# Did the Princes in the Tower survive 1483?

Published page: https://historical-mysteries.netlify.app/princes-in-the-tower/report
Local copy: `report.html` (same content).

## The mystery
King Edward IV of England died in April 1483, leaving two sons: Edward V, aged twelve, and Richard, Duke of York, aged nine. Their uncle Richard, Duke of Gloucester, took charge of the young king, lodged both boys in the Tower of London, had them declared illegitimate, and was crowned Richard III on 6 July. Over that summer the boys' servants were dismissed and, in the words of Dominic Mancini, an Italian visitor who wrote up what he had seen for a French patron that December, they were seen "more rarely behind the bars and windows, till at length they ceased to appear altogether". No one recorded a dated sighting of either boy again. By October, men who had plotted to free them had heard that they were dead and rose instead for Henry Tudor, who killed Richard at Bosworth in August 1485 and became Henry VII.

## Why it is a mystery, and the two sides
No one who wrote at the time claims to know whether the boys died, or when, or how. Richard III never said; Henry VII never produced bodies or a published confession; the first full account of a murder is Thomas More's, written about thirty years later; and two children's bones found under a Tower staircase in 1674 have never been dated, sexed or tested. Most historians of the period conclude that the boys died in the Tower in 1483, most probably on Richard's orders, because that is what the conduct of everyone around them implies. Richard III's defenders (Ricardians) answer that no death is recorded and that two later claimants to the throne, the boy crowned "King Edward" at Dublin in 1487 (called Lambert Simnel by Henry VII's government) and the "Richard, Duke of York" of 1491–97 (called Perkin Warbeck), may have been the princes. In 2023 Philippa Langley, the writer who led the search that found Richard III's grave in 2012, put four documents forward as evidence that they were: a receipt for pikes in the accounts of Maximilian's artillery at Lille (December 1487) calling the Dublin king "son of King Edward"; a first-person account of Richard of York's escape in a manuscript of about 1500 at Arnhem in the Netherlands; a pledge by "Richard of England" to the Duke of Saxony at Dresden (October 1493); and a letter of Maximilian's naming three bodily marks by which the claimant could be known.

## How it was judged
Every document on both sides was read in print or online, in its own language with a translation where an edition could be reached, and tagged by what it can bear: for the survival documents, whether it shows a prince alive or only that a claimant or his backer said so; for the murder sources, whether the writer knew, inferred or reported a rumour, and whose side he was on. The evidence was then sorted into five clusters and weighed in a ledger (`code/ledger.py`): each cluster is given a number for how likely that evidence would be under each of five outcomes, and the numbers are combined by Bayes's rule from a stated starting weight. Each number is a judgment about the evidence, and another reader might set it differently. The ledger exists to make the weighing checkable and to show how far the verdict moves when any number changes.

## Verdict
Both Edward V and Richard, Duke of York, most probably died in the Tower in 1483, in Richard III's custody: about 85 in 100. One or both survived to adulthood: about 6 in 100. Something else: about 9 in 100. The evidence label is Probable.

- **Died in 1483, in Richard III's custody: about 85%** (ledger 0.848). Whether on his order, by an unauthorised act (the Duke of Buckingham, Richard's chief ally who turned against him that autumn, has been proposed) or by a natural death of the elder boy is not separated by the ledger; the sources point to his order but do not establish it.
- **One or both survived to adulthood (as "Simnel", "Warbeck" or otherwise): about 6%** (ledger 0.063: Edward 0.021, Richard 0.022, both 0.020). Nearly all of this is the 10% allowance for evidence not yet found; the raw ledger gives 0.003.
- **Something else (died 1484–85 in Richard's keeping; died under Henry VII; died abroad as children): about 9%** (ledger 0.090).

## How sure this is
The evidence label is Probable: the evidence clearly favours one answer over all the others, but does not prove it, and the percentage moves when the same evidence is weighed differently. Every figure depends on the five outcomes and the cluster weights set in `code/ledger.py`. The case for murder is a case from how people behaved, and it does not amount to proof. Nothing in the report is a new document or a new fact.

Sensitivity (all from `code/ledger_output.txt`): raw, no allowance 0.920 / 0.003 / 0.077; flat prior 0.820 / 0.071 / 0.110; drop the 1483 cluster entirely 0.674 / 0.072 / 0.254; drop More, Thornton and the bones 0.818 / 0.068 / 0.114; halve the strength of the three clusters that cut hardest against survival 0.773 / 0.077 / 0.150. A reader who assigns different weights gets a different verdict.

Sub-findings, also judgments: the 1487 Dublin king claimed to be the boys' cousin Edward, Earl of Warwick, or at any rate a new "King Edward", not Edward V, about 95% on the records tabulated; Richard of York's survival in any form is about 4% on the ledger's weights, and his survival as the 1490s claimant is a part of that.

## Where it stands against historians
The verdict agrees with the standard academic studies of the reign, by Charles Ross (1981), Rosemary Horrox (1989), A. J. Pollard (1991), Michael Hicks (2003, 2024) and Sean Cunningham (2003), and with the recent biographies by David Horspool (2015) and Chris Skidmore (2017). Their books are known here through summaries in the secondary literature; Hicks's 2024 article was read in abstract. The report agrees with them and changes nothing in their position (for the site's badge: agrees with the studies named; nothing here overturns what historians already think). The survival case it answers is Philippa Langley's (Missing Princes Project, 2023; revised 2025), with the research pages of the Richard III Society.

## What is new
- New evidence: none. No document was seen in manuscript and none is newly identified.
- Not new, credited in the report: the verdict (the studies above); the regnal-year argument, that the Dublin king's own documents date his reign as beginning in 1487, which Edward V's could not have done (Randolph Jones, writing in the Richard III Society's journal *The Ricardian*, 2014, 2021, 2023; Chris Phillips, an independent researcher, 2024); the Lille receipt's 1895 printing and impossible June 1487 chronology (Phillips 2024; Michael Bletzer 2013 for the earlier quotation); the identification of the 1490s claimant as a native of Tournai (James Gairdner, the Victorian historian of the reign, 1898; Ann Wroe's *Perkin*, 2003, on the Setúbal depositions of 1496); the age error in the claimant's 1493 letter (Gairdner 1898, p. 268 n.); the reading of the Arnhem text as propaganda that either hypothesis needs (Hicks 2024, anticipated by the Utrecht professor D. Th. Enklaar's question in 1951) and its tracking of the 1493 letter and 1497 confession (the historian Nathen Amin; Richard III Society pages); the overlap between the Arnhem text and More on Sir James Tyrell and Miles Forest (Hicks 2024; the historian Tim Thornton 2024); Forest's sons at court, Henry VII's whereabouts in 1502 and the chain "which was yonge kynge Edward the Vth" in a 1516 will (Thornton 2021, 2024); the argument from Henry VII's silence (the Ricardian writer Jeremy Potter 1983 and older).
- New analysis of old evidence, all of form: (1) a table of every 1487 record by jurisdiction (England, Ireland, the Burgundian Netherlands), built on Phillips's inventory with Gordon Smith 1996 and Jones 2014, arranged to test the claim that the English records are Tudor propaganda; (2) a sharpening of the Arnhem point (the stations in that text absent from the 1493 letter are exactly the stations of the Setúbal depositions and the 1497 confession, in order); (3) the murder sources set side by side in their original words and classed as rumour, inference or claimed knowledge, with one small textual note (the printed text of the French chancellor's 1484 speech reads "impune trucidari", not the often-quoted "trucidatos"); (4) the five-cluster ledger with sensitivity tests. None changes the answer.

## What would overturn it
A record from 1483–91 of either boy's whereabouts made by someone not later paid to say so; radiocarbon, sex and DNA results on the 1674 bones (children of the right ages in the wrong century would end the murder story as told; a Yorkist maternal-line match with a 1480s date would end the survival story); an Irish or Burgundian record of 1487 dating by "Edward V" or "the fifth year"; proof that the Setúbal depositions or the 1487 confession of the priest William Symonds were fabricated.

## What was not read
- Mancini's Latin (the 1936/1969 edition by C. A. J. Armstrong is lending-restricted); the Dutch of the Arnhem manuscript in full (no edition exists; read in Langley's translation as quoted); the Setúbal depositions themselves (Wroe 2003, read in summary); Hicks 2024 (abstract only); the *Ricardian Bulletin* notes 2023–25 (via intermediaries); Langley 2023/2025 in full (extracts and reviews); the Great Chronicle of London in the 1938 edition (secondary quotation); the standard academic monographs and recent biographies (Ross, Horrox, Pollard, Hicks 2003, Cunningham, Horspool, Skidmore), whose positions are taken from secondary summaries.
- The Nuremberg charters, the Dresden pledge and Maximilian's "three marks" letter are known only from the Missing Princes Project's descriptions.
- The status of the bones question was checked to March 2026, the latest report found, and not beyond.

## Figures on the page
One: the verdict bar at the top, captioned as Figure 1, drawn from the ledger's reported shares (85, 6 and 9 per cent). There are no maps or reproduced images. The five tables each carry their sources.

## Files
- `report.html` — the published page (self-contained HTML; Google Fonts optional).
- `notes/00-notes-1487-claimant.md` — Lille receipt (Finot 1895 French), Phillips's inventory, Henry VII to Innocent VIII (Latin), attainders, papal indult.
- `notes/01-simnel-lille.md` — every 1487 record by jurisdiction with original wording (Convocation, Heralds' Memoir, York House Books, André, Molinet, de But, Irish records, coins, Annals of Ulster).
- `notes/02-gelderland-warbeck.md` — the Gelderland MS (catalogue, 1951 letters, content as published), the 1493 letter to Isabella (Latin), the 1497 confession and Exeter letter, Tournai registers, Setúbal, the backers' despatches (CSP Spain/Venice/Milan).
- `notes/03-murder-sources.md` — Stallworth, Cely, Mancini, Crowland (Fulman 1684 Latin), Rochefort (Masselin 1835 Latin), Rennes oath, Ricart, Fabyan, Great Chronicle, Rous, Commynes, Divisiekroniek, Weinreich, Valera, Vergil, More, attainder, the 1674 bones; summary table; unverified list.
- `notes/04-modern-scholarship.md` — Thornton 2021/2024, Hicks 2024, the reception of Langley 2023, the bones and the DNA question to 2026, the consensus and the alternatives.
- `code/ledger.py` — the evidence ledger (pure Python 3); `code/ledger_output.txt` — its output. To change a judgment, edit the cluster likelihoods at the top and re-run.