Scenario tests and the ledger

The question the ledger models is narrow on purpose: where was the principal seat of the main body of the colonists in the years after they left Roanoke, roughly 1588 to 1607? Five outcomes cover it. Croatoan; the Albemarle mainland, meaning the head of the sound, the Chowan mouth and the Roanoke River country, with a Croatoan stage allowed; the Chesapeake; the sea; and anything else. Parties elsewhere, a division of the colony, a killing and later descendants are judged separately below, because the record supports combinations of outcomes that one question cannot represent.

How each scenario meets the evidence

Test	Croatoan	Albemarle mainland	Chesapeake	Sea
"50 miles further vp into the maine" (1587)	against: not the main, not 50 miles	fits in every reckoning of the mile	against: 130 miles by Lane, north across the sounds	against
CROATOAN without a cross; boats gone (1590)	fits exactly	fits as a stage or a party	strained: a stage far off the route	fits as a stage
Croatoan had "but little" corn (1587)	strained for 116 people over years	neutral	neutral	neutral
Harbour empty by July 1588; no vessel ever reported	neutral	neutral	neutral	against
Clothed men at Ocanahonan and Panawicke; "4 men ... from Roonock to Ocanahowan" (1608)	strained: inland, south-west	fits	strained: far south of the bay	strained
Slaughter of people "twenty and od yeares ... intermixt", "out of his territory"; survivors up the Chowan (1609 to 1612)	possible: far from Powhatan	fits	strained: the Chesepians were inside his reach	against
Chowanokes "little hope", Mangoags "all dead" (1609)	fits	strained: the Chowanokes should have known	fits	fits
Chesepians destroyed for a prophecy; a boy with yellow hair seen on the James by a Jamestown settler in 1607	neutral	neutral	fits	neutral
Hatteras ancestors "could talk in a Book" (1709)	fits	fits as a stage or a party	strained	strained
English ceramics on Salmon Creek (2012 to 2020)	strained	fits, if the date holds	strained	strained
European objects and hammerscale at Cape Creek	fits, if the date holds	fits as a stage	strained	strained
No sixteenth-century material at Great Neck	neutral	neutral	against, weakly	neutral

The ledger

Each row is a cluster of evidence that shares a source or a channel, and each carries one judgement of how likely that cluster is under each outcome. The numbers are judgements, and their purpose is to make the weighing explicit and to show how far the result moves when any one of them is changed. The prior, the starting weight given to each outcome before any evidence is counted, gives the mainland {{PRIOR_ALBEMARLE}}, Croatoan {{PRIOR_CROATOAN}}, the Chesapeake {{PRIOR_CHESAPEAKE}}, the sea {{PRIOR_SEA}} and other {{PRIOR_OTHER}}, on grounds stated in the notes to the published code.

Cluster	Croatoan	Albemarle	Chesapeake	Sea	Other	Why these values

The raw product gives the mainland {{RAW_ALBEMARLE}}, Croatoan {{RAW_CROATOAN}}, the Chesapeake {{RAW_CHESAPEAKE}}, the sea {{RAW_SEA}} and other {{RAW_OTHER}}. Blending in {{MIX}} parts in 100 of an even split across the five outcomes, to allow for evidence not in the ledger, gives the reported values: mainland {{P_ALBEMARLE}}, Croatoan {{P_CROATOAN}}, other {{P_OTHER}}, Chesapeake {{P_CHESAPEAKE}}, sea {{P_SEA}}.

What happens when the weights change

The table runs the ledger again nineteen ways: with each cluster left out in turn, with no starting preference, with the two Jamestown clusters counted as one, with the evidence against each rival halved, with every weight halved, and with a larger allowance for the unknown. The mainland leads in every case. Its lead over Croatoan is narrow when the plan (cluster C1, the first row of the ledger) is left out and when the strength of four clusters (C1, C4, C5 and C9) is halved, the case labelled Croatoan-friendly in the table; those two cases give the lowest values for the mainland.

Case	Croatoan	Albemarle	Chesapeake	Sea	Other

The separate findings

These are judgements on their own evidence, not products of the ledger, and they are what makes the leading scenario a combination rather than a single place.

Finding	In 100	Range	Basis

Reading the numbers

The probabilities in the ledger are judgements about hearsay, not measurements. Each depends on the five outcomes chosen and on the weight given to each cluster, and the range across the alternative weightings is the more informative figure. The table above shows the size of that dependence for nineteen alternatives; more alternatives would not test the judgement against evidence outside the ledger.
The clusters are not fully independent even after grouping. The three Jamestown informants of 1607 to 1608 and the London informant of 1609 drew on the same body of report; the case that counts them as one is in the table.
The two archaeological clusters would each move the result by ten to fifteen points if their dates were published and held, and by as much the other way if they failed.
"Other" is not a hypothesis anyone has argued for in detail. It stands for the interior of the Pamlico, Tar and Neuse, where the Zúñiga chart's Pakerikanick and Strachey's Peccarecamek are placed, and for an early annihilation that left no report.