Sourcing Ladder¶

Rank the sources backing a claim by directness, independence, expertise, and accountability, then resolve disagreements by letting the higher rung win and cap a chain's confidence at its weakest load-bearing link.

Core Idea¶

A sourcing ladder is a ranked ordering of evidence sources for a given claim, organised by directness, independence, expertise, and accountability, used to assign differential weight to competing claims and to resolve conflicts between sources when they disagree. The canonical instance in evidence-based medicine is the GRADE framework and the Oxford Centre for Evidence-Based Medicine levels: systematic reviews and meta-analyses of randomised controlled trials sit at the top; individual RCTs below them; observational cohort and case-control studies below those; case series and expert opinion at the base. The journalism instance ranks named on-the-record primary sources above anonymous sources above secondary reporting; the legal instance ranks original documents and direct testimony above copies and hearsay. Intelligence analysis formalises this as the admiralty grading system, which rates source reliability and information credibility on separate axes independently.

The mechanism is a conflict-resolution protocol: when sources at different rungs disagree, the higher-rung source wins by default, because the higher rung controls for confounders (randomisation in RCTs), removes conflicts of interest (independence), and requires accountability for error (publication with peer review, on-the-record attribution, sworn testimony) in ways the lower rung does not. The protocol also governs claim-confidence propagation: a claim that is load-bearingly supported only by a lower-rung source carries that rung's confidence ceiling even if higher-rung sources elsewhere corroborate adjacent claims. This is the claim-chain principle — confidence is limited by its weakest load-bearing link, not by its strongest.

The ladder is claim-specific, not source-specific in the abstract: a manufacturer's data sheet is the highest-rung source for a chip's electrical specifications (it is the primary authority) but a low-rung source for a claim about the chip's safety in an unvalidated application. A systematic review is the highest-rung source for an aggregate clinical effect size but not for the particularised experience of a patient whose characteristics exclude them from trial populations. Applying the ladder requires identifying what kind of claim is being made, not only what kind of source is being cited, because the appropriate hierarchy is defined by the epistemic architecture of the claim type.

Structural Signature¶

Sig role-phrases:

the claim — the specific proposition for which evidence is being assessed; its epistemic architecture, not the byline, fixes the ordering
the candidate sources — the multiple competing evidence providers (meta-analysis, trial, eyewitness, data sheet, deposition, press release) to be ranked
the reliability dimensions — directness, independence, expertise, accountability: the distinct assessment axes a rung may be high on some and low on others
the ranked ordering (the ladder) — the partial order on sources by those dimensions, encoding the controls (randomization, peer review, on-the-record attribution) each rung secures
the claim-specific re-selection — the ordering is fixed by the claim type, so the analyst first classifies the claim and that selects which ladder applies; source prestige is not portable across claims
the higher-rung-wins conflict rule — when sources at different rungs disagree, the higher rung prevails by default, reducing every disagreement to a one-step lookup
the lowest-load-bearing-rung confidence rule — a conclusion inherits the rung of its weakest load-bearing link, so high-rung corroboration of adjacent claims cannot launder a chain resting on one low rung
the top-down search strategy — start at the highest plausible rung and descend only when forced, reaching a defensible answer with least descent
the domain-calibration / exception boundary — each domain calibrates its own rungs (so cross-domain comparison requires translation, not reading rungs across ladders), and the default rung order is temperable by domain exceptions (a meta-analysis of bad trials below a single strong RCT)

What It Is Not¶

Not a fixed ranking of sources in the abstract. The ladder is claim-specific: the same document is top-rung for one claim and bottom-rung for another. A manufacturer's data sheet is the primary authority on a chip's electrical specifications but near-worthless on its safety in an unvalidated application. The first question is not "is this a good source?" but "good source for what claim?" — the claim's epistemic architecture fixes the order, not the byline.
Not a single-dimensional "credibility" scale. The ordering runs on distinct axes — directness, independence, expertise, accountability — that can pull apart: an eyewitness is high on directness, low on expertise; a meta-analysis high on expertise, low on directness. A rung is not a scalar; the evaluator must ask which reliability a given rung actually buys.
Not an absolute conflict rule. Higher-rung-wins is a default, not a law. Domain exceptions temper it — a meta-analysis built from bad trials can be worse than a single high-quality RCT — so rung position resolves conflicts by default while remaining overridable on quality grounds.
Not merely the primary-versus-secondary distinction. That two-level cut is one slice of the ladder; the full instrument has more rungs and more dimensions, plus the two default rules (conflict resolution and confidence propagation) and claim-specific re-selection that the bare primary/secondary contrast does not carry.
Not confidence set by the strongest link. A conclusion inherits the rung of its weakest load-bearing link, not its best-supported one. High-rung corroboration of adjacent claims cannot launder a chain that load-bearingly depends on a single low rung; the confidence ceiling is read off the floor.
Not a single ladder whose rungs read across domains. Each field calibrates its own ladder to its own evidence ecology — an RCT-topped medical ladder, a documents-and-testimony legal ladder, a reliability×credibility intelligence grid. Cross-domain comparison requires translating between separately calibrated ladders, not reading a rung in one as equivalent to a rung in another.

Scope of Application¶

The sourcing ladder, as the evidence-evaluation construct, lives across the journalism-practice and fact-checking subfields of communication and media studies, wherever competing sources for a claim must be ranked and conflicts resolved; its reach is within that domain. The genuine co-instances of the same partial-order-plus-two-rules machinery in medicine, law, research synthesis, and intelligence travel under the parent (a composition of primary_vs_secondary_sources, triangulation, stratification, and weighting), each field calibrating its own rungs.

Journalistic evidence evaluation — the home: the named-on-the-record over anonymous over secondary-reporting ordering, and the editorial conflict-resolution that prefers higher-rung attribution by default.
Science journalism and fact-checking — peer-reviewed publication over preprint over press release over blog post over social media, used to adjudicate competing claims about studies and to set a story's confidence ceiling.
The primary-versus-secondary distinction in reporting — the two-level slice the full ladder extends, embedded in attribution conventions that prefer documents and direct testimony over copies and derived accounts.
Source-credibility and source-protection practice — the sibling source-handling constructs the ladder sits beside: the perception-side weighting and the channel-preservation shielding that together govern how a ranked source is used.
Claim-confidence reporting — the lowest-load-bearing-rung discipline applied to published conclusions, disclosing to downstream consumers that a claim resting on a low rung carries that rung's ceiling.

Clarity¶

Naming the sourcing ladder converts a vague intuition that "some sources are better than others" into a usable instrument with explicit rungs and explicit axes. Without it, a reviewer or fact-checker who distrusts a citation can only gesture; with it, the objection becomes precise and answerable — "this rests only on a press release," "this is hearsay," "this is observational where an RCT exists" — because the ladder supplies the vocabulary and the default resolution rule. It also separates dimensions that bare "credibility" fuses: a source can be high on directness yet low on expertise (an eyewitness), or high on expertise yet low on directness (a meta-analysis), so the ladder forces the evaluator to ask which reliability axis a given rung actually buys.

Its sharpest clarifying move is to make the hierarchy claim-specific rather than source-specific. The same document is top-rung for one claim and bottom-rung for another — a manufacturer's data sheet is the primary authority on a chip's electrical specifications but near-worthless on its safety in an unvalidated application; a systematic review is authoritative for an aggregate effect size but not for a patient excluded from the trial populations. Recognizing this blocks the common error of treating a source's prestige as portable across every claim it might touch, and reframes the practitioner's first question from "is this a good source?" to "good source for what claim?" — the epistemic architecture of the claim type, not the byline, fixes the ordering. The ladder additionally makes confidence propagate legibly: a conclusion inherits the rung of its weakest load-bearing link, so corroboration of adjacent claims by higher-rung evidence cannot launder a chain that depends on a single low rung.

Manages Complexity¶

A claim can be backed by a dozen sources of wildly different kinds — a meta-analysis, a single trial, a cohort study, an eyewitness, a manufacturer's data sheet, a press release, a sworn deposition, a copied document, a blog post — each strong on some reliability dimension and weak on others, and any two of them may flatly disagree. Evaluated source-by-source, dimension-by-dimension, this is a high-dimensional scoring problem: for every citation the analyst would weigh directness against independence against expertise against accountability, then somehow combine those vectors across competing sources to decide what the body of evidence actually says. The sourcing ladder compresses that by collapsing the multi-dimensional assessment into a single ranked ordering and attaching two default rules to it. The first rule, conflict resolution, reduces every disagreement between sources to a one-step lookup: locate each on the ladder, and the higher rung wins by default, because the rung position already encodes the controls — randomisation, independence, peer review, on-the-record accountability — that the lower rung lacks. The second rule, confidence propagation, fixes how a whole chain inherits a number: the conclusion carries the rung of its weakest load-bearing link, so the analyst need not recompute confidence from the full citation graph but only find the lowest rung the argument actually depends on. Together these turn "synthesise all this conflicting evidence" into two tracked quantities — the highest rung on which the claim has support, and the lowest rung it load-bearingly rests on.

The compression is sharpened, not weakened, by the ladder's claim-specificity. Because the appropriate ordering is fixed by the epistemic architecture of the claim type rather than by the source's prestige, the analyst's first move is to classify the claim — aggregate effect size, particular eyewitnessed event, primary electrical specification, individual patient excluded from trial populations — and that single classification selects which ladder applies, after which the same two default rules run. So the parameters the practitioner actually tracks are few: what kind of claim is being made (which selects the ladder), the highest rung with supporting evidence, the lowest load-bearing rung, and whether any domain-specific exception (a meta-analysis built from bad trials, say) overrides the default rung order. From those, the qualitative outcome reads off directly — which of two conflicting sources prevails, and what confidence ceiling the conclusion may claim — and the search strategy follows too: start at the highest plausible rung, descend only when forced. The branch structure is correspondingly simple. When sources at different rungs conflict, the higher rung carries it; when a chain's support bottoms out at a low rung, no amount of high-rung corroboration of adjacent claims can lift the conclusion above that floor; when the claim type changes, the ladder itself is re-selected and the same machinery re-applied. An open-ended "is this well-sourced?" problem becomes a small ordered set of categories and two lookups.

Abstract Reasoning¶

The sourcing ladder licenses a set of reasoning moves over bodies of evidence, all generated from its ranked ordering of sources, its two default rules (higher rung wins conflicts; confidence follows the lowest load-bearing rung), and its claim-specific calibration.

Diagnostic — locate a source on the ladder, identify the load-bearing rung, and read which reliability axis a rung buys. The signature inference places each cited source on the ladder for the claim at hand and reads its weight off the rung. The decisive diagnostic move is to find the lowest load-bearing rung of an argument: a conclusion inherits the rung of its weakest load-bearing link, so the analyst infers a chain's confidence ceiling by locating where its support bottoms out, not where it peaks — and infers that high-rung corroboration of adjacent claims cannot launder a chain depending on a single low rung. A second diagnostic decomposes "credibility" into the axes a rung actually buys: a source can be high on directness but low on expertise (an eyewitness) or high on expertise but low on directness (a meta-analysis), so the analyst infers which reliability the rung secures rather than treating the rung as a scalar. The construct also diagnoses the claim-specificity error: confronted with a prestigious source, the analyst asks "high-rung for what claim?" — a manufacturer's data sheet is top-rung for a chip's electrical specifications but bottom-rung for its safety in an unvalidated application — and infers that source prestige is not portable across the claims a source might touch.

Interventionist — resolve conflicts by rung, report confidence by the floor, and search top-down. The construct's interventions are rule-driven. When sources at different rungs disagree, the intervention is to default to the higher rung, with the predicted justification that the higher rung controls confounders (randomization), removes conflicts of interest (independence), and enforces accountability for error (peer review, on-the-record attribution, sworn testimony) that the lower rung lacks — so the resolution is a one-step lookup rather than an ad hoc weighing. When reporting a conclusion, the intervention is to assign it the confidence ceiling of its lowest load-bearing rung and to disclose that ceiling to downstream consumers. For evidence gathering, the construct prescribes a search strategy: start at the highest plausible rung and descend only when forced, with the predicted payoff of reaching a defensible answer with the least descent. It also licenses targeted citation criticism as an intervention — "this rests only on a press release," "this is hearsay," "this is observational where an RCT exists" — each a precise, answerable objection the ladder's vocabulary makes available.

Boundary-drawing — re-select the ladder by claim type, and temper the default with domain exceptions. The construct's central boundary move is that the ordering is claim-specific: the appropriate hierarchy is fixed by the epistemic architecture of the claim type, so the analyst's first act is to classify the claim (aggregate effect size, particular eyewitnessed event, primary specification, individual patient excluded from trial populations), and that classification selects which ladder applies before any rule runs. It bounds the higher-rung-wins default with domain-specific exceptions — a meta-analysis built from bad trials can be worse than a single high-quality RCT, so rung position is a default, not an absolute. And it marks where the ladder does not reach: it does not apply where there is no conflict to resolve and no confidence to propagate, and cross-domain comparison requires translating between each domain's separately calibrated ladder rather than reading rungs across ladders directly.

Predictive reasoning. From the load-bearing-rung rule the construct predicts a conclusion's defensible confidence ceiling in advance — a claim resting on a low rung cannot be argued above that floor regardless of surrounding high-rung material. It predicts which of two conflicting sources will prevail under the default and when a domain exception will override it. And via claim-specificity it predicts that the same source flips between authoritative and near-worthless as the claim type changes, so the analyst forecasts a source's standing not from its byline but from the match between its rung and the claim's epistemic architecture.

Knowledge Transfer¶

Within communication and media studies the sourcing ladder transfers as mechanism, intact, across the evidence-evaluation and journalism-practice cluster. From its evidence-evaluation home it ranks named-on-the-record over anonymous over secondary reporting, sits alongside its sibling source-handling constructs — source credibility (the perception-side weighting), source protection (channel-preservation shielding), off-the-record/on-background (attribution tiers), original reporting (the primary-source practice) — and its full apparatus moves without translation across reporting and fact-checking contexts: the multi-dimensional ordering (directness, independence, expertise, accountability), the two default rules (higher rung wins conflicts; confidence follows the lowest load-bearing rung), the claim-specific re-selection of the ladder, and the top-down search strategy. This is genuine within-domain mechanism transfer.

Beyond communication, the sourcing ladder is best understood as the evidence-evaluation codification of a more general mechanism that genuinely recurs across structurally distinct fields — the cross-domain reach is broad and substantive, and it carries a useful honesty wrinkle. The portable structural core is a claim-specific partial order on evidence sources by reliability dimensions, with a higher-rung-wins conflict rule and a lowest-load-bearing-rung confidence rule — itself a composition of catalog primes: primary_vs_secondary_sources (a two-level slice of the ladder), triangulation, stratification, evidence, and weighted aggregation, plus the not-yet-promoted claim-confidence-propagation rule. That core recurs as genuine co-instances, not metaphors: GRADE and the Oxford CEBM levels in evidence-based medicine; the best-evidence rule, the hearsay hierarchy, and Daubert/Frye in law; Cochrane methodology and primary/secondary/tertiary distinctions in research synthesis; the admiralty grading system (reliability and credibility on separate axes) in intelligence analysis; the eyewitness-over-interpretation-over-textbook ordering in history; and vendor-advisory-over-CVE-over-rumor in software security disclosure. Across all of these the same conflict-resolution and confidence-propagation machinery transfers as real discipline — which is why hierarchy-based conflict resolution, the directness/independence/expertise/accountability checklist, and search-down-the-ladder strategy port between them. But two things stay home. First, the rungs themselves do not transfer: each domain calibrates its own ladder to its own evidence ecology (an RCT-topped medical ladder, a documents-and-testimony legal ladder, a reliability×credibility intelligence grid), so cross-domain comparison requires translating between separately calibrated ladders, not reading rungs across them. Second, the named codifications — GRADE, CEBM, the best-evidence rule, admiralty grading — are domain-particular vocabulary on top of the shared composition. So the honest cross-domain lesson should carry the structure (the partial-order-plus-two-rules mechanism, composing primary_vs_secondary_sources, triangulation, stratification, and weighting) while recognizing that "sourcing ladder" names the evidence-evaluation instance and that each field's specific rung order is its own. Where that line falls is the subject of Structural Core vs. Domain Accent below.

Examples¶

Canonical¶

Evidence-based medicine's handling of hormone replacement therapy (HRT) and heart disease is the textbook demonstration of the higher-rung-wins rule. For years, large observational cohort studies (notably the Nurses' Health Study) found that women taking HRT had markedly lower rates of coronary heart disease, and clinical practice widely adopted HRT for cardioprotection. But observational cohorts sit below randomized controlled trials on the ladder because they cannot control for confounders — here, the "healthy-user" bias, since women who took HRT were also healthier and wealthier on average. When the Women's Health Initiative randomized trial reported in 2002, it found HRT actually increased cardiovascular and stroke risk. The claim about cardioprotection had been load-bearingly supported only by the lower rung; the higher rung overturned it, and practice reversed.

Mapped back: "HRT prevents heart disease" is the claim; the cohort studies and the WHI trial are the candidate sources differing on the reliability dimension of confounder control. The RCT's superior control places it higher on the ranked ordering, so the higher-rung-wins conflict rule resolves the disagreement against the observational finding. The reversal shows a chain resting on a low rung could not be argued above that floor.

Applied / In Practice¶

Military and NATO intelligence formalize a sourcing ladder in the Admiralty (or NATO) grading system, which scores every incoming report on two independent axes: source reliability graded A ("completely reliable") through F ("cannot be judged"), and information credibility graded 1 ("confirmed by other sources") through 6 ("cannot be judged"). A report might be rated "B2" — a usually-reliable source, with information probably true and corroborated. Analysts deliberately keep the two axes separate so that a trusted source relaying an unverified rumor (say "A5") is not conflated with an unproven source whose claim is independently confirmed ("F1"). Conflicting reports are weighed by these codes rather than by ad hoc impression, and a low grade caps how far a conclusion built on that report can be pushed.

Mapped back: The two-axis code operationalizes the reliability dimensions explicitly — separating independence/track-record from corroboration. Each graded report is one of the candidate sources; the letter-number grid is the ranked ordering, and analysts resolving contradictions by grade enact the higher-rung-wins conflict rule. A low credibility grade sets the lowest-load-bearing-rung confidence rule ceiling on any assessment resting on it.

Structural Tensions¶

T1: Higher-rung-wins default versus domain-quality exceptions (rank as heuristic, not law). The conflict rule defaults to the higher rung because rung position already encodes the controls the lower rung lacks — randomisation, independence, peer review, on-the-record accountability — reducing every disagreement to a one-step lookup. But it is a default, not an absolute: a meta-analysis built from bad trials can be worse than a single high-quality RCT, so a rung is a presumption overridable on quality grounds. The tension is that the instrument's speed comes from treating rank as decisive, while its correctness sometimes requires suspending that very treatment — and knowing when to override is exactly the judgment the one-step lookup was meant to spare. Diagnostic: Is the higher-rung source actually of higher quality here, or does a domain-specific quality exception invert the default?

T2: Claim-specificity versus source prestige (good source for what claim). The ladder's sharpest move is that the ordering is fixed by the epistemic architecture of the claim, not the byline: a manufacturer's data sheet is top-rung for a chip's electrical specifications and near-worthless for its safety in an unvalidated application. This blocks treating prestige as portable across every claim a source might touch. But it also means the analyst's very first act — classifying the claim to select which ladder applies — happens before any rule runs and is not itself supplied by the ladder; a mis-classified claim selects the wrong hierarchy and every downstream lookup inherits the error. The discipline's precision rests on a prior judgment it does not itself discipline. Diagnostic: Has the claim type been correctly classified to select the ladder, or is a prestigious source's standing being read as portable across the claims it touches?

T3: Confidence by the floor versus corroboration by the peak (the weakest load-bearing link). A conclusion inherits the rung of its weakest load-bearing link, so no amount of high-rung corroboration of adjacent claims can launder a chain that load-bearingly depends on one low rung — the ceiling is read off the floor. This is a powerful anti-laundering discipline, but it hinges entirely on correctly identifying which link is actually load-bearing. Mislabel a decorative or redundant citation as load-bearing and the ceiling is set too low; miss the single low rung the argument truly rests on and it is set too high. The rule is only as good as the dependency analysis that feeds it, and that analysis is not mechanical. Diagnostic: Which link does the conclusion genuinely depend on, and is that rung the one being used to cap confidence?

T4: Multi-axis reliability versus scalar rung (which reliability a rung actually buys). Reliability decomposes into distinct axes — directness, independence, expertise, accountability — that pull apart: an eyewitness is high on directness but low on expertise, a meta-analysis the reverse. Collapsing these into a single ranked ordering buys the one-step lookup that makes the ladder tractable, but the collapse hides which axis a rung secures, so two sources nominally at "the same rung" may be reliable in entirely different ways, and a claim needing directness is ill-served by a high rung that bought only expertise. The compression that makes the ladder usable is the same compression that can misroute a claim to the wrong kind of reliability. Diagnostic: Is the rung being treated as a scalar, or is it explicit which reliability axis it actually secures for the claim at hand?

T5: Domain-calibrated rungs versus cross-domain reading (translation required). Each field calibrates its own ladder to its own evidence ecology — an RCT-topped medical ladder, a documents-and-testimony legal ladder, a reliability×credibility intelligence grid. This makes each ladder sharp at home, but it blocks reading a rung in one as equivalent to a rung in another; cross-domain comparison requires translating between separately calibrated ladders, not reading rungs across them. The shared name "sourcing ladder" and the shared two-rule machinery tempt exactly the illegitimate direct comparison the calibration forbids. Diagnostic: Are two rungs being compared across separately-calibrated ladders directly, or first translated through each domain's own evidence ecology?

T6: Autonomy versus reduction (an evidence-evaluation construct or the partial-order-plus-two-rules parent). The sourcing ladder is a codified evidence-evaluation and journalism construct — GRADE, the Oxford CEBM levels, the best-evidence rule, admiralty grading are its named instances. But the portable core is the parent composition: a claim-specific partial order on sources by reliability dimensions, with a higher-rung-wins conflict rule and a lowest-load-bearing-rung confidence rule, composing primary_vs_secondary_sources, triangulation, stratification, and weighted aggregation. That machinery recurs across medicine, law, research synthesis, intelligence, history, and security disclosure, while the specific rung orders and named codifications stay home. Diagnostic: Resolve toward the partial-order-plus-two-rules composition when carrying the machinery across fields; toward the sourcing ladder when ranking competing sources for a claim in evidence evaluation.

Structural–Framed Character¶

The sourcing ladder sits at the mixed midpoint. Its evaluative_weight is present but structural rather than moral: the ladder does render weight — it ranks sources and declares which wins — yet that weighting is a partial order over evidence quality, an epistemic rule, not a normative verdict about persons or acts. It is human_practice_bound in that the instrument is exercised by evaluators adjudicating claims — the conflict-resolution and confidence-propagation rules are things analysts do — though what they track (directness, independence, expertise, accountability, and the controls each rung encodes) are partly features of the evidence itself. Its institutional_origin is high at the codification layer: GRADE, the Oxford CEBM levels, the best-evidence rule, and admiralty grading are named artifacts of specific disciplines, and each domain calibrates its own rung order. On vocab_travels it is bounded twice over — both the named codifications and, critically, the specific rungs stay home, so cross-domain comparison requires translation rather than rung-reading — yet on import_vs_recognize the underlying machinery is recognized as genuine co-instance across medicine, law, research synthesis, intelligence, and history, not carried by analogy.

The portable skeleton is a claim-specific partial order on evidence sources with a higher-rung-wins conflict rule and a lowest-load-bearing-rung confidence rule — a composition of primary_vs_secondary_sources (its two-level slice), triangulation, stratification, and weighted aggregation, plus a not-yet-promoted confidence-propagation rule. That partial-order-plus-two-rules machinery is what the sourcing ladder instantiates from its parents and what recurs field to field; the cross-domain reach belongs to it, while the calibrated rung orders and the GRADE/CEBM/admiralty names are the domain accent that does not travel. Its character: an evaluatively-loaded-but-epistemic, practice-exercised evidence instrument whose partial-order-plus-two-rules skeleton genuinely recurs as mechanism, structural in that shared machinery and domain-specific in the calibrated rungs and named codifications that pin it home.

Structural Core vs. Domain Accent¶

This section decides why the sourcing ladder is a domain-specific abstraction and not a prime, and it carries the case for its domain-specificity — a case sharpened by a second layer of domain-binding the other source-handling constructs lack: not only the codification but the rungs themselves stay home.

What is skeletal (could lift toward a cross-domain prime). Strip the evidence-evaluation apparatus and a thin relational structure survives: a claim-specific partial order ranks candidate evidence sources by reliability dimensions, and two default rules run over it — when sources at different ranks disagree, the higher rank wins; and a conclusion's confidence is capped at the rank of its weakest load-bearing link. The portable pieces are abstract — a claim whose epistemic architecture fixes the order, a set of competing sources, a partial order over them, a conflict-resolution rule, and a confidence-propagation rule. This core is genuinely substrate-portable — indeed it is itself a composition of catalog primes the entry names: primary_vs_secondary_sources (a two-level slice of the ladder), triangulation, stratification, and evidence under weighted aggregation, plus the not-yet-promoted claim-confidence-propagation rule. That is why the machinery recurs, as genuine co-instances rather than metaphors, across medicine, law, research synthesis, intelligence, history, and security disclosure. But it is the core the entry shares, not what makes the sourcing ladder distinctive.

What is domain-bound. Two layers of content stay home, and this is the distinctive point. First, the named codifications are evidence-evaluation furniture that does not survive extraction: GRADE, the Oxford CEBM levels, the best-evidence rule, the hearsay hierarchy, Daubert/Frye, Cochrane methodology, and the admiralty reliability×credibility grid. Second — and this is what separates the sourcing ladder from a construct like source credibility — the rungs themselves are domain-calibrated and do not transfer: an RCT-topped medical ladder, a documents-and-testimony legal ladder, a reliability×credibility intelligence grid are each tuned to their own evidence ecology, so even the co-instantiating fields cannot read a rung in one as equivalent to a rung in another; cross-domain comparison requires translating between separately calibrated ladders. The decisive test: the partial-order-plus-two-rules machinery recurs intact in a courtroom, but the medical ladder's top rung (a meta-analysis of RCTs) has no place on the legal ladder, whose top rung is an original document or direct testimony. Remove the calibrated rung order and the named codification and what is left is not "the sourcing ladder" but the bare ranking machinery each field re-instruments locally.

Why this does not clear the prime bar. A prime is a relational structure whose vocabulary travels and whose cross-domain transfer is recognition of the same mechanism, not analogy. The sourcing ladder's transfer is bimodal. Within communication and media studies the full apparatus travels intact as mechanism — the multi-dimensional ordering, the two default rules, the claim-specific re-selection, and the top-down search strategy move without translation across the evidence-evaluation and journalism-practice cluster. Beyond journalism, what recurs in medicine, law, research synthesis, intelligence, history, and security disclosure is the parent partial-order-plus-two-rules machinery, not "the sourcing ladder" as named — those fields carry the mechanism but supply their own rungs and their own codifications, so the named construct and its specific rung order do not travel, only the composition beneath it does. And when the bare cross-domain lesson is needed, it is already carried in more general form by the primes the entry composes: primary_vs_secondary_sources, triangulation, stratification, and evidence under weighted aggregation. The cross-domain reach belongs to that composition of parents; "the sourcing ladder," as named, is its evidence-evaluation instance, carrying calibrated rungs and named codifications that should stay home.

Relationships to Other Abstractions¶

Current abstraction Sourcing Ladder Domain-specific

Parents (3) — more general patterns this builds on

Sourcing Ladder is a kind of Hierarchy Prime

A Sourcing Ladder is a hierarchy specialized to claim-specific source reliability, organizing sources into ordered rungs that govern conflict and confidence.
Sourcing Ladder presupposes Evidence Prime

A Sourcing Ladder presupposes Evidence because its ranks have meaning only relative to how strongly sources support a particular claim.
Sourcing Ladder is part of Primary vs. Secondary Sources Prime

A Sourcing Ladder contains Primary vs. Secondary Sources as its direct-versus-derived two-level slice inside the wider multi-rung ordering.

Not to Be Confused With¶

Source credibility. The perception-side sibling the ladder sits beside: how believable an audience finds a source, a judgment about received trustworthiness. The sourcing ladder ranks sources by objective reliability dimensions (directness, independence, expertise, accountability) and resolves conflicts by rung; source credibility is the audience's subjective read of a source, which may track the rung or diverge from it (a low-rung source can be widely believed). Tell: is the question which of competing sources should prevail for a claim on reliability grounds (sourcing ladder), or how persuasive an audience perceives a source to be (credibility)?
Source protection. The channel-preservation sibling — shielding a source's identity so the channel of information is not burned. It is a different function entirely: the ladder weighs and ranks what a source supplies, protection conceals who the source is. Tell: is the concern adjudicating which source wins a disagreement (sourcing ladder), or keeping a source's identity confidential to preserve access (protection)?
Primary vs. secondary sources. A parent prime and a part of the ladder, not the whole: the two-level directness cut (original document / direct testimony above copies and derived accounts). The sourcing ladder extends that single slice into more rungs, several distinct reliability axes, two default rules (higher-rung-wins conflict resolution, lowest-load-bearing-rung confidence), and claim-specific re-selection. Tell: is only the direct-versus-derived binary in play (primary/secondary), or the full multi-rung, multi-axis apparatus with conflict and confidence rules (sourcing ladder)?
Triangulation. A composing prime, and a genuinely different move: corroboration by convergence across multiple independent sources, treating agreement as the warrant. The ladder's conflict rule instead lets the single highest-ranked source prevail by default when sources disagree. Tell: does the resolution privilege the top-ranked source over the others (sourcing ladder's higher-rung-wins), or seek mutual agreement among several independent sources regardless of their rung (triangulation)?
GRADE, the Oxford CEBM levels, the best-evidence rule, the admiralty grid. These are domain-particular instances of the ladder, each with its own calibrated rungs — not the construct itself. Treating any one as "the sourcing ladder" invites reading a rung calibrated for medicine (a meta-analysis of RCTs on top) as equivalent to a rung on the legal ladder (an original document on top), which the entry forbids: rungs do not read across separately calibrated ladders. Tell: are you naming one field's specific rung order (an instance) or the partial-order-plus-two-rules machinery common to all of them (the construct)?
The partial-order-plus-two-rules parent composition (umbrella). The substrate-portable machinery the ladder instantiates — a claim-specific partial order over sources with a higher-rung-wins conflict rule and a lowest-load-bearing-rung confidence rule, composing primary_vs_secondary_sources, triangulation, stratification, and weighted aggregation. This composition is what recurs across medicine, law, research synthesis, intelligence, and history; the sourcing ladder is its evidence-evaluation instance with journalism-calibrated rungs. Tell: is the claim about the machinery that travels field to field (parent, treated more fully in the transfer sections), or the named evidence-evaluation instrument with its specific rungs and codifications (sourcing ladder)?

Neighborhood in Abstraction Space¶

Sourcing Ladder sits in a sparse region of the domain-specific corpus (69^th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.

Family — Argument Structure & Evidentiary Reasoning (12 abstractions)

Nearest neighbors

Cherry Picking — 0.85
Ad Hominem — 0.83
Toulmin Model — 0.82
Review — 0.82
Benjamini–Hochberg Procedure — 0.82

Computed from structural-signature embeddings · 2026-07-12