Becoming the Author, Not the Subject

Every chapter has documented how control gets imposed and sustained. This final chapter documents the counter-mechanism — how a pattern of coercion, built from the ground up to be deniable, trivial incident by trivial incident, and endlessly reclassifiable, gets converted into a record that can actually survive contact with a court, a journalist, or simply one’s own memory under stress. Exit is not a single act. It is the point at which a target stops being the subject of someone else’s story and becomes the author of her own.

The evidentiary problem this book’s entire method has to solve was named precisely by Evan Stark: criminal law is built to evaluate discrete incidents — a slap, a threat, a broken phone — while coercive control is not an incident at all. It is a pattern, a micro-regulation of daily life through isolation, financial control, and surveillance, where each individual act looks minor or even entirely legal, and the crime only becomes visible in its cumulative, structural effect.¹ Evidence built one incident at a time will always look weak, because it was never designed to be assessed one incident at a time. Evidence built as a timeline — frequency, escalation, context, a target’s visibly shrinking behavioral freedom over months or years — is what actually proves the mechanism. The legal system has begun, unevenly, to catch up to this insight directly: the UK’s Serious Crime Act made “controlling or coercive behaviour” itself the first major common-law offense to criminalize a pattern rather than its constituent acts,² and the Crown Prosecution Service’s own guidance instructs building a cumulative timeline rather than charging isolated incidents.³ Most jurisdictions still lack an equivalent statute, which means the same documentation discipline has to do double duty everywhere else — building a case under whatever charge happens to be available while simultaneously preserving the fuller pattern record for a future proceeding, or a future jurisdiction, that might actually recognize it.

Sally Challen, the woman whose marriage opened this book’s argument at human scale in the first chapter, is the case that shows what this costs when the record and the recognition both arrive too late. The pattern that eventually freed her, the decades of isolation, financial control, and humiliation this book has been calling coercive control, existed in full the entire time it was happening. What did not exist, in 2010, was any means of making it legible: coercive control was not yet a criminal category in England and Wales, and Challen had kept no contemporaneous record of a pattern she had never been given the language to name. So it had to be reconstructed afterward — by her son, by lawyers, by expert witnesses, from fragments, years later, on the far side of a killing and a murder conviction — before any court would agree to see the shape that had been there the whole time.⁴ When the Court of Appeal finally quashed her murder conviction in 2019, it did so on fresh evidence — psychiatric diagnoses never presented at her trial — with the decades of coercive control admitted, at last, as the context that made her state of mind legible. Coercive control was not, the court held, a defence to murder in itself; what its reconstruction changed was whether the rest of her could be understood at all. That is the argument proven in the negative: the reconstruction that saved her, nine years and one tragedy late, is exactly the reconstruction this chapter exists to help someone build while it is still happening, contemporaneously, and in safe custody, so that it never has to be salvaged from wreckage at the far end of one.

The most counterintuitive finding in this entire literature is that leaving is the most dangerous moment, not the safest one. Jacquelyn Campbell’s Danger Assessment, developed at Johns Hopkins and now the standard validated instrument used by advocates, clinicians, and courts, established that separation is the period of highest lethality risk — not its resolution.⁵ For an abuser — or, at any scale this book has examined, an operation whose core resource is a target’s compliance — the act of leaving reads as the control system itself visibly failing, and escalation at that exact moment is not irrational. It is often the last lever still available to reassert an equilibrium that is slipping away in real time.

Separation, though, is only one item on that instrument, and a book that quotes it for one finding owes the reader the rest — because the Assessment’s real work is not the general warning but the distinction it draws between one situation and another. It exists to tell apart two people who are both frightened and are not in the same danger, and it does that on exactly this book’s principle: not by how the situation feels from inside it, but by markers a stranger can check. The eleven-city case-control study the instrument was built on found the sharpest of them to be an abuser’s access to a gun, a previous threat with a weapon, and estrangement from a controlling partner — that last being why the separation finding above is not a free-standing fact but a multiplier on everything standing beside it.⁶ One marker has since been isolated, and it is worth a reader’s whole attention. In later case-control work from the same group, women who had been strangled by a partner and survived it carried roughly seven and a half times the odds of later being murdered by him, set against abused women who had not been. A hand on the throat — once, briefly, leaving no mark — is among the most powerful single predictors anyone has measured in this literature.⁷ It is why England and Wales made non-fatal strangulation a criminal offence in its own right in 2022, deliberately requiring no proof of injury, rather than leaving it to be charged as a common assault and vanish into the background noise of a file.⁸ Threats to kill, escalating frequency or severity, stalking, forced sex, an abuser’s own threats of suicide, and violence during pregnancy each carry their own weight in the scoring. None of these is a reason to act alone on a self-assessment: the Assessment is free, it takes minutes, and any domestic-violence advocate is trained to administer and interpret it. But a reader who has been strangled and does not know what that predicts is missing the one fact in this book most likely to save their life, and the undifferentiated warning that leaving is dangerous — delivered to someone at low risk and someone at the highest alike — is not a substitute for knowing which of the two they are. Calibration is the point. The general caution, uncalibrated, can just as easily keep a person in.

The Assessment measures the danger of being killed. It does not measure the other death, and for a long time neither did anyone else. The police in England and Wales now count both, and the count unsettles something this chapter would otherwise have let stand: that the machine’s lethal endpoint, at this scale, is the abuser killing the target. In the year to March 2023 the Vulnerability Knowledge and Practice Programme recorded ninety-three suspected victim suicides following domestic abuse, against eighty intimate-partner homicides — the first year in which the suicides outnumbered the killings, and not, on the programme’s own reading, because the toll had risen, but because the counting had improved.⁹ The machine has always produced this death. What changed is that someone finally wrote it down.

That it can be charged at all is newer still. In 2017 Nicholas Allen pleaded guilty at Stafford Crown Court to the manslaughter of Justene Reece, and to stalking and controlling or coercive behaviour, and was sentenced to ten years. Reece, forty-six, had left him and gone to a refuge; across the six months that followed, he attempted to reach her and her family roughly three and a half thousand times. She left a note saying she had run out of fight. Women’s Aid called the outcome a historic first.¹⁰

Those four words are not a figure of speech, and they are not despair. They are the exact condition Chapter 1 named and Chapter 2 traced back to its engine — not a person who declined to fight, but a person taught by long instruction that fighting changed nothing. Learned helplessness, in that note, stops being a laboratory finding. It is a cause of death, written in the victim’s own hand, in the vocabulary this book has used since its first chapter.

Her case belongs in this chapter rather than an earlier one because of what the Independent Office for Police Conduct found afterward. Between September 2016 and her death in February 2017, thirty-four incidents involving Reece or connected to her reached Staffordshire Police. The force did not cross-reference them. Each was worked as its own small thing, and the scale of what was being done to her was therefore never assembled by anybody — the watchdog’s finding was that the failure to join the incidents was itself the failure.¹¹

That is this book’s thesis, reached independently, by a statutory investigator, at the cost of a life. A pattern engineered to be deniable is deniable, one incident at a time; it becomes undeniable only laid end to end. The police held the record. What nobody held was the timeline. The discipline this chapter asks for is not clerical fussiness, and it is not only for courts: it is the act of assembling, on your own behalf, the thing that institutions with every resource and every duty have repeatedly failed to assemble on yours.

This is also why documentation and exit planning can never be sequential — leave, then document — and instead have to run concurrently and, critically, covertly, because the surveillance infrastructure this book has documented at every earlier scale is frequently still active during exit, and often grows more sensitive to detection at exactly the moment departure is being planned. Standard advocate guidance now bakes evidence preservation directly into the safety plan itself — copies of documents stored outside the home, an emergency bag, a trusted contact holding duplicates — not as an afterthought reached once safety has already been achieved, but as part of achieving it.¹²

What actually makes a record hold up, once it exists, follows four principles this book’s own method has applied throughout. Contemporaneity beats recall: courts, clinicians, and journalists alike weight records made at the time far higher than reconstructed memory, not because reconstructed memory is dishonest, but because trauma-affected memory is documented to compress, reorder, and lose specific detail under sustained stress.¹³ A dated, specific note made in the moment — this day, this time, this is what was said — is evidentiary. A narrative summary reconstructed six months later is testimony, which is weaker and directly attackable. Corroboration defeats the single-witness problem that DARVO depends on entirely: a friend texted in real time, a neighbor who heard something, a medical record made for an unrelated reason that happens to note an injury, a financial record showing a pattern of transactions — each moves the record out of the he-said-she-said frame this book has shown collapsing in DARVO’s favor every time it stays a two-person dispute. Originals matter more than summaries, especially in digital form: standard forensic practice defines a four-phase process of collection, examination, analysis, and reporting specifically to keep digital evidence admissible,¹⁴ which means preserving the original artifact rather than a paraphrase, preserving metadata alongside content rather than content alone, exporting and backing up before access is lost — since shared devices and joint accounts are exactly how evidence disappears before a target even realizes it needs preserving — and storing copies with an independent custodian outside the household entirely.

One point here is safety-critical and easy to skip past. Because the surveillance described throughout this book is often still live, and most alert, precisely during separation, this copying must be done on a device the other person cannot see or reach — a friend’s phone, a library or workplace computer — never on a shared or monitored phone, and never into a cloud account they may control. The act of exporting can itself raise the alarm the whole effort depends on not raising. One corollary is safety-critical and easy to get backwards: preserving evidence is not the same as hunting for it. Do not try to find, open, or remove suspected monitoring software on your own device — removing it can alert the person who installed it as surely as exporting can, and probing for it can do the same. If you believe a device is compromised, that is a job for a specialist, not for the person the device is watching; the Resources appendix at the back of this book lists where to turn.

And the record has to be written for a reader who does not already believe you: the habits that make a private diary useful, interpretation, editorializing, emotional processing, are exactly the habits that make a record easy to attack under skeptical or adversarial reading. The evidentiary register looks more like an incident log — date, time, what was directly observed or said, who else was present, what happened immediately after — with interpretation and emotional context kept in a separate, clearly labeled companion record, useful for a survivor’s own processing without weakening the evidentiary one underneath it. That companion record is itself discoverable, and if found it becomes ammunition — the private fears and interpretations in it are exactly what a reversal narrative feeds on. It has to be kept as securely as the evidentiary log, on the same unmonitored device or none at all; the same is true of the physical fact of this book, which is its own trace if left where the wrong person finds it.

This is, precisely, the direct counter to DARVO, the mechanism this book named early and has returned to at every scale since. DARVO functions because at the moment of confrontation there is no fixed record to anchor against — a perpetrator’s denial and reversal only need to outcompete memory, and memory recalled under pressure is always contestable in real time. A contemporaneous record predates the reversal narrative by simple construction: it was made before the confrontation ever existed to attack, which is precisely the one thing memory recalled under pressure can never claim for itself. This is the load-bearing reason documentation functions as a counter-mechanism to a specific, named tactic in this book, rather than as generic advice dispensed at the end of a difficult story. It defeats DARVO because of when it was made, not merely because of what it says.

Recovery and documentation are not separate processes running side by side. They are sequenced together, and Judith Herman’s three-stage model — examined in this book’s opening chapter as the clinical capstone that first proved this project’s entire thesis — maps directly onto when documentation work is safe to do and when attempting it causes further harm.¹⁵ The first stage, safety and stabilization, is when documentation has to stay covert and minimal, maximizing operational security rather than narrative completeness; this is triage, not the moment for full processing. The second stage, remembrance and mourning, is where the fuller record gets built out — the timeline, the pattern, the bearing witness to one’s own story that Herman describes clinically — once physical safety is no longer the binding constraint, and the record begins doing double duty: simultaneously therapeutic, in externalizing and sequencing what happened, and evidentiary, because it is the same timeline a court or an advocate would eventually need. The third stage, reconnection, is where the record’s function shifts again, from a private survival tool into something that can finally be shared — with an attorney, an advocate, or, at the scale this book has operated at throughout, converted into documented case material. This chapter has an honest limit: forcing evidentiary rigor during the first stage, treating someone in acute danger as though they were already a witness building a case file, is itself a documented secondary harm in the advocate literature. Documentation exists to serve safety and, eventually, justice. It should never be imposed as an obligation competing against either one.

The same discipline this chapter has described at interpersonal scale turns out to be structurally identical to what protects a whistleblower exposing an institution or a state — contemporaneous records, independent corroboration, secure custody outside the compromised system, and a record built specifically to survive a credibility attack rather than merely to persuade an already-sympathetic reader. Whistleblower-protection guidance recommends precisely these same four elements, derived completely independently from an entirely different starting point.¹⁶

Read one scale up, this book’s earlier cases show what each principle costs when it fails. Leymann’s mobbing research supplies the same warning this chapter drew from DARVO: authority adopts the group’s narrative before ever hearing the target’s side, which means a record has to predate the confrontation to mean anything — whether the confrontation is a partner’s denial or a manager’s closed-door meeting with everyone except the person being discussed.¹⁷ Karen Silkwood was driving to hand a folder of documents and photographs to a journalist and a union official, proof of the safety violations she had spent months compiling, when she died in a car crash under circumstances this book has held honestly as unresolved. The folder was never found.¹⁸ Hers is this project’s starkest illustration of why the fourth principle — secure custody with an independent party, before it is needed rather than after — is not bureaucratic caution. It is the difference between a record that outlives its author and one that doesn’t. The Senate Select Committee’s Torture Report, built from more than 6.3 million pages of the CIA’s own contemporaneous cables, emails, and internal records, is the same principle at the largest scale there is.¹⁹ And William Binney had already done what this chapter recommends, filing through the Department of Defense’s own Inspector General rather than waiting to be asked, and was raided at gunpoint anyway — proof that proper documentation does not guarantee protection, only that it remains the precondition for anyone to ever prove what happened, however long that proof takes to matter.²⁰

The tools are identical. Only the scale of what they’re used against, and how long the record has to survive before anyone is willing to read it, ever changes.

The Framework in the Wrong Hands

A book that hands the reader a vocabulary for coercive control has to answer for who else can pick it up. The vocabulary of victimhood is not access-controlled. An abuser can learn the words isolation, gaslighting, coercive control as readily as anyone, and the most advanced form of DARVO — the tactic this book named at the start and has tracked at every scale since — is precisely the perpetrator who has read the literature and reaches the victim label first, recasting the person they harmed as the real aggressor in fluent, up-to-date clinical language. If the framework could be run by whoever deploys its vocabulary most persuasively, it would be worse than useless: it would be a weapon handed to the party already better at controlling the story.

It cannot be, and the reason is the whole of this chapter. The discipline this book runs does not adjudicate by vocabulary, and it never has. It does not decide who is the victim by who sounds most like one, who reaches the language first, or who deploys the terms most fluently — every one of which the more practiced operator tends to win. It adjudicates by the same evidentiary tests the record was built to survive: which account predates the confrontation rather than being assembled after it, which claims are independently corroborated rather than merely asserted, and which direction the pattern of control actually ran when measured over time rather than performed in a single charged moment. Those tests are indifferent to fluency. An abuser can learn to say coercive control by tomorrow morning. What no one can do is retroactively manufacture a contemporaneous, independently corroborated record showing that the isolation, the financial control, and the surveillance ran the other way — because that record, had it existed, would have needed to be made before there was any confrontation to win, by someone who was not, in fact, the one running the machine.

This is the same firewall the book turns on its own material, pointed now at its own reader. The framework does not license concluding abuse from resemblance, from vocabulary, or from who tells the more sympathetic story; it requires the evidence, held to the same three tiers as every claim in this book. That requirement is exactly what the bad-faith user cannot satisfy, which is why the framework in the wrong hands is self-limiting rather than self-amplifying. It is the discipline, not the words, that does the work — and the discipline is the one thing the person weaponizing the words does not have.

Two limits keep this honest. The discipline protects the analysis — it will not let this book, or a careful reader, license a conclusion of abuse from fluency, sympathy, or vocabulary. It cannot by itself protect a victim standing in a biased forum, and this book has spent its length arguing that the forums are often biased: credibility is decided by position as much as by proof, and the party with more resources is also the party better placed to build, or to fake, a record that predates the confrontation. The record shifts the odds, sometimes decisively; it does not guarantee the outcome against a well-resourced bad-faith actor in a room already inclined to believe him. It is the strongest instrument the weaker party has. It is not, and this chapter should not pretend it is, a certainty.

---

None of this is the end the word recovery usually promises, and it should not be sold as one. The record is not, finally, the point of building it. It is the instrument; the point is the person who is no longer inside the machine. Herman’s third stage is named reconnection for a reason the evidentiary frame can obscure: the record’s last function is not to convict anyone but to make the story finally shareable — to return the person who lived it to people who believe them, which is the exact repair for the isolation the whole mechanism opened with. And for the reader who will never see a courtroom, never file a complaint, never hand the record to anyone at all — the reader for whom there is no proceeding coming and never will be — the reconstruction still does its first work, the work that needs no audience to be real: it is how she takes her own account of what happened back from the version that was installed in its place. Being believed by others is the repair for the isolation; believing herself again, on the evidence she assembled, is the repair for the part of the machine that had gotten inside. The record can do that second thing even if no one ever reads a line of it. Sally Challen was not made free the day her conviction was quashed; she moved toward free at the point her account stopped being the unreliable one and became the one the record supported. That is what it means to move from subject to author — not that the harm is undone, which no record can do, but that the person it happened to is the one now holding the pen. The machine’s first move, at every scale in this book, was to take that pen away. The last thing this chapter has to say is that it can be taken back.

Notes

1. Evan Stark, Coercive Control: How Men Entrap Women in Personal Life (Oxford University Press, 2007; 2nd ed. 2023). Stark’s central legal argument: criminal law evaluates discrete incidents, while coercive control is a pattern — micro-regulation of daily life through isolation, financial control, and surveillance — that becomes visible only in its cumulative, structural effect.↑

2. UK Serious Crime Act 2015, s.76 — “Controlling or coercive behaviour in an intimate or family relationship” — the first major common-law statute to criminalize the pattern itself (repeated or continuous behaviour, serious effect on the victim, defendant knew or ought to have known) rather than constituent acts. https://www.legislation.gov.uk/ukpga/2015/9/section/76/enacted.↑

3. Crown Prosecution Service, prosecution guidance on controlling or coercive behaviour in an intimate or family relationship, which instructs building a cumulative timeline (frequency, context, pattern of control) rather than charging isolated incidents. https://www.cps.gov.uk/prosecution-guidance/controlling-or-coercive-behaviour-intimate-or-family-relationship.↑

4. Sally Challen’s 2011 murder conviction was quashed by the Court of Appeal in 2019 (R v Challen [2019] EWCA Crim 916). The decisive fresh evidence was psychiatric — conditions not diagnosed at her original trial — supporting diminished responsibility; the coercive control she had lived under, a category introduced into English criminal law only by the Serious Crime Act 2015 (see note 2), after her trial, was admitted as relevant context, the court holding that coercive control is not itself a defence to murder. At retrial the prosecution accepted a plea of manslaughter by reason of diminished responsibility and she was released on time already served. This case is introduced in full in Chapter One; see also Centre for Women’s Justice, “The Case of Sally Challen” (2019), https://www.centreforwomensjustice.org.uk/news/2019/12/16/the-case-of-sally-challen, and D. Storey, “Coercive Control: An Offence but Not a Defence,” Journal of Criminal Law 84, no. 1 (2020).↑

5. Jacquelyn Campbell, Danger Assessment (developed 1986, Johns Hopkins School of Nursing) — a validated 20-item lethality-risk instrument used by advocates, clinicians, and courts, establishing separation as the period of highest intimate-partner-homicide risk. https://www.dangerassessment.org/; peer-reviewed validation: “The Danger Assessment: Validation of a Lethality Risk Assessment Instrument for Intimate Partner Femicide,” https://pmc.ncbi.nlm.nih.gov/articles/PMC7878014/.↑

6. Jacquelyn C. Campbell, Daniel Webster, Jane Koziol-McLain, et al., “Risk Factors for Femicide in Abusive Relationships: Results From a Multisite Case Control Study,” American Journal of Public Health 93, no. 7 (2003): 1089–1097. An 11-city study comparing proxy interviews for 220 intimate-partner femicide victims against 343 abused control women. Preincident risk factors associated with increased risk in multivariate analysis included the perpetrator’s access to a gun, a previous threat with a weapon, a stepchild in the home, and estrangement — particularly from a controlling partner. https://doi.org/10.2105/AJPH.93.7.1089↑

7. Nancy Glass, Kathryn Laughon, Jacquelyn Campbell, et al., “Non-fatal strangulation is an important risk factor for homicide of women,” Journal of Emergency Medicine 35, no. 3 (2008): 329–335. Case-control design: 506 completed and attempted homicides against 427 abused controls. Prior non-fatal strangulation was reported by 10% of abused controls, 45% of attempted-homicide victims, and 43% of homicide victims, and was associated with odds of 6.70 (95% CI 3.91–11.49) of becoming an attempted-homicide victim and 7.48 (95% CI 4.53–12.35) of becoming a homicide victim. https://pubmed.ncbi.nlm.nih.gov/17961956/↑

8. Domestic Abuse Act 2021, section 70, in force 7 June 2022, inserting section 75A into the Serious Crime Act 2015 — the same Act whose section 76 created the controlling-or-coercive-behaviour offence relied on throughout this book. The offence covers strangulation and any act affecting the victim’s ability to breathe, requires no proof of serious physical injury, and carries a maximum of five years’ imprisonment on indictment. See Home Office Circular 2022/01. https://www.legislation.gov.uk/ukpga/2021/17/section/70↑

9. Vulnerability Knowledge and Practice Programme, Domestic Homicides and Suspected Victim Suicides, Year 3 report, published 13 March 2024, covering April 2022 to March 2023. Of 242 domestic-abuse-related deaths recorded in England and Wales: 93 suspected victim suicides following domestic abuse, 80 intimate-partner homicides, 31 adult family homicides, 23 unexpected deaths, 11 child deaths, 4 other. The report attributes suspected victim suicides overtaking intimate-partner homicides for the first time to “increased awareness and improvements in recording information by officers” rather than to any rise in the underlying toll — a counting artefact, and the direction of the artefact is the point. https://www.vkpp.org.uk/news/report-reveals-scale-of-domestic-homicide-and-suicides-by-victims-of-domestic-abuse/↑

10. Nicholas Allen pleaded guilty at Stafford Crown Court in June 2017 to the manslaughter of Justene Reece, 46, and to stalking and engaging in controlling or coercive behaviour; he was sentenced in July 2017 to ten years’ imprisonment with an extended licence period. Reece had left Allen and entered a Women’s Aid refuge in November 2016; Allen attempted contact with her and her family approximately 3,500 times in the six months before her death in February 2017. Katie Ghose, then chief executive of Women’s Aid, said of the sentence: “the result in court today has not been seen before, it is a historic first.” Widely reported; see ITV News Central, 28 July 2017, and Express & Star, 27 June 2017.↑

11. Independent Office for Police Conduct investigation into Staffordshire Police’s handling of the case, reported April 2019. The IOPC examined the police response to 34 incidents either directly involving or connected to Justene Reece between September 2016 and her death in February 2017, and found that a failure to cross-reference those incidents led to them being treated in isolation, and to a lost opportunity to recognise the scale of Allen’s offending.↑

12. National Domestic Violence Hotline safety-planning framework (thehotline.org), which builds evidence preservation directly into standard safety planning — copies of documents stored outside the home, an emergency bag, a trusted contact holding duplicates.↑

13. Judith Herman, Trauma and Recovery (1992). Herman’s three-stage recovery model — safety and stabilization, remembrance and mourning, reconnection — and her clinical documentation of how sustained traumatic stress compresses, reorders, and strips detail from memory underlie this chapter’s argument that contemporaneous documentation outperforms reconstructed recall.↑

14. NIST Special Publication 800-86, “Guide to Integrating Forensic Techniques into Incident Response,” which defines the standard four-phase digital-forensics process — collection, examination, analysis, and reporting — used to keep digital evidence admissible. https://csrc.nist.gov/pubs/sp/800/86/final.↑

15. Judith Herman, Trauma and Recovery (1992) — the three-stage recovery model (safety and stabilization, remembrance and mourning, reconnection) applied here to when documentation work is safe to attempt versus when it causes further harm. Same source as note 6.↑

16. Government Accountability Project whistleblower guidance, recommending contemporaneous records, independent corroboration, secure custody outside the compromised system, and a record built to survive a credibility attack. https://whistleblower.org/whistleblower-guides/; see also “Make a Note to the Record” (GAP + Union of Concerned Scientists), via https://www.ucs.org/resources/resources-federal-scientists.↑

17. Heinz Leymann, “Mobbing and Psychological Terror at Workplaces,” Violence and Victims 5 (1990): 119–126, https://pubmed.ncbi.nlm.nih.gov/2278952/. Leymann’s four-stage mobbing model documents authority figures adopting a group’s narrative before hearing the target’s side. This case is examined in full in Chapter 32.↑

18. Karen Silkwood, a lab technician and union safety investigator at Kerr-McGee’s Cimarron plutonium plant, died in a car crash on November 13, 1974, while driving to meet a New York Times reporter and a union official with a folder of documents and photographs she said would prove safety violations; the folder was never found, and marks on her car’s bumper were consistent with being forced off the road. Police ruled it an accident; the cause remains genuinely disputed (Silkwood believed she had been deliberately contaminated; Kerr-McGee alleged she had contaminated herself). Silkwood v. Kerr-McGee Corp., 464 U.S. 238 (1984); Richard Rashke, The Killing of Karen Silkwood (1981). This case is examined in full in Chapter 29.↑

19. Senate Select Committee on Intelligence, Committee Study of the Central Intelligence Agency’s Detention and Interrogation Program (executive summary declassified December 9, 2014), built from a roughly six-year review of more than 6.3 million pages of the CIA’s own internal cables, emails, and records. https://www.intelligence.senate.gov/sites/default/files/publications/CRPT-113srpt288.pdf. This case is examined in full in Chapter 43.↑

20. William Binney, a thirty-year NSA veteran, filed a complaint through the Department of Defense’s own Inspector General in 2002 alleging waste, fraud, and mismanagement; in July 2007 a dozen armed FBI agents raided his home in an unannounced early-morning operation, and he was never charged with anything. Jane Mayer, “The Secret Sharer,” The New Yorker, May 23, 2011. This case is examined in full in Chapter 29.↑