# Roanoke: what became of the 1587 colonists

Report: https://historical-mysteries.netlify.app/roanoke/report
Built page: report.html (from report-src/ via code/build_report.py).

## The story

On 27 August 1587 John White sailed from Roanoke Island, on the coast of what is now North Carolina, to fetch supplies from England. He was governor of a colony of about 115 men, women and children sent out that spring by Sir Walter Ralegh, the first English settlement in America to include families; his daughter Eleanor and her newborn child Virginia stayed behind. War with Spain kept him away for three years. When he landed again on 18 August 1590 the houses were down, the site was walled with tree trunks, the boats were gone, and the word CROATOAN was carved on a post: the name of an island fifty miles to the south and of the friendly people who lived there. By a code agreed before he left, a cross over the word would have meant trouble, and there was none. A storm cost his ships their anchors before he could reach the island, and no Englishman ever saw any of the colonists again.

## Why it is a mystery, and what each side says

No search reached them, no house, grave or bone of theirs has been identified, and every later report is hearsay from Indians, heard through interpreters at Jamestown, the English colony founded on Chesapeake Bay in 1607. The colonists had told White they would move fifty miles inland, to the west, while the word they carved names an island to the south. David Beers Quinn, the historian who edited the colony's documents, held that the main body went to Chesapeake Bay and was killed by Powhatan's people in April 1607. Since 2010 the usual reading, set out most fully by James Horn, is that the colony split between Croatoan and the mainland at the head of Albemarle Sound and was absorbed; the First Colony Foundation, which digs on that shore, announced in June 2026 that it had found the colony. Scott Dawson and the archaeologist Mark Horton, who dig on Hatteras, hold that everyone went to Croatoan. A carved stone of 1937 claiming to be a message from Eleanor Dare has its own defenders.

## How the case was judged

Every surviving document was read in full and in the original spelling: the fifty pages in Hakluyt's Principal Navigations, the Spanish documents, the Jamestown reports of 1607 to 1625 traced each to its first teller, and the later testimony of 1650 to 1888. The archaeology of three sites was checked against what has been published, as of September 2026. The possible fates were then set side by side in a ledger, a table of weights that anyone can change, and the ledger was run nineteen ways to show how far the result depends on the weights. The result is a ranking. The mainland is ahead of each rival, and the order can change when one judgement changes.

## Verdict and confidence

Label: leading candidate. Several answers remain possible and are put in order. The first is ahead of each rival, and the order can change when one judgement changes.

The colonists left Roanoke Island in an orderly removal by water. The word they cut, CROATOAN, named by the agreed code the place where they were or meant to be seated, without distress. White read it as a token of their safe being there; he never reached the island, and no witness recorded their arrival. Read together, the surviving reports point to a main body that moved on into the Albemarle mainland at the head of the sound, the fifty miles west it had announced, and dispersed among the peoples there. That sequence is a reconstruction, and the figures below are judgements that depend on the weighing set out in section 6 of the report, which anyone can change. Some colonists were possibly killed by Powhatan's people about 1607 (45 in 100). Lawson's Hatteras Indians of 1709 with white ancestors are more likely than not a trace of them (60 in 100).

Principal seat of the main body, about 1588 to 1607, on the weighing in section 6 (code/scenario_model.py; each figure blends in an even split across the five places, at 15 parts in 100, to allow for evidence nobody has seen; the range is across nineteen alternative weightings):

| Outcome | In 100 | Range |
|---|---|---|
| Albemarle mainland (Chowan mouth, Roanoke River country) | 68 | 41 to 77 |
| Croatoan (Hatteras) as the principal seat | 17 | 11 to 38 |
| Somewhere else (Pamlico, Tar, Neuse interior; early annihilation) | 7 | 5 to 14 |
| Chesapeake (Quinn) | 4 | 2 to 9 |
| Tried to sail home | 3 | 0 to 6 |

Separate findings, each judged on its own evidence and each depending on the readings stated in section 6 of the report: some of the colonists went to Croatoan after August 1587, 80 (65 to 90), where the carving names Croatoan as their seat and arrival is inferred; the colony divided, 75 (60 to 85); some were killed by Powhatan's people about 1607, 45 (30 to 60), so "possibly"; descendants lived among the Hatteras Indians in 1709, 60 (45 to 75); some tried to sail home, 10 (5 to 20); the first Dare Stone is a sixteenth-century inscription, 20 (15 to 25), which makes it about 80 in 100 a fabrication of 1937.

The leading scenario is a reconstruction from hearsay and from digs that have not been published. No source names the place, no structure or grave has been found, and the two excavations that could decide it (Salmon Creek Sites X and Y; the Cape Creek hammerscale) have no published data as of 14 September 2026.

## Confirms, shifts or overturns

Earlier accounts: Quinn (1974, 1984, 1985), followed by Kupperman, Oberg (2008) and Horn (2005), held that the main body went to the Chesapeake and was killed by Powhatan in April 1607. Since 2010 (Parramore 2001; Horn, A Kingdom Strange, 2010; the First Colony Foundation; Lawler 2018) the usual reading is that the colony split between Croatoan and the Chowan and was absorbed. Dawson and Horton (2016 to 2026), endorsed by DuVal (2025), hold that all of them went to Croatoan. Ewen and Shields (2024) hold that no hypothesis is more probable than another.

The report agrees with Horn's split-and-absorption reading, at lower confidence than he states: the sources support the direction and the dispersal, and not his route, season, headcount at Croatoan or raid of 400 warriors. It moves the weight of the slaughter story: the Council's instructions of May 1609, Hakluyt's dedication of April 1609 and Strachey's Historie are one London report from one Powhatan informant (Machumps), which itself dates the killing to spring 1607 and places it outside Powhatan's territory, so it does not support the Chesapeake; and Powhatan's "confession" with the musket barrel and brass mortar is a 1625 marginal note by Purchas, not in Smith. It overturns no earlier account.

## What is new

There is no new evidence: every document used was in print by 1888 and every excavation is someone else's. What follows is analysis of old evidence, and each item stands only on its own terms (the weights in the ledger are the report's own; the mile calibration is arithmetic on Lane's figures; nobody else has run the spelling test). Quinn, Barbour and Horn read the documents closely first, and section 8 of the report credits each step (Quinn for "at Roanoak" meaning the colony; Horn for Strachey's geography against the Chesapeake; Rountree for the dating of the Chesepians' destruction; Barbour for the 1609 searchers not crossing the watershed; Miller and Kupperman for the propaganda reading; Oberg for the Croatoan subsistence point; Ewen for the stratigraphic objection at Cape Creek; Fullam for the Spaniards at Ocanahowan). New here:

1. A ledger, published with its code, that shows how far the mainland reading depends on the announced plan and the Jamestown hearsay, and how close Croatoan runs when those are discounted.
2. A check of the colonists' "50 miles further vp into the maine" against Lane's own "130 miles" to the Chesepians and to Chawanook: in any reckoning of the mile the phrase stays inside the Albemarle basin and does not by itself pick the Chowan mouth over the middle of the sound.
3. A check of every archaeological claim of 2024 to 2026 against what has been published, which is nothing: no site report for Site X or Y, no sherd count for Site Y, no section or radiocarbon table for the hammerscale, and a First Colony Foundation web statement of 4 June 2026 announcing the colony "found" with no new data behind it.
4. A test of the first Dare Stone's spellings against Hakluyt, Smith and Strachey (notes/D_dare_stones.md): its Elizabethan surface is made of the devices an imitator reaches for, and La Vere's claim that "salvage" was used "during only a few years" is wrong.
5. Two small observations: the objects in Purchas's note, if real, could have been picked up on the abandoned Roanoke site where White saw iron lying in the grass; and Bland's 1650 order to enquire for "an English woman cast away long since" among the Tuscarora is a living tradition not discussed in any of the works read here.

## What could not be read

- The modern monographs (Quinn's Roanoke Voyages and Set Fair for Roanoke; Horn; Lawler; Kupperman; Oberg; Miller; Rountree; Fullam; Dawson) are in copyright and were read in search-inside snippets (page-numbered where Google Books supplied them), the authors' own articles, and full reviews, not cover to cover. Parramore (2001) and Oberg's review of Lawler were not obtained; Parramore's argument is known from the journal's own page-level annual index for volume 78 (sources/NCHR_2001_78_index_IA_ocr.txt) and from Fullam. Klingelhofer ed. 2023 was read in unpaginated Google Books snippets (59 terms), contents and reviews; Ewen and Shields 2024 as publisher descriptions, contents and interviews.
- The Zúñiga chart was read in three published transcriptions (Brown 1890, Barbour 1975, Encyclopedia Virginia) which disagree in wording; Barbour's 1969 diplomatic text was not obtained.
- Oré's Spanish text of the 1588 voyage was not obtained; Geiger's translation is used and Quinn's difference noted.
- The 1609 instructions were read in Kingsbury's edition, not the Ashmole manuscript; Strachey in the Sloane text (Major 1849), with the Princeton readings only as Barbour quotes them.
- No fieldwork, no laboratory result, no new document.

## What would overturn the ranking

A published, sealed 1580s deposit at Cape Creek with a hearth and slag (Croatoan to first place); a sixteenth-century English burial or house floor on the Lynnhaven or Elizabeth River (Quinn restored); a demonstration that the Salmon Creek ceramics are of 1607 to 1650 or later (mainland and Croatoan at parity); Rayner's lost relation of 1621 or Strachey's second decade; a laboratory result on the first Dare Stone.

## Figures

Both figures in the report were made for it. Figure 1 (the bar chart of the five outcomes with their ranges) is drawn from data/model_results.json by code/make_figures.py. Figure 2 (the map of straight-line distances from Roanoke Island) is drawn by the same script from data/distances_from_roanoke.csv, whose coordinates were computed as great-circle distances in statute miles by code/geography.py. No image is reproduced from any other source; the first Dare Stone's text is set in type from Pearce's 1938 reading.

## Files

- report.html, report-src/ (00_head, 10_header, 20_record, 30_geography, 40_jamestown, 45_later, 50_archaeology, 60_model, 70_darestone, 80_debate, 90_closing, 95_end; figs/ SVGs)
- code/build_report.py (assemble the page), code/geography.py (distances), code/scenario_model.py (ledger and its alternatives; writes data/model_results.json and data/tokens.json), code/make_figures.py (SVG figures)
- data/distances_from_roanoke.csv, data/model_results.json, data/tokens.json
- notes/A_roanoke_voyages.md (1584 to 1603, Hakluyt 1600 and the Spanish documents), notes/B_jamestown_sources.md (1607 to 1888), notes/C_archaeology.md, notes/D_dare_stones.md, notes/E_scholarship.md (modern scholarship, with a page-numbered snippet addendum), notes/F_chronology_and_findings.md
- sources/ (about 300 raw texts, each with URL, edition and fetch date in its header; raw_downloads/ for the Hakluyt and Spanish scans)

## Headline for the site

Roanoke, 1587: the Albemarle mainland is the leading candidate. In August 1587 John White, governor of Sir Walter Ralegh's colony of about 115 men, women and children on Roanoke Island, sailed to England for supplies. When he returned in 1590 the settlement was empty and the word CROATOAN, the name of a friendly island fifty miles to the south, was carved on a post. No one from England ever saw the colonists again. They had also said they would move fifty miles inland, to the west, so the two signs point different ways. Historians have argued for Chesapeake Bay (David Beers Quinn), for a split between Croatoan and the mainland at the head of Albemarle Sound (James Horn, the First Colony Foundation) and for Croatoan alone (Scott Dawson and Mark Horton). This report reads every surviving document in full and weighs the possibilities in a ledger that anyone can change. It puts the mainland first at about 68 in 100 (range 41 to 77), Croatoan second at 17 and Chesapeake Bay at 4; the sequence is a reconstruction, and no witness recorded it. Some colonists were possibly killed by Powhatan's people about 1607 (45). It agrees with the split-and-absorption reading at lower confidence, moves the slaughter tradition to one London report of 1609 that does not support the Chesapeake, and finds the first Dare Stone, a carved rock produced in 1937, about 80 in 100 a fabrication. What it adds is analysis of old evidence and none of it new evidence: the ledger with its alternatives, the mile calibration, the 2024 to 2026 archaeology check, the Dare Stone spelling test.