Reasons and Inclusion: The Foundation of Deliberation*


  • *

    Address correspondence to: Shamus Khan, Department of Sociology, Columbia University, 413 Fayerweather Hall, 1180 Amsterdam Avenue, New York, NY 10027. Tel.: 212-854-2489. E-mail: This article is equally co-authored. We wish to thank Charles Camic, John DeLamater, Jeremy Freese, David Laitin, Alair MacLean, James Montgomery, Pamela Oliver, and especially Erik Olin Wright, Mustafa Emirbayer and Myra Marx Ferree for their extensive help. Thanks also to our five undergraduate administrators, Amanda Hinze, Danya Bader-Natal, Britin Wagner, Sarah Potter, and Leanna Trunzo. We also thank the three anonymous reviewers for their comments on a previous draft of this article.


This article provides two empirical evaluations of deliberation. Given that scholars of deliberation often argue for its importance without empirical support, we first examine whether there is a “deliberative difference”; if actors engaging in deliberation arrive at different decisions than those who think on their own or “just talk.” As we find a general convergence within deliberation scholarship around reasons and inclusion, the second test examines whether these two specific mechanisms are central to deliberation. The first evaluation looks at outcomes within a laboratory setting; the second at videotapes of decision-making processes within this setting. Our results show two things. First, in terms of outcomes, deliberation differs from other forms of interaction. Second, reasons and inclusion are central to the deliberative process. The more reasons provided within each group, the more likely participants were to change their position; similarly, the more inclusive groups were, the more likely participants were to change their position. We conclude by arguing that more work needs to be done, both in evaluating the deliberative difference and in disaggregating deliberation and examining its central explanatory mechanisms.

“Deliberation” is an increasingly popular concept within the social science literature and has gained political weight as democratic movements have argued for its importance to the political process. Yet social scientists have done relatively little to examine empirically how deliberation might vary from other forms of decision making, or if the posited key constitutive elements of deliberation are central to explaining the “deliberative difference” in interactions. We argue that as deliberation is asked to bear more political and theoretical weight, knowing the “how and why” of deliberation becomes increasingly important. Specifically, we ask (1) if deliberation differs from other forms of decision making and (2) if reasons and inclusion explain this “deliberative difference.”

While the literature1 is not in agreement about what counts as “deliberation,” we argue that there is a general convergence around the explanatory importance of reasons and inclusion for deliberation. We argue that for almost all scholars they constitute the essence of deliberation; as such they form the basis of the hypothesis testing in our study. Using the experimental method, we test the effects of deliberation, which we operationalize as interaction marked by reason-giving and inclusion, against two other treatments: (1) essay, by which we mean simply thinking for oneself, and (2) discussion, by which we mean interaction without being guided by a process that emphasizes reason-giving and inclusion. We operationalize individual choice as the dollar amount each subject wishes to pay for university segregated fees: a program that is effectively a tax on students used to fund student services and organizations.2 While we find no real difference between the essay and discussion treatments, there is a significant difference between the deliberation and the essay treatments. When we control for administrator effects, there is also a significant difference between the deliberation and discussion treatments. This suggests that when asked to deliberate, subjects are more likely to change their position than they are when they are simply thinking for themselves or “just talking.”

We then move beyond the holistic approach (deliberation as variable) and examine deliberation as an interactive process marked by two mechanisms—reason-giving and inclusion—that can be tested as variables in their own right. To do this we use direct observation of videotapes of all group interactions, looking more closely at exactly what happens within our groups. Our findings indicate that across treatments, the more a group provided reasons and fostered an ethic of inclusion—the more groups robustly engaged in deliberation—the more likely the participants were to change their position. In sum, when reasons are given and people are included, outcomes differ more significantly. We acknowledge that our findings come from the exploration of a very small set of phenomena and that our narrow conceptualization of deliberation leaves much unconsidered.3 Yet we believe that they are interesting and robust enough to provide motivation for further studies into both the effects of different modes of decision making and the mechanisms behind them.


We are not the first scholars to point to the Kantian-inspired theoretical foundations upon which the idea of deliberation is built: reasons, inclusion, and justice or legitimacy (Bohman 1996; Gutmann and Thompson 1996, 2004; Steiner et al. 2004). Gutmann and Thompson (2004), for example, argue that reasons and inclusion are two central aspects of deliberation.4“[Deliberative democracy's] first and most important characteristic is its reason-giving requirement … A second characteristic of deliberative democracy is that the reasons given in this process should be accessible to all the citizens to whom they are addressed”(2004:3–4, italics in the original).5 In his theoretical work on “public deliberation”Bohman (1996) similarly highlights the importance of “exchanging reasons” and inclusive “interpersonal coordination and cooperation” (1996:27).

The modern builders of these ideas are John Rawls and Jürgen Habermas; their thinking provides the theoretical core of our article. Rawls (1971) and Habermas (1984, 1990) believe that deliberation can help us realize just or legitimate ends, and the process is not one that occurs within subjects but rather is realized through interactions between subjects. Rawls's project is a philosophical one: he seeks a universal theory of justice. Rawls points out that actors are embedded within relations where past and present experiences greatly affect their parameters of understanding, their recognition of positions other than their own, and their interests. Abstraction from such positioning within the world is impossible, making a theory of justice unrealized. His solution is the “veil of ignorance.” Imagine that we could place all actors behind a veil wherein they are ignorant of their conditions in life—a situation Rawls calls “the original position.” What kind of decisions would such actors make about the social world? Rawls argues that these decisions can and should serve as the basis of a theory of justice. When Rawls thinks about what happens behind this veil of ignorance, he imagines a situation where actors are included, offer reasons, and employ a particular kind of rationality: deliberative rationality. Rawls is unable to imagine what might happen behind a veil of ignorance without thinking of groups deliberating. “To say that a certain conception of justice would be chosen in the original position is equivalent to saying that rational deliberation satisfying certain conditions and restrictions would reach a certain conclusion” (1971:138). One of these conditions is “equal participation,” another the provision of reasons (1971:231–34, 139).

Habermas does not look for a theory of justice, but rather seeks to find a social process by which legitimate decisions can be reached. Throughout his work Habermas argues that discourse and deliberation are central to the process by which we generate general values and norms. Whereas for Rawls such discourse is enacted through deliberative rationality, Habermas has a much different process in mind for his actors, that of “communicative action.” Within this process, he describes a discourse ethic in which “only those norms can claim to be valid that meet (or could meet) with the approval of all affected in their capacity as participants in a practical discourse” (Habermas 1990:66). There are, then, two central conditions for Habermasian deliberation: first, that actors offer up reasons in attempts to provide “justification of norms,” and second, that all those affected are “participants in a practical discourse.” This emphasis on the practical marks a clear difference between Habermas and Rawls; Habermas argues, quite forcefully, that deliberation is not simply a form of rationality, but rather is a kind of social process. Further, “the communicative model of action does not equate action with communication” (Habermas 1990:101). Instead, he understands it as an interactive process where actors collectively reason through validity claims with all affected actors in order to arrive at those that are universally valid.6

For all the differences between Rawls and Habermas, these two thinkers have much in common. They both develop a model of deliberation in their work; for both, this model focuses on reasons and inclusion; and finally, they both argue that deliberation is central for realizing “just” or “legitimate” outcomes.7Young (1999) and Fraser (1991) correctly point out the problems behind the idea of inclusion presented by Rawls and Habermas, and argue for wider representativeness and a critical evaluation of the biases implicit within the field of interaction.8 These critiques, however, do not argue against the importance of reasons or inclusion for the deliberative process; rather, they claim that those within the Kantian tradition have not applied the concepts fully enough (either in terms of the breadth of acceptable reasons or the inclusiveness the deliberation). We share this concern.


Subsequent scholars have picked up on the theoretical ideas provided by Rawls and Habermas to build an agenda for empirical research. Of the recent empirical work we have examined, we think that the best articulation of this agenda is from Fung and Wright (2003). They take Habermas as a theoretical starting point, and drawing upon several empirical studies of deliberation argue that:

In deliberative decision-making participants listen to each other's positions and generate group choices after due consideration. Participants ought to persuade one another by offering reasons that others can accept… . Real-world deliberations are often characterized by heated conflict, winners and losers. The important feature of genuine deliberation is that participants find reasons that they can accept in collective actions, not necessarily ones that they completely endorse or find maximally advantageous. (Fung and Wright 2003:17)

This approach searches less for social ideals and acknowledges more the real-life aspects of these actual processes, which yield winners and losers. Fung and Wright come as close to creating a dialogue between empirical and theoretical work as any working within the deliberation tradition. Drawing on a number of case studies, from Porto Alegre (Baiocchi 2001), to Chicago (Fung 2001), to the U.S. Forestry Service (Thomas 2001), and Kerala (Heller and Isaac 2003), they give us a sense of what happens in real-world deliberations across a variety of contexts; this informs both their arguments that deliberation can produce emancipatory effects and their speculation as to how this happens.

While we appreciate the real-world context provided by Fung and Wright's close attention to empirical work, we have concerns about the work they draw upon. These concerns are twofold. First, because of their holistic approach, looking solely at deliberation as a variable, these studies largely ignore mechanisms within deliberation. None of the empirical work on deliberation seeks to evaluate whether or not reasons and inclusion (or any other deliberative mechanisms, for that matter) are truly explanatory. Rather, these mechanisms are either axiomatic or ignored. Second, all of the work that Fung and Wright draw upon is qualitative. One of the challenges for qualitative research is in assigning causal effects to a single variable (or set of variables). We cannot help speculating that there are confounding variables in most, if not all, of the studies that Fung and Wright cite. For example, if we take Baiocchi's work on Porto Alegre,9 we cannot be sure if it is the deliberative process, the form or constitution of the government or citizenry, the local political sensibility, or the realm of civil society that produces the kinds of results we observe. Chances are that all of these things are causally relevant, but as other cities in Brazil have been less successful at realizing Porto Alegre's results while using the same procedures, it is difficult to argue that deliberation explains the effects observed in the Porto Alegre case.10 Thus, we are in a position to question the appropriateness of the causal assignment of outcomes to deliberation.11 Furthermore, these qualitative studies are often of people who already have a commitment to participatory democracy. This no doubt clouds the results.

The problems mentioned above are not simply confined to qualitative approaches; Luskin et al.'s (2002) work on deliberative polling12 has similar issues. Luskin et al. (2002:456) argue that one of the most difficult problems for citizens making decisions (or answering questions about potential decisions) is that “not many respondents answering any given question have very well considered or informed opinions about the issue.” Deliberative polls seek to explore “what the public would think about issues, if it thought more earnestly and had more information about them” (2002:458). They brought together a national probability sample of British citizens for a weekend (N= 301) and had them listen to experts and discuss (in small groups) the issue of crime. The purpose was to make subjects “more like ideal citizens” (2002:460). They show that after this weekend citizens did indeed make different decisions; in other words, deliberation produces different outcomes.

Yet we find that this work still leaves fundamental questions unanswered. Luskin et al. point out that multiple variables are part of the overall process of the deliberative poll. And so, again, we cannot be sure that it is deliberation that explains the outcomes they observe. This critique would come as no surprise to Luskin et al., as they freely acknowledge that their empirical work does not explain

how much of the information gains and changes in policy preferences came from the briefing materials, versus talking, reading and thinking about the issues in the gestation period between recruitment and deliberation, versus the small group discussions, versus the large group discussions with policy experts, versus the large group discussions with politicians, etc. (2002:484)

We believe that knowing the causal weight of talking and information gains is of critical importance. If the observed change is simply a matter of information gains, then social interaction may not matter. If it is a matter of talking, then subjects do not require experts to inform them of aspects of a situation; they simply need to interact with one another. We also argue that what is lost in all the above work is close attention to what has traditionally been seen as the central mechanism of the deliberative process: the provision of reasons and an ethic of inclusion.13

We are not the first scholars to note problems with the empirical sociological literature on deliberation. As Polletta and Lee (2006:700) point out in their recent work on deliberation, most good sociological work on deliberation is either historical (Fraser 1992; Ryan 1992; Schudson 1992, 1997) or ethnographic (Eliasoph 1998; Eliasoph and Lichterman 2003; Hart 2001; Lichterman 1996). While we have learned much from this work, it does not address what we find to be the two central challenges to this literature: “deliberation” is so vaguely defined in much of the above empirical work or so densely integrated with other variables that we cannot be sure (1) if it explains the observed outcomes, or (2) if the continually posited mechanisms of reasons and inclusion truly matter. We take on both of these tasks by first looking at a holistic model of deliberation and seeing how it affects decision making in the lab. Second, we explore the actual interactive process of groups, examining if the posited disaggregated mechanisms of reasons and inclusion truly matter to generating the deliberative difference. This focus on looking inside “the process of deliberation” is a starting point for recent empirical agendas (Polletta and Lee 2006; Rosenberg 2005) and serves as an important part of our work. Yet we feel that before looking inside the process, we must be sure that there is a deliberative difference.

Simon and Sulkin's (2002; Sulkin and Simon 2001) laboratory work at first glance seems to be what we are looking for. Their project of testing “Habermas in the lab” and examining the importance of discussion in group decision making is promising. Using ultimatum and divide-the-dollar games, they show how deliberating subjects are more likely to distribute money equally and perceive outcomes as fair, as compared to those who do not deliberate. Yet the work has some fundamental problems. First, what they consider “deliberation” would be barely recognizable by most deliberation theorists. Their experiments consist of subjects sitting at computers typing to one another. This is far from the dynamic process that Habermas imagines, nor does it necessarily ensure that subjects employ reason-giving or include all participants. While their results provide support for the idea of communication affecting outcomes, they generate no insight into how deliberation might differ from any other form of communication. In fact, for Simon and Sulkin nearly all communication is deliberation; as such, deliberation is a fairly meaningless concept. Following Fearon (1998), they use discussion as a proxy for deliberation. “Although ‘mere communication’ may not be deliberative, discussion is a necessary prerequisite of the deliberative process because, without discussion, there can be no deliberation. Moreover, deliberation itself is likely impossible to manipulate in the way required in experimental designs because experimenters cannot guarantee that participants will engage in the principled conduct that deliberation requires” (2002:404). We disagree that deliberation cannot be manipulated—we believe we have found a way to do so while preserving the strength of the experimental method. And by coupling this method with direct observation, we are able to turn their perceived methodological limitations into a research question. Even further, it is unclear whether it is the communication (transfer of information) that generates the change in position, or the emergence of a social relationship that has causal significance. A final major difficulty is that Sulkin and Simon's laboratory setting is so far removed from actual deliberative processes, with the exception of possible chat-room-style deliberative processes, that their results are hard to apply beyond a laboratory setting.14 For example, how do we extrapolate from two to five students sitting in front of a computer to, say, a school board meeting or a legislative committee debate? While all experiments have problems of external validity, in this case the difficulties are extreme.


Deliberation theorists posit that deliberation affects outcomes. We find a general convergence around reasons and inclusion as central mechanisms for explaining why. Rational choice theorists are less than clear on answering the “why” question, but might explain the effects of reason and inclusion relative to the provision of information about the situation or the effects on others. Those who do empirical work on deliberation still have a long way to go before they demonstrate the effects of deliberation or the mechanisms by which such effects might be explained. We use the experimental method to assess whether there is a deliberative difference against other forms of interaction. We then turn to direct observation of group interactions to disaggregate deliberation and assess the importance of reasons and inclusions as variables in their own right.

We first draw upon the experimental method because it provides the researcher with two main advantages over other methods: the ability to control for factors other than the independent variables of interest and random assignment of participants to treatment and control groups. (Aronson et al 1998; Marwell 2000; McDermott 2002) Almost all of the causal weight can be assigned to the condition(s) that vary between treatments, providing tremendous power for making statistical and causal inference. These advantages come at a cost, for while every researcher attempts to replicate “real-world” situations in the lab there is a big difference between a group talking at a table as part of an experiment and a group of citizens arguing over a school budget at their town meeting. Devine et al. (2001) cast this problem in terms of the cost of “structural verisimilitude,” or the relevance of experimental findings for actual deliberation. Sulkin and Simon (2001) provide an answer to this concern by pointing out that “natural” deliberative settings may provide increased external validity but make it difficult to determine causal relationships because of their complexity. Our experimental design allows us to cut through the complexity and isolate the effects of greatest interest to our study, while avoiding what Devine et al. (2001) call the “plague” of confounding variables. As with all experiments, however, it has the problem of lower external validity. Unlike other experimental works on deliberation, such as that of Sulkin and Simon (2001), one of the primary concerns for our design was to replicate real-world processes as closely as possible without losing the ability to assign causal weight. We do not believe that it is an either-or situation: while we concede that we have relinquished some external validity in order to isolate specific variables of interest, we also believe that our design does approximate, in a meaningful way, group interactions that occur in the social world.

The question for our experimental data is whether the kind of social interaction matters for individual decision making. We used three different treatments in our experiment, summarized in Table 1.

Table 1.  Experimental Treatments
 Interactive Context of Communication
Communicate with OneselfCommunicate with Others
Rules of communicationUnspecifiedIII
Spontaneous essayDiscussion
Reason-giving, inclusive ethic——15Deliberation

In the first treatment, essay, participants consider ideas on their own, through the project of writing an essay. This is, in essence, an operationalization of undirected, individual thinking and serves as our control in the experiment. While we acknowledge that asking a student to write an essay is a very particular operationalization of individual thinking, we believe that it is justified and does not threaten the internal validity of the experiment. We would also point out that by designing the experiment in this way, it is not our intention to assert that humans are incapable of self-deliberation. In defining deliberation as a social process involving two or more individuals, we are able to ask if just thinking about an issue for a period of time is enough to change someone's position, or if the interpersonal interaction matters. Our second treatment, discussion, is operationalized in the experiment as group interaction (verbal and nonverbal) with no input from an administrator.16 In the discussion treatment, reason-giving and inclusion may spontaneously occur but are not part of the administered treatment. The third treatment, deliberation, is operationalized as a group interaction in which reason-giving and an ethic of inclusion are actively encouraged through clearly stated procedural guidelines and the limited input of the administrator. The guidelines are grounded in six central principles that were explained to the participants:17

  • 1Participants listen to one another.
  • 2Participants do not just offer opinions, but rather provide reasons.
  • 3Conflict is okay.
  • 4Participants should find reasons they can accept.
  • 5Participants should be open to new proposals.
  • 6All participants should be included in the process.

We understand these principles to be based on reason-giving and an ethic of inclusion.18 It should be clear that our experimental design rejects the idea that deliberation is pace Habermas—an organic phenomenon that cannot be exogenously imposed. If we are wrong, then any experimental manipulation, including those mentioned above, would have no external validity. As we shall see from our direct observation, we have some reason to feel protected from this critique.

At this point we wish to squarely address those readers who might object to our parsimonious operationalization of deliberation. The strength of our chosen method—the isolation of a causal effect—is also its weakness insofar as only a few variables can be tested at a time. We take seriously the claims of deliberation theorists that this process is more complex and dynamic than our simple variant allows. However, for the purposes of such empirical testing—especially when such testing is in its infant stages—such parsimony is necessary and it provides our work with a level of internal validity that other empirical work often lacks. Further, our second observational method allows us to explore deliberation beyond the simplistic terms that the experimental method demands, allowing us to both disaggregate reasons and inclusion within deliberation and see deliberation as a more interactive dynamic social process.

We also wish to anticipate two other criticisms, namely, that our administrators are confounding variables, or that they represent a flaw in our design. How can we be certain that any effects from the deliberative treatment are due to reason-giving and inclusion, and not simply the added input of the administrator? One could even argue that our third manipulation is not deliberation at all, even in our own sense of the term, and that the treatment is really just about cueing participants. We use the term administrator conscientiously because they were not facilitators. The administrators were under strict guidelines to ask specific scripted questions designed to foster inclusion and reason-giving. They did not offer opinions, mediate, summarize points, or attempt to generate consensus. They were there to administer the treatment and attempt to ensure that “deliberation” took place. Further, administrators were blind to the object of research. The administrators for both interactive treatments did not sit at the table with the participants. In the discussion treatments, administrators sat at a table in one corner of the room with their back to the group. In the deliberation treatment, administrators walked around the table, shifting their position every few minutes. We argue that the presence and participation of the administrator was simply our best effort to operationalize the deliberation guidelines provided at the outset of the treatment. We did not believe that the guidelines in and of themselves would be enough to stimulate a threshold of deliberation; the gentle nudging and reminders from an administrator to be inclusive and offer more than just opinions were required as well. Upon watching videotapes of all sessions, we find our belief to be correct: groups did need reminders from the administrators to be inclusive and offer reasons. In fact, this is what the administrators did and not much else. Further, we wanted administrators in our design because in many “real-world” deliberative processes, administrators are present. Town meetings have moderators, community discussions have committee chairs, and so on. We argue, then, that administrators, are not confounding variables; their presence in real-world deliberative processes makes them an important part of our design.

Our experimental design is meant to determine if deliberation differs from other decision-making processes. Our deliberation treatment embodies this question, as participants were provided with a set of deliberative principles and an administrator was present in an attempt to ensure that these principles were adhered to. However, the provision of a set of principles and the active efforts of an administrator do not necessarily indicate that deliberation actually took place. A group could ignore the initial guidelines and also the input from an administrator, engaging in an interaction devoid of reasons and excluding one or more members. Conversely, the absence of initial guidelines and an administrator does not preclude the possibility that deliberation occurred; that is, just because some groups were administered the discussion treatment does not mean that they did not spontaneously engage in deliberation. Our results, which we present later in the article, explore these points and show that concern over the confounding effect of the administrator is unfounded.

Like the experimental method, direct observation has its own unique strengths. It allows us to engage in unobtrusive measurement of our variables of interest, in other words, to look inside the process of deliberation. By taping and later watching the interactions in all of our groups, we could evaluate the quality of our operationalization, look at what administrators were actually doing in groups, see the degree of reason-giving and inclusion of groups, and carefully observe the complex processes of interactions that we were so interested in. While our experiment gives us a nice ability to assign causal weight to different treatments, it also reduces a dynamic social process to a single indicator. We took seriously other scholars' arguments that deliberation is a complex dynamic interactive social process (Gutmann and Thompson 2004) and we wanted a way for our study to be able to explore this dynamism. We knew that we wanted to see if deliberation happened, and when it did, what key elements emerged to explain it. While our experimental outcomes allow us to evaluate the existence of a “deliberative difference,” our observational analysis allows us to disaggregate deliberation and examine why this difference might exist. In short, we could examine whether or not the general convergence of deliberation scholars around reasons and inclusion is useful in explaining why deliberation might differ from other forms of interaction. Further, rather than being bound to treatment groups we could look at our groups as engaged in dynamic emergent interactions.

We videotaped our group interactions (using multiple cameras so as to clearly see the expressions of all actors). There were five main questions we sought to answer with our direct observation: (1) Were our administrators operationalizing our treatments as indicated? (2) What affect did our administrators have on groups? (3) Knowing that we could not force our treatment upon groups (Simon and Sulkin 2002), could we explain our outlier groups by our variables? (4) What really happened in these different group interactions? and most importantly (5) Was it really reasons and inclusion that were explaining our outcomes? Direct observation was truly the only way we could address these important questions. In our direct observation we coded interactions relative to their inclusiveness and reason-giving. This coding was done by both authors independently, before the experimental data were analyzed.


Our data collection was conducted over a three-week period in the fall of 2003 with participants drawn from introductory courses on human sexuality and race relations offered at our university. Our sample size emerged as a compromise between the two methods we employed. On the one hand, we needed to recruit enough subjects to give weight to our statistical analysis. On the other hand, we wanted to watch and code the videotapes of all interactive sessions; for every group we recruited we would need to do several hours of coding and analysis. Thus, we decided to recruit no more than 150 subjects for the study. For the entire experiment, 30 lab sessions were run in the three weeks, with 122 participants.

Five white female undergraduate student administrators ran the experiment as part of an independent study project. These administrators were blind to the aims of the study. All groups were the same size (five subjects) and had either one or two men. For each session, subjects were randomly assigned to one of five seats at a round table. The deliberation and discussion treatments were randomly assigned to each of the 30 groups. In the 15 groups where not all five subjects came during their scheduled time, the administrators gave the essay treatment (N= 46), which did not require any social interaction. Six of the groups received the deliberation treatment (N= 30), and nine received the discussion treatment (N= 45).19

The topic of consideration across all groups was the issue of whether there should be segregated group activity fees at the university. We chose this topic because it is controversial on campus and because it affects all subjects.20 Students, including the participants in the research, pay these fees as part of their tuition bill each semester. In essence, these fees are like a tax that goes to support student programs and organizations. The segregated fee system has also been the subject of national controversy, due to a recent Supreme Court challenge by several university law students.21 The topic is not so controversial as to polarize people from the outset; however, it is an issue important to all subjects insofar as it affects student services and pocketbooks.

Across all treatments, subjects were first provided with an information sheet on segregated fees, which they were given five minutes to review. This sheet was a comprehensive summary of the segregated fee program and provided budget information that participants could refer to during the entire experiment. This was an attempt on our part to control for information; in short, actors were provided with all the information about the program so as to minimize the information gains that could be provided by any particular members of each group. This information sheet is similar in both size and scope to information one might receive at the start of a board meeting. The review time is consistent as well; it is not uncommon for members of a board or other decision-making body to hear a brief presentation or receive a packet of information and then immediately begin to talk about it. Subjects in all treatment groups were then given a baseline survey that asked them a number of questions relevant to segregated fees, most importantly, our dependent variable, which was the amount of segregated fees they thought should be collected per student each year.22 For this first portion of the experiment, the administration scripts were the same across treatments.

The treatments diverged upon completion of the baseline survey: those subjects who were in the essay treatment groups were given brief instructions and asked to begin writing an essay on what they thought of the segregated fee system; those in the discussion treatment groups were asked to begin talking about the system without any input from the administrator; and those in the deliberation treatment groups received written guidelines and limited input from the administrator.

All treatments lasted 35 minutes. At the end of all treatments, subjects across groups were asked to complete another survey, identical to the initial baseline survey. Finally, all subjects, regardless of treatment, were asked to complete a survey of basic background demographic characteristics and political attitudes, and, in the case of the discussion and deliberation groups, to evaluate their group members and their group interaction. One might argue that 35 minutes is not enough time for a deliberation. We believe that it is enough time and is, in fact, consistent with most real-world deliberations. Meetings of school boards, academic departments, planning boards, city councils, and a variety of other groups typically follow an agenda with a suggested time for each topic to be considered. It is not uncommon for these groups to discuss and vote on important topics after less than one hour of consideration. Furthermore, as we shall show in our observational data discussion, subjects rarely required more time to engage the topic.

An examination of the video recordings of the experiment treatments revealed that one of our administrators did not follow our protocol in the deliberation treatment. Administrator 4 differed from the other administrators in terms of her performance of the administration (she only administered one deliberation group: Group 7). While the other four administrators engaged the subjects early on, having at least three or four interactions with their groups within the first 10 minutes of the deliberation, Administrator 4 did not make any comments until 10 minutes had elapsed.

Administrator 4 also differed significantly in the content of her interactions with her group; seven of her eight interactions over the total 35 minutes came in the form of questions that were not scripted: these questions did not realize either inclusion or reasons. Only once did the administrator actually draw on any of the principles of deliberation mentioned in the script, when she followed up on a participant's comment by asking, “Why do you say that?” This operationalization of the treatment is in significant contrast to those of the other administrators, who repeatedly asked participants to give reasons. A final difference between Administrator 4 and the other administrators is in the inclusion of all participants in the discussion. In Group 7, one participant did not speak at all for the last 25 minutes of the deliberation and Administrator 4 did not attempt to draw her into the conversation.23 A second member of the group was relatively silent as well. A count of the number of times each participant spoke revealed that two members of Group 7 each individually spoke more than the other three members combined. In other groups the administrators frequently urged participants to join the discussion and present ideas—as was directed by our design. In short, Administrator 4 differed notably from the other four administrators in her operationalization of deliberation, simply putting up questions to her group; the other administrators acted as active catalysts for more robust and inclusive discussion—drawing on our principles of reasons and participation. For this reason, we dropped Group 7 from our quantitative analysis.24


As indicated in Table 2, 72 of the subjects, or 60 percent of our sample, did not change their position25 on how much money they would allocate to the segregated fee system after having gone through their treatment. However, a closer examination by treatment reveals that many of these cases—33 subjects—were part of the essay treatment: that is, those subjects who neither discussed nor deliberated. Whereas 72 percent of the essay treatment group did not change the amount of money they would allocate, for the other two treatments the percentages were notably lower: 50 percent of the discussion group and 57 percent of the deliberative group, respectively.

Table 2.  Difference in Dollar Allocations (Pre- and Posttest) by Subject
TreatmentDollar Amount by Which Subjects Changed Their Position
<−400−301 to −400−201 to −300−101 to −200−1 to −10001 to 100101 to 200201 to 300301 to 400>400
 No. of subjects1 0 3 1 1 33  1 2 1 0 3 
 % of treatment 2% 7%2%2%72% 2%4%2% 7%
 No. of subjects 0 1 2 3 2 22  6 4 3 0 1 
 % of treatment  2%5%7%5%50%14%9%7% 2%
 No. of subjects 2 1 0 2 2 17  2 1 1 1 1 
 % of treatment 7%3% 7%7%57% 7%3%3%3%3%

Subjects both increased and decreased the amounts they would allocate across the discussion and deliberation treatments; there was no general directional tendency in any treatment condition or within any group. As a result, the difference between these mean changes by treatment type did not prove illuminating (they were often likely to cancel each other out). Yet, given that our study only seeks to ask whether or not subjects arrived at different outcomes as a result of social interaction (an individual-level observation), we are indifferent to this aspect of the data. Whereas some theorists speculate that deliberation leads to more just, progressive, or systematic outcomes, our project was simply to test whether it produced outcomes at all. The absolute value27 of the difference in proposed segregated fees before and after the treatments is a measure that gets at whether subjects changed their decision;28 it revealed that there were differences between the treatments. As Table 3 shows, the mean absolute value of the change in view is $73 in the essay treatment (our control), $99 in the discussion treatment, and $129 in the deliberation treatment. Subjects were increasingly likely to change their positions if they received the discussion and deliberation treatments, respectively.

Table 3.  Change in Subjects' Positions by Treatment Type
TreatmentAbsolute Value of Subjects' Change in Position (in Dollars)
Essay (N= 46) $73
Discussion (N= 44)26 $99
Deliberation (N= 30)$129

We conducted a linear regression comparing the amount by which subjects in different treatment groups changed their views. In our simplest model (without any controls) we found that deliberation groups were more likely to change their positions than essay groups ( p= 0.0493).29 In contrast, discussion groups were not more likely to change their positions than essay groups ( p= 0.471). However, the difference between deliberating and discussing groups did not prove to be significant ( p= 0.244). The results of this model show a difference between the deliberation treatment and the control, no difference between the discussion treatment and control, but no difference between the two forms of “social interaction” treatments. In other words, deliberation differs from just thinking for oneself, “just talk” does not differ from just thinking for oneself, but there is no clear difference between deliberation and “just talk.”

We then controlled for administrator effects. When these controls were added, we found a significant difference between subjects in deliberating and essay groups ( p= 0.015) and between subjects in deliberating and discussion groups ( p= 0.037).30 Finally, subjects in discussion groups were not more likely to change their positions than essay groups ( p= 0.948).

These models provide us with two key findings: First there is no significant difference between the discussion and essay treatments but there is a difference between the deliberation and essay treatments. While this is certainly interesting, it is not robust enough to allow for strong arguments to be made about the difference between kinds of communication. However, through the control of administrator effects, we find that it is not just that deliberative processes are different from noncommunicative ones but also that they are different from other forms of communication. In short, there is something different about communicative processes that are inclusive and wherein subjects provide reasons. Figure 1 provides an illustration of our experimental results, establishing strong support for what we believe is a deliberative difference.

Figure 1.

Changes in position relative to essay treatment (control).

None of the subjects' background characteristics had effects on our analysis (gender, race, class, etc.). We would not argue that this is because these variables do not matter for deliberation. Rather, our population simply did not vary greatly along these lines (particularly race and class). Where at all possible, we attempted to keep the gender composition of our groups constant (with one or two men in all groups). Again, this gender imbalance was due to a gender imbalance in the population upon which we drew (introductory sociology courses). Differences in the gender composition of the group did not have observable effects—but again, our sample size was relatively small, making it nearly impossible to observe any. We had such little racial variation that these variables had no observable effects. The class position of the subjects did not affect either the amount of money they were willing to allocate in the first place, nor whether or not they were likely to change their position. Finally, we added a series of questions to our final survey asking who participants thought was the most influential group member, who was the most leader-like, who was the most likeable; in short we asked about group dynamics to see if single, influential subjects explained changes in position and not the content of the social interaction itself. None of these variables had significant effects on individual-level decisions.


The analysis of our experimental manipulations shows that deliberation is indeed a unique kind of social interaction that produces effects. This answered our first question—does deliberation matter—affirmatively. Our observational analysis corroborates these findings and directly addresses our second question: If we disaggregate deliberation into reasons and inclusion do we find that these two mechanisms, as variables in their own right, explain the deliberative difference? We watched the videotapes of all interactive groups in order to ascertain whether or not there was a correlation between outcomes and the provision of reasons in an inclusive setting, regardless of the administered treatment. In order to avoid biasing our coding, we did not look at the outcomes for any groups until after we had finished watching and coding all 15 sessions. The two authors also coded independently.

We coded each group session for its level of inclusion and the number of reasons given. We believe inclusion means more than simple participation. Clearly, one must speak to be included in a debate and so we decided to keep a running count of who spoke and for how long. But while talking is one part of including oneself in a group interaction, it is also important that one's comments be on topic. For this reason we also took notes on the content of what participants actually said.31 Inclusion is also reciprocal; the efforts of group members to draw others into the debate are just as important as the response they may engender from others. Thus, we counted the number of times individuals asked the opinions of others or referenced previous points made by another in the group. We used all of these data to develop an inclusion score for each group, ranging from one to five, with one representing a low level of inclusion and five representing a high level of inclusion.32 After watching each videotaped session, we reviewed our notes and independently scored the session for its level of inclusion. We then compared scores; nearly all of our scores were within 0.5 of each other. When we differed, we used the mean of our two numbers as the final inclusion score. Table 4 summarizes our coding. There was little question in our minds as to which groups were the most and least inclusive: we gave identical scores of five to the three most inclusive groups and a score of one to the least inclusive group.

Table 4.  Summary of Coding and Scores
GroupTreatmentNumber of ReasonsReason Score34Inclusion ScoreChange in Position
14Deliberation23 5 5 $320
28Discussion16 3.55 $250
18Deliberation18.54 3 $180
 8Discussion19.54.25 $160
 6Discussion12 2.62.5$140
 5Deliberation19.54.24 $130
 2Deliberation15 3.34 $100
23Discussion 4  .92  $70
 3Discussion 8.51.82  $61
21Discussion 5.51.22  $60
 1Discussion13.52.91  $57
10Discussion10 2.22  $40
13Discussion11.52.52  $38
 7Deliberation11 2.42.5 $15

We also examined the degree of reason-giving in each group. Reasons are more than just opinions; they offer a justification for a stated position related to the topic under debate, an answer to the question: “Why do you say that?” From our point of view, reason-giving is not about people talking more, it is about people talking differently. The “difference” of reason-giving is justification of one's position. We did not make judgments concerning the validity of reasons offered. We also counted reasons even if they had already been offered earlier in the interaction. One might argue that not all reasons are created equal. This is a fair point but we are convinced that even the most skeptical of critics would acknowledge that the reasons offered were typically reasonable; no one invoked psychics, elves, or vague sentiments in making their points. Further, if one is to be inclusive, one must respect the variety of ways that subjects reason; we are weary of models that impose external normative criteria for “good” reasons. While watching each group session, we independently counted the number of reasons given and compared our numbers at the end. As in the case of the inclusion score, our counts were very consistent with one another; in the few cases where we differed, we took the mean of the two numbers. The highest number of reasons given by any group was 23, while the lowest number of reasons given by any group was four.

Our results are clear and support our findings in the previous section; more importantly, they point to the importance of reasons and inclusion for explaining the deliberative difference. As Figures 2 and 3 and Table 4 show, regardless of treatment the more a group deliberated (provided reasons and fostered an ethic of inclusion), the more participants were likely to change their dollar allocations for segregated fees. When groups did not engage in robust deliberation, regardless of whether or not an administrator was present, participants were less likely to alter their allocations.33

Figure 2.

Reason-giving and change in position.

Figure 3.

Inclusion and change in position.

This finding also helps explain the outliers in the analysis of our experimental manipulation. Discussion treatment groups with relatively high changes in fee allocations typically had high levels of reason-giving and inclusion. Deliberation treatment groups with small changes in fee allocations typically had lower levels of reason-giving and inclusion. Finally, it was not just the presence of an administrator that produced our effects. Rather, reasons and inclusion are central to explaining our outcomes. If we look at our two discussion groups wherein subjects were most likely to change their position (Groups 8 and 28), we find that these two groups have the highest reason-giving and inclusion scores of all discussion groups, and similar scores to “deliberating” groups. Similarly, if we look at the deliberation group in which subjects were unlikely to change their position (Group 7), we find that this group had low levels of reason-giving and low levels of inclusion. If we look at the discussion group where subjects were least likely to change their position (Group 13), we find that this group had the lowest level of inclusion and a low level of reason-giving. Finally, if we look at the deliberation group that had the highest amount of position changes (Group 14), we find that this group had the highest amount of reason-giving and was the most inclusive (see Table 4 and Figures 2 and 3). The variables that we identified, observed, and measured (reason-giving and inclusion) were instrumental in explaining the difference between the groups that failed to deliberate (and should not have) and the ones that succeeded at deliberating (and should not have). While the administrator was instrumental in operationalizing deliberation, she could not force this social process upon groups, nor could the lack of her presence prevent it from organically emerging. To make this point starker: it is reasons and inclusion that explain the differences in our results, not the presence or absence of an administrator. This result also points to some of the dynamism of deliberation. While we believe our work directly challenges Sulkin and Simon's view (2001) that deliberation cannot be manipulated in an experimental setting, we also believe that this observational analysis provides clear support to those theorists who emphasize the dynamic emergent character of deliberative processes. It also provides support for the general convergence around reasons and inclusion for explaining the deliberative difference.


Finally, in order to understand how individuals thought through our topic outside of interpersonal interaction, we did a content analysis of the essays that the control subjects wrote. The content and direction of the essays were almost all unilluminative. For the most part, subjects stated and then defended their positions; we found little correlation between number of reasons given (or speculation of others' positions) and changes in position. Furthermore, four of the subjects accounted for almost all of the difference in outcomes. As such, we can offer little support for Goodin (2000, 2003) or Dryzek's (2000) belief in autodeliberation. Admittedly, we did not instruct subjects to engage in a “personal decision process, in which the individual mulls things over in her or her mind,” as Dryzek suggests can/should be the case with what he calls “discursive democracy.” However, in reviewing the content of essays we found that almost no subjects mulled over the topic or had an internal debate. No doubt this is affected by the fact that our subjects were college students. When asked to write an essay, many defended rather than evaluated their position—this is consistent with what we would expect from students who are trained in this kind of argument defense within college. We are not in a position to argue against Dryzek and Goodin as we did not directly test their ideas; this could certainly serve for a future project. The near complete lack of evidence of autodeliberation within our essays leads us to believe that such a process is unlikely to yield different decisions. As we noted above, spontaneous deliberation was possible within our interactive treatments, yet we found no evidence of autodeliberation in our essays. Given that this was not the primary focus of our empirical investigation, this point should not be taken strongly.


This article asked two questions: (1) Is there any deliberative difference—that is, does deliberation differ from other forms of social interaction in terms of outcomes? (2) Are reasons and inclusion important—that is, is the general convergence around reasons and inclusion within the deliberation literature warranted for explaining the deliberative difference? We found evidence that deliberation does produce outcomes different from other kinds of social interaction. Further, when reasons are given and people are included, participants in the group interaction are more likely to change their positions. This has multiple implications for the theoretical and empirical approaches to deliberation.

First, our work provides support for the general convergence around reasons and inclusion for explaining the deliberative difference, notably in the work of Rawls and Habermas and most recently in the work of Gutmann and Thompson (2004) and Steiner et al. (2004), among others. But we also found that it is not enough simply to impose those conditions on individuals in a discursive context. There are informal norms and practices that allow a group designed for one kind of communication to follow another, at least to some degree.35 This provides support for observations about the emergent quality of deliberative democracy. This “emergence” as well as the evidence on the specificity of reason-giving and inclusivity for the deliberative process provides valuable evidence for many deliberative scholars who explicitly or implicitly rely upon these ideas (Baiocchi 2001, 2003; Bohman 1996; Eliasoph and Lichterman 2003; Fishkin 1991, 2003; Fung 2001; Gutmann and Thompson 1996, 2004; Heller and Isaac 2003; Lichterman 1996; Luskin et al. 2002; Mansbridge 1999; Polletta and Lee 2006; Thomas 2001; Young 1999, 2000).

Second, we would encourage those who are interested in strengthening the empirical foundations of deliberation to move beyond a holistic approach to deliberation. As our results show, we can indeed identify key mechanisms within modes of interaction and test them as variables in their own right. This may be the most important point to emerge from our study. Empirically oriented social scientists, regardless of their theoretical orientation, typically have been all too willing to treat the social interaction they call “deliberation” as a black box; we hope to have shown that this is a mistake. Modes of social interaction are complex but not necessarily undecipherable. If we think different social interactions might produce different outcomes and we identify mechanisms that we think produce such differences, then we should do the work of social science and test these ideas. Such testing does not need to happen through experimentation; social scientists have many tools to evaluate theoretical propositions. Using both experimentation and direct observation, we are able to claim that as reasons and inclusion increase, so does the likelihood of actors changing their positions.

Third, we think that our work should cause those who situate themselves within the rational choice camp to take note. There are scholars who deploy rational choice theory when engaging with deliberation (Austen-Smith 1992; Johnson 1991, 1993; Knight and Johnson 1994; Shapiro 1999). We do not take the position that they must go searching for a new theoretical frame.36 But we do think we are in a position to ask what it is about reasons and inclusion that makes actors more likely to alter their positions (especially given our attempts to control for information). For rational choice theorists of deliberation, a position change may result from the acquisition of additional information about the situation concerning preferences or potential consequences (Austen-Smith 1990:125, 1992). New information may change what an actor does insofar as it presents the actor with a new opportunity structure. Through the communicative process, actors also may become cognizant of effects on others and act accordingly. For example, in their quasi-experiments on deliberative polling, Luskin et al. (2002:484) suggest “some participants [in deliberation] may come to see the alternatives through different eyes.” Participants, in becoming aware of the effects of decisions on others, may change their positions. In this model, deliberation is a way to increase information gains through reason-giving and inclusion and therefore affects individual decision making. We acknowledge that reasons and inclusion could simply be folded into (or stretched to fit) the rational choice framework; a participant in a deliberation provided with a reason (i.e., information) from a variety of actors (inclusion increasing information) may come to some sort of realization about the situation and/or possible effects of future action. This is an empirical question that could—and we argue should—be studied. In general, we argue that these two factors must be taken much more seriously by those adopting a rational choice frame. Overall, we find the literature to be less than clear on the similarities and differences between inclusive reasoning and simple provision of information in a rational choice framework. Our results indicate that this lack of clarity needs to be addressed, and the importance of reasons and inclusion explained within this framework.

Fourth, although this article engages with those who work with the idea of deliberation, there are those (Shapiro 1999; Stokes 1998) who think this concept has been overused and that in fact it explains very little. We hope to have shown skeptics that there is something to the deliberative process, that it is something worth paying attention to, and something that is of potential significance. While a quick look at our variance shows that we are not explaining a lot of what is going on in the social processes we are exploring, we still find that deliberation differs from other forms of interaction in terms of position formation. In short, there is a deliberative difference. We have been silent on what this difference is (only asserting that it exists). The content of the difference is certainly a realm for future empirical inquiries.

Fifth, we feel it important to address the policy implications of our results. In terms of outcomes, simply discussing an issue is not very different from not discussing it at all. The political implication is that in social gatherings, talk is cheap. Discussion unmotivated by reason-giving and inclusiveness is unlikely to produce different results than just allowing members to consider the matter on their own. For those who have sat through meetings ruled by “just talk,” this may come as little surprise. We agree with Conover and Searing (2005) that “everyday political talk” is a significant element of our democracy and that people are “wary” of reasons that differ from their own views. This is precisely the point we are trying to make. Positions change when participants are asked to provide reasons and be inclusive of all members; these processes make subjects think. While we cannot make claims about the content of what these changes should (or are likely to) look like, we do not believe that this is a bad thing. Social scientists have a less than stellar record of accomplishment in generating “should” statements about social processes. But we do feel confident in claiming that groups that seek to meet and make decisions ensure their meetings are marked by reason-giving and an ethic of inclusion; if they are not, our evidence suggests that their decisions are unlikely to differ from the ones members would have made had they not met at all. Given that our study is a relatively small empirical exploration, we do not wish to overstate this point. Before strong arguments can be made against modes of interaction that are not inclusive and do not include reasons, we must do more work. However, this first study has begun this work, and we do not want to underplay the implications it has for democratic processes. While just talk may be cheap, encouraging interactions that are inclusive, where included members provide reasons, and where all members listen to a variety of reasons for differently situated actors, is valuable.

Finally, given what we have learned from this relatively small study on decision-making processes, we feel that we are in a position to do what all scholars tend to do: ask for more work. Let us be more concrete. We are not alone in arguing that a stronger connection between theoretical and empirical work would certainly strengthen both fields (Bachtiger and Steiner 2005; Rosenberg 2005). In our case, philosophers suggested that particular processes mattered and that we would be likely to observe particular (just) outcomes; empiricists tended not to test the very foundation of the philosophical claims, often ignored the processes, and were likely to share the normative view of the emancipatory potential of the claims. We found that, contrary to their practice, social scientists could evaluate mechanisms like reasons and inclusion, that the foundational claim of deliberation's importance did matter, and that the normative view of “increased justice” needs to be more fully examined. In short, the dialogue between empirical and theoretical or philosophical work proved fruitful; it provided both camps with evidence that requires evaluation. In many ways our work has provided the social science community with more questions than answers, and we hope that other scholars will join us (and debate us) as we continue to answer some of the questions around deliberation. Given the importance of different decision-making processes to human communities, we feel that the impact of continuing such work should not be underestimated.


  • 1

    When referencing the deliberation literature, we include neither the “juries” literature, nor the “argumentation” literature grounded in semiotics, both of which ask fundamentally different questions.

  • 2

    Our dependent variable is consistent with other studies of deliberation and decision making (e.g., Sulkin and Simon 2001) that use dollar allocations to measure outcomes. This kind of decision also has strong external validity, as deliberative bodies such as school boards and budget committees are often forced to make decisions about “how much” to allocate to certain programs.

  • 3

    For example, Steiner et al. (2004) offer up to six key characteristics for their ideal type of deliberation.

  • 4

    They add two additional conditions beyond reasons and inclusion: (3) that decisions are implemented and binding, and (4) that such decisions can be changed if need be. As we are concerned with the decision-making process (deliberation), and not the implementations of that process (deliberative democracy) these later two conditions are beyond the concerns of this project. Steiner et al. (2004) go further, developing an even more complex index that comprises essential characteristics for deliberation. Again, some elements are beyond the concerns of our project (content of justifications, and constructive politics) and the rest (participation, level of justification, and respect) converge in content to our own conceptualizations.

  • 5

    By accessible, Gutmann and Thompson mean both understandable and inclusive. Other scholars have emphasized reason-giving and “story-telling” (Mansbridge 1999; Polletta and Lee 2006; Young 1996, 2000). We appreciate these contributions and simply note that “story-telling” is used in these arguments as a way for otherwise silenced or disadvantaged groups to be heard. In short, it is a form of inclusion.

  • 6

    Habermas's model is, admittedly, more complex than this simple formulation but we believe that our formulation is fundamentally sound. Some of the complexities of the Habermasian model that we have minimized are so because of the difficulties empirically of evaluating these ideas, particularly “the three world relations of actors and the corresponding concepts of object, social, and subjective truth” (1990:137).

  • 7

    For the relationship between Habermas and Rawls (and their reliance on Kant), see McCarthy (1994). See also their debate in Habermas (1995) and Rawls (1995).

  • 8

    We agree with their critiques, but it is beyond the scope of our study to address them. For more on these critiques, see Baker (1995), Okin (1989), Seidman (1997), Young (1996, 2000), Sanders (1997), Bickford (1996), and Mansbridge (1999).

  • 9

    Baiocchi (2001, 2003) explores the ways in which city budgets are generated within the city of Porto Alegre, Brazil. In brief, districts within the city hold community forums to decide how to spend portions of the budget. These broad inclusive meetings, Baiocchi shows, invoke deliberative processes. Baiocchi thus outlines the ways in which deliberation over budgets (or participatory budgeting) can produce new vibrant democratic forms.

  • 10

    For example, participatory budgeting in Belo Horizonte has not had the effects that we observe within the Porto Alegre case; it is the conditions in Belo Horizonte, not the participatory process, that differ.

  • 11

    Steiner et al. (2004) is one of the few works to point out the lack of quantitative analysis in empirical studies of deliberation. Their work on developing a discourse quality index (DQI) is a significant step in developing a reliable measure of the deliberative process. Their conclusion calls for experimental evaluation of deliberative processes.

  • 12

    There are several recent works on deliberative polling; we deal primarily with Luskin et al. (2002) as we think it offers the most comprehensive theoretical treatment of deliberation. For more on deliberative polling, see McCombs and Reynolds (1999), Fishkin (1991, 1995, 2003), Ackerman and Fishkin (2004), and Fishkin and Luskin (2005).

  • 13

    Ackerman and Fishkin (2004) devote one page to the importance of “deploying the arts of political reasonableness” in order to be “effective” in deliberation. We are not quite sure what they mean by “effective” in this case. Furthermore, their evidence for the importance of reason-giving comes not from any of their work on deliberation, but rather from “evidence from the operation of the Anglo-American jury system” (2004:184). While we agree with their assertion that deliberators can be reasonable in a political context, we question their call for reasons (and inclusion, to some extent) without empirical basis. Again, these characteristics of “good” deliberation seem to be axiomatic.

  • 14

    We do not simply object to the computer setting of Sulkin and Simon. Rather, we find Polletta and Lee's work (2006) on “virtual” interactions to be a convincing form of interaction. The objection is that not all forms of discussion are deliberation; our results prove this concern to be legitimate.

  • 16

    The division of deliberation and discussion into separate treatments challenges the arguments of Fearon (1998).

  • 15

    The empty cell in the table could be the subject of another test. For example, Dryzek presents the concept of “discursive democracy” in which an individual can engage in a personal decision process. Goodin (2000:79–107, 2003:170–92) makes a similar point with his call for internal deliberation (within each subject) as central to the deliberative process. Goodin seeks to move the literature beyond deliberation as a process and argues within an “internal reflective mode” we can better realize ends. “Deliberation is supposed to have an end it is supposed to resolve something” (2003:192). The empirical evidence on the value of internal deliberation is unclear; the experimental work of Dijksterhuis et al. (2006) challenges Goodin's ideas on ends. We argue that deliberation is both about ends and processes; we test both. The question of autodeliberation, or an “internal-reflective mode” is indirectly addressed by our essay treatment. In analyzing the content of our essays we found little to suggest any internal deliberation took place.

  • 17

    We adapted these principles, which have their origins in the work of Rawls and Habermas, from Fung and Wright (2003:17). They embody what we identify as central tenets of the literature on deliberation.

  • 18

    Shapiro (2002:203) argues that “significant though reason-giving is to legitimacy (particularly in unelected institutions in a democracy), it does not capture the essence of deliberation … Deliberation is about getting the right answer.” Shapiro sees deliberations that do not reach the “right answer” as being a waste of time. We are both very close and very far away from Shapiro. He does not completely remove reasons from the equation. He argues that reason-giving is not where the action is; however, reason-solicitation is essential. We are open to this possibility, but ask that such a distinction be tested. Our results simply speak to reasons—they do not isolate whether or not they are given or solicited. When we speak of “reason-giving” we also mean “reason-solicitation.” Our qualitative counts of “reasons provided” included both of these phenomena. Unlike Shapiro, we have no expectations about a “right answer.” We are not even sure we could identify one over others—this requires the imposition of a secondary normative theory that goes well beyond deliberative processes. Finally, Shapiro is silent on the inclusive aspects of deliberation. From our position, this is a problem.

  • 19

    The number of treatment groups differs because of random assignment. One essay and one discussion subject did not fully fill out the instruments we used (hence these numbers add to 120, not 122). The discussion group that this member was in was not thrown out, but this one subject was not included in the quantitative analysis; the subject was included in the observational analysis.

  • 20

    Hence we take seriously Young's (1999:155) point that deliberation should include representatives of all those in the public who “might be affected by decisions” made by the group.

  • 21

    Board of Regents v. Southworth, case number 98–1189. The court, on March 22, 2000, upheld the university's student fee system.

  • 22

    The exact wording of the question was: “How much money do you think students should have to pay per year in segregated fees (Range: $0–2,000)?” All students were informed of the amount they presently pay each year: $506. We provided an upper bound of $2,000 because this is approximately what the fees would be if all groups received all the funding they requested. No subject neared the upper bound in their allocation; very few subjects selected the amount they presently pay—indicating that many believed they pay either too much or too little.

  • 23

    Again, we had a scripted statement for administrators to use so that they could draw silent members into the discussion.

  • 24

    As is explained below, this group is not dropped from our qualitative analysis.

  • 25

    “Change their position” refers to whether or not (and to what degree) subjects altered their monetary allocation for the segregated fee program. That is, it simply looks at the difference in the amount of money subjects thought should be allocated to the system before the treatment, and then again after the treatment.

  • 26

    As one subject did not fully fill out the instrument, this subject was removed from the analysis. The group this subject was in, however, was kept.

  • 27

    This absolute value was calculated by taking the dollar amount the subject put on our instrument after the treatment, and subtracting from this the dollar amount the subject put on our instrument before the treatment (these before/instruments were identical), then taking the absolute value of this number. It marks a change in position and does not allow opposite individual changes to cancel each other out within groups. This absolute value for each subject is our primary dependent variable for this study.

  • 28

    Our language is again deliberate. We make no claims about whether or not people changed their minds (or preferences); this is beyond the scope of our design. The choice of a specific dollar amount constitutes a decision, or outcome.

  • 29

    All reported p-values are for two-tailed tests.

  • 30

    The reader may note from Table 2 that there are some subjects who are outliers. If these outliers are removed from the analysis, the results are still significant.

  • 31

    In taking notes, we were primarily interested in whether or not the speaking was on topic. For example, one participant spoke frequently but about unrelated issues, such as taking his roommate to lacrosse practice. We do not consider this reason-giving, nor do we consider it participation, let alone inclusion.

  • 32

    This was not a subjective scale. It involved counting the number of times each member talked and/or referred to another's idea. An inclusive group would have each member having roughly the same level of activity and/or speaking directly to other members' comments; a noninclusive group would have either some subset of members dominating the conversation, one member being relative silent, and/or having “talk” that was not directly related to other members' comments or the topic of consideration.

  • 33

    Group 7, which was dropped in the quantitative section, was included in this analysis. It proved to have a low level of inclusion (a score of 2.5 out of 5) and a low number of provided reasons (11), and few changed their positions significantly.

  • 34

    Reason count converted to a five-point scale.

  • 35

    Thanks to an anonymous reviewer for making this clear to us.

  • 36

    We understand that the Kantian-inspired literature on deliberation, with its emphasis on just outcomes and convergence, is some distance away from work on decision making relying on a rational choice frame. That said, we choose to emphasize common ground across the two approaches, particularly in terms of the process of deliberation. Even those who are far from being rational choice advocates look to information. Manin (1987) emphasizes information insofar as no actor can envision the variety of perspectives presented within political decisions, and no actor can possess all the information necessary to make a decision that affects all.