This article begins with an empirical finding: U.S. voters are more “competent” than research into opinion formation once took them to be (e.g., Carmines and Kuklinski Reference Carmines, Kuklinski, Ferejohn and Kuklinsky1990; Lupia Reference Lupia1992, Reference Lupia1994; Lupia and McCubbins Reference Lupia and McCubbins1998; Popkin Reference Popkin1991; Sniderman Reference Sniderman, Lupia, McCubbins and Popkin2000; Sniderman, Brody, and Tetlock Reference Sniderman, Brody and Tetlock1991).Footnote ¹ They can form opinions and hold preferences that are coherent and stable enough to be represented. However, the reasoning process by which they form these preferences depends on communications put forward by political elites in their bids to forge a winning majority in an election or policy contest.Footnote ² Elites educate constituents as they recruit them to positions that work to elites’ own advantage in an interparty struggle for power, typically without avowing the dual motive (Gerber and Jackson Reference Gerber and Jackson1993). Political learning takes place by means of communication that is at once explicitly oriented to constituents and silently enmeshed in a struggle for power.

This finding sits uneasily with two prominent democratic intuitions that belong to two different schools of democratic theory. First is the idea that representative democracy is defined by the “continuing responsiveness of the government to the preferences of its citizens” (Dahl Reference Dahl1971, 1).Footnote ³ This idea, whose orientation toward preexisting preferences makes it one aspect of what Mansbridge (Reference Mansbridge2003, 518) calls the “traditional model of promissory representation,” not only makes responsiveness the signature feature of democratic representation, but also prescribes that it be unidirectional. Pitkin (Reference Pitkin1967, 140) captures this when she writes that, in democratic representation, the “representative must be responsive to [the constituent] rather than the other way around.”

Pitkin expresses here what I term the “bedrock” norm, the common-sense notion that representatives in a democratic regime should take citizen preferences as the “bedrock for social choice” (Page and Shapiro Reference Page and Shapiro1992, 354; cf. Achen Reference Achen1975, 1227; Bartels Reference Bartels, McKuen and Rabinowitz2003, 62).Footnote ⁴ According to this norm, which configures the representation process as linear and dyadic, legitimacy turns on voters and representatives being oriented in the proper direction. As Mansbridge (Reference Mansbridge2003, 518) aptly characterizes it, the “voter's power works forward” to hold representatives to the promises they made at election time, whereas the “representative's attention looks backward” to the previously expressed preferences of the constituency. Empirical findings about preference formation sit uneasily with this norm because they defy this static portrait of preferences together with this linear dyadic model of influence. They reveal, instead, that the representative process is dynamic and interactive. Representatives look backward to preferences that have been expressed, and orient themselves forward in a speculative mode toward what their constituency might want or be induced to want at the next election. In short, empirical research reveals political representation to be constitutive: legislators do not simply respond to constituent preferences but are “active . . . in searching out and sometimes creating them” (Mansbridge Reference Mansbridge2003, 518; cf. Manza and Cook Reference Manza and Cook2002; Squires Reference Squires2008; Williams Reference Williams1990.

This research shows that responsiveness does, indeed, turn “the other way around.” Does this inversion irredeemably compromise democratic ideals? For the most committed formal modelers and theorists of rational choice, the answer would have to be “yes” because preference endogeneity in itself troubles their model of the rational individual. In contrast, deliberative democratic theorists allow that political elites can shape constituent preferences as long as their interventions facilitate “mutual education, communication, and influence,” rather than manipulation (Mansbridge Reference Mansbridge2003, 520). For them, the problem is not endogeneity per se but the opening it affords to what they deem strategic action by elites.

This belief that preferences, which are necessarily endogenous to politics, must be deliberatively formed is the second democratic intuition. It is even more at odds with the scholarship on opinion formation by virtue of its intellectual inheritance from discourse theories of democracy. Specifically, it derives from the Habermasian analytic separation of “communicative and strategic action,” which aims to distinguish (counterfactually) social coordination that is achieved noncoercively, by the give and take of reasons among speakers oriented toward mutual understanding, from that which either results from the overt use of threats or bribes or occurs under conditions where a party conceals an orientation toward nonmutual advantage (Habermas Reference Habermas, Lenhardt and Nicholsen1990, 58; cf. Cohen Reference Cohen, Hamlin and Pettit1989).Footnote ⁵ The discourse ideal casts suspicion on preferences that are formed in information contexts where power is at stake and where unstated motives exist—the very conditions of preference formation according to empirical scholarship.

Scholars who are influenced by this ideal propose to judge representation by a distinction between education and manipulation that, as Mansbridge (Reference Mansbridge2003) acknowledges, is “not easy to operationalize.” They define education as (1) lacking intent to deceive, (2) serving to clarify voters’ “underlying interests and the policy implications of those interests,” (3) leading voters to make choices in their interests, and (4) garnering support for changes that voters approve in retrospect (519; cf. Page and Shapiro Reference Page and Shapiro1992, 354; Zaller Reference Zaller1992, 313). One problem with these criteria is that they beg the question “what is and what is not in an individual's interests,” another point that Mansbridge (Reference Mansbridge2003, 519) acknowledges to be open to contest. Thus, identifying manipulation requires the analyst to posit the “counterfactual” of how voters would define their interests “under other, often ill-defined circumstances” (Druckman Reference Druckman2001a, 233). A second problem is that if one adheres to these criteria, then one is bound to find manipulation everywhere.

This is not because deception is everywhere, although elites certainly engage in it (e.g., President Johnson's claims regarding North Vietnamese attacks in the Tonkin Gulf or President Bush's assertions regarding the existence of “weapons of mass destruction” in Iraq). Much of the time, political discourse is neither straightforwardly false nor intentionally misleading. It consists, rather, of what Jacobs and Shapiro (Reference Jacobs and Shapiro2000, 27) call “crafted talk”—messages that political elites systematically develop through the technologies of opinion polling and focus groups “in order to attract favorable press coverage and ‘win’ public support for what they desire.” Such messages have a twofold motivation and a twofold effect: to declare a stand and to “shift . . . the line of cleavage” in a bid for power (Schattschneider [Reference Schattschneider1960] 1975, 61). Put simply, whereas elite discourse serves at once to inform potential voters and to recruit them into a winning majority, elites avow only the first of these goals. To disclose the second self-interested aim, as Habermasian deliberation requires, would make their speech acts less likely to succeed.Footnote ⁶

Empirical research shows that in the “political environment,” “communicative” and “strategic” action are linked inextricably.Footnote ⁷ Individuals form coherent and relatively stable preferences not in spite of but by means of messages that political elites deploy in pursuit of unavowed competitive goals. This sets up what I term the “dilemma of democratic competence”: citizens’ capacity to form preferences depends on the self-interested communications of elites. Research that aimed to vindicate representative democracy ends up provoking the discomfiting sense that “one form of incompetence simply replaces another. Specifically, while people who rely on party cues avoid basing their preferences on arbitrary information, they also expose themselves to the possibility of elite manipulation” (Druckman Reference Druckman2001a, 239). Thus, communication that discourse theory regards as manipulative turns out to be intrinsic to the learning process.

What are researchers to make of the fact that they can affirm citizens’ capacity for preference formation only at the cost of revealing their susceptibility to the self-seeking rhetoric of competing elites? Must they despair of citizens’ capacity to live up to democratic norms? Must they denounce the communications apparatus of mass democracy? Or is it time to call the norms themselves into question? I argue for the latter. To rest democratic representation either on the “bedrock” of citizen preferences or on a cognitivist model of deliberation is to misunderstand what preferences are and how they form. It is also to deprecate crucial features of democratic representation: figuration and mobilization.Footnote ⁸ Democratic political representation neither simply reflects nor transmits demands; it creates them as it actively recruits constituencies.Footnote ⁹

The argument continues over four sections. I begin by specifying the challenge that empirical findings on opinion formation pose to preference-based conceptions of democratic representation. I turn next to the new wave of democratic theory and show how its allegiances to a vestige of foundationalism prevent it from answering this challenge. Following that, I make a brief return to Pitkin's (Reference Pitkin1967) classic text, which I argue both initiated a bolder move toward representation as mobilization than her successors have managed and opened the gap between normative expectations and empirical research. I build my conception of representation as mobilization out of Pitkin's most radical insights, bringing that conception within the compass of deliberative democracy by drawing on the work of scholars of “rhetorical deliberation” (Garsten Reference Garsten2006; cf. Bickford Reference Bickford1996). I propose “reflexivity” as the measure of its legitimacy. A term that I take up from Mansbridge (Reference Mansbridge2003, 518), I elaborate “reflexivity” as a systemic capacity rather than an individual subjective one.

THE “CONSTRUCTIVIST TURN” IN OPINION FORMATION AND PREFERENCE-BASED DEMOCRACY

Beginning with Gamson and Modigliani (Reference Gamson and Modigliani1989), research in public opinion and political psychology has taken what researchers call a “constructionist” or “constructivist” turn (Kinder and Nelson Reference Kinder, Nelson, Callaghan, Schnell and Entman2005, 103) from a rational individualist to an environmental or contextual account of preference formation. The former holds individuals to define their preferences according to needs, desires, and values that are fixed prior to political contests and relatively impervious to political influence. The latter contends that preferences are constituted in the communication that occurs during decision making, implying that choice is as much something that institutions effect as it is something that an individual makes. Insofar as this is a move from an individualist to an environmental account of preference formation, it might seem more appropriate to label this position “contextualist” rather than “constructivist.” I contend that it can indeed be viewed as a “constructivist” turn by virtue of the commitment to the notion that opinion is not simply there to be “discovered or intuited” but must be built either by researchers or by the individual him- or herself (O'Neill Reference O'Neill and Freeman2002, 348).Footnote ¹⁰

Findings regarding the context dependence of preferences have significant implications for the “bedrock” norm, which holds that democratic representation is necessarily preference based. For those who subscribe to this norm, preferences that fluctuate in response to politicians’ and pollsters’ “choice of language” lack the “nice properties of global coherence and consistency that would allow them to play the role of preferences” as “most liberal theorists of democracy (and their cousins, the economists)” have defined the term (Bartels Reference Bartels, McKuen and Rabinowitz2003, 67, 56, 49). Their expectation is that citizens should hold “definite, preexisting preferences regarding the underlying issues [that] any reasonable choice of language might elicit . . . equally well” (51, 67). I argue for a reframing of this expectation. Is it the case that preferences must display the constancy of bedrock to count as such? Must political representation take preferences as its starting place and ground in order to be democratically legitimate? Constructivist research in opinion formation says “no” to both.

As to the first expectation, Chong and Druckman (Reference Chong and Druckman2007, 652) counter that the decisive question is not whether preferences change but whether they do so arbitrarily, in response to logically equivalent information.Footnote ¹¹ In political discourse, opinion shifts in response to what Druckman (Reference Druckman2001a, 235; Reference Druckman2004, 672) calls “issue” framing, instances where a speaker brings “a substantively different consideration . . . to bear on the issue at hand.” Such change, which is not necessarily arbitrary (although it may be so in some cases), is an ever-present feature of democratic politics, which “involves battles over how a campaign, a problem, or an issue should be understood” (Druckman Reference Druckman2001a, 235). Issue framing is one means of waging what Schattschneider ([Reference Schattschneider1960] 1975, 64) termed the “conflict of conflicts.”

This sensitivity of preferences to contexts of conflict need not compromise their integrity as preferences, except to the methodological individualist who holds that preferences must have the “nice properties of global coherence and consistency” in and of themselves, that is, independent of or prior to political contests (Bartels Reference Bartels, McKuen and Rabinowitz2003, 56). For Druckman and Lupia (Reference Druckman and Lupia2000), who have mustered the insights of cognitive science against rational individualist notions of rationality, preferences simply do not attain consistency in advance. As “comparative evaluation[s]” or “ranking[s] over a set of objects” that conflict brings into relationship with one another, they necessarily develop over the course of a political contest (2). This view of preferences takes the normative charge out of the distinction between preference “endogeneity” and “exogeneity” because it suggests that preferences are always endogenous to some context that is beyond an individual's control.

For these researchers, so-called endogenous preferences are preferences. Features of the environment such as cues and frames help individuals form preferences that are consistent and relatively stable. Such preferences can be represented; they cannot serve as a basis for responsiveness as the “bedrock” norm requires because they cannot have “independent causal import” (Manza and Cook Reference Manza and Cook2002, 657). In sum, there is an emerging consensus that political representation need not and cannot take preferences as its starting place and ground. It remains in dispute whether—given the context dependence of preferences—representation can be democratically legitimate.

In exploring this question, it is noteworthy that researchers do not assess the rationality of voter preferences against the counterfactual of what their interests “really” are. They look at the interaction among messages, rather than at their content, focusing on the “external forces” that make it more likely for political learning to take place (Lupia and McCubbins Reference Lupia and McCubbins1998, ch. 2). Framing research, for example, has begun to study the competition among frames and between frames and cues (e.g., Druckman et al. Reference Druckman, Hennessy, St. Charles and Webber2010; Sniderman and Theriault Reference Sniderman, Theriault, Saris and Sniderman2004).Footnote ¹² In effect, this research studies deliberation in the actual public sphere of electronically mediated, highly funded, two-party mass democracy. Taking political communication as it is, neither oriented toward mutual understanding nor regulated by the principle of the unforced force of the better argument, but competitive, self-interested, and oriented toward winning elections, it asks: can competition prompt people to think twice about their preferences rather than absorbing them from elite propaganda? Studying frames, Chong and Druckman (Reference Chong and Druckman2007) find three ways in which competition fosters political judgment in this sense. First, it can moderate ideological extremes. Second, exposure to multiple frames prompts people to consider which is the most applicable. Third, people make that judgment based on a frame's “relative strength,” not merely on its frequency of repetition (639, 651).

That “strong” frames win out over loud and insistent ones is no guarantee of democratic legitimacy, however. For a frame to be “strong” means only that it is credible—that it is proposed by a respected source, taps familiar concerns, and applies to the matter at hand. A strong frame is not necessarily a strong—i.e., disciplined and well-reasoned—argument. It may derive that credibility “as much [from] its source and cultural values and symbols it invokes as [from] its causal logic” (Chong and Druckman Reference Chong and Druckman2007, 652).Footnote ¹³ It follows that competition among frames cannot satisfy the Habermasian counterfactual of producing a reasoned position that warrants the assent of all affected. Competition may raise the quality of public debate by improving the “odds that germane considerations will be publicized and discussed” (652). Druckman (Reference Druckman and Keren2010) cautions, however, that a frame may prevail over its rivals by tapping insecurity and prejudice rather than by appealing to logic and fact. Some researchers who focus on cues rather than frames are more optimistic. Lupia and McCubbins (Reference Lupia and McCubbins1998, 201) maintain that democratic institutions render people capable of drawing knowledge from the political environment that enables them to be “selective about whom they choose to believe.”

Noteworthy here is researchers’ turn from the bedrock of autonomous preference to the communication process. This shows, as Mansbridge (Reference Mansbridge2003, 518) argues, that empirical research is newly open to the “deliberative side” of political representation. Yet, I argue that normative theory is not fully equipped to exploit this opening. There remains a mismatch between the constructivism of empirical research and the “deliberative theory of democracy,” which Urbinati (Reference Urbinati2006, 118–19) rightly characterizes as “hesitant to face” the “ideological as rhetorical process of representation.” This hesitancy is true of even the most politically astute defenders of deliberation insofar as they hold out for an ideal of independence from partisanship that puts their work out of synch with the findings on preference context dependency.

THE DEMOCRATIC “REDISCOVERY” OF REPRESENTATION

Following the wave of academic theory that embraced “participatory” democracy in the wake of the Civil Rights, student, and antiwar movements (e.g., Barber Reference Barber, Benjamin1984; Miller Reference Miller1987; Pateman Reference Pateman1970), normative scholars have proclaimed a “democratic rediscovery of representation” (Urbinati Reference Urbinati2006, 5; see Dovi Reference Dovi2002; Manin Reference Manin1997; Mansbridge Reference Mansbridge1999, Reference Mansbridge2003; Plotke Reference Plotke1997; Schwartz Reference Schwartz1988; Seitz Reference Seitz1995; Urbinati and Warren Reference Urbinati and Warren2008; Young Reference Young, Shapiro and Kymlicka1997, Reference Young2000). This rediscovery begins by questioning the assumption that democracy need have its origins in a constituency, as a democratic constituency exists “at best potentially” (Young Reference Young2000, 130). Being large, dispersed, and vaguely defined, it “rarely brings itself to affirm a common will,” but requires “representative institutions and the process of authorization themselves [to] call its members into action” (130). In effect, then, as public opinion researchers have come to recognize the context dependency of preferences, theorists of political representation have taken a similar turn to conceive of the “people,” democracy's political subject, as endogenous to the process of representation.

The rediscovery of representation alters the very valence of “representative democracy.” This phrase that once struck participatory democrats as an “oxymoron” (Urbinati Reference Urbinati2006, 4) for putting representatives in the place where the people should be, now strikes theorists of representation as “in fact a tautology” (Näsström Reference Näsström2006, 330). It is only through representation that a people comes to be as a political agent, one capable of putting forward a demand. Thus, representation cannot be regarded as either supplementary or compensatory; it is “the essence of democracy” (330).

This, too, is a constructivist turn, one motivated by a critical awakening to what Young (Reference Young2000, 125) calls democratic political theory's “metaphysics of presence.” Presence, a concept that Young borrows from Derrida ([Reference Derrida and Allison1967] 1973), names the fantasy of a reality that is self-evident, unmediated by social processes, and sovereign so that it can be imagined to provide an origin and point of reference for assessing the accuracy and faithfulness of any attempt to represent it. To reject such a fantasy is precisely to refute the assumption that representation is a “descriptive and mimetic” process, one that merely transmits “something preexisting it, like for instance a single or collective sovereign that seeks pictorial representation through election” (Urbinati Reference Urbinati2006, 46, 33). For proponents of the “rediscovery” of representation, “democratic politics is constituted partly through representation” (Plotke Reference Plotke1997, 31; emphasis added).

There remains a conceptual gap between normative and empirical bodies of work that opens at the most basic question of what is to be represented. Many normative theorists would agree with Plotke (Reference Plotke1997, 32) that “interest representation” is the “starting point in a democratic view of representation.” Plotke's choice of the term “interest” rather than “preference” marks something more than a semantic disagreement with empirical scholars. Mansbridge (Reference Mansbridge2003, 519–20) affirms a qualitative difference between the two, asserting that “questions regarding voters’ interests, in contrast to their preferences, are not susceptible to certain resolution.” Whereas Mansbridge (517, n. 6) regards preferences in behavioral terms, she takes pains to gloss “interest” as including “identity-constituting[,] ideal-regarding commitments as well as material needs.” Mere preferences, she suggests, are unreflexive, but interests are “enlightened preferences,” refined dialogically in light of not just “simple cognition” but “experience and emotional understanding” (517, n. 6). The question at stake between normative and empirical scholars is: would the context-dependent “preferences” theorized by Druckman, Lupia, and other such scholars count as what Plotke and Mansbridge term “interests” or “enlightened preferences”?

Not according to the most rationalist version of interest, which is modeled by Habermas’ communicative action/strategic action distinction. Deliberative democrats have attempted to translate that ideal into politics by means of “citizens’ juries,” “mini-publics,” and other such experiments that insulate “participants from the usual sources of information and persuasion and the usual conditions under which they respond to polls about their preferences” (Rosenblum Reference Rosenblum2008, 300).Footnote ¹⁴ Such experiments are at odds with empirical research on opinion formation because they juxtapose deliberation against “partisan contestation” (300). Not only do they participate in a long tradition of antipartisanship, as Rosenblum argues, but they also open a gulf between normative models of the way citizens should reason and empirical accounts of the way they do reason (300).

Even those deliberative democrats who have criticized the rationalism of the Habermasian model persist in an intersubjective model of dialogue and an attachment to independent interests as a point of departure for the representative relationship. Plotke (Reference Plotke1997, 32) envisions a dialogic process in two stages where citizens first “aim to clarify their own preferences” among themselves, then “seek to select representatives who will try to produce suitable results.” Young (Reference Young2000, 132) also envisions a two-stage process, one that begins with “citizen participation” in a context of conflict to produce demands that are specific to that context, then seeks representatives to carry those demands forward. Even though Young describes demands emerging out of political conflict, she nonetheless insulates the process of their formation from party politics. Hers is a world where “citizens . . . form themselves into [constituencies]” and engage “in debate and struggle over the wisdom and implications of policy decisions” independently of elite cues and frames (131).Footnote ¹⁵

On such accounts, it is not elite communication but a “public sphere of discussion” that orients citizens in conflict by setting “an issue agenda and the main terms of dispute or struggle” (Young Reference Young2000, 130). To be sure, Young does not adhere strictly to the bedrock norm; she does not understand interests to be fixed, but rather conceives them to be arrived at dialogically. But she does preserve the temporal logic of bedrock, together with the requirement of unilateral responsiveness, by stipulating that it is only once citizens have “organize[d] and discuss[ed] the issues that are important to them” that they “call on candidates to respond to their interests” (130). As Squires (Reference Squires2008, 190) notes, the “process here flows from the constituents to the representatives.” In sum, although both Young and Plotke recognize that interests are not fixed prior to politics, they make them endogenous to politics in an idealized way: they are formed by practices of public reason to secure the independence and autonomy of citizens’ judgment against the opportunistic communications of elites.

These models of two-stage processes conflict with empirical accounts of context-dependent preferences in two ways. First, they hold onto an ideal of independence from partisan communication as a touchstone for democratic legitimacy, departing from liberal/economic models of politics by making interest (rather than preference) its site and dialogue (rather than preference exogeneity) its guarantor. This is effectively a deliberative rewriting of the bedrock norm. Second, they model deliberation out of the context of interparty conflict, as occurring first among citizens and then between citizens and their representatives. In contrast, as I argue, empirical research on preference context dependency shows preferences to evolve in the context of interparty competition and to be responsive to communications from representatives and other elites. These communications, although addressed to constituencies, are primarily oriented toward that competition. Put simply, citizens learn from, not in spite of, the frames and cues that political elites deploy to gain an edge in partisan contests. In short, empirical research recognizes political communication as twofold: addressed explicitly to constituencies and silently enmeshed in interparty struggle. What would happen to dialogic models if they had to confront this twofold character?

Mansbridge (Reference Mansbridge2003, 515) grapples with precisely this question in a path-breaking article where she takes up the normative implications of empirical studies of “‘retrospective voting,’” in which voters assess their representatives’ performance based on priorities that they developed over the course of the term, rather than on those that they held at the moment of election. Mansbridge explains that retrospective voting effects a “shift in temporal emphasis” that gives rise to “unexpected normative changes” (518). In the traditional promissory model, the representative is bound by a common-sense moral imperative “to do what he or she had promised the voters at [election time]” (515, 518). In retrospective voting, legislators act not only according to the “actual [past] preferences of the voter,” but also according to their own “beliefs” about “the future preferences of the voter” at reelection time. When voting is retrospective, representation becomes “anticipatory,” creating “a prudential, not a moral, relationship to those voters” (518). Whereas promissory representatives have only one course of action (to do as they said or to justify any deviation in terms of the principles they professed to hold), anticipatory representatives, who aim to please “future voters,” have two. They attempt to gauge what a voting majority will want at the next election, turning to “public opinion polls, focus groups, and gossip about the ‘mood of the nation’” so as to adapt themselves to the electorate's movement (517). They also make themselves agents of that change, attempting to shape public opinion so that voters “will be more likely to approve of the representative's actions” (517).

Mansbridge (Reference Mansbridge2003) draws two important conclusions from this new landscape. First, anticipatory voting opens up the possibility that “voters can change their preferences after thinking about them” (517). In turn, because representatives will not merely follow those changes but will actively seek to influence them, it follows that anticipatory representation “encourages us to think of voters at [re-election time] as educable (or manipulable)” (517). Mansbridge insinuates into her text (by means of the parentheses) an all-too-familiar distinction. The question is: how does she propose to tell the difference between them?

This question brings Mansbridge (Reference Mansbridge2003, 518) to her second important conclusion, that anticipatory representation is “in most instances interactive and more continually reflexive” than the promissory model allows. Consequently, the “appropriate normative criteria for judging [anticipatory and other] more recently identified forms of representation are systemic, in contrast to the dyadic criteria appropriate for promissory representation” (515). In other words, we can no longer assume that democratic representation is secured by the match (or “congruence”) between constituent policy preferences and legislative votes. Nor can we emphasize, as Young and Plotke do, the quality of deliberation among constituents and between them and their legislators. Although Mansbridge notes that it is necessary to evaluate the “quality of mutual education between legislator and constituents,” she cautions that this communication “depends only in small part on the dyadic efforts of the representative and constituent” (518–19). It is impersonal and systemic, relying on the “functioning of the entire representative process—including political parties, political challengers, the media, interest groups, hearings, opinion surveys, and all other processes of communication” (519).

Mansbridge (Reference Mansbridge2003) moves much closer to the competitive political environment than either Plotke or Young is willing to venture. She recognizes that political reasoning is mediated by “opinion polls and focus groups,” as well as by “opposition candidates, political parties, and the media” (520). These need not be mere “tools of manipulation” but may contribute to a genuinely deliberative process of interest formation, one that need not be unidirectionally responsive (as in promissory representation) but would be “reflexive” (520). What would reflexivity entail? Despite her innovative moves beyond the foundationalist model of democratic representation, here Mansbridge falls back on a familiar communicative ideal, defining reflexivity subjectively as a property of an individual's judgment rather than as a systemic capacity. For Mansbridge, a reflexive “representative system [should contribute] to ongoing factually accurate and mutually educative communication” (519).

The difficulty is that what would count as education for Mansbridge rarely occurs in actual political discourse. Education, as Mansbridge (Reference Mansbridge2003, 519; emphasis added) defines it, is a “form of influence” that meets two criteria: it “works through arguments on the merits and is by definition in the recipients’ interests.” Druckman's (Reference Druckman2001a) arguments regarding issue framing—that politics involves conflicts over precisely what considerations should be brought to bear on a particular question—cast doubt on the feasibility of this first criterion. Although it is possible to identify deceitful, irrelevant, and misleading claims, there is often no neutral standpoint from which to resolve the question what is or is not on the merits of the case.Footnote ¹⁶ As to the second criterion, Mansbridge (Reference Mansbridge2003, 519) herself has conceded that being “in the recipients’ interests” is similarly contested. Once again, in a political environment where political elites never speak exclusively as educators of constituents but use political communication to gain an edge in interparty competition, education is not an orientation but a side effect of the battle.

Despite recognizing the need to give normative theory a “systemic” turn, Mansbridge's normative vision of political communication sets aside precisely what I take to be definitive of its systemic aspect—that representatives and opinion shapers are not only (or even primarily) in relationship to potential voters, but also in competition with each other. Her definition of education falls back on criteria better suited to assess an intersubjective relationship than a representative system. Why should this be so? The pervasiveness of the Habermasian paradigm of deliberation has something to do with it. Even those who, like Mansbridge, would not endorse Habermas’ rationalism, nonetheless retain a vestige of his urge to separate “communicative” from “strategic” action. In addition, these scholars are following a path blazed by Hanna Pitkin who anticipated more of today's debates than she is credited with doing—both the insights and the impasses.

HANNA PITKIN'S “MOBILIZATION” CONCEPT OF REPRESENTATION

In 1967, Hanna Pitkin initiated a bolder break with the traditional dyadic model of representation than many of her successors managed. This claim might come as a surprise to the many scholars who would credit the most mainstream understanding of representative democracy to the summary definition that Pitkin (209) formulated: “representing here means acting in the interest of the represented, in a manner responsive to them.”Footnote ¹⁷ As Rehfeld (Reference Rehfeld2006, 3) notes, “few historical treatments have been so completely accepted as a standard account of a concept in all areas of political science” as Pitkin's. This “standard account” reduced Pitkin's central thesis to the nutshell formulation that “responsiveness is what representative government is all about” (Kuklinski and Segura Reference Kuklinski and Segura1995, 4). And this short phrase made Pitkin's work easy to assimilate to the empirical paradigm of the moment, which held “constituency influence” to be the hallmark of democratic political representation and made “congruence”—the literal match between a legislator's votes and the preferences of his or her constituents—its index (Miller and Stokes Reference Miller and Stokes1963).Footnote ¹⁸ That assimilation, in turn, covered over her intricately argued assault on a fundamental premise: that representation can be understood on the model of a principal–agent relation.

Pitkin refutes the principal–agent model by opposing the intuition, definitive for late twentieth-century liberalism, that citizen preferences are and ought to be the “principal force in a representative system” (Sunstein Reference Sunstein1991, 6–7; Wahlke Reference Wahlke1971, 272–73). She argues that legislators respond to too “great a complexity and plurality of determinants” for citizen preferences to be a driving force in legislative decisions (Pitkin Reference Pitkin1967, 214 or 220). Such decisions are multidimensional; they cannot be reduced to a “one-to-one, person-to-person relationship” between a principal and an agent (221). Even if a legislator were to want to accord the constituency pride of place, constituents seldom hold articulate and well-formed preferences on the bills that actually come before Congress. When they do, the discrete preferences of the members of a district would not add up to a “single interest,” and so lack the unity of a principal (221).Footnote ¹⁹

Pitkin (Reference Pitkin1967) sets the dyadic model aside to propose that political representation should be conceived as a “public, institutionalized arrangement,” one where representation emerges not from “any single action by any one participant, but [from] the over-all structure and functioning of the system” (221–22). For representation to take place, there does not need to be a meeting of the minds between representative and constituency, or even so much as a meeting. Pitkin contends that legislators’ “pursuit of the public interest and response to public opinion need not always be conscious and deliberate,” and that “representation may emerge from a political system in which many individuals, both voters and legislators, are pursuing quite other goals” (224).

This is Pitkin's own constructivist turn. She effectively redefines democratic representation from an interpersonal relationship to a systemic process that is anonymous, impersonal, and not seated in intent.Footnote ²⁰ The process will be judged representative so long as it “promot[es] the interests of the represented, in a context where the latter is conceived as capable of action and judgment, but in such a way that he does not object to what is done in his name” (Pitkin Reference Pitkin1967, 155). As Runciman (Reference Runciman2007, 95) insightfully argues, the measure of interest here is “negative.” The representativity of the system turns not on the match between the opinions of the represented and the votes of the representative, but rather on what he terms Pitkin's “non-objection criterion”: the “ability of individuals to object to what is done in their name” (95).

Pitkin's nonobjection criterion not only robs congruence of its substantive ground, but it also casts it under suspicion. She suggests that a congruence finding can be trusted as an indicator of democracy only when a constituency has the capacity to object. As provision for such a capacity, Pitkin calls for competitive elections, the guarantee that those who win genuinely have the power to govern, universal franchise, and the protection of opposition and its extension to all. I suggest that this list is inadequate, even on Pitkin's own terms, because it fails to make good on a significant clause of the nonobjection criterion: that the represented be “conceived as capable of action and judgment.” This, which I call the “judgment” clause, is crucial because without it there is no trusting that citizens’ objection or nonobjection has not simply been framed or primed out of the debate by habit, ignorance, or stereotype. As Druckman (Reference Druckman and Keren2010) cautions, the most successful frames may not stimulate judgment but foreclose it by tapping exactly those features of the political environment (cf. Chong Reference Chong2000, ch. 4). I come back to the question what it would take to satisfy the “judgment” clause at the end of this article.

A few of Pitkin's early readers picked up on the radicalism of her arguments. Registering this move to “understand representation as a systemic property,” they warned that it could upset many “conventional assumptions” (Prewitt and Eulau Reference Prewitt and Eulau1969, 431; cf. Hansen Reference Hansen1975, 1186). Not least among these was the reigning wisdom that representative government is democratic because it realizes not majority rule but “minorities rule” (Dahl Reference Dahl1956, 132)—the representation of a plurality of groups, understood as “context-independent entities” (Lavaque-Manty Reference Lavaque-Manty2006, 6). In contrast, Pitkin (Reference Pitkin1967, 215; emphasis added) holds democratic representation to involve acting for an “unorganized group.”Footnote ²¹ To the empirical democrats, who regarded self-organizing as prerequisite to political representation, an “unorganized” group would not only be an oxymoron, but its political representation would be next to impossible. Pitkin understands representing to participate in defining group identities. Speaking of the “national unity that gives localities an interest in the welfare of the whole,” she insists that it is “not merely presupposed by representation [but] also continually re-created by the representatives’ activities” (218; emphasis added). Her point is not that representation invents constituencies out of whole cloth but that it draws them together: it imputes to them a unity that they discover only through being represented. This makes representing an activity without a model, without certainty, and—in Pitkin's words—without “guarantee” (163).

At its most radical and unique, then, Pitkin's is what I term a “mobilization” conception of representation.Footnote ²² She holds the process of representation to participate in forming demands and social cleavages, not merely to reflect them. Thus, for Pitkin, as for Urbinati (Reference Urbinati2006, 37), political representation does “not simply allow the social to be translated into the political, but . . . facilitates the formation of political groups and identities.” It aims, then, not to reproduce a state of affairs but to produce an effect: to call forth a constituency by depicting it as a collective with a shared aim.

A mobilization conception of representation is anticipatory, in Mansbridge's sense. It both seeks to attract potential constituencies and aims, prudentially, to please future voters. Pitkin (Reference Pitkin1967, 164) even sounds like Mansbridge when she writes, “legislators often pattern their actions not on what their constituents ought to want but on what they anticipate their constituents will want (in all their ignorance).” What neither Mansbridge nor Pitkin recognize is that this anticipatory aspect makes the mobilization conception of political representation analogous to aesthetic and literary models of representation that emphasize that representations are performative: representing is an activity that produces ontological effects while seeming merely to follow from an existing state of affairs (cf. Butler Reference Butler, Benhabib, Butler, Cornell and Fraser1995, 134).Footnote ²³

A mobilization conception of representation accords with “poststructuralist” or “post-Marxist” pluralist theories that have emphasized how political identities and demands do not emerge directly from social divisions, but rather that social differences and the politics to which they give rise are influenced by elite discourse (Laclau Reference Laclau1996; Laclau and Mouffe Reference Laclau and Mouffe1985; McClure Reference McClure and Mouffe1992).Footnote ²⁴ The idea is not that political elites create constituencies arbitrarily and in words alone.Footnote ²⁵ Instead, the claim is that there is an inescapably figurative moment in the emergence of a democratic constituency.Footnote ²⁶ Several theorists have proposed ways of characterizing this moment. Laclau (Reference Laclau1996, 98; Reference Laclau2005, 155) writes of the “impurity” in political representation that “does not simply reproduce . . . a fullness preceding it” but is primary “in the constitution of objectivity.” Saward (Reference Saward2006, 300) proposes the concept “representative claim” to underscore that “at the heart of the act of representing is the depicting of a constituency as this or that, as requiring this or that, as having this or that set of interests.” Ankersmit (Reference Ankersmit2002, 115; emphasis added) explains that without representation, “we are without a conception of what political reality—the represented—is like.”Footnote ²⁷ Representing rouses a constituency to action by giving it a picture of itself that enables it to recognize itself in terms of a “generality”—a common enemy, shared problem, shared virtue—that is neither given nor self-evident but must be narrated into being (Rosanvallon Reference Rosanvallon2008, 11).

To link Pitkin to this radical strand of contemporary democratic theory is to put her work at odds with the standard account of (unidirectional) responsiveness with which so many readers have associated it.Footnote ²⁸ Are those readers simply wrong? No. Pitkin retreated from the most challenging implications of what I have termed her “constructivist turn.”

The retreat begins at the end of the book, where Pitkin (Reference Pitkin1967) sums up her unconventional argument in these conventional terms: “representing here means acting in the interest of the represented, in a manner responsive to them” (208). Read out of context of her critique of the principal–agent model and endorsement of representation as a systemic process that is anonymous and impersonal, this could sound like (and was taken to be) a return to the standard dyad. Pitkin quickly counters this, however, writing that, as she understood it, responsiveness is not a “constant activity” but a “condition . . . of potential readiness to respond” (233). This feature, that it “requires only potential responsiveness, access to power rather than its actual exercise,” makes her conception of representation “perfectly compatible with leadership and with action to meet new or emergency situations” but “incompatible . . . with manipulation or coercion of the public” (233).

Here is the language that so bedevils contemporary normative theories of democratic representation. Pitkin recognizes, as does Mansbridge, that as soon as representation becomes anticipatory (rather than merely responsive) it opens up the possibility for political elites to change voters’ preferences. Pitkin, like Mansbridge, is inclined to parse this possibility in terms of what she presents as an opposition between “leadership” and “manipulation.” Although Pitkin (Reference Pitkin1967, 223) acknowledges that the “line” between these two “is a tenuous one, and may be difficult to draw,” she leaves no doubt that normative theory should find ways of doing so. As to just what this might entail, she does not say. Instead, Pitkin finishes by asserting, as if italics could make it so, “there undoubtedly is a difference, and this difference makes leadership compatible with representation while manipulation is not” (233; emphasis original).

Just what underlies Pitkin's italicized conviction? Nothing less that what she terms the “etymological origins” of the word representation, which she reads off its prefix: “re-presentation, a making present again” (Pitkin Reference Pitkin1967, 8). That is, Pitkin invokes etymology to secure the distinction between democratic leadership and authoritarian manipulation. In a memorable passage, she writes: “as the ‘re’ in ‘representation’ seems to suggest, and as I have argued in rejecting the fascist model of representation, the represented must be somehow logically prior; the representative must be responsive to him rather than the other way around” (140). Pitkin imposes here a stricture of unidirectionality that I term the “etymological protocol.” This protocol is Pitkin's return to the bedrock norm. It ensures that there is a “one-way flow of influence from public opinion to policy” (Manza and Cook Reference Manza and Cook2002, 639), establishing, as if it followed inevitably from the prefix to the word, that when responsiveness is functioning properly, it is properly democratic. Thus does Pitkin, by a linguistic sleight of hand, fuse representation to democracy in a way that turns one of her genuinely radical suggestions in reverse: she has moved from demonstrating that democracy is intrinsically representative to asserting that representation is intrinsically democratic (cf. Rehfeld Reference Rehfeld2006).

Scholars who came away from Concept with the common-sense notion that “responsiveness is what representative government is all about” at once got less from Pitkin's text than they might have and took on more than they bargained for. Whereas they missed her assault on the principal–agent model, they internalized the etymological protocol of unidirectionality. This protocol ensured that contemporary empirical findings, which leave no doubt that responsiveness has indeed turned “the other way around,” would seem to betray a fundamental democratic norm. Furthermore, Pitkin left contemporary scholars with the urge to rehabilitate that norm by way of the leadership/manipulation difference. Yet, it is precisely this difference that proves so elusive in the face of current empirical findings about preference context dependency. First, as I argue, citizens learn from communication that recruits them to a side in interparty conflict. Second, there is no “bedrock”—unadulterated preference or enlightened interest—on which to ground a determination as to which of these has occurred. In short, by its fall back on unidirectional responsiveness and its normative orientation to a difference that proves difficult to discern, Pitkin's text left normative theory without a purchase over the empirical findings that have established elite cueing, framing, and other modes of influence as preconditions for democratic competence.

RETHINKING RE-PRESENTATION

I propose to change the terms of this debate by enlisting Derrida's ([Reference Derrida and Allison1967] 1973) Speech and Phenomena, which, incidentally, was published in France in 1967, the same year that Pitkin's book appeared in the United States. This text, which demonstrates that the meaning of signs is not derived from the ideas that they are taken to represent but rather generated in their relations to other signs, casts suspicion on this habit of deriving governing protocols for re-presentation from its etymology. Against the presumption that a re-presentation must follow from something that has already been present (“primordial presentation”), Derrida proposes that it is from repetition that reality acquires the attribute of originality, the quality of seeming to be both logically and temporally prior to its repetition (45, fn. 4). As he puts it, “the presence-of-the-present is derived from repetition and not the reverse” (52). On Derrida's account, representation is not a “reduplication that befalls a simple presence” (53). It is an activity that creates its own reference points and then affects, by “etymological” feint, to have done nothing at all.

I take Derrida's argument to call into question the etymological protocol of unidirectionality. Why assume that the “re” is a temporal “re,” the “re” of return? It might just as well be an iterative “re,” the “re” of repetition. When Pitkin defines representation as “re-presentation, a making present again,” she puts both into play. One could follow the etymological protocol to emphasize making present again, the “re” of return (temporal and recapitulative). Yet, as I show, there is much in her text that violates this protocol and warrants a Derridean reading. Such a reading would put the emphasis on the making present, activating the “re” of repetition (iterative and active) and bringing out what I call its figurative and mobilizing aspects.

Derrida's account has the merits of explaining at once why violations of unidirectionality in responsiveness need not stir anxiety and why they will do so nonetheless. The very word representation perpetrates a ruse by the ambiguity that is built into its prefix. Representing is a making (the iterative “re”) that affects fidelity to something prior (by the “re” of return). I suggest that the dilemma of democratic competence is, in part, an effect of this ruse. For to find politicians framing, cueing, and priming, and to find citizens forming preferences in response to that activity, is merely to find both exercising the practice of representation, understood in the iterative sense. At the same time, it is to find a breach of that practice insofar as the etymology of the word seems to promise not a making present but a making present again.

Understood as making present again, representing will inevitably give rise to a normative urge for fidelity to a popular mandate. If representing is a making present, as the mobilization conception would have it, then that mandate cannot be trusted; the risk will always be that it testifies not to the deliberative competence of the people but to the duplicity of the representative who seduced them into voicing a demand. This does not mean that a mobilization conception of representation cannot be democratically legitimate, only that it cannot be legitimate on the terms that the bedrock norm defines.

In fact, it is no simple matter to bring existing democratic norms to bear on representation as mobilization, which is at odds with the model of interest representation to which leading proponents of deliberative democracy subscribe. On the mobilization conception, representing is “not meant to make a pre-existing entity—i.e., the unity of the state or the people or the nation—visible” (Urbinati Reference Urbinati2006, 24). Nor can it be regulated by the Habermasian distinction between communication and strategy. Representation as mobilization aims to persuade: its modus operandi is rhetorical and anticipatory. Urbinati (46; emphasis original) explains, representatives “prefigure courses of action and project their deliberation in the future, which is, unavoidably, a dimension inhabited by things that have only a hypothetical or fictional nature.”Footnote ²⁹

Representation as mobilization is at odds with deliberation on the Habermasian model, which aims to “legitimiz[e] one shared and authoritative perspective from which to judge all public controversies” (Garsten Reference Garsten2006, 190).Footnote ³⁰ As Bohman (Reference Bohman1988, 187) argues, this model denigrates rhetoric by “asserting the primacy of the literal use of language and often of the argumentative, if not logical, structure of discourse,” requiring deliberators to argue their case in terms with which all reasonable people could agree.Footnote ³¹ In contrast, scholars who are building a model of “rhetorical deliberation” from Aristotle begin from the unabashed avowal that rhetoric aims to influence a particular audience. They argue that the constraints on the practice of rhetoric can only be immanent to its own orientation to success. Rhetors who want to persuade cannot invent frames out of whole cloth. They must engage the sympathies of a specific audience that holds particular value commitments at a particular place and time (Garsten Reference Garsten2006, 190). As Garsten explains, a good rhetorician targets an audience “where they stand” and seeks to bring them “to thoughts or intentions they might not otherwise have adopted” (3, 6). Rhetoric does not succeed without the active participation of the audience, who “change their own beliefs and desires in light of what has been said” (7; emphasis added). Bickford (Reference Bickford1996, 42) adds that Aristotle's rhetoric is an art of attunement “whose function is ‘to see the available means of persuasion in each case’” and thereby to identify what kinds of appeals are likely to set a particular audience thinking. This is the aim of deliberation on the Aristotelian model. Not to produce a justifiable general understanding but to “dra[w] out good judgment” in a time- and place-bound audience (Garsten Reference Garsten2006, 190).

In Aristotle's terms, the finding that citizen preferences respond to such rhetorical techniques as issue framing need not indicate a pathology. It simply confirms that persuasion has occurred. As Lupia and McCubbins (Reference Lupia and McCubbins1998, 40) argue, “in settings where reasoned choice requires learning from others, persuasion is [its] necessary condition.” If citizens are persuaded by a rhetor who respects their art as something that should provoke thinking rather than merely tap prejudice, the resulting preferences can be affirmed as reasoned choices. Yet, this notion of respecting the art brings us to the limit of this Aristotelian position. To establish, as Aristotle does, that there is an art to rhetoric, hence that it is not in itself instrumental, is no guarantee that every rhetor will respect the constraints that are immanent to its practice as an art. In short, whereas these Aristotelian theorists of deliberation do make a place for representation as mobilization in deliberative democracy, they leave open the crucial normative question of how it is to be evaluated.

FROM RESPONSIVENESS TO REFLEXIVITY

To characterize representation as mobilizing is to call attention to its creative effects and, thereby, to alter expectations about what it ought to do. If democratic representation neither takes the social as its ground, nor relies on good reasons alone to recruit supporters, then it is not best assessed by its congruence with group interests because the very notion of congruence assumes that representatives do and should carry forward demands that “belong” to groups in a prefigured social field. In characterizing representation as mobilization, I emphasize how acts of representation work together with political practices to configure the social field and to frame the terms of conflict within which the pertinence and cogency of arguments are judged.

These claims suggest that representation as mobilization is not well suited to be judged by an ideal model of argumentation that forces a distinction between communication and strategy. For the democratic representative, as for the social critic, legitimate political communication can be simultaneously “oriented to understanding and oriented to success” (Bohman Reference Bohman1988, 195). Both the critic and the representative will recruit supporters as they educate them, employing rhetorical practices that aim to effect “changes in beliefs, desires, and attitudes” not by the unforced force of argumentation alone, but by appeal to identity, emotion, and bias (195). If elite communication is inescapably twofold, at once oriented toward constituents and enmeshed in interparty struggle, then does it follow that manipulation is inevitable? Is it no longer possible to tell the difference between a popular mobilization that uses its constituency as a pawn in elite partisan warfare and one that activates incipient concerns to stage a new and potentially transformative conflict?

Even without congruence, responsiveness, and other measures that rely on some version of bedrock, it should be possible to assess whether a democratic political system is more or less representative. But normative theory in the wake of the constructivist turn needs to break with typical assumptions about the relationship between democratic representation and popular sovereignty. It requires what Garsten (Reference Garsten, Shapiro, Stokes, Wood and Kirshner2009, 91) rightly identifies as “counterintuitive” thinking about what political representation is supposed to do. Its purpose, Garsten argues, drawing on liberal thinkers in eighteenth-century France and the United States, is not to respond to popular demands but to “multiply and challenge governmental claims to represent the people” (91). By his emphasis on contestation, I believe that Garsten gets closer to naming the conditions that would actualize Pitkin's “non-objection criterion” than she did herself.

I propose “reflexivity” as the normative standard for evaluating political representation understood in these (nonsovereign) terms. I elaborate this concept not primarily as a quality of individual judgment, but rather as a system capacity. As I use the term “reflexivity,” it is not purely descriptive. It is not a synonym for feedback, the interference that occurs when authoritative predictions and pronouncements become an active force within the systems they are supposed merely to regulate or observe.Footnote ³² Nor is it normative in the strong sense, naming the mode of argumentation that produces warranted decisions, thereby securing the “moral right” of each individual “not to be subjected to certain actions or institutional norms that cannot be adequately justified to them” (Forst Reference Forst2010, 712). It would not make sense to say of the design or outcomes of a representative system, as one might say of a principle of justice, that they are “right” in the sense of justifying the assent of all affected. Finally, reflexivity, as I use it, is not that consciousness of self in time or in relation to others that makes possible a modern sense of history or ethics, one ordered without reference to a transcendental principle (Koselleck [Reference Koselleck and Tribe1979] 2004). It is, instead, the measure according to which a representation process can be judged as more or less democratic insofar as it does more or less to mobilize both express and implicit objections from the represented.

Building on Garsten, I try to imagine what it would mean to extend reflexivity not only to government institutions, but also to the representation process as a whole—“including political parties, political challengers, the media, interest groups, hearings, opinion surveys, and all other processes of communication” (Mansbridge Reference Mansbridge2003, 519). For a representation process to be reflexive, it would have to encourage contestation. First, no official or unofficial body could claim to speak for the people absolutely and definitively, so that dissent would be a norm rather than a betrayal. Second, the represented would enjoy both formal and informal means of communication and action to contest government and party initiatives or, equally important, to protest government and party inaction where initiative should be. Finally, the political communications of advocacy groups, mass media, and opinion shapers would be in competition with one another so as to mitigate passive absorption of elite communications.

For a system as for an individual, however, reflexivity means more than the mere fact of contestation. Official and unofficial representatives must have regular, structured ways of taking objections into account. In the case of official representatives (e.g., government or party), reflexivity would require provision for a formal response that at least registers (if not necessarily incorporates) popular challenges. In the case of unofficial or “self-appointed” representatives, reflexivity is difficult to mandate but could be observed in the representative's response to challenges to his or her (or its) reputation or to a decrease in following or contributor base (Grant and Keohane Reference Grant and Keohane2005; Montanero Reference Montanero2010, ch. 5). As for the mass media and opinion shapers, it would be important to assess the degree to which audiences expose themselves to diverse sources or remain in discrete ideological “silos.”

Although it is beyond the scope of this article to develop it in full, I draw an example of reflexivity from the constitutional design proposed by French revolutionary Condorcet, whom both Rosanvallon (Reference Rosanvallon2008) and Urbinati (Reference Urbinati2006) have rediscovered as a theorist of democratic representation. Condorcet's model of representation is reflexive because it is not “binary,” premising democratic legitimacy on a link between the people and its representatives that is “zippered” by elections, but iterative, involving a dynamic movement between authoritative acts and opinion in process (176). Although Condorcet's model was never adopted and was limited to formal government institutions, it is nonetheless possible to extract principles from its features that could guide a more expansive democratically representative practice today.

The design begins with a system of primary assemblies. These were to be organized at the district level, composed of 450 to 900 members each, run by an elected bureau of 50 members, and federated—first into communes and then into the 85 departments that were represented in the National Assembly (Condorcet Reference Condorcet1793, Title 7, Article 2). The primary assemblies were not sites of direct democracy, but rather entry points into representative government. In them, citizens would select nominees, vote on candidates, and, most important, set in motion the iterative process by which opinion and authoritative acts were made to engage each other (Title 8, Article 1). They would be the means by which, in Pitkin's terms, citizens object to government acts.

The objection process would begin with a motion to repeal a law, submitted by an individual citizen to the bureau of his or her primary assembly in the form of a petition signed by 50 members of his or her district (Title 8, Article 3). As long as the signatures were valid, the bureau would convene the District Assembly to hear the petition. Following a week's deliberation, Assembly members would be required to vote—yes or no—on whether the petition warranted consideration at the communal level. If a majority of its members voted yes, then the primary assemblies across the commune would be required to convene to decide whether to submit it to the consideration of the department as a whole. If majorities in a majority of the assemblies across the commune put it forward, then the departmental administration would have to order a general convocation, submitting the petition to the judgment of all assemblies in its communes.

I draw two principles from these aspects of Condorcet's design. First, reflexive institutions are interlocking so that a ruling by one triggers a review by another. Second, they are structured to incrementally broaden what Schattschneider ([Reference Schattschneider1960] 1975, 3) terms the “scope of conflict.” At the lowest levels of the structure, a commune simply heard what the majority in one district referred to it and decided whether the petition warranted further consideration, whereas a department would both hear and rule on the merits of a proposal that a majority of its communes deemed worthy of its attention (Condorcet Reference Condorcet1793, Title 8, Article 10). In the move from the communal to the department level, then, there would be an expansion of the scope of conflict and an intensification of its stakes. At the departmental level, if voting majorities in a majority of the communes voted for repeal, then that judgment would extend the scope of conflict and raise its stakes once again, triggering reconsideration of the law by the National Assembly (Title 8, Article 13). In short, Condorcet's design would permit voting majorities of the Assemblies in just 1 of the 85 national departments to call the legitimacy of a national law into question.

Urbinati (Reference Urbinati2006, 196) argues that, by its “complex system of time delays” and multiplication of the “sites of debate,” this structure increases the possibility of producing a judgment that “command[s] rational conviction” (Urbinati 2009, 196). This is, indeed, possible. But the more certain and, hence, striking feature of Condorcet's design is that it would steadily and automatically bring new participants into a conflict. It lends momentum to dissent. That Condorcet designed this momentum to build through deliberative public forums, rather than instantaneously (or “virally” by today's vehicle of blogs and 30-second advertising spots), may encourage greater rationality. Reflexivity is, however, a necessary but not a sufficient condition: although it protects against the spontaneous and immediate mobilization of citizen objections, it cannot guarantee that objections that make it through the process will be reasonable.

Finally, Condorcet ensured that no site would be privileged as the locus from which to express the popular will. As Urbinati (Reference Urbinati2006, 183) argues, Condorcet “set up political processes that presumed dissenting interpretations of the meaning of the ‘common opinion,’” creating the means to put any statement of popular will—whether by majority vote, popular initiative, legislative action, or executive order—up for revision or even repeal. This is evident from the next step in the process, the ruling of the National Assembly, which would be required to happen within two weeks of a vote to repeal by the department (Condorcet Reference Condorcet1793, Title 8, Article 13). The National Assembly would have the option to affirm the departmental repeal or oppose it and stand by its law. If it voted against repeal, then that ruling would not be final; on the contrary, it would perpetuate the conflict by occasioning a general convocation of every primary assembly in the 85 departments (Title 8, Article 20). The scope of conflict would expand again because, if the National Assembly stood by its law in the face of one department's challenge, the question of repeal would be referred to the citizenry as a whole.

Remarkably, that a majority joined in the opposition would still not suffice to repeal the law. For Condorcet, even a popular referendum is not synonymous with the voice of the people: it cannot make a substantive claim in the people's name. It can, however, set in motion what amounts to a recall of their representatives. If the general convocation results in a “yes” for repeal, then there would be new national elections in which members who stood by the law would be prohibited from competing. Once reconstituted, the National Assembly would take up the question of repeal again, with its decision subject to the same process of censure, beginning anew with a petition to the district (Condorcet Reference Condorcet1793, Title 8, Articles 22 and 26). Condorcet's constitution makes it possible to call any statement of the popular will into question. Because no site is privileged for its expression, the process of its articulation is always open to amendment.

Condorcet's design exemplifies what reflexivity might look like as a capacity of formal government institutions. His account is limited by that focus and by the fact that it institutionalized deliberation by means of interlocking face-to-face arenas. Today, the principles of his design would have to be extended to the representative process as a whole, including opinion surveys, lobby groups, and other organized forces, and take account of the mediation of political communication by television, the Internet, and political advertising (together with the structure of its financing).Footnote ³³ However, these principles—interlocking assemblies, systematically broadening the scope of conflict, and privileging no site as the locus of the popular will—foster the kind of competition that Chong and Druckman (Reference Chong and Druckman2007) suggest may activate citizens’ judgment. These principles, then, begin to fill in the details of what it would take to satisfy Pitkin's “non-objection criterion” together with its “judgment” clause. Expanding the scope of conflict mobilizes objections; competition among the sites aims to ensure (but cannot guarantee) that objections raised will be well reasoned. Insofar as it is reflexive, a representative process may be democratically legitimate in spite of the fact that it includes and even requires the participation of self-seeking elites.