What Do We Really Know About First-Person-Shooter Games? An Event-Related, High-Resolution Content Analysis



Multiple studies have been addressing effects of playing violent video games. However, most such studies neglect users' individual experiences. In fact, each player's gameplaying choices creates his or her own specific game content. Within this study we analyzed the individually generated content of a typical first-person-shooter game with high temporal resolution and associated them with physiological response measures (heart rate, skin conductance) collected during the game play. Thirteen experienced game players played the multiplayer first-person-shooter game “Tactical Ops: Assault on Terror” for 50 minutes on average. Playing phases and events in between were analyzed on both an intraplayer and interplayer level. Results indicate varying gaming experiences in the users and distinct arousal levels over time and for different game events.

Video and computer games are the fastest growing segment of the entertainment industry with shooter games accounting for 11% of the video and computer game sales (National Institute on Media and the Family, 2008; Entertainment Software Association, 2007). First-Person-Shooter games (FPSGs) are designed to closely engage players in violent virtual activities (Jansz & Tanis, 2007). They feature forms of violence presented in contexts that have elicited criticism from youth advocates and researchers around the world who fear antisocial effects as a result of these violent portrayals (e.g., Children Now). Although a number of studies suggest that violent video games are related to aggressive cognitions, affect, and behavior (cf. Weber, Ritterfeld, & Kostygina, 2006), there is still a controversial discussion about the validity of these findings (cf. Ferguson, 2007). Compared to the body of research in the area of television violence exposure (e.g. Gunter, Harrison & Wykes, 2003) and the steadily growing literature on the (negative) effects of violent video games (Anderson & Bushman, 2001; Sherry, 2001; Anderson, 2004), there is a paucity of research investigating the content of violent video games and FPSGs, and there is almost no research that examines individual differences across players.

The way games are played and the resulting content a user is exposed to varies with player skill level (Smith, 2006) and personality traits (Lachlan & Maloney, 2006). We proffer that content is also likely to vary as a function of different tactical approaches that users apply to a game. The existence of discernible tactics should produce predictable patterns of violent content that result from a player's use of these tactics in games. As such, our knowledge of common player tactics may aid attempts to delineate those players who are more or less likely to be exposed to violent content and the theoretically relevant contexts in which the violence may occur.

Few studies have systematically analyzed the content of violent video games in general, and FPSGs in particular. Most existing studies apply a methodology developed for non-interactive media, concentrating on the analysis of discrete content attributes found in short samples of game play derived from a large array of violent games. By contrast, the present study observes dynamic patterns of play in a typical FPSG by different users over an extended period of time and examines the amount of violence that is generated within different gaming patterns. The aims of this approach are to identify a set of common player tactics, and ascertain the varying amounts and types of violence found within these different player tactics. Continuous psycho-physiological response measures are applied to investigate immediate effects of different game events in more detail.

Content analyses of violent video games

Most content analyses of video games provide global indicators of typical content. They tell us, for example, that 79% of all video games feature physical aggression (Dietz, 1998), or that 64% of the “E”-rated console games feature violence (Thompson & Haninger, 2001). With regard to FPSGs in particular, Smith, Lachlan, and Tamborini (2003) found substantial amounts of realistic violent portrayals in contemporary video games of that time, concluding the standard game narrative featured “a human perpetrator engaging in repeated acts of justified violence involving weapons that result in some blood shed to the victim” (2003, p. 71).

Such reports have been criticized for using methods developed for linear media like television and film (Lachlan & Maloney, 2006). Disregard for player differences is found in scientific studies of content that provide quantitative indicators of game features (for overview see Smith, 2006). The most notable challenges to the validity of content analyses conducted with methods from traditional media stem from problems associated with the sampling of time frames and variance in player skills (Smith, 2006).

The content created in video games varies considerably over time and across different players (Southwell & Doyle, 2004) jeopardizing the capacity of content selected from the play of individual users to produce a representative sample. The sample of content generated by individual users can be expected to differ systematically based on characteristics of the gamers and the length of time they spent playing–two determinants often overlooked. To avoid these influences, some content analyses rely on stable content like game covers (Provenzo, 1991; Brand, Knight & Majewski, 2003), reviews in game magazines (Miller & Summers, 2007; Dill & Thill, 2007), or introductory films of games (Brand, Knight & Majewski, 2003; Jansz & Martis, 2007). These attempts cannot contribute to an understanding of typical playing patterns nor do they acknowledge differences in game content generated in each player.

Some authors relied on time frames ranging from as little as 10 minutes (Smith, Lachlan & Tamborini, 2003; Brand, Knight & Majewski, 2003) to frames as long as an hour (Haninger & Thompson, 2004), 90 minutes (Thompson & Haninger, 2001) or 2 hours (Shibuya & Sakamoto, 2004). This strategy seems still limited given that the effects of violent video games vary with the time played (Sherry, 2001) and that contemporary video games take about 20 hours of play time for an experienced gamer to complete (Jansz & Martis, 2007). The problem of short time frames seems most apparent with offline single-player games that incorporate a narrative to be followed. Online FPSGs like Counter-Strike without narratives are usually played in a defined number of short consecutive rounds. Thus, sampling strategies for content analyses of FPSGs should correspond to the rules of the games in question and should include relevant time frames.

In addition to their short time frames, most previous content analyses have relied on single players (Smith, 2006) to produce content samples. In cases where games were played by multiple players, the overlapping material was used to check for coding consistency (Shibuya & Sakamoto, 2004; Brand, Knight & Majewski, 2003; Smith, Lachlan & Tamborini, 2003; Children Now, 2001) With rare exception (see Lachlan & Maloney, 2006) differences in play between gamers were ignored. Moreover, none of these studies have provided a longitudinal perspective. This study tries to fill this gap by means of event-related, intra- and interplayer content analyses of FPSG game playing behavior with high temporal resolution. Therefore, our first two research questions are as follows:

RQ1: What type and extent of violent content do experienced players of a typical first-person-shooter game generate over a typical period of game play?

RQ2: What are typical playing patterns applied to first-person-shooter games by experienced game players?

Video games and physiological arousal

To better understand the relationship between content and its impact, continuous response measures tracking immediate effects on players during video game play are required. Associating content and psychological response over time with high temporal resolution seems to be a suitable pathway for short-term effect studies in interactive media. As a demonstration, this paper focuses on the well-developed psychological construct of arousal. Physiological measures like heart rate (HR) and skin conductance response (SCR) are considered valid and reliable indicators for arousal (Ravaja, 2004).

Several studies applied physiological measures to assess immediate responses to media messages (e.g. Lang, 1995; Lang, Zhou, Schwartz, Bolls & Potter, 2000), and to video game play in particular. Segal and Dietz (1991) found increased heart rate, blood pressure, and oxygen consumption among 32 players, aged 16-25, playing “MS Pac Man”. Murphy, Alpert, and Walker (1991) found increased blood pressure and heart rate during the use of violent video games compared to nonviolent video games. Recent studies confirm early findings of increased arousal as a result of playing violent games compared to non-violent games and indicate that the outcome has become accentuated in the latest-generation games (Ivory & Kalyanaraman, 2007; Barlett, Harris & Baldassaro, 2007; Tafalla, 2007; Carnagey, Anderson & Bushman, 2007; Arriaga, Esteves, Carneiro & Monteiro, 2006; Bushman & Huesman, 2006; Anderson et al., 2004; Baldaro et al., 2004; Ballard & Wiest, 1996; Murphy, Stoney, Alpert & Walker, 1995).

Yet, while this research is informative about the influence of game play on arousal, these studies exclusively focused on game playing sessions as a whole. Given the variety of events during a gameplay session which make or can make an FPSG differentially exciting, event-related analyses of arousal variations during game play should provide deeper insights into the effects of individual event types. Ravaja et al. (2006) examined physiological responses to different events in the video game Super Monkey Ball 2 (Sega Corporation). They defined eight events which were automatically coded from recorded video-game play, examined physiological responses, and found reliable valence- and arousal-related physiological responses for each type of event.

Our study attempts to analyze arousal responses to events and event patterns in a FPSG defined by an event-related content analysis with high temporal resolution. The continuous measurement of both heart rate and skin conductance response provides an appropriate and reliable indicator of arousal. It may be used to better understand what players experience when playing FPSGs over time. Consequently, we pose our third research question:

RQ3: How do arousal levels of experienced FPSG players respond to individually generated game events over an extended period of game play?


We conducted an inductive, event-related content analysis of recorded and digitized FPSG playing sessions with a temporal resolution of one second. Additionally, we measured heart rates and skin conductance as indicators for arousal during play.

FPSG players

Thirteen experienced male German FPSG players aged 18 to 26 were recruited in local video game shops. On average they had started playing video games by the age of 12 and played about 15 hours per week (SD = 9.0). We decided on male players since FPSGs are predominantly played by young male players (Jansz & Tanis, 2007). None of the participants were familiar with the game used for the study. Participants practiced until they felt comfortable with the game mechanisms before game sessions were recorded. After practice and a break for two hours, the volunteers played the game for 50 minutes (SD = 6.8 min) on average which corresponds to typical playing times of FPSGs (Electronic Sports League, 2006).

Stimulus material

To analyze a typical FPSG, Tactical Ops: Assault on Terror (Infogrames Europe; U.S. edition; http://www.tactical-ops.to/) was used as the experimental game. In this game two groups fight each other as terrorists or SWAT (Special Weapons And Tactics) team. The game provides different location settings (maps) such as inside an industrial building or at a beach area. The plot and missions are typical for the FPSG genre. For example, terrorists have to hold hostages or place bombs whereas the SWAT team has to rescue the hostages or stop the bomb countdown. The players have the visual impression of holding a weapon in their own hands. The game is played in several rounds per map. A round is over if one of the teams has fulfilled its mission, killed all adversaries, or if a predetermined time limit is reached. If players are (virtually) killed before the round is over, they can observe the further game play from a viewer perspective (so-called ghost mode), but have no opportunity to interact with other players until the next round.

The game mode described above is very typical for FPSGs. It was introduced in 1997 as a user-generated modification called Clan Arena for the 1996 FPSG Quake (id software) (Wright, 2000). The same game principles as in Clan Arena are applied in Counter-Strike and in Tactical Ops: Assault on Terror. Tactical Ops was first published in 2000 as a user-made modification for the FPSG Unreal Tournament (Epic Games). Tactical Ops: Assault on Terror (a sequel of Tactical Ops) provides a sophisticated single player mode in which all game characters except the player's avatar are generated by the computer. This setting guaranteed that a players' performance was not influenced by other human players' skills or actions. Players could freely choose among 15 different maps that were comparable in terms of size and player mission. The maps had been preselected by the researchers. We assumed that the possibility to choose among several maps increased the external validity of the gaming situation while, at the same time, keeping the basic gaming principles applied in the selected maps rather constant. Per default setting of the game, every map could be played to a maximum of 12 minutes and every round within a map was limited to 2 minutes.

Definition of violence, coding scheme, and units of analysis

Definition of violence

We developed a coding scheme to identify various player actions, game playing situations, and game events. The coding scheme was initially designed for an experimental study on violence in first-person-shooter games (Weber, Ritterfeld & Mathiak, 2006; Mathiak & Weber, 2006). Therefore the main goal of the content analysis was to differentiate between violent and nonviolent player actions. In Tactical Ops: Assault on Terror, every violent interaction is performed by using weapons. The option of using verbal violence against other players is not built into the game. Thus we defined player violence as those periods of playing time in which a player fires his gun2.

Coding scheme

We defined seven characteristic playing phases in order to describe the most important player actions during the game along with 18 codes for events that describe transitions between the playing phases. In addition, we recorded the start and end time of playing phases.

FPSGs usually consist of various situations where players do not engage in virtual violence. Game descriptions and strategy guides for FPSGs (e.g., http://whisper.ausgamers.com/warstrats.htm#warstrats; see footnote 1) tell us that FPSG play follows a kind of game plot with different characteristic playing phases. At the beginning of a new round players are located at the opposite ends of a map. They prepare for the next match by buying weapons, ammunition, and other equipment. Then players start approaching each other trying to accomplish different tasks such as reaching and taking control of strategic places in a map that would provide them a defensive advantage or allow them a good hiding spot. By contrast, the players may chose to rush up on the adversary to set the stage for a major attack. After these first strategic moves, players will catch sight of their opponents. Finally the players engage in combat. Cycles of play patterns such as this may appear more than once during a round.

We designed our coding scheme to resemble this variety of player actions. Besides differentiating between violent and nonviolent player actions, we defined six additional playing phases in order to identify the nonviolent player actions during the game. In sum, seven phases describe the characteristic player actions: 1) waiting time between the rounds due to computer's loading time (not considered for later analysis), 2) use of equipment menu, 3) safe–no opponents/strangers in the player's visual field, 4) danger–opponents/strangers are in the player's visual field, but no violent actions are directed to or committed by the player, 5) combat–player fires gun, 6) under attack–the player is attacked by an opponent before or after using his own weapon, and 7) ghost mode–player is killed before the end of a round. Phase 2 differs from the other game phases in that the equipment menu is displayed on the screen instead of the gaming environment which may disturb the level of immersion into the game. However, as the equipment menu can be accessed during the game at any moment, this phase is considered an integral part of game play. Based on these phases, the amount of typical FPGS content created by various players can be analyzed. Characteristic game playing patterns can be examined by an analysis of consecutive playing phases.

We also analyzed several game events that mark the beginning of a new phase. For example, a safe-phase can start after combat due to different events: The player could have killed his opponent in combat, or a team member of the player could have killed the opponent. Both interactions lead to a safe-phase if there are no further opponents around. But the player's emotional responses to these events may differ. Players may experience pride after the first event, and gratitude due to the second event. Therefore, we defined a total of 18 codes (labeled as event codes) for different events that lead to the beginning of a new phase (see Table 1). These codes were identified by examining the game rules, by playing the game, and by examining the recorded material played by our participants. In the process of analysis, every event that occurred in the recorded material could be assigned to one of the predefined codes. We did not find events that we were unable to cover with our coding scheme. We concluded that the 18 event codes covered the events in the recorded game play sufficiently.

Table 1.  Description of 18 event codes
Phase 2: Use of Equipment Menu
Phase 3: Safe, no opponent/stranger on the screen
300Player starts new round
310Player returns from phase 2 (equipment menu)
320Player killed opponent in phase 5 (combat), no attention to killed opponent
322Player killed opponent in phase 5 (combat), attention/shooting of dead opponent
324After phase 5 (combat), opponent disappears from the screen
330After phase 4 (danger), opponent disappears from the screen
340After phase 4 (danger), player disappears from opponent's visual field
350After phase 4 (danger), player identifies stranger as team member
360After phase 4 (danger), player identifies stranger as hostage
Phase 4: Danger–opponent/stranger in players' visual field
400Opponent/stranger appears in player's visual field
410Player withdraws from phase 5 (combat) without having killed opponent
415After phase 5 (combat), player runs out of ammunition
Phase 5: Combat–player fires gun
500Player uses weapon
Beginning of Phase 6: Under attack–player is attacked by opponent
Phase 7: Ghost mode–player is killed during a round
700Player is killed during a round and switches to third person perspective

Units of analysis

For every phase, the beginning and ending was captured with the precision of a single video frame (25/s). After merging the content analysis data with the physiological data (see below), the time resolution was 1 Hertz - for every second of the entire game play we collected data of both the generated content and physiological response. We consider and define this as high temporal resolution.

Physiological measures

We recorded heart rate (HR) and skin conductance response (SCR) during game play. HR was extracted from peripheral pulseoxymetry (oxygen/pulse monitor; Nonin Medical Inc., Minneapolis, MN) that was mounted to the left index finger. The game was controlled by a trackball and button device with the right hand. The left hand, including the left index finger, was not used to control the game. Several test runs indicated that the HR sensor did not interfere with controlling the game and with game performance. Automatic pulse curves peak detection allowed the measurement of beat-to-beat interval and the calculation of instantaneous HR. To avoid artefacts from hand movements, SCR was measured at the left foot above the abductor hallucis muscle about midway between the proximal phalanx of the first toe and a point below the ankle medial to the sole of the left foot. Silver-silver chloride electrodes and unibase electrolyte were used. The signal was sampled with an ambulatory digital recorder (Vitaport II; Becker Meditec, Karlsruhe, Germany) at 80 Hertz. Data were bandpass filtered (cutoff frequencies, 0.05 and 10 Hz) to reduce signal drifts and artefacts. SCR was defined as squared fluctuations of the signal.

Playing procedure

The participants played the game during an experimental procedure already described elsewhere (citation redacted). Prior to the experiment, the physiological measures were set up, the participants tested the video game controller, and became used to the playing environment.

A postplaying questionnaire (scale: 1 “totally disagree” to 9 “totally agree”) indicated high means for compliance: “The study was fun” (M = 7.4, SD = 1.6); “The study was interesting” (M = 7.9, SD = 1.4); “I felt bad during the measurement” (M = 2.2, SD = 1.5); and “I would participate in a similar study again” (M = 7.6, SD = 2.1). Immersion in the game play was rated above scale mean: “I felt like I was acting in the environment rather than controlling a game” (M = 4.7, SD = 2.4); “I felt present in the game environment” (M = 5.7, SD = 2.3); “From time to time I was not aware of my real environment” (M = 5.6, SD = 3.5); “The game required all of my attention” (M = 5.6, SD = 2.3). Overall, the participants felt comfortable in the experimental setting.

Coder training and reliability

Two coders (male graduate students from a private western U.S. university) and a member of the research team as supervisor coded the recorded game play videos. The coders received 16 hours of training in which they discussed the different playing phases with experienced players of video games and learned to rate events and violent interactions. The training was based on a video of one player's recorded game play (not included in the final analysis). The entire coding procedure took 120 hours per coder. Two-thirds of the material were coded by both coders to check for coding consistency which yielded an overall intercoder reliability of 0.81 (Cohen's Kappa for two coders and multiple codes). Inconsistent codings were discussed with the coders after coding and after calculating intercoder reliability. Based on these discussions, the supervisor corrected obvious errors (e.g. typos). Inconsistent codes that could not explained by obvious coding errors were not considered in the analysis.


We proposed three research questions. Our first research question was: What type and extent of violent content do experienced players of a typical first-person-shooter game generate over a typical period of game play? Our second research question asked for playing patterns: What are typical playing patterns applied to first-person-shooter games by experienced game players? In order to answer research questions one and two, we first identified various game playing patterns in an interplayer and intraplayer analysis. We then used the identified game playing patterns to examine differences in arousal responses between players and to answer research question three: How do arousal levels of experienced FPSG players respond to individually generated game events over an extended period of game play?

Type and extent of violent content

Our first research question focuses on the type and extent of violent content that is generated in FPSGs. To answer this question the frequency of occurrence and the duration of the main playing phases were analyzed both across all players and for each individual player. Violent player actions were defined as those periods of playing time when a player fires his weapon (combat-phase). The analysis revealed that these violent player actions occurred in 15% of all events and accounted for 7% of the total time played. In addition, experience as the target of video game violence (when players were attacked by opponents without engaging in violent behavior themselves) was also analyzed. This was covered by the under-attack phase of our content analysis, occurring in 7% of all events and corresponding to 1% of the time played.

The safe phase (no opponents/stranger in the player's visual field) was generated most frequently with 36% of all events and accounted for almost half of the total time played (45%). The danger phase (opponent/stranger in the player's visual field) showed the second most frequent occurrence with 28% of all events, but accounted for only 8% of the total time played. The ghost-mode phase was generated in 5% of all events, but represented 26% of the total time played. Use of the equipment menu occurred in 10% of all events and with duration of 14% of the total time played. In summary, the most frequently game playing situations were safe, danger, and combat, whereas the safe, ghost-mode, and equipment-menu phases accounted for the majority of time.

We further analyzed percentages of playing time for each phase per each individual player in order to examine differences between players (Table 2).

Table 2.  Percentage playing time by phase and player
  1. Note: Framed cells mark values below the lower limit (LL) of the confidence interval across all players, shaded cells mark values above the upper limit (UL) of confidence intervals

Ghost Md.18.4815.2937.7024.1439.2720.5319.5125.3923.5025.2324.8935.9024.4425.620.4224.7126.54
Total (%)100100100100100100100100100100100100100    

Confidence intervals for small samples with a confidence level of 95% (with a t-value of t(df(12)= 2.18), served as criterion to report statistically different values between players. For example, player 1 showed an above average value for the combat-phase, he spent about twelve percent of his playing time in combat. He also showed below average values for the phases of danger and ghost mode. This indicates that participant 1 played successfully–he engaged in extensive combats, killed his opponents effectively, and was in the ghost mode after being killed for short periods of time only. In contrast, player 2 displayed above average durations for safe and danger, and the shortest duration for ghost mode. He spent much time in the safe-phase and little time in combat or in the ghost mode. These results suggest that player 2 acted rather cautiously. Likewise, all other players could be categorized. These analyses demonstrate that although all players possessed comparable past FPSG playing experiences, they applied different playing tactics and hereby generated varying intensities of exposure to video game violence.

Frequency and duration of game playing patterns

Our second research question asked for typical playing patterns applied to FPSGs by experienced game players. Thus, we computed progressions of phases and event codes. The shortest sequences comprise two consecutive game events, e. g. event 300 followed by event 200. In this situation, a player started a round in the safe phase and then opened the equipment menu. The analysis of two-event sequences yielded 143 different codes. A similar analysis was performed for all sequences of three consecutive event codes, resulting in 483 different three-event-sequences, and for all patterns of four consecutive event codes (N = 1.057).

The goal was to find an optimal match between the complexity and explanatory power of the patterns. Two consecutive events do not seem to represent typical cycles of player actions. The playing time covered by two-event sequences are rather short (M = 13.17 seconds, SD = 15.22). The three-event sequences cover actions patterns such as “player in combat-phase–kills an opponent and returns to safe-phase–player in danger-phase after another opponent appeared”. On average, each three-event sequence lasts for 19.74 seconds (SD = 18.34). The 10 three-event sequences that occur most often account for 41.1% of all three-events sequences; the 16 most frequent sequences account for 50.9% of all three-event sequences. The four-event sequences have an average duration of 26.32 seconds (SD = 21.76), and the 10 most frequent sequences account for 28.3 % of all events.

For further analyses, we decided to use the three-event sequences, because they represent player actions on a more complex level than two-event sequences do. Compared to the four-event sequences, which represent highly individualistic playing patterns, three-event sequences still allow for the identification of typical patterns of game play. Of the 483 three-event sequences, the 16 most frequently occurring sequences account for 50.90% of all sequences and comprise 58.51% of the total playing time. Frequency and duration of the three-event sequences are highly correlated (r = .922, p < .01) which means that frequently occurring sequences also account for large amounts of playing time. Thus–and for reasons of simplification–we focus on these 16 three-event sequences (Table 3) in order to answer our second research question.

Table 3.  Frequency of occurrence and duration of play time for 16 three-event sequences across all 13 players
three-event sequenceDescriptionFrequency (%)Playing time (%)
700-300-200ghost mode–new round/safe–equipment4.5510.44
500-320-400combat–player killed opponent/safe–danger2.881.76
400-500-320danger–combat–player kills opponent2.841.97
400-330-400danger–opponent disappears–danger2.530.93
400-500-415danger–combat–player runs out of ammunition2.470.86
400-350-400danger–stranger identified as team member/safe–danger2.391.57
500-700-300combat–ghost-mode–new round/safe2.384.98
600-700-300under attack–ghost mode–new round/safe2.163.91
400-600-500danger–under attack–combat1.950.36
330-400-500opponent disappeared/safe–danger–combat1.480.64
500-300-200combat–new round/safe–equipment1.450.96
400-500-700danger–combat–ghost mode1.422.79
500-415-600combat–player runs out of ammunition–under attack1.360.40
Total 50.9158.51

The most frequent three-event sequence was 200-310-400, which occurred in 8.13% of all three-event sequences and accounted for 10.81% of the total playing time. This combination indicates that a player used the equipment menu, returned to the safe phase and entered the danger phase because he spotted a potentially dangerous character. The second most frequent three-event sequence is 300-200-310 (safe-phase, use of the equipment menu, return to the safe-phase). This sequence occurred in 8.10% of all three-event sequences and accounted for 10.77% of the total playing time. Both event sequences are typical for the beginning of a new round in a FPSG like Tactical Ops: Assault on terror: The players enter the equipment menu to buy weapons and ammunition and then start searching for opponents. Having returned to the safe phase from the equipment menu and spotting an opponent in their visual field, the players enter the danger-phase and often begin to use their weapons. This action pattern is represented by the third most frequently three-event sequence, 310-400-500, which occurred in 4.81% of all three-event sequences and accounted for 5.34% of the total playing time. In most combats, the players managed to kill their adversaries (codes 500-320-400, 400-500-320), approaching the next opponent afterwards.

These results suggest that there are typical patterns of play for FPSG. Players start a new round, buy equipment, look for opponents, and engage in combat. In combat, they either manage to kill their opponents, approaching another opponent afterwards, but sometimes the opponents hide from the players, attack them first or while they ran out of ammunition. Being shot before the end of a round and waiting for the next round to begin is also not uncommon. The results show that despite a variety of 483 occurring three-event sequences, typical playing patterns can be identified.

Physiological responses to game events

With the third research question we asked: How do arousal levels of experienced FPSG players respond to individually generated game events over an extended period of game play? In order to answer this question, we analyzed HR and SCR as physiological indicators for arousal. We conducted two analyses of variance in a repeated measurement design (one for SCR and one for HR). We used time and game events as within-player factors: We compared the first four maps played by each player which covers approximately the first 40 minutes of playing time, and we used the event codes from the content analyses above to focus on physiological responses to game events. Within the 16 three-event sequences that accounted for half of the total time played, only 11 out of our 18 previously defined event codes occurred. This indicates that the remaining seven events are of minor importance for typical FPSG game play. For our analysis of psycho-physiological responses, we therefore reduced the number of game events by merging similar codes: We merged code 322 (player killed opponent and pays attention/shoots dead body) with code 320 (player killed opponent, does not pay attention to the dead body), codes 324 (opponent disappears during combat) and 340 (player hides from opponent) with code 330 (opponent disappears after danger-phase) and code 360 (stranger identified as hostage) with code 350 (stranger identified as team member). All codes that were merged with another code accounted for less than 1% of all events. Two event codes (410, 416) were deleted from the subsequent analyses, because they did not occur among the most frequent three-event sequences and could not be merged with one of the other events. In sum, we analyzed the psycho-physiological responses of our participants across 11 game events.

We aggregated HR and SCR for 11 of the 13 players for the first four maps (sessions) played, and for the mentioned 11 game events. Two players had to be excluded because the physiological data were incomplete.

For SCR, the analysis revealed decreasing SCRs over time and distinct responses to game events. Both the effects of time (F(1.2,12.0) = 2.153, p = 0.17, η2partial = 0.18) and of game events (F(2.1,21.2) = 1.32, p = 0.29, η2partial = 0.12) did not reach significance. This can be in part explained by the rather small sample size of 11 players.

With respect to the players' HR, the analysis revealed significant differences for both time and game events. Over time, HR averages decreased significantly (F(1.9,19.0) = 10.20, p < 0.01; Figure 1) with an effect size of η2partial = 0.50. The influence of game events on heart rates also reached significance (F(3.6,35.6) = 4.49, p = .006, η2 = 0.31; Figure 2).

Figure 1.

Heart rate by time.

Figure 2.

Heart rate by game events.

Events with highest heart rates were 300 (player started round) and 415 (player ran out of ammunition while under combat) indicating that starting a new round always goes along with high arousal, presumably due to uncertainty of the players regarding the upcoming events and their performance. Event 415 (player runs out of ammunition while under combat) produces the highest heart rates over time, reflecting the players' arousal in these situations of imminent danger to the player's character. The event with the lowest heart rates was 400 (opponent appears on the player's visual field, danger), indicating that observing opponents produces less arousal than combat, and that players concentrate on the game and what will happen. The second lowest heart rates are produced by event 200 (equipment menu) indicating that players are very focused while they prepare for combat through buying equipment, even if starting a new round produced high arousal before. In summary, the results answer research question four. Experienced players of FPSGs vary in their psycho-physiological responses to different game events over time.


The present study examined player generated, individualized content in a typical FPSG and physiological responses to game events over time. We conducted an event-related, high-resolution content analysis of 13 experienced first-person-shooter game players with 50 minutes of playing time per player on average. We analyzed frequency and duration of defined game phases and three-event sequences over all players as well as for each player separately in order to describe both general player generated content and specific individual game playing patterns. Afterwards we combined game events derived from the content analysis and psycho-physiological data to examine arousal in different game playing situations.

In our first research question we asked for the type and extent of violent content generated by experienced players of a typical FPSG over a typical period of game play. What stands out is that the players spent an average of 45% of their playing time in the safe-phase exploring the game's environment, searching for opponents, and working on combat strategies. Moreover, the players spent an average of 26% of their playing time in the ghost-mode phase, which is a common feature among Counter-Strike and other FPSGs. During these times, the players were not engaged in violent interactions. Violent player actions, i.e. the use of the weapon, accounted on average for only 7% of the total playing time. This is not due to low player skills: all players were experienced video game players and had time to practice before their game play was recorded. Also, Tactical Ops is a prototypical FPSG. Even though the participants were not familiar with the game prior to this study, they should not have experienced surprise or anxiety far beyond their normal level which could have interfered with our results. The participants were also exposed to violent content during under attack-mode play, but these situations accounted on average for only 1% of the total time played. This also indicates the players' expertise with FPSGs and supports the notion that the small amount and duration of combat situations is typical for this game genre and game mode. Results may differ for games like Unreal Tournament, in which players return to the game after being killed instead of switching to a ghost mode.

Our second research question asked for typical patterns of playing behavior as applied by experienced players of FPSGs. The examination of sequences of three consecutive game events revealed that 16 of these event sequences account for 51% of all three-event sequences and for 59% of the total playing time. The most frequent three-event sequences correspond to the game plot described in strategy guides: Players start a new round, buy equipment, head towards their opponents, and engage in combat. If players manage to kill their opponents, they engage in further combats, but in many cases they are killed and switch to an observational mode and have to wait for the beginning of the next round. The findings indicate FPSG are indeed played as one would intuitively expect, and that playing FPSGs follows a typical plot. Further research is needed on how this plot affects the player's entertainment experience, feelings of presence, and aggressive reactions. Nevertheless, we found 483 three-event sequences in total, indicating that there is a variety of individual playing patterns besides typical patterns of player actions. These playing patterns might define different playing strategies which in turn might be related, for example, to personality traits and the player's experience and could also mediate video game effects.

Our third research question was: How do arousal levels of experienced FPSG players respond to individually generated game events over an extended period of game play? We measured heart rate and skin conductance responses as indicators for arousal and analyzed these measures for 11 of the 13 players for different game events. We found clearly interpretable and meaningful differences in HR for 11 distinct game events. Although mean differences in SCR were not significant, the clear pattern and the large effect size suggest that this result is due to the small sample size. We found the highest arousal levels when a player was forced to stop shooting in combat because he ran out of ammunition. Arousal is rather low when a player observes an opponent or uses the equipment menu. These low arousal levels during enemy observations seem somewhat surprising at first sight, since these situations are more dangerous to the player's character compared to, for example, simply exploring the game environment. We assume that in this case, arousal is a function of uncertainty and perceived control. Uncertainty about where enemies might hide and from where they will attack is rather high as long as they are invisible. This uncertainty should go along with reduced levels of perceived control and thus increased levels of arousal. Uncertainty about threats is considerably lower in situations when opponents can be clearly located by the player, which should lead to increased perceptions of control and, in turn, decreased levels of arousal.

We also found increased levels of arousal at the beginning of the game, but decreasing arousal levels when playing a FPSG for an extended period of time. These findings correspond to previous research on violent video game play and arousal. In previous experimental studies that found increased arousal levels during play, the period of play time was typically as short as 10 minutes (e. g., Ivory & Kalyanaraman, 2007; Ballard & Wiest, 1995), 15 minutes (Barlett, Harris & Baldassaro, 2007), or 20 minutes (Anderson et al., 2004). The decrease of arousal levels after an extended period of play time mirrors the finding that aggression is negatively related to play time (Sherry, 2001).

Our content analysis has two limitations. First, the coding scheme was initially designed for an experimental study on violent interactions in FPSG (Weber, Ritterfeld, & Mathiak, 2006; Mathiak & Weber, 2006). A more sophisticated coding scheme considering the context of violence (cf. Smith, 2006) might have revealed even more detailed information on player interactions. However, the coding scheme allowed for the identification of violent interactions and other important player actions. Our findings hereby provide first insights into typical and individual FPSG playing patterns. Second, only 13 male players were analyzed with a rather laborious content analysis procedure which clearly restricts generalizability. Because each player played about 50 minutes, the study design provided enough information for a demonstration of an event-related, longitudinal content analysis with high temporal resolution.

Our results have several implications for research on video game violence. Players of FPSGs seem to experience rather small amounts of violence compared to the time they spend in nonviolent gaming situations. Under the light of previous findings on increased aggressive reactions to violent video games (Weber, Ritterfeld & Kostygina, 2006), questions arise concerning to what extent these reactions can be traced back to actual violent behavior, and to what extent the context of violence or the “dramaturgy” of game playing contributes to aggressive reactions. When investigating the psychological mechanisms leading to increased aggression, we need to go beyond the macrolevel of game content as a whole and account for the effects of game content on the microlevel of specific events.

With regard to individual game content as created by various players, the possible consequences of different emotional reactions should be considered. Appraisal and decision processes (Anderson & Bushman, 2002) of users applying very violent tactics might be different compared to users who play the game in a more cautious way. Moreover, we should expect that player interpretations of the video game content they are exposed to influence their judgments of the violence and resulting aggressive reactions.

The methodology applied in this study revealed detailed insights into how FPSGs such as Counter-Strike are typically played. Although previous content analyses are helpful in determining stable characteristics of video games like character appearances, their representation of the content typical (FPS) game players are exposed to is superficial. Our study shows that media effects research using psycho-physiological measurements can be illustrative. If individual differences in game play are considered, great insight on the effects of video games can be gained from studies applying content analyses and psycho-physiological measures. Moreover, our results raise questions of treatment design in media effects studies of violent video games. Usually, effects of violent video games are examined comparing groups of participants who played a violent versus a nonviolent game. These comparisons are tied to problems regarding internal validity, because violent and nonviolent games may vary in terms of suspense, immersion, entertainment, etc. Instead of using different games in effects studies, one could use content analysis of game play and compare groups of participants who played the same violent video game in terms of the amount of violence they actually created. When comparing violent and nonviolent games, it is also questionable whether the amount of time is meaningful for comparison. The number of violent game events as committed by each player could be a more valid predictor for amount of violence.

Yet, we should caution. Our findings indicate that experienced FPSG players generate rather small amounts of violent content and their arousal decreases after prolonged game play. These results, however, do not indicate that latest-generation FPSGs have less harmful effects. Event-related fMRI analyses of experienced players and similarly generated violent content (Weber, Ritterfeld, & Mathiak, 2006; Mathiak & Weber, 2006) have shown a link between small amounts of violent interactions in a FPSG and brain activity patterns that can be considered as characteristic for aggressive cognitions and affects. Nevertheless, the results of our analyses emphasize the importance of examining video games under consideration of their interactive nature. Due to this unique characteristic, every player creates his or her own specific content that may moderate potentially harmful effects.


1 Examples of playing strategies for Counter-Strike in general: http://guillaume.tournand.com/project_cs_guide/counter-strike_strategy_guide.pdf, http://strategywiki.org/wiki/Counter-Strike:_Source/Game play_strategies, http://whisper.ausgamers.com/warstrats.htm. Examples of strategies for particular Counter-Strike maps: http://planethalflife.gamespy.com/View.php?view=CSStrategies.Detail&id=2&game=5, http://www.csnation.net/view.php/strategy/maps/, http://www.counterstrikestrats.com/

2 Uses of weapons with no need to score were coded, but not treated as violent behavior. These actions differ from the use of weapon in that they are due to e.g., curiosity or strategic reasons instead of trying to kill adversaries.

3 Greenhouse-Geisser sphericity correction was applied for all F-Tests.

About the Authors

René Weber, Ph.D. (Dr.rer.nat.), M.D. (Dr.rer.medic.), is an Associate Professor in the Department of Communication at the University of California Santa Barbara. His research interests include the cognitive and emotional effects of television and new media technology.

Katharina-Maria Behr is a Ph.D. candidate in mass communication at the Institute of Mass Communication and Media Research, University of Zurich. Her research interests include the use of new media technologies, media entertainment, video and computer games, and the modification of computer games.

Ron Tamborini (Ph.D., 1982, Indiana University) is the Director of Doctoral Studies and a Professor in the Department of Communication at Michigan State University where he teaches courses in communication research methods and media influence processes. His research focuses on traditional and new media in educational and entertainment settings.

Ute Ritterfeld, Ph.D., is professor of Media Psychology at the VU University Amsterdam, co/founder and director of interdisciplinary research at the Center for Advanced Media Research Amsterdam (CAMeRA).

Klaus Mathiak, Ph.D., M.D., is head of the Behavioral Psychobiology group at the RWTH Aachen University in Germany since 2004. His current research focus is on functional brain imaging of complex cognitive functions and the neuronal basis of media experience as a model for social behavior.

Address: René Weber, Ph.D., M.D. University of California Santa Barbara, Department of Communication, Ellison Hall 4020, Santa Barbara, CA 93106, U.S.A. Phone: +1-805-893-2156

Email: renew@comm.ucsb.edu