Source Hierarchy

How CLCI Hub ranks the evidence behind every claim — what counts as primary, secondary, and tertiary; how court records, government reports, academic work, journalism, books, and survivor testimony are weighed; and how we handle contested evidence and the distinction between allegation and finding.

This page is a longer, more detailed companion to the Source Policy. The Source Policy is the canonical short-form document covering tiers and confidence weighting; this page adds explicit guidance on contested evidence, allegations versus findings, and the boundaries between “reported” and “documented.”

The three reliability tiers

Primary sources

Primary sources are first-order public-record documents produced through a process with adversarial review or institutional accountability. They are the strongest evidence available and a single substantive primary source can move a profile's confidence to High on its own.

Court records. Criminal verdicts, civil judgments, sentencing transcripts, deposition transcripts, and exhibits entered into evidence. We cite the case name and jurisdiction; where a stable public URL exists (PACER, CourtListener, government court portals), we link to it. Sealed material is excluded until unsealed.
Government investigations. State attorney-general reports, royal commission findings, parliamentary inquiry reports, regulator enforcement actions, charity-commission decisions, tax-court rulings. These are public-record documents authored by institutions with statutory authority.
Sealed-then-released documents. Material disclosed under Freedom of Information Act, Subject Access Request, or equivalent statutory process. We treat these as primary once they are public.
Organisational public statements. The group's own published material is primary evidence of what the group says about itself. We cite the URL and date and quote sparingly in our own voice.

Secondary sources

Secondary sources are independently produced material reviewing or analysing primary material. They carry substantial evidentiary weight and a profile supported by multiple corroborating secondary sources can also reach High confidence.

Peer-reviewed academic work. Journal articles, university-press books, doctoral theses, and conference papers from established disciplines (religious studies, sociology of religion, cult studies, social psychology). Peer review is itself a form of institutional accountability.
Long-form investigative journalism from outlets with documented editorial standards, corrections policies, and post-publication review processes. Examples include the BBC, The New York Times, The Guardian, Reuters, Le Monde, Der Spiegel, Christianity Today, The Hindu, Asahi Shimbun, Politico, ProPublica, and equivalent. Single investigative pieces from less-established outlets are weighted lower than the staffed-investigations-desk pieces from major outlets.
Independent documentary films from production houses with editorial review and a track record of accurate reporting (Netflix Originals where they have editorial oversight, HBO Documentary Films, BBC Storyville, PBS Frontline). Single-creator documentaries are weighted lower than studio-produced material.
Books from major trade publishers that undergo editorial fact-checking. Memoirs from named former members fall here when published by such houses.

Tertiary sources

Tertiary sources are aggregators or commentary built on top of primary and secondary material. They are useful pointers but rarely sufficient alone.

Aggregator and encyclopaedia entries (Wikipedia, Apologetics Index, cult-information websites). Used to locate primary and secondary sources; not used as primary support for specific claims.
Single-author opinion pieces in reputable outlets. These can frame a debate but rarely settle it.
Self-published memoirs and ex-member blogs. Sometimes the only source for a specific personal experience; we cite them with named attribution and treat them as personal-testimony tertiary sources.
Podcast episodes. Used as pointers to the underlying source the podcast is reporting on.

Ex-member testimony

Ex-member testimony is the most contested category in cult-research sourcing. Critics of the field have argued that ex-member accounts are inherently unreliable; defenders have argued that they are often the only window into organisations that operate under strict internal secrecy. The honest position is that ex-member testimony varies enormously in evidentiary weight, and treating it as a single category is itself misleading.

We apply a consistent four-level framework:

Court-attested testimony. A named former member who has given sworn evidence in court, signed an affidavit, or testified in a government inquiry is being held to a higher accountability standard than someone posting online. Where that testimony has survived cross-examination, we treat it as a secondary source for the specific facts the witness personally observed.
Named, public ex-member statements. A former member named in a journalistic investigation, a published memoir, or an on-the-record interview is staking their own reputation on their account. We cite by name and treat as a secondary source for personal-experience claims, weighted against corroboration from other sources.
Survivor-network materials (Telling The Truth for Two by Twos, Holding Out Help for FLDS, Mental Health Resources for Anabaptist Exits, exJW survivor networks, etc.) are treated as secondary where the network has editorial review and links to primary documents; otherwise as tertiary.
Anonymous online testimony (Reddit, forums, social media, anonymous blog posts) is treated as Low-reliability tertiary. We use it to confirm patterns already documented in primary or secondary sources, never as the sole support for a specific factual claim about a specific person or incident.

Allegation versus finding

Distinguishing allegation from finding is one of the most important editorial moves on CLCI Hub. The language we use signals which we are reporting.

Finding. A court verdict, a government investigation conclusion, or a peer-reviewed academic claim that has survived disciplinary review. We can write: “Convicted of X in Y court, sentenced to Z.”
Allegation. A claim that has been made publicly but not yet adjudicated, or that has been disputed. We write: “Alleged by X (year) to have done Y; the organisation disputes this; the matter remains unresolved in public record.”
Contested allegation. A claim where reasonable observers disagree on the underlying facts. We describe the dispute rather than taking a side. We may write: “X has been alleged by multiple former members and one investigative journalist; the organisation has issued a public denial and threatened litigation; no court or government body has ruled on the claim.”

We never present an allegation as a finding. Where a court has explicitly declined to find on a question (acquittal, dismissal for lack of evidence, statute-of-limitations bar), we say so. Where reporting on an allegation has been retracted, we treat that retraction as a primary source about the unreliability of the original report.

How conflicting evidence is handled

When primary and secondary sources disagree, we report the disagreement and explain how we weighted the evidence. We do not silently pick a side. Standard patterns we apply:

Court finding contradicts journalistic claim. Court finding carries the higher weight; we cite both, dated, and explain the discrepancy.
Multiple independent investigations corroborate; one source dissents. We follow the corroborated account and cite the dissenter with attribution.
Organisation's public statement contradicts ex-member testimony. Both are cited. We weight by the broader source mix and explicitly note where the organisation has the stronger documentary record and where the ex-member account does.
Older finding superseded by newer finding. The newer finding governs the current rating; the older is noted with date.
Reformed practice contradicts historical record. The historical record is preserved in the timeline and the score reflects current practice. See Score Updates for the reform-credit mechanism.

Why anonymous reports are not treated as conclusive

Anonymous reports come from a wide range of motivations: genuine survivors protecting themselves from retaliation; current members defending the group without identifying themselves; bad-faith actors pursuing personal grievances; researchers protecting their own identity from a litigious organisation. We cannot, from outside, distinguish these cases reliably enough to base a published claim on a single anonymous source. We can and do use anonymous accounts as a pointer: if a pattern emerges across many anonymous reports we will look for primary or secondary sources that document the same pattern.

This is a deliberately conservative posture. We accept that it sometimes means a real pattern of abuse remains uncaptured on a profile until a primary source emerges. We believe the alternative — publishing serious claims about named individuals on anonymous evidence — would do more harm to the field's credibility, to survivors' long-term ability to be believed, and to innocent third parties than the conservative posture costs.

Sources we will not use

Leaked private documents we cannot independently authenticate.
Sealed court material that is not yet public.
Anonymous testimony as the sole support for a specific factual claim about a named person or incident.
Material from interested parties without disclosure of that interest.
Reporting that has been retracted by its original publisher.
AI-generated text presented as a primary source.
Doxxing material, even where it contains accurate facts, because we are not in a position to verify provenance.
Information obtained by deception of the source organisation (undercover infiltration journalism is treated case-by-case, weighted by the journalistic outlet's editorial process).

How sources flow into confidence ratings

The source mix on a profile determines its overall confidence rating. Briefly: High confidence requires multiple primary sources plus secondary corroboration. Medium confidence is credible secondary work with limited primary support. Low confidence is fragmentary anecdotal reports, single-source claims, or active controversy. Low-confidence ratings are explicitly provisional and marked as such on the profile.