On Background¶

The middle attribution tier where a claim may be published but the source is named only by a generic class descriptor — separating evidentiary content from source identity, with descriptor granularity a tunable knob trading warrant against identifiability.

Core Idea¶

"On background" is the middle tier of the journalism attribution ladder, between on the record and off the record: information may be published, but the source may be identified only by a generic class descriptor — "a senior administration official," "a person familiar with the matter," "a Western diplomat" — and never by name or specific role. The structural commitment is a negotiated attribution boundary that separates the evidentiary content from the source identity. The published claim is freed; the source's identity remains shielded. The class descriptor performs a specific credibility-warrant function for the reader: it conveys that the speaker occupies a role or proximity appropriate to the claim (senior enough to know, involved enough to be reliable) without exposing the individual to retaliation or identification. The key design parameter is descriptor granularity — how precisely the class is named — which sets a noise-tolerance trade-off: a descriptor specific enough to carry genuine warrant ("a senior NSC official familiar with the arms-control negotiations") risks identifying the source if only a few people fit the description; a descriptor too generic ("a person with knowledge of the situation") empties the warrant of meaning. Three distinct breach modes follow from this: over-specific descriptor (source effectively identified), over-generic descriptor (warrant hollow and reader-misleading), and wrong-class descriptor (warrant misleading about positional authority). When publications over-rely on background attribution, the class descriptors proliferate until they lose discriminatory power — "senior administration official" fails to distinguish cabinet members from mid-level staff, and the warrant structure collapses. The on-background convention is closely paired with off the record: off the record prohibits publication entirely; on background permits publication under identity restriction. Diplomatic background briefings, corporate pre-announcement calls, qualitative-research pseudonymisation, and intelligence analytic attribution under redacted sourcing all instantiate the same content-versus-identity separation in adjacent professional contexts.

Structural Signature¶

Sig role-phrases:

the publishable content — the evidentiary claim freed for public use, the item the tier permits to appear
the shielded source identity — the individual's name and specific role withheld, the restricted item separated from the content
the negotiated attribution boundary — the explicit middle-tier license (between on the record and off the record) permitting publication under an identity restriction, so the question is how to attribute, not whether to run
the class descriptor — the generic role label ("a senior administration official," "a person familiar with the matter") attached in place of the name
the credibility-warrant transfer — the descriptor's evidentiary function, telling the reader the speaker holds a role or proximity adequate to the claim
the granularity knob — how precisely the class is named, the single tunable parameter
the warrant-versus-identifiability trade-off — precision buying warrant but spending source safety, generality protecting the source but emptying the warrant, with a usable band bounded at both ends
the three breach modes — over-specific (source named by elimination), over-generic (warrant hollow, reader misled), wrong-class (warrant misrepresents positional authority)
the aggregate warrant collapse — descriptors proliferating under overuse until "senior administration official" no longer discriminates, the warrant structure failing newsroom-wide before any single attribution looks wrong

What It Is Not¶

Not off the record. Off the record bars publication of the content entirely; on background permits publication under an identity restriction. The reporter's question is therefore not "may I run this?" but "how may I attribute it?" — they are adjacent tiers of one ladder, differing in degree, and confusing them treats a permission as a prohibition.
Not full anonymity. The content is published, and it carries an attribution — a generic class descriptor ("a senior administration official," "a person familiar with the matter") rather than a name. The source identity is shielded, not the claim, so this is a restricted attribution, not the absence of attribution that a bare anonymous tip would be.
Not decoration. The class descriptor does real evidentiary work — a credibility-warrant transfer telling the reader the speaker holds a role or proximity adequate to the claim (senior enough to know, close enough to be reliable). "A senior NSC official familiar with the negotiations" is a warrant to be assessed (is this the right class of speaker for this claim?), not a throwaway gesture toward a hidden source.
Not "more generic is safer." Descriptor granularity is a tunable knob with a usable band bounded at both ends: too specific and it names the source by elimination when only a few people fit; too generic and it empties the warrant, leaving the reader misled about how much to credit the claim. Defaulting to maximal vagueness is its own breach (over-generic), not a safe choice — and aggregate overuse drifts the whole descriptor vocabulary toward non-discrimination ("senior administration official" ceasing to separate a cabinet secretary from a mid-level aide).
Not data anonymization. It is a negotiated attribution license between specific parties at a specific moment, not a data-processing operation that strips identity from a dataset. The granularity knob is a weak, qualitative shadow of the re-identification problem that data privacy formalizes (k-anonymity, differential privacy) — recognizable structure, but far thinner than the quantified machinery.
Not responsibility attribution. Responsibility attribution assigns credit or blame; on background restricts source-identity attribution downstream in a publication. It is the dual concept — a constraint on naming who said it, not an allocation of who is answerable for it.

Scope of Application¶

"On background" lives across the reporting contexts of journalism where a source is published under a class descriptor rather than by name — wherever the attribution ladder governs how a shielded source's claim is attributed; its reach is within that domain. The structurally identical content-versus-identity separations elsewhere — diplomatic background briefings under Chatham House Rule-style conventions, corporate pre-announcement calls and technology embargoes, qualitative-research pseudonymization, intelligence analytic products under redacted sourcing — are fellow instances of the general content-versus-identity-separation pattern (carried by confidentiality, access_control, contract, and responsibility_attribution, with the granularity knob a weak shadow of the privacy re-identification machinery), not uses of this journalism tier, so they are analogy and stay out of the map.

Political and government reporting — the canonical home: "a senior administration official," "a person briefed on the matter," where the class descriptor warrants the source's proximity to the claim without naming them.
National-security and diplomatic reporting — "a Western diplomat," "an intelligence official familiar with the assessment," shielding sources whose naming would carry severe consequences.
Business and financial reporting — "a person familiar with the negotiations," "a Wall Street trader," conveying market-relevant positional authority under identity restriction.
Whistleblower and source-protected reporting — anonymous-source protections functioning as on-background-at-publication, the descriptor carrying warrant while the identity stays shielded.
The attribution ladder as surrounding apparatus — the middle tier between on the record and off the record (its sibling off_the_record), negotiated at the start of a source interaction across all beats.

Clarity¶

Naming the on-background tier makes one separation legible above all: the split between a claim's evidentiary content (publishable) and its source identity (shielded), negotiated as an explicit attribution boundary rather than left to ad hoc anonymity. That clears up where on background sits relative to its neighbors on the ladder — it is not off the record, which bars publication outright, but a permission to publish under an identity restriction — so the reporter's question is no longer "may I run this?" but the sharper "how may I attribute it?" The frame also reveals what the generic class descriptor is actually for: it is not decoration but a credibility-warrant transfer, telling the reader that the speaker holds a role or proximity adequate to the claim (senior enough to know, close enough to be reliable) while withholding the name. Seen that way, the descriptor is doing evidentiary work for the reader, and "a senior administration official" can be evaluated as a warrant — is this the right class of speaker for this claim? — instead of read as a mere gesture toward a hidden source.

The decisive clarity is that the descriptor's granularity is a tunable parameter with a built-in trade-off, and naming it converts a vague worry about anonymous sourcing into a precise design question. Too specific a descriptor ("a senior NSC official familiar with the arms-control negotiations") can name the source by elimination when only a handful of people fit; too generic a one ("a person with knowledge of the situation") empties the warrant of meaning. Holding that trade-off in view splits what would otherwise look like a single "anonymous source problem" into three diagnosable breaches: over-specific (the source is effectively identified), over-generic (the warrant is hollow and the reader misled about how much to credit the claim), and wrong-class (the warrant misrepresents the speaker's positional authority). It further exposes a slow, aggregate failure invisible at the level of any one story: when a newsroom over-relies on background, the class descriptors proliferate until they lose discriminating power — "senior administration official" ceases to separate a cabinet secretary from a mid-level aide — and the warrant structure quietly collapses. The label lets an editor ask not just "is this source protected?" but "does this descriptor still carry warrant, and at what cost to the source's safety?"

Manages Complexity¶

Sourcing under anonymity looks, untreated, like one amorphous "anonymous source problem" — a reporter weighing case by case whether to run an unnamed tip, how to phrase the attribution, whether the shielding is adequate, with no fixed handle on what is being traded against what. The on-background tier compresses that diffuse worry onto a single design parameter and a small branch structure. The first compression is locating the tier on the attribution ladder: on background is publication under an identity restriction, distinct from off the record (no publication), so the reporter's question collapses from "may I run this?" to the sharper "how may I attribute it?" — a difference of degree along a fixed scale rather than an open judgment. The second, decisive compression is recognizing that the generic class descriptor does evidentiary work — a credibility-warrant transfer telling the reader the speaker holds a role or proximity adequate to the claim — and that its granularity is a single tunable knob with a built-in trade-off: precision buys warrant but spends source safety, generality protects the source but empties the warrant. So the whole negotiation reduces to setting one parameter (how precisely to name the class) against one trade-off (warrant value versus identifiability), and the qualitative outcome reads off where that knob sits. That single parameter also fixes a clean three-way classification of what would otherwise be an undifferentiated "bad sourcing": over-specific (descriptor names the source by elimination), over-generic (warrant hollow, reader misled about how much to credit the claim), and wrong-class (warrant misrepresents the speaker's positional authority) — three diagnosable breaches, each tied to a specific setting or mis-setting of the granularity knob, rather than a vague sense that the attribution is off. The frame further exposes a slow aggregate failure invisible at the single-story level and makes it readable as the same parameter drifting in bulk: when a newsroom over-relies on background, descriptors proliferate and lose discriminating power — "senior administration official" ceases to separate a cabinet secretary from a mid-level aide — so the warrant structure collapses across the publication even when each individual descriptor looked acceptable. Instead of an open-ended anonymity problem, the practitioner tracks one ladder position, one granularity parameter, and one warrant-versus-safety trade-off, and reads off how to attribute, which of three breaches threatens, and whether the descriptor still carries warrant at acceptable cost — at the level of a single story and of the newsroom's aggregate practice alike.

Abstract Reasoning¶

With the negotiation reduced to one ladder position, one granularity knob, and one warrant-versus-safety trade-off, the tier licenses inferences keyed to the descriptor as a tunable credibility instrument.

Diagnostic — evaluate the class descriptor as a warrant, and classify a sourcing failure by how the granularity knob is mis-set. The signature inference treats the generic descriptor not as a gesture toward a hidden source but as a credibility-warrant transfer doing evidentiary work, and asks of it the warrant question: is this the right class of speaker for this claim — senior enough to know, close enough to be reliable? "A senior NSC official familiar with the arms-control negotiations" is read as a warrant to be assessed, not a hidden name to be guessed. From the single granularity parameter the analyst classifies any sourcing failure into one of three diagnosable breaches: over-specific (the descriptor names the source by elimination because only a handful fit), over-generic (the warrant is hollow and the reader misled about how much to credit the claim), and wrong-class (the warrant misrepresents the speaker's positional authority). Each maps to a specific setting or mis-setting of the knob, so "the attribution is off" resolves into which way it is off. A second-scale diagnostic reads an aggregate signature invisible in any one story: when descriptors proliferate until "senior administration official" no longer separates a cabinet secretary from a mid-level aide, the analyst infers a newsroom-wide warrant collapse from over-reliance on background, even where each individual descriptor looked acceptable.

Interventionist — set the descriptor granularity to balance warrant against exposure, and predict the trade each setting makes. The lever is the granularity knob, and the intervention is to set it against a built-in two-sided trade-off with a forecast for each direction: making the descriptor more precise buys warrant but spends source safety (predicted to raise identifiability when few people fit), while making it more generic protects the source but empties the warrant (predicted to leave the reader unable to gauge how much to credit the claim). The prescribed move is to find the granularity that carries genuine warrant at acceptable identification risk for the specific claim and source pool — neither maximizing precision nor defaulting to the safest vagueness. At the aggregate scale the intervention is to ration background attribution itself, on the prediction that overuse drifts the whole stock of descriptors toward non-discrimination regardless of any single story's care.

Boundary-drawing — content is separated from identity, the tier sits between its neighbors, and the descriptor must clear a warrant floor. The tier draws its defining boundary by splitting a claim's evidentiary content (publishable) from its source identity (shielded) as an explicit negotiated boundary, so the reporter's question is bounded away from "may I run this?" to "how may I attribute it?" Its position on the ladder bounds it against off the record (which bars publication outright) and on the record (named attribution): on background is publication under an identity restriction, a difference of degree along a fixed scale. The granularity parameter draws an inner boundary at both ends — below a precision floor the descriptor carries no warrant and ceases to do evidentiary work; above a precision ceiling it identifies the source and breaches the shield — so the usable band is bounded on both sides, and a descriptor outside it has failed the tier's purpose even if publication proceeded.

Predictive — the descriptor's granularity predicts both source exposure and reader credence, at single-story and newsroom scale. From where the granularity knob is set the analyst predicts two outcomes at once: how identifiable the source is (precise descriptors against a small eligible pool predict exposure) and how much warrant the reader receives (generic descriptors predict under-credence or misplaced credence). The construct also predicts the aggregate trajectory: sustained over-reliance on background is forecast to erode the discriminating power of the descriptor vocabulary across a publication, so the warrant structure is predicted to collapse in bulk before any single attribution looks wrong — a failure anticipated from the rate of background use rather than diagnosed story by story.

Knowledge Transfer¶

Within journalism the tier transfers as mechanism, intact, across the sourcing situations that use the attribution ladder. The content-versus-identity separation (the evidentiary claim freed, the source identity shielded behind a generic class descriptor), the reframing of the descriptor as a credibility-warrant transfer (assess "a senior NSC official familiar with the negotiations" as a warrant — is this the right class of speaker for this claim? — rather than as a hidden name to guess), the single tunable granularity parameter with its built-in warrant-versus-identifiability trade-off, the three breach modes (over-specific, over-generic, wrong-class), and the aggregate-scale failure (descriptor proliferation collapsing discriminating power across a newsroom) all carry without translation. A reporter who has internalized the background NSC briefing applies the same instrument to every class-descriptor attribution, reading off how to attribute, which breach threatens, and whether the descriptor still carries warrant at acceptable source-exposure cost — at the level of a single story and of the publication's aggregate practice alike. (It is the middle tier of one ladder with its sibling off_the_record: off the record bars publication entirely, on background permits publication under an identity restriction.)

Beyond journalism the honest account is a shared-abstract-mechanism one, and it tracks its sibling closely: on background factors into the same catalog primes as off_the_record — the broader normative class of confidentiality, access_control (the structural cousin governing access where this governs downstream use and attribution), contract and commitment (the license-as-agreement layer), and responsibility_attribution (the dual concept — here a restriction on source-identity attribution rather than the assignment of credit or blame). The same content-versus-identity separation recurs across adjacent professional contexts as genuine co-instances: diplomatic background briefings (Chatham House Rule-style conventions), corporate pre-announcement calls and technology embargoes, qualitative-research pseudonymization (quotes attributed to "a mid-career nurse manager, hospital A"), and intelligence analytic products with class-descriptor attribution under redacted sourcing. But what travels across these is the general content-versus-identity-separation pattern and its constituent primes, not the journalism tier; what stays home-bound is the on-background cargo — the journalism class-descriptor ladder, the attribution-rule conventions, the source-protection practices, the specific "senior administration official" vocabulary. So the cross-domain reasoner should carry the general pattern (and confidentiality, access_control, contract, responsibility_attribution), and treat a diplomatic background briefing or a research pseudonym as a fellow instance rather than as "on background"; using the journalism label across that boundary is analogy to be marked.

The standout transfer note is the descriptor-granularity parameter, which is itself a clean shared-abstract-mechanism case and shows exactly where the deeper portable lesson lives. The trade-off it encodes — a descriptor specific enough to carry genuine warrant risks identifying the source when only a few people fit; one generic enough to protect the source empties the warrant — is a recognizable shadow of the general re-identification problem formalized in data privacy as k-anonymity and differential privacy (and more broadly as pseudonymization): granularity of a quasi-identifier traded against re-identifiability of the individual. That structure genuinely recurs wherever an evidentiary contribution must be preserved while identity is blurred at a controlled granularity — but it recurs there in full, quantified form, while on background instantiates it in a much weaker, qualitative form bound to negotiated class descriptors. So the portable lesson at the granularity knob belongs to the general re-identification / privacy machinery, not to "on background," whose version is the journalism-specific, much-thinner realization. The tier stays distinct from its neighbors accordingly — it is not off_the_record (which prohibits publication, where this permits it under attribution restriction), not data-processing anonymization in the privacy sense (it is a negotiated attribution license between specific parties at a specific moment, not a data operation), and not responsibility_attribution in the credit/blame sense (it restricts source-identity attribution downstream). Strip the journalism idiom and what remains is the content-versus-identity-separation pattern carried by confidentiality, access_control, contract, and responsibility_attribution, with the granularity knob a weak shadow of the general re-identification structure — the boundary between this domain-specific abstraction and the substrate-independent mechanisms it composes (see Structural Core vs. Domain Accent).

Examples¶

Canonical¶

The paradigm instance is the Washington background briefing. Before a policy rollout or a foreign trip, the White House or State Department gathers reporters and a named official speaks under the ground rule that they may be quoted only as "a senior administration official." The reporters may publish everything substantive the official says — the policy, the reasoning, the numbers — but must attach that generic label instead of the name. Readers learn that the account comes from someone senior enough to speak authoritatively for the administration, which warrants the claim, while the individual is shielded from being personally pinned to it. The descriptor is calibrated: "senior administration official" is specific enough to signal authority but generic enough that dozens of people could fit, keeping the source unidentifiable. It is publication under an identity restriction — the defining shape of the on-background tier.

Mapped back: The policy account is the publishable content and the official's name the shielded source identity, split by the negotiated attribution boundary. "Senior administration official" is the class descriptor performing the credibility-warrant transfer, and its calibration to signal authority yet fit many people is the granularity knob set within the warrant-versus-identifiability trade-off.

Applied / In Practice¶

Newsrooms operationalize the granularity trade-off in their sourcing standards. The Associated Press's news-values guidance, for instance, directs journalists to describe an anonymous source as specifically as possible — enough for readers to judge the source's reliability and vantage ("a person who was in the room during the negotiations") — while stopping short of detail that would identify the individual. Editors thus tune the descriptor per story against the same two-sided constraint the tier defines: push toward precision to give readers real warrant, but not so far that a small eligible pool exposes the source. Guidance also cautions against overuse of anonymous sourcing, precisely because a flood of "people familiar with the matter" erodes the descriptors' meaning. The abstract knob becomes an enforced editorial discipline applied claim by claim.

Mapped back: "Describe as specifically as possible without identifying" is exactly the granularity knob set within the warrant-versus-identifiability trade-off, with the descriptor doing the credibility-warrant transfer for the reader. The caution against overuse targets the aggregate warrant collapse — descriptors proliferating until they no longer discriminate — enforced as newsroom policy.

Structural Tensions¶

T1: Warrant versus identifiability (a usable band that can be empty). The granularity knob is billed as tunable — set it precise enough to carry warrant, generic enough to shield the source. But the two constraints are not always jointly satisfiable: when the eligible pool is small and the claim needs specific positional authority to be credible, the descriptor precise enough to warrant the claim ("a senior NSC official familiar with the arms-control negotiations") is exactly the one that names the source by elimination, and any descriptor generic enough to protect them empties the warrant. The usable band the tier presumes can collapse to nothing, leaving no setting that both credits the claim and shields the source. The tension is that the knob is presented as a dial with a good position, when for the highest-stakes, smallest-pool sources — the ones most needing protection — there may be no admissible setting at all. Diagnostic: For this claim and this eligible pool, does any descriptor exist that carries warrant without narrowing to the source, or is the band empty?

T2: Single-story care versus aggregate collapse (a commons no editor defects on). Each background attribution can be individually defensible — the descriptor fits many people, warrants the claim, shields the source — while the newsroom's whole stock of descriptors drifts toward non-discrimination as background sourcing proliferates, until "senior administration official" no longer separates a cabinet secretary from a mid-level aide. The failure is aggregate and invisible at the level of any one story: no single editor made a wrong call, yet the shared vocabulary's discriminating power erodes in bulk. The tension is a genuine commons dynamic — locally rational descriptor choices deplete a collective resource (the warrant value of the class-descriptor vocabulary) that no individual story is responsible for protecting. Policing story by story cannot catch it; only rationing the rate of background use can. Diagnostic: Is the concern whether this descriptor is acceptable, or whether the newsroom's cumulative background use is draining the descriptor vocabulary's discriminating power?

T3: Credibility conveyed versus warrant the reader cannot verify (the wrong-class breach is invisible to its victim). The descriptor's whole function is a credibility-warrant transfer — it tells the reader the speaker holds a role adequate to the claim. But the reader cannot check the characterization: only the journalist knows whether "a senior administration official" is genuinely senior or a mid-level aide dressed up, and the wrong-class breach — a descriptor that misrepresents positional authority — is undetectable by the very reader it misleads. The tension is that the instrument built to convey credibility is unverifiable by its audience, so its warrant rests entirely on trust in the journalist's honest class assignment. The reader is asked to credit a claim on evidence they are structurally barred from auditing, which means the tier's credibility function works only as far as the reader's trust in the newsroom extends. Diagnostic: Does crediting this claim require trusting the journalist's unverifiable assignment of the source's class, and is that trust warranted here?

T4: Source protection versus public accountability (the shield that also lets the powerful message unnamed). The tier's protective purpose is real — it shields sources from retaliation and buys reporters access they could not otherwise get. But the same shield lets a powerful source float a policy, test a message, or attack a rival without being personally answerable for it, converting anonymity-for-safety into anonymity-for-spin. In the canonical Washington background briefing, the official who speaks as "a senior administration official" gains the authority of the podium while shedding the accountability of a name. The tension is that the identity shield cannot distinguish the whistleblower who needs protection from the official who exploits it, so the instrument that safeguards vulnerable sources equally insulates powerful ones from the accountability that on-the-record attribution would impose. Diagnostic: Is the shield here protecting a source who would face genuine consequences if named, or sparing a powerful source the accountability of owning their words?

T5: Negotiated convention versus enforceable rule (a boundary held only by shared understanding). On background is a negotiated attribution license agreed between specific parties at a specific moment — its meaning depends on both source and reporter sharing the ground rules, which are professional conventions, not contracts with external enforcement. The adjacent tiers blur in practice: sources and journalists routinely disagree about whether a conversation was "on background" or "off the record," and there is no arbiter to settle it after the fact. The tension is that a boundary carrying real consequences — publish-under-restriction versus do-not-publish — rests on a convention that can be misremembered, differently understood, or unilaterally reinterpreted, with the source's protection and the reporter's credibility both hostage to an unwritten mutual understanding. The tier looks like a fixed rung on a ladder but is actually a fragile, per-interaction handshake. Diagnostic: Was the tier explicitly negotiated and mutually confirmed at the outset, or is it an assumed ground rule that source and reporter might recall differently?

T6: Autonomy versus reduction (a journalism tier or a composition of substrate-general primes). "On background" is a named, richly conventionalized journalism instrument — the class-descriptor ladder, the "senior administration official" idiom, the source-protection practices. Yet strip the journalism idiom and its portable structure is not proprietary: it composes confidentiality, access_control, contract/commitment, and the dual of responsibility_attribution (a restriction on source-identity attribution), sits beside its sibling off_the_record on one ladder, and its granularity knob is a weak, qualitative shadow of the re-identification structure that data privacy formalizes fully in k-anonymity and differential privacy. Diplomatic background briefings, corporate embargoes, and research pseudonymization are fellow instances of the general content-versus-identity separation, not uses of this tier. The tension is between a named professional convention worth studying in its own idiom and the recognition that its cross-domain cargo already belongs to those general primes. Diagnostic: Resolve toward the general content-versus-identity primes (confidentiality, access_control, contract, the re-identification shadow) when reasoning across professional domains; toward on background when governing source attribution inside journalism.

Structural–Framed Character¶

"On background" sits at the framed end of the structural–framed spectrum — framed-leaning: a journalism-particular attribution tier constituted by a source-reporter practice, the sibling of off the record, more a tunable technique than a normative verdict. On evaluative_weight it is largely neutral as an instrument (a granularity knob traded against a warrant), though its three breach modes (over-specific, over-generic, wrong-class) carry a corrective charge, so it sits above a bare mechanism while short of a censure-label. On human_practice_bound it is high: the tier is constituted by the practice of publishing a source's claim under a class descriptor on the attribution ladder and dissolves without it — with no publication, no reader to warrant, and no source to shield, there is only a person speaking, nothing being attributed at a tuned granularity. Institutional_origin is pronounced: the class-descriptor ladder, the "senior administration official" idiom, and the source-protection conventions are journalism-ethics furniture negotiated per interaction, an apparatus of a professional tradition rather than a fact of nature. On vocab_travels it scores low: the attribution-ladder idiom is pinned to reporting. And on import_vs_recognize the transfer is bimodal — within journalism it ports as mechanism across political, security, and business reporting, but diplomatic background briefings, corporate embargoes, and research pseudonymization are co-instances of a shared pattern, not imports of "on background."

The portable content is a content-versus-identity separation — the evidentiary claim freed while source identity is shielded at a controlled granularity — factoring into catalog primes the entry genuinely rests on: confidentiality, access_control (governing access where this governs downstream use and attribution), contract/commitment (the agreement layer), and the dual of responsibility_attribution (a restriction on source-identity attribution rather than an assignment of credit or blame). Distinctively, its granularity knob is a weak, qualitative shadow of the re-identification structure that data privacy formalizes fully in k-anonymity and differential privacy. Each of these travels on its own terms. But together they do not pull "on background" off the framed pole, because that content-versus-identity template is exactly what the tier instantiates from its umbrellas, not what makes "on background" itself travel: the cross-domain reach belongs to the confidentiality/access-control/contract/responsibility-attribution constellation and, at the granularity knob, to the full re-identification machinery, while the class-descriptor ladder and the "senior administration official" idiom stay home. Its character: a practice-constituted journalism attribution tier, structural only in the content-versus-identity-separation template it borrows from its umbrellas and in the re-identification structure its granularity knob faintly shadows.

Structural Core vs. Domain Accent¶

This section decides why "on background" is a domain-specific abstraction and not a prime — why its portable content is a composition of general primes while the newsroom tier stays home.

What is skeletal (could lift toward a cross-domain prime). Strip the journalism and a thin relational structure survives: a claim's evidentiary content is released while the source's identity is shielded behind a class descriptor whose granularity is tuned to carry credibility-warrant without permitting re-identification — a content-versus-identity separation with a tunable precision knob. The portable pieces are abstract: publishable content, a shielded identity, a class descriptor that transfers warrant, and a granularity trade-off (precision buys warrant but spends identifiability). This skeleton factors cleanly into catalog primes that each travel on their own terms — confidentiality (the normative class), access_control (governing access where this governs downstream use and attribution), contract/commitment (the agreement layer), and the dual of responsibility_attribution (a restriction on source-identity attribution rather than an assignment of credit or blame). Distinctively, the granularity knob is a weak, qualitative shadow of the re-identification structure that data privacy formalizes fully in k-anonymity and differential privacy. But these are the cores "on background" shares with a diplomatic briefing or a research pseudonym, not what makes the tier distinctive.

What is domain-bound. What is distinctive is journalism-ethics furniture and does not survive extraction. The attribution ladder position (the middle tier between on the record and off the record, sibling to off the record); the specific class-descriptor idiom ("a senior administration official," "a person familiar with the matter"); the three breach modes (over-specific, over-generic, wrong-class) named against newsroom practice; the aggregate warrant collapse (descriptor proliferation eroding discriminating power newsroom-wide); and the source-protection conventions negotiated per interaction — these are the ladder, the idiom, and the failure taxonomy, all specific to publishing a source's claim under a class descriptor. Even the granularity knob is a qualitative, negotiated realization, far thinner than the quantified k-anonymity/differential-privacy machinery it shadows. The decisive test: diplomatic background briefings, corporate embargoes, and qualitative-research pseudonymization are structurally identical content-versus-identity separations — but each carries its own apparatus, and calling any of them "on background" borrows the journalism label for a sibling. Remove the attribution ladder and the publication practice and there is no on-background tier, only the bare content-versus-identity template.

Why this does not clear the prime bar. A prime's vocabulary travels and its transfer is recognition of the same mechanism, not analogy. "On background"'s transfer is bimodal. Within journalism it moves as full mechanism — the content-versus-identity separation, the descriptor-as-warrant reframing, the granularity knob with its warrant-versus-identifiability trade-off, the three breach modes, and the aggregate-collapse dynamic carry intact across political, national-security, and business reporting, because each is a source-reporter exchange under the same ladder (recognition, not analogy). Beyond journalism the cross-domain cases are genuine co-instances of the general content-versus-identity-separation pattern, not imports of the tier: diplomatic briefings, corporate pre-announcement calls, research pseudonymization, and redacted intelligence attribution all instantiate the same separation, sharing the pattern but not the newsroom apparatus — and, at the granularity knob, the deeper portable lesson belongs to the full re-identification / privacy machinery, of which on background is the much-thinner qualitative shadow. The genuinely portable structure is that pattern and its constituent primes (confidentiality, access_control, contract, responsibility_attribution, plus the re-identification structure). So the cross-domain reach belongs to those parents; the disciplined move is to carry them when reasoning across professional domains, and reserve "on background" for governing source attribution inside journalism. It clears the domain-specific bar comfortably for journalism, but its only substrate-spanning content is already carried, in more general form, by the template it composes.

Relationships to Other Abstractions¶

Current abstraction On Background Domain-specific

Parents (3) — more general patterns this builds on

On Background is a kind of Anonymous Sourcing Domain-specific

On Background is the class-descriptor species of Anonymous Sourcing: the claim is published, the source is known inside the newsroom, and identity is hidden from readers.
On Background is a kind of Information Use License Prime

On Background is a strict specialization of Information Use License.
On Background is a decomposition of, conditional Minimum-Necessary Disclosure Prime

A newsroom can project at publication only the least identifying class descriptor that still gives readers the warrant needed to assess the claim.

Not to Be Confused With¶

Off the record. The adjacent, stricter tier — and the entry's sibling — which bars publication of the content entirely. On background permits publication under an identity restriction. The reporter's question shifts from "may I run this?" (off the record) to "how may I attribute it?" (on background); confusing them treats a permission as a prohibition (the drift breach). Tell: may the substance be published at all under a class descriptor (on background), or is publication forbidden until the source upgrades (off the record)?
On the record / deep background (the other rungs). On the record permits full named attribution; deep background is stricter still — the information may be used but attributed to no source at all, or only in the reporter's own voice. On background sits between, permitting a generic class descriptor. Tell: is the attribution a name (on the record), a class descriptor (on background), or no source attribution whatsoever (deep background)?
Data anonymization / k-anonymity / differential privacy. The formal, quantified re-identification machinery of data privacy — stripping or coarsening identifiers in a dataset so individuals cannot be re-identified. On background's granularity knob is a weak, qualitative shadow of this, but it is a negotiated per-interaction attribution license, not a data-processing operation. Tell: is identity being blurred by a formal transformation of a dataset (anonymization), or by a class descriptor agreed between a source and reporter at a moment (on background)?
Responsibility attribution. The assignment of credit or blame — who is answerable for an act. On background is the dual concept: a restriction on naming who said something downstream in a publication, not an allocation of who is responsible for it. Tell: is the question who is accountable for the deed (responsibility attribution) or how the source of a statement may be named (on background)?
Confidentiality (the broad duty). The general normative obligation not to disclose entrusted information. On background is a specific attribution-restriction license within that class — content published, identity shielded behind a tuned descriptor — not a blanket duty of discretion. Tell: is it a general obligation of non-disclosure (confidentiality) or a specific publish-under-class-descriptor tier with a granularity knob and breach modes (on background)?
The content-versus-identity-separation template (parent) and its co-instances. The substrate-general pattern the tier instantiates — evidentiary content freed while identity is shielded at a controlled granularity — carried by confidentiality, access_control, contract, and the dual of responsibility_attribution, with the granularity knob shadowing the re-identification machinery. Diplomatic background briefings (Chatham House Rule), corporate embargoes, and research pseudonymization are fellow co-instances, not uses of the journalism tier. Tell: is the newsroom class-descriptor ladder the apparatus (on background), or a content-versus-identity separation in another institution (a co-instance of the template)? (Treated more fully in Structural Core vs. Domain Accent.)

Neighborhood in Abstraction Space¶

On Background sits in a crowded region of the domain-specific corpus (31^st percentile for distinctiveness): several abstractions share nearly its structure, so a description that fits it tends to fit its neighbors too.

Family — Authority Control & Identity (11 abstractions)

Nearest neighbors

Anonymous Sourcing — 0.87
Off the Record — 0.87
Toulmin Model — 0.85
Corrections Policy — 0.84
Ad Hominem — 0.84

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