Any theory of democracy must be capable of explaining sudden change in the behavior of both the political elite and the populace. Taking long-term social, economic or cultural change as their independent variable, theories of democracy in comparative politics do not address the short-run reasons why people begin taking sides in politics. That failure should make it unsurprising that standard theories of democracy also provide no reason to expect democracy, or even "protodemocracy," in Russia, where democracy was never expected to begin. To explain why people take sides in politics, any theory must explain why they neglect the benefits of confining themselves to private pursuits. A theory solving that riddle also explains, of course, why it ever makes sense for politicians to invite them into the contest.

Long-term theories of social change abstract from human individuals by merging them into groups, while cost-benefit theories abstract from the psychology of human individuals by reducing them to nothing more than evaluators of expected utility. A theory that explains why individuals neglect their incentives needs to begin by replacing these abstractions with assumptions that describe real human agency (or, if one prefers, with assumptions at a lower level of abstraction). Construction of such a theory in political science need not be detained by epistemological qualms about whether "reality" is knowable in advance of assumptions. Herbert Simon explains how to avoid this trap:

A theory attributing democracy to discourse takes as its assumptions the conclusions of linguistics and psychology about the impact of linguistic cues on human behavior. Deriving its assumptions from these lower-level disciplines, the theory I propose qualifies as emergent in the sense of Simon’s recommendation. This theory is in no sense a normal theory of discourse. Normal discursive theories attempt to account for properties of discourse. By contrast, I use known properties of discourse in an attempt to account for properties of democracy.

This theory explains either the collective action that defines citizens or the collective abstention from politics that defines subjects as an emergent property of encounters between persons as individuals and cues presented by political discourse. To preview the theory, some citizens engage in politics because the political texts they encounter cause them to identify with the speaker. Identification counteracts the temptation to exploit gains from free-riding. Identifiers therefore take the speaker's side in the political contest. Most subjects withdraw from politics for two reasons. First they do not encounter the kind of texts that prompt citizens to identify and to take sides. Second, discursive cues of the kind that they do encounter under despotism stimulate a minority to identify with the rulers and become agents of repression, whose identification with the rulers overcomes their temptation to exploit gains from free-riding and motivates them to engage in the form of collective action known as repression that prevents anyone from emitting the discursive cues capable of turning other subjects active. Rulers deny freedom of speech because speech itself imperils their rule.

Because the same cues that preclude identification by most subjects stimulate identification by the minority of agents of repression, my discursive theory of democracy problematizes "audience response": how the same cue can produce opposite effects on the behavior of different persons. The theory addresses the problem of audience response by relying on research by linguists, who have never discovered uniformity within or across human beings in the production or interpretation of spoken or written forms at any level. Even phonemes, the smallest unit of speech, differ noticeably from one individual to the next and within the same individual from one occasion to another. While linguists can generalize about the relative frequency of responses to a given textual cue, they always find variation in individual discourse processing that produces exceptions, deviations and outliers among minorities of respondents.

People vary in how they read texts because any text underdetermines its meaning. Not containing any meaning, a text is merely a patterned arrangement of symbols, a set of insufficient cues from which an audience must attempt to reconstruct meaning. Presented inadequate guidance by the text, audience members are necessarily inconsistent in their reconstruction of meaning. Their inconsistency is capable of cumulating over time into widely divergent patterns of response within any given audience. Of course it is mystery why the response by any particular individual falls into a majority or minority category. But for my purposes that mystery requires, fortunately, no solution. It is enough that some individuals can be counted on to diverge from the norm.

Like any theory, this discursive theory of democracy is an abstraction built from assumptions about units of analysis from which I infer hypotheses about a universal association between separately observable dependent and independent variables. The theory tries to explain why agents ever neglect costs associated with their behavior. While relative cost influences all human choices, sometimes agents will accept costs to themselves that they need not bear to receive the same benefit. Besides evaluating costs, the theory assumes, agents also possess identities and perform behaviors. Identities and behaviors interact with each other and with discursive cues. The interaction among identities, behaviors and cues is triangular. An observer can slice into the triangle along any axis that the observer chooses. Agents with identities decide among behaviors. Each agent uses its own behavior, chosen at any moment in time, to infer an updated identity at the next moment in time. Besides behavior, in inferring the updated identity the agent also considers new discursive cues. The updated identity shapes the next round of the agent’s behavior by altering the agent’s evaluation of relative costs even if the costs do not change. The process then repeats itself. Both identities and behaviors occur only if the agent encounters texts whose composition and whose interpretation are shaped by discourses, or procedures for composition and interpretation learned by the agent from past encounters with texts. To the extent that the discourses remain constant, identities and behaviors display stable patterns as well. Because agents engage in further learning about discourses from every new encounter with any text, neither identity nor behavior can remain constant. Discourses manifest continuous, small-scale, cumulative change.

It is this property of cumulative small change that gives discourse potential as an explanation for the suddenness of outbursts of democratic activity. When discourse accumulates sufficient change, an observer can recognize new patterns of identity and behavior predictable from the change in discourse. Developments like democracy in Russia come as surprises because the continuous minor change in the procedures guiding composition and interpretation of texts occurs below the horizon visible to observers, who consequently overlook the imminence of change in construction of identities and in resulting behaviors.

Because the theory I offer in this chapter concerns short-term change in the behavior of multitudinous human agents, it deliberately avoids one methodological move prominent in contemporary political science: the so-called "agent-structure" distinction. "Structure" is a metaphor, and like all metaphors it serves to highlight some objects by hiding others. A structural metaphor like "modernization" serves to highlight the substitution of professional politicians for rulers by hiding—within socioeconomic "structure"—the activities of other agents. "Modernization" simply means that many agents change their behavior, moving into factory labor and out of subsistence farming, acting as financiers and factory owners instead of merchants or moneylenders, and proliferating novel occupations. Even a favorite structural metaphor like "Gross Domestic Product per capita" is simply a number assigned by some agent acting as an economic estimator to valuations by all agents of other agents’ economic activities, divided by the number of agents. Diverting attention to some agents from others, each of these metaphors also transforms into a lasting social fact what may equally be understood as transient individual activities. While structural metaphors may conceivably (although I doubt it) be useful for some theoretical enterprises in the social sciences, a theory for the purpose of explaining sudden changes in the behavior of many individuals cannot afford to introduce structural metaphors that smooth out variations over time in individual behavior or that relegate large numbers of agents to an unexamined background condition.

The unit of analysis in this discursive theory of democracy is the individual. Democracy can, of course, be theorized at the level of analysis of the whole society, as in the case of civil society theory or modernization theory. It can also be theorized at the level of analysis of the class or social group, as Przeworski advises in a speech act that masquerades as a simple, although astonishing, assertion of fact: "Democratic societies are populated not by freely acting individuals but by collective organizations." Theories of democracy at the level of groups or societies instantiate Simon’s point about the reconcilability of bodies of theory at different hierarchical levels. I use discourse to explain how individuals combine into groups. There is no necessary incompatibility between such a theory and various theories of democracy that use group interests to explain how electoral procedures remain in force or that use long-term social change to explain improvement in the prospects for electoral competition to take hold.

Indeed a discursive theory of democracy could supply a lack wanting in the theories at higher levels: a reduction theory to "show how unanalyzed elementary structures"—e.g., Przeworski’s "collective organizations"—can exist at all. For a theory of democracy in Russia, where, as Fish cogently observed, democracy had to be built "from scratch" because the collective organizations did not exist, a reduction theory is particularly essential. In the conclusion to this book I will reconcile a discursive theory with those valid empirical correlations which have lent an illusory credibility to higher-order theories.

Theories taking the group or society as the unit of analysis, rather than the individual, offer to the theorist the advantage of abstracting from the problem of how individuals can combine in collective action. Instead of considering concurrent decisions by millions of individuals deciding separately, the theorist of groups can contemplate an interaction among only a few large organizations, few enough to monitor each other’s choices and each assumed capable of summoning action by its supporters on demand. The theorist of society can ignore the problem of interaction entirely. Neither option is open to an individualistic theory, which must confront the problem squarely. Any theory faces a far more challenging task of aggregation when it starts with choices by millions of individual agents and proceeds to propositions about a whole state and a whole society. Discourse, as the unifying feature of the vast number of cues which not only prompt individuals’ choices of behavior but also enable individuals to infer their identities, offers a principle of aggregation capable of reconciling individual diversity into collective patterns. This kind of theory enables observers to investigate the question of how many millions of human beings can make decisions that converge on either despotic or democratic institutions. The dependent variable is democratic action, while the independent variable is the kind of discourse that individual agents encounter, leading them to construct different kinds of identities that either prompt or impede action. It is to the construction of the independent variable that I now turn.

A discourse builds shared identity that motivates action. When people encounter a text, perhaps they may be able to assign a semantic interpretation to it, but certainly they can decide whether they share an identity with the author. The vast majority of messages in French communicate semantic meaning only to speakers of French and perhaps of closely related languages such as Italian, Spanish or Romanian, but the same message tells anyone whether the author shares the listener's identity as a member of some speech community. The French speaker discovers that the author does; everyone else, even the Italian, Spanish or Romanian speakers or others who might understand the message, discovers that the author does not. Even within a natural language, variation in usage signals difference of social identity. In Labov's famous example, New Yorkers tell each other whether they share a class identity by pronouncing or muting post-vocalic /r/ in phrases like fourth floor. The utterance transmits a signal of social identity regardless of whether its message says "I’m on the fourth floor," "I’m not on the fourth floor," or "What I'm saying has nothing to do with the fourth floor." The discipline of sociolinguistics exists to study how variation within a natural language connects to difference in social identity among its speakers.

The inevitability of signaling social identity transforms texts into powerful determinants of political behavior. A standard finding in social psychology reveals that individuals made aware of their social identity are willing to bear costs to the self if bearing those costs spreads some perceived difference between their own group and some opposed group. Moreover, awareness of social identity also reduces their ability to differentiate among individuals, both in their own group and in opposing groups. Because awareness of social identity produces this pair of consequences, changes in political discourse that alter social identities are capable of transforming political behavior and the associated institutions. Distancing discourse (of which the ultimate form is a foreign language) causes a minority to separate itself from the rest of the population and to discriminate against the majority. Deprived of discussion of politics in their own discourse, the majority becomes passive, often quite sullenly so. The consequence is an undemocratic regime. Ordinary discourse causes everyone to distinguish between political speakers, making choices available that motivate people to sort themselves into political allegiances ("parties") and to bear the individual costs of voting. The consequent contest for power, resolved by inviting more people to join, produces a democratic state.

In short, discourse forms social identities, social identities make people act, distancing discourses form minorities that oppress majorities that fail to resist, ordinary discourse sorts people into contending parties whose competition promotes equality of political rights. How do each of these happen?

The possibility of stimulating social identification by presenting linguistic cues arises from the multiple meanings carried by any linguistic cue. Every text or utterance is potentially a "speech act" that cues audiences to recognize three kinds of meanings: (a) what the text says, (b) what social convention the text performs, and (c) what relationship the text establishes between author and audience. Sentence (1), which appeared in at least two speeches by the Republican candidate Bob Dole during the 1996 US presidential campaign, exemplifies a text in which the three meanings converge:

(1) Give us your votes and your support.

Dole’s text says that he wants his listeners to vote, his utterance represents a conventionally formulated request or invitation to vote, and Republican voters hearing Dole understood him as joining with them in a relationship based on a common effort to occupy the White House.

Not every text need be as convergent as Dole’s among the three kinds of meaning. Consider sentence (2), a revision of Dole’s utterance:

Sentence (2) says what sentence (1) said, and it is grammatically accurate English, but no politician would say it, because it does not meet the criteria for the social conventions of requesting or inviting. Such a stilted sentence would be ineffective in establishing a relationship of identification between Dole and potential voters. Instead it would make voters think Dole strange, odd, inarticulate or clumsy. (A text like sentence [2] is said to be "infelicitous.") Recognition of social conventions and acceptance of the relationship established by a text are known as "uptake" of the text by its audience. The effect of sentence (2) is to preserve what Dole says while disrupting hearers' uptake of the other two kinds of meaning.

As the contrast between sentences (1) and (2) suggests, the completion of a speech act depends not only on the speaker's intention but also on the audience's uptake. Uptake does not depend solely on the form of an utterance. While all three meanings may converge in a text like sentence (1), individual listeners need not converge on a common uptake, especially of the third meaning. Most listeners may have heard Dole's expression of desire for their votes, many listeners may have recognized a request or invitation, but not every listener gave Dole the requested vote. For uptake to occur, listeners must conclude that the speaker has formed the intent to commit the speech act and that the speaker is an appropriate person to do the speech act in question. Many listeners, especially among Democrats, did not consider Dole the appropriate speaker to ask for their votes. Yet even Democrats varied: for a smallish minority, he was. Uptake also depends on attributions by the audience, which vary across individuals within an audience.

Texts less artificial than sentence (2) can easily be imagined that feature divergence among what a text says, its performance of social conventions, and its establishment of relationships. In speeches given by the winner in the 1996 US presidential campaign, explicit invitations to participate comparable to sentence (1) are notable for their absence. The outcome of the election leaves little doubt that citizens "took up" an invitation to vote from Bill Clinton’s speeches or that what he said established the supportive relationship that Dole achieved only among a smaller number of voters. Lacking any explicit request for support or invitation to participate, Clinton’s campaign speeches exemplify texts in which what is said diverges from what convention is performed and what relationship is established.

The divergence between candidate Clinton and candidate Dole reveals a further general trait of speech acts: not everyone is eligible to perform a given act. Having previously been recognized as the candidates of the two major parties, Clinton and Dole were more able than anyone else to complete the performance of the speech act of inviting a vote by establishing the relationship constructed by that act, but neither of them was the appropriate person to perform the act for all or even most American citizens. Even though Clinton won the election, a large majority of potential voters failed to take up the invitation expressed in his speeches. Millions of other persons were in no position to perform for anyone the act that Clinton and Dole performed for their respective voters.

Since J.L. Austin broached the possibility of "doing things with words," the use of texts to perform conventional acts and to establish social relationships has drawn extensive attention from linguists. A standard textbook identifies distinctions by a wide variety of linguists between the "function which language serves in the expression of ‘content’... and that function involved in expressing social relations and personal attitudes."

The ability of texts to carry meanings beyond "what is said" has drawn linguists’ attention to the regularity with which encounters with texts cue inferences about the social identity of speakers and writers. One speech act performed by any utterance achieves the effect of positioning the speaker in society relative to other persons. Every utterance bears what are known as social markers in speech: "A particular pronunciation is also an index of a region, social class, and so on. But note that such an index is already a symbol, since it means ‘New Yorker,’ ‘Texan,’ ‘educated person,’ ‘man,’ ‘woman,’ etc." While these examples concern sound, the same is true of writing. Indeed the very ability to write has been a carefully monitored social marker: when slavery remained constitutional in the United States, state laws forbade teaching slaves to write, although some learned anyway. With the spread of the ability to write, grammar, spelling, vocabulary, or kind of script have often served as social markers.

Sociolinguists often present variation in linguistic behavior as a consequence of differences in social position. Michael Scotti-Rosin takes the explicit position that "language is a social phenomenon and therefore a product of society and lexical events and changes mirror [widerspiegeln] social and ideological conditions." Richer or more educated persons or persons of higher social status speak differently from those of less wealth or education or social standing. However, while varieties of speech do indeed correlate with social variables such as position in hierarchies or occupation, correlation is not causation. Sociolinguists have designated society as the independent variable because their concern is to explain variation in language use: why some people speaking the same language pronounce post-vocalic /r/ in when others mute it, in Labov's example. But it is not at all clear that the stream of influence runs exclusively or even primarily from social difference to speech variation.

Some prominent sociolinguists recognize that the relationship between society and discourse is more complex. One pair of authors presents social differentiation as a potential that discourse can switch on or off:

Penelope Brown and Stephen Levinson go even further. They reject any simple theory that "linguistic variables act as direct signals of the correlated social variable." Instead, speakers and hearers or writers and readers use linguistic markers to establish social identities and social relationships, such as relative status, with different markers being chosen and different identities and relationships developing contingent on who is addressing whom, in what circumstances, for what purposes. Labov himself rejected any direct assignment of linguistic behaviors to social status: "The same sociolinguistic variable is used to signal social and stylistic stratification. It may therefore be difficult to interpret any signal by itself—to distinguish, for example, a casual salesman from a careful pipefitter."

Linguistic variables are held to depend on social variables because the latter are deemed more real. Yet when social variables are examined closely, they turn out to themselves to be, in some cases to a lesser, in other cases to a greater extent, discursive objects talked into existence. While income differences express real relations of scarcity and abundance, they also express possession of money or other symbolic forms of wealth. These represent claims on non-discursive services performed by actual people, but they do so in a discursive form: money is printed or engraved, and it is a symbol of a claim, of a speech act. While income is partly a product of physical scarcity and abundance, social standing and social position are metaphors to physical conditions—erectness, location—not physical conditions in their own right. Two people cannot simultaneously occupy the same location but certainly can simultaneously occupy the same social position. While standing or position often does enable a person to make claims on other people, social standing and social position are talked into existence.

Discourse constructs social difference in part by misrepresenting it. There is an observable discrepancy between social difference and the linguistic forms that signal it. Variables like income, education, social standing and even race that correlate with linguistic variables are continuous, but the linguistic variables that depict these dimensions are dichotomous. People either pronounce or mute post-vocalic /r/, they either use a dialect form or a standard form, at the end of a question their intonation rises or it falls, they do or do not subject final consonantal clusters to simplification. Dichotomous variables invariably misrepresent continuous variation by sorting certain people into opposing groups who are closer to each other on the continuous dimension than to most members of their own group as defined by the dichotomous variable. If the distribution of a population along some distribution such as income were highly bimodal, a dichotomous linguistic variable, with each value confined to persons grouped around one mode, might be a reasonable representation of the distribution, but dichotomous speech markers are found where distributions are unimodal or even flat, and their values are not confined to people in one part of the distribution. While speakers’ propensity to vary the relative frequency of any linguistic marker, rather than its absolute presence or absence, might increase the ability of linguistic markers to represent continuous variation along a social dimension, it is generally not the case that individual occupants of intermediate positions on social scales are individually moderate in the frequency of their usage of linguistic markers. Instead certain individuals occupying intermediate social positions generally use the markers identified with the upper group, while others habitually use the markers identified with the lower group. Speakers also exacerbate their misrepresentation of continuous social variables by collapsing the multiple dimensions of social variation onto a single "bipolar dimension of formal versus informal" linguistic behavior. Social markers in speech are used to categorize others and to maintain awareness of a person's own social identity. When people use social markers to form conceptions of society, any dichotomous representation of continuous variables necessarily misleads them about actual relationships between themselves and others.

A potential instance of self-deception introduced by dichotomous linguistic representation is the very class contradiction between employers and employees so often taken to be fundamental to linguistic variation. Employers as a group do have interests that compete with those of employees as a group, for example concerning wage rates. But equally each employer with that employer’s employees as a group has interests that compete with those of other employers with their respective employees as groups, for example market share and profitability. No economic reason can be given why, or even that, a horizontal commonality of interests among employers versus employees outweighs the vertical commonality of interests linking each employer to that employer’s employees. Yet sociolinguists regularly observe that speech variation tends to divide employers from employees, despite other cases that can be cited in which speech forms link people vertically. Any military organization, for example, develops distinctive markers shared between higher and lower ranks and distinguishing them from civilians, as in the adaptation, depicted in sentence (3), within the US military of the noun "type" to serve as a generalized adjectival suffix:

(3) He’s an outstanding-type individual.

Of course it could be relevant that "soldier" exists in US English as a noun embracing both "officer" and "enlisted man," while no English noun embraces both "employer" and "employee." The dichotomous contrast between the agentive suffix -er and the passive suffix -ee exemplifies how representation can oversimplify a far more complex and puzzling social relationship. The significance of an inclusive usage of "soldier" is also evident in the differences between command relationships in the U.S. Armed Forces and those in the former Soviet military, where enlisted personnel were far more subject to commanders’ arbitrary discretion. Correspondingly, at least in formal usage during the Soviet era, the Russian translation soldat mentioned only the enlisted personnel below the rank of sergeant, ofitser being compulsory when officers below the rank of general were mentioned, and general being compulsory to mention generals.

My argument is not that the contrast between inclusive "soldier" and exclusive soldat causes the difference in the command relationships or that the absence of a comprehensive noun for employer and employee causes the tension between the two groups. Instead the presence or absence of an inclusive term expresses how speakers conceive a social relationship. It is by no means impossible for an English speaker to conceive of employer-employee alliances competing rather than employers competing against employees; I just did, and I am by no means the first to do so. But as the clumsiness of certain sentences in the preceding paragraph indicates, English terminology does make it more difficult to write about a collectivity comprising an employer with employees than about employers or employees as collectivities. The collectivities in question differ correspondingly in how easy they are for the reader to visualize. Because social relationships (in contrast to physical ones) are consequences of how they are conceived, the terms that express the conception spread a culture's understanding of the relationship among the bearers of that culture. While the terms change if the bearers of the culture alter their understanding, once they have arrived at a certain understanding they show some propensity to keep it. In this sense the terminology lends it stability independent of any underlying social causation.

Discourse is further removed from any social rootedness by individuals’ use of discourse to manipulate their social standing. When people identify with a social group, they observably accentuate their identification by increasing the divergence of their linguistic usages from norms prevalent outside the group. US soldiers’ conversion of "type" to an adjectival suffix is an instance of this accentuation. People who seek entry into another group copy its linguistic practices, causing those who want to preserve their group’s distinctiveness to react by further elaborating markers of separateness. Both kinds of people make random errors in reproducing social markers that further proliferate signs of difference. People who seek to establish solidarity across a group boundary emulate each other's speech practices. Experiment has shown that when people of different status engage in friendly conversation in English, the conversation partner with higher status speaks more slowly and increases the frequency of non-standard usages, while the lower-status partner accelerates the rate of speech and increases the frequency of standard forms. It is particularly significant that neither partner knows about this change in linguistic behavior by the self or the other. Instead of knowing their relative social status and deliberately modifying their linguistic behavior to signal mutual convergence, they unknowingly construct a relationship of relative equality by the unattended production of convergent linguistic signs. Because people change their discourse depending on their purposes in an interaction, observed discourse does not depend solely on social position.

Anyone suggesting that cause could flow from social markers to society, instead of only from society to social markers, bears an obligation to offer at least a sketch of how discourse could produce social differentiation. I make no pretense, of course, of proving that this occurs, a task that would demand a far more comprehensive investigation. Still, a general observation by Brown and Colin Fraser is extremely suggestive. They examine highly stratified societies where the language presents a formal and an informal form of the second-person singular (like French vous and tu, although modern French society is not nearly as stratified as the ones they are describing). In these societies, they notice, upper strata tend receive the formal form and lower strata receive the informal form. This pattern might be attributed to the greater independence enjoyed by members of upper strata (by virtue of their possession of additional resources) and to the greater dependence of the lower strata on mutual aid (to compensate for their poverty).

But the reasoning could just as well run the other direction. In languages that possess such forms, formals of the second person singular are ordinarily plurals (e.g., French vous, German Sie, Russian vy) while informals are singulars also used in addressing family members. The general proposition is advanced that people use the speech markers they receive as cues to determining their own identities. If so, a person who most often hears plural forms addressed to the self might infer an identity entitled to a larger share of available resources than someone who ordinarily hears singular forms. Conversely, someone who more often hears forms used within the family addressed to the self might feel more restrained in competition with others—would one be as indifferent to sharing with one’s brother, sister, or child as to sharing with a stranger? It has been shown experimentally that different pronouns produce asymmetric effects on the unattended, automatic attributions by which people ascribe identity to others and to themselves. Moreover, it is an ordinary claim among discourse analysts that variation in pronoun usage alters audiences’ uptake of speech acts. In experimental settings, linguistic cues that divide people into dominant and subordinate groups cause the members of the dominant group to view themselves as more competent and creative, while the members of the subordinate group view themselves as more friendly, likeable and cooperative.

Most important of all, pronouns of address are far from the only difference between the discourse encountered by upper strata and that encountered by lower strata. Pronouns are merely easily studied indicators of a broader range of acts encompassing both many verbal cues and a variety of non-verbal behaviors that present contrasts between the discourses of dominant and subordinate groups. "All over the world," it is reported, people whose forms of address signal linguistic deference to social superiors also display a variety of behaviors marked by "bumbling, by the kinesics, prosodics and language of slow-wittedness or buffoonery." A person who feels constrained to behave so deferentially is seldom an effective competitor for economic or other resources in societies where competition is relatively open. Indeed in such societies both social markers in speech and deferential behaviors are used as an index of occupational potential. This indexing to speech and behavioral markers is observed even where the second personal singular is unmarked for formality or intimacy, for example wherever modern English is spoken. Where people discount occupational potential for deferential speech markers or behaviors, people who display those markers or behaviors will experience economic discrimination.

If speech practices influence both individuals’ eagerness to compete and others’ willingness to let them compete, discourse is capable of constructing the economic differentiation normally seen as the base of social difference. Stratification of speech is known to pre-exist the economic differentiation held to be its cause. When Lincoln Bloomfield studied the Menomini, a group of Native Americans living in Wisconsin in 1925, he investigated a language without dialects or writing spoken by fewer than 2,000 rural residents who were hardly distinguishable in standard of living or occupation. Nevertheless he described three registers: one that was "illiterate, childish, stupid; the second is normal; the third elevated, poetic, archaizing." The relationship need not be one-way from society to discourse.

Of course I do not pretend to have demonstrated which way the relationship flows. That demonstration would require a much larger project than I can undertake here. I do claim to have given reason to suspect that discourse might be independent of and influential on social relationships without the necessity of supposing any underlying social causation. If audiences rely on discursive cues to categorize society into groups, speakers who transform the discursive cues can alter the audiences' composition and definition of social groups. Of course the power of discourse to work social change might ultimately be bounded by social difference (although it should be stressed that this limitedness is presumed rather than known). And it may not be every speaker who can achieve the transformation. Perhaps this capacity belongs only to certain especially influential speakers capable of commanding extremely large audiences already habituated to attributing particular importance to whatever such a speaker says. In any case it is an empirical question whether social differentiation produces discursive variety or whether variation in discourse produces group identity. When we observe that social differentiation flowered in Russia following the collapse of communism, an empirical finding that a discursive transformation preceded its collapse would adduce evidence in favor of the proposition that society depends on discourse as well as, or perhaps rather than, the reverse.

If discourse constructs social identity, whether along or across social cleavages, it provides a powerful motivation to political action. Political action, I argued in chapter 1, requires neglecting gains available from letting others put forth the effort necessary to achieve collective goals. Research on social identity supplies an answer to why individuals neglect the gains of concentrating on private pursuits. A person may be said to recognize shared social identity with another when the person categorizes him- or herself and the other together with "some class of stimuli as the same (identical, similar, equivalent, interchangeable, and so on) in contrast to some other class of stimuli." Speech markers that people recognize as their own offer an instance of such a "class of stimuli"; so do other signs such as earmarks of relative affluence, education or deference from others.

Social identity theory (in all its variants) portrays every human’s self-concept as dual. People know both a personal self contrasted to all others and a social self categorized together with certain others and contrasted to third parties. These selves are mutually incompatible. Which self takes over a person's choices at a given moment depends on the presence of cues to similarity and difference. This view of identity partakes of Deborah Schiffrin’s characterization of both the personal and the social self as "locally situated interpersonal processes." In constructing a self at any moment, each person does so in a real or imagined conversation with other persons. The self, Schiffrin claims, is neither prior to nor consequent upon linguistic interaction, but "arises within conversation." As she concludes, "We are continuously locating and relocating ourselves, defining and redefining ourselves and our worlds…." People continuously update themselves, moving back and forth between a changing personal self and a changeable social self.

Experiments reveal two consequences of moving to the social self. One consequence is perceptual, the other behavioral. First, people cued to recognize an opposing group identity perceive fewer differences between individuals than do people to whom an opposed social identity has not been made salient. This result is known as "depersonalization." Rather than perceiving individuals as varying, people perceive members of their own group as more uniformly bearing favorable traits and members of opposed groups as more uniformly bearing unfavorable traits. This change in perception is known as "ingroup favoritism." Second, people made aware of shared identity voluntarily pay costs if acceptance of those costs widens the perceived difference between members of the person’s own group and members of other groups. In the original experiment, a subject is informed that he or she belongs to some arbitrary group. The subject is then asked to choose between two allocations of money between members of the subject's group and members of another group. One allocation grants equal amounts to members of either group. The other allocation discriminates—at a price! Members of the subject’s own group receive less than they would from the first allocation, but more than members of the other group. Most subjects choose the second allocation, discriminating against the other group at cost to their own group. This experiment has been repeated more than two dozen times with subjects of both sexes from a variety of cultures and across the distribution of various ages.

The original experiment prevented subjects from choosing their own individual rewards. That decision would have drawn the subjects’ attention to the individual self when the purpose of the experiment was to test the consequences of a salient social self. Experiments on "social dilemmas" have been used to demonstrate individuals’ acceptance of personal costs. In a social dilemma, gains to the individual conflict with gains to the group, as they do in decisions to vote, demonstrate or repress. It is of course impossible in the laboratory to replicate the context of large elections in which an individual makes a decision jointly with tens of millions of others. Yet experiments can replicate the trade off between personal and group goals, and in such experiments, subjects are more likely to sacrifice personal gains when social identity is made salient. Such experiments are interpreted as indicating that "that the general process underlying mutually cooperative intentions and expectations is the extent to which players come to see themselves as a collective or joint unit, to feel a sense of ‘we-ness,’ of being together in the same situation facing the same problems."

George Quattrone and Amos Tversky conducted experiments that directly, although inadvertently, tested the impact of social identity on willingness to accept costs to the self, one of them in the context of voting. While Quattrone and Tversky designed their experiment on voting to investigate the sources of "voter’s illusion" (the self-deception common among voters that their individual votes affect the outcomes of large elections), their experimental design inadvertently manipulated social identity instead. Subjects were told that they were supporters of an arbitrarily designated political party and that the party was facing an election that it might either win or lose by a margin of several hundred thousand votes, too many for the individual subject’s own vote to change the outcome. There were two experimental conditions. In one condition subjects were told that turnout by voters unlike themselves ("non-aligned voters") would decide the outcome of the election; in the other condition they were told that the election would be decided by the turnout among party supporters like themselves. They were also told that voting was costly. The subjects who had received the second set of instructions, that they were members of the decisive group, were more likely to express readiness to pay the individual cost of voting, even though they had been instructed that they were individually indecisive. In a related experiment Quattrone and Tversky also showed that subjects were willing to tolerate individual cost in the form of increased physical pain (prolonging voluntary immersion of their own arms in icy water) when they had been told that this action signified membership in a group with the positive characteristic of prolonged life expectancy.

In the capacity of social identity to motivate willingness to bear costs to the personal self, even physical pain, lies the connection between discursive cues and political action. Discourse can cue social identity that motivates actions costly to the personal self. In these experiments, variation in average willingness to bear costs depends solely on what subjects have been told (or in some cases, shown). It is enough to inform a subject of the sharing of social identity for both the perceptual and the behavioral effect to appear. The information need not be true and, in fact, is ordinarily a deception by the experimenter. Nor need the information be explicit or deliberate. Quattrone and Tversky, in particular, did not intend to inform people of different social identities. The information on contrasting group membership was an inadvertent consequence of their clumsy experimental design. Social identities and even their other human bearers need not exist in any reality outside the cue (linguistic or other) encountered by the subject.

Judging by the record of experimental research on social identity, if political discourse is observed that informs people, deliberately or not, whether they share social identity with the speaker, those persons who respond by sharing the speaker’s identity will display at least some willingness to ignore their individual costs of action. The probability that costs will be ignored rises as the magnitude of the costs falls.

Political identity differs systematically in despotism and democracy. Despotism divides people between a positive and a negative identity that motivates collective action by bearers of the former and directs bearers of the latter to private pursuits. Undemocratic rule bifurcates society along a horizontal cleavage separating the positive political identity of the rulers and their enforcers above this cleavage from the negative identity borne by everyone else below the cleavage. Democracies replace the horizontal cleavage with one or more vertical cleavages that divide people among two or more positive identities (metaphorically, located to the right or the left of each cleavage) capable of motivating their bearers to pay the cost of forgoing private pursuits in favor of collective action. The vertical cleavages separate political candidates with their voters from each other, but these vertical cleavages are much more permeable than the horizontal cleavage of despotism.

Identity plays an often overlooked part in Joseph Schumpeter's definition of democracy as "that institutional arrangement for arriving at political decisions in which individuals acquire the power to decide by means of competitive struggle for the people’s vote." Although Schumpeter's definition of democracy is sometimes criticized as minimalist for neglecting features said to make democracy desirable for its citizens (such as guarantees of their rights or of their influence on government), his definition does distinguish democracy from those forms of government where persons gain "the power to decide" by some means other than a popular vote. Schumpeter's definition attributes the choice of leaders to a political identity called "the people." In any undemocratic government, leaders are still chosen, but those eligible to make the choice compose some minority much smaller than the whole people.

Who belongs to this minority varies across types of undemocratic regimes. In monarchies nobles choose, in bureaucratic authoritarianism officials choose, in Confucian states literati choose, in military dictatorships soldiers choose, in communism cadres named through the system of nomenklatura choose. In "exclusive republics" (exemplified by the early United States), where voting is often the institutional mechanism by which rulers are chosen, those eligible form a larger minority or may even be a majority, but still some limitation of the franchise restricts those with the power to choose -- to males only, to persons bearing certain ethnic or racial markers only, to persons possessed of a certain amount of wealth or income only, or to other limited groups.

How they choose also varies. While voting sometimes takes place, the minority exercising the right to vote is more likely to decide by acclamation. More important than the vote in expressing support for a ruler is willing obedience. When nobles, officials, literati, soldiers, or nomenklatura cadres choose, those who choose the rulers are also the enforcers of rule on the population as a whole. If they become willing to enforce a ruler’s decisions, that ruler becomes empowered. If not, he or she departs the ruling group, usually voluntarily, sometimes by expulsion. One of the rulers’ decisions that they enforce is repression, the policy of excluding the population from exercising any political rights.

The political identity that unifies agents of repression is defined by a privative opposition, i.e., a pair of opposites in which one of the pair is given positive definition and the other is undefined except by the absence of the quality defining the first member. This political identity is recognizable because it bears a label. Noble, official, literati, soldier, nomenklatura are all labels attached to a minority of enforcers of rule. While the population as a whole sometimes bears a contrasting label, this label never denotes agency.

In every case the quality that defines the collectivity of enforcers is the same as the principle purported, often on remarkably incoherent grounds, to entitle the rulers to power. Nobilities are defined by (pretended) descent where kings rule by right of birth. Even though everyone is born, only nobles claim "birth." Where a dynastic ruler is not only sultan but also caliph, leader of Islam ("submission"), he rules by exemplifying submissiveness to Allah, and his enforcers may therefore be mamluk or yeni cheri (Mamelukes or Janissaries), "Slaves" whose title expresses the acme of submissiveness even as their rebellions replace sultans. Officials are distinguished by promotion received from bureaucratic superiors, whose own series of promotions entitles them to seats on the ruling council. Soldiers earn military rank justified by claims to especial competence and patriotism, qualities by which the highest-ranking soldiers affirm themselves qualified to rescue the nation in its putative peril. Confucian literati pass tests of their competence to reproduce sacred texts. This competence enables them to achieve correct performance of li, the divine ritual credited with sustaining the social order by enacting paternal authority and obligating filial obedience. As the "son of Heaven" owes filial obedience only to Heaven, he alone is obligated to obey no earthly father, and therefore he is the residual claimant of obedience from all others. Submitting to him and joining him in the performance of li, the literati thereby entitle themselves, in their own understanding, to compel obedience from everyone else by the use of fa, the set of legal decrees backed by corporal punishments, often extremely painful, for failure to comply.

The Soviet Union evidently was a case of this generalization that the quality defining enforcers of rule also establishes, at least in their minds, their right to rule. Philip Roeder’s classic institutional analysis describes an unwritten "constitution of Bolshevism" that insulated selection of leaders (a "directorate" called the Politburo) from influence by the Soviet population. The Politburo chose heads of large bureaucracies, who in turn acted as a "selectorate," called the Central Committee and ultimately numbering about three hundred persons, with the power to elect and remove members of the Politburo. In their bureaucratic capacities these senior cadres supervised the activities of subordinate officials, who were in turn chosen for their offices through the institution of nomenklatura, the list of official posts and of persons eligible to occupy them, by a subset of cadres reporting through a separate hierarchy (the party committees’ "sections for organizational-party work") ultimately to the Politburo. Subordinate cadres constituting the nomenklatura acted as the agents of repression, using the state’s near monopoly of employment and its police powers to compel members of the general populace to refrain from politics. The cadres qualified for office by carrying out tasks set by party committees (whether the superordinate bodies called "committees" or the subordinate "primary party organizations") to which cadres became increasingly likely to be assigned as they became eligible for more powerful posts.

In short, in a communist despotism, nomenklatura cadres voluntarily obeyed a General Secretary chosen by an election in which the vote was cast only by certain senior cadres, known collectively as the Central Committee, because fulfillment of committee service was the principle that distinguished between the minority who bore the identity nomenklatura from the vast majority who lacked committee assignments. The privative opposition was expressed in terminology that overlapped rather than coincided with the division between committee cadres and the population as a whole. All Soviet citizens were divided between "communists" with party cards and "non-party persons" (bespartiinye), those lacking the essential trait for official assignment. In Russian the denial of social identity to the populace was explicit. According to an officially approved 1985 publication intended to train the political awareness of the populace, the adjective partiinii denoted "expressing the interests, attitudes, aspirations of any social group." If so, to be bespartiinyi, "without party," was to fail to express interests, attitudes or aspirations of any social group. A characteristically more pithy statement of the same privative opposition was offered by the founder of the nomenklatura Stalin: "Either you are nothing in the eyes of the party or you are a party member with full rights."

In contrast to the privative opposition ubiquitous in undemocratic rule that contrasts a positive versus a negative identity, democracies feature oppositions between two or more positive identities. In the United States, for example, nearly all voters, even those who describe themselves as "independents," repeatedly cast their ballots for either Democratic or Republican candidates. People may occasionally defect to a candidate of the opposite party or to a third party, but over the long term their voting manifests a prevalent sympathy with one or the other of the major parties. While in the United States Democrat may therefore be the residual of Republican, and vice versa, neither is defined as the negation of the other. Instead Democrat and Republican bear contrasting positive definitions, as shown by the positive form of both parties' names. The names are even meaningful: Democrats in the contemporary United States are more associated with the idea of government by the people understood as the less privileged, while Republicans are more associated with idea of government as pursuit of a common good understood as not distinguishing between more and less privileged individuals. When democracy replaces despotism, identities shift from the privative opposition between the defined and the undefined to the positive opposition among various definitions that are not residuals of one another.

While despotism and democracy are alike in featuring contraposed political identities, they differ not only in the polarity of identities but also in the intensity of identification. When a privative opposition gives way to an opposition of positive self-definitions, the new opposition is weaker and more labile. In a despotic regime the rulers may well distinguish between identities within the group of enforcers, but individuals or families are mobile across the boundaries of those identities. While in the Frankish kingdom the enforcers bore identities of abbots and bishops as well as magnates, the abbots and bishops were ordinarily drawn from magnate families. Acceptance of a monk's tonsure was the normal, if humiliating, means of escape from execution or blinding as a penalty for defeat in the power struggle. Under Soviet communism nomenklatura cadres moved freely back and forth between assignments as Communist Party administrators and as state bureaucrats. The term nomenklatura applied to the selection procedures for both kinds of post. While circulation within the multiple identities of the nomenklatura might be frequent, however, movement across the privative opposition separating nomenklatura from the population normally occurred only once, with acceptance into the nomenklatura. With rare exceptions no one ever crossed this barrier more than twice, first when admitted to and second if expelled from the nomenklatura. Only in a few cases during World War II did anyone expelled from the nomenklatura ever recover membership.

By contrast the partisan identities of a democracy are far more labile. While some partisans would never even consider voting for a candidate from some other party, many people may occasionally desert their parties’ candidate only to return to the fold in later elections. People sometimes even change party identification over the course of their lifetime, although it is more common for people to move from strong party identification into calling themselves "Independents" (often not very convincingly) rather than moving from one party into another. The privative opposition definitive of despotic rule is relatively fixed, the identities of democracy much less so.

A distinction between democracy and despotism in terms of identity restores to the theoretical foreground the puzzle perceived by David Hume:

Hume’s surprise at the ability of the few to prescribe the conduct of the many has been virtually lost in contemporary political science. In part Hume's puzzlement has been lost because of an unspoken conventional assumption that, in contradiction to Hume, despotism can be distinguished from democracy by dependence on force. When Lipset, for example, limits by a prepositional phrase the scope of his assertion that "Political stability in democratic systems cannot rely on force," he implies that stability in despotisms can rest on force. Even Schumpeter, normally so clear sighted, distinguished between democracy and "those cases in which all competition with the established leader is prevented by force." Yet force, while explaining why the subjects obey the ruler, cannot, as Hume pointed out, explain how the ruler secures compliance from the agents of repression. Who is there to enforce rule on the enforcers of rule? Any explanation of the stability of despotism in terms of its ability to use force is trapped in an infinite regress.

Hume’s puzzlement became lost among political scientists because his question, like so many others, appeared to have received a convincing answer from Karl Marx. When Marx analyzed Louis Bonaparte’s coup d’etat of 1851, he posed in specific terms precisely the same question that Hume had posed in general: "It remains to be explained how a nation of thirty-six millions can be surprised and delivered unresisting into captivity by three chevaliers d’industrie." As a solution to the puzzle of how the many ("thirty-six millions") could be ruled by the few ("three," although Marx really thought it was one), Marx proffered his concept of class interest. He assumed that people in general (as opposed to individuals) must act so as to preserve their "material interests," by which he meant the means of livelihood peculiar to their class (note that this is not maximization of individual income, and therefore avoids the paradox of collective action).

In France in 1851, Marx argued, there were four classes, each with its own particular means of making a living. The bourgeoisie lived by hiring others to work on its property in large landholdings, factories or banks. The urban petty bourgeoisie lived by working on its own property in small shops. The rural petty bourgeoisie, or peasantry, lived by working on its own property in small farms. The proletariat lived by selling its labor to the bourgeoisie. In addition there were many persons without fixed means of livelihood, who collectively composed the lumpenproletariat.

The four classes and the lumpenproletariat submitted to Louis Bonaparte’s dictatorship for different reasons. The three propertied classes submitted because the dictator upheld their interest in property against the threat of the proletariat by employing the lumpenproletariat, ultimately paid from the funds of the bourgeoisie, to serve in the state administration (especially the army) where its members repressed first the proletariat and then the urban petty bourgeosie. Trying to explain why a people remains passive ("unresisting") before a dictator, Marx attributed the passivity of the proletariat to military defeat and the passivity of the rest to actions of the dictator that tended to preserve their "material interest." In Marx’s story, what Hume called "opinion" played no role. Indeed, the opinions of the various actors Marx dismissed as what later came to be called "false consciousness":

Although Marx's solution to the problem that puzzled both himself and Hume has commanded very widespread assent among political scientists, the collapse of the Soviet Union shows that Marx’s answer cannot offer a general solution to how minorities rule. By Marx’s categories, the Soviet population consisted of one uniform lumpenproletariat. (Russians in the early 1990s, having been educated in Marx’s terminology, sometimes referred to themselves, with a tinge of sad irony, as liumpeny). Until 1986 nearly all legally employed persons in the Soviet Union, with only rare exceptions, received their wages from the state. Collective farmers lived by sale of their own product, but mainly to the state, and then only in excess of the requisition demanded by the state as ground rent. There was little difference between the means of livelihood of a collective farmer and employment as a "worker" on a state farm. In other words, almost no one in the Soviet Union belonged to a class with a means of livelihood, i.e., a material interest, distinguishable from the state.

If a lumpenproletariat can be considered a class and people generally act in politics to preserve a means of livelihood peculiar to their class, as Marx asserted, then Russians should have acted to preserve the Soviet state, not demonstrated for its overthrow. If a lumpenproletariat does not qualify as a social class by Marx’s definition, because its members have no established means of livelihood and therefore must depend on the state for their livelihood, then Russians still should have upheld the Soviet order, not overthrown it.

While it would be wrong to credit Soviet communist officials with having been consistent theoretical Marxists, many officials did expect the dependence of the Soviet people on the state for their livelihood to resign them to continued passivity in the face of dictatorship. In an interview in October 1990, the Leningrad party official who had conducted the city committee’s disastrous electoral campaign in the previous year expressed bewilderment to me about the behavior of employees in the city’s defense industry. He had warned them that by supporting the democrats they would ultimately abolish their own jobs, but they had rejected his advice. The Soviet state’s programmatic reliance, which Fish has highlighted, on "control and repression in the workplace" to maintain the passivity of the state employees indicates that this Leningrad official’s expectations were anything but exceptional. It is Soviet citizens’ renunciation of passivity that empirically refutes Marx’s imputation of "reality" to material interest, understood as the means of livelihood peculiar to a class, at the expense of identity as a determinant of people’s behavior. With the collapse of organized Marxism the time has come to discard theoretical Marxism. Of course a rejection of material interest as explaining political behavior does not mean that what people or political parties say about themselves can taken uncritically as true, but it does mean that their discourse can be taken as material to analysis of their behavior.

If material interest does not relegate the many to passive toleration of rule by the few, then Hume’s puzzle remains unsolved. While Hume used the word "opinion" to describe what enables the few to rule the many, his concept of opinion has the same elements as the concept of identity that is central to the theory I present here. Hume perceived, just as I have argued, that "the most despotic and most military governments, as well as the most free and most popular" operate according to "the same maxim." Hume used a privative opposition to define despotic or military governments, juxtaposing persons "led... like men, by their opinion" against "harmless subjects, [driven] like brute beasts, against their sentiments and inclination...." This opposition is privative because the reader knows that these "harmless subjects" really are human beings, yet Hume’s description of them as "brute beasts" negates a human identity. If "men" is a positive description of human beings and "brute beast" a negative one, then Hume is saying that despotism is typified by the imposition of a negative identity on everyone except the agents of repression epitomized by "Mamelukes and praetorian bands." In contrast to the "harmless subjects," the agents of repression bear a name.

Conceived in Hume’s terms, despotism and democracy are ends of a continuum rather than a dichotomy, and ends of a single continuum of inclusion rather than of a mapping onto Dahl’s two dimensions of contestation and inclusion. There is evidence that this continuum, rather than electoral method, is how Schumpeter understood democracy, despite his statements about procedures and method. He comments, for example, that the reduction of democracy to electoral method "excludes methods which should not be excluded, for instance, the acquisition of political leadership by the people’s tacit acceptance of it or by election quasi per inspirationem. The latter differs from election by voting only by a technicality." He also states explicitly: "there is a continuous range of variation within which the democratic method of government shades off into the autocratic one by imperceptible steps."

When despotism and democracy are understood as ends of a continuum rather than a dichotomy, democratic transition is reconceived as movement from despotism toward, not to, democracy. Movement from despotism toward democracy occurs whenever positive political identity is extended to encompass a larger proportion of some population and, conversely, the proportion denied political identity decreases. This means, of course, that undemocratic rule can exist, and movement toward democracy can occur, within electoral institutions. Indeed, movement toward democracy has occurred in the United States, and other contemporary electoral states, with every expansion of the suffrage. If the institution of suffrage is an expression of positive political identity, initial establishment of a suffrage and further expansion of the suffrage are both manifestations of a single common process, the wider extension of political identity. If electoral institutions are compatible with privative oppositions between a positive identity for agents of repression defined by the right to vote and a negative identity imposed on those ineligible to vote, moreover, electoral institutions can be no bar to reversals of the process of extending positive political identities. In such a theory, democratic reversals are to be expected.

Despotism is maintained by repression. Being a form of collective action, repression can occur only if enforcers of undemocratic rule are willing to neglect their gains from shirking. Repression can be motivated if undemocratic rulers supply the enforcers of their rule with positive political identity by addressing the population in a distancing discourse of rule. By separating the rulers from the population, a distancing discourse of rule opens a conceptual gap available for occupancy by agents of repression who select themselves from the population as a whole. An interesting question, although not one requiring an answer here, concerns why a given individual responds to distancing discourse by identifying with the speaker. It might be plausible to suggest that some people feel separate from most people around them and respond to a distancing discourse by perceiving similarity between themselves and speakers whose discourse presents them as also separate. In any case, conceiving themselves as more like the rulers than like other members of the population, the agents of repression adopt a collective label, establishing a separate social identity that supplies a motivation for its individual bearers capable of partly offsetting the continuing lure of private gain. Where the motivation is strong enough, it disciplines both the ultimate rulers and their agents (Hume's "Mamelukes and praetorian bands"), empowering them to establishing undemocratic states. Suppressing other forms of political discourse, the agents preclude the formation of other political identities. As an emergent consequence, the privative opposition typifying undemocratic rule appears as a distinction between the positive identity borne by the rulers with their enforcers and an absence of any political identity for the population as a whole.

A distancing discourse may be a foreign language, the ultimate signal to a hearer that a speaker does not share the hearer’s social identity. Historically (as I will show in chapter 5) undemocratic rulers have often deliberately adopted foreign languages as their form of address to their subjects. Other undemocratic institutions, particularly colonialism, have sprung into existence because conquerors bring with them a language foreign to their new subjects. Yet even within a natural language shared between rulers and ruled, the rulers can find ways of distancing themselves. The rulers can exploit the universal feature of language that linguists call iconicity, the expression of meaning by resemblance of form.

Distancing discourse can be built from two kinds of icons: diagrams and metaphors. A diagram is an icon in which meaning is communicated by some property of syntactic arrangement that matches the meaning to be expressed. To communicate distance from the hearer, speakers across cultures elongate the utterance. A speaker expressing distance may protract any unit of the utterance: the whole, its component clauses, lexical items, phonemes, pauses. Non-literate Amerind and Austronesian cultures distance speakers from hearers in ritual (including political ceremony) by doubling semantic elements, producing elongated clauses. Certain Australian cultures divide people into moieties whose members must avoid approaching, gazing at, or addressing anyone from the opposing moiety. Speech in the presence of taboo persons symbolizes the speaker's deliberate inattention to them by replacing the words used between interlocutors from the same moiety with synonyms. The avoidant synonyms are generally longer than the words they replace. The same is true of the secret jargons of rhyming slang that sometimes define low-status groups (e.g., Cockneys) and of the professional registers that define occupational specialties. In symbolizing interpersonal distance by the length of words, these Austronesian cultures display a striking similarity with the culture of English speakers. One trait of distancing in English is choice among synonymous triplets of English, French or Latin origin, e.g., ask-question-interrogate. This example is typical of the tendency for word length to increase as the relationship features more disparity of authority. English aristocrats drawled, prolonging vowels. German chancellery officials doubled or even tripled certain letters when writing. Javanese courtiers deliberately protracted pauses that in ordinary Javanese conversation would be interpreted as signaling the interlocutor’s turn to speak. As John Haiman comments, "Gobbledygook is an icon, pure and simple, of non-linguistic formal behavior." Elongation evokes a feeling of distance from a speaker, Talmy Givon argues, by requiring hearers to exert additional cognitive effort in reconstructing the semantic meaning of a text (given, of course, that the same semantic meaning is expressed in the shorter and the longer version). The same extra cognitive effort would be required to reconstruct the meaning of verbal cues attenuated by the weakening of sound waves over physical distance.

To express interpersonal distance, speakers also use metaphor. Where a diagram communicates meaning by resemblance of form, a metaphor communicates meaning by positing an unspoken resemblance between a target domain presupposed to be unfamiliar and a source domain presupposed to be familiar. Eve Sweetser observes that metaphorical communication "allows people to understand one thing as another, without thinking the two things are objectively the same." To call discourse distancing is itself an instance of metaphor. As the familiar experience of physical distance represents the unfamiliar abstraction of difference, so interpersonal distance is used to conceive membership in different groups. Besides physical distance, people also use the vertical orientation of their bodies to conceive the difference between themselves and those who exercise power over them. As Sweetser notes,

Russian, also being Indo-European, predictably forms the identical metaphor nadziratel', combining nad- ("over") plus zirat' ("look," "see") plus -el' ("-er"), although Russian nadziratel' is used more narrowly than its English counterpart, and translation, supervisor. To call someone a "superior" performs the speech act of enabling that person to exercise authority; whether the speech act succeeds depends, as always, on uptake by the audience.

Diagrams motivate metaphors, while metaphors motivate diagrams. Linguistic diagrams exploit the temporal sequence of words. In English and Russian, temporal sequences are understood by the experience of objects in motion, which in human experience is ordinarily horizontal. Consequently, a diagram of interpersonal distance formed by elongating the temporal sequence of words constructs a horizontal diagram in the mind of listener. In English, a horizontal diagram of authority corresponds to the metaphor leader, someone who horizontally precedes others, and in Russian to the metaphor nachal'nik, "boss," which literally means "one at the beginning." Vertical metaphors of authority exploit what Lakoff and Johnson call "complex coherences across metaphors." Up is a metaphor not only for authority but also for quantity. Metaphorically those who are up in authority are also up in quantity – larger, weightier. In everyone's childhood experience – that is, their language-learning experience – larger and heavier people are also more powerful. Moreover, "more is better": the rulers are not only a more powerful group but also a preferred group, an "aristocracy" of rule by the best, or as the Soviet communists used to describe themselves, luchshie iz luchshikh, "the best of the best." In this phrase the metaphor of superiority motivates a reduplicative diagram intensifying the claim.

Separated from the population by distancing diagrams and metaphors into a distinct social identity bearing a positive label, the enforcers of rule experience both the perceptual and behavioral consequences of a salient social identity. The perceptual consequence is the indistinctness of members of the population in the minds of the enforcers. Members of the population become a mass. In Soviet discourse, individuality lacked expression. Rather than presenting people as individuals, official discourse construed the Soviet people as a set of kollektivy, collectives, much as its predecessor under tsarist rule had assembled individual peasants into village communes (obshchiny) with joint liability for taxes.

The behavioral consequence is motivation to discriminate against the population even at cost to the enforcers of rule. Repression imposes two different ways of life, one materially abundant in proportion to standing within the identity of rulers and enforcers, the other more uniform in its relative privation for the remainder of the population. The Soviet Union's prized egalitarianism was a leveling down into poverty for the population that coexisted with privilege for the nomenklatura. The extraction of material resources from the population by and for the enforcers that characterizes every despotism cannot be the purpose of despotism, since such a purpose would require collective action to be motivated by contemplation of individual costs and benefits. Instead, extraction of material resources from the population must be the consequence of despotism’s dependence on definition of identity by privative opposition. This effect of confining positive identity to the enforcers of despotism is of course also a material condition for despotism’s survival, but at the same time, to the extent that poverty imposed on a population reduces its members’ incentive for economic effort, agents of repression may even decrease the aggregate wealth from which they draw their own incomes relative to what they would receive under more efficient economic institutions. In accepting a reduction of their incomes for the sake of enlarging the observable difference between their incomes and the incomes of the population, the agents of repression conform precisely to the expectation of social identity theory that persons informed of a social identity will accept costs to themselves to enlarge the perceptual difference – in the Soviet case, visible, tangible, even gustatory – between themselves and members of the population under their rule.

This expectation of social identity theory helps to explain how nomenklatura officials could award themselves superior living conditions while stubbornly maintaining that they themselves were selfless. In 1990 several officials of the Leningrad university communist organization, intending to ask me to find US research funds that would augment their declining incomes from the university, invited me to visit the dacha originally built for the famed Leningrad party chief Sergei Kirov and allegedly used, after Kirov’s assassination, by Stalin himself. We drove into a small park surrounding a modest single-story home with five or six spacious rooms plus an attic that they said Stalin had used as an office when in residence. See, they said, no luxury here! This remark expressed their sensitivity to the avalanche of public criticism that had recently descended on privileges enjoyed by the nomenklatura. And indeed Kirov's dacha was modest – except by comparison to the dachas with three miniscule rooms crowded together on tiny plots of land for use even by relatively privileged Leningraders.

The peculiar feature of undemocratic regimes is the absence of popular opposition. Although this absence is often attributed to policies that buy quiescence by affording relative material security or to successful repression that intimidates people eager to throw off the dictator’s yoke, that eagerness often seems to be missing. Scholars visiting the Soviet Union were struck by the readiness of Russians (as well as persons bearing other ethnic markers) to accept and even to endorse Soviet rule. The effects of a distancing discourse on identity explain why subjects of a tyranny remain unresisting for so long.

Distancing discourse affects the perceptions and behavior not only of the enforcers but also of the populace. When a privative opposition between the identities of rulers and ruled is made salient, the enforcers do not distinguish among individuals in the population; conversely, made aware of the rulers’ separate identity, members of the population fail to distinguish among the rulers as individuals. Perceiving the rulers as identical, the population believes that the political contest offers no choices. In fact, since rulers are perceived not to differ, members of the population are likely to believe that no political contest is occurring. Even if the rulers were to issue invitations for the populace to take sides in their internecine contest, a distancing discourse would disrupt uptake by precluding the text from corresponding in form to the social conventions, embedded in ordinary discourse, for expressing an invitation. Seldom drawn from the ranks of trained linguists, rulers may not know that distancing discourse safeguards them from the danger that one of their number will appeal for popular intervention in the struggle, but their rule persists where they adopt discourse of this kind.

Perceiving themselves as lacking political choices, members of the population see no reason to intervene in a political contest that they are inclined to think does not happen. Denied a positive political identity by the privative opposition that emerges as a result of distancing discourse, members of the population also lack any motivation to neglect the personal costs of political action. This absence of motivation is unaffected by increases in the amount of information available to the population. Quite often the upsurge in popular activity during the Gorbachev years is attributed to the increased information that his policy of glasnost', usually (mis)translated "openness," made available to the public. Increased information persuaded Russians and others that the social problems facing the Soviet Union were more severe than they had realized. Yet increased appreciation for the severity of social problems cannot explain a change in popular political activism, since the opportunity to free-ride remains. Awareness that the magnitude of collective problems has increased merely enhances the gain that can be obtained, while the individual continues to receive the same enlarged gain regardless of whether he or she chooses to act or to free-ride, and action remains more costly than free-riding. While undemocratic regimes invariably engage in censorship, it is a mistake to regard censorship solely from the population's point of view, seeing it as a device to conceal the magnitude of social problems. While censorship does result in hiding the truth from the population, its cause is instead what social identity researchers call "ingroup favoritism," motivated by the distinctive social identity that the enforcers of rule construct in response to distancing cues emitted by the rulers. Censorship is the enforcers' device to avoid encounters with any messages that conflict with their unduly favorable depiction to themselves of themselves and their accomplishments.

Reports of experimental research on social identity, which has concentrated on the effects of making positive social identity salient, do not contain information about people denied cues to social identity. But sociolinguists have researched this question. W.E. Lambert developed "matched guise" tests to evaluate attitudes towards speakers of English and French in the bilingual community of Quebec province. In a matched guise experiment, subjects hear tape recordings of speakers talking in different languages or in different registers of a single language. The subjects are asked to evaluate the speakers’ personal qualities (e.g., intelligence, friendliness, sincerity). The subjects do not know, however, that each person whom they evaluate is heard twice, once in each language or register. Lambert’s original experiments in Quebec showed, and others have since confirmed in other communities, that speakers of the politically dominant language or register evaluate others more favorably when the others are also speaking the dominant language. This result confirms the ingroup favoritism found by experiments on social identity. Surprisingly, Lambert and his successors also find that speakers of the subordinate language or register also evaluate others more favorably when hearing them speak the dominant form. People cued by politics that their own identity is negative fail to discriminate against bearers of the positive identity.

Since Lambert’s original study, moreover, the Parti Quebecois has actively campaigned in favor of making French dominant in Quebec, ultimately succeeding in passing a law requiring French education. The result has been a shift in linguistic behavior: while native-French bilinguals formerly accommodated to native-English bilinguals by speaking English, increasingly native-English bilinguals are accommodating native-French bilinguals by speaking French. This behavior is of course costly to both sides. Because each is better at passive recognition of the other language than at communicating in it, the optimum for communication between French and English bilinguals would be for the native-French bilinguals to speak French and hear English while the native-English bilinguals hear French and speak English. Instead the dominant group discriminates, with persons on both sides paying costs in the form of impaired communication but the side that pays the greater cost depending on the political cues that both kinds of people encounter.

Commenting on bilingual interactions within Great Britain, the linguistic historian Dick Leith summarizes the consequences of linguistic subordination:

If the speaker’s self-evaluation is one of powerlessness and inability to express his or her feelings, lack of resistance against exploitation is to be expected.

There is also a second way in which bifurcation of identities in undemocratic polities deprives the populace of motivation to neglect its gains from free-riding. Because bifurcation of identity motivates the enforcers of rule to impose material privations on the population, it results in leveling the differences in income, education and social standing around which political identities form that motivate action in electoral societies. Centralization of the economy in the hands of the Soviet state enabled its rulers to impose a relative equality of earnings that produced a characteristic "flattening" of society outside the communist hierarchy, where individual rewards were differentiated in strict proportion to rank. This flattening of society was not deliberately designed to perpetuate the exclusion of the populace from politics but contributed nevertheless to that objective of the regime by limiting social differentiation around which political identities might otherwise develop.

While most people therefore remain unresisting, their steadfast pursuit of private goals is unsatisfying to them. One variant of social identity theory ascribes to human beings underlying needs for both attachment to groups and differentiation from others. An alternative version sees social identity as the product of needs for self-esteem. Yet a third version sees social identity as a consequence of the organization of the brain that produces an innate tendency to categorize. It might be added that human cognition finds positives easier to process than negatives. It is therefore cognitively less demanding for people to represent the self by a positive than by a negative identity. The denial of positive political identity by despotic regimes fails to satisfy the motivational needs while allowing no scope for the cognitive propensities. Invariably, therefore, in an undemocratic regime at least some persons will attempt to fill the void by voicing a positive political identity. Possessing no other linguistic means by which to express this positive political identity, such persons voice it in the language ordinary to themselves. If left unhindered, expressions of positive political identity in ordinary language would elicit recognition of shared political identity by the population, stimulating people to begin taking the side of such persons against the rulers. Because the natural propensity of human beings to seek positive self-definition consequently poses a danger to the exclusion of the populace from politics essential to undemocratic rule, for any despotism to survive the agents of repression must subdue any such expression of positive popular identity.

It is for this reason that wherever despotic rulers are found, they invariably deny freedom of speech, freedom of assemblies where expressions of positive popular identities might be heard, freedom of the press where they might be read, and freedom of organization by which they might be formulated and disseminated. Contrary to Dahl’s assertion, such freedoms are not requisites for democracy. Instead, they are privative opposites of restrictions essential to the maintenance of despotism. It is by the coupling of positive identity for the agents of repression with denial of identity to the remainder of the population that tiny minorities can maintain rule in despotisms. Force need be used not against all the population but only against those few persons bold enough to espouse a positive identity for the populace. If rebellious individuals are suppressed before repeated encounters with their texts over time can build a discourse capable of representing a shared positive identity with the population, then the rest will remain quiescent. Repression can be selective, and therefore efficient.

It is a paradox that among the populations of dictatorships and monarchies are to be found the rational egocentric individuals of formal theories in political science, who abstain from collective action in favor of devoting themselves solely to private gain. Of course this fact has been noticed before, under such labels as "amoral familism" or "atomization," but under these labels it appears as a product of undemocratic rule that counteracts a supposed natural tendency to associate, when in reality neither the self-seeking behavior of subjects nor the collective action of citizens is natural. Both are products of discourse.

While distancing discourse constructs some minority as repressive, ordinary discourse promotes majority voting. When politicians address citizens in the citizens' own language, they supply cues to a shared identity linking the politicians to the citizens as members of the same speech community. Political texts in ordinary language induce both the perceptual and the behavioral consequences of sharing social identity. Not alerted to any opposing identity, the citizens become more sensitive to differences between individuals. Perceiving politicians as individuals, the citizens automatically rank the various politicians as more or less similar to themselves. Categorizing some politician as most like themselves, the citizens begin to identify with that politician in opposition to those categorized as unlike themselves. The resulting identification motivates the citizens to neglect their individual gains from free-riding and instead intervene in the political contest on the side of the politician with whom they identify. This intervention may take the form of protests or demonstrations but ultimately produces a demand for voting as the least costly way of taking sides. The pattern of social identity shifts from the privative opposition characteristic of despotism to an opposition in which all opposing groups bear positive labels and membership is widespread.

Elections reproduce on a grand scale the conditions presented by a social identity experiment conducted on a small scale. The parallels enable social identity theory to resolve the apparent paradox of voting as a behavior that costs the voter without achieving any offsetting gain. The experiment on social identity begins by presenting subjects with cues to their identity. An election begins with a campaign that reminds potential voters of their political identity. The experiment invites people to make a choice. An election presents potential voters with a choice. In the experiment the subject must weigh the costs of various choices. In an election, voting is costly to the voter in time and attention. In the experiment, when a subject is informed that he or she shares social identity with others, the subject becomes more perceptive of individual differences. In an election, conduct of the campaign in ordinary discourse informs voters that they share an overarching identity with all candidates that encourages them to perceive individual differences among candidates. Those citizens who recognize a larger difference between individual candidates are more likely to vote. In an experiment the subject becomes willing to bear costs to the self if the cost enlarges the perceived difference between the subject’s own group and an opposing group. Crucially, the groups need not exist in any objective reality. The perception within the experimental subject's own mind provides the subject with the unconscious motivation behind a deliberate decision to pay the cost. In an election the individual’s vote is highly unlikely to influence the objective reality of the electoral result, and quite often voters are confident that the election will be a landslide in which no individual vote makes the difference. Without affecting objective reality, the casting of a vote does widen the perceptual contrast between the voter’s own group and other groups by attaching to each voter’s own candidate—in the voter’s own mind at the moment of voting, not later in the electoral outcome—the positive act of an endorsement and to the opposing candidate the negative act of a rejection. If social identity theory has merit, accomplishment of change in each voter.s own perceptual world motivates each voter to overcome the disincentive to voting. Alexander Schuessler has formulated this connection between identity and the act of voting: "the voter [in the United States] votes Democratic not to effect a Democratic electoral victory, but to be a Democrat."

While the enormous empirical literature on voting has not previously attributed (not even Schuessler) the behavior of voting to social identity constructed by discourse, the theory of voting offered here preserves the data and has the additional advantage of unifying the bewildering variety of causes that are the bane of the empirical literature. Grounding his admirable study of Russian voting in the 1995 and 1996 elections, the distinguished Russianist Timothy Colton quotes approvingly a conclusion from a study of Canadian voting that "virtually every imaginable element in the calculus of voting… mattered in the final choice." A study of US voting that Colton takes as a model to be emulated identifies no fewer than "eight clusters of variables" assembled into "six stages or blocs."

Voting looks so complicated only because this literature has from the very beginning posed a misleading question. Empirical researchers studying voting have begun by assuming that the object of their inquiry consists of "two fundamental decisions: should I vote, and (if so) for whom?" Of course it is possible that some voters actually make those two decisions in that sequence, although the finding that voters who call themselves "independents" in the United States are in fact highly partisan in their voting decisions casts some doubt on how many voters ever need to decide for whom they will vote.

Representing voters as fundamentally neutral and therefore needing to decide how to vote, the empirical students of voting rely on a metaphor that misdirects their attention. Instead of describing voting as taking sides in the political contest, they construe it as taking part, and they label the object of their study "democratic participation." Now of course this metaphor makes no difference in the truth value of the description. It is logically impossible to take sides without taking part, and in reference to voting (at least where voting is limited to expressing a preference for one candidate for each elective position), it is also logically impossible to take part without taking sides. Yet like all metaphors, taking part and taking sides highlight and hide, and they highlight different things while correspondingly hiding the alternative. Taking part highlights the difference between voters and non-voters, while taking sides highlights the difference between the voters on one side and the combination of non-voters with voters for other sides. Guided by the metaphor of taking part, the literature on democratic participation attempts to understand who votes as opposed to who does not. The investigators’ posing of the question conflicts with the actors’ self-understanding. Taking part is neutral. Investigators conceive Democrats and Republicans as doing the same act in contrast to non-voters. But voting is anything but neutral. Democrats and Republicans understand their own casting of votes as doing opposite acts, and Democratic and Republican voters each understand themselves as more similar to non-voters than to each other. Voters understand themselves to be taking sides.

If voting were conceived as taking sides rather than taking part, the study of voting would presume that each potential voter makes a single decision, usually without needing to think about it: do I have a side? Every eligible voter who does, votes for that side; if not, the eligible voter abstains. This rule can account for all the bewildering variety of differences between voters and non-voters that the empirical literature has accumulated. All these differences fall into two simple categories: (1) reasons for voters to perceive themselves as similar to candidates and (2) greater likelihood of exposure to political discourse that presents reasons to perceive similarity to candidates.

There are two main branches of the empirical literature (each sprouting many twigs), one that ascribes voting primarily to social and psychological causes, the other primarily to institutional context. According to one branch of the empirical literature, people become more likely to vote when endowed with more resources – affluence, education, social standing – or when characterized by more psychological engagement with politics. Although party identification provides the strongest single explanation of who votes, a main line of argument in this literature presents affiliation with politicized groups (not only with parties, but with racial minorities, pressure groups, churches, and others) as only modifying the effects of the putatively more basic social and psychological conditions for voting. This branch of the literature provides no coherent argument why resource endowment or psychological engagement should motivate voting and other forms of political participation. In their most recent exposition, its proponents comment only that political activity is too much trouble for anyone who lacks at least a "modicum" of resources and of engagement with politics. While the statement is plausible enough, it is curiously at variance with its authors’ own empirical finding that those who take sides in politics possess not a modicum but an abundance of resources and psychological engagement.

If the dependent variable "who votes" were replaced by the dependent variable "which side a voter takes," the effects of social and psychological determinants, putatively fundamental, would emerge as greatly attenuated. If the average Democratic voter in the United States is more educated or more affluent or possessed of higher social standing than the average of Republican voters combined with non-voters, the difference is very slight. While the difference is greater when Republican voters are contrasted to Democrats combined with non-voters, it is still much smaller than when all voters are contrasted to non-voters.

All these weaker effects are understandable as forms of identification developing through discursive interpretations. As discourse analysts occasionally remark, observers and agents are interchangeable. Observers try to understand their own actions, while agents condition their actions on their own observations. Predictors of voting for the scientific observer are markers of shared social identity for the agent. When scientific observers attribute voting to resource endowment or psychological engagement, they begin by constructing scales of income, education, social standing, and interest in politics. They find that those who rank "high" on these scales are more likely to vote. Of course, persons who rank "high" on these scales also rank themselves "high" on their own versions of the same scales. High income, higher education, and high social standing are expressions in ordinary language that some people would use to describe themselves, as well as terms in the discourse of scholarly observers describing those same persons. If Lakoff and Johnson are correct to argue that people judge similarity by evaluating "complex coherence across metaphors," then people's self-positioning on social scales shapes their choices about whether to engage in politics. People form judgements of similarity, Lakoff and Johnson argue, by matching concepts that are expressed by the same metaphor, even when they know the target concepts to be dissimilar. Even in the absence of any reason other than the shared metaphor, complex coherence across metaphors would lead occupants of high positions on the social and psychological scales to detect similarity with political candidates, who typically also occupy high positions on the same scales as those most likely to vote. (For example, the 2000 US presidential contest featured a pair of millionaires, a Yale graduate and a Harvard graduate, sons of former holders of national political office, both of whom are psychologically highly engaged with politics, and in 2004 both psychologically engaged millionaires had graduated from Yale.) Moreover candidates compete for high office. In the United States, occupants of high positions on scales of income, education, and social standing have often attained their positions by competing. If metaphorical coherence guides their reasoning, competing for political office would seem to them a behavior of the candidate like their own past competitive behaviors. High office would seem to them a goal of the candidate like their own goals of high income, high education or high social standing. At the same time, all these traits differentiate the candidate from those lower on any of these scales who are more likely not to vote.

An alternative approach in the empirical literature emphasizes instead the institutional context of voting. Across countries turnout has been higher where the dividing line between parties coincides with social cleavages such as class, religion or ethnicity. "Catchall" parties depress turnout. In the United States, where citizens must register before they can vote, turnout is higher in states where the deadline for registration is closer to election day. In addition, membership in social networks has been found to encourage both voting and other forms of democratic participation.

A theory that attributes voting to identification with the candidate achieved through discursive interpretation explains these effects of institutional contexts as consequences of variation in exposure to political discourse. The consequences of exposure may be most obvious in the case of late registration deadlines. By adding to the cost of voting, registration requirements demand stronger identification on the part of the potential voter. Identification develops through exposure to discursive cues from politicians. Where people can register later, they have longer experience of a political campaign that has exposed them to more discursive cues. Because identification cumulates with exposure, people who have received more exposure should be more likely to have built identification strong enough to move them across the threshold of voting. Exposure can also account for the effects of party identities bounded by social or economic or religious cleavages. Sociolinguists have gone to great lengths to demonstrate that members of every social group index themselves by a discourse peculiar to the members of that group, a so-called "sociolect." Where political campaigners confine their addressees to the members of a particular class, religion, or ethnicity, they can use that group's sociolect, with the result that their hearers are exposed to a familiar discourse. While familiarity of the discourse presents a similarity capable of developing identification independently of the message, it also influences the likelihood that the message will be remembered and trusted. More likely to be remembered and trusted, a politician’s message is more likely to be repeated by members of the audience. Increased repetition by some members increases exposure among the other members of the politician’s audience. Finally, incorporation into social networks increases the likelihood that people converse with others, and people who converse more in general are more likely to expose themselves to conversations about politics and political candidates.

In sum, the findings of empirical research on democratic participation are readily understandable as consequences of (a) exposure to political discourse and (b) discursive processes of political identification. Electoral politicians who want people to take sides in politics should compose texts that citizens will interpret as signifying the sharing of social identity with potential voters. In other words the speech of democracy should be the ordinary, everyday language of the people. It is the very ordinariness of political language in democracy that has kept researchers from attending to the discursive roots of democratic participation. Being ordinary, political language in democracy is not noticeable, and being unnoticeable, it has tended to go unnoticed.

Of course the cues that generate identification and political action are not exclusively linguistic. Much campaigning relies on visual cues: pictures of a candidate, symbols associated with a party such as the drawing of an apple used by the Russian party Iabloko ("Apple"), cartoons and, since the popularity of television, short films. In every case, however, these visual cues are accompanied by linguistic texts on which the potential voter relies to interpret them. Another non-linguistic cue is the candidate’s physical appearance. Lawrence Bobo and Franklin Gilliam have shown for African-Americans that the physical appearance of candidates serves as a cue to social identification, finding that their trust in government, sense of political efficacy, and knowledge about politics all increase when someone of the same race occupies electoral office. While their demonstration of the impact on participatory attitudes of the extra-linguistic cue of racial identity applies only to one racial minority, their work has obvious implications for the racial majority’s higher propensity to participate in a country where elected officials preponderantly belong to the racial majority. Since many African-Americans speak a distinctive English, it is possible that race is conflated in this study with unobserved linguistic cues.

Also, within ordinary language, cues of different kinds can increase or decrease willingness to take sides. Stephen Ansolabehere and Shanto Iyengar have shown how negative advertising decreases readiness to participate among U.S. citizens. This effect presumably results from the dependence of voting on the cuing of a positive political identity. By causing the potential voter to dissociate from the negative identity attributed to an opposing candidate without providing the voter with a positive referent, exposure to negative advertising deprives the voter of the cue that stimulates identification and voting. In accord with this interpretation of their data, Ansolabehere and Iyengar found that negative advertising depressed voting primarily among those voters with the fewest other earmarks (e.g., partisanship, education, income, interest in politics) of someone likely to have reasons for sharing positive identity with a candidate.

"Ordinariness" in language, of course, begs the question of what linguistic traits will induce uptake as "ordinary" by a population. These traits can be specified by introducing Halliday’s concept of register, which he defined as a complex of linguistic markers tied to a particular situation of use. Any natural language consists of multiple registers. The ubiquity of register is visible in the examples which Halliday had in mind: technical language, thieves’ cant, various scholarly registers, casual registers, sporting registers, and so on. Registers vary in vocabulary, syntax and phonetics. Among variations in situation that change usage is the degree of formality of the communication. Sentences (4a-c) represent utterances from registers in American English decreasing in order of formality.

(4a) It will not rain.

(4b) It isn’t going to rain.

(4c) Ain’t gonna rain.

Formality is in turn a product of the transformation of language, learned from and optimized for face-to-face communication in dyads, to serve the purpose of communication with groups. Because political communications address large groups whose members are mainly accessible through print and broadcast media, the public language of electoral politicians will closely resemble the formal written standard for any population with whom they seek to identify. Because the populations whom they address master the formal standard imperfectly, electoral politicians will pitch their texts just below the formal standard, generally complying with the standard while incorporating a few elements of casual speech.

Identification cued by ordinary discourse explains not only the behavior of voting but also the institution of voting. When ordinary discourse cues potential voters to differentiate among candidates and to identify with the candidate judged as more similar to the self, many of the resulting identifications are weak. When all the candidates campaign in ordinary language, to many voters no candidate looks all that different from the self. While people in democracies often feel passionately about politics, their feelings are much less intense than those of people living under oppression. When was the last time a Democratic voter took a sledgehammer to a building used as a Republican campaign office, or a Republican firebombed a Democratic office? Because the contrasts that form political identification in democracies are vague and the resulting identifications are weak, politicians can expect few citizens to engage in political action if its costs are high. The politicians therefore introduce the institution of voting as a device for reducing the cost of action to the citizens. Compared to other institutions by which a whole citizenry might choose national political leaders, voting demands remarkably little time, energy or attention. For example, it would be just as democratic to convene all the citizens in the capital every four years for a debate that would last until they reached consensus on a new slate of officials, much as the Church convenes the cardinals in Rome to choose a new Pope or the Communist Party used to convene Central Committee members in Moscow to choose a new General Secretary. But such a procedure is obviously too costly if the whole citizenry (as opposed to a few high officials) are to take sides, and it offers no certainty of outcome. Because voting is a cheap way to exercise influence, demonstrators ready to give their time and even their lives in protest against oppression are instructed by their leaders to demand the introduction of voting as a means to reduce the cost to the demonstrators and thereby to increase the number of people whose identification is strong enough to motivate paying the reduced costs. If rulers anticipate that protesters will demand elections, they can preempt demonstrations by offering elections in advance. Moves toward democracy can be achieved by pacts or protests, and the path may conceivably make a sharp difference in the future of a people, but democratization is the same process either way.

In undemocratic polities only some subset of the population is eligible to take sides in the contest to rule. Undemocratic choice of rulers bifurcates a population along a horizontal cleavage separating a positive political identity borne by rulers with the enforcers of their rule from a negative identity ascribed to everyone else. In the Soviet Union the positive identity was the nomenklatura, the negative identity the bespartiinyi, "non-party person." Positive political identity motivates the enforcers of rule to impose institutions that discriminate against the subjects even at the cost of reducing the enforcers' own welfare relative to more efficient economic institutions. Instances of such institutional choices are decisions in traditional polities to preserve subsistence agriculture or in communist regimes to establish a command economy. Negative political identity deprives everyone else of any comparable motivation to take sides in the political contest. The populace becomes unlikely to perceive any contest among rulers whose distinctive collective identity makes them seem indistinguishable as individuals, and members of the populace lack any motivation to sacrifice gains from private pursuits in favor of collective resistance against the rulers' discriminatory practices.

Democracies open the possibility of taking sides in the contest by sharing political identity between politicians and citizens. Sharing of political identity prompts each citizen to distinguish among politicians as individuals, to identify with the politician deemed most similar to the self, and to pay the private costs of voting because in the citizen's own perceptions, not in the later electoral outcome, the vote discriminates against opposing candidates in favor of the citizen’s own candidate. In democracies politicians agree among themselves to resolve their contest by holding competitive elections because voting reduces the cost of political action below the threshold where even relatively weak political identification still suffices to motivate people to take sides.

Like all social identities, political identities are inferred from encounters with discourse. Rulers build a discursive barrier between themselves and the population under their rule. A distancing discourse that establishes the barrier cues most people to recognize the rulers as separate from their own speech community but cues a minority to identify with the rulers on the basis of their similar alienation from the majority. For democracy to appear, someone must rupture this barrier. Either the rulers must fail to repress some outsider who begins to address the population in its own ordinary discourse, or one or more of the rulers must begin to dismantle the discursive barrier.

This theory implies two hypotheses testable by observing any polity. First, if the polity is undemocratic, its political discourse must be distancing; if it is democratic, it must be ordinary. Second, where democracy replaces undemocratic rule, political discourse must begin to change before democracy appears.