Essay · by Frater Custos · 12 minutes

Who invented Samhain

In 1786 a British military engineer wrote that the Irish worshipped a god of the dead called Saman. That god exists in no Irish text. Two hundred and forty years later it is still the opening line of almost everything written about Halloween.

Charles Vallancey was an English military engineer sent to Ireland to carry out surveys, and he ended up devoting his life to proving that Irish descended from Phoenician. In his Vindication of the Ancient History of Ireland, printed in Dublin in 1786, he writes that «the Irish deity Saman» was the judge of the souls of the dead, and likens him to the Persian Asuman, «the angel of death».

There is not one Irish text that says anything of the sort. There is no god Samhain, in no manuscript, in no gloss, in no annal. Vallancey invented him, along with the rest of his Phoeno-Scythian system, which philology demolished in the nineteenth century and which nobody has defended for a hundred and fifty years.

And yet that sentence has crossed two centuries intact. You find it, almost word for word, in the articles that explain every 31 October that Halloween is «the ancient Celtic festival of the dead». It is worth going to see what the sources actually contain, because the answer is neither «it's Celtic» nor «it's American»: it is more interesting than either.

What the Irish texts say

The passage everyone cites, when they cite anything, is in Tochmarc Emire — «The Wooing of Emer» — in the version translated by Kuno Meyer in 1888. It is a glossary of the four quarter-days of the year. It says this:

«Samuin, that is, summer's end. For formerly the year was divided in two, summer from Beltaine (the first of May), and winter from Samuin to Beltaine.»

And just below, under the next heading:

«Beltine, a favourable fire. For the druids used to make two fires with great incantations, and to drive the cattle between them against the plagues, every year.»

It is worth reading twice. The fire rite is attached to Beltine, the first of May. Not to Samuin. The text cited as evidence for the Samhain bonfires, at the exact point where it speaks of bonfires, is speaking about a different date.

The other texts confirm a picture that is coherent and sober. In Serglige Con Culainn, preserved in Lebor na hUidre, the Ulaid hold a week-long assembly at Samain — the three days before, the day itself, the three days after — with «games, markets and splendours», contests, feasting, the display of trophies. In Cath Crinna Samain is a judicial assembly. In Lebor Gabála it is the day tribute is collected. In the Scéla Conchobuir attendance is compulsory and there is a penalty for absence.

Samain, in the sources, is a deadline: the cut between summer and winter for a pastoral economy, and with it the fair, the court, the feast, the accounts that fall due. At that date the barrier with the síde — the dwellings of the Otherworld — is loosened, and this much is attested in several tales.

Four things that are not in the sources

Listing the absences is more useful than listing the presences, because it is on the absences that everything else has been built.

On the etymology, for completeness: the medieval Irish glossators explained samain as sam «summer» + fuin «end». That is evidence about what they believed, not about what the word meant. Modern Celtic scholarship — Matasović in his etymological dictionary of Proto-Celtic, Delamarre for Gaulish — reconstructs rather *samoni-, «gathering, assembly», with parallels in Old Norse saman and Sanskrit samāna. That is: the word probably means exactly what the texts describe. A gathering.

The Christian date, and why the story runs backwards

The other half of the legend says the Church put All Saints on the first of November in order to cover Samhain. Here the documentation exists, and it can be followed.

The soundest and earliest evidence is a letter. In the year 800, before 19 March, Alcuin of York writes to Bishop Arno of Salzburg: «Kalendis Novembris solemnitas omnium sanctorum» — on the kalends of November the solemnity of all the saints — and urges him to celebrate it every year on that date, with a three-day fast, prayers, masses and alms. By 800 the first of November feast is already normal in the Anglo-Frankish world.

The famous decree of Gregory IV and Louis the Pious of 835, by contrast, needs handling with care: the source is Sigebert of Gembloux, writing in the twelfth century. We do not have a papal document to that effect. It is a tradition recorded three hundred years after the fact.

And now the point that reverses the argument. The Irish churches celebrated All Saints on 20 April. The Félire Óengusso, the martyrology of Óengus of Tallaght, dated to 828-833, records the first of November but also «all the saints of Europe» on 20 April and «all the saints of Africa» on 20 December; and the Irish martyrology of the late eighth century puts the feast in April.

The documentary trail, in short, runs from Rome towards Northumbria and Francia, and if anything reaches Ireland from the continent, not the other way round. Why the date moved from 13 May — the dedication of the Pantheon to the martyrs, in 609 or 610 — to the first of November, scholars do not know: the New Catholic Encyclopedia writes that on the question there is «no scholarly consensus» and that the reason «remains unexplained». What can be said is that no text of Gregory III, Gregory IV, Alcuin, Ado or Sigebert names Samain. Ever. The idea of a deliberate overlay does not come from a source: it comes from Frazer.

Two professors, and a conditional lost along the way

The version we all know is born between 1886 and 1914, in two English universities.

John Rhŷs, at Oxford, argues in the Hibbert Lectures of 1886 that the first of November was the Celtic new year. The evidence he offers, and repeats in Celtic Folklore in 1901, is this: on the Isle of Man, on that date, «the tenancies of land terminate, and servants go to their places. In other words, it is the beginning of a new year». From a nineteenth-century Manx contractual custom he derives a pre-Christian new year for all the Celts. Ronald Hutton calls this evidence, in a single word, flimsy.

James Frazer, at Cambridge, adds the dead in The Golden Bough. He is worth quoting exactly, because the most interesting thing is how he writes:

«Halloween seems to have been of old the time of year when the souls of the departed were supposed to revisit their old homes…» — and shortly after — «it was perhaps a natural thought that the approach of winter should drive the poor shivering hungry ghosts… to the shelter of the cottage»; and on the new year: «we may with some probability infer that they dated the year from Halloween».

Seems to have been. It was perhaps. We may with some probability infer. Frazer conjectures, and declares it in every line. What happened next is that popularisation removed the conditionals and kept the sentences. A century of repetition turned three cautious hypotheses into a fact that is now in school textbooks.

❦

What is documented — and it is all recent

There is a remarkable quantity of material on Halloween. It is just two or three centuries old, not two thousand years.

The divination games. The first dense, dated description is a poem: Halloween by Robert Burns, written in 1785 and printed at Kilmarnock in 1786 — twenty-eight stanzas listing the Ayrshire rites analytically, with the author's own notes. Before him, John Mayne in 1780. In 1833 Daniel Maclise paints Snap-Apple Night after attending a Halloween party at Blarney, County Cork.

The turnip lantern. The Oxford English Dictionary dates to 1837 the first use of jack-o'-lantern for a lantern made from a hollowed vegetable; before that, from 1658, the word means the will-o'-the-wisp. Hutton documents turnips and beets hollowed and carved as lanterns «common in parts of Ireland and the Scottish Highlands in the nineteenth century». There is no ancient or medieval attestation whatsoever.

Guising. The first detailed account is from November 1895, in «Frank Leslie's Popular Monthly»: in Scotland, masqueraders carrying hollowed turnip lanterns visit houses and are rewarded with cakes, fruit and money. In 1897 the «Arbroath Herald» hosts a readers' controversy over guising, «a silly and reprehensible practice».

The pumpkin. The earliest printed attestation I know is from 1 November 1866, in the «Daily News» of Kingston, Ontario: pumpkins with «transparent heads and faces», lit by the usual two inches of tallow candle. The replacement of the turnip by the pumpkin is an agricultural matter — the pumpkin is American, large and easy to carve — not a symbolic one.

«Trick or treat». Here a date that circulates everywhere, including in the Canadian press, needs correcting: it is not 1927 at Blackie, Alberta. Lexicographical research has pushed it back ten years. The earliest known printed attestation is from 1 November 1917, in the «Sault Daily Star» of Sault Ste. Marie, Ontario: «Tricks or treats, you could hear the gangs call out, and if the householder produced the coin for the treats his property was safe from attack». From there the formula crosses the Canadian prairies through the 1920s and appears in the United States on 1 November 1928, in the «Bay City Daily Times» of Michigan.

Note something that usually escapes attention: the earliest attestations of both guising and «trick or treat» are Canadian, and they come from areas of heavy Scottish settlement. The chain «Irish emigrants hand Halloween to America» is a reasonable and universally repeated reconstruction, but I have not found it documented point by point in any source. I report it for what it is.

What this does not prove

Having got this far, the temptation is to turn the thing over and conclude that Halloween «has nothing to do with the Celts», that it is an American festival, that the whole business is a fabrication. That would be the symmetrical error, and it would be equally false.

Ronald Hutton, who is the historian who took the dossier apart in 1996 and whom everybody cites to say it is all false, actually says three distinct things, and popularisation always flattens at least one of them:

First: yes. «There seems no doubt that the opening of November was the time of a major pagan festival», celebrated at the least in all those parts of the British Isles with a pastoral economy. Samain existed. It is not a nineteenth-century invention.

Second: but. «There is no evidence that it was connected with the dead, and no proof that it opened the year — but it was certainly a time when supernatural forces were especially to be guarded against or propitiated.»

Third: and anyway no. The Christian date does not appear to have been chosen to cover it, because the trail runs the other way.

Whoever cites Hutton to say «Halloween has nothing to do with the Celts» misrepresents him as much as whoever ignores him.

And there is a live disagreement, which should be reported. Tok Thompson, an anthropologist at the University of Southern California, argues for continuity: that the celebration of the dead at Halloween is «one of the oldest surviving traditions», and that it is on Samhain eve «that the dead are at their most powerful». The divergence is not only about facts: it is disciplinary. Historians ask for a documentary chain and, not finding one, stop; anthropologists reason about continuity of practice and of structure, and do not regard the written chain as the only admissible proof. These are two different ways of deciding what counts as evidence, and neither of them is stupid.

What can be said with the sources in hand, and what this atlas is willing to defend, is narrower and less satisfying than either headline version: at the beginning of November, in the Gaelic areas with a pastoral economy, a festival called Samain is documented from the early Middle Ages — a seasonal cut, an assembly, a fair, a feast, and a time when the síde open. It is not documented as a festival of the dead, nor as a new year. The Christian date does not appear to have been chosen to replace it. And the festival celebrated today — the lanterns, the costumes, the door-to-door, the formula — is documented from the eighteenth century in Scotland and Ireland, and takes its present shape between Ontario and the United States in the early twentieth.

Which leaves the hardest question open, and I leave it open: whether between the Samain of the manuscripts and the carved turnip of Ayrshire there is a thread the documents do not record, or whether they are two different things that meet on the same date for reasons of the agricultural calendar. I don't know, and nobody does. Anyone who says they know, in either direction, is adding something to the sources.