Heated policy: policy actors’ emotional storylines and conflict escalation

Policy conflict is gaining attention in policy studies. In this paper, we explore the relation between emotional storylines and policy conflict escalation in the case of the Dutch gasquakes in the north of the country. Based on a longitudinal analysis of emotional storylines in 1308 newspaper articles and additional empirical data we find that Dutch subnational governmental actors as well as citizen action groups discursively express emotional storylines about anxiety/fear, anger, and contempt in relation to discursive expressions of trustworthiness/distrust. Over time, specific combinations of these emotional storylines shape the interpretation of the problem and point toward responsible actors. Also the way in which specific sequences of emotional storylines develop (particularly from fear to anger) suggests a discursive escalation. In addition, discursive escalation can be found in the increased intensity of specific emotional storylines. We conclude that the combinations, sequencing and increasing intensity of the emotional storylines suggest a process of emotionally expressed escalation, which we have only just begun to explore.


Introduction
In the policy sciences, conflict often appears as a background concept (Weible & Heikkila, 2017), or as something to settle as quickly as possible (Forester, 2013;Susskind & Cruikshank, 1987). More recently, policy analysts have put conflicts center stage, such as behavioral approaches in the Policy Conflict Framework (PCF) (Weible & Heikkila, 2017), and a variety of interpretive approaches (Laws and Forester n.d.;Cuppen et al., 2015; 1 3 Dodge & Metze, 2017;Durnova, 2018;Griggs & Howarth, 2013;Wolf & van Dooren, 2017;Yuana et al., 2020). All these policy studies make some contribution to understanding how policy conflicts escalate and de-escalate. Some of them point toward the role of discursively expressed emotions, but how they contribute to the escalation of policy conflicts is still an important, unsolved piece of the puzzle. The PCF acknowledges that threats to and from policy positions are at the core of different intensity levels of policy conflicts (Weible & Heikkila, 2017, 29), but the framework does not further unpack how this works. Various interpretive studies indicate that emotions are important elements of policy conflicts, but have hardly analyzed how they relate to (de-)escalation processes (Laws et al., 2014;Durnova, 2018;Wolf & van Dooren, 2017;Ransan-Cooper et al., 2018;Verhoeven & Duyvendak, 2015).
In conflict studies, a common understanding of conflict escalation can be found in the dual concern model by Rubin et al. (1994), in which actors may have concerns about their own outcome, understood as interests, needs and values, or about the other's outcome. These concerns, that can be discursively expressed, may lead actors to four basic choices in conflict strategies: problem solving, yielding, contending, and avoiding. Contending entails that actors only have strong concerns for their own outcomes, which encourages them to impose their preferred solutions on others at their expense. If other actors also contend, conflicts will escalate through a 'set of moves and countermoves that intensifies conflict' (Rubin et al., 1994, 9; see also Glasl, 1982). Some of these escalating moves and countermoves involve transformations in issue complexity, the number of actors, the resources spent, the use of tactics, and psychological distancing (Glasl, 1982;Rubin et al., 1994). In addition, emotions can be seen as 'powerful engines' of conflict behavior (Halperin, 2016, 5). Hence, we can expect escalating moves and countermoves to involve transformations in the expression of emotions (Wolf & Van Dooren, 2017), as contending policy actors may express emotional orientations (Halperin, 2016) based on their appraisal of the consequences that moves by others have for their outcomes, or because they choose to discursively express the emotions of others they represent (Rubin et al., 1994).
In this paper, our aim is threefold. First, we aim to conceptualize relations between the expression of emotions and policy conflict escalation by drawing on insights from both the PCF (Weible & Heikkila, 2017), and the policy controversies perspective (Schön & Rein, 1994), and then, within the latter perspective, we theorize how conflict escalation by policy actors may be traced in the discursive expression of emotions. Second, we aim to empirically explore how discursively expressed emotions such as emotional storylines relate to the (de-)escalation of policy conflicts. Third, we aim to develop arguments on how empirical insights into emotional storylines not only contribute to interpretive studies of policy conflict but are also of value to behaviorist studies based on the PCF, in line with recent attempts to cross epistemological divides between behavioralism and interpretivism (Durnova & Weible, 2020).
Our empirical exploration is based on the heated case of the gas production-induced earthquakes in the Netherlands, known as 'gasquakes'. From 2012 onward, several earthquakes induced by natural gas production led to large-scale material and psychological damage in the northeastern province of Groningen (Perlaviciute, 2018;Postmes et al., 2018). A policy conflict emerged which spread from the local and regional policy subsystem to the entire Dutch political system. Our empirical question is: how did the expression of emotions in storylines by subnational governments and citizen action groups relate to the escalation or de-escalation of the gasquakes conflict in the Netherlands? We conducted a longitudinal media analysis of 1308 articles from one regional and five national newspapers between September 2012 and April 2015. In addition, we conducted three in-depth 1 3 interviews with citizen activists, and we analyzed web content, such as blogs, stories of protests, and other materials used by subnational governmental actors and citizen activists. Based on these data, we explore the storylines that these actors use to express emotions over time.
In the next section, we draw on the PCF and the policy controversies perspective to conceptualize the relations between the expression of emotions and policy conflict (de-) escalation, and we theorize how discursive expression of emotions and conflict escalation may relate. This is followed by an explanation of our case study and methods, the results section, and the conclusions.

Insights from the PCF and the policy controversies perspective on emotional expression
Both the PCF (Weible & Heikkila, 2017), and the policy controversies perspective by Schön and Rein (1994) have a specific view on what policy conflict and controversies are, and what role emotions may play in these. The approaches both take divergence on policy positions as a starting point for the escalation of policy conflict. However, they do so in two distinct ways.
In the PCF, the basis for conflict is differences in deep core beliefs, which are more or less stable or even resistant to change. Deep core beliefs function as cognitive filters through which politicians, civil servants, policy advisors, civic action groups and others (hereafter called policy actors) perceive threats to their policy positions and through which they determine their willingness to compromise on their own policy positions. Such threats may be perceived by other actors' use of strategies and tactics inflicting potential costs or losses on one's own policy position. Individual policy actors sharing the same beliefs or perceiving the same threats may join forces in advocacy coalitions, which oppose other coalitions (Weible & Heikkila, 2017). In the interpretive tradition, particularly those with roots in Schön and Rein's (1994) work on policy controversies, it is argued that conflicting frames consisting of underlying structures of belief, perception, and appreciation shape divergent policy positions. Similar to the deep core beliefs in the PCF, such frames can be seen as filters through which policy actors observe policy positions to determine which ones they accept as legitimate. The difference with the PCF is that divergence in policy positions is not cued by perceived threats on the basis of strategies and tactics, but by the different interpretations that actors make of facts and evidence based on their frames. It is competing discourse coalitions with conflicting storylines that oppose each other (Metze & Dodge, 2016).
The behaviorist inclination of the PCF and the interpretive disposition of the policy controversies perspective have different implications for how to study emotions during policy conflicts. The PCF builds on an individualist ontology, emphasizing individual policymakers, their core beliefs and their assessments of policy positions. In this ontology, deep core beliefs are seen as essences-as attributes of individual policy actors. Through this ontological lens we may see emotions primarily from a psychological point of view. The most dominant psychological standpoint is that emotions are elicited in cognitive appraisal processes of other people, objects, or events in specific situations informed by values, beliefs, or perceptions, and experienced through facial expressions, 1 3 behavior, and discursive expressions (Halperin, 2016;Turner & Stets, 2005;Ten Houten, 2007). Such a psychological perspective fits well with the PCF, since the deep core beliefs of individual policy actors can serve as normative benchmarks during the cognitive appraisal processes that elicit emotions. Similar to deep core beliefs, emotions thus also become essences that individual policy actors experience.
In contrast to the PCF's individualist ontology, interpretivists adopt a relational ontology with respect to policy conflicts. Instead of studying deep core beliefs or emotions as essences held or experienced by policy actors, a relational ontology assumes that we need to focus on the storylines that are produced in relationships, and during interactions between policy actors (Fischer, 2003;Hajer, 1995). As Fischer (2003, 102) explains, such storylines '(…) symbolically reflect the concerns of core beliefs rather than the beliefs themselves.' Likewise, storylines might symbolically reflect the gist of emotions that may circulate as collective emotional orientations (Halperin, 2016) rather than the emotions themselves. Such a relational ontology opens up the study of emotions from a sociological point of view, which sees emotions as socially constructed through language according to culturally determined linguistic labels, and according to cultural constraints on what counts as legitimate emotions to experience and express (Turner & Stets, 2005). By foregrounding culture, the sociological perspective moves in the direction of groups of policy actors as important interpreters and signifiers of emotions: of what the emotional meaning of a situation is, of which emotional concerns are relevant or appropriate, and in which circumstances. Hence, emotions are no longer only considered to be individual, nor are they static. Rather, emotional orientations are relational, and produced and reproduced in the interactions between diverse policy actors. These policy actors express emotional orientations in shared and contesting emotional storylines that can change over time. Storylines are 'a generative sort of narrative that allows people to draw upon various discursive categories to give meaning to specific physical or social phenomena' (Hajer, 1995, 56). Likewise, emotional storylines can serve as condensed versions of larger emotional narratives through which policy actors interpret a policy conflict.
As we can see, there are differences in focus between the PCF perspective and the policy controversies perspective. The PCF perspective focuses on policy actors and their core beliefs. As a consequence, a study of emotions in policy conflicts might include an appraisal of facial expressions, behavior, and discursive expressions as filtered through individuals' core beliefs. The policy controversies perspective focuses on groups of policy actors who, by producing and reproducing storylines throughout their relationships and interactions, arrive at a discursive expression of emotions. Both perspectives are of value for studying the role of emotions in the escalation of policy conflicts. On the one hand, studies that focus on the emotions of individuals, and how these become part of escalating policy conflicts through particular strategies, or through leadership styles, may benefit from the PCF perspective. On the other hand, studies that want to shift attention to understanding the discursive expression of emotional storylines by interacting groups of policy actors in relation to escalating policy conflicts are better off with the policy controversies perspective. In this latter perspective it is possible to study how emotional storylines may be contested or become dominant in relation to other emotional storylines, and how these interactions may influence policy escalation or de-escalation. Our empirical analysis in this paper draws on the policy controversies perspective by exploring the relation between emotional storylines of groups of policy actors and (de-)escalation of policy conflicts. To do so, we need to theorize how conflict escalation by groups of policy actors may be traced in the discursive expression of emotions.

Tracing emotional storylines and their influence on conflict escalation
Policy actors use storylines as shorthand constructs, enabling a connection of a variety of broader and deeper narratives through which they make sense of what is happening around them by interpreting ongoing relationships, interactions, and events. Storylines 'serve to position social actors and institutional practices in ongoing, competing narratives' (Fischer, 2003, 87), and to order and re-order their understandings of the situation (Hajer, 1995, 56). They serve as condensed versions of narratives or framings that suggest unity in discursively varying and complex narratives, by making concise statements that allude to patterns or connections. To this we can add that storylines reflect the intertwined cognitive and emotional interpretations by policy actors of what goes on around them. By making sense of relations, interactions, and events, policy actors develop shared storylines carrying joint expressions of emotions, and as such can form coalitions. These coalitions around storylines represent different political realities (Hajer, 1995), and when these storylines are emotional ones they contain representations of the emotional aspects of political realities.
When employing emotions in their storylines, what might we expect policy actors to express? According to Ten Houten (2007, 3), emotions can be seen as 'ways of coping and adapting to the social situations that life presents'. Emotions are often acute, elicited by a situation, episodic in nature, focused on external objects (persons, organizations), and made public through actions and movements shown by facial expressions, behavior, and stories. Consequentially, emotions are short-term reactions to events and information in social situations (Ten Houten, 2007;Jasper, 2018). Policy actors may discursively express such short-term reactions by talking about anger, fear, disgust, surprise, shock, disappointment, and joy. In addition, these short-term emotional reactions may be accompanied or replaced by long-term positive or negative feelings about others or objects (affective commitments), which can be found in utterances of love, hate, (mis) trust, respect, or contempt. All of these can also be accompanied or replaced by longterm feelings of approval or disapproval that stem from moral intuitions or principles (moral emotions), which can be recognized in policy actors referring to shame, guilt, pride, indignation, outrage, and compassion (Jasper, 2018, 4). According to Halperin, it matters a great deal which emotions are communicated in conflicts, because the expression of specific, discrete emotions can form collective emotional orientations, which, depending on the discrete emotion at stake, can prolong and further escalate conflicts, or contribute to de-escalation and resolution (Halperin, 2016, 27).
The discursive expression of discrete emotions may take three different forms in relation to conflict (de-)escalation. First, research in social movements has shown that discrete emotions change, and are uttered by actors in specific combinations at fixed points in time. Jasper (2011, 291), for instance, discusses 'moral batteries', which are combinations of expressed negative and positive emotions such as anxiety and hope, or shame and pride, which motivate action of a movement's activists at fixed points in time. Second, Williamson (2011) theorizes that sequences of expressed emotions may form chains over time, as one or several discrete emotions transform into other ones. She shows empirically how a religious movement managed to keep people returning to a camp by transforming confusion, fear, and anger into courage, sadness, and mostly hope. In the same vein, Kleres and Wettergren (2017) demonstrate how hope and guilt help to transform fear into anger as a driver for action in climate activism, and Ransan-Cooper et al. (2018) show how anger drove the mobilization of the anti-coal seam gas movement in Australia, while the combination of anger and joy helped to sustain the movement over time. Third, another way in which emotional storylines may change is through the uttered intensity of discrete emotions. Turner and Stets (2005, 15-16) distinguish low, moderate, and high levels of intensity that can be traced in the specific words through which emotions are expressed. Fear may, for instance, be expressed as concern, anxiety, or horror, ranging from low to high intensity. Another example would be expressions of low to high intensity anger, ranging from irritation, to frustration, and fury. Based on a close examination of language use, we explore if the emotional storylines of groups of policy actors increase in intensity over the course of an emerging policy conflict.
By studying the combinations, sequences, and intensity of emerging emotional storylines, we aim to form an empirical understanding of the transformations in emotional storylines, and their relation to the moves and countermoves of groups of policy actors in the escalation of policy conflicts.

The case
Ever since natural gas production started in Groningen in 1963, the gas fields have been owned by the Dutch state and exploited by the Nederlandse Aardolie Maatschappij (NAM), a joint venture of Shell and ExxonMobil. The economic and political stakes of gas production are enormous, with total revenues of 288 billion euros for the national government and 29 billion euros for the NAM (Scholtens, 2018), a great dependence of most households on gas for cooking and heating, and large interests in the export of gas (Mulder & Perey, 2018).
All of this was rather uncontested until August 16, 2012, when the small village of Huizinge became the epicenter of an earthquake of 3.6 on the Richter scale. For decades, experts knew about the relationship between gas production and seismic activity, but predicted only minor damage (Mulder & Perey, 2018). The Huizinge earthquake and subsequent smaller ones proved these experts wrong, with a total number of 170,000 people experiencing damage to their houses, of which 85,000 experienced multiple instances of damage from consecutive earthquakes (Postmes et al., 2018). Soon after the Huizinge earthquake, the policy conflict escalated. Municipalities and the regional authority, together with action groups Groninger Bodem Beweging (GBB) and Schokkend Groningen (SG), started to contest national decision-making and the way in which the NAM dealt with claims by individual citizens. Citizen action group GBB started on 6 November 2009 as an interest group which assisted citizens with their claims for damages. After the Huizinge earthquake, GBB grew exponentially from about 200 members to approximately 3300 by April 2016 (interviews GBB1, interviews GBB2). Schokkend Groningen (SG) emerged from GBB on April 28, 2013, when two active members decided to form a more radical action group.
Since 2012 the conflict has escalated time and again, making it one of the longest-running policy conflicts in the Netherlands. Due to the impact on so many people's lives, the conflict is laden with emotional storylines, which are expressed by subnational governmental actors and the action groups as part of their contestation. Given the number of affected people, the high economic and political stakes and the prevalence of emotional storylines over a longer time period, the case is very suitable for our study.

Methods
We conducted a newspaper analysis of five national newspapers: NRC Handelsblad, De Volkskrant, Trouw, Algemeen Dagblad, and De Telegraaf, covering a broad political spectrum. In addition we analyzed the well-read regional newspaper Dagblad van het Noorden for more detailed coverage. Our analysis focused on the period between September 2012, shortly after the Huizinge earthquake, and April 2015, when the policy conflict started to cool off for a long period. From the LexisNexis academic NL database we retrieved 1053 national articles and 1771 regional ones. In Excel we created a dataset per year, per newspaper, and per number of articles, allowing us to map media attention for the earthquakes in Groningen (see Fig. 1).
We identified three peaks in media coverage: January-April 2013, November 2013-February 2014, and December 2014-February 2015. We checked if the media coverage between the peaks contained important developments in the case related to conflict escalation, which it did not. To limit our data set and to trace change over time, we analyzed all articles from these peaks, which amounted to a total of 1308 articles, 651 from the five national newspapers and 657 from the regional paper. The two authors coded inductively with Atlas.ti with regular crosschecking of how text and codes were interpreted. We used 1308 administrative codes for types of articles, 1002 administrative codes for subnational governmental actors and action groups, and 315 content codes for emotional storylines by these actors. These storylines were coded by looking for words and phrases that directly mentioned emotions such as: 'people are afraid that…' or 'the government should be ashamed of…'. We also coded for metaphors indirectly referring to emotions, for example by using 'heat' to express anger: 'people are hot under the collar', or 'they will be in for a hot summer' (Edwards, 1999;Verhoeven & Duyvendak, 2015). In the coding process, only direct quotes by subnational governmental actors and activists were included, leaving out emotions attributed to them by journalists. From this inductive analysis we further analyzed: (1) combinations of emotional storylines, (2) sequences of emotional expressions, and (3) increased or decreased intensity of expressed emotions through time. The combinations were analyzed per media peak, to see which emotional storylines were employed in a specific time period. The sequences can only be observed by comparing the emotional storylines between the different media peaks. Increasing or decreasing intensity of emotional storylines can especially be observed by comparing the media peaks, although changes of intensity within a peak are also conceivable. We used axial coding to label emotional storylines related to combinations, sequences and intensity.
In addition, we analyzed web content, such as blogs, stories of protests, and other materials used by subnational governmental actors, GBB, and SG, and we conducted three in-depth interviews with four members of the two action groups (which mainly serve as background information). The coding as described above was also applied to these sources. Since we conducted an exploratory study in which we aimed to identify patterns in the specific language used to express emotions, we did not further quantify our coding process to establish dominance in the patterns. Our main aim was to explore the possible patterns of expressed emotions. The absence of quantitative analysis is a limitation of this study. Our findings and coding scheme might be used for quantification in new research projects. The quotes chosen in the results are exemplary quotes for the emotional storylines expressed by governmental actors, and citizen action groups in this period. We do not present all coded quotes per media peak, but use the ones that are most telling about or representative of all these quotes.

Media peak 1: storylines about anxiety, anger, and questionable trustworthiness
The most prominent emotional storyline to be found during the first media peak is about anxiety. In this storyline, most of all subnational governmental actors but also action groups expressed the concern of citizens in the area about unsafety caused by the unexpected magnitude of the earthquakes and the ensuing damage, and they used expressions of anxiety to argue what should be done. For example, the regional deputy expressed this storyline as follows: "I am scared by the risks for north Groningen. Earthquakes of 5 on the Richter scale. That is quite something. (…) The minister should come to explain why north Groningen is exposed to these risks. He should take away people's anxiety that things will get out of control" (Regional deputy, DvhN, 26-01-2013). 1 Or, as the mayor of one of the cities in the region expressed it in a national newspaper: "There are concerns here about heavier earthquakes caused by gas production in Groningen. People are getting scared, asking: how does this end?" (NRC, 7-02-2013).
The same storyline was also expressed by action group GBB, for example by using the emotional language of 'not feeling safe': "I have talked a lot with people and seen what it does to them. It paralyzes them. They cannot do anything and do not feel safe any longer. Gradually it controls their whole life. I know people who sleep with their clothes on, and who run into the street when they feel something" (DvhN April 9, 2013). This quote also illustrates how GBB activists tried to make sense of what the people in the region were feeling. Similar to the subnational governmental actors, GBB expresses anxiety and emphasizes the expectation of something bad happening (the fear of earthquakes), and shows how it affects people (they sleep with their clothes on).
In addition to this prominent storyline about anxiety, two others emerged during this period, which would become much more important in later stages of the conflict. The first storyline is about anger, which was initially voiced by the SG activists. In this storyline, the core message was that the earthquakes are induced by gas production, and that the damages caused by the earthquakes but also the way governmental actors are responding, raises anger. The SG activists for example said that they most of all felt anger: "most of it is plain anger about what they inflict upon us, and not only us but in fact all Groningers" (interview SG). The anger storyline tied together personal frustrations and injustices inflicted on the citizens of Groningen by the gas producing company NAM and the Minister of Economic Affairs.
The second storyline that emerged was one about trust. In this storyline, the line of reasoning was that to prevent anxiety and anger from further deepening, the national government needed to start acting in a trustworthy manner. This storyline was for example uttered by one of the regional authorities: "It cannot be the case that citizens and businesses suffer indirect damages or are affected by drawbacks as such. A trustworthy and proactive approach to this issue is required" (DvhN, 9 February 2013). The overall story that emerged in peak one was that in order to make citizens less anxious or angry, the national government needed to start acting in a more trustworthy way.

Media peak II: storylines about fear, distrust, and contempt
During the second media peak, the emotional storyline of anxiety intensified into fear. In this storyline the core idea is that citizens are no longer just worried but are starting to experience fear, and are feeling very unsafe. For example, the action group GBB launched a video campaign based on the expression of fear. GBB's narrative in these videos was that gas production leads to severe earthquakes and that the unknown risks cause fear. In the production of three videos, exemplary narratives were told: a young girl that felt unsafe, a man who could not sell his house in order to leave the area, and an older woman who could not sleep due to emotional problems (interview GBB1, interview GBB2). 2 In the example of the young girl, we hear a woman's voice saying: "This is Mieke. Mieke does not sleep well. Not due to hunger, sadness or loneliness, but due to fear. Fear of earthquakes through gas extraction in Groningen" (YouTube, accessed February 7, 2017). Compared to the first media peak, when anxiety was expressed in terms of 'concerns', 'being scared', or 'taking away anxiety', the direct usage of the term fear in relation to earthquakes suggests an increase of the intensity of this type of emotional storyline. Concern and being scared can be considered low to moderately intense expressions, whereas the term fear is associated with a high intensity expression related to an imminent threat (Ten Houten, 2007;Svendsen, 2008, KNMI 2013. In this peak, the storyline of anger transformed into contempt. This was most of all visible in discursive expressions by the radical action group SG. From early 2014 onward, SG started to make judgments about 'the other' as immoral, not willing to correct their wrongdoing, and to be looked down upon (Ten Houten, 2007;Fischer & Giner Sorolla, 2016;Jasper, 2018). For example, in January 2014, during a hearing of a parliamentary committee on location in Groningen, a core member of SG, when asked how trust could be restored, cried out: "Help people with problems, and first people with psychological problems. Do you want to have it on your conscience if people are driven to suicide? 3 " Not helping people with psychological problems was cast as immoral, and as something which nobody wanted to have on their conscience. The contempt-ridden storyline stretched beyond the minister and members of parliament, by also including the NAM. An example can be found in an email to departing NAM CEO Bart van de Leemput which SG published on their website. Part of this email read as follows: "there are pending crime reports against you and Henk Kamp [the minister], how about those? Have you been to a police station? Did they interrogate you? If so, how does it make you feel? Do you feel guilty as well? Do you sleep well? It is the sleep of the innocent?" 4 In the email, the NAM's CEO was sketched as incompetent and immoral because he had, in their opinion, risked many lives. With the expression of contempt, there is no longer a call upon those considered responsible to redress their wrongdoings, as would be the case when anger is expressed, but instead contempt is expressed by threatening the responsible persons, distancing from them, and looking down on them (Fischer & Giner-Sorolla, 2016).
Meanwhile, the trust storyline from the first media peak, which talks about a 'trustworthy approach' being needed to make people feel less anxious and angry, was transformed into a distrust storyline by alluding to 'the breach of trust' that is 'growing'. For example, the leader of the regional authority expressed how current decision-making and policy-making falls short and 'feeds' distrust: "The breach of trust only grows the longer the answer by the government to uncertainties concerning the safety of gas production and proposals to maintain livability fail to appear. It feeds the thought that nothing will come from safety and compensation" (DvhN, December 21, 2013). Hence, in this second episode of the conflict, the overall story intensified in emotions and this was connected to a growing breach of trust. For those who were expressing anxiety in peak one, this now became fear. The discursive expression of anger had transformed into contempt. In connection to this intensification of emotional storylines, a third storyline about growing distrust became more important: the responsible authorities and industrial actors had not dealt with the issues in a trustworthy manner, which made distrust in them grow.

Media peak III: storylines about anger and contempt
In our analysis we see that the two predominant emotional storylines during the third media peak further shifted toward anger and contempt, and that these were related to distrust in national government and the NAM. In response to the Dutch Safety Board, for example, subnational governmental actors expressed their anger -and not only that of their citizens. For example, the Mayor of the City of Groningen argued that the people in the region of Groningen were being 'shamefully misled', and called for a parliamentary investigation (De Volkskrant February 19, 2015). Another regional official was very annoyed, calling for the recognition of people that complained about the government, and for drastic measures against prevailing forms of decision-making: "Their remarks were put aside as nonsense. They should be given recognition in retrospect. (…) Safety was subordinate to financial interests. The gas bastion needs to be forced open" (DvhN February 19, 2015). Both the mayor and the regional deputy focused on the unjustifiable acts by the national government and the NAM. The deputy added a recommendation to restore the violation of the moral order by forcing open the 'gas bastion', a metaphor referring to a fortification or a stronghold which would take quite some effort to subjugate. Such unjustifiable acts and moral violations are strongly connected to anger in emotion literature (Turner & Stets, 2006, Jasper, 2018, Ten Houten, 2007. Similar to subnational governmental actors, the rather moderate GBB activists in this period started to employ anger storylines by addressing the minister. For example, GBB's secretary Dick Kleijer remarked in the regional newspaper: "Even when the State Supervision of Mines gave the recommendation to substantially lower the gas extraction in January 2013, the minister did not follow up on that. We blame him very much for that. Trust is further away than ever" (DvhN January 12, 2015). Here we see that also the action groups started connecting anger storylines to trust or distrust storylines related to the national government. They blamed the minister for unjustifiable acts, which is a clear sign of referring to anger. In addition, this blaming of the minister was connected to distrust.
This distrust was also expressed by the other action group, SG, that gatecrashed the Province of Groningen's New Year's Reception in January 2015 and took to the stage uninvited to make a speech, in which they decried that the Netherlands 'is no longer a democracy but a dictatorship'. 5 One more example of how contempt got connected to distrust in this episode of the conflict: in a parliamentary debate in The Hague with the minister, a core member of SG screamed from the public gallery: "Henk [first name of the Minister], if you want to have my wife's death on your conscience you will have a problem" (DvhN February 13, 2015). In this third episode of the conflict, the main story emotionally intensified from fear to anger to contempt about the consequences of the induced gasquakes, and these emotions led to distrust most of all in national government, and for some even in the whole democratic system.

Conclusions
The Groningen case shows that emotional storylines expressed in a case of heated policy conflict comprise the following negative emotions: anxiety/fear, anger, and contempt. Within each media peak, one overall story emerged in which these emotions were connected in different ways to a problematization of the trustworthiness of government actors and the NAM. A pattern emerged of moving from anxiety to fear, to anger and contempt in connection to increasing distrust. The combinations and sequencing of the emotional storylines, as well as their increasing intensity, suggest a process of emotionally expressed escalation.
The combinations of storylines in each peak all convey negative emotions: anxiety, anger, and questioning trustworthiness (peak 1); fear, contempt, and a warning for distrust (peak 2); and anger, and contempt, with distrust (peak 3). When interpreting these specific negative emotional storylines with the help of existing literature on emotions, we see that anxiety and fear are usually related to anticipated risks. With anxiety/fear storylines, policy actors express the expectation of 'evil' or of an imminent threat, and there is not a concrete wrongdoer (Ten Houten, 2007;Jasper, 2018;Svendsen, 2008). Anxiety and fear in our case study were about the magnitude of the induced earthquakes. The anger storyline indeed pointed to specific actors that commit unjustifiable acts which require a repairing of the moral order, something which is also recognized in the literature (Turner & Stets, 2006, 553-554). Contempt storylines convey a judgment of specific actors that looks down on them as 'stupid, incompetent and immoral', and not willing to correct their wrongdoing (Fischer & Giner Sorolla, 2016, 346-347;Ten Houten, 2007;Jasper, 2018). From the literature, we also know that trustworthiness/distrust storylines signal a lack of legitimacy caused by actors not operating according to formal or informal rules (Jasper, 2018, 113). Hence, in the anxiety/fear storylines the issue at stake is a threat in the future that may be difficult to control. In the anger and contempt storylines the problem lies with specific actors that are to blame (e.g., Benford & Snow, 2000). Finally, storylines of trustworthiness/distrust point to the issue being a lack of legitimacy for those considered responsible for causing or solving the problem.
A second element in the emotionally expressed escalation that is relevant for understanding how escalation takes place, is the sequences of emotional storylines that develop over time. The main change between the three media peaks is that anxiety (peak 1), and fear (peak 2), were followed by anger (especially during peak 3), and contempt. We know from other research that the change from anxiety/fear to anger can provide a strong motivation for people to engage in protest (Jasper, 1998;Verhoeven & Duyvendak, 2015). In particular, moderate people can be triggered by the anger storyline to take action. The literature also shows that shifting from anger to contempt can offer a motivation to people with more radical views to also act in more radical ways (Tausch et al., 2011).
A third element in the emotionally expressed escalation was that over time we see indications of the growing intensity of emotional storylines. The first important change of emotional intensity is between media peaks I and II, in which the anxiety storyline shifted to fear, suggesting a more intense expression of this emotion. Second, over time the intensity of emotional storylines also changed from anger to contempt, and we observed a shift from the questioning of trustworthiness of those considered responsible, to a distrust storyline.
These findings give in-depth insights into how emotional storylines are a form of policy conflict escalation in itself (cf. Wolf & van Dooren, 2017) that develops in relation to and on top of other well-known forms of escalation such as increasing issue complexity, use of heavier tactics, use of resources, support by allies and psychological distancing (Glasl, 1982;Rubin et al., 1994). Policy conflict escalation in terms of emotional storylines takes place in specific combinations of expressed discrete emotions; through changes in sequences of emotional storylines; and through an increased intensity of policy actors' expressed negative emotions.
Within this paper, we have initiated an exploration of the relationship between emotional storylines and policy conflict escalation. Analysis of the media peaks indicates that specific decisions of the Dutch national government may have fueled the escalation in terms of emotional storylines. For example, in the first peak, the national government decided to reduce gas production in the area but citizens and regional officials demanded more reduction, while in peak III a report was published that concluded that economic interests had always been more important in decision-making on gas production than people's safety, which stirred anger storylines (De Volkskrant February 19, 2015). However, the relations between these types of decisions by proponents of a policy, and the expression of emotional storylines by opponents, need further investigation.
Our study also has limitations in that the empirical part is based on a single case. To overcome this limitation, future interpretive studies can develop other case studies, preferably comparative ones, to further confirm or reject our findings. In addition, future interpretive studies can quantify emotional storylines through big data analysis, without losing the nuance of discursive expressions involved in the growing emotional intensity of these storylines. This would require context-sensitive machine learning by, for example, developing elaborate libraries of emotion words and frequently used metaphorical expressions.
Our study contributes to the PCF approach to policy conflict in at least two ways. First, the different combinations and sequences of emotional storylines that we found contribute to the idea of policy threats in the PCF. Policy actors may interpret developments as threatening, and our study suggests that this interpretation can lead to an emotional escalation not only at the behavioral level of individuals but also at the social level through shared storylines of anxiety, anger, contempt or distrust. Second, the PCF suggests that the intensity of policy conflict depends on whether policy actors are willing to compromise on their policy positions. Our findings indicate that sequences of emotional storylines that develop over time from anxiety, to fear, anger and contempt, can contribute to policy actors' willingness to engage in contention. Similarly, the words used to express such emotional storylines can become more intense over time. By studying sequences and intensity of emotional storylines, more detailed insights can be developed on why policy actors may become ever more deeply convinced of defending their policy position and increasingly engage in contentious behavior. Future research that attempts to integrate our approach to emotional storylines with the PCF could try to combine the analysis of media material with extensive in-depth interviews among individual policy actors, as well as surveys to get a better impression of the role emotions play in contending policy positions. In relation to policy conflict escalation, it would be interesting to investigate: (1) if there are differences between the emotional storylines of individual policy actors and groups of policy actors; (2) how these storylines influence each other; (3) what role core beliefs and frames play in the construction of these storylines; and (4) how emotional storylines by individuals or groups interact with behavioral aspects such as the use of strategies and tactics. To capture the current state of knowledge on the role of emotions in policy conflict escalation, a 1970s song by the Carpenters is brought to mind: 'We've only just begun'.