Identifying Deliberation in Social Movement Assemblies: Challenges of Comparative Participant Observation

Contemporary social movements can serve as a critical case for the empirical study of deliberation. In countless face-to-face meetings activists often discuss long hours before a decision is reached. In this context, we try to analyse the conditions under which deliberation is successfully employed as a method of discursive conflict resolution. As we develop participant observation in a comparative approach we encounter three methodological challenges which this paper addresses. First, we look at some characteristics of the global justice movements, briefly addressing the different settings in which controversial discussions occur. Second, we give a rationale for applying a semi-standardised multi-level participant observation in order to allow the collection of comparable data by various researchers in several countries. Focusing on participant observation on the level of controversial discussions we thirdly conceptualise competitiveness, power, and asymmetry as three theoretical dimensions to identify eight different practices of discourse, one of them being deliberation. We are currently implementing this model for regular observations of group meetings on a local, national and European level. First results should be available in the near future.


Introduction: Deliberative Democracy and the Global Justice Movement
Deliberation as symmetric, cooperative problem solving through arguments is on a rising tide in studies of democracy. While being discussed as a normative utopia in the first stages, many empirical studies now search for potentials of deliberation in the real world (Chambers 2003;Ryfe 2005). Major political changes have contributed to this trend. Liberal representative democracy as the predominant model of western politics has long been criticised for in fact being a "polyarchy" (Dahl 1971). Since the 1990's political and social changes fuelled this criticism and representative democracies increasingly suffer from a lack of citizen's participation. Throughout the western world, formal political organisations like parties and trade unions continue to suffer dwindling membership. Voter participation decreases or stagnates (Norris 2002). Additionally, economic globalisation undermines democratic control through the nation state (Brand et al. 2000: 74ff).
In the context of such developments, deliberative democracy has become an appealing idea, since it places a participatory process of opinion formation and the quality of public discourse, not formal voting procedures, in the heart of democracy (Chambers 2003: 307). However, it is disputed whether deliberation is the only democratic form of discourse. Some authors claim that protest, partisan movements and political struggle challenge the idea of deliberation (Sanders 1997;Snyder 2003;Young 2001) and that consensus -which is the normative aim of deliberation -may not always be the most democratic method to take decisions (Mansbridge 2003). Others have argued that the cooperative design of deliberative politics disguises the true struggles which are necessarily at the core of 'the political' (Mouffe 1996). Furthermore, research in small-groups has shown that deliberation is "hard work" (Holt 1993(Holt , 1999 and that equality amongst discussants is hard to achieve even within groups which explicitly reject hierarchies (e.g. Gastil 1993a).
Of course, these problems should be taken seriously, but the basic notion of deliberation is still worth considering: Using arguments to influence collective decision-making rather than sanctions that are based on economical or physical force does have a democratic effect. This idea bears great potential in view of the above mentioned crisis of real existing democracy. The international research project "Democracy in Europe and the Mobilization of Society" (DEMOS) conducted in six European countries, in which the authors of this paper are actively involved, combines the emphasis on discourse in democratic theory with research on social movements which are often considered not only as promoters but also as protagonists of democracy (della Porta 2005;Ibarra 2003).
Given the fact that representative democracy fails to include a growing part of the populace, we propose to focus on those arenas where citizens engage to articulate their visions of politics. Social movements are certainly the most important forums for an immediate expression of political claims. Thus, they can be considered to be ideal cases to analyse the potential of deliberation. Social movements have a long tradition of discursive decision-making and identity building (Polletta 2002). It has been argued, that especially the global justice movements (henceforth GJMs) which evolved in recent years have developed a special deliberative culture, characterised by a high acceptance of difference, willingness to listen and learn from each other (della Porta 2005: 79-91) as well as a critical attitude towards power politics.
However, the potential for deliberation in social movement groups cannot be taken for granted. It might well be that other forms of communication such as bargaining play an important role. Our approach is to look at these political groups and networks engaged in the GJMs as laboratories of participatory and discursive democracy. It is then an empirical question to which extent these real world experiments can be considered as deliberation. Methodologically speaking, the GJMs are a "critical case" (Yin 1989: 38-40). If deliberation is not possible under the very advantageous conditions of the movement culture described above, democratic theory will have to be more critical in promoting deliberative democracy (West & Gastil 2004: 2). We suggest using the method of participant observation to identify the potential for deliberation within the GJMs and to assess the relevance of other types of communication.
In researching the discursive practices of groups within these movements empirically, we encounter a number of methodological problems and practical obstacles. In this paper we try to deal with these by developing a semi-standardised form of participant observation suitable for a comparative research design and applicable to any type of controversial discussion. We will address three major methodological challenges: First, we address the question where the discourse practices that we are interested in can be observed within the GJMs. Since already the notion of a social movement is cloudy we take a closer look at the arenas within movements where deliberation takes place. This leads to our second question: How can we compare the results of participant observation conducted by different researchers in various groups in a transnational enterprise such as the DEMOS project? In section 3, we suggest to tackle this problem by using a semi-standardised form of participant observation. The focus of our effort -to distinguish and define different types of discourse -is conceptualized in section 4. We ask: What type of interaction has to occur so that we can speak of deliberation as opposed to other forms of discourse? Section 4 describes the concepts of power, asymmetry, competitiveness and participation and argues that they can be regarded as the core theoretical dimensions of previous efforts to define deliberation. Section 5 finally faces the task of operationalising these dimensions so that participant observers can recognise them when observing discussions in the field.

Mapping the field: Communicative Spaces in Global Justice Movements
The new social movements that emerged in Western Europe and North America during the second half of the 20 th century have been identified as an important locus to develop a "democracy from below" (Koopmans 1995). In these movements and the succeeding GJMs citizens have been organizing themselves to put forward problems that are neglected in the institutional political process. The interpretation of political problems and potential ways to gain leverage are discussed here with an emphasis on democratic norms such as equality, participation and recognition of the other. In so far, the aim to analyse the activity of the GJMs in search of deliberative practices is a promising approach. But what does that mean exactly? Who are the GJMs? Where does deliberation take place in this context? And how do different contexts affect the actual deliberative practices?
The GJMs are characterized by a double structure. One component of the GJMs consists of a set of single movements (e.g. environmentalist, women's rights, and international solidarity movement) converging selectively for mobilization campaigns against neo-liberal policies. Thus, they form what has been called a movement family (della Porta & Rucht 1995). These movements "have different specific objectives but share a similar world view, overlap in membership and frequently work together in protest campaigns" (della Porta & Diani 1999: 148). Each movement itself is a network of groups and individuals, held together by collective identity and mobilising for protest in order to support or oppose political or societal change. For these activists, 'global justice' is usually not a self-contained concept with a binding identity but just one aspect of their agenda or a notion to (master-)frame their individual commitment. As a second component, new organisations and forums have been developed during the last decade, which define themselves primarily as "critical to globalization" (in German) or "altermondialist" (in French), e.g. People's Global Action, Attac and Social Forums. 1 In these arenas the notion of global justice is a prime mover and a likely basis to form a collective identity.
The first step to identify loci of deliberation in the GJMs is to take a look at those incidents to promote global justice that are publicly visible, mainly protest events, public education and movement events such as counter-summits or social forums. By going public, social movements try to convey their messages to a larger public. This process of feeding in alternative knowledge to a broader discursive setting can be seen as an attempt to deliberate. In fact, this has been the perspective of most theories of deliberative democracy emphasising the role of civil society actors in public deliberation (e.g. Habermas 1996 Ch. VIII;Dryzek 2000; for a focus on social movements see Medearis 2004). But social movements do notoriously have a weak standing in public discourse and arguments might be presented in a confrontational manner so that the notion of deliberation as an exchange of arguments inter pares can be applied to this public activity only in an abstract manner (Gerhards 1997: 30-32). In addition, parts of the GJMs reject the very idea of dialogue with a hegemonic culture that in their view integrates dissident voices at the cost of their radical impetus (Glasius 2005).
As a consequence, we chose to seek for the deliberative substance of social movements on a different level than the general public. To find the laboratories of participatory democracy one has to access the backstage 2 of the mentioned public events. The events forming the public image of the GJMs are organised by hundreds of political groups. They serve as the movements' 'backbone' with their personal networks, meetings, mailing lists and alternative media. These communicative arenas evolving in their everyday practices are not only backstages with regard to the general public but can also be regarded as frontstages in their own right (Haug 2006). In this respect, face-to-face group meetings form the main stages for deliberation within the movements (West & Gastil 2004: 2). More spe-1 We follow Rucht (2002: 18) in referring to the global justice movements in plural because of this polymorphic structure. The term "anti-globalization movement" does not make sense to us, because these movements do not reject globalization as such. They rather conceive themselves as part of another globalization that does not follow neo-liberal imperatives.
2 For the distinction of front-and backstage see Goffman 1959. cifically, it is likely that the marrow of deliberation -defining norms, developing ideas and making decisions -evolves particularly during controversial discussions within these arenas and not in interactions with the authorities, mass media or other audiences.
In order to understand how the conditions for deliberation vary in movement meetings we want to briefly point out some contextual parameters we consider relevant as independent variables when we study deliberation as a dependent variable. These contextual conditions will have to be considered when choosing the arenas to be studied and compared so that sufficient variance of these factors can be achieved. To start with, the distinction of micro-and mesomobilisation (McAdam 1988;Ohlemacher 1992;Gerhards & Rucht 1992) can be helpful to describe the arenas of discourse in a social movement. Discussions are likely to be different in homogeneous (micromobilisation) groups than in mesomobilisation groups composed of representatives of different movement groups. Secondly, group research (e.g. Schäfers 1994; Neidhardt 1983) can bring helpful categorisations to distinguish between different group properties. Besides the mere size of the group, questions of the stability, intimacy and cohesion of a group, thirdly, also procedural and cognitive variables are likely to effect group discussion. Some empirical studies on deliberation, especially in social psychology, systematically take up the question of moderation, decision-making-rules and other procedural norms (e.g. Gastil 1993b;Ryfe 2005;Trénel 2006).
Moreover, different organisational ideologies (or cultures) can be considered as context factors. The meetings of formal and bureaucratic organisations like trade unions may lead to discourse between officials representing different sets of client interests. By contrast, the meetings of networks adherent to grassroots organisation may consist of self-selected individuals, which speak for themselves and not for others. Also, the question of political ideology, most notably the difference between radicals and reformists, is an important structuring force that should be considered in the analysis of discussions within the GJMs (For a combination of the organisational and ideological dimension see Rucht, Teune & Yang 2007: 165-166). Lastly, language diversity and cultural differences will affect the quality of discourse (Doerr 2006(Doerr , 2007 so that an analysis of deliberation within the GJMs without cross-country and -language comparison is obviously limited in scope. This multitude of factors influencing the character of discussions in social movement arenas is reflected in relatively persistent characteristics on the group level. This means that at least some of those factors can be controlled through the selection of the groups to be studied. If we furthermore include the various conditions which are subject to change from session to session (of the same group) and from controversy to controversy (within each session), we get a vast number of factors which can hardly all be studied systematically with a rather limited number of cases. 3

Soaking and Poking: Comparative Participant Observation
If we consider the face-to-face meetings of groups and networks active in the GJMs as an important site for a deliberative democracy elaborated from below, we are confronted with the methodological question of how to gather relevant data. Two questions are of paramount importance in our research: (1) How do participation, deliberation and decision-making look like in different kinds of groups, different countries and at different levels -from the local to the transnational? And (2) what factors facilitate and restrict deliberative/participatory practices? The debates taking place in face-to-face meetings are transient events so that there is a need to record these instances in one way or another. Only some social movement actors produce minutes of their debates, and as much as interviews with activists they are arguably (partially) biased by the involvement of the minute-taker (or interviewee). For the given purpose, it makes sense to get in closer touch with the subjects under study and collect the required information at the spot. As a form of gathering data "in which the researcher observes and to some degree participates in the action being studied, as the action is happening" (Lichterman 2002: 120), participant observation seems to be the appropriate method. The immediate inclusion of the researcher in the field, famously labelled as "soaking and poking" by Richard Fenno (1978: 249), allows discovering unexpected aspects of the object 3 The authors of this paper are currently studying groups on the local (Berlin), national (Germany) and transnational (Europe) level, e.g. the Berlin Social Forum, Attac Berlin, the German Social Forum, Attac Germany, the European Preparatory Assemblies for the European Social Forum, meetings of the European Attac network and others. In the DEMOS-project teams in six European countries are studying local and some transnational groups. When choosing the groups, the ideological and organisational ideology as well as the micro-/meso-mobilisation distinction (heterogeneity) has been taken into account. Since the research is bound to groups within selected countries in Europe it should be made clear that it cannot be taken as typical examples for other parts of the world. under study. Participant observers can use all their senses and register behaviour that is not written down in documents or recorded on video or audiotape. This is particularly important as we assume that an analysis of varieties of communication has to include both a non-verbal component and the knowledge of a particular group style (Eliasoph & Lichterman 2003). Early tests in our project have shown that the interpretation of communication is highly dependent on the observers familiarity with internal habits and conflicts.
Participant observation as used in anthropology is a qualitative method in which data collection and data analysis continuously influence each other allowing the individual to adapt his or her methodology depending on the necessities of the field. However, our aim is to compare various groups in different contexts in order to identify potentials for deliberative democracy. With an adoptable design, this will be hard to achieve, especially when there is more than one researcher involved. Even in cultural anthropology -a long-standing domain of participant observation -comparative designs are "largely neglected and thus underconceptualised both theoretically and methodologically" (Kaschuba 2003: 341, own translation).
Nevertheless, to develop a method of participant observation of deliberation suitable for a cross-country comparative research we can rely on empirical research on deliberation and studies applying participant observation. Part of the former work helps to conceptualise a contextual framework, namely a set of variables influencing the occurrence and quality of deliberation (e.g. Steiner et al. 2004). In addition, tools have been developed to analyse the process of deliberation itself (cf. several contributions in Bächtiger & Steiner 2005). However, deliberation as a symmetric, cooperative and open-ended exchange of rational arguments has not yet been analysed in vivo but rather on the basis of off-or online documents (for an overview see Trénel 2004;Janssen & Kies 2005).
Although we can draw on insightful categorisations from this sort of empirical research, the particularities of studying deliberation as it actually happens draws a line to the methodology of participant observation. Single-case studies have been dominating the use of participant observation not only in anthropology but also in the study of social movements. 4 For participant observation that deals with one case, Lichterman distinguishes between the inductive "field-driven" approach and an deductive "theory-driven" design (2002: 121-125). The objective of the former as an explorative method is to "elucidate an empirical unit or subject matter" that has been terra incognita before (ibid: 122). By contrast, the latter treats the object in the light of a pre-existing theoretical framework. This is obviously the approach that we are using if we look for a pre-existing concept of deliberation in the communication of social movement groups. When dealing with a single case, insights are drawn from this in an "extended case" logic to explain a general socio-cultural pattern (Burawoy 1998).
While such observations limited to a specific field/group have taught us a lot about the way meaning is produced collectively and the specific manner social groups adapt to their environment (Lichterman 1996, Eliasoph & Lichterman 2003, purely qualitative methods are ambivalent for our purpose. On the one hand a thick description based on an intimate knowledge of the group is indispensable to fully understand its dynamics. This is why a group description based on field notes and interviews with group members will be part and parcel of our analysis. But on the other hand these approaches are not appropriate to analyse a specific sector of group interaction in varying contexts. With this interest in mind, a methodological component with a clear structuring seems to be more accurate.
Scholars from various disciplines have developed different levels of formalisation to organise the data gathered in participant observation (cf. Schöne 2003). While the common understanding of field notes (produced most prominently in ethnographic field work) certainly implies a less restricted way to structure the observations, elaborated schemes are used in psychology and political sciences to reach a high degree of generalisation and reliability (cf. Friedrichs & Lüdtke 1973). Our approach lies somewhere in between these poles. Because we want to study the way political groups and networks communicate not only in different national contexts but also observe these groups with a number of researchers, we need a 4 Participant observation has been used repeatedly to analyse social movements. Still, methodological reflection about this method is scarce in this context. While several recent studies rely on participant observation (e.g. Lichterman 1996, Eliasoph 1998, Leach 2006, methodological concerns are neglected in most of these publications. But see separate articles by Eliasoph & Lichterman (1999Lichterman 1998Lichterman , 2002. structured observation that concentrates on certain aspects of the interaction without leaving much room for random interpretations. Audio or video tape recording, which could compensate for the limited capacities of a single observer, is not always accepted in these groups. One example of a standardised coding scheme are the socio-psychological SYM-LOG categories by Robert Bales et al. (1979; see also Bales 1950). The model was developed to observe interaction processes in small groups by coding every single speaking turn in the discussion. However, the elaborated SYMLOG coding scheme with a multitude of categories tends to demand too much from the observer. This is why SYMLOG categories which proofed imperfect as an in vivo coding scheme were subsequently enhanced and adapted to the needs of small group research (e.g. Fisch 1994;Fietkau & Trénel 2002).
Having tested derivates of the SYMLOG-model, we found that it was impossible to code both the behaviour of discussants as well as the content of what they say. Since we thought it would be important to be able to at least roughly reconstruct the course of the discussions when analysing the data, we decided to refrain from coding every turn and developed a coding scheme on the basis of controversies. Thus, our coding unit is a controversial discussion about a conflictual issue which starts as soon as at least two people contradict each other. But also with this approach, it was necessary to find a limited number of variables to facilitate a systematic and reliable analysis of the discourse-practices observed.

Defining deliberation as a specific practice of discourse
Empirical research is far from a standard list of criteria for what should be considered "ideal deliberation" (West & Gastil 2004: 2). A number of scholars have compared existing literature and extracted the most common criteria used to define deliberation. Trénel (2004: 5) looked at the criteria which have been used in previous empirical research. He extracts five core dimensions: equality, rationality, respect, constructiveness, and interactivity (ibid: 3-4). Della Porta (2005: 74) draws on theoretical literature and characterizes deliberative democracy on the basis of preference (trans)formation, orientation to the public good, rational argument, consensus, equality, inclusiveness, and transparency. Burkhalter et al. (2002) integrate both empirical and theoretical literature, in their set of criteria: availabil-ity of information, considering a range of solutions which represent the underlying diversity of views, making evaluative criteria explicit and recognizing existing differences, evaluating solutions and reaching decisions, granting sufficient opportunities to speak, adequate comprehension and consideration, accepting different styles of communication and of reasoning, mutual listening, employing empathy, and creating a shared language. Finally, Mansbridge et al. (2006: 1) have investigated the normative criteria held by professional facilitators and find that "practitioners [of deliberation] value good emotional interaction alongside good reasongiving, interpret common good as 'common ground' , conceptualize freedom as the 'free flow' of ideas in the discussion, and view inequality as a multifaceted obstacle to deliberation" (ibid: 1-2, our emphasis).
The mentioned studies of deliberation take largely different perspectives. The problem is not that some of them emphasize certain aspects while neglecting others but that many studies are 'diagonal' to each other: Their indicators are not directly comparable because they are partially overlapping or lie on a different level of abstraction. 5 A lot of the confusion and imprecision of existing concepts is due to a lack of differentiation between deliberation as a specific practice of discourse (or of communicative interaction) from both the goals it is supposed to fulfil (e.g. reaching a consensus or a legitimate/ reasonable decision) as well as from the conditions under which it is supposed to take place (e.g. a productive atmosphere, willingness to listen, availability of all relevant information).
Our approach to defining deliberation was to look for a very limited number of theoretical dimensions which on the one hand represent the core aspects of what most authors consider as deliberation. On the other hand these dimensions should also leave room for other practices of discourse such as bargaining, consultation or pressuring. Table 1 illustrates the connection of our dimensions to the above (1) Transformation of preferences (5) Orientation to the public good (6) Transparency (1) Dialogue as mutual listening, employing empathy, creating a shared language (2) Personal testimony as (one) source of relevant information (5) Consider a range of solutions (6) Availability of relevant information (7) Evaluating solutions and reaching a decision (1) Identifying a common ground while preserving plurality (3) Good emotional interaction (4) Self-facilitation rather than facilitator control (as an aspect of equality) (4) Maintaining a positive atmosphere (5) Common good as 'common ground' (7) Making progress Additional aspects / causes and results of deliberation mentioned works. 6 Each row refers to indicators for deliberation that are similar in the cited articles.

Cooperation vs. Competitiveness
The first dimension we consider important for the definition of deliberation separates cooperative from competitive behaviour. The scholarly concepts of consensus, the common good/ common ground, constructiveness and (willingness for a) transformation of preferences all refer to notions of commonality and the will to bridge differences as a main feature of deliberation. We would argue that their common denominator is in fact cooperative behaviour. In order to also grasp differing forms of communication, our model opposes cooperation and competitiveness. This polarization is inspired by agonistic critiques of deliberative democratic theory (e.g. Connolly 1991;Honig 1996;Mouffe 2000;Tully 1999;Young 1996Young , 2001.
These critics address the very nature of political communication. While deliberative democrats prefer rather cooperative discussions oriented towards consensus, their agonistic critics underline the importance of struggle and competition between different political positions. They argue that a cooperative discourse always implies an a priori consensus amongst the participants which limits the scope of positions to be debated in the discussion, thus excluding certain political actors and their opinions. According to this position, the cooperative 'consensus' which deliberative democratic theory aims to achieve through the process of deliberation thus rests on exclusionary conditions that challenge the substance of democracy. Taking this position serious, we also expect to find discussions which are a competitive struggle for the own position to prevail. It is important to apprehend however, that competitive behaviour is independent of the means that are used to 'win the competition' . These means are the core of our second dimension, the dimension of power.

Soft Power vs. Hard Power
The question which inspires the dimension of power is: What kind of resources do speakers employ in order to make participants agree (or at least not disagree)? As pertains deliberation, all studies mentioned above refer to criteria for reasongiving: giving rational arguments, careful weighing of views, or giving good reasons (cf. table 1). Many scholars have criticized the rationalistic bias of studies relying only on reasoned arguments and argue that also other forms of speech such as testimony or narratives are legitimate and frequently used in group discussions (e.g. Young 1996;Ryfe 2006). This has lead to long lists of different styles of communication complementing the idea of the "forceless force of the better argument" (Habermas 1973: 137). It remains unclear however, what these have in common.
It seems that because of the emphasis on inclusiveness in deliberative democracy its proponents where eager to rid the exclusion of any style of communication that participants might want to use to express themselves. Obviously, for a clear definition of deliberation, we need to explain, what kind of substantiation should not be included and why. In our view, it is the source of power on which speakers rely that defines the second criterion for deliberation: Do speakers draw on the "discursively produced and intersubjectively shared beliefs … [as] a motivating force" (Habermas 1996: 147) or rather on their capacity to impose sanctions or grant rewards (cf. e.g. French & Raven 1960). In order to avoid the conceptual ambiguities related to communicative power (cf. Flynn 2004: 444-451) and its close linkage with the ideal of rational discourse we prefer to speak of soft power including not only arguments but also symbols, metaphors, narratives and other forms of speech as far as they do not curtail "the cognitive aspects of using one's communicative freedom to say 'yes' or 'no' to a validity claim for reasons" (Flynn 2004: 445;emphasis removed). In other words, the main resource of soft power is the doxa or 'public opinion' of a specific situation; it is actualised by speakers (e.g. through arguments) in order to motivate their audience 7 to accept their claim.
Hard power, by contrast, is defined as the capacity of a speaker to impose sanctions or grant rewards based on one-sided or mutual dependencies. This includes the use of expert power or other forms of social power where social dependencies are actualised. Looking at discursive practices it is not the actual balance or imbalance of dependencies that counts, but whether or not these dependencies are mobilized and thus become relevant in the discussion. Power structures which exist beyond their actualisation in a specific situation will be dealt with on the level of a more holistic group-analysis, not in the situational analysis of discourse (see below in section 5.3).

Symmetry vs. Asymmetry
All texts on deliberation (cf. table 1) refer to equality as a characteristic of deliberation. However, the notion of equality is defined in very different ways, either in the sense of equal participation, inclusiveness and equal opportunities to speak or as mutual respect and listening (which does not necessarily imply equal participation). 8 Since we want to regard deliberation strictly as a practice of discourse, i.e. of communicative interaction, we focus on the relation which speakers constitute between themselves and others as they speak. In deliberation, these relations will be symmetric in the sense that other discussants are recognised and treated as equals. They are not devaluated or valorised by the speakers (e.g. because of existing differences). Other forms of discourse, by contrast, may be characterised by asymmetry. In such a situation, speakers treat the other as inferior or less important.

Participation as condition and result of deliberation
Most literature on deliberation and deliberative democracy also refers to such criteria as equality, inclusiveness, extensive participation, representation of all views (cf . table 1). This participatory characteristic of deliberation is perhaps the most consensual one compared to the others mentioned above. It is unclear however, if it refers to active participation in the sense of taking part in the discussion itself or to passive participation, understood as listening to the discussion but having the opportunity to intervene. In the case of group meetings the first refers to actively making claims in the discussion while the latter simply means attending the meeting, which of course requires accessibility of the meeting or its mediated forms.
Passive participation is obviously relevant for concepts of participatory and deliberative democracy but since our aim is to conceptualise deliberation as a discourse practice and not as a comprehensive democratic procedure we consider passive participation as not relevant for our definition. Nevertheless, participation is part and parcel of our research (see figure 1). We gather data to analyse how the number of people in the audience as well as their characteristics influence the type of discourse amongst the active participants (arrow C in figure 1). Similarly, we can analyse if meetings which are known for a specific type of discourse attract more or fewer participants (arrow D).
Active participation is more closely related to concepts of deliberation but we also exclude it from our definition of deliberation as a practice of discourse, primarily because there is no theoretical reason to define deliberation per se as a democratic practice. By distinguishing between deliberation and participation we can study empirically whether or not (or in which situations) deliberation is a democratic form of discourse in the sense that it fosters participation (arrow D in figure 1). Burkhalter et al. (2002: 413) argue that this is the case and even speak of a "reproduction circuit", which reinforces "expectations, attitudes, beliefs, and habits that, in turn, are conducive to the original behaviours [i.e. deliberation]" (ibid.). This might be true for individuals on a general level but if we look at the actual discursive practices as they occur in concrete situations, the question arises if the circuit is really reinforced in practice because with a higher number of active participants deliberation might become more difficult so that participants get frustrated or fall back to "familiar cognitive routines and scripts" (Ryfe 2005: 61).
In other words, there is some reason to assume that high active participation does not facilitate deliberation (arrow C in figure 1) and it remains to be studied under which conditions (arrows A and B in figure 1) deliberation takes place and can be sustained even under conditions of high participation. As figure 1 indicates, we suggest studying these conditions on three levels: the level of group structures, the setting of the observed meeting and -since we are especially interested in the capacity of deliberation to resolve conflict -the nature of the specific controversy.

The 3-Dimensional Model of Discourse Practices
Having identified deliberation as a specific practice of discourse with three core characteristics and the relation of this concept to participation, we will now illustrate how the above dimensions can be combined to form a 3-dimensionsonal model of discourse practices. It provides a tool to distinguish eight practices of discourse, with deliberation being one of them. This way, we may broaden our horizon when analysing group discussions and acquit ourselves from an normative obsession with deliberation as the ultimate art of discussion.
Our model can be visualised as a three-dimensional space in which any discourse practice, which a speaker may make use of, can be located (figure 2). The origin of the coordinate system is defined as "ideal deliberation" -a symmetric and cooperative discussion where only soft power is used. It serves as our analytical point of reference, thus measuring the real world practices of discourse as deviations from this point. It is important to note however, that it is not our aim to put a normative value on these deviations. It is simply the most coherent way of combining the three dimensions to form a three dimensional space. Different democratic theories have different norms of public discourse (Ferree et al. 2002) and in every controversial situation it depends on the respective institutional setting how competitive the discussion, how symmetric the relations between participants ought to be, and what sources of power may be legitimately used.

Conditions of discourse on various levels:
-group-level -meeting-level -controversy The three axes do not represent the absolute amount of hard power, asymmetry, or competitiveness but their relative importance compared to soft power, symmetry and cooperation respectively. Every communicative interaction which we observe can be located somewhere in this three dimensional cube with its coordinates indicating its relative amount of power, asymmetry and competitiveness. However, an exact measurement of these dimensions is -at least for comparative participant observation -practically impossible. For our purposes, it is perfectly sufficient to split every axis into two parts with one of the antagonistic concepts prevailing. Each of the resulting eight spaces labelled A to H in figure 2 represents a specific practice of discourse. We explicitly refrain from labelling these cubes because they are not filled exhaustively by common terms for discursive interaction such as guiding, teaching or arguing. Nevertheless, to make the idea more tangible, we will give one example for each of these analytic spaces. Some of the names assigned to each discursive practice are more appropriate than others so they should be regarded as examples roughly describing the different discourse-practices but not defining them.

Degree of participation
As already mentioned, deliberation (A) takes place when discussants treat each other as equals, view the discussion as a cooperative search for a common solution and rely mainly on the power of arguments. Coordination (B) differs from deliberation in that arguments are obviously less important. Discussants cooperatively try to combine their different positions in order to build a common solution based on their respective hard power. However, possible differences do not play a role in the process (symmetry). In opposition to such discourse, guidance (C) is characterised by the asymmetric relation between discussants. While still aiming at a common solution for all, there is clearly one leader (or several leaders) who does not treat others as equal (e.g. they are considered less important or less knowledgeable) and the others accept their (partially) inferior position. In the process of guidance, the potential sanctions possibly imposed by the leader(s) matter more than their arguments because the guided are dependant on the guide (or believe to be). Consultation (D) could also be considered as some kind of guidance but it differs from it as hard power is not relevant here. It is the quality of the advice that matters to the advised because there is no sanction to be imposed by the consultant if the advice is not followed. However, other than in deliberation, arguments do not fully speak for themselves but are considered valuable because the consultant who advances them is considered superior (at least with regard to the subject matter).
Looking at the competitive forms of discourse, we will find a dispute (E) similar to deliberation. The difference is marked by the fact that discussants do not aim at a common solution but at promoting their own position. However, they do not use hard power to enforce their position. In bargaining (F), there is a high relevance of hard power, which the discussants use in a competitive way. Unlike in coordination (where the aim is to build a common solution) discussants only make concessions where they are forced to or where they gain an advantage. They regard each other as equals however, either because they have to (expectations of the group, formal rules etc.) or simply because they prefer such relations. Whenever this is not the case, we are witnessing pressuring (G). In this kind of discourse practice, discussants do not usually regard the others as equals but as inferiors. While guides also treat the guided as inferiors, they still aim at a common goal. To the contrary, pressuring is done in order to push forward the own position and any available sanctions are used to reach this goal. If hard power is not available or there is a taboo on using it, then discussants will try to influence the others by persuasion (H), often in an agitatory or inflammatory way. Other participants are treated as objects of indoctrination, rather than equal subjects. Other than in processes of consultation where the consultant listens to those seeking advice in order to find a common solution, agitators do not listen or if they do, they do so in order to optimise their strategy for their own victory.

Operationalisation for Comparative Participant Observation
In order to produce comparable data about our participant observations, a codebook was developed in the framework of the DEMOS-project, which gives detailed instructions how the theoretical dimensions described above can be measured in practice by research teams in different countries. 9 Ensuring reliability of the coding in such an enterprise is a challenging task though. This is surely one of the reasons why comparative participant observation involving data from more than one observer has rarely been put into practice (cf. Schöne 2003). Ideally, all observers should participate in the same meeting, code it independently and then compare their codings, discuss inconsistencies and repeat the procedure until all observers share the same idea of each category. When doing the actual fieldwork two observers should ideally attend each meeting and discuss inconsistent coding not only between them but also with the other observers. In practice however, we will have to cope with a number of curtailments of this ideal situation.
Firstly, a rather practical problem can be getting access to the field. It may not be easy to find a group, which is willing to be observed in their meetings by several researchers. We can say from our experience that social movement groups are generally less doubtful against researchers than they used to be (some even welcome being observed and getting feedback about their own practices). Nevertheless, one should not underestimate the scepticism often present at least in parts of the group. Apart from the absurdity of a meeting with, in our case, ten test-coders, it may also be difficult to continuously have two observers in the same group because trust needs to be established toward both observers.
Secondly, the joint pre-tests require significant resources in terms of travel costs and time. In our case, a copy of the movie "12 Angry Men" (1957 by Sidney Lumet) turned out to be a helpful surrogate. 10 However, it also became clear that watching a movie is far different from observing a real social movement assembly, both in terms of the type of meeting and the form of presentation (cameraperspective and cuts). It was therefore essential that at the time of test-coding (parts of) the movie, all observers had already done some test-coding in the field so that during the discussion about the coding of the movie everyone knew what kind of situations would really be coded afterwards.
Thirdly, cultural and language differences are likely to be responsible for misunderstandings and thus miscodings when researchers from different countries come together to observe meetings in one specific country (a country where a language is spoken that is a lingua franca for all researchers). The cultural differences can be relevant in two ways. When comparing across continents, but even across Europe or the U.S.A., a basic knowledge of local habits and customs plus the political context is needed to understand practices of discourse adequately. Nevertheless, ensuing discussions amongst the researchers about misunderstandings can be instructive for a deeper understanding of their categories.
As a solution to the above mentioned problems a continuous exchange amongst all observers should be established during the observation period. The value of an email list (or even periodical phone conferences) dedicated to the ambiguities of coding complex interactions in a reliable way should not be underestimated. Discussing uncertainties of the code book as they arise helps to create a common understanding of the categories and ensure adequate reliability 11 of the data. 10 The whole movie captures a very controversial meeting of a citizens jury, discussing whether or not a boy accused of killing his father is guilty or not. 11 One of the anonymous reviewers was insistent that we should "provide some mechanism to test the reliability" of the coding. We can understand that demand from the point of view of a classic 'quantitative' approach. However, for the reasons discussed above, we found that although comparative participant observation needs some degree of standardization and -of coursecommon understanding of those standardized categories, it cannot (yet?) reach the standards of, for example, quantitative content analysis (which by the way is rarely ever done by several teams in different countries -otherwise they would also experience similar problems). Besides not aiming at a purely standardized approach, we also believe that reliability can also be 'measured' discursively through continuous discussion within and between the research teams dealing with different cases. Furthermore, it is essential to view the standardised coding of theoretically inspired dimensions as only one part of the data set. Field notes are nevertheless vital part of the research process, also in comparative participant observation. They serve as a framework to ensure adequate interpretations of the standardised data. For a systematic approach to comparing contextual data (left box in figure 1), we distinguish three levels of observation and accordingly use three types of reports to be given by the participant observer: a general group portrait, a session report and a controversy protocol.
The group portrait is based on the participant observer's accumulated knowledge about the group and only given once for each group at the end of the observation period. The session report is filled in after each observed session and one controversy protocol is coded for each controversial discussion. Both the session report and the controversy protocol are based on notes taken by the observer during the session. As a result of this three-layered procedure of data collection, we are able to compare our data not only on the micro-level of controversial situations but also on the contextual level of sessions (as a context in which the controversies take place) and on the group level (as a context in which the sessions take place). Figure  3 illustrates this structure of the data. The data on the controversy level is coded in a standardised way, the data on the session level also contains some standard codes but also some text descriptions of the agenda as well as special incidents and peculiarities during the session. Finally, the group description is a free flowing text, written along a given number of keywords to ensure that a number of essential aspects have been dealt with by all observers.
In the following, we will focus on the controversy protocol since our theoretical model has been developed with regard to discursive practices in controversial discussions. The other two levels will be briefly described.
One possible result of this process of deliberation amongst the researchers can be the insight that no common understanding of a certain category can be reached, despite detailed description and examples. This can be due to diverging interests regarding the main research question but more likely this will happen when the category is too complex or too abstract. However, as regards the main categories suggested in this paper, more than 13 researchers in six countries did not reach that point of complete disagreement. Whenever we experienced differences in coding or unclear understanding of how to apply a category, we were able to solve this through collective deliberation (quite close to the ideal point in figure 2) leading to further specifying -for example -the difference between hard power and soft power.

The Controversy Protocol
For each controversy, a total of 28 variables are coded, including the dimensions introduced above, are coded. 12 The codebook contains examples and detailed coding instructions defining the meaning of the codes/categories and giving examples of how they should be used. In the following we will briefly describe the operationalisation of our main dimensions.
The main difficulty in operationalising each dimension of discourse is that the basis for the coding is a whole controversial discussion comprising a number of speaking turns that might be unique with respect to power, asymmetry, and competitiveness. They all have to be subsumed under one single code for each dimension when we want to characterise the controversy as a whole, i.e. we are not coding individual behaviour but 'group behaviour' (as an aggregate of individual behaviours). The observer has to estimate something like an average across all speaking turns within that discussion but also consider the discussion as a whole, i.e. the observer should consider not only the mere number of statements showing a certain tendency but also their importance for the course of the controversy. As a general rule, the intensity of single speaking turns can be neglected as long as they do not obviously influence the rest of the discussion to a large degree. For example, if someone makes a very angry and disrespectful intervention but the others just ignore his anger and continue with their discussion in a calm way, then we note that some asymmetric behaviour was present in the discussion, but certainly not prevailing. In ambivalent situations the participant observer uses as an orientation the hegemonic definition within the group of what the situation is. In other words, the perspective of the observed actors themselves has to be taken into consideration.
In order to ensure reliability, this type of estimated measurement can only be rather rough. Measuring our dimensions of discourse in more detail than a fourpoint-scale 13 would be assuming an accuracy which is simply not possible. In any 13 The four categories can be understood as 0-25 percent, 25-50 percent, 50-75 percent and 75-100 percent of whatever the scale measures.

Figure 3:
The three-levelled structure of the data from comparative participant observation case, it is essential for observers to take notes during their observation that help estimating the codes and also to allow later the reconstruction of the coding decision.

Competitiveness
Indicators for competitiveness are speaking turns, in which speakers give no indication that they are ready to change their position, e.g. when they seem quite convinced that there is little to make them change their mind or to compromise. This might be indicated by an assertive speech style as well as through persistent repetition of the same position. An indicator for a high degree of cooperation, by contrast, would be speakers indicating that their aim is not to 'win' the discussion but to find a common solution, e.g. when they seem unsure of whether their opinion is right or even when they are convinced about their position but are willing to compromise with the others. This might be explicitly indicated by speakers or signified by tone of voice, thinking pauses, etc. as well as asking others for their opinion.
The coding of the discussion-process should not be inferred from the envisaged outcome of the discussion, i.e. if speakers argue in favour of a compromise, this does not necessarily mean that they behave in a cooperative way. For example, if they continuously uphold a specific compromise, arguing that this is the best compromise and everybody should agree, then this is a competitive behaviour.

Power
Typical speaking turns addressing hard power are offers, demands and threats. All of them draw on the possibility of the speaker to either harm or reward other participants or the group as a whole. More generally, hard power is used when references to social dependencies are made in order to motivate others to accept a certain position. It is not necessary to check if the relevant resources are really available to the speaker since we look at practices of discourse and not structural balances of power. However, if a statement is obviously an idle threat and most participants appear to know that, we do not regard this as use of hard power. Also in this dimension the perspective of the participants is relevant.
Excessive shouting, intimidating gesticulation or using more speaking time than granted by the moderator are not acts of hard power as long as they are not suitable to directly enforce the position of the speaker. Such behaviour is usually only suitable to inhibit others from participating in the discourse but not to influence the outcome of the discussion directly. Its effects are thus indirectly measured in the dimension of participation. 14 Shouting and speaking over time are nevertheless expressions of asymmetric relations (see below).
One of the main challenges in measuring the power dimension is to distinguish between power structures within the group and the actualisation of these structures in a specific situation. For example, it makes an important difference if an expert is invited to a meeting to speak as an expert and discuss with the participants or if the same expert sits in the audience and (parts of the) participants do not know about his status as an expert. In the former case, the expert status is publicly actualized (or even created) in that very situation and therefore needs to be considered when coding a controversial discussion in which this expert takes part. 15 In the latter case, the expert status remains irrelevant for the coding as long as it is not actualized in the situation, for example by the expert himself mentioning that he has been working quite a long time on this subject. This limitation of the observer's perspective to the evidence available in a given situation is necessary in order to ensure reliable coding on the level of controversies. 16 In order to recognise the discursive use of power in practice, observers have to understand the nature of hard power which relies on social dependencies rather than publicly proclaimed beliefs and norms. In the case of material resources and working time, it is quite obvious how threatening an announcement to give or to withdraw such 'goods' may be. Experts hold a monopoly on a specific knowledge which the others in the group need. They depend on the willing participation of the expert who can thus effectively threaten to withdraw this knowledge from the group. The exit of 'ordinary' participants from the discussion will usually not affect the discussion in the same way. The power of non-expert participants lies within their collective organisation enabling representatives to speak for them and using the exit-threat to exert power in the discussion. Here the power of a speaker 14 We have not systematically included explanatory variables for participation since our dependent variable is "type of discourse". 15 In fact, many experts try to neutralize their elevated position in such discussions by downplaying their achievements while others deliberately use the authority given to them in this situation as a source of power in the discussion. 16 The accumulated knowledge of the observer about power relations within a group is part of the group portrait and will be taken into consideration in the process of data analysis.
representing others results from the dependence of the opponent on the represented and their exit-option.
Furthermore, we need to mention that the degree of hard power used in the process of discussion should not be inferred from the result of the discussion or from the mode of decision. Even though a decision might be finally taken on the basis of hard power (e.g. majority vote), hard power does not necessarily have to be involved during the process of discussion and vice versa. Our aim is to code to what degree discussants employ hard power in the discussion, i.e. they use it trying to achieve a goal 17 and that is independent of actually achieving it. In other words, it is well possible that even though a major exit threat at the beginning of a controversy has an important weight at the end when the decision is finally taken, most of the discussion leading to this decision was free of hard power. For example, the discussants might have discussed the advantages and disadvantages of the some people leaving the group and other alternatives.

Asymmetry
This dimension aims at measuring to what degree speakers do not consider/treat each other as equals. It does not refer to the quantitative distribution of speaking turns amongst the participants (this is coded in the participation dimension) but to the way the speakers relate to each other. We are witnessing asymmetry not only when a speaker absolutely and fully despises some or even all others. We need to consider much lower levels of asymmetry, e.g. where others are regarded as different and that difference is regarded relevant for the issue at stake in the controversy. Such a difference might be inequality in knowledge: We witness asymmetry when a speaker regards others as less or more knowledgeable with regard to the subject matter (while perhaps admiring him for his cooking skills or as a good friend). Another way of expressing an asymmetric relation can be shouting at others or not reacting to their questions, demands or arguments.
By contrast, we may witness rather symmetric discussions, where there is a high awareness for possible inequalities and attempts to compensate these inequalities. For example, before entering (or continuing) a debate, some basic information about the issue and perhaps a summary of previous discussions might be given, so that all participants have the same background knowledge. Another behaviour displaying symmetry is openness for different styles of debate and openness for completely different approaches to the issue.
Note that for a speaking turn to be considered asymmetric, the speaker does not need to actively show disrespect for others. Like in the other two dimensions, the context of the situation is crucial; otherwise coding will not be valid.

The Session Report
The session report gathers data about every group session attended by the observer. Its purpose is to provide data about the context that frames controversies within the group. The session report structures all communication during a group's meeting chronologically be it controversial or not. For each phase (agenda item) of the meeting we record the duration, the dominant form of interaction (input statement, separate contributions, discussions, brainstorming, go-round etc.), a rough measure for participation (number of male/female participating actively/passively) and possible decisions as well as the mode of decision (unanimity, majority vote, tacit agreement etc.). Other particularities of the meeting are also noted (e.g. who prepared the meeting? who is the facilitator? why? suspected hidden agendas etc.). Providing these context data, the session report allows us to answer several questions: How much of the meeting is actually devoted to controversies? In which situations does conflict emerge? Does participation differ during controversies? And do decisions and related constraints trigger discussions that otherwise would have been unheard? Beyond keeping hold of the sequence of a meeting, the report is designed to record particularities of the session. This is to notice differences in comparison with precedent meetings that might be a sign of particular circumstances or a general change in the group's behaviour and/or composition.

The Group Portrait
On a more general level a group portrait complements the information gathered in specific sessions of the group. Knowing about the group's ideology, its history, social composition and strategic preferences is a prerequisite to understand the controversies that are observed and classified in a more formalised manner. Longstanding personal or ideological conflicts, for instance, might have shaped the group communication and continue to structure internal debates. The purpose of the portrait is to convey a basic knowledge about the group to the other researchers who should then be able to understand better group processes as they are recorded in the protocols during a session. Ebbs and tides in the membership, the composition of participants and specific roles within the group are some of the peculiarities that can be reconstructed on the basis of documents and short interviews with long time members of the group. Particularly relevant for the study of deliberation is information about shared ideas about democracy in the group. On the internal level, the group's vision of democracy helps to understand why the group chooses specific routines and styles of discussion.

Summary
In this paper, we have addressed three methodological challenges which we encountered when trying to study practices of discourse within the global justice movements. First, we looked at some characteristics of these movements, briefly addressing the different settings in which controversial discussions occur and pointed out that these various conditions of discourse should be considered when selecting the cases to be studied. Second, we gave a rationale for applying a semistandardised multi-level participant observation in order to allow the collection of comparable data by various researchers in several countries. Focusing on participant observation on the level of controversial discussions we thirdly conceptualised competitiveness, power, and asymmetry as three theoretical dimensions to identify eight different practices of discourse, one of them being deliberation. Lastly, we operationalized the theoretical model providing some examples of how various behaviours in face-to-face discussions should be coded. We are currently implementing this model for regular observations of group meetings on a local, national and European level. First results should be available in the near future.