguments/ . Read: SEP §§2–4; for the book, only the chapter abstract and the review excerpt.\n\n## 11. Norton, *The Material Theory of Induction* (2021), ch. 4 \"Analogy\"\n**Claim.** Analogy is a form of inductive inference, but formal schemas for it (Joyce's *bare analogy*, Hesse's two-dimensional account, Bartha's articulation model) are either too permissive or patched without end. Scientists treat the analogy as a fact to be investigated, not as an argument form. An analogical inference is good to the extent a local warranting fact backs it; no general principle of similarity stands behind it, and stated similarities are mostly shorthand (§4.7, *snippet*). Norton quotes Jevons (1879, p. 110, public domain): \"we ought in fact never to rest satisfied with mere analogy, but ought to try to discover the general laws governing the case\".\nhttps://ucp.manifoldapp.org/read/the-material-theory-of-induction/section/daf5d24b-d0ee-4f5d-9bd1-dafd4e05b687 . Read: §§4.1–4.3 in full; §4.7 and the case studies only as snippets.\n\n## 12. Davies and Russell, \"A Logical Approach to Reasoning by Analogy\" (IJCAI-87)\n**Claim.** The background knowledge that licenses an inference from one case is a *determination rule*: P determines Q (make, model and year determine a used car's value). It is weaker than *all P are Q*, so the source still supplies information (the non-redundancy requirement). A single instance can then license a valid conclusion without induction over many. They note that most work on analogy assumed an instance gives only inductive support for a rule, and suggest rule formation and analogical projection are instead guided by higher-level domain knowledge about relevance (discussion section).\nhttps://www.ijcai.org/Proceedings/87-1/Papers/053.pdf . Read: text searched; abstract, introduction and discussion read.\n\n## 13. John Wisdom, *Proof and Explanation* (Virginia Lectures 1957; published 1991)\n**Secondary sources only.** Wisdom distinguishes deductive argument, inductive or analogical argument, and the case-by-case procedure, an argument by parallels that may use imagined as well as actual cases (Kroes). In his example a child asks the father to ground each general principle; the father cites ever more general ones until he calls one self-evident and then gives instances, as the mother did. Wisdom concludes that the father's procedure does as much as the mother's but no more, so induction and deduction both come down to comparing cases (Kroes, citing pp. 47–50). Kroes also reports (p. 56) that all reasoning, good or bad, comes in the end to the case-by-case procedure. Milkov, quoting Ayers (ODNB), says Wisdom extends to all reasoning what Mill said of induction, and gives Wisdom's remark that the final appeal is to cases (1961, p. 102). Govier says her *a priori* analogies are what Wisdom called case-by-case reasoning (2002 reply).\nhttps://eprints.illc.uva.nl/id/document/11461 (MSc thesis, not peer-reviewed); https://philarchive.org/archive/MILWWYv1 (Milkov); https://informallogic.ca/index.php/informal_logic/article/download/2580/2021 (Govier). Read: the thesis's Wisdom passage and Milkov §5. **Wisdom's own text was not read.**\n\n## 14. Argumentation theory: is analogy deductive with a hidden universal premise?\n**Claim.** *Waller (2001)* argues that *a priori* analogies (law, ethics) are deductive because they carry an underlying principle; but the principle is not logically or epistemically prior to the cases, and the two are adjusted together (what Waller calls thoughtful mutual adjustment). His claim covers a priori analogies only, not inductive ones. *Govier* replies that a universal claim is implied but often cannot be stated and need not be; the reasoning is a priori, not deductive (2002). *Shecaira (2013)* defends a qualified version: one deductive step plus a non-deductive sub-argument. *Guarini (2004)* defends non-deductive reconstruction, partly because strength varies with degree of similarity. *Juthe* (2005 on) holds that analogy is *sui generis*, and (2019 abstract) that inductive and analogical inference differ in ways that cannot be bridged. *Botting (2017)* answers Juthe's objections to deductivism, concedes the cumulative-force argument, and prefers a confirmation-theoretic account. Brewer (n. 95, 1996) names the camps *deductivist*, *inductivist* and *nonreductionist*.\nhttps://informallogic.ca/index.php/informal_logic/article/view/2246/1690 (Waller); https://informallogic.ca/index.php/informal_logic/article/download/2580/2021 (Govier 2002); https://apcz.umk.pl/LLP/article/view/LLP.2019.025 (Juthe); https://apcz.umk.pl/LLP/article/view/LLP.2017.011 (Botting); https://exa.ai/library/publication/dhyw5c1dnsn (Shecaira). Read: Waller mostly; Juthe 2019 abstract page; Govier, Shecaira, Botting and Guarini (via a 2016 Guarini text) as snippets. **Not read:** Guarini 2004, Juthe 2005, Govier 1989.\n\n## 15. Hájek, \"The reference class problem is your problem too\" (*Synthese* 2007)\n**Claim.** An item belongs to many classes and its probability changes with the class chosen. Versions of the classical, logical, propensity and subjectivist interpretations meet their own variants. Taking conditional probability as primitive dissolves the metaphysical problem; the epistemological one, which class should guide inductive reasoning and decision, remains. He quotes Venn (1876, public domain), the earliest statement Hájek knows: \"every individual thing or event has an indefinite number of properties or attributes observable in it, and might therefore be considered as belonging to an indefinite number of different classes of things\". He finds Reichenbach's rule (take the narrowest class with reliable statistics) inadequate. For the founder: choosing which earlier cases count as alike is choosing a reference class.\nhttps://fitelson.org/probability/hajek_rc.pdf . Read: abstract, §1, closing sections; middle searched.\n\n## 16. Indian logic\n**Gillon (SEP \"Logic in Classical Indian Philosophy\").** Early texts use a five-member syllogism that is *analogical*: the Caraka-saṃhitā's argument that the soul is eternal because unproduced, like space, with the application that just as space is unproduced, so is the soul; and Vātsyāyana's versions by similarity and by dissimilarity (Nyāya-sūtra 1.1.32–39). Later canonical forms (Dignāga, Praśastapāda) are deductively valid, with a universal in the example; the Rúshí lùn rejects the analogical form. Dignāga's canonical argument gives a similarity example (pot) and a dissimilarity example (space). Pervasion (*vyāpti*) is defined by occurrence: wherever the second occurs the first occurs. How the universal is known is a problem: it seems to require surveying every instance, which would make inference superfluous. Īśvarasena appealed to non-perception; Dharmakīrti to causation or identity; Gillon says that does not work (Steinkellner dissents). The three marks of a good reason include being found in similar cases and absent from dissimilar ones.\n**Phillips (SEP \"Epistemology in Classical Indian Philosophy\").** Dignāga's tests: the prover is known on the subject, known at least once with the probandum, and no counter-case is known. As a nominalist he treats inference as running from particulars to particulars. Pervasion is typically known by generalising from cases, presupposing knowledge of relevant similarity. *Upamāna* (comparison) is for Nyāya only a way of learning a word's meaning (the *gavaya* is like a cow). In the stock inference (smoke on the mountain) the kitchen hearth counts as the example because it is relevantly similar to the mountain.\nhttps://plato.stanford.edu/entries/logic-india/ ; https://plato.stanford.edu/entries/epistemology-india/ . Read: Gillon §§3–4 and the induction passage; Phillips on inference and §7.1. **No primary text** (Nyāya-sūtra, Dignāga); IEP Nyāya not read.\n\n## 17. Law and ethics\n- **Levi, \"An Introduction to Legal Reasoning\" (1948 article; 1949 book).** The basic pattern of legal reasoning is reasoning by example, case to case: similarity is seen, the rule inherent in the first case is announced, and it is applied to the second. There is no fixed prior rule saying which similarity is decisive, so the rules are made in the act of comparing, and the classification changes as it is made. Levi adds that in a sense all reasoning is of this type, and footnotes Prior Analytics 69a. https://chicagounbound.uchicago.edu/cgi/viewcontent.cgi?article=5687&context=journal_articles . Read: pp. 501–503.\n- **Sunstein, \"On Analogical Reasoning\" (1993).** Analogy is bottom-up and incompletely theorized, with principles developed by constant reference to particular cases; he rejects the view that it is merely a crude version of the search for reflective equilibrium. http://www.horty.umiacs.io/courses/readings/sunstein-1993-analogy.pdf . Read: introduction and §I.A.\n- **Sunstein, \"Incompletely Theorized Agreements\" (1995).** People agree on outcomes and low-level explanations without agreeing on fundamental principle. A copy of the text says a judge cannot go from one particular to another without identifying a principle or reason (*snippet*). https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2995488 ; https://exa.ai/library/publication/bv2mhppkkss . Read: abstract and excerpts only.\n- **Brewer, \"Exemplary Reasoning\" (1996).** An *analogy-warranting rule* states the logical relation between the shared and the inferred characteristics. It is found by abduction, then tested by *reflective adjustment* among the rule, the examples and explanatory rationales; the rule plus the new case then yields the conclusion. He cites Goodman and Rawls for the device. http://www.horty.umiacs.io/courses/readings/brewer-1996-analogy.pdf . Read: taxonomy, model summary, rule definition.\n- **Sceptics (via Schauer and Spellman 2017).** Alexander and Sherwin (2008, pp. 76–83, 88): similarities are infinite, so a rule or principle is needed to pick the important ones, and analogy is ordinary moral, empirical and deductive reasoning. Posner (2008, p. 182): consulting precedent means looking for policy insights. Schauer (2008, abstract): precedent binds even a judge who disagrees, so it is not analogy. Schauer and Spellman themselves: analogy begins with an initial perception of similarity shaped by expertise, and both camps are partly right. https://lawreview.uchicago.edu/sites/default/files/12%20SchauerSpellman_SYMP_IC.pdf ; https://journals.sagepub.com/doi/10.1111/j.1745-6924.2008.00090.x . Read: S&S abstract and §§II–III. **Alexander, Sherwin and Posner not read directly.**\n- **Lamond, SEP \"Precedent and Analogy in Legal Reasoning\".** It is widely agreed that an analogy depends ultimately on the justification of the earlier decision; the principles-based and reasons-based accounts are set out. https://plato.stanford.edu/entries/legal-reas-prec/ . Read: §§3 (end)–5.\n- **Jonsen and Toulmin, *The Abuse of Casuistry* (1988).** Maxims are understood through the paradigm cases they cover; no rule interprets itself; they call for a methodical map of morally significant likenesses and differences. Members of the National Commission agreed on cases while disagreeing on why (*snippet*). https://www.scribd.com/document/728139093/Jonsen-Et-Toulmin-The-Abuse-of-Casuistry . Read: Prologue pp. 1–15 in a Scribd copy.\n\n## 18. Carnap (secondary only)\nCarnap's last work leaves inductive logic sensitive to analogy by proximity and by similarity as open tasks (Skyrms, abstract). Analogy by similarity uses distances between predicate regions in an attribute space and falls outside the λ-γ-continuum (Springer chapter, *snippet*). Skyrms's text says Johnson's sufficientness postulate just is the absence of analogy by similarity (*snippet*), so Carnap's continuum, which keeps that postulate, ignores similarity. https://projecteuclid.org/ebooks/institute-of-mathematical-statistics-lecture-notes-monograph-series/Statistics-probability-and-game-theory/chapter/Carnapian-inductive-logic-and-Bayesian-statistics/10.1214/lnms/1215453580.pdf ; https://link.springer.com/chapter/10.1007/978-3-662-72937-3_52 . Read: Skyrms abstract and §§1–2. **Carnap not read.**\n\n## 19. Watanabe's ugly duckling theorem (secondary only)\nWithout a bias privileging some predicates, any two objects share as many extensional predicates as any other two. A philosophy paper draws the moral that calling two things similar for sharing properties requires privileging some properties (*snippet*). Wikipedia dates the result to Watanabe 1969, pp. 376–377. https://docslib.org/doc/9376091/objective-similarity-and-mental-representation ; https://en.wikipedia.org/wiki/Ugly_duckling_theorem . **Watanabe not read**; the scan link on Wikipedia gave 404.\n\n## 20. Wittgenstein (via SEP)\nFamily resemblance (PI 65–67) replaces a common core with a web of overlapping likenesses among particular cases; the SEP stresses that the looking is done at particular cases, not generalisations. Rule-following (PI 185–242; the paradox at 201) says no course of action is fixed by a rule, since any course can be made to accord with it; the way out is a way of grasping a rule that is not an interpretation but shows in cases of obeying and going against it (IEP *snippet*). https://plato.stanford.edu/entries/wittgenstein/ ; https://iep.utm.edu/wittgens/ . Read: SEP on family resemblance and rule-following. *Investigations* itself not read.\n\n## 21. Papers from 2020 to 2026\n- **Walliser, Zwirn and Zwirn, \"Analogical Reasoning as an Inference Scheme\", *Dialogue* (online 2021).** Nearest to the founder's thesis. §4 says single-case induction and analogy share one form, so every step of enumerative induction is analogical, and single-case induction is analogy in a normalised setting resting on entrenched over-hypotheses. But it adds that this does not make analogy reducible to induction, and that analogy needs an over-hypothesis linking two sets of properties, with belief in the conclusion tracking belief in it. https://www.cambridge.org/core/services/aop-cambridge-core/content/view/2F91715A8DA23DB57207BD73D5A2D7B9/S0012217321000226a.pdf/div-class-title-analogical-reasoning-as-an-inference-scheme-div.pdf . Read: §4 in full, abstract, framing.\n- **Nappo, \"Confirmation by analogy\", *Synthese* 2022.** A Bayesian account on which analogies give inductive support; he denies this is a self-standing justification (*snippet*). https://link.springer.com/article/10.1007/s11229-022-03545-w . Read: abstract, opening.\n- **Alhambra, \"A Particularist Approach to Arguments by Analogy\", *Argumentation* 2023.** Against *generalism* (analogy needs a universal principle): such principles are by-products of good analogies, not requirements. He cites Beardsley's worry that once the generalisation is explicit the analogy is idle (*snippet*). https://link.springer.com/article/10.1007/s10503-023-09616-7 . Read: abstract, introduction.\n- **Kroes, MSc thesis (Amsterdam 2021).** Analogy underlies both induction and deduction, building on Wisdom and Turner. Not peer-reviewed. https://eprints.illc.uva.nl/id/document/11461\n- **Bartha 2022** (entry 9): Hesse's later work (1974, 1988) grows more pessimistic about inductive logic.\n\n## Closing: how the sources sort (only from what I read)\n\n**(a) Support \"rules rest on cases, and cases on likeness\"**\n- Hume (1): all experience-based reasoning is \"founded on the similarity\" among objects; yet no reasoning bridges one instance and many, so custom does.\n- Aristotle (3): the example runs \"like to like\" and is an induction; the universal arises from many remembered cases.\n- Mill (2): \"All inference is from particulars to particulars\"; general propositions are \"registers\"; one formula covers induction and analogy.\n- Kant (4): induction and analogy are twin inferences from particulars; lecture notes fold analogy into induction.\n- Peirce 1868 and CP 1.65 (5b, 5c): analogy combines induction with hypothesis (retroduction). A compound, not a foundation.\n- Keynes (6): Hume's eggs rest on likeness plus number; extra instances count by enlarging the negative analogy.\n- Locke (1): analogy is \"the only help we have\" beyond the senses.\n- Quine (8): every reasonable expectation rests on similarity.\n- Wisdom (13, second-hand): all reasoning comes in the end to the case-by-case procedure.\n- Levi, Jonsen and Toulmin (17): reasoning by example; maxims learned through paradigm cases.\n- Indian sources (16): the early five-member syllogism is analogical; Dignāga goes particulars to particulars; pervasion is generalised from cases and presupposes relevant similarity.\n- Walliser et al., Kroes (21): each enumerative step is analogical.\n- Wittgenstein via SEP (20): particular cases and family resemblance replace general explanation.\n\n**(b) Deny it, or reverse it (likeness needs a rule, theory or fact first)**\n- Goodman (7): inductive practice supplies canons of similarity; similarity gives induction no guidance.\n- Quine (8): the similarity standard is subjective and animal; theory-determined similarity replaces it.\n- Peirce 1892 (5d): induction, hypothesis and analogy rest on sampling and self-correction; a postulate of like circumstances is \"a blunder\".\n- Mill (2): analogy covers only properties \"merely not known to be connected\"; known laws put it \"out of court\"; it points toward induction.\n- Keynes (6): both methods rest on a postulate about the world, limited independent variety.\n- Norton (11): a local warranting fact comes first; similarities are shorthand.\n- Davies and Russell (12): a determination rule (relevance knowledge) must exist; then one case suffices.\n- Bartha, Hesse (9, 10): similarity is judged relative to a model of the source; causal vertical relations constrain which similarities count; her postulate supports induction, not analogy.\n- Alexander and Sherwin, Posner, Schauer (17): similarities are infinite, so a rule or policy comes first; precedent is not analogy.\n- Waller, Shecaira, Beardsley via Alhambra (14, 21): a hidden principle carries the inference.\n- Watanabe, Hájek, Wittgenstein (19, 15, 20): no similarity without privileged predicates, a reference class, or a practice of application.\n- Indian logic (16): the Rúshí lùn replaces the analogical syllogism with one stating a universal; Dharmakīrti grounds pervasion in causation or identity.\n- Juthe 2019 (14): analogical inference is not inductive inference. Carnap (18, second-hand): analogy by similarity needs a metric over predicates.\n\n**(c) Rules and cases depend on each other**\n- Goodman (7): rules and inferences are adjusted against each other until they agree.\n- Brewer (17): the analogy-warranting rule is abduced from examples, then tested against them and its rationales.\n- Levi (17): rules are discovered while determining similarity and difference.\n- Sunstein (17): principles developed with constant reference to cases; a judge needs a reason to move between particulars.\n- Waller (14): the principle is not prior to the cases; mutual adjustment.\n- Lamond (17): an analogy depends on the justification of the earlier decision.\n- Schauer and Spellman (17): perceived similarity reflects the principles the perceiver brings; sceptics and celebrants are each partly right.\n- Jonsen and Toulmin (17): no rule interprets itself; paradigms and maxims are learned together.\n- Walliser et al. (21): the step is analogical but needs an entrenched over-hypothesis.\n- Peirce 1878 (5a): rule, case and result are symmetric; each is inferable from the other two.\n- Mill (2): a generalisation is \"a security for good reasoning, not a condition of all reasoning\".\n\n## Corrections to your brief, and what I could not verify\n**Corrections.**\n1. Peirce's 1878 paper has no analogy-as-compound; that is 1868 and CP 1.65. In 1878 resemblance-inference counts as hypothesis.\n2. Keynes writes \"limitation of independent variety\"; the capitalised \"Principle of Limited Independent Variety\" was not found in his text.\n3. Levi: the 1949 book reprints a 1948 *University of Chicago Law Review* article, which is what I read.\n4. Waller's deductivism covers a priori analogies only; Shecaira's is mixed; Guarini argues non-deductive reconstruction, so \"own kind\" fits Govier and Juthe better.\n\n**Not verified, or second-hand.**\n- Kant §84 in English; Hesse 1966; Bartha's book; Wisdom 1991; Carnap; Watanabe (scan link 404); Wittgenstein's text; Dignāga and the Nyāya-sūtra; IEP Nyāya.\n- Guarini 2004, Juthe 2005, Govier 1989/1999; Botting and Shecaira (abstract only); Alexander and Sherwin, Posner, Dworkin (via S&S); Schauer 2008 (abstract).\n- Sunstein 1995 full text and page numbers; Jonsen and Toulmin (Scribd copy; Commission passage second-hand).\n- That Peirce's \"popular writer\" is Mill (my inference); the date of CP 1.65.\n- Texts are Gutenberg and web editions; Quine and Goodman pagination is second-hand.\n- Copyright handling: in-copyright works are paraphrased, so exact wording must be taken from the URLs.\n"}