The Quality of and About Reasoning: Democratic Justification Beyond the Discourse Quality Index

Section 1: Straightforward and Problematic Democracy

Democracy is both straightforward and problematic. It is straightforward in that it produces winners and losers according to rules that are, for the most part, publicly understood: votes are counted, seats allocated and governments formed through recognised procedures. Which electoral system is best may be disputed, but that a government formed under such rules holds power is not. It is problematic because the losers remain subject to a government they did not choose. What makes this acceptable is not the count alone. An election is the visible tip of a larger structure of justification. Campaigns, manifestos and public debate form reasons, open them to contestation and carry them into the vote. Those subject to a democratic order are, in Forst's terms, also its normative authorities, whose standing as justificatory equals entitles them to reasons for the power exercised over them and to the means of contesting those reasons (Forst, 2017, p. 49). Where that standing holds, losing can be acceptable. Not every proposal for policy or national direction will be adopted. A democracy succeeds when those whose proposals fail can regard the outcome as justified, even if begrudgingly. The relevant standard is not agreement with the result, but the possibility of accepting the terms under which it was reached. Such acceptance is a mark of a stable democracy. When losers reject the result itself, the fault may lie not in the count but further down. The legitimacy of the electoral result depends on the quality of what lies beneath it.

Electoral losses are visible. They are counted, declared and recorded, with the loser retaining their standing to contest the result. However, formally equal electoral procedures can operate within unequal social conditions. As Bonvin et al. (2018) and Sen (2001, pp. 74, 91) show, differences in resources, information, social position and effective opportunity can shape who is able to participate and influence collective decisions. Being outargued in an open contest leaves a person’s standing as a participant intact. Lacking the means to argue at all is a loss of standing, which I treat as taking two forms. Some citizens are excluded from the space of reasons altogether. Others are present within it but lack effective opportunity to influence how others think (Young, 2000, pp. 53–55). Such losses have several sources. A citizen may hold a formal right to participate that the conditions of their life never convert into a real one (Bonvin et al., 2018, p. 954). They may be dominated into silence. Or they may find that the space of reasons has already been determined, occupied or sealed off by others (Forst, 2017, p. 42). None of this is easily discernible in the result. More worryingly, much of it does not surface in debate, lobbying and associated democratic processes. Where the capacity to take part is foreclosed, there is nothing to record. In Diamond and Morlino's terms, it is the justification stage of vertical accountability that fails (2004, p. 25).

If justificatory losses do not surface in debate, instruments designed to assess debate, such as the Discourse Quality Index (DQI) (Steenbergen et al., 2003), will not find them. The DQI measures selected features of the reasoning expressed in speech. Its unit of analysis is the speech. Only those parts of a speech that make a demand are coded (Steenbergen et al., 2003, p. 27). The authors are clear about the limits of their test case. In parliaments, restrictions on participation are mild and procedure is rule-governed. Outside a formal chamber, access is a greater barrier, with some citizens excluded altogether. The DQI does not capture this (Steenbergen et al., 2003, pp. 43–44). Bonvin et al. (2018, pp. 956–957) go further, reading the DQI as an instance of Habermasian proceduralism that neglects substantive inequality. Following Sen, they hold that democracy does not reduce to a vote held every four or five years (p. 963) and they set out five dimensions along which what they call "real democracy" (p. 962, drawing on Sen) can be assessed (pp. 964–966). In effect, they propose an index of their own, though, unlike the DQI, they leave its codification to future participatory research (p. 966). Where the DQI codifies how well reasoning is conducted, and Bonvin et al. what democracy should contain, I address a thinner question: what must hold beneath an election and beneath the debates the DQI codes for their outcomes to be justified to those who lose? The DQI codifies the quality of reasoning, particularly within a controlled environment such as a parliament. It cannot codify the quality about reasoning: whether the space in which reasoning takes place admits those it should and whether that reasoning is addressed to the right democratic object rather than a substitute for it.

Forst’s discussion of political domination identifies two related problems: rule by unjustifiable norms and, reflexively, rule without adequate discursive arenas and institutional structures in which existing justifications can be contested and better ones constructed (Forst, 2017, pp. 50–51). The first problem concerns the quality of reasoning: whether the reasons offered meet the criteria of reciprocity and generality by which Forst assesses justification (Forst, 2017, p. 49). This partially overlaps with what the DQI codifies through indicators of respect, justification and appeals to the common good (Steenbergen et al., 2003, pp. 28–29). The second problem concerns the quality about reasoning. It takes two forms, corresponding to the ways in which power over the space of reasons is exercised (Forst, 2017, p. 42). Where the space is sealed off, the question is admission, asking who may enter it. Where it is determined or occupied, the question is substitution, asking whether what is debated is the justification of power itself or a substitute for it. The gap between the two qualities can be wide. A debate may be reciprocal, respectful and oriented to the common good among those admitted, while others remain outside it, or while its object has quietly been exchanged for another. Because the second quality settles who reasons and about what, it is normatively prior; the first can only ever be a quality of the reasoning the second allows. For Forst, the question of power is "the first question of justice" (2017, pp. 51–52). As with the first quality, the second is a matter of degree rather than kind, since power over the space of reasons is exercised "in different degrees" (Forst, 2017, p. 42). It can therefore be assessed, though not by coding speech alone. Forst describes the critique of relations of justification as a survey of the positions from which discursive power is generated and exercised across social and political spheres (2017, p. 51), which is the territory Bonvin et al.'s five dimensions begin to map. I begin from the limit the DQI identifies and set out what an assessment of the second quality must capture.

In what follows I make and support three related claims. First, democratic assessment must distinguish the quality of reasoning from the quality about reasoning: the difference between how well reasons are exchanged within a deliberative setting and whether that setting admits those entitled to contest power and addresses the relevant object of justification. Second, drawing on Forst's account of noumenal power, I explain why these qualities may diverge. Capacity and domination, which together with justification form mutually constitutive dimensions, shape not only who participates but also what may count as a politically intelligible reason or question. Third, I identify the limit of DQI-style coding and specify the conditions that an assessment of the second quality must capture: admission, standing, uptake and resistance to substitution.

Section 2: What the DQI Codifies

The DQI's strength is its explicitness. Derived from Habermas's discourse ethics, it translates normative principles into seven coding categories applied to observable discourse behaviour (Steenbergen et al., 2003, pp. 23, 27). Read through Forst, several of these categories concern the first quality. The level of justification records whether reasons are given for demands at all, which is the minimum the right to justification requires. Respect toward counterarguments approximates to reciprocity in that opposing reasons are acknowledged and engaged rather than ignored or degraded. The content of justification, which distinguishes appeals to group interest from appeals to the common good in utilitarian or difference-principle terms (Steenbergen et al., 2003, p. 26), approximates to generality. The approximations are not direct translations. An appeal to the common good, for example, is not the same as a reason that none of those affected could reasonably reject. The DQI records, with high intercoder agreement (Steenbergen et al., 2003, p. 37), that such a claim was made; it does not assess whether the claim would meet Forst's criterion of generality. Nonetheless, the DQI is a partial codification of the first quality, showing that this quality can be measured reliably.

Every instrument selects an informational basis and that selection is itself a normative choice (Bonvin et al., 2018, p. 960). The DQI's selection is governed by observability. A measure of deliberation must tap into observable phenomena (Steenbergen et al., 2003, p. 23). The authors apply this criterion consistently. Authenticity, though central to discourse ethics, is excluded because true preferences cannot be observed and any judgment about them would introduce systematic error (Steenbergen et al., 2003, p. 26). The cost is knowingly accepted. The same criterion, however, excludes justificatory losses. As argued in Section 1, what is not said cannot be observed.

This produces a tension within the instrument. Steenbergen et al. state that every element of Habermas's discourse ethics except authenticity finds a place in the DQI (2003, p. 26). Yet the first of those elements, open participation, requires not only that every competent individual be free to take part but that the rules and procedures of discourse themselves be open for discussion (Steenbergen et al., 2003, p. 25). In the DQI, open participation becomes a single binary code: whether a speaker was interrupted (p. 27). The second-order demand, that the terms of discourse be open to contestation, has no code at all. This is not an error. It follows from observability, since exclusion from an arena and the fixing of its terms are not events within a speech. However, it means the DQI's own theoretical source contains a demand focused on the quality about reasoning that the instrument cannot register. To understand what that demand involves, and why it has priority, I apply Forst's account of power over the space of reasons.

Section 3: Justification as Noumenal Power

Power, in Forst's sense, is not a matter of physical force. Nor is it simply a synonym for strength, might or the ability physically to overpower another person. Consider a person who refuses to leave a building. I could pick them up and carry them out. In that case, they are moved as an object, against their will, and their agency is not engaged through a reason. For Forst, force alone does not capture the specifically noumenal dimension of power, because it moves a person as an object rather than through a reason (Forst, 2017, p. 41). Alternatively, I could tell the person that security will be called if they remain, implying that force will follow, or I could explain that the building is closing and that everyone must leave. In both cases, the person leaves because they recognise a reason for doing so, a threatened consequence in the first and an explanation of a general rule in the second. Both are exercises of noumenal power, because both work through the space of reasons (p. 41). "Noumenal" is borrowed from Kant, but Forst does not mean Kant's metaphysics of “things in themselves”. He means that power operates in the space of reasons, the realm in which persons justify what they think and do (Forst, 2017, p. 38). The threat and the explanation can both be effective, even if the reasons I give are unjustified.

The offering of a reason does not itself justify the power exercised. I might tell the person that they must leave because of their race, gender or religion. The instruction may be intelligible and effective. They will understand what I say and, possibly, follow my instruction without objecting, especially if the instruction is supported by a law. However unjust the instruction, the person may understand the authority being applied and leave in anticipation of its consequences. However effective it is, the reason fails as a justification. It does not treat the person as an equal normative authority. The reason claims for some a privilege it denies to others and it excludes the affected person's objection from the terms on which the instruction is defended. It therefore fails both criteria through which Forst assesses justification (Forst, 2017, p. 49): reciprocity (no one may claim what they deny to others) and generality (the reasons must be ones that all those affected could share). Where such an instruction is backed by law, it exemplifies the first problem in Forst’s account of domination: rule by unjustifiable norms (Forst, 2017, pp. 50–51). Legality provides the instruction with an appearance of legitimacy. This is how domination operates, by making asymmetrical relations appear legitimate or unalterable to those subject to them (pp. 49–50). In Diamond and Morlino’s terms, this bears on the justification stage of vertical accountability: elected leaders must be answerable for their decisions when citizens demand an account (Diamond and Morlino, 2004, p. 25). Where a relation appears legitimate or unalterable, that demand may be suppressed before it becomes an explicit political question. A law may be validly enacted, even by a democratic majority, and still fail as a justification to those it governs.

The contrast is not between power exercised through reasons and power exercised without them. The contrast is between reasons that may be justified to those subject to power as equals and reasons that reproduce domination by denying them equal standing. Forst's concept of power is deliberately neutral between these cases (Forst, 2017, p. 40). Identifying an exercise of power does not evaluate it. The evaluation comes later. It asks whether the reasons through which power works can be reciprocally and generally justified to those it governs. In the building, that is the question the person is entitled to ask: “why should I leave?” Here the quality of justification enters my analysis.

Understanding what noumenal power represents changes what democratic assessment must examine. With power working through reasons, the object of analysis is the justifications that move persons, meaning the reasons offered as justifying, whether or not they succeed (Forst, 2017, p. 50). The DQI examines such justifications, but only those offered as demands within a formal arena (Steenbergen et al., 2003, p. 27). The authors’ application shows its precision and limit. Coding a 1998 House of Commons debate on women's issues, they identified 56 speeches that made demands and graded the justification in each. An argument that separate taxation would help women leave abusive households, for example, was coded as a qualified justification because it linked the demand to a reason (pp. 31–32). Every one of the 56 speeches also received the code for normal participation, since no speaker complained of interruption (pp. 31–32). By design, the instrument could not register concerns that never appeared as demands in the coded speeches. Forst locates power struggles more widely, in the discursive realm where interests and preferences are formed, articulated and contested (Forst, 2017, p. 46). Locating the struggle there is not simply a matter of looking beyond formal arenas, as Steenbergen et al. (2003, pp. 43–44) and Bonvin et al. (2018, p. 963) propose. The discursive realm is not a further place alongside the chamber, as though there were several rooms. It runs through every arena, the chamber included, wherever justifications are formed, tested or sealed off. By the time a demand is voiced in a chamber, much of the struggle over what can count as a reason has already taken place. What the DQI codes is the outcome of that struggle, not the struggle.

Forst’s notion of narratives of justification explains how that prior contest shapes what follows. Social relations and institutions appear justified through narratives that frame them. These narratives can be ideological: they justify asymmetry and subordination with bad reasons perceived as good (Forst, 2017, p. 42). This is where substitution, the second form of the quality about reasoning, connects to Forst’s reflexive account of domination. Suppose the building is not closing, but I tell the person that it is, rather than state the discriminatory reason. Any discussion of opening hours then addresses a substitute. The reason moving me is never placed into contestation. Substitution does not mean that the proximate question is irrelevant. Rather, the proximate question displaces the one to which the exercise of power is answerable. The same can occur at the level of government. Debate may proceed fluently and respectfully while the justification of the power being exercised is never brought into question. An arena of discussion may formally exist, yet the structure of justification fails in the reflexive sense Forst identifies (Forst, 2017, pp. 50–51). Admission concerns who may enter the space. Substitution concerns whether, once inside, the right question can be asked.

Lukes objects that Forst's account is too cognitivist. Much significant power, he argues, works below or around reasons rather than through them: through unconscious intuitions and cultural schemas and through what he calls the power of occlusion. Occlusion is the ability to control the terms of debate, to influence what counts as a valid argument and what is taken seriously as a description of the world, so that certain ideas are excluded before questions about them can even be asked (Lukes, 2018, pp. 53–54). The objection names what the DQI cannot see, but it does not take us outside the space of reasons. Forst allows that noumenal power ranges from reflective acceptance of reasons to almost blind compliance (Forst, 2017, p. 41). Occlusion, I argue, is better understood as sealing at its most complete than as power that bypasses reasons. The alternative never presents itself as a possibility. Lukes himself adds two points that support the argument here. Occlusion is never wholly effective, and is less effective in open democratic societies, where contrary views are marginalised rather than silenced. It can be studied empirically, for example through topic modelling (Lukes, 2018, p. 54). Occlusion is a matter of degree and it leaves traces. It can also operate at two points: it can prevent a view from forming at all, or leave it formed but unable to gain a hearing. The first is a failure of capacity, the second a failure of uptake.

Formal position does not settle uptake on its own. Forst notes that institutionally weak persons or groups can generate considerable noumenal power, while some who hold strong positions are hardly taken seriously; they lack noumenal power because they cannot manoeuvre within, or influence, the public space of reasons (Forst, 2017, p. 51). For example, in June 2020 the footballer Marcus Rashford, holding no political mandate, wrote an open letter to MPs urging the government to keep funding free school meal vouchers over the summer holiday. Within days the government reversed its decision (House of Commons Library, 2020). The opposite is equally familiar. A minister whose announcements are routinely discounted holds a strong position but little noumenal power, especially when removal from office appears imminent. Position and uptake can therefore come apart in both directions: Rashford had uptake without office, the minister office without uptake. Formal position, like formal participation, does not guarantee that reasons will register, and uptake must be assessed in its own right. The analysis must also remain a matter of degree. Even the most powerful cannot determine or close off the space of reasons entirely (Forst, 2017, p. 51), so exclusion and uptake are never all or nothing.

My central question can now be addressed: can the quality about reasoning be assessed? My answer, drawing on Forst, is that it can be and must be. Doing justice to the power formations in a society requires that "an appropriate matrix must be worked out", covering both the plurality of reasons for which orders are accepted and the positions of discursive power across the media, religion, politics, education and elsewhere (Forst, 2017, pp. 50–51). He adds a caution. As Kant observed, we cannot look into people's heads to discover which reasons actually move them, so any analysis of noumenal power must accept ambivalence and contestation and can never be final or completely objective (p. 51). This caution mirrors the DQI's exclusion of authenticity on grounds of observability (Steenbergen et al., 2003, p. 26). Both accept that inner motivation cannot be codified. The difference is where each then looks. The DQI confines itself to what is said in an arena. Forst's matrix looks to positions, structures and dominant narratives, which can be investigated even when motives cannot be directly observed. An assessment of the quality about reasoning therefore need not claim an objectivity it cannot have. Any claim must be comparative, revisable and open to contestation. That is not a defect in a democratic measure. It is what we should expect of an assessment of the conditions of justification that is itself subject to justification.

Forst's account also supplies the normative ground for the priority of the second quality. Democratic rule exists where those subject to a normative order are also the normative authorities who co-determine it through democratic procedures of justification, their standing as justificatory equals secured by rights and institutions (Forst, 2017, p. 49). A basic structure of justification among free and equal persons is, for Forst, the first demand of justice: fundamental justice (p. 51). This supports a constitutive rather than merely instrumental reading of legitimacy. The quality about reasoning is not simply a precondition that makes good reasoning more likely. It is a key part of what makes an outcome justified to those who lose. It also sets a threshold rather than an ideal. Fundamental justice asks for a structure in which justification is possible for all, not a society in which every question has been fully justified. That is the thin standard I hold. The standard is transcendental in its demand but not in its application. Assessment is comparative and revisable, which meets the concern raised by Bonvin et al., following Sen, that critique should not depend on a transcendental ideal (2018, p. 966).

Assessing the quality about reasoning in practice requires analysing power on two levels (Forst, 2017, pp. 48–50). The first is discursive: the dominant reasons and narratives of justification and their specific content. The second concerns positions, the different "normative powers" or "noumenal capital" that agents have to generate and use discursive power, including status, competence and institutional structures (p. 50). The questions at this second level are sociological, asking who has which possibilities to influence the dominant order of justification and how relations of justification are arranged across social spheres and in political life (pp. 48–49). The DQI works at a narrow part of the first level, coding the content of demands in one arena, not the dominant narratives within which those demands make sense. Capacity and domination operate across both levels, but they become especially important at the second, the level of the positions and structures that determine who can generate, circulate and contest reasons. Capacity concerns the noumenal capital agents can bring to the space of reasons and domination concerns the arrangements that close the space against them.

Section 4: Capacity

By capacity I mean the practical and communicative ability to give, receive and contest reasons: the noumenal capital agents bring to the space of reasons (Forst, 2017, p. 50). In the building, it is whether the person can ask "why should I leave?" at all. They must understand the instruction, have the words to reply and have the confidence and means to question it. These three elements correspond to receiving, giving and contesting reasons. A deficit in one weakens the others. Someone who cannot understand an instruction cannot contest it. Whether they can afford the risk of asking is a separate question, which I address in Section 5. Capacity belongs to persons, individually and as groups. Institutions matter to capacity, not as capacities themselves but as what Bonvin et al., following Sen, call conversion factors, the individual and collective conditions that turn formal rights to participate into real ones (Bonvin et al., 2018, p. 965). A citizen with a vote but without time, information or the means of organising with others holds a right that is formally equal but practically diminished. Converting that right into practice depends on material resources, since participation costs time and energy and is rarely paid, and, for the most disadvantaged, on the ability to organise collectively and so counterbalance asymmetries of status and competence (p. 965). Here I differ from Bonvin et al.'s contrast between formal and "real" democracy (2018, p. 962). The vote is not an abstraction awaiting substance. Losing a vote under fair conditions is what democratic losing ordinarily means. A loss becomes wrongful when formal rights to make claims are not converted into real ones. Capacity marks one side of that boundary; domination, addressed in Section 5, marks the other.

Bonvin et al. draw on Knight and Johnson to specify the capacities that political influence requires (2018, pp. 958–959). The first is cognitive: access to information and to decision-making, together with the ability to define the agenda or frame of collective decisions. The second is the capacity to form one's own preferences rather than adaptive ones, shaped by the limits of one's situation. The third is the capacity to make effective use of recognised cultural resources, which is especially demanding when minorities must express their views in the language of dominant groups. In the building, someone who does not know that the closing time is posted, or that they could ask to see it, lacks the first capacity. Someone who has been asked to leave so often that they no longer think to object has adapted their preferences to their situation and lacks the second capacity. Someone who must object in a language or register that is not their own, to a doorman fluent in it, lacks the third. These capacities are unequally distributed, which calls into question the apparent inclusiveness of deliberative processes even where inclusion is formally recognised (p. 959). The first capacity overlaps with noumenal power. The ability to frame a question is a capacity when it describes what an agent can do. When it is exercised over others, so that they must argue within a frame they did not set, it is noumenal power, whose effect I identified as substitution in Section 3. The same phenomenon appears on both sides of the relation: one group's capacity to frame is another's exposure to being framed. As with noumenal power generally, neither framing nor being framed is a wrong in itself. Every debate has a frame and someone must set it. The evaluative question is whether the frame could be justified to those who must argue within it and whether they could contest it.

Capacity can fail at a level deeper than individual skill. Fricker's account of hermeneutical injustice describes a lack of shared interpretive resources: some significant area of a person's social experience is obscured from collective understanding (Fricker, 2007, p. 155). Her central example is Carmita Wood, who left a job after persistent unwanted advances from a senior colleague. Asked why she had left, she could only answer that her reasons had been personal, and the claim was denied (pp. 149–150). The concept that would have made her reasons intelligible, sexual harassment, had not yet been named. Wood did not lack intelligence or articulacy. She lacked a concept the collective resource had failed to supply. What she had to say could not register as a reason. Capacity shapes, in this sense, what can count as a politically intelligible reason, not only who participates. Wood's case illustrates the first form of occlusion identified in Section 3, a view prevented from forming. Its counterpart, a view formed but denied a hearing, is Fricker's testimonial injustice (2007, p. 28), the failure of uptake that Section 3 showed can occur regardless of formal position.

The case study also shows how capacity relates to domination. Hermeneutical injustice, Fricker argues, is not inflicted by any agent (Fricker, 2007, p. 168). No one silenced Wood before the claims investigator. Fricker traces such gaps instead to hermeneutical marginalisation, the unequal participation of powerless groups in the practices by which social meanings are generated (pp. 152–153). In a single encounter, capacity and domination can diverge. Wood lacked capacity although no one before her in that encounter dominated her. Over time, however, domination can produce incapacity. The relation runs deeper than cause and effect. What Wood could say depended on what could count as a reason, so her capacity was partly constituted by the space of reasons in which she spoke. Domination, in turn, works through that space, as Section 3 showed, by making asymmetrical relations appear legitimate, and it is sustained by the incapacity it produces. Capacity, domination and justification are in this sense mutually constitutive. Each partly shapes what the others are. They must be assessed separately, because in particular encounters they can diverge, and together, because none is what it is without the others. When the concept of sexual harassment was named, all three shifted together. Those in Wood's position gained a capacity she had lacked, the domination she had suffered became visible as a wrong and the space of reasons admitted a reason it had previously excluded. Such gaps are often visible only in hindsight, once the missing concept has been supplied. An assessment cannot identify concepts that do not yet exist, but it can examine the conditions Fricker identifies, asking who participates in the practices through which social meanings are generated.

The hindsight problem also clarifies the distinction I drew in Section 3 between a standard that is transcendental in its demand and an assessment that is comparative in its application. The demand that those subject to power be able to call for its justification does not wait on a society's recognising a particular wrong, though what can be measured does. Before unequal treatment of women was widely named as discrimination, it could not be measured as such, but women's exclusion from the institutions in which social meanings are made could be. Assessment must also remain open to revision as such concepts emerge. Within these limits, capacity is the dimension most readily assessed. Its conditions are largely observable: material resources, education, access to information, the presence of associations through which the disadvantaged can organise and the existence of public spaces in which less conventional forms of expression are recognised (Bonvin et al., 2018, pp. 965–966). Following Forst, I treat capacity as the most observable part of the matrix of positions that an analysis of noumenal power requires (Forst, 2017, pp. 50–51). The capability approach offers a way of framing such assessment. It asks not what people achieve but what they are really able to do: the freedom to lead "the kind of life he or she has reason to value" (Sen, 2001, p. 89). Critique proceeds by asking how much "capability for voice" is left to all stakeholders (Bonvin et al., 2018, p. 962). Because the dimensions are mutually constitutive, however, a measure of capacity cannot stand alone. Greater capability for voice means little where domination closes the space in which it would be used. In the building, a person may understand the instruction, have the words to reply and the confidence to question it, yet know that anyone who argues with the doorman is barred for good. The capacity to ask "why should I leave?" is intact. Using it has been made too costly. What the capability approach cannot supply on its own is the normative account of why a shortfall in capability for voice is a democratic wrong rather than a disadvantage. That account comes from Forst. A shortfall in capacity denies persons their standing as justificatory equals: those to whom reasons are owed and who must be able to contest them.

Section 5: Domination

Domination does its work before it is exercised. It rests on one group's ability to impose a cost. That ability is effective merely by being known: one group's power to sanction is another's reason not to ask. Section 4 ended with a person whose capacity to ask "why should I leave?" is intact but whose use of it has been made too costly. Lukes charges Forst with missing this, calling it the "exercise fallacy": the most effective power secures compliance without lifting a finger, through anticipated reactions and deference to status (Lukes, 2018, p. 49). The charge is only partly correct. Forst describes the phenomenon himself: within a patriarchal structure, women may conform to patriarchal rules even where the patriarch leaves things implicit or is absent (Forst, 2017, p. 45). What Lukes identifies is a matter of classification. Forst reserves "power" for intentional exercise and calls such cases "influence" (pp. 45, 49). For assessment, the label matters less than the effect. In both cases, the demand for justification is withheld.

The doorman need not bar anyone. It is enough that everyone knows he can. This does not take domination outside the space of reasons. The anticipated sanction is recognised. Recognition gives a reason not to ask, just as the explicit threat in Section 3 gave a reason to leave. What changes is that the threat need not be voiced. Domination, in this sense, is an enduring condition in which demands for justification are withheld because their cost is anticipated.

At the scale of a polity, domination by anticipation produces silence rather than conflict. These are the losses that Section 1 described as leaving nothing to record. Diamond and Morlino locate part of the work of vertical accountability in civil society, which sustains it by monitoring, questioning and demanding justification from those who govern (Diamond and Morlino, 2004, p. 25). Section 3 showed one way the justification stage fails: a relation that appears legitimate prompts no one to ask. Anticipation adds a second. The demand is recognised as possible, even as necessary to reduce the impact of domination on capacity, but withheld as too costly. Diamond and Morlino note that vertical accountability can be undermined not only by fraud or intimidation but by more subtle means, such as unequal access to finance and the media (p. 29). Domination by anticipation is subtler still. Nothing is prevented, because nothing is attempted.

Silence is difficult to measure, but anticipation leaves traces. The V-Dem Institute records media self-censorship as a growing problem in 39 countries, including established democracies such as the United Kingdom and the United States, describing the practice as a pre-emptive measure to avoid direct censorship or persecution (Nord et al., 2026, p. 13). The same report tracks the repression of civil society organisations, which is deteriorating in 39 countries (p. 13). A measure of self-censorship is a measure of domination by anticipation. It records statements not made and attributes their absence to anticipated cost.

The use of non-disclosure agreements in workplace harassment and discrimination cases gives a concrete instance of such statements. A House of Commons committee found that clauses unlikely to be enforceable were nonetheless widely used to deter disclosure (House of Commons Women and Equalities Committee, 2019). They silenced by being believed, not by being enforced. Unlike Carmita Wood, those bound by such clauses had the concept of harassment; what they lacked was the freedom to use it. The case also shows that the conditions of anticipatory domination are themselves observable. They include relations of dependency, such as employment, housing or benefits, in which dissent carries a cost. They also include the sanctions available to those who hold power, and the legal protections, or their absence, for those who contest it. As with capacity, and for the reason given in Section 3, that motives cannot be directly observed, the assessment looks to observable factors and traces, and its findings are matters of degree.

Domination is not exercised only by those in power. In Section 1 I noted that when losers reject an electoral result itself, the fault may lie not in the count but further down. The reverse can also hold. A rejection of a clean result can be an attempt to occupy the space of reasons, persuading supporters, for example through unfounded claims that votes were rigged, that a justified outcome is illegitimate. The difference lies between contesting the reasons for a decision, which is what standing entitles losers to do, and denying that the decision was reached through legitimate procedures at all. Forst's criteria provide the test. A contestation that could be justified reciprocally and generally to those who won is an exercise of standing. A rejection that could not be so justified is itself a bid for domination.

Anticipation also deepens the mutual constitution I describe in Section 4. A demand never made is never refined. Likewise a concept never used in public is never developed. Domination suppresses the use of capacity. Disuse weakens capacity further. Both narrow the space of reasons in which justification could be demanded. In Section 6 I draw these threads together as the conditions that an assessment of the quality about reasoning must capture.

Section 6: The Quality About Reasoning

In Sections 3 to 5 I have examined justification, capacity and domination in turn, showing them to be mutually constitutive. Here I draw them together as four conditions that an assessment of the quality about reasoning must capture: admission, standing, uptake and resistance to substitution. The building case study illustrates each. Admission is whether the person may enter at all. Standing is whether, once inside, they may ask "why should I leave?" without the question being made too costly. Uptake covers whether the answer they receive responds to what they have said. Resistance to substitution tests whether that answer gives the real reason or a substitute for it.

Young's distinction between external and internal exclusion organises the four conditions. Democratic norms, she argues, make inclusion a criterion of legitimacy: outcomes are legitimate only if those who must abide by them had a part in forming them and those who disagree with an outcome must accept its legitimacy if it was reached through an inclusive process (Young, 2000, p. 52). That is the legitimate loser of Section 1. External exclusion names the ways in which people who ought to be included are left out of the fora of discussion and decision-making (p. 53). Internal exclusion names the ways in which people lack effective opportunity to influence the thinking of others even when they have access to those fora (p. 55). The distinction yields three distinct positions. The legitimate loser is present, has standing and has been outargued or outvoted. The internally excluded are present but lack effective influence. The externally excluded are absent. The DQI is built for the first position, reaches partly into the second through its indicators of respect but cannot see the third at all. Young adds that external exclusion impedes political equality more fundamentally than internal exclusion (p. 55). External exclusion, which the DQI cannot see, is the more fundamental.

Admission concerns external exclusion. Its barriers may be formal, but they are more often practical. Young points to the location and timing of meetings, and to back-door brokering, in which agendas are settled privately and presented to the public as accomplished facts (Young, 2000, p. 54). Its sources are the two dimensions I examined in Sections 4 and 5. A lack of capacity means formal rights are not converted into real ones. Domination by anticipation means the costs of taking part outweigh its value. Admission is assessed by presence: who takes part, across formal arenas and in the institutions in which social meanings are made, and who does not.

Standing concerns whether those who are present can act as normative authorities, entitled to demand justification and to contest it. It requires both capacity and freedom from anticipatory domination. The person who has signed a non-disclosure agreement is present in the workplace and may well have the concept and the words to describe what happened, but lacks standing to use them. Standing is assessed through the indicators I set out in Section 5: the costs attached to dissent, the sanctions available to those who hold power and the protections available to those who contest it.

Uptake concerns internal exclusion. Young describes people who, once formally included, find their claims ignored, dismissed or patronised and their ways of expressing themselves treated as unworthy of consideration (Young, 2000, p. 55). In Section 3 I showed that uptake does not follow formal position: a footballer without office moved government policy, while ministers in office can find their announcements discounted. Bonvin et al. ask a closely related question: whether there are bridges by which concerns raised in informal settings reach formal ones (2018, p. 965). Uptake is assessed by whether contributions change the reasoning of others. The DQI's respect indicators capture part of this within an arena. Whether what is said in one arena is taken up in another lies beyond them.

Resistance to substitution concerns whether, once inside, the right question can be asked. Among the forms of internal exclusion, Young lists terms of discourse that make assumptions some do not share (Young, 2000, p. 52). Bonvin et al. warn that a debate framed too narrowly, in its agenda and themes, excludes concerns before discussion begins (2018, p. 964). Lukes's power of occlusion names the limiting case, in which the alternative never presents itself (2018, p. 54). Resistance to substitution is the hardest of the four conditions to assess, because a displaced question leaves only an indirect trace. It is assessed by comparison: between the questions a debate addresses and the power actually being exercised, and between the dominant narratives of justification and the reasons that could be justified to those subject to them.

The four conditions can diverge from the quality of reasoning. A debate may score well on every DQI indicator: its speakers justify their demands, respect their opponents and appeal to the common good. Yet it may take place among the admitted, about a substitute question, while those most affected remain outside or unheard. The 1998 debate that Steenbergen et al. coded may or may not have been such a case; the DQI's findings cannot tell us. Because the four conditions settle who reasons and about what, they are normatively prior. A high DQI score is a quality of the reasoning the conditions allow. Its democratic significance therefore depends on an assessment the DQI cannot make, of the conditions under which it was achieved. Read alone, a high score is ambiguous. Read alongside an assessment of the four conditions, it can be interpreted.

An assessment of the quality about reasoning is therefore a profile across four conditions rather than a single score. A democracy may admit widely but grant little uptake or protect standing while tolerating substitution. Because the dimensions that produce them are mutually constitutive, the conditions must be read together. A rise in admission means little if those admitted cannot ask their questions. Those questions mean little if they are not the questions to which power is answerable. Each condition is a matter of degree and each is assessed through observable factors and traces. The assessment is comparative, revisable and open to contestation, as Section 3 required. It will not be complete, since hindsight will continue to reveal exclusions that present concepts cannot name. However, it can say more than the DQI can about whether the losers of a democratic contest lost as justificatory equals.

Section 7: Conclusion

Democracy produces winners and losers and it is the losers who test it. I began with the claim that what makes losing acceptable is not the count alone but the structure of justification beneath it. An election is the visible tip and its legitimacy depends on the quality of what lies below. I have argued that assessing democratic justification therefore requires attention to two qualities, not one.

The quality of reasoning concerns how well reasons are exchanged among those taking part. The DQI codifies it with real precision and its indicators of justification, respect and the common good approximate Forst's criteria of reciprocity and generality. The quality about reasoning concerns the space in which that exchange occurs: who may enter it, whether they may demand justification without the demand being made too costly, whether what they say is taken up and whether the question debated is the one to which power is answerable. Forst's account of noumenal power explains why the two qualities diverge. Power works through the space of reasons and it can determine, occupy or seal off that space before any demand is voiced in a chamber or other arena. Capacity and domination, which together with justification form mutually constitutive dimensions, shape not only who takes part but what can count as a reason.

Three claims follow. First, the two qualities must be distinguished. A debate can score well on every DQI indicator while those most affected remain outside it, or while its question has been displaced. Second, the quality about reasoning is normatively prior. It settles whose reasoning the first quality measures and so determines what a high score means. Third, it can be assessed, though not by coding speech alone. Admission, standing, uptake and resistance to substitution leave observable traces. These traces are seen in patterns of presence, in the costs attached to dissent, in whether contributions change the reasoning of others and in the gap between what is debated and what power is doing. The assessment is a profile of degrees, comparative and revisable, and hindsight will continue to reveal what present concepts miss.

The standard behind the assessment is thin. It does not ask for a society in which every question has been fully justified, only for a structure in which justification can be demanded by all. That is what distinguishes the two kinds of loser with which the paper began. The legitimate loser has been outargued or outvoted while keeping their standing. A win over them is not a domination. The justificatory loser has lost something prior to the vote: the standing to ask and to be answered.

The next stage is to develop a framework of indicators general enough to hold the four conditions and precise enough to withstand disputes about what counts as evidence. Much of what it needs already exists. The DQI assesses the first quality (of reasoning), while indicators such as V-Dem's measures of self-censorship and civil society repression, together with the capability approach's attention to voice, already track parts of the second (about reasoning). The four conditions need developing into indicators that can be applied across arenas and compared across cases. Relations between them need empirical as well as conceptual investigation. The direction, however, is clear. Return one last time to the building. Whether democracy is working there does not depend only on how courteously the closing time is explained. It depends on whether the person may ask "why should I leave?", whether asking is safe, whether the answer responds to what they say and whether it gives the real reason. That is the quality about reasoning and it is what democratic assessment must learn to see.