David Butterfield on the joys and the challenges involved in editing Lucretius

Lucretius’ De rerum natura (“On the nature of everything”), written in the mid-first century BC, was the most ambitious poem ever written in Latin. Quite possibly, I may even venture, it is the most ambitious poem ever written in world literature. Yes, that is quite the claim. But I will only say that if there are other poems that seek not just to describe the totality of the physical world, and to explore the universe in which it spins, but also to remove all the pains and problems of human existence by showing us the ultimate rationale by which everything around us, and indeed within us, happens, then I will be happy to compare notes. The good news, for me at least, is that this particular claim isn’t what I am seeking to defend in this piece. Instead, as someone currently rounding off the new Oxford Classical Text of Lucretius’ epic poem, I want to introduce you to the sorts of problems that confront the editor seeking to present this universe-roving, time-transcending, life-solving poem to readers in the 21st century.

The task of editing Lucretius De rerum natura presents a situation that mixes the good and the bad, the reassuring and the unnerving. To set out the scope of the challenge: we have six books, of nearly 7,500 hexameter verses, which aim to convert the Roman reader to the philosophy of the Hellenistic hedonistic atomist Epicurus (341–271 BC). The logical progress of the work – moving from the microscopic movement of atoms, through the workings of our human bodies, up to the macroscopic survey of subterraneal regions and the heavens above – is carefully graduated, and grindingly relentless. Despite being a technical poem based on physics, it is filled with the poet’s passion, and often attains to stunning beauty.

Sounds good. So what’s the bad news? Well, Lucretius’ autograph copy of this poem is long lost, presumably disappearing around the time of his death in the 50s BC. More problematically, not a single manuscript from the Roman world survives. In fact, there is little evidence that the poem found any philosophical success among Roman readers. Instead, its chief influence, evidenced most notably in the Augustan poets Virgil and Horace, was literary. Yet no surviving author from the first hundred years after Lucretius’ death cited the work directly, and no Roman testimony survives praising the DRN as an intellectual venture. On the contrary, the Christian Church Fathers, most prominently Lactantius, mocked and vilified the work as the dangerously misconceived undertaking of a delirious author, and malicious tales of his suicide by a love potion tainted his later biography. 

In order to survive, Lucretius’ poem passed though a remarkable bottle-neck – one so narrow that of all of the hundreds, perhaps thousands, of rolls and codices of his work that once existed in antiquity seemingly only one both survived the fall of the Roman Empire and managed to be transcribed in a medieval scriptorium. A copy of this fortunate survivor, made perhaps in Northern Italy but now also lost, found its way to the court of Charlemagne (740s–814, ruled from 768). And it is to his transformational rule, and the decades that followed, that our three oldest surviving manuscripts of the poem date. Two of these (called the “Oblongus” and “Quadratus” for their rectangular and square shapes relative to each other) survive effectively complete, and are now in the University Library of Leiden, Holland. Along with the fragments of the third manuscript, spread between Copenhagen and Vienna, these are the only texts that a Lucretian editor needs to look at. While it’s true that there also exist more than 50 other manuscripts of De rerum natura, mostly written in fifteenth-century Italy, these are ultimately derived from the three ninth-century manuscripts just described. They are therefore of no independent value for preserving the authentic text of what the poet wrote but merely serve as a repertory of the conjectures made by enterprising Renaissance readers.

The opening of Lucretius’ poem, as preserved in a 15th-century manuscript (Vatican Library Lat. 1569, f.1r). The first line of poetry begins with the cherubic A and descends down the inscription-like capitals, before the second line appears on the cherub-suspended parchment.

Some mixed news now. The Carolingian manuscripts are not three independent books: they in fact share the mistakes and omissions that prove they all descend from a single lost manuscript – the 8th-century book mentioned above, which we term the “archetype”. That book had no special status per se, but because it is the oldest shared ancestor whose text we can reconstruct by comparing the readings of the manuscripts that still do exist, it bears this technical name. But here’s the rub: this archetype was written some 750 years after Lucretius’ death, and yet it is the closest concrete representative of the poem’s text as put to papyrus in the 50s BC.

stemma codicum, showing the relationship of the three surviving witnesses that an editor must consult.

Two problems loom large. The first is that over the course of three quarters of a millennium Lucretius’ poem was copied unknowably many times in sequence: perhaps ten or more copies stand between his own copy and the archetype itself. Since humanum est errare – it’s only human to make mistakes – it is inevitable that at every stage an uncertain number of errors were made in uncertain places during the manual process of transcription. Leaving aside this kind of accidental human harm, each of these individual manuscripts could also have suffered physical damage (from water spillages, tears, dislocation of leaves) or deliberate change (wilful deletion, alteration or addition) that affected the text permanently before the next copy could be made. At each stage, then, truthful readings of the poem could be lost – perhaps irredeemably.

So, if we take the Codex Oblongus – a remarkably  lavish manuscript of c.800 AD, and the oldest witness to Lucretius’ poem – we find a text that is seriously corrupt: it contains well over a thousand mistakes. Many are trivial, some are tricky, and some are utterly baffling—more on these soon.

A page from the oldest manuscript of Lucretius, the Codex Oblongus (f.10r), in which verses 1.364-8 at the bottom of the page have been added by the Irish scholar-monk Dungal, employed in Charlemagne’s court.

The second problem may be even more alarming. When Lucretius died in 55 or 54 BC, his poem seems not to have been known to the general public. Indeed, there is plenty of evidence that the great work was unfinished when he died, and was presumably put into circulation by a friend. We may dismiss St Jerome’s claim that (the!) Cicero was Lucretius’ editor, not least because he loathed Epicureanism. But presumably some other associate was then faced with a mixture of papyrus rolls (one for each book?) along with various loose leaves awaiting incorporation into a polished manuscript version. Close observation of the state of the poem in our ninth-century manuscripts does in fact reveal that earlier and later versions of the same passage survive in tandem, one of which Lucretius intended to replace with the other. Presumably, however, the first scribe was commissioned to make a complete copy, including all the verses that survived. Had Lucretius lived to polish the poem in full, he doubtless would have refined these awkward parallel versions into a single fair-copy form.

Now the question confronts the editor: is some apparent irregularity or oddity in the argument or phrasing of the poem a result of damage it has suffered during phases of copying, or is it rather the result of the poet himself having not lived to smooth it out? Panning out the perspective, what in fact is the purpose of an editor? To recreate the poem as Lucretius left it on his death, or somehow to forge the finished poem that he never quite rounded off? (Cards on the table, I restrict myself to the former.)

So, to summarise: we can reconstruct the reading of the archetype pretty mechanically by comparing our oldest manuscripts. But what do we do when the reading we have in front of us simply can’t be right? Well in such a scenario we have to move to “conjecture”, where we use our knowledge of Lucretius language, style, metre and method of argument, and of Epicurean philosophy in general, and of common forms of scribal error across all Latin manuscripts, to “emend” the text to something that seems more probably Lucretian. At all times we are not simply seeking the “best” possible reading – a question-begging venture that could risk emending the author himself. Instead we are asking whether it is more probable that (i) conjecture X is Lucretian and (ii) has been corrupted to Y over 750 years of transmission, than that (iii) reading Y, odd or surprising or flat though it is, nevertheless is what Lucretius chose to wrote instead.

That is enough theory for now. So let’s take a direct look at some examples of error in Lucretius’ poem. As we do, we may bear in mind that many thousands of minds have stared at these problems over many a century: the text has been carefully worked on since its rediscovery and introduction to Italian society in the early 15th century. And even 600 years before that it was in the Oblongus, where he makes very many successful corrections – alongside some misguided guesses.

Let us begin with some larger-scale kinds of error linked with the exercise of copying. A scribe wrote by looking at his exemplar (the book he was copying), remembering a block of text (be it a word, phrase, or whole verse), and then turning to copy it in his apograph (the book he was in the slowand careful course of making). This sounds simple enough – but it was all too possible for the scribe to look back to the wrong part of the text, usually by skipping further down a column of text and therefore unwittingly missing out a number of lines. If he, or his supervisor in the monastic scriptorium, failed to notice the omission, then we are left with a gap in our text.

Lucretius poem certainly has a number of gaps – or lacunae. Major editors typically mark around 20 gaps (usually of unknowable size!) throughout the work. In some cases these are easy to spot, as sentences break off mid-flow and we have to assume an accidental loss in transmission. In other instances we have to infer that gaps are Lucretius’ own omissions. In Book 5, for instance, the poet promises that he will prove ‘with extensive discussion’ (sermone largo probabo, 155), his account of the nature of the gods, a promise which the poem never fulfils, even in part. That such a treatment has been accidentally lost in its entirety is a drastic supposition; his never having lived to write it such a (difficult!) section is in fact an easier inference.

If we return to the scribe engaged in the act of copying, it is possible that some monk whose eves have skipped accidentally down the page realises his error. Rather than cross out the accidentally anticipated verses (something medieval scribes were very keen not to do) or start a new leaf afresh (something that medieval bursars really did not want done, given the great expense of parchment), the scribe would insert the omitted text, either later in the column of text itself, or (if the error was discovered appreciably later) in the margin. Small signs would indicate where the verses should properly be read by diligent readers; but if these were not noticed or not understood by a later scribe, the random transposition would become permanent – lurking undetected until an editor comes along and says, “Hang on a minute.”

At times an evident transposition comes from physical damage. In Book 4, for instance, the verses traditionally numbered 299–322 occur in the wrong place, before verses 323–47 which manifestly must precede them. What had happened was that, in the archetype, the leaf which contained verses 299 through to 347 of Book 4 had become disbound, like a loose page in any modern printed book, and was inserted into the codex the wrong way round. The result was that a scribe naturally copied the text of the original verso (back) of that leaf before coming to the recto (front): and so it appears in our manuscripts, all of which are copied from that damaged book. In fact, we have other evidence to show that four further leaves fell out of this same archetype in the next few decades and were haphazardly inserted at the back of the book: Lucretius was well-thumbed indeed! It was this combination of errors that allowed the great German scholar Karl Lachmann to reconstruct with remarkable clarity the actual pagination of the Lucretian archetype, a manuscript that regularly had 26 lines per page, and a total of 152 leaves. Although we can never see this book, for it surely perished many centuries ago (but do please check your attic now and then!), we can through such physical transpositions reconstruct its material form with a very high degree of probability.

We may now turn to textual errors, beginning with the commonest instance – where the manuscript reading is ungrammatical Latin, or non-existent Latin. Changing the case (e.g. lumina to lumine) or gender (e.g. magnum to magnam) of a word is very often a necessary and certain emendation, and such corrections have been accepted by all for many centuries. But others are much more complex and debate continues – at times fiercely! Even in 2022, considerable work remains to be done on the text, and anyone is welcome to come join the discussion.

What about this problem, then? Towards the beginning of his account of early mankind in Book 5, Lucretius describes how humans drank natural sources of water:

at sedare sitim fluuii fontesque uocabant,

ut nunc montibus e magnis decursus aquai

†claricitatiate† sitientia saecla ferarum. (5.945-7)

But streams and springs invited them to slake their thirst, just as now the downpour of water from great mountains ???? the thirsty races of beasts.

947 claricitatiate Ω : claricitat late Bosius : clarigitat late Lachmann : clarus citat late Forbiger : largus citat late Ritschl

claricitatiate, the string of nonsense opening line 947, remains stubbornly unsolved. While Simeon Bosius was surely right to see the adverb late (“far and wide”) at the close of this monstrous univerbation, it remains a challenge to restore a verb to the clause. The unattested verb claricito is itself a most implausible frequentative form (< clarico) or compound (< clari- + cito); no better is Lachmann’s neologism clarigito (< clarigo). Numerous scholars have been attracted to Forbiger’s clarus citat but the metrical licence of ignoring the presence of –s (“sigmatic ecthlipsis”) is not found elsewhere in the first two feet of Lucretius’ verses, so it is highly risky to claim that the poet needed to invoke it here alone. It should also be added that no instance of citare is attested in Lucretius’ poem, notwithstanding his common employment of ciere.

To find a different route forward, I have suggested clare accit, “loudly summons,” thereby introducing a verb which, though rare, Lucretius himself uses 50 lines later: posterius tremulas super ulcera taetra tenentes / palmas horriferis accibant uocibus Orcum (“afterwards they held trembling hands over their foul sores and called on Orcus with terrible cries,” 995-6). It also serves as a natural companion to uocabant in 945, unifying the comparison between past and present. The asyndetic combination of two adverbs (clare... late, “loudly far and wide”) is not unlucretian: see elsewhere in Book 5 pariter tantundem at 494, and temere incassum frustra at 1002.

In making a claim that this is what Lucretius, an editor should suggest how the gobbledegook arose in our manuscripts. Although we can only guess, perhaps -cc- in clareaccit became -ct- (as happens commonly in early minuscule scripts); clareactit is very close to, and indeed a simple anagram of, clarecitat, so it could have become misread or rewritten as the transmitted claricitat for both metrical and phonetic reasons. Other possibilities of course exist. And perhaps Ad Familiares readers have better ideas as to what Lucretius wrote?

Let us move to an instance where the transmitted text doesn’t scan – i.e. it fails to fit the rigorous constraints of the dactylic hexameter. If we turn to the last page of the poem we find the chaotic scene of Athenians burying their dead in the terrible plague of 430 BC. Lines 1281–2 run as follows:

            quisque suum pro re ??? maestus humabat;

            multaque ??? subita et paupertas horrida suasit

Each griever buried his own according to his circumstances ????; and sudden ??? and poverty encouraged many terrible things.

The sense is clear enough in both lines, but each verse is metrically deficient to make up the full hexameter. In 1281 something could have been lost before suum, or after it, or after pro re. There is no way of knowing the precise location, let alone what has been lost. So how to supplement the line when the sense is relatively complete? Various suggestions have been suggested: crudely supplying a noun or participle to go with suum, such as cognatum (“relative”, Avancius) or compostum (“person laid out”, Lachmann), even though these give a decidedly rough rhythm. Or providing a second adverbial/participial phrase before pro re, such as propere (“hurriedly”, Housman), or after it, such as et pro tempore (“and as time allowed”, Martin Ferguson Smith). My own suggestion is the Lucretian adverb properanter, “hurriedly” (used at 5.300; he does not use propere). Perhaps a scribe – himself hurrying to the end of the work just a few lines later – looked back not to pro re but to properanter in his exemplar, and unwittingly moved onto the next work maestus, thus causing the omission?

            The second line, 1282, presents a simpler problem: a monosyllabic feminine noun is required to stand before subita, “sudden”, which agrees with it. 15th-century readers suggested res (in the sense of “need”) or vis (“force”), and in the 19th century Jacob Bernays suggested mors (“death”). What suggestion, as a counterpart to “poverty” in causing “many terrible things” for plague-stricken, grief-racked mourners, seems most compelling to you? Are there in fact some other unsuggested and untested options? Could we have lost instead a word beginning with a vowel and shaped short-long, which stands after multaque? There are possibilities, but I leave it with you...

A depiction of the Plague at Athens, appearing as an engraving before the beginning of De rerum natura Book 6 in the lavish edition of Jacob Tonson (London, 1712).

At times the corruption is quite baffling. Consider this instance from near the end of Book 5, when the achievements of mankind are nearing Lucretius’ present day (1442–3):

                        tum mare ueliuolis florebat propter odores

                        auxilia ac socios iam pacto foedere habebant...

Then the sea was in bloom with sail-fliers [=boats] on account of smells; [people] now had allegiances and allies under formal treaty...

The close of 1442 is a famous locus uexatissimus, or “stinker of a problem”, described by the great Lucretian commentator Cyril Bailey as “perhaps the most desperate textual crux in the poem”. propter odores (“on account of smells”) makes no sense in context; it can hardly be an offhand condensed mention of the incipient perfume trade. Most curiously, though, the words propter odores happen to close a verse three books earlier, at 2.417, but there propter is an adverb (“nearby”), not a preposition. Well over 50 emendations of the latter half of 1442 have been made, and none has found general favour. You will be relieved to learn that this is not the place for me to object to each of them individually. Instead, I will set out my own thinking: I believe that propter odores is a corruption of words that bore a similar appearance; that ueliuolis is used in the Lucretian fashion as a florid substantive – “sail-fliers” = boats; and that the presence of propter odores at 2.417 is one of those freak coincidences that can happen in a poem of 7,500 verses. My own suggestion is propter et omnes: propter is employed adverbially, “nearby” (as indeed it is at 2.417); omnes (“all”), or possibly urbes (“cities”), provides an explicit subject for the verb habebant of 1443. Lucretius thus emphasises, after his discussion of human success on land, that the seas nearby these peoples were then enjoying much maritime activity and societal cooperation.

In all of these cases there is no scope for a certain restoration. So an editor should mark his or her or another’s conjecture clearly as an editorial supplement or intervention, lest the reader fail to see that the words being read cannot be guaranted as Lucretius’ own. Textual criticism is only a partial science: while one may be certain in diagnosing where there is an error in the manuscript tradition, which necessarily requires correction, a significant minority of suggestions about what should replace these errors can be nothing more than good guesses. And you, as the reader, have as much right as anybody else – past, present or future – to say “No, that won’t quite do. Here is a better idea.”

Dr David Butterfield is Senior Lecturer in Classics and Fellow and Director of Studies at Queens’ College, Cambridge. His work on The Early Textual Transmission of Lucretius’ De rerum natura was published by Cambridge University Press in 2013 and his new Oxford Classical Text of Lucretius is eagerly awaited. He also edits the Classics website Antigone.