Multivocality and robust action dynamics in political discourse
Introduction
How do politicians gain an upper hand in political discourse? Political discourse is an arena where power is maintained and generated (van Dijk & Teun, 1993; Fairclough, 2013; Thornborrow, 2016), and this process can be understood as a framing contest (Cornelissen & Werner, 2014; Dewulf et al., 2009; Kaplan, 2008), where political actors dynamically adjust their behavior to respond to speech acts from their opponents (Krebs & Jackson, 2007). In this article we adopt a relational robust-action perspective to discourse, as we argue that political actors engage in framing contests by choosing between two kinds of behaviors: First, speakers are interested in exploring the repertoires of their opponents, aiming to make them shift through their discursive positions and thus reveal the full repertoire of their framings (Leifer, 1988; Steinberg, 1999). Second, beyond such exploration political actors are ultimately interested in gaining the upper hand in discourse by taking a position that their opponents have no viable answer to (Koopmans & Olzak, 2004; Nelson & Kinder, 1996; Schröter and Taylor 2018), thus speakers aim to switc from exploration to exploitation. In sum, political actors engage in discourse with a behavioral portfolio similar to that of a boxer, starting with an exploratory dance around the opponent, looking for the opportunity to deliver a punch (Wacquant, 1995).
Understanding political discourse requires a model of discourse dynamics. One perspective to understand the connection between political parties and discourse relies on the concept of issue ownership: relatively stable affiliations of parties with issues that align with their voter base (Petrocik, 1996; Rokkan et al., 1999). However, there is much evidence for a relationship between parties and issues that is much more dynamic and competitive (Meguid, 2005; Sniderman et al., 2004), with challenger parties bringing about disruptive innovations in framings (De Vries & Hobolt, 2020; Hobolt & Tilley, 2016). In line with broader transformations, as the political field shifted from class-based party politics to issue-based party politics over the past half century, understanding dynamics in issue representation beyond issue ownership is now a key task (Green-Pedersen, 2019).
This article presents a sociological analysis of political discourse dynamics by relying on a rich literature in political and cultural sociology that re-constructs discourse as a temporal symbolic network (Carley, 1994; Foucault, 1972; Mohr, 1998), to understand relations of power in the political field (Benford, 1993; Bourdieu, 1991). Our key contribution is that we shift focus to robust action in discourse dynamics as relational drama (McFarland, 2004), rather than adopting the more usual focus on the reception of political frames by the audience (De Sio & Weber, 2014; Vasi et al., 2015): in our study discourse positions are both the dependent and independent variables (Snow et al., 2004), as we aim to understand strategic interaction of political rivals (Krebs & Jackson, 2007; Mische & Pattison, 2000; Steinberg, 1999).
Robust action perspectives carry a lot of promise to conceptualize political discourse, but their key concepts are often difficult to operationalize, such as opaqueness, ambiguity, and seizing the moment (Ferraro et al., 2015; Padgett & Ansell, 1993). We rely on the concept of local action1 (Leifer, 1988) and symbolic network analytic methods to provide a clear operationalization for robust action in political discourse. We distinguish two modes of acting: local action, when actors explore role repertoires of adversaries, and role claim, when an actor attempts to occupy a dominant discourse position. We argue that this idea of local action is a fruitful way to combine a structural analysis of symbols and speech acts with a relational analysis of political organizations struggling for power (Emirbayer & Goodwin, 1994; Erikson, 2016).
Our empirical case is the economic policy discourse of consolidation in Hungary, over the spring of 1997. During this period government side political actors (and other expert and journalist discourse participants mostly on the left) shifted from defending austerity measures to an offensive built on the idea of GDP growth. With this they defeated a stigmatizing offensive launched by young Viktor Orban and the FIDESZ party, that recently veered to the right. Using a two-mode blockmodeling approach to identify discourse positions, we show that this innovation relied on the most multivocal position in the discourse, combining several frames in a way, that silenced opposition positions. While all other government and opposition side positions had a clear answer in the discourse, and did not significantly reduce the probability of speaking, the position on GDP growth significantly reduced the probability of speaking, regardless of the identity of actor. In a fashion that paralleled the world economic policy stage (Schmelzer, 2016), growth discourse in our case connected macroeconomic indicators with individual prosperity in a uniquely effective way2.
We translate our textual corpus into a network, building from ties among statements (the elementary symbols in this article) that are made when a political actor uses two or more statements in the same speech act. Taking the resulting two-mode network of statements connected with speech acts, we use a two-mode blockmodeling approach to reveal frames and positions – clusters of structurally equivalent statements and speech acts, respectively. A frame is a typical kind of building block, while a position is an actual construction out of frames. A position can build from one frame only (with little multivocality), or it can be composed of multiple frames, and thus be multivocal.
We then analyze the relational dynamics of using these positions to test hypotheses about the presence of a local action dynamic in the discourse. First, we test a weak hypothesis of dynamics, H1, that the prevalence of discourse positions significantly differs from expected in the relevant temporal window following other positions. Second, we test stronger hypotheses about local action, taking a pooled time series approach with actor-days as units of analysis. The first set of stronger hypotheses are about exploratory local action, stating the expectation that politicians will alter their discourse positions in response to what had been said in a preceding time window. We expect that it is the amount said on the other side (rather than multivocality) that induces position shifts. The second set of stronger hypotheses expect significant reduction in the number of speech acts (become silenced) when the opposing side takes a multivocal position, bridging frames. We expect that it is the multivocality (rather than the amount of speech on the other side) that silences politicians.
Section snippets
A network approach to discourse frames and positions
A key conceptual goal of this article is to bridge between symbolic regularities of a discourse and dynamic interplay among political actors, as they struggle to dominate the agenda in the public sphere. Others have recognized the fundamental duality of analyzing discourse as a system of representation, and as an arena of interaction (Dewulf et al., 2009; Gonos, 1977). Analytically, cultural and social relations are to be mapped separately, but conceptually, the dimensions of meaning and agency
The case of economic policy discourse, Hungary 1997
We test hypotheses of local action dynamics in political discourse using a case of high-stakes political struggle, that brought about a discourse innovation of government actors that silenced the opposition. We discuss two aspects of selecting our case: specifying the general outline of the discourse we study, and the specific demarcation of the boundaries of our case. The general location of our case is Hungary, early 1997, when Gyula Horn was prime minister. This period can first of all be
Blockmodel analysis to identify frames and positions
The first step in our analytic strategy is to map frames and positions. We cluster statements into frames and speech acts into positions based on patterns of ties among them, taking a two-mode blockmodeling approach by structural equivalence: If two statements are mostly used together (or are mostly omitted together) then we see them to be structurally equivalent. A typical network blockmodel builds from a one-mode network (White et al., 1976), and has recently been extended to consider two
Relational discourse dynamics
In the preceding section we have taken a network approach at the symbolic level of frames and positions. Now we turn to relational dynamics among actors, as the use these discourse positions. We have clustered each speech act into positions, and since speech acts are time-stamped (we know the day when they were made), we can trace the dynamics of positions take daily. However, a model that builds form the positions taken on a given day would not be a realistic model of discourse, as it would
Discussion and conclusions
We have observed relational regularities in discourse dynamics on multiple levels of analysis. First, there are pronounced frames and positions in the discourse. Positions are usually centered on one frame, but some positions are more multivocal, and span multiple frames. On the government side a frame-spanning position is about GDP growth: the most multivocal (multi-frame, multi-statement) position in the discourse.
We have also observed these positions in use. The way in which positions evoke
Balazs Vedres is associate professor at the Oxford Internet Institute, University of Oxford, where he directs the new PhD program on social data science. Vedres' research furthers the agenda of developing data science and network science with social theoretical insight. His research results were published in the top journals of data science, network science, and sociology, with two recent articles in the American Journal of Sociology developing the pragmatist notion of structural folds:
References (83)
- et al.
Becoming a Nazi: A model for narrative networks
Poetics
(2000) A tool kit for practice theory
Poetics
(2000)- et al.
Working together but in opposition: An examination of the ‘good-cop/bad-cop’ negotiating team tactic
Organizational Behavior and Human Decision Processes
(2000) Extracting culture through textual analysis
Poetics
(1994)- et al.
Generalized blockmodeling of two-mode network data
Social Networks
(2004) - et al.
Formal studies of culture: Issues, challenges, and current trends
Poetics
(2018) - et al.
Relating social and symbolic relations in quantitative text analysis. A study of parliamentary discourse in the weimar republic
Poetics
(2020) Soldiers, mothers, tramps and others: Discourse roles in the 1907 New York City charity directory
Poetics
(1994)- et al.
Introduction—Topic models: What they are and why they matter
Poetics
(2013) - et al.
From affirmative action to outreach: Discourse shifts at the university of California
Poetics
(2000)
Lattices and dimensional representations: Matrix decompositions and ordering structures
Social Networks
The strength of weak culture
Poetics
Distance and cosine measures of niche overlap
Social Networks
The judge, the cop, and the queen of carnival: Ethnography, storytelling, and the (Contested) meanings of protest
Theory and Society
The fringe effect: Civil society organizations and the evolution of media discourse about Islam since the September 11th attacks
American Sociological Review
Frame disputes within the nuclear disarmament movement
Social Forces
Latent Dirichlet allocation
The Journal of Machine Learning Research
A model of robust positions in social networks
American Journal of Sociology
Language and symbolic power
The duality of persons and groups
Social Forces
Putting framing in perspective: A review of framing and frame analysis across the management and organizational literature
Academy of Management Annals
Issue yield: A model of party strategy in multidimensional space
American Political Science Review
Game changer: The topology of creativity
American Journal of Sociology
The rise of challenger parties
Political Insight
Disentangling approaches to framing in conflict and negotiation research: A meta-paradigmatic perspective
Human Relations
Principles of critical discourse analysis
Discourse & Society
Culture in interaction
Source: American Journal of Sociology
What is agency?
American Journal of Sociology
Network analysis, culture, and the problem of agency
American Journal of Sociology
Publics in history
Theory and Society
Formalist and relationalist theory in social network analysis
Sociological Theory
Industry Induces Academic Science to Know Less about More
American Journal of Sociology
Critical discourse analysis: The critical study of language
Tackling grand challenges pragmatically: Robust action revisited
Organization Studies
The archaeology of knowledge
The meaning structure of social networks
Sociological Theory
Seizing the moment: The problem of conversational agency
Sociological Theory
Participation shifts: Order and differentiation in group conversation
Social Forces
Taking turns and talking ties: Networks and conversational interaction
American Journal of Sociology
Community structure in social and biological networks
Proceedings of the National Academy of Sciences
‘Situation’ versus ‘Frame’: The ‘Interactionist’ and the ‘Structuralist’ Analyses of Everyday Life
American Sociological Review
Cited by (2)
Holding a position: Public opinion as cognition in a disorganized field
2022, PoeticsCitation Excerpt :Moreover, rather than asserting the pro-life position on the “hard” cases, the pro-life camp dismissed their relevance, arguing that modern medicine virtually eliminated situations where the mother's life is pitted against that of the child and that “something biological happens to rape victims that precludes the possibility of pregnancy” (Luker 1985:235)—a widely discredited argument that nonetheless was reiterated by pro-life Senate candidates as recently as 2010. Another example that displays both of these tactics—switching to a more favorable sub-issue within a domain, or pivoting to an entirely different domain—comes from Vedres's (2021) examination of political messaging in 1997 Hungary, where the governing party's communications marshalled positive economic indicators to argue that their policies are decreasing inflation and bringing GDP growth. The opposition first responded by focusing on a different aspect of the economy.
A Tamed Transformation. Debating Digitalisation in Research and Higher Education Policy in Switzerland, 1998-2020
2023, Swiss Journal of Sociology
Balazs Vedres is associate professor at the Oxford Internet Institute, University of Oxford, where he directs the new PhD program on social data science. Vedres' research furthers the agenda of developing data science and network science with social theoretical insight. His research results were published in the top journals of data science, network science, and sociology, with two recent articles in the American Journal of Sociology developing the pragmatist notion of structural folds: creative tensions in intersecting yet cognitively diverse cohesive communities. Vedres' recent research follows entrepreneurs, video game developers, jazz musicians, programmers, and graphic designers as they weave collaborative networks through their projects and recording sessions, analyzing questions of the sources of creativity, gender inequality, and the historical sustainability of innovation systems. In another line of work, Vedres has analyzed historical network evolution in the areas of transnational civic activism, politicized business groups, and the evolution of global economic flows.
I am most indebted for Peter Csigo for collecting and coding newspaper articles. Research for this paper was supported by a grant from the National Science Foundation, # SES-1123807 and the European Research Council grant ERC-ADG-2015- 695256. I am thankful for comments of Peter Bearman, Daniel Beunza, Ronald Breiger, Laszlo Bruszt, Zsolt Enyedi, Bruce Kogut, John Krinsky, John Mohr, David Stark, Duncan Watts, and Harrison White. I am grateful for the support by the Center for Organizational Innovation and the Institute for Social and Economic Research and Policy in providing a fellowship during the preparation of this manuscript.