Skip to content

Reasoning From What We See To What We Cannot: Philodemus' On Methods Of Inference And The Epicurean Answer To The Stoics

The Stoics Said Only Logic Can Give Certainty. The Epicureans Answered That Logic Itself Was Learned From Experience.

Epicurus taught that the universe is made of atoms and void, that the atoms are infinite in number, that worlds without number form and dissolve, and that the blessed beings he called gods live without trouble and cause none. No one has ever seen an atom, the void, another world, or a divine being with his eyes. How, then, could Epicurus claim to know any of these things?

His answer is short. “As regards the imperceptible we must draw inferences from phenomena. For all thoughts have their origin in sensations by means of coincidence and analogy and similarity and combination, reasoning too contributing something” (DL X.32). And again: “that bodies exist, sense itself witnesses in the experience of all men, and in accordance with the evidence of sense we must of necessity judge of the imperceptible by reasoning” (DL X.39). Epicurus shows the method at work when he reasons about the smallest parts of the atom “on the analogy of” the smallest things we can see: “the affinity which the least parts of the atom have to the homogeneous parts of sensible things is sufficient to justify our conclusion to this extent” (DL X.58–59).

Two centuries later, the Stoics attacked that method directly, and the Epicureans answered them at length. The record of that debate survives in Philodemus’ On Methods of Inference, often called On Signs (Greek Peri Sēmeiōseōn, “On Inferences from Signs”). Philodemus of Gadara (about 110 to about 40 BC) led an Epicurean school at Naples and Herculaneum in the time of Cicero. His treatise was buried by Vesuvius in AD 79 in the library of the Villa of the Papyri at Herculaneum, recovered as a charred roll in the eighteenth century, and translated into English by Phillip and Estelle De Lacy in 1941. The full translation, with De Lacy’s introduction and supplementary chapters and an outline linked to every column of the papyrus, is on our page Philodemus, On Methods of Inference (De Lacy).

This article sets out what the debate was about, what the Epicureans answered, and how their answers fit with the rest of what Epicurus taught about knowledge.

On Methods of Inference is not a treatise in the usual sense. Philodemus collected the arguments of three Epicurean teachers, Zeno of Sidon (his own teacher at Athens), Bromius, and Demetrius of Laconia, and of a fourth whose name is lost. Each of them reports the Stoic objections and answers them. As De Lacy puts it, Philodemus “merely reports to his readers the views of the leading Epicureans on the problem of empirical inference” (De Lacy’s Introduction).

The chief Stoic opponent is named Dionysius, probably Dionysius of Cyrene, a Stoic of the second century BC. The Stoics drew a line between two kinds of signs. A “common” sign can exist whether or not the hidden thing exists, as wealth can belong to a good man or a bad one: “anyone who thinks that a certain man is good because he is rich is using an unsound and common sign” (Col. I). A “particular” sign exists only when the hidden thing exists. And the Stoics held that a particular sign can be established only by one formal test, which De Lacy translates “contraposition” (Greek anaskeuē, “elimination”): imagine the hidden thing away, and if the visible sign must disappear with it, the sign is valid.

The Epicureans agreed that common signs prove nothing. What they rejected was the claim that the formal test is the only road to knowledge. Their method was inference “by analogy” or “by similarity” (Greek homoiotēs): what holds without exception in all the cases we have observed, in all their variety, holds also in cases we cannot observe, so long as nothing in our experience conflicts.

The debate turns on a handful of words that are easy to confuse, especially “mark” and “sign.” The Latin words are the ones Cicero and Seneca chose when they explained the Greek debate to Roman readers. They are often closer to English than the Greek, and they show how the ancients themselves understood the terms.

EnglishLatinGreekHow the term is used in this article
impression, presentationvisum (“a thing seen”; Cicero, Academic Questions II.18)phantasiaWhat the senses present to the mind. Bury’s Sextus translates “presentation.”
kataleptic or apprehensive impressionvisum impressum effictumque ex eo, unde esset (“an impression stamped and molded from what it came from,” Cicero’s rendering of Zeno’s definition, II.18)katalēptikē phantasiaThe Stoic class of impressions that are supposed to be true beyond any possibility of error.
comprehension, graspcomprehensio, perceptio, cognitio (II.17)katalēpsisThe Stoic act of grasping an object through a kataleptic impression.
stamped, sealedimpressum et signatum et effictum (“impressed and sealed and molded,” II.77); signatum (Seneca, Letters 113.16)enapesphragismenē, enapomemagmenē (“sealed in,” “imprinted”)The Stoic picture of an object pressing its likeness into the mind, as a seal ring presses its design into wax.
craftsman, designerartifex, opifex (“craftsman,” “maker”); ignis artificiosus (“craftsmanlike fire”), Zeno’s nature (On the Nature of the Gods II.57–58, II.142); divinus artifex (“divine craftsman,” Seneca, Letters 113.16)technitēs; pyr technikon (“craftsmanlike fire”)The Stoic view of nature, or God, as a designing craftsman who gives each thing its own pattern and builds the senses to perceive it. It is the final guarantee of the marks. Epicurus denied that the world has any designer.
marknota (“mark, brand,” II.33–34); insigne veri (“the badge of truth,” II.36); proprietas (“special character,” Seneca; “singularum rerum singulas proprietates,” II.56); certa iudicandi et adsentiendi nota (“an infallible mark to guide judgement and assent,” On the Nature of the Gods I.12)idiōma (“peculiar feature”)A feature of an impression that is supposed to show that the impression is true. It belongs to the theory of perception.
common / peculiarcommunis / proprius (“common to true and false” / “peculiar to truth,” II.33–34)koinos / idiosWhether a mark, or a sign, could belong to a falsehood as well as to the truth.
clear evidenceperspicuitas, evidentia (Cicero’s coinages, II.17)enargeiaWhat is plainly seen. Epicurus’ word for the presentation itself (Sextus, Against the Logicians I.203).
opinionopinio (II.45)doxaThe judgment the mind adds to an impression. For Epicurus, the only place where error lies.
anticipation, preconcept, notionanticipatio (“anticipation”) and praenotio (“prior notion”), Cicero’s coinages for Epicurus’ term: “anticipationem quandam deorum, quam appellat πρόληψιν Epicurus” (On the Nature of the Gods I.43–44); notitia (Academic Questions II.22; Lucretius IV.476–479, “notitiem veri,” “the concept of the true”)prolēpsis, ennoiaThe general idea formed from many experiences, by which we recognize things. De Lacy translates prolēpsis as “anticipation” (Col. XXXIV, Fragment I). Rackham’s Cicero has “preconception.” We say “preconcept.”
signsignum (“sign,” II.36); id quod significatur (“the thing signified,” II.36)sēmeionSomething perceived from which we infer something not perceived, as smoke signifies fire. It belongs to the theory of inference.
inference from signs—sēmeiōsisReasoning from what is seen to what is unseen. The subject of Philodemus’ book.
similarity, analogysimilitudohomoiotēsThe Epicurean method of inference: what holds without exception in experience holds also where we cannot see.
contraposition, elimination—anaskeuēThe Stoic test: suppose the hidden thing away, and see whether the sign must disappear with it.

The two words in bold are the ones to keep apart. A mark (nota) is a feature of a perception that is supposed to certify the perception itself as true. A sign (signum) is something perceived that leads the mind to something not perceived. The Latin shows how close they stand: Seneca’s leaves are each signatum, “sealed,” with a special character, and Zeno’s impression is signatum, “sealed,” by its object, from the same root as signum. That closeness is not an accident, as the next section shows.

Marks Before Signs: The Stoic Kataleptic Impression

Section titled “Marks Before Signs: The Stoic Kataleptic Impression”

The Stoic demand for “particular” signs came from the Stoic theory of perception.

The Stoics held that some impressions, and only some, are a reliable test of truth. Cicero gives Zeno’s definition in Latin: such an impression is “impressum effictumque ex eo, unde esset, quale esse non posset, ex eo, unde non esset”, an impression “stamped and molded from what it came from, of such a kind as could not come from what it did not come from” (Academic Questions II.18, our translation). Sextus gives the same definition in Greek: “An apprehensive presentation is one caused by an existing object and imaged and stamped in the subject in accordance with that existing object, of such a kind as could not be derived from a non-existent object” (Against the Logicians I.248). Such an impression “has a special characteristic of its own, like the horned serpents as compared with all other serpents” (I.252).

That special characteristic is the mark, in Cicero’s Latin the nota, a word for a brand or identifying mark. The Stoic-minded Lucullus says we know true things “by a mark belonging peculiarly to truth, and not common to what is true and false” (“propria veri, non communi veri et falsi nota,” Academic Questions II.33–34). Without such a mark, he says, no one can have “any mark at all by which he can know the truth” (“ullam omnino veritatis notam”). Cicero also calls it the insigne veri, “the badge of truth” (II.36). And when such an impression meets no obstacle, the Stoics said, it “lays hold of us, almost by the very hair, as they say, and drags us off to assent” (Sextus I.257).

Where Do The Marks Come From? Every Hair, Every Grain, Every Leaf

Section titled “Where Do The Marks Come From? Every Hair, Every Grain, Every Leaf”

This is the crucial point, and it is easy to grasp. A seal can leave a mark that belongs to it alone only if no other seal is exactly like it. If two seals were identical, their impressions in the wax would be identical, and no one could tell from the wax which seal had made it. So the Stoic theory of marks required that no two things in the whole world be exactly alike.

The Stoics accepted that consequence and asserted it as literal fact. Cicero reports their claim in plain words: “nullum esse pilum omnibus rebus talem, qualis sit pilus alius, nullum granum”, “there is not a single hair or a single grain in every respect like another hair or grain.” And he adds his own verdict: “Stoicum est istud quidem nec admodum credibile”, “That is a Stoic doctrine, and one not very credible” (Academic Questions II.85).

Seneca, a Stoic himself, tells us where this uniqueness was supposed to come from. It was the work of God: “Inter cetera propter quae mirabile divini artificis ingenium est hoc quoque existimo esse, quod in tanta copia rerum numquam in idem incidit … Tot fecit genera foliorum: nullum non sua proprietate signatum,” “And among the other reasons for marvelling at the genius of the Divine Creator is, I believe, this,—that amid all this abundance there is no repetition; even seemingly similar things are, on comparison, unlike. God has created all the great number of leaves that we behold: each, however, is stamped with its special pattern” (Seneca, Moral Letters 113.16).

Cicero’s Stoic speakers make the same claim in the same Latin words. Lucullus, defending the Stoic theory against the twins, appeals to “that doctrine which is laid down very clearly by the more refined natural philosophers, that everything has its own separate property” (“singularum rerum singulas proprietates esse,” Academic Questions II.56). That is Seneca’s word, proprietas, the special character stamped on each leaf. And Balbus, the Stoic speaker in Cicero’s On the Nature of the Gods, shows who does the stamping. Zeno, he says, defined nature as “a craftsmanlike fire, proceeding methodically to the work of generation” (“ignem esse artificiosum ad gignendum progredientem via”), and the nature of the world itself “is styled by Zeno not merely ‘craftsmanlike’ but actually ‘a craftsman,’ whose foresight plans out the work to serve its use and purpose in every detail” (“non artificiosa solum sed plane artifex … consultrix et provida,” On the Nature of the Gods II.57–58). The same craftsman built our senses: “what artificer but nature, who is unsurpassed in her cunning, could have attained such skilfulness in the construction of the senses?” (“Quis vero opifex praeter naturam, qua nihil potest esse callidius, tantam sollertiam persequi potuisset in sensibus?” II.142). Seneca’s divinus artifex is Zeno’s nature, the artifex and opifex, under another name.

The Latin merits a close look. The “Divine Creator” is the divinus artifex, the divine craftsman, an artisan who makes things by design. Each leaf is signatum, “sealed” or “stamped,” with its own proprietas, its own special character. The same word, signatum, appears in Cicero’s version of Zeno’s definition of the kataleptic impression: “impressum et signatum et effictum”, “impressed and sealed and molded” by its object (II.77). The chain of reasoning is complete:

  1. God the craftsman seals every object with a character of its own, so no two things are alike.
  2. Each object therefore seals the mind with an impression that no other object could produce.
  3. That impression carries a mark, a nota, peculiar to the truth.
  4. By that mark, the wise man knows that his impression is true.

Lucullus completes the picture by appealing to “with how much artificial skill, as it were, nature has made, first of all, every animal; secondly, man most especially,” and to “the mind of man,” which is “admirably calculated for the science of things” (Academic Questions II.30–31). For the Stoics, nature is a providential, rational power. The certainty of the kataleptic impression rests, in the end, on divine design.

That is why the skeptics’ attack was so effective. They held that “a false one exactly similar to the apprehensive presentation can be found” (Sextus I.252): twins such as Publius and Quintus Geminus, eggs, identical statues, the dreams and visions of madmen. Anyone can see that two eggs look alike. If even two impressions can be identical, the mark is worthless, because, in Cicero’s Latin, “proprium in communi signo notari non potest”, “nothing peculiar can be noted in one sign common to two things” (II.34). Cicero, speaking as an Academic at the opening of On the Nature of the Gods, puts the skeptic case in the same terms: all true sensations are joined to false ones “so closely resembling them that they contain no infallible mark to guide our judgement and assent” (“ut in is nulla insit certa iudicandi et adsentiendi nota,” I.12). The Stoic claim that every hair and grain is unique was not a curiosity. It was the foundation the whole theory of marks stood on, and it is a claim about the world that no one could ever check. Our article Twins, Eggs, and Duplicate Worlds follows that fight.

Did The Epicureans Hold The Opposite? Limited Kinds Of Atoms, Democritus, And Free Will

Section titled “Did The Epicureans Hold The Opposite? Limited Kinds Of Atoms, Democritus, And Free Will”

If the Stoics said that no two things are ever alike, did the Epicureans say that exact duplicates are everywhere? No. That was Democritus, and Epicurus corrected him on two points that matter here.

First, the kinds of atoms are limited, so what is possible is limited. Democritus held that the shapes of atoms are infinite in number. Epicurus argued against him that the atoms come in a limited number of kinds, “not quite infinite, but only incomprehensible in number,” while “in each shape the atoms are quite infinite in number” (Letter to Herodotus 42). David Sedley states the point plainly: “we know that Epicurus went to considerable lengths to argue against Democritus’s thesis that there are infinitely many types of atom. … The Epicurean conclusion is that … there are only finitely many types of atom in the universe. Although the number of atomic types is ‘unimaginably’ large, it is not actually infinite” (Sedley, Creationism and Its Critics in Antiquity, 2007, pp. 161–162). A limited alphabet of atoms means a limited range of things that can be built from them. That is why Lucretius can say that “a fixed limit to things marks the extreme on either side” (Lucretius II.507–521), and why Bromius can dismiss the “men with a nature of iron, who go through walls” (Col. XXI). Limited kinds put limits on what nature can produce anywhere in the infinite universe. They also guarantee that every kind of thing that can exist, and every kind of world that can sustain life, recurs again and again, as our article The Power of Infinity explains. What recurs, in Epicurus’ account, is the kind: “in the universe there is nothing single, nothing born unique and growing unique and alone, but it is always of some tribe” (Lucretius II.1077). Every man has fellow men, every sun has fellow suns, every world has worlds like it.

Second, perfect doubles of worlds and men were Democritus’ doctrine, not Epicurus’. Cicero makes the attribution explicit. The skeptic, Lucullus complains, argues that “Democritum dicere innumerabilis esse mundos et quidem sic quosdam inter sese non solum similis, sed undique perfecte et absolute ita paris, ut inter eos nihil prorsus intersit … itemque homines”, “Democritus says that there are innumerable worlds, and some of them not only similar to one another, but in every way so perfectly and absolutely equal that there is no difference at all between them … and the same of men,” so that other worlds hold “innumerable” copies of Quintus Lutatius Catulus (Academic Questions II.55, our translation). Lucullus replies, “you call me to Democritus, with whom I do not agree” (“me ad Democritum vocas, cui non adsentior,” II.56). Epicurus’ own words are more careful: “there are infinite worlds both like and unlike this world of ours” (Letter to Herodotus 45). Like, not identical.

Sedley explains why the difference is no accident: “by contrast with Democritus and, in due course, the Stoics as well …, Epicurus seems not to have spoken of two worlds being perfect doublets of each other, down to the minute detail of containing identical inhabitants with the same names, histories etc.” Democritus and the Stoics leaned toward determinism, the doctrine that the same starting conditions must produce the same events, human actions included, and identical worlds with identical histories are that doctrine written large. Epicurus rejected determinism. He introduced “a small degree of genuine indeterminacy into the motions of atoms, the ‘swerve,’” and so “even if world A at some given moment were atomically indistinguishable from world B at some given moment, nothing would follow about those two worlds and their inhabitants also having identical histories” (Sedley, pp. 164–165). Sedley adds that Epicurus “is not a Democritean reductionist,” so that for him “even two worlds with identical atomic histories might not contain psychologically indistinguishable people” (p. 165 n. 61).

This is where free will enters. Epicurus held that of the things that happen, “some happen by necessity and some by chance, and some are within our control” (Letter to Menoeceus 133). Lucretius ties that control to the swerve, which breaks “the decrees of fate”: “whence comes this free will for living things all over the earth … this will whereby we move forward, where pleasure leads each one of us” (II.251). A living being with a mind is shaped not only by the atoms it is born with but by its own choices, training, and history: “however much training gives some of them an equal culture, yet it leaves those first traces of the nature of the mind of each,” and yet reason can reduce those traces so far “that nothing hinders us from living a life worthy of the gods” (Lucretius III.307–322). A man who chooses is not a fixed pattern of atoms that must repeat itself wherever the same atoms meet. He is in part his own making. Lucretius adds that even our very atoms gathered again in the same order “would not concern us at all, when once the remembrance of our former selves were snapped in twain” (III.847–851). A person includes the unbroken chain of his own life and memory, which no copy shares.

So the Epicurean answer to “Are any two things exactly alike?” is neither the Stoic “never” nor Democritus’ “everywhere, down to men.” It runs like this:

  • Kinds recur throughout the infinite universe, because the kinds of atoms are limited and the atoms of each kind are unlimited.
  • What is possible is limited, for the same reason. There are no men of iron anywhere.
  • Simple concrete things, such as two eggs, two coins from one die, or two grains, may well be alike in every way the senses can detect. Lucretius himself notes that grains of corn usually show “some difference between their forms” (II.371), but nothing in Epicurean physics forbids close or exact likeness among such things, and the skeptics’ eggs do not trouble the Epicurean at all.
  • Living beings with minds are individuals, shaped in part by their own free choices. Epicurus did not teach that they have exact doubles living the same lives.

Most important for the theory of knowledge, Epicurus’ canonics does not depend on any of these answers. The Stoic needed every hair and grain to be unique, because his whole theory of knowledge rested on the mark. The Epicurean needed nothing of the kind. A sensation does not claim to identify an individual. Whether the man across the street is Publius or his twin is a question for opinion, and it is settled the ordinary way, by further evidence.

The vocabulary of the debate over marks, a feature communis, “common” to true and false, against one proprius, “peculiar” to the truth, is the same vocabulary the Stoics use about signs in On Methods of Inference. A “common” sign “can exist whether the unperceived object exists or not,” while a “particular” sign exists only with the hidden thing (Col. I). De Lacy’s note on that passage cites Cicero’s Lucullus 33–34, the same discussion of the nota “not common to what is true and false.” In effect the Stoics carried their demand for a mark from perception over into inference. A valid sign had to be like a kataleptic impression: one that could not exist unless its hidden object existed, certified by a test that ruled out every possibility of error. Contraposition was that test.

The Epicureans rejected the whole apparatus, and they did it at the root. Epicurus did not divide impressions into a privileged class that carries a mark of truth and a remainder that may be false. He held that every presentation is true. Sextus reports that the presentation, “which he also terms ‘evidence’” (Greek enargeia, which Cicero renders perspicuitas or evidentia, II.17), “is constantly true,” for “if a presentation is termed ‘true’ whenever it arises from a real object and in accord with that real object, and every presentation arises from a real presented object and in accord with that object, then every presentation is necessarily true” (Sextus I.203–205). The Stoic Dionysius in On Methods of Inference makes the same report as a charge: the Epicureans “consider that every appearance and notion is necessarily true” (Col. VII). A fragment of the book says of the images that reach us that they “cannot be false” (Fragment VI).

If every presentation is true, no nota is needed to pick out the true ones. Falsehood does not lie in the visum at all. It lies in the opinio, the opinion the mind adds to it: “falsehood and error always lie in the addition of opinion with regard to what is waiting to be confirmed or not contradicted, and then is not confirmed or is contradicted” (Letter to Herodotus 50). Cicero states Epicurus’ position in the same Latin terms: Epicurus taught that it is the wise man’s task “opinionem a perspicuitate seiungere”, “to separate opinion from evident knowledge” (Academic Questions II.45). The twin who looks like Publius produces a true impression of what is there. The mistake is the opinion “This is Publius,” and it is corrected by further evidence, not by inspecting the impression for a hidden mark.

Cicero shows the positions side by side. Epicurus’ maxim, as Cicero reports it, is “If anything that appears to the senses be false, then nothing can be perceived” (Academic Questions II.101). Epicurus “brings the matter down to this point, that if once in a man’s life one of his senses has decided wrongly, none of them is ever to be trusted” (II.79). The Stoics and their ally Antiochus conceded that the senses can deliver false impressions, and then had to find a nota that would sort the true from the false. The Academics accepted the same concession and concluded that no such mark exists. Epicurus refused the concession. That is why the skeptics’ twins and eggs could defeat the Stoic mark and could not touch Epicurus. Lucretius gives the Epicurean answer to the skeptic in a line built on the same Latin root as nota: “invenies primis ab sensibus esse creatam notitiem veri neque sensus posse refelli”, “you will find that the concept of the true is begotten first from the senses, and that the senses cannot be gainsaid” (Lucretius IV.478). The notitia of truth, the concept of truth itself, comes from the senses. It is not a badge that some impressions wear and others lack.

Truth, for Epicurus, does not mean that every impression tells us everything. DeWitt insists on this: the distant view of a square tower is true to the distance, but “it fails to reveal the whole truth about the tower,” and the value of sensations as evidence varies “all the way from totality to zero” (DeWitt, Epicurus and His Philosophy, 1954, p. 139). The Stoics too urged a man to draw close, to rub his eyes, to seek a clearer view (Sextus I.258). The difference is in what the closer view does. For the Stoic, the clear impression certifies itself by its mark. For the Epicurean, the closer view is more evidence, and the opinion is confirmed when the evidence agrees and nothing contradicts it.

And the Epicureans needed no divine craftsman to guarantee their senses. Nature has no plan and no opinions, so it has no reason to deceive. Our impressions are true because they are caused by what is really there, the images and effluences that actually reach the senses (Letter to Herodotus 51), not because a divinus artifex sealed every leaf, hair, and grain with a pattern of its own.

So the answer to the question whether signs are the same as marks is no, though they are connected. The Epicureans rejected marks entirely, because they rejected the premise that some impressions are false. They accepted signs, but they denied that a sign needs anything like a mark to be trusted. A sign is established the same way every opinion is: by observing what holds without exception across the variety of experience, and by finding nothing that conflicts.

The surviving text begins in the middle of the Stoic case, and the objections are strong ones.

Pomegranates everywhere. If analogy were valid, “it is possible to say that since there are pomegranates and figs in our experience, they exist everywhere” (Col. Ia). Things that exist here need not exist elsewhere.

Unique cases. The magnet is the only stone that draws iron, amber the only thing that attracts chaff, and the square of four the only square whose perimeter equals its area. “How, then, can we say that there is not some exceptional race of men who do not die when the heart has been pierced?” (Cols. I–II).

Rare cases. Experience itself contains freaks: a dwarf in Alexandria “with a colossal head that could be beaten with a hammer,” a person in Epidaurus “who was married as a maiden and then later became a man,” a giant in Crete, the pygmies Antony brought from Syria. “For certainly some things in this world deviate from the normal” (Col. II).

Analogy assumes what it proves. “Since men in our experience are mortal, all men are” is valid only “if we assume that men who are in unperceived places are similar to men in our experience in all respects” (Cols. II–III). The conclusion is smuggled into the premise.

The atoms. “According to the method of analogy one ought to infer that since all bodies in our experience have color, and atoms are bodies, atoms have color” (Col. V). By their own method, the Epicureans should give the atoms color and make them perishable.

Identity or similarity? If the hidden thing is identical with the visible one, there is nothing hidden to infer. If it is only similar, “how shall we be able to say that the unperceived object does not, by virtue of its distinct nature, differ from the apparent object from which we make the inference?” (Col. VI).

Never enough cases. The Epicureans say analogy holds when nothing conflicts. “For how will it be possible to establish that no appearance or previously demonstrated fact conflicts?” (Col. VIII). We can never examine every case.

Behind all of these lies one demand. The Stoics wanted necessity, a connection that could not have been otherwise, and they held that only the formal test could deliver it: “if we proceed by the method of contraposition we shall establish that it alone has certainty” (Col. VII).

The Epicureans’ central reply turns the Stoic position upside down. “If the method of analogy is not cogent, the method of contraposition is not cogent” (Col. VIII).

Their example is Epicurus’ own argument for the void. “If there is motion, there is void” looks like a perfect case for the formal test: imagine the void away, and motion goes with it. But how do we know that motion needs void? “We do not establish in any other way than by the method of analogy, by proving that it is impossible for motion to take place without void. Establishing by induction all the constant conditions for things that are moved within our experience, apart from which we see nothing moved, we judge by analogy that all objects that are moved are moved in every case under these conditions” (Cols. VIII–IX). So the formal test, “which is entirely confirmed by and through analogy,” is “not by itself cogent.”

The last section of the book makes the same point about smoke and fire, the standard ancient example of a sign. To assert “If there is or has been no fire, there is no smoke,” “we contend that always in all cases smoke has been observed to be given off by fire” (Col. XXXVI). The connection is not discovered by thinking about the words “smoke” and “fire.” It is learned by watching fires.

This is the deepest point in the book. The Stoics treated logic as a source of knowledge prior to experience, which experience can only confirm. The Epicureans answered that the connections logic works with were learned from experience in the first place. As De Lacy summarizes the fourth section, “contraposition can be reduced to analogy, but not vice versa” (Cols. XXXI–XXXII). And the Stoics refute themselves every time they argue. Their objections from variation and unique cases are analogies from experience (Cols. XXIX–XXX). They know that there are “no Centaurs or Pans” only “by the method of analogy” (Col. XXXI). “Those who reject the method of inference by analogy make all unperceived objects unsignified. For there is no other correct method of inference besides this” (Cols. XXX–XXXI).

Inconceivability: Necessity Learned From Experience

Section titled “Inconceivability: Necessity Learned From Experience”

The Epicureans did not give up on necessity. They located it somewhere else. A sign is valid, they said, “whenever it is inconceivable that the appearance exists or is such as it is and the unperceived object does not exist … just as it is impossible to conceive that Epicurus is a man and Metrodorus not a man” (Col. XIV). “The best test of the hypothetical proposition and the particular sign is established whenever we are not able to conceive that the first exists and the second does not” (Col. XXXIII).

This is not an appeal to pure reason. We cannot conceive a man who is not mortal because we have never met one, and because nothing we have ever met points that way. Bromius states the principle behind it: “an object deprived of all similarity with appearance is inconceivable” (Col. XXI). What we can and cannot conceive is itself a product of experience. Necessity is learned, not assumed.

The final section of the book takes up a Stoic offer that sounds harmless. The Epicureans could make their argument valid, the Stoics said, by adding “in so far as”: “Since the men in our experience, in so far as and according as they are men, are mortal, Men everywhere are mortal” (Cols. III–IV). Mortality would then belong to the definition of man, and the inference would pass the formal test.

The anonymous Epicurean answers by distinguishing four meanings of “as such,” “according as,” and “in so far as” (Cols. XXXIII–XXXIV). One of them is “that this is the particular definition of that, i.e., this is the anticipation of that.” The word “anticipation” here is prolēpsis, the Epicurean preconcept. And on every one of the four meanings, the Epicurean gives the same answer about how we know the connection: “we Epicureans take this to be necessarily connected with that from the fact that this has been observed to be a property of that in all cases that we have come upon, and because we have observed many varied living creatures of the same genus who have differences in all other respects from each other, but who all share in certain common qualities” (Col. XXXV).

This fits what Epicurus taught about meaning. We do not begin with definitions. We begin with examples met again and again and set against each other. The preconcept is “a recollection of what has often been presented from without,” and it lets us ask of a distant shape, “Is that standing yonder a horse or a cow?” (DL X.33). Epicurus “does not hold with giving a definition of the thing in question,” as Cicero complained (On Ends II.4), and Torquatus asked, “Who is there who does not know what pleasure is? Who needs a definition to assist him to understand it?” (On Ends II.6). The Stoic who wants mortality to be part of the definition of man has it backwards. Whatever goes into our notion of man was put there by the men we have met. A definition that is not drawn from examples is not knowledge. It is a decision about words.

The Stoic charge about pomegranates assumes that the Epicureans infer from any likeness at all. They did not. The inference “ought not to be made from any chance common quality to any other, but from that quality which does not admit a spark to the contrary or exhibit even a breath in opposition to appearances” (Col. XIII). Beheaded men everywhere stay dead, because men everywhere we have met are alike in that respect. But plants differ from region to region “in odor, color, form, size, and other characteristics,” so their presence here says nothing about their presence elsewhere.

The same care answers the objection about the atoms. “Bodies in our experience are destructible not as bodies, but in so far as they partake of a nature opposed to the corporeal and non-resistant. Similarly, bodies in our experience have color, but not as bodies … Indeed, bodies in the dark have no color, yet still are bodies” (Cols. XVII–XVIII). What experience shows to go with bodies as bodies is resistance and weight. “As a consequence of the fact that bodies in our experience rebound, we judge that atoms are heavy and solid, for we cannot conceive that objects in our experience are of this kind and that similar ones (unperceived) do not have this quality” (Col. XXXVII). Color and perishing come and go with circumstances, so the analogy does not carry them to the atoms. Lucretius makes the same argument at length when he explains why the atoms have no color (Lucretius II.730–841).

“Inferences should be made from objects that are most closely related and from those that are as similar as possible” (Col. XVIII): from men to men, from a species to its nearest kin, from bodies to bodies. Knowing which similarities count is itself something learned by watching which qualities stay constant across all the variety of experience and which do not.

The Stoic objections from unique and rare cases draw the most revealing Epicurean answers, and they connect On Methods of Inference to the Epicurean doctrine of limits.

First, uniqueness does not break analogy. It supports it. “No one of our school denies such peculiarities, and yet the method of analogy does not become invalid because one class of stones draws iron … If other stones were similar to or identical with the stones that draw iron, and some drew iron, and some did not, the method of analogy would be destroyed. Since this does not happen … in no way is the method of analogy shaken” (Cols. XIV–XV). The magnet is unique, but it is uniquely constant. Every magnet draws iron, so we may infer that magnets “in the infinite universes” do too (Col. XV).

Second, variation is real, but it has limits. Bromius answers the Stoic question whether things beyond our experience might vary without end: “Let no one oppose us or be so stubborn as to question whether there are, beyond such differences, men with a nature of iron, who go through walls as we go through air” (Cols. XXI–XXII). There will be differences among men elsewhere “analogous to the variations in our experience,” but never “a peculiarity of such a kind, that it will appear possible in any way that a man be invulnerable or that we see any finite object not bounded by another object” (Col. XXIV). There is “a determinate limit to what men can eat,” so “we shall not admit that there are men who eat hay, easily digest it, and are nourished” (Col. XXVI).

This is the same principle that runs through Epicurean physics. Lucretius argues that the atoms come in a limited number of shapes, which is why nature has fixed boundaries: “a fixed limit to things marks the extreme on either side,” as heat and cold are “marked off at either end by the twin points” (Lucretius II.507–521). Because the kinds of atoms are limited while their number is infinite, every kind of thing we see has its counterparts throughout the infinite universe, and nothing beyond the limits of those kinds can arise anywhere. That doctrine, and the principle of isonomia (equal distribution) that goes with it, are set out in our article The Power of Infinity. On Methods of Inference shows the same principle at work in the theory of knowledge. The limits we observe in nature are what make it safe to reason from here to everywhere.

The phrase “any finite object not bounded by another object” points back to Epicurus’ own definition of a limit: “that which is bounded has an extreme point: and the extreme point is seen against something else” (Letter to Herodotus 41). A limit is known by contrast, by setting one thing against what lies beyond it. That is why the Epicureans could treat unique cases as evidence rather than embarrassment. The magnet’s peculiarity is visible only because it is set against every other stone, and the contrast is constant. Meaning, and the limits of what we can expect, both come from examples set against each other.

Phillip De Lacy later traced this theme through the whole of Epicurean philosophy in his essay “Limit and Variation in the Epicurean Philosophy” (Phoenix, 1969), arguing that limit is the Epicurean answer to the old demand for fixed standards in a world of change. On Methods of Inference is one of the clearest places to see why.

The Stoic dilemma, identity or similarity, also gets a direct answer. Bromius says that “sometimes we shall infer from identical objects, whenever there is a distinction in accompanying circumstances,” and “sometimes we shall infer from non-identical objects, in so far as they share the same community of similar attributes” (Col. XXII). An object seen in one place and an identical object unseen in another differ in their circumstances, and that is enough. Two different objects can still share the qualities the inference is about.

Bromius’ example is a striking one. We may reason about divine beings from the living beings we know, “when we consider that nothing prevents god from being similar in body to man, since man alone of living beings in our experience is capable of thought. For god cannot be conceived apart from thought” (Col. XXII). And against the Stoic charge that Epicurean gods and atoms are unlike anything we see, the Epicureans answer that “the gods and the first elements of things appear to be indestructible and unborn, since this is a condition of their being such as they are, and since inductive inference from appearances has proved it” (Col. XXVII).

The Stoic dilemma has the same shape as the Academic argument about twins and identical eggs, which Cicero reports in the Academic Questions. There the Epicurean answer is that sameness is a judgment the mind makes, not something a sensation claims to report. Our article Twins, Eggs, and Duplicate Worlds follows that debate.

It is often said that the Epicureans claimed infallible and necessary knowledge of the unseen. That description gives away the very premise they rejected. The Stoics, and the Academic skeptics who opposed them, shared one assumption: knowledge must be certain beyond all possible error, or it is not knowledge at all. The Stoic concluded that only his wise man knows anything. The skeptic concluded that no one does. Both accepted the all-or-nothing demand.

On Methods of Inference does not accept it. The Epicureans say plainly that knowledge comes in degrees and that the degree depends on the evidence. “Those who make declarations about anything whatsoever sometimes affirm confidently a universal proposition, and sometimes use probability. Certainty and probability are both derived from the observation of appearances” (Col. XXV). In some matters, “it will be sufficient for us to base our belief on probability … just as we do in regard to what is learned from trial; for example, that we shall be safe sailing in summer, since we have had experience of favorable winds in that season” (Col. VII). “Sometimes when you chance on one instance you say that a thing is of such and such a character, and often you use two cases and sometimes more” (Col. XXVI). And when the evidence is strong and nothing conflicts, the inference deserves full confidence: “nor mistrust the facts proved through apparent objects according to analogy, but trust them just as one trusts the facts from which the inference is made” (Fragment II).

We need not examine every case to know. “We must consider many homogeneous and varied ones, so that from our experience of them and from the accounts of history concerning them we may take the inseparable constituent of each of them” (Cols. XX–XXI). We may use the observations of others, too: “Do those who have never been there doubt that Crete and Sicily are islands?” (Col. XXXII). And those who reason badly are not proof that reasoning fails. They “are corrected by the facts in individual cases” (Col. XXV), as the man who decides from his own neighbors that all men are white is corrected by the Ethiopians (Col. XXX).

Norman DeWitt made the same point about the senses themselves. Epicurus held that sensations are true, in that they faithfully report what reaches us, but DeWitt insists that this is not a claim that they cannot mislead. Those who stretch “all sensations are true” to mean that the senses are infallible “are confusing the concept of truth with the concept of value.” A distant view of a square tower is true to the distance, but “it fails to reveal the whole truth about the tower.” The value of sensations as evidence, “apart from their truth, varied all the way from totality to zero” (DeWitt, Epicurus and His Philosophy, 1954, p. 139). That is why Epicurus urged attention to all the evidence and checked one sensation by another.

This is the same rejection of the all-or-nothing error that we trace in Almost Is Not Nothing. Knowledge, like pleasure and like progress toward wisdom, is real at every degree. It is not worthless because it falls short of an impossible perfection.

The Epicurean condition, that nothing may conflict with the inference, is not a blank check, as Dionysius claimed. It is the core of Epicurus’ own test for opinions. “Falsehood and error always lie in the addition of opinion with regard to what is waiting to be confirmed or not contradicted, and then is not confirmed or is contradicted” (Letter to Herodotus 50). Diogenes Laertius gives the standard example: an opinion is “awaiting confirmation” when we wait “to come near the tower and see how it looks to the near view” (DL X.34). Sextus Empiricus, an opponent, reports the Epicurean rule fully: true opinions are those “which testify for, and not against, the evidence of sense” (Sextus, Against the Logicians I.211).

On Methods of Inference applies that rule to hidden things. Analogy is sound “with this condition, that no other appearance or previously demonstrated fact conflicts with the inference” (Col. XXXII). It applies the rule even to Epicurus’ own doctrines: “it is not sufficient to admit the very slight swerves of the atoms because of chance and free will; but we must also prove that this swerve does not in any way conflict with any other appearance” (Col. XXXVI). And it warns against using the wrong test for the wrong kind of thing. We must not “think that we know objects temporally unperceived by the absence of conflicting evidence” (Fragment IV). The tower can be approached, so we must go and look. Only for things unseen by nature, like the void, does the absence of conflicting evidence settle the question. Our article Is The Wise Man Guided By Opinion? follows these tests in more detail.

The same rule governs the famous Epicurean practice of multiple explanations. Epicurus would have liked to know the true cause of every event in the sky, but where the evidence cannot decide between explanations, “when one accepts one theory and rejects another, which harmonizes as well with the phenomenon, it is obvious that he altogether leaves the path of scientific inquiry and has recourse to myth” (DL X.87). Every explanation must fit the facts, and “only superstition must be excluded” (DL X.104). For basic physics, where only one account fits the evidence, as with atoms and void, the Epicureans held that account to be known.

There is one more reason the Epicureans could trust analogy, and it sets them apart from the Stoics at the root. The Stoics believed the world was governed by a divine reason that made everything for a purpose, and they looked to that providence to guarantee that the mind’s impressions could be trusted. The Epicureans needed no such guarantee. Nature has no plan and no opinions, so it has no reason to deceive us. Its uniformity comes from the limited kinds of atoms and the fixed limits they impose, not from the will of a designer. Everyone who infers well, says the fourth Epicurean teacher, “considers it impossible that the nature of things and their generation from each other should be inconsistent with appearances. Thus he expounds the origin of the cosmos” (Col. XXXIII).

That is how the theory of knowledge in On Methods of Inference connects to the first question of Epicurean philosophy, how the world came to be. Epicurus came to philosophy, Diogenes Laertius reports, because his teachers could not explain where Hesiod’s Chaos came from (DL X.2). His answer, worlds formed by atoms without a designer, rests on inference from what we see to what we cannot, and On Methods of Inference is the fullest surviving defense of that method. Our article The Epicurean Alternative To Divine Creation follows the physics.

The Stoics were not the only critics. Sextus Empiricus, writing two centuries after Philodemus, accepted signs that remind us of things observed together, like smoke and fire: “we ourselves also are of the same mind and infer fire from smoke, and a previous wound from a scar, and death from a previous puncture of the heart” (Sextus, Against the Logicians II.157). What he rejected were signs that claim to reveal what is never seen, like the void. He compared those who inquire into hidden things “to men shooting at a mark in the dark” (II.325), and he singled out Epicurus’ argument “If motion exists, void exists; but in fact motion exists; therefore void exists” as a disputed opinion (II.329).

On Methods of Inference had already answered him. The skeptic who infers death from a pierced heart is using analogy from experience, as Philodemus’ Stoics were when they argued from unique cases. The line between “reminding” signs and “revealing” signs is not a line between two methods. It is the same method applied to things that can be checked directly and things that can be checked only by the absence of conflicting evidence. Cicero, who claimed the Academic skeptic’s right to follow only “what seems to me to be probable” (On Duties II.7–8), never answered the Epicurean question: from what did he learn what is probable, if not from the evidence of the senses? Epicurus had put it in one line: “If you fight against all sensations, you will have no standard by which to judge even those of them which you say are false” (Principal Doctrine 23). Our article Absurdities Of The Skeptics shows what happened to skeptics who tried to live by their own doctrine.

The fragments of On Methods of Inference list “the criteria of the unperceived: namely, perception, anticipations, mental perceptions, and feelings” (Fragment I). That is four criteria, not the three of Epicurus’ Canon. Diogenes Laertius records the same addition: “in The Canon Epicurus says that the tests of truth are the sensations and concepts (prolepsis = pre-concept) and the feelings; the Epicureans add to these the intuitive apprehensions of the mind” (DL X.31).

DeWitt explains why Epicurus himself did not rank these apprehensions as criteria. When we recognize that the thing we taste is honey, “a secondary reaction in the form of a recognition involving intelligence has taken place,” which Epicurus called “a fantastic perception of the intelligence.” These “were not given the rank of criteria by Epicurus,” DeWitt writes, because the intelligence has already come into play. “It is on record, however, that later Epicureans did so” (DeWitt, p. 140). The fourth criterion of the fragments is a later school development, and On Methods of Inference is a document of that later school. Its value lies in how faithfully it applies Epicurus’ method, not in any claim that every term in it comes from Epicurus.

On Methods of Inference records the moment when Epicurean philosophy met its most serious logical challenge and answered it. The Stoics demanded a certainty that only formal logic could give. The Epicureans answered that the connections logic uses were learned by watching the world, that the limits we observe in nature make it safe to reason from the seen to the unseen, that unique cases are evidence rather than exceptions, and that knowledge is real at every degree the evidence supports. They insisted on the same tests Epicurus set for every opinion: confirmation where we can look, and no conflicting evidence where we cannot.

These answers are not museum pieces. Nearly every science that reasons from observed regularities to unobserved cases, from the laboratory to the stars, uses the method Philodemus defended. The Epicureans’ claim that experience is the source of the rules we reason by, and not the other way around, is as important now as it was when Zeno of Sidon taught it at Athens. Epicurus started from what every living thing perceives and feels. On Methods of Inference shows that the same starting point can carry us, step by careful step, all the way to the atoms and the infinite universe.