I am grateful to Jane Mansbridge for her clarifying and generous response to my previous work. (Mansbridge Reference Mansbridge2011; Rehfeld Reference Rehfeld2009). I agree with her that political representation is complex and relational. I also agree that the traditional concepts that scholars have used are no longer adequate to the task. As she indicates in four of her five critiques, our disagreements are not so much about the nature of political representation per se as about the concepts we use to study it.Footnote ¹ These disagreements center on three issues: (i) the standards we use to guide concept design; (ii) how concepts can best model the relational nature of the forms of political representation; and (iii) whether we must build substantive views of representation into the concepts we use to study it. On each of these questions, I argue that a more parsimonious approach to concept formation would yield more useful results.

The conceptual disagreements between us speak to a larger issue that Henry Bertram Mayo prefigured 50 years ago when he suggested that the term “representation” had become so complex and shifting as to cease to be useful (Mayo Reference Mayo1960). I believe that he was correct and that the study of representation, particularly among normative scholars, has continued to suffer from the attempt, as Hanna Pitkin put it, to show that the term “does have an identifiable meaning” and that it is a “single, highly complex concept” (Pitkin Reference Pitkin1967, 8). Rather than formulating these debates in terms of one concept of representation, let alone the concept of representation, as Pitkin put it, I believe it would be more useful to develop concepts of representation to study the broad array of phenomena that we often imprecisely classify as “representation.” These concepts would usefully explore what a representative is and what activity we think is properly denoted by “representing,” and separately explain what it means for one thing, or activity, to be “representative” of another. These concepts would further be developed by reference to a range of normative ideals of authority, accountability, consent, interests, responsiveness, recognition, sovereignty, and policy correspondence, to name just a few. There is simply no reason to presume that these very different ideas must share some common covering theme, or that the creation of a concept to do so would be of much use to normative and empirical scholarship, rather than serving merely to obfuscate and confuse. Indeed, the attempt to discern or create a single covering concept of representation has lead to some of the deepest confusions surrounding this topic since Pitkin's seminal work was first published over four decades ago.

Despite the titles of our original articles, and even Mansbridge's claim to be capturing the complexity of the concept itself, Mansbridge and I are not really attempting to “rethink representation.” We are, rather, trying to understand the nature of one particular aspect of representation in a highly constrained arena: the relationship between those represented and their representatives in the activity of democratic law making, and democratic decision making more generally. This is why I think Mansbridge's work in this area is so potentially fruitful and illuminating: She is clarifying critical components of this relationship, even though I resist her own moves to combine them into more complex relational forms. One of the insights that came from my original analysis (Rehfeld Reference Rehfeld2009) is that the “trustee”/“delegate” distinction emerged from questions primarily about decision making rather than principal–agent questions that were thought to be at the heart of political representation itself. This was surprising, and although Mansbridge does not mention this view in her response, it was to me the most important contribution of the article.

In what follows I will take up a defense of conceptual clarity and disambiguation by responding to the particular substantive disagreements that appear in Mansbridge's response. In the second section, I argue against the view that empirical regularity should guide concept formation; in the third, I argue against building relational complexity into the concepts we use; in the fourth, I explain why including any substantive view of representation in “surrogate representation” minimizes its conceptual utility and also illustrate the disutility of including any notion of hierarchy in Mansbridge's otherwise helpful “selection model.” I conclude by indicating the importance of developing concepts that stretch beyond the democratic contexts that feature prominently in her response.

WHY CONCEPTS SHOULD NOT BE (OVERLY) CONSTRAINED BY EMPIRICS

In my original article (Rehfeld Reference Rehfeld2009), I argued that the historic debate about the proper relationship between representatives and their constituents collapsed three kinds of decisions that representatives were making into the binary trustee/delegate framework. These three descriptive features of their decision-making process when they voted on laws were as follows:

1. (i) Representatives’ source of judgment: Were they self-reliant, or did they depend upon their constituents’ views about how to vote?

2. (ii) The aims of legislation: Were they promoting the good of all or the good of a part?

3. (iii) Representatives’ own responsiveness to sanction: Were they more or less responsive to the prospect of re-election or other sanction?

With the alternatives framed in this way, a representative who acted as a trustee was usually described as (i) relying on his own judgment (ii) to promote the good of all (iii) in a manner that was relatively nonresponsive to electoral sanction. In contrast, a representative who acted as a delegate was usually described as (i) relying on his constituents’ judgment (ii) to promote their more narrow good (iii) in a manner that was extremely responsive to electoral (or other) sanction. Once we separate out these three conceptual dyads we will develop eight underlying ideal types; when we force them into the trustee/delegate dyad alone we ignore six of these possibilities (Rehfeld Reference Rehfeld2009, Table 2).

Mansbridge objects to this typology because empirically two of its conceptually distinct features—responsiveness and source of judgment—tend to go hand in hand. Representatives who are more responsive to sanction tend to rely more on their constituents’ judgment. And representatives who are less responsive to sanction tend to rely more on their own judgment instead. Because they are correlated empirically, Mansbridge believes it would be more useful to simply combine them in her concepts: A “gyroscopic representative,” as she has described it, thus refer to a representative who is both nonresponsive to sanction and self-reliant as well. And because two of the cells in my original table describe unobserved phenomena there is little use in developing them: the nonresponsive representative who follows on her constituents' judgment and the responsive representative who nevertheless follows on her own judgment.

The first disagreement between Mansbridge's view and my own thus centers on our divergent views of how best to develop concepts for the study of social phenomena. As this example illustrates, Mansbridge claims that a concept is less useful if we cannot find many empirical instances (Mansbridge Reference Mansbridge2011, 629) of its logically independent features, and that our purpose is to choose concepts “for their empirical utility in particular contexts” (Mansbridge Reference Mansbridge2011, 621). By these standards, concepts should reflect fidelity to the world as it is, has been, or is likely to be, rather than the ways the world might be. As Mansbridge rightly argues, we do not usually see representatives who simultaneously are “nonresponsive to sanction” and follow their constituents’ judgment rather than their own. Nor do we often see its corollary, representatives who are “responsive to sanction” but who nevertheless follow their own judgment. Even though she concedes that my three distinctions are analytically precise and possibly useful in other contexts, she rejects their usefulness in studying democratic legislatures because, in those contexts, they produce empirically empty combinations.

By contrast, I do not think that concepts should be limited by their current or past utility in describing the empirical world, but rather by our theories about what might matter normatively or causally.Footnote ² Because causal and normative theories often require us to imagine counterfactuals or mere possibilities as ideals or ideal types, we should develop concepts to cover entities beyond what we observe or are likely to observe. Limiting concepts to the empirically likely would also unjustifiably reify existing normative relationships and practices. As a way of illustration—though importantly not what Mansbridge had in mind for reasons I will explain in a moment—consider that if in 1850 we had to fashion a conceptual map of a “representative” in a way that reflected “many empirical instances” [or one in which we were required to “produce . . . examples . . . .of this sort” (Mansbridge Reference Mansbridge2011, 624)] we would have rejected the concept of the “female legislator,” or worse, built “male” into the very concept of “legislator.” Indeed, it is quite likely this is how people did conceive of “legislator” at that time. Yet it would have been a mistake to build “male” into the concept despite its fidelity to the observed cases of “legislator,” because it would have then been difficult to ask whether or to what extent gender does or should affect a legislator's decisions, deliberations, etc.

The fact that concepts should not be limited by the way things are does not tell us how we should develop them. Nor does it explain why the distinctions I drew were useful in any way. Indeed, in my original article I merely asserted without any demonstration that the distinctions were “enough to provide a useful conceptual space . . . [providing] a rubric by which to assess the rest” (Rehfeld Reference Rehfeld2009, 225). Mansbridge is right to object, then, not because the distinctions I drew do not conform to empirical regularities but because they simply appear to be pointless. So let me turn to defend this view of concept formation, and then demonstrate the value of these particular distinctions in thinking about political representation.

Concepts express the underlying idea that variables seek to measure; they form the connective tissue between the variables we use and what we think is causally, descriptively, or normatively important. Consider, for example, that Americans who approve of the U.S. President are often found to think that the economy is doing well. If we wanted to explore this frequently observed correlation it would be unhelpful to have a concept and variable like “presconomy” whose meaning was “approve of the president and approve of the economy,” because we want to test what the relationship between those two components is.Footnote ³ At still other times we will want to separate and then ignore other distinctions because we think they are irrelevant to the thing we want to study, again, for either normative or empirical reasons. To use an example from Plato close to my heart (or more precisely my head), we could develop a concept of “ruler” that collapsed a person's “fitness to rule” with “baldness,” but such a distinction would not be useful, because we have no good reason to think that these two features are causally related.

Turning to the particular examples under discussion, we first note, then, the empirical usefulness of keeping responsiveness and source of judgment as independent variables if we wish to test their causal relationship to one another. Mansbridge's own claim that “Nonresponsiveness to sanction and self-reliant judgment occur together in practice because they are causally linked” (Mansbridge Reference Mansbridge2011, 624) is impossible to make, let alone test, unless one separates responsiveness from source of judgment and lets each vary independent of the other. Does nonresponsiveness to sanction cause someone to rely on his or her own judgment, or the other way around? Or is there a prior cause for both (such as “arrogance and selfassuredness”)? It thus must be possible for the variables to take on either of the values (or range between more or less of each, because they are continuous) for us to systematically study this relationship. This is why I still think it would be more useful for “gyroscopic representative” to denote only a representative's nonresponsiveness to sanction, allowing his or her source of judgment to vary independently. Indeed, it is only by separating these two variables, and thus these two concepts, that we can pursue the following question: Why does the activity of representation in democratic legislatures so often exclude some combinations of these components that are not null sets in other contexts? That returns us to the causal relationship between judgment and responsiveness that Mansbridge stipulates without argument, but that needs to be explored, and would require distinguishing judgment from responsiveness for analysis.

Let us turn, then, directly to the two seldom seen cases that Mansbridge raises in her objection and ask whether it is useful to combine them into separate ideal types. In the first, we find a representative who is responsive to sanction but still relies on his or her own judgment. Here I had put “Madisonian lawmaker” as one illustration (along with aiming at republican ends). Mansbridge was right that I had not fully developed the dynamics of Madison's view in my earlier work and that the citation I provided was incomplete; thus that I had not made my case that such a combination was useful. Indeed, in my book I claimed only that self-reliant representatives who used their own judgment to aim at the public good were more likely to be elected, without any indication of why or how this would come about.Footnote ⁴ The details are worth sketching out here, because they illustrate the usefulness of retaining this category to think about the highly responsive representative who relies on his or her own judgment, even if we cannot point to one today.

In my reading,Footnote ⁵ Madison presumed that most representatives would be very responsive to electoral sanction, leading him to endorse the extremely large and heterogeneous electoral constituency to incentivize representatives to rely on their own judgment. The Madisonian lawmaker, in my view, is someone who finds himself or herself in this situation, in a district with high heterogeneity of interests and great difficulty in having any one interest coalesce into a majority position. One town (or group) cares to keep butter prices low, another town wants road improvements, still another town wants to fight the creation of a national bank, and another wants a strong defense for the sake of the public good. Because the district is so large, Madison thought, each town would find it very difficult, if not impossible, to “communicate and coordinate” with others to build coalitions; candidates would discover that they could not do so either. In the face of this high heterogeneity of interests, where there is no singular “constituency judgment” to be had, candidates would quickly realize that promising to follow any one particular group's judgment about what to do would be a path to electoral defeat.

Why would these representatives then come to rely on their own judgment instead? Because voters would also come to realize that their preferred candidate (the one who would follow any subgroup's judgment about policy) would be unlikely to win election. So voters would opt for their second-best option: the man who would use his own good judgment to consider the best policy for all. (Madison also thought this dynamic would free up the candidate to be “republican”—aiming at the good of all—rather than “pluralist”—aiming at the good of the constituency or other subgroups.) Candidates would begin to campaign as if they were filled with republican virtue whether or not they actually were. They would begin to emphasize their qualities as men of good (self-reliant) judgment prone to act as if they cared about the public good, and not any partial view of it, even as they were highly motivated by sanction. What I took Madison's insight to be is not a defense of American-style pluralism but of what I called the Madisonian lawmaker: an individual motivated by sanction (future electoral success) who responds to the dynamics of the large, highly heterogeneous constituency by relying on his own judgment to pursue the common good.

Of course, Madison's plan did not work, because he failed to anticipate the rise of political parties, which made communication and coordination between subgroups much easier than he predicted. But the account illustrates the virtue of modeling the representative who is responsive to sanction, and self-reliant in judgment, because it gives us a view of the empirical and normative constraints we might face in trying to achieve it. The model might also inspire us to create the institutional conditions to realize Madison's vision for us today, perhaps by eliminating territorial districts and randomly assigning voters to permanent electoral constituencies instead. It is a plan I defend elsewhere (Rehfeld Reference Rehfeld2005, Reference Rehfeld2008) and would solve the perennial problems of gerrymandering and local “pork” spending. Most importantly for this discussion, it provides an illustration of where these conceptual distinctions might usefully lead.

If I have demonstrated the usefulness of thinking about representatives who are self-reliant but responsive to sanction, and thus of the need to keep these features conceptually distinct, what of its fellow traveler, representatives who rely on their constituents’ judgment instead of their own but are not responsive to sanction? We can see the usefulness of this dyad by imagining a “true” populist who enters politics dedicated to enacting the people's will, no matter what it is. Let us imagine a college professor—let us call him James “Jimmy” Chaplink—whose life work has been the study and promotion of participatory democracy. Chaplink cares more about the participatory process than election and re-election, and has always hated “begging for votes,” as Plato once put it. So when some of his former students approach him to run for Congress to promote his ideas, he initially declines.

But then Chaplink reads an article about a different kind of representative who has not been seen before—someone who relies on the judgment of his constituents, but who is not motivated by the prospects of sanction (winning—or losing—an election). And then he realizes, “wait a minute, that's me, that's what I've stood for my whole life.” After mulling over the possibilities, he realizes how this ideal type could bridge the gap between participatory and representational politics: If elected he would commit himself to cultivating participation among his constituents, educating them on the issues, bringing decisions to them as much as possible, and then relying on their judgments to cast his vote in Congress, whether or not he won re-election. He wants only to embody a pure “transmission belt” view of what a representative should do (Schwartz Reference Schwartz1988)

Because Jimmy's academic expertise lies at the intersection between normative and empirical political science, he knows that simply asking his constituents for their views about policy is not what true democratic participation entails, nor is it likely to produce good answers. So he proposes to increase their opportunities for deliberative and educative participation, planning to hold daylong forums on public policy and legislation, his own “deliberation days” as it were (Ackerman and Fishkin Reference Ackerman and Fishkin2004), and promises to do so as often as necessary and practicable. He will construct citizen juries (Warren Reference Warren, Warren and Pearse2008) and use his expense account and campaign funds to pay transportation costs and provide small stipends to make these sessions accessible to the less affluent.

Having made this his campaign platform, and in something of a fluke, Chaplink wins the election. He spends his next two years implementing his participatory program, running seminars, paying for public affairs shows to be broadcast via broadband Internet access in every one of his constituents’ homes, getting people involved—in short, bringing the legislature to the people. And the initial results are promising. Chaplink's constituents are excited to participate in politics in this new way, treated as mature citizens whose opinions matter by a representative not motivated by his prospects for re-election, but nevertheless reliant on their judgment.

But then reality sets in. The initial interest that surrounded Chaplink's seminars, workshops, and deliberative forums wanes, giving way to disinterest as people go back to leading their own lives. It seems that for all their initial enthusiasm, having tasted of the participatory fruit, voters no longer want such engagement and would now prefer to have someone who is less committed to those participatory values, but who generally shares their other values and motivations about policy, someone they might also set off on their own internal “gyroscope,” as Chaplink had once put it, without the need for constant monitoring.

As poll numbers come in showing a precipitous decline in support, Chaplink remains undeterred. He keeps the faith even in the face of increasingly hostile letters from his constituents urging him to show some independent judgment and leadership. “If it isn't your function to use your own judgment,” one of them angrily writes, “then what the hell is your function?”Footnote ⁶ Nevertheless, Chaplink continues to follow their judgment about how to vote and remains completely nonresponsive to their threat of sanction. Like most representatives who ignore all threat of constituent sanction, he loses office in the next election, becoming a one-term member of Congress, a mere footnote to history.

Despite being entirely reliant on his constituents’ judgment about policy, Chaplink is the quintessential gyroscopic representative, but only if we limit the scope of that concept to Chaplink's nonresponsiveness to sanction: He tells the voters what he is going to do and sets about doing it, completely nonresponsive to their threat of sanction.Footnote ⁷ Further, as with Mansbridge's description of the selection model, voters originally chose this gyroscopic representative to pursue their own values of participatory democracy, but at the next election they chose a different gyroscopic representative when their values changed. Mansbridge is right that we do not see characters such as Chaplink often, or perhaps even at all in democratic politics. More realistically, as Mansbridge and I have both noted, people are more or less reliant and dependent upon their own or other people's judgment, so the extreme case presented here is simply an ideal type. But that really is the point: I would not want to preclude a character who moved in the direction that Chaplink represents simply because we have never seen him before, or because he is unlikely to be seen, and providing even a rough sketch helps us think about the underlying causal and normative relationships that we would want to study or promote.

Chaplink is the example in the legislative sphere that illustrates what would be lost from ignoring these distinctions, and the value of conceptualizing the political world based on what could be, even though it never has been and likely never will be. This is why I think it would be more useful to limit Mansbridge's concept of “gyroscopic” only to describe nonresponsiveness to sanction and let it vary independent of the source of a representative's judgment. Of course sometimes categories will be empty and hard to imagine.Footnote ⁸ But it was precisely the formation of the conceptual category that, in this case, preceded and caused the imaginative exercise itself.

COMPLEXITY AND RELATIONAL ASPECTS OF REPRESENTATION

Our second point of disagreement arises because, as Mansbridge describes it, representation is a complex relational phenomenon; “representation” always entails one entity that represents, and another entity that is represented (Mansbridge Reference Mansbridge2011) Mansbridge wants concepts that reflect that relational complexity; thus she has embedded the relational structure of representation in the four kinds of representation she identified and developed—promissory, anticipatory, surrogate, and gyroscopic (Mansbridge Reference Mansbridge2003)—each of which is meant to describe the activities of representatives and their constituents simultaneously as they relate to each other. In contrast, Mansbridge claims that by focusing on a representative's “individual characteristics,” I miss the relational quality of representation: Because “‘surrogate’ representation is a relationship, not a characteristic of an individual representative, it is missing from Rehfeld's analysis” (Mansbridge Reference Mansbridge2011, 621). Mansbridge's critique thus raises two distinct issues. The first concerns whether I have described characteristics of individual representatives or something else. The second, more important critique concerns how best to approach the relational nature of representation. I will take each in turn.

Mansbridge is right that my tripartite distinctions focused only on the representative's perspective as he or she made decisions. However, these distinctions were not meant to describe characteristics of the representatives but the kind of activity in which those representatives were engaged: making decisions about how to vote on laws. The difference between representatives and representation-as-activity presses on the noun and verb senses of representation, which, along with its adjectival form, I think are critical to distinguish for the purpose of normative and empirical analysis. They tend to be lost when scholars attempt, as both Pitkin and Mansbridge have, to form “single highly complex concepts” (Pitkin Reference Pitkin1967, 8; quoted previously). And, as I said earlier, the failure to distinguish these parts into separate concepts has only contributed to the confusions and ambiguities that are at the heart of a good deal of scholarship on this topic.

Let us start by distinguishing the three different senses of representation that correspond to the noun, verb, and adjective forms of the word. We can point to “representatives” (noun) as entities in the world: “Susan is my representative.” We speak of “represent” (verb) as a kind of activity: “Susan represents me.” Finally, we often describe two entities that share features with each other as “representative” (adjective) of the other—people who share the race, gender, or ideological commitments of other members of a group are sometimes said to be representative of them in this adjectival sense. Any theory of political representation, explanatory or normative, will have to account for the relations between these three forms, and that will require us to keep them conceptually distinct. For example, must one engage in the activity of representingFootnote ⁹ (verb) to be a representative (noun)? Does representing (verb) plus being a representative (noun) entail that one is representative (adjective) of a group? Probably not: A farmer of one ethnicity may be the representative (noun) of an urban constituency of another ethnicity, and represent (verb) them and their interests in Congress, without in any way actually being representative (adjective) of them.

Even if there is a minimal core shared sense that defines a Wittgensteinian family resemblance around concepts such as “representation,” there is no reason to treat a concept's cognate forms—“represents,” “representative,” etc.—as referring to the same underlying idea. Consider, by analogy, H.L.A. Hart's The Concept of Law, in which Hart set forth the question, “What is Law?” (Hart Reference Hart1961, 1). Had he asked instead “what is legislation?” he would have had to clarify first whether he meant its noun or verb forms, whether, that is, he wanted to inquire into the concept of law (as he did) or the concept of “legislate” that its verb form covers, and then to explain how those two related to each other. Although each may have been related to some overlapping idea (whether having to do with rules, as Hart believed, or fidelity to moral norms, as Fuller and Dworkin had thought) a concept of legislation that failed to clearly distinguish the linguistic forms of the concept, or attempted to conceptualize a “single, highly complex concept” of “legislation,” would have been less useful, and perhaps doomed to failure.

The underlying concepts of “legislation” are easier to identify because we use distinct words to track them: law (noun), legislate (verb), and legal (adjective). In contrast, the related forms of “representation” [representative (noun), represent (verb), representative (adjective)] bear a far greater similarity to one another (its noun and adjective forms being identical). But the fact that history has given us words whose underlying forms are harder to distinguish does not tell us much about whether their forms bear a greater or lesser similarity to one another than the forms of any other concept. Like “legislation,” the word “representation” refers ambiguously to noun and verb forms. For example, “I did not receive representation in that case,” might mean that I was not represented (verb) by anyone [whether or not by a representative (noun)], or it might mean that I had no representative (noun) in some context, without reference to the kind of activity in which that person engaged. This may explain in part why it is difficult to gain traction on some key issues of group representation, because it is not always clear whether that term is used to refer to a representative (noun) a group can control, being represented (verb) in a way that advances the group's interests, or a person who is representative (adjective) in the sense of “bearing some similarity” to the group in question. Each of these may in some sense be related to the other, and all of them might be important. (Phillips Reference Phillips1995) But they are certainly not the same set of issues, and treating them by reference to a singular concept of “group representation” cannot but confuse research into the topic.Footnote ¹⁰

I certainly do not mean to decide these issues here. The point is only to demonstrate the value in keeping these forms distinct so that we may ask how each of the parts relate to one another and to further press the difficulty with attempting a conceptualization of a term such as “representation” that is supposed to span all these forms. With these distinctions in mind, my original article (Rehfeld Reference Rehfeld2009) meant to unpack what representing (verb) is often said to involve and was not meant to be speaking to characteristics of representatives (noun) per se, as Mansbridge has claimed.

This first point about separating the noun, verb, and adjective forms of representation is meant as clarification. Mansbridge's more important criticism was that I had failed to capture the relational nature of representation in my suggestion that our concepts should separate the activity of representatives (in terms of their expectations or judgments about voters) from the activity of voters (their expectations or judgments about their representatives). And here she is right in one sense: I am not considering directly the relationship between representatives and those whom they purportedly represent, but considering instead one side of that relationship at a time: the sources of representatives’ judgment, their referent points, and the motivations that go into describing how they make decisions. However, I agree with her that representation is a relational concept, so that any of its forms (noun, verb, or adjective) always indicates two sides of a relation. Our disagreement thus centers on whether to design concepts that emphasize one side of the relationship at a time, as I prefer, or whether, as Mansbridge prefers, to design concepts to identify their simultaneous interactions.Footnote ¹¹

Let us start with our point of agreement and ask what it means to say that representation is a relational concept. Relational concepts are concepts that refer simultaneously to two entities that stand in relation to one another. So, for example, the relational concept of “parent” is impossible to understand without the concept “child” in hand, because one cannot be a parent without standing in relation to a child. Representation is the same way: One cannot be a representative (noun), or represent (verb) or be representative (adjective), without two entities being involved: an object of representation, and something else standing-in-for that object. The person or group represented may exist prior to being represented, but it is not formally a represented entity without a representative in-for-whom it stands. Thus, for example, an electoral constituency might exist prior to the election of their representative, but it is not an object of representation and cannot be until it is being represented (verb) or at the point where a representative (noun) of it exists.Footnote ¹²

All relational concepts pose a particular challenge to concept formation, because there is no way to describe one side of the relationship without in some sense implicating the other. Non-relational concepts such as “table,” “bus,” or “pencil” have no such implications, for we can think of all these things on their own, as it were, without reference to other specific entities (even if they are related more generally to the world around them). In contrast, there is no way to conceive of one part of a relational concept without implicating the idea of its fellow traveler: Asking what a parent ought to do implicates the idea of “child,” because one cannot be a parent or act in a parental way except in relation to a child. Even when we use the terms euphemistically, we imply both sides of the relation. For example, when we say, “The teacher acted parentally toward her students,” we mean, “she treated them like children in some sense.” Yet, as much as the concept “parent” implies the concept “child,” the parent as an entity is conceptually distinct from the child; similarly, parenting as an activity is conceptually distinct from the child's activity, even if one side implicates the other. Emphasizing one part of the relationship never denies that it is somehow related to the other; it simply directs our focus to one side of that relationship at a time.

Returning to representation, the question before us is whether concepts that emphasize one side of the relationship, fixing values to the variables they identify, are more or less useful than concepts that emphasize and fix both sides of the relationship at once. Mansbridge insists that our concepts should emphasize not just the representative's view of keeping promises, or anticipating voters, but also simultaneously the judgment of constituents who vote retrospectively or prospectively for keeping, or failing to keep, the promises they made. Though both perspectives are not always included in her definitions, they often are, as in this description of “promissory representation”:

In her response, Mansbridge more forcefully explains that these concepts are meant to describe features of what both representatives and their constituents do, not just a description of a “representative's individual characteristics.” Continuing, she explains,

By this relational view, as we might call it, promissory representation would seem to apply only when the representative currently and the voter later looks back to past promises, whereas anticipatory representation would apply only when the voter and representative are both looking forward to what the representative or voters might do.

Yet if we used “promissory” and “anticipatory” to set both sides of the relation simultaneously, it would be incomplete, and leave us unable to classify cases where representatives acted in one way but voters acted another way. Consider the case of U.S. President George H.W. Bush's 1988 promise to raise “no new taxes,” a promise he later broke. Let us presume for the sake of this example that Bush broke this promise anticipating that he could change voter minds in the subsequent election. So on his side of the relationship this is shaping up to be a case of anticipatory representation. Yet when he ran for re-election in 1992, many voters were firmly in the promissory camp; that is, they held him to account for having broken his promise. So according to Mansbridge, if we have to include both voter and representative perspectives simultaneously, when Bush signed a tax increase in 1990 it was neither a case of promissory representation (Bush broke his promise) nor a case of anticipatory representation (voters voted retrospectively). The named concepts are thus unhelpful precisely because they do not let each side of the relationship vary independent of (or more precisely, in reaction to) the other side.

Now consider how useful it would be to keep each side of the relationship independent of the other, one in which “honoring promises” and “anticipating future elections” were features that described different ways of representing (verbs), and, consistent with the existing literature, “prospective” and “retrospective” described different ways that constituents voted. Do or should representatives respect the promises they made? Do or should they anticipate what they believe the voters will want at the time of the next election, and if so, do or should they actively attempt to change their constituents’ views? And those questions concerning what representatives do would then stand in relation to voter judgment. Do or should voters vote prospectively based on what they believe their representative will do, or retrospectively based on sanctioning a representative for past performance? These questions require us to have concepts that emphasize one side of the relationship, rather than fixing both sides at once.

In fact, if we used these terms only to emphasize the representative's side of the relationship, and used the terms “retrospective” and “prospective” (already in wide use in many literatures of political science) to describe what voters do, we would move beyond the binary “promissory/anticipatory” space to one in which promissory and anticipatory representing (verb) vary along with prospective and retrospective voting:

• • Promissory representing of prospective voters: A representative who keeps promises representing a constituency that votes based on its expectations of future performance. Arguably illustrated by the example of Jimmy Chaplink presented earlier, and described by Mansbridge's selection model (Mansbridge Reference Mansbridge2009).

• • Promissory representing of retrospective voters: A representative who keeps promises representing a constituency that votes as a sanction or reward for past performance. Arguably what Mansbridge means by “promissory representation.”

• • Anticipatory representing of prospective voters: A representative who tries to anticipate what a constituency will want representing a constituency that votes based on its expectations of future performance. Arguably what Mansbridge means by “anticipatory representation.”

• • Anticipatory representing of retrospective voters: A representative who tries to anticipate what a constituency will want representing a constituency that votes as a sanction or reward for past performance. Arguably the case of George H.W. Bush's broken “no new taxes” promise.

Either way, if there is a relation of promising that involves looking back, and one of anticipating that involves looking forward, and we are emphasizing the relationship between constituents and representatives, we ought to let each of them be forward-looking or backward-looking independent of the other. And whatever we want to call them, these features of anticipation and promising are worth isolating from all the other features that Mansbridge suggests including in the terms.

SURROGATE/VIRTUAL REPRESENTATION AND ELITISM

Our last disagreement begins with a correction. In my original article, I described Mansbridge's “surrogate representation” as identical to Burke's “virtual representation”—each emphasizing the lack of an electoral connection between a represented group and their representative. I further claimed that the only reason Mansbridge resisted Burke's term was because she found his views elitist. This was a mistake. As Mansbridge (Reference Mansbridge2011) has made clear, the term “surrogate” was meant to emphasize different substantive views of representation from Burke's, not merely the absence of an electoral connection. And Mansbridge's objection to Burke's elitism lead her away from Burke's term “trustee” in place of which her selection model “eschews such hierarchy” (623).

However, in defending these choices Mansbridge (Reference Mansbridge2011) makes clear that her concepts are meant simultaneously to describe relational features and substantive views of representation, rather than keeping each distinct. “Surrogate representation” is meant to describe both a surrogate relationship between two parties and apply to aggregative and deliberative representing (verb) during that relationship. “Gyroscopic representation” and its corresponding “selection model” is meant to describe both an unmonitored relationship between two parties, and a relationship in which both sides eschew hierarchy. Our final disagreement then arises because I think these concepts would have far greater utility if they focused only on their relational features rather than building into them other substantive views. I take each in turn.

First, what exactly is the difference between Burke's view of virtual representation and Mansbridge's view of surrogate representation, and more generally what is the value of either modifying concept (“surrogate” or “virtual”)? As I had written, “for Burke (and the many others who use the term), the critical point of ‘virtual’ was precisely, only, and no more than Mansbridge's own definition of the surrogate: ‘representation by a representative with whom one has no electoral relationship’” (Rehfeld Reference Rehfeld2009, 221, note 17; quoting Mansbridge Reference Mansbridge2003, 522) The reason Mansbridge disagrees with that characterization is that she believes the use of “virtual” carries with it Burke's substantive view of representation, which the term modifies. As she described, “Edmund Burke had a version [of surrogate representation] he called ‘virtual’ representation, but Burke's concept focused on morally right answers, wisdom rather than will, relatively fixed and objective interests and the good of the whole” (Mansbridge Reference Mansbridge2003, 522). Furthermore, Mansbridge insists that surrogate representation involves aggregative and deliberative elements, whereas the virtual representation that Burke described “applies only to deliberation. Surrogate representation applies to both the aggregative and the deliberative functions of democracy” (Mansbridge Reference Mansbridge2011, 627). So although “surrogate” and “virtual” both capture the lack of an electoral connection between those represented and the representative, surrogate representation is meant to differ from Burke's views of virtual representation in that it includes a particular substantive account of representation that differs from Burke's substantive view of representation.

Note that the four qualities Mansbridge lists of Burkean representation (“morally right answers . . . wisdom . . . objective interests . . . and the good of the whole”) are features of Burke's substantive view of good representation, but do not distinguish virtual or actual types. In other words, emphasizing those four qualities does not capture the work that “virtual” was meant to do in Burke's own theory. And if, as Mansbridge claims, deliberative and aggregative aspects of representation make surrogate representation different from Burke's virtual representation, it is unclear why she thinks that virtual representation was a version of surrogate representation at allFootnote ¹³ (Mansbridge Reference Mansbridge2003, 522).

What Burke really meant by “virtual” is of course something that Burke scholars may want to debate. And if Mansbridge or other scholars find it helpful to limit “surrogate representation” to those cases where representation-as-activity is both deliberative and aggregative, they should of course use the term in that way. But the substantive issue here is that it would not be particularly useful to adopt a concept of surrogate representation that covers lack of an electoral connection, plus deliberative and aggregative elements, minus morally right answers . . . wisdom . . . objective interests . . . and the good of the whole (i.e., Burke's key substantive views of representation). For if we were to employ such a concept, we would need many other concepts to cover cases of “lack of electoral connection” when paired with any number of alternative substantive views of representation, such as Urbinati's (Reference Urbinati2000) “representation as advocacy,” Young's (Reference Young2000) “representation as identity,” or Stimson, Mackuen, and Erikson's (Reference Stimson, Mackuen and Erikson1995) “representation as policy correspondence.” More importantly, it would be far more useful to have a concept that denotes the transformation of whatever substantive conception of representation-as-activity one uses to cases where the person acting has no electoral connection to those on behalf of whom he acts, for that would be the only way to treat the lack of an electoral connection as an independent variable for normative and empirical analysis. That is all that I believe Burke meant when he used the term “virtual”; that is the nub of what I believe is valuable from Mansbridge's idea of surrogate representation; and it allows us to investigate the very interesting causal questions of whether this transformation alone has other effects, and whether normatively it is worth pursuing.

As for Burke's elitism, Mansbridge (Reference Mansbridge2011) objects to my use of “Burkean trustee” to illustrate the overlap of the republican, self-reliant, and nonresponsive representative. Having ignored Burke's elitist views, she argues, “this term does not and should not define that cell” (625). Though I had only used “Burkean trustee” as an example of the cell and explicitly not a definition (Rehfeld Reference Rehfeld2009, 223), Mansbridge is right that the three distinctions and the eight cells of my original table did not capture whether voters thought their representatives were superior to them or not. Though I generally agree that Burke had what Mansbridge describes as elitist views of representatives (or, rather, dismissive views of voters) and I think it is of great importance to explore dimensions of political equality, I do not see the value of differentiating along these lines as part of the map of decision making I was trying to capture. Further, and consistent with the previous discussion, if we think it is useful to capture views of hierarchy between voters and representatives, we ought to differentiate clearly between the perspective of representatives toward their constituents (“swinish multitude”) and the perspective of voters toward their representatives (“I want someone who's my equal”), a distinction that Mansbridge ignores.

However, the reason that Mansbridge believes we should prefer her selection model that “eschews such hierarchy” (Mansbridge Reference Mansbridge2011, 623) is again that it reflects one seemingly common regularity: the case where voters want representatives who are “like them.” But even if this were the regular case, there will be greater value in having a concept of selection that allows perceptions of superiority to vary apart from the views either side has of the other. Is the utility of the selection model not chiefly, as Mansbridge described, that voters select representatives who “are internally motivated and have goals closely aligned with those of [the voters themselves]” (Mansbridge Reference Mansbridge2011, 622)? If so, it would be more useful to apply the model whenever voters select representatives who are internally motivated to seek those shared goals, no matter what hierarchical status voters attribute to their representatives or vice versa. Maybe, with Manin (Reference Manin1997), voters want or should want representatives who they believe are distinctive in many ways, so that those representatives are more likely to achieve those shared goals. Maybe voters want or should want representatives whom they perceive as equals because they want to feel a stronger personal connection to them. Or maybe they want or should want representatives whom they view as inferiors, because they want to feel powerful and in control. I see no utility in burdening her selection model—valuable as it is to describe “selection of gyroscopic representatives” by voters—with substantive dimensions of hierarchy or equality at all. It is another example of the value at the core of Mansbridge's concepts, but only when separated from the complexity that she wishes to build into them.