Abstract
Political actors increasingly use the microblogging service, Twitter, for the organization, coordination, and documentation of collective action. These interactions with Twitter leave digital artifacts that can be analyzed. In this article, we look at Twitter messages commenting on one of the most contentious protests in Germany’s recent history, the protests against the infrastructure project Stuttgart 21. We analyze all messages containing the hashtag #s21 that were posted between May 25, 2010, and November 14, 2010, by the 80,000 most followed Twitter users in Germany. We do this to answer three questions: First, what distinguishes events that resulted in high activity on Twitter from events that did not? Second, during times of high activity, does the behavior of Twitter users vary from their usual behavior patterns? Third, were the artifacts (retweets, links) that dominated conversations during times of high activity indicative of tactical support of the protests or of symbolic association with it?
Through a Glass, Darkly: The Online Traces of Collective Action
Political actors use the microblogging service Twitter increasingly to organize, coordinate, document, or comment on collective action, such as issue campaigns, protests, or strikes. In these contexts, Twitter enables, or at least facilitates, collective action and serves as a space for participants and bystanders to communicate about the events itself and corresponding issues. Although Twitter clearly has practical value in the coordination, organization, and documentation of collective action, it is also a communication space where the meaning of issues, campaigns, or events is negotiated between participants, sympathizers, critics, political elites, and bystanders. The nature of this space therefore leads to the formation of issue publics on Twitter, which can be clearly identified by specific hashtags. Through the actions of participants in these issue publics (e.g., retweets, links to external content) artifacts (e.g., tweets, links), rise in prominence and even come to dominate a communication space. In this article, we use the term communication space to refer to all messages referencing a common topic (e.g., Stuttgart 21) on one Internet service. We do this to differentiate from the term sphere, which focuses on all communication on a device or service (Rogers, 2013, p. 33), and the term issue public, which could refer to communication on a given topic on multiple devices or services. We use the term communication space in this article to refer to all communication commenting on Stuttgart 21 on the Internet service Twitter. The aggregate behavior of users in a communication space becomes a filter through which artifacts emerge that are of collective relevance to the participants in an issue public. Twitter thus becomes a tool for the organization and tactical support of collective action, the public negotiation of meaning, and a space in which the aggregate behavior of participants emphasizes artifacts of collective significance.
When examining how Twitter is used for collective action, we can identify two general uses. First, there are users who use Twitter for the coordination and organization of protests. These users also link to outside sources, informing other users about the reasons and contexts of the protests. Users may also make use of Twitter for documenting protests, as they are unfolding by linking to alternative media or photographs. We think these uses qualify as tactical support of collective action. Other users use Twitter for public association with protests, phatic commentary, or for the publication or linking of satirical and ironic content. We consider these uses as forms of symbolic association of Twitter users with collective action. Depending on the dynamics of specific campaigns, protests, or discussions, one or the other usage practice may dominate the online public discourse, or there might be a mix of both usage practices.
We will show that both these usage categories were simultaneously present during different phases of the protests in our case study. Furthermore, we demonstrate how their respective prominence changed when specific events led to a sudden increase in public attention. Considering these changes, we will work off the preliminary hypothesis that, the two uses are connected to different user types. Tactical support seems a usage pattern more prominent among core supporters of a protest. Core supporters are actively concerned with the success of the protest and thus disseminate potentially crucial information. We will call these users digital core supporters. Meanwhile, symbolic association seems to be a usage pattern more prominent among users who sympathize with the protests, but are not necessarily actively involved in the organization and coordination of protesters. These users frequently choose the format of phatic or satirical commentary to express their emotional or symbolic association with the protests. We call these users digital bystanders.
In this article, we analyze Twitter messages commenting on one of the most controversial protests in recent German history, protests against the infrastructure project Stuttgart 21. We analyze all messages containing the hashtag #s21 posted by the 80,000 most followed Twitter users in Germany between May 25, 2010, and November 14, 2010. With our analysis, we aim to answer three questions: First, what distinguishes events that resulted in high activity on Twitter from those events that did not? Second, during times of high activity, does the behavior of Twitter users vary from their usual behavior patterns? Third, were the artifacts (retweets, links) that dominated the communication space during times of high activity indicative of tactical support of the protests or of symbolic association with it?
Stuttgart 21: The Protests and Their Relevance
Stuttgart 21 is an ambitious infrastructure project in Germany’s southwest. Since the autumn of 2010, the term is also synonymous with one of the longest series of heavy protests in Germany’s recent history. In cooperation with federal, state, and municipal governments, Germany’s biggest railway company (Deutsche Bahn) planned an overhaul and extension of its railway network in the region. One of the project’s keystones was the partial demolition and reconstruction of the railway station in Stuttgart, the state capital of Baden-Württemberg. The project and the construction works on Stuttgart’s railway station were highly controversial. Critics pointed out that the project would foreseeably run over budget, with the taxpayer likely being coerced into picking up the tab. Another highly emotional point of contestation was that, in order for the construction of the new segments of the railway station to go ahead, old trees in the immediate proximity of the construction site had to be cut down. These issues led to month-long protests in Stuttgart. During this period, established political actors and parties positioned themselves in support or opposition of the project. The protests reached a high point on the night of September 30 to October 1, when violence between police and protesters escalated. Roughly 400 protesters were hurt that night in clashes with the police. A public outcry that followed condemned the police response and the perceived role of Baden-Württemberg’s state government in the violent escalation of the dispute.
Figure 1 shows a detailed timeline of the major events of Stuttgart 21 between February and December 2010 (all graphs in this article were produced using R, R Core Team, 2012 and ggplot2, Wickham, 2009). This account of the events is based on newspaper reports and activists’ websites. The planning stages of Stuttgart 21 go back to the 1990s. For much of the time, the project was controversially discussed and sometimes even actively protested against. In 2010, these protests regularly counted participants in tens of thousands. The protests gained national attention in July 2010, when the construction work at Stuttgart’s train station proceeded to demolish the station’s north wing and cut down adjacent trees. The protests escalated on September 30. After the events of September 30, mediation talks between opponents of the project and developers were scheduled and lasted from late October to early December 2010. The results of the talks were seen as largely backing the developers and supporters of Stuttgart 21, while exhausting public interest in the project and the accompanying protests. The following year, a statewide referendum was held in Baden-Württemberg in which the supporters of the project won (Gabriel, Schoen, & Faden-Kuhne, 2013). We will focus on Twitter messages commenting on Stuttgart 21 from May 25 to November 14, 2010, thus documenting the early phases of the protests, the escalation on September 30, and the direct aftermath.
Key events during the protests against Stuttgart 21.
The Internet played a significant role in the organization, coordination, documentation, and the crafting of the public narrative of the on-sight protests against Stuttgart 21. Activists used the Internet and Twitter as tactical tools for the coordination of the protests. For example, activists at the construction site in Stuttgart regularly used Twitter to inform supporters when the construction workers prepared for new stages of the construction process (e.g., the arrivals of diggers to demolish the station’s north wing or the preparations to cut down trees). In response to these updates, activists mobilized to block or protest these developments. Activists also used Twitter to link to live video streams documenting the protests. Thus, Twitter became an access point to alternative media, information sources that activists used to document the protests from their own point of view, bypassing the selection process of traditional media (Atton, 2002; Couldry & Curran, 2003; Poell & Borra, 2013). Also the Internet, and specifically Twitter, was an important tool for nonparticipants, digital bystanders, to follow the events in Stuttgart and comment on them. Media stories about the protests often included references to their nature as grassroots, bottom-up, or Internet enabled, and went on to elaborate on the role of the Internet for the organization of the protests (Kuhn, 2010; Mader, 2010). Stuttgart 21 must also be seen as a crucial step in the evolution in Internet-enabled activism that we have seen in other political arenas in Germany in recent years (Jungherr & Schoen, 2013).
Twitter in Collective Action: A Tool for Curation, Tactical Support, and Symbolic Association
Twitter is an online service that allows the publication of short text messages (up to 140 characters in length). This encoded restriction quickly led to the emergence of usage conventions that enabled users to interact and to establish the context of their messages without using many characters. These conventions (e.g., @message, @mention, retweet, hashtags, link shorteners) enable researchers to identify usage patterns by automated text analysis. Some of these usage conventions can be connected to functions of Twitter for collective action. Also, researchers have shown that usage patterns of these cultural conventions systematically differ depending on the type of topic or event Twitter users are commenting on (Bruns & Stieglitz, 2012, 2013; Lin, Keegan, Margolin, & Lazer, 2013). Differences in usage patterns have also been documented during different phases of an event that was commented upon (Segerberg & Bennett, 2011). In Table 1, we document how we connect these usage conventions to various functions of Twitter for collective action.
Functions of Twitter Based on Usage Conventions.
On Twitter, users can anchor their messages in a given context using topical key words and by preceding them with a hashtag (#) sign. The service can be searched for all messages with a given hashtag. Thus, users can read messages by others on given topics and, if they choose, interact. These attributes of the platform lead to the emergence of issue publics that constitute communication spaces.
In these communication spaces, users, messages, or linked content can rise to prominence. This is the result of a collective filtering process by the aggregate use of @messages, @mentions, retweets, and links by participants in a given issue public. If many users retweet a given message, this indicates that they see the message as relevant to the issue public and thus want to spread it. The same goes for prominently linked content. This collective filtering process has also been called curating, gatekeeping, or gatewatching (Bastos, Raimundo, & Travitzki, 2013; Bruns, 2005; Meraz & Papacharissi, 2012; Papacharissi & de Fatima, 2012; Segerberg & Bennett, 2011). Regardless of the label one chooses to apply, this is a highly relevant practice since it determines which messages and links come to dominate the Twitter discourse on a given hashtag and thus influences heavily how participants and outsiders perceive this discourse.
Several studies have shown that the volume of Twitter messages commenting on a given political topic rises when events relevant to the topic are happening (Jungherr & Jürgens, 2013a, 2013b; Segerberg & Bennett, 2011). There seems to be a correlation between collective action events offline and Twitter coverage. Still, as of now this correlation has not been shown to be deterministic. Some political events lead to high activity on Twitter while others do not.
In this article, we will be looking at the events in the #s21 communication space that resulted in high activity on Twitter and identify shared characteristics. We approach this analysis informed by news value research, an approach that identifies key characteristics of news and events that attract media coverage and audience attention (Donsbach, 1991; Eilders, 1997; Schulz, 1976). Like modern news value research, we understand these characteristics to mirror the criteria of relevance that Twitter users employ before tweeting on any given topic.
In a next step, we will analyze those messages and links that dominated the #s21 communication space during times of high activity. We thus rely on the collective curating process described above to find messages that held collective meaning for the participants in the communication space. We focus on the content of the most retweeted messages and the destinations of the hyperlinks that were used most frequently during time intervals of high activity. For categorization of these messages, we introduce two concepts for Twitter use in issue publics commenting on collective action: tactical support and symbolic association.
One of the characteristic functions of Twitter and other information and communications technologies (ICT) for collective action is their support in the organization, coordination, and documentation of campaigns and protests (Bimber, 2003; Bimber, Flanagin, & Stohl, 2005, 2012; Earl & Kimport, 2011). Twitter has been identified as a tool for political organizations and activists to distribute information (Howard & Hussain, 2011, 2013; Lotan et al., 2011), to widen an active audience for collective action (González-Bailón, Borge-Holthoefer, & Moreno, 2013; González-Bailón, Borge-Holthoefer, Rivero, & Moreno, 2011), or as a tactical tool for the coordination of collective action (Gerbaudo, 2012; Segerberg & Bennett, 2011; Verma et al., 2011). We collect these uses under the term tactical support. To identify this usage pattern in our data, we analyzed the content of messages and links that were dominant in the #s21 communication space during time intervals with exceptionally high message volume.
Another function of Twitter for collective action is the symbolic association of Twitter users with issues, campaigns, or protests by attaching topically relevant hashtags to their tweets. Using hashtags related to collective action users can voice support, critique, or comment on the issues in question. In doing this, they do not necessarily become supporters or active participants of an organization, issue, campaign, or protest but they publicly interact with it, be it in support or in opposition. We call this usage pattern symbolic association. Similar practices have been called in other contexts active spectatorship (Kreiss, Meadows, & Remensperger, 2013), connective action (Bennett & Segerberg, 2012), or the personalization of collective action (Bennett, 2012; Bennett & Segerberg, 2011). To identify this usage pattern in our data, we analyzed the content of the messages and links that were dominant in the #s21 communication space during time intervals with exceptionally high message volume commenting on Stuttgart 21.
The Data Set
The observations discussed above seem to hold strong potential in the analysis of Twitter messages commenting on Stuttgart 21. The case offers a unique hashtag related to the protests that was used over a period of several months. During this time, the protests underwent various transformations and gained significant public attention.
Our analysis is based on Twitter messages posted by the 80,000 most followed Twitter users in Germany. Of all the messages these users posted, we concentrated only on the messages commenting on Stuttgart 21 (messages were identified by their common key word #s21). To generate our original sample of users, we queried Twitter’s streaming application programming interface for random messages. These messages were checked against a series of tests (i.e., locations, time zones, presence of letters typical to German) to determine whether their authors were posting in German from Germany. Based on this initial sample, we collected the accounts these users subscribed to (in Twitter parlance: followed) or were subscribed by. The resulting list of accounts was ordered based on the number of their followers. Thus, we collected a sample of the 80,000 Twitter users who, based on their follower counts, were the most popular users in Germany during the time interval in question (Jungherr & Jürgens, 2013b; Jürgens, 2010).
We used automatic as well as manual analysis procedures to examine the data. The calculation of total message counts and the use of different communication conventions (@messages, @mentions, retweets, hyperlinks) were done by computational procedures. The content of the most popular retweets and the destination of the most popular hyperlinks were coded manually.
Tactical Support and Symbolic Association in the Stuttgart 21 Communication Space
Of the 80,000 users in our sample, 12,916 used the hashtag #s21 at least once during this time span. These messages are the basis for the following analysis. In Table 2, we document the conventions used in messages containing the hashtag #s21.
Conventions in Use in Messages Containing s21 Between May 25, 2010, and November 14, 2010.
Figure 2 shows the distribution of messages per day; we see that messages containing #s21 appear in clusters centered on a few prominent dates, the most obvious being September 30 and October 1 when 46,546 messages were posted containing #s21. Nearly 30% of all mentions of #s21 over nearly 6-month period occur in a time span of just 2 days. Other obvious spikes are visible between August 25 and September 1 (12,163 messages), September 6 to September 9 (5,877 messages), October 2 to October 7 (28,658), and October 10 to October 17 (17,636 messages).
Number of s21 messages per day.
Another perspective on the events of Stuttgart 21 and the distinct phases of the protest offers the analysis when individual users used #s21 for the first time. Figure 3 shows how many users started to use #s21 on any given day. We identify four distinct phases of the protest: May to late July marks the beginning of the protests. By August 1, only 311 Twitter users had used the hashtag #s21 in their tweets. From August to September 30, we observe a significant increase of users joining the conversation: Between August 1 and September 29, another 3,058 users had used #s21. On September 30 and October 1, the day and the direct aftermath of the clashes between police and protesters, 5,821 users used the hashtag for the first time. The last period in our graph between October 2 and November 14 is marked by a steep increase of users posting the hashtag #s21 for the first time followed by a somewhat flattened increase from late October onward. These phases correspond with the phases apparent in the events themselves as shown in Figure 1. Up until August, the protests found little public attention. This started to change between August and September when the protests rose in number and frequency. When the events in Stuttgart finally escalated on September 30 and October 1, the protests became a topic of highest media attention and national interest. Figure 3 shows these phases.
Number of users posting s21 for the first time per day.
These data indicate a significant change in the composition of the communication space initiated by the events of September 30. Before September 30, Stuttgart 21 was a topic that generated some interest on Twitter. The events of September 30 massively increased the activity on Twitter as the media picked up on the escalating events in Stuttgart. It is safe to assume that users commenting on #s21 on this day for the first time were differently motivated than users tweeting about the events before this. Users contributing to the #s21 communication space before this date did not need massive media attention to be active participants. For us, this is an indicator that these users were motivated by the events themselves. They are digital core supporters. Users contributing on September 30 for the first time to the communication space #s21 reacted to media and online attention and voiced their support or critique of the protests. They are digital bystanders. The difference between these user groups can also be seen in their ongoing participation in the #s21 communication space after October 1.
The sudden increase in users tweeting about Stuttgart 21 on September 30 and October 1 raises the question of whether these new users, who were clearly attracted to the topic by the sudden increase in traditional media and online coverage, continued to actively comment on Stuttgart 21 after the first wave of public attention ebbed. Of the 5,821 users using the hashtag #s21 for the first time on September 30 and October 1, only 42% posted at least one message using #s21 during the remaining days in our data set. In comparison, the 3,359 users who had used #s21 prior September 30, as well as on September 30 and October 1, roughly 61% continued to do so after October 1. While the events thus led to a significant increase in users actively commenting on Stuttgart 21, the initial attention of most of the users gained on September 30 and October 1 did not translate in prolonged activity in the communication space.
Let us now focus on four points on the timeline of the protests that led to heavy activity on Twitter. August 25 and 26 (4,415 messages), September 6 and 7 (3,280 messages), September 30 and October 1 (46,546), and finally October 16 and 17 (3,849 messages). We identified these days by comparing the day-to-day volume of messages containing #s21 and looking for sudden volume increases. In order to determine the composition of the communication space during these time intervals with greater accuracy, we included the day directly following each day that showed sudden increase in volume in our analysis.
As expected, on all days with high activity in the #s21 communication space events took place on sight that were clearly connected to the protests in Stuttgart. Two of these events are connected to crucial moments in the protests. August 25 was the day of the demolition and subsequent occupation of the station’s north wing; September 30 was the day of heavy clashes between protesters and police. On both these occasions, big offline protests translated to high activity on Twitter. The other two events are also connected to the protests, but on- and offline-significance do not correlate as clearly as in the two cases before. On September 6, a regularly scheduled demonstration took place in Stuttgart. During this protest, an alteration between a policeman and a protester took place, resulting in an alleged beating of a woman by the policeman. Also a member of the local Pirate Party was arrested. Both these events had no greater significance for the protests at large but created strong reactions online. On October 16, protesters barricaded themselves in the south wing of the train station. The police had to break down a metal door that blocked the area occupied by the protesters. This event was documented by video, images, and ironic commentary online and thus generated high activity on Twitter. Again, the relevance of these events for the overall protests against Stuttgart 21 is negligible; yet, they triggered very high online activity.
The discussion of these four events shows that the analysis of messages with hashtags connected to collective action allows for qualified assessment of the corresponding offline events; yet, these assessments do not provide us with a complete picture. For instance, if we compare Figures 1 and 2, we observe that virtually no step taken by officials in response to the protests against Stuttgart 21 resulted in strong online reactions. Also, not every protest event found a strong echo on Twitter. While patterns in online data are therefore clearly connected to offline events, significant offline events are not necessarily echoed by high activity online.
What do the events, which triggered strong reactions online, tell us about the criteria of relevance that led to high activity in the #s21 communication space? Each of the events offered new and surprising escalations of the protests: on August 25, protesters physically blocked the beginning of the demolition works; on September 6, the conflict between police and protesters escalated by the alleged beating of a woman by a policeman; on September 30, large-scale clashes between police and protesters made the events in Stuttgart a national news event; and October 16, offered surprising pictures of policemen breaking down a metal door protesters were barricaded behind. Beyond that, each event offered pictures or testimonies of individual protesters clashing with police forces, with protesters sometimes even suffering injuries. Many of the messages referring to these key episodes linked to related images and offered phatic or ironic commentary on the events. The events that led to heavy activity on Twitter all shared attributes that are also very influential in the decision of media consumers to select news items: surprise, harm, controversy, and personalization (Donsbach, 1991; Eilders, 1997; Eilders & Wirth, 1999). In the case of #s21, Twitter data offer no unbiased account of events as they happened but instead document a version of the events, an edited view guided by the collective attention of participants in the communication space with a bias that allowed for strong emotional responses, visualization (Ruhrmann, Woelke, Maier, & Diehlmann, 2003), and ironic commentary.
What about the user behavior during these events? Was it roughly similar, or can shifts in user behavior tell us something about the nature of the events Twitter users were commenting on? Table 3 shows that August 25/26 and September 6/7 show roughly comparable activity in the #s21 communication space. As was to be expected, September 30 and October 1 show much higher activity. On October 16/17, the activity ebbs somewhat, but is still higher than on the days before the events of September 30. Table 3 documents the frequency in which communication conventions were used during the different time spans under examination. Here, strong difference in usage patterns become apparent between September 30/October 1 and the other three time spans under examination. First of all, we see that the average number of daily messages that users posted is massively higher during the time spans of peak activity when compared to the average as calculated over the whole time span. Second, we see that September 30 and October 1 showed still higher average activity when compared to the other three time spans. On those days, retweets made up a much bigger percentage of all messages posted than during the other three time spans. Also, @messages and @mentions were much less common than during the other time spans. These patterns show that on September 30 and October 1 the #s21 communication space became much more active, more focused on distributing information and less focused on interaction between users than during other time spans. This observation coupled with the sudden influx of new users commenting on Stuttgart 21 on September 30 and October 1 makes it safe to assume that on these days the sudden increase in attention led to a significant change in the composition of users participating in the communication space. In the following section, we will examine whether this shift can also be seen in the content analysis of the tweets contributing to the #s21 communication space.
Usage Patterns During Different Time Spans.
In the next step of our analysis, we focus on those artifacts that through a selective curating process came to dominate the #s21 communication space during the four time intervals of high activity. To do this, we selected the 100 messages that were retweeted most often and the 100 hyperlinks that were linked to most often during the time intervals in question. We then examined the content of these messages and the destination of these hyperlinks for evidence of tactical support and symbolic association.
In this analysis, we made two choices that have to be justified. First, we focused only on popular retweets and links and second, we restricted our analysis only to the 100 most retweeted tweets and posted links. In concentrating on retweets and links, we only cover the content of messages in the #s21 communication. Alternatively, we could have also used @messages or @mentions to determine users that were frequently referred to or mentioned and code the content of their messages. This approach would have allowed us to assess the role that tactical support and symbolic association of #s21 would have played in the content of their messages. Instead of this user-centric approach, we focused on the content of messages and links that dominated the #s21 communication space. For us, retweets and hyperlinks are strong indicators that users posting them on Twitter at least implicitly thought their content or their sources relevant to their followers. Thus, coding their content allows us to assess which type of content—tactical support or symbolic association—emerged as relevant through the collective curating process.
Furthermore, the choice of our cutoff point at 100 messages and hyperlinks might seem somewhat arbitrary. However, closer examination showed that by coding the top 100 artifacts in each category during each of the four time intervals, we could account for a significant portion of all retweeted messages and all posted hyperlinks during each interval. Our rationale for choosing this approach was to focus on those artifacts that rose to prominence through the collective curating process; in this manner, we could determine whether messages and hyperlinks that participants in the communication space thought relevant could be characterized as tactical support and symbolic association. The trade-off is that, we cannot offer an analysis of the relation between the two uses in all messages by all users but only of their relation in those messages that came to dominate the #s21 communication space during the time intervals in question.
We coded these artifacts by hand for indicators of tactical support and symbolic association. In coding the most retweeted messages, we differentiated between tweets based on whether they contained information about the protests or their cause, calls for the coordination of the protests, evidence of artifacts documenting the proceedings of the protests or disinformation. These uses we took as evidence for tactical support. If messages in our sample contained phatic commentaries on the protests, claims of association, and solidarity with the protesters or satirical and ironic content, we took the messages as indicators of symbolic association. Table 4 documents our results. We also coded the URLs most prominently linked from the tweets in our sample. In doing so, we followed the above described coding patterns adding information on the authors of the sources of these linked content. We differentiated between links to content provided by activist organizations, traditional political organizations (e.g., political parties, government sources, websites of the project developers, unions), traditional media (e.g., newspapers, TV stations), new media (e.g., websites, news blogs, wikis), and alternative media (e.g., video or audio streams by activists, personal blogs commenting on the events, tweets, images taken by participants of the protests). We document these results in Table 5.
Retweets.
Hyperlinks.
When examining the tweets dominating the communication space during the four time intervals (i.e., those messages that were retweeted most often), we indeed find a change in the ratio between retweets coded as tactical support and symbolic association (see Table 4). We found that the most retweeted messages on August 25/26 and September 6/7 were predominantly indicative of tactical support. This ratio declines sharply on September 30/October 1 and October 16/17. On these days, we find much more messages indicative of symbolic association. When examining the usage practices more closely, we see this increase largely driven by tweets with satirical or humorous content referencing the events. This is an interesting result, since—as we showed above—on September 30 and October 1, a significant amount of new users started using the hashtag #s21, effectively expanding the corresponding issue public. This increase seems to have led to a change in the nature of the messages gaining prominence in the communication space.
When analyzing the hyperlinks most prominent in the communication space during the selected time intervals, we see much more stable usage patterns (see Table 5). In all four time intervals, tactical support is clearly the dominating usage pattern. Although its prominence somewhat falls during the run of the protests. Hyperlinks in messages containing #s21 largely led to outside sources that contain background information about the protest and related issues and sources that document the developments of the protest itself. This is also visible if we look at the creators of these linked sources. We see here that sources that dominate come from traditional media (e.g., newspapers, TV stations) and alternative media (e.g., photographs by activists, video streams by the protesters). Fluctuations in their presence in the communication space seem unrelated to the different phases of the protests.
In analyzing the messages and hyperlinks that dominated the #s21 communication space during four distinct time intervals, we showed that both uses—tactical support and symbolic association—were prevalent. Large changes in the usage patterns during different phases of the protests were only visible with retweets. Thus, the use of retweets in communication about collective action seems much more sensitive to changes in the communication space. The use of hyperlinks remained much more stable and was clearly biased toward the use of tactical support.
What Has Been Shown, What Remains to Be Seen?
In this article, we asked three questions: First, what distinguishes events that resulted in high activity on Twitter from those events that did not? Second, during times of high activity, does the behavior of Twitter users vary from their usual behavior patterns? Third, were the artifacts (retweets, links) that dominated the communication space during times of high activity indicative of tactical support of the protests or of symbolic association with it?
The first part of the analysis showed that during times of high activity on Twitter, offline events happened that were connected to the protests against Stuttgart 21. The reverse was not true. Many events that were relevant to the protests did not result in heavy activity on Twitter. Spikes in online activity seemed to correspond with events offline, but offline events did not necessarily find an echo online. Furthermore, when examining the events that resulted in heavy activity on Twitter, we found that they all shared attributes like surprise, harm, controversy, and personalization. Also, these events allowed for strong emotional reactions, visualization, and ironic commentary. Thus, in the case of Stuttgart 21, Twitter data offer no unbiased account of events, but instead document a view of the events guided by the collective attention of participants. There is significant research potential in establishing the determinants for this relationship. What are the necessary characteristics of an offline event to lead to an online echo? Is online activity only dependent on characteristics of the event referred to (comparable to the news value of news items) or are other mediating factors (e.g., media coverage, early comments by online multiplicators, etc.) important in determining the scale of online reactions to offline events? So, while it seems evident that there is a relationship between offline events and online activity, the rules and mediating factors determining this relationship remain unclear.
Second, we looked at shifts in the behavior of Twitter users during different phases of high activity on Twitter. Here, we found that there were indeed interesting differences. During these phases, the average number of messages per users rose significantly when compared to the whole time period. Also, we found that on August 25 and September 30 (days of heavy protests), the number of messages containing hyperlinks dropped. Maybe this decline is an indication that users tweeted more about the events itself and did not refer to outside sources as strongly as normal. These pattern shifts are interesting and can be interpreted after the events. Still, these pattern shifts seem not stable enough to allow for the systematic detection of specific events during the protests. The most stable pattern seems to be the sudden rise in the average number of daily messages posted by users.
With regard to the question if Twitter was predominantly used as a tool for tactical support or symbolic association by users publicly commenting on Stuttgart 21, the results show that both uses were prominent in the messages and hyperlinks dominating the #s21 communication space during the four time spans examined by us. We found that most artifacts, be they messages or hyperlinks, dominating the communication space could be described as products of tactical support. They contributed to the organization, coordination, or documentation of the protests. Especially hyperlinks linked predominantly to the content on the web that fulfilled this function. Still, messages and hyperlinks were also very visible that could be best described as products of symbolic association. Their authors associated themselves publicly with supporters or critics of the protesters and they contained phatic or satirical comments. These communication artifacts did not contribute to the organization or coordination of the protests and thus served little function to the protesters, instead they served a function for their authors. These messages worked like traditional buttons with which people could symbolize their support of a party, candidate, or cause. This usage pattern was much more visible in retweets than in hyperlinks. The strong presence of both usage patterns shows that it is probably reductive to speak of Twitter as a medium of either tactical importance for the organization of collective action or a medium for the simple expression of symbolic support or critique by social media users. Recent research studies on political uses, functions, and effects of the Internet have begun to conceptualize the Internet as a hybrid medium with various, potentially contradicting, mechanisms and effects (Chadwick, 2013). It might be beneficial to explore this approach further with regard to collective action.
Our analysis focused only on the content of messages. We were thus able to determine whether the content of a message could have counted as tactical support in the organization, coordination, or documentation of the protests. We have no information on whether these messages actually fulfilled these intended material functions. Just because someone tweeted about the appearance of the police in a certain area does not mean that the organizers of the protest where actually using this information to coordinate protesters accordingly. The real tactical support Twitter provided for protesters has to be examined differently. This offers vast research opportunities for case studies focusing on the actual social media usage patterns among the organizers of collective action, participating protesters, and supporting activists (Tufekci & Wilson, 2012).
This limitation is also true with regard to symbolic association: We could show that this use of Twitter is very visible in the messages and hyperlinks dominating the communication space at different times. What we do not know, however, is what motives or expectations the individuals posting these messages connected with them. Also, potential effects on their readers cannot be accounted for. Here, more research is needed to understand which role this symbolic association with collective action online plays in the participation repertoires of these users. Is this a form of personalized politics (Bennett, 2012) potentially leading to traditional political participation further down the line? Or is this a form of politicized active spectatorship (Kreiss et al., 2013) where the eyes of the people (Green, 2010)—or at least of those people with a Twitter account—find an outlet to signal political elites and activists that they are watching and are evaluating their actions? If true, this would make Twitter, and other social media tools, public arenas in which an interested and vocal part of the audience of collective action would comment on, link to or debate about the actions of those directly participating in collective action. Twitter, social media, or the Internet in general would thus become another public stage, on which the meaning and norms of political events and processes would be negotiated between political elites, activists, and onlookers. Political events generating high public attention online, like those in Stuttgart on September 30 and October 1, and the reactions to them in social media channels could be seen as new media events. Events that, like the big media events (Couldry, Hepp, & Krotz, 2010; Dayan & Katz, 1992) of old, focus public attention and serve a crucial function in the political life of a society by the public documentation and negotiation of the meaning of political events.
Part of gaining a deeper understanding of the role of online services for collective action lies in the analysis of data created by the interactions of users with the respective services. As shown in this article, these data hold potential for the analysis of user behavior online and also on the development of the events or topics offline that users chose to comment on. The connection between online data and offline events is far from deterministic, but there is a connection. It remains to be seen whether mechanisms and mediating factors influencing this correlation can be defined more closely in the future. A deeper understanding of the relationship between online data and offline phenomena seems crucial; otherwise researchers looking for the reflections of offline events in online data might only see their own reflection through a glass darkly.
Footnotes
Authors’ Note
We presented an early version of this article at the 2013 ECPR Joint Sessions in Mainz, Germany, in the workshop “Collective Action Online” organized by Helen Margetts and Lance Bennett.
Acknowledgments
We want to thank the organizers and participants of said workshop for valuable comments. We also want to thank Harald Schoen and two anonymous reviewers for extensive comments and Damien Schlarb for proofreading the article.
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.