Gary Schober, Åbo Akademi University, gschober@abo.fi
Kaj Bjorkqvist, Åbo Akademi University, Kaj. Björkqvist@abo.fi

Abstract

This pilot study examined the relationship of self – reported direct physical and indirect aggression displayed in the past to the number of times a participant responded aggressively in a computer game known as Mimics (1997).  Aggressive responding in the computer game was defined as the simulated punching of orbs (i.e., computer controlled game characters) with an avatar (i.e., player controlled game character). Hierarchical regression was used to formally test for sex differences, showing that the relationship of aggressive responding in the computer game for males was positively associated with direct physical aggression and negatively associated with indirect aggression. Conversely, aggressive responding for females was found to be positively associated with indirect aggression and no association was found with direct physical aggression. The associations mirror documented findings on sex differences and direct physical and indirect aggression (c.f., Björkqvist, Lagerspetz & Kaukiainen, 1992). Results are discussed within the context of direct and indirect aggression and Fisler’s (2006) theory of the performed player.

Introduction

The interaction of human and computer game has been studied using a variety of techniques derived from an assortment of disciplines. As serious academic communication, studies concerning computer games largely began as an assessment of whether they were causes of real life aggression in humans (Frasca, 2003). The discipline is characterised by exclusively examining the medium as a form of digital text, images, and storylines no different from television, and no different from playing with war toys or engaging in aggressive sports (Frasca, ibid.). This research philosophy describes narratology which is a literature research method focusing on how stories are told, along with the linguistic and representational processes that are involved (Dillon, 2005; Frasca, 2003; Küchlich, 2003). According to Egenfeldt - Nielsen (2003), narratology in relation to computer games has its strengths in explaining the dynamics of the adventure game genre, where a meaningful engaging game is created through a carefully constructed story, but fails when it comes to alternative game genres. Some researchers (c.f. Dillon, 2005; Frasca, 2003; Juul, 2000) have rejected this over - emphasis on narrative explanations which do not acknowledge computer games as a distinctive medium from that of television. Computer game research is thus characterised by a theoretical tension between two lines of thinking on the subject, i.e., narratology, and ludology (c.f. Egenfeldt-Nielsen, 2003; Frasca, 2003; Juul, 2000; Küchlich, 2003).

Traditional literature theory and semiotics fail to deal with electronic texts, adventure games, and textual based multi-user environments, because these types of technology are not just made of sequences of signs, but rather behave like machines or sign-generators (Frasca, 2003). This fact gave rise to the academic study of ludology - the study of games in all its forms, particularly computer games. The impetus for computer game studies was largely inspired by Aarseth in the late 1990’s (Frasca, ibid.). Aarseth (1997) created a typology of texts and showed that hyper text is just one possible dimension of these systemic texts, which he called cyber texts. As Frasca explains, the reign of academic representation was contested paving the way for simulation, and game theory. The spirit of ludology can be articulated by a statement made in Juul’s (2000, p.1) paper, “we need a ludology - a theory of games - and to get one, academia must learn from the game development community”.  The objective of this study was to refine a methodology aimed at shedding light on the narratology- ludology debate by examining the relationship of playing behaviour, particularly aggressive responding in a computer game to aspects of personality relating to aggressive behaviour with a particular interest in direct physical and indirect aggression.

Direct physical aggression has been defined as physical aggression carried out by the attacker in person in a face-to-face manner (e.g., hitting somebody) (Björkqvist, Lagerspetz, & Österman, 1992b). Indirect aggression has been operationalised as aggression perpetrated by the attacker in such a manner that their identity and/or aggressive intention is unknown to the victim and is also characterised by using others as vehicles to deliver harm (Björkqvist et al., ibid.) (e.g., spreading insinuations about somebody, without direct accusation) which has been described in the literature as a kind of social manipulation (Lagerspetz, Björkqvist, & Peltonen, 1988). All behaviours are operationalised in line with a traditional view of aggression which defines the construct as behaviours intended to harm another living being who is motivated to avoid such treatment (c.f., Baron, 1997, p. 7). From the definitions presented above, the constructs of direct and indirect aggression are in essence contrasting strategically based classifications of aggressive behaviour all defined within the confines of behaviours that have the quality of causing intentional harm to another living being, inherent in a standard definition of aggression.  Since Lagerspetz et al. (1988), results have consistently indicated that males indeed make greater use of direct physical aggression than females, while females on the other hand have been found to use indirect styles of aggression more often than males (c.f., Björkqvist, Lagerspetz & Kaukiainen,1992; Björkqvist , Österman & Kaukiainen, 1992; Österman, Björkqvist, Lagerspetz, Kaukiainen, Landau , Fraczek & Caprara, 1998).

To further extrapolate on the objectives of this study, we were not just interested in how aggressive responding in a computer game related to self reported direct physical and indirect aggression but also how sex differences on the aforementioned constructs of aggression affected this relationship. The following study is distinct from the traditional computer game aggression research which proposes that the simulation of aggressive responding in a computer game by a player leads to personality change (i.e., increased real life aggressive behaviour) as predicted by social learning theory (c.f. Griffiths, 1999 for a review) but rather aims to examine the interaction of features of personality relating to direct and indirect aggressive behaviour to aggressive responding in a computer game. A theory proposed by Fisler (2006) derived from the area of ludology has attempted to explain the interaction between personality and playing behaviour in computer games.

Fisler (2006) postulates that the human - computer game interaction between digital representations (i.e., avatar), and human activity is a mediated interaction which is in contrast to views that have examined the connection as a purely direct one or as a purely passive digital performance (i.e., no connection whatsoever) (Fisler, ibid.).  In this process, Fisler (ibid.) argues that this kind of interaction generates a specialized narcissism which Klein (1946) has called positions, which are the projection of polar or conflated understandings of self into idealised objects (e.g., the avatar in a computer game). From this process, emerging from the medium is the performed player, generated by the interaction between the digital representation on screen, and the human controlling it (Fisler, ibid.). According to Fisler (ibid.), the psychological process for which playing behaviour becomes a mediated representation of real world behaviour is explained by the ego projecting into the object (e.g., avatar) certain qualities of the self and then identifying the object with those qualities. Thus, the process is best explained by a specialized type of transference of the ego, and a specialized type of electronic narcissism (see Fisler, 2006 for a more comprehensive discussion of the theory).

Fisler’s performed player metaphor can be illustrated with the computer game Grand Theft Auto (1998). In this game, the player is rewarded points for killing pedestrians; a direct view point of projection for a player using the game strategy of killing pedestrians to win the game would seem to suggest that the player may be an unsafe driver or worse, or may simply be physically aggressive. However, a mediated view point may suggest that projections of power may come at the expense of the player’s fairness or mercy or put another way this type of playing behaviour may be an indicator of someone who may possibly harm others emotionally along the way to achieve success in many of life’s pursuits (Fisler, ibid). From this example, we can see that the mediated view point is potentially broader in its interpretation of playing behaviour as reflections of real world personality or behaviour. It is very clear that the interpretation is not anchored by narrative thinking, i.e., as a projection, simulated physical aggression in a computer game does not necessarily equal physical aggression displayed in the real world.

A second example of personality interaction with simulated environments can be illustrated with the computer game Civilization (1991). In Civilization, there are several ways to win; we refer only to the military and diplomatic victory choices available within the game, which equate to simulated world conquest and simulated election to the head of the United Nations. The two victory styles can be seen as a direct and indirect approach to achieving the task goal of winning the game respectively.

The indirect style of game play, may, perhaps, compare to individuals who in lived reality adopt an indirect approach to life. It may describe individuals who may resort to manipulative indirect means to achieve success in real life. This could describe a politician paying bribes to get ahead or someone engaging in aggressive gossip to elevate his/her social status. On the other hand, the direct style of game play, the military victory, might describe someone who in real life approaches situations in a direct manner. This could possibly include things like face - face physical or implied physical aggression or verbal aggression. For example, the school bully that takes someone’s lunch money or simply hitting someone because they made him/her angry. It does not necessarily mean that someone who prefers a military victory in Civilization (1991) is a megalomaniac who would engage in taking over the world in real life.

Fisler (2006) suggests that the advantage of the performed player metaphor is that it provides a paradigm for the study of playing behaviour, and lived reality which can examine the subjective experiences of the player constructed on the specialized logic of the medium, examining both connectivity and division between the digital extension, and the player. The metaphor offers a new bridge between an already comprehensive collection of extrinsic and intrinsic studies on computer games. Research models that take into consideration the subjective experiences of the player constructed on the specialized logic of the medium provides a plethora of research opportunities. The video game literature is the discipline that has dominated the extrinsic literature and is considered by Fisler (ibid.) to be one of the research areas that may benefit from the metaphor being applied to a more demanding methodology of impact, acknowledging simultaneously the connection and division between the digital extension and the player.

A research area that can be considered related to ludology, due to the emphasis of the research on playing behaviour, known as performance-based-testing, has used computer game simulations as assessments of personality attributes such as risk taking behaviour, spatial intelligence etc (c.f. Graham, Cook, Cohen, Phelps, & Gerkovich., 1985). Within the area, a number of games have been used and developed for research in the field. Mimics (1997) is one such game designed by Aidman and Shmelyov (2002) which they describe as a social interaction simulator. Mimics (1997) incorporates a technique characteristic of this line of research, which is the use of embedded measures. This technique simply involves using the software of the computer game to record player moves without the knowledge of the player (Graham et al., 1985), theoretically overcoming image management problems inherent within many self-report assessments of personality. Using the Mimics (1997), Aidman and Shmelyov (2002) examined the number of times a simulated aggressive action was used by the player, and successfully correlated this with self-report measures of aggression. Visually, the game is based on Ekman’s (1999) schematized cross - cultural facial universals covering a range of expressions (i.e., orbs and the player avatar are capable of three facial expressions). The game is played on a 9 by 9 grid, similar to a chess board, where only a 3 by 3 segment of the grid is visible to the player at any one time. During play, the participant inevitably encounters these orbs, which are similar in appearance to the avatar contained within the grid, in fixed positions, that block the player’s path to the goal of winning the game, i.e., reaching a symbolic apple. The avatar is used by the player to explore, match the right expression, or respond aggressively (i.e., hit the orb) to find an effective route to pass through the orb grid position.

Operationally, the game’s objective is to elicit and assess strategic and tactical choices in a conflict/cooperation choice paradigm. Aidman and Shmelyov (2002, p. 85) do not offer a theoretical explanation as to why they consider the game to be a measure of real world personality, except they do say, “the player is free to choose between these three main tactics (referring to tactics available to the player in the game) in any individual interaction (referring to the avatar to orb interactions). Mimics can thus be seen as a form of projective game”.                  

Results published by them using a correlation procedure between game play strategies and two self-report measures of aggression, suggest that an aggressive playing style (i.e., hitting player characters within the game paradigm) may be some sort of representation of real world physical aggression, as indicated by significant correlations with only the physical aggression subscales of the two independent self – report measures and not other types of aggression (i.e., verbal aggression or indirect aggression).  They used a measure of direct and indirect aggression based on the work of Lagerspetz, Björkqvist, and Peltonen (1988), and Buss and Perry’s (1992) Aggression Questionnaire. Their findings also seem to be supported by what they seem to operationalise as a representation of verbal aggression (i.e., threatening facial expression of the avatar) correlating significantly with a measure of verbal aggression, as measured by the Aggression Questionnaire, and not any other types of aggression.

In relation to the previously mentioned goals of this study we intended to examine Fisler’s (2006) proposition by investigating whether the avatar in the Mimics (1997) functioned as an extension of the self, similar to the Aidman and Shmelyov (2002) study. However, in contrast to their work, we had a particular interest in sex differences, which was something not examined by them. In relation to our interest in sex differences we speculated whether this would yield any information on the nature of the human - avatar relationship, if any; i.e., is it a direct or mediated one? To facilitate this objective, personality attributes relating to real life aggression of the individual measured by self-report were correlated with applicable strategies of game play.

It has been suggested that the results obtained by Aidman and Shmelyov (2002) supported a direct projection of player attributes validated by an implied narrative interpretation; we wish to say that they do not make an explicit statement on this. For example, direct projection implies that participant’ aggressive response styles, e.g., hitting player characters in the game equals physical aggression displayed by the participant in the real world. However, support for a mediated or performed player perspective is theorised to be represented by significant relationships between aggressive responding in the game (e.g. using the avatar to hit player characters), and two or more types of aggression (e.g., direct physical, and indirect aggression). This would be considered to be particularly evident if the characteristic aggressive styles of sex, and direct and indirect aggression outlined above, were found to both be associated with aggressive responding in the game (i.e., hitting player characters), as a function of sex. Such a finding would mean that the hitting function was not necessarily a synonym for physical aggression representing narratological thinking, but rather a general indicator of real world aggressive response styles measured as direct and indirect aggression, identified by past research as a characteristic of the sexes; therefore, representing a mediated projection by showing that the hitting function holds duel meanings in relation to aggression. Thus, we expected aggressive responding in the Mimics (1997) to be positively associated with direct physical aggression in males and with indirect aggression in females.

Method

Participants   

The study used two samples enrolled at the University of Ballarat, Victoria, Australia who were eligible for course credit for attending. The sample consisted of 14 males (M = 27.4, SD = 8.8 years) and 33 females     (M = 20.7, SD = 6.2 years), who completed a self-report assessment of aggression and then played a computer game.

Apparatus

Mimics: Mimics (1997) is a computer game, which requires the player to navigate an avatar to reach a target by passing through a grid in a prescribed amount of time, with limited resources (i.e., power needed for aggressive responding). Decision making in the Mimics centres on conflict or cooperation with the opponent orbs present in the game that block the player’s path to winning the game. The player is therefore required to decide on a method either aggressive or cooperative to pass through the orb grid position or search for a new route to the goal.

An aggressive solution available to the player in the paradigm to pass through the orb grid is to hit the orb. Embedded measures are calculated in three distinct situations within the paradigm. The first variable assesses the number of hits delivered by the participant as an intrinsic choice (Hit₁). The second variable calculates aggressive responding by the player after the avatar is hit by the orb (Hit₂). The third variable calculates aggressive responding when the avatar is blocked by the orb (Hit₃).  

Alternatively to aggressive responding, cooperative based strategies are also available to players to pass through the orb grid position. These strategies involve participant’s using the avatar to match the right facial expression with the orb. There are four available options; a player may use a happy, neutral, sad, or angry face (Threatening). When successfully adopting one of these strategies, the player is rewarded power to the avatar, giving the participant more playing time, thus scoring more points in the game, and countering the effect of aggressive responding draining power from the avatar. Thus, Mimics balances through player rewards the use of either conflict or cooperative based solutions, which theoretically results in the projective function of the computer game.

Self-reports

Direct and indirect aggression measures: To measure direct and indirect aggression, it was necessary to use an adaptation of the original peer estimation measure pioneered by Lagerspetz et al. (1988), designed to assess school children in the classroom which requires a social network. The original inventory includes measures of direct physical and indirect aggression. As the present study was examining an adult population where no social network was present, it became necessary to adopt an adult self-report version of the original peer estimation inventory. Campbell, Sapochnick and Muncer’s (1997) adaptation of the instrument for use among adults into a self report format was used in the study. Respondents answered items within the context of, “Circle the response closest to how you usually respond when you are angry with or get into conflict with somebody” measured on a five point Likert scale ranging from never to always. The α – values obtained for the direct physical and indirect aggression subscales used in the study were .82 and .75 respectively.

Procedure

While recruiting participants, it was necessary to under - inform them about the true purpose of the study. The study was presented as an investigation of intelligence measured with a computer game in order to alleviate any effects of self-presentation, particularly relevant to the study of aggression (e.g., Björkqvist et al. 1992). The requirement that participants play the computer game first was considered fundamental to the study as the self-report inventories used may alert participants about the true purpose of the study (i.e., aggression), which the knowledge of may have resulted in them altering their playing behaviour on the game. After the conclusion of testing, participants were thanked for their participation and encouraged to take part in a debriefing session outlining the true purpose of the study.

Computer-based assessment methodology:  Prior to actual testing with the Mimics (1997) (i.e., recording scores), participants played through a series of practice trials on the software. The practice trials were necessary to assess whether the participants’ tactical choices were fully informed. The embedded scoring system assessed this by automatically monitoring tactics during the practice trials by using an algorithm, which was incorporated internally within the software as part of the embedded scoring system. The algorithm was tuned to detecting the participants’ use of the space bar for attacking orbs. The vital stage upon which the algorithm was activated occurred at the 8^th diagonal upon which the participants’ avatar was always attacked by an orb. The only way the participants could get through the 8^th diagonal was to attack the orbs, thereby activating the algorithm. The software then assumed that all the tactics in the game had been sampled at least once by virtue of the player getting through to the 8^th diagonal. From then on, formal testing with the game began. The participants played through a set of four games. After completion of the four games, a full record of player movements was obtained.

Results

Variable screening

Prior to any formal testing, condition of the variables was examined using box plots and histograms for 47 participants. The Threatening variable was found to be positively skewed to the point of extreme. This was also found to be the case with Hit₃. Approximately half of the participants had obtained a zero score on these variables. Since too many zero scores on a variable can cause problems with the statistics especially in an extreme case such as this, the Threatening variable was dropped from any further analysis. The problem with the Hit₃ variable was treated in a different way.

Hit₁, Hit_2, and Hit₃ are all measures of aggressive responding in three distinct situations in the game paradigm. In order to overcome the problem with zeros on Hit₃, it was decided to create an aggregated variable by combining all three variables (i.e., Hit _Total). The combination of all three variables amounted to a total count of participant aggressive responses in the game paradigm over a set of four games. Box plots also identified two persistent univariate outliers on the Hit _Total and Hit₁ variables which were also found to be multivariate outliers on the direct and indirect measures, these outliers (i.e., one male and female participant) were dropped from the analysis.

Regression Procedure

In relation to one of the goals of this study, it was decided to formally test for any sex differences by coding sex as a dummy variable (i.e., males = 1 and females = 0), and by creating interaction variables between sex and the constructs of aggression. The analyses examined the relationship of participant scores on self - reported direct physical and indirect aggression to Hit _Total and Hit₁ as predictors of aggressive playing behaviour in the game. This type of analysis offered the opportunity to assess the relationship of self - reported direct physical and indirect aggression to aggressive responding in the Mimics (1997) in unison, as opposed to individual examinations of Hit _Total and Hit₁ to the constructs of aggression.

A hierarchical regression procedure was used. The first model examined the constructs of aggression as main effect predictors followed by a second model that entered sex as a third variable, and then a third model adding the two interaction terms to the model. Analysis was conducted as a set of three models in this order examining the ∆ R²   statistic to determine whether the extra terms significantly contributed to the model. This methodology is advocated by Garson (2006).         

Assumptions of Regression

Recommended by several researchers is a minimum case per ratio of 5 – 10 cases per predictor variable with more acceptable ratios that are considered good, ranging from 10 to 40 participants per predictor variable (c.f., Hair,  Anderson., Tatham, & Black, 1995). The ratios in the following analysis ranged from 22 – 9 cases per predictor variable. Thus, the requirements of the assumption were considered met.

Multicollinearity was assessed by examining the condition index (c.f., Garson, 2006; Tabachnick & Fidel, 2005). A problem with multicollinearity was encountered with the third models described above, which was related to the dummy variables, this is not uncommon (c.f., Garson, 2006; Tabachnick & Fidel, 2005). To overcome the problem, the predictor variables were centred. This procedure is recommended when using dummy variables in an analysis (c.f. Garson, 2006; Tabachnick & Fidel, 2005).

Assumptions concerning linearity, normality (re-examined), and homosdeacity of regression can be examined post analysis by investigating the predicted y-values against the error residuals (Tabachnick & Fidel, 2001). There were no violations of any of the aforementioned assumptions applicable to regression. In addition, the Durbon Watson coefficient was also examined to determine if there was a problem with autocorrelation (Garson, 2006); none was encountered.

Multivariate outliers can also be assessed by examining expected and detrended normality plots of the residuals (Tabachnick & Fidel, 2001). No unusual cases were identified after examining the residuals, which confirmed that our pre-analysis data preparation was successful. This was also the case for univariate outliers where none were identified after examining case-wise diagnostics.

The Regression Analyses and Interpretation of Regression Coefficients

The third model was significant for both the Hit _Total and Hit₁ variables. Table 1 illustrates the pattern of associations. The significant direct physical aggression by sex interaction highlights a positive association of aggressive responding for males with direct physical aggression and a negative association with indirect aggression. The equation for males is also negatively moderated by a significant main effect for sex. In the case of females, due to the dummy variable coding, the interactions and sex main effect drop out of the equation leaving only the significant main effect of indirect aggression, highlighting a positive relationship of the construct to aggressive responding. For all the final models presented, inclusion of the sex variables and associated interaction terms was supported by significant results obtained on the ∆ R² statistic. This indicates that sex differences in self reported direct and indirect aggression was a major factor in determining the relationship between aggressive responding in Mimics (1997) and the aforementioned constructs of aggression.

Table 1

Predicting Playing Behaviour with Measures of Direct Physical and Indirect Aggression

*p < .05. **p < .01.

Note: The sex variable is coded 1 for males and 0 for females.

Discussion

The results of the statistical analysis were in line with the expectations of this pilot study which was to show that the aggressive responses of participants in the computer game described as the hitting of orbs (i.e., a simulated punch) with a player controlled avatar was a reflection of a participants’ previously displayed real world direct physical or indirect aggression. Sex differences on these constructs of aggression and their relationship to aggressive responding in the computer game mirror well documented findings on sex differences in direct and indirect aggression (c.f., Björkqvist et al., 1992a; Björkqvist et al., 1992c; Österman, et al., 1998). Despite the requirement that some reservation must be exercised concerning the small sample of males in the study and size of the sample overall, the results of the analysis would appear strong even for a sample this size. A preferred prerequisite of some researchers (c.f. Fisher, 1970; Milton, 1986; Calvert, 2003) other than arbitrary sample size limitations being applied to regression is to consider the size of the correlation (R in regression) with respect to sample size. This method requires the computation of t-variate scores based on the sample size and correlation (R) obtained. The t-variate score obtained for this analysis was 4.53 which equates to a 1 in 1000 chance of being wrong (one - tailed) or 5 in 10,000 chance of being wrong (two - tailed). The size of R exceeds .5 for calculations with both the Hit_Total and Hit₁ variables with both parameters explaining approximately 25 percent of the variance in scores (i.e., aggressive responding) on the Mimics (1997) based on previously displayed past direct and indirect aggressive behaviour. Cohen and Cohen (1986) have also reported a number of calculations for regression to achieve significant associations of power .8 which are calculated with respect to sample size and the level of significance of the regression model while also taking into account the number of predictor variables. According to the table presented by Cohen and Cohen (ibid.) this particular analysis corresponds to a power of .8. It also has to be pointed out that the way in which the associations presented above mirror past work on sex differences in direct and indirect aggression make it less likely that the results are simply due to chance. As a pilot study designed to iron out problems with methodology, we would simply state that the results seem interesting and encouraging even for a sample this size.

The study would seem to illustrate that the relationship of human to a particular playing strategy in a computer game may be mediated rather than a direct projection supporting a narrative theoretical perspective. The mediated view point appears to be demonstrated as significant associations of both direct physical and indirect aggression to aggressive responding (i.e., simulated punching within the game paradigm). In other words if the association was solely significant with only the direct physical construct of aggression the narratological direct projection perspective would have been supported. However, the results of this analysis seem to show that simulated hitting within the game paradigm may take on dual meanings as demonstrated by the additional significant association of self -reported indirect aggression to simulated hitting within the Mimics (1997) paradigm. Hence, simulated hitting within the computer game does not appear to be a simple synonym for direct physical aggression. This result would appear to support a broader theoretical perspective on playing behaviour and its relationship to personality, not anchored by narrative perspectives on the aggressive features contained within many computer games. These preliminary findings would appear to be more in line with Fisler’s (2006) theory of the Performed or Mediated player. However, a much larger sample size is needed in a future study to illustrate this with more confidence.

It needs to be mentioned that this type of study is not just about using any game; we are of the opinion that a game needs certain characteristics to be an adequate projective assessment of personality. This is highlighted in the methods section, but is worth restating. First, the game needs to offer the participants choice (s). For example, a typical shoot em-up game virtually only requires a participant to shoot, thus comparisons of aggressive responding among participants are meaningless. It needs to contain a paradigm such as conflict versus cooperation, with comparable rewards for successful game completion as is the case in the Mimics (1997). A suitable game may need to offer choices to the participant, thereby yielding projective potential relating to the personality attributes of the individual. Second, the participant’s tactical choice in any game needs to be fully informed. The researcher needs to know that a participant has not gravitated toward a particular strategy because they are unaware of the alternatives, thus the inclusion of a practice trial in this study. Participant experience with the game needs to be adequate but minimal, a participant can not be an expert player that has learned in a sense to beat the game designer, discovering through experience the best way to score the highest points, which may affect the projective value of the game. There are probably several more issues that concern game design and projection but last but not least, the game has to be a game creating tension through scoring systems, alternative strategies, or exploration of the game universe. These characteristics are considered conducive to projecting personality through preferred game play strategies.

Conclusion

The study seems to have illustrated that aggressive responding in a computer game simulating physical aggression may not be a direct projection of physical aggression but rather multifaceted with dual meanings relating to direct physical and indirect aggression strongly influenced by sex differences on these constructs. What may be said with some confidence is that the research does not support a simplistic narratalogical view when sex differences are taken into consideration. The study would also seem to indicate that the direction of the human – to - computer game interaction effect may also be from human – to - computer game rather than solely in the game to human direction as previously described by research in the area. However, the relationship may be a little of both. Further investigation is needed to examine connectivity and division between the digital extension (i.e., Avatar) and the human controlling it, taking into account that the effects of the relationship may be in two directions. This may have implications to our understanding of the impact of media related violence on the individuals exposed to it, at least within the context of computer games.

Acknowledgements

The authors wish to gratefully acknowledge the creators of Mimics (Aidman & Shmelyov, 1997) without this paper would not have been possible and the expertise of Aidman in guiding the research throughout the years 1999- 2002 initially as a part of the University of Ballarat’s Honours program. Particular gratitude is given to Glen Jolozides as the inspiration for this study. Appreciation is also given to Åbo Akademi University for the support given to the first author in the form of a research grant to complete this project.

References

Aarseth, E. (1997). Cyber text: perspective on ergodic literature. Baltimore: John Hopkins.

Aidman, E.V., & Shmelyov A. G. (1997). MIMICS: Interaction simulator and social skills  game [Computer Software]. InterMind: Melbourne, Vic.

Aidman, E.V., & Shmelyov, A.G. (2002). MIMICS: A symbolic conflict/cooperation simulation program, with embedded protocol recording and automatic psychometric assessment. Behaviour Research Methods Instruments & Computers, 24, 83-89.

Baron, R. A. (1997). Human aggression. New York: Plenum Press.

Bartle, R. (1999). Hearts, clubs, diamonds, spades: players who suit muds. Retrieved September 13, 2006, from http//www.mud.co.uk/Richards/hcds.htm.

Björkqvist, K., Lagerspetz., K. M., & Kaukiainen A. (1992a). Do girls manipulate and boys fight? Developmental trends in regard to direct and indirect aggression. Aggressive Behaviour, 18, 117–127.

Björkqvist, K., Lagerspetz, K., & Österman, K. (1992b). The Direct and Indirect Aggression Scales. Vasa, Finland: Åbo Akademi University, Department of Social Sciences.

Björkqvist, K., Österman K., & Kaukiainen A. (1992c). The development of direct and indirect aggressive strategies in males and females. In: K. Björkqvist , P. Niemelä,  (Eds.), Of mice and women: Aspects of female aggression (pp. 51 - 64). San Diego, CA: Academic Press.

Björkqvist, K., Österman, K., & Kaukianen, A. (2000). Social intelligence – empathy =     aggression. Aggression and Violent Behaviour, 5, 191 – 200.

Calvert, J. B. (2003). Linear regression: understanding what goes on when you fit a straight line to data or find a correlation coefficient. Retrieved October 11, 2006, from http//www.du.edu/~jcalvert/econregress.htm.

Campbell, A., Sapochnik, M., & Muncer, S. (1997). Sex differences in aggression: does social representation mediate form of aggression? British Journal of Social Psychology, 36, 161 - 171.

Cohen, J., & Cohen, P. (1983). Applied multiple regression/ correlation analysis for the behavioural sciences. Hillsdale: Lawrence Erlbaum Associates.

Dillon, T. (2006). Computer game theory. Future lab. Retrieved September 05, 2006 from           http//www.futurelab.org.uk/viewpoint/art56.htm.

Egenfeldt-Nielsen, S. (2003). Exploration in computer games. Game studies: The International Journal of Computer Game Research. Retrieved, from http://www.Game-research.Com/art Exploration in Games. ASP.

Ekman, P. (1999). Facial expressions of emotion. In T. Dalgesh & M. J. Power (Eds.), Handbook of cognition and emotion (pp. 301-320). Chichester; U K: Wiley.

Fisher, R. (1970). Statistical methods for research workers. Edinburgh: Oliver and Boyd.

Fisler, B. (2006). Digital extensions: a player theoretical model for the video/computer game. Reconstruction: studies in contemporary literature. Retrieved September 13, 2006, from http://reconstruction. eserver. org/061/Fisler.html.

Frasca, G .(2003). Simulated versus narrative: Introduction to ludology. In M. J. P. Wolf & B. P. Perron (Eds.), The video game theory (pp.221-235). New York: Routledge.

Garson, D. (2006). Multiple regression. Retrieved March 29, 2006, from http://www2.chass.ncsu.edu/garson/pa765/regress.htm

Graham, G., Cook, M., Cohen, M., Phelps, M., & Gerkovich., M. (1985). STAR: A unique embedded performance assessment technique. Behaviour Research Methods, Instruments & Computers, 17, 224-228.

Griffiths, M. (1999). Violent video games and aggression: a review of the literature. Aggressive and Violent Behaviour, 4, 203–212.

Hair., J.,  Anderson, R., Tatham, R., & Black, W. (1995).  Multivariate data analysis with readings. New Jersey: Prentice Hall.

Houser, S., Donovan, T., Houser, D., King, J., & Foreman. G. (1998). Grand theft auto. Rockstar Games: PC CD-ROM.

Küchlich, J. (2006). Perspectives of computer game philology. Game studies: the international journal of computer game researchers. Retrieved September 05, 2006, from http://www.gamestudies.org/0301/Küchlich.

Juul, J . (2000). What computer games can and can’t do. Paper presented at the 2000 Digital arts and culture conference. Retrieved September 5, 2006, from http://www.jesperjuul. net/textwcgcad.html

Lagerpetz, K. M. J., Björkqvist, K., & Peltonen, T. (1988). Is indirect aggression typical of females? Gender differences in aggressiveness in 11-12-year old children. Aggressive Behaviour, 14, 303-315.

Mier, S. (1991). Civilization. Microprose Software: PC CD-ROM.

Milton, S. (1986). Sample size formula for multiple regression studies. The public opinion quarterly, 50,      112-118.

Österman, K., Björkqvist, K., Lagerspetz, K. M. J,, Kaukiainen A., Landau SF., Fraczek A., & Caprara, G.V. (1998). Cross-cultural evidence of female indirect aggression. Aggressive Behaviour, 24, 1-8.

Tabachnick, B. G., & Fidell, L. S. (2001). Using multivariate statistics. Boston: Ally and Bacon.