Is "First, Do No Harm" Actually in the Hippocratic Oath?

Somebody will correct you on this eventually, in the mild, precise way people correct each other about things they are sure of. First, do no harm, the phrase everyone says medicine is built on, is not in the Hippocratic Oath. That much is easy to check, and true. What is harder to check, and what almost everyone who corrects you will also get wrong, is where it actually came from.

The claim

The claim is that medicine rests on a single Latin imperative, primum non nocere, and that the imperative reaches us from Hippocrates by way of the oath many medical schools still administer at graduation. It is repeated in commencement addresses, in bioethics textbooks, in journalism about pharmaceutical scandals, and in courtroom cross-examinations of physicians. It is repeated so consistently that questioning it feels like a small act of vandalism.

The evidence

The evidence assembles in three pieces.

First, the Oath does not contain the phrase, nor any Greek from which it could be translated. It contains no word on the βλάπτ- root at all; its harm vocabulary is δήλησις, ἀδικίη, φθορή. The closest it comes, in W. H. S. Jones's 1923 Loeb translation, is "I will abstain from all intentional wrong-doing and harm." Harvard Medical School's own health publication concedes the point in a piece titled, without irony, "First, do no harm," noting that although the phrase is attributed to Hippocrates, "it isn't a part of the Hippocratic Oath at all."

Second, the closest source in the Hippocratic Corpus is not the Oath but Epidemics, Book I: As to diseases, make a habit of two things — to help, or at least to do no harm. The Greek is ἀσκεῖν περὶ τὰ νοσήματα δύο, ὠφελεῖν ἢ μὴ βλάπτειν. This is what the tradition is quietly paraphrasing. But the passage orders its priorities in the opposite direction from the Latin. Hippocrates puts help first, and not-harming second, as the fallback when help is not possible.

That reading is not merely mine. Galen, commenting on this exact line, glosses the priority explicitly: μάλιστα, most of all, benefit the sick; εἰ δὲ μή, failing that, at least do not harm them. Littré, Jones and Jouanna all translate it that way. It is not unanimous — Francis Adams renders the disjunction flat, and elsewhere in the Corpus the same pair appears as a flat taxonomy. But the fallback reading has ancient authority behind it, and the Latin dictum reverses the ranking. The reversal, not the vocabulary, is the thing that needs explaining.

A note that belongs in an essay about careless citation: the Greek that circulates in medical-ethics papers, commas and all, is Littré's 1840 French edition, not "the original Greek." Littré himself argued in his apparatus that the sentence's first word is a gloss that leaked out of Galen's commentary, and proposed deleting it. Nobody quoting the original mentions that its opening word is under suspicion from the man who established the text.

Third, the Latin phrase appears nowhere in the ancient record. Not in Hippocrates. Not reported in Galen, who also wrote in Greek. Not in the Corpus. And not, so far as any searchable corpus shows, anywhere in Latin before the nineteenth century: the exact string returns zero hits across all 17,098 Latin texts of Corpus Corporum, spanning the Patrologia Latina, the CSEL, the Digest, Gratian, the Corpus Thomisticum and the neo-Latin collections, from 148 BC to 1902.

The scholarly disagreement

For more than twenty years the standard citation has been Cedric M. Smith's 2005 paper in the Journal of Clinical Pharmacology, which identified the earliest print appearance as an 1860 book by the English physician Thomas Inman. Inman attributed the maxim to Thomas Sydenham, who never wrote it down.

Daniel Sokol stated the same position in the BMJ in 2013, citing Inman directly, in a commissioned column rather than a peer-reviewed paper. He granted that the maxim is a useful injunction against overtreatment, then called it "a crude piece of advice" and, in his closing line, "a flawed dictum," proposing "first do no net harm" in its place. That is the version the medical literature still cites, and the version Wikipedia still carries today.

Then there is the medical historian Thomas Morris, who has an inconvenient counter-argument. Far more of the medical literature is searchable now than when Smith wrote, and it took Morris very little time to find the phrase, in Latin, in print, decades before Inman. His earliest instance is the French surgeon Alexis Boyer, household surgeon to Napoleon, using primùm non nocere in his Traité des maladies chirurgicales, in a chapter on cataract surgery, warning against operating too early: by operating at that stage, one would act against the first precept of the art.

Morris dates that to 1814. It is 1816. The passage sits in tome cinquième, page 511, and the title page of that volume says 1816. The Traité is an eleven-volume set running from 1814 to 1826, and the libraries stamp the set's opening year on every volume — the Wellcome catalogue labels volumes four, five and six all "1814," which cannot be true of all three. Morris took a catalogue date for a volume date, and everyone repeating him has inherited it.

That is a small correction with an awkward implication. The newest date in a debate about dates is itself a cataloguing artifact.

Morris finds more instances before Inman — Dance in 1829, Deval and Heyfelder in 1838, von Gützeit's typhus treatise in 1842 with the phrase on the title page, Leroy d'Etiolles in 1845 — and notes these are only a handful of many. A title page is where you print what everyone in your field already recognizes. So Smith is not wrong that the Latin is non-Hippocratic. He is wrong by nearly half a century about when it entered print.

The deeper trail

Morris did not stop at Boyer. He found the formula in earlier company, and eventually in a place nobody expects: an early Christian writer.

Around 310 AD, Lactantius, a North African rhetorician who became tutor to Constantine's son, composed the Divinae Institutiones. He later abridged the work himself as the Epitome divinarum institutionum. In chapter 55 of the Epitome — numbered 60 in the older editions, which is where the familiar citation comes from — on the duties of justice, he wrote: Primum est enim non nocere, proximum prodesse. The first thing is to do no harm, the next is to do good.

Get the citation right, because almost nobody does. The sentence is in the Epitome, not the Divine Institutes, which is seven books numbered per book and has no chapter 60 at all.

Three things complicate the story that this is where medicine got its maxim, and they make it better rather than worse.

Morris is not the first to notice the parallel. Jean Jouanna, the leading Hippocratic philologist, published the identification in the standard Budé edition of Epidemics I and III in 2016, and Marquis Berrey restated it in English four months before Morris's post. The identification belongs to a classicist, printed in the one book a historian of this question ought to open.

The famous wording did not exist in print before 1712. The antithetical reading depends on a manuscript found at Turin in 1711 and published in Paris the year after. Every earlier printed edition reads Primus est enim non nocere proximo, sed prodesse — the first thing is not to harm one's neighbour, but to benefit him. No antithesis, no detachable maxim. The eighteenth-century editor John Davis warned the celebrated reading might be a copyist's repair. The argument that generations of medically-minded readers met this crisp sentence in their Lactantius does not survive contact with the editions they would have been reading.

And no one who used the phrase is known to have cited Lactantius. Boyer credits les anciens. Henry Maunsell, whose 1847 paediatric textbook gives the closest wording, credits "a physician of old" who had it written over his consulting-room door, which Lactantius, not being a physician, did not have. Morris himself hedges. Only his last sentence overreaches.

Lactantius is the closest surviving ancient Latin wording. He is a parallel, not a pedigree.

What nobody has looked at

There is a better candidate, sitting in plain sight in a literature nobody in this debate reads: the Latin literature of justice.

The two-limb formula is a commonplace of Roman ethics and Roman law long before it is anything medical. Cicero gives justice two offices. Ulpian's three precepts of law, which every educated European met in the opening pages of Justinian's Institutes, include alterum non laedere, harm no one else. And in 1673 Samuel Pufendorf, in the ethics textbook half of Europe's universities assigned, ranked the absolute duties and wrote that primum locum obtinet, ne quis alterum laedat — first place is held by the duty that no one harm another.

That is a Latin "first, do no harm," ranked first, in a book every university student read, one hundred and forty-three years before Boyer reached for the phrase in a surgical treatise.

It also explains the one thing the Lactantius story cannot. If the maxim descends from the Epidemics, why is it a ranking when the Greek it supposedly translates is not one? Because the ranking did not come from medicine. It came from the law, where ordering duties is the whole enterprise. A French surgeon trained in the classics, reaching for a Latin tag to justify surgical restraint, was not misremembering Hippocrates. He was reaching into the nearest available drawer, and the nearest drawer was jurisprudence.

I cannot prove that, and I am not going to pretend otherwise. Nobody has done the work. That is rather the point.

The judgment

The phrase is not in the Hippocratic Oath. Every source agrees. If you have said it was, the sentence was wrong, and that is settled.

The Epidemics passage does exist, and it does put helping first. When the tradition claims Hippocratic authority for the ordering do no harm first, it is inverting the Hippocratic hierarchy rather than transmitting it.

And the transmission history is not what medicine tells itself. It is not a straight line from Greek into modern practice. It is a Latin ethical commonplace with at least two ancestries, one Christian and one juristic, absorbed into medical writing in the eighteenth and nineteenth centuries, popularized by a French surgeon in 1816, and attributed afterward to Hippocrates or Sydenham because those were the authorities the field wanted it to carry.

Two different questions get asked when we ask where primum non nocere comes from. Did Hippocrates write this phrase? No, unambiguously. Where did the underlying moral commitment come from? A tangled multilingual inheritance in which Hippocrates is one node, Lactantius another, Roman jurisprudence a third, and none of them wrote what medicine now says they wrote.

Ancient authority is one of the most valuable currencies a profession can print. The AI ethics community is printing it right now; so is the wellness industry. Somewhere, someone with a whiteboard is turning a policy preference into a Latin motto to give it the authority of a rule Hippocrates never wrote.

Twenty-one years after Smith and seven after Morris, Wikipedia still gives 1860. The medical literature has not corrected itself. The two people who have actually done the work — Jouanna in a French critical edition, Richard Suss in an unreviewed 2024 preprint — are invisible to the citation chain that keeps repeating the date. Most physicians never learn any of this, and the ones who learn the first half, that it is not in the Oath, usually stop there, satisfied, and pass along a date that has been wrong since 2005.

The trail is longer than anyone told us. Most of what is confidently said about its beginning is wrong.