The Bigot List Leak Bench
A memo in chapter 21 of In Plain Sight says the people read into a program, the 'Bigot list', numbered 400 to 800 and had kept its secret since 1947; the admiral it quotes denies the meeting. Take those 400 to 800 people, and give each of them the paper's lowest chance of a leak, p = 4.09 × 10⁻⁶ a year, back-solved from the 30,000 NSA staff it counts for PRISM: the list has a 12 to 23 percent chance of a leak by 2026, and reaches even odds only after 212 to 424 years. Give them instead the p of the one case of the list's size, the FBI laboratory's 500 people (2.45 × 10⁻⁴), and a leak passes 95 percent within 17 to 32 years of 1947, before the 1997 audit the memo describes. The paper's 5 percent threshold, solved for 79 years, allows 158 people. Count the leavers who keep knowing (the bench's own extension), and a list of 800 means 1,515 people by 2026. The paper's Table 4 prints 2,531 where its own text, and its Eq. 3 solved for N, give 2,521; its correction's 1,328 comes back as 1,327.5. Whether the memo was the leak the model counts is not in the model: two readings hang at equal size.
Keep the signal
The station is free. No ads, no tracking cookies. If the files are worth your time, following is the support.
Open the interactive ▸ What you're looking at
The Clock is Grimes’s L, the chance of at least one leak, from 1947 to 2026 for a list of 400 (dashed) and 800 (solid) at the calibration on the dial, with the other calibrations as faint bands behind it. The chapter’s own dates sit on the time axis with the list’s L at each: the 1997 audit and Pentagon meeting, the memo’s meeting of October 2002, the memo online in June 2019, Wilson’s denial of June 2020, the book in 2021. A switch reads the memo as the model’s leak, and the question becomes how likely a first leak was by then.
The Calibration shows where p comes from: the NSA’s PRISM program, the Tuskegee study and the FBI laboratory, each drawn as its own clock crossing L* in the year it came out, and the paper’s alternative count for the FBI from 1932. Below them a log axis of p per person per year carries the four values and two bands: the p a list of 400 to 800 needs for even odds, and for a 95 percent chance, of no leak by the year on the dial.
The Map is L over the number who know (10 to a million) and the years kept (0 to 100) at the chosen p, with the paper’s 5 percent threshold, even odds and 95 percent drawn as contours, the paper’s Table 4 as gold rings on the threshold, the three cases at their own N and t, and the list as a box. The List counts who knows three ways: Grimes’s constant list, the bench’s list with its alumni, and the paper’s single 1947 cohort. The File runs the cards, the chapter and the paper quoted, and two readings at equal size.
Why it is here
The book has a culture note on the site, the Culture 08 source note on In Plain Sight, which records how Coulthart handles this memo: Wilson denies it, and the book says it 'has to be treated as a hoax document until proved otherwise'; the wiki has an entry on the Wilson memo. Chapter 21 carries one number an instrument can take over: the memo says the people read into the program 'consisted of just 400-800 people', and that the secret had been kept since 1947. Coulthart's own sentence, that such a secret 'strains my cynical journalist's brain' in a bureaucracy he calls 'notoriously leak-prone and gossipy', is a statement about a rate. In 2016 David Robert Grimes published a model for exactly that kind of question in PLoS ONE, with its calibration.
So the instrument does one thing: it runs the list through Grimes's clock and prices both readings. It takes the list's two ends as given (the question it asks is conditional on them, not about whether they are true), takes the paper's model and its three cases as published, and reproduces the paper's Table 1, its Table 4 but for one figure, and its correction; the model's choices are dials: which case sets p, the L* at which a case came out, who counts as knowing, the threshold, the year. It does not say whether the program exists, whether the memo is genuine, or which calibration fits. The record is the record's; the reading is the reader's.
How it works
Everything on the bench is arithmetic on the paper’s equations and on the chapter’s two numbers. The script that ships with the site re-derives every displayed number, and reproduces the paper’s own printed figures, before the build is allowed to pass.
L = 1 − exp(−∫₀ᵗ (1 − (1 − p)^N(s)) ds) · p = 1 − (1 + ln(1 − L*)/t)^(1/N) · N_max = ln(1 + ln(1 − μ)/t) / ln(1 − p)
The clock. If each of N people has a chance p a year of a leak that ends the secret, independently, the chance of at least one leak in t years is L = 1 − e^(−t(1 − (1 − p)^N)) (Grimes’s Eqs. 1 to 3). For a number who know that changes over time the paper’s March 2016 correction replaces t(1 − (1 − p)^N) with its integral; the paper’s conclusions, all drawn for a constant list, stand, and the correction redraws Figs 1 and 4 and rewrites one sentence. For 400 and 800 people over 79 years at p = 4.09 × 10⁻⁶: L = 0.121 and 0.228.
The calibration. A case of N people that came out after t years, taken as having reached L* by then, gives p = 1 − (1 + ln(1 − L*)/t)^(1/N). At the paper’s L* = 0.5 this is its Eq. 8, and the bench gets back Table 1: 4.09 × 10⁻⁶ for PRISM (30,000, 6 years), 4.20 × 10⁻⁶ for Tuskegee (6,700, 25 years), 2.45 × 10⁻⁴ for the FBI laboratory (500, 6 years), and 2.11 × 10⁻⁵ for the FBI counted from 1932. The same inversion with survival S in place of L* gives the p the list needs: even odds to 2026 at 1.10 to 2.20 × 10⁻⁵, 95 percent at 8.12 × 10⁻⁷ to 1.62 × 10⁻⁶.
Who counts. Grimes’s constant form (Eq. 4) keeps the number who know at N; his single-event form (Eq. 5) lets one cohort of average age 40 die by Gompertz constants α = 10⁻⁴ and β = 0.085, and the bench reproduces his correction’s ‘around 1328 participants’ with it (1,327.5). The bench adds a form of its own from the same parts: members leave after a tenure, keep what they know, and die by the same constants from the age they leave; with a 20-year tenure a list of 800 means 1,515 people by 2026, and L at PRISM’s p rises from 0.228 to 0.366.
Nothing is fitted. The only choices are the dials: the calibration or a p of your own, L*, the year, who counts and the tenure, the reading and the threshold; every finding is quoted at the defaults and across them.
The dials that decide the result
The calibration and L*; the year; who counts; the reading and the threshold. Together they draw every version of the clock the paper’s model allows.
- The calibration. PRISM (the default, the value the paper carries into its results), Tuskegee, the FBI laboratory, or the FBI counted from 1932; or a p of your own from 10⁻⁷ to 10⁻³ a year.
- L* when a case came out. 0.05 to 0.95; the paper takes 0.5. Lower means each case came out early for its size, which lowers every p; at PRISM, 0.05 gives 2.86 × 10⁻⁷ and 0.95 gives 2.31 × 10⁻⁵.
- The year. 1950 to 2026, counted from the start of 1947, with the chapter’s dates as presets: 1997, 2002, 2019, and now.
- Who counts. A constant list (Eq. 4, the default), the list with its alumni (the bench’s, with a tenure of 5 to 40 years, default 20), or one 1947 cohort (Eq. 5).
- The reading and the threshold. No leak counted, or the memo as the leak; the paper’s threshold μ from 1 to 50 percent, default its 5.
The claims, as they stand
The chapter’s figures and the paper’s one by one, and where each lands when the paper’s model is run.
| The people read into the program, the 'Bigot list', 'consisted of just 400-800 people' proposed by the memo attributed to Eric Davis, as In Plain Sight reports it (p. 263) | REPORTED | Entered as given, both ends; nothing between them is assumed. Wilson denies the meeting the memo records (pp. 263 to 266). The bench asks a question conditional on the figure, not whether the figure is true. |
| 'ever since the Roswell crash in 1947' proposed by the memo, as the book reports it (p. 263) | REPORTED | The clock starts at the start of 1947: 79 years to 2026, about 56 to the memo’s meeting in 2002, about 72 to the memo online in 2019. |
| 'It also strains my cynical journalist's brain to contemplate the possibility that such a preposterous secret could have been kept inside the notoriously leak-prone and gossipy bureaucracy frequenting the Potomac Washington swamp.' proposed by Coulthart (p. 272) | READING | A statement about a rate. At the paper’s lowest p the list has a 12 to 23 percent chance of a leak by 2026; at the FBI laboratory’s, above 99.9 percent. Which rate the sentence has in mind, it does not say. |
| 'The UFO leak of the century' proposed by Richard Dolan, as the book reports him (p. 261) | READING | The model counts one kind of event, a leak from inside that ends the secret. Read the memo as that leak, and at PRISM’s p a first leak by 2019 had an 11 to 21 percent chance; read it as not, and the list’s chance of none by 2026 is 77 to 88 percent. The bench prices both and chooses neither. |
| p = 4.09 × 10⁻⁶ (PRISM), 4.20 × 10⁻⁶ (Tuskegee), 2.45 × 10⁻⁴ (FBI laboratory) proposed by Grimes 2016, Table 1 | EXACT | Reproduced from each case’s N and t at L = 0.5, to the printed three figures. The paper carries the lowest into its results and calls it generous to the secret; the highest, roughly 60 times greater, it sets aside. |
| Table 4: at most 2,531 people for 5 years under a 5 percent chance of a leak proposed by Grimes 2016 | EXACT | Eq. 3 solved for N and the paper’s own Discussion give 2,521; every other row of the table comes back within one person. For 79 years it gives 158. |
| 'the odds of failure exceed 5% within 10 years at around 1328 participants' proposed by Grimes, the correction of March 2016 | EXACT | Eq. 5 as printed, with the correction’s integral, gives 1,327.5. |
| The list at the two large cases proposed by this bench | MODELLED | 12 to 23 percent by 2026 at PRISM’s p, the same to the percent at Tuskegee’s; even odds only after 212 to 424 years. The list is 2.5 to 5.0 times the largest group the paper’s 5 percent threshold allows for 79 years. |
| The list at the case of its size proposed by this bench | MODELLED | At the FBI laboratory’s p (500 people, 6 years), a 95 percent chance of a leak within 17 to 32 years of 1947; counted from 1932 instead, 49 to 73 percent by 2026. |
| Who counts as knowing proposed by this bench (the alumni form is its own) | MODELLED | With leavers who keep knowing and a 20-year tenure, a list of 800 means 1,515 people by 2026, and PRISM’s 12 to 23 percent becomes 20 to 37. One 1947 cohort with none added gives 5 to 10 percent. |
| What the bench amounts to proposed by this bench | READING | One model, two calibrations: a 77 to 88 percent chance of no leak by 2026 on the large cases, a leak all but certain decades before the memo on the case of the list’s size. Which calibration fits, and whether the memo was the leak, the chapter and the paper do not settle, and the bench does not. |
Try this
- Start on the clock. At PRISM’s p the band sits low: 12 to 23 percent by 2026. Click the FBI laboratory’s row and watch it climb to the top before 1980.
- Switch the reading. Read the memo as the leak and the violet chip gives the chance of a first leak by 2002 and by 2019; switch back and it is the chance of none by now.
- Open the calibration. Slide L* to 0.05, then to 0.95, and watch all four marks move on the p axis while the two bands the list needs stay put.
- Read the map. Find the gold ring at 5 years sitting off the threshold line; then find the FBI laboratory’s diamond, drawn with PRISM’s p.
- Count who knows. Switch to the list with its alumni and shorten the tenure to 10 years.
- Finish in the file. The list, the denial, the leak the chapter names, the three cases, then the two readings.
Accuracy
The honest line between what is quoted, what is documented, what is arithmetic, what is a choice, and what is a reading:
| Feature | Status | What that means |
|---|---|---|
| The chapter | Reported | The list's 400 to 800 and its 1947 start, as In Plain Sight reports the memo (p. 263); Wilson's denial (pp. 263 to 266); Coulthart's own lines (p. 272); the chapter's dates (pp. 261 to 273). Quoted as printed; the bench takes the list's two ends as given and asks a question conditional on them. |
| The three cases | Documented | That PRISM (2013), the Tuskegee study (1972) and the FBI laboratory (1998) came out, and through whom, as the paper describes them. Nothing is added from outside the paper. |
| The paper's estimates | Reported | Each case's N and t, the assumption of L = 0.5 when it came out, Table 1's p, Table 4's limits and the correction's Gompertz figure: the paper's own, which it calls a best case for the secret. The bench reproduces each; one, Table 4's 2,531 at 5 years, differs from the paper's own text and from Eq. 3 solved for N (2,521). |
| The arithmetic | Exact | L = 1 − e^(−t(1 − (1 − p)^N)) and its integral form; the inversion for p; the largest N under a threshold; the Gompertz cohort. No parameter is fitted. |
| The dials | Modelled | Which case sets p; L*; the year; who counts as knowing (the alumni form and its tenure are the bench's own); the threshold. The findings are quoted at the defaults, and the dials show where they part. |
| What it amounts to | Reading | Which calibration fits a program like this one, and whether the memo, or anything else in the chapter, was the leak the model counts, are readings, and the bench declines to make them for you. |
In one line: The chapter’s two numbers are taken as given, the paper’s model and cases as published and checked against its own printed figures, every choice is named as a dial, and the two readings are priced at equal size; whether the program exists, and whether the memo was the leak, are left where the chapter and the paper leave them. The reading is yours.
Sources
- Coulthart, R. (2021; revised edition 2023) In Plain Sight: An Investigation into UFOs and Impossible Science, HarperCollins Australia: chapter 21, "Not Made by Human Hands" (pp. 260 to 273 of the revised edition).
- Grimes, D. R. (2016) "On the Viability of Conspiratorial Beliefs", PLoS ONE 11(1): e0147905, doi:10.1371/journal.pone.0147905: Eqs. 1 to 8, Table 1, Table 4.
- Grimes, D. R. (2016) "Correction: On the Viability of Conspiratorial Beliefs", PLoS ONE 11(3): e0151003, doi:10.1371/journal.pone.0151003: the integral form and the rewritten single-event sentence.
- Signals from the Periphery, Culture 08, the source note on In Plain Sight; the wiki entry on the Wilson memo.