Epstein Data Wiki

Ellen Marmur

Properties2
tagsperson, physician, dermatology, medical-network, new-york, mount-sinai, vendors
aliasesDr. Marmur, Dr Marmur, Marmur Medical

37 min read

ProfessionDermatologist. Her own e-mail signature sets her name and then, on the line beneath it, “President/Founder, Marmur Medical”12
Practice”Marmur Medical”, trading as “Marmur Dermatology and Cardiology PLLC” on its own card receipt. Two addresses print in the documents read here: “12 East 87th Street, Suite 1A” and “1050 PARK AVE STE 1A”341
Titles she signs”Associate Clinical Professor, The Department of Dermatology” and “Associate Clinical Professor, The Department of Genomics & Genetic Science”, “The Mount Sinai Medical Center”, in 2016 and in 201812
Relationship to the Epstein officeTreating dermatologist to a patient in Epstein’s household. Her practice’s details circulate inside his office and her charges stand on his card account. No document read for this page records Epstein himself as her patient562
Earliest and latest dated documents read here27 July 2016 and 3 July 201956
Her own acts in the recordShe accepts a referral and offers appointment times; she refuses to accept an equivocal answer from another physician and requires it in writing; she answers a patient query by pointing to a published reference127
Covered elsewhere on this wikiThe letter of 11 December 2018 that a second physician was asked to sign, from that physician’s side: Woodson Merrell. The wider set of doctors around Epstein’s household: the Epstein medical network

Ellen Marmur is a New York dermatologist who treated a patient in Jeffrey Epstein’s household between at least 2016 and 2019. Documents released under the Epstein Files Transparency Act record her accepting a referral, setting appointment times, her practice being paid through Epstein’s card account, and, on one afternoon in December 2018, declining to act on another doctor’s informal assurance and requiring him to put it in writing instead. The same documents record Epstein’s office spending that afternoon trying to produce the document she had asked for. Epstein wrote to two physicians himself. One of them declined in his own words; of the other, Epstein’s assistant reported back that she had been told he would not sign. She then carried a drafted letter to that physician’s office with the line that it was “the letter Jeffrey would like Dr. Merrell to sign”; that office asked back for “the exact wording”; and the same evening that physician wrote that he had composed and sent it.

Nothing on this page is a finding of wrongdoing, and appearing in the released files is not evidence of it. A dermatologist who treated a patient, billed for the treatment and insisted in writing on the documentation her prescribing required has done nothing wrong, and on the face of these documents she is the person in the sequence applying the brake rather than releasing it.

This page does not print the patient’s name, and that is a claim a reader can test by searching this page for it. Where a document read here turns on that person’s health, each footnote says for itself what it leaves out. The page on Woodson Merrell, which cites the signed letter of 11 December 2018, handles the same episode the same way from the other side.

No statement by Ellen Marmur about this correspondence was found in the material searched for this page. An outside search was made, and what it covered is set out at the foot of this page.

How far the record reaches, and how the name is spelled in it

Measured against the full-text index on 18 October 2026, as a count of distinct documents over the page text:

form querieddocumentsof which outside the bare surname
"Ellen Marmur"150
"Dr. Marmur"580
union of those two650
"Marmur Medical"190
bare marmur94n/a
Mannur21
the practice’s own printed addresses and handles, run as keys204

Floor reached for this page: 99 documents. It is the bare surname’s 94, the one document that only the damaged spelling Mannur reaches, and four further documents that no spelling of the surname reaches at all, found by running the practice’s own printed addresses and handles as search keys. One of the 99 is a different person entirely, so 98 concern this subject or her practice. It is a floor and not a population.8

Five things in that arithmetic are worth stating, because each of them is a trap that has cost other pages in this batch:

  • The full name is a sixth of the record. "Ellen Marmur" reaches 15 documents and the bare surname reaches 94, so a page built on the full name would have seen 16 per cent of what the bare surname alone reaches. The household almost never wrote her forename: "Dr. Marmur" alone reaches 58, and the two phrase forms are very nearly disjoint, sharing only 8 documents.9
  • "Dr. Marmur" and "Dr Marmur" are one query, not two. The index tokenizer splits on the full stop, so the two strings search identically and return the same 58. Neither is corroboration of the other.9
  • "Marmur" NOT "Ellen" returns 75, and that is not a second person. Of the 94 bare-surname documents, 19 carry the token “Ellen” somewhere and 75 do not, and the 75 are overwhelmingly the household’s honorific or bare usage. 75 plus 19 is 94, which is the whole set.9
  • There is a second Marmur in the release, and he is a lawyer. One document, carried on three of its sheets, is a court service list naming “Nathaniel Z Marmur” at a law-firm address of his own name. He is not written about on this page and is subtracted from the figure above. Outside the release there is also a second physician of the same surname in her family; no document read for this page carries his name.10
  • A variant generator could not have found the form that mattered, and the form that mattered was a bar, not a misreading. Mannur is the extraction’s rendering of a correctly printed signature: the sheet reads “Dr. Ellen Marmur” at 400 dpi and the text layer reads “Dr. Ellen Mannur”. The clipped form Marmu, which on another subject in this batch reached documents nothing else did, returns 0 here. Marmer returns 11 documents and every one read is the extraction’s rendering of the ordinary word “manner” or a different person of that name; Marmr returns 1 and is a badly damaged sheet about nothing related. Both are excluded.11

The surname searched here is her married name, as an alumni profile published by her medical school records; that profile was fetched and read for this page on 18 October 2026. Every arm in the table above is therefore an arm on a married name.

What no method used here can reach. An appointment booked under a staff member’s name; a message that calls her “the dermatologist” or “her derm” without the surname; a sheet carrying only her office staff; and her maiden name, which the outside record read for this page does not give, so that hole is named here and is not sized. One such message is in the floor only because the surname happens to sit beside it.

The practice, as the documents print it

Her e-mail signature, in the form used in November 2016 and again in December 2018, reads:

Dr. Ellen Marmur President/Founder, Marmur Medical 12 East 87th Street, Suite 1A NY, NY 10128

followed by two lines the producing party has blacked out, and then “Associate Clinical Professor, The Department of Dermatology”, “Associate Clinical Professor, The Department of Genomics & Genetic Science”, “The Mount Sinai Medical Center”, “New York, New York 10029”.12

A card receipt of October 2017 gives the practice’s trading name as “Marmur Dermatology and Cardiology PLLC”, under the heading “Marmur Medical”, at “1050 PARK AVE STE 1A / NEW YORK, NY, 10028”. (“PLLC” is a professional limited liability company, the form a New York medical practice ordinarily takes.) Two contact lines on that sheet are not reproduced here. The signature box on it is empty.3

The earliest document read for this page is a short message to Epstein of 27 July 2016 under the subject line “contact info”. It lists three doctors. The first entry reads “Dr. Marmur”, then a blacked-out line, then “12 East 87th Street, Suite 1A”.5

November 2016: a referral accepted

On Friday 11 November 2016 Marmur answered a correspondent who had referred a prospective patient to her. The producing party has blacked out that correspondent’s name at every point it appears on the sheet read here, and no name is supplied in its place; a producer’s mark also covers the greeting and part of one sentence.

She thanked the correspondent for the referral, said she was away chairing a professional meeting, and offered two times on each of two named weekdays. The next sentence of her message is cut off by a mark. She then said the practice was heavily booked and asked the correspondent to apologise in advance for any delay, and signed with her forename. The correspondent answered a minute later in three short lines, saying the patient could surely come and that they were checking. The producing party bars that correspondent’s name, and this page supplies nothing about them, their sex included.1

That is the only document read for this page in which she sets an appointment time herself, and the exchange is an arrangement: nothing on this sheet records whether the appointment took place.

January 2017: a gift to a skin-cancer campaign

Two documents a day apart concern fundraising rather than treatment. On 3 January 2017 the American Academy of Dermatology issued a confirmation headed “Thank You for Your Gift!” with a second sheet reading:

Event/Donation Form: Skin Cancer, Take a Hike! - Patagonia In Support Of: Ellen Marmur Date: Jan 3, 2017 10:22:56 AM Amount: $670.00

This page does not identify the donor. On the covering message the producing party has blacked out the addressee line and the name in the greeting, and the form has no donor field at all: its “In Support Of” line names the person the sponsorship supports, not the person who gave. The form’s tracking code is not reproduced here.12

The following day a message sent from Marmur’s own address carried the subject “Thank you for your gift!” and the text “Dr. Marmur sends her heartfelt thanks for your meaningful contribution to skin cancer prevention!” and “From the entire team at Marmur Medical, THANK YOU for your donation to the American Academy of Dermatology’s Skin Cancer, Take a Hike!”. It speaks of her in the third person and carries a fundraising platform’s own footer, so it is a generated acknowledgement rather than something she sat down and wrote.13

The practice in the payment record

Charges to the practice run through Epstein’s American Express account statements. This paragraph rests on the text layer of a 9,824-sheet statement run and on no image: the run was not rendered for this page. On that text layer, “MARMUR MEDICAL” appears as a merchant on 55 dated charge lines, 54 of them in the statement run itself and one in a shorter extract of the same statements, across 42 distinct dates, the earliest 14 November 2016 and the latest 3 July 2019.6

The lines fall into two descriptors, which is a fact about how the practice took money rather than about any patient. 16 of the 55 carry a payment-processor tag pointing at a card-reader service’s receipt page; the other 39 carry the merchant name alone, and one of those is categorised on the statement as “DOCTOR & PHYSICIAN”. Two receipts in the release match the split: one is a card-reader receipt itemising a retail product with a New York sales-tax line, and the other is a card receipt for a service with no tax line.43 No amount from the statement run is reproduced here, because the extraction of these sheets mangles currency figures badly and none of them was read at an image.

No document read for this page establishes who was treated on any given date, and this page names no cardholder.14

11 and 12 December 2018: a letter one doctor was asked to sign and send to her

The sharpest sequence in this file runs over one afternoon and evening in December 2018, and it is readable without naming anybody’s condition, diagnosis or medication. Every time below is taken from a header that prints its own offset; where a sheet prints no zone, no clock time is given.

18:22:11 UTC. Epstein wrote in his own name to Woodson Merrell, asking him to send a letter to Marmur. The message has no subject line. Its closing words are “we will change drs this month , itsx crazy”.15

18:41:03 UTC, eighteen minutes and fifty-two seconds later. Lesley Groff, Epstein’s assistant, reported back to him that “I was told Dr. Merrell says he will not sign off on the letter to Dr. Marmur …”, that “he is in with a patient”, that she had asked “if I could please connect him to you”, and that “I will follow up in 30 minutes.” Her message reports what she had been told, not something she heard him say, and no document read for this page records that physician’s refusal in his own words. Part of that message is under a producer’s mark and is not supplied here.16

18:54:45 UTC. Epstein put the same request to Bruce Moskowitz by text message, in a thread recovered from a device and stamped by the export in UTC.17

19:24:28 UTC. Marmur wrote to Moskowitz and copied the message to others. The body is two paragraphs, each setting out in the second person what she wanted him to confirm, and each ending with the same three words. She opened “Dear Dr. Moskowitz,” and “Thank you for your email. To clarify and for medicolegal purposes,” and then twice: “Please confirm yes.” She signed “Sincerely, Dr Marmur”. (“Medicolegal” is the ordinary term of art for the point at which a clinical decision has to be defensible in a legal forum; the sentence says plainly why she wanted the answer in writing.) Nothing from the substance of those two paragraphs is reproduced here.2

19:28:52 UTC, four minutes and twenty-four seconds later. He declined. Moskowitz replied to her that he could not tell her any more, because “I did not do a full exam”, and recommended a full examination.18

20:30:22 and 20:32:26 UTC. A word-processor file whose name says it is the letter for Dr. Merrell reached Groff from the patient, and two minutes and four seconds later Groff forwarded it to Merrell’s office under the subject “Attn: Kashmire-Letter for Dr. Merrell to send to Dr. Marmur”. Her covering line, entire, was:

Hello Kashmire. This is the letter Jeffrey would like Dr. Merrell to sign.

Thanks, Lesley

Assistant to Jeffrey Epstein

19

Later the same evening. Merrell’s administrative assistant, Kashmire Gilmore, answered Groff: “I am in receipt of your email. I will print it out and give to Dr Merrell. Thank you for your patience.”20 Epstein then forwarded onward a message reporting that “Kashmire from Dr. Merrell’s office would like to k=ow the exact wording you would like on the letter you want Dr. Merrell to =ign and send to Dr. Marmur. She said she will then put the letter in front=of Dr. Merrell and ask him to sign.” (The stray equals signs are on the image itself: they are quoted-printable damage carried in the producing party’s own rendering of the message, not an artefact of the machine reading.)21

Quoted at 7:20 in the evening on a sheet that prints no zone, Merrell wrote: “Sorry for delay---I had to leave for a (rare) hospital emergency this afternoon. I just composed (needed rewording) and sent the letterof medical clearance out to Dr. Marmur for the Rx. Fax receipt confirmed. …” (“Rx” is the ordinary abbreviation for a prescription; “letterof” is run together on the sheet.) Epstein had written to him nine minutes earlier, “sorry for the trouble. did the lettr go out.”22

The next morning Merrell was told that the practice had said it did not have a fax from him, and fifty-four minutes later that the practice had called back and the fax had been found.23

An undated, unsigned, un-letterheaded draft letter is in the release, addressed to her by name and written in the first person for a signatory it does not name. At 400 dpi solid producer marks cover part of each of its first two sentences. Its two paragraphs are set in two different typefaces, the first sans-serif and the second serif, so the sheet is an assembly rather than a single composition. No sentence of its body is reproduced here.24

And this page does not say that Merrell signed anything. What the documents read here record is that he wrote that he had composed and sent a letter, and that the practice was reported the next morning to have called back and found the fax. The signed letter itself is cited and read on his own page and is not cited here.

December 2018 and January 2019: who was writing the patient’s letters

Two sheets fifty-five minutes and thirty-eight seconds apart, each carrying its own UTC offset, record the same body of text twice. At 16:26:41 UTC on 28 December 2018, Epstein wrote to the patient a message that is itself a letter: it opens with a salutation to Marmur and closes “thnk you”. At 17:22:19 UTC that text reached Marmur from the patient, broken into paragraphs, with the sign-off tidied and a signature added. The body is not reproduced here; the finding is who composed it.2526

Marmur answered on 2 January 2019 at 08:31:53 EST. She opened “Happy New Year.”, pointed the correspondent to a citation on a United States government medical database, and asked whether they still had their own copy of a programme agreement they had signed. Nothing further from that message is reproduced here. Epstein’s reply, forwarded to the patient the same morning, argues the manufacturer’s own prescribing information back at her, in the first person, as though he were the patient.7

2019

A message of 7 May 2019 in Epstein’s files is the only one read for this page that records somebody on the way back from the practice rather than on the way to it. The producing party has withheld the sender on that sheet, and this page attributes the message to nobody and supplies no attribute of any kind for its author.27

The last dated appearance of the practice in the material read for this page is a charge line of 3 July 2019 on the statement run.6

What was searched outside the released files

WebSearch was run four times on 18 October 2026, on her name alone, her name with Epstein, her practice with the release, and her name with the names of others in this correspondence. Her Mount Sinai faculty profile, her practice’s own biography page and an alumni profile published by her medical school were then fetched and read. The faculty profile gives her position as a clinical professor of dermatology, with hospital affiliations at The Mount Sinai Hospital and Mount Sinai Queens, and gives her practice as Marmur Medical at the East 87th Street address above. That is a later and different title from the two she signs on the documents above, and this page does not reconcile them.

No statement by her about this correspondence, or about the release, was found in the material searched for this page or in that outside search, and no request for comment was made by this desk. None of the outside pages read mentions Epstein. The search was four queries and three fetched pages; it was not exhaustive, and a statement may exist that it did not reach.

Sources and how they were read

Twenty sheets across seventeen documents were rendered locally at 400 dpi and read at the image, one of them also at 600 dpi; everything else cited here was read only at the machine-read text layer, and each footnote says which. This page cites 40 documents, every one of them inside the floor of 99 stated above, and the seventeen read at the image are seventeen of those 40. The other 59 of the 99 are cited nowhere on this page and none of them has been read at an image for it: they were read only in the text-layer window around each occurrence of the surname that the floor footnote counts. That hole is named and sized here rather than left to be inferred. Where a name, a value or a word is blacked out by the producing party, nothing is supplied in its place.

Related pages on this wiki: Jeffrey Epstein; Woodson Merrell; Bruce Moskowitz, who declined to give the confirmation she asked for; Lesley Groff, who carried the drafted letter between the two offices; the Epstein medical network; and Steven Victor, another dermatologist in the same files.

Footnotes

  1. Reply to Ellen Marmur, Subject: Re:, dated 11 November 2016 at 17:55:00 UTC, the header carrying its own zero offset, carrying her own message quoted beneath it, which the sheet stamps a minute earlier in New York local time. EFTA00628302 p. 1 of 2, with the tail of her signature block running onto p. 2. Both sheets read at the image at 400 dpi on 2026-10-18. ⚠ An earlier version of this footnote said p. 2 was not rendered and that nothing on this page rested on it. That was wrong on both limbs and is withdrawn: the three closing lines of her signature block quoted above stand on p. 2, so the November 2016 form of that block rests on it. The correspondent’s name is blacked out in the sender field, in the greeting of her message, in the attribution line of the earlier message and in that message’s sign-off, and the body of the earlier message is under a single wide mark; nothing is supplied in place of any of them, and no other copy is cited for this exchange. A mark also covers the remainder of one sentence of her own message. Two lines of her signature block are blacked out and are not reproduced. ⚠ The text layer of this sheet is damaged at three points inside the passages this page quotes from it, counted over the signature block reproduced above and over the paragraph her quoted sentence comes from: it renders Suite 1A as Suite IA, it replaces the comma of NY, NY 10128 with a full stop, and it mangles the opening of a later sentence. The image reads all three correctly and this page follows the image. ↩ ↩2 ↩3 ↩4 ↩5 ↩6

  2. Ellen Marmur to Bruce Moskowitz, with copies, dated 11 December 2018 at 19:24:28 UTC. EFTA00550992 p. 1 of 1. Read at the image at 400 dpi on 2026-10-18. All five e-mail addresses on the sheet are blacked out, and no correspondent on the sheet is identified here beyond those already named in the body above. The body’s two paragraphs are not reproduced here beyond the fourteen words printed above, which carry no particular. Two lines of her signature block are blacked out and are not reproduced. ⚠ The text layer of this sheet renders her own signature as Dr. Ellen Mannur; the image reads Dr. Ellen Marmur, and the page follows the image. A second filing carrying the same message inside the reply to it is at EFTA00550989. ↩ ↩2 ↩3 ↩4 ↩5 ↩6

  3. Card receipt, Subject: Your Signed Receipt - Marmur Medical, dated 27 October 2017 at 14:57:34 UTC, with the receipt itself on the second sheet. EFTA00569409 p. 2 of 2. Read at the image at 400 dpi on 2026-10-18. The sheet is headed Marmur Medical over Marmur Dermatology and Cardiology PLLC, gives the address 1050 PARK AVE STE 1A / NEW YORK, NY, 10028, is stamped 27 October 2017 at 10:56 in the morning, Eastern daylight time, and totals $820.00 with no tax line. The billed-to field is blacked out. Two contact lines, the approval, transaction, terminal and invoice identifiers, and the card’s last four digits are not reproduced here. The signature panel is empty: the image shows a broken-image placeholder inside the box and no mark of any kind, so this page makes no claim that anything was signed. ↩ ↩2 ↩3

  4. Card-reader receipt, Subject: Receipt from Marmur Medical, dated 29 March 2018 at 16:24:10 UTC. EFTA00540402 pp. 1 and 2 of 2. UNVERIFIED, from a search hit: this document was not rendered. It gives the practice at 1050 Park Avenue, Suite 1A, itemises one retail product, carries a Sales Tax (8.875%) line and totals $157.87, and the second sheet records the card entry and a local timestamp. This page names no recipient. A retail sale carrying New York sales tax is what separates this descriptor from the service charges at [^payjunction]; no clinical inference is drawn from the item and the item is not described further here. ↩ ↩2

  5. Message to Epstein, subject contact info, dated 27 July 2016 at 13:46:52 UTC, the header carrying its own zero offset, listing three doctors. The first entry reads Dr. Marmur, a blacked-out line, and 12 East 87th Street, Suite 1A. The sender’s name and address are blacked out. EFTA00691490 p. 1 of 1. Read at the image at 400 dpi on 2026-10-18. A contact value for a third doctor on this sheet is not reproduced here. ↩ ↩2 ↩3

  6. American Express statement run. This footnote rests on the text layer only. The run itself is openable: it was fetched whole on 18 October 2026 and its length confirmed at 9,824 sheets, which is the number of sheets held for it in the full-text index. But none of the charge lines counted below has been read at an image, and an earlier attempt to render a single sheet of it was killed by its own time limit at 900 seconds. Measured 2026-10-18 by a word-boundary scan of the stored page text, counting only lines where a date immediately precedes the merchant name so that the pairing comes from a contiguous string rather than from column position: 55 dated MARMUR MEDICAL charge lines over 42 distinct dates, the earliest 11/14/16 and the latest 07/03/19; 16 of the 55 carry a card-reader processor tag on the same extracted line as the merchant name and 39 do not. ⚠ The extraction renders that processor descriptor four different ways, so a scan for one literal spelling of it undercounts the tag. Measured 2026-10-19 over those same 55 lines, and the figures that follow count LINES, not charges and not documents: the commonest of the four renderings reaches 12 lines on its own, the other three reach 4 between them, and all four together reach the 16; an earlier version of this footnote and of the body gave 14 and 41, which is what three of the four renderings return, and those figures are withdrawn. All sixteen are paired in the extraction with the processor’s own receipt-page address, itself appearing in 14 distinct renderings, and that pairing is what establishes the sixteen are one descriptor rather than several. That pairing is a fact about the extracted text and not about the printed sheet, which this footnote has not opened: in the extraction the address never stands on the same line as the tag. 54 of the lines are in EFTA01671962 and 1 is in EFTA01731336 p. 20, which duplicates one of them, so the distinct-charge figure is at most 54. No amount is reproduced here: on the sheets inspected the extraction of this run drops digits from currency figures, loses decimal points, substitutes letters for digits and in one case replaces a figure with two punctuation characters, and a figure taken from such a layer is not a measurement. ⚠ A desk report of 14 September 2026 gives a table of amounts for this merchant drawn from the same sheets; this page does not adopt it, and not one of the sheets it rests on has been read at an image. Three of the run’s 9,824 sheets have since been rendered at 600 dpi for a different subject on this wiki, and none of the three carries a charge line counted here. ↩ ↩2 ↩3 ↩4

  7. Epstein to the patient, dated 2 January 2019 at 13:54:08 UTC, carrying Marmur’s message of the same morning at 8:31:53 Eastern standard time, which is 13:31:53 UTC and, beneath it, the message at [^dec28sent]. EFTA00550654 p. 1 of 2. UNVERIFIED, from a search hit: this sheet was not rendered. Two particulars in Marmur’s message are not reproduced here; the page states only that she pointed the correspondent to a published reference and asked about a signed programme agreement. Epstein’s own paragraph on this sheet is written in the first person as though he were the patient and is not reproduced here either. No address and no correspondent from this thread is reproduced here. ↩ ↩2

  8. Floor: 99 documents, of which 98 concern this subject or her practice. It is the bare surname’s 94, plus the single document reachable only by the damaged form Mannur, plus four documents reached only by running the practice’s own printed addresses and handles as keys, and minus one document for the different person at [^other] when the subject-specific figure is given. The 95 reached by the surname arms were read at the text layer for this page, in a window around each of the 227 occurrences of the surname or its variants. That figure counts OCCURRENCES, not sheets and not documents, and its vocabulary and its scope are given here so that a reader can reproduce it. Measured 2026-10-18 by a case-insensitive word-boundary scan of the stored page text over every sheet of those 95 documents, counting the six forms that [^variants] records as reaching documents inside the set: Marmur 216, Mannur 2, Marnur 1, mamwr 1, marmurmedical 6 and marimirmedical 1, which sum to 227. Each form was also run on its own, and the six separate counts sum to the same 227 as the single combined scan, so no occurrence is counted twice and none is hidden inside a longer form. 157 of those sheets carry at least one occurrence, and confining the scan to those 157 returns the same 227, so the figure does not turn on whether the scope is whole documents or matched sheets. Adding the clipped form Marmu, or either of the two spellings [^variants] excludes, changes it by nothing, because none of the three reaches a sheet inside the 95. ⚠ An earlier version of this footnote, and of the table above, gave the floor as 95. That figure was drawn over spellings of one surname only and did not include the forms the page itself prints; it is withdrawn. The four further documents are cited nowhere on this page and nothing on it rests on them. This is a floor and not a population: the surname searched here is a married name and no arm reaches a maiden one; no method used here reaches a message naming her only by office or role, nor an appointment booked under another name; and the statement run at [^amex] was not opened at an image. ↩

  9. Measured 2026-10-18 against /mnt/d/dump/AI-investigator/Epstein-research-data/full_text_corpus.db, as count(distinct efta_number) over select efta_number from pages_fts where pages_fts match 'text_content:"<form>"', no LIMIT and no join. "Ellen Marmur" 15; "Dr. Marmur" 58; their intersection 8; their union 65; "Marmur Medical" 19; bare marmur 94. "Dr Marmur" returns the same 58 as "Dr. Marmur", because the tokenizer splits on the full stop and they are one query. marmur AND ellen 19 and marmur NOT ellen 75, which sum to the 94. The returns were grouped by sheet before any total was read: the largest sheet carrying the surname is 3,489 characters, so no oversized container row sits in the set, and a control query that does reach such a row showed the guard firing on nine of them. ↩ ↩2 ↩3

  10. Court service list naming “Nathaniel Z Marmur” with a law-firm e-mail domain of his own name, printed on three sheets of one document, among several dozen other counsel. EFTA00044963 pp. 214, 222 and 230. p. 214 read at the image at 400 dpi on 2026-10-18, where the sheet is a notice of electronic filing in an unrelated criminal case and the name stands in its list of counsel; pp. 222 and 230 were not rendered. He is a different person from the subject of this page and nothing here concerns him. Separately, the outside search at the foot of this page returned an alumni-magazine item naming a second physician of the same surname in her family; no document read for this page carries that name, and he is not written about here. ↩

  11. scripts/ocr_variants.py was not relied on as the form test, because the forms that matter here are not misreadings. Measured 2026-10-18: Marnur 1 document, inside the surname set; Mannur 2, of which 1 is outside it and is a second copy of the message at [^dec11epstein] with the surname damaged; mamwr 5, of which 4 are the extraction’s rendering of the ordinary word “manner” in an unrelated financial prospectus and 1 is inside the set; marmurmedical 4 and marimirmedical 1, all inside the set; Marmur Dermatology 1; Marmu, the clipped form that on a sibling subject in this batch reached documents no other form did, 0. Excluded after reading: Marmer 11, every one the extraction’s rendering of “manner” or a different person named Marmer in a list of web links, and Marmr 1, a sheet so badly damaged that the surrounding words are unreadable. A non-zero damaged form is not an identification and each of these was read before it was counted. ↩

  12. American Academy of Dermatology, Subject: Thank You for Your Gift!, dated 3 January 2017 at 16:22:56 UTC, with the donation form on the second sheet. EFTA00571607 pp. 1 and 2 of 2. Both sheets read at the image at 400 dpi on 2026-10-18. The form’s own local timestamp is the same instant as the message header’s 16:22:56 UTC read in the Academy’s Illinois time zone. On p. 1 the addressee line and the name in the greeting are covered by producer marks, and so is the first word of a quoted testimonial and a web address in the footer. A telephone line and the form’s tracking code are not reproduced here. This page identifies no donor. ↩

  13. Message from Ellen Marmur’s address, Subject: Thank you for your gift!, dated 4 January 2017 at 14:01:42 UTC. EFTA00571609 p. 1 of 1. UNVERIFIED, from a search hit: this sheet was not rendered. Its text refers to her in the third person, offers a link to “my personal page” and carries a fundraising platform’s unsubscribe footer, so it is a message generated by that platform rather than one she composed. This page names no recipient. ↩

  14. A shorter statement extract in the release carries a cardholder column. EFTA00797912 pp. 1 to 14. UNVERIFIED, from a search hit: this document was not rendered. This page names no cardholder and draws no inference about who was treated. ↩

  15. Epstein to Woodson Merrell, no subject line, dated 11 December 2018 at 18:22:11 UTC. EFTA01013884 p. 1 of 1. Read at the image at 400 dpi on 2026-10-18. The body is one sentence and carries no producer mark, and only the closing clause is quoted on this page; nothing else from it is reproduced here. The sheet sets several of these words with double spaces between them, which this page does not reproduce. Four further copies of the same message are in the release at EFTA00550986, EFTA02312775, EFTA02610653 and EFTA02611224, and ⚠ two of those print its send time five hours apart from each other, one in local New York time and one in UTC, which is why this page takes every clock reading from a header that carries its own offset. A fifth further copy at EFTA01013870 renders the surname mannur and is the one document in the floor that no spelling of the surname reaches. ↩

  16. Lesley Groff to Epstein, Subject: Dr. Merrell, dated 11 December 2018 at 18:41:03 UTC. EFTA00487061 p. 1 of 1. Read at the image at 400 dpi on 2026-10-18. Part of the message is under a producer’s mark and is not supplied here, and a further mark covers the sign-off. The ellipsis in the first quotation above marks the remainder of that sentence, which runs on at the image and is not reproduced here. ⚠ The text layer of this sheet begins at r. Merrell says: it has dropped the opening words I was told D entirely, with nothing to show that anything is missing. The image carries them, and the quotations above are taken from the image. Two further filings of the same message are at EFTA01013883 and EFTA02267106. ↩

  17. Text message from Epstein to Bruce Moskowitz in a device export, stamped by the export at 18:54:45 on 11 December 2018, with the zone named. EFTA00508911 p. 15, with further copies at EFTA00782720 p. 15 and EFTA01211211 p. 2. UNVERIFIED, from a search hit: none of these sheets was rendered. Nothing from the message is quoted here; it is cited only for the fact and the time of the approach. The same export page carries a contact value for the practice, which is not reproduced here. ↩

  18. Bruce Moskowitz to Ellen Marmur, same copy list, dated 11 December 2018 at 19:28:52 UTC, four minutes and twenty-four seconds after the message it answers. EFTA00550989 p. 1 of 2. Read at the image at 400 dpi on 2026-10-18. Only the seven quoted words, which carry no particular, are reproduced from it. The page’s characterisation of what he declined to do is taken from the sheet as a whole and the reader can check it there. ↩

  19. Lesley Groff to Merrell’s office, Subject: Attn: Kashmire-Letter for Dr. Merrell to send to Dr. Marmur, dated 11 December 2018 at 20:32:26 UTC, with one attachment, a word-processor file whose name says it is the letter for Dr. Merrell, and forwarding a message the sheet stamps two minutes and four seconds earlier in New York local time, which is 20:30:22 UTC. EFTA00487097 p. 1 of 1. Read at the image at 400 dpi on 2026-10-18. No correspondent on the sheet is named here beyond those already named in the body above; all four addresses on the sheet are blacked out. ↩

  20. Kashmire Gilmore, administrative assistant to Woodson Merrell, to Lesley Groff, Subject: RE: Attn: Kashmire-Letter for Dr. Merrell to send to Dr. Marmur, dated 11 December 2018 on a sheet whose export header carries no time zone, so no clock time from it is given here. EFTA02268933 p. 1 of 2. Read at the image at 400 dpi on 2026-10-18. The sheet carries Groff’s quoted message beneath the reply with the local stamp 3:32 PM, which agrees with the UTC header at [^dec11letter]. A companion message from Groff giving the practice’s fax route is at EFTA02268849 p. 1, and a third filing of the same exchange at EFTA02312829 p. 1; neither was rendered, and the fax number on it is not reproduced here. ↩

  21. Epstein forwarding to the patient a message he had received, Subject: Fwd: D. Woody Merrell, dated 11 December 2018. EFTA02312744 p. 1 of 1. Read at the image at 400 dpi on 2026-10-18. ⚠ The two clocks on this sheet cannot both be read the way the clocks on its siblings are read: the export header gives 8:10 PM and the forwarded message inside it gives 3:58 PM, and no zone is printed on either, so no clock time from this sheet is given on this page and the sheet is used only for the date and the wording. The quotation is given exactly as the sheet prints it, including three points where a character has been replaced by an equals sign: that is quoted-printable encoding damage present in the producing party’s own rendering, confirmed at the image, and not an artefact of the machine reading. The inner message’s sender is blacked out and is not supplied here. A second filing of the same message is at EFTA02312874. ↩

  22. Thread carrying Woodson Merrell’s message quoted at twenty past seven in the evening of 11 December 2018 and Epstein’s nine minutes before it, inside an enclosing message dated 12 December 2018 at 19:19:47 UTC. EFTA00550912 p. 2 of 3. Read at the image at 400 dpi on 2026-10-18. Neither quoted stamp carries a zone, so this page gives those two times in words and takes no offset from them. The image sets the first three words of his message over a triple hyphen where the text layer gives a double one, and reads letterof run together; the quotation follows the image. E-mail addresses in the thread are blacked out and none is reproduced here. The quotation drops nothing from the middle of the passage; the ellipsis marks one further sentence at the end of his message, which is not reproduced here. ↩

  23. Same thread, p. 1 and p. 2, carrying two messages to Merrell stamped 8:36 and 9:30 on the morning of 12 December 2018, the first reporting that the practice had no fax from him and the second that the fax had been found. EFTA00550912 pp. 1 and 2 of 3. Read at the image at 400 dpi on 2026-10-18. The stamps carry no zone; the page gives only the interval between them. Neither message is quoted here. Further filings of the same thread are at EFTA00550921, EFTA02313174 and EFTA02313396, none of which was rendered. ↩

  24. Undated, unsigned, un-letterheaded single sheet opening with a salutation to her and written in the first person for a signatory it does not name. EFTA00298454 p. 1 of 1. Read at the image at 400 dpi on 2026-10-18. Solid producer marks cover part of each of the first two sentences. At the image the first paragraph is set in a sans-serif face and the second in a serif face, which is a measurement of this sheet and of no other. No sentence of the body is reproduced here. The sheet carries no date, no letterhead and no signature line, so this page does not assert when it was written or by whom, and does not assert that it is the file attached at [^dec11letter]. ↩

  25. Epstein to the patient, no subject line, dated 28 December 2018 at 16:26:41 UTC, the body of which is a letter addressed to Marmur and closing thnk you. EFTA00550668 p. 1 of 1. Read at the image at 400 dpi on 2026-10-18. The body is not reproduced here. A second filing is at EFTA02313926, which carries quoted-printable damage at sixteen points, counting every equals sign on that sheet, not one of which is a legitimate one; it was not rendered. ↩

  26. The patient to Ellen Marmur, dated 28 December 2018 at 17:22:19 UTC, fifty-five minutes and thirty-eight seconds after [^dec28draft], opening with the same salutation and signed. EFTA00551680 p. 1 of 1. Read at the image at 400 dpi on 2026-10-18. The subject line and the body are not reproduced here, and no correspondent is named here beyond those already named in the body above. The recipient’s address is blacked out. The comparison stated on this page is between the two bodies read side by side at the image: the words are the same, the sentence breaks differ, the sign-off thnk you is corrected, and a signature is added. ↩

  27. Message in Epstein’s files, Subject: Re:, dated 7 May 2019, quoting Epstein’s own one-character reply of early that afternoon and, beneath that, a message about a housekeeper interview. EFTA02315012 p. 1 of 1. Read at the image at 400 dpi and again at 600 dpi on 2026-10-18. This page attributes the message to nobody and supplies no attribute of any kind for its author, and nothing is supplied from any other source. No clock time from this sheet is given on this page. ↩

Backlinks