The Demarcation Problem

Censorship scholarship has long exceeded a model centred on overt prohibition. Jansen (1988) places censorship within relations of power and knowledge rather than treating it solely as a set of discrete acts of suppression. Müller (2004) maps censorship as cultural regulation, while Freshwater (2004) argues for its redefinition beyond conventional prohibitive forms. Bunn (2015) surveys the resulting “new censorship” tradition. The breadth of censorship beyond bans is therefore not the novelty claimed here.

One strand of this expansion is structural or constitutive. Bourdieu (1991) shows how the structure of a field can govern access to expression and impose the forms in which expression becomes acceptable. Butler (1998) likewise treats censorship as implicated in the production and delimitation of speakable vocabularies. On these accounts, censorship can be built into communicative conditions rather than appearing only as a later prohibition. A second strand identifies indirect mechanisms. Roberts (2018) distinguishes techniques such as friction and flooding, which burden access or redirect attention rather than simply deleting speech. By making source availability and the knowledge base itself part of the censorial problem, De Baets (2026) similarly shows that censorship may operate before a present utterance becomes available for evaluation.

This breadth creates a demarcation problem. As Schauer (1998) argues, the restriction of communicative possibility cannot by itself identify censorship, because ordinary human behaviour both constitutes and restricts communicative possibilities. Moore (2013) likewise warns against accounts so expansive that censorship becomes difficult to distinguish from the conditions under which communication is always selected, structured, and constrained. A theory capable of explaining structural and indirect censorship must therefore preserve a boundary between censorship and ordinary organisation.

This article proposes that censorship is a shortcut around contest: it occurs where participation is controlled in a way that substitutes for, prejudges, or forecloses contest over something that would otherwise remain open. The proposal is descriptive before it is evaluative. It identifies a censorial structure without determining whether that structure is justified, beneficial, harmful, proportionate, lawful, or professionally appropriate. Its scope is censorship, not a general theory of morality, social control, political legitimacy, propaganda, persuasion, or social shortcuts.

The proposal has both an explanatory and a demarcational function. It explains why censorship can appear as direct prohibition, indirect structural control, altered representation, anticipatory self-censorship, source-base control, or computational mediation. At the same time, it preserves distinctions between censorship and practices that may constrain participation without shortcutting contest. Criticism can defeat a claim within contest; selection and ranking allocate attention; editing changes presentation; moderation enforces forum rules; organisation structures process; and error can produce censor-like effects without creating a censorial relation. Constraint alone is insufficient.

The Lineage of the Shortcut Account

The idea that censorship or speech control can function as a shortcut has substantial prior history, although the work being bypassed differs across the literature. In his analysis of prior restraint, Jeffries (1983) characterises administrative preclearance as an “administrative shortcut to suppression”: unlike an injunction, it can make legality turn on prior official permission without requiring the government to identify a particular case for suppression, bear its justificatory burden in court, and remain subject to public scrutiny. This is a procedural-bypass account tied to prior restraint. In the context of hate-speech regulation and silencing, Heinrichs (1998) describes censorship as a shortcut to a more receptive audience, because suppressing competitors can substitute for persuading a broader public. Cohen-Almagor (2012) presents a related objection within the hate-speech dilemma: prosecution may be an ill-advised shortcut around the lengthy and costly work of education and raising awareness, although he weighs that objection against reasons for legal restriction. These accounts respectively concern procedure, persuasion, and education rather than a general demarcation of censorship.

The closest general theoretical predecessor is Steward (2014), who treats censorship as the frustration or foreclosure of inquiry, describes it as criticism gone wrong, and warns that apparent consensus may reflect a short-circuit of inquiry rather than its successful completion. The present account does not claim novelty for the proposition that censorship can bypass inquiry. Its departure is to generalise that structure beyond inquiry and use it to demarcate censorship across domains.

Adjacent theories identify further objects of shortcutting. Lafont (2020) develops a theory of democratic shortcuts directed at institutional proposals that bypass citizens’ own processes of opinion- and will-formation; contestation and demands for justification form part of the participatory alternative. While criticising Lafont, Shoemaker (2026) makes the connection to censorship explicit through a “propaganda shortcut” in which government silences dissenters and floods the media environment so that public preferences align with policy without the deliberation that would otherwise produce that alignment. This joins exclusion and saturation as means of bypassing deliberation, although its object remains democratic legitimacy. In content moderation, Douek (2025) argues that platforms cannot moderate their way around the harder work of public education and trust building, and warns that short-term suppression may undermine longer-term truth formation and trust.

These works identify speech control as a shortcut around procedure, persuasion, education, inquiry, deliberation, or trust formation. They establish substantial prior art for the shortcut intuition. What they do not individually or collectively supply is the generalisation proposed here: these are domain-specific instances of contest, and the demarcational question is whether control of participation substitutes for, prejudges, or forecloses the contest that would otherwise bear on the matter at issue. The contribution lies in that generalisation and its explanatory compression, not in the observation that censorship can be shortcut-like.

Two further lines of work help specify the account. First, shortcutting may occur upstream of substantive assessment. Atkinson (1994) shows that scientific gatekeeping can shape outcomes by controlling what receives review before assessment is complete, while Bachrach and Baratz (1962, 1963) show more generally how power operates through decisions and nondecisions that control what reaches decision at all. Second, the immediate on-ramp for the present argument was Lukianoff and Schlott (2023), whose account of cancellation foregrounds attempts to “win” disputes without winning the underlying argument, including by removing an opponent’s platform or career rather than refuting the position. In the present framework, cancellation in that form is a straightforward censorial move: control of practical access to the relevant arena substitutes for contest over the underlying matter. This connection explains the starting intuition, while the preceding lineage establishes that neither the shortcut idea nor its contrast with persuasion originated here.

Generalising inquiry to contest is necessary because not every domain in which censorship operates is well described as inquiry. Scientific credibility may turn on reasons and evidence, whereas political or cultural matters may concern legitimacy, recognition, interpretation, or visibility and proceed through representation, response, artistic expression, or self-presentation. Questions of admissibility and authority may themselves be contested when they determine whether a contribution receives standing. Participation must therefore mean more than permission to utter words. Democratic theory uses contestability to denote the practical possibility of challenge (Girard 2015; Trantidis 2017); content-moderation research applies the same idea to appeal and challenge within platform governance (Vaccaro et al. 2021); and relational accounts locate speech freedom in relations among participants rather than formal non-interference alone (Bonotti and Seglow 2022; Bishop forthcoming).

Work on silencing and inclusion identifies more specific participation failures. Speech may be present yet fail to perform its intended illocutionary act (Hornsby and Langton 1998), and testimonial practices may effectively silence speakers (Dotson 2011); Young (2000) provides a broader democratic account of inclusion. At the level of argumentative standing, Anttila and Domínguez-Armas (2025) analyse argumentative exclusion, while Aikin and Casey (2023) identify free-speech fallacies as meta-argumentative errors. Bejan (2020) places these problems within the wider tension between free expression and equal speech. The closest participation-side antecedent is the account of argumentative exclusion developed by Anttila and Domínguez-Armas (2025). The present proposal is broader because the relevant contribution need not be an argument or even realised speech, but narrower because impaired participation is not sufficient for censorship. Participation must be controlled so as to bypass the contest through which the matter would otherwise be worked out.

Algorithmic and infrastructural scholarship extends these concerns upstream. Cobbe (2021) emphasises the reach, interventionist capacity, private ordering, and structural power of algorithmic censorship. Elkin-Koren (2020) treats the ability to challenge filtering decisions as part of the institutional problem created by artificial-intelligence filtering. Xin, Zheng, and Ni (2026) describe “discursive preclusion,” in which infrastructure conditions what becomes visible or expressible before post-hoc moderation, while warning that proprietary mechanisms may remain causally underdetermined even when discursive patterns are observable. Implementation, timing, and opacity do not independently establish that contest has been bypassed.

The account therefore occupies a middle position between overt-act theories and accounts on which constitutive constraint readily becomes censorship. Structural conditions can matter without all being censorial. The decisive question is whether control of participation settles, prejudges, or forecloses something that would otherwise remain open to contest.

Contest, Participation, and Shortcut

Contest and Practical Participation

Contest is the process through which a matter at issue is established, defended, or changed. In some domains, this process occurs primarily through reasons, evidence, criticism, and response. In others, interpretation, comparison, representation, or challenges to standing do substantive work. Contest is intentionally broader than debate. Artistic expression, self-presentation, and struggles over what receives attention can participate in contest when they bear on the matter at issue.

The matter being contested must be specified in the terms of each case. Scientific credibility differs from journal fit; political legitimacy differs from compliance with a forum rule; historical interpretation differs from the prior availability of the sources on which interpretation depends. Identifying the matter at issue prevents the theory from treating every obstacle to participation as a bypass of the same contest.

Participation is practical rather than merely formal. A contribution must first be available and able to enter the relevant process. Once present, it may need to remain available, be encountered rather than effectively hidden, and be represented faithfully enough that others can respond to what it actually contributes. Mere presence is therefore not equivalent to effective participation. Source-base censorship makes this point especially clear: control may occur before present speech or debate through the records from which later contest proceeds (De Baets 2026; Tiffert 2019).

The account can also address unrealised expression and absent records where an established practice, process, or duty makes clear what would otherwise have been available. A merely imaginable absence is insufficient. More generally, what would otherwise remain open to contest must be identifiable independently of the disputed interference. Without that independent specification, the claim that contest has been short-circuited becomes circular.

From Structure to Shortcut

Every real practice structures participation. Rules define remit and format, set evidentiary and conduct standards, and impose ordinary burdens of compliance. Selection and ranking allocate scarce attention, while jurisdiction identifies the arena in which a matter will be considered. Restriction, burden, exclusion, unequal treatment, and outcome dependence therefore cannot by themselves distinguish censorship from ordinary organisation.

The longstanding selection–censorship debate in librarianship illustrates the problem. Asheim (1953) defends a professional distinction between selection and censorship, whereas Kidd (2009) critiques the ethic of selection and the effects of anticensorship practice. Their disagreement shows that the boundary has itself been contested and cannot be settled merely by applying an institutional label.

The central distinction is functional. A rule, decision, or practice may apply grounds relevant to the matter being contested, even when it excludes or defeats a contribution. It shortcuts contest when control of participation prevents those grounds from operating on the contribution, or permits them to operate only on a substituted or materially altered version of it. Casadevall and Fang (2009) frame this problem directly in relation to peer review. Dormandy and Grimley (2024) distinguish legitimate scientific gatekeeping from gatekeeping failure, while Hauswald (2026) analyses how failures of gatekeeping can produce degenerative consensus dynamics.

The same intervention may therefore be ordinary with respect to one question while shortcutting another. A journal-scope rule can decide fit for that journal while leaving scientific credibility open. Venue choice may alter strategic prospects without itself resolving the substantive issue, as Pralle (2003) shows in the context of venue shopping. Gieryn (1983) similarly demonstrates that allocations of authority and membership can themselves become objects of contest through boundary work. Control of relevant venues becomes censorially significant only when it takes over something that would otherwise remain for substantive assessment.

Control may act directly on a contribution, but it may also operate through the identity of its speaker, its source, its position on the matter at issue, or the channel through which it would participate. What matters is the relation between the control and the contested matter. A generic obstruction that happens to impede contest is not censorial merely because contest becomes harder or impossible. The relevant control must bear on the contribution, its source, speaker, standing, or channel in relation to the matter at issue.

Censorial intent is unnecessary. A shortcut may be explicit or emerge from distributed practices, and it may result from action or selective non-action. Nor must participation control secure its intended downstream result. A local relation may prejudge or foreclose a contribution’s participation yet fail to settle the wider matter, induce the expected behaviour, produce consensus, or persist over time. An attempted shortcut can therefore fail in its downstream purpose even when the participation-control relation itself is censorial. These distinctions concern how the relation is produced and what follows from it, not separate criteria of censorship. Several ordinary conditions may also combine to create a shortcut even when no condition independently does so. Accumulation is not sufficient: the combined operation must take over something that would otherwise remain to contest.

The evidential question follows from this relational structure. Evidence must support the alleged bypass, not merely an unwanted outcome. Labels such as scope, relevance, or quality may name substantive grounds of assessment, while safety, conduct, or compliance may name forum rules. None of these labels establishes whether the relevant contest was engaged or displaced. Aggregate disparity likewise does not reveal its mechanism. Process evidence and reasons actually given can bear directly on the relation; temporal changes and comparators may reveal differential operation; and appeal routes or alternative forums may show whether substantive contest remained practically available. Some cases will remain empirically underdetermined.

A shortcut may be episodic, but it may also alter later participation so that maintaining the result requires less fresh contest or intervention. This persistence is a possible consequence of an already-censorial relation, not an additional criterion for identifying censorship.

Forms and Effects of Participation Control

Familiar censorship vocabulary remains useful when it describes how participation changes. The terms considered here are overlapping descriptions rather than a six-part taxonomy or an exhaustive classification. Some concern changes to participation, others concern responses or persistence, and a single case may involve several.

Exclusion blocks or removes a contribution. It may operate on present speech, but it may also remove records that would have supported later contest. Historical and source-base censorship therefore extend exclusion upstream from present expression (De Baets 2026; Tiffert 2019).

Attenuation leaves a contribution present while making it harder to encounter or sustain. Gillespie (2022) analyses recommendation reduction as a moderation practice distinct from removal, while Howard and Kira (2026) examine demotion as a rights-relevant alternative to deletion. These practices show why practical participation cannot be reduced to binary presence or absence.

Transformation alters the version through which a contribution participates. Zheng (2025) documents politically directional revisions alongside ordinary translational corrections in Chinese translations. The coexistence of the two makes the distinction between unavoidable and censorial transformation theoretically important.

Saturation changes the surroundings of contest rather than the contribution itself. Roberts (2018) develops distraction and diversion as censorship strategies. King, Pan, and Roberts (2017) provide empirical evidence of large-scale fabricated posting oriented toward strategic distraction rather than engaged argument. These accounts identify burden and displacement as different ways in which an informational environment can interfere with contest.

Self-censorship and chilling are responses to surrounding conditions, not independent criteria for censorship. Penney (2022) analyses chilling effects as behavioural conformity produced by perceived legal or social consequences, and Kuran (1995) shows how anticipated costs can separate private from public expression. More recent studies find that expected social sanctions and social influence can suppress political expression (Weeks, Halversen, and Neubaum 2024; Yuen and Lee 2025). The response need not be silence. Dotson (2011) describes testimonial smothering, in which speakers alter what they say in anticipation of audiences unable or unwilling to receive fuller testimony. Wang and Mayer (2023) document deletion, account restriction, and reduced sensitive posting around Hong Kong’s National Security Law. Withholding and reshaping can thus be effects of a censorial environment without every prudent decision not to speak becoming a distinct act of censorship.

Finally, censorship may persist without continuously repeated censoring decisions. Foucault (1977) shows how disciplinary effects can become embodied in routines and expectations. Bowker and Star (1999) demonstrate the durable consequences of classification, while Star (1999) explains how infrastructure becomes embedded and taken for granted. A shortcut may become self-reinforcing when the standing or visibility it helped create becomes an input to later decisions about participation. Yuen and Lee (2025) provide a more immediate example of social influence sustaining self-censoring behaviour. Such dynamics can reduce dependence on fresh intervention, but they do not make every inherited or persistent condition censorial. The later condition must still perform shortcutting work.

Cases and Conceptual Contrasts

The following cases are deliberately asymmetric. Some are strong positive applications; others are boundary cases or cases of evidential underdetermination. Each removes a different intuitive assumption from the theory. They are not offered as a political programme or historical survey, and they need not instantiate the same features in symmetrical form.

Scientific Review and Withdrawal

Scientific evaluation supplies a clear initial contrast because exclusion can occur both through scientific assessment and around it. Casadevall and Fang (2009) pose the issue directly by asking when peer review is censorship. The Schön case provides the limiting contrast. The Bell Labs investigation addressed the validity of the measurements themselves (Beasley et al. 2002), and the subsequent retraction recorded the resulting change in the paper’s standing (Schön, Meng, and Bao 2003). Because measurement validity is a scientific ground bearing on the claims, the exclusion occurred through scientific evaluation rather than by bypassing it.

Gatekeeping scholarship sharpens this boundary. Peer review can regulate science by controlling access to evaluation (Atkinson 1994). Yet rejecting a paper after relevant scientific reasons have been applied against it may remain part of scientific contest. Preventing theories, objections, or evidence from receiving serious consideration for reasons that substitute for those grounds is structurally different (Dormandy and Grimley 2024). Such failures may also sustain degenerative consensus dynamics (Hauswald 2026).

The withdrawal of Bruce Gilley’s “The Case for Colonialism” isolates circumvention after review. The article passed double-blind peer review and was subsequently withdrawn (Gilley 2017). The publisher’s notice attributes the withdrawal to serious and credible threats made against the editor (Taylor & Francis 2017). Whatever the article’s scholarly merits, the threat did not engage them. It determined the article’s continued availability after scholarly evaluation. The case therefore illustrates circumvention without implying that peer review is non-selective or free of gatekeeping.

Translation: Presence Without Faithful Participation

Translation demonstrates why censorship need not require disappearance. A source may remain available while participating through an altered representation. Translation Studies has confronted the accompanying demarcation problem directly. A forum in Translation Studies considers whether broad, constitutive conceptions of censorship risk collapsing translation’s unavoidable selection and transformation into censorship itself (Baer et al. 2012). Karoubi (2025) similarly argues that constitutive theories can make exclusion and control indiscriminately censorial and proposes a Bourdieusian field account to disambiguate censorship in translation. The present proposal shares this demarcational aim but locates the boundary elsewhere: transformation as such is not censorship; the question is whether selective intervention changes what a source can contribute before readers can engage it.

Zheng (2025) reports ordinary corrections alongside directional political revisions in Chinese translations concerning EU–Russia relations. Criticism of Russia was generally softened, while negative evaluation of the European Union was strengthened. The asymmetry distinguishes the case from transformation merely as an unavoidable feature of translation. The translated source is presented as bearing on readers’ judgments of legitimacy and blame concerning Russia and the EU. It remains present, but politically selective revision changes what it can contribute before readers encounter and assess its presented position. This is a local claim about the represented contribution; it does not require showing that the translations determined the wider downstream standing of Russia or the EU.

Saturation: Burden and Displacement

Saturation shows how a shortcut may operate on the surroundings of contest without removing or modifying a contribution. One form is burden: filtering, checking, answering, or recovering the matter at issue becomes costly without the matter being resolved. The relevant effect is not volume alone but whether the added work prevents or materially frustrates the effort needed to recover and answer what is at issue.

Litigation provides a structural comparison. Sharpe (2009) discusses practices in which discovery costs may become strategic burdens. In Arbor Realty Funding v. Herrick, Feinstein, document production was criticised for burying responsive material within a much larger mass (“Arbor Realty Funding, LLC v. Herrick, Feinstein LLP” 2016). The responsive material can remain technically present while practical access to it is degraded. The comparison illustrates burden; it does not independently establish censorship in the litigation context.

A second form is displacement, which redirects attention and response to another matter. King, Pan, and Roberts (2017) estimate that the Chinese government fabricates roughly 448 million social-media comments per year and find that the operation largely avoids arguing with critics. Their evidence supports an inference of strategic distraction around collective-action discussions and events. Roberts (2018) places such distraction within a wider repertoire of censorship through friction and flooding. This is strong evidence of a strategy fitted to displacement, but it does not prove that every deployment successfully displaced practical attention.

Counter-speech supplies the limiting contrast. Even overwhelming rebuttal can remain within contest when the difficulty arises from substantive engagement with the matter at issue. Saturation becomes shortcut-like when surrounding conditions substitute for engagement rather than intensify it.

Before Speech or Evidence Reaches Contest

Anticipated criticism or embarrassment may rationally discourage expression while remaining part of ordinary contest. Kuran (1995) provides a general model of expression changing under anticipated social costs, and Penney (2022) develops chilling as behavioural conformity under perceived consequences. Neither phenomenon is automatically censorship on the present account. The question is what produced the anticipated cost and whether that surrounding condition itself shortcuts contest.

The response may consist of altered speech rather than silence. Testimonial smothering occurs when speakers truncate or reshape testimony because they expect audiences to be unable or unwilling to receive it (Dotson 2011). A censorial environment may therefore affect what reaches contest without producing literal non-speech.

Population-level evidence further illustrates the distinction between a surrounding condition and individual response. Wang and Mayer (2023) report increased deletion or account restriction and reduced politically sensitive posting around Hong Kong’s National Security Law. Fear of social sanctions can suppress political expression (Weeks, Halversen, and Neubaum 2024), and social influence can foster self-censorship under democratic backsliding (Yuen and Lee 2025). Together, these studies support population-level responses to surrounding conditions without establishing every individual’s motive or making every decision not to post a separate censoring act.

Control before contest need not operate on a speaker. Tiffert (2019) documents stealth redaction of Chinese digital historical databases and reverse-engineers a politically sanitising pattern of removals. De Baets (2026) places control of historical sources within the wider problem of censorship of history. In such a case, records that would bear on historical claims are selectively removed before scholars can contest their credibility or significance. This is a stronger source-base case because the evidence reaches the intervention itself rather than merely establishing a downstream absence.

Ranking and Recommendation

Computational mediation may increase the scale and speed of control, make its operation persistent or opaque, and individualise what users encounter. Algorithmic censorship can combine reach, interventionist capacity, and private ordering (Cobbe 2021); automated filtering also makes challenge and appeal institutionally important (Elkin-Koren 2020). None of these implementation properties is independently a criterion of censorship. The question remains whether software controls participation in a broader contest over what is seen or receives practical standing.

Explicit political filtering provides the strongest positive mechanised case. Knockel, Kato, and Dirks (2023) identify more than 60,000 censorship rules across eight search platforms accessible in China. These include hard censorship, which returns no results, and soft censorship, which restricts results to selected or authorised sources. Query-sensitive political and religious rules determine source eligibility before retrieval and ranking. Disallowed sources therefore cannot bear on public encounters with the issue.

Ambient feedback explains why algorithmic inequality alone is insufficient. Cho and Roy (2004) show that popularity-first ranking can reinforce existing popularity and delay the discovery of new pages. Introna and Nissenbaum (2000) identify systematic prominence and exclusion in search that may arise by design or without an explicit censorial decision. Existing prominence can feed back into future visibility without censorial intent, but unequal attention is not censorship merely because it is unequal.

Personalised curation supplies a second boundary. González-Bailón et al. (2023) show that Facebook’s computational curation materially changed the ideological composition of political news actually encountered relative to what was potentially available. This establishes the structuring of an informational arena. It does not by itself establish censorship, downstream belief change, or the bypass of a particular contest.

The concept of discursive preclusion sharpens the evidential limit. Strong political–non-political and cross-platform differences in visible discourse may reveal infrastructural structuring while leaving proprietary filtering mechanisms underdetermined (Xin, Zheng, and Ni 2026). Mechanisation can make participation control powerful without making every mediated outcome censorship.

Boundary Contrasts

Jurisdiction and venue illustrate the need to analyse forum control without assuming either that relocation is censorial or that it is harmless. Actors may move disputes across venues to improve their political prospects (Pralle 2003), while institutional boundary work allocates authority and membership (Gieryn 1983). Arena choice can thus change strategic prospects without shortcutting contest. The relevant question is whether control of the forum takes over something that would otherwise remain for substantive contest.

Moderation shows how one intervention can decide one question while affecting another. Chandrasekharan et al. (2017) find that Reddit’s 2015 bans reduced hate speech in the communities studied. Moderation may also operate through visibility changes short of removal, including recommendation reduction and demotion (Gillespie 2022; Howard and Kira 2026). None of these effects settles the censorial classification. A conduct rule may decide what behaviour a forum permits while also controlling whether a position participates in a separate contest over public credibility or legitimacy. Those questions must remain distinct.

Settled norms and constitutive structures supply a more difficult boundary. Identity concealment may respond to heteronormative professional or workplace climates (Beagan et al. 2022; Holman, Ogolsky, and Oswald 2022). Such cases demonstrate that social expectations shape disclosure even without a discrete censoring act. Fields and vocabularies may more deeply structure what forms of expression become possible or acceptable (Bourdieu 1991; Butler 1998), while routines, classifications, and infrastructure may embed constraints in ordinary practice (Foucault 1977; Bowker and Star 1999; Star 1999). These effects may be socially important, but pervasive structuring cannot itself be the classification rule. As Moore (2013) warns, otherwise censorship becomes indistinguishable from ordinary communicative conditions. Where what has been bypassed and how surrounding conditions performed that work cannot be identified, the effect remains outside the present classification.

Evidential and Conceptual Limits

The shortcut account is not a universal decision procedure or a quantitative score of censoriality. Counterfactual reasoning may clarify what would otherwise have remained open; comparators may reveal differential treatment; process traces may show how participation was controlled. These are explanatory probes rather than individually necessary or sufficient tests.

Hard cases are to be expected where the practices setting the terms of participation are entangled with the contest itself, or where the available evidence establishes an outcome without establishing how it was produced. The account should locate that uncertainty rather than force a binary classification. Labels, disparities, opacity, and downstream outcomes may all be relevant, but none substitutes for evidence of the relation between participation control and the matter at issue.

Interacting conditions may produce censorship at the system level even if no component is independently censorial. Yet accumulation alone remains insufficient: the combined conditions must perform the relevant shortcutting work. Conversely, the level at which a shortcut occurs need not match the level at which downstream discourse is measured. A local reduction in participation may coexist with compensation elsewhere, organised resistance, amplification of the targeted contribution, or backfire. A censorial participation-control relation may therefore fail to secure downstream settlement, behavioural conformity, or persistence. Simpson (2024) distinguishes chilling from heating effects, underscoring that downstream behavioural response is not a simple measure of the local intervention.

The account also separates identification from evaluation. Structural description does not determine lawfulness or ethics. A practice may be censorial in the descriptive sense developed here while still requiring distinct analysis of proportionality, necessity, professional appropriateness, benefit, harm, or legitimacy.

Conclusion

Censorship can intervene at different points along the path by which a contribution reaches and participates in contest. It may prohibit or remove, reduce visibility, alter representation, reshape surrounding attention, induce anticipatory withholding, control the source base, or mediate participation computationally. Control of participation provides a common basis for analysing these forms without making them identical.

The account’s central distinction is between what happens within contest and interventions that take over work the contest would otherwise have to do. Relevant grounds may exclude or defeat a contribution without bypassing contest. Conversely, participation may remain formally possible while its practical contribution is removed, attenuated, transformed, displaced, or prejudged. Exclusion, attenuation, transformation, saturation, chilling, persistence, and mechanised filtering are therefore descriptions of mechanisms or effects, not substitute definitions of censorship.

The strongest result is explanatory compression. Difficult cases become comparable without every constraint becoming censorial, and uncertainty remains visible where evidence does not support classification. The proposal should accordingly be judged by whether it clarifies more with less theoretical machinery, not by whether it closes every possible objection or functions as a universal decision procedure.