Cosmicist fiction · A novella in ten chapters

The Book That Learned to Exist

An imaginary book acquires an impeccable past. The scholars who reconstructed it discover that correcting the evidence only gives that past more history.

By Therion Ware Ten chapters Approximately 79,377 words

A fictional development of the Necronomicon reconstruction, The Book That Should Not Exist.

Chapters and reading options
  1. IAn Erratum Concerning an Imaginary Book
  2. IIThe Problem of Provenance
  3. IIIThe Seventh Division
  4. IVThe Probability of Yesterday
  5. VAn Error in Alhazred
  6. VIThings That Have Always Been There
  7. VIIThe Geometry of the Gate
  8. VIIIThe Failure of Correction
  9. IXThe Arrival That Has Already Happened
  10. XA Minor Bibliographical Correction

Chapter I 1 of 10

An Erratum Concerning an Imaginary Book

The first indication that something was wrong arrived in the form of a bibliographical correction. Dr Miriam Vale received it at 08:43 on a Thursday morning, between a reminder concerning an overdue journal review and an invitation to contribute to a symposium on the cultural afterlife of medieval demonology. Neither communication interested her particularly. The third, however, concerned an article she had helped prepare for a website called City of Dis, and was sufficiently peculiar to interrupt her breakfast.

The article was entitled The Book That Should Not Exist. It concerned the Necronomicon, the imaginary grimoire invented by H. P. Lovecraft and subsequently furnished with so elaborate a fictional history that generations of readers had persuaded themselves, or attempted to persuade others, that a genuine historical work lay behind it. Miriam had enjoyed the research.

Her contribution had consisted largely of checking the bibliographical claims against Lovecraft's original writings, separating his inventions from those of later imitators, and ensuring that none of the supposedly historical details had acquired an undeserved appearance of authenticity through careless citation. It was the kind of work she found immensely satisfying: scholarship devoted to establishing, with absolute precision, the properties of something that had never existed. The correspondence was less satisfying.

Two evenings earlier, returning from the shops, Miriam had found her neighbour Ruth Bennett struggling with the latch on the garden gate. Ruth was a retired pharmacist who had little interest in medieval literature and considerably more in the upkeep of the shared fence. Miriam held the gate while she freed a trailing rose stem.

"What's occupied you this week?" Ruth asked.

"An imaginary book. I'm helping to write its history."

"That should save you the expense of buying it."

"You'd think so. It's acquiring a considerable bibliography."

Ruth laughed, retrieved her shopping bag and went indoors. Miriam remembered the exchange because it was the first time all week she had successfully explained her work without also explaining Lovecraft. No written record was made. The conversation had served no purpose beyond filling the half-minute it took to free the gate.

Subject: Correction concerning the Wormius printing

Dear Dr Vale,

I have been reading your recent reconstruction of the Necronomicon with considerable interest, particularly your discussion of the surviving Latin editions.

There appears, however, to be an error in your treatment of the fifteenth-century printing.

You describe this edition as entirely fictional, attributing its supposed preservation in the British Museum to Lovecraft's invented bibliography.

This is difficult to reconcile with the copy presently held in the British Library's historical collections, formerly belonging to the British Museum.

The volume is catalogued as a German black-letter printing, provisionally dated to the final decade of the fifteenth century. Its provenance can be traced through the Museum's nineteenth-century accession records.

I appreciate that the publication history of the Necronomicon has been complicated by modern fabrications. Nevertheless, the existence of this particular volume is not, so far as I am aware, seriously disputed.

I would be grateful if you could clarify the basis of your assertion.

Yours sincerely,

Oliver Stroud
Rare Books and Early Printed Collections

Miriam read the message twice. Then she laughed. Not because it was particularly amusing, but because it represented such an excellent example of the problem the essay had been written to address.

There was a substantial industry devoted to pretending that the Necronomicon existed. Some participants were frauds. Others were enthusiasts. A few appeared to have difficulty distinguishing between a literary invention and the historical background constructed to make that invention convincing. The more elaborate efforts occasionally produced forged catalogue records, manufactured manuscript fragments and purported translations attributed to otherwise respectable historical figures. The name Olaus Wormius was a particular favourite.

Lovecraft had assigned his fictional Latin translation to a scholar bearing that name, although the historical Ole Worm belonged to a considerably later period. The confusion was useful to hoaxers because it provided the appearance of a genuine scholarly connection without requiring much familiarity with medieval intellectual history. Miriam opened the article. The relevant section was precisely as she remembered it.

Lovecraft's imaginary chronology placed the Arabic original, Al Azif, in the eighth century. A Greek translation followed in the tenth. Wormius's supposed Latin version appeared in 1228, and later printings were assigned to the fifteenth and seventeenth centuries. The British Museum was said to possess one of the earlier Latin copies. The article made the distinction between literary invention and historical evidence unambiguously clear. Miriam composed a courteous reply.

Dear Mr Stroud,

Thank you for your message.

I suspect there has been some misunderstanding.

The fifteenth-century Latin edition to which you refer is part of Lovecraft's invented bibliographical history. It is not a genuine incunable, and no historical copy of the Necronomicon is known to exist.

The attribution to Wormius, the dates of the supposed translations and the locations of the surviving copies derive from Lovecraft's short fictional essay, "History of the Necronomicon".

If you have encountered a catalogue entry purporting to document such a volume, I should be interested to examine it. Similar entries have occasionally appeared in connection with literary hoaxes.

Kind regards,

Miriam Vale

She sent the message, finished her coffee and returned to a manuscript concerning the transmission of an obscure Byzantine commentary on Aristotle. Forty-three minutes later, Stroud replied. There was no sign of embarrassment.

Dear Dr Vale,

Thank you for responding so promptly.

I am familiar with Lovecraft's bibliographical fiction. That is precisely why I found your article interesting.

I attach the current catalogue record and a photograph of the volume's opening leaf.

Our holdings also include nineteenth-century documentation of its acquisition. I can provide those references if necessary.

The book has not been available for ordinary consultation for some years because of its condition, but I can confirm that it is physically present.

I confess that I am rather puzzled by your insistence that it cannot exist.

Best wishes,

Oliver Stroud

Miriam opened the first attachment. It was a catalogue record. Not an imitation composed in a word processor, nor a crudely assembled screenshot. It appeared to be an ordinary export from an institutional bibliographical database. The entry described a Latin work attributed to Abdul Alhazred, translated by Olaus Wormius, printed without an identified publisher in the German-speaking lands around 1490. The title appeared as Necronomicon. The record included physical dimensions, collation, notes concerning damaged gatherings and a description of the binding.

There were references to earlier catalogues, to conservation work undertaken during the twentieth century, and to the book's transfer from the British Museum collections. Miriam examined the details carefully.

There was nothing intrinsically impossible about the physical description. A late fifteenth-century printed book of that size and construction was perfectly plausible. The terminology was correct. The collation followed recognised conventions. The sophistication of the forgery, if forgery it was, exceeded the usual standards of occult enthusiasm. She opened the second attachment. It showed the first surviving printed leaf of an old book resting in a conservation cradle.

The paper was yellowed, though not extravagantly so. There were small areas of staining near the inner margin. The typography was dense and rather ugly, with the irregularities characteristic of early printing. There were no elaborate sigils. No human faces pressed into leather. No theatrical symbols of forbidden knowledge. The heading was printed in black-letter type. Beneath it, in a smaller hand, stood a Latin rendering of Alhazred's name. Miriam enlarged the image.

She had examined thousands of photographs of early printed books. She knew the difference between an authentic-looking reproduction and an authentic object could not be established from a photograph alone. Nevertheless, the image was professionally made. The lighting was appropriate. The page had the correct slight curvature. The shadows fell naturally across the binding. The photograph included a colour calibration strip and a ruler. The document was, in every respect, ordinary. This irritated her more than a grotesque forgery would have done.

She opened her browser and searched the British Library's catalogue independently. Not through the link Stroud had supplied. She entered the institution's address herself and navigated to the catalogue interface. She searched for the title. There were several results concerning Lovecraft, modern occult publications and literary studies. And one concerning a fifteenth-century Latin printing. She stared at it. Then she checked the address bar. The domain was correct. The connection was secure. The catalogue behaved normally. Other searches returned precisely the results she expected. The entry was there.

She opened it. Its bibliographical description agreed with Stroud's attachment. A linked authority record associated the work with Abdul Alhazred, described as an author traditionally assigned to the eighth century. The explanatory note distinguished between the Arabic original, the Greek transmission and the Latin translation. Miriam noticed something interesting. The record contained no reference to Lovecraft. That was odd. A genuine historical copy of the Necronomicon, had one somehow been discovered, would have attracted extraordinary attention. It would represent an almost unparalleled coincidence in literary history.

Yet the record treated the volume as an obscure medieval work of uncertain authorship. Nothing suggested that the cataloguer had encountered the slightest difficulty with its existence. She searched for articles discussing it. There were several.

One concerned the relationship between the Latin text and medieval Arabic cosmological speculation. Another examined the history of the Wormius translation. A third disputed the dating of the surviving German printing. The abstracts were blandly academic. The authors appeared to be real scholars with ordinary institutional affiliations and publication histories. Miriam recognised one of their names.

Professor Eleanor Marsh had written extensively on medieval Latin translations. Miriam had met her at a conference in Oxford some years earlier and remembered an exceptionally tiresome disagreement concerning a manuscript attribution. The article was dated 1998. Its title was The Alhazred Tradition and the Problem of the Latin Necronomicon. Miriam sat back. There were several possible explanations, none yet requiring a revision of her understanding of reality. An unusually elaborate hoax was one. A compromised catalogue was another.

A coordinated literary experiment involving real academics was a third, though she found the last rather improbable. The possibility that the Necronomicon had genuinely existed for centuries and had somehow escaped the attention of Lovecraft scholars was not one she took seriously. It would require too many independent failures of scholarship. More importantly, Lovecraft's own correspondence established the fictional nature of his invention. The book had been invented by a writer of weird fiction. That was not a disputed historical proposition. She telephoned the British Library.

After a brief conversation with a member of staff, she was transferred to the department responsible for early printed collections. A woman answered. Miriam explained that she was trying to verify the existence of a particular volume. She supplied the catalogue reference. There was a pause, accompanied by the faint sound of typing.

"Yes," the woman said. "I can see it."

"Would you mind confirming the title?"

"Certainly. Necronomicon. Latin translation attributed to Olaus Wormius. Fifteenth-century printing."

"And the volume itself?"

"Excuse me?"

"Is it physically present? I'm not asking whether the catalogue says it exists. Has anyone actually examined the object?"

Another pause.

"I believe Mr Stroud is responsible for that material. Would you like me to ask him to contact you?"

"He already has."

"Then I imagine he'll be able to answer your question."

"One final thing. Does the catalogue mention H. P. Lovecraft?"

"Who?"

"The American writer."

"I can check."

More typing.

"No. Should it?"

Miriam considered the question.

"Yes," she said. "Rather prominently."

By lunchtime, she had assembled a small collection of apparently independent evidence for the existence of a book that did not exist. The situation was becoming intellectually offensive. The catalogue record might conceivably be fabricated. The photograph might be manipulated. The academic articles could be counterfeit or the work of an elaborate group of enthusiasts. But their interconnections required explanation. The published bibliographies cited one another. The references led to older catalogues. Those catalogues referred to acquisition records, binding descriptions, institutional correspondence and earlier scholarly disputes.

A paper published in 1963 criticised an interpretation advanced in 1937. An article from 1937 cited a nineteenth-century catalogue. The catalogue referred to an acquisition made in 1891. Each document possessed the sort of mundane contextual detail that serious historians valued precisely because forgers often neglected it. There were disagreements about the dating. There were corrections to earlier transcriptions. There were competing suggestions concerning the location of the printing workshop.

There was even an obscure argument about whether a damaged initial represented the letter P or the letter R. Nobody was trying to make the book mysterious. That, Miriam thought, was the most suspicious feature of all. The documents behaved like ordinary scholarship concerning an ordinary object. Their authors appeared to regard the Necronomicon as a rather unimportant example of late medieval textual transmission. The book's more extravagant claims were treated as curiosities of intellectual history.

One scholar suggested that Alhazred's descriptions of pre-human civilisations represented an unusually developed form of ancient speculation about the age of the world. Another had argued, apparently with some success, that passages concerning unusual geometrical relations between places reflected a primitive attempt to reconcile contradictory cosmological traditions. The arguments were sensible. They were also impossible. Miriam opened the draft of the City of Dis essay she had saved on her computer.

She had contributed comments and corrections over several days, preserving the working files because she habitually retained research material long after publication.

In the section concerning the Latin translations, she found the sentence she remembered reviewing:

Lovecraft gave his imaginary book the sort of textual transmission normally possessed by genuine works of antiquity.

She checked the following paragraphs. They described the translations, the dates of the supposed editions and the institutional locations of the surviving copies. The text was identical to the published version. Miriam felt a small measure of relief. At least one document remained where it ought to be. She checked the file's creation date. Then she opened the folder containing her research notes.

There were references to Lovecraft's fiction, to the History of the Necronomicon, to the publication dates of relevant stories and to the appearances of the book in successive works. There were notes on John Dee, on the historical Ole Worm, and on the various modern books falsely marketed as genuine editions of Alhazred's work. There was nothing about an actual medieval Necronomicon. It would have been an extraordinarily conspicuous omission. She sent Stroud another email.

Dear Mr Stroud,

I have now examined the catalogue entry and several of the associated references.

Would it be possible to arrange an inspection of the volume?

I should also be grateful for copies of the nineteenth-century acquisition documentation you mentioned.

There is an unusual bibliographical problem here which I would prefer to investigate in person.

Kind regards,

Miriam Vale

His reply arrived almost immediately.

Dear Dr Vale,

Of course.

I should be happy to arrange an examination, subject to the usual conservation restrictions.

I attach an image of the relevant entry from the Museum's accession register. The volume was acquired from the estate of a private collector in November 1891.

I have also located a reference in a printed catalogue issued the following year.

There is one matter I should perhaps mention before your visit.

Until reading your article, I had always understood Lovecraft to have borrowed the title and some of the alleged history from a genuine medieval text.

Your account of the book's literary development is very persuasive, and I find myself unable to reconcile the two histories.

It would be useful to discuss this.

Miriam read the final paragraph several times. It was the first genuinely interesting thing Stroud had written. He wasn't simply defending an obscure institutional holding against the assertions of a literary historian. He had encountered the same contradiction from the opposite direction. For him, the book belonged to history, and Lovecraft had borrowed it. For her, the book belonged to fiction, and history had no place for it. Both positions appeared to be supported by extensive documentation. And neither could accommodate the other.

That evening Miriam visited her study, a narrow room at the rear of the house which contained more books than its dimensions reasonably permitted. The shelves were arranged by subject, though the organisation had gradually deteriorated into a system intelligible chiefly to herself.

Medieval intellectual history occupied the wall opposite the window. Literary criticism and textual scholarship were arranged beside the desk. A smaller collection of books about supernatural fiction stood near the door. She was looking for a particular volume of Lovecraft correspondence. It contained a discussion of the Necronomicon which she remembered using while checking the City of Dis article. The letters made it clear that Lovecraft regarded the book as his own invention. She found the volume and sat down. The relevant correspondence was there.

She read it carefully, checking the wording against her notes. Nothing had changed. Lovecraft had invented the name. He had invented Alhazred's literary history. He had invented the translations and the locations of surviving copies. Of that there seemed no reasonable doubt. Miriam turned back to her computer. The catalogue record was still open. It documented the existence of the book decades before Lovecraft's birth. Both things could not be true in the ordinary sense.

A writer could invent a fictional account of a real book, certainly. He could embellish an obscure historical source with imaginary details. But the particulars of the alleged historical book were precisely those Lovecraft had invented. Even its textual imperfections matched the fictional record. It was as though a historian had discovered a genuine eighteenth-century manuscript containing the dialogue of a character from a novel not published until the following century. One or two coincidences might be explained away. The entire structure could not.

Miriam looked at the accession-register image Stroud had supplied. The handwritten entry was dated 1891. It described a volume acquired from a private collection, with notes concerning its binding and incomplete state. She enlarged the photograph. The handwriting was consistent with other entries on the page. The ink appeared uniform. There were no obvious signs of later insertion. None of this established authenticity, of course. A photograph could not demonstrate that the underlying document was what it claimed to be. Nevertheless, something had begun to disturb her.

Not the suggestion that the Necronomicon existed. That remained an absurdity. What disturbed her was the growing number of entirely reasonable people who appeared to have always known that it did. She returned to the printed bookshelves. A few years earlier she had purchased several bound volumes of nineteenth-century British Museum acquisition catalogues at an antiquarian bookshop.

They were of little practical use in an age of searchable digital records, but she had acquired them because she liked the physical continuity they provided with an earlier generation of scholarship. The relevant volume stood on the upper shelf. She took it down. Its binding was faded blue cloth, with a narrow leather label bearing a date range in gilt lettering. There was an old bookseller's price pencilled onto the inside cover. She remembered buying it. She remembered the absurd weight of the parcel when it arrived.

She even remembered complaining that the seller had described the condition as very good when the spine was plainly separating from the text block. There was no reason to doubt the book's provenance. She opened it. The catalogue was arranged by accession date. It took her some minutes to locate November 1891. There were entries for theological works, travel accounts, medical treatises and several collections of pamphlets. Then, close to the bottom of a page, she found the name.

Alhazred, Abdul.

Beneath it was a description of a Latin printing of the Necronomicon. The bibliographical details agreed with the modern catalogue. Miriam did not move. She could feel the roughness of the old paper beneath her fingers. She examined the printing through a magnifying glass. The type was consistent with the rest of the page. The impression and paper texture showed nothing unusual. The entry occupied its natural place in the alphabetical sequence. There was no pasted insertion. No evidence that the page had been replaced.

She turned to the preceding leaf, then the following one. Everything was ordinary. The book had been sitting on her shelf for years. She was certain of that. And throughout those years she had been equally certain that the Necronomicon had never existed. Miriam went back to her desk and opened her research notebook. It was a hardback volume with numbered pages, filled with the untidy handwriting she used when working from printed sources. She found the section devoted to the City of Dis essay.

At the top of one page she had written:

NECRONOMICON: SOURCE RECONSTRUCTION.

Below it were references to Lovecraft's stories, dates of composition, and brief observations about their fictional bibliography.

Halfway down the page, underlined twice, was a sentence she remembered writing:

No historical original. All alleged surviving copies derive from Lovecraft's invented chronology.

She had written that sentence herself. She could remember doing so. She had been sitting at the same desk, irritated by yet another website claiming that the book had originated in ancient Sumer. Her notes continued onto the following page. There she found a reference to the British Museum copy. It contained the expected citation to Lovecraft's History. But immediately beneath it, in her own handwriting, was a second reference. An abbreviated bibliographical citation. A date. A page number. Miriam looked at it for a long time.

The reference was to the nineteenth-century accession catalogue she had just examined. She had no recollection of writing it. She checked the ink. It appeared to be the same black ink she had used throughout the notebook. The handwriting was hers, including the peculiar way she formed the capital B. There was no question of a forged hand or a substituted page. She turned back to the previous sheet. The assertion that no historical original existed was still there. The later citation contradicted it.

Both were in her handwriting. Both apparently belonged to the same afternoon's research. She examined the dates at the top of the pages. There was nothing unusual about them. For several minutes she attempted to reconstruct the sequence of work that might have produced such a contradiction. Perhaps she had found the catalogue reference and dismissed it as a hoax. Perhaps she had intended to investigate it further. Perhaps the citation referred to something else entirely.

Yet none of these possibilities explained why she had no memory of encountering the entry, particularly when the existence of a real fifteenth-century Necronomicon would have been the most extraordinary discovery of her professional life. There was something else. A small pencil mark in the margin beside the accession reference. It was an asterisk. Miriam habitually used asterisks to identify matters requiring further investigation. She turned to the final pages of the notebook, where she collected unresolved questions. There was only one entry concerning the Necronomicon.

It had been written in pencil, in the abbreviated style she used when reminding herself of a task.

Check accession date against composition of HPL History. Chronology impossible?

Below it was a second sentence.

Why does nobody else notice?

Miriam stared at the words. She could not remember writing them. She turned the notebook sideways, almost expecting the angle to expose some unfamiliar habit in the letters. The small hooked tail on the y, the compressed capitals, the pencil indentation visible on the following page: all were hers. She pressed a thumb into the groove and felt the pressure of a hand she could not remember moving.

Why does nobody else notice?

Someone had noticed, and had apparently asked the question in Miriam Vale's handwriting. She did not sleep particularly well.

At half past three in the morning she returned to the study and checked the catalogue once more. The record remained unchanged. She checked Lovecraft's correspondence. Again, unchanged. She checked her notebook. The anomalous references were still there. She had, by this point, developed a list of hypotheses. The first was that she was mistaken. The second was that she was being deceived. The third was that the historical records had been corrupted through some process she did not understand. She did not add a fourth.

There are some explanations which do not become intellectually respectable merely because the alternatives are inconvenient. At 04:17 she sent a final email to Oliver Stroud.

Dear Mr Stroud,

I should like to examine the volume at the earliest opportunity.

There is also a question I would appreciate your answering as precisely as possible.

Before reading the City of Dis article, had you ever encountered the suggestion that the Necronomicon was an entirely fictional work invented by Lovecraft?

If so, what did you make of it?

Kind regards,

Miriam Vale

She did not expect a response until morning. It arrived eleven minutes later.

Dear Dr Vale,

I was about to write to you on a related matter.

Yes, of course I had encountered that suggestion. I teach occasional classes on imaginary bibliography, and the Necronomicon is a familiar example of the difficulties surrounding invented authorities.

What puzzles me is that I can distinctly remember teaching a session last year in which I explained that no such historical work existed.

I have checked my lecture notes.

They now contain a section discussing the fifteenth-century printing in our collection.

I have no recollection of preparing that section.

I should very much like to speak to you.

Oliver

Miriam read the message, then returned to her notebook. The contradiction had not been hers alone. That was, in one respect, reassuring. In another, it removed the most comfortable explanation. She looked through the study window. Beyond the garden, the town lay beneath a low November sky. A streetlamp illuminated the pavement. Somewhere in the distance a car passed along the main road. The world remained obstinately ordinary. There were no strange lights. No disturbances in the heavens. No suggestion that the laws of nature had been suspended.

There was simply an old book in a library, a catalogue entry in a volume published more than a century earlier, and two scholars who remembered a history their documents now contradicted. Miriam returned to the essay.

Near its conclusion, the author had written that Lovecraft's imaginary book was dangerous not because it contained spells, but because it connected otherwise unrelated observations into a coherent account of reality. She remembered admiring the argument. A single strange fossil was merely a curiosity. A puzzling mathematical result was merely a problem. An obscure historical tradition was merely folklore. The danger began when those fragments were recognised as parts of the same underlying structure. The Necronomicon, according to the essay, was an engine of correlation.

Miriam looked again at the surviving edition's catalogue record. She wondered, for the first time, whether the process might work in both directions. Whether the act of constructing a sufficiently coherent description could somehow do more than explain an object. Whether it could provide what an object required in order to exist. She immediately rejected the thought. It was a category error, and a rather spectacular one. Descriptions did not create their subjects.

A sufficiently detailed biography of an imaginary person did not produce a living individual. A complete account of an invented civilisation did not deposit its ruins beneath the earth. A fictional cosmology did not oblige the universe to rearrange itself to accommodate the entities it described. That was simply not how reality worked. Miriam closed the browser. In the silence of the study, her eyes returned to the final sentence of the article.

It concerned the discovery that the Necronomicon, for all Alhazred's errors and misunderstandings, might have been substantially correct about the universe. She had read the sentence many times during the preparation of the essay. Now a rather different possibility occurred to her. Suppose the act of establishing the book's history had made the book real. That would be extraordinary enough. But the reconstructed text did not exist in isolation. It contained descriptions of places, beings, civilisations and events. It possessed a history of translations, discoveries, suppressions and rediscoveries.

Its contents referred to other objects whose existence, in turn, would require explanation. If the book had somehow acquired a genuine past, that past might not stop at the circumstances of its printing. It might continue backwards through the supposed Latin translation, through the Greek manuscripts, through Alhazred himself. And beyond Alhazred, to the things he claimed to have seen. Miriam opened a fresh page in her notebook.

At the top she wrote:

Assuming the physical book exists, what else must have existed for its contents to possess a coherent history?

She considered the question for several minutes.

Then she drew a line beneath it and added:

Distinguish historical existence of text from truth of claims.

It was an elementary precaution. The existence of a medieval account of dragons did not establish the existence of dragons. A genuine ancient text could contain invented beings, erroneous cosmologies and wholly imaginary events. She knew that perfectly well. Still, the question would not leave her. There was a difference between a text which happened to contain errors and a text whose newly acquired history appeared to be assembling itself around a specific fictional description. If the process required consistency, where did that requirement end?

She began listing the book's supposed sources. Ancient records beneath a nameless desert city. A civilisation predating humanity. Unfamiliar biological forms. Locations whose geometry did not conform to ordinary space. Entities associated with the structure of time itself. She stopped writing. The list was absurd. It was also, she noticed, almost identical to the sequence of topics in the City of Dis reconstruction. Miriam turned to the final page of the essay's printed draft.

For the first time, she wondered whether scholarship could constitute an experiment without its author knowing what was being tested. And whether publication might, under circumstances no one had yet imagined, be something more consequential than the dissemination of a conclusion. She looked at the date of the article. Then at the dates of the earliest anomalous catalogue records. A century separated them. Her research had been completed only days earlier.

If there was any causal relationship between the two, it did not run in the direction to which ordinary experience had accustomed her. She wrote one final sentence.

The evidence precedes the event which may have caused it.

Then she put down her pen. It was nearly five in the morning. Outside, the first suggestion of daylight had begun to dilute the darkness above the rooftops. Somewhere in London, in a temperature-controlled store, an old volume lay in its protective enclosure.

Its paper had been manufactured centuries before Miriam's birth. Its type had been set by printers long dead. Its binding had been repaired, its pages examined, its ownership recorded and its existence documented by generations of scholars. It possessed an entirely respectable history. The only difficulty was that, until very recently, none of that history had happened. And if the same process was still continuing, the world's libraries might soon be the least interesting places in which to look for evidence.

Chapter II 2 of 10

The Problem of Provenance

The British Library was not, Miriam reflected, an institution designed to accommodate the unexpected. Its procedures assumed that objects existed, that they possessed histories, and that those histories could, at least in principle, be reconstructed from documentary evidence. A book might be misidentified, misdated, incorrectly attributed or even deliberately falsified. Such problems were familiar to librarians, conservators and textual historians. Indeed, a considerable portion of their professional lives was devoted to resolving them.

What was not contemplated was the possibility that a book might possess a perfectly satisfactory history despite having had no existence whatsoever until the previous Tuesday. Miriam arrived shortly after ten.

She had travelled down from Cambridge that morning, bringing a laptop, two notebooks, a copy of the City of Dis essay and a growing collection of photocopies whose mutual incompatibility was becoming difficult to ignore. She had also brought her own memory, though she was beginning to regard it as the least reliable item in her possession. Oliver Stroud met her near the entrance to the reading rooms.

He was a narrow, rather colourless man in his early fifties, with thinning hair, rimless spectacles and the distracted courtesy of someone accustomed to treating books with considerably more interest than people. He introduced himself without unnecessary ceremony.

"Dr Vale. Thank you for coming."

"Miriam, please."

"Oliver."

They shook hands. He glanced at the folder beneath her arm.

"I gather you've brought some material."

"Enough to demonstrate that one of us is wrong."

"Only one?"

She looked at him. For the first time since receiving his correspondence, she smiled.

"I was hoping you'd say something reassuring."

"I can offer you a book cradle and a catalogue. Reassurance was never in the departmental budget."

He led her through the building. They spoke little until they reached a small room adjoining the department's conservation facilities. It contained a large table, several chairs, adjustable lamps and a collection of book supports arranged on a wheeled trolley. There were no ceremonial precautions, no atmosphere of mystery and certainly no suggestion that they were about to examine anything other than an old printed volume. Miriam was grateful for that. She had encountered too many people who considered unusual historical objects improved by theatrical treatment.

Oliver placed his folder on the table.

"I've arranged for us to examine the book here rather than in the public reading room. Its binding is fragile, and there are restrictions on opening it beyond a certain angle."

"Is it genuine?"

He looked mildly offended.

"That rather depends upon what you mean."

"Is it a fifteenth-century printed book?"

"Every indication says yes."

"And the text?"

"That's a separate question."

"Quite."

Oliver opened his folder.

"I've been through the acquisition records. The provenance is better than I initially suggested. Not complete, certainly, but better than we usually have for an anonymous printing of this period."

He spread several documents across the table.

"The Museum acquired it in 1891 from the estate of a private collector. Before that, it belonged to an antiquarian who purchased it from a bookseller in Leiden. We have correspondence concerning the purchase, an invoice, and a description of the binding."

"All contemporary?"

"Apparently."

"Apparently?"

Oliver adjusted his spectacles.

"I thought it best not to overstate the case."

Miriam examined the first document. It was a nineteenth-century letter, written in the small, disciplined handwriting of someone who had learned penmanship before the invention of the typewriter. The author described a collection of miscellaneous medieval and early printed material, including several theological works, an incomplete astronomical treatise and a Latin volume attributed to an Arabian writer. The relevant entry was unremarkable. It mentioned that the book was defective, having lost several preliminary leaves, and that its contents were principally concerned with cosmological speculation.

There was a note about its poor binding. Nothing about demons. Nothing about monstrous entities. Nothing even faintly suggestive of the reputation the Necronomicon would subsequently acquire in twentieth-century fiction.

"This is the sort of thing I find particularly difficult," Oliver said.

"Why?"

"Because it's so ordinary."

"Forgery generally benefits from ordinariness."

"Yes. But forgers are usually trying to establish something. This material doesn't appear to be trying to establish anything at all."

He indicated a second document.

"The valuation was modest. The dealer originally thought the volume might be a defective copy of an Arabic astronomical work translated into Latin. He wasn't particularly interested in it."

"Which makes him an unusually discriminating occultist."

Oliver smiled faintly.

"Quite."

Miriam turned the documents over.

"How far back does the provenance go?"

"With confidence? The middle of the nineteenth century. With reasonable probability, considerably earlier. There are ownership inscriptions in the volume itself."

"Any mention of the title before that?"

"Several. Unfortunately, some are difficult to interpret."

"Because they're damaged?"

"Because the title is not always given in the same form. The Latin spelling varies. One seventeenth-century inventory appears to describe the book simply as a work concerning the names of things."

Miriam looked up.

"The names of things?"

"That's my translation. The Latin is ambiguous."

"May I see it?"

Oliver passed her a photocopy. She read the relevant sentence. The phrase was indeed ambiguous. It could refer to classification, nomenclature or the philosophical problem of universals. In a medieval theological context, it would not have been particularly remarkable.

"What interests me," Oliver continued, "is that none of these descriptions seems to recognise the book as especially important."

"Which supports the possibility that we're dealing with an ordinary medieval text later appropriated by Lovecraft."

"Yes."

"Except that Lovecraft explicitly said he invented it."

"Yes."

"And the book's alleged publication history is the same as the one he invented."

"Yes."

"And neither of us has an adequate explanation for that."

"No."

Oliver folded his hands.

"I thought we might begin by establishing what we actually have, rather than what either of us remembers."

Miriam regarded him for a moment.

"An excellent suggestion."

The volume arrived in a shallow conservation tray. It was smaller than Miriam had expected. The Necronomicon had acquired, through a century of imaginative elaboration, the popular appearance of a monstrous object: enormous, iron-bound, covered in symbols, perhaps wrapped in human skin. The actual book resembled nothing so much as an elderly and rather neglected collection of sermons.

Its boards were covered in dark brown leather, cracked along the hinges. The spine had been repaired with a material slightly different in colour and texture from the original covering. A small paper label, bearing a modern shelfmark, had been attached near the lower edge. There were no metal clasps. No decorative bosses. No evidence that anyone had ever considered it especially frightening. Miriam leaned forward.

"That's disappointing."

Oliver looked at her.

"Did you expect something more spectacular?"

"Given its reputation, perhaps a faint smell of brimstone."

"Old leather and repaired paste. Very little brimstone in our acquisitions budget."

He removed the protective covering. There was an ordinary smell of dust, ageing paper and the faintly acidic odour of deteriorating adhesive. Miriam found it immensely reassuring.

The book was real.

That much, at least, could no longer be denied. It occupied space. It had weight. Its binding was worn in ways consistent with prolonged use and indifferent storage. The surface of its leather carried minute scratches, abrasions and darker patches where hands had handled it over generations. Oliver adjusted the supports.

"Clean hands are preferable to gloves for this sort of examination. Just don't touch the ink unnecessarily."

Miriam nodded. He opened the front board. On the inside was an old ownership inscription, partly obscured by a later bookplate. The inscription was dated 1614. Beneath it were two additional names, one apparently from the eighteenth century and another from the nineteenth. The bookplate belonged to the collector from whose estate the British Museum had acquired the volume. Miriam studied the paper. There was nothing unusual about its appearance.

The bookplate showed the faint edge staining one would expect from a nineteenth-century adhesive. Its corners had lifted slightly. The older ink had penetrated the paper irregularly, producing the familiar feathering seen in historical manuscripts.

"Has anyone analysed these inscriptions?"

"Not chemically. They've been examined under magnification and appropriate illumination. Nothing suggests modern additions."

"And the binding?"

"Partly seventeenth-century, with later repairs."

"Which means the book existed in the seventeenth century."

Oliver hesitated.

"Which means the binding and inscriptions are consistent with that history."

"Very careful."

"I've become rather attached to careful distinctions."

He turned the first surviving leaf. The opening was damaged, with a narrow strip missing along its outer margin. There was no title page. The surviving text began partway through a Latin sentence. The typography was dense, irregular black-letter printing, with abbreviated forms characteristic of late medieval production. Miriam recognised the general style immediately. She had examined enough incunabula to know that a convincing forgery of such a volume was no trivial undertaking.

The paper showed chain lines and laid lines. Several leaves bore parts of watermarks. The type displayed the slight inconsistencies resulting from repeated use and the imperfect casting of early printing equipment. Even more persuasively, the book possessed printing errors. Some letters had been imperfectly inked. Others were displaced. There were occasional corrections made by hand. One leaf contained a small piece of type printed at an angle. Miriam found herself becoming increasingly irritated.

"Whoever produced this," she said, "understood early printing."

"Or was an early printer."

"That's precisely what bothers me."

Oliver turned another leaf. The text was Latin, though not particularly elegant Latin. It contained awkward constructions and a number of transliterated words which appeared to derive from Arabic. There were passages concerned with the motions of the heavens, the measurement of time and the relationship between celestial observations and terrestrial events. In isolation, much of it might have belonged to an eccentric medieval cosmological treatise. Miriam read a paragraph. Then another. The translation was imperfect, but she could follow its argument.

It concerned the difficulty of establishing whether an observed sequence of events represented the actual order in which those events had occurred. The author distinguished between the order in which things were perceived and the order in which they possessed being. That was hardly unusual in medieval philosophy. What followed was less familiar. The text appeared to suggest that the existence of an event might depend not merely upon its causes but upon the existence of later events requiring it. Miriam read the passage again.

"That's interesting."

Oliver leaned forward.

"The passage about causes?"

"Yes."

"Professor Marsh discussed it."

"Did she?"

"Her interpretation was that Alhazred was attempting to reconcile Aristotelian causality with a rather eccentric form of Neoplatonism."

"That sounds plausible."

"I thought so."

Miriam translated the sentence aloud.

"'For a thing is not secured in being by the beginning of its causes alone, but also by the completion of those things which require it.'"

She paused.

"That's an awkward construction."

"Most of the translation is."

"It almost sounds as though he's proposing a reciprocal relationship between past and future."

"An anticipation of backwards causation?"

"Not necessarily. It might simply be a metaphysical observation about the completeness of divine knowledge."

She looked at the next paragraph. There was a description of stars which had ceased to exist before their light reached observers. Again, the idea was not wholly impossible as ancient speculation. Philosophers had occasionally entertained remarkably sophisticated conjectures without possessing the means to test them. Nevertheless, the language was peculiar.

Alhazred, or whoever had written the passage, seemed less concerned with the delay of light than with the possibility that an observed star might acquire a different history depending upon what subsequently became known about it. Miriam stopped reading.

"Can you show me the collation?"

Oliver reached for his notes.

"Of course."

They spent the next two hours examining signatures, gathering boundaries, watermarks and the physical construction of the volume. It was a deliberately tedious exercise. Miriam insisted upon it. Spectacular claims could wait. Before considering the implications of the text, she wanted to establish that the object itself behaved as an ordinary book. It did. The gatherings were consistent with the printing practices of the period.

The paper showed variations of thickness and texture that were natural rather than artificially uniform. There were occasional traces of earlier repairs, including narrow strips of paper used to reinforce weakened leaves. The volume contained imperfections that no sensible forger would have deliberately included unless determined to expend an absurd amount of labour. There were wormholes. Not one or two decorative holes added to suggest antiquity, but a pattern of actual insect damage passing through successive gatherings.

The tunnels began in one leaf, widened slightly through several others and terminated where the paper had evidently become less accessible. Their trajectories corresponded to the arrangement of the gatherings.

"One could reproduce this," Miriam said.

"Certainly."

"With enough effort."

"Yes."

"And an unreasonable amount of expertise."

"Yes."

"But not impossible."

"No."

Oliver sounded almost pleased. Miriam looked at him.

"You do realise that I am attempting to find reasons not to believe the evidence in front of me?"

"I gathered."

"I'd prefer not to be humoured."

"I'm not humouring you. I'm trying to establish what it would take to manufacture the object."

"And?"

"An extraordinary amount of effort, considerable specialist knowledge, access to authentic historical materials, and a convincing documentary history extending across several centuries."

"Not impossible."

"No."

"Good."

He considered her for a moment.

"May I ask you something?"

"Of course."

"Why is it so important that this particular book should be false?"

Miriam looked at the open volume.

"Because I know where it came from."

"You know where the literary idea came from."

"No. I know where the specific bibliographical history came from. Lovecraft invented it. We can trace the invention through his stories and correspondence."

"Could he have been mistaken?"

"About inventing it?"

"People misremember their sources."

"Certainly. But this isn't a vague recollection of something he might once have read. The invention is documented. He discussed it with other writers. He elaborated the chronology for literary purposes."

"Perhaps the original book influenced him without his remembering."

"That might explain a name, a title or an isolated detail. It doesn't explain why the book's actual history should reproduce the particulars of his imaginary history so precisely."

Oliver said nothing. Miriam continued.

"If Lovecraft had borrowed an authentic medieval title, we'd expect to find an independent textual tradition, however obscure. Instead, we have an historical record which appears to reproduce his fiction."

"Yet the physical evidence predates him."

"Apparently."

"You've used that word rather frequently."

"Under the circumstances, I think it's earned its keep."

At half past twelve, a conservator named Dr Helen Mercer joined them. She was a brisk woman with short grey hair and the air of someone who had very little patience for romantic ideas about old books. Oliver introduced her as the specialist responsible for the volume's most recent examination.

"Dr Vale is interested in whether the book could be a modern fabrication."

Helen looked at Miriam.

"Anything could be a fabrication in principle. What would you like to know?"

"Whether the materials are consistent with the assigned date."

"Yes."

"Whether you've found any evidence of modern construction."

"No."

"Whether the paper has been tested."

"Non-destructively. The results are compatible with the expected period."

"And the ink?"

"Likewise. But those results aren't a precise dating method."

"Quite."

Helen examined the volume.

"There's something you may find useful. We have photographs from the restoration carried out in 1968."

Oliver nodded.

"I mentioned those."

"I don't think you've shown her the originals."

She left the room and returned several minutes later with a folder containing photographic prints mounted on stiff backing. They were monochrome photographs, with the soft tonal range and slight imperfections characteristic of their period. One showed the volume before restoration. Another showed the damaged spine. A third displayed several separated leaves during repair. Miriam examined them under the lamp. The damage corresponded precisely to the book in front of her.

A tear in one leaf had been repaired with a narrow strip of tissue. The same repair appeared in the photograph and in the present volume. Several wormholes were visible in exactly the expected positions. There was even a small dark stain near the lower margin of one page. It appeared in the old photograph. It remained on the paper.

"These prints are original?" Miriam asked.

"As far as the records establish."

"Not modern reproductions?"

"No. They're part of the conservation archive."

"And the negatives?"

"We retain some negatives from that period. I can check whether these particular ones survive."

"Please."

Helen studied her.

"May I ask what you're trying to establish?"

Miriam hesitated. It was difficult to explain without sounding either foolish or alarmingly unstable.

"I'm concerned about the possibility of a highly sophisticated literary forgery."

Helen looked at the book.

"One would have to be extraordinarily committed."

"I agree."

"And what would be the purpose?"

"That's what I'm trying to understand."

The conservator considered this answer for a moment. Then she shrugged.

"People have done stranger things."

Miriam was inclined to agree. After lunch, Oliver took Miriam to examine the associated documentary material. The most useful item was an accession register from 1891. Unlike the photocopy he had supplied, the original made the peculiar chronology unavoidable. It was a substantial bound ledger, its pages ruled into columns for dates, descriptions, sources and acquisition details. There were entries for hundreds of objects. Miriam turned to November. The Necronomicon appeared among a group of books acquired from the same estate. Its entry occupied three lines.

There was no sign of insertion or alteration. The ink had the same general character as the surrounding entries. The handwriting belonged to the same clerk. The spacing was ordinary. A small correction had been made to the recorded price. The correction was initialled. Miriam found this particularly persuasive. The forgery hypothesis now required someone not merely to fabricate a book but to reproduce an entire chain of mundane administrative accidents. The collector's estate had existed. The dealer had existed. The Museum officials had existed.

Their correspondence referred to transactions with entirely unrelated parties. A forged Necronomicon would have to be embedded in this network without disturbing the rest of it. That was conceivable. But the complexity was growing.

"There's an additional problem," Oliver said.

"What?"

"The photographs."

"You showed me those."

"Not the ones I'm thinking of."

He opened another folder.

"The Museum conducted a photographic survey of selected rare printed works in the 1930s. The Necronomicon was included, presumably because of its unusual typography."

He placed a photograph on the table. It showed several volumes arranged together, with their spines facing outward. The Necronomicon was one of them. Miriam recognised the distinctive repair to its lower spine.

"The image was used in a departmental report," Oliver said. "We have a printed copy."

He produced it. The photograph was reproduced on a page surrounded by text concerning the preservation of early printed books. The book appeared incidentally. No attention was drawn to it. There was no caption identifying it individually. It was simply one old volume among several others. Miriam examined the report. The paper was yellowed and slightly brittle. Its binding was consistent with institutional publications of the period. The text included references to conservation practices long since abandoned.

"Where did you obtain this copy?"

"From the departmental archive."

"Other copies?"

"Presumably. It's a printed report."

"Can we locate one outside the Library?"

"I imagine so."

"Please do."

Oliver wrote a note. Miriam returned to the photograph. The volume was unmistakable. And yet she could think of no sensible reason for a forgery to include such an incidental reference. The purpose of a forged document is to convince someone of something. This photograph appeared to be convincing nobody of anything. It had existed, apparently, for nearly ninety years without attracting the slightest attention. The more evidence they uncovered, the less the book resembled a deliberate deception. It resembled an ordinary historical object.

That was precisely the problem. At three o'clock, Oliver suggested they take a break. They found a quiet table in the café. Miriam ordered coffee. Oliver chose tea and a sandwich, then seemed to forget about the latter almost immediately. For several minutes neither spoke. Eventually Oliver said, "I think I should tell you something." Miriam looked up.

"I rather hoped you would."

"When I first read your essay, I was annoyed."

"Understandable. It contradicted your catalogue."

"More than that. It contradicted something I regarded as elementary professional knowledge."

"The book's existence."

"Yes."

He stirred his tea.

"I'd worked with the volume before. Not extensively, but enough to know it was genuine. I'd referred to it in lectures. I'd seen the conservation reports."

"And yet you remembered teaching that it was fictional."

"Yes."

"Can you account for that?"

"No."

Miriam waited. Oliver continued.

"I've been trying to reconstruct my own understanding of the book. The difficulty is that I seem to possess two different histories of it."

"Two memories?"

"Not quite. The memories don't simply contradict one another. They have different degrees of completeness."

He looked uncomfortable.

"I remember learning about Lovecraft's invention as a student. I remember being amused by the number of people who believed the Necronomicon was real. I remember discussing it with colleagues as an example of fictional bibliography."

"And the other history?"

"I remember handling the actual book. I remember preparing a lecture on its Latin transmission. I remember an argument with a colleague about the date of the printing."

"Which memory feels more convincing?"

"Both."

"Equally?"

Oliver considered.

"No. That's what disturbs me."

"Go on."

"The memories concerning the physical book are becoming more detailed."

Miriam put down her cup.

"What do you mean?"

"Yesterday I remembered examining it several years ago, but I couldn't remember the circumstances. This morning I recalled that I'd been asked to identify an inscription on the front pastedown. I remembered the colleague who asked me. I remembered where we were sitting."

"Could you verify any of that?"

"Yes. There's an email exchange."

"From when?"

"2017."

"And the fictional history?"

"That's becoming less distinct."

He stared into his tea.

"I know I taught that the book was imaginary. I can remember saying so. But I can no longer remember which lecture it was, or who attended."

Miriam felt an unpleasant tightening in her stomach. She had experienced something similar that morning. On the train, she had attempted to reconstruct her research on Lovecraft's fictional bibliography. Certain memories had remained clear. Others seemed increasingly uncertain. She could recall looking up the British Museum reference in History of the Necronomicon. She could recall making notes about the Wormius translation. But she had also found herself remembering a short discussion of the genuine Latin printing in a textbook she had used during her postgraduate work.

She had dismissed the recollection as confusion. Now she was less certain.

"Did you check the textbook?" Oliver asked when she described it.

"No."

"Why not?"

"Because I was frightened it would be there."

Oliver looked at her without speaking. It was the first time either had admitted to being frightened. Miriam continued.

"I know how ridiculous that sounds."

"Not especially."

"I'm a textual historian, Oliver. I spend my professional life examining how unreliable records become accepted as authoritative. I know perfectly well that memory is reconstructive, that people misremember sources, that confidence is not evidence."

"So do I."

"Yet I have a notebook containing references I don't remember writing, and you've got lectures you don't remember preparing."

"Yes."

"And we're sitting here treating the possibility of some extraordinary alteration of historical reality as though it were a reasonable hypothesis."

Oliver gave a small, humourless laugh.

"When you put it like that, I admit it sounds rather bad."

"It's worse than bad. It's intellectually disgraceful."

"Do you have a better explanation?"

"Not yet."

He nodded.

"Then perhaps we should continue being disgraceful until we find one."

They returned to the archive. Miriam wanted to examine the conservation correspondence, particularly the records surrounding the 1968 restoration. Oliver retrieved a box containing folders of typed reports, photographs, letters and handwritten notes. The documents were exactly what she expected. The binding had been weakened by repeated opening. There was evidence of earlier repairs. Several leaves had suffered damage along their outer margins. A conservator had recommended stabilisation rather than extensive restoration. The work had been authorised, recorded and completed. Nothing about the procedure was remarkable.

Miriam examined the original conservation report. The final paragraph mentioned that the volume was defective at both beginning and end, and that several internal leaves appeared to be missing. A collation had been attempted, though the unusual structure of the text made reconstruction difficult. There was a handwritten note in the margin.

Compare with Marsh transcription?

Miriam frowned.

"Eleanor Marsh?"

Oliver leaned over.

"That would be impossible."

"Why?"

"Marsh didn't publish her transcription until 1998."

"Perhaps a different Marsh."

"Possibly."

They examined the handwriting. The note had been added in pencil. It was not part of the original typed document. There was no date. Miriam considered it.

"Someone could have added that at any time."

"Yes."

"There's nothing anomalous here."

"No."

She looked again. The pencil was faint, but the words were legible. The note appeared to refer to a transcription produced thirty years after the conservation report. An entirely ordinary later annotation would explain it. She was almost relieved. Then she turned the page. The following report, dated 1971, contained a typed reference to Marsh's work. Miriam stopped.

"Oliver."

He looked at the page. The reference occurred in a section discussing the identification of a damaged passage. It gave a title, a publication date and a page number. The title was that of Marsh's 1998 article. The date was correct. The page number was correct. Oliver read the sentence twice.

"That isn't possible."

"Exactly."

He examined the paper. The reference was not a handwritten annotation. It formed part of the original typescript, with the same typeface and alignment as the surrounding text. Miriam looked at the document under the lamp. There were no signs of erasure, overwriting or physical insertion. The paragraph had been typed continuously. She read it aloud.

"'The disputed reading should be compared with the reconstruction proposed by Marsh in her study of the Alhazred tradition.'"

She turned to Oliver.

"Who typed this?"

"The conservation office, presumably."

"Can we establish the name?"

"There should be a signature."

He examined the final page. The report had been prepared by a conservator named Margaret Ellison, who had retired decades earlier.

"Is she still alive?"

"I don't know."

Miriam returned to the reference. There were several possible explanations. A later replacement page might have been inserted into the file. A modern transcript might have been mistaken for an original. Someone might have altered the report deliberately. But the paper appeared to belong with the surrounding sheets. The typewriter characteristics were consistent. The staple holes aligned. And the reference was embedded in a perfectly ordinary discussion of damaged text. She looked at Oliver.

"This is the first thing we've found that isn't merely improbable."

He nodded.

"It's impossible."

"Only if the document is what it appears to be."

"Which we haven't established."

"No."

She closed the folder.

"Let's not rush into inventing new physics because someone may have substituted a page in an archive."

"Agreed."

But neither sounded particularly reassured. They spent the next hour checking the document's provenance. The report was listed in a departmental register from 1971. Its reference number appeared in a later conservation summary. There were carbon copies of related correspondence. They located an inventory prepared in 1984 which listed the same report. None of this demonstrated that the anomalous reference had been present in the original. It did, however, make deliberate insertion more difficult to explain. Finally, Oliver found a photocopy made in 1986.

The disputed paragraph was visible. The reference to Marsh's 1998 article was present. Miriam placed the photocopy beside the typescript. They agreed in every detail.

"Could the photocopy have been made later?" she asked.

"Certainly."

"Could the date be wrong?"

"Yes."

"Could someone have fabricated the entire sequence?"

"Yes."

She looked at him.

"You're becoming very accommodating."

"I thought that was what you wanted."

"I want an explanation."

"So do I."

They examined the surrounding paperwork. The photocopy belonged to a bundle of documents assembled for an internal audit. The other papers carried dates, signatures and references consistent with the mid-1980s. Again, the entire collection could conceivably have been manipulated. But why? Who would have devoted such extraordinary effort to fabricating a conservation history for a fictional book, then buried the evidence in an institutional archive where almost nobody would examine it? And how would such a person have anticipated a scholarly article published decades later?

Miriam opened her laptop. She found Eleanor Marsh's paper. The disputed reference was genuine. The title, page number and argument all matched. She checked the publication history. The journal had appeared in 1998. The article's submission date was given as November 1997. There was no indication of an earlier version. Oliver was standing beside her.

"Could Marsh have drawn upon Ellison's report?"

"That would explain the similarity of the argument, not the citation."

"Unless the report was retrospectively revised."

"Yes."

"By whom?"

"I don't know."

"Nor do I."

Miriam closed the document. She had spent the day attempting to establish whether the Necronomicon possessed a genuine historical provenance. Instead, she was beginning to suspect that the provenance itself was the phenomenon requiring explanation. The book was not an isolated anomaly. It was the centre of an expanding network of evidence. And some parts of that network appeared to have been constructed in the wrong temporal order. At five o'clock, Helen Mercer returned. She had located the negatives associated with the 1968 conservation photographs.

They were stored in archival sleeves, each labelled with a reference number and description. Several corresponded to the prints examined earlier. Miriam inspected them. There was no obvious sign of alteration. The images showed the damaged volume before restoration, during treatment and after completion. The same stains, wormholes and tears appeared in the expected positions. The physical correspondence was exact. Helen watched her examination with increasing curiosity.

"Is there something particular you're looking for?"

"A discrepancy."

"Between the photographs and the book?"

"Yes."

"Why?"

Miriam considered her answer.

"Because I'm trying to establish whether the evidence has a common source."

Helen frowned.

"Of course it has a common source. It's the same book."

"That's one possibility."

"What would the alternative be?"

Miriam said nothing. The conservator looked at Oliver. He appeared no more inclined to explain. Eventually Helen said, "Well, if you're concerned about the authenticity of the negatives, I can tell you that they appear consistent with the period. But photography isn't magic. An image can be manipulated, and the age of the material doesn't guarantee the authenticity of its contents."

"Quite," Miriam said.

"Although I must say I'd be interested to know what possible motive anyone would have for such an undertaking."

"So would I."

Helen gathered the materials. Before leaving, she paused.

"There's one thing."

Miriam looked up.

"The conservation file is rather unusual."

"In what respect?"

"The number of examinations."

Oliver frowned.

"What do you mean?"

"Most books in this condition are examined when there's a particular need. A repair, a request for consultation, a concern about deterioration."

"And this one?"

"It's been examined repeatedly."

"By whom?"

"Different people. Different periods. Sometimes without an obvious reason."

Miriam leaned forward.

"How often?"

Helen consulted her notes.

"There's a conservation review in 1968, another in 1971, one in 1984, another in 1996, then several more recent assessments."

"That doesn't sound particularly extraordinary."

"No. But the reports keep returning to the same question."

"Which question?"

Helen looked at them.

"Whether the volume is genuine."

Miriam felt a sudden interest.

"Do the reports explain why its authenticity was doubted?"

"Not consistently. Some mention the unusual attribution. Others discuss the absence of satisfactory textual parallels."

"And the conclusions?"

"Every report seems to conclude that the book is genuine."

"Then why repeat the examination?"

Helen considered.

"That's what I find odd."

She opened the most recent report.

"The same thing happened last year. There was a request for an examination, apparently prompted by concerns about the book's existence."

"Whose request?"

"I haven't established that."

Miriam looked at Oliver.

"Can we find out?"

"Of course."

Helen returned to her folder.

"There's another curious detail."

"What?"

"The requests become more frequent as we approach the present."

She read from the list.

"1996. 2008. 2017. 2022. Then three separate enquiries this year."

Miriam said nothing.

"One was in March. Another in September. The third was last week."

"Last week?" Oliver repeated.

"Yes."

"Who requested it?"

Helen checked the record.

"Apparently someone involved in a literary research project."

She handed the document to Miriam. The request was dated four days before the publication of the City of Dis essay. The applicant's name was Miriam Vale. For a moment, none of them spoke. Miriam examined the document. It was an ordinary electronic request, printed and filed with the conservation material. It gave her name. Her professional affiliation. Her email address. The subject line referred to the Necronomicon and the history of its Latin editions.

The request asked whether the Library could confirm that a particular fifteenth-century printing was physically present. Miriam read it carefully. Then she read it again.

"I didn't send this."

Oliver took the page.

"Are you certain?"

"Absolutely."

He examined the address.

"It's yours?"

"Yes."

"The affiliation?"

"Correct."

"And the project?"

"Yes."

"Could you have forgotten?"

Miriam gave him a look that would have silenced a doctoral candidate at twenty paces.

"I might forget a telephone conversation or an informal enquiry. I do not forget submitting a formal request about the existence of a book whose non-existence was central to an article I was preparing."

Oliver returned the page.

"Is it possible someone else submitted it using your details?"

"Certainly."

"Can you check your sent messages?"

"Of course."

She opened her laptop. The Library's guest network took a few moments to connect. She searched her email archive. There was the message. Sent four days before publication. Addressed to the relevant department. The subject line and body were identical to the printed request. Miriam stared at the screen. Oliver leaned over.

"Does it show as sent?"

"Yes."

"Any indication of forwarding or an automated process?"

"No."

She opened the message headers. Nothing suggested an obvious anomaly. The email appeared to have originated from her own account through the usual mail service. There was no sign that anyone had impersonated her. She searched the days surrounding it. The request sat among ordinary correspondence concerning the City of Dis essay. Before it was a message discussing a correction to the description of the Greek translation. After it was a note concerning the historical Ole Worm. The sequence was entirely plausible.

Except that Miriam remembered writing both surrounding messages. She did not remember the one between them. She clicked the date. The message had been sent at 14:32. She opened her handwritten diary. On that afternoon she had been working at home. She remembered the work clearly. She had spent several hours examining Lovecraft's published correspondence, checking whether the invention of the Necronomicon could be traced to a particular story. She had concluded, as she had always concluded, that the book was fictional.

There was a brief diary entry concerning the research. Beneath it, in her own handwriting, was a line she had not noticed before.

BL enquiry submitted. Await confirmation.

She closed the diary. The room had become very quiet. Helen spoke first.

"Dr Vale, are you all right?"

"Yes."

"You look rather pale."

"I'm beginning to feel that my professional habits have become a liability."

Oliver sat down beside her.

"There's something we need to decide."

"What?"

"Whether we trust documents more than memory."

Miriam considered the question. It ought to have been straightforward. Historians had always understood that memory was unreliable. Documents were not infallible, but they provided material evidence that could be examined, compared and challenged. The entire discipline depended upon the distinction. Yet what if the documents themselves were being altered? What if the evidence was not merely wrong, but possessed a different history from the one the investigator remembered? Then the normal hierarchy of evidence would become a trap.

The better the documentary record, the more effectively it would conceal the alteration.

"We need something independent," she said.

Oliver looked at her.

"Independent of what?"

"Independent of the book's history. Something that records the discrepancy rather than simply participating in it."

"Such as?"

She considered.

"A controlled observation. A prediction. Some way of establishing a change before it occurs."

Helen looked between them.

"I think perhaps one of you should explain what this is actually about."

Miriam turned to her.

"Dr Mercer, have you ever encountered an object that appeared to have acquired a history it didn't previously possess?"

The conservator stared at her.

"No."

"Nor have I."

"Then I'm afraid I don't understand."

"Neither do we."

By the time Miriam left the Library, she had copies of the relevant documents, photographs of the volume and permission to return for a more detailed examination. She also possessed something she had not expected to acquire. A methodological problem. It was no longer sufficient to ask whether the Necronomicon existed. The evidence established that an object with that title, physical characteristics and bibliographical history was present in the Library.

The problem was that the evidence also appeared to establish its existence at times when, according to Miriam's recollection and the literary history she had studied, no such object had existed. And her own records were beginning to accommodate the contradiction. This was not merely an issue of memory. Miriam could imagine having forgotten sending an email. She could imagine having misremembered a lecture or overlooked a catalogue entry.

What she could not easily explain was how a preservation report written in 1971 could contain a reference to a scholarly article published in 1998. That anomaly might still have an ordinary explanation. A substituted page. A modern transcript mistaken for an original. An elaborate deception. The task was to eliminate those possibilities systematically. She took the train north. The carriage was crowded with commuters. People read newspapers, watched films on their telephones and discussed the minor frustrations of their working days.

Nothing in their behaviour suggested the possibility that the historical record might have become unstable. Miriam watched the passing landscape. Fields, industrial estates, roads and railway embankments moved past the window beneath a darkening sky. Every feature possessed a history.

The railway had been constructed at a particular time, using materials produced by particular processes. The buildings had been designed, financed, erected and altered. The fields had been cultivated, divided and repurposed. Their present existence was the outcome of innumerable earlier events. The world was not merely a collection of objects. It was a collection of histories. And if those histories could change while leaving the present apparently coherent, how would anyone know? Miriam opened her laptop. She began drafting a memorandum.

Preliminary observations concerning the Wormius printing

1. A physical volume exists whose characteristics are consistent with a fifteenth-century Latin printing.

2. Independent documentary sources assign the volume a provenance extending at least to the nineteenth century.

3. The bibliographical tradition agrees in remarkable detail with a history demonstrably invented by H. P. Lovecraft.

4. Certain archival records contain apparent references to publications or events occurring after the records' stated dates.

5. At least two investigators report memories incompatible with their own documentary records.

6. These discrepancies are not presently sufficient to establish any extraordinary conclusion.

She considered the list. It was restrained. It was reasonable. It contained nothing that would embarrass her if read by a colleague, provided the colleague was willing to tolerate a certain amount of eccentricity. She added a seventh point.

7. The quantity of mutually supporting evidence appears to be increasing.

Then she stopped. Was that actually true? She had encountered more evidence during the day because she had been looking for it. The growing documentary record might reflect nothing more mysterious than a widening investigation. It was an elementary distinction. She deleted the sentence. Then she opened the original City of Dis essay. There was a passage discussing the extent to which Lovecraft had furnished his imaginary book with a convincing textual history.

The essay explained how successive editions, translations, missing copies and scholarly disputes gave the fictional object the appearance of independent existence. Miriam reread the paragraph. Something about its language troubled her. Not because it had changed. It hadn't. She recognised every sentence. What troubled her was that the research had followed precisely the method by which the book's fictional history now appeared to have become real. The essay had established a coherent structure. A supposed Arabic original. A Greek translation. A Latin translation. Printed editions. Institutional holdings.

Conflicting textual witnesses. Scholarly disputes. And finally, an attempt to reconstruct the relationship between the book's contents and the nature of the universe it described. The research had not merely listed fictional claims. It had connected them. It had determined which claims were compatible, which contradicted one another, and which could be reconciled by assuming missing intermediate sources. In doing so, it had constructed something resembling a historical model. Miriam looked at the memorandum. She added a final observation.

8. The newly documented historical object corresponds closely to the coherent reconstruction produced by the research, rather than to the full range of inconsistent popular versions.

She stared at the sentence. That was potentially important. The multitude of modern Necronomicons sold to occult enthusiasts did not appear in the institutional record. Neither did the more extravagant inventions of later writers. The historical book corresponded, as far as she could establish, to the narrower structure reconstructed from Lovecraft's own material. If someone had fabricated the evidence, they had done so with an exceptionally precise understanding of the essay. If the phenomenon was real, the implications were considerably worse. She closed the laptop.

The train was approaching Cambridge. At home, Miriam made herself tea and returned to the problem of the conservation report. She had spent the journey considering possible ways to establish its authenticity. The most obvious approach was to locate independent copies of the 1971 report. If they existed in other archives, their physical characteristics could be examined. The same applied to the 1930s departmental publication. She drafted an email to a colleague specialising in the history of the British Museum's collections. Then she remembered Eleanor Marsh.

The scholar whose 1998 article had apparently been cited in 1971 might be able to clarify the textual relationship. Miriam had not spoken to her for several years. She found the relevant contact details and began writing.

Dear Eleanor,

I hope you're well.

I'm investigating a rather peculiar bibliographical issue concerning the Latin Necronomicon tradition.

I have encountered a conservation report dated 1971 which appears to cite your 1998 paper on Alhazred.

I assume there is a straightforward explanation involving a later annotation or replacement transcript, but I wondered whether you might have encountered the report during your research.

I'd also be interested to know whether you examined the British Library's fifteenth-century copy directly.

Best wishes,

Miriam

She hesitated before sending it. The phrase Latin Necronomicon tradition appeared perfectly natural. That disturbed her. Only days earlier she would have considered the same phrase a description of Lovecraft's fictional bibliography. Now she was using it to refer to an apparently genuine body of medieval texts. She sent the message. Then she retrieved her original working notes. There was a particular folder containing early drafts of the City of Dis essay and the sources used to check its chronology. She opened the oldest file.

The document had been created several weeks before publication. It contained the initial outline, several bibliographical questions and a list of uncertainties to resolve. She scrolled through it. The references to Lovecraft were familiar. The discussions of Alhazred's supposed chronology were familiar. Then she found a section she did not remember.

It was headed:

Independent witnesses and the problem of textual priority

Beneath the heading was a short paragraph.

The historical evidence for the Wormius printing remains difficult to reconcile with the established literary chronology. Particular attention should be given to the possibility that the physical witnesses are dependent upon the fictional reconstruction rather than the reverse.

Miriam stared at the screen. The paragraph expressed exactly the suspicion she had developed that afternoon. Yet the file had been created weeks earlier. She checked its modification history. There were several revisions, all apparently predating publication. She opened the earliest saved version. The paragraph was present. She checked a backup copy stored on an external drive. It was there too. She had no memory of writing it. But the prose was hers. The style, the phrasing, even the preference for the term textual priority were entirely characteristic.

She sat back. The room seemed suddenly smaller. If the paragraph had genuinely existed weeks earlier, then she had already considered the possibility that the physical Necronomicon was somehow dependent upon Lovecraft's fictional reconstruction. She had not merely discovered the contradiction today. She had apparently investigated it before completing the essay. And then forgotten. Or something had altered the record of her work. Miriam took out her notebook. She turned to the page containing the inexplicable citation to the 1891 accession register.

There was now a second reference beneath it. She was almost certain it had not been there the previous evening. It referred to Eleanor Marsh's 1998 article. There was a page number. An abbreviated quotation. And a short comment in the margin.

This cannot be right.

Miriam stared at the words. They were in her handwriting. She had written them, or something had produced a document indistinguishable from one in which she had written them. The distinction was becoming increasingly difficult to formulate. She turned the page. There was another note.

Check earliest version of essay. Did we introduce the Wormius material, or merely recover it?

She read the sentence twice. The word introduce had been underlined. Beneath it, in smaller writing, was a question.

If the evidence is retrospective, what fixes the direction of causation?

Miriam closed the notebook. For several minutes she sat motionless. There was no sound except the quiet hum of her computer and the occasional noise of traffic outside. Eventually she reached for her telephone. She called Oliver. He answered on the third ring.

"Dr Vale?"

"Miriam. Sorry to call so late."

"Not at all."

"Have you found anything else?"

There was a pause.

"Yes."

His voice had changed.

"What?"

"I've been looking through the historical catalogue descriptions."

"And?"

"The oldest descriptions are less stable than the recent ones."

"What do you mean by stable?"

"They disagree with one another in ways I hadn't noticed before."

"About the date?"

"About the contents."

Miriam sat forward.

"Explain."

"The earliest descriptions are vague. They identify the book, its language and its physical characteristics, but say very little about what it contains."

"That's normal."

"Yes. But the later catalogues become progressively more specific. They describe passages, divisions and subjects."

"Also normal."

"Except that the details correspond with the literary reconstruction."

"Which reconstruction?"

"The one in your article."

Miriam felt her stomach tighten.

"You're suggesting the catalogue descriptions became more accurate over time."

"Not exactly."

"Then what?"

Oliver hesitated.

"It's as though the earlier records establish that the book exists, while the later ones establish what sort of book it is."

Miriam said nothing.

"And there's something else."

"What?"

"I checked the original 1891 accession register again."

"Yes?"

"The entry has a marginal note."

"I saw it."

"No. Not the price correction. Another note."

"What does it say?"

Oliver was silent for a moment. Then he answered.

"It refers to a missing manuscript."

"Which manuscript?"

"The Arabic original."

"Al Azif?"

"Yes."

Miriam looked at her notebook.

"What does it say?"

"Only that the printed Latin text appears to preserve material from a source not represented in the surviving Greek tradition."

"That seems unremarkable."

"Perhaps."

"Oliver, what are you trying to tell me?"

He took a breath.

"The note refers to a particular passage. A passage concerning a city beneath the desert."

Miriam waited.

"It gives a cross-reference to another work."

"Which one?"

"An archaeological report."

"From when?"

"1934."

Miriam closed her eyes.

"The accession register is from 1891."

"Yes."

"And the reference is written in the same hand?"

"As far as I can tell."

"Could it be a later annotation?"

"Of course."

"Then have it examined."

"I intend to."

There was another silence. Miriam could hear papers moving at the other end of the line.

"There's something I haven't told you," Oliver said.

"What?"

"The archaeological report is real."

"That doesn't establish anything."

"I know."

"Where was it published?"

"A specialist journal. The excavation was in the Empty Quarter."

Miriam felt a cold, unpleasant sensation at the base of her neck.

"An archaeological excavation?"

"Yes."

"Of what?"

Oliver hesitated.

"A settlement."

"What sort of settlement?"

"That's the difficulty. The report is very brief. It describes structures uncovered beneath deposits which the excavators thought much older than the settlement itself."

"That doesn't sound especially remarkable."

"No."

"Then why are you concerned?"

"Because the report gives the name of the site."

Miriam said nothing. Oliver spoke again.

"The archaeologists called it Iram."

She closed her eyes. The name belonged to a tradition older than Lovecraft. It appeared in Arabic and Islamic accounts of a lost city associated with the people of ʿĀd. Lovecraft had incorporated the legend into his fictional history of Alhazred, connecting it with the poet's discoveries in the desert. The appearance of Iram in an archaeological report proved nothing. There were countless reasons why researchers might associate a desert site with a legendary city. But the pattern was becoming difficult to dismiss.

"Send me the reference," Miriam said.

"I will."

"And Oliver?"

"Yes?"

"Don't assume the report is genuine merely because you can find it."

"I know."

"Don't assume the site exists."

"I know."

"And don't assume that the name means what we think it means."

There was a long pause. Then Oliver said, very quietly, "Miriam, I've already checked the photographs." She said nothing.

"There are photographs of the excavation."

"That still doesn't prove the interpretation."

"No."

"Or the date."

"No."

"Or the existence of anything described by Alhazred."

"I understand."

Miriam pressed her fingertips against her eyes.

"What do they show?"

Oliver hesitated.

"Not very much. Stone structures. Some inscriptions. The report says the excavation was abandoned because of difficulties with the terrain."

"That sounds familiar."

"Yes."

"Too familiar."

"Yes."

"Send me everything."

"I will."

She ended the call. For several minutes she sat looking at the dark screen of her telephone. Then she opened her research notes once more. The City of Dis reconstruction had examined Lovecraft's account of Alhazred's travels, including his supposed encounters with abandoned cities and remnants of pre-human civilisations. These had been treated, correctly, as elements of a fictional biography. The story had given the imaginary author experiences sufficient to explain how he might have acquired knowledge of things beyond human history.

That was the purpose of the invention. Alhazred needed sources. The Necronomicon needed a history. And that history required the existence of things for Alhazred to have discovered. Miriam opened a fresh document.

She wrote:

THE PROVENANCE PROBLEM

An object may be assigned a history by reference to evidence.

But if the object and its history are both anomalous, the existence of mutually supporting evidence does not necessarily establish which came first.

She stopped.

Then she added:

If the book is becoming real, its provenance is becoming real with it.

She considered the sentence. It was an absurd hypothesis. There was no established physical mechanism by which a fictional object could acquire a past. There was no reason to suppose that historical evidence could be generated retrospectively. Yet the alternative explanations were becoming progressively more elaborate. Every attempt to account for the book as a forgery required a larger conspiracy, a more sophisticated deception and an increasingly improbable degree of coordination among documents separated by centuries. She looked at the next blank line.

Then she typed:

The process may not be confined to the book itself.

The words remained on the screen. Miriam stared at them. An object could acquire a history. A history could require earlier objects. Those objects could require earlier events. And those events might require the existence of things that had never belonged to the world at all. She opened the essay once more. There was a list of the Necronomicon's supposed subjects. The pre-human inhabitants of the Earth. The monstrous products of an ancient biological science. Entities whose existence preceded familiar geological and evolutionary history.

Things said to inhabit regions of space and time inaccessible to ordinary human perception. The text did not merely describe them. Within Lovecraft's fiction, it claimed to record knowledge obtained from their traces, their works and their effects upon the physical world. If the reconstructed book's history now included genuine sources, those sources might demand their own explanations. Miriam remembered a sentence from the Latin passage she had examined that morning.

For a thing is not secured in being by the beginning of its causes alone, but also by the completion of those things which require it.

She had dismissed it as medieval metaphysics. Now she wondered whether she had misunderstood the direction of the argument. Perhaps the author had not been speculating about causation. Perhaps he had been describing an observed property of reality. She glanced at the clock. It was nearly midnight. Her email notification sounded. A message from Oliver.

The subject line read:

Iram excavation - 1934

She opened it. There were several attachments. The first was a scan of a journal article. The second was a photograph showing an excavation trench beneath a desert escarpment. The third appeared to be a close-up of an inscribed stone. Miriam examined the first two. They were unremarkable. The article described a small expedition which had identified possible remains of an ancient settlement. Its conclusions were cautious, its evidence fragmentary. There was no suggestion of supernatural significance. She opened the third attachment.

The photograph showed part of a weathered stone slab. The inscription was damaged and difficult to interpret. Beneath the photograph was a typed caption.

Fragmentary inscription recovered from the lowest accessible level. Translation uncertain.

Miriam enlarged the image. The signs were not familiar to her. She could not determine whether they formed a coherent script, and she was not prepared to accept the excavators' tentative interpretation without independent analysis. She returned to the article. In a footnote, the author suggested that certain features might preserve a tradition associated with the lost city of Iram. Again, nothing extraordinary. Legends were routinely attached to archaeological sites. But the final paragraph contained a statement she found deeply unpleasant.

The excavators had apparently abandoned their work after encountering structures which they could not reconcile with the expected sequence of occupation. The stonework was described as lying beneath deposits of much greater antiquity than any plausible human settlement in the region. The report cautiously suggested a geological explanation. Then it added that further investigation might establish whether the structures were natural formations which merely resembled architecture. Miriam read the passage again. It was carefully written. The author had not claimed to discover a pre-human civilisation.

He had encountered an apparent anomaly and proposed a perfectly sensible explanation. The report ended with a recommendation for additional excavation. She searched for subsequent publications. There were none. Instead, she found a later reference in a geological survey. The survey concerned deep sedimentary structures in the Arabian Peninsula. It mentioned the 1934 excavation as an early observation of a much larger formation. There were diagrams. Measurements. Stratigraphic interpretations. The feature was now recognised, apparently, as an unusual but natural geological structure. Miriam examined the diagrams.

The report was dated 1972. It cited several later studies. She followed those references. The structure had been examined repeatedly. Its geometry was unusual. Not impossible. Not supernatural. Merely difficult to reconcile with familiar processes of sedimentation and erosion. She found a more recent paper. It included computer-generated models. One figure showed a reconstruction of the formation beneath the desert. Miriam stared at it. The model was incomplete. Large regions were labelled uncertain.

But the visible structures formed an arrangement of intersecting planes and curved surfaces which gave the impression of an architecture designed according to principles very different from those familiar to human builders. The authors were careful not to interpret it as artificial. They described it as a complex geological formation. Miriam agreed that this was the only reasonable conclusion. Then she noticed the publication date. The paper had appeared six months earlier. Its references included a preliminary geological model published three years before that.

She opened the older paper. It contained a less detailed reconstruction. The overall pattern was already visible. She searched further backwards. The oldest diagrams were simpler, but the same general structure appeared. At each stage, the evidence became more extensive, more coherent and more difficult to interpret as an accidental resemblance to architecture. She sat in silence. The book had acquired a history. Now one of the places mentioned in its fictional history appeared to possess an increasingly elaborate geological record. The relationship might be coincidental.

It might be the result of selection bias. She might simply be noticing connections because she had begun looking for them. These were all possibilities worth taking seriously. But she could not escape the impression that the process she had witnessed in the Library was continuing elsewhere. Not in the present. In the past. The evidence was not appearing as new objects falling from the sky. It was appearing as old observations, forgotten reports and increasingly precise reconstructions of things that had apparently always been there.

Miriam opened her memorandum. She added a ninth observation.

9. Historical claims associated with the reconstructed text may be acquiring independently documented physical referents.

Then she deleted the word may. She looked at the sentence. After a moment she restored it. There were limits to what the evidence justified. She intended to respect them. Her telephone vibrated. A message from Oliver.

One further thought. If the earlier sources are being supplied retrospectively, how would we distinguish a newly acquired history from one that was genuinely there all along?

Miriam read it. She began typing a reply.

We couldn't, unless some evidence of the previous state survived.

She stopped. That was the difficulty. Their memories appeared to preserve fragments of a history which the documents increasingly denied. But memory was unreliable, and documents were normally considered stronger evidence. The phenomenon, if there was one, attacked the very distinction by which historians established the past. She sent the message. Then, after a moment, she added another.

We need to stop asking whether the Necronomicon is authentic. We need to discover whether authenticity itself is changing.

Oliver replied almost immediately.

And if it is?

Miriam stared at the question. She thought of the physical volume in London. Its paper. Its binding. Its wormholes. Its stains. The perfectly ordinary evidence of centuries of existence. She thought of the Arabic original which had supposedly preceded it, the lost Greek translation, the obscure sources and forgotten cities. She thought of Alhazred's purported discoveries. And of the entities whose presence in the universe his book had been invented to explain.

Finally, she typed:

Then the book is probably the least important thing that's going to become real.

She sent the message. Outside, the night was still. There was no movement among the stars. No sign of approaching catastrophe. No indication that anything had changed. And that, Miriam realised, was precisely what she ought to expect. If her hypothesis was correct, the universe would not announce the arrival of its newly established inhabitants. It would simply possess the history in which they had always existed.

And somewhere, in a discipline she had never studied, a scientist might already be preparing a paper describing the first inexplicable traces. Not traces of something arriving. Traces of something that had been present long before humanity evolved. Something whose existence, until a few days ago, had been confined to a work of fiction. Miriam closed the laptop. She had spent the day trying to determine whether a book was genuine. She went to bed wondering how much of the rest of the world was.

Chapter III 3 of 10

The Seventh Division

The trouble with a changing text was deciding what had changed. Miriam had spent her career separating scribal error from deliberate revision, an editor's conjecture from an author's wording. Every variation demanded a history of its own. Now two photographs of the same leaf demanded incompatible histories.

On Saturday morning she placed the British Library's 2008 conservation image beside the photograph taken the previous afternoon. Both showed the forty-seventh surviving leaf of the fifteenth-century Necronomicon: the narrow tear from the upper margin, three wormholes passing through the lower text, the stain that had spread into a partial word. The damage coincided down to its smallest irregularities. The printing did not. Miriam enlarged the images until the individual letters were clearly visible.

The earlier photograph read:

Quae extra ordinem temporum sunt, neque praecedunt neque sequuntur.

Those things which exist outside the order of time neither precede nor follow. The recent photograph contained an additional clause.

Quae extra ordinem temporum sunt, neque praecedunt neque sequuntur, sed omnibus successionibus simul adsunt.

Those things which exist outside the order of time neither precede nor follow, but are simultaneously present to all successions. She checked the transcription. The additional words appeared there too. She checked the catalogue description. A summary of the passage mentioned the relationship between temporal succession and a class of entities supposedly capable of existing independently of ordinary chronology. The summary was dated 1984. Miriam returned to the photographs.

She had expected, or at least hoped, to discover a difference in physical condition. A repaired leaf, a substituted page, some indication that the images represented different states of the same object. There was none. The additional clause fitted the line perfectly.

The spacing of the surrounding words had adjusted to accommodate it. The type remained consistent with the rest of the page. The line endings had changed, but the paragraph retained exactly the same number of lines. It was not an insertion. It was a different setting of type. And yet the damage to the paper was identical. Miriam sat very still. She understood what the photographs appeared to show. The difficulty was that she had no intention of accepting it.

Oliver Stroud arrived at eleven, carrying a folder of transcriptions and looking as though he had slept even less than she had.

They had arranged to meet in Cambridge rather than London. Miriam wanted to compare the Library's digital images with her own research materials, and Oliver had agreed to bring copies of the relevant documentary records. He entered her study, looked at the bookshelves and smiled.

"An admirable lack of available floor space."

"Books are rather like cats. One starts with a manageable number and eventually discovers that the house belongs to them."

"I'd always thought cats were less demanding."

"You've never tried to establish the provenance of one."

He put his folder on the desk. Miriam pointed to the monitor.

"Before we discuss anything else, look at these."

Oliver leaned forward. She displayed the photographs side by side. He examined them for perhaps half a minute.

"Different transcription?"

"Different text."

"Could one image have been retouched?"

"Certainly."

"Have you checked the original files?"

"Yes. The later photograph is the one you sent me yesterday. The earlier one comes from your department's image archive."

"Then we need to verify both against the physical volume."

"We do. But first I want you to tell me what you see."

Oliver studied the screen.

"An additional clause."

"Anything else?"

"The spacing has changed."

"Yes."

"The line endings are different."

"Yes."

He looked at the margins.

"But the tear is the same."

"And the wormholes."

Oliver moved closer.

"That's peculiar."

"An unusually restrained description."

He continued examining the images.

"The additional text is quite natural in context."

"That's the problem."

"I don't follow."

"The longer reading doesn't interrupt the argument. It completes it."

Miriam brought up a transcription of the surrounding passage.

"The shorter version makes a general claim about things existing outside temporal succession. The longer version explains how such things might relate to successive events."

"So?"

"So the additional clause resolves an ambiguity. It supplies precisely the distinction the paragraph otherwise lacks."

Oliver straightened.

"Are you suggesting that someone corrected the text?"

"I'm suggesting that the text appears to have acquired a more coherent form."

"Which is what editors do."

"Editors don't normally alter photographs of the original printing."

"Not deliberately."

Miriam looked at him.

"You're determined to remain sceptical."

"I thought that was the purpose of the exercise."

"It is."

She opened a third image. This was a photograph from the 1968 conservation archive. The relevant portion of the page was partly obscured by a paper support, but the beginning of the paragraph was visible. The longer reading appeared to be present. Oliver frowned.

"Then perhaps the 2008 image was altered."

"Perhaps."

"Or depicts another state of the printing."

"Perhaps."

"But?"

Miriam enlarged the damaged portion.

"The wormholes are identical in all three images. The tear is identical. The staining is identical."

"Which implies the same physical leaf."

"Yes."

"And the 1968 and 2026 photographs contain the longer reading."

"Yes."

"While the 2008 image contains the shorter."

"Exactly."

Oliver considered this.

"An error introduced into the 2008 digital image, then."

"That would be a reasonable working hypothesis."

"Have you checked the corresponding transcription from 2008?"

"I have."

"And?"

"It contains the shorter reading."

He looked at her.

"The transcription and photograph agree?"

"Yes."

"And the earlier and later records agree on the longer reading?"

"Yes."

Oliver sat down.

"That is inconvenient."

"Very."

He removed his spectacles and cleaned them.

"Have you considered that we're approaching this in entirely the wrong way?"

"Frequently."

"We keep trying to establish which document is genuine."

"Yes."

"But perhaps we should ask a different question."

"Which?"

"What determines which version survives?"

Miriam regarded him.

"That's what I wanted to discuss."

She had spent much of the previous night constructing a table of textual witnesses. The task was straightforward in principle.

A medieval work survives in multiple manuscripts or printed editions. The historian compares their readings and attempts to reconstruct the relationships between them. Shared errors can indicate common ancestry. Distinctive omissions may reveal dependence upon a damaged exemplar. Corrections and interpolations can sometimes be assigned to particular branches of transmission. The result is a provisional family tree of texts, conventionally called a stemma. It rarely produces certainty. But it imposes discipline upon speculation.

Miriam had assembled every reference to the Necronomicon which she could verify, or at least locate, in the newly available historical literature. There were fewer witnesses than the surrounding mythology suggested. The supposed Arabic original was lost. The Greek translation attributed to Theodorus Philetas was likewise absent, though fragments were cited in several later works. The Latin version associated with Wormius survived in the British Library's fifteenth-century printing and in a defective seventeenth-century edition held elsewhere. Several catalogues mentioned other copies, but their present whereabouts were uncertain.

This was not, by medieval standards, an especially rich textual tradition. It was nevertheless sufficient to identify patterns. Miriam showed Oliver her table.

"I've divided the readings into three categories."

She highlighted the first column.

"Stable readings. Passages that are consistently represented across the surviving witnesses."

"The cosmological sections?"

"Mostly. Those discussing the motions of the heavens, numerical relationships and certain forms of classification."

She highlighted the second.

"Unstable readings. Passages that vary substantially between witnesses."

"Which ones?"

"Descriptions of places, historical events and alleged observations."

Oliver considered that.

"More vulnerable to scribal interference."

"Possibly."

"And the third category?"

"Unresolved readings. Passages known only from references, damaged witnesses or incomplete descriptions."

She opened another document.

"The most serious discrepancies occur in those sections."

"Because we have less evidence."

"Yes. But not merely because we have less evidence."

She pointed to a sequence of entries.

"In several instances, the documentation has become more complete since the reconstruction was published."

Oliver frowned.

"New material has been discovered?"

"That's what the records say."

"But you think the records themselves may have changed."

"I think we should test that possibility."

She opened a scanned bibliography which she had consulted during preparation of the City of Dis essay.

"This copy was downloaded three weeks ago."

She showed him a reference to a lost manuscript.

"The bibliography describes it as unattested."

Then she opened the current online version. The same entry now mentioned a seventeenth-century inventory and a nineteenth-century sale catalogue. The publication date of the bibliography had not changed. The page numbers had not changed. Its revision history recorded no amendment. Oliver leaned forward.

"Could the online file have been silently replaced?"

"Yes."

"Could your earlier copy have been incomplete?"

"Yes."

"Could the original author have discovered new evidence and updated the reference without recording the change?"

"Yes."

He looked at her.

"You're going to say those explanations become less plausible when repeated."

"Precisely."

She showed him a second reference. Then a third. Each concerned a different unresolved detail in the book's textual history. Each had acquired additional documentary support. The newly available evidence was not spectacular. There were no complete manuscripts emerging from sealed tombs, no astonishing discoveries announced in scholarly journals. There were catalogue descriptions, footnotes, references to private collections and minor corrections to established bibliographies. Individually, they were unremarkable. Together, they suggested something more disturbing.

The uncertain portions of the book's history were acquiring the evidence needed to make them less uncertain. Oliver looked at the table.

"If someone were manufacturing an authentic textual tradition, this would be a very thorough way to do it."

Miriam shook her head.

"Not manufacturing an authentic tradition."

"What, then?"

"Resolving an incomplete one."

He considered the distinction.

"Explain."

"A forgery begins with an object someone wishes us to believe is genuine. The forger constructs evidence to support it."

"And here?"

"Here the supporting evidence seems to be accumulating around the gaps in an existing reconstruction."

"You think the essay provided the structure."

"I think that's one possibility we have to consider."

Oliver looked towards the window.

"And the missing sources are being supplied."

"Perhaps."

"By what?"

"I haven't the slightest idea."

He gave a small, exhausted smile.

"That's becoming a familiar answer."

"At present it's the only intellectually respectable one."

The seventh division had become Miriam's principal concern because it was the least complete. In the London printing several leaves were missing, and later descriptions disagreed about what they contained. The City of Dis reconstruction had treated the gaps as literary puzzles. Now there was a medieval text, or what appeared to be one, and gaps into which the historical record was steadily introducing new witnesses.

Oliver spread the transcriptions on her table. The surviving Latin opening concerned the survival of form after the destruction of a living body. Its terms were closer to memory and organisation than to the conventional vocabulary of an immortal soul.

"Listen to this," Miriam said. "Non manet quod sentit, sed ordo quo sentiri potest. What persists is not that which feels, but the order by which feeling is possible."

"A universal intellect?" Oliver suggested.

"Perhaps. But here the writer distinguishes a preserved pattern from the person whose pattern it was. A copy of your memories could insist it was you. The insistence wouldn't establish the continuation of your experience."

Oliver read the next surviving line. "Where the order is reproduced, the form may be recognised, but the one who was present need not therefore be present again."

"That's a better distinction than some modern discussions manage," Miriam said.

The leaf following it was missing. Three later references claimed to describe its contents. One concerned speaking to the dead. A second described something capable of retaining the forms of vanished minds. A third concerned existence outside the distinction between past and present.

"Different interpretations of the same argument?" Oliver asked.

"Or different texts. We don't know."

She showed him the original reconstruction, which had proposed that the three accounts might derive from a single underlying idea: persistence without ordinary continuation through time. She had enjoyed making the conjecture. The surviving fragments fitted it almost too neatly.

"A dead mathematician's work remains available to readers," she said. "But the mathematician doesn't survive in the books. Now imagine a preserving relationship which isn't simply later than the thing preserved."

"I can't."

"Neither can I, clearly. That's why it remains a conjecture."

Oliver reached for the pencil and began writing a possible connective clause in the margin of his transcription. Miriam watched the words appear. They were entirely sensible. They would also complete the gap in precisely the direction she had proposed.

"Could we leave that unwritten?" she asked.

He stopped. "You think the act of reconstruction matters."

"I don't know whether it does. But I know our earlier reconstruction now has a physical object answering to it. I'd prefer to examine the witnesses before supplying another missing sentence."

Oliver erased the clause. A faint indentation remained in the paper, legible when he held it towards the light. He folded the sheet and put it away. They devised an experiment that afternoon.

Miriam disliked the word. It suggested a degree of methodological sophistication which their circumstances did not justify. They had no established theory, no calibrated instrument capable of detecting the alleged phenomenon, and no reliable estimate of the variables involved. Nevertheless, they could begin with an elementary observation. They would record the present state of several textual witnesses, preserve independent copies of the records, and compare them over time. The selection of passages was important. Miriam chose twelve.

Four came from portions of the Necronomicon whose textual history was relatively secure. Four came from passages with substantial variation between the surviving witnesses. The final four concerned incomplete or unattested sections, including the seventh division. Oliver suggested adding a thirteenth passage from an unrelated medieval astronomical treatise.

"Something that has nothing to do with Alhazred."

"An excellent control."

"And perhaps another from a modern printed work?"

"Yes. Preferably something with an uncomplicated publication history."

He selected a twentieth-century critical edition of an Aristotelian commentary. Miriam approved. Their procedure was deliberately simple.

They printed the available images and transcriptions. Each sheet was numbered and dated. They generated digital copies and calculated cryptographic hashes, recording the resulting values separately. They photographed the printouts with two different cameras. They then placed one set of documents in Miriam's study and prepared another for Oliver to take back to London. Oliver proposed keeping a copy in a sealed envelope.

Miriam agreed, although she was doubtful that a sealed envelope offered meaningful protection against a phenomenon which might retrospectively alter the circumstances of its own creation.

"That's rather the difficulty," she said.

"What is?"

"If a historical alteration changes the records along with the event, we may find that all our copies agree."

"Then the experiment fails."

"Yes."

"Unless our memories disagree."

"Which would leave us with precisely the problem we started with."

Oliver considered this.

"Perhaps we could ask someone else to make an independent record."

"Someone who knows nothing about the book?"

"Yes."

"That would help establish whether our observations are being influenced by expectation. But I don't want to recruit unsuspecting colleagues into something we don't understand."

"Do you genuinely believe there's a risk?"

Miriam looked at the photographs.

"I don't know. That's the difficulty with an unrecognised phenomenon. One doesn't know which precautions are sensible."

They completed the records shortly after four. The process produced a reassuring quantity of paperwork. There were fourteen passages, each represented by several independent copies. Their digital fingerprints were recorded. The source files were catalogued. A dated memorandum described the procedure. Nothing changed. The photographs remained identical. The printed passages remained stable. Miriam examined the seventh division's damaged opening once more. There was no additional text. Oliver looked faintly disappointed.

"Perhaps the universe isn't interested in being tested."

"Or perhaps our hypothesis is wrong."

"That would be a relief."

"An enormous one."

Oliver reached for the packet containing the 2008 photograph of the disputed leaf. Its upper corner bore a thin magenta printer streak. Across the back Miriam had written 11: shorter reading, then dated and initialled it. Oliver added his initials. There was only one signed print. They had made it as a control for their copies, not as an artefact anyone might wish to preserve.

"We've been very busy observing what happens when evidence is discovered," Miriam said. "What if we do something which should leave a simple, irreversible record?"

Oliver turned the sheet over. "Something we control. Not the book."

"This print. We can destroy it and retain the others."

"I dislike destroying evidence."

"So do I. It's a working copy. We have the source image, the file hash and two witnesses."

They recorded the procedure before beginning: the unique print, its magenta streak, the two sets of initials, the intended method of destruction. Oliver took a photograph with his telephone. Miriam took another. They checked that each showed the annotation, then placed the images in separate folders.

In the study stood a small cross-cut shredder used for confidential correspondence. Miriam switched it on. Oliver fed the sheet between the rollers, face up. The corner with the magenta streak vanished last. Miriam remembered the sound precisely because it caught and stopped for a moment. Oliver reversed the mechanism, freed the paper, then fed the remainder through. She could still see the two initials disappearing into the blades.

They emptied the container into a bowl, added water and reduced the strips to an unreadable pulp. Oliver tipped the pulp into a clear specimen bag, sealed it and wrote the time on the label. Miriam photographed his hands holding it. A short nick from the corner of a paper strip had left a line of blood beside her thumbnail. She washed it in the kitchen.

They retained the bag, the photographs, and a one-page record signed by both of them. The experiment proved only that they had destroyed a sheet of paper. But it was an event, not an interpretation of an ancient document. Both had observed it. For once they could describe exactly what should no longer exist.

They returned to the desk. Oliver put the remaining packets into their envelopes. Miriam checked her thumbnail and found the shallow cut still open. She felt faintly foolish. The procedure had been a disproportionate response to several inconclusive photographs. Nevertheless, she was glad to have done it.

She saved the final document. A message appeared on her screen. It was from Eleanor Marsh. The email was brief.

Dear Miriam,

How extraordinary to hear from you about this particular text.

Yes, I examined the London printing while preparing my 1998 paper, although my principal interest concerned its relationship to the fragmentary Greek tradition.

I have no recollection of the 1971 conservation report you mention. The citation is certainly peculiar and I'd be very interested to see it.

There is, however, a more immediate complication.

You may be aware that a previously unidentified manuscript in Paris has recently been recognised as preserving substantial material from the Latin Necronomicon.

I haven't examined it personally. From the preliminary description, it appears to be a fourteenth-century witness and may therefore be of considerable importance.

I am particularly interested in its treatment of the seventh division, which is much more complete than the British Library printing.

Do send me the conservation reference.

Best,

Eleanor

Miriam read the message. Then she handed the laptop to Oliver. He read it in silence.

"A fourteenth-century manuscript," he said.

"Apparently."

"That would predate the printed copies."

"By a century or more."

"And it contains the seventh division."

"According to Eleanor."

He looked towards the documents spread across her desk.

"The very section we've been examining."

"Yes."

"That's an extraordinary coincidence."

Miriam shook her head.

"Not yet. Manuscripts are identified all the time. A text can survive unnoticed for centuries, particularly if it's miscatalogued or incorporated into a composite volume."

"But this particular text?"

"Even that isn't impossible."

She opened the browser.

"Let's find out what she means by recently."

They searched the principal French manuscript catalogues. At first nothing appeared. There were modern publications concerning Lovecraft, references to imaginary books, and several studies of medieval Arabic intellectual history. Then Miriam found an entry in a specialist database. It described a parchment manuscript held in Paris, containing a collection of Latin translations from Arabic sources. The date was provisionally assigned to the fourteenth century. The manuscript's title did not mention the Necronomicon.

Its catalogue description listed several astronomical and philosophical treatises, followed by an anonymous work concerning temporal relations, post-mortem consciousness and the nature of existence. A recent note identified that final work as a partial Latin witness to the Necronomicon. Miriam opened the record. It had apparently been created in 1987. The identification note was dated 2019.

"Not terribly recent," she said.

Oliver leaned forward.

"Perhaps Marsh means the identification has only now become widely known."

"Perhaps."

They followed the references. A short article published in 2021 discussed the manuscript's relationship to the Wormius tradition. Another, dated 2024, examined several unusual readings in its seventh division. A third, published earlier that year, proposed that the manuscript preserved a more complete form of the text than any previously known witness. Miriam opened the third article. Its author had included a transcription of several passages. She read the abstract. Then the introduction.

The manuscript had supposedly been recognised through a comparison of its unusual terminology with known descriptions of the Necronomicon. Nothing about the discovery was intrinsically implausible. A medieval manuscript could remain incompletely identified for generations, especially if its constituent texts lacked titles or clear authorial attributions. What was disturbing was the precision with which the newly available material addressed the problems Miriam and Oliver had been discussing. The seventh division was no longer merely a collection of fragmentary references. It now had an independent manuscript witness.

That witness predated the surviving printed copies. And it supplied passages precisely where the textual reconstruction was weakest. Oliver looked at Miriam.

"Do you want to read it?"

She hesitated.

"Very much."

"Then perhaps we should."

"That's exactly what worries me."

"You think reading the manuscript could make more of it exist?"

"I think we have no idea whether our investigation is passive."

Oliver considered this.

"Suppose it isn't."

"Then the act of resolving the text's uncertainties might itself be relevant to the phenomenon."

"Which means that any attempt to understand it could make matters worse."

"Potentially."

He looked at the laptop.

"And refusing to investigate?"

"Would leave us ignorant."

"While whatever is happening continues."

"Yes."

They sat in silence. At length Miriam opened the transcription.

"At least," she said, "we should avoid supplying our own missing passages."

The Paris manuscript was designated Latin 11462. It was a composite volume, containing works copied by several hands. Its parchment leaves varied in size and quality, suggesting that the constituent texts had not originally belonged together. The portion identified with the Necronomicon occupied the final gatherings. Unlike the London printing, it contained a reasonably complete seventh division. There were omissions, corrections and damaged passages, but the overall structure was recoverable. Miriam examined the digital images.

The script was consistent with the assigned period. The ruling, abbreviations and letterforms appeared plausible. Several pages contained marginal corrections by a second hand. There was nothing obviously fraudulent. She read the first paragraph. It concerned the difference between the destruction of a living organism and the loss of the information constituting its organisation. The argument was more elaborate than the surviving London fragments. It distinguished three forms of persistence. The first was material continuity. The second was the preservation of a pattern through successive material arrangements.

The third was something the author described as the persistence of relations independently of their particular manifestations. Miriam paused.

"That's interesting."

Oliver looked over her shoulder.

"What is?"

"The third category."

"Explain."

"The first two are relatively comprehensible. A body persists through material continuity. A pattern can persist when reproduced in different materials."

"And the third?"

"The text seems to suggest that a relationship can possess a kind of existence without being continuously represented by any particular object."

Oliver frowned.

"An abstraction?"

"Possibly. Mathematical relationships, perhaps."

"Those don't need physical objects."

"That depends on what you mean by existence. A mathematical Platonist might agree with you. A nominalist might not."

"What does Alhazred say?"

"He's not doing conventional metaphysics."

She read the next sentence. The Latin was awkward. The author appeared to claim that some relations were not merely abstract descriptions imposed by thinking beings, but constitutive features of reality. The objects participating in those relations could change or disappear. The relations themselves did not necessarily depend upon the continued existence of the objects. Miriam translated a short passage.

"'The pattern is not preserved because the things endure. Rather, the things are comprehensible because the pattern may be completed.'"

Oliver considered it.

"That sounds like the passage about causes and their completion."

"Yes."

"Does it mean the future determines the past?"

"Not necessarily. It could be expressing a philosophical doctrine concerning the logical dependence of particulars upon universals."

"But you don't think so."

"I think we should stop trying to decide what it means before we've established what the text actually says."

She continued reading. The next section concerned the persistence of memory after death. Here the author distinguished between the recreation of a dead person's thoughts and the continuation of their subjective experience. A reconstructed mind, he argued, might possess every memory and disposition of its predecessor without being numerically identical to it. Oliver recognised the argument.

"That's the passage from the London printing."

"Yes. But considerably longer."

"Does it resolve the problem?"

"In one respect."

"How?"

"It says that the persistence of information about a mind isn't equivalent to the survival of that mind."

"Which seems sensible."

"Entirely."

"Not much of a necromantic revelation."

"No."

Miriam read on. Then she stopped. Oliver waited.

"What?"

"The next paragraph is different."

"In what respect?"

"It describes the possibility of obtaining information from a structure that no longer exists."

"A record?"

"No. Something more peculiar."

She read silently for several minutes. The author was discussing a hypothetical arrangement of physical objects which had once existed but had subsequently been destroyed.

He suggested that if the relations necessary to establish that arrangement were sufficiently constrained by other facts, the former arrangement might be recoverable without possessing an ordinary material record of it. Miriam considered this.

"It sounds rather like a claim about inference."

"Reconstructing an extinct organism from fossils?"

"Something along those lines. Except the author seems to distinguish between reconstruction as a human intellectual activity and reconstruction as a property of reality."

"Meaning?"

"He appears to be saying that the world itself may determine a missing structure when enough of its relationships are fixed."

Oliver looked at her.

"Like our manuscript?"

Miriam did not answer immediately. She read the paragraph again. Then she translated one sentence.

"'That which has not been may nevertheless be required by that which is.'"

Oliver said nothing. She continued.

"'And when the requirement admits no alternative, the distinction between what was and what must have been cannot be maintained.'"

There was a long silence. Oliver finally said, "That's rather close to your hypothesis."

"Uncomfortably."

"Could you be interpreting it that way because you've already formed the hypothesis?"

"Of course."

"But the words are there."

"The words are there."

"And the manuscript is supposedly fourteenth-century."

"Yes."

"Then somebody was considering this idea centuries ago."

"Or somebody has constructed a very elaborate forgery."

"Or?"

Miriam looked at him.

"Or we're examining a document whose history is being supplied retrospectively."

Oliver sat back. He seemed, for a moment, to have aged considerably.

"Do you think Alhazred understood what was happening?"

"We don't know whether there was an Alhazred."

"Suppose there was."

"Then we have no reason to believe he understood the phenomenon any better than we do."

"But he described it."

"A medieval author could describe a concept without understanding its physical implications."

"Even a concept like this?"

"Especially a concept like this. Philosophers have spent millennia making extraordinarily sophisticated arguments about things they couldn't possibly test."

Oliver looked at the manuscript image.

"What if this one could be tested?"

Miriam considered the question. Then she returned to the text. The next section contained a reference to the division's title. It had been partly obscured by damage. The first word was legible.

De.

The second could be interpreted in several ways. The third was absent. Miriam examined the surrounding text. The common editorial reconstruction was De restitutione formarum, concerning the restoration of forms. It was plausible. But she was not persuaded.

"There's an alternative reading," she said.

Oliver leaned forward.

"What?"

"The second word might be constitutione, not restitutione."

"Constitution rather than restoration?"

"Yes."

"Does it matter?"

"Considerably."

She pointed to the damaged word.

"Restoration implies that something previously existed and is being brought back. Constitution suggests that its existence is being established."

Oliver looked at the manuscript.

"Which would change the meaning of the entire division."

"Potentially."

"Can you establish the reading?"

"Not from this image."

"Could a specialist examine the original?"

"Certainly."

Miriam paused. Then she closed the document. Oliver looked surprised.

"Why stop?"

"Because I'm not going to propose a definitive reading merely to see whether reality obliges me."

He considered that for a moment.

"A sensible precaution."

"I wish I were more confident that it would accomplish anything."

They returned to the photographs of the London printing. Miriam wanted to see whether the newly discovered manuscript could clarify the shorter and longer readings. The Paris witness contained a passage corresponding to the altered sentence. It supported the longer version. That was unsurprising. A fuller manuscript might well preserve material omitted in a later printing. What was not so easily explained was the shorter reading in the 2008 photograph. Miriam opened the file. It still contained the shorter text. She checked its hash.

The value agreed with the one they had recorded earlier that afternoon.

"Good," Oliver said.

"Not necessarily."

"At least the file hasn't changed."

"Since we recorded it."

She opened the 1968 image. The longer reading remained. Then the recent photograph. Again, the longer version. Their records were stable. The contradiction persisted. Miriam considered what they had established. They possessed three images of apparently the same physical leaf. Two contained one reading. One contained another. The shorter reading could be a digital manipulation. The longer one could be an interpolation. The photographs could have been incorrectly labelled. There were many ordinary possibilities. But there was now a fourth image to examine.

Oliver had brought a scan from the 1996 conservation assessment, which neither of them had yet studied in detail. Miriam opened it. The page was visible beneath a transparent support. The relevant paragraph occupied the lower half. She enlarged the image. The longer reading was present. Oliver leaned forward.

"Another for the longer version."

"Yes."

She examined the damage. The tear was identical. The wormholes aligned. The stain occupied the expected position. Then she noticed something.

"Look at the lower margin."

Oliver leaned closer.

"What am I looking for?"

"The staining."

"It seems the same."

"The shape, yes. But there's a small extension towards the inner edge in the recent photograph."

He examined the images.

"You're right."

"Could be lighting."

"Or moisture damage between 1996 and the present."

"Yes."

They compared the 2008 photograph. Its stain was intermediate between the two. Miriam frowned.

"That's interesting."

"Why?"

"The physical damage follows an ordinary chronological progression."

"Of course it does."

"But the text doesn't."

She arranged the four images by date. The 1968 and 1996 photographs contained the longer reading. The 2008 photograph contained the shorter. The recent photograph contained the longer again.

Meanwhile, the paper's physical deterioration appeared entirely consistent with the passage of time. Oliver examined the sequence.

"So the text changes, but the paper ages normally."

"Apparently."

"How can that happen?"

"I don't know."

He considered the photographs.

"Wouldn't a change to the printed text require changes to the type, the line spacing and perhaps the position of the ink on the paper?"

"Yes."

"And wouldn't those changes affect the relationship between the printing and the wormholes?"

"They should."

Miriam enlarged the images again. The wormholes passed through different portions of the text in the shorter and longer versions. But their physical trajectories remained unchanged. The relationship between the printing and the damage was internally consistent in every photograph. It was as though each image represented the same physical leaf produced under a different historical arrangement of type, subsequently subjected to the same pattern of deterioration. Miriam felt an unpleasant sensation of recognition. The evidence did not suggest that words had been added to a page.

It suggested that the page had always been printed differently. The alteration, if alteration there had been, reached backwards into the original act of printing. She wrote a sentence in her notebook.

The textual variants may represent alternative histories of the same physical object.

Then she crossed it out.

"Too strong?" Oliver asked.

"Much too strong."

"What would you write instead?"

Miriam considered.

The available photographic record is not presently compatible with a single, straightforward account of the page's production.

Oliver read the sentence.

"Better."

"Less satisfying."

"Considerably."

"But accurate."

She closed the images. It was nearly seven. Oliver had intended to return to London that afternoon, but they had become too absorbed in the investigation to notice the time. Miriam offered him the spare room. He accepted with the evident relief of a man who had been contemplating the journey home through an increasingly unwelcome series of railway connections. They ordered food and continued discussing the seventh division over dinner. Miriam insisted that the principal danger was interpretative.

They were approaching the text with a hypothesis already in mind. It would be easy to mistake ambiguous philosophical language for descriptions of precisely the phenomenon they believed themselves to be observing. Oliver agreed. But he raised an objection.

"What if the ambiguity is itself the important feature?"

"In what sense?"

"Suppose the text isn't changing randomly."

"Go on."

"Suppose the uncertain passages are becoming more definite as we investigate them."

"That would be a form of selection bias unless we could demonstrate actual change."

"Yes. But imagine we could demonstrate it."

"Then we'd have a much more interesting problem."

"And suppose the changes consistently resolved uncertainty in the same direction."

"Which direction?"

Oliver gestured towards her notebook.

"Towards the history described in the reconstruction."

Miriam considered.

"That would suggest some relationship between the model and the emerging evidence."

"Would it establish causation?"

"No."

"But it would make the relationship difficult to dismiss."

"Yes."

He looked at her.

"And if the seventh division is describing the same process?"

Miriam put down her fork.

"Then we need to be very careful about treating it as a set of instructions."

"Why?"

"Because instructions imply agency and purpose. The text might describe a natural phenomenon without its author understanding it. We shouldn't assume that the book is attempting to do anything."

Oliver considered this.

"Nor that the process is attempting to do anything."

"Exactly."

"And yet something is happening."

"Perhaps."

He smiled faintly.

"You do rather cling to that word."

"It's one of the few things keeping us intellectually respectable."

At 21:16, Miriam received another email from Eleanor Marsh. It contained a link to a newly digitised manuscript catalogue and a short explanatory note.

Dear Miriam,

I've found the earlier Paris catalogue description.

You may find the identification history interesting. The manuscript was apparently recognised as an Alhazred witness some years before the attribution became widely accepted.

One curious point: the catalogue's oldest description identifies the seventh division as incomplete, whereas a later description calls it substantially intact.

This may simply reflect the discovery of additional leaves in the binding, but I haven't found any published account of such a discovery.

Best,

Eleanor

Miriam opened the link. The digitised catalogue was an old printed volume, apparently published in the nineteenth century. Its pages were yellowed and irregularly scanned. She located the manuscript reference. The description was brief. It mentioned several treatises, an anonymous work concerning the nature of time, and a defective section discussing the persistence of forms. The final leaves were described as damaged. There was no attribution to Alhazred. No title identifying the Necronomicon. And no suggestion that the seventh division was complete.

Miriam compared the record with the modern catalogue. The newer description stated that the manuscript preserved almost the entire division. That discrepancy could have an ordinary explanation. Perhaps the leaves had been misordered. Perhaps later scholars had recognised that apparently unrelated fragments belonged to the same text. Perhaps improved photography had revealed writing previously thought illegible. But something about the descriptions troubled her. She examined the modern digitisation. The seventh division occupied seventeen leaves. The older catalogue referred to only twelve. Five additional leaves. Miriam checked the foliation.

The missing leaves did not form a separate gathering inserted at the end. They occurred within the text. The manuscript's present foliation was continuous, and the relevant leaves showed no obvious signs of later insertion. She checked whether the older catalogue used a different numbering system. It did. The original description had apparently counted only the leaves it could identify as belonging to the particular treatise. That might explain the discrepancy. She began to relax. Then she found a note in a modern scholarly article.

It explained that the Paris manuscript had been rebound in 1874. During the restoration, five leaves previously thought lost had supposedly been discovered folded into the spine. The discovery was recorded in the conservator's report. Miriam opened the cited report. It was dated 1874. The five leaves were described. Their text was identified as belonging to an anonymous Latin philosophical work. The report noted that they had been found within the binding material, apparently having been reused during an earlier repair. That was entirely plausible.

Old manuscript leaves were frequently reused in bindings. Miriam read on. The report stated that the recovered leaves had subsequently been restored to their presumed original positions. Again, plausible. Then she noticed the date of the earlier repair. 1698. The relevant leaves had apparently been removed from the manuscript and incorporated into its binding at the end of the seventeenth century. They had remained hidden until 1874. That would explain why an earlier inventory described the text as defective.

But it did not explain why the nineteenth-century catalogue, published after the restoration, still described the seventh division as incomplete. Perhaps the cataloguer had relied upon an older description. Perhaps the recovered leaves had not yet been identified. Both possibilities were ordinary. Miriam opened a further reference. It was an inventory dated 1812. The description included a note concerning the five missing leaves. It stated that they had been recovered during a restoration undertaken in 1874. Miriam stopped. She checked the date. 1812. She checked the reproduction.

The words were not a marginal annotation. They formed part of the printed description. The inventory had been published sixty-two years before the supposed restoration. Miriam closed her eyes. She had encountered the same pattern in London. A document referring to an event which, according to its own date, had not yet occurred. She opened the record again. The reference was still there. It was possible that the scanned volume was a later reprint incorporating amendments.

It was possible that the publication date referred only to the original edition and that the surviving copy had been revised. It was possible that the catalogue had been misidentified. She began listing those explanations. Then she noticed the bibliography attached to the modern record. The 1812 inventory was cited as evidence that the five leaves had been known before the restoration. Another scholarly article used the same reference to argue that the manuscript had been complete by the beginning of the nineteenth century.

The contradiction was not isolated. It had already been incorporated into the historical literature. Miriam looked at the clock. It was nearly eleven. Oliver had gone to bed. She considered waking him, then decided against it. There was nothing he could do that she could not do herself. She downloaded the images, saved copies of the catalogue entries and recorded their hashes. She printed the relevant pages. Then she placed them in a folder. She was becoming increasingly conscious of the ritualistic quality of these precautions.

The documents might change. Their copies might change. Their memories might change. Every attempt to preserve evidence assumed some stable medium which the phenomenon could not retrospectively alter. She had no reason to believe such a medium existed. Nevertheless, she continued. It was what scholars did. When confronted with uncertainty, they made records.

The next morning, Miriam found Oliver in the kitchen reading one of her books. He looked up as she entered.

"Morning."

"Did you sleep?"

"Reasonably."

"That's more than I managed."

He closed the book.

"Something happened?"

"I found an additional manuscript inconsistency."

He listened while she explained the Paris catalogue. Then he asked the question she had been expecting.

"Could the dates simply be wrong?"

"Of course."

"Have you checked the editions?"

"Not yet."

"Then perhaps we should."

Miriam nodded. They spent the morning tracing the publication history of the relevant catalogues. The 1812 inventory existed in several copies. Its publication date was well established. The particular scan Miriam had examined was not a modern reprint. The page containing the reference to the 1874 restoration belonged to the original printed gathering. At least, that was what the available images showed. They found references to the inventory in later works. Some quoted the passage. Others described the manuscript as incomplete.

The evidence was contradictory in precisely the way Miriam had begun to expect. It suggested not merely that the manuscript's history had changed, but that different parts of the documentary record were accommodating the alteration at different rates. She disliked the metaphor as soon as she thought of it.

Rates implied a process unfolding through time.

Yet the supposed alteration involved events which had occurred, or now appeared to have occurred, centuries earlier. What would it mean for a past event to change gradually? Gradually according to whose clock? The question was not merely difficult. It threatened to become incoherent. Miriam opened a new document.

Temporal inconsistency in documentary witnesses

She began writing.

If two records disagree about a historical event, the ordinary assumption is that at least one record is inaccurate.

If multiple records contain references to events occurring after their apparent dates of production, several explanations remain possible, including later alteration, incorrect dating, mistaken attribution and deliberate fabrication.

The present evidence does not justify abandoning those explanations.

She paused.

Then added:

Nevertheless, the discrepancies appear to be concentrated in material associated with the reconstructed Necronomicon tradition.

The concentration may be significant.

She read the sentences. They were cautious. They established nothing. But they defined a problem which could, at least in principle, be investigated. Oliver looked at the screen.

"There's something missing."

"What?"

"You haven't mentioned the seventh division."

"I thought it best to keep interpretation separate from observation."

"Very sensible."

He pointed to the manuscript images.

"But there's a further question. What if the newly recovered leaves don't merely explain a missing section?"

Miriam looked at him.

"What else would they do?"

"What if they're necessary for the text to make sense?"

She considered that.

"Many manuscript discoveries clarify previously obscure passages."

"Yes. But these particular passages describe the establishment of something whose prior existence was uncertain."

"That's our interpretation."

"Agreed."

"And?"

Oliver hesitated.

"Suppose the text itself is describing the process by which its missing portions are acquiring a history."

Miriam looked at him for a long moment.

"Then we have a very uncomfortable problem of self-reference."

"Precisely."

She turned back to the Paris manuscript. The passage concerning the constitution of forms was incomplete at the beginning. Its title remained uncertain. She enlarged the damaged heading. The second word could be read as restitutione. Or constitutione. The third was missing. The editor had supplied formarum. It was a reasonable reconstruction. But it was still a reconstruction. Miriam opened the modern scholarly article. The author had discussed the ambiguity and concluded that constitutione was marginally preferable. The argument was based on a comparison with another passage.

She followed the reference. The supporting passage came from a manuscript in a private collection. It was described as a late medieval copy of an otherwise unidentified Latin treatise. Miriam frowned.

"Oliver."

He looked up.

"What?"

"There's another witness."

"Where?"

"Apparently in a private collection."

"Which one?"

"The article doesn't say."

"Perhaps a reference to an auction catalogue?"

"Possibly."

She followed the footnote. It cited a manuscript description published in 1956. The description was brief. The relevant passage contained the phrase constitutio formarum. Constitution of forms. Miriam read the context. It was not obviously a quotation from the Necronomicon. But the terminology was similar. The editor had treated it as a possible parallel. Oliver considered the reference.

"Could the phrase be common?"

"Certainly. It's not inherently distinctive."

"Then perhaps it's irrelevant."

"Perhaps."

She examined the manuscript description again. The private collection was identified only by its owner's initials. The manuscript's present location was unknown. Its contents were described as philosophical and cosmological, with particular emphasis upon the relation between possible and actual existence. Miriam stared at the words. They were almost too convenient. A newly established manuscript witness had supplied a reading that supported the interpretation she and Oliver had begun considering. Then another, still more obscure witness had appeared to corroborate it. Each reference reduced uncertainty.

Each also expanded the network of historical objects required to support the text. A private manuscript implied a collector. A collector implied a chain of ownership. A chain of ownership implied catalogues, transactions and previous witnesses. Where would it stop? She closed the article.

"We need to speak to someone who isn't involved in textual scholarship."

Oliver looked surprised.

"Why?"

"Because we're approaching this with the assumptions of historians. We're trying to establish the authenticity of a documentary record."

"Which seems reasonable."

"Yes. But suppose the phenomenon isn't primarily historical."

"What would it be?"

"I don't know."

He smiled.

"Again."

"We need someone familiar with causality, probability and information. Someone who can help distinguish a genuine anomaly from our increasingly elaborate interpretations of one."

"A physicist?"

"Perhaps."

"Would a physicist take us seriously?"

"Not if we tell the story as we've been telling it."

"How would you present it?"

"As a problem involving inconsistent records and the apparent retrospective determination of missing information."

Oliver considered.

"That sounds rather like a description of the same absurdity in more respectable language."

"Which is sometimes the beginning of scientific progress."

"And sometimes the beginning of a grant application."

Miriam laughed. It was the first time she had done so without effort since the correspondence began. Then she opened her address book. There was someone she wanted to contact. Dr Yasmin Qureshi had trained as a mathematical physicist and subsequently worked on problems involving probability, information and the structure of causal inference. Miriam knew her through a series of interdisciplinary seminars at Cambridge. Yasmin possessed a formidable talent for identifying unsupported assumptions and was entirely unimpressed by appeals to scholarly reputation.

She was also unusually patient with strange questions, provided those questions were formulated precisely. Miriam began composing an email.

Dear Yasmin,

I have encountered a problem which I suspect you will initially regard as evidence that historians should be kept away from mathematical physics.

I am investigating a set of documentary discrepancies involving the apparent dependence of historical records upon information established at later dates.

The ordinary explanations, including misdating, interpolation and fabrication, remain possibilities. Nevertheless, the pattern is sufficiently peculiar that I should value your opinion on whether the underlying causal assumptions can be formulated coherently.

I am not asking you to believe that the past has changed.

I am asking whether there is a meaningful distinction between a history which has been discovered and one whose specification is constrained by later information.

Would you be willing to discuss it?

Best,

Miriam

She read the message. Then she sent it. Oliver looked at her.

"Do you think she'll help?"

"If she thinks we're making an elementary mistake, she'll tell us."

"And if she doesn't?"

"Then I expect she'll be extremely unpleasant about the implications."

They returned to the seventh division one final time before Oliver left. Miriam wanted to establish the relationship between the damaged heading and the earliest surviving descriptions. The oldest catalogue mentioned a work concerning the names and constitution of things. The London printing contained fragments discussing the persistence of forms. The Paris manuscript appeared to preserve a more complete argument about how relational structures might become physically instantiated. The evidence suggested that the seventh division was not primarily concerned with raising the dead.

It was concerned with the conditions under which something could be said to exist. The distinction mattered. Necromancy, in its conventional form, presupposed an established boundary between the living and the dead. The dead had existed. They had ceased to live. The practitioner sought to restore contact with them. But the seventh division seemed to contemplate a more radical possibility. It concerned structures which had never been physically realised, but whose existence might nevertheless be required by a sufficiently complete set of relationships.

Miriam found the concept deeply disturbing. Not because it was mystical. Because it was impersonal. The text did not suggest that anyone intended the process. There was no divine intervention, no supernatural will, no invocation of spirits. There was only a proposed relationship between consistency and existence. Oliver examined the concluding fragment.

"There's something curious here."

"What?"

"The final sentence."

Miriam looked at it. The Latin was incomplete. A damaged word interrupted the clause, and the final line had been lost.

The surviving text could be rendered approximately:

For those things which are constituted by necessity do not await the will of...

The remainder was missing. Oliver looked at her.

"Of God?"

"Possibly."

"Of the operator?"

"Possibly."

"Of the observer?"

"Also possible."

He considered the passage.

"Doesn't that make rather a difference?"

"An enormous difference."

"If it says God, it's metaphysics. If it says the operator, it could be a warning against ritual intervention. If it says the observer..."

"Then it may be making a claim about the independence of the process from human knowledge or intention."

Oliver looked at the damaged page.

"Can the missing word be recovered?"

"Not from the surviving image."

"Perhaps the original manuscript..."

Miriam raised a hand.

"Don't."

He stopped. For a moment neither spoke. Then Oliver smiled ruefully.

"We've become frightened of completing a sentence."

"An admittedly ridiculous position for two people who work with manuscripts."

"Do you think we're being irrational?"

"Possibly."

"Then why stop?"

Miriam looked at the text.

"Because until we understand whether our reconstructions are merely descriptive, I'd prefer not to supply any more of the missing information than necessary."

"That sounds rather like superstition."

"It may be."

She closed the image.

"But superstition consists in attributing causal powers without adequate evidence. At present, I'm not attributing any such power. I'm withholding an intervention because the evidence doesn't yet exclude an unfamiliar causal relationship."

Oliver considered.

"A distinction some people might find rather fine."

"Then they can write a paper explaining why it's invalid."

Before Oliver left, Miriam went to the desk drawer for the sealed bag of pulp. It was there. The seal was intact. The label gave the time they had written.

Beside it lay the signed 2008 photograph.

The magenta printer streak was visible at the top. Her initials and Oliver's were on the back. The sheet was uncreased. In the bottom margin, in Miriam's small writing, stood a sentence neither remembered: Retain marked original. Shred unsigned spare for disposal test; do not compromise the witness set.

Oliver took the paper without speaking. He had watched the initials enter the shredder. He could still describe the jam and the sound of the motor reversing. Miriam remembered the damp shreds catching against the side of the bowl. The nick beside her thumbnail was still sore.

They opened the bag. Its contents were the expected grey pulp. The label now said unsigned duplicate. Nothing about the seal suggested it had been disturbed.

Oliver checked his telephone photographs. One showed the intact signed sheet, set aside on the desk. Another showed him feeding a plain copy into the shredder. The printer streak was absent from the corner under his hand. The photographs' files were ordinary and their stored fingerprints agreed with the values in the revised record. On Miriam's computer the signed manifest described the destruction of the spare. Her signature appeared at the bottom; she recognised her impatient upward flick on the final letter.

"We both remember the marked sheet," Oliver said.

"Yes."

"Yet we've left ourselves an impeccably sensible chain of custody in which it was never destroyed."

Miriam examined the annotation. She had not merely changed her mind. According to the records, she had explained exactly why she should change it. The explanation was more responsible scholarship than the act she remembered.

Oliver laid the print on the desk as carefully as if it were the fifteenth-century volume itself. "I don't want another destructive test."

"Neither do I."

He photographed the surviving print again, then closed his case. No one proposed a new hypothesis. They had lost their only secure account of an action both could remember performing. The paper was a trivial object, but Miriam could not stop looking at the unfamiliar line in her hand.

He gathered his documents. It was almost time for his train. Before leaving, he placed one of their sealed envelopes on Miriam's desk.

"This is your copy."

"Yes."

"Mine is in my bag."

"Good."

He hesitated.

"If the records change, what do we do?"

"We document the change."

"And if the documentation changes?"

"We document that too, if we can."

"And if our memories change?"

Miriam looked at him.

"Then I suppose we'll have to hope that some part of the previous history remains accessible."

Oliver picked up his coat.

"An uncomfortable dependence upon hope."

"Very."

He moved towards the door. Then stopped.

"Miriam?"

"Yes?"

"Do you think the book knows we're reading it?"

She considered the question.

"No."

"You're sure?"

"Not remotely. But I see no reason to attribute awareness to a collection of paper and ink."

"Even this collection?"

"Especially this one."

He looked relieved.

"Good."

Miriam continued.

"If there is a process here, I suspect it's no more concerned with us than gravity is concerned with the people who fall from buildings."

Oliver's expression changed.

"That's considerably less reassuring."

"I didn't intend it to be."

After Oliver left, Miriam returned to the study. The house was quiet. She had intended to put the investigation aside for the evening, but the seventh division continued to trouble her. Not its more extravagant suggestions. Those could be dismissed, or at least suspended, pending evidence. What disturbed her was the structure of its argument. The text appeared to treat existence not as an elementary property possessed by objects, but as something established through relationships. That was not entirely foreign to modern physics or philosophy.

Objects were identified by their properties and interactions. Events occupied positions within causal structures. Certain theoretical descriptions privileged relations over independently existing substances. But the seventh division seemed to go further. It suggested that the distinction between actual and possible structures might not be fundamental. A sufficiently constrained possibility could become indistinguishable from a realised history. Miriam disliked the phrasing. It invited more speculation than it justified. She opened her notebook. There was a new entry beneath the list of textual witnesses. She stopped. The handwriting was hers.

The ink was black. The entry referred to the Paris manuscript. It included the shelfmark, the date of the supposed discovery and a brief description of the seventh division. Miriam could not remember writing it. That alone was no longer especially surprising. What troubled her was the date. She had written the date beside the entry, as she habitually did when making research notes. It was three weeks earlier. Before she had first received Oliver's email. Before she had begun investigating the physical Necronomicon.

Before the publication of the City of Dis essay. She examined the page. The entry was not squeezed into an existing space. It formed part of the original sequence of notes. The surrounding references concerned Lovecraft's fictional bibliography. One referred to an invented Greek translation. Another to the supposed Wormius edition. Then came the Paris manuscript. Miriam read the entry carefully. It described the manuscript as a possible witness to an earlier version of the seventh division. There was a short comment in the margin.

Important if authentic. May resolve title.

She turned to the previous page. The notes were familiar. She remembered writing them. She turned forward again. The Paris reference was still there. Below it was another sentence.

Do not reconstruct missing conclusion without independent witness.

Miriam stared at the words. The warning was precisely the precaution she had taken that afternoon. Yet the note appeared to have been written weeks earlier. She felt, for the first time, a distinct reluctance to continue reading. Not fear of the book. Not fear of the entities described within it. Something more immediate. The possibility that her own decisions were being incorporated into a history she had not yet lived. She closed the notebook. Then she opened it again.

There was no reason to assume the warning was prophetic. It might be an altered record. It might be a forgotten note. It might be part of an elaborate deception. And if she began treating every unfamiliar entry as evidence of an impossible process, she would rapidly lose the capacity to distinguish observation from interpretation. She reminded herself of that several times. Then she checked the date of the notebook's earliest photographs.

She had photographed several pages during the original research, sending the images to a colleague while discussing a bibliographical question. She found the email. It was dated three weeks earlier. The photographs were attached. She opened the relevant image. The Paris entry was present. The warning beneath it was present. She enlarged the photograph. The handwriting matched the physical notebook. There was no obvious discrepancy. Miriam sat back. The evidence was becoming painfully consistent. She had no independent record establishing that the entry had ever been absent.

Only her memory. And memory, as she had spent her professional life explaining to students, was not a reliable archive of the past. She looked at the photograph again. The warning seemed almost absurdly practical.

Do not reconstruct missing conclusion without independent witness.

It did not say that reconstruction was dangerous. It offered no explanation. It merely advised restraint. Perhaps she had written it because she recognised the ordinary scholarly risk of conjectural emendation. Perhaps she had forgotten doing so. Perhaps the note had acquired a past because it was now a coherent part of her investigation. Each possibility remained open. She returned to the Paris manuscript. The digital image displayed the damaged final sentence.

For those things which are constituted by necessity do not await the will of...

The last word remained unreadable. Miriam resisted the temptation to supply one. Instead, she examined the surrounding page. There was a marginal note in a different hand, probably added by a later reader. It consisted of two short Latin words.

Non invocantur.

They are not invoked. Miriam leaned closer. The note had been written beside the damaged passage. It appeared to be a correction or warning. She checked the scholarly transcription. The marginal note was recorded. Its date was uncertain. The editor suggested that it belonged to a later stage of the manuscript's use, perhaps the fifteenth century. Miriam read it again.

Non invocantur.

Not invoked. She considered the implications. If the note referred to the entities discussed in the seventh division, it might be an attempt to distinguish them from spirits called by magical ritual. That would be consistent with the passage's emphasis upon impersonal necessity. But it was also possible that the note had nothing to do with such entities. A medieval reader might have been correcting a theological interpretation, or objecting to a translator's terminology. Miriam refused to decide. She looked at the next marginal note.

This one had been written in smaller letters, close to the edge of the page. The script was difficult to read. She enlarged the image. There were perhaps six words. The first was unclear. The final three were more legible.

... iam semper fuerunt.

They have already always existed. Miriam stared at the phrase. It was awkward Latin, but the meaning was plain enough. She checked the transcription. The note appeared there too. The editor had translated it as an expression of the timeless existence of certain spiritual beings. Miriam looked back at the text. The phrase could certainly be understood that way. But it suggested something more disturbing in the context of her investigation. Not entities waiting to be summoned. Not creatures approaching from some inaccessible region of space.

Entities whose histories might become established retrospectively. Entities which, once included within a coherent account of reality, would appear to have existed all along. She closed the manuscript. The room was very quiet. Miriam thought of the Iram excavation. The archaeological reports. The geological surveys. The peculiar structures beneath the desert. She thought of the lost Greek translation and the increasingly detailed evidence concerning the Arabic original. Each new witness required additional history. Each additional history required earlier events.

And some of those events, if the Necronomicon's descriptions were taken seriously, required the existence of things which humanity had previously encountered only in fiction. She could no longer avoid the question. If an imaginary book could acquire a past, what would prevent the universe from acquiring the things described in that book? And if those things belonged to periods long before humanity existed, what form would their emergence take? Not an arrival. Not an invasion. Not an awakening. An alteration of what had always been true.

Miriam opened her notebook to a fresh page. The old text did not describe things coming into existence at a particular moment. It claimed, or might claim, that certain relationships could be required by a complete history. If the book's supposed entities possessed physical referents, traces of them would lie in earlier rock, earlier astronomical observations, earlier records. No one would witness an arrival.

She wrote to Yasmin Qureshi, a mathematical physicist whose seminars she had attended in Cambridge. Her message asked for a rigorous distinction between a changing physical history and a changing account of one. She deliberately omitted the word Necronomicon from the opening paragraph. Yasmin replied within the hour.

Your question is either badly formulated or considerably more interesting than you realise. Are you claiming that the underlying physical state changes, or merely that the available description becomes more complete? Those are radically different propositions. Bring evidence, not an interpretation.

Miriam smiled. It was the first sensible invitation she had received in days.

Before preparing the papers, she reopened the sealed experimental records. The prints and their digital fingerprints agreed. The fourth extract from the seventh division, however, contained a sentence which she had not seen when it was prepared. It completed the damaged passage she and Oliver had refused to reconstruct:

For those things which are constituted by necessity do not await the will of the one who names them.

The line appeared in the photograph taken before the envelope was sealed. Its digital hash matched the recorded value. No tampering had been established. The contradiction was between the record and Miriam's recollection, a poor basis for announcing a new physical law. Nevertheless she sat for a long time looking at the sheet.

The original Latin leaf remained damaged; no surviving witness supplied those exact words. She checked the Paris transcription again, then the London photograph. She had not introduced the completion into either. She could not establish where it had come from.

At five she went to the kitchen. The kettle had boiled dry enough to leave a pale mark on its heating plate. She could not remember switching it on. On the worktop lay the note Oliver had made before leaving, his proposed clause erased so thoroughly that only its pressure remained.

Miriam touched the scored paper. She had prevented him from completing a sentence. Her own archive now contained a completion instead. The difference between the two facts was not a theory; it was a small, shameful uncertainty about her powers of attention. She placed the disputed sheet in a fresh envelope and wrote the date across the seal. Then she turned over the earlier notebook page, which she knew almost by heart.

Do not reconstruct missing conclusion without independent witness.

It bore a date three weeks earlier. Miriam left the envelope on the desk. Beyond the window, the stars had begun to appear. If a history was being made self-consistent, it had not asked whether its human investigators approved of the corrections.

Chapter IV 4 of 10

The Probability of Yesterday

Dr Yasmin Qureshi had three objections to the proposition that the past was changing. The first was that it contradicted the ordinary causal structure of every physical theory which had survived serious experimental scrutiny.

The second was that the evidence presented in its support was indistinguishable, at least initially, from the consequences of mistaken attribution, imperfect record-keeping, selective attention and the notorious unreliability of human memory. The third, which she considered the most serious, was that nobody had explained what the proposition actually meant. She delivered these objections within eleven minutes of Miriam Vale's arrival.

They were sitting in Yasmin's office, a narrow room in Cambridge containing two desks, a blackboard, several computer monitors and a collection of books arranged according to a system Miriam could not immediately discern.

On the wall opposite the window hung a framed reproduction of a nineteenth-century astronomical photograph. Beside it was a small notice bearing the words:

What observation would demonstrate that you are wrong?

Miriam approved of the sentiment. She was less enthusiastic about being required to answer it.

"Let's begin," Yasmin said, "by getting rid of the word 'changing'."

Miriam looked at her.

"Why?"

"Because you haven't established a change. You've established that several records disagree and that some people remember things differently from what their documents suggest."

"That's rather more than a minor disagreement."

"Perhaps. But it's not a change in the past. Not yet."

Yasmin was a small woman in her early forties, with dark hair cut just above her shoulders and the concentrated expression of someone who found imprecise arguments a personal inconvenience. She spoke quietly, but with a degree of exactness which made interruptions difficult. Miriam had known her for several years.

They had met at an interdisciplinary seminar concerning the relationship between mathematical description and physical reality, during which Yasmin had reduced an elaborate philosophical argument to three unsupported premises and a conclusion which did not follow from any of them. Miriam had liked her immediately.

"Very well," Miriam said. "What would you prefer?"

"Anomalous historical records."

"That seems rather evasive."

"It describes what you've actually observed."

Yasmin opened the folder Miriam had brought.

"Now, this book."

"The Necronomicon."

"Yes. An unfortunate choice of subject if you're trying to establish scientific credibility."

"I didn't choose it."

"Someone did."

"Lovecraft."

"Precisely. Which means we begin with an unusually large population of people predisposed to create elaborate forgeries, invent scholarly references and misunderstand their own sources."

"You're preaching to the converted."

"Good. Then we can dispense with the obvious."

She removed the first photograph. It showed the London printing resting in its conservation cradle.

"Your claim is that this is a genuine fifteenth-century book."

"The physical evidence is consistent with that conclusion."

"Fine. And your claim is also that the textual history associated with it was invented by Lovecraft."

"That is documented."

"Then either Lovecraft's invention was less original than you believe, or the historical evidence has been fabricated, or something else has happened."

"Yes."

Yasmin looked at her.

"That's not a paradox. It's a list of possibilities."

"I agree."

"Then why am I here?"

Miriam opened the second folder.

"Because the possibilities don't adequately account for these."

She laid out the photographs of the disputed leaf. Yasmin examined them.

"The text differs."

"Yes."

"The physical damage appears the same."

"Yes."

"Have you established that the images came from the same physical object?"

"The catalogue records say so."

"That's not what I asked."

"No. We haven't established it independently."

"Could the image files have been altered?"

"Yes."

"Could the catalogue records be wrong?"

"Yes."

"Could the photographs have been incorrectly dated?"

"Yes."

Yasmin placed them aside.

"Then I don't see why we're discussing the structure of spacetime."

"Because there's more."

"This photograph was taken before we destroyed its subject," Miriam said, placing the signed print on the desk.

Yasmin examined the magenta streak, the initials and the note at the foot of the page. Miriam described the shredder jamming, the waterlogged pulp and the cut beside her thumbnail, now almost healed. Oliver's account had been written separately. It agreed with hers about the physical act. The photographs and the signed manifest now told a different story: an unsigned duplicate had been pulped, the annotated print retained for comparison. The records made the alternative decision look prudent, even inevitable.

Yasmin did not ask whether the print had materialised. She laid out both accounts on a sheet of paper, separating what each witness remembered from what the surviving objects and images recorded.

"You have two recollections of destroying the same marked object," she said. "The present record supports a deliberate decision to preserve it. There is no remaining observation of a vanished sheet returning. But there is a contradiction between independent testimony and the chain of custody."

"It costs us more than a misdated catalogue," Miriam said.

"Yes. It costs you confidence in your own procedural record." Yasmin turned the sheet over. "And perhaps that's the first thing we should try to predict, rather than another medieval book."

Miriam produced the conservation report dated 1971, with its reference to a paper published in 1998. Yasmin read it. Then she examined the associated photocopy and catalogue references. For the first time, she stopped asking questions. She read the report again.

"This is peculiar."

"I thought you'd find it so."

"A reference to an article twenty-seven years before publication."

"Yes."

"Have you contacted the author?"

"Eleanor Marsh. She's confirmed the article's history, but doesn't remember seeing the conservation report."

"And the report's provenance?"

"Good, but not unassailable."

Yasmin looked at her.

"Nothing is unassailable."

"Quite."

She returned to the document.

"A retrospective alteration of a physical record would be one explanation."

"Yes."

"So would a forgery."

"Yes."

"So would an error in its dating."

"Yes."

Yasmin smiled faintly.

"I'm beginning to suspect you're rather disappointed that I haven't rejected the evidence outright."

"On the contrary. I'd have been disappointed if you had."

"Good. Because what you've brought me is a problem in historical evidence, not yet a problem in physics."

She set down the report.

"Now explain the part about your own research notes."

Miriam spent the next half-hour describing the alterations, or apparent alterations, in her notebooks, emails and working files. She explained the manuscript references she did not remember writing. She described Oliver's conflicting memories of teaching the Necronomicon as both an imaginary work and a genuine medieval text. She outlined the history of the seventh division, the Paris manuscript and the increasingly elaborate network of documentary sources supporting the book's supposed existence. Yasmin listened without interruption. When Miriam finished, she turned to the blackboard.

"Let's separate three things."

She wrote:

PHYSICAL STATE

RECORDED INFORMATION

REMEMBERED INFORMATION

"These are not interchangeable," she said.

"Agreed."

"A physical object has properties. A record contains information about those properties. A memory contains information which may or may not correspond to the record."

"Yes."

"Ordinarily, discrepancies arise because one or more representations are inaccurate."

"Yes."

"But you're entertaining a fourth possibility."

She added another line.

HISTORICAL CONSISTENCY

Miriam regarded it.

"Go on."

"You're suggesting that the apparent physical history of an object might become constrained by information which was established only later."

"Something like that."

"Then we need to distinguish two very different hypotheses."

She wrote them on the board.

H1: Information about the past changes.

H2: The past itself changes.

"Under the first hypothesis," Yasmin said, "nothing extraordinary has necessarily happened. We discover new evidence. We correct errors. We reinterpret records. Our account of the past improves."

"Yes."

"Under the second, the events which constitute the past are somehow different from what they previously were."

"Yes."

"And the difficulty is that you have no direct access to a previous physical past."

Miriam looked at her.

"Only to records and memories."

"Exactly. If those records are themselves part of the alteration, then you have a serious observational problem."

"Which is why I came to you."

Yasmin sat down.

"Then let's be precise. If every physical trace of an earlier history were replaced, including every observer's memory, there would be no observation by which you could establish that a change had occurred."

"I had reached much the same conclusion."

"Good. And if you can't establish a difference between two hypotheses by any possible observation, you need to be very careful about claiming that they describe different physical processes."

Miriam considered this.

"But some discrepancies survive."

"Apparently."

"The photographs."

"Yes."

"The conflicting catalogue references."

"Yes."

"Our memories."

"Those I would treat with considerable suspicion."

"Naturally."

Yasmin looked towards the board.

"Then perhaps the important question isn't whether the past changes."

"What is it?"

"Why some records remain inconsistent while others appear to agree."

She paused.

"If there's a genuine phenomenon, the discrepancies may be more informative than the coherent evidence."

Miriam looked at her.

"That's an interesting reversal."

"Not really. In physics, anomalies are often where a model's assumptions become visible."

She turned back to the photographs.

"Tell me something. Does the uncertainty seem randomly distributed?"

"No."

"Where is it concentrated?"

"In the portions of the text whose history was least completely established."

Yasmin looked up.

"That is interesting."

By lunchtime they had constructed a table containing sixty-three documentary claims. Each claim was classified according to its source, the date assigned to that source, its degree of independent corroboration and its relationship to the reconstructed history of the Necronomicon. Miriam insisted that the sources be assessed using ordinary historical standards. A printed catalogue entry was not equivalent to a surviving manuscript. A modern scholarly citation was not independent evidence if it merely repeated an earlier author's assertion.

An institutional record was not necessarily authentic because it appeared on an institutional website. Yasmin agreed with all of this. She was particularly interested in distinguishing independent evidence from evidence which appeared independent only because its common source had been obscured.

"Otherwise," she said, "you'll count the same assertion twenty times and persuade yourself that you've acquired twenty confirmations."

"One of the oldest scholarly mistakes."

"And not confined to scholarship."

They identified twenty-one claims supported by reasonably secure physical evidence. Another twenty-six depended primarily upon documentary references. The remainder were uncertain, contradictory or supported only by later commentary. Yasmin entered the relationships into a simple graphical model. Each node represented a proposition. Some concerned physical objects: the London printing, the Paris manuscript, particular sheets of paper, binding repairs or surviving inscriptions. Others concerned events: translations, acquisitions, discoveries, scholarly examinations and the preparation of catalogues. Still others concerned textual content. The connections represented dependencies.

If a catalogue described a particular manuscript, the catalogue entry depended upon that manuscript having existed, or upon the cataloguer having been mistaken or dishonest. If a seventeenth-century commentator quoted a Latin translation, the quotation depended upon some source of the quoted material being available. If two manuscripts shared an unusual error, their relationship required explanation through textual transmission or coincidence. The resulting network was complicated. But it was not incomprehensible. Yasmin studied the screen.

"This is a familiar kind of problem."

Miriam looked surprised.

"In what discipline?"

"Several. Constraint satisfaction, probabilistic graphical models, information theory. Different methods, but related ideas."

"Explain."

"We have a collection of propositions linked by conditions. Some combinations are mutually compatible. Others aren't."

"Like a textual stemma."

"Yes, though a general graph needn't have the branching structure of a manuscript family tree."

She pointed to two nodes.

"Suppose a document refers to another document. We need some explanation for that relationship. The second document may be genuine, misidentified, fabricated or something else."

"Yes."

"Now suppose the second document is independently cited by five other sources."

"Then its existence becomes better supported."

"Ordinarily. But we must discount common dependence and possible fabrication."

"Of course."

"The important point is that the network becomes more constrained as we add independent relationships."

Miriam considered this.

"And if a proposed object is required to satisfy those relationships?"

"Then the model may assign it a high probability."

"Even if the object doesn't exist?"

Yasmin shook her head.

"Don't confuse a mathematical model's preferred explanation with physical existence. A model can strongly favour a nonexistent planet, a forged document or an imaginary cause if its assumptions are wrong."

"Quite."

"But suppose, purely hypothetically, that physical reality itself admitted some process resembling constraint satisfaction."

Miriam said nothing. Yasmin continued.

"Then a sufficiently interconnected description might favour the realisation of a particular structure."

"That's the possibility I've been considering."

"I know. But I want to stress that we're now constructing a speculative model, not deriving an established physical law."

"Understood."

Yasmin returned to the blackboard. She drew a series of circles connected by lines.

"Imagine that each circle is a claim about the world. Some claims can be true together. Others cannot. A coherent history is an assignment of physical events which satisfies the relevant constraints."

"And an incoherent history?"

"An assignment which fails to satisfy them."

"Such as a fifteenth-century book whose bibliographical history originated in the twentieth century."

"Possibly. Although you mustn't assume that Lovecraft's invention was necessarily original merely because he believed it was."

"I know."

Yasmin continued drawing.

"Now here's the interesting part. Suppose the system is allowed to resolve inconsistencies by admitting additional historical events."

She added another circle.

"For example, the existence of an earlier manuscript."

Miriam leaned forward.

"Which supplies an explanation for the later printing."

"Yes."

"And the manuscript itself requires a history."

"Exactly."

Yasmin added more circles.

"An earlier translation. A source text. A chain of transmission."

"And the history expands."

"Potentially."

She put down the chalk.

"But this only becomes physically interesting if the process has observable consequences which differ from ordinary discovery or fabrication."

Miriam examined the diagram.

"Could such a system have a preferred direction?"

"What do you mean?"

"Could it prefer certain ways of resolving contradictions?"

"Any mathematical model can be constructed with preferences. The question is whether those preferences correspond to anything real."

"Suppose the least disruptive resolution is favoured."

"Least disruptive according to what measure?"

Miriam hesitated.

"Perhaps the fewest changes."

Yasmin smiled.

"Changes to what? If you're claiming the previous history is inaccessible, you've no objective baseline against which to count them."

Miriam looked at the board.

"Then what would you use?"

"Perhaps the complexity of the additional information required to make the network consistent."

"An information-theoretic measure?"

"Possibly."

"Would that be physically meaningful?"

"Not automatically. But it might give us a testable ranking of possible outcomes."

She returned to her computer.

"Let's see what happens."

The first model was deliberately simple. Yasmin gave each documentary claim a node and marked the dependencies between them. The 1891 accession, the London printing and the 1968 conservation photographs supported a medieval physical object. Lovecraft's letters supported a twentieth-century invention of the very same bibliographical particulars. The model could not accept every claim without qualification.

"The sensible answer is a mistake somewhere," Yasmin said. "That's our baseline. But indulge me in a different calculation."

She imposed an artificial condition: retain the disputed claims and find the smallest network of additional propositions that would make them coexist. The program supplied a Latin exemplar, a Greek intermediary, a chain of manuscript transmission, and a history connecting Lovecraft's invention to an older text. Miriam read the screen. "That's alarmingly close to what the archives have been supplying."

"Because we gave it the archives. Don't applaud a model for recognising its own inputs."

Yasmin turned the screen away. "What interests me is whether it can commit to a document we haven't found yet. Before we look."

Miriam thought of the gaps in the Greek-Latin tradition. For the first time, the question was not what evidence might accommodate the book's history, but what evidence should be found at a particular place, in a particular source, if the reconstruction had any predictive value. The difficulty lay in deciding what to predict. There were hundreds of possible missing documents. The text's supposed history contained numerous uncertainties, and almost any medieval source could conceivably acquire some indirect connection with it.

A vague prediction that another reference would be discovered was worthless. Given sufficient time and a sufficiently permissive interpretation, someone would almost certainly find one. Yasmin insisted upon narrowing the possibilities. They selected a single problem. The relationship between the Greek translation attributed to Theodorus Philetas and the later Latin version attributed to Wormius. The surviving documentary history contained a gap. The supposed Greek translation belonged to the tenth century. The Latin translation was assigned to the thirteenth.

There were references to Greek manuscripts, but few independently verifiable details concerning how the text had passed into Latin scholarly circulation. The gap was not itself unusual. Many medieval translations possessed fragmentary histories. But within the model it occupied a particularly constrained position. The Greek translation was needed to explain several features of the Latin text. The Latin version was, in turn, supported by the physical printing and associated historical records. If the entire network was to remain coherent, some route of transmission was necessary.

Miriam examined the model's output.

"It favours an intermediary source."

"Yes."

"Which is hardly surprising."

"No."

"Can we be more specific?"

Yasmin considered the network.

"The model identifies several possible solutions. The least complicated is a Greek manuscript available to Latin scholars before the alleged Wormius translation."

"Again, obvious."

"Perhaps. But look at the geographical constraints."

Miriam leaned forward. The documentary references suggested connections with southern Italy, Constantinople and several collections associated with the movement of Greek manuscripts into western Europe. Yasmin highlighted the relevant links.

"If we accept these references, an intermediary in the eastern Mediterranean is easier to accommodate than one originating in northern Europe."

"That makes historical sense."

"Good. Now we need a specific archive where such a reference would be expected but is not presently recorded."

Miriam thought for several minutes. Then she remembered a collection of medieval inventories associated with a former monastic library in southern Italy. It had been catalogued repeatedly, including in a substantial nineteenth-century publication. The relevant volumes were digitised. She knew the collection because it had appeared in an unrelated research project. There were several surviving inventories from the thirteenth and fourteenth centuries. They included Greek and Latin works, some identified only by abbreviated titles. Miriam opened the catalogue. They searched the relevant sections. Nothing. No Alhazred. No Necronomicon.

No plausible abbreviation. No work whose description suggested the missing Greek intermediary. Yasmin nodded.

"Good. We have a negative observation."

"Only of this catalogue."

"Yes. We mustn't pretend it proves no such source exists elsewhere."

They downloaded the catalogue. They recorded its bibliographical details and file hash. They photographed the relevant pages on a separate display. They made printed copies. Then Yasmin entered the prediction into a dated document.

If the proposed consistency process is real, and if the reconstruction of the Greek-Latin transmission is becoming more fully supported, a presently unattested reference may emerge within the documentary network connecting southern Italian Greek collections to the thirteenth-century Latin tradition.

Miriam read it.

"Still rather vague."

"Yes."

"Could we specify this catalogue?"

"We can, but that would be a much stronger prediction than the model justifies."

"Then perhaps we should."

Yasmin looked at her.

"Why?"

"Because otherwise we'll be tempted to count any future discovery as confirmation."

"Fair."

She amended the statement.

Primary target: the nineteenth-century published catalogue of the selected southern Italian collection, presently containing no identifiable reference to the proposed intermediary.

Predicted alteration: an additional entry, cross-reference or textual description connecting a Greek Alhazred work to the later Latin tradition.

Time frame: assessment during the coming week.

Controls: comparable entries concerning unrelated Greek philosophical and astronomical texts.

Miriam examined the document.

"That's better."

"Still not a precise physical prediction."

"No."

"But it's sufficiently narrow to make a positive result interesting."

"And a negative result?"

"Would count against this particular model and prediction. Not necessarily against every conceivable version of your hypothesis."

Miriam smiled.

"You're making it difficult to be right."

"That's generally how one finds out whether one is."

Oliver joined them by video call in the afternoon. Yasmin explained the experiment without mentioning speculative physical mechanisms. She described it as a test of whether a preselected documentary gap would acquire new supporting evidence. Oliver listened carefully.

"You're asking whether the catalogue will change?"

"Not necessarily," Yasmin said. "I'm asking whether a prediction based on the structure of the evidence will identify a previously unrecognised source."

"But the catalogue currently contains no such reference."

"Correct."

"And if it appears?"

"We investigate ordinary explanations first."

"Such as?"

"A different edition, a scanning error, a revised catalogue, a mistake in our original search, or an unnoticed abbreviation."

"And if none applies?"

"Then we have an anomaly worth considerably more attention."

Oliver considered this.

"Should I examine the original printed catalogue?"

"Yes."

"When?"

"As soon as practical. But record everything before you begin."

He nodded.

"I can arrange access tomorrow."

"Excellent. And don't look for evidence outside the specified target until we've completed the test."

"Why?"

"Because I don't want our expectations influencing which discoveries we count."

Miriam looked at Yasmin.

"There's another reason."

Yasmin turned.

"Yes?"

"If the process really is responsive to the construction of coherent descriptions, investigating additional sources might alter the network."

Yasmin considered this.

"That's speculation."

"Of course."

"But if you're concerned about intervention effects, we should record the scope of the intervention."

"Exactly."

Oliver looked between them.

"I must say, this is becoming the strangest bibliographical enquiry of my career."

Yasmin smiled.

"I sincerely hope it remains the strangest."

They had not expected the first result that afternoon. It arrived at 16:38. Oliver sent a short message.

I've checked our institutional holdings. We possess a copy of the southern Italian catalogue. The entry isn't present in the digitised version, but there appears to be a discrepancy in the printed index.

Miriam read the message aloud. Yasmin looked up sharply.

"What discrepancy?"

Oliver sent a photograph. It showed a page of the catalogue's alphabetical index. There was an entry under A.

Alhazred, liber graecus, 214.

Miriam stared at it. The reference was to page 214. That was one of the pages they had examined earlier. The digitised copy contained no such entry on that page. Yasmin leaned towards the screen.

"Interesting."

Miriam opened their saved image. The index entry was absent. She checked the original download. Absent. She checked the hash. It agreed with their recorded value.

"Different edition?" Yasmin asked.

"Possibly."

They contacted Oliver. He confirmed that the printed volume bore the same publication date and edition statement as the digital version.

"Check the pagination," Yasmin said.

"The same."

"Printer's marks?"

"Apparently identical."

"Could the index page have been replaced?"

"I'll examine it."

"Do that. And don't assume the entry is genuine until you've established the physical details."

Oliver agreed. Yasmin ended the call. Miriam looked at her.

"Well?"

"Well what?"

"Isn't that rather striking?"

"Certainly."

"One of the exact kinds of reference we predicted."

"Yes."

"In the selected catalogue."

"Yes."

"Within hours."

Yasmin looked at her.

"And now we must try very hard to explain it without inventing a new property of the universe."

Miriam smiled.

"I thought you might say that."

The original printed index was examined the following morning. Oliver reported that the entry formed part of the original typography. There was no sign of a replacement page or later insertion. The paper, type and binding were consistent with the surrounding gathering. More interestingly, the page contained several references to unrelated Greek manuscripts which appeared in both versions of the catalogue. Only the Alhazred entry differed. The first difficulty was that page 214 still contained no corresponding description.

The index referred to something which the main catalogue did not include. That was unusual, but hardly impossible. Indexes were frequently prepared from earlier drafts. Entries could survive after the deletion or rearrangement of their associated text. Yasmin accepted that explanation as a serious possibility. Miriam agreed. Then Oliver located another copy of the same edition in a university collection. Its index contained the Alhazred reference. Its page 214 contained a brief description of a Greek manuscript attributed to an Arabian author. The description was incomplete.

It gave no full title. But it mentioned a work concerning the names of things, the order of temporal events and the relation between life and persistence. Miriam read the photograph.

"That's the missing connection."

Yasmin frowned.

"Perhaps."

"It's exactly the kind of source the model predicted."

"Yes."

"In the specific collection."

"Yes."

"And the second copy contains material absent from the digitised version."

"Apparently."

Miriam looked at her.

"What more do you require?"

"Evidence that these really are copies of the same printing, and a careful examination of the possibility of different states or issues."

"Of course."

"Early printed books can contain cancel leaves, corrected gatherings and variant states."

"Naturally."

"And nineteenth-century catalogues can be republished with alterations while retaining much of the original title-page information."

"Yes."

Yasmin leaned back.

"But I agree it's interesting."

"Interesting?"

"Very interesting."

Miriam laughed.

"From you, I suppose that's a declaration of astonishment."

"Don't become accustomed to it."

The next result was more difficult to explain. At 11:42 Oliver sent them a photograph of a handwritten note on the inside cover of the university copy. The note referred to a correction issued by the catalogue's editor. It explained that the description on page 214 had been inadvertently omitted from some copies during printing. A replacement gathering had been distributed to subscribers. This was an entirely ordinary explanation for the discrepancy. Miriam felt a sudden wave of relief.

"There," she said. "Different states of the edition."

Yasmin examined the note.

"Yes."

"Nothing extraordinary."

"Not necessarily."

"Surely that's the explanation."

"Probably. But let's check."

They located the publisher's correspondence. There was a notice concerning the replacement gathering. It was dated several months after the catalogue's original publication. The notice explained that an omission had been discovered during the preparation of the index. It instructed subscribers to replace the defective gathering. Everything fitted. The digitised copy represented the defective state. The university copy contained the corrected gathering. The institutional copy retained the original gathering but had somehow acquired the revised index.

There were minor complications, but nothing beyond the ordinary possibilities of nineteenth-century printing and bookbinding. Miriam looked at Yasmin.

"So the prediction succeeded because the source was already there, but we'd overlooked the variant edition."

"That's one interpretation."

"A much more plausible interpretation than historical alteration."

"Certainly."

Yasmin returned to the model.

"But there's still a question."

"What?"

"Why did the model select this particular gap?"

"Because we supplied the historical constraints."

"Yes. And why did the archival record contain precisely the kind of intermediary required?"

"Because the historical tradition may be genuine."

"Exactly."

She looked at Miriam.

"If we accept the ordinary explanation, we've discovered evidence that the Necronomicon's Greek-Latin transmission has a more substantial historical basis than you believed."

"Which remains difficult to reconcile with Lovecraft's invention."

"Yes."

"So our ordinary explanation solves one problem by making another worse."

Yasmin nodded.

"That's what interests me."

They spent the remainder of Monday examining the consequences of the newly identified source. The Greek manuscript described in the catalogue was no longer in the original collection. It had apparently disappeared during the dispersal of the library's holdings. Several later catalogues mentioned it. One eighteenth-century scholar had discussed a possible relationship between the manuscript and an obscure Latin translation. A nineteenth-century editor had disputed the identification. The references formed a small but coherent textual tradition. Miriam found them increasingly persuasive. Yasmin found them increasingly inconvenient.

By evening the model had been updated to include the new material. Its preferred reconstruction now required fewer unsupported connections between the Greek and Latin traditions. The historical account had become more coherent. But the most striking feature lay elsewhere. When Yasmin compared the updated model with the one constructed before the prediction, she found that several propositions previously classified as weakly supported had acquired additional corroboration. The new references did not merely fill the selected gap. They also supported other parts of the reconstructed textual history.

A quotation attributed to a Greek source now provided an explanation for a distinctive Latin mistranslation. An obscure manuscript description accounted for a passage whose wording had previously appeared anomalous. A seventeenth-century catalogue entry supplied a plausible route of transmission between two collections. The network had become denser. And the new evidence had precisely the character Miriam had been describing. Not spectacular discoveries. Not objects appearing from nowhere. Mundane scholarly references, each possessing an apparently respectable history. Yasmin studied the results.

"There's a pattern here."

Miriam looked up.

"What sort?"

"The additional evidence isn't randomly distributed."

"That's what I've been saying."

"I know. But now I can formulate it more precisely."

She displayed the graph. The original network contained several poorly supported connections. Some could be removed without affecting the overall reconstruction. Others were essential to maintaining the relationship between major branches of the historical account. The newly identified sources clustered around the latter.

"They're resolving the most constrained gaps," Yasmin said.

Miriam examined the display.

"Which is what your model predicted."

"Broadly."

"And what does that suggest?"

"At present? That our search strategy is effective."

Miriam smiled.

"You're determined not to make this easy."

"Yes."

Yasmin continued.

"But if the pattern persists under controlled conditions, it might suggest that the apparent discovery of evidence is systematically related to the informational constraints of the reconstruction."

"Which would be extraordinary."

"Very."

"Could you quantify it?"

"Not responsibly yet. We need a defensible null model for ordinary discovery, including our selection methods."

She paused.

"But I can tell you what troubles me."

"What?"

"The additional evidence doesn't simply support the book's existence."

"Go on."

"It supports the particular historical structure which your essay reconstructed."

Miriam said nothing.

"And if that structure is becoming more fully realised, we need to ask how far its dependencies extend."

On Tuesday morning, Yasmin called Miriam before breakfast.

"I've been thinking about the model."

"That sounds ominous."

"It may be."

"What have you found?"

"Nothing new. But I think we've been making an assumption."

"Which one?"

"That the book is the object being explained."

Miriam sat down.

"Isn't it?"

"That's how we've built the network. We've treated the physical Necronomicon as the central node and its historical sources as supporting evidence."

"Reasonable enough."

"Yes. But it's a choice we've made."

"Explain."

"Suppose the book is not the central object. Suppose it's merely one part of a much larger consistency structure."

Miriam thought of the Iram excavation.

"The things described in it."

"Possibly."

"The pre-human civilisations."

"Perhaps."

"The entities."

Yasmin hesitated.

"Let's not run ahead of the evidence."

"Of course."

"But your graph contains references to physical things which are not books."

"Yes."

"Archaeological sites. Geological structures. Astronomical observations."

"Yes."

"Those references are not all equally important. But some may be necessary to make particular passages intelligible."

Miriam looked towards her study window.

"And if the same process extends beyond documentary evidence?"

"Then it might generate predictions outside textual history."

"Which would be rather more convincing."

"Potentially."

There was a pause. Then Yasmin said, "I want to be careful about something."

"What?"

"You keep describing the phenomenon as though the essay caused the book to exist."

"That's the working premise."

"Perhaps. But the mathematical model doesn't establish any direction of causation."

Miriam frowned.

"The essay came first."

"First according to what chronology?"

"The book was fictional until the reconstruction."

"That's what you remember. But the physical evidence says the book existed centuries before."

"Which is the contradiction."

"Yes. And your model doesn't tell us which side of that contradiction is fundamental."

Miriam considered.

"You mean we can't establish whether the reconstruction produced the historical book or whether the book somehow influenced the reconstruction."

"Exactly."

"But Lovecraft invented it."

"According to one body of evidence."

"And the medieval printing existed."

"According to another."

She paused.

"Our model can describe the constraints linking those claims. It cannot yet tell us which is cause and which is effect."

Miriam looked at her notebook.

"Then perhaps causation isn't the right relationship."

"Perhaps not."

"Could the structure be time-symmetric?"

Yasmin was silent for several seconds.

"Mathematically, certainly. Physically, that's a much larger claim."

"But possible?"

"Not ruled out merely by the form of the equations. That isn't evidence that nature behaves that way."

"Understood."

Yasmin continued.

"I've been trying to think of a name for the hypothetical process."

Miriam smiled.

"Oh dear."

"Something provisional. Something which doesn't imply magic or deliberate intervention."

"What do you suggest?"

"Consistency tunnelling."

Miriam considered the phrase.

"Rather evocative."

"Possibly too evocative. I don't mean quantum tunnelling in the established physical sense. It's an analogy for a transition between alternative consistent descriptions."

"From one history to another?"

"Perhaps. But I'd rather say between structures which differ in what they require to exist."

"Consistency tunnelling," Miriam repeated.

"It may turn out to be complete nonsense."

"Most interesting hypotheses begin their lives in some danger of that."

"Yes. The distinction is what happens when we test them."

By Wednesday Yasmin had a less comfortable result. As she incorporated additional independent references, the model accepted fewer historical reconstructions. That was expected: evidence normally removes possibilities. Yet the remaining reconstructions were also becoming harder to distinguish. One treated the book as medieval and Lovecraft as a borrower. Another treated the modern reconstruction as the local event from which the apparent provenance was retrospectively constrained. Both now required almost the same observable records.

She printed the two networks on separate sheets and marked the points where they differed. Most lay beyond the reach of surviving evidence. The black lines converged so completely that Miriam had to hold the sheets against the window to distinguish them.

"More evidence should help," Miriam said.

"Not if the evidence is what the hypotheses disagree about." Yasmin took the sheets back. "If all accessible records eventually support one consistent account, there may be no observation left which identifies an earlier one."

"And the records in which we remember otherwise?"

"The observations are our memories. We can preserve testimony. We can't turn it into a physical baseline by wishing it were one."

Miriam touched the signed photograph from their failed destruction test. The paper was intact. Her handwriting now supplied the reason it had been spared. The model could accommodate the memory of a different action as an error in recollection; it could also accommodate a hypothetical replacement of the relevant history. The present evidence did not choose between them.

Yasmin restored the earlier version of the graph. "We need predictions that fail, not just a more elegant explanation for why everything looks ordinary. The physical claims in the text might help. But I won't treat literary descriptions as biological data."

Their next prediction concerned the limits of the documentary network. Yasmin wanted to identify a physical observation which could not reasonably be explained as a newly discovered manuscript or catalogue entry. Miriam was less enthusiastic. The first experiment had been relatively contained. Its outcome had concerned the identification of a textual source.

A second experiment, if their hypothesis was correct, might encourage the appearance of something considerably less manageable. Yasmin challenged the assumption.

"You're treating the model as though it responds to our intentions."

"Not intentionally."

"Then why should investigating it make anything worse?"

"Because the act of constructing a coherent description might be part of the constraint structure."

"That's possible within your hypothesis, but we don't know whether it's true."

"Agreed."

"And doing nothing wouldn't necessarily prevent the process."

"No."

"Then perhaps the most responsible course is to understand it before the evidence becomes inaccessible."

There was a certain grim logic to the argument. But it required assuming that their investigation would not itself accelerate whatever was happening. They had no way to establish that. Yasmin appeared to recognise the problem.

"Let's avoid generating new speculative descriptions," she said. "We'll use existing ones and seek independent evidence."

Miriam nodded.

"That seems sensible."

"Which portions of the reconstructed Necronomicon contain claims about physical objects?"

"Several."

"Objects which could leave evidence?"

"Yes."

"Preferably evidence outside human history."

Miriam looked at her.

"Why outside human history?"

"Because it would remove many ordinary explanations involving forgery, misattribution and literary influence."

"Quite."

She thought for a moment.

"The text refers to pre-human civilisations."

"That's rather broad."

"And to biological forms supposedly existing long before familiar terrestrial life."

"Better."

"Some of those descriptions derive from Lovecraft's fictional account of the Antarctic expedition."

Yasmin considered.

"Wouldn't that create a problem? We know the literary source."

"Yes."

"Then a discovery resembling the fiction might simply reflect someone interpreting natural evidence through Lovecraft's imagery."

"Certainly."

"Could we identify a prediction specific enough to avoid that?"

Miriam hesitated.

"Perhaps by looking for physical structures whose properties are difficult to reconcile with established biological categories."

"Such as?"

"Organisms whose morphology is not constrained in the usual way."

Yasmin looked sceptical.

"That's not a prediction. That's a description broad enough to include a great deal of ordinary biology."

"You're right."

Miriam opened the relevant section of the reconstruction. The text discussed Alhazred's supposed descriptions of pre-human organisms, including entities whose bodies appeared capable of radical structural alteration. It also noted that the descriptions were inconsistent and possibly erroneous. Miriam reread the passage. Then she stopped.

"Perhaps the error is the important part."

Yasmin looked up.

"What error?"

"Alhazred's descriptions of the organisms."

"Go on."

"Suppose the text contains an incomplete or mistaken account of a real biological system."

"That's possible in principle."

"Then a coherent history might require not only the organisms, but evidence explaining why the author misunderstood them."

Yasmin considered.

"Which would require independent physical properties of the organisms, not merely agreement with his descriptions."

"Exactly."

"And those properties might be discoverable."

"Yes."

Yasmin turned to the computer.

"Now that is interesting."

She opened the model.

"What does the reconstruction actually constrain?"

Miriam began listing the relevant features. Not the theatrical details of the fiction. Not the names or imagined behaviours. Only the claims which might plausibly be translated into physical observations. A body plan capable of substantial reorganisation. A biological system not easily accommodated within familiar evolutionary lineages. Evidence of ancient existence. Possible relationships with geological structures outside the known record of human civilisation. Yasmin listened. Then she shook her head.

"It's still too vague."

"I know."

"And we're in danger of treating a literary invention as a biological hypothesis."

"Yes."

"But there's one point worth retaining."

"Which?"

"The prediction isn't necessarily that we'll discover an organism matching the description."

"What is it?"

"That independent evidence may emerge which makes the description appear to have had a physical referent."

Miriam considered.

"A difference between finding a shoggoth and finding evidence that someone once encountered something later described as a shoggoth."

"Exactly."

"And the latter could be much less obvious."

"Indeed."

Yasmin saved the model.

"Let's not publish a biological prediction yet. We don't have enough specificity."

Miriam felt relieved.

"Agreed."

"But we can establish something else."

"What?"

"The order in which the network is likely to acquire support."

She displayed a new diagram. The book's physical history was now densely connected. The Greek and Latin transmission had acquired additional witnesses. The Arabic original remained largely unsupported. The claims concerning pre-human observations were even more weakly constrained.

"If the model has any predictive value," Yasmin said, "we might expect the next substantial developments to occur where the existing history requires additional independent sources."

"Such as the Arabic tradition."

"Yes."

"Or the physical evidence supposedly underlying it."

"Possibly."

Miriam looked at the graph.

"And if the evidence emerges outside textual scholarship?"

"Then we'd have something considerably harder to dismiss."

On Thursday afternoon, Miriam received an email from Oliver. It contained a revised catalogue reference for the London printing. The Library had apparently located an early description of the book in a collection of seventeenth-century scholarly correspondence. The reference was genuine, according to the preliminary assessment, and appeared to supply additional evidence for the Latin text's circulation. Miriam forwarded it to Yasmin. Her reply came within minutes.

Another dependency satisfied.

Miriam stared at the words. There was something almost mechanical about the process. Not in the sense of gears turning or machinery operating. In the sense of a system resolving its own incompleteness. A gap appeared. A relationship was required. Evidence emerged. The resulting network became more coherent. And the newly coherent history supplied the basis for further relationships. She opened the latest version of Yasmin's model. The graph contained more nodes than it had on Monday. Some represented newly discovered documents. Others represented previously unattested events.

Still others represented hypothetical objects required to explain the documents. Miriam examined the outermost branches. They extended beyond the history of books and manuscripts. One branch led towards the supposed observations made by Alhazred. Another towards ancient geological formations. A third towards references concerning the physical structure of space and time. The book's textual history was becoming increasingly secure. But the greater structure remained unfinished. And the more she examined it, the less reasonable it seemed to treat the book as the central object.

Perhaps it was merely the first component to become sufficiently constrained. A comparatively simple structure. Paper. Ink. A binding. A few centuries of scholarly transmission. Those requirements were modest. The requirements of the things described within it were not. Yasmin telephoned that evening. The newest source had been entered into the network, and the apparent dependencies now reached beyond manuscripts. She did not ask Miriam to accept another speculative consequence. She asked what assumption they would be testing next.

"We've imposed an order on our discoveries," she said. "The essay first, the book next, then the historical witnesses. But the model contains no observation establishing that as a causal sequence."

"Our memories establish the order in which we encountered them."

"Yes. Not the order in which the physical relations became true. A time-symmetric constraint model can be written over a complete history, but mathematics doesn't tell us whether the universe uses it. Nor does it justify saying that physical laws are changing."

Miriam looked again at the outer branches of the graph, where historical claims depended upon supposed observations of pre-human things. The book was an old object with a binding and an accession number. Its content was quite another matter. The distinction had become easy to forget precisely because the book's provenance had grown so persuasive.

"The seventh division says some things are not invoked," she said. "Perhaps its author was trying to distinguish objects from the relations by which objects become possible."

"Perhaps. Don't let the translation do the work of an experiment."

Miriam smiled faintly. She heard Yasmin set down a cup. The conversation ended without either woman proposing a method to halt the process. In the study the pages of the withdrawn reconstruction occupied one side of Miriam's desk; on the other lay the surviving signed photograph she remembered destroying. She could feel the edge of the paper against her finger. There was no need to invent a more theatrical threat. She already had one event which the present world could no longer corroborate.

On Friday morning, Yasmin arrived at her office to find a message from the university's information services department. It concerned a backup archive she had requested for the experiment. The department had located an earlier copy of the southern Italian manuscript catalogue in an institutional research collection. The file was dated several years before Miriam's essay. Yasmin downloaded it. The Alhazred reference was present. She checked its metadata. The file appeared to have been created during a digitisation project in 2014. She checked the associated records.

The project was genuine. There were invoices, technical reports and correspondence concerning the digitisation. The catalogue had been scanned as part of a large collection of nineteenth-century printed reference works. There was nothing extraordinary about its inclusion. Yasmin opened the relevant page. The description of the Greek manuscript was present. The reference matched the recently discovered printed copy. She compared the file with the version she and Miriam had downloaded on Monday. The earlier file lacked the entry. The newer archival copy contained it.

Both appeared to represent the same nineteenth-century publication. Both possessed plausible histories. Neither showed obvious signs of manipulation. Yasmin sat very still. She had anticipated that a positive result would be difficult to interpret. She had not anticipated quite how quickly the evidence would become self-supporting. The newly discovered reference no longer depended upon a single printed copy. It possessed a digitisation history. That history was supported by administrative records. The records were themselves embedded in a larger institutional archive.

The additional evidence did not merely establish the manuscript reference. It established that the reference had been available to researchers for years. Yasmin opened the model. She added the new source. The network became more coherent. Several previously uncertain relationships were now better supported. The model assigned a lower complexity cost to the existence of the Greek intermediary. She stared at the screen. There was something profoundly unsatisfactory about the result. The mathematical model treated the additional evidence as a reduction in uncertainty. That was its purpose.

But Yasmin's own uncertainty was increasing. The more historically secure the evidence became, the less confident she felt that the documentary record could be treated as a passive account of past events. She looked at the framed notice on her wall.

What observation would demonstrate that you are wrong?

For the first time, she found the question difficult to answer. If the hypothesis was false, they might simply be discovering previously overlooked evidence. If it was true, they might be witnessing the establishment of that evidence. And if the process altered not merely the evidence but the entire chain of events required to produce it, the two outcomes might become observationally indistinguishable. Yasmin opened a new document.

She wrote:

A problem of identifiability

The proposed process may admit observational states which are indistinguishable from ordinary historical discovery.

Where the prior state is not independently preserved, a change in the set of physically realised events cannot be distinguished from a change in available information.

The phenomenon, if real, may therefore become progressively less detectable as the resulting history acquires independent support.

She paused.

Then added:

The strongest evidence may consist not of the completed historical record, but of the residual inconsistencies left during its formation.

She read the sentence. It was speculative. But it suggested a practical consequence. They needed to preserve the inconsistencies. Not merely the documents which appeared to change. The contradictions themselves. The dates which could not coexist. The references to publications which had not yet appeared. The memories of a fictional book whose historical existence was becoming increasingly difficult to deny. Those inconsistencies might represent the only accessible evidence of the transition. And transitions, in physical systems, sometimes revealed properties which stable states concealed. Yasmin saved the document.

Then she telephoned Miriam.

"I think we need to change our approach."

"How?"

"We've been concentrating on the book and the evidence supporting it."

"Yes."

"We should concentrate on the contradictions which haven't yet been resolved."

"Because they may disappear?"

"Exactly."

Miriam was silent for a moment.

"How much time do you think we have?"

"I don't know."

"But you think there's a limit?"

Yasmin looked at the increasingly coherent graph on her screen.

"If the model is even approximately right, every additional independent relationship may reduce the number of histories compatible with the evidence."

"And eventually?"

"Eventually there may be only one."

Miriam said nothing. Yasmin continued.

"At that point, the historical record would be completely consistent."

"Which sounds desirable."

"Ordinarily."

"And in this case?"

Yasmin looked at the screen.

"In this case, it might mean that nobody could demonstrate that the universe had ever been any different."

That afternoon Miriam opened the working draft of the City of Dis essay. It had provided the book with sources, translations, disputed readings and a rationale for its errors. Yasmin's model now contained the same links as constraints. Neither of them had established that the links did anything outside the mathematics. She wanted a test that did not depend upon paper.

The supposed Arabic tradition referred to ancient observations, including unfamiliar biological forms. Some of those descriptions had been developed through Lovecraft's Antarctic fiction. They were too imprecise to function as scientific predictions. Mutable anatomy, for example, could describe a great deal of ordinary life, a pattern of mineral growth or nothing at all.

Miriam made a list of features that could be tested without allowing a researcher to see the literary account: age of host deposits, structure of preserved compartments, chemical composition, evidence of contamination, and whether the apparent organisation persisted across unlike physical shapes. The list was less exciting than a lost city or a monster. That was its principal advantage.

Her telephone rang. Yasmin had found a series of Antarctic sediment samples collected long before their investigation. Different laboratories had examined them for different purposes. Some of the raw data remained available.

"Why those samples?" Miriam asked.

"Independent collections, old accession records, physical material which can be re-examined. And a paper using the phrase structural plasticity. It may have nothing to do with our question."

"Has the author read Lovecraft?"

"I haven't asked. I intend to request her account of the material before she sees any of ours."

"Who is she?"

"Dr Lena Voss. She works on ancient biological structures."

Miriam looked at the list beside the telephone. The paper's terminology overlapped with only one of its proposed observations. That resemblance could be accidental, or could reflect the familiar hazard of seeking an image of a fictional object in ambiguous data.

"Send her the questions," she said. "And don't tell her what we're comparing them with."

After the call she walked to the corner shop. At the crossing she found herself unable to remember whether the light had already changed when she stepped off the pavement. A bus passed close enough to stir the hem of her coat. She stood for several seconds after it had gone, frightened more by the ordinary failure of attention than by anything in the Necronomicon.

Back at home she wrote Lena Voss beneath the test plan. The book, if it was involved at all, would not be entitled to interpret the evidence. The sediment would have to speak through its chemistry, structure and geological context. For a few hours that division of labour was enough to let Miriam sleep.

Chapter V 5 of 10

An Error in Alhazred

Dr Lena Voss had been examining the same fragment of Antarctic sediment for eleven days before she began to suspect that there was something wrong with the photographs.

Not the specimen itself. There were several perfectly ordinary reasons why a fossil-like structure might resist identification. Mineral replacement could produce astonishingly convincing imitations of biological tissue. Pressure and heat could distort microscopic features. Chemical precipitation could create regular geometries which looked designed, and biological material could be transported into deposits considerably older than the organisms responsible for it. The history of palaeontology was littered with embarrassing examples of investigators discovering life where none had existed. Lena had no intention of adding her name to that particular literature.

The specimen was designated RB-17/642, a small fragment of dark sedimentary rock recovered from beneath the Antarctic ice during an earlier geological drilling programme. It had spent several years in storage before being selected for detailed examination as part of a project investigating the preservation of complex organic structures in ancient marine deposits. The project was unremarkable. Its funding was modest, its ambitions carefully restricted, and its likely publications destined for journals whose readership seldom extended beyond a few hundred specialists. Lena liked it that way.

The rock was provisionally assigned to the late Neoproterozoic, although its precise depositional history remained uncertain. Several surrounding horizons had been dated using volcanic material, and the resulting geological model placed the sedimentary sequence at approximately 630 million years old. That estimate was not a direct measurement of the age of the structures within the specimen. Reworking, contamination and later mineral infiltration all remained possibilities.

Lena had repeated that distinction so frequently to her younger colleagues that one of them had once suggested printing it on a notice above the laboratory door. The unusual material appeared in thin sections as a network of interconnected compartments. Some resembled elongated cells. Others formed irregular chambers. Still others looked like branching channels whose dimensions varied in ways not readily explained by ordinary sedimentary deformation. Under polarised light, the structures displayed a peculiar mixture of mineral and carbonaceous material. Nothing about them was impossible.

Nothing, individually, required an unfamiliar form of life. The difficulty was their organisation. Lena had examined dozens of sections taken from the original fragment. The structures differed greatly in appearance, yet their internal connections repeatedly exhibited the same topological pattern. Not the same shape. The same pattern of relationships. She had first noticed this while examining a series of three-dimensional reconstructions produced from X-ray microtomography. One region resembled a loose aggregation of rounded compartments. Another formed long, tubular structures.

A third consisted of flattened sheets intersecting at irregular angles. The shapes were different. But when the reconstructions were reduced to networks of connected regions, certain relationships remained unexpectedly consistent. It was as though the material possessed no preferred anatomy, yet retained some deeper principle of organisation. Lena regarded the observation with considerable suspicion. Algorithms were perfectly capable of discovering patterns that existed chiefly in their own assumptions. She had therefore repeated the analysis using a second segmentation method. The result was similar.

She then asked a colleague to examine the images without being told what she expected to find. He identified much the same relationships. That was interesting. It was not yet persuasive.

The geological literature contained numerous examples of self-organising mineral systems. Regular structures could arise without biological intervention, and the preservation of apparent relationships across differently shaped regions might reflect nothing more than the physics of mineral growth. Lena knew all this. What she could not explain was why the most conspicuous structures appeared to possess features associated with entirely different biological functions. One region contained a network resembling channels for transporting fluid. Another formed densely folded surfaces, greatly increasing the area exposed to its surroundings.

A third possessed repeated cavities whose arrangement suggested, though certainly did not establish, a capacity for detecting environmental differences. The structures were not organs. There was no evidence that they had ever functioned as organs. But their organisation looked uncomfortably like a catalogue of possible anatomical solutions. As though a single material system had explored several incompatible body plans without committing itself to any of them. Lena enlarged the most recent reconstruction. Then she opened the corresponding image from an earlier examination. The two did not agree.

She checked the filenames. The specimen identifier was correct. The acquisition parameters were different, but not sufficiently different to explain the discrepancy. A branching channel visible in the earlier reconstruction was absent from the later one. In its place was a flattened arrangement of connected compartments. She checked the underlying tomography data. The discrepancy appeared to originate in the reconstruction, not in any demonstrated physical alteration of the rock. The original projections were noisy, and different reconstruction settings could produce different apparent boundaries. That was the sensible explanation.

She wrote it in her notebook.

Apparent morphological variation probably attributable to segmentation instability. Repeat using fixed reconstruction parameters and independent acquisition.

Then she closed the files. It was nearly six. She had spent most of the day investigating a pattern which might prove to be nothing more than an artefact of image processing. There was no reason to suppose it was anything else. Nevertheless, before leaving, she opened the earliest description of the specimen. It was a preliminary laboratory report dated 1986, prepared during the original investigation of the drilling samples. The report described the unusual material as a probable mineral pseudomorph.

There was a handwritten annotation beside the relevant paragraph.

No stable cellular morphology.

Lena read it twice. That was a perfectly reasonable observation. But the next sentence was more peculiar.

Organisation appears conserved despite substantial variation in constituent structures.

She frowned. The report had been prepared decades before modern three-dimensional imaging made such an interpretation plausible. She examined the accompanying photographs. They were ordinary optical micrographs, blurred and poorly contrasted by contemporary standards. There was very little detail. Certainly not enough to establish the relational patterns she had been investigating. Lena looked again at the annotation. The handwriting belonged to the original investigator. He had apparently recognised something which the available images did not adequately demonstrate. She made a note to locate his working files.

Then she turned off the monitor and went home. Yasmin Qureshi telephoned the following morning. Lena knew her only slightly. They had encountered one another at several university seminars, and Lena remembered an argument concerning the interpretation of probability in biological models. Yasmin had been correct. Lena had found this irritating at the time.

"Dr Voss?"

"Speaking."

"Yasmin Qureshi. I hope I'm not interrupting."

"No. What can I do for you?"

"I've been looking at some of the published material concerning your Antarctic samples."

"Which material?"

"The unusual structures in the RB-17 series."

There was a brief silence.

"Why?"

"I'm investigating a problem involving historical records and the interpretation of incomplete physical evidence."

"That sounds rather distant from Antarctic geology."

"At present, yes."

Lena looked towards the laboratory window.

"Are you interested in the age of the deposits?"

"Partly. But I'm more interested in the structures themselves."

"Then you should understand that their biological interpretation is highly uncertain."

"I do."

"The published papers are preliminary."

"I understand."

"And some of the reconstructions are dependent upon rather questionable assumptions."

"That's precisely why I'd like to speak to you."

Lena considered this.

"What are you actually investigating?"

Yasmin hesitated.

"Something which may turn out to be an exercise in identifying false correlations."

"That sounds more promising."

"I thought you might approve."

"What's the correlation?"

"I'd prefer not to tell you yet."

Lena frowned.

"Why?"

"Because I don't want to influence your description of the material."

"You're proposing some kind of blind comparison?"

"Exactly."

"With what?"

"An independently prepared set of claims about physical organisation."

Lena thought for a moment.

"Are these claims from a biological model?"

"Not originally."

"Then where?"

"That's part of what I need to keep separate until we've recorded your observations."

Lena smiled despite herself.

"You do realise this sounds rather suspicious."

"Entirely."

"Are you involved with one of those people who think geological structures are evidence of ancient technology?"

"No."

"Or extraterrestrial intervention?"

"No."

"Or some elaborate theory about consciousness being fundamental to the universe?"

"Certainly not."

"Good. Those are the usual callers."

Yasmin laughed.

"I promise not to waste your time with a theory of universal consciousness."

"Then send me the questions."

"I'd rather discuss them in person. There's a colleague I'd like to bring."

"What sort of colleague?"

"A textual historian."

Lena stared at her telephone.

"A what?"

"A specialist in historical texts."

"Yasmin, have you changed your field?"

"Not yet."

"Then why does a textual historian want to examine Antarctic sediment?"

"That is a considerably longer story."

Lena considered refusing. She had no shortage of work, and the prospect of explaining microscopic geological structures to someone interested in medieval manuscripts was not particularly attractive. But Yasmin was a serious researcher. And the request for a blind comparison suggested, at least, an awareness of the dangers of confirmation bias.

"All right," Lena said. "Tomorrow afternoon. But I'd like the questions written down before we meet."

"Of course."

"And I don't want to hear any interpretation until I've finished explaining the evidence."

"Agreed."

"Good."

She ended the call. Then she returned to the 1986 laboratory report. The annotation still troubled her.

Organisation appears conserved despite substantial variation in constituent structures.

She looked up the original investigator. Dr Peter Haldane had been a sedimentary geologist with an interest in the preservation of organic material in ancient deposits. He had retired many years earlier and died in 2014. His published work was conventional. There was nothing suggesting any remarkable discovery. The original report had never led to a substantial publication. The specimen had been classified as an ambiguous mineral structure and returned to storage. Lena found this reassuring. Scientists routinely encountered oddities which did not justify further investigation.

Not every unresolved observation concealed a revolution in natural history. Most concealed inadequate data. She opened the archived correspondence. There were several letters concerning the difficulty of preparing reliable sections from the sample. One mentioned contamination during handling. Another suggested that the structures were probably formed by mineral precipitation along pre-existing fractures. Then she found a memorandum written by Haldane himself. It was dated 12 September 1986.

The relevant paragraph read:

The principal difficulty is not the absence of cellular organisation but its apparent multiplicity. Different portions of the material exhibit structures analogous to several distinct biological arrangements, yet the relationships between them remain unexpectedly regular. It is difficult to establish whether we are examining the remains of one organism, a colony of organisms, or an entirely non-biological system.

Lena read it carefully. It was a remarkably sophisticated description of the problem she had been investigating. Haldane had apparently recognised the central difficulty without possessing the imaging techniques needed to explore it. She continued reading. The final sentence was underlined.

It may be an error to suppose that the material possessed a single fixed morphology.

Lena sat back. The observation was entirely compatible with her own. And it had been made forty years earlier. She was impressed. She was also puzzled. Why had nobody followed it up? The correspondence provided a partial answer. The samples were limited. Their provenance was uncertain. The structures could not be demonstrated convincingly with the available techniques. The project had been discontinued. Nothing mysterious. Nothing extraordinary. But Lena found herself wondering whether the old report contained observations worth revisiting.

She sent a message to the laboratory's collections manager requesting access to the remaining material from Haldane's investigation. Then she began preparing for Yasmin's visit. Miriam Vale arrived with Yasmin shortly after two the following afternoon. Lena met them outside the laboratory. She was tall, with short fair hair and an expression of professional impatience which became more pronounced whenever someone attempted to explain her own research to her. Miriam immediately liked her. Not because Lena was welcoming. She wasn't.

But because she appeared to regard the entire visit as a potential waste of time, and seemed determined to establish whether it was one.

"Dr Vale?"

"Miriam, please."

"Lena."

They shook hands.

"I understand you're a textual historian."

"Yes."

"And you're interested in Antarctic microstructures."

"Indirectly."

Lena looked at Yasmin.

"That's not especially reassuring."

"I'm aware," Yasmin said.

They entered the laboratory. The room contained several microscopes, imaging workstations, sample cabinets and the usual assortment of instruments whose purposes were not immediately apparent to a visitor. There was nothing dramatic about it. No sealed chambers. No warning notices concerning unidentified life. No suggestion that its contents had anything to do with the more extravagant speculations of medieval occult literature. Miriam found this reassuring. Lena led them to a workstation.

"Before we begin, I want to establish something. The material we're discussing is not confirmed biological tissue."

"Understood," Yasmin said.

"It's not even necessarily a fossil."

"Yes."

"And the age of the host sediment doesn't automatically establish the age of every structure within it."

"Of course," Miriam said.

Lena looked at her.

"You'd be surprised how often that distinction gets lost."

"I spend my professional life making similar distinctions about manuscripts."

"Then perhaps we'll get along."

She opened the first image. It showed a polished section of dark rock containing pale, irregular structures.

"This is RB-17/642. The sample comes from a sedimentary sequence beneath the Antarctic ice. The dating model places the relevant horizon in the late Neoproterozoic, but the precise history of the material is still under investigation."

She pointed to a cluster of elongated cavities.

"These are the features which originally attracted attention."

Miriam examined the image. They resembled nothing in particular. Several branching forms lay embedded in a darker matrix. Some appeared tubular. Others were flattened or rounded. Lena continued.

"Initially, the structures were interpreted as mineral replacements along fractures. That's still a serious possibility. But their internal organisation is unusual."

She opened a three-dimensional reconstruction. The screen displayed a network of connected chambers. It rotated slowly.

"Here we're looking at one region of the specimen. The reconstruction is derived from X-ray tomography."

She changed the image. A second region appeared. Its shape was very different. Instead of elongated chambers, it contained broad flattened surfaces connected by narrow channels.

"Same specimen?" Yasmin asked.

"Yes."

"Same geological layer?"

"Yes."

"Same reconstruction method?"

"Initially, no. We've since repeated the analysis using matched parameters."

"And the result?"

"The difference in overall morphology remains."

Miriam studied the images.

"What do you mean by morphology?"

"The form and arrangement of the structures. Their shape, distribution and relationship to one another."

"But you said the internal organisation was similar."

"Some aspects of it."

Lena displayed a simplified network diagram.

"These are connectivity graphs extracted from the reconstructions. They describe which compartments connect to which others, without preserving their exact shapes."

The two diagrams were surprisingly similar. Not identical. But similar enough that Miriam could recognise corresponding branches. Lena pointed to one group.

"The gross anatomy changes. The internal relationships remain relatively stable."

"Could that happen in an ordinary mineral system?" Yasmin asked.

"Certainly."

"How?"

"Growth constraints, fracture networks, chemical gradients, repeated branching processes. There are several possibilities."

"Have you tested them?"

"Some. None reproduces all the observed features, but that's hardly surprising. We haven't established the correct formation mechanism."

Miriam examined the screen.

"Could it be biological?"

"Possibly."

"What kind of organism?"

"That's the question."

"Multicellular?"

"Perhaps. Or colonial. Or something simpler whose preservation produces a misleading appearance of complexity."

"What about a single organism capable of changing its structure?"

Lena turned towards her.

"Where did that come from?"

Miriam hesitated.

"A possibility suggested by the different morphologies."

"That's not an observation."

"No."

"We have a fossil-like structure. We don't have a living organism. We don't know whether the different regions ever possessed different shapes during life, or whether the shapes reflect mineralisation, damage or growth."

"Quite."

"So we certainly can't conclude that it changed its anatomy at will."

"I wasn't suggesting we could."

"Good."

Lena returned to the images.

"There's an additional difficulty. The structures don't conform comfortably to any particular known group."

"Because they're too old?" Yasmin asked.

"Partly. But age is not the principal problem. The structural organisation is."

She enlarged another reconstruction.

"Look at this region. The arrangement resembles a transport network. Here we have something more like an exchange surface. And here a cluster of compartments with repeated peripheral openings."

Miriam looked at them.

"They suggest different functions."

"Perhaps. But apparent function is not evidence of actual function. A mineral formation can resemble a lung without ever having been used to breathe."

"Naturally."

"That's why I'm reluctant to call this an organism."

Yasmin nodded.

"What would persuade you?"

"Independent chemical evidence for biological origin, convincing structural continuity, exclusion of contamination and a formation model which accounts for the material better than the available geological alternatives."

"And if those conditions were satisfied?"

"Then we'd have an interesting fossil."

"Only interesting?"

Lena smiled faintly.

"Potentially extraordinarily interesting. But I'd prefer to earn the adjective."

Miriam found herself becoming increasingly impressed. The woman was exactly what they needed. A researcher with no interest in the literary material, no reason to recognise its supposed implications, and a professional commitment to eliminating extraordinary interpretations before entertaining them. She had not been told about the Necronomicon. She had not been told about shoggoths. She had independently described a material system whose apparent organisation bore an uncomfortable resemblance to one of Lovecraft's more disturbing inventions. That did not establish any connection.

But it made the next part of the investigation necessary. They spent the next hour examining the original data. Yasmin was particularly interested in the relationship between the reconstructed images and the measurements from which they had been derived. She asked whether the apparent similarities between the different regions survived changes in segmentation criteria. Lena demonstrated that they did, though the strength of the resemblance varied. She asked whether independent analysts had obtained comparable results. Two had.

She asked whether the same procedure produced similar patterns in ordinary mineral specimens. Sometimes. That last answer interested Yasmin most.

"Then the pattern isn't unique."

"No."

"How much more strongly does it appear in this specimen?"

"That depends upon the metric."

"Which metric did you select before examining the data?"

Lena smiled.

"You've identified one of the weaknesses of our analysis."

"Which is?"

"We didn't. The interesting structural relationships emerged during exploratory work."

"Then the apparent significance may be overstated."

"Quite possibly."

Miriam listened. The conversation was reassuringly hostile to the sort of reasoning upon which bad apologetics, dubious archaeology and pseudoscience depended. It was also accomplishing something valuable. They were beginning to distinguish the observed properties of the material from the interpretations suggested by its appearance. Yasmin asked one final question.

"If you had to describe the most unusual feature of the specimen in a single sentence, what would it be?"

Lena thought for some time. Then she answered.

"The apparent preservation of organisational relationships across regions which possess no consistent overall anatomy."

Yasmin wrote the sentence down.

"Are you satisfied with that?"

"As an observation, yes. Though 'apparent' is doing important work."

"Understood."

Miriam opened her folder.

"May we show you something now?"

Lena looked at her.

"I was beginning to wonder whether you ever would."

Miriam removed a printed sheet. It contained a carefully prepared description drawn from the reconstructed Necronomicon tradition. The text had been stripped of proper names, supernatural terminology and references to the fictional circumstances in which the descriptions appeared. It was presented as a series of propositions concerning the physical organisation of an unfamiliar biological system. Lena read it. The first proposition concerned the absence of a stable body plan. The second described the development of temporary structures in response to environmental requirements.

The third proposed that apparently distinct anatomical forms might represent different configurations of a common underlying organisation. The fourth concerned the relationship between such an organisation and the physical substrate in which it was instantiated. Lena read the sheet again.

"Where did this come from?"

"A historical text," Miriam said.

"What sort of historical text?"

"One whose authenticity is disputed."

Lena looked at her.

"How old?"

"That is part of the dispute."

She examined the propositions.

"They're extremely vague."

"Yes."

"You could apply the first three to various known biological processes."

"Certainly."

"And the fourth is scarcely biological at all. It's a philosophical statement."

"I agree."

"Then why are you showing me this?"

Miriam hesitated.

"Because the material you've described appears to possess some of the properties anticipated by this account."

Lena gave a short laugh.

"That's a very dangerous word, 'anticipated'."

"Yes."

"Do you mean the author predicted these structures, or that you've selected descriptions which resemble something subsequently discovered?"

"The latter remains a serious possibility."

"Good."

She returned to the sheet.

"There's nothing here that requires an unknown organism. Even if my specimen proves biological, the resemblance could be coincidental."

"Quite."

"And the description doesn't mention any physical measurement, chemical composition or specific anatomical feature."

"No."

"Then it's not a scientific prediction."

"Not yet."

Lena looked at Miriam with slightly greater interest.

"Not yet?"

Miriam took out a second page.

"This is the fuller description, with its original context."

Lena read it. Her expression changed. She looked at Yasmin. Then back at Miriam.

"You're joking."

"No."

"This is Lovecraft."

"Yes."

"The creatures from his Antarctic story."

"Yes."

Lena put down the paper.

"Shoggoths."

Miriam nodded. For several seconds Lena said nothing. Then she began to laugh. It was not a particularly pleasant laugh.

"Good God. You've brought me a description of a fictional monster and you're asking whether my fossils resemble it."

"Not quite."

"That's certainly what it looks like."

"Which is why we asked you to describe the evidence before seeing the text."

Lena looked at Yasmin.

"Did you know about this?"

"Yes."

"And you agreed to participate?"

"I agreed because we needed an independent account of the specimen's properties."

"Independent of a story by H. P. Lovecraft."

"Exactly."

Lena shook her head.

"Yasmin, I expected better."

"So did I," Yasmin said. "Which is why I haven't suggested that your specimen is a shoggoth."

"Then what are you suggesting?"

"At present, nothing about the specimen beyond what you've established. But we have an anomalous historical problem, and this literary description is part of it."

Lena looked towards Miriam.

"What historical problem?"

Miriam considered how to answer. She had no intention of beginning with a claim that the past was changing. Instead, she described the London printing. Its apparently authentic physical construction. Its provenance. The contradictory literary history. The documentary references which predated the events they supposedly described. She explained the Paris manuscript and the emergence of additional textual witnesses. She did not attempt to persuade Lena that reality had been altered. She simply outlined the evidence and the difficulty of reconciling it. Lena listened with increasing scepticism.

When Miriam finished, she said, "There must be a mistake."

"Almost certainly," Miriam replied.

"Then find it."

"That's what we're attempting to do."

"By examining Antarctic sediment?"

"Because some of the reconstructed text's claims concern physical phenomena rather than books."

"And you think those claims are becoming true."

"I think we have reason to investigate whether previously unsupported claims are acquiring independent physical referents."

Lena stared at her.

"That's a very elaborate way of saying something impossible."

"Yes."

"Do you believe it?"

Miriam considered.

"I believe we have encountered evidence which I cannot presently reconcile with an ordinary historical account."

"That isn't the same thing."

"No. It isn't."

Lena looked at Yasmin.

"And you?"

"I think the model is interesting enough to test, and insufficiently supported to believe."

Lena sat back.

"Well. At least one of you is still behaving like a scientist."

Miriam smiled.

"I shall try not to take that personally."

Lena might have ended the conversation there. Later, she would wonder why she had not. There were several reasons. The first was that Yasmin's involvement made the affair difficult to dismiss as an exercise in occult enthusiasm. The second was that Miriam had been unusually careful to distinguish evidence from interpretation. The third was less defensible. There was something about the description which troubled her. Not the resemblance between Lovecraft's fictional creature and her specimen. That was far too general to carry much evidential weight.

It was the emphasis upon a lack of fixed anatomy. Most fictional monsters were recognisable creatures. They possessed characteristic shapes, structures and behaviours. Even imaginary organisms assembled from incompatible biological features generally retained enough anatomical consistency to remain identifiable. Lovecraft's shoggoths were different. Their defining property was not a particular form, but the capacity to produce forms. Temporary eyes. Temporary mouths. Structures appearing and disappearing as required. The creature was not so much an organism with an anatomy as a material system capable of generating anatomical arrangements.

Lena recognised the conceptual distinction. It was not proof of anything. But it was interesting. She reopened the three-dimensional reconstruction.

"There's something I should clarify."

Miriam looked up.

"What?"

"Your literary description concerns an organism capable of changing its shape during life."

"Yes."

"My specimen doesn't establish that."

"Understood."

"But the distinction between fixed anatomy and conserved organisation is relevant to what we're observing."

"How?"

"Suppose, entirely hypothetically, that these structures are biological. We would ordinarily attempt to identify the organism by its morphology."

"Yes."

"But what if morphology isn't the stable characteristic?"

Miriam leaned forward.

"Then what is?"

"The organisation of the system. Its internal relations. The way its constituent parts interact."

"Which might survive changes in external form."

"Potentially. But we haven't shown that any change occurred during life."

Yasmin examined the diagram.

"Could the same organisation produce radically different structures under different conditions?"

"In principle. Developmental biology already provides plenty of examples of organisms producing different structures in response to environmental conditions. But this would be unusually extensive."

"Could it occur without central control?"

"Certainly. Complex organisation doesn't require a conscious designer or executive intelligence."

Miriam looked at the screen.

"Then the apparent anatomy might be a temporary expression of something more fundamental."

Lena considered.

"That's a reasonable hypothetical description."

"And the fossil record?"

"Would preserve whichever structures happened to be present when the material was deposited and mineralised."

"Not necessarily a characteristic body plan."

"Exactly."

Yasmin looked at the two reconstructions.

"So if two portions of the same organism possessed radically different shapes, a palaeontologist might initially classify them as different organisms."

"Yes."

"Even though their underlying organisation was related."

"Yes."

Miriam thought of Alhazred's supposed descriptions. A mass of living substance producing eyes and mouths wherever required. A body without a permanent shape. An organism whose apparent anatomy consisted of transient adaptations. She felt an unpleasant sense of recognition. The ancient description was absurdly anthropomorphic. It interpreted unfamiliar biological forms in terms of mouths, eyes, limbs and other recognisable structures. But beneath those interpretations lay an idea which was not inherently ridiculous. An organisation capable of assuming different physical forms.

The literary description might be wrong in almost every particular. Yet a real structure could conceivably possess enough of the underlying properties to explain why such a description had arisen. That was the direction of inference she had been missing. They did not need to establish that shoggoths existed as Lovecraft had imagined them. They needed to ask what sort of physical phenomenon could give rise to a description which a later author would interpret in those terms.

And whether the increasingly coherent history of the Necronomicon was beginning to require such a phenomenon. Lena retrieved a second sample. It was a narrow section of dark rock mounted in a small transparent holder.

"This comes from an adjacent portion of the core. It's not the same material you've been looking at."

Miriam examined it. To her eyes, it was simply a piece of stone. Lena placed it beneath a microscope.

"Have a look."

Miriam adjusted the focus. The image was surprisingly delicate. Pale chambers lay within the darker rock, their boundaries forming thin, irregular walls. Some structures resembled networks of vessels. Others appeared as rounded compartments. They were not conventionally cellular. Nor were they wholly unlike the preserved remains of biological tissue.

"What am I looking at?" Miriam asked.

"Possibly mineral replacement. Possibly a biological structure. Possibly a combination of processes."

"How old?"

"The host deposit is approximately 630 million years old, subject to the qualifications I've already explained."

Miriam looked again.

"And these structures?"

"We haven't established their age independently."

"Could they be much younger?"

"Yes. Material can enter older rock through fractures. Microbial activity can occur long after deposition. Contamination during collection and preparation is another possibility."

"Have you found evidence for that?"

"Not yet. But absence of evidence isn't a licence to ignore the possibility."

Miriam smiled.

"I couldn't agree more."

Lena adjusted the microscope.

"There's something else. The surrounding matrix contains structures similar to these, but they're distributed across several different layers."

"Could that be transport?"

"Possibly."

"Or repeated growth?"

"Possibly."

"Or a single organism extending through the sediment?"

"That's much less likely, and we have no evidence for it."

Miriam stepped away. Lena removed the specimen.

"One of my colleagues suggested that the material might preserve an unusual colonial organism."

"What did you think?"

"I thought it was premature."

"Naturally."

"But it raised an interesting question."

"Which?"

"If these are biological structures, how would we distinguish a colony of separate organisms from a single distributed organism?"

Yasmin looked interested.

"By genetic evidence?"

"If any suitable material survived, perhaps. But in specimens of this age, original genetic information is not something we should expect to recover."

"Then by structural continuity?"

"Possibly. By growth patterns, internal transport systems, chemical relationships and other evidence."

"And here?"

"We don't have enough."

Lena returned the specimen to its case.

"The difficulty is that our biological categories are derived from organisms whose organisation we understand reasonably well. We recognise individuals because their development, reproduction and physical boundaries give us reasons to distinguish them."

"And this material?"

"We don't know what constitutes an individual, assuming it's biological at all."

Miriam thought about that. It was a more disturbing proposition than the existence of some gigantic prehistoric monster. A monster was at least an object. It possessed a location, a body and a boundary. One could imagine observing it. One could, in principle, discover its anatomy and understand its behaviour. But a biological system in which individuality was not clearly associated with a stable physical body would challenge assumptions far more fundamental than the existence of an unfamiliar species. Miriam looked at Lena.

"Could there be an organisation without a fixed individual?"

"Many living systems already complicate the distinction between individual and colony. Some fungi, plants, colonial animals and microbial communities make individuality less straightforward than people imagine."

"But this would be more radical."

"If your interpretation were correct, possibly. But at present we have a small piece of rock and an alarming quantity of speculation."

"Quite."

Lena smiled.

"Please don't forget that."

They returned to the historical material after tea. Miriam had brought copies of the relevant passages from the reconstructed Necronomicon, including those associated with the pre-human civilisations described in Lovecraft's Antarctic fiction. She arranged them on the table.

"The traditional account presents the shoggoths as artificial organisms," she said.

"Artificial?" Lena asked.

"Constructed by an ancient non-human civilisation. They were supposedly used as labourers and could assume different forms according to the purposes of their makers."

Lena looked at her.

"That's very convenient."

"In what sense?"

"It explains every possible anatomy by appealing to intentional design."

Miriam smiled.

"An explanation with which I've had occasion to become rather familiar in other contexts."

"Yes. But scientifically it's useless unless you can identify independent evidence of manufacture."

"Agreed."

"Otherwise you're simply saying that an unknown intelligence produced whatever feature you don't understand."

"Precisely."

Lena examined the text.

"There's another problem."

"What?"

"Your fictional civilisation supposedly engineered these organisms."

"Yes."

"When?"

"In a very remote period of terrestrial history. The fictional chronology is not always consistent."

"Then the relationship is impossible to test without additional evidence."

"Yes."

Lena pointed to the reconstructed description.

"And look at this. The creatures are said to develop sensory structures wherever necessary."

"Yes."

"That assumes the structures are sensory organs."

"Of course."

"But an unfamiliar observer might interpret an opening as a mouth, a cavity as an eye, or a branching channel as a nervous system."

Miriam looked at her.

"Which would be a mistake."

"Potentially."

"And the mistake might persist through successive translations."

"Certainly."

Miriam examined the passage. For the first time, she considered it not as a description of a fictional monster, but as a hypothetical account of someone trying to understand an unfamiliar natural phenomenon. An observer encountered a material system whose morphology varied dramatically. He possessed no conceptual framework for its organisation. He therefore interpreted its temporary structures as familiar organs. He assumed that the changes occurred intentionally. He attributed the organism's existence to an intelligent maker. He described it as a servant. The mistakes were entirely comprehensible.

Indeed, they were exactly the mistakes one might expect from a writer who believed that every complex organisation required a controlling intelligence. Miriam looked up.

"There's something significant about that."

Lena frowned.

"What?"

"The errors."

"What errors?"

"Alhazred's supposed errors. If the book is merely a description of fictional monsters, the details are arbitrary."

"Yes."

"But if the text is acquiring a genuine historical basis, its errors may need explanations."

Lena considered.

"I don't follow."

Miriam pointed to the passage.

"Suppose a historical writer describes an unfamiliar animal incorrectly. The errors themselves can provide evidence concerning what he actually observed."

"Certainly."

"He might describe a rhinoceros as a kind of armoured horse. The description is wrong, but the mistake is intelligible."

"Yes."

"And if we encountered a previously unknown medieval account of such an animal, we might be able to identify the real organism behind it."

"Of course."

"Now suppose the same principle applies here."

Lena stared at her.

"You're suggesting that the text requires a real organism for its author to have misunderstood."

"Within the hypothetical consistency process, yes."

"That's absurd."

"Under ordinary assumptions, absolutely."

"More than that. It's circular. You're using a fictional text to specify what evidence ought to exist, then proposing that its existence would explain the text."

"I know."

"Then why entertain it?"

"Because the documentary history appears to be acquiring precisely this sort of independent support."

Lena shook her head.

"You're connecting two speculative propositions."

"Yes."

"And neither has been adequately established."

"Correct."

Yasmin intervened.

"Which is why we need a prediction that distinguishes the hypothesis from ordinary coincidence."

Lena looked at her.

"What prediction?"

Yasmin opened the graphical model.

"If the text's apparent history is becoming more coherent through the addition of physical evidence, then the most informative discoveries may be those which explain the text's errors rather than simply resemble its descriptions."

Lena considered that.

"Meaning?"

"Suppose a newly recognised physical structure explains why an observer might have mistaken one natural process for another."

"Yes."

"That would establish a more specific relationship than superficial resemblance."

"Possibly."

"But only if the interpretation were fixed before the evidence was examined."

"Exactly."

Miriam looked at the passage again. There was one error which seemed especially important. The shoggoths were supposed to have been manufactured. Their existence was explained through intelligent intervention by an earlier civilisation. Yet if Lena's specimen proved to be biological, and if its age could be established, the material might represent a natural lineage of extraordinary antiquity. A real organism would not require the existence of the fictional engineers. It might instead explain why an observer, encountering unfamiliar traces, had imagined such engineers. Miriam considered the distinction.

The Necronomicon did not need to be right. It needed to have been wrong about something real. And that requirement might be considerably more dangerous. Because a correct description could be satisfied by a single object. An incorrect description required an object, an observer, a mistake and a history in which the mistake became comprehensible. The error demanded more of reality than the truth. The collections manager arrived shortly before five. He carried a small archival container.

"Dr Voss? The Haldane material you requested."

Lena looked surprised.

"Already?"

"It was in the adjacent storage series. The index had been updated."

"Updated when?"

"Recently, I think."

He placed the container on the table.

"There are four prepared sections and one unprocessed fragment."

Lena frowned.

"Four?"

"According to the inventory."

"I thought there were only two."

"Perhaps the older catalogue was incomplete."

"Possibly."

He left. Lena opened the container. Inside were several small specimen holders, each labelled with an accession number. The labels were old, written in faded ink. There was nothing unusual about them. She removed the first section. Then the second. The third. And the fourth. She stared at the final specimen.

"That's odd."

Yasmin looked up.

"What?"

"This one wasn't in the original inventory."

"Could it have been overlooked?"

"Certainly."

Lena examined the label. It bore the same collection reference as the other material. The handwriting appeared consistent. The preparation was mounted using an older method, and the adhesive showed signs of age. She turned it over. There was a second label.

RB-04/119.

Lena stopped.

"That's not from the same core."

Miriam leaned forward.

"Where is it from?"

"An earlier drilling programme. Different location."

"How much earlier?"

"The sample was collected in 1974."

Yasmin looked at the specimen.

"Could it have been stored with the wrong collection?"

"Yes. Archival misfiling is common."

Lena examined the original report. It mentioned RB-04/119 in a footnote. The description was brief.

Comparable reticulate structures identified in earlier material. Interpretation uncertain.

She read the sentence twice. Then opened the associated catalogue entry. The earlier specimen had been collected hundreds of kilometres from the later drilling site. The deposits were assigned to broadly similar geological periods, though their stratigraphic relationships were not straightforward. Lena looked at Miriam.

"Now that is interesting."

"In what sense?"

"If the material is genuinely comparable, then we may be dealing with a structure distributed across a considerably larger region than we thought."

"Biological?"

"Possibly. Or a widespread geological process."

"Could you compare them?"

"Of course."

She mounted the older section beneath the microscope. The image appeared on the monitor. Miriam recognised the pale branching compartments immediately. They were not identical to the earlier specimen. But their general organisation was strikingly similar. Lena adjusted the focus. Then increased the magnification. For several minutes she said nothing. Finally, she opened the analysis software. She selected a small region and began reconstructing its network of connected compartments. Yasmin watched. The resulting graph appeared beside the earlier model. The resemblance was substantial. Lena frowned.

"That's more similar than I'd expected."

"Could it be a common mineral formation mechanism?" Yasmin asked.

"Certainly."

"Or contamination from the same preparation process?"

"Also possible."

"Can we distinguish them?"

"Not from these images alone."

She examined the labels.

"The most important question is whether the samples were genuinely collected from independent locations."

Miriam looked at her.

"You doubt the records?"

"I doubt everything until the evidence justifies accepting it."

"Excellent."

Lena smiled faintly.

"I thought you'd approve."

She returned the specimen to its holder.

"We'll need to check the original drilling logs, sample transfers, preparation records and geological context. If the provenance is secure, then the distribution is interesting."

"And if it isn't?"

"Then we have a storage problem."

Miriam looked at the four prepared sections. The container possessed an ordinary archival history. Its labels were old. The reports were old. The samples appeared to have been collected decades earlier. Every element was exactly what one would expect from a genuine historical collection. And that was precisely what frightened her. The physical evidence was beginning to acquire the same character as the documentary record of the Necronomicon. An obscure item previously regarded as incomplete now contained additional material.

The additional material supplied a connection to an earlier source. The earlier source possessed its own independent provenance. And the resulting structure became more coherent. She looked at Yasmin. Yasmin was already looking at her. Neither spoke.

That evening, Lena agreed to examine the two collections more systematically. She did not agree that they represented unfamiliar biological organisms. She certainly did not agree that they had anything to do with Lovecraft. But the possible relationship between the samples was worth investigating. Yasmin proposed a formal separation of responsibilities. Lena and her colleagues would examine the physical material, establish its provenance and test its geological interpretation. Miriam would continue analysing the historical text.

Yasmin would maintain the mathematical model and evaluate whether the resulting observations supported any predictions beyond those expected from ordinary research. The three investigations would remain separate wherever possible. Interpretations would be recorded before results were exchanged. Lena approved.

"Good. Because I don't want my students being told that they're investigating prehistoric monsters."

"Neither do I," Miriam said.

"Or that they're discovering evidence of some ancient civilisation."

"Agreed."

"Or that the universe is rewriting itself to accommodate a fictional book."

Miriam hesitated.

"That one might be more difficult to avoid indefinitely."

Lena stared at her.

"Then I suggest you avoid it for as long as possible."

Yasmin smiled.

"We're in complete agreement."

Before leaving, Miriam asked one final question.

"Suppose the structures are biological."

Lena sighed.

"Go on."

"And suppose their original age is established."

"Yes."

"And suppose we find that the same underlying organisation occurs in substantially different forms."

"Yes."

"Would that be compatible with a single lineage?"

"Potentially."

"Even if the morphology varied dramatically?"

"Yes."

"And could such a lineage exist for immense periods without producing recognisable conventional fossils?"

"Possibly. But the question is too general to answer usefully."

Miriam nodded.

"Thank you."

Lena looked at her.

"What are you really asking?"

Miriam considered.

"Whether a form of life could remain recognisably the same without preserving a recognisable form."

Lena was silent for a moment. Then she answered.

"Biological identity isn't reducible to shape. We already know that. The harder question is what features of an organisation must remain continuous for us to call it the same living system."

"And if we couldn't identify those features?"

"Then we might not recognise it as life at all."

Miriam looked towards the specimen cabinets. That answer remained with her long after they had left the building. On the train back to Cambridge, Yasmin opened the latest version of the consistency model. She entered the existence of the earlier Antarctic sample as a provisional claim, assigning low confidence until its provenance could be verified. The new node carried provisional links to the geological age of the specimens, their apparent organisation, the two collections, possible formation processes and the historical descriptions of organisms without fixed anatomy. When Yasmin updated the model, a weakly supported branch became more coherent. That was a property of the network she had constructed, not proof of a physical change. Yasmin examined it for several minutes. Then she looked at Miriam.

Yasmin pointed to the biological branch. "I've treated the specimens as though they were evidence for separate organisms. Where did we establish individuality?"

"We haven't," Miriam said.

"The historical descriptions assume discrete creatures. The specimens might instead be manifestations of a distributed organisation, if they're biological at all."

Miriam considered the maps. "A single system extending through a large region?"

"Or a lineage whose members possess no stable morphology. Those possibilities aren't interchangeable, and we have evidence for neither yet."

Miriam looked at the diagram. The supposed shoggoths of Lovecraft's fiction were discrete beings, however amorphous. They were manufactured, controlled and employed. They possessed individual bodies, even if those bodies could change shape. But Lena's material did not yet justify any such interpretation. It might not represent individuals at all. The category could be fundamentally mistaken. Miriam remembered the seventh division.

What persists is not that which feels, but the order by which feeling is possible.

She had assumed the passage concerned consciousness. Perhaps it was more general. Perhaps the Necronomicon's supposed natural philosophy was concerned with forms of organisation whose identity was not tied to individual physical objects. The same underlying structure could be instantiated in different materials. It could acquire different shapes. It might persist through discontinuous physical manifestations. And an observer lacking the appropriate conceptual framework would mistake those manifestations for separate creatures. Miriam looked at Yasmin.

"Suppose Alhazred didn't understand what he was describing."

"That seems almost guaranteed, assuming he existed."

"Suppose he thought he was describing animals."

"Yes."

"But the underlying phenomenon wasn't an animal in any conventional sense."

Yasmin considered.

"Then his observations might correspond to local manifestations of something much larger."

"Exactly."

"That's speculative."

"Entirely."

"But it would explain why the descriptions are so inconsistent."

Miriam looked at the graph.

"And why the supposed creatures are said to assume different forms."

"Perhaps."

She thought for a moment.

"And why they were thought to be manufactured."

Yasmin looked at her.

"How?"

"Because an observer encountering a highly organised structure without recognising any familiar developmental process might infer deliberate construction."

"An argument from apparent design."

"Yes."

"Which wouldn't establish design."

"No."

Miriam leaned back.

"Alhazred may have made the same mistake humanity has made repeatedly. He encountered something organised, failed to understand its natural history, and invented an intelligence responsible for it."

Yasmin smiled faintly.

"That would be a satisfyingly ordinary error."

"Except for the thing he encountered."

"Which we have yet to establish."

"Quite."

The train continued north. Outside the windows, the landscape was dark. Miriam returned to the graph. The new evidence had not established that shoggoths existed. It had done something much subtler. It had provided a possible physical explanation for why a historical observer might describe an unfamiliar natural structure in terms resembling the fictional account. And if the process of retrospective consistency was real, that might be enough. Not enough to establish the entire fictional cosmology. But enough to make one portion of it historically intelligible.

One error corrected. One missing referent supplied. One more part of the universe's past becoming coherent. Lena remained in the laboratory long after the visitors had gone. She had intended to return home, but the earlier sample continued to trouble her. RB-04/119. Collected in 1974. Stored in an archival collection. Apparently examined by Haldane during his investigation in 1986. Its structural similarities to the newer material were real enough to warrant further study, though far from sufficient to establish a shared biological origin.

She opened the original drilling report. The relevant section described a sequence of sedimentary rocks recovered beneath the ice. The report contained the usual technical material: depths, lithological descriptions, drilling conditions and sample identifiers. She located RB-04/119. The sample was listed. Its depth was correct. The surrounding geology was consistent with the later catalogue. Everything was ordinary. Then she found an appendix containing the original microscopic observations. The appendix was dated 1976. It described several unusual mineral structures. One entry referred to a branching network of compartments.

Another mentioned a complex reticulate arrangement whose apparent morphology varied between prepared sections. Lena read the entries. They were consistent with the specimens she had examined. Then she noticed a reference to a supplementary report. The title concerned the origin of peculiar carbonaceous structures in the Antarctic deposits. She searched the departmental archive. The report appeared in the catalogue. It was dated 1978. She downloaded it. The document was forty-seven pages long. Most of it concerned mineralogical analysis and the interpretation of ambiguous microscopic features.

The authors had concluded that there was insufficient evidence to regard the structures as biological. Lena found this entirely reasonable. She continued reading.

Near the end was a short section headed:

Possible biological affinities

It discussed the difficulty of distinguishing a mineral pseudomorph from the preserved remains of an organism lacking conventional cellular organisation. The authors considered several possible analogies with known biological systems. They rejected them. The final paragraph was cautious.

The material exhibits no consistent anatomical form, although certain structural relationships recur across specimens. It is possible that the apparent diversity of morphology reflects preservation rather than original biological variation. Alternatively, the material may represent an organisation for which presently recognised biological categories provide no satisfactory comparison.

Lena paused. It was remarkably close to her own assessment. That was not impossible. Different researchers examining the same material could reasonably arrive at similar conclusions. But it was unusual that the report had received so little attention. She checked its citation history. Several later papers mentioned it. One discussed mineral pseudomorphs. Another examined the preservation of complex organic structures. A third, published in 1991, proposed that the material might represent a previously unrecognised form of colonial organisation. Lena frowned. She knew the 1991 paper.

Or thought she did. She had cited it in an earlier review. She remembered it as a discussion of unusual Precambrian microfossils, not as a paper directly relevant to the Antarctic material. She opened the publication. The abstract mentioned the specimens. The introduction discussed the earlier drilling programme. The figures included photographs from RB-04/119. Lena looked at them. She had no recollection of those images. Yet the paper's title, authors and publication details were familiar. She checked her own review. The citation was there.

The surrounding paragraph referred explicitly to the Antarctic structures. She had apparently discussed them several years before beginning the current project. She stared at the page. Her memory said otherwise. But memory was unreliable. She knew that. There was no reason to assume an extraordinary explanation merely because she could not remember reading something. She closed the review. Then she opened her laboratory notebook from the beginning of the current project. The first entry concerning RB-17/642 was dated several months earlier. She remembered writing it.

The second described preliminary imaging. The third contained a note about possible mineral replacement. Then came a reference to the earlier Antarctic collection.

Compare RB-04/119. Similar network topology, uncertain biological affinity.

Lena stopped. She had not known about RB-04/119 until that afternoon. At least, she had believed she had not. The entry was in her handwriting. The date was six months earlier. She looked at it for a long time. Then she turned to the next page. There was a short note, written in the margin.

Do not assume common morphology implies common ancestry. Check whether conserved relations can arise independently.

That was perfectly sensible. It was exactly the sort of precaution she would have written. Indeed, she recognised the phrasing. But she could not remember writing it. She turned back to the earlier pages. Her research had apparently begun with knowledge of a specimen she had believed she was discovering for the first time. Lena sat motionless. Then she reached for her telephone. She called Yasmin. It was nearly ten. Yasmin answered after several rings.

"Lena?"

"Yes."

"Is something wrong?"

"I've found an earlier paper."

"Concerning the specimens?"

"Yes."

"Good. Does it clarify their origin?"

"Not especially."

"Then what's the problem?"

Lena looked at the notebook.

"I seem to have cited it before."

"That's not unusual."

"No."

"Then?"

"I don't remember doing so."

There was a pause. Yasmin said nothing. Lena continued.

"And I've found references to the earlier sample in my own notes."

"When were they written?"

"Six months ago."

"Could you simply have forgotten?"

"Of course."

"Then why are you calling?"

Lena looked at the page.

"Because the references are too specific."

"Explain."

"They mention the same structural relationships I've only just begun to investigate."

"Perhaps you investigated them earlier and forgot."

"Possibly."

"Or perhaps the earlier work influenced your current interpretation."

"Yes."

Yasmin waited. Lena took a breath.

"Do you remember what you said about records acquiring more coherent histories?"

"Yes."

"I think I understand why you found that disturbing."

There was a long silence. Then Yasmin asked, "Have you checked the original data?"

"Some of it."

"Any physical discrepancy?"

"Not yet."

"Then don't infer one."

"I'm not."

"Good."

Lena looked towards the specimen cabinet.

"But Yasmin, there's something else."

"What?"

"The earlier samples."

"Yes?"

"If their provenance is correct, the structures occur in deposits from two geographically separated locations."

"That would support a more extensive distribution."

"Potentially."

"And?"

"The more I examine the earlier literature, the more evidence I find that the distribution was already recognised."

Yasmin said nothing.

"Not recently. Decades ago."

"That could simply mean the work was overlooked."

"Yes."

"But?"

Lena hesitated.

"I've spent eleven days trying to establish whether this material belongs to an unfamiliar biological system. Now I'm discovering an increasingly extensive research history explaining why it might."

"That sounds familiar."

"Yes."

"Can you document the discrepancy?"

"Only imperfectly."

"Then preserve what you have."

"I intend to."

Lena ended the call. She returned to the specimen cabinet. The samples were still there. Their labels were old. Their histories were documented. Their apparent age was immense. They required no supernatural explanation. And yet she was beginning to experience the same difficulty Miriam had described. The more evidence she found, the less certain she became that the history she was uncovering had always been available to her.

The next morning, Lena received the preliminary results of an independent geochemical examination. The analyses did not establish a biological origin. They did, however, indicate that some of the carbonaceous material was closely associated with the original mineral fabric rather than being an obvious modern contaminant. Further tests were required. The result was modest. Scientifically, it justified continuing the investigation. Nothing more. Lena forwarded the report to Yasmin with a short note.

No evidence of modern contamination in the examined region. Biological interpretation remains uncertain.

Yasmin replied:

Thank you. Please continue without reference to the literary material.

Lena smiled. She intended to do precisely that. Then she opened the oldest geological description of the relevant Antarctic deposits. It was a report published shortly after the original drilling programme. The stratigraphic diagrams were crude by modern standards, but the observations were careful. She studied the layers surrounding RB-04/119. The sampled material appeared to be enclosed within a sequence consistent with the late Neoproterozoic dating model. If the interpretation was correct, the structures belonged to a world vastly older than any human civilisation.

Older than the familiar animal communities of the Cambrian. Older than most of the biological forms upon which humanity's intuitive understanding of life was based. The organism, if organism it was, would have existed in an Earth whose oceans, atmosphere and ecology bore little resemblance to the present. It would have required no human observer. No name. No myth. No ancient book. It would simply have existed. Lena looked at the microscopic images. She could not yet decide whether they represented life.

But she was becoming increasingly convinced that their organisation deserved a more serious explanation than the existing literature had provided. She enlarged the most complete reconstruction. Its connected compartments formed a structure without a recognisable body plan. There was no head. No centre. No obvious distinction between interior and exterior beyond the local arrangement of surfaces. No anatomical feature which could confidently be called permanent. The organisation was recognisable. The form was not. For a moment she remembered the fictional description Miriam had shown her.

The suggestion of temporary organs. The absence of fixed anatomy. The repeated reconfiguration of a common underlying structure. Then she dismissed the comparison. The resemblance was not evidence. She knew that. Nevertheless, one question remained. What would such a system have looked like when alive? Would it have possessed a definite shape? Would its boundaries have remained stable? Would it have been possible to distinguish one individual from another?

Or would its organisation have extended through the environment in ways which made the ordinary distinction between organism and surroundings difficult to maintain? Lena looked at the image. There was something profoundly uncomfortable about the possibility. Not because the material was monstrous. It was not. It was a few centimetres of dark rock containing microscopic structures of uncertain origin. But if those structures represented a living system, then life might once have existed in forms which ordinary biological classification could scarcely accommodate.

And the fossil record might contain traces of those systems without anyone recognising them. Not because the evidence was hidden. Because the categories used to interpret it were wrong. She made a note.

Consider whether morphological variability is fundamental rather than preservational.

Then she paused. The sentence was almost identical to a note in Haldane's report from 1986. She checked. It was not identical. But the conceptual resemblance was unmistakable. She wondered how many times the same observation had been made. How many researchers had noticed the problem. How many had dismissed it. And how much of the apparent continuity between their investigations was genuine.

Lena had sent the laboratory comparison to two colleagues before she gave it to the visitors. She marked both datasets origin unestablished and refused the technician's suggestion that the reconstructed compartments be labelled cells. The printed labels were removed from the slide tray. She wanted the rock to be allowed to remain rock.

That afternoon a co-author returned her draft with a comment on the first page: Please acknowledge the established Antarctic series and the neglected literature. We cannot present this as the first recognition of the organisation. A proposed replacement paragraph contained references to Haldane and to the earlier specimens, including RB-04/119. Lena recognised her colleague's courteous, slightly exasperated phrasing. It also appeared in an old review response she apparently had sent herself months before the present investigation.

She reopened the preserved original data. The measured structures were not more biological than they had been the day before. But the history of their examination had become inconveniently complete. The wording first recognition could no longer be defended. She removed it, marked the older sources as requiring independent verification, and kept the conclusion of uncertain origin.

"I won't give this an ancestry to make a good paper," she told the technician, who had asked why the title was changing again. "If there's an ancestry, the specimens must earn it."

That evening she stood alone by the sample cabinets with her hands in her coat pockets. The tray contained the same prepared sections. Their labels now described the comparative series she had believed herself to be assembling for the first time. She locked the cabinet and checked the key twice, an irritation with her own hands that had nothing to do with whether the structures were biological.

That evening, Miriam received Lena's preliminary report. It was twelve pages long and contained no literary allusions. In the photographs RB-17/642 lay under the microscope as it had for decades, a few centimetres of dark rock with pale chambers lodged in the mineral matrix. The chemical analysis had not established life. The structural comparisons were promising but did not exclude a recurring mineral process. Lena's conclusion was scrupulous: the origin remained uncertain, while the apparent continuity of organisation across unlike forms deserved further testing.

Miriam printed the report and carried it into the kitchen. Her hands trembled enough to blur the first image. She realised she had eaten nothing since breakfast. The refrigerator light illuminated a carton of milk she could not remember buying. Its receipt was among the papers on the worktop, dated the previous day. She put both away. An unexplained household lapse was not an anomaly in physical history; it was evidence that she was exhausted.

Back at the desk she laid Lena's images beside the reconstructed passage concerning the shoggoths. The resemblance remained too general to identify any organism. A mineral process could still account for the samples. Lovecraft's mutable creatures remained fictional as far as the evidence established. But the text's errors suggested a more demanding possibility.

The supposed creatures had been made, allegedly by a civilisation older than humanity. Their flexible structures were described as organs produced for a master's purposes. An unfamiliar natural organisation had been interpreted through the categories of anatomy, manufacture and obedience. If an actual observer had once made those mistakes, the history would require not merely the material he observed but the circumstances of his misunderstanding.

A correct description could be satisfied by a corresponding object. A mistaken description demanded an object, an observer and an explanation for the mistake. She marked the sentence in Lena's report: The observed relationships remain more stable than the morphological classifications applied to them.

It would have been comforting to find a monstrous creature in the rock. A creature had a body and a location. One could determine when it lived, how it reproduced, what conditions it required. The evidence instead concerned an organisation whose boundaries might not coincide with any recognisable individual. Even its possible biological nature remained unsettled.

Miriam returned to the list of dependencies. If an account of artificial servants required a history of manufacture, the next question would be whether the supposed makers had left evidence of their own. An error about natural origins might call an entire imagined civilisation into question, or furnish the world with circumstances under which the attribution had been made. She did not add the civilisation to Yasmin's model. She wrote the question on paper and left it unanswered.

At the laboratory Lena was still working. She placed the older Antarctic section beside the newer one, switched off the room lights and watched their pale networks appear under the instrument lamp. No coherent creature emerged. The samples remained rock. At midnight she found herself holding one of the prepared slides longer than necessary, feeling its weight through the cardboard mount. She had never previously been disturbed by a specimen that offered no clear result.

Miriam finally shut down her computer. The original question had been whether the things described in the Necronomicon would follow the book into existence. The accumulating evidence suggested a less theatrical formulation: if they possessed physical referents, their histories would lead away from humanity, into periods when no person existed to name them. Nothing would arrive. A geological sample would simply turn out to have been there all along.

Chapter VI 6 of 10

Things That Have Always Been There

Gabriel Sato disliked maps which looked more definite than the measurements beneath them. The ocean floor had supplied generations of amateur antiquarians with straight edges, convenient angles and cities created by interpolation software. A hydrographic chart was a record of soundings and assumptions, not a photograph. There was no particular reason to expect a hundred kilometres of seafloor to offer a tidy answer to a question framed by someone on dry land.

His old colleague Yasmin had sent him three documents: a 1927 hydrographic survey of the central Pacific, a 1974 geological interpretation, and a modern bathymetric reconstruction. She wanted to know whether the features were genuine and whether their published history was consistent.

He recognised the formation, provisionally called the Pelagic Complex: intersecting submarine ridges, enclosed depressions, terraces produced or modified by faulting and volcanic processes. The formation was real. Its architectural resemblance was beside the point.

What bothered him was the 1927 chart. Sparse soundings could establish the broad relief. They could not readily justify every contour drawn on the sheet, especially beyond the survey vessel's recorded tracks. Yet several of those contours corresponded to features resolved decades later.

Gabriel opened the original navigation tables. The figures were small, inked into columns whose orderly appearance concealed the rough weather, drifting instruments and human judgement needed to produce them. One sounding was annotated sea doubtful; a second had been struck through and restored in pencil. Such imperfections were more persuasive than a perfect diagram.

By lunchtime he had found older papers, each pointing to the same formation. One recalled a disagreement over whether the central depression was volcanic; another insisted the surrounding ridges belonged to separate tectonic episodes. The technical history was becoming richer as he searched it, but a few maps still appeared to know more than their instruments could have measured. He wrote to Yasmin without mentioning the resemblance to ruins. If she had been sent to him by another lost-city enthusiast, she could enjoy the disappointment without assistance.

Yasmin received his reply the following morning.

Dear Yasmin,

The formations appear to be genuine bathymetric features, although some of the older reconstructions are much less reliable than their illustrations suggest.

I've found several problems with the historical survey data.

In particular, certain early publications contain remarkably detailed descriptions of structures which appear to have been measured using methods incapable of resolving them.

This may reflect later redrafting, retrospective reinterpretation or errors in the attribution of the source data.

I wouldn't draw any unusual conclusion without examining the original survey records.

However, I agree that the documentary history deserves investigation.

Gabriel

Yasmin forwarded the message to Miriam. Miriam read it twice. Then she telephoned.

"Another one."

"Perhaps," Yasmin said.

"An historical record becoming more complete."

"Or a problem with the way published maps were compiled."

"Of course."

"But Gabriel has identified a discrepancy independently."

"Has he?"

"Yes. I gave him the references without mentioning the Necronomicon."

"Good."

"And he was already familiar with the general geological formation."

Miriam considered.

"Did he remember the earlier maps differently?"

"He says one of the published illustrations is more detailed than he recalls."

"That's not strong evidence."

"No."

"Could simply be mistaken memory."

"Entirely."

There was a pause. Miriam asked, "What exactly are the formations?"

"A complicated arrangement of submarine ridges and depressions."

"Natural?"

"Almost certainly."

"Then what is the relevance?"

Yasmin hesitated.

"Their history is becoming more coherent in much the same way as the book's."

"That's the interpretation."

"Yes."

"And the model?"

"It identifies the region as a possible physical dependency of one of the book's geographical claims."

Miriam said nothing. Yasmin continued.

"I haven't told Gabriel which claim."

"Good."

"But we should probably examine it ourselves."

"Yes."

"And Miriam?"

"What?"

"There's something unusual about the dimensions."

"In what sense?"

"The structures are much larger than I expected."

"How large?"

"Some of the individual features extend for tens of kilometres. The broader region under discussion is several hundred kilometres across."

Miriam stared at the wall.

"That's not a city."

"No."

"Not even a very large city."

"No."

"Then perhaps we shouldn't think of it as one."

"That," Yasmin said, "is precisely what I was hoping you'd say."

Oliver Stroud had never visited the Pacific. His professional life had been largely confined to libraries, archives, reading rooms and the occasional conference in a city with sufficiently interesting second-hand bookshops to justify the journey. He was therefore mildly surprised to discover that his investigation of an obscure medieval printing had begun to involve marine geology. He was less surprised to discover that the connection arose through an apparently insignificant bibliographical note. The note appeared in a seventeenth-century compilation of geographical curiosities.

It referred to a passage in the Latin Necronomicon describing a region of the ocean where the floor possessed an unusual arrangement of elevated and depressed surfaces. The original Latin was ambiguous. The compiler had translated the description as a reference to the remains of ancient buildings beneath the sea. A later commentator had interpreted it as evidence for a vanished island civilisation. A nineteenth-century antiquarian had dismissed the whole account as a fanciful elaboration of classical stories concerning Atlantis. None of those interpretations was especially remarkable.

Historical literature was filled with imaginary islands, submerged kingdoms and legendary cities. The difficulty lay in the reference's subsequent history. Oliver found it cited in an early twentieth-century geographical journal. The author mentioned the old account while discussing a newly reported submarine formation. The passage was brief. Its tone was sceptical. The writer considered the supposed correspondence between the medieval description and the hydrographic survey an interesting coincidence. That, in itself, would not have attracted Oliver's attention. But the journal article was dated 1908.

The survey it discussed had supposedly been conducted in 1927. Oliver checked the dates carefully. Then he checked the journal's publication history. The relevant issue had appeared in 1908. The article's title was recorded in the annual index. Copies were held in several libraries. The author had died in 1931. Oliver examined the reference again. The sentence was unambiguous. It discussed observations made during the 1927 survey. An event nineteen years in the future. He telephoned Miriam.

"Another retrospective citation."

"Where?"

"A geographical journal."

"Date?"

"1908."

"And the reference?"

"A hydrographic survey conducted in 1927."

Miriam sighed.

"Have you checked for later revisions?"

"Not yet. The copy I have is a scan."

"Then we need the physical volume."

"I've requested it."

"Good."

Oliver hesitated.

"There's something else."

"What?"

"The passage from the Necronomicon."

"Yes?"

"It refers to a structure beneath the Pacific."

"That's hardly surprising. Lovecraft's fiction makes extensive use of submerged cities."

"Yes. But this description is rather different."

"How?"

"It doesn't actually describe buildings."

Miriam sat forward.

"What does it describe?"

"An arrangement of elevations, depressions and boundaries."

"Could still be architectural."

"Certainly. But the original language doesn't require that interpretation."

"Then why did the compiler describe a city?"

"Because he apparently assumed that anything possessing a regular arrangement must have been constructed."

Miriam said nothing. Oliver continued.

"There's a phrase concerning the joining of surfaces in a place without ordinary directions."

"That sounds familiar."

"From the book's other passages?"

"Yes. Particularly those concerning spatial relations."

"Well, the compiler interpreted it as a description of strange architecture."

"And you?"

"I think he may have misunderstood the text."

"In what way?"

Oliver hesitated.

"I think he was attempting to describe a geometrical relationship using architectural vocabulary."

Miriam looked at her notes. A regular structure beneath the ocean. A supposed city. An interpretation imposed by a historical observer who lacked the conceptual framework to understand what he was describing. The same problem had appeared in Alhazred's account of the shoggoths. The text's errors were not random. They appeared to follow from attempts to understand unfamiliar phenomena through familiar categories. A structure became a city. A complex organism became an artificial servant. An unusual relationship between physical events became a magical operation.

The Necronomicon might be acquiring a history in which its extravagant descriptions were not literally true. They might instead become intelligible mistakes about real things. And those real things could be considerably less accommodating to human intuition than the monsters of fiction.

"Send me the original Latin," Miriam said.

"I will."

"And Oliver?"

"Yes?"

"Don't translate it as architecture unless the vocabulary requires it."

"I haven't."

"Good."

She ended the call. Then she opened the reconstruction. The relevant passage belonged to a section concerning places where the ordinary relationships between direction, distance and location were said to become unreliable. Miriam had treated it as part of Lovecraft's fictional geography. Now she wondered whether that had been too simple. Perhaps the supposed city was not a city. Perhaps the buildings were not buildings. Perhaps the whole description was an attempt to express some unfamiliar physical structure through the metaphor of architecture.

A metaphor which subsequent readers had mistaken for a literal account. That possibility was not reassuring. It merely substituted a poorly understood natural phenomenon for an imaginary archaeological site. And the natural phenomenon might be much more consequential. Gabriel Sato joined the investigation reluctantly. He agreed to participate only after Yasmin assured him that she was not attempting to establish the existence of a lost civilisation.

"The problem," he explained during their first video conference, "is that people see regularity and immediately infer construction."

Miriam nodded.

"A familiar error."

"Quite. Geological systems can produce extremely regular patterns. Columnar jointing, fault systems, sedimentary layering, volcanic structures. Geometry isn't evidence of intelligence."

"Agreed."

"And the ocean floor is particularly vulnerable to imaginative interpretation because the data are often incomplete."

Yasmin leaned forward.

"Could you explain the survey problem?"

"Certainly."

Sato shared his screen. A bathymetric map appeared.

"Most people imagine that we possess something resembling a photographic image of the entire ocean floor. We don't. Detailed measurements are unevenly distributed."

He indicated several regions.

"Here we have direct depth measurements from ships. Elsewhere we have broader estimates derived from satellite observations of the sea surface and gravitational field. These methods operate at different resolutions and have different uncertainties."

Miriam examined the map.

"Some of the structures may be artefacts?"

"Certainly. Interpolation can produce features which look more definite than the underlying data justify."

"And the historical maps?"

"That's where it becomes interesting."

He opened a second image.

"This is the 1927 survey."

The map showed a series of ridges surrounding a broad depression.

"At the time, the available depth measurements were sparse. The survey could reasonably have identified the general relief, but not every detail represented here."

He opened another map.

"This is a modern reconstruction based on later measurements."

It contained a much more elaborate arrangement of terraces and ridges.

"Notice the similarities."

Miriam looked at the two.

"They're rather striking."

"Yes. Some can be explained by the earlier survey having correctly identified large-scale features. But the agreement extends into regions where the original data coverage appears inadequate."

"Could the old map have been redrawn later?"

"That's the first possibility."

"Have you checked?"

"Not completely. But I've located reproductions published before the modern surveys."

"And they contain the detail?"

"Some do."

Yasmin looked at the screen.

"Could the early surveyors have possessed more measurements than the surviving records indicate?"

"Certainly."

"Then the discrepancy might be documentary rather than physical."

"Exactly."

Sato changed the image.

"This is a reconstruction derived from seismic profiles."

The map now showed a series of broad, intersecting structures beneath the sediment cover.

"These features are not buildings. They're associated with deformation of the oceanic crust and overlying sediments."

"Could they be ancient?"

"Certainly. Some structures developed over tens of millions of years."

"Could they have formed before the present ocean basin?"

"Not in their current physical location without a very complicated tectonic history. Oceanic crust is created and destroyed. We must be careful about assigning impossible ages to structures simply because they resemble something old."

Miriam approved of the correction.

"Then what troubles you?"

Sato hesitated.

"The organisation."

"In what sense?"

"The structures don't form a straightforward volcanic or tectonic pattern. Several plausible geological processes are involved, but their arrangement is unusually correlated."

"Correlated how?"

He opened a simplified diagram.

"Imagine a large region containing ridges, depressions and fault-bounded blocks. In an ordinary geological system, their arrangement reflects local stresses, material properties and the history of deformation."

"Yes."

"Here, certain relationships persist across regions which should have experienced substantially different local conditions."

Yasmin looked interested.

"Which relationships?"

"Relative orientation, spacing ratios and some aspects of connectivity. Not perfectly. But sufficiently often that several researchers have proposed a common underlying structural control."

"Have they identified one?"

"No satisfactory model accounts for everything."

Miriam examined the diagram. It reminded her of Lena's biological reconstructions. The shapes differed. The relationships persisted. She looked at Yasmin. Yasmin had noticed the same thing. Sato continued.

"One hypothesis invokes inherited weaknesses in the crust. Another proposes a complicated sequence of tectonic events. A third treats part of the apparent regularity as an artefact of the reconstruction."

"All plausible?"

"To varying degrees."

"And which do you favour?"

"I don't know."

He looked directly at the camera.

"That isn't a confession of geological incompetence. It's the appropriate conclusion when the evidence is incomplete."

Miriam smiled.

"I couldn't agree more."

They asked Sato whether the formation could be described as artificial. He laughed.

"No."

"Not even provisionally?" Oliver asked.

"No. There is no credible evidence of construction."

"What about the regularity?"

"Regularity isn't sufficient."

"The scale?"

"Scale tells us nothing about intention."

"The apparent terraces?"

"Natural processes produce terraces."

"And the intersecting boundaries?"

"Faulting can produce intersecting boundaries."

Oliver nodded.

"Then the architectural interpretation is unjustified."

"Entirely."

Miriam looked at her notes.

"Would you say the underlying formation is nonetheless unusual?"

"Yes."

"Unusual enough to require an unfamiliar physical process?"

"Not necessarily. An unusual outcome can arise from familiar processes interacting in complicated ways."

"Quite."

Sato paused.

"But the historical data do contain something genuinely peculiar."

"What?"

"The apparent precision of the early descriptions."

"Could that be a consequence of later redrafting?"

"Perhaps. But some of the surviving records appear to predate the relevant measurements."

Miriam looked at him.

"Are you certain?"

"No. That's why I'm investigating."

He opened another image. It showed a page from an early twentieth-century hydrographic publication. The page contained a diagram of several submerged features. A caption identified them as part of the Pelagic Complex.

"The diagram is dated 1908."

Yasmin frowned.

"The same year as Oliver's geographical article."

"Yes."

"But the relevant survey wasn't conducted until 1927."

"Correct."

"Could the diagram derive from an earlier survey?"

"That's possible. There were sparse soundings and route observations. But the degree of agreement with the later reconstruction is difficult to account for from the available records."

Miriam examined the page. There was something almost malicious in its ordinariness. The diagram was not an occult illustration. It was not accompanied by warnings, prophecies or claims of forbidden knowledge. It was a technical representation of the ocean floor. A handful of lines, measurements and annotations. Yet it appeared to contain information obtained decades later. The same kind of inconsistency had appeared in the conservation reports concerning the Necronomicon. It had appeared in the Paris manuscript records. It had appeared in Lena's laboratory archive.

And now it appeared in the history of marine surveying. Different disciplines. Different institutions. Different physical objects. The common feature was not the subject matter. It was the increasingly complete relationship between present evidence and an apparently revised past. Miriam thought of Yasmin's model. A local set of constraints requiring additional historical support. A structure becoming more coherent through the incorporation of previously unsupported events. The process, if it existed, appeared to be spreading beyond anything which could reasonably be described as the history of a book.

The next records were less useful than Gabriel hoped. A chart which apparently anticipated a modern survey proved to be a later compilation. Another inconsistency resulted from an incorrectly labelled plate. These explanations strengthened his confidence that the ordinary method could still identify errors.

The 1908 drawing remained troublesome. Its level of detail was unsupported by the surviving soundings, yet the old sheet belonged to the expected publication. Gabriel sealed the comparison of measured and inferred contours in his own data archive, with the uncorrected observations attached. He would not allow the image to acquire sharper edges through another round of interpolation. For him the unresolved part was not an ancient city but a limitation of the evidence by which scientists claimed to know the shape of the sea.

It was Lena who noticed the resemblance.

She was visiting Cambridge to discuss the Antarctic material with Yasmin when Gabriel Sato's latest reconstruction appeared on the screen. The diagram showed the connectivity relationships between the principal features of the Pelagic Complex. Lena stopped speaking.

"Where did that come from?"

Yasmin looked up.

"The Pacific survey."

"May I see it?"

"Of course."

Lena examined the diagram. It consisted of a network of connected regions, stripped of their physical dimensions and geographical coordinates. There were no mountains. No valleys. No ocean. Only relationships between features.

"That's odd," Lena said.

"In what way?"

"It resembles one of our structural models."

Yasmin frowned.

"From the Antarctic material?"

"Yes."

Miriam looked towards the screen. The resemblance was not immediately obvious. The geological network contained far more nodes than the biological one. Its physical scale was incomparably greater. Its underlying processes, so far as anyone knew, had nothing in common with biological development. Lena opened her laptop. She located a simplified graph derived from RB-17/642. Then she placed it beside the oceanic reconstruction.

"Look at the central relationships."

Yasmin leaned forward.

"They're similar."

"Yes."

"How similar?"

"I don't know."

"Then we shouldn't attach any significance to it."

"Agreed."

But Lena continued examining the diagrams. The problem was not that the shapes were identical. They weren't. It was that certain patterns of connectivity recurred. A central region connected to several subsidiary regions through intermediate structures. Some paths formed loops. Others terminated. A small number of nodes occupied positions through which disproportionately many connections passed. Such features were common in networks. Transport systems, river basins, biological tissues and human infrastructure could all display superficially similar arrangements. The resemblance might mean nothing. Nevertheless, Yasmin decided to test it.

She selected a series of comparison networks drawn from unrelated natural systems. They included geological fracture patterns, biological transport networks and several artificially generated graphs. She then compared the Antarctic and Pacific structures using a measure designed to identify topological similarity while accounting for differences in size and sampling. The result was not spectacular. The two networks were more similar than some controls and less similar than others. There was no reason to conclude that they possessed a common origin. Lena accepted this. So did Miriam.

But Yasmin noticed a different problem. The biological and geological models had been constructed using different methods. Their apparent similarity was sensitive to how the underlying features were defined. A particular choice of segmentation could make the networks resemble one another more closely. Another choice weakened the resemblance. This was an ordinary analytical hazard. But it raised a question. What determined the boundaries of the relevant objects? In the Antarctic sample, the investigator had to decide which compartments belonged to a single structure.

In the Pacific survey, the investigator had to decide which ridges and depressions belonged to the Pelagic Complex. Neither boundary was given unambiguously by nature. Both depended partly upon the model used to interpret the evidence. Yasmin looked at the two diagrams.

"We may be comparing artefacts of classification."

"Very possibly," Lena said.

"But there's a more general question."

Miriam looked at her.

"Which?"

"What if the physical structures are not the fundamental objects?"

Lena frowned.

"Explain."

"We're identifying objects by their boundaries. An organism here, a geological formation there."

"Yes."

"Then we're comparing their internal organisation."

"Yes."

"But perhaps the organisation is the more stable feature, and the apparent object is merely one physical manifestation of it."

Lena considered this.

"In biology that can sometimes be a useful perspective. But it doesn't establish a common phenomenon across unrelated systems."

"I know."

"Then what's the relevance?"

Yasmin looked towards the graph.

"The seventh division."

Miriam felt a sudden chill. The disputed passage.

What persists is not that which feels, but the order by which feeling is possible.

The supposed distinction between a physical object and the relations through which its organisation was constituted. The argument that a pattern might remain identifiable despite changes in material form. She had assumed the text concerned the persistence of consciousness. Perhaps it did. But the broader principle was not restricted to minds. A structure might possess relationships which could be instantiated in different physical systems. Biological tissue. Geological formations. Perhaps even the arrangement of events in spacetime. Lena looked at Miriam.

"You're thinking about that ridiculous book again."

"Yes."

"Then don't."

Miriam smiled faintly.

"I'll endeavour to restrain myself."

Lena returned to the diagrams.

"These are very different natural systems. Any resemblance could be coincidental or produced by our analytical methods."

"Agreed."

"And even if the same mathematical pattern appeared in both, that wouldn't imply a common cause."

"Of course."

Yasmin said nothing. She was already examining another possibility. The model she had constructed concerned the relationships necessary to maintain a coherent history. It treated physical objects as nodes within a wider network of constraints. But perhaps the assumption that the nodes represented stable, independently existing objects was too simple. Perhaps some of the apparent objects were manifestations of structures defined at a more general level. That possibility was mathematically respectable. Its application to their evidence remained entirely speculative.

Nevertheless, it suggested that the distinction between a biological system and a geological formation might not be the most useful one for investigating the phenomenon. The question was not whether the two structures were physically similar. They plainly weren't. The question was whether they participated in a common relational structure which their separate disciplines had failed to recognise. And if so, what else might participate in it? Oliver's discovery came two days later. It concerned a marginal annotation in a seventeenth-century copy of a geographical commentary.

The note referred to the submerged structures of the Pacific. It described them as the remnants of a place whose original name had been lost. A later hand supplied a name.

R'lyeh.

Oliver had initially dismissed the annotation as an obvious modern addition. The name belonged to Lovecraft's fiction. Its presence in an apparently seventeenth-century volume was precisely the sort of thing one might expect from a literary hoax. He therefore arranged for the relevant page to be examined. The ink appeared old. The handwriting was consistent with other annotations in the volume. There was no obvious evidence of modern insertion. None of this established that the note was genuinely seventeenth-century. But it complicated the simplest explanation. Oliver contacted Miriam.

"I've found the name."

"Which name?"

"R'lyeh."

There was a long silence.

"Where?"

"In a geographical commentary."

"Date?"

"Seventeenth century, apparently."

"Apparently?"

"The annotation may be later."

"Then treat it as later until you have evidence otherwise."

"Of course."

Miriam looked at her notes.

"Does the passage describe a city?"

"Not exactly."

"What does it describe?"

"A place beneath the sea."

"That could mean almost anything."

"Yes. But the wording is peculiar."

"How?"

"It says the place is not reached by travelling towards it."

Miriam frowned.

"What?"

"The Latin is difficult. The commentator appears to distinguish between a location which can be approached through ordinary movement and one which can be occupied only under particular spatial conditions."

"That sounds like bad metaphysics."

"Quite possibly."

"Or an attempt to explain an obscure geographical tradition."

"Yes."

"What else?"

Oliver hesitated.

"There's a reference to the relationship between distance and position."

"Go on."

"The writer seems to suggest that two places may be separated by an immense distance while nevertheless sharing some more fundamental relation of location."

Miriam thought of the seventh division. The passage concerning things which existed outside ordinary temporal succession. The damaged text about the constitution of forms. The marginal warning that certain entities were not invoked. She looked towards Yasmin's model.

"Send me the passage."

"I will."

"And Oliver?"

"Yes?"

"Don't tell the others about the name yet."

"Why?"

"Because I want Gabriel to finish his analysis without knowing what we're looking for."

"Understood."

She ended the call. Then she sat for several minutes, considering the implications. R'lyeh. The name carried an enormous weight of literary association. A sunken city. Impossible architecture. A sleeping entity. The possibility that familiar geometrical relationships might fail in its vicinity. But those associations were precisely what made it dangerous as evidence. The name could contaminate every subsequent interpretation. A geologist who knew the supposed connection might unconsciously select features resembling architecture. A physicist might search for geometric anomalies.

A historian might interpret ambiguous passages as references to Lovecraft's mythology. The investigators would find precisely what they had been primed to expect. Miriam had no intention of allowing that. She opened a separate document.

Restricted literary correspondences

She entered the name, source and provisional interpretation. Then she closed the file. The scientific investigation would continue without it. For now, the name was nothing more than a suspicious historical annotation. And one of the oldest traps in scholarship was the tendency to mistake a recognisable name for a genuine explanation. The following Monday, Gabriel sent Yasmin the results of his examination of the earliest hydrographic records. They were disappointing in precisely the way serious research often was. Several supposed anomalies had ordinary explanations.

Some of the early depth measurements had been incorrectly assigned to the relevant survey. A few detailed maps were based on later compilations. One apparent discrepancy resulted from a poorly documented revision to a printed chart. But the central problem remained. The oldest surviving representation of the Pelagic Complex contained relationships between submarine features which were not adequately supported by the measurements attributed to it. Sato had examined the original survey records. The relevant soundings were sparse.

The map nonetheless represented several structures whose existence had only been confirmed through much later work. He was reluctant to call this impossible. There might be missing records. There might be undocumented measurements. The surviving materials might represent a revised compilation whose production history had been misunderstood. All these explanations remained open. But the anomaly was real enough to justify further investigation. Yasmin added the result to the model. Then she examined the wider documentary network. The same pattern was now visible across several disciplines.

The Necronomicon possessed an increasingly complete bibliographical history. The Antarctic material possessed an increasingly extensive geological and scientific history. The Pelagic Complex possessed an increasingly detailed record of observation. And in each case, the most troubling inconsistencies concerned the relationship between the date of a record and the information it appeared to contain. The evidence did not establish retrospective alteration. But it was becoming more difficult to dismiss the concentration of anomalies as coincidental. Yasmin constructed a new graph.

This time she did not use the Necronomicon as the central node. Instead, she represented the three investigations independently. The textual history. The biological material. The submarine formations. She then added the known relationships between them. The result was sparse. There was no direct physical connection between Antarctic sediment and Pacific bathymetry. No demonstrated common mechanism. No evidence that the biological structures had anything to do with the formation of the ocean floor. The apparent similarities remained suggestive at most.

But the documentary dependencies converged upon a common set of claims. Claims concerning the distinction between stable objects and persistent relationships. Claims concerning the possibility of multiple physical manifestations of a single organisation. Claims concerning the dependence of historical existence upon a larger structure of consistency. Yasmin looked at the graph. She had been treating the phenomenon as a process by which missing objects acquired histories. Now she wondered whether that description was still adequate. Perhaps the process was not primarily concerned with objects.

Perhaps it was concerned with relationships. And objects, along with their histories, were being supplied where necessary to realise those relationships. The distinction was subtle. Its implications were not. An object could be imagined as a thing appearing in the world. A relationship might require changes to the structure of the world itself. Yasmin stared at the screen. Then she opened the mathematical model. She removed the assumption that every node represented a fixed physical entity. She allowed some nodes to represent relationships whose realisations could vary.

The model became more flexible. It could accommodate the same abstract structure through different physical arrangements. A particular relationship might be represented by one set of objects in one history and a different set in another. The change did not prove anything. It was a mathematical experiment. But it produced a result Yasmin found deeply unpleasant. The number of compatible histories increased dramatically. Some differed only in minor documentary details. Others required substantially different physical objects.

A few differed in the number and identity of the entities supposedly involved. Yet the underlying relational structure remained unchanged. Yasmin examined the solutions. If the model possessed any correspondence with reality, the distinction between an entity and a relationship might be less fundamental than she had assumed. An apparent object could be one local manifestation of a larger organisation. Its physical form might depend upon the conditions under which the organisation became realised. That would explain why the biological material resisted a fixed anatomy.

It might explain why the submarine formations displayed persistent structural relationships across different geological regions. And it might explain why certain passages of the Necronomicon appeared to discuss entities which could not be assigned ordinary positions in space and time. Yasmin stopped. She had crossed a line. Not from scepticism into belief. But from investigating anomalous historical evidence into constructing a model in which the universe's ordinary inventory of objects was no longer necessarily fundamental. She looked at the framed notice on her wall.

What observation would demonstrate that you are wrong?

For several minutes she could think of none which would distinguish the enlarged model from a sufficiently flexible collection of ordinary explanations. That was a serious problem. A model capable of accommodating anything explained very little. She reduced its freedom again. Then she began looking for a relationship which would produce a specific, risky prediction. Something not already contained in the evidence. Something which could fail. It was the only way forward. Miriam received Yasmin's preliminary findings that evening. She read them carefully.

The mathematical details were beyond her professional expertise, but the general argument was comprehensible. The evidence might be organised more naturally around persistent relationships than around fixed physical objects. The distinction reminded her of the seventh division. She opened the Paris manuscript transcription. There was a passage she had previously considered largely philosophical. It concerned the relationship between the names of things and the structures by which those things could be identified. The Latin was difficult. Miriam translated it slowly.

The author appeared to distinguish between a name assigned to a visible object and a name assigned to the relations which made the object possible. The first kind of name depended upon an observer recognising an individual thing. The second referred to something which might possess multiple manifestations. Miriam considered the argument. It could be interpreted as a discussion of universals. A familiar medieval problem. But the context suggested something stranger. The text appeared to claim that some entities were not properly described as objects at all.

Their apparent forms were temporary expressions of relationships which did not depend upon those particular forms. She remembered the shoggoths. The supposition that their bodies were flexible while their underlying organisation persisted. She remembered the Pacific formations. The assumption that their regularity implied architecture. The errors were similar. In both cases, an observer had mistaken a local manifestation for a stable object. Perhaps the book's supposed author had encountered phenomena for which ordinary physical categories were inadequate.

Perhaps he had interpreted them in the only language available to him. Creatures. Cities. Spirits. Gods. The names belonged to familiar human categories. But the things being named might not. Miriam looked at the passage again. There was one sentence she had overlooked. It concerned the danger of assigning familiar names to unfamiliar relations. The wording was awkward. She translated it tentatively.

The ignorant call them beings because they encounter them as things. The learned err more greatly, for they suppose that the things exhaust what is.

Miriam stared at the sentence. It was an unusually sophisticated distinction. The ignorant encountered an unfamiliar phenomenon and called it a being. The learned constructed an elaborate account of the being's properties. Both might be wrong. Not because the phenomenon was imaginary. But because the category of a being was itself inadequate. She read the passage again. Then she returned to the reconstruction. The essay had treated Lovecraft's cosmic entities as things possessing properties beyond ordinary human experience. That was a reasonable literary interpretation.

But perhaps even that was too anthropomorphic. Perhaps the entities were not creatures occupying the universe. Perhaps they were organisations which could be manifested through the universe. The distinction would explain why their supposed locations and forms were so inconsistent. It might also explain why some of the descriptions concerned space and time themselves rather than objects within them. Miriam felt a slow, unpleasant recognition. The book might not be acquiring a pantheon of monsters. It might be acquiring a cosmology. And the difference was considerable.

A monster could exist in the world without changing the world's fundamental structure. A cosmology could not.

The next morning, Oliver sent her the full transcription of the seventeenth-century geographical commentary. Miriam read it before breakfast. The passage concerning the submerged region was short. It mentioned a place beneath the ocean, described through a sequence of spatial relationships rather than a conventional geographical position. The later annotation supplied the name R'lyeh. The original text did not. That was important. The name might be a modern literary interpolation. The underlying description could belong to an entirely different tradition. Miriam examined the Latin.

The language was not particularly elegant. Several terms appeared to have been translated from an earlier source whose vocabulary the compiler had only imperfectly understood. A phrase concerning the conjunction of distant places had been rendered as a reference to walls meeting at impossible angles. Another expression described a region whose extent could not be determined by ordinary measurement. The compiler interpreted this as a statement about the enormous size of the supposed city. But Miriam was not convinced.

The underlying terminology could refer to a failure of the measuring system itself. A third passage concerned the relationship between the place and those who approached it. It appeared to suggest that the location was not simply reached by travelling a particular distance. Instead, the conditions under which it became accessible were described through relationships between events. Miriam stopped reading. That was not geography. At least, not geography as normally understood.

It was an attempt to describe spatial relations without assuming that distance and direction were sufficient to define them. She opened Yasmin's latest memorandum. The mathematical model had begun exploring precisely that distinction. The relationship between physical objects and the structure of spacetime in which their locations were defined. Miriam considered the coincidence. It could be interpretative contamination. She had read Yasmin's memorandum before examining the Latin. She might be imposing its concepts upon an ambiguous text. That was a serious possibility.

She therefore prepared a deliberately conservative translation. She avoided supplying modern mathematical terminology. She retained the ambiguities. She included alternative readings where the Latin permitted them. Then she sent the result to Oliver and asked him to examine it independently. She did not mention Yasmin's model. Oliver replied several hours later. His translation differed in several particulars. But the central difficulty remained. The passage did not describe an ordinary geographical location. It described a set of spatial relationships which the compiler had apparently struggled to express.

Miriam forwarded both translations to Yasmin. The reply was brief.

We need to stop treating the Pacific formation as a map of an object and start treating it as evidence about the geometry of the region.

Miriam read the sentence twice. Then she called.

"What do you mean by geometry?"

"Not architecture," Yasmin said.

"I understand that."

"Nor merely the shape of the ocean floor."

"Then what?"

"The relationships between measured distances, positions and paths."

Miriam frowned.

"Those are ordinary geographical properties."

"Usually."

"Usually?"

Yasmin hesitated.

"Suppose two locations are separated by a particular distance according to one set of measurements, but another independent method gives a systematically different result."

"Measurement error."

"Probably."

"Or an incorrect model."

"Yes."

"And?"

"If the discrepancy persists after those explanations are eliminated, it could indicate that our assumed geometrical framework is inadequate."

Miriam looked towards the window.

"You're suggesting that the Pacific formation might contain evidence of non-standard spatial geometry."

"I'm suggesting that we should test whether the measurements are mutually consistent under the ordinary geometrical model."

"Which is a considerably more defensible proposition."

"Exactly."

"And the historical passage?"

"May be irrelevant."

"Naturally."

"But it provides a reason to ask a question which the geological descriptions alone might not have suggested."

Miriam considered.

"Isn't that dangerously close to allowing the book to direct the investigation?"

"Yes."

"Then why do it?"

"Because a hypothesis isn't invalid merely because its source is peculiar. What matters is whether the prediction is precise, independently testable and capable of failing."

Miriam smiled faintly.

"Very well."

"But there's another concern."

"What?"

"If the geometry itself is part of what we're investigating, we need to be careful about defining our measurements in terms which already assume the conclusion."

"Explain."

"Distance isn't a single elementary observation. Different methods infer it through physical models. We need to identify what each instrument actually measures."

"And compare independent methods."

"Exactly."

"Could Gabriel help?"

"Yes. But I want to formulate the test before we tell him what we're looking for."

Miriam looked at the translation. A place not reached by travelling towards it. A location whose extent could not be determined by ordinary measurement. The joining of distant places. These were familiar motifs in supernatural fiction. They were also, in sufficiently abstract form, questions about the mathematical structure of space. Miriam wondered whether the ancient text was describing an unfamiliar geometry through the only metaphors available to its author. Or whether Yasmin, armed with modern mathematics, was simply discovering new ways to misunderstand an old piece of nonsense.

Both possibilities remained open. She intended to keep them that way until the evidence justified a conclusion. The first measurement inconsistency appeared in a survey that had nothing to do with the Pelagic Complex. Gabriel had been comparing satellite-derived bathymetric estimates with shipborne measurements across a wider region, using the surrounding ocean floor as a control. Most discrepancies followed familiar patterns. The satellite estimates were less precise. Some regions had poor data coverage. Other differences reflected the limitations of the gravitational models.

There was nothing surprising about any of this. But a small set of observations resisted straightforward reconciliation. They concerned the relative positions of several submarine features near the edge of the Pelagic Complex. The differences were not enormous. They were within ranges which could plausibly reflect errors in navigation, sound-speed correction or data processing. Sato was not alarmed. He was, however, dissatisfied. The discrepancies appeared correlated. When one feature's position was adjusted to improve agreement with a particular survey, the fit to another survey became worse.

A different correction improved the second agreement but degraded the first. This could result from systematic errors. Indeed, that was the most likely explanation. Sato began checking the navigation records. Then the instrument calibrations. Then the processing history. The records were unusually extensive. Several surveys contained independent position estimates derived from different methods. Some of the older data were less reliable, but the discrepancies persisted in a subset of modern measurements. Sato prepared a memorandum.

Preliminary assessment of survey inconsistencies

The present discrepancies are not evidence of anomalous geometry. Their magnitudes remain compatible with several conventional error sources.

Nevertheless, the residuals display a structured pattern which is not adequately explained by the simple correction models examined so far.

Further independent measurements would be required before any physical interpretation could be justified.

He sent the memorandum to Yasmin. Then he returned to his ordinary work. There was no reason to make more of it. Yet Yasmin recognised the significance immediately. The discrepancy was precisely the kind of observation she had intended to investigate. It was small. It was uncertain. It possessed several plausible conventional explanations. But it concerned the consistency of spatial measurements rather than merely the interpretation of an unusual geological structure.

For the first time, the investigation had produced a possible connection between the book's strange spatial descriptions and a modern physical measurement which could, in principle, be tested. Yasmin opened her model. She entered the discrepancy as a provisional observation. Then she began constructing a test based on independent measurement methods. The ordinary explanation remained overwhelmingly preferable. But now there was something concrete to investigate. Not a monster. Not a lost city. A disagreement between measurements. An almost insultingly small discrepancy.

And perhaps, if the impossible hypothesis survived, the first evidence that the structure of space itself was less secure than anyone had supposed. Miriam received the memorandum late that evening. She read it carefully. The technical details were unfamiliar, but Yasmin had attached a short explanation. The measurements could not yet be reconciled under the simplest models. Ordinary instrumental or processing errors remained likely. Further tests were needed. That was all. Miriam found the modesty of the result reassuring. Then she opened the historical material. The submerged formation.

The supposed city. The name R'lyeh. The descriptions of impossible architecture. The suggestion that a place might be defined by relationships other than ordinary distance and direction. She thought of Lovecraft's fiction. The sleeping entity beneath the Pacific. The ancient city whose geometry violated familiar expectations. The idea that certain arrangements of space and time might permit contact with something otherwise inaccessible. Within the fiction, these things were expressed as horror. But stripped of their literary imagery, they became a different kind of question.

Could the universe contain geometrical relationships which human beings had not yet recognised? Of course. Modern physics had already taught that ordinary Euclidean intuition was not universally adequate. Could an unfamiliar geometrical structure produce observations which were misinterpreted as impossible architecture? Possibly. Could such a structure possess relationships to locations and events which ordinary geographical descriptions failed to capture? That remained speculative. But it was not necessarily a supernatural question. Miriam looked at the name in the seventeenth-century annotation.

R'lyeh.

She had expected the word to frighten her. Instead, she found it increasingly irritating. It was too familiar. Too burdened with literary associations. The name encouraged precisely the wrong kind of thinking. It suggested a place. A city. An object possessing a definite location beneath the sea. Something which might eventually rise from the depths and confront humanity. But the evidence was beginning to suggest that the underlying phenomenon, if there was one, was not a place in that ordinary sense. It might be a relationship.

A feature of the geometry through which places were defined. And if that were true, the entire idea of something rising from the ocean would be a misunderstanding. A spatial structure would not need to travel. It would not need to awaken. It would not need to emerge. It might already be part of the relations through which the world possessed its familiar dimensions. Miriam closed the commentary. She thought again of Alhazred's errors. The shoggoths had been mistaken for artificial servants.

The submarine formation had been mistaken for a city. And perhaps the entities associated with the strange geometry had been mistaken for inhabitants of that city. Each interpretation imposed a familiar human category upon something which might possess no corresponding feature. The more the historical record became coherent, the less literal the old descriptions seemed. And the less literal they became, the more disturbing their possible physical referents appeared. A monster was comprehensible. A city was comprehensible.

Even a god, understood as a tremendously powerful mind, remained within the conceptual range of human experience. But a structure which altered the relationships between locations, events and physical existence was something else entirely. It could not be negotiated with. It could not be commanded. It might not even be capable of recognising the distinction between a living organism and an inanimate object. It would simply be part of the world's architecture.

And if the world's architecture was changing, there was no reason to assume that anything occupying it would remain unchanged.

The following morning Oliver sent a corrected transcription of the seventeenth-century commentary. He had located an earlier manuscript version without the name R'lyeh. The word had been added by a later reader, and its date remained uncertain. He was relieved: at least that particular correspondence could be put aside.

The older text contained something less easily dismissed. A marginal gloss described the submerged region as not among the places of the world, but among the conditions by which the world possesses places.

Miriam set the two translations on her desk, the Latin beside Gabriel's map. The old writer had apparently struggled to express a relationship between positions using the only available analogy: a place, a city, walls meeting at impossible angles. Gabriel's report described a quite different difficulty. Some routes through the measurement network gave residuals which resisted the simplest corrections. Miriam telephoned Yasmin. "I think we should stop asking where the city is."

"There is no city," Yasmin said.

"Exactly. The text may be misdescribing a relationship, not a location."

Yasmin was quiet for a moment. "I've asked Gabriel for the raw measurements. If the anomaly concerns the assumed geometry, the residuals should change when we choose different measurement loops."

"And if they don't?"

"Then the model loses. Which would be a pleasant result."

Miriam put down the telephone. In the study the door stood open onto the hallway. She knew the distance to the stairs without having measured it for years. The assurance was so ordinary that she had never understood it as an empirical expectation.

She closed the door. The latch settled into place with a small, conclusive click. Neither the old text nor the new measurements justified anything dramatic. But for the first time she had begun to question not what occupied a place, but the physical relationships through which a place was constituted.

Chapter VII 7 of 10

The Geometry of the Gate

Yasmin crossed out the word gate before Miriam had finished taking off her coat.

"It prejudges what we're measuring."

Miriam examined the title of her memorandum. The Geometry of the Gate. "It's the text's term, not mine."

"That's worse."

Rain blurred the Cambridge windows. Miriam had brought Oliver's Latin transcriptions and the commentary in which a supposed entity was associated with connections between remote places. Yasmin had brought a residual plot from Gabriel's survey, a less impressive object but one she trusted more.

Miriam showed her the passage naming Yog-Sothoth. Its grammar did not securely distinguish an entity from the relation through which it became perceptible. A later reader had rendered the relation as a gate. The word carried all the wrong associations: something standing between two established places, capable of being opened by someone who wished to pass.

"Can the Latin mean that the relation is the thing?" Yasmin asked.

"It can. It may simply be clumsy metaphysics."

"Good. We won't build a cosmology on the adjective 'clumsy'."

She moved Miriam's manuscript to one side and placed Gabriel's measurements on the desk. Three survey regions, labelled A, B and C, had individually satisfactory pairwise relations but left a structured residual when combined in a closed network. There were many likely sources of error, some already corrected. The question was whether the residual followed an instrument, an environmental influence or the arrangement of the measurements themselves. Miriam touched the edge of the map. "The old writers called this impossible architecture."

"They were entitled to be wrong in Latin. Gabriel's instruments will have to be wrong in numbers."

The refusal made Miriam smile. She had not come to ask whether Yog-Sothoth existed. She had come to see whether a specific measurement, undertaken without knowledge of that name, could discriminate between an ordinary error and an inadequate geometrical model. Gabriel Sato's latest report arrived shortly before lunch. He had been comparing three independent sets of observations concerning the Pelagic Complex. The first consisted of historical bathymetric measurements, corrected where possible for known limitations in navigation and depth estimation. The second drew upon more recent shipborne surveys.

The third incorporated satellite-derived gravitational data and modern reconstructions of the ocean floor. The agreement was generally good. Most of the apparent anomalies identified during the earlier investigation had disappeared after correcting for known systematic errors. Miriam regarded this as encouraging. Yasmin regarded it as indispensable. Gabriel regarded it as the ordinary consequence of competent scientific work. Nevertheless, a small residual inconsistency remained. It concerned three regions near the western boundary of the Pelagic Complex. For convenience, Gabriel had designated them A, B and C.

Each region contained identifiable submarine features whose relative positions could be estimated through several independent methods. The problem arose when the measurements were considered together. The relationship between A and B appeared consistent. So did the relationship between B and C. The relationship between C and A was less satisfactory. A conventional adjustment to the positions could reduce the discrepancy, but doing so degraded the agreement elsewhere. This was not, in itself, extraordinary. Survey networks frequently contained inconsistent observations.

Measurement uncertainty, systematic errors and inadequate physical models could all produce precisely such results. Gabriel had spent several days investigating those possibilities. He had found several. He had corrected them. The remaining discrepancy was smaller. But it had not disappeared. Yasmin examined the latest residual plot.

"How significant?"

Gabriel appeared on the video screen.

"Statistically? More than I'd like. Physically? Still within the range of mechanisms we haven't excluded."

"Could it be navigation error?"

"Possibly."

"Errors in the gravity model?"

"Possibly."

"Ocean currents affecting acoustic measurements?"

"Some of the relevant methods are sensitive to sound-speed structure and currents, yes."

"Then we haven't established anomalous geometry."

"Certainly not."

Miriam watched the exchange with interest. The two scientists appeared to derive a peculiar satisfaction from reducing the significance of their own findings.

"What's unusual about the residual?" Yasmin asked.

Gabriel opened another diagram.

"It's the structure of the discrepancy rather than simply its magnitude."

"Explain."

"If the problem were a straightforward position error, we'd expect the observations to become more mutually consistent after adjusting the relevant coordinates."

"Yes."

"But here the best-fitting correction depends upon which measurement relationships we prioritise."

"Could that reflect correlated systematic errors?"

"Of course."

"And have you modelled them?"

"Several forms. Some improve the fit."

"But not completely."

"No."

Yasmin studied the diagram.

"Have you considered a geometrical model in which the region isn't represented adequately by the coordinate relationships used in the survey?"

Gabriel smiled.

"Yes. Every geophysicist does, whenever the standard models become inconvenient."

"I don't mean an arbitrary deformation."

"What do you mean?"

"A model in which the relationship between local measurements and global position contains an additional geometrical degree of freedom."

Gabriel stopped smiling.

"That's a rather ambitious interpretation of some imperfect survey data."

"I know."

"What's the physical justification?"

"At present, none."

"Then why propose it?"

"Because I want to establish what observation would distinguish it from conventional error."

Gabriel considered.

"That's reasonable."

He opened the underlying measurements.

"All right. Let's assume, provisionally, that the local observations are reliable but their combination requires a different spatial model. What would you expect?"

Yasmin thought for a moment.

"The discrepancy should exhibit a reproducible structure across independent measurement methods."

"Yes."

"And it should persist when the ordinary environmental corrections are applied."

"Yes."

"It should also vary in a predictable way with the geometry of the observation network."

Gabriel nodded.

"That's the important part."

Miriam leaned forward.

"Why?"

Gabriel answered.

"Because an arbitrary error can explain almost anything after the event. A geometrical model should tell us how the discrepancy changes when we alter the arrangement of the measurements."

"Which would allow an independent test."

"Exactly."

Yasmin opened her own model. She had extended it to include a hypothetical additional relationship between the three survey regions. The result was mathematically simple, though its physical interpretation remained unclear. It was possible to reproduce the observed inconsistency by assigning the network a small additional transformation when measurements were carried around certain closed paths. Miriam looked at the diagram.

"Doesn't ordinary curved space already do something like that?"

Yasmin smiled.

"Yes. In differential geometry, transporting a direction or other geometrical quantity around a closed loop can produce a result dependent upon the curvature enclosed. That's not exotic. Nor is it evidence of a violation of general relativity."

"Then what's unusual here?"

"The size and structure of the inferred effect, if it survives proper analysis. It doesn't resemble the corrections we expect from the known gravitational field or the conventional geophysical models."

Gabriel added, "And we haven't demonstrated that it survives. That's rather important."

"Of course," Yasmin said.

Miriam examined the triangle. A to B. B to C. C to A. An arrangement whose local relationships appeared consistent but whose global interpretation remained troublesome. It reminded her of the old descriptions of impossible architecture. Walls whose angles could not be reconciled. Distances which depended upon the route taken. A place whose geometry did not conform to familiar spatial assumptions. The resemblance was striking. It was also dangerously seductive. She said nothing. Gabriel continued.

"I can design a survey comparison that would test the model. We have enough independent measurements to begin, although a definitive test would require new observations."

"How long?"

"That depends upon data availability and the cooperation of several institutions."

"Then begin with the existing material," Yasmin said.

"Agreed."

Miriam looked at the screen.

"And Gabriel?"

"Yes?"

"Would anything in your current analysis justify describing the formation as a city?"

He laughed.

"Absolutely nothing."

"Or an artificial structure?"

"No."

"Or a place possessing impossible geometry?"

"Not remotely. At present it justifies describing some of our measurements as inadequately reconciled."

"Good."

Gabriel looked puzzled.

"Why do you keep asking?"

"Because I'm becoming rather concerned about the damage done by inappropriate metaphors."

"Then you have my sympathy."

The first unexpected result came not from the Pacific but from a university building in Cambridge. Yasmin had been examining the mathematical properties of her provisional model when she realised that its predictions did not depend upon the physical scale of the observation network. That did not mean the hypothetical phenomenon should be observable everywhere. Most physical theories contain parameters determining the circumstances under which particular effects become significant. But the model's geometrical structure could, in principle, be tested without sending a research vessel into the Pacific.

She wanted a controlled environment. A location in which distances, signal paths and timing relationships could be measured through independent methods. A place where the instruments could be calibrated, their uncertainties quantified and their behaviour checked against known physics. She approached a colleague who worked on precision measurement. Dr Daniel Harcourt was an experimental physicist whose principal research concerned optical frequency transfer and the comparison of highly stable timing references.

He was a patient man, though not particularly tolerant of theoretical proposals which failed to specify what an instrument was supposed to measure. Yasmin liked him. Their working relationship consisted largely of her proposing mathematically interesting possibilities and his explaining why the available apparatus would not demonstrate them. He regarded this as a productive arrangement.

"What do you want to measure?" he asked.

"Whether certain spatial relationships remain mutually consistent under independent methods of distance and time transfer."

"That describes most of metrology."

"I know."

"Are you looking for a gravitational effect?"

"Not specifically."

"An anisotropy?"

"Perhaps."

"A failure of clock synchronisation?"

"Possibly."

Daniel looked at her.

"You're being remarkably unhelpful."

"Because I don't want to bias your interpretation."

"Then give me the model and let me tell you whether it predicts anything observable."

Yasmin supplied a simplified version. She removed all references to the Necronomicon, the Pacific structures and the supposed historical anomalies. Daniel read it. Then he read it again.

"This is an unusual geometrical construction."

"Yes."

"Does it arise from a physical theory?"

"Not yet."

"Then what are the parameters?"

"Partly unspecified."

He looked at her.

"Wonderful."

"I thought you'd appreciate that."

"Are you proposing a new spacetime metric?"

"Provisionally, a family of possible geometrical relationships."

"That's not the same thing."

"I know."

"And you're assuming that two physically distinct locations may possess an additional relation not captured by their ordinary separation?"

"Yes."

Daniel considered.

"Then the question is how that relation couples to something measurable."

"Exactly."

He looked at the equations again.

"If it doesn't affect signal propagation, clock rates, phase or some other observable, then we can't test it."

"Agreed."

"And if it does, we need a quantitative prediction."

"Yes."

He paused.

"There's a possible experiment using our existing timing-transfer infrastructure, but I wouldn't interpret the result geometrically unless the conventional error sources were exhausted."

"I wouldn't ask you to."

Daniel looked at her.

"That sounds suspiciously reasonable."

"I'm capable of it occasionally."

He smiled.

"All right. Show me what you've got."

The experiment involved three timing references distributed between university facilities. The apparatus had not been constructed for the investigation. Its purpose was ordinary precision metrology. Signals were transferred through known paths, and the resulting timing differences were recorded using calibrated equipment. The system was capable of detecting extremely small discrepancies. It was also capable of producing misleading results when environmental effects, fibre delays, hardware asymmetries or synchronisation assumptions were inadequately modelled. Daniel insisted upon beginning with those limitations. Yasmin agreed.

They designated the three reference stations A, B and C. Each possessed an independent timing record. Signals could be compared along more than one route. The geometry of the arrangement was not especially interesting in itself. That was deliberate. A university laboratory was a much better place to investigate a strange measurement than an allegedly cursed region of the Pacific. They recorded the existing calibration data. They established the expected uncertainties. They selected several control intervals during which no unusual behaviour was anticipated.

Then they examined the timing relationships around the three-station network. Most observations agreed with the established model. A few contained residual discrepancies. Daniel immediately identified several likely causes. Temperature-dependent delays. Instrumental asymmetries. Small differences in cable length. Corrections applied at different stages of processing. The discrepancies diminished after these were addressed. Yasmin watched without comment. The investigation was proceeding exactly as it should. Then Daniel found a residual which did not respond to the expected correction. It was small. Very small. But its structure was peculiar.

When the timing relationships were compared around the closed network, the residual depended upon the direction in which the comparisons were combined. Daniel examined the calculations.

"That's annoying."

Yasmin looked up.

"Why?"

"Because it suggests an unmodelled asymmetry."

"Could it be in the transfer hardware?"

"Almost certainly."

"Then we should test that."

"Yes."

They exchanged several components. The residual changed. That was encouraging. It suggested an instrumental contribution. Then they repeated the measurements. The residual returned, though with a different magnitude. Daniel frowned.

"That's less encouraging."

"Could it be environmental?"

"Possibly."

They checked the temperature records. The residual did not follow the expected thermal behaviour. They examined the signal paths. They changed the order of comparison. They repeated the calibration. The effect remained intermittent. Yasmin refused to interpret it. Daniel became increasingly irritated. Not because he believed they had discovered a new property of spacetime. Because he could not immediately explain an instrument which was misbehaving in a particularly systematic manner. After two days, he produced a preliminary report. It identified a small direction-dependent residual in the timing-transfer network.

The report explicitly stated that instrumental or processing explanations remained likely. It recommended further calibration and independent replication. Yasmin read it.

"Excellent."

Daniel stared at her.

"Excellent?"

"Yes."

"We've found a nuisance."

"We've found something we can test."

"Which will almost certainly turn out to be a nuisance."

"Quite possibly."

He considered.

"You're enjoying this far too much."

"Only because you're refusing to overinterpret it."

Daniel looked at the graph.

"Well, I suppose someone has to maintain standards."

Miriam visited the laboratory on Thursday. She had no intention of pretending to understand the apparatus. Daniel explained the experiment in terms she could follow. The stations possessed independently maintained timing references. Signals travelled along known paths. The researchers compared the resulting observations against a model incorporating the expected physical delays. The residual discrepancies were small. They did not establish anything exotic.

"But if they persisted," Miriam asked, "what would they imply?"

Daniel considered.

"That our present model of the apparatus was incomplete."

"And if every ordinary explanation were excluded?"

"Then we'd have a genuinely interesting physical anomaly."

"Could it involve the geometry of space?"

"Possibly. But that would be one of many interpretations, and not the first I'd choose."

Miriam nodded. She liked Daniel. He treated the possibility of a revolutionary discovery with approximately the same emotional enthusiasm he might have displayed towards a defective power supply. Yasmin showed her the provisional geometrical model. It represented the timing references as points connected by measured relationships. Under the conventional model, the relevant differences could be made mutually consistent through the expected corrections. Under the alternative model, a small additional relational term was permitted. The model could reproduce the observed residual. Miriam looked at it.

"Have you fitted the effect after seeing the data?"

"Yes," Yasmin said.

"Then it isn't a prediction."

"Correct."

"Good."

Daniel smiled.

"She's been teaching you."

"Rather effectively," Miriam said.

Yasmin displayed a second graph.

"This is the important part. The alternative model predicts a particular relationship between the residual and the arrangement of the signal paths."

"Which can be tested independently?"

"Yes."

"Have you done so?"

"Not yet."

Miriam examined the diagram. Three stations. Three connecting relationships. A discrepancy which appeared only when their measurements were combined. It was almost embarrassingly modest. Nothing had appeared. No opening had formed in the air. No object had vanished. The apparatus continued to occupy the same room, and its instruments behaved with the ordinary mixture of precision and imperfection characteristic of experimental physics. Yet the mathematical model suggested something peculiar.

If the residual were physical rather than instrumental, it might represent a relationship which could not be reduced to the ordinary distances between the stations. Miriam thought of the seventh division. The distinction between positions and the conditions by which positions were constituted. The apparent identity of Yog-Sothoth with the relationships connecting regions of space and time. She looked at the three stations.

"Could two locations be separate under one description and coincident under another?"

Daniel considered.

"That depends upon what you mean by coincident."

"Occupying the same position."

"Position is defined within a geometrical model. Different coordinate systems can assign different labels to the same event, but they don't ordinarily turn two physically separate events into one."

"Of course."

"However, more general mathematical spaces can contain identifications or relationships which aren't represented by ordinary Euclidean separation."

"Such as?"

"Topological identifications, for example. Or geometrical structures in which the shortest connection between regions isn't represented by the path you'd expect from their apparent embedding."

Miriam looked at him.

"Would that mean the locations were actually the same?"

"Not without a physical definition of what 'same' means."

Yasmin nodded.

"Exactly."

Daniel continued.

"If two events can be connected through some additional physical relation, that doesn't necessarily make them identical. We'd need to specify what measurements establish the identification."

"And if the measurements did?"

"Then we'd have to revise our geometrical description."

Miriam considered.

"Could a medieval observer have described such a relationship as a gate?"

Daniel laughed.

"Perhaps. But I'd rather not use medieval terminology to interpret precision measurements."

"Quite."

Yasmin looked at Miriam.

"That's the problem with the name."

"What name?"

"Yog-Sothoth."

Miriam felt a sudden discomfort. She had not expected Yasmin to mention it in front of Daniel. He looked between them.

"I'm sorry?"

"An entity in a fictional text," Yasmin said, "associated with the relations between places and events."

Daniel stared at her.

"You've been modelling Lovecraft?"

"No."

"Then why have you got a name from his fiction attached to your geometry?"

"Because the historical material contains descriptions whose underlying relationships are surprisingly similar to the mathematical problem."

Daniel looked at Miriam.

"And you're the historian."

"Unfortunately."

He sat down. For several seconds he said nothing. Then he began laughing. Miriam waited patiently. Eventually he stopped.

"Well," he said, "that explains the secrecy."

Yasmin looked slightly offended.

"I haven't concealed the experimental assumptions."

"No. You've merely omitted the fact that your theoretical inspiration is a fictional cosmic entity."

"An inspiration isn't evidence."

"Quite."

"And the experiment is valid or invalid independently of the source of the hypothesis."

"Also quite."

Daniel examined the graph.

"But if you start interpreting every instrumental discrepancy as the influence of Yog-Sothoth, I shall resign from the project immediately."

Miriam smiled.

"So shall I."

Yasmin looked at them.

"Then perhaps we can return to the measurements."

The independent test began on Friday. Daniel altered the timing-transfer configuration according to a procedure agreed before the new measurements were taken. The experiment was designed to distinguish between three possibilities. The first was that the residual arose from ordinary instrumental asymmetry. The second was that it reflected an environmental effect shared by the apparatus. The third was that it depended upon the relational arrangement of the measurement network in the manner predicted by Yasmin's provisional model. The distinction was important.

An instrumental error should ordinarily follow the relevant instrument or signal path. An environmental effect should correlate with physical conditions. A genuine geometrical relationship, if such a thing were involved, should remain associated with the arrangement of the measurement locations rather than with a particular piece of equipment. Daniel insisted upon a blind analysis. The instruments were assigned identifiers which concealed their physical positions from the person performing the first stage of the data reduction. The predicted residuals were recorded in advance. Several control configurations were included.

Yasmin was not permitted to adjust the model after seeing the results. Miriam found the arrangement admirable. It also made her uncomfortable. For the first time, the investigation was approaching a test whose outcome might genuinely discriminate between their competing explanations. The earlier discoveries had been frustratingly retrospective. A document appeared to contain information from the future. A historical source acquired a newly recognised connection. A geological specimen possessed a more extensive research history than its investigator remembered.

Each anomaly could be explained, at least in principle, by some combination of error, incomplete knowledge or deception. The present experiment was different. The prediction had been recorded. The apparatus had been calibrated. The possible outcomes had been defined before the data were collected. If the residual followed the instruments, the geometrical hypothesis would lose support. If it followed environmental conditions, the same would be true. If it followed the predicted relational structure, there would be something more difficult to explain. Miriam spent the afternoon in the laboratory.

She had no useful role in the measurements. She therefore sat at a table with Oliver's latest manuscript transcriptions, attempting to discover whether the distinction between positions and their underlying relations had appeared elsewhere in the textual tradition. The relevant passages were numerous. They were also maddeningly ambiguous. One described an entity which could be present wherever certain conditions were satisfied. Another referred to an existence which was not divided by the succession of events.

A third distinguished between the movement of a thing and the alteration of the relation by which its location was established. Miriam resisted the temptation to construct a coherent doctrine from them. Medieval texts were perfectly capable of containing contradictions. Their authors were not necessarily systematic philosophers. And the transmission of unfamiliar material through successive languages could produce apparent relationships which had never existed in the original. Still, the recurrence of the same distinction was striking.

The book did not consistently describe Yog-Sothoth as a creature occupying a particular location. At times it appeared to identify the name with the very possibility of connection between locations. The entity was not beyond the gate. It was not waiting to pass through. It was, according to one interpretation, the gate itself. Miriam had always considered that expression an effective piece of supernatural imagery. Now she wondered whether it might conceal a more precise, and considerably less anthropomorphic, idea.

A gate did not necessarily have to be a physical opening. It could be a relation which made passage possible. A relation which might exist independently of the objects connected by it. And if the universe's spatial relationships were becoming more complicated, then the appearance of such a relation might be indistinguishable from an alteration in the structure of space. There would be no door. Nothing would visibly open. Two regions previously regarded as separate might simply prove to possess a connection which had not been recognised before.

Or which, until the process reached a sufficient degree of consistency, had not existed at all. Miriam looked up from the manuscript. Daniel was examining the first set of results. He appeared irritated.

"Something wrong?" she asked.

"Possibly."

Yasmin looked towards the screen.

"What?"

"The residuals aren't following the hardware as closely as I'd expected."

"Could the recalibration be incomplete?"

"Certainly."

He opened a second display.

"But there's something else."

The timing differences had been plotted against the configuration of the measurement network. Most points followed the expected relationships. A smaller group departed from them. The deviations were modest. But their signs and relative magnitudes followed the pattern Yasmin had recorded before the experiment. Daniel stared at the plot.

"That's inconvenient."

Yasmin said nothing. He checked the analysis. Then he checked the calibration records. Then he asked the colleague responsible for the blinded data reduction to repeat the calculations independently. The result remained. Daniel sat back.

"Well."

Miriam looked between them.

"Is that good?"

"No," Daniel said.

"Bad?"

"Not necessarily."

"Then what?"

"Annoying."

Yasmin leaned forward.

"Does it favour the alternative model?"

"Under the present uncertainty assumptions, yes."

"Significantly?"

"Enough to justify repeating the experiment. Not enough to justify rewriting the textbooks."

Miriam smiled faintly.

"A familiar distinction."

Daniel looked at the graph.

"The conventional model remains possible if we've underestimated some correlated error source."

"Which we may have," Yasmin said.

"Quite."

"But the predicted relationship survived the configuration change."

"Yes."

"Without adjusting the model."

"Yes."

Yasmin sat very still. Miriam recognised the expression. It was not triumph. It was the look of someone who had just discovered that a carefully designed experiment had failed to remove an unwelcome possibility. The second test took place the following week. This time Daniel introduced an independent timing reference and altered one of the transfer paths. He also arranged for a separate laboratory to analyse a subset of the measurements without access to Yasmin's interpretation. The results were less clear. Some of the original residual disappeared.

A different discrepancy emerged. One control behaved unexpectedly. Another agreed almost perfectly with the conventional model. Daniel was relieved.

"Good," he said.

Miriam looked surprised.

"Good?"

"Yes. It means we aren't simply obtaining the result we want every time."

Yasmin nodded.

"Which would be suspicious."

"Precisely."

The team spent two days identifying ordinary causes for the new discrepancies. Several involved temperature-dependent effects. One arose from an incorrect delay correction. Another resulted from a software assumption concerning the order in which timestamps were combined. The last interested Yasmin particularly. The software had treated the timing references as though they could be placed within a single globally consistent ordering. For ordinary laboratory conditions, this was a reasonable approximation. But the mathematical model she was investigating did not necessarily admit such an ordering under every possible configuration.

Daniel examined the relevant calculation.

"That's not a physical anomaly. It's a modelling assumption."

"I know."

"Then we correct it."

"Of course."

They did so. Most of the discrepancy disappeared. But the correction raised a conceptual question. What did it mean to assign a temporal order to events occurring at different locations?

Under ordinary relativity, there was no unique invariant ordering for events separated in such a way that neither could causally influence the other. Their apparent order could depend upon the reference frame. That was established physics. It did not imply that events could simply cause their own past. Nor did it allow arbitrary contradictions in local causal sequences. But the distinction between temporal succession and causal relationship was fundamental. Yasmin had been modelling the network as a set of events connected by measured physical relations.

The resulting structure did not always support a unique global ordering. That, by itself, was not extraordinary. What troubled her was that some of the preferred relationships differed from those predicted by the ordinary spacetime geometry used to interpret the measurements. The alternative model admitted configurations in which events conventionally regarded as widely separated could occupy a different relationship within the underlying structure. Not necessarily the same event. Not necessarily directly connected by a signal.

But connected in a manner which could not be expressed entirely through the familiar spatial distances and timing relations. Yasmin looked at the graph. She had spent days attempting to identify a physical interpretation for its additional geometrical term. Now a possibility occurred to her. Perhaps she had been treating it incorrectly. She had assumed it represented a correction to the ordinary geometry. A small additional effect superimposed upon the familiar structure of space and time. But the mathematical construction did not require that interpretation.

It could instead describe a more general relational structure from which the familiar geometry emerged as a particular approximation. The distinction was considerable. A correction implied that ordinary space remained fundamental. An emergent geometry implied that ordinary space was merely one way in which deeper relationships became physically expressed. Yasmin looked at Daniel.

"Could you do something for me?"

"What?"

"Analyse the measurements without assuming that the standard spatial coordinates are the primary variables."

He frowned.

"Then what would you use?"

"The measured relationships themselves."

"That sounds like a relational reconstruction."

"Exactly."

Daniel considered.

"We can attempt it. But we'll lose some of the constraints that normally make the analysis stable."

"I know."

"And the result may be radically underdetermined."

"Yes."

"And any apparent improvement could simply reflect the additional freedom of the model."

"Which is why we'll penalise complexity and test it against held-out data."

Daniel smiled.

"You're learning to anticipate my objections."

"Years of practice."

He turned back to the apparatus.

"Very well. Let's see whether your geometry can survive without its familiar coordinates."

Before Daniel began the relational reconstruction, Yasmin prepared a four-page technical note: Closed-Loop Residuals in Timing-Transfer Networks: Limits of a Geometrical Interpretation. It contained the observed residuals, the proposed alternative model, the control failures and the calibration uncertainties. The conclusion explicitly declined to identify a new physical effect. She circulated it to Daniel and to Dr Priya Sen, a metrologist who knew nothing of the book or the documentary investigation.

"You intend to put your name on this?" Daniel asked.

"On the measurements and their limitations. Not on an apparition in a timing network."

He checked the graph captions and agreed to let the note circulate without his name as co-author. Yasmin registered a dated copy with the department's ordinary research archive. She put the signed print in her laboratory notebook, photographed the page on her phone and recorded the file fingerprint. At last she had a claim narrow enough to be professionally responsible for.

Priya's reply came the next morning: Why are you circulating this as a new note? We discussed the same closed-loop construction in your seminar three years ago. I've attached my comments from that meeting.

The attachment showed Yasmin's equations, written in the same hand, above diagrams of a three-station timing network. Priya had objected to fitting and testing a model on the same residuals. Yasmin remembered having learned that objection from her. She did not remember giving the seminar. The department's list of earlier talks contained her title and a room number, along with a brief abstract so close to the new note that it might have been an earlier draft.

She called Priya. "Are you certain I gave that talk?"

"You drew the triangle twice because the board was too narrow. You were annoyed with me about the controls."

"I wouldn't have been annoyed about that."

"You were. You bought me coffee afterwards."

Yasmin thanked her, ended the call and found the room booking in her old calendar. The signed print in her notebook remained exactly where she had put it. There was no contradiction in the current record, only a history of her own intellectual work which she did not remember conducting.

She moved the note from new work to reanalysis in the departmental register. It was a professional correction she could justify from the evidence. It also meant she could no longer present the mathematical model as an independent response to Miriam's problem. According to her records, she had possessed its central construction for years.

For the first time she understood why Miriam had looked ill when she found an unfamiliar citation in her own handwriting. Yasmin could defend the equations. What she could not defend was the provenance of her thought.

The result appeared three days later. It was not a picture. There were no impossible angles, no distorted rooms and no visual representation of a doorway opening into an unfamiliar dimension. There was a mathematical structure. A network of events and relationships. Most of it corresponded closely to ordinary spacetime. The familiar geometry emerged naturally across the great majority of the observations. But one small portion required an additional identification.

Two experimentally distinct regions, separated by an ordinary measurable distance, occupied positions in the reconstructed relational structure which were much closer than their conventional spatial separation suggested. Yasmin examined the result. The model did not establish that the regions were literally identical. It did not imply that a person could step from one into the other. It did not predict a usable shortcut.

But it suggested that some of the measured relationships might be represented more economically if the two regions were treated as sharing a deeper geometrical connection. Daniel disliked the result.

"It's underconstrained."

"Yes."

"And the improvement may be a consequence of the model's additional flexibility."

"Yes."

"Then we shouldn't conclude anything."

"Agreed."

He looked at Yasmin.

"But you're interested."

"Very."

"Why?"

She opened the prediction recorded before the experiment. The additional identification had been anticipated by the form of the mathematical model. Not its exact location. Not its magnitude. But its relational character. The independent analysis had selected a structure from the predicted family. Daniel examined the comparison.

"That's interesting."

Miriam, who had been listening quietly, looked up.

"From you, I gather that's quite a statement."

Daniel ignored her.

"We'll need a genuinely independent replication."

"Yes."

"With different hardware."

"Yes."

"And ideally a different physical method."

"Agreed."

He looked towards the screen.

"Because if this survives, we have a serious problem."

Miriam leaned forward.

"What sort of problem?"

Daniel considered.

"Either our model of the measurements is missing something important, or our model of the geometry is."

"And if it's the geometry?"

He looked at her.

"Then the distinction between two separated locations may not be as straightforward as we've assumed."

Miriam felt an unpleasant sensation of recognition. The old manuscript. The place not reached by travelling towards it. The distinction between positions and the conditions by which positions were constituted. She looked at the relational reconstruction. Two locations. Separate in ordinary geometry. Connected differently in the more general model. For a moment she understood what the old writers might have been trying to describe. Not a supernatural passage. Not a corridor through another universe. Something more fundamental.

A relationship under which the ordinary distinction between here and there was not sufficient. She thought of the word gate. She began to understand why it might have been chosen.

On Wednesday Oliver requested a controlled photographic examination of the forty-seventh surviving leaf. He had not been persuaded that previous differences between images proved any change in the physical book. He wanted to establish whether the imaging process introduced them.

Helen Mercer agreed to supervise. She insisted on the ordinary conservation rules: the volume remained in its cradle, its opening angle fixed, and no leaf was touched during the sequence. The laboratory supplied a colour target, a calibrated support and a camera set to manual exposure. A second camera recorded the apparatus rather than the text. A control leaf from an unrelated early printing was photographed in the same session.

At 10:14, Helen captured the relevant passage and inspected the preview. The heading of the disputed section contained a word Miriam read as restitutione. She wrote it on the observation sheet. Oliver checked the screen and initialled the entry. Neither claimed the reading was secure; the lettering was cramped, and the form could be confused with a similar word.

At 10:19, after an unchanged exposure and without opening the volume any further, Helen captured a second image. It was displayed beside the first. The two photographs showed the same tear, stain and wormholes. But the second heading read constitutione. Its type spacing had been redistributed along the line; the adjacent letters aligned differently, although the paper damage occupied the same positions.

Helen inspected the physical leaf through the magnifier. She read constitutione. She examined the first image again, checking for a reflection or an imaging artefact. It continued to show restitutione. The control leaf agreed across both exposures. On the second camera's record Helen had not moved the book between shots.

"I won't permit another exposure until we've checked the equipment," she said. "Nor will I permit someone to lift a leaf to satisfy a theory about it."

They changed the light source, checked the sensor and repeated the photograph. The later reading remained. The camera had recorded two different arrangements of ink without any corresponding alteration anyone had seen. Miriam could not say when, or even whether, the physical sheet had changed. She had spent the entire interval looking at the image monitor rather than at the leaf.

The original raw files were copied to independent storage and hashed. Each file's fingerprint matched the value printed on its own acquisition sheet; neither photograph had silently become the other. Their timestamps were five minutes apart. The mismatch was between the depicted readings, not between an image and its current checksum. Helen signed a report limited to what the procedure established: two exposures under the recorded conditions, incompatible printed readings, cause undetermined.

Oliver carried the volume back in its cradle. He would not call the incident a manifestation. He would not allow the book to become an experimental prop. The conservation report would describe it as a discrepancy in the imaging sequence until the ordinary alternatives had been pursued.

Miriam took the copies home. Under a bright desk lamp she compared the two lines until the letters seemed to lose their meaning. She had seen neither word change. That was the important restraint. Yet the disagreement existed in two physical records made under direct supervision, not in a manuscript whose catalogue she had consulted years too late. She could no longer make the evidence vanish by conceding that she had misremembered a research note.

Yasmin received the images and the signed report. She read the headings, ran the checks and asked Helen for the unprocessed files. She wrote one sentence beneath the data: The experiment has not identified a changing physical state; it has identified incompatible records produced under a controlled observation protocol. She then placed the sentence in the part of her laboratory notebook reserved for results which her model had not predicted.

On Thursday evening, Yasmin asked Miriam to visit her office. It was nearly nine when she arrived. The building was quiet. Only a few rooms remained illuminated, and the corridors possessed the slightly abandoned appearance common to academic institutions after the day's ordinary business had finished. Yasmin was sitting before her monitor. She looked exhausted.

"Have you eaten?" Miriam asked.

"Not properly."

"That seems to be becoming a habit."

"Yes."

Miriam placed a sandwich on the desk. Yasmin looked at it.

"Thank you."

"Now tell me what's happened."

Yasmin opened the latest relational reconstruction.

"I've been examining the time-ordering problem."

"The one concerning the three timing references?"

"Yes."

"And?"

"I think we've been asking the wrong question."

"Again?"

"Again."

She displayed a network of events. Each event was represented by a node. Connections indicated measured relationships, including signal propagation and timing comparisons. The ordinary spacetime model provided one interpretation. The alternative model provided another.

"Under conventional assumptions," Yasmin said, "we assign positions and times to the events, then calculate the relationships between them."

"Yes."

"But we can also begin with the relationships and ask what geometrical structure would support them."

"A relational reconstruction."

"Exactly."

"And the result?"

"The familiar spacetime description remains an excellent approximation across most of the network."

"Most?"

"Yes."

She highlighted a small group of nodes.

"But these events can be embedded in the relational structure in more than one way."

"Meaning?"

"The measurements don't uniquely determine their positions within the ordinary geometrical model."

"That sounds like insufficient data."

"It may be."

"But?"

"The alternative embeddings aren't arbitrary. They're related by a specific transformation which also appears in the Pacific model."

Miriam looked at her.

"The same mathematical structure?"

"A closely related one."

"Could that be because you've used the same assumptions?"

"That's a serious possibility. I've been trying to separate the model's imposed structure from relationships genuinely demanded by the data."

"And have you succeeded?"

"Not completely."

Miriam sat down. Yasmin continued.

"There's another difficulty. Under one embedding, two events are separated by an ordinary spatial interval. Under another, they occupy a much closer relationship in the underlying structure."

"Could either interpretation be physically correct?"

"We don't know."

"Could both?"

Yasmin hesitated.

"That's the question."

Miriam looked at the graph.

"If both are mathematically consistent, how would we choose?"

"Through additional physical observations."

"And if the observations support both?"

"Then the distinction may not be physically meaningful in the way we're assuming."

Miriam thought for a moment.

"You're suggesting that the same events could admit different spatial descriptions without either being merely a coordinate convention."

"Possibly. But we'd need to establish that the descriptions make distinct, testable predictions."

"And time?"

Yasmin looked at the highlighted nodes.

"The same difficulty appears in the ordering of certain separated events."

"Isn't that ordinary relativity?"

"Sometimes. Relativity already tells us that spacelike separated events have no unique invariant temporal order. That isn't the anomaly."

"Then what is?"

"The relational model changes which event pairs fall into that category."

Miriam considered.

"Events we ordinarily regard as causally ordered might become differently related under the alternative geometry."

"Yes. But I want to be careful. We have not observed a signal travelling into its own past, nor established any violation of local causality."

"Understood."

"Nevertheless, the model suggests that our assignment of causal relationships may depend upon a geometrical structure which the present measurements don't uniquely determine."

Miriam looked at the screen.

"Could we have been using the wrong geometry all along?"

Yasmin did not answer immediately. Then she said, "That's one of the possibilities I can no longer dismiss as merely incoherent." For several minutes neither spoke. Rain moved against the windows. Somewhere in the building a ventilation system started, produced a low mechanical hum, and stopped again. Nothing else happened. There were no disturbances in the room. No peculiar shadows. No suggestion that the mathematical model was anything more than an abstraction on a computer screen. Miriam found herself wishing for something dramatic. An impossible angle.

A doorway opening onto another location. Some unmistakable event which would justify abandoning the ordinary assumptions through which she had understood the world. But the universe offered no such reassurance. There were only measurements. Small discrepancies. Competing mathematical descriptions. And a growing body of historical evidence whose provenance was becoming increasingly difficult to reconcile with the chronology of its discovery. Miriam looked at Yasmin.

"Do you think the book caused this?"

Yasmin shook her head.

"I don't know."

"Does the model favour that explanation?"

"No."

"Then what does it favour?"

"It doesn't privilege any particular direction of causation."

Miriam waited. Yasmin continued.

"The structure is compatible with the book having acquired a history through the reconstruction. But it's also compatible with a more general history becoming established, one in which the book is simply a consequence."

"Which would mean the research wasn't necessarily the cause."

"Exactly."

"Then what was?"

Yasmin looked towards the graph.

"Perhaps that's the wrong question."

"Why?"

"Because we're still thinking in terms of an event occurring at a particular time and producing effects afterwards."

"Which is how causation ordinarily works."

"Yes. But the mathematical structure we're exploring doesn't necessarily admit that kind of description globally."

Miriam felt a sudden chill.

"Then the book might not have caused the geometry."

"No."

"The geometry might have caused the book."

"Perhaps."

"Or both might be parts of the same consistency structure."

"Exactly."

Miriam looked at the notes concerning Yog-Sothoth. The supposed entity associated with gates. With the conjunction of places. With a mode of existence not confined to temporal succession. She had interpreted the descriptions as imaginative mythology. Now they seemed increasingly like attempts to name something which possessed no ordinary object-like existence at all. A relationship. A structure. A condition under which space and time acquired their familiar organisation. And if that structure became part of the universe's physical history, the question of when it arrived would be meaningless.

It would not arrive. It would become one of the conditions under which arrival and departure were possible. Yasmin turned to the blackboard.

"I want to show you something."

She drew a circle.

"This represents a region of ordinary spacetime."

Miriam nodded. Yasmin drew another circle beside it.

"And this represents another region."

"Separated?"

"Under the familiar geometry, yes."

She drew a line between them.

"Now suppose the relationship between these regions isn't exhausted by their ordinary separation."

She added a second connection.

"This represents an additional identification in a more general relational structure."

Miriam examined the diagram.

"A shortcut?"

"Not necessarily. A shortcut suggests a path which can be traversed."

"Then what?"

"A relationship which alters the conditions under which separation is defined."

Miriam considered.

"Could we use it to travel?"

"There's no evidence of that."

"Could we communicate through it?"

"Again, no evidence."

"Could it affect measurements?"

"That's what we're investigating."

Yasmin stepped back.

"Now suppose the additional relation is not created at a particular moment."

"What do you mean?"

"Suppose it belongs to the complete structure of spacetime."

"Then it has always existed."

"Under that description, yes."

"But if the structure is becoming established retrospectively..."

"Then there may be no physically privileged moment at which the relation began."

Miriam looked at the diagram. A gate which did not open. A connection which did not require travel. A relationship which might be part of the world's geometry without ever having been an event within the world. She thought of the old phrase.

They have already always existed.

Perhaps the author had not been making a theological claim about eternal beings. Perhaps he had been attempting to describe a structure whose existence could not be assigned a beginning within the time it helped constitute. Miriam looked at Yasmin.

"And Yog-Sothoth?"

Yasmin put down the chalk.

"If the text's description corresponds to anything real, I doubt we're dealing with an entity in the ordinary sense."

"What, then?"

"A name assigned to a relationship the author didn't know how to describe."

"One which concerns the structure of spacetime."

"Possibly."

"And which might not possess a single location."

"Yes."

"Or a beginning."

"Perhaps."

Miriam considered.

"Then the name is misleading."

"Almost certainly."

"The mythology is misleading."

"Probably."

"And the idea that something is waiting beyond a gate..."

Yasmin shook her head.

"May be a misunderstanding of what the gate actually is."

There was a long silence. Then Miriam said, "That's considerably worse than a monster." Yasmin looked at her.

"Why?"

"Because a monster occupies the world. However powerful it may be, it remains one thing among others."

"Yes."

"But this..."

Miriam looked at the diagram.

"This might be part of what makes the world possible."

Yasmin said nothing. Miriam continued.

"If the book's history is becoming real, then perhaps the entities described within it aren't all objects being added to the universe."

"Go on."

"Some may be properties of the universe which are becoming part of its consistent description."

"That's close to my concern."

"And if one of those properties determines the relationships between places and events..."

"Then it could affect everything occupying spacetime."

Miriam closed her eyes. The scale of the implication was difficult to grasp. They had begun by investigating a physical book. A minor bibliographical anomaly. Then manuscripts. Then geological samples. Then submerged formations. Each step had extended the investigation into a larger part of the world's physical history. Now they were approaching something which might not be confined to any particular part of that history. A relationship governing the structure within which history itself occurred. Miriam opened her eyes.

"Can we stop it?"

Yasmin looked at her.

"We don't know that there's a process to stop."

"But suppose there is."

"Then we'd need to establish how it operates."

"And if the geometry is part of the mechanism?"

Yasmin hesitated.

"Then our attempts to describe it may already be taking place within the structure we're trying to understand."

Miriam looked at the blackboard.

"Which makes the investigation self-referential."

"Yes."

"And possibly causally ambiguous."

"Yes."

"Wonderful."

Yasmin smiled faintly.

"I thought you'd appreciate that."

Miriam walked home after midnight. The rain had stopped. A crossing signal changed, and she stepped off the pavement only when she saw the light and the cars had halted. The road, the junction and the opposite kerb formed an arrangement she knew so well she scarcely looked at it. The measurements at Cambridge were no justification for mistrusting a pedestrian crossing. Yet she caught herself counting the steps between the two sides.

At home she opened the seventh division. The old sentence about the gate could describe an object connecting two places, or a relation necessary for those places to be distinguished. She wrote both translations in the margin, resisted choosing one, and closed the book.

For a few seconds she stood in the study doorway with one hand on the frame. The hallway beyond was narrow, lined with familiar pictures. Nothing about its shape had changed. She was ashamed to discover that she wanted it to change, because a visible impossibility would have been easier to defend than a small residual in a timing network. She switched off the light. The house answered with an ordinary settling creak.

At 01:17 Daniel Harcourt sent Miriam an older timing-network paper. It reported a modest residual in a three-station comparison, attributed provisionally to instrument asymmetry and imperfect synchronisation. Nothing about that conclusion was extraordinary. The paper was dated 1994.

The footnote belonged, indirectly, to a different century. It cited a mathematical treatment published in 2011, complete with title, page range and journal reference. The citation was printed in the same type and spacing as the rest of the page. Daniel was checking whether a later revision had been incorporated during digitisation; the PDF alone could not establish more.

Miriam laid the printout beside the photograph of the London conservation report. In 1971, a typed page had cited Marsh's 1998 study. Now a 1994 experimental paper apparently cited work from 2011. The historical anomalies were no longer restricted to the transmission of one eccentric book. She sent Yasmin the comparison with a short message: The measurement problem is acquiring its own earlier literature. We need to know which editions of these papers we actually possess.

Then she returned to the seventeenth-century commentary. The author wrote of a gate, but described no door, threshold or traversable passage. The words concerned relations by which apparently separate places could be identified. If the passage possessed any physical referent, the name Yog-Sothoth might designate a structure rather than a person.

Miriam went downstairs to collect a glass of water. She stopped at the study door. The open doorway connected the room to the hall; it did not create the spatial relation between them. She could have explained this to a child. Until that week, she had never imagined needing to explain it to herself.

A gate might merely be the wrong name for a condition of connection. It might also be nothing more than a metaphor imposed on a badly translated text. She closed the door, hearing the latch settle, and returned the manuscript to its folder. The timing residual was small. The paper was old. Neither thing had changed visibly. Their histories, however, were becoming harder to disentangle from the geometry they were supposed to describe.

Chapter VIII 8 of 10

The Failure of Correction

Miriam Vale had always considered the correction of an error a relatively straightforward intellectual obligation. One discovered that a proposition was false, established the grounds for rejecting it, and amended the record accordingly. Where the error had been published, a correction should be published. Where subsequent arguments depended upon the mistake, their authors should be informed. Where the error was sufficiently serious to invalidate the entire work, withdrawal might be necessary. The process could be embarrassing. It could damage reputations.

It could require the abandonment of years of research. But none of these inconveniences altered the underlying principle. A false proposition did not become true because it had acquired a respectable bibliography. That was what Miriam had taught her students. It was what she believed. And it was the principle upon which she intended to act when she decided to withdraw The Book That Should Not Exist. The decision came on a Sunday morning, after a particularly unpleasant examination of the latest version of Yasmin's consistency model.

The model no longer treated the Necronomicon as an isolated historical object. It represented the book as one component within a much larger network of relationships: textual witnesses, translations, archaeological observations, biological structures, geological formations and the increasingly troublesome measurements of spatial relationships. Most of the connections remained speculative. Some depended upon interpretations Miriam distrusted. Others represented observations for which ordinary explanations had not yet been excluded. Nevertheless, the network had acquired a property which disturbed her.

It was no longer necessary to assume that the physical Necronomicon had appeared through a single extraordinary event. Its existence could now be supported through multiple independent routes. The London printing possessed a documentary history. The Paris manuscript possessed another. The Greek and Arabic traditions had acquired additional witnesses. The biological and geological investigations were developing histories of their own. The spatial anomalies had begun to acquire a scientific literature extending decades into the past. The original research had supplied a coherent description.

But that description was no longer the only structure supporting the book's apparent existence. Miriam examined the graph. Then she closed it. She had made a mistake. Not necessarily a mistake of scholarship. The original essay had been careful, qualified and, as far as she could establish, correct according to the evidence available when it was written. The mistake had been more fundamental. She had assumed that a sufficiently rigorous reconstruction of an imaginary object was a harmless exercise. She was no longer confident that it had been.

The essay had collected disparate fictional details and arranged them into a mutually intelligible structure. It had reconciled contradictions where possible, identified irreconcilable traditions, supplied a coherent account of textual transmission and considered the natural philosophy implied by the book's contents. It had given the imaginary Necronomicon something resembling a complete historical identity. And if Yasmin's hypothesis was correct, that might have been sufficient to establish relationships which physical reality had subsequently resolved through the appearance of an apparently genuine past. Miriam had no proof.

But she had enough evidence to question whether continuing to maintain the reconstruction was prudent. She opened the published article. Then she began writing to its author. Her message was considerably more difficult to compose than she had anticipated. She could hardly begin by explaining that the essay might have caused a fictional book to acquire a fifteenth-century printing history. Nor could she reasonably suggest that removing the article might prevent the appearance of pre-human biological structures and geometrical relationships associated with Lovecraft's imaginary cosmology.

Such claims would require evidence vastly stronger than anything she possessed. She therefore confined herself to the scholarly problem.

I've been examining the documentary history of the Necronomicon following publication of our reconstruction.

There are serious discrepancies between the fictional chronology and a growing body of apparently independent historical evidence.

Some of the relevant sources contain chronological contradictions which I cannot presently explain.

I think we should temporarily withdraw the article pending further investigation.

This isn't an admission that the original literary analysis was necessarily wrong. It is a precaution against continuing to present a settled reconstruction when the status of some of its underlying claims has become uncertain.

She read the message. It was accurate. It was also an almost comically inadequate description of the situation. She sent it. Then she telephoned Yasmin.

"I'm withdrawing the essay."

There was a pause.

"Why?"

"Because I think it may be contributing to the phenomenon."

"That's a hypothesis."

"Yes."

"Not an established conclusion."

"I know."

"Then what exactly do you expect withdrawal to accomplish?"

Miriam looked at the graph.

"If the reconstruction supplied the original constraints, removing it might reduce their influence."

Yasmin was silent for several seconds.

"Do you believe the historical record depends upon the continued availability of the essay?"

"I don't know."

"Then let's examine the assumption."

Miriam sighed.

"I rather hoped you'd simply tell me it was a sensible precaution."

"It may be. But I don't want you mistaking a precaution for an intervention whose effects we understand."

"Fair."

Yasmin continued.

"The book now possesses an extensive documentary history. Removing one recent article won't remove the manuscripts, catalogues, laboratory reports or measurements."

"No."

"Nor will it erase references already made to the article."

"No."

"And if the process is genuinely retrospective, deleting the original description may not affect the relationships which have already become established."

Miriam considered.

"Then perhaps we should publish a correction."

"A correction to what?"

"The reconstruction itself."

"Which parts?"

"The supposed chronology. The textual relationships. The claims which make the book a coherent historical object."

Yasmin hesitated.

"Miriam, I'm not sure that's wise."

"Why?"

"Because you're assuming that introducing contradictions will reverse whatever consistency has already developed."

"Wouldn't it?"

"Not necessarily."

"Explain."

Yasmin took a moment.

"A contradiction in a historical record doesn't automatically destroy the history. Historians deal with contradictions constantly."

"Of course."

"One manuscript gives a different date. Another supplies a different author. A later commentator rejects an earlier attribution."

"Yes."

"Those disagreements don't make the underlying object cease to exist."

Miriam looked at the bookshelves.

"They become part of its textual history."

"Exactly."

There was a long silence. Then Miriam said, "Nevertheless, the article should be withdrawn."

"On scholarly grounds, that may be reasonable."

"And the corrections?"

"I'd prefer to discuss them before you publish anything."

"Very well."

Yasmin hesitated.

"There's something else."

"What?"

"I think you should consider the possibility that the original essay is no longer causally necessary."

Miriam looked at the graph.

"I have."

"Then removing it may accomplish nothing."

"Perhaps."

"And publishing elaborate contradictions may simply add more information."

Miriam felt a sudden irritation.

"Then what are you suggesting? That we do nothing?"

"I'm suggesting that we stop treating a speculative model as though it came with an instruction manual."

Miriam closed her eyes. Yasmin continued more gently.

"We need to understand whether the process responds to descriptions, to the relationships those descriptions establish, or to something neither of us has recognised."

"And in the meantime?"

"In the meantime, we can stop publishing additional speculative material."

Miriam considered.

"That, at least, seems sensible."

"Yes."

"But I still intend to withdraw the essay."

"Then preserve the present version first."

"Why?"

"Because if the records are changing, we need to know what we're comparing."

Miriam looked at the open webpage. The article lay before her. Its words were familiar. Its argument was coherent. Its conclusion remained an admirable expression of cosmicist philosophy. She found herself strangely reluctant to remove it. The work had been done carefully. The research had been intellectually honest. Nothing about its original purpose had been malicious. And yet the possibility remained that the most important thing it had accomplished was not a contribution to literary criticism.

It might have supplied the universe with a history it had previously lacked. Miriam began preparing the archive. The withdrawal took place at 14:06. The live article was replaced with a short notice explaining that the reconstruction had been temporarily withdrawn pending further bibliographical verification. The author retained the original files. Miriam retained her working drafts. Yasmin preserved a copy of the last published version, together with its digital fingerprint and a dated record of the withdrawal. Oliver downloaded the notice from London.

Daniel Harcourt, who had been persuaded to assist with the preservation procedure, retained an independent copy of both versions. The arrangements were deliberately mundane. No elaborate cryptographic ritual. No attempt to isolate the files in a supposedly invulnerable container. No assumption that a particular storage medium would resist an alteration which, if real, appeared capable of reaching into the documented past. They simply recorded what they had done. Then they waited.

Nothing happened.

The British Library's catalogue continued to list the fifteenth-century printing. The Paris manuscript remained available in the French collections. The geological reports did not disappear. The Antarctic specimens remained in their storage cabinets. Gabriel's measurements of the Pacific formation remained unchanged, so far as the available records established. Yasmin examined the consistency model. The removal of the article altered the graph only if the model was explicitly instructed to remove the relationships derived from it. But those relationships were now supported through other sources. The underlying structure remained.

Miriam called her that evening.

"Anything?"

"No."

"Nothing at all?"

"Nothing which we can attribute to the withdrawal."

Miriam looked at the notice on the website.

"Perhaps it takes time."

"Perhaps. But that's an assumption without evidence."

"Yes."

"Don't begin inventing a delay mechanism merely because the expected result hasn't occurred."

"I wasn't."

"You were about to."

Miriam smiled reluctantly.

"Possibly."

They agreed to leave the article withdrawn while continuing the investigation. It seemed a reasonable precaution.

By the following morning, Miriam had begun to feel a little better. The public reconstruction was no longer available in its original form. At least they were no longer actively encouraging readers to assemble its claims into a coherent whole. Then Oliver telephoned.

"There may be a problem."

Miriam closed her eyes.

"What sort?"

"The withdrawal notice."

"What about it?"

"It's being cited."

"By whom?"

"A scholar who has written a short commentary concerning the historiography of the Necronomicon."

Miriam stared at the telephone.

"Already?"

"Apparently."

"That's absurd. The notice was published yesterday."

"Yes."

"And the commentary?"

"Well, that's the difficulty."

"Oliver."

"It's dated 1976."

The commentary was an obscure academic paper concerning the reception of medieval texts in modern literature. Its author had discussed the Necronomicon as an unusual example of a historical work whose existence had been repeatedly questioned by later scholars. There was nothing extraordinary about the central argument. The paper distinguished between the physical existence of medieval manuscripts and the imaginative elaborations introduced by twentieth-century writers. It criticised occult enthusiasts who treated every supposed translation as authentic.

It also criticised literary historians who, in the author's view, had gone too far in treating the entire bibliographical tradition as fictional. Miriam found the argument irritatingly plausible. The paper contained a footnote referring to an episode in which researchers had withdrawn a reconstruction of the Necronomicon after encountering contradictory historical evidence. The footnote did not identify the researchers by name. It described the withdrawal as a characteristic example of the difficulties caused by treating literary invention and historical transmission as mutually exclusive categories.

Miriam examined the publication date. 1976. The article had apparently been written half a century before her own. The reference did not reproduce the wording of the withdrawal notice. It contained no impossible date. No direct citation to a future publication. It was merely a description of an episode which, in its general character, resembled the one now unfolding. That made it more difficult to interpret. Oliver was cautious.

"It could refer to an earlier controversy."

"Of course."

"There have been other arguments concerning the supposed authenticity of the book."

"Yes."

"Then it may have nothing to do with us."

Miriam examined the footnote.

"Does the author identify the withdrawn reconstruction?"

"No."

"Then we have no reason to assume it's ours."

"Agreed."

"Good."

She was almost relieved. Then she reached the final paragraph. The author had written that attempts to deny the Necronomicon's existence often contributed to its apparent authenticity by stimulating more careful examination of the available evidence. Miriam read the sentence twice. It was perfectly sensible historiography. It was also an unpleasantly accurate description of what she had just done. She closed the paper.

"Oliver?"

"Yes?"

"Find the earliest surviving discussion of the controversy."

"Which controversy?"

"Whether the Necronomicon exists."

There was a pause.

"That could be a substantial undertaking."

"I know."

"How far back?"

"As far as the documentary record permits."

"Very well."

She ended the call. Then she opened Yasmin's model. The withdrawal had not removed a constraint. It had supplied a new event. An event which could be situated within a wider history of scholarly disagreement. And that disagreement was apparently acquiring precedents. The deliberate contradictions began two days later. Miriam would subsequently describe the decision as the least defensible act of her professional life. Not because it was irrational to test a hypothesis.

But because she had permitted her desire to reverse the phenomenon to outrun her understanding of its possible consequences. She and Yasmin had spent a long afternoon discussing whether the consistency model could be deliberately destabilised. The question was simple. If the original reconstruction had established a coherent informational structure, could introducing sufficiently serious contradictions undermine it? Yasmin's answer was not encouraging.

"That depends upon what you mean by undermine."

"Make the structure impossible to realise."

"Then you'd need contradictions which couldn't be accommodated through alternative interpretations."

"Such as?"

"Mutually exclusive requirements concerning the same object under the same conditions."

Miriam considered.

"The seventh division."

"What about it?"

"We have several conflicting descriptions of its contents."

"Yes."

"Suppose we introduced a new description which directly contradicted the established reading."

Yasmin frowned.

"That might simply become another textual variant."

"Then something stronger."

"Such as?"

"Suppose we claimed the seventh division never existed."

Yasmin thought for a moment.

"That's a normal historical disagreement. A scholar might argue that the division was a later invention."

"And if we published evidence supporting that conclusion?"

"Then the model would have to accommodate a new scholarly claim."

"Which could reduce confidence in the division's authenticity."

"Perhaps. But confidence isn't the same thing as physical existence."

Miriam looked at the graph.

"Then we need a contradiction which can't be reconciled."

Yasmin shook her head.

"I don't think you've understood the difficulty."

"Explain."

"You're assuming that the historical system must treat every published statement as true."

"Not every statement."

"But enough of them to produce a contradiction."

"Yes."

"Yet the history of scholarship is full of false statements. A false attribution doesn't make a manuscript impossible. It merely requires an explanation for why the attribution was made."

Miriam was silent. Yasmin continued.

"If the process favours coherent histories, it may be able to accommodate your contradiction by making it part of the history of an error."

"Which would require further evidence."

"Possibly."

"And make matters worse."

"Potentially."

Miriam looked towards the window.

"Then perhaps we should test it under controlled conditions."

Yasmin considered.

"Not by publishing arbitrary claims about the physical world."

"No."

"Nor by inventing new entities."

"Certainly not."

"Then what do you propose?"

Miriam thought for some time.

"A contradiction concerning the textual organisation of the book itself."

"Something limited."

"Yes."

"Something already known to be disputed."

"Exactly."

Yasmin remained doubtful. But the proposal had one advantage. It could be designed without introducing a new object or specifying a previously undescribed feature of the universe. They would work with the existing history. They would create an intentionally false textual claim. They would observe whether it was simply rejected, ignored or incorporated into the documentary network. It was a poor experiment by ordinary scientific standards. There was no established mechanism linking the intervention to its supposed effects. The number of possible outcomes was enormous.

And the researchers could not exclude ordinary explanations for any subsequent discoveries. Nevertheless, it was more disciplined than simply publishing contradictions and hoping the universe would cease to make sense. Yasmin agreed to supervise the procedure. Reluctantly. They selected the seventh division. The choice was not arbitrary. Its surviving evidence was incomplete. Its title remained uncertain. The relationship between the Paris manuscript and the London printing was disputed. There were already several competing reconstructions of its organisation. Miriam prepared a deliberately incorrect statement.

It claimed that the division conventionally numbered seventh had originally occupied the sixth position, and that the familiar numbering resulted from a later editorial mistake. The statement was plausible. It did not require inventing a new manuscript. It did not attribute impossible powers to any entity. It merely contradicted the established arrangement of the text. Yasmin required Miriam to record the claim's falsity explicitly in a separate memorandum. They also prepared a second, deliberately mundane false statement concerning an unrelated medieval text.

A third document contained a genuine correction to a minor bibliographical error. The intention was to distinguish between the consequences of an ordinary false claim, an ordinary true correction and the supposed special behaviour of the Necronomicon tradition. The three statements were placed in separate, clearly labelled research files. They were not publicly circulated. Yasmin insisted upon that.

"We have no reason to involve other scholars until we know whether there's anything worth investigating."

Miriam agreed. They recorded the existing manuscript descriptions. They preserved the relevant digital files. Then Miriam inserted the false claim into a private working copy of the reconstruction. The document was saved.

Nothing happened.

They checked the manuscript catalogues. Unchanged. They checked the London printing's collation. Unchanged. They checked the Paris transcription. Unchanged.

The following morning, nothing had changed. After three days, the results remained negative. Miriam felt an unexpected mixture of relief and disappointment. The intervention had not produced any identifiable effect. Yasmin was satisfied.

"Good. That weakens the simplest version of the hypothesis."

"Which version?"

"The idea that any sufficiently coherent statement about the book automatically produces corresponding historical evidence."

Miriam nodded.

"Perhaps the process requires wider dissemination."

"Perhaps. But that's another assumption."

"Or greater consistency."

"Possibly."

"Or more independent constraints."

"Yes."

Yasmin looked at her.

"But I want you to notice something. Every time the hypothesis fails, you find another condition which might preserve it."

Miriam considered.

"You're right."

"Then we need to establish those conditions independently, not invent them after each negative result."

"Agreed."

They terminated the experiment. The false claim remained in the private working document, marked as deliberately incorrect. No publication was made. No unusual consequence was established. For several days, Miriam began to think the affair might finally be approaching the limits of its apparent significance. Then Oliver sent her a manuscript description from Bologna. The manuscript was a late medieval compilation of Latin philosophical and cosmological texts. It contained fragments attributed to several authors, together with anonymous material whose original sources were uncertain.

One section had been identified as a possible witness to the Necronomicon tradition. The identification was not new. A preliminary catalogue from 1989 mentioned it. A later study had examined its relationship to the Paris manuscript. The relevant material was fragmentary. But its arrangement was unusual. The passages conventionally assigned to the seventh division appeared among material associated with the sixth. Miriam stared at the description. She checked the date. Then the catalogue history. Then the published transcription. Everything appeared ordinary.

The manuscript had supposedly been examined years before the City of Dis reconstruction. Its unusual arrangement had been noted in the scholarly literature. Miriam called Yasmin.

"I think we have a problem."

"What kind?"

"The sixth division."

There was a pause.

"Explain."

Miriam described the Bologna manuscript. Yasmin listened. Then she asked, "Did we know about this source before the experiment?"

"No."

"Could we have overlooked it?"

"Certainly."

"Does it establish that the original division was sixth rather than seventh?"

"No. It could represent a scribal rearrangement."

"Then it's consistent with an ordinary textual variant."

"Yes."

"And your deliberately false claim?"

"Corresponds to the interpretation proposed by one of the manuscript's editors."

Yasmin was silent. Miriam continued.

"The editor argues that the conventional division numbering may preserve a later rearrangement."

"Exactly the claim you introduced."

"Almost exactly."

"Publication date?"

"2003."

There was a long pause. Yasmin asked, "Could the claim have originated from a source you encountered during your original research?"

"I don't remember it."

"That's not an answer."

"I know."

"Then we need to trace the source independently."

"Agreed."

Miriam opened the relevant paper. The editor's argument was carefully constructed. It compared manuscript arrangements, textual transitions and several damaged headings. The conclusion was qualified. The author did not claim certainty. He merely suggested that the conventional numbering of the divisions might reflect a later stage of the text's transmission. It was exactly the sort of argument a competent textual historian might make. Miriam felt a sudden despair. She had introduced a contradiction in the hope of destabilising the reconstruction.

Instead, she had discovered an apparently respectable historical tradition in which the contradiction was an ordinary scholarly hypothesis. The false statement had not become demonstrably true. It had become historically plausible. And its plausibility depended upon further documents, further transcriptions and further interpretative disputes. The network had grown. Yasmin examined the updated model. She added the Bologna manuscript as a provisional witness. The number of possible textual relationships increased. But so did the number of independent constraints upon the history of the seventh division.

The manuscript did not destroy the reconstructed tradition. It enriched it. Yasmin looked at Miriam.

"That's what I was afraid of."

Miriam sat down.

"The contradiction becomes a variant."

"Yes."

"And the variant acquires a provenance."

"Apparently."

"And the provenance supports the existence of the book."

"Potentially."

Miriam looked at the graph.

"Then our attempt to undermine the history may have made it more complete."

Yasmin hesitated.

"We haven't established that our intervention caused the discovery."

"I know."

"But within the hypothetical process, that's a possible interpretation."

Miriam closed her eyes.

"An unpleasant one."

"Very."

The unrelated control produced no comparable result. The deliberately false statement about the other medieval text remained unsupported. No newly identified manuscript appeared to corroborate it. No forgotten scholarly controversy emerged. The genuine minor correction behaved exactly as expected. It corrected an ordinary bibliographical mistake. Nothing more. Yasmin considered this cautiously interesting. It suggested, if the effects were real, that the phenomenon was not simply responding to arbitrary statements. But there were other explanations. The Necronomicon material was already the focus of an unusually intensive investigation.

The researchers possessed strong expectations. Their searches were not perfectly independent. The control text belonged to a different scholarly tradition with different levels of documentation and accessibility. It would be easy to mistake a difference in research activity for a difference in physical behaviour. Yasmin insisted that the result remain inconclusive. Miriam agreed. But the episode altered her understanding of the problem. She had imagined that contradiction might reverse the process. Now she saw that contradiction could be assimilated.

An inaccurate description did not necessarily destroy a coherent history. It could become part of that history. An error could acquire an author. A false attribution could acquire a tradition. A rejected reading could acquire manuscript witnesses. The attempt to refute a proposition could itself become evidence that the proposition had been sufficiently important to provoke refutation. And if the underlying process concerned the construction of consistent relationships, then disagreement might provide more opportunities for consistency rather than fewer. Miriam opened the original essay.

It had discussed precisely this feature of the fictional Necronomicon. The book possessed conflicting editions. Missing passages. Disputed translations. Contradictory descriptions. Those imperfections made the fictional history more convincing. A perfectly consistent ancient book would have looked suspicious. A text with a complicated transmission, imperfect witnesses and centuries of scholarly disagreement looked real. Miriam stared at the paragraph. The thought was almost unbearable. She had reconstructed the Necronomicon by treating its contradictions as ordinary features of textual history.

Now those contradictions were becoming ordinary features of an apparently genuine textual history. The research had not merely supplied a coherent description. It had supplied a coherent account of incoherence. And that was considerably harder to undo. Oliver next located a seventeenth-century sceptical treatise which rejected the Necronomicon's Arabic attribution. Its author treated the Latin text as a compilation assembled from unrelated sources, the supposed Greek intermediary as unsupported, and its more extravagant reputation as a product of careless transmission. The argument resembled Miriam's own.

Miriam read the original under the lamp. The ink had faded at the fold; the scholar had placed corrections beside two of his references. His objections were sensible. They also presupposed that a textual tradition existed to be attacked. Refutation did not remove the book. It supplied an earlier history of refutation.

"He wouldn't have liked Lovecraft much," Oliver observed.

"Or us."

Oliver closed the treatise. He had no wish to see it disbound or sampled. The object mattered to him independently of the hypothesis it complicated. "Do you think we should stop correcting the record?"

"I think we should stop expecting correction to reverse it."

He returned the book to its support. It had survived centuries of dispute without needing to win any of them.

Yasmin tried to define a boundary. The original model treated the Necronomicon as a network of constraints. It could not become a theory of everything simply because every unrelated observation was eventually connected to it. She assembled a set of control records from unrelated textual disputes, geological revisions and historical measurements. Their provenance was established in advance; the tests and exclusions were recorded before the searches began.

For three days most controls behaved as expected. Misdated facsimiles were identified. A supposed anticipatory citation proved to be a later edition. An apparent change in a survey diagram resulted from an incorrect plate number. Yasmin felt better each time one of these explanations succeeded.

Then the separation began to weaken. A historical treatment of causal order had been cited in a study of medieval natural philosophy. A control survey included structures resembling those in the Antarctic samples. A supposedly unrelated Latin manuscript belonged to a collection whose owner had also examined the Wormius printing.

None of these links was extraordinary. That was the difficulty. Intellectual history contained many such connections, and once investigators knew where to look, they found them. In the model, however, each link drew another ostensibly independent record into the same connected structure.

Yasmin marked the graph's outer boundary in red. On Monday it enclosed forty-seven nodes. On Thursday, after the records had been checked, it enclosed ninety-one. The numbers were properties of her chosen representation, not measurements of a growing physical process. Still, she disliked how the red line retreated whenever she tried to make it defensible. She phoned Miriam. "I can't give the model an edge without choosing one by fiat."

"Could the affected history extend indefinitely?"

"The mathematics permits it. The evidence doesn't yet demand it. Those are different propositions, and I intend to keep them different."

Miriam heard a drawer close on Yasmin's end of the line.

"Then our corrections?"

"They're becoming another class of records. That's the part we can actually see."

For the first time Yasmin sounded angry. Not frightened exactly, but offended that a study designed to discover a limit had produced only a better catalogue of dependencies. She cleared her desk of the graphs before going home. The next morning she found the archive copy of her earlier seminar pinned above the blackboard. The annotations were familiar now. That disturbed her more than finding them had.

Miriam took down her private reconstruction. The shelf held the earliest typed pages, the first photographs and a folder of notes composed while the book was safely imaginary. She could still feel the paper nick beside her thumb from Saturday's destruction test. The cut had healed; the annotated sheet remained intact in its envelope. The record now described her decision to spare it, an excellent scholarly decision which she had never made.

She had brought the folder to the shredder before she remembered the signed photograph returning. She stood there a moment with the pages against her chest. Destroying them would provide another event for the record to explain, and she had already seen how readily the history acquired a version in which her better judgement prevailed.

The oldest digital draft contained a paragraph she had no recollection of composing: The act of reconstruction may establish relationships whose consequences cannot be confined to the text. The note appeared in an early printout as well. In a drawer she found a second folder labelled, in her hand, Initial reconstruction: errors and objections. She did not open it.

At midnight she telephoned Yasmin. "I took the drafts to the shredder," she said. "I didn't put them through."

"Good. The first attempt established that the records of your actions aren't a fixed point. We don't need to repeat it."

"The correction has already become part of the history. The disputed division has a manuscript variant, the withdrawal a precedent, and the controls are finding connections we hadn't suspected."

Yasmin sounded more tired than frightened. "I keep removing dependencies. The physical branches remain connected. I'm not going to pretend another dramatic gesture would be a controlled experiment."

Miriam replaced the drafts on the shelf. She had thought that refusing to act would preserve one uncontaminated choice. Yet the note explaining her restraint had apparently been written before she considered it. Her handwriting was no longer a reliable witness to her intentions.

There was a long silence. Then Yasmin said, "There's a more serious implication."

"What?"

"If the original reconstruction is no longer necessary, then destroying it won't reverse the apparent history."

"Yes."

"And if the process has become self-supporting through independent relationships, we may not be able to identify any single intervention capable of reversing it."

Miriam closed her eyes.

"I know."

Yasmin continued.

"Perhaps we should stop thinking about reversal altogether."

Miriam opened them again.

"What would you suggest?"

"Understanding the conditions under which the resulting history remains compatible with the world we know."

"And if it doesn't?"

Yasmin did not answer immediately.

"Then we need to establish how the incompatibility would become observable."

Miriam looked towards the shelves. The books stood in their accustomed positions. Her notebooks lay on the desk. The original essay had been withdrawn. Nothing in the room appeared different. Yet the documents on her computer now described an investigation in which she had apparently anticipated the very danger she was attempting to escape. She felt a sudden exhaustion.

"Yasmin?"

"Yes?"

"Do you think it's possible that we never initiated this?"

There was a pause.

"Yes."

Miriam said nothing. Yasmin continued.

"The mathematical model doesn't privilege that interpretation."

"Then the essay may have been necessary for the larger history to become coherent."

"Possibly."

"Not the cause of the alteration."

"Not necessarily."

"Merely one of its consequences."

"Yes."

Miriam looked at the warning in her earliest draft. The sentence appeared to describe knowledge she had not possessed when the document was supposedly written. She wondered whether the research had been inevitable. Not through destiny. Not through the intention of some supernatural intelligence. But because the emerging structure of reality required a particular sequence of events, and her investigation happened to be one way of satisfying those requirements. The thought was appalling.

It suggested a universe in which the ordinary distinction between discovery and creation might be unstable. A universe in which human intentions were not privileged causes but local features of a larger consistency structure. She had always rejected the idea that reality existed for humanity's benefit. Now she was beginning to suspect something more radical. Reality might not even possess a particularly stable distinction between what human beings caused and what they merely participated in.

The next morning, Oliver sent her a final document. It was a catalogue description of a seventeenth-century work entitled Contra Necronomicon. The author was a minor scholar whose name appeared in several contemporary bibliographies. The book was described as a systematic critique of the supposed Arabic origins of the Necronomicon. Its central argument was that the work's historical claims had been misunderstood by successive generations of readers. The surviving Latin text, the author suggested, was a compilation of speculative philosophy, distorted geographical traditions and misunderstood natural observations.

The attribution to Alhazred was doubtful. The supposed chronology was unreliable. The reputation for occult knowledge was undeserved. Miriam read the description. It was almost painfully familiar. The seventeenth-century author had apparently undertaken the same project as the City of Dis reconstruction. He had attempted to separate genuine textual evidence from accumulated fiction. He had examined the contradictions. He had questioned the attributions. He had proposed a more coherent account of the surviving material. And his work had become part of the Necronomicon's historical tradition. Miriam telephoned Oliver.

"Have you found a copy?"

"Not yet."

"Does the catalogue identify its contents in detail?"

"Only partially."

"Any independent references?"

"Several."

"Then it may be genuine."

"Apparently."

Miriam looked at the title.

Contra Necronomicon.

Against the Necronomicon. She considered the irony. A work intended to expose the book's supposed falsehoods had become another document establishing its historical existence. The refutation had survived. The object refuted had survived. The disagreement possessed a history. And that history was now more coherent because the disagreement had occurred.

"Oliver?"

"Yes?"

"I think we should stop trying to correct it."

There was a long silence.

"Why?"

"Because every correction becomes another part of the record."

"Then what do we do?"

"We investigate the physical consequences."

"The biological material?"

"Yes."

"The Pacific formations?"

"Yes."

"And the geometrical measurements?"

"Especially those."

Oliver hesitated.

"Do you think the book is finished?"

Miriam considered the question.

"No."

"Why not?"

"Because its history is still acquiring independent support."

"Then perhaps the text itself will become complete."

"Perhaps."

"And the missing sections?"

"Possibly."

He was silent. Then he asked, "What happens when there's nothing left to reconstruct?" Miriam looked at the shelves.

"I don't know."

But she had begun to suspect. The original research had supplied a coherent description of an imaginary book. The emerging historical record was supplying the objects, events and relationships required to make that description physically intelligible. The process might continue until every relevant constraint had been satisfied. Until the book possessed a complete history. Until its supposed observations had acquired real referents. Until the universe contained everything necessary to make the resulting account consistent. And then there might be no anomaly left to detect.

Only the world as it had always been.

That afternoon, Miriam wrote a final correction. She did not publish it. She did not send it to Oliver, Yasmin or the author of the essay. She wrote it in her private notebook, on a page whose number she recorded separately. It consisted of a single sentence.

The Necronomicon does not exist.

She looked at the words. They were plainly false according to the present physical evidence. A book bearing that title existed in London. Other manuscript witnesses appeared to exist elsewhere. Its textual tradition was supported by an extensive documentary record. But the sentence expressed a different claim. It asserted that the historical object was not independent of the fictional reconstruction. That it had not belonged to the world before the research. That its apparent past was an acquired property rather than an ordinary sequence of events.

The problem was that none of those claims could be demonstrated through the surviving evidence. Miriam stared at the sentence. Then she drew a line beneath it.

She wrote:

It did not exist before we described it.

She considered the statement. It was closer to her original hypothesis. But it also contained an assumption she could no longer justify. It privileged the research as the cause. It presumed a previous history in which the book was absent. It assumed that the distinction between before and after remained meaningful at the level where the supposed alteration occurred. She crossed out the sentence.

Then she wrote:

We remember a world in which it did not exist.

That, at least, was defensible. She and Oliver had memories incompatible with the documentary record. Those memories might be mistaken. They might be incomplete. They might preserve traces of an earlier history. But their existence was itself an observation. Miriam looked at the page. Three statements. The first apparently false. The second unsupported. The third carefully limited to what could actually be claimed. She closed the notebook. For the first time in several weeks, she felt that she had written something intellectually honest. Then she opened it again.

Beneath the three statements was a fourth. It was in pencil. The handwriting was hers. She did not remember writing it.

The distinction will not survive the completion of the history.

Miriam stared at the words. She checked the paper. The pencil mark was faint. The pressure varied naturally. There was no indication that the sentence had been added through any unusual physical process. She could have written it earlier. She might simply have forgotten. The explanation was entirely plausible. But the sentence expressed precisely the conclusion towards which the investigation had been moving.

Once the historical record became fully coherent, the distinction between a world in which the book had always existed and one in which it had acquired a past might become impossible to establish. The evidence would support only the completed history. The alternative would survive, if at all, as an unsupported memory. Miriam closed the notebook. She had attempted to correct the past. The past, apparently, was acquiring explanations for her corrections. There was nothing more she could do by editing the text.

Yasmin found the next development in an entirely different place. She had been reviewing the independent timing experiments when she noticed that the oldest apparent anomaly had acquired a new interpretation in the scientific literature. A paper published several years earlier proposed a geometrical model capable of reproducing part of the unexplained residual. The model was not identical to hers. It made different assumptions. Its predictions were limited. But the mathematical structure was similar enough to warrant comparison. Yasmin examined the references.

The paper had been cited by several later studies. One discussed relational approaches to spacetime. Another examined the interpretation of timing networks. A third proposed that certain measurements ordinarily attributed to instrumental error might, under highly restricted conditions, reveal limitations in the geometrical assumptions used to interpret them. None of the authors claimed to have discovered impossible geometry. The work was serious, cautious and mathematically sophisticated. Yasmin found it deeply troubling. She remembered searching the same literature before developing her model. She remembered finding no directly comparable treatment.

Now the relevant papers appeared in ordinary databases, with conventional publication histories and citations. She could not establish that they had been absent. She had only her memory. The papers were there. Their equations were coherent. Their authors appeared to be genuine researchers. The theory was becoming part of a recognised scientific discussion. Yasmin updated her model. Several previously unsupported assumptions could now be replaced by independently published mathematical results. The network became more constrained. The speculative geometry acquired a scientific pedigree.

And Yasmin began to understand something Miriam had already recognised. The historical process, if it existed, was not concerned with preserving the originality of their discoveries. It was making their discoveries unnecessary. The physical evidence would eventually possess enough independent support that no one would need to refer to the original anomalous investigation. The Necronomicon would become an obscure historical text. The Antarctic material would become an unfamiliar but legitimate subject of palaeobiological research. The Pacific formations would become complicated geological structures.

The spatial measurements would become part of a developing physical theory. Each would possess its own history. Its own specialists. Its own disputes. Its own ordinary explanations. And the connection between them might become no more than an eccentric episode in the history of interdisciplinary scholarship. Yasmin looked at the graph. The model had begun with a fictional book. Now the book was only one node among thousands. She attempted to remove it. Not physically. Not historically. Only from the mathematical representation.

She deleted the central node and recalculated the network using the remaining observational relationships. The model remained connected. Some constraints weakened. A few relationships became uncertain. But the larger structure survived. Yasmin stared at the result. She repeated the calculation. Then she removed the original essay as a source. The structure remained. She removed the literary references to Lovecraft. Again, the structure remained. She removed every connection which depended solely upon the fictional Necronomicon. The network became smaller. But its major physical branches persisted.

The biological observations retained their relationships. The geological observations retained theirs. The mathematical model of spatial relations remained independently supported. Yasmin sat back. The original reconstruction was no longer necessary. The phenomenon, if real, had become independent of the information which appeared to have initiated it. There was no longer any reason to assume that suppressing the book would alter the physical world. The world now possessed the evidence required to explain itself.

And the explanation was becoming increasingly difficult to distinguish from an ordinary history of scientific discovery. She telephoned Miriam.

"I've removed the book."

There was a pause.

"What?"

"From the model."

"And?"

"The structure survives."

Miriam said nothing.

"I removed the essay as well."

"Yes?"

"Then the literary references."

"And?"

"The physical relationships remain."

There was a long silence. Finally Miriam said, "Then we can't undo it by correcting the text."

"No."

"Or by withdrawing the article."

"No."

"Or by destroying the manuscripts."

"Not according to the model."

Miriam closed her eyes.

"Then the book isn't necessary anymore."

"Exactly."

She sat in silence. Yasmin continued.

"And there's something else."

"What?"

"The model is becoming more coherent as the independent physical evidence accumulates."

"That sounds familiar."

"Yes. But I've begun to suspect that we've misunderstood the direction of the process."

Miriam opened her eyes.

"Explain."

"We've been imagining that the book is becoming real and causing the world to acquire the things it describes."

"Yes."

"But the physical relationships can now be modelled without the book."

"Which means?"

"It means the larger structure doesn't require the book as its cause."

Miriam stared at the wall.

"Then perhaps the book is merely one way the structure becomes intelligible."

"Exactly."

"An interpretation."

"Possibly."

"A historical representation of something much larger."

"Yes."

Miriam thought of the seventh division. The distinction between objects and the relations which constituted them. The error of treating a geometrical relationship as an entity. The tendency to interpret unfamiliar physical structures through the language of gods, monsters and forbidden knowledge. Perhaps the Necronomicon had never been the important part. Perhaps it was merely the first human description to become sufficiently coherent for the underlying relationships to be recognised.

Or perhaps the process had always been larger, and the research had simply supplied one of the local histories through which it could be expressed. Miriam looked towards the window. The evening was approaching. Nothing in the familiar world appeared disturbed. She found that increasingly difficult to regard as reassuring. The final message arrived from Lena Voss shortly before midnight. It contained no reference to the Necronomicon. No suggestion of retrospective history. No speculation concerning the nature of existence. It was a short scientific update.

We've completed the first independent comparison of the Antarctic specimens.

The structures in RB-17/642 and RB-04/119 remain closely similar under several measures of network organisation, although common geological formation processes have not been excluded.

The stratigraphic evidence is consistent with considerable antiquity, but further work is needed to establish the origin and age of the structures themselves.

There is also a new complication.

The earlier collection appears to contain references to additional material recovered during drilling programmes in the 1960s.

I've located the relevant accession records.

If the provenance is correct, the distribution may be considerably more extensive than we thought.

I'm checking the original samples before drawing any conclusion.

Lena

Miriam read the message. Then she opened the attached map. It showed the locations of the known Antarctic samples. There were several points. Some lay hundreds of kilometres apart. The newly identified records extended the distribution further. The map was provisional. The interpretations were cautious. Nothing on it established the existence of an unfamiliar organism. Nevertheless, the pattern was developing. The material now possessed a longer research history. A wider distribution. Additional specimens. Independent observations. And increasingly detailed explanations of its unusual organisation. Miriam looked at the map.

She thought of the shoggoths. The supposed artificial servants of a pre-human civilisation. The creatures whose bodies possessed no stable form. The fictional beings whose descriptions might have arisen from the misunderstanding of a natural system. The evidence was becoming more substantial. But it did not resemble the arrival of monsters. It resembled the gradual discovery of an ancient biological history. One which, if genuine, had existed long before humanity. And one which might remain entirely indifferent to whether human beings recognised its significance. She closed the attachment.

The original essay was withdrawn. The private corrections had failed. The deliberate contradiction had acquired a possible manuscript witness. The sceptical refutations had become part of the book's history. And the physical evidence continued to develop independently. Miriam had attempted to remove the initiating description. The larger structure no longer required it. She opened her notebook. At the bottom of the page containing the three corrections, she wrote one final sentence.

The reconstruction is no longer necessary.

She read it. Then she crossed out reconstruction. She replaced it with researchers.

The sentence now read:

The researchers are no longer necessary.

Miriam stared at the words. It was not a prophecy. It was not a threat. It was a conclusion drawn from the increasingly independent history of the evidence. The book had acquired a past. The past had acquired sources. The sources had acquired physical referents. The referents were acquiring explanations. And the explanations were becoming part of a world in which no extraordinary intervention would eventually be required. The universe would not need Miriam Vale. It would not need Oliver Stroud.

It would not need Yasmin Qureshi, Lena Voss or Gabriel Sato. It would not even need Lovecraft. Their work, their memories and their disagreements might survive as minor episodes in the history of a phenomenon which had never required human understanding. Or they might not survive at all. The distinction, Miriam realised, would become increasingly difficult to establish. She closed the notebook. For the first time, she no longer wished to correct the essay. The attempt had been based upon an assumption she could not justify.

That the universe possessed a preferred history in which humanity's familiar world existed, and that the researchers might restore it by removing an error. But the consistency model contained no such preference. It did not distinguish between a history which preserved human civilisation and one which did not. It did not assign particular value to the continuity of individual minds. It did not privilege the biological arrangements from which the investigators had emerged.

And it certainly did not regard a corrected scholarly article as more important than the physical relationships of an entire universe. Miriam turned out the light. The study remained familiar. The books stood on their shelves. The door opened onto the hallway. The computer held its ordinary collection of documents. Nothing had changed in any way she could demonstrate. Yet she had the increasingly uncomfortable conviction that the world was approaching a state in which nothing would ever have changed.

A state in which the Necronomicon had always existed. In which the Antarctic structures had always occupied their ancient deposits. In which the Pacific formations had always possessed their unusual geometry. In which the spatial relationships now troubling Yasmin had always belonged to the structure of reality. And in which the researchers' memories of a different history would eventually become no more than inexplicable private convictions. Miriam thought of the first question she had asked. Why did the book exist?

She no longer believed that was the important problem. The important problem was why the universe required a history in which it existed. And what else that history would require. She looked once more at the notebook.

The researchers are no longer necessary.

She did not cross out the sentence. There was no point. Somewhere in the expanding network of records, measurements and physical relationships, the universe was acquiring a coherent account of the things the Necronomicon described. The process no longer needed the book's human reconstruction. It might never have needed it as a cause. And if the history became sufficiently complete, there would be no remaining distinction between a universe which had acquired those things and one in which they had always existed.

The correction had failed because the error was no longer localised. It was distributed across reality. Or perhaps there had never been an error. Perhaps the familiar world had merely been an incomplete description. And the process now underway was not an intrusion into that world, but the completion of a structure whose eventual form had never been obliged to include humanity. Miriam went upstairs. She slept badly. In London, the Necronomicon remained in its protective enclosure. In Paris, the seventh division possessed an increasingly coherent textual history.

In Antarctica, an ancient material system was acquiring a more extensive distribution through the geological record. Beneath the Pacific, the measurements of an unfamiliar spatial structure continued to resist complete reconciliation. And in Cambridge, a mathematical model had ceased to require the literary object from which the investigation had begun. Nothing was arriving. Nothing was being summoned. Nothing was waiting for permission. The history was simply becoming complete. And whatever that completion required would, when the process was finished, have always been there.

Chapter IX 9 of 10

The Arrival That Has Already Happened

Lena Voss arrived carrying two microscope slides and a letter from a journal editor.

She had brought the slides in a rigid case, cushioned with foam. The letter was folded into a jacket pocket. She placed the case on Yasmin Qureshi's desk first, as though the order mattered.

"The earlier specimen isn't an illustration," she said. "It's a specimen. If we talk about it, we talk about the preparation, the collection record and what the mineral fabric permits us to infer."

Miriam sat opposite her. Gabriel Sato was joining them by video from his laboratory, where it was already evening. He had sent a map stripped of place names and a table of measurement residuals. Yasmin had agreed to compare the datasets without showing either specialist the other's preferred interpretation.

Lena opened the case. Each slide contained a thin section of dark Antarctic rock, so unremarkable in ordinary light that Miriam had to look twice to distinguish the prepared surfaces. One was RB-17/642, the newer specimen; the other came from the earlier collection and bore an old accession label, RB-04/119.

"They are from different drilling programmes," Lena said. "The chain of custody for the older slide now looks sound. That doesn't make its contents biological."

"What have you established?" Yasmin asked.

"A related organisation appears under several independently chosen segmentation methods. We can reproduce parts of it from the raw imaging data. We cannot reproduce every apparent boundary, and the chemical evidence still permits a mineral origin."

She opened a laptop and displayed the reconstructed networks. In one specimen the chambers were flattened and irregular; in the other they were narrow and branching. Their outlines differed considerably. When the shapes were suppressed and only their connections retained, several correspondences became visible.

Lena placed a finger against the screen. "Here. This isn't an organ. It isn't a nerve or a vessel merely because the graph looks familiar. It's a relation between regions of preserved material. That's all I will name it."

Gabriel's voice came through the speaker. "That's more restraint than I encounter in most interpretations of the ocean floor."

He shared his survey reconstruction. The Pelagic Complex remained a geological formation whose ridges and depressions were largely explicable through tectonic and volcanic processes. The old suggestion of a submerged city had no supporting evidence. Several apparent straight edges had disappeared when the original soundings were reprocessed.

"So the city is gone," Miriam said.

"It was never there," Gabriel replied. "An interpolation routine supplied most of its streets. The sea can keep its geology without inheriting our architecture."

What remained was a smaller problem, and for that reason a more useful one. Three measurement regions, A, B and C, could be reconciled pairwise within the expected uncertainties. A closed comparison of their measured relationships retained a structured residual. Corrections for several known errors had reduced it but had not eliminated it in every dataset.

Gabriel displayed the control cases. Several regions unrelated to the Pelagic Complex showed no such pattern. Others showed larger discrepancies traceable to poor navigation or incompatible processing conventions. He had published those negative results alongside the positive one.

"We can describe the residual," he said. "We cannot name its cause. If it's an unrecognised common error, an independent survey should eventually expose it. I would prefer that outcome."

Yasmin set Lena's graph beside Gabriel's. Their sizes were different, their variables different, their physical origins unknown. A common mathematical description could fit selected relationships from both, but a flexible network model could do the same for many unrelated natural systems. Yasmin displayed the control comparisons without comment.

The Antarctic and Pacific structures were not identical. They did not prove a shared physical process. What they shared, at present, was a resistance to classification by the visible outlines of their constituent parts. Lena looked at the curves and shrugged.

"A useful mathematical description," she said. "Not a biological conclusion."

"And not a new ocean," Gabriel added. "If your model has a place for bad sonar, I might begin to trust it."

Yasmin dismissed the coloured overlays and left only the measurements on screen. It was a small act, almost comic after the disproportionate expectations which had assembled around the investigation. Miriam was grateful for it. For an hour the room contained nothing more mysterious than two prepared slides, imperfect surveys and people who knew how little those materials permitted them to say.

Then Lena took the journal letter from her pocket.

"The editor wants a different title."

Miriam looked up. "Why?"

"We called the work a preliminary investigation of an unclassified Antarctic structure. The referees say the organisation has been discussed for decades. They want Reassessment of the Antarctic reticulate series."

"Do you disagree?" Yasmin asked.

"The references exist. They're not all reliable, and some describe material we haven't inspected, but I can't honestly call this the first recognition. The paper I thought I was writing has acquired a literature before I've finished it."

She had already marked the altered title on the proofs. The word reassessment annoyed her. It suggested that she had returned to a problem she remembered encountering as a new one. But she would not pretend to originality merely to rescue a private recollection. The specimen was real enough to require that concession, even if it proved in the end to be an ordinary mineral system.

Lena closed the slide case. "There's the cost of this, if you want one. I've lost the right to describe the finding in the terms in which I found it. That doesn't make the rock alive."

Miriam looked at the case's worn corner, then at the folded letter. Both were perfectly ordinary. Each belonged to a history in which the research had been proceeding for longer than Lena recalled.


After Lena left, Yasmin showed Miriam the current version of her own technical note.

Closed-Loop Residuals in Timing-Transfer Networks: Limits of a Geometrical Interpretation. The title remained modest, the claims narrow. It reported the laboratory observations without attributing them to a new physical geometry. Daniel Harcourt had insisted that the calibration uncertainties remain prominent. Yasmin had complied gladly.

The departmental archive now contained an earlier seminar abstract in which she had apparently introduced the central mathematical construction years before Miriam contacted her. Priya Sen remembered the seminar. Daniel remembered hearing about it. Yasmin had the programme, the room booking and a set of annotated slides in her handwriting. Her own account of developing the model over the previous weeks was unsupported.

"I can still reproduce the calculation," she said. "That's my consolation. If I discovered it in the wrong year, it remains the same mathematics."

"And the note?"

"It's a reanalysis now, not a new result. I changed the departmental description. Priya was right to query the provenance."

Miriam studied her face. Yasmin had lost weight, and there was a faint mark where the frame of her glasses rested against her nose. She looked more irritated than frightened. Her anxiety had a specific professional shape: she might be able to defend the results, yet no longer establish which observations had led her to them.

"Are you going to publish?" Miriam asked.

"Not the geometry. The timing residuals, perhaps, when Daniel is satisfied with the controls. A theory which explains every record by revising the history of its own discovery isn't yet an experimental theory."

She closed the article and opened the consistency model. Over several weeks she had removed speculative relationships, downgraded weak archival claims and discarded any inference supported only by the original literary reconstruction. The graph was smaller and harder to satisfy. Its strongest physical branches still stood independently: the London printing's demonstrable material properties, Lena's observations, Gabriel's survey residuals, and the laboratory timing measurements.

"It was easier when I could blame the book," Yasmin said.

Miriam smiled. "Was it?"

"A book has a location. One can ask who printed it, what paper they used, whether the ink belongs to the period. This has no single object left to blame."

She selected the node representing the City of Dis essay and removed it. The network changed, but the physical observations remained connected through other relationships. She restored it and removed the fictional descriptions attributed to Lovecraft. The same essential measurements survived. Neither deletion was an experiment upon reality; it was a test of which dependencies the mathematical account actually required.

"The book is no longer a necessary assumption," Yasmin said. "Not for these branches. It may have been the way we noticed their relationship. That isn't the same as being their cause."

Miriam thought of the signed photograph they had fed into the shredder, then found intact with an annotation explaining why it had been preserved. She remembered the sound of the motor catching. The cut beside her thumbnail had healed without leaving a mark. The surviving photographs showed them destroying an unsigned copy. She could prove none of what she remembered.

"We thought we'd made the book real," she said.

"We still don't know that anything became real."

"But the history we're finding doesn't begin with our research."

"No. That was an assumption imported from the order of discovery. The mathematical constraints don't select a privileged first event."

Yasmin turned the screen so Miriam could see it properly. The network's longest paths reached into human intellectual history, geological time and the mathematical assumptions by which events were ordered. Nothing about their representation required the essay to have preceded the objects it described.

Miriam looked for the point at which the supposed entities would enter the network. There was none. The Antarctic specimens, if the material proved biological, belonged to ancient deposits. The Pacific geometry, if the residual were eventually shown to have physical significance, would concern relationships already presupposed by measurements. The medieval book, if authentic, belonged to a long textual tradition. The interpretations remained uncertain; the histories were ordinary in form.

"There was no arrival," Miriam said.

Yasmin looked at the graph. "No arrival that the model can identify. A complete history would contain the conditions for the things it described. It wouldn't need a moment at which they crossed from one world into another."

Miriam remembered the note from the seventh division: They have already always existed.

For once neither woman tried to improve the translation.


Oliver telephoned that evening, principally to complain about a request for further invasive examination of the London printing.

"Someone wants to remove a fragment from the new gathering for destructive analysis," he said. "The conservation case hasn't been made. We have photographic methods, and the binding's already fragile."

"Have you refused?"

"For now. It's a book, Miriam. An old, damaged book. I refuse to let its value be reduced to the possibility that a famous fiction writer may have borrowed its title."

He sounded like himself. Miriam could almost see him placing the protective tissue over a leaf, checking the support under its spine, writing a note in an ordinary conservation register.

"Do you remember your first email to me?" she asked.

"Certainly. I objected to your treatment of the printing as fictional."

"And do you remember telling me you'd taught the Necronomicon as an imaginary bibliography?"

A pause. "I remember that you've said I did."

"But not saying it."

"No. I've been through the lecture notes. I discussed its genuine transmission and Lovecraft's subsequent fictional elaboration. What else would I have taught?"

Miriam had no answer which did not privilege her memory over his records. Oliver continued, explaining which repair materials would be acceptable and why repeated opening would do more damage than another year's scholarly dispute. He cared about the object before him. His professionalism had become, in the completed history, evidence that the object had always deserved such care.

After the call Miriam stood in the kitchen without switching on the light. She wanted to hear a voice unconnected with the investigation, someone who had not read the book, handled the manuscripts or seen the model. She thought of Ruth Bennett and the rose stem caught in the garden gate.

She went next door.

Ruth answered in slippers, holding a folded tea towel. The conversation began with ordinary politeness. The fence needed another coat of paint; a parcel had been delivered to the wrong house; the weather might improve. Miriam waited until Ruth put down the towel.

"Do you remember a Tuesday evening some weeks ago? You were having trouble with the gate, and I told you I was helping to write the history of a book that didn't exist."

Ruth considered. "The old book in London?"

"No. I said it was imaginary."

"Did you? I remember you saying you'd been looking into a rare old book. I said you'd need to be careful with it. They're expensive, aren't they?"

"You said it would save me the expense of buying it."

Ruth smiled uncertainly. "Perhaps. I don't remember that. I remember the rose catching my sleeve, though. You were good enough to hold the gate while I freed it."

Miriam looked towards the gate. The latch was still stiff. The rose had been tied back with green garden twine. She remembered holding it open. She remembered Ruth's exact little joke about an imaginary book requiring no shelf space. Those memories felt equally secure. Ruth possessed only one of them.

"I'm sorry," Ruth said. "Is it important?"

"I thought it might be."

Ruth offered tea. Miriam declined, then thanked her for remembering what she could. Walking back through the short gap between their houses, she touched the latch. The metal was cold and slightly rough beneath her fingers. It moved exactly as it always had. The garden offered nothing which could testify to the words spoken beside it.


Miriam returned to the study and placed the oldest notebook on the desk. She opened it at the page where she had written, in heavy black ink, No historical original. Beneath it stood the accession citation she could not remember adding. On another page was the failed destruction experiment, now documented as a decision to preserve the signed print. There were notes on the Paris witness and on the fragment of the seventh division she had refused to complete. Each had acquired a respectable history.

The notebook remained the one she had owned for years. Its corners were rubbed; the binding opened naturally at the pages she had used most often. It bore the pressure of her hand. But it no longer supported the sequence of research she remembered.

She opened the archived article. The Book That Should Not Exist. The argument was still recognisably hers in the passages she had checked. Lovecraft's imaginary chronology still existed in the correspondence. The physical book existed in London. Both histories now seemed capable of being told without embarrassment, provided one accepted that Lovecraft had elaborated a tradition which earlier scholars had already studied.

She could not remember accepting that account. Oliver did. The author of the article had withdrawn the original reconstruction. Yasmin's records placed her mathematical work years earlier. Lena's paper had become a reassessment of known material. Gabriel's geological model required no lost city. Ruth remembered Miriam discussing a real antique.

There was no conspiracy among them. Nothing had instructed the investigators to forget. Each had revised an account of the past in the direction of evidence that appeared independently sound. The resulting histories had room for scholarly disagreement, but increasingly little room for a world in which the Necronomicon had never existed.

Miriam put the notebook beside the bed and went upstairs. She did not try to resolve the distinction one more time. The house was quiet. From the road came a brief wash of tyre noise and then silence.

She lay down remembering a universe in which the book had been invented.

By morning, she might not.

Chapter X 10 of 10

A Minor Bibliographical Correction

The letter from the British Library arrived on a Tuesday. It was not an email. Someone had sent an actual letter, printed on institutional stationery, enclosed in a cream-coloured envelope and delivered with the ordinary morning post. Miriam Vale found this disproportionately disturbing.

There was no reason for it. The Library maintained physical correspondence with researchers, particularly concerning conservation work and the examination of restricted material. The envelope bore a conventional postal mark. Her address was correct. The letter had been signed by Oliver Stroud. Everything about it was ordinary. She opened it.

Dear Miriam,

During a recent examination of the Wormius printing, we discovered an additional gathering which had previously been mistaken for material associated with a later repair.

The leaves are extremely fragile, and their relationship to the original printing remains uncertain.

The text appears to contain a conclusion to the seventh division, followed by a short bibliographical appendix.

There is also a reference which I think may interest you.

I would appreciate your opinion before we proceed with any formal identification.

Best wishes,

Oliver

A second sheet contained a photograph. It showed a printed page from the Necronomicon. The typography was consistent with the rest of the volume. A narrow portion of the lower margin had been lost, and several words were obscured by damage. The text was Latin. Near the bottom, however, was a short reference written in an abbreviated form resembling a bibliographical citation. Miriam read it. Then she read it again. The reference concerned a modern study of the Necronomicon's historical status. The title was given in English.

The Book That Should Not Exist.

Beneath it was a date. 2026. Miriam sat down. The fifteenth-century printing now apparently contained a reference to the essay whose research had, according to her original hypothesis, caused the book to acquire a physical existence. A book printed around 1490 cited a publication from 2026. She examined the photograph. The reference was not handwritten. It was part of the printed text. The letterforms, spacing and apparent impression corresponded to the surrounding material. There was no sign of an obvious modern insertion. But appearances were not evidence enough.

She knew that. A skilled forger could produce remarkably convincing material. A later gathering could have been inserted into an old binding. A photograph could have been manipulated. There were ordinary possibilities. She began listing them. Then she stopped. The problem was no longer the existence of a single impossible citation. They had encountered those before. The problem was its position. The reference appeared in a bibliographical appendix attached to the newly recovered conclusion of the seventh division. It was not presented as a warning.

It was not marked as a prophecy. It was treated as an ordinary source. Miriam examined the surrounding Latin. The appendix discussed competing accounts of the book's origin. Some attributed it to an Arabian author. Others treated it as a compilation assembled from older traditions. Still others regarded portions of the text as later inventions imposed upon an otherwise genuine historical work. The discussion was familiar. It reproduced, in outline, the scholarly controversies which Miriam had spent weeks investigating. Then came the modern reference.

It appeared to identify the City of Dis reconstruction as one of the authorities relevant to establishing the book's historical relationships. The author of the appendix, whoever that might have been, had treated a twenty-first-century essay as evidence concerning a fifteenth-century printing. Miriam stared at the page. She had spent most of the investigation trying to determine whether the book's existence depended upon the reconstruction. Now the book appeared to acknowledge that dependence.

But in a form which made the distinction between cause and consequence almost impossible to maintain. The essay described the book. The book cited the essay. Each appeared within the other's history. There was no longer an obvious place to begin. Yasmin Qureshi arrived at Miriam's house shortly after noon. She examined the photograph without speaking. Then she asked for the original file. Miriam supplied it. Yasmin checked the metadata. It appeared to have been created during the Library's recent examination. The image was high resolution.

The lighting was appropriate. The page showed the expected signs of age and handling. Nothing established that the printing was genuinely fifteenth-century. Nothing established that the citation had been present before the photograph was taken. Yasmin put down the laptop.

"This isn't sufficient."

"I know."

"We need the physical leaf."

"Oliver is arranging an examination."

"And the provenance of the gathering."

"Yes."

"And an independent assessment of the printing."

"Naturally."

Yasmin looked at her.

"Good."

For several minutes neither spoke. Then Miriam asked, "What if it's genuine?" Yasmin considered.

"Then we have a document whose physical age is incompatible with the ordinary chronology of the information it contains."

"We've encountered that before."

"Yes."

"But this is more direct."

"More suggestive, certainly."

Miriam looked at the photograph.

"The book cites the article."

"Apparently."

"And the article reconstructs the book."

"Yes."

"Then the relationship is circular."

"Only if both references are causally relevant. A circular documentary relationship isn't necessarily a causal loop."

Miriam smiled faintly.

"I was hoping you might have become less exacting."

"Under the circumstances, I'd regard that as a serious deterioration."

She opened the latest consistency model. It no longer possessed a central node. The original Necronomicon reconstruction was one small component within a much larger network of documentary, physical and mathematical relationships. The biological and geological branches were now supported through substantial independent evidence, though their more extraordinary interpretations remained unproved. The spatial model had developed into a serious, if highly speculative, investigation of relational geometry. Several of the original chronological inconsistencies remained unresolved. Others had acquired ordinary explanations.

And the memories of the investigators had become increasingly unreliable as independent evidence for any earlier state. Yasmin added the new reference. The model changed slightly. Then it settled. Miriam looked at the screen.

"That's all?"

"What did you expect?"

"I don't know."

"The new citation adds another relationship. It doesn't fundamentally alter the mathematical structure."

"But it's an explicit reference from the supposed past to the future."

"Yes."

"And the model accommodates it."

"Because we've constructed a model which permits time-symmetric constraints. That doesn't establish that nature does."

Miriam considered.

"Then what would establish it?"

"A physical observation which cannot be explained under the ordinary causal framework and which the alternative model predicts independently."

"Have we got one?"

"Not yet."

There was a long silence. Then Yasmin said, "But I think the question may be becoming less useful." Miriam looked at her.

"Which question?"

"Whether the book became real."

"Why?"

"Because the physical history now possesses enough independent support that the distinction between an acquired past and an ordinary past is becoming observationally inaccessible."

"I know."

"And the model doesn't privilege the essay as the initiating event."

"Yes."

Yasmin looked at the graph.

"So perhaps we should stop asking what caused the book to exist."

"What should we ask?"

"Why the larger structure contains the particular relationships it does."

Miriam examined the screen. The graph represented an immense network of dependencies. Some were textual. Some physical. Some mathematical. A few concerned the basic relationships through which space and time were defined. The Necronomicon was no longer particularly important. It was simply one object whose history participated in the larger structure. Miriam looked at Yasmin.

"Have you found a way to determine which history is fundamental?"

Yasmin shook her head.

"No."

"Not even in principle?"

"Under the current model, no. The distinction isn't identifiable from the evidence."

"Then we can't establish whether the book entered our universe."

"No."

"Or whether our universe entered a history in which the book existed."

Yasmin hesitated.

"That's an evocative way of putting it."

"But mathematically?"

"Both descriptions may correspond to the same complete relational structure."

Miriam looked again at the photograph. The modern essay cited by the medieval book. The medieval book reconstructed by the modern essay. A perfectly circular history. She felt, unexpectedly, a small measure of relief. At least the problem was no longer centred upon her. The examination took place on Thursday. Oliver had arranged for the newly identified gathering to be studied under controlled conditions. Helen Mercer, the conservator, was present. So was a specialist in early printing who had no prior involvement in the investigation.

His name was Dr Adrian Bell. He was a man of considerable professional experience, and he found the purported fifteenth-century reference to a twenty-first-century publication more irritating than astonishing.

"Obviously later," he said.

Miriam looked at him.

"Obviously?"

"Yes."

"On what grounds?"

"Because the alternative would require us to abandon the ordinary chronology of printing."

"That's a reason for suspicion, not an examination of the object."

Bell looked at her.

"Fair."

He adjusted the lamp. They examined the leaf. The paper appeared consistent with the main volume. The type was similar. The printing technique was compatible with the assigned period. There were minor differences. But those could reflect the use of a separate setting of type or a different printing session. The gathering's relationship to the rest of the book was difficult to establish. Its outer leaves bore signs of having been folded and reused as binding material. Several edges had been trimmed. There was evidence of later repair.

Bell examined the fibres. Then the ink. Then the impression of the type. He made notes. Finally he sat back.

"I don't like it."

Oliver looked at him.

"Why?"

"Because there's no obvious modern component."

"Does that establish authenticity?"

"Certainly not."

"Then what does it establish?"

"That if this is a modern insertion, it's an unusually sophisticated one."

Miriam considered.

"Could the paper be genuinely old but the printing recent?"

"In principle."

"Could the type have been reproduced?"

"Yes."

"Could the ink have been formulated to resemble historical materials?"

"Of course."

"Then we still have ordinary explanations."

Bell looked at her.

"We do."

He examined the page again.

"But there's a difficulty."

"What?"

"The damage."

Miriam leaned forward. Bell pointed to the lower margin.

"Several wormholes pass through the gathering."

"Yes."

"Their trajectories correspond to damage in adjacent leaves."

"Meaning?"

"If the gathering was inserted recently, someone would have had to reproduce the insect damage in a way consistent with its position in the bound volume."

"Possible?"

"Certainly."

"Likely?"

"Not especially."

He turned another leaf.

"And the binding repairs appear to accommodate the gathering as though it had been present during earlier treatments."

Miriam looked at Oliver.

"Which treatments?"

"At least the seventeenth-century repair, according to the preliminary assessment."

Bell nodded.

"Though I'd want to examine the conservation history before making that claim."

Miriam sat back. Another apparently impossible object. Another plausible provenance. Another historical explanation. The pattern had become familiar. What disturbed her was no longer the possibility that the page might be genuine. It was the increasing difficulty of imagining how any ordinary evidence could establish that it wasn't. The final passage of the seventh division occupied the reverse of the leaf containing the modern citation. Miriam had not seen a complete photograph of it. The surviving text was damaged. Several words were uncertain.

Nevertheless, enough remained to establish the general structure of the argument. It concerned the distinction between the existence of a thing and the conditions under which its existence could be specified. The author argued that a sufficiently complete account of relationships might admit more than one apparent history. In one description, an object could be understood as arising from a sequence of earlier events. In another, the events themselves could be understood as necessary consequences of the object's position within a larger structure. The passage was philosophical.

Its terminology was not mathematical in any modern sense. But its central distinction was remarkably similar to the problem Yasmin had been investigating. Miriam translated slowly. The author appeared to reject the assumption that a thing must possess an independently established past before it could participate in the structure of reality. Instead, the text treated the past as one aspect of a completed set of relationships. Miriam stopped. Oliver looked at her.

"What?"

"The final sentence."

"Is it legible?"

"Mostly."

She read it aloud.

"'Those who ask when the world became such as it is assume that the world possessed a former condition from which the present may be distinguished.'"

She paused. The following words were damaged. Then came the conclusion.

"'But where the distinction is not preserved, there is no former world to which the question may refer.'"

Yasmin said nothing. Miriam read it again. The sentence was not proof of anything. A medieval philosopher could have formulated such an argument without possessing the slightest understanding of modern physics. The proposition might concern divine creation, the nature of eternity or the logical structure of possible worlds. But within the context of their investigation, it expressed the central difficulty with extraordinary clarity. If the historical record became complete, there might be no evidence of an earlier state.

The question of when the change occurred would presuppose the very distinction which the process had eliminated. Miriam looked at Yasmin.

"That's your identifiability problem."

"Yes."

"In a fifteenth-century printing."

"Apparently."

"And what do you make of it?"

Yasmin examined the page.

"I think the historical author, whoever they were, was considering a distinction which remains philosophically interesting."

"Nothing more?"

"Nothing I can responsibly conclude from this evidence."

Miriam smiled.

"I rather thought you'd say that."

Yasmin looked at her.

"But there is something I find troubling."

"What?"

"The passage doesn't describe an intervention."

"No."

"It doesn't describe a god or an entity choosing between possible worlds."

"No."

"Nor does it describe an act of creation in the ordinary temporal sense."

"Exactly."

Yasmin looked towards the manuscript.

"It treats the complete history as the primary object."

Miriam considered.

"And the things within that history?"

"Consequences of its structure."

There was a long silence. Then Miriam said, "Which would include us." Yasmin did not answer. She didn't need to. They examined the bibliographical appendix next. Most of it concerned the complicated transmission of the book. The supposed Arabic original. The Greek translation. The Latin printing. The disputes over authorship. The uncertainties surrounding the seventh division. It was an ordinary piece of scholarly apparatus, except for its impossible chronology. Then Miriam found something she had not expected.

A short paragraph near the end discussed the relationship between the text and the history of the species which had transmitted it. The Latin was obscure. Several words appeared to be transliterations of terms unfamiliar to the translator. The passage distinguished between the biological existence of a species and the persistence of the information associated with its intellectual activity. Miriam initially assumed it was another discussion of the survival of consciousness. But the context was different.

The author was discussing the historical conditions under which a text could exist. A text required authors, copyists, readers and material means of transmission. Those conditions, the author argued, might belong to a history which was not preserved in every complete account of the larger structure. Miriam read the passage again.

It appeared to suggest that a work could retain a coherent history even where the species responsible for its composition did not occupy the same position in the completed physical world. She looked at Oliver.

"Could this be a discussion of extinct civilisations?"

"Possibly."

"Or the preservation of texts after the destruction of their authors?"

"That would be a natural interpretation."

Miriam examined the next sentence. There was a reference to humanity. Not a general term for mankind. A technical designation in the later commentary. The surrounding passage had been annotated by a modern scholar, who identified the subject as Homo sapiens. Miriam frowned. The note was not part of the fifteenth-century printing. It belonged to an accompanying scholarly transcription. She checked its date. 2021. The editor's interpretation was conventional. The passage, according to the commentary, concerned the historical disappearance of a particular biological lineage. Miriam read the translation.

Then she stopped. The species was described in the past tense. Not as a hypothetical future extinction. Not as a moral warning. As an event belonging to the completed history of the world. She looked at Oliver.

"What's the editor's explanation?"

He examined the commentary.

"He treats it as an example of the text's speculative natural philosophy."

"Does he identify the species?"

"Yes."

"How?"

Oliver read the note. Then he frowned.

"That's odd."

"What?"

"The identification is explicit."

Miriam looked over his shoulder. The note described Homo sapiens as an extinct terrestrial species of the late Quaternary. She stared at the words. The phrase was absurd. Human beings were not extinct. The room contained four of them. The Library was filled with them. The cities beyond its walls were occupied by billions more. The entire documentary tradition of the Necronomicon had been created, examined and transmitted through human institutions. The note was plainly wrong. Miriam felt an unexpected relief.

At last, an error which could not be accommodated without abandoning the most elementary evidence of her own existence. She pointed to the commentary.

"That's incorrect."

Oliver examined it.

"Obviously."

"Then why has nobody corrected it?"

"Perhaps it's a typographical error."

"An unusual one."

"Or a mistaken identification of the species referred to in the passage."

"More likely."

Miriam looked at the Latin. The original text did not contain the modern biological name. The editor had supplied it. There was room for error. She examined the supporting references. The commentary cited a geological work concerning the history of terrestrial life. The work was recent. Its author appeared to be a genuine palaeontologist. The relevant section discussed several hominin lineages and the uncertainties surrounding their dates of extinction. Miriam followed the reference. The scientific discussion was cautious. It did not claim that modern humanity was extinct.

She felt reassured. Then she noticed a footnote. It referred to the distinction between the present biological classification of human populations and a proposed earlier lineage sharing the name Homo sapiens. Miriam frowned. She read it carefully. The terminology was confused. The author appeared to distinguish between the historical species and its later ecological successors. She could not determine whether the distinction represented a genuine taxonomic proposal or a mistake. She showed the passage to Yasmin. Yasmin looked at it for several minutes.

"This isn't my field."

"Nor mine."

"But the classification seems peculiar."

"Yes."

"Could be an unconventional taxonomic argument."

"Certainly."

"Or an error."

"Of course."

Yasmin looked at Miriam.

"Don't infer anything extraordinary."

"I wasn't planning to."

But the passage disturbed her. Not because it established that humanity had become extinct. It did nothing of the sort. It disturbed her because it suggested a route by which the contradiction might eventually be accommodated. The species named in the historical text could be reclassified. Its history could be distinguished from that of the present population. A new account of human evolution might arise. The apparently impossible statement could become a normal scholarly dispute.

And if the consistency process was real, the mistake might acquire a history in which it was no longer mistaken. Miriam looked at the page. For the first time, she understood why the investigators had been so eager to interpret the physical evidence as evidence of ancient monsters. The monsters were reassuringly separate. They belonged to another part of the world. Their existence might be frightening, but it did not necessarily require the disappearance of humanity. The greater danger lay in the relationships.

The possibility that the completion of the larger history might alter the conditions under which human beings had evolved, lived and remembered their past. The book's supposed entities were not necessarily hostile. They might not even possess awareness. But their histories could be incompatible with hers. And the universe possessed no obligation to privilege her version. The examination ended shortly before five. Miriam asked for copies of the newly recovered leaves and the associated commentary. Oliver agreed. He seemed almost embarrassed by the anomalous biological reference.

"I'll ask the editor for clarification."

"Please do."

"It may be a simple mistake."

"Very possibly."

He looked at her.

"You seem rather troubled."

Miriam considered.

"Do you remember when I first insisted that the book didn't exist?"

Oliver smiled.

"Vaguely."

"And you thought I was mistaken."

"Yes."

"Now the evidence is on your side."

"Apparently."

"Do you ever wonder whether we might have been right at the beginning?"

Oliver looked puzzled.

"Right about what?"

"That the book was fictional."

He considered.

"Lovecraft certainly transformed the tradition into fiction."

"Yes."

"But the historical material appears well established."

"Yes."

"Then perhaps the mistake was treating the two as mutually exclusive."

Miriam looked at him. A perfectly sensible answer. A perfectly coherent answer. One which no longer required him to remember a history in which the book had not existed. She smiled sadly.

"Perhaps."

They said goodbye. Oliver returned to his work. Helen carried the volume back to the conservation facilities. The newly recovered gathering was placed in protective storage. The book's history had become slightly more complete. And Miriam left the Library with the uncomfortable suspicion that the last contradiction in which she could place any confidence concerned the continued existence of her own species. Yasmin accompanied her to the station. They walked in silence for several minutes. The streets were crowded.

People moved around them, carrying bags, speaking into telephones and attending to the minor obligations of ordinary life. Miriam watched them. They appeared entirely real. Their conversations were ordinary. Their movements occupied space. Their lives possessed histories. Nothing suggested that humanity was becoming obsolete within some larger structure of reality. She found that reassuring. Then she remembered the marginal commentary.

Homo sapiens, extinct.

The statement was plainly false. Yet she had spent weeks watching other apparently false statements acquire documentary support. She looked at Yasmin.

"Do you think that passage will become correct?"

"Which passage?"

"The one concerning humanity."

Yasmin considered.

"I don't know."

"But the model permits histories in which human beings don't exist."

"Of course."

"And histories in which they do."

"Yes."

"Could it distinguish between them?"

"Only if we supplied additional constraints."

Miriam looked at the crowd.

"Such as the fact that we're here."

Yasmin nodded.

"That is a very substantial constraint."

"Then perhaps we're safe."

Yasmin did not answer immediately. Finally she said, "Only if the completed history is required to preserve the present observations in the way we understand them." Miriam felt a sudden chill.

"What do you mean?"

"The model doesn't necessarily treat every presently recorded fact as invariant across possible complete histories."

"But our existence..."

"Is an observation within one history."

Miriam stopped walking. Yasmin turned towards her.

"That doesn't mean we're about to disappear," she said. "It means the mathematical model doesn't guarantee the preservation of every feature of the history from which the observations were made."

Miriam looked around. The city continued its ordinary business. The buildings stood where they had always stood. The station clock displayed the expected time. A train arrived. People boarded. Nothing was happening. Yet she could no longer escape the possibility that the world she was observing was merely one consistent arrangement among others. And that the process, if real, might eventually leave no accessible evidence of this arrangement. She looked at Yasmin.

"Would there be a moment?"

"A moment of what?"

"Of disappearance."

Yasmin considered.

"Not necessarily."

"Then how would it happen?"

"That assumes a process unfolding within an already fixed time."

Miriam closed her eyes.

"Of course."

Yasmin continued.

"If the relevant history is completed as a whole, there may be no transition which an observer could experience."

"Because the observer might not exist in the completed history."

"Yes."

Miriam opened her eyes.

"And there would be no one to remember."

"Not in that history."

She looked towards the arriving train. A world without humanity. Not a world in which human beings had died. Not a world in which they had been destroyed by monsters. Not a world which remembered their extinction. A world in which the particular history giving rise to their existence had never been among the completed relationships. There would be no catastrophe to witness. No final scream. No ruins. No evidence that anything had been lost. There would simply be a universe.

And the universe would possess no reason to regard the absence of humanity as an omission. Miriam returned to Cambridge with a copy of the species commentary. The text identified an extinct population; a modern editor had supplied Homo sapiens, although the Latin was ambiguous and the biological identification unsupported. It would have been an easy correction in an ordinary scholarly argument.

At home she opened the old notebook. The first statement, The Necronomicon does not exist, was crossed out. So was the second: It did not exist before we described it. The third remained: We remember a world in which it did not exist.

She had written all three. She remembered crossing out the first two, and the reasons she had done it. On the next page she found a reference to the newly recovered London gathering, dated three weeks before Oliver reported the discovery. Beside it was a pencil warning in her own hand: Do not correct the species entry.

She set the notebook under the desk lamp. The pencil pressure had left a faint groove in the paper beneath. The warning was not evidence that scholarly correction could alter biology. It was an unexplained entry in a record which had repeatedly contradicted her recollection. She had sufficient training to know the difference. She wished training made the difference easier to endure.

Miriam opened a document headed A Minor Bibliographical Correction. She prepared two alternative readings of the damaged Latin and a note explaining why neither justified the modern species identification. The argument was conventional. She could have submitted it to the editor that afternoon. Then she remembered the fourth sentence in the older draft: Where the existence of the observer is among the disputed conditions, correction cannot be assumed to be an external act.

She deleted nothing. She saved nothing. She left the proposed correction on the screen and went upstairs to lie down, still dressed, with the curtains open to the grey evening. When she returned an hour later, the room was dark and the screen had locked. She did not reopen the document.

The next morning, Yasmin telephoned.

"Have you seen the latest version of the model?"

"No."

"There's something important."

Miriam sat down.

"Go on."

"I've been comparing the time-symmetric reconstructions."

"And?"

"The mathematical distinction between the book entering our history and our history becoming part of a larger structure is not merely difficult to establish."

"What is it?"

"Under the current assumptions, it's absent."

Miriam considered.

"Explain."

"Both descriptions correspond to the same complete set of relationships. Neither possesses a privileged causal direction."

"Then the idea of tunnelling from non-existence into existence..."

"May be a useful metaphor for the way we experienced the discrepancy, but it doesn't describe an established physical process."

"Quite."

Yasmin hesitated.

"There's something else."

Miriam waited.

"The model admits histories in which the species responsible for the original reconstruction doesn't occupy the same position in the completed structure."

Miriam felt a sudden coldness.

"Humanity."

"Yes."

"Does the model predict our extinction?"

"No. Absolutely not."

"Then what does it predict?"

"Nothing so specific. It shows that our present observational constraints don't uniquely determine a history preserving every feature of human existence."

Miriam considered.

"Which means we aren't guaranteed a place in the completed history."

"Correct."

"Nor are we guaranteed to remember an earlier one."

"No."

"Then the final state might contain the book, its sources, the physical structures and the underlying geometry..."

"Yes."

"But not us."

"That's mathematically possible within the model."

Miriam looked towards her desk. The unfinished correction was gone. She had not saved it.

"Yasmin?"

"Yes?"

"Do you remember why we began this?"

There was a long silence.

"Because you encountered a discrepancy concerning the historical Necronomicon."

"Before that."

"What do you mean?"

"The book was fictional. We knew it was fictional. The original research was an exercise in reconstructing something which did not exist."

Yasmin did not answer immediately. Then she said, "I remember you telling me that."

"And do you remember believing it?"

Another silence.

"Miriam, I don't know."

She closed her eyes. Yasmin continued.

"The historical evidence is now substantial. I can't reconstruct a previous physical state in which the book was absent."

"I know."

"And the mathematical model doesn't establish that such a state existed."

"I know."

"Then perhaps our original interpretation was mistaken."

Miriam sat very still. It was a reasonable conclusion. An intellectually respectable conclusion. One supported by the increasingly coherent historical record. And one which would eventually make the entire investigation unnecessary. She looked at her notebook. Three statements. Two crossed out. One remaining.

We remember a world in which it did not exist.

But Yasmin no longer remembered. Oliver no longer remembered. Lena had never known the original contradiction. Gabriel had never been told the complete literary history. Daniel regarded the connection with Lovecraft as an eccentric source of theoretical inspiration. The original author had withdrawn the essay, but the historical tradition now possessed an extensive independent record. Miriam was becoming the last person who still remembered the book's non-existence as something more than an ordinary scholarly mistake. She looked at the sentence.

Then she realised that it was no longer correct.

We remember.

No. Not we. She took up her pen. For several seconds she hesitated. Then she crossed out the first word.

Above it she wrote:

I.

The sentence now read:

I remember a world in which it did not exist.

Miriam stared at it. A single surviving contradiction. An unreliable memory. A proposition unsupported by the completed documentary record. The last apparent trace of a history which might never have existed. She wondered whether the correction had made the sentence more accurate. It had. And she wondered whether accuracy had become the wrong objective.

She considered asking Ruth to remember their conversation at the garden gate again. She had already asked once, and Ruth's reply had been unremarkable: Miriam had been talking about a rare old book, not a fictional one, and Ruth had made a joke about what such things cost. The fence latch was still stiff; the rose still caught at her sleeve. There was nothing left outside the paperwork which remembered the remark Miriam had heard. She did not ask a second time.

By evening, the latest commentary on the seventh division had appeared online. Miriam opened it reluctantly. The editor had issued a small correction concerning the extinct species reference. The correction did not remove the claim. It clarified the terminology. The species formerly identified as Homo sapiens was now assigned a separate designation, representing a hypothetical ancestral population whose classification remained disputed. The editor emphasised that the original Latin was ambiguous and that the proposed identification should not be treated as established. Miriam read the correction.

It was entirely reasonable. It resolved the most obvious contradiction. Humanity was no longer described as extinct. The disputed historical population belonged to an uncertain earlier lineage. The modern human species remained exactly where it belonged. She felt a wave of relief. Then she read the final footnote. The editor referred to an alternative interpretation of the passage. According to that interpretation, the text did not concern the extinction of a biological species at all.

It concerned the disappearance of a historical configuration in which a particular kind of observer had existed. The commentary described this as a speculative philosophical reading. Miriam stared at it. The correction had not removed the deeper problem. It had merely relocated it. The text no longer claimed that modern humanity was extinct. It suggested that a particular history of human existence might not be preserved within every complete account of reality. The contradiction had become less obvious. More sophisticated. More difficult to dismiss.

And more deeply integrated into the philosophical structure of the seventh division. Miriam closed the commentary. She understood now why the attempted corrections had failed. The process did not require every statement to be true. It required every statement, including its errors, to possess a coherent place within the history. A false attribution could become a scholarly controversy. A disputed passage could become a textual variant. An apparent impossibility could become a problem of interpretation.

And an attempt to correct the historical record could become another episode in the development of that record. The larger structure could accommodate almost anything, provided the necessary relationships were supplied. Almost anything. But perhaps not every history. Perhaps some arrangements were incompatible. Perhaps some observers existed only in configurations which the completed structure did not preserve. Miriam looked at the notebook.

I remember a world in which it did not exist.

The sentence was now a statement about herself. Nothing more. It did not establish an earlier universe. It did not demonstrate retrospective causation. It did not prove that the Necronomicon had tunnelled into existence. It merely recorded a memory which the available evidence no longer supported. And perhaps that was all that remained of the original history.

The following morning, Miriam returned to the British Library. She had not arranged another formal examination of the book. She wanted only to see Oliver. He met her in the same small room where they had first examined the Wormius printing. The conservation supports were arranged on the trolley. The table was empty. The familiar smell of old paper and cleaning materials lingered in the air. Oliver looked pleased to see her.

"Miriam. I wasn't expecting you."

"I know."

"Is something wrong?"

She considered.

"I wanted to ask you one final question."

"Of course."

"Do you remember the first email you sent me?"

He smiled.

"About the catalogue entry?"

"Yes."

"Certainly."

"And why did you send it?"

"Because your article treated the historical printing as fictional."

"Yes."

He looked puzzled.

"Is there something significant about that?"

Miriam considered.

"At the time, I was absolutely certain the book didn't exist."

"A perfectly understandable error."

"Was it?"

"Given Lovecraft's elaborate fictional treatment, yes."

She looked at him.

"Oliver, do you remember thinking the same thing?"

He hesitated.

"No."

"Not at all?"

"Not clearly."

She nodded.

"Thank you."

He looked concerned.

"Miriam, are you all right?"

She considered the question. Then she smiled.

"I don't know."

Oliver appeared uncertain how to respond. Miriam looked towards the empty conservation table. The physical Necronomicon was elsewhere in the building. Its paper had been manufactured centuries before her birth. Its type had been set by printers long dead. Its binding had been repaired. Its ownership recorded. Its contents examined. Its history documented by generations of scholars. Every detail now possessed a plausible place within the world.

The book was real.

That much she could no longer reasonably deny. What she could not establish was whether it had always been real. Nor whether the world in which it was real had always been hers. She thought of asking Oliver to retrieve the volume. Then she decided against it. There was nothing more the book itself could tell her. Its history was becoming complete. And the more complete it became, the less room remained for the contradiction she remembered. She thanked him. Then she left. Outside, London was entirely ordinary.

The streets were crowded. The traffic moved. People crossed the roads, entered shops, spoke into telephones and hurried towards appointments whose importance was confined to the brief circumstances of their individual lives. Nothing suggested the imminent disappearance of humanity. Nothing suggested that the universe's history was approaching a different configuration. The sky was pale. A few clouds moved above the buildings. The world looked precisely as it had always looked. Miriam walked towards the station. She had spent weeks investigating an impossible book.

She had discovered a physical volume whose history contradicted her memory. She had encountered manuscripts containing references to future scholarship. She had watched scientific records acquire increasingly elaborate histories. She had examined ancient biological structures and unusual geometrical relationships. And she had attempted, unsuccessfully, to reverse the process by correcting the text. Now she possessed no compelling evidence that anything extraordinary had ever happened. The physical anomalies remained unresolved scientific questions. The documentary contradictions remained historical problems requiring further examination. The Necronomicon possessed an apparently genuine textual tradition.

The mathematical model remained speculative. And the only surviving evidence of another history was a memory which she could not independently verify. She found the conclusion intellectually intolerable. It was also unavoidable. A historian could not privilege her own recollection over a substantial, mutually supporting body of physical evidence merely because the evidence contradicted what she believed she remembered. She had spent her career explaining that principle. She was not about to abandon it. Yet something in her refused to let the original memory go.

She remembered the book as fictional. She remembered researching Lovecraft's invention. She remembered the ordinary satisfaction of constructing a coherent account of something which had never existed. She remembered Oliver's first email. Her laughter. Her irritation. The catalogue entry. The first impossible reference. The beginning of an investigation whose conclusion she could no longer formulate without contradicting the evidence. Miriam stopped walking. The crowd moved around her. For a moment she felt profoundly alone. Not because she was the only person who understood the phenomenon.

She did not understand it. Not because she had been chosen to preserve some forbidden truth. Nothing in the evidence suggested that the universe possessed any interest in her. But because she might be the last surviving observer who remembered a version of history which the completed world no longer required. She looked at the people passing her. Their existence was ordinary. Their histories were coherent. Their memories were presumably consistent with the world around them. Nothing marked them as provisional. Nothing marked her as different.

And perhaps there was no difference. Perhaps she was simply mistaken. That possibility remained. It was, she reminded herself, still the most reasonable conclusion. She continued walking. Back in Cambridge, Miriam made one final entry in her notebook. She did not attempt another correction. She did not propose a new hypothesis. She did not try to persuade Yasmin to alter the mathematical model. She simply recorded what she remembered. The original essay. The fictional book. The first correspondence. The discovery of the London printing. The accumulating historical evidence.

The biological and geological investigations. The spatial anomalies. The failed corrections. The increasingly coherent world. She wrote carefully. She distinguished observations from interpretations. She marked uncertainties. She avoided claims she could not support. Where her memory conflicted with the documentary record, she said so explicitly. The account occupied several pages.

At the end, she wrote:

I cannot establish that the world I remember ever existed.

She paused.

Then added:

I cannot establish that it did not.

She considered the two sentences. They were unsatisfying. They were also accurate. She looked at the notebook. The final page contained the brief correction concerning the extinct species reference. The passage had been misidentified. The modern editor had amended it. The apparent contradiction had been resolved. But the philosophical question remained. Could a historical configuration containing a particular class of observer fail to survive in a completed structure of reality? The mathematical model permitted such possibilities. It did not establish that they occurred.

Nor did it establish that any human action could influence them. Miriam knew that. She also knew that her own existence was not a privileged constraint upon the universe. It was one fact among others. One biological arrangement. One temporary organisation of matter. One conscious perspective arising within a history whose larger structure she could not determine. The possibility that her history might not be preserved was no more cosmically significant than the extinction of an ancient species or the disappearance of a star.

The universe had never promised otherwise. Miriam looked at the final correction. She considered the ambiguous Latin passage. The editor had resolved its most obvious error. She could propose a more rigorous reading. She could distinguish the biological terminology from the philosophical claims. She could identify the unsupported inference. It would be a useful contribution to scholarship. And it might, if the consistency hypothesis was correct, supply another relationship to the growing network. Miriam hesitated. Then she closed the document. She would not correct it.

Not because she believed the text possessed magical power. Not because she imagined that the book could punish her. But because she could no longer be certain that the correction was merely descriptive. And because, for the first time in her professional life, she faced a question whose consequences might not be confined to the truth of the statement being corrected. She looked at the notebook. The original contradiction remained.

I remember a world in which it did not exist.

The sentence was unsupported. It was probably mistaken. It might be the last residue of an alternative history. Or it might be nothing more than a false memory. She could not decide. And she no longer intended to pretend otherwise.

That night Miriam left the notebook open on the desk. She had written a correction to the species entry and then withheld it. The two Latin readings were still beside one another: one referring to an extinct biological population, the other to a history in which the population's existence was no longer required. Both depended upon damaged words. Neither had earned her certainty.

At three in the morning she came downstairs for water. The page was where she had left it. Her final sentence, I remember a world in which it did not exist, stood beneath the earlier crossed-out versions. She recognised the pressure of the pen on the paper. She remembered making the correction from we to I. The memory was intact, and entirely unsupported by anyone else's.

She began to turn the page, then stopped. On the reverse was a citation to the newly recovered gathering in London, written in her own hand and dated three weeks before Oliver had reported finding it. She put a finger against the paper, not to test its age but to steady the slight tremor in her hand.

A historian could point out that the date might be mistaken. A physicist could observe that no mechanism had been established. Both would be right. Neither observation would supply the missing earlier world.

She looked at the proposed correction. To make the ambiguous passage more precise would be responsible scholarship. It would also introduce one further relation into a history which had proved disturbingly capable of supplying the rest. She could not establish that the act mattered. For once, she could not establish that it did not. Miriam closed the notebook without correcting the passage. Beyond the window, the stars continued their ancient business. They had never belonged to her universe. The universe had merely possessed her for a while.

End of The Book That Learned to Exist