Method
How a record qualifies
A record qualifies when the victim was subject to the perpetrator's spiritual authority, and that authority is what made the harm possible or unrefusable.
Everything on this page exists to make that one sentence checkable, case by case. The single test applied to every candidate record is: did the belief system give this perpetrator power over this victim? Not "is this tradition associated with harm," not "did this person claim religious motivation" — specifically, did the victim's position under this perpetrator's claimed spiritual authority make the harm possible, or make refusal impossible.
Scope
What is excluded, and why
Exclusion is not a judgment that these things don't matter. It is a claim that they are a different phenomenon than the one this corpus tracks, and blurring them together would corrupt the criterion.
Religious persecution
The inverse category: a state or majority prosecuting, jailing, or harming people for their religious belief or practice. This corpus tracks a perpetrator's spiritual authority over a victim — persecution runs in the opposite direction and is out of scope.
Outsider-directed violence and religious terrorism
Attacks carried out in the name of a belief system against people outside it — including terrorism that uses religious imagery or rhetoric — fall outside the criterion, because the victim was not under the perpetrator's spiritual authority to begin with.
Family-only nexus
Ordinary family violence that happens to occur inside a religious household does not qualify on family membership alone. The exception: cases where doctrine or claimed spiritual authority is what enabled or compelled the specific harm — the doctrinal-coercion exception — which do qualify.
Mere affiliation
A perpetrator's membership in a religious group, or a victim's shared faith with a perpetrator, is not sufficient on its own. The authority relationship — not shared identity — is what the criterion tests for.
The heart of the method
Evidence tiers
Every record is labeled by the strength of what supports it. The tiers are never blended into a single undifferentiated "documented" status.
Tier 1
Criminal convictions that stood
A conviction qualifies only after appeals are checked. If a conviction was overturned, vacated, or the sentence reduced to acquittal on appeal, the case does not qualify as a standing conviction — full stop. A reversed conviction never re-enters the record under this tier.
Tier 2
Official commission and inquiry findings
Findings from royal commissions, government inquiries, parliamentary reports, and UN or state anti-trafficking assessments. These are official findings, not criminal verdicts, and are labeled as such.
Tier 3
Civil findings and court-supervised settlements
Civil judgments, and settlements reached under court supervision or as part of a legal process (bankruptcy-supervised abuse settlements, for example). Labeled as civil outcomes, never conflated with criminal guilt.
Unlabeled tier
Allegations — never presented as findings
Credible, sourced allegations that have not (yet) produced a conviction, official finding, or civil resolution are tracked separately and marked as allegations. They are never counted in the core tally and never presented as though a finding exists.
Taxonomy
Record classes
Core
1,023 records
The defining class: the victim was under the perpetrator's spiritual authority, and that authority made the harm possible or unrefusable. This is the number the site counts and leads with.
Context
95 records
Victim-directed ritual — cases where religious imagery or ritual apparatus was turned into the instrument used against a victim, but the case sits adjacent to the core test rather than squarely inside it. Distinct from religious terrorism, which is out of scope entirely (see Excluded, below).
Institutional-event
130 records
Institutional findings — a commission report, an internal review, a pattern finding — that document an institution's structure or failure rather than a single victim's case. Shown as map annotations; excluded from the core count.
Separate register
16 records
A separate register, held to its own sourcing standard and excluded from the core count. It is not a chronological class — the register runs from 1906 to 1997, and the historical-class records inside the main corpus run from 1937 to 2017. Most of the latter are moral-panic prosecutions in which the accusation itself was false and the harm was done by the state: McMartin, Kern County, Wenatchee, Little Rascals, Orkney, Rochdale, Oak Hill, the San Antonio Four, the West Memphis Three. They are kept because the panic is part of this subject's history and because a database of ritual-abuse claims that omitted the discredited ones would misrepresent the record. The remainder are cases whose spiritual-authority nexus is real but whose facts sit outside the core test.
Candor, not confidence
Known gaps and uncertainty
Being candid about what this corpus cannot show is the point of this page, not a caveat tacked onto it. A record count without its coverage limits is a claim this archive will not make.
The record skews recent, and that is a searchability artifact
The archive covers 1819 to 2026. Within that span the distribution is steeply uneven: roughly 80% of records postdate 2010, and only 45 predate 1990. That is not a claim that spiritual abuse is a recent phenomenon — it reflects what is digitized, indexed, and findable in English-language search. Older cases exist in paper archives, untranslated local press, and jurisdictions with no searchable court database. The earliest year here is a limit of the record, not of the phenomenon, and the archive is actively pushing it back.
Christian records are roughly half the corpus, and that is a publishing artifact
Christian-context cases make up about half of all records. That proportion tracks which jurisdictions publish court decisions and inquiry findings in indexed, English-accessible form (the US, UK, Ireland, Australia) — not a claim about which tradition produces more harm.
Nineteen countries have zero records, and that is usually an access wall, not an absence of crime
The countries with no record in this corpus are: Ethiopia, Estonia, Uzbekistan, Turkmenistan, Armenia, Azerbaijan, Myanmar, Laos, Afghanistan, Iran, Syria, Jordan, Lebanon, Somalia, Rwanda, Burundi, Ivory Coast, Cuba and Trinidad. A further six hold exactly one: the Democratic Republic of the Congo, Mongolia, Tajikistan, Cambodia, Iraq and Cameroon. For most of these the honest read is an access wall — PDF-locked court databases (Uzbekistan publishes decisions as unsearchable PDFs), anti-bot-blocked legal portals, and thin digitization of routine trials — not that nothing happened there. Four are empty for a different reason: in Afghanistan, Iran, Syria and Somalia most documented religious harm is state persecution, which this archive excludes by design, so an empty cell there reflects the scope of the criterion rather than a gap in searching. Estonia is the one case where a genuinely low base rate is the credible explanation, and even that is inference. An empty cell in this corpus is a research gap to disclose, not a finding to report.
The largest gap is not in this archive's research — it is that sexual violence is almost never reported at all
Every record here began with someone reporting a crime, and most victims never do. Across jurisdictions, victimization surveys consistently find that the majority of sexual assaults are never reported to police, and that reported cases fall away again at each subsequent stage — charge, prosecution, conviction — so that only a small fraction of assaults ever produce the court record this archive is built from. The reasons are structural, not incidental: reporting exposes a woman to disbelief, to invasive forensic examination, to cross-examination on her sexual history, and to the prospect of facing the person who assaulted her in a courtroom. That is retraumatization delivered by the process itself, and declining it is a rational act, not a silence to be explained away. Children report less still, and later. A child abused by a spiritual authority is typically dependent on that authority, is often told the abuse is sacred or deserved or their own fault, and frequently discloses only in adulthood if at all — which is why so many cases in this archive surface decades after the fact, and why the year attached to a record is often the year of the trial rather than the year of the harm. Add the specific pressures of a religious community — the fear of destroying a congregation, of losing family and standing, of being counted a traitor to the faith — and underreporting compounds. Every number on this site should be read against that: this archive counts prosecutions, and prosecutions are a small and unrepresentative sample of what happened. It is a floor, never an estimate.
Convictions are easier to find than acquittals or cases never prosecuted
Search indexes, news coverage, and appellate records concentrate on cases that were charged and won. That structurally biases any court-record-based corpus — this one included — toward the prosecuted. Non-prosecutions and acquittals are undercounted by construction, not by editorial choice.
Room for error
Corrections and standing traps
Sentences reduced or suspended on appeal are stated as what they are, not repeated as the headline figure from the original trial. Where an appeal changed the outcome, the current status is what the record shows.
Standing traps — specific case identifiers and dates for convictions that were later overturned — are documented internally precisely so an overturned conviction can never quietly re-enter the record through a future update. This is a discipline against the corpus contradicting itself over time, not a promise that no error has ever occurred.
The corpus also records honest empties: countries, decades, or traditions with zero qualifying records are left at zero rather than padded, and the gaps above explain why an empty cell is not evidence of an empty history.
If you find an error — a misclassified record, an overturned conviction not yet reflected, a source that no longer supports its citation — submit a correction through the same encrypted intake used for new records. Every submission enters a private verification queue; nothing publishes automatically.