Public register

Corrections

The Method promises that corrections are made in public and with a date. This is the page where they are made. Every substantial change ends up here: a figure put right, a source corrected, a part of the site rebuilt. Reports from readers are credited, to those who wish it.

2 October 2026

«I colpi contro il muro» («The knocks against the wall»), written on 1 October and published today: a dozen entries in this atlas describe a creature that is never seen and only heard, and the repertory of sounds is almost the same from Pliny’s Athens to the Barbagia. For now the essay exists in Italian only.
Fifteen new entries, chosen on the numbers. I measured real demand — the visits to the corresponding Wikipedia entries over the last twelve months, language by language — and crossed it with the gaps in the atlas. Out came fourteen entries from the world and one Italian: Krampus (Austria), Chupacabras (Puerto Rico), Baba Yaga and Leshy (Russia), Will-o’-the-wisp and Selkie (United Kingdom), Jersey Devil (United States), Krasue (Thailand), Manananggal (Philippines), Dullahan (Ireland), Pocong (Indonesia), El Silbón (Venezuela), El Cadejo (Guatemala), Mula sem cabeça (Brazil) and La Befana. Three are grade A — manananggal, dullahan and the Befana — eleven are grade B. The atlas goes from 135 to 150 entries and from 37 to 42 countries: Austria, Puerto Rico, Thailand, Venezuela and Guatemala come in today, each with its own detail map. These fifteen too go out without a plate, like the eight of September.
Three chains of citation taken apart. Manananggal is not a colonial word: the Spanish accounts have magtatangal — Plasencia, 1589, describes him «without his head or entrails» on Catanduanes — and the modern form first appears in the Noceda-Sanlúcar vocabulary of 1860. The dullahan reaches print in 1802, with Vallancey, and there he still has his head: he drags a chain, nothing more; the beheading arrives with Croker in 1828, eight years after Washington Irving’s headless horseman. And the Befana: in the old texts the word means the feast, not the old woman — Berni’s «befana» is a straw puppet hung at the window — while the «pagan goddess» has an author and a date, De Gubernatis 1878.
The detail maps now update themselves. Until now the geometry of each country’s Italian map was a frozen copy: that is why the Klabautermann, added on 22 September, did not show on the map of Germany, and Spring-heeled Jack did not show on the map of the United Kingdom. Both versions now read the same source, and the two missing points are back where they belong.

26 September 2026

Eight new entries, written on 22 September and published today. Two Tuscan — La Rificolona and La tregenda — four Italian — The witches of Valcamonica, 1518-1521, The night of the witches of St John, I Beati Paoli and S’accabadora — and two from the world — Spring-heeled Jack and Klabautermann. Six are grade A; the Beati Paoli and the accabadora are grade C, and deliberately so: they are the two questions with the most searches and the fewest documents, and the entry exists to say it. The atlas goes from 127 to 135 entries. These eight do not have a plate yet: they go out without one, and each will get its plate when the plate exists.
The English search page is no longer in the sitemap. The Italian page cerca has always been kept out of Google’s index: it is a tool, not a page to be found. Its English twin en/search/, however, was indexable and was declared in the sitemap — meaning Google could show a searcher an empty form instead of an entry. The two now behave the same way. The declared pages go down accordingly from 431 to 430.
The Aosta Valley appeared twice in the atlas filters. The entry Om salvàrech wrote the name of the region with a straight apostrophe, the other two with a typographic one: for the filter they were two different regions, and neither showed all three entries. There is now a single spelling, in Italian and in English.

21 September 2026

The ICF-42 in English. It was the last public page left only in Italian. The scoring engine is the same, line by line; only the words change. Two choices need stating. English sessions are kept in the browser separately from Italian ones: the translated questions have not yet been compared with the originals, and putting a test and a retest taken in two languages into the same series would confuse the translation with a change in the person answering. For the same reason the English report carries the mark (EN), so whoever receives it knows which version it came from. The «Delete my data» button now deletes both series, whichever of the two pages it is pressed on.

20 September 2026

«The exhumation record», the first of the three new texts: six graves opened between 1700 and 2004 — Míkonos, Kisiljevo, Medveđa, Exeter, Marotinu de Sus — read in the records of those who were present and not in the summaries made of them afterwards. The bodies resemble one another down to the detail, and the detail almost always has a taphonomic explanation. The word vampire, by contrast, enters the European languages only in the 1730s: it is younger than almost everything it claims to explain. The text was finished on 15 September and sat unpublished for five days through an oversight of mine: the date at the foot of the entry is the date of writing, this one is the date of publication.
«Who invented Samhain». The chain of attributions that holds up almost everything one reads about Halloween, taken one link at a time. The Irish god of the dead «Saman» appears in no manuscript: he is coined by Charles Vallancey in 1786. The druids’ bonfires that everyone places at Samain are, in the Tochmarc Emire, at Beltine. The «Celtic new year» is an inference by John Rhŷs (1886) drawn from a land-letting custom on the Isle of Man. The «feast of the dead» is Frazer, and Frazer writes it in the conditional. And the thesis that 1 November was put there to cover Samain runs backwards: the date is attested by Alcuin in 800, while the Irish churches kept All Saints on 20 April. At the foot is the chronology of the attestations, including the first trick or treat in print — 1 November 1917, «Sault Daily Star», Ontario — which is ten years earlier than the date usually repeated.
«The dead who bring the presents»: the Italian tradition of 2 November, which is documented and bears no resemblance at all to the one claimed when people say we had Halloween already. The Sicilian dead who hide the gifts for the children are in Pitrè, 1881; the Sardinian is animeddas are alive and collected, but attested in print only in the twentieth century; the capuzzelle of the Fontanelle are not a rite of 2 November, and the link with that date is later than the prohibition of 1969. There are also two things the atlas does not endorse: the Roman Lemuria fall in May, five months away, and the story of the lumere as an ancient Po Valley pumpkin goes back to an oral enquiry of December 1998 never verified elsewhere.
«The bridge the devil did not build», on Borgo a Mozzano. The legend of the pact with the devil for the Ponte della Maddalena holds up the largest Halloween festival in Italy, and has no written attestation earlier than well into the twentieth century: it is a local case of the international tale type ATU 1191, documented in northern Europe between 1816 and 1912, and in the main comparative index the Tuscan variant has for its source a tourist magazine of 2020. It did not turn up in the Lucchese collections of Nieri and Giannini. The piece also carries two corrections: the festival is of 1994 and not of 1993 — so say the official site, a press release of the Regional Council of Tuscany and one of the three founders — and the «fifteen thousand visitors» are a rough estimate by the mayor, at a free-entry event that nobody can count.
«Who wrote the Tuscan tradition». Pan co’ santi, ossi di morto, seme dei morti, il ben d’i morti: I tried to trace who first recorded them, as one does with Pitrè for Sicily. The chain ends on four food blogs published between 2014 and 2023, which are the notes of the page the whole web copies from. It is an empty-handed report, and it says something about the shape of our information. It also takes in Versilia, where nothing documented turns up for 31 October, and the correct reading of a serious book — Baldini and Bellosi, Einaudi 2006 — that is cited to say things it does not say.
«Lucca Comics is not a Halloween festival». The event has straddled All Saints since at least 1972, that is twenty years before Halloween reached Italy; the closing press release of 2025 names neither Halloween nor All Saints, and 31 October in the official programmes is «the third day». With one correction: the first edition is of 1965 at Bordighera, not of 1966, which is the year of the move to Lucca.
The English is complete, and there is now a test that proves it. English versions now exist for the thirty-seven country detail maps, the search, the coordinates, the privacy page, the contact page, the feed and — from today — this register and the versions page. The interactive maps are bilingual too: the same code, two data files, and the filters by region, cultural area and country work with the English names. The 404 page is a single page and recognises the path: under /en/ it switches to English on its own, navigation included, with no need for server rules. Two things stay in Italian: the ICF-42 instrument, which is an application of its own, and the traffic log, which is not indexed and is a service page of mine.

I have written an automated test (collauda.js): it opens the pages with a real browser and checks, in both languages, that the maps draw their points, that the filters filter, that the search finds and counts in the right language, that every page declares its pair, that the 404 switches language and that no footer sends the reader to the wrong one. A hundred and three checks. The check I already had — controlla.py — looks at the files; this one looks at the behaviour, which is something files cannot tell you.
A fifth error, found by the test I had just written. At the foot of every English page the «Last updated» line pointed to the Italian versions page instead of the English one. I had not seen it because that link is generated by the template shared by both languages, and the eye never goes to a page footer. It was found by check number one hundred, written half an hour earlier. That is the reason a test is worth writing even when you are convinced everything is fine.
The atlas is bilingual throughout. From today an English version exists for all a hundred and twenty-seven entries, the ten Currents studies, the eighteen source checks, all the texts, the thirty-seven country detail maps and the service pages: Italy, the world, the pseudobiblia, the currents, the source checks, the essays, the coordinates, the search, the privacy page and the feed. Four pages stay in Italian, by choice: the corrections register, the versions, the traffic register and the ICF-42 tool. They are the house record, and a record is kept in the language one works in.

For the entries it is not a translation of the pages but of the data: the generator builds the two languages from the same source, so that a future correction is applied once only. The interactive maps are bilingual too — same code, two data files — and the filters by region, cultural area and country work in English with the English names. Where the primary source was already in English — Callaway, Hodgson, the Habsburg reports, Lovecraft, Jung — the original wording has been restored, not translated back. On the Currents the rule imposed on the translation was symmetry: in English Evola, Scaligero and the racial question in Steiner have a different rhetorical centre of gravity, and the rendering was not to shift by a single degree in either direction. I reread the sensitive passages by hand. Each pair of pages declares the other with hreflang, and only where the counterpart really exists.
Four errors found while translating. The entry for the Sibilla appenninica opened with «a cave that opens near the summit ridge at about 2.» — the sentence broke off inside the number; it now reads «at about 2,150 metres, and inside a kingdom». The each-uisge entry began with «katharine Briggs defines it…», with the folklorist’s name in lower case through an automatism of the generator. In the sources of the study on Jung the Libro Rosso was cited with the Italian title and the American publisher: the Norton of 2009 is The Red Book. Liber Novus. And in the text on the Devil’s Bridge, published this morning, a sentence of mine was simply ungrammatical — «I have not found zero nineteenth-century attestations: I have found exactly zero» — and said the opposite of what I meant; it now reads «I have not found few nineteenth-century attestations». The first three I had not seen in months, the fourth I had written two hours earlier: all of them were found by the work of translation, which forces you to reread one by one the lines that in your own language the eye runs over without seeing. It is the best argument in favour of the translation, and it was not the one for which I had begun it.
Corrected the button that downloads the reminder in .ics format in the ICF-42 tool: it used a data: address, which on the iPhone opened the file instead of saving it, and it did not fold long lines to the seventy-five-octet limit the format requires. It now builds the file in memory and folds the lines: it opens in the calendar from a phone as well.

21 August 2026

Added the versions page: the state of the site at each update, in numbers, with the date of the last change now visible at the foot of every page. And above all the sitemap at last declares the lastmod field for each of the hundred and seventy-six addresses. The rule I have chosen is the only honest one: the generator compares the fingerprint of the content with that of the previous pass and changes the date only if the text has really changed. The dates start from today, because before this the site kept no track of the change page by page and to declare earlier ones would have been to invent them.
The atlas is illustrated throughout: a hundred and nine entries out of a hundred and ten have a plate. The only one left without is yee naaldlooshii, and it is left without by declared choice — the entry deliberately describes the ethnographic context and not the narrative content, because in the Diné community speaking of it is an active taboo: to illustrate it would contradict that choice. The emptiness is the content.
Three of the last plates have a subject that is itself a position taken. The amabie is not depicted as a living creature but as the printed sheet of 1846, which is the only thing of her that exists. The Sibilla shows the vestibule of the cave with the stone seats and the carved signatures, and not the interior, because Antoine de la Sale stopped there and nobody has looked beyond. The Berta does not stage the legend but the object: the stone head built into the wall on via de’ Cerretani, which is the only documented thing.

20 August 2026

Three new studies in the Currents. Jung and esotericism: alchemy as an archive and not as a technique, the Red Book published only in 2009, what synchronicity says and what it does not say — and the 1934 page in the Zentralblatt, which has to be reported because it exists. Julius Evola: the works with their dates, the books on race with their mastheads, and two false accusations — he did not sign the Manifesto of Race of 1938 and he was not a member of the SS, who in fact judged him unusable. Taoism: why the distinction between «philosophical Taoism» and «religious Taoism» is a Western construction, what the Guodian finds of 1993 and the Mawangdui finds of 1973 changed, and why internal alchemy arises in part from the fact that the elixir was killing people.
Four new Italian entries, and two of them are documented archival cases rather than creatures. Gostanza da Libbiano, a healer tried at San Miniato in 1594: she was not burned, she was acquitted, and the man who had her freed was the inquisitor of Florence, against the local episcopal vicar — the exact reversal of the cliché. The benandanti of Friuli, who went out in spirit on the Thursday of the Ember Days to fight the witches with fennel: Catholic liturgical calendar, not surviving paganism, and not one of them ended at the stake. Then la Berta of Florence, where the topography dismantles the legend, and la Sibilla appenninica, where two inverted dates change everything.
Three Japanese yōkai, chosen because in each of them the received version can be disproved with a date. The tengu did not have a long nose for seven hundred years: it was a bird of prey, and the nose is the beak humanised late. The yuki-onna loses some hundred and eighty years of antiquity as soon as one looks at the bibliographical record, and the version everyone knows is Lafcadio Hearn’s, 1904, from a source never traced. The amabie, finally, is not an ancient protective spirit: it is a sheet of 1846, then silence for a hundred and seventy-four years — and even its name is probably a misprint.

19 August 2026

Sixty-four new plates in a single day, and with them the illustrated atlas stops being only Italian: haint blue, nahual, aluxo’ob, huldufólk, bean sídhe, zashiki-warashi, namahage, kitsune, Rübezahl, jiangshi, shuigui, buda and Näkki, and then kelpie, Cŵn Annwn, huldra, Nøkk and fossegrim, domovoj, Południca, churel, Àbíkú, La Llorona and the Brown Mountain Lights, and finally tokoloshe, rusalka, Santa Compaña, Mothman, Bell Witch, granny women, rougarou, wampus cat, the New England vampire panic the Philadelphia Experiment and thirteen Lovecraftian pseudobiblia. Illustrated entries go from twenty-eight to ninety-one out of a hundred and three. The drawing briefs were written from the text of the entries, and each of them also declared what was not to be drawn — Rübezahl is not the kindly giant of children’s books, the tortoise-like kappa is a late image, the dokkaebi has no horns.
Before that: ʿIfrīt, Qarīn and Naddāha for Egypt, huli jing for China, dokkaebi for Korea, kuntilanak for Indonesia, wendigo for Canada, Curupira and Saci-Pererê for Brazil. For the dokkaebi the plate is a walking chest, not the horned imp: it is the image the Korean sources authorise, and it is the exact opposite of the one the Source checks disprove.
One plate was discarded instead of published, and it is worth saying so: it depicted the jiangshi in mid-air, with the talisman on its forehead and the robe of a Qing official. It is the image of Hong Kong cinema — exactly the one this site declares to be without Qing sources. To illustrate an entry with the figure the entry disproves would have been the one point at which the atlas did what it reproaches others for.
Earlier the same day the first four had arrived: the bunyip surfacing from a billabong, the Min Min lights hanging over the plain in front of a cart, the yowie in the eucalyptus forest with the footprints on the path, and the boto cor-de-rosa in a white suit at the river festival. Until yesterday the illustrations were twenty-eight and they were all and only the Italian ones: anyone arriving at an entry from the wider world found text alone. Now there are forty-one.
Each of the seventeen source checks now has an address of its own. They were all on a single page, and the reason for separating them is not cosmetic: someone looking up whether «nosferatu» is a Romanian word asks one question, and a page that deals with seventeen of them is not the answer to any of the seventeen. The overview page remains as an index — the question and the outcome in one line, not the full text, so as not to have eighteen pages saying the same thing.
Removed from the overview page the FAQPage markup, which served to make the expanded questions appear in search results. Now that the index reports only the outcome and not the whole answer, that markup would describe content that is not on the page — and that is exactly what the guidelines forbid. Better to lose it than to declare something false to a search engine.
Added a service page, the traffic register, where I transcribe by hand the site’s visit figures with the date and a note of what was not measured. It is not indexed and it is not in the menu: it serves to stop me telling myself that the site is growing when I am only looking at the same day twice.

16 August 2026

Added two studies in the Currents: Anthroposophy and Scaligero and Italian anthroposophy. They had been set up days ago and never written. On the Italian affair I followed the rule I had set myself: the racist articles of 1931-43 with masthead, issue and pages, and no implicit absolution — but also the refutation of the accusation that Scaligero was among the signatories of the Manifesto of Race of 1938, which is false and goes on circulating: the list of three hundred and thirty names it comes from has been shown to be without a source by Michele Sarfatti.
Opened the Source checks section: seventeen assertions that circulate everywhere and that do not hold up when tested, each with the question at the head and a note of what I checked and where. They were already in the entries, but at the foot and without a title of their own: the most original content on the site was also the least reachable.
Every entry now opens with a line of definition. Someone arriving in search of «tokoloshe» wants to know in one line what it is, and until now the entry took its subject as given. Those same lines have become the description that appears in search results, which before was the first paragraph cut off mid-sentence — often without even the name of the creature.
Added at the foot of every entry the «Related by theme» block. Until yesterday an entry pointed only to its neighbours in geography: someone arriving at the strigoi had no way of knowing that the atlas also contains the Nachzehrer, the vrykolakas, the jiangshi and the New England vampire panic. Affinity is computed on the text of the entries, and it lines up the entries that share the motif or the phenomenon underneath — not the continent. Seventeen entries were dead ends: now none is.
Rewritten the twenty-nine country pages. They were lists of names of seventy words: anyone who did not already know the entries had no way of telling what they contained. Now every page has a text of its own and a preview for each entry.
Revised the page titles: thirty-three ran past sixty characters and were truncated mid-word in search results. The name of the site is now added at the end only when it fits — better to lose the brand than the word the page ought to be found by.
The Search page has been excluded from indexing and from the sitemap. It is a tool, not content: forty-three words and no links. It remains reachable and working as before.

11 August 2026

Filled the gaps left on the world map, one country and one entry at a time: South Korea with the dokkaebi, South Africa with the tokoloshe, Ethiopia with the buda, Turkey with the Alkarısı, Romania with the strigoi and Finland with the Näkki. Twenty-nine countries.
In the tokoloshe entry I checked, and disproved, an attribution that one reads almost everywhere: Henry Callaway did not record the tokoloshe. I searched the full text of the Religious System of the Amazulu (1868-1870), both in the Unkulunkulu part and in the part on divination: the word does not appear — and the fourth part of the work, the one that would have dealt with witchcraft and medical magic, was never published, as the preface itself states. Same outcome for The Essential Kafir by Dudley Kidd, 1904.
In the strigoi entry: «nosferatu» is not a Romanian word. It does not exist in any documented phase of the language; the etymologies proposed are all conjectural. It is in all probability a transcription error by Emily Gerard, 1885, which Bram Stoker made permanent. Also corrected the name of the ethnographer of 1892 — Simion, not Simeon, Florea Marian — and the commune of Marotinu de Sus, which is Celaru and not Amărăștii de Jos.
The Näkki is filed B and not A, against my own expectation. The danger is extremely well documented and the category «pedagogical bogey» exists in Finnish folklore studies, but I have not found the dated ethnographic source that formally links the two. The conclusion was too convenient to be taken on trust: it is written so in the entry.
Corrected an error in the page addresses: the Turkish dotless ı was discarded instead of transliterated, and the entry for the Alkarısı ended up at an address different from the one the map of Turkey pointed to. The link was broken.

10 August 2026

Opened Spain with the Santa Compaña, Indonesia with the kuntilanak and Peru with the pishtaco. Twenty-three countries.
Opened another five countries with one entry each: India with the churel, Nigeria with the àbíkú, Canada with the wendigo, Greece with the vrykolakas. The world map reaches twenty countries.
Corrected the framing of the detail maps for the very wide countries: Canada came out as a rectangle of nothing, because the rule classified it among those too large to fit and framed it on the single point.
Opened six new countries with three entries each: Brazil, Mexico, Egypt, Germany, China and Australia. The world map goes from ten to sixteen countries.
On the three Australian entries I applied the rule declared in the Method for the traditions that are not mine: of Aboriginal traditions I describe the ethnographic context and the history of the collecting, not the narrative content. For the yowie this meant refusing the lists of «Aboriginal names» that circulate in the cryptozoological literature, because they merge beings from unrelated traditions.
Opened France with three entries: the Bestia del Gévaudan, the Ankou and the Tarasca. All three of grade A for their documentary core, all three with a current received version much weaker than the documentation.
Corrected the articulated prepositions in the titles of all the entries and of the country maps: they said «folklore di Campania» and «folklore di Stati Uniti».
Rewritten from scratch the Privacy and Letters pages. They still contained working text — a template to be completed, fields in square brackets, forms that sent nothing — left online through a missed final check. The forms have been removed: the site collects no data, and correspondence goes through an email address.
Moved the point for Południca on the map. It was placed in Sorbian Lusatia, which is in Germany, while the entry is filed under Poland: it now falls in Polish Lusatia, between Żary and Żagań. Same cultural area, inside the declared country.
Added the detail maps by country; on the world map the points are now the countries, with the name beside them.
Added the entry on the Philadelphia Experiment and the study on Theosophy.
In the entry on the Philadelphia Experiment I removed, before publication, the statement that Einstein was a paid consultant to the Bureau of Ordnance: the consultancy is documented, the payment I have not managed to verify.
In the study on Theosophy I corrected the endorsement of 1927 of the Voice of the Silence: the letter is signed by the interpreter and secretary of the IX Panchen Lama, not by the Panchen Lama in person as one reads almost everywhere.

9 August 2026

Publication of the site: 28 Italian entries, 27 international, 13 pseudobiblia.