Under the bridge: An in-depth examination of online trolling in the gaming context.

Trolling is a subject of apparent academic confusion; the few studies conducted thus far yielded a variety of conflicting definitions regarding what constitutes trolling behaviour and little information regarding trolling motivations. In order to shed further light on this phenomenon, the present study aimed to (1) determine which behaviours actual trolls consider as trolling, (2) explore the motivations behind trolling, and (3) examine the online community's response to trolling as perceived by the troll. After performing semi-structured interviews with 22 self-confessed trolls, we found that there is a variety of behaviours trolls consider trolling which can now be put in clear categories based on target and method. Three key motivations to troll emerged: personal enjoyment, revenge, and thrill-seeking. Trolling also appears to be a cyclical, self-perpetuating phenomenon enabled by the online community at large. Theoretical implications for future trolling research are also discussed.

In the world of online gaming, undesirable behaviour is commonplace. Players will kill teammates, verbally abuse their peers, and misdirect new community members, spreading chaos and disorder (see Riot Games, 2015). These people are called ‘trolls’ and their behaviour ‘trolling’. However, despite its prevalence in cyberspace, trolling as a subject of academic study is a confusing space, with different researchers using different criteria to describe the same phenomenon. This is likely due to the fact that it is such a new field of study: existing studies are few and far between, and nearly all of them have been atheoretical due to a lack of empirical basis upon which to build any theories (Herring et al., 2002; Shachaf and Hara, 2010; Thacker and Griffiths, 2012). Some researchers treat any deceptive action online as trolling (Buckels et al., 2014), while deception is not always required by other researchers (Fichman and Sanfilippo, 2014). Other negative behaviours with a perceived hostile intent are also sometimes grouped into trolling, while other researchers treat them as separate phenomena, such as griefing and flaming (Coyne et al., 2009; O’Sullivan and Flanagin, 2003; Thacker and Griffiths, 2012). With the concept of trolling in this fractured state, it is difficult at best to determine what to analyse when examining online behaviour in or out of game.

In addition to what trolls do, limited research has been conducted into the motivations and goals behind their actions – why and who they choose to troll and what gratifies them. Those that do address the community itself only indirectly (Herring et al., 2002; Luzón, 2011), examine only the victims’ or bystanders’ perspective (Maltby and et al, 2015; Shachaf and Hara, 2010), or lack an in-depth interview method (Buckels et al., 2014; Thacker and Griffiths, 2012). Finally, there is almost no available research pertaining to the influence the online community might have on trolling, despite multiple studies suggesting that the actions and attitudes of other netizens can have an important impact on both on- and offline behaviour (Ridout and Campbell, 2014; Whitty and Carr, 2006; Young and Jordan, 2013). The present study aims to rectify these research gaps by moving past the survey method and instead performing in-depth interviews with actual trolls to (1) determine which behaviours actual trolls consider as trolling, (2) explore the motivations behind trolling, and (3) examine the online community’s response to trolling as perceived by the troll.

Trolling behaviour

As previously mentioned, there is a lack of academic consensus on the subject of trolling behaviour. Most researchers agree that trolling can fall into two large categories of verbal and behavioural trolling, with behavioural trolling being largely relegated to the gaming sphere. Beyond this basic agreement, discrepancies abound. A variety of negative statements, such as personal insults and exclusion tactics have been considered ‘trolling’ by some researchers (Herring et al., 2002; Luzón, 2011; Shachaf and Hara, 2010; Thacker and Griffiths, 2012), but other researchers will treat some of these options as separate from trolling (e.g. Alonzo and Aiken, 2004; Douglas and McGarty, 2001; Hardaker, 2010). Hostile intent is meant to be one of the few underlying threads that connect these verbal behaviours (Buckels et al., 2014; Fichman and Sanfilippo, 2014), but in the case of false ignorance (pretending to be ignorant of game mechanics or lore), the target of the trolling is the troll – they do not attack anyone else (Thacker and Griffiths, 2012). In terms of behavioural trolling, the examples most commonly given in the few studies that address this topic are team-killing and team-blocking, in which the troll either kills off other members of their team in-game or hinders their progress in completing objectives some other way. There are, however, other kinds of behavioural trolling that are not covered in the literature: feeding (allowing oneself to be killed by the opposing team to disadvantage one’s own team and advantaging the opponents), spamming, and going ‘AFK’, or ‘away from keyboard’ (Riot Games, 2015). Each of these behaviours consists of the abuse of a game mechanic that results in a disadvantage for opponents or teammates in-game.

In the past, researchers have often classified behaviours as trolling either based on reported incidents from community members (see Shachaf and Hara, 2010) or via observations of interactions in natural online settings and labelling parts of these interactions as trolling (see Suler and Phillips, 1998). This more bottom-up approach is both valid and useful in directing future trolling research. It remains, however, incomplete without its top-down counterpart. Although studies exist in which trolls are consulted, these studies use a survey methodology (see Buckels et al., 2014). This is a reasonable approach, but it is less exploratory, thus potentially missing phenomenological insights. The present study will fill this gap by talking directly to trolls, thus giving a community insider’s perspective on which behaviours should be classified as trolling, in which environments, and under which circumstances. In this way, we can solidify our understanding of trolling on a variety of levels and complete our picture of trolling behaviour.

Trolling motivations

In addition to determining which kinds of behaviours are considered trolling, we also aim to uncover the intent behind these questionable behaviours, as this element of intent appears key in the categorization of trolling as successful or failed. Typically, the success of a troll is determined by (1) examining second-hand accounts for emotive bystander or victim responses (see Herring et al., 2002) or (2) talking to victims or bystanders directly to see if a given behaviour had negative effects (see Fichman and Sanfilippo, 2014). However, this limits the researcher to looking at the interaction from a maximum of two perspectives: victim or bystander. Trolls as a population of interest allow for the unique research opportunity of speaking to one person from all perspectives in a trolling situation – the victim, the bystander, and the perpetrator. This allows us to ask questions from several angles that are simply not possible with a different population, thus adding a new level of profundity to our understanding of trolling intention.

Integral to the question of intent, however, is that of motivation – what precedes the hostile intent and brings forth action? Although some studies have begun to explore various possible motives, these studies are few in number (Buckels et al., 2014; Shachaf and Hara, 2010; Thacker and Griffiths, 2012). More importantly, however, is their apparent homogeneity; in nearly all of these studies, trolls are presented as uniformly hostile and antisocial (Fichman and Sanfilippo, 2014; Herring et al., 2002; Shachaf and Hara, 2010). For example, Buckels and colleagues (2014) took a personality psychology approach to examining trolling antecedents. They found that trolling behaviour correlated positively with three out of the four components of the Dark Tetrad: sadism, psychopathy, and Machiavellianism. Because sadism correlated the strongest with trolling execution, Buckels et al. (2014) concluded that online trolling seems to be ‘an Internet manifestation of everyday sadism’ (p. 1). This conclusion effectively reduces trolling to a single, personality-driven cause. If trolls are more than a personality typology, however, it seems unlikely that they are all consistently antisocial. In addition to this, previous research has shown that not all trolling is antisocial; false ignorance, for example, does not necessitate harm done (Thacker and Griffiths, 2012). Given the variety of documented trolling behaviours (see Suler and Phillips, 1998), it seems much more likely that different motivations guide different behaviours in different situations. Making this distinction is critical for the field’s advancement, as the fields of personality and community dynamics differ considerably. The present study will address this question of motivational homogeneity or heterogeneity by asking trolls directly what they intend by what they do and what triggers the problematic behaviours.

Role of the community

Thus far, we have addressed two of our primary aims: determining which behaviours constitute trolling, and exploring the various motivations behind said behaviours. Although academic forays into trolling behaviour and motivation have been limited in number, even fewer studies address the role of context and community in the phenomenon. Is trolling in fact normative in the online community? Traditional psychology and sociology suggest that norms are formed and solidified over time (MacNeil and Sherif, 1976) and are largely based on three key factors: (1) inclinations (how a person thinks they should act), (2) regulatory interests (how that person thinks others should act), and (3) enforcement resources (the capacity of the group to enforce the rules; Heckathorn, 1988). These three factors interact over time to create norms in various settings, from experimental studies (Martin et al., 1974) to large-scale cultural conceptions of justice (Stolte, 1987). Thus, the question of trolling’s deviance or normativity should be determined by examining these three factors. If trolling is deviant, as postulated by Fichman and Sanfilippo (2014), then trolls should be statistically abnormal in their inclinations, and the community should be essentially uniform in their regulatory interests, with medium to high levels of enforcement resources.

However, norm formation online is a little-studied topic, and to our knowledge, it has not been examined in the online gaming context prior to the present study. Thus, to begin to test the foundations of Fichman and Sanfilippo (2014) assertion, we are required to examine the gaming community’s online presence directly. For example, by looking at the gaming community on YouTube, we can see that regulatory interests of the group appear divided. Trolling channels seem to be a popular form of entertainment on YouTube, with one such channel alone, videogamedunkey, having over 3,000,000 subscribers at the time of writing. By searching ‘trolling’ in YouTube’s video search function, over 20,000,000 hits are produced in seconds (YouTube, 2017). Yet, on the fora of popular online game League of Legends, there is an entire section dedicated to player behaviour. Its opening page is filled with questions such as ‘Riot why can people that do this get away with it’ and ‘Why is a player with a history of trolling not banned?’ (Riot Games, 2017). This latter example also suggests a lack of power on the part of the gaming community to regulate in-game behaviour, as the enforcement resources lie with the game administrators. Players are free to report, but the final say in terms of punishment belongs to the company. All of this gives reason to doubt the deviance of trolling within the community, but it remains anecdotal evidence. Here again the multiple perspectives trolls afford as online gamers will enable us to better answer this question on the community’s understanding and perception of trolling behaviour, thus allowing for the first scientific examination of these norms.

Method

Data collection

For this study, in-depth interviews were conducted with self-confessed trolls, meaning they are aware of their trolling history and their continuance in the behaviour. Participants were allowed to have a history of trolling in any type of game, be it PC, console, or otherwise, in any genre. This was decided in order to reach the widest audience and have the most variety possible in our sample due to the exploratory nature of the study, and previous reports that trolls are an extremely difficult population to reach (see Shachaf and Hara, 2010). This self-identification as a troll was our primary concern. The only other participation criterion was age: participants must be 19 years of age or older due to ethical constraints. All participants were recruited using a combination of web-based advertisements, paper flyers, network sampling, and snowball techniques.

A total of 22 semi-structured interviews were conducted via Skype with a native-English speaking researcher (the first author) between February 2016 and May 2016. Three continents were represented in this sample, with 68% of participants living in North America (10), 36% living in Europe (8), and 18% of participants living in South America (4). Of these participants, three identified as having a differing country of origin: one Indian participant and one Colombian participant were living in the Netherlands at the time of the study, and one South Korean participant was living in Canada. All other participants were residing in their home country. Participants were on average 23.6 years old (standard deviation [SD] = 2.4) and all had at least a high school diploma or its equivalent, while 32% had also completed some level of post-secondary education (one college certificate, four bachelor’s degrees, two master’s degrees). Participants had been gaming from 3.5 to 23 years, with an average of 14 years overall (SD = 5.78). Only 9% of the sample was female (2), which is a common finding in the extant literature (Buckels et al., 2014; Thacker and Griffiths, 2012), and prevented us from making gender comparisons. By the 22nd interview, saturation had been reached; irrespective of cultural background, stories were remarkably similar. Interviews lasted between 24 and 90 minutes (M = 51.06, SD = 15.42).

Data analysis

Interviews were transcribed and it was these transcriptions that were analysed. Throughout this process, common themes or keywords were taken in note. Keywords were chosen to reflect themes present in the literature, such as trolling’s apparent inherent negative nature (see Buckels et al., 2014), or for their frequency of use in question responses. At this point, questions were arranged by the three aforementioned themes (trolling behaviour, trolling motivation, role of the community) and a codebook was created. The primary researcher then coded the interview transcripts using the codebook. After this, a second coder experienced in gaming and game terminology was given the codebook and the interview transcripts and asked to code them a second time. At the initial coding, inter-rater agreement was at 63%. Recognizing that this was low for exploratory research (see Lombard et al., 2002), both coders met to discuss differences in their respective codes. After this discussion, coding was adjusted to reflect what they had agreed upon (e.g. if a game glitch was exploited, it was considered contrary play). It was the final coding consensus that was used to rank the categories in frequency of use.

Results

Trolling behaviour

In the questions regarding trolling behaviours, our sample described their own personal definition of trolling, as well as the various kinds of trolling they see exhibited in-game. They also expounded upon their own trolling behaviours.

Trolling definitions

Consistent with the extant literature, trolls themselves also give a variety of different trolling definitions. However, these definitions can be generally split into three categories based on the elements of trolling they stressed: (1) attack, (2) sensation-seeking, and (3) interaction-seeking. These elements are not mutually exclusive. There can be elements of sensation-seeking within the interaction-seeking group or vice-versa, for example. However, each definition did have a primary stress, and it was by this emphasis that they were categorized.

According to the participants who stressed the element of attack in their definitions, trolling is a direct attack on the other players’ enjoyment of the game or gameplay. This type of definition was the most common seen in the sample, and tends to be the view of trolling commonly presented in the extant literature, gaming-focused or otherwise (Buckels et al., 2014; Herring et al., 2002; Shachaf and Hara, 2010). Participants holding this viewpoint called trolling ‘ruining gameplay for other people’ (P15, 24, female) and described it as ‘intentional loss, or people playing with the intent to piss other players off’ (P12, 24, male). In this definition category, trolling is presented as purely anti-social and antagonistic.

Participants emphasizing sensation-seeking in their definitions painted a more asocial picture of trolling: it is neither inherently good nor bad, but simply a behaviour which leads to enjoyable consequences for the troll. Participants described it as ‘the creation of drama’ (P22, 20, male) and as a way to get attention. The victim’s reaction figured heavily as a source of thrill or enjoyment in these definitions, as it is this drama that satisfies the troll. Typically, the more outrageous the reaction, the better (P11, 23, male; P20, 23, male). This definition category is typified by its hedonistic flair and its emphasis on the other players as a source of pleasure, be it sadistic or otherwise.

Finally, participants who emphasized interaction-seeking defined trolling as an unorthodox method of communication designed to make players get involved in both the conversation and the game. These definitions can present trolling as either prosocial or asocial, but never antisocial. The following passage is typical of a definition in this category:

For me, [trolling is] mostly when somebody is doing well but then doesn’t … hmm … want to play the game just for the game but more for like … hm … interaction purposes. Trolls generally interact a lot. So we like participation from the other side too. It gets really boring when you are the only one trolling. It has to be a few trolls to be fun. (P4, 27, male)

Like the sensation-seeking category, this definition type also emphasizes other players and their reactions to trolling. However, the desired reaction is radically different between the two. Based on this definition, a positive response is much more highly valued than an outrageous one. The desire is not sensation, but friendship. In fact, some participants even argue that trolling between friends actually makes the game more enjoyable (P10, 23, male: ‘Trolling with fun and friends, I think they laugh about it and it’s ok’). Whether it occurs between friends or friends-to-be, this definition of trolling places particular importance on amusing, albeit unorthodox, interaction between players.

In-game behaviour

Based on the participants’ reports, trolling behaviours can be largely divided into two groups: verbal trolling and behavioural trolling. Table 1 presents a summary of both types, their sub-types, and an explanation of these. Within the verbal category of trolling, trash-talking is the most commonly mentioned, and is typically comparable to what you might see in a traditional sports arena:

Table 1.

Trolling types and sub-divisions.

Trolling type	Explanation
Verbal Trolling	Using a chat function in-game to troll another player.
Trash-talking	Putting down or making fun of other players
Flaming	Presenting emotionally fuelled or contrary statements with an instrumental purpose.
Misdirection	Spread false information among targeted or general players
Spamming	Repeating game-unrelated chat either textually or audibly in-game.
Inappropriate Roleplaying	Pretending you are a different person (non-game-related) to obtain some kind of specific reaction.
Behavioural Trolling	Using existing game mechanics to troll another player.
Inhibiting team	Actively hampering your teammates’ in their goals.
Contrary play	Playing the game outside of what is intended by most players.
Aiding the enemy	Disregarding strategic play to make it easier for the opposing team to win.

Oh, so um, while the game is loading, at first you can write messages to each other. So they’re like ‘Yeah, we’re gonna win! We’re gonna defeat you!’ Like sometimes just talk with their … like for example, ‘I’ve won this many times!’ For example with my cousin or my brother, we’re like ‘Yeah you’re the worst! My army is so good. You’ll always lose!’ and that kind of stuff. (P16, 21, male)

However, more negative trash-talking is also present in the sample. This consists of direct and often unjustified criticism of another player on a personal or gameplay level, or in some more extreme cases, even insulting or degrading the player’s family members (P3, 24, male ‘Your mom is a B, like a B-word’).

In terms of behavioural trolling, individual behaviours were grouped into the following three categories: (1) inhibiting your team, (2) contrary play, and (3) aiding the enemy team (see Table 1 for descriptions). Although all three were reported regularly by participants, inhibiting your team was the most popular of the sample-described trolling methods. Essentially, team inhibition consists of playing the game with the goal of ruining the gameplay of your teammates. It can be as complicated as throwing your character in front of the attacks of friends in order to make it look like they are killing their own team members and incur a penalty, or as simple as blocking a path required for your teammates to reach their goals:

Well um … for COD for instance, you could uh … corner-trap your teammates. They’ll be just camping in the corner and then if you stand in front of them, they can’t escape … and they get really mad. (P9, 22, male)

Examples of contrary play would be exploiting a game glitch to gain an advantage over other players, or pushing another player’s avatar into a body of water, while aiding the enemy team could be broadcasting team positions or ‘feeding’. Interestingly, all of these behaviours were also instrumental. If there was no goal or ulterior motive attached, for example, in a flaming or contrary play situation, then it was not classified as trolling. When referring to trolling, participants consistently included a goal, such as obtaining a reaction from others or distracting the enemy players. Instrumentality figured strongly as a key indicator of trolling classification in our sample.

Trolling motivations

In the questions regarding trolling motivations, our sample described why they exhibit the aforementioned behaviours by clarifying their ‘triggers’ and associated goals – essentially, why they troll. Although goals and triggers may seem to overlap – both have clear connections to motivation – it is important to note that they were treated separately during the interviews. Triggers were referred to as a catalyst to begin trolling, while goals refer to the ultimate achievement desired by the troll. Despite some similarities in response, we consider these to be conceptually distinct.

Trolling triggers

Despite some claims that it happened ‘randomly’, all members of the sample were able to identify one or more triggers that typically preceded their trolling. These events could be broadly categorized into the following three types: (1) social triggers, (2) internal triggers, and (3) circumstantial triggers.

Social

In our most popular and broad category, there was one clear forerunner for type of trigger: being trolled. Of all the social triggers listed, being trolled first was the single most popular reason to begin trolling:

It sounds super silly, but more often than not, if I’m trolling it’s because I’ve been trolled, and it’s kinda in the hopes that they stop, and they realize that it’s super annoying, and that they wanna win the game, and that it’s way easier to win the game if there are 5 people participating instead of 3 people participating. So you know, I hope that it stops, or they at least leave me alone so that I can play my best game. (P15, 24, female)

Trolling appears to breed trolling, with the behaviour seemingly becoming a social contagion among gamers. Other social triggers typically involved noticing weakness in other players, either poor gameplay or general gullibility. However, these border on internal triggers, as the player has to be seeking signs of weakness or vulnerability in order to spot them quickly and act upon them.

Internal

Eleven members of the sample mentioned internal triggers as the catalyst to their trolling. In fact, with the exception of the two female trolls in the sample, all participants who mentioned more than one trigger mentioned at least one internal trigger. Of these internal triggers, two emerged as the most common overall: being ‘on tilt’, referring to a negative emotional state in which gameplay typically suffers, and boredom:

Why did you start to troll yourself?

Personal enjoyment.

Something to do?

Yeah, you get bored of the game sometimes, and sometimes it makes the game more exciting. (P22, male)

Those who troll due to boredom often seem to treat the trolling as a sort of meta-game. They see themselves as being ‘beyond’ the game, having seen all there is to see, and thus try to ‘win’ at trolling instead. Remarkably, the other primary internal trigger, being on tilt, appears to be primarily caused by consecutive loss:

Yeah, so for you [trolling is], it’s a reaction based on anger, it’s lashing out in anger.

Yeah.

And the two things that happen beforehand, it’s losing, it’s triggered by losing, or …

Yeah, but losing on a streak, like, like lots of times in a row and that just … pissed me off. (20, male)

Thus, both veteran players and newbies can fall victim to internal triggers: if you are experienced in the game, you risk boredom; and if you are brand new, you risk consecutive loss against more experienced players. It is worth noting, however, that trolling itself is also listed as a reason someone may be on tilt, or one of the factors playing into a loss-streak. This strongly suggests a vicious-circle-like element to trolling.

Circumstantial

Least-popular among the triggers are the circumstantial triggers – only five members of the sample even mentioned them. There are two different circumstances that the sample listed as potential trolling-triggers: the pre-game, and a winning start to a match. In most online games, there is a ‘pre-game lobby’ in which players select their characters or avatars. Players can also chat in this pre-game lobby. The pre-game seems to set the tone for the rest of the game, and it is where trolling initiates in games with circumstantially triggered trolls. Despite the few trolls who mentioned them, this seems to suggest that the earliest part of the game is a partial determinant of trolling. For at least some trolls, simply having the opportunity to communicate pre-game is enough to initiate trolling.

Trolling goals

A variety of individual motivations emerged from the interviews, but these can be grouped into three broader categories in order of sample frequency: (1) personal enjoyment, (2) revenge, and (3) thrill-seeking.

Personal enjoyment

Based on the sample, trolls motivated by personal enjoyment can either derive their joy from the sheer pleasure of trolling itself, or can use trolling strategically to disable or weaken their opponents in-game and enjoy winning the game:

It’s kinda like a game, almost? It’s the game within the actual game. Like, if you, if all five of you go in and you can manage to win with shields, it’s like, even more fun than just normally winning the round by playing the game. (P14, 19, male)

Trolling thus becomes a meta-game, an added challenge that heightens the gaming experience. This goal is closely linked with the boredom trolling trigger, as most participants in this category mention being disenchanted with the actual game and wanting something more from the gameplay (P14, 19, male: ‘Eventually you just get bored and you’re like, let’s like, let’s make it fun, let’s make it funny’.)

Revenge

Revenge-motivated trolls, who are uniformly people who have first been trolled themselves, seek either the misery and/or failure of the initial troll, or the reformation of said troll’s behaviour by showing them how their behaviour affects others negatively:

The only reason why I would troll is if someone does it to me. It’s not that I would go do something like that because I really want a lane and I’m just going to insta-lock – that’s not me. I troll someone if he trolls me, so it’s just a response. (P8, 19, female)

The above participant falls into a previously unknown trolling archetype – the vigilante. These trolls prey exclusively on other trolls to ‘give them a taste of their own medicine,’ so to speak, in the hopes of either reforming them or scaring them away from the online community. Other revenge-motivated trolls are purely ‘reactionary’ (P10, 23, male: ‘It’s a response to something, and most the times, if it’s something stupid, I react about it normal, but if they act stupid again, I might flame them or something’.), who troll instinctually when trolled, as opposed to with a specific goal in mind. Both, however, are considered revenge-motivated due to the requirement of being trolled first before taking action themselves.

Thrill-seeking

Thrill-seeking trolls seem to most resemble the trolling depictions presented in the extant literature (Buckels et al., 2014; Fichman and Sanfilippo, 2014; Thacker and Griffiths, 2012). These trolls in our sample had a noted disregard for the potential impact of their behaviour, single-mindedly seeking the most outrageous reaction possible by any means necessary, verbal or behavioural:

Uh, you annoy other people. The same thing – why would you play against other players? It’s, well you can kill their avatar. If you do it in a way that works, that’s kind of nice too, but if you can do it in a way that they get pissed off even more, that’s even more fun. (P19, 27, male)

Like their enjoyment-seeking counterparts, these trolls also frequently cite fun as a key goal in their enterprises. However, there is an additional goal in this category that does not appear in the personal enjoyment sphere: satisfaction of curiosity. Several trolls in our sample treat trolling as a ‘social experiment’ (P3, 24, male) designed to ‘gauge’ (P1, 24, male) other players. They satisfy their curiosity by, in their mind, empirically testing their preconceptions of other players. In either case, thrill-seeking trolls appear to be the most aggressive of the troll types.

Role of the community

In the final set of questions regarding the role of the community, our sample discussed the kinds of responses to trolling they see in the community, both when they are the perpetrator and when they are a simple bystander. They also examined their own thoughts regarding the online community’s opinions of trolling.

Trolling responses

When questioned regarding typical responses given by victims and bystanders to trolling, once more our sample gave a variety of answers. These were condensed into five categories: rage – verbally expressing negative feelings towards the troll; ignore – taking no action related to the trolling; troll – either joining in the troll’s victimizing or trolling the troll back; prevention – muting or reporting the troll; and participation – joining in the conversation surrounding the trolling without trolling or raging oneself (bystanders only). These categories were then ranked for victims and bystanders according to the frequency with which responses falling into these categories were mentioned by participants. For bystanders, participation was the most popular reaction, followed by ignoring the troll, trolling themselves, raging, and finally prevention. Victims show a similar but differing pattern, with rage as the most popular response, followed by ignoring the troll, trolling back, and finally prevention.

Thus, it seems as though victims most often become angry and respond in kind when trolled, further entrenching the cyclical nature of trolling. According to our sample, rage and trolling back are two of the top reactions to trolling among victims. As discussed previously, thrill-seeking trolls are motivated by strong reactions, rage (and flaming) included; thus, by raging, victims motivate thrill-seeking trolls to continue trolling, while trolling back further entrenches trolling behaviour in the community. Interestingly, prevention behaviours such as reporting the troll or muting them are the least popular response, despite the fact that four members of our sample list ‘getting reported’ as a negative consequence and deterrent of trolling. This seems to indicate that victims of trolling also tend to be enablers of trolling, reacting in such a way that trolling is encouraged, and not taking advantage of built-in systems like the mute button and reporting to prevent the behaviour from happening again in the future. That said, ignoring is still a relatively popular option, and being ignored is also listed by nine of our trolls as a negative response to trolling, meaning that it discourages the troll. Still, the vast majority of victims seem to engage in responses that encourage further trolling in the community.

Based on our sample, bystanders are just as likely to engage in trolling prevention as victims, and also have their share of enabling behaviours. Participation in the trolling conversation, which typically entails either defending the victim or trying to ‘talk down’ the troll, emerged as the most popular bystander response to trolling. This, although perhaps not as direct as with the victim, provides trolls with further reactions to their behaviour, motivating the thrill-seekers and potentially even the personal enjoyment-motivated trolls. Ignoring was also a popular option, suggesting that bystanders are not constantly engaging in enabling behaviour. Typical enablers, such as trolling back or raging, were less popular among bystanders than among victims. That said, preventative behaviours remained the least popular response to trolling, despite its aforementioned capacity to reduce trolling in online communities. Thus, it appears that although bystanders do not seem to react as strongly as victims to trolling situations, they are still prone to troll-enabling by neglecting preventative measures and providing additional reactions to the trolling behaviour.

Trolling: normative? Before addressing trolling’s normativity, it is crucial to understand the degree of nuance presented in our trolls’ responses

There were few if any cut and dry answers regarding the normativity of trolling. When asked what they think the gaming community feels about trolling, 19 members of the sample said that it was negatively perceived. Participants used terms like ‘necessary evil’ and ‘guilty pleasure’ when describing trolling from the community’s perspective, indicating its negative nature. Other participants described it as ‘toxic’ and ‘annoying’, cementing it as a darker side of gaming. A few participants also mentioned that trolling had evolved into its current negative state, and that it was the new generation of trolls that was twisting its original purpose (P20, 23, male; P21, 24, male). According to one participant in particular, trolling started as a way to play mind games and trick other players in games, but now the term has come to include all negative behaviours online (P20, 23, male). Trolling is thus exposed as a dynamic, evolving phenomenon. Participant 12 may have summed all of these thoughts up best when he described trolling as ‘a problem that will never go away’ (24, male).

This said, ten members of the sample also mentioned at some point in the interview that trolling was a part of gaming, an inextricable piece of the activity. In addition to this, some sample members cited YouTube channels as proof of trolling’s normalcy:

Um … … …. well there’s a lot of Youtube channels dedicated toward trolling people, so I’d say a lot of people just think it’s funny. They enjoy it when it’s not them. But I would say, if they weren’t being trolled themselves, they could enjoy someone else being trolled. Then, everyone seems to like it. (P9, 22, male)

This suggests that trolling is not only normative in the community, but even celebrated by some of its members, trolls and everyday gamers alike. Other participants made the distinction that, among friends, trolling is completely acceptable, and that it only treads into negative territory when it takes place among strangers (P10, 23, male: ‘So … yeah, trolling for fun … it’s ok, but trolling to um, influence other people that you don’t know is not ok’.). As previously mentioned, there is a high degree of nuance in our trolls’ responses, and context seems to be an important factor.

All of these varying responses seem to suggest that trolling is neither normative, nor deviant, but rather somewhere in between. It is clear that trolling is considered a negative phenomenon, but it is also an expected phenomenon, a ‘rite of passage’ (P4, 27, male) within the online gaming community. Were trolling normative, there would be repercussions for not engaging in trolling behaviour. In other words, the regulatory interests of the group would be activated by non-trolls. However, were trolling deviant, it would not have such a high prevalence, nor would it be expected to the current degree it is in the community. Thus, we propose trolling as an a-normative phenomenon – neither deviant nor prescribed, but an active part of the community and largely tolerated. Whether considered positive or negative, however, nearly the entire sample agreed that it was normal behaviour within the community.

Discussion

Conclusions and implications

At this study’s outset, we aimed to examine what constituted trolling behaviour, what kinds of motivations and goals were associated with trolling, and how the community impacts trolling. Prior to this study, there were multiple, occasionally contradictory, definitions of what constitutes trolling floating in academia. We were able to confirm that this trend extends into the trolls’ world as well – definitions provided by our sample were varied, though they generally fell into a few key thematic categories: attack, sensation-seeking, and interaction-seeking. This suggests that trolling is not a uniform phenomenon, but rather an umbrella term for certain types of instrumental online interactions. Attack trolls want their victims’ misery, interaction-seeking trolls want friendship or conversation, and thrill-seeking trolls want sensation for themselves. This finding of instrumentality in trolling contradicts what is suggested by Buckels et al. (2014) trolling definition: trolling is actually characterized by its instrumentality, and not a wanton nature. In fact, negative or controversial online behaviours that are not typically considered trolling are often categorized as trolling once they develop an instrumental purpose. Take the example of flaming, which O’Sullivan and Flanagin (2003) define as ‘hostile and aggressive interactions via text-based computer mediated communication’ (p. 69). We found that flaming only entered the realm of trolling if it was used specifically to obtain an outrageous reaction from victims in the pursuit of sensation or thrills. This is a crucial insight into trolls’ perception of trolling, and this instrumentality should be taken into careful consideration in future studies.

There remained also the question of what constituted trolling. Once more, there were several examples that had been documented (see Herring et al., 2002 and Shachaf and Hara, 2010), but they had not been categorized in any systematic fashion and were typically deemed ‘trolling’ by either researchers or laymen. There was no previous research classifying trolling behaviour according to the actual perpetrators. Through our interviews, we were able to develop a clear classification system of trolling behaviour from the trolls’ perspective: verbal (trash-talking, flaming, misdirection, spamming, and inappropriate roleplaying) and behavioural (inhibiting your team, aiding the enemy team, and contrary play). We also found that these behaviours appear to be dispersed unevenly across generations of gamers, creating a generational gap between trolls. Veteran gamers take on a trickster archetype when they troll, and tend towards misdirection and subterfuge, while new and younger gamers go for a more abrasive approach, engaging in behaviours such as trash-talking and killing teammates. This has caused veteran gamers to renege on the term ‘troll’, as they perceive it to have a different meaning today. While it was once a badge of honour symbolizing their mastery of intellect and gameplay, it has turned into a sign of shame reviled by most gamers. This finding suggests a hierarchical aspect of the online gaming community, and particularly for trolls, only hinted at previously (see Thacker and Griffiths, 2012), and never before attributed to gaming experience and age.

In addition to the question of trolling behaviour, we also sought to uncover the motivations behind said behaviour. The present study allowed for an in-depth examination of trolling motivation, revealing a variety of possible motivations and goals, ranging from prosocial to antisocial, and including personal enjoyment, revenge, and thrill-seeking. In particular, the importance of personal enjoyment and boredom as presented by the extant literature (Buckels et al., 2014; Thacker and Griffiths, 2012) was confirmed, as these also emerged as some of our sample’s most popular motivations and catalysts to trolling. However, we also uncovered otherwise unknown motivations, such as interaction-seeking and looking for friendship via trolling. Another previously undocumented phenomenon also emerged from the interviews: the ‘vigilante troll’ – trolls who target other trolls with the goal of reforming their behaviour or exacting revenge. Interestingly, the two women trolls that we interviewed fell into this trolling motivation type. Although not generalizable to all women trolls due to the small sample size, it is still an important phenomenon to note, specifically in the light of GamerGate, an online event in which gamers of all stripes banded together against what they perceived as invasive feminist rhetoric (Chess and Shaw, 2015; Massanari, 2015; Mortensen, 2016; Vermeulen et al., 2016). Much of the literature on this topic is in the feminist tradition (Chess and Shaw, 2015; Massanari, 2015; Parkin, 2014; Todd, 2015) and espouses that despite increased presence of women in the gaming community and gaming industry, what Chess and Shaw (2015) call a ‘hegemonic masculinity’ (p. 218) pervades online gaming. However, another line of research focuses on the emotions involved in GamerGate – specifically a sense of victimization and rage on the part of gamers (Mortensen, 2016). It is these feelings that Mortensen (2016) asserts led to the GamerGate scandal, with gamers organizing themselves as a community to stand up to what they saw as their persecutors. This idea of vigilante justice is remarkably similar to the motivations our female trolls displayed when trolling other, statistically male, trolls. Although this motivation is not relegated exclusively to female trolls, further research is required to determine gender’s true role in trolling.

These motivational findings also link in to the idea of trolling being fundamentally instrumental. Some have alleged (see Buckels et al., 2014) that trolling is an aimless pursuit. However, trolls consistently list multiple associated goals and motivations behind the behaviour when asked (i.e. personal enjoyment, revenge, reformation, fun, etc.). We found that friendships can be formed and cemented via trolling, and that vigilante trolls seek to reform or remove other trolls from their game-space. Trolls have goals, and these vary dramatically from troll to troll. We also identified several trolling catalyst-events, called ‘triggers’ here, for modelling purposes. Many of the trolling triggers identified in this study are social, and should be detectable via analysis of chat logs and the like. However, some of the most common triggers and motivations were internal, with boredom being of particular importance. Trolling researchers must be careful to include variables such as mood and state of mind in future studies to ensure that they are taken with proper consideration in any modelling or empirical testing.

In terms of community and its impact on trolling, a major implication from the present study is the fact that there is a trolling community at all. Were there no form of community, it would have been impossible to recruit using a network sampling technique. As it stands, trolls are aware of one another and are often connected to other trolls, forming at the very least a loose community. In fact, many members of our sample reported trolling more often in groups than alone. Historically, trolls have been treated largely as individuals (see Buckels et al., 2014). However, this finding opens the way for group-level analyses and the exploration of these trolling communities. In addition to uncovering a trolling community, however, we also explored the normativity of trolling in the overall gaming community. Yet again, our findings contradict those presented by the extant literature (Fichman and Sanfilippo, 2014); trolling is, in fact, a normal, expected event, sometimes even described as a rite of passage. No one escapes it, and it thus becomes a shared, common experience between gamers, cementing the community. It is also anormative – tolerated, but not encouraged. In order to be considered deviant, trolling must enter the realm of cyberbullying or cybercrime, typically by persistently targeting a single person or entity repeatedly or by breaking a written law. It is important to note, however, that we did not find that trolling is considered positive. It may be common and expected, but our sample was unanimous in saying that it is seldom an enjoyable experience to be the victim of a troll. Thus, it seems to trolls and gamers alike that trolling is normal, but negative.

However, what is perhaps the most important and novel finding from this study is that trolling appears to be a cyclical, self-perpetuating phenomenon – the community’s role in trolling appears to be enabling and perpetuating behaviour. Every single member of our sample reported having at some point been a victim of trolling themselves. Based on their responses, the cycle is strongly reinforced by the community and its response to trolling. We found that victims will more often than not respond to trolls by trolling them back. Our own trolls frequently reported carrying negativity forward into future games, suggesting that these initial victim responses can easily translate into future trolling experiences in which the initial victim becomes the perpetrator. In addition to this, bystanders tend to fuel the flames by jumping into the conversation between troll and victim, giving the troll an even larger reaction. By contrast, both victims and bystanders are relatively unlikely to engage in preventative action, thus supporting trolling by both omission and commission, however indirectly. These findings together form the greater finding that bystanders and victims are, however unwittingly, complicit in the trolling process. This could have potential connections to cyberbullying, as this too is reportedly a cyclical phenomenon (Vandebosch and Van Cleemput, 2009). Other studies have suggested that internet identity and cyberbullying identity is fluid, with bystanders and victims and perpetrators all interchanging roles over time (Park et al., 2014). In either case, both phenomena appear to be self-perpetuating and negatively perceived.

Limitations and future directions

As with all studies, this one is not without its limitations. One such limitation is in terms of the sample. Due to ethical constraints, we were only able to interview participants aged 19 or over. However, many of our participants suggested that trolling is even more common in younger audiences, and that there is a generational gap between veteran and young trolls. Thus, it would have been ideal to have more trolls and a wider age variance in our sample to better determine whether or not this allegation is accurate, or merely a perceptual bias on the part of veteran gamers. Future studies could contrast age groups, or perform a similar study targeting a different age range to explore this apparent trend in-depth; with a larger sample available, they could contrast cultures as well. We also interviewed trolls specifically. This in of itself is not problematic, and was in fact the goal of the study. However, we discussed not only their experiences trolling, but also those of the bystanders to and victims of their trolling. Once more, this provides a different perspective on trolling, but should be taken carefully, as perpetrators within community settings have been previously shown to misperceive said community’s norms and values (Young and Weerman, 2013). Thus, although the trolls’ perception is a novel finding, further studies are required to be sure that this evaluation of community norms is not a perceptual bias on the trolls’ part.

In addition, there remain some outstanding questions regarding the homogeneity or heterogeneity of the trolling community. The present study examined gamers, but there are other places that trolls can practice their craft – social media websites, forums, comment sections and the like. The differences and similarities between trolls and trolling behaviours on these different platforms has yet to be compared and contrasted. Answering this question could even potentially explain the differing results found by trolling researchers thus far (Buckels et al., 2014; Fichman and Sanfilippo, 2014; Shachaf and Hara, 2010; Thacker and Griffiths, 2012) if it is due to platform differences. Culture may also come into play here. The present study examined many different cultural groups, touching on three continents. Due to the difficulty in finding trolls willing to be interviewed, this happened naturally in the recruitment process. However, these groups were too small to make a truly generalizable cultural comparison. By examining how different cultures troll and how people troll on different platforms, we can enrich our understanding of the phenomenon as a whole. The present study is a foundation of things to come. There is still much to be done before trolling as a subject of academic study can reach its full potential.