What the collection says about itself

A library is not a neutral window onto its subject. It is a pile of things that happened to be written, happened to be printed, happened to survive and happened to be digitised, and the shape of the pile is itself evidence. This page reports what can be measured about these 2,756 volumes — and, just as carefully, what cannot.

Nobody has read them all. Not the person who assembled this library and certainly not the machine that generated this page. Everything below is counted rather than recalled, every figure is recomputed each time the library is rebuilt, and where a method is known to fail it is said so on the spot.

When these books were printed

1,873 of 2,756 volumes state a date the library can corroborate — a year standing beside a printer, a publisher or a city, rather than a year merely mentioned somewhere on the page. The rest are undated here, which is not the same as undatable.

Volumes per decade, 1800–1959. Peak: the 1910s, with 401.

The rise is steep and the fall is a cliff. That shape is partly real: the Theosophical Society was founded in 1875, the Hermetic Order of the Golden Dawn in 1888, and the decades on either side of 1900 genuinely were a publishing boom in this subject.

But most of the cliff is copyright, not history. A library that publishes only what it may lawfully publish cannot hold much after the 1920s, because most of it is still in copyright. The collapse after that peak is very largely a fact about law, and any reading of it as the movement dying out would be an artefact of how this collection was assembled. It is the clearest example on this page of a shape that looks like a finding and is not.

Counting words is not counting ideas

The obvious way to trace an idea through a corpus is to search for its name. It does not work, and it fails in a way that produces confident nonsense rather than obvious errors.

Term	Books	Spread
reincarnation	303	1.20	term of art
etheric	168	1.57	term of art
karma	344	1.61	term of art
planchette	73	2.64	term of art
mesmerism	217	1.36	term of art
astral	456	0.98	leaning technical
talisman	344	0.81	leaning technical
vibration	402	1.13	term of art
sympathy	1,387	0.38	ordinary English
correspondence	1,119	0.29	ordinary English

Spread is how unevenly a word falls across the subject shelves. Ordinary English is spread evenly, because ordinary English is everywhere; a term of art clusters.

Sympathy occurs in more books than any occult term in this library, and correspondence is not far behind. Neither is evidence of the doctrine of sympathies or the law of correspondence: they are people feeling sympathy and people receiving letters. A study built on raw counts would have announced both as the central organising ideas of Western esotericism, with thousands of citations, and been wrong. Only the clustered terms are used below.

One vocabulary replaces another

Each bar is the share of that decade's dated volumes using the word, so the growth of the collection itself cannot masquerade as an idea spreading. Each row is scaled to its own peak.

Fading

Absent, then everywhere

Already present, then intensifying

Borrowed back

1840s on the left, 1920s on the right. Decades holding fewer than 30 dated volumes are left out of these rows: too few books to take a share of.

Mesmerism peaks in the 1850s, in more than a quarter of that decade's volumes, and falls away to under a tenth by the 1910s. It is the only word measured here that clearly goes.

What arrives is sharper. Reincarnation is in none of the dated volumes of the 1840s and 1850s and about a fifth of them by 1900. Etheric is later still — effectively absent until the 1860s, and climbing right through to the end of the range. Karma follows the same path. All three are Theosophical vocabulary, and all three take hold within roughly two decades of the Society's founding in 1875. Planchette is the sharpest of all: nothing, nothing, nothing, and then a tenth of the 1870s, which is exactly when the device was a drawing-room craze.

Two words refused the story. Astral is in six to eleven per cent of the volumes of the 1840s to 1860s — before the Society existed — and rises to about a quarter. It was not introduced by Theosophy; it was absorbed by it. Talisman is broadly flat across the whole century, neither displaced nor revived. Both had been grouped the other way in a first draft of this page, because that is what the familiar account predicts, and the counts would not have it.

Vibration is the odd one. It appears early, in eighteenth and nineteenth-century natural philosophy, where it means what a physicist means; it then surges to nearly half of the 1930s volumes in a quite different sense. The word did not arrive. It was borrowed back, out of physics and into metaphysics, a century after it left. That last column rests on 34 volumes and 16 of them, so it is the thinnest claim here — the rise through the 1890s to 1920s, on hundreds of books a decade, is the solid part.

Where they were printed

796 volumes name a city of publication.

LONDON	456
NEW YORK	167
BOSTON	39
PHILADELPHIA	25
CHICAGO	20
EDINBURGH	15
CAMBRIDGE	13
WASHINGTON	11

London and New York between them account for most of what is here. That is partly the subject — the nineteenth-century occult revival was an Anglo-American affair with a strong Indian traffic — and partly this library: it holds English-language texts, so an English-language publishing geography is what it can possibly show. What the collection cannot tell you is what was being printed in French, German or Bengali on the same subjects at the same time.

The library is several libraries

Nothing here was collected at once. The volumes carry id numbers in the order they were accessioned, and those numbers fall into blocks that barely overlap — so the id is not a shelfmark so much as a timestamp of how the collection was assembled.

Where it came from	Volumes	Id range
the 2004 disc	90	1–169
EPUB editions	215	174–773
PDF scans	342	172–838
Project Gutenberg	1,619	839–3040
declassified releases	322	1844–2237
other	168	2847–3049

The 2004 CD-ROM is the founding layer and occupies the lowest ids. PDF and EPUB restorations came next. Then a very large Project Gutenberg import, which is now most of the library. The declassified files arrived last and sit in a block of their own. A shelf built mainly from one layer will look and read differently from a shelf built from another, and several do — which is worth holding in mind before reading any subject as though it were a considered selection.

The same work, arriving twice

A collection assembled in layers will acquire the same book more than once — the 2004 disc and a later ebook edition of the same text, catalogued separately because nothing compared them. That has happened 57 times in the archive.

None of them is published twice. 5 of them still show both copies, which is a fault and is listed for repair. The duplicates are real and the archive keeps them; the shelf you browse does not show them.

Shelf by shelf

The 28 subjects holding twenty volumes or more, largest first — 2,609 of the 2,756 books between them. For each, the words that turn up markedly more often there than in the rest of the library. Nothing is chosen by hand: every shelf gets the same test, and the percentage is the share of that shelf's volumes using the word.

Some of these read as a subject index and some do not, and the difference is itself worth noticing. Astrology returns the working vocabulary of a craft — sextile, cusp, benefic, succedent — in most of its volumes; so does Secret Societies, whose distinctive words are the furniture of Masonic ritual. Theosophy returns devachan and Dzyan, terms Blavatsky either coined or introduced, which is a movement writing its own dictionary. Witch-Craft returns names — Tituba, Kyteler, Cullender — because its books are trial records, and a trial record is about people.

Mysticism is the surprise. Its distinctive words are vastation, inmosts, conjugial — not the common property of mystical writing but the private vocabulary of one author, Emanuel Swedenborg. He wrote 25 of the shelf's 92 volumes, 27% of it, and conjugial appears in 25% of them, vastation in 21%. One man's coinages are the sharpest marker of a shelf named for a whole tradition.

And Folk-Lore & Mythology, the largest shelf of all, returns almost no subject terms — its distinctive words are archaic narrative verbs. What separates it from the rest of the library is not what it is about but how it is written: it is the shelf of old storytelling prose. A shelf can be defined by register rather than by topic, and this one is.

How the shelves stand to each other

Every shelf can be described by the share of its volumes using each word. Comparing those descriptions says how alike two shelves are in language — not in what anyone believes, only in what gets written down. 54,970 words are common enough to be shared and rare enough to discriminate; the figures below run from 0 to 1.

Shelf	Most resembles
Comparative	Judaism	0.92
Judaism	Comparative	0.92
Esoteric & Occult	Comparative	0.91
Islam & Sufism	Comparative	0.91
Folk-Lore & Mythology	Paganism	0.90
Paganism	Folk-Lore & Mythology	0.90
Secret Societies	Comparative	0.90
Life After Death	Comparative	0.90
Self-Help & New Thought	Life After Death	0.89
Theosophy	Comparative	0.89
Philosophy & Ethics	Comparative	0.87
Apocrypha	Christianity	0.87
Christianity	Apocrypha	0.87
Mysticism	Comparative	0.87
Paranormal	Life After Death	0.87
Meditation & Yoga	Theosophy	0.86
Buddhism	Comparative	0.85
Gnosticism & Hermetica	Mysticism	0.85
Witch-Craft	Esoteric & Occult	0.84
Prophecies	Comparative	0.83
Symbols & Sacred Signs	Esoteric & Occult	0.82
Fortune Telling & Divination	Esoteric & Occult	0.80
Ancient Egypt	Paganism	0.80
Hinduism	Buddhism	0.79
Alchemy	Gnosticism & Hermetica	0.78
Astrology	Fortune Telling & Divination	0.76
Hieroglyphs & Ancient Scripts	Ancient Egypt	0.73
Declassified Records	Self-Help & New Thought	0.45

The median pair of shelves in this library scores 0.73.

Most of it is unsurprising and that is the point — the measure is behaving. Folk-Lore sits next to Paganism, Judaism next to Comparative, Paranormal next to Life After Death. Three results are worth stopping on.

“Ascended Masters (2004 disc)” is not a subject. It is the nearest neighbour of Esoteric & Occult, Theosophy, Mysticism, Meditation & Yoga and Gnosticism & Hermetica — it sits in the middle of the whole occult half of the library because it is a note about where a book came from, wearing a subject heading. The measure recovers that without being told.

Declassified Records is an island. It is the furthest shelf from almost every other one in the collection. Its closest relative is UFOs & Aliens at — — still below the median for any two shelves here. And UFOs & Aliens is itself close to Paranormal, which makes it the hinge: the one shelf with a foot in the occult library and a foot in the government files.

A claim from the readings, measured

The readings argue that Gnosticism & Hermetica, Paranormal and Declassified Records are one argument in three buildings — temple, parlour, secure room — the same unauthorised faculty under different letterhead. Here is that claim as an angle between vocabularies.

Pair	Similarity
Gnosticism & Hermetica beside Paranormal	0.69
Gnosticism & Hermetica beside Declassified Records	0.33
Paranormal beside Declassified Records	0.45
any two shelves, for comparison	0.73

As a trio it does not hold. Gnosticism & Hermetica and Paranormal are ordinary neighbours, a shade below the median. Both are a long way from the declassified files. If these three shelves are one argument, the argument is not visible in the words.

Which is, oddly, the readings' own point turned into a number. They say the metaphysics gets smuggled through each era in new clothes — clairvoyance, then telepathy, then perceptual augmentation — and that what continues is a method rather than a doctrine. If that is right, then a machine comparing vocabularies is exactly the instrument that cannot see it. The measurement does not refute the reading. It marks the edge of what counting can reach, and the reading is making its case on the other side of that edge.

Where the books came from, and what each source cost

Every volume arrived from somewhere, and the somewheres do not behave alike. The honest measure of a source is not how much it delivered but how much of it the library then had to withhold, because it turned out to be somebody’s property.

source	items	withheld	rate
Project Gutenberg	1,643	4	0.2%
scanned PDFs	398	49	12.3%
declassified releases	362	40	11.0%
donated EPUBs	270	41	15.2%
the 2004 disc	170	71	41.8%
other	168	0	0.0%
everything	3,011	205	6.8%

The spread is the finding. Project Gutenberg supplies more than half this library and has produced almost nothing that had to come down: its texts were cleared before they were ever typed. The 2004 CD-ROM the collection started from is the opposite. Two of every five items on it could not be published. It is the kind of disc that was pressed by the hundred in that decade - whatever fitted and looked like ancient wisdom, with no one asking who owned it - and every rights problem this library has had, it has had because of that habit.

The numbering shows it plainly. The disc numbered its contents 001 to 169, and 90 of those numbers are still in the library. The other 79 were not moved to other shelves: not one of them was. Of those, 65 are held back on rights grounds and 14 were never good enough to publish. The gaps in the numbering are not a filing history. They are the withholding rule, visible as an absence.

Where the pages are missing

Two kinds of hole run through this collection, and they look identical in a text file. A lacuna is a fact about history: the tablet broke, the leaf is gone, and nobody will ever repair it. A transcription gap is a fact about whoever typed the electronic copy - the printed page still says it. Confusing the two would let a library present its own carelessness as the ravages of time.

26 of 2,756 volumes carry three or more places where an editor marked the manuscript itself as damaged - 379 marks in all. Only a handful of those are genuinely fragmentary works rather than books that mention a gap in passing. The richest is Thompson’s Devils and Evil Spirits of Babylonia, whose editor did the rare and useful thing of measuring each hole as he came to it: [Hiatus of about four lines.] Those gaps are set out one by one, with what two independent readers made of each, at the argument about the holes.

The copies are damaged far more often than the sources. 688 volumes carry a marker where a transcriber recorded leaving something out, and the count separates into kinds that need different answers:

what the transcriber left	count	can it be put back
Greek deleted outright, marked [------]	1,353	only from the printed original
Greek written out in Latin letters with its accents, [Greek: a)pori/ai]	4,263	yes, and it has been
Greek written out without its accents, [Greek: gnosis]	2,462	no — the breathings and vowel lengths are gone
a picture the library does not hold	16,152	no

The second row is the one that could be fixed, and was. That scheme spells the Greek out in Latin letters and carries the breathings and accents as punctuation after the vowel, so nothing was lost - only made unreadable. It converts exactly, and the reader now sees ἀπορίαι where the file says a)pori/ai. The third row is the one that cannot: strip the accents and γνῶσις and γνωσις become the same string, and no amount of processing will tell you which was meant. An early version of the converter did those too, and produced Greek that looked right and was wrong in a quarter of everything it touched.

The archival text is untouched by any of this. What the transcriber wrote is what book.txt still says, and the search index is built from that, so a reader looking for Heraclitus can type logos on an ordinary keyboard.

What this page cannot tell you

• The dating is imperfect and fails in a particular direction. A book about an earlier period can be dated to it. Frank Podmore's study of mesmerism, published in 1909, is dated here to 1784, because its preface discusses the royal commission of that year and the word "Society" sits nearby. Aggregate curves survive a few hundred such errors; a claim about any single book does not.
• Only what survived, was scanned, and is out of copyright is here. The collection is a sample with known biases, not a census. Cheap pamphlets, private printings and anything still in copyright are systematically absent.
• Presence is not belief. A book using the word karma may be arguing against it. Nothing here distinguishes an advocate from a critic.
• Publisher names are recorded for too few volumes to support any claim about which houses shaped the subject, so none is made.

Every figure above is recomputed from the books each time this site is built. If a volume is added or withdrawn, the numbers move with it. The reasoning behind what is and is not in the collection is set out at what we publish.

Search the libraryBrowse by subjectWhat we publish