Stimulus Materials

Anderson pointed out in 2002:

In the early days of digital games, Pac-Man (...) had some parents concerned about potential consequences of playing this "violent" game. So it should come as no surprise that the violence of the "high violence" conditions in early studies is very different from the high violence games in more recent studies. (p. 110)

That means, aside from the shifts in perception of (media) violence in society, there have been vast changes in violent content of digital games over the last two decades. Still, a lot of researchers use games that are heavily outdated, even considering the time it takes to administer an experiment, write a paper, have it reviewed and published. Although it is not really considered a first-person shooter as they claim but rather a rail or gallery shooter (since the player cannot use controls to move, only shoot), Barlett, Harris, and Baldassaro (2007) used Time Crisis 3 for their study, published in 2003 by Nextech, so it was fairly up to date. Anderson and Carnagey (2009) used, among other games, Madden NFL 2004 (EA Tiburon, 2003), at the time of the article publication 6 years old.

However, they also used NFL Blitz (Midway Games), a game from 1998, so there were 11 years of game development between the game and article publications. In the frequently cited study by Anderson and Dill (2000), the games Myst (Cyan, 1993) and Wolfenstein 3D (id Software, 1992) were respectively 7 and 8 years older than the publication. The stimulus material in the study by Anderson et al. (2004) was even more antique, Marathon 2 (Bungie Software, 1995), and Glider PRO (Calhoun, 1991). The latter was published 13 years before the article. Tafalla used Doom 1 (id Software, 1993) for his study from 2007, but the record, if you want to call it that, is held by Bushman and Anderson (2009), whose "latest" game was Future Cop: LAPD (Electronic Arts, 1998), and the oldest Duke Nukem (Apogee Software, 1991), published astounding 18 years before the article, so that some of the participants in their study might likely not even have been

born when the game hit the shelves. Why the authors did not use a more recent game, like e.g. Counter-Strike: Source (Valve Corporation, 2004a), Far Cry (Crytek, 2004), or Unreal Tournament 2004 (Digital Extremes & Epic Games, 2004), all three violent digital games released in 2004 and thus 5 years older than the publication, leaves room for speculation.

Considering the technological improvement in computers, we have to ask ourselves as researchers: Can we learn anything about the effects of today’s games from studies on games that are 5, 8, 12 or even 18 years old? Are those games really the same in principal as new games? Or have games changed in graphics (see figure 3.1 for an example), sound, gameplay, and many other ways so drastically that they cannot be taken as one homogeneous group with the games that were created more than a decade ago? There has been little research on the psychological effects of technological advancements in digital games, but results indicate e.g. that higher image quality leads to a significantly higher immersion (Bracken & Skalski, 2009), presence, involvement and even physiological arousal (Ivory & Kalyanaraman, 2007). Even if it was true that the games commonly used in research cause increases in aggression, this would not be a big problem since those games have scarcely been played for many years anyway. Therefore, in this work, I used a currently popular, latest-generation first-person shooter.

It is hard enough to find adequate violent games for research, though: There are no clear answers to the questions, whether or not ice hockey games are violent, or if Super Mario is acting aggressively when he jumps on a Goomba. And even when two games are clearly about fighting, the violence of one game does not necessarily equal the violence of another. However, even more problematic is the selection of stimuli for control groups, so called "non-violent games". In experimental laboratory psychology, it is absolutely vital to manipulate the variable a researcher is interested in, while controlling all other possibly confounding variables so that they do not interfere with any effect that should be explained by the manipulation. However, this is rarely taken into consideration in digital game research. It is very convenient to divide games into two groups (violent and non-violent), but one should not forget that the occurrence of violence is unlikely to be the only difference between two games. Those differences can be conflating variables that might confound any result if they are not controlled. For example, a first-person shooter is likely to be more violent than a regular racing game. However, a first-person shooter is also played from the first-person perspective (hence "first-person"), probably has a narrative, intriguing characters, and is controlled with keyboard and mouse, all features very unlikely to find in a racing game, but that might have an impact on taken measures in an experiment. While Carnagey and Anderson (2004) are making a good first step when they say that "the obvious solution for future studies is to do more pilot testing

Figure 3.1: Duke Nukem, used in the study by Bushman and Anderson (2009); Counter-Strike: Source, an available alternative; Unreal Tournament 3, used in this work.

or manipulation checks on such aggression-relevant dimensions" (p. 9), this statement neglects two problems: Firstly, it might be hard for a researcher to even find all the aggression-relevant aspects and control for them via a simple questionnaire that asks how exciting, frustrating, or fun a game was (e.g. Carnagey et al., 2007). Secondly, not even the purely aggression-relevant aspects might be interesting to control, but also those that change the overall experience of a game. The question really is: What is really found when we discover differences between a group that played Grand Theft Auto: Vice City (Rockstar North, 2002) and another that played Tetris? (Cicchirillo & Chory-Assad, 2005). This work’s section on modding (see 5.1.2) offers a suggestion how to address some of the problems commonly encountered in experimental digital game research.