26 November 2008 Measurement of Aggression in Laboratory Settings Aimée E. M. Capello Utrecht University Abstract This article represents an evaluation of the construct validity of contemporary laboratory aggression research. The notion of severe limitations in major aggression paradigms was raised by Tedeschi and Quigley (1996) and largely disputed by Giancola and Chermack (1998). Several important recommendations were proposed, and presently, the extent to which these have been taken into account was evaluated. The two major paradigms, the Taylor Competitive Reaction Time Task and the Point Subtraction Aggression Paradigm, still partly suffer from the same limitations. While theory development has occurred in which attention to the subjective motivation and intention has increased, this has not yet been fully adopted in practical laboratory assessment. Still too often, subjective goals for variations in response patterns are ignored, and limited response alternatives are provided, as a result of which laboratory measures still under represent the construct of aggression. Moreover, empirical evidence is poor, and it is unclear why these recommendations have been ignored. Fortunately, meaningful attempts have been made, and sometimes followed by other researchers. There is a development to better operationalisations, but overall, the main conclusion as first stated by Tedeschi and Quigley remains largely unaltered: contemporary laboratory aggression research lacks construct validity. “Research without theory is blind” - Immanuel Kant- 2 Contents INTRODUCTION 5 CHAPTER ONE AGGRESSION 7 1.1 Aggression 7 1. 1. 1 Aggression is… 7 1. 1. 2 Aggression is not… 8 1.2 Classification and modelling of aggressive behaviours 9 1. 2. 1 Hostile and Instrumental aggression 9 1. 2. 2 A new taxonomic system 9 1. 2. 3 A social-cognitive model of aggression: the GAM 10 1.3 Concluding Remarks 10 CHAPTER TWO TRADITIONAL PARADIGMS AND THE VALIDITY DEBATE 12 2.1 Traditional Paradigms 12 2.1.1 Teacher-Learner and Essay Evaluation Paradigms 12 2.1.2 The Taylor Competitive Reaction Time Task 13 2.1.3 Point Subtraction Aggression Paradigm 13 2.2 Validity Threats 14 2.2.1. Theoretical concerns 14 2.3.2. Non-aggressive response options 15 2.3.3. Social interaction 16 2.3 Empirical evidence of construct validity 16 2.3.1 Convergent validity 17 2.3.2 Discriminant validity 17 2.3.3 Indirect evidence 17 2.4 Concluding Remarks 18 3 CHAPTER THREE CONTEMPORARY LABORATORY AGGRESSION RESEARCH 19 3. 1. Modifications to existing paradigms 19 3. 1. 1 The Taylor Competitive Reaction Time Task 19 3. 1. 2 The Point Subtraction Aggression Paradigm 19 3. 2 Other Viewpoints 20 3. 2. 1 The Hot-Sauce paradigm 20 3. 2. 2 Competitive Spatial Tracking Task 21 3. 2. 3 Bungled Procedure Paradigm 22 3. 2. 4 Concluding remarks 22 3. 3. Limitations 22 3. 3. 1. Theoretical concerns 23 3. 3. 2. Non-aggressive response options 24 3. 3. 3. Social interaction 26 3.4 Concluding Remarks 26 CHAPTER FOUR RECENT EMPIRICAL EVIDENCE FOR CONSTRUCT VALIDITY 27 4. 1 Convergent validity 27 4. 2 Discriminant validity 28 4. 3 Indirect evidence 28 4. 4 Related issues 28 4. 5 Concluding Remarks 29 GENERAL CONCLUSION 31 REFERENCES 33 Attachment 1 38 4 INTRODUCTION The aim of the present paper is to discuss and evaluate the construct validity of contemporary laboratory aggression research. Approximately ten years ago, the validity of aggression paradigms was heavily debated by Tedeschi and Quigley (1996; 2000), and Giancola and Chermack (1998). According to Tedeschi and Quigley, most of research on physical aggressive behaviour in laboratory settings depended on a limited set of paradigms lacking construct validity. Giancola and Chermack replied by argueing that some paradigms were sufficiently adequate for measuring a construct as complex as aggression, but agreed that some improvements were necessary. Tedeschi and Quigley (1996; 2000) present with some important recommendations for advancing aggression research. Therefore it was considered necessary to re-evaluate these issues and examine adequacy of current laboratory measures of aggressive behaviour. Some important limitations have been brought forward by Tedeschi and Quigley (1996; 2000). They propose that traditional paradigms are aimed at “simple attack-retaliation situations” (Tedeschi & Quigley , 1996, p. 163) while efforts should be aimed at examining subjective motivations and intentions to be able to judge whether adherence to the definition of aggression has successfully been achieved. They present with several significant recommendations. As opponents of construct validity of laboratory aggression research, Giancola and Chermack (1998) aim to demonstrate that some of these criticisms are unwarranted. They agree with the recommendations as suggested by Tedeschi and Quigley (1996; 2000), but disagree with their statement that “these measures are severely limited in their generalisability to real world aggression” and consider it to be a “gross overstatement”, and remind that laboratory measurements are an “invaluable component of a multimethod approach”. Finally, in a reply by Tedeschi and Quigley (2000) these authors argue the Giancola and Chermack (1998) paper, in their view, has not adequately contributed to supporting validity on both conceptual as well as methodological grounds. Currently, the validity of research paradigms is still under investigation (Ferguson, 2007; Ritter & 2005; Giancola & Parrott, 2008). Two paradigms in particular are widely used and many inferences about aggressive behaviour have been based upon their findings. Given that aggression voorkomt in zowel de normale populatie als verschillende, meerdere klinische populatie, it is important this behaviour be systematically and, reliably, studied. To advance scientific knowledge on the processes and factors involved in physical aggressive acts, it is paramount that validity of measures be assessed and positively confirmed. The aim of the present is to examine to what extent these methodological and theoretical issues have been taken into consideration in current investigations of controlled aggressive responding. Objectives 5 are to examine to what extent traditional paradigms have been modified, whether new paradigms have been developed, and whether their validity has been assessed. The present article will be divided into three chapters. Chapter one will present the reader with a brief discussion of researchers’ attempts to delineate what aggression exactly is. Chapter two will describe the more traditional paradigms used to measure aggressive behaviour in laboratory settings and will provide with a summary of the validity debate between Tedeschi and Quigley (1996; 2000), and Giancola and Chermack (1998). Finally, in chapter three, contemporary laboratory research into physical aggression will be discussed. Focus will be on modifications of traditional paradigms and more recently developed laboratory designs. Empirical evidence for construct validity will be discussed. 6 CHAPTER ONE AGGRESSION 1.1 AGGRESSION Before we can re-examine laboratory measurement procedures in aggression research, we need to have a clear understanding of what it is that makes that a certain act or behaviour is perceived as aggressive. Consider for instance, a farmer chopping the head off a chicken. Or a boxer who knocks out his opponent. Situations like this are generally not conceived aggressive. But what about a shopkeeper who shoots a man stealing his takings? Many difficulties arise when trying to explicitly state what aggression entails, and also, what it does not entail. Or, as Parrott & Giancola (2007) put it, “[i]ndeed, the existence of so many different conceptualizations of aggression reflects the struggle by researchers to define this construct.” (p. 282). The difficulties of defining aggression can be traced to the fact that there are many different kinds of aggressive behaviour, and the manner and extent of aggression that will be exhibited depends on a multitude of factors which all need to be considered when attempting to explain aggressive behaviour. 1. 1. 1 Aggression is... Several definitions and variations thereof have been offered to explain aggressive acts. One influencing definition of aggression came from Buss (1961). He defined an aggressive act as “a response that delivers noxious stimuli to another organism” (pg. 1). Other definitions that have been put forward are “behaviour aimed at causing either physical or psychological pain” (Aronson et al., 1997, p.324) and “the intent to harm or otherwise injure another person, an implication inferred from events preceding or following that act of aggression” (Kaplan & Sadock, 2003). As has often been noted (e.g. Parrott & Giancola, 2007) components such as the intent of the actor and the motivation of target are important to include in the conceptualisation. That is, harming another living being without the intent to do so is not considered aggressive. Similarly, situations in which the target consents with the harm-doing (e.g. inside the boxing-ring) are not considered aggressive either. The above mentioned definitions are limited in that they neglect either one or both of these components. 7 In line with these issues, Baron (1977) and more recently Baron and Richardson (1994) extended the definition to include the state of the target, and described the concept as “any form of behaviour directed toward the goal of harming or injuring another living being who is motivated to avoid such treatment (Baron & Richardson, 1994; pg. 7). A more recent modification ws suggested by defining aggression as “any behaviour directed toward another individual that is carried out with the proximate intent to cause harm. In addition, the perpetrator must believe that the behaviour will harm the target, and that the target is motivated to avoid the behaviour” (Bushman & Anderson, 2001; Anderson & Bushman, 2002; Tedeschi & Felson, 1995). In this definition, a distinction is made between proximate, immediate goals and ultimate or primary goals. The important difference here is that intent is necesary only as a proximate goal, and that different types of aggression emerge only at the ultimate level. For instance, this way, the shopkeeper who shoots at the man stealing his takings is performing an aggressive act. The proximate intent or immediate goal is to harm the target, but the ultimate or primary goal is to defend oneself and his belongings. As this example illustrates, it is a comprehensive definition. Contemporary research might not yet have fully adopted these recent definitions, but in reviewing recent laboratory aggression research, where applicable, this definition will be kept in mind. 1. 1. 2 Agression is not... There are several constructs that are more or less related to aggression. Violence can be defined as a destructive act against a person or object (www.northern.ac.uk, 2003) and is one of the ways in which aggression is expressed. This form of aggression is called physical aggression and is one of the most widely investigated form of aggressive behaviour. Hostility is an attitudinal/cognitive construct comprised of enduring cognitions (Parrott & Giancola, 2007). Hostility does not need to be expressed for it to be there, it is an underlying tendency, while aggression is a form of behavioural expression. However, expressed hostility is probably similar to aggression (Parrott & Giancola, 2007). Anger is said to be different from aggression for several reasons (cf. Parrot & Giancola, 2007). First, it is an experiential, emotional state whereas aggression is an expressive, behavioural process. Second, anger is not necessary for aggression to occur (e.g. instrumental aggression) nor is anger necessarily expressed by aggression. Finally, unlike the expression of aggression, the experience of anger does not require an explicit target (Berkowitz, 1993). In an attempt to better grasp the complete array of behaviours that can be considered agressive, researchers have offered numerous ways of classifiying, and modelling, aggression and aggressive acts. To these we will turn next. 8 1.2 CLASSIFICATION AND MODELLING OF AGGRESSIVE BEHAVIOURS 1. 2. 1 Hostile and Instrumental aggression Numerous subtypes have been put forth to classify aggressive acts. However, the most frequently mentioned, adopted, and well-known dichotomous conceptualisation of aggression, is the distinction between hostile and instrumental aggression (Parrot & Giancola, 2007). Instrumental aggression is the expression of aggression as a means to an end rather than as an act of retribution or defence and is expressed to obtain some objective (power, money, sexual gratification, or some other goal) other than inflicting injury on the victim (Anderson & Bushman, 2002; Parrot & Giancola, 2007; Berkowitz 1993). Key features of instrumental aggression are goal-directedness and planning/preparation. It usually involves little or no provocation and is driven by goals, not emotions; emotional arousal (e.g. anger, hostility) is relatively low or is secondary to the act (Cornell et al., 1996). Key feature of hostile aggression is that is is driven by emotional motivation instigated by provocation (Berkowitz, 1993; Cornell et al., 1996; Anderson & Bushman, 2002). Other frequently used terms are affective, impulsive, reactive, retaliatory (Parrott & Giancola, 2007). While the idea of clear difference between hostile/reactive motivated and instrumentally motivated aggression is supported by forensic evidence (McEllistrem, 2004) and was recently demonstrated to contribute to a clearer understanding of the motivations for aggressive behavior (Polman, 2008), the functionality of the distinction has been argued (Bushman & Anderson, 2001). The distinction is not considered elaborate enough to capture the whole of aggression and fails to consider aggressive behaviour that is motivated by multiple goals. It is reminiscent of the distinction between controlled (i.e. planned, goal-directed instrumental aggression) and automatic (i.e. impulsive, reactive, unplanned) processing, which is conceived of as being a continuum, and the Hostile/Instrumental distinction should be perceived similar (Cornell et al., 1996; Bushman & Anderson, 2001) . That is, hostile aggressive actions are not always completely unplanned or reactive, and not all instrumental acts are solely goal-directed and unaffected by emotion or provocation. Support for this notion comes from a study suggesting that approximately half of the aggressive incidents cannot be categorized as hostile or instrumental (Barratt & Slaughter, 1998). 1. 2. 2. A New Taxonomic System In replacement of the Hostile/Instrumental distinction, Parrott and Giancola (2007) present a new taxonomic system based on two levels of expression and several distinct subtypes. The authors argued 9 that the operational definition of aggressive behaviour still lacks adequate conceptual transparency and simplicity, causing variations in measurement tools of aggression, and ineffectiveness of aggression research. They present with a system to describe and classify aggression along two continuums or axis of expression, specifically, active versus passive and direct versus indirect. Five subtypes of aggression (physical, verbal, postural, damage to property and theft) can exist anywhere along these dimensions. The advantage of te taxonomy is that the different types of aggression that can be classified are a ten-fold of possibilities with the more traditional distinction of hostile and instrumental aggression. 1. 2. 3 A Social Cognitive Model of Aggression: the GAM Bushman and Anderson (2002) and Joireman et al. (2003) propose a mediational, social-cognitive model of aggression, the General Aggression Model. In this framework, personality is defined as a collection of stable knowledge structures by which the social world is interpreted and behaviour is guided. These knowledge structures influence processes of perception, decision-making, and action. Briefly put, the emphasis is on person and situation inputs; heavily interconnected affective, cognitive and arousal routes; and the outcomes of information processes (appraisal and decision). These informational processes are placed on a continuum from automatic to controlled and they are expressed in actions ranging from thoughtful to impulsive action (Anderson & Bushman, 2002, p. 34). In a feedback loop, the final (behavioural) outcomes become part of the person and situation inputs for the next episode. Berkowitz’s (1993) theory had a major influence on aggression research, and at least some of the limitations in aggression measurement originate form this theory (Tedeschi & Quigley, 1996; 2000). Has laboratory research adapted to modifications of definition and accompanying theory? 1.3 CONCLUDING REMARKS Aggression is a complex construct because there are various different motives for resorting to aggression and several different expression modes, together creating numerous subtypes of aggressive behaviour. Developments in past years have lead to a more comprehensible definition of human aggression, better corresponding to the depth of detail of constructs of human behaviour. In the literature, aggression has been classified in several ways. One frequently adopted distinction is between hostile and instrumental aggression. However, its contributive value has been argued, and other classification systems that capture a wider range of aggressive exhibitions have been put forward. One such system is a new 10 taxonomy which is more transparent and includes more dimensions and is therefore argued to possess superior capability to grasp the complexity involved in human aggressive behaviour. Most of aggression research is based on assumptions coming from Berkowitz’ neo-cognitivists’ theory of aggression. Adjusted models have been developed, one such model that seems promising is the General Affective Model. Interesting now is to consider to what extent these changes have positively affected or influenced practical execution of laboratory measurement. 11 CHAPTER TWO TRADITIONAL PARADIGMS AND THE VALIDITY DEBATE Before we can evaluate changes in laboratory aggression measurement in the past years, the traditional paradigms and their limitations will be reviewed. This chapter will first present a brief description of the traditional paradigms that were the subject of the Tedeschi and Quigley (1996; 2000) and Giancola and Chermack (1998) validity discussion. Next, the main critique following the debate will be summarised. And lastly, the empirical evidence supporting validity introduced by the authors will briefly be critically evaluated. 2.1 TRADITIONAL PARADIGMS 2.1.1 Teacher-Learner (TLP) and Essay Evaluation Paradigms (EEP) Teacher-Learner Paradigm (TLP) One well-known measure of physical aggression was the Buss Aggression Machine, or the Teacher/Learner paradigm (Buss, 1961; cf. Giancola & Parrot, 2008). Subjects are told that “the research is concerned with the effects of punishment on learning” and that they are to engage in a learning task in which they have the role of teachers. The learner, a confederate, has to learn a list of word pairs. In the test phase, when the ‘teacher’ calls the first word, the learner has to reproduce the second. Both are in different rooms and communicate via intercoms. Correct responses are reinforced verbally, while incorrect responses are punished by means of delivering shocks with the ‘aggression machine’ in front of them to which learner is connected by a finger electrode. Dependant variables are shock intensity, duration and sometimes frequency. Intensity can be varied depending on which button is pressed (pressing 1 delivers the lowest intensity, pressing 10 the highest). Duration is manipulated by how long the button is held down. Sometimes subjects are given a shock of intermediate intensity so they can experience its unpleasantness. In reality, the learner never receives shocks. Essay Evaluation Paradigm (EEP) The EEP (Berkowitz, 1962) has been developed as an improved measurement tool relative to the TLP. In this protocol, subjects are informed that they are to evaluate another person’s (a confederate) written product by delivering shocks (Tedeschi &Quigley, 1996). The subjects will also be evaluated themselves. After the short essays have been written, the evaluation begins. The subjects are provoked by having the confederate evaluate their essay first. Physical aggression is operationalized as shock frequency, ranging from 1 to 10 (max), and the confederate 12 gives the subject on average one to seven shocks. Typically, subjects evaluate the confederate’s essay similar and subjects deliver a similar amount of shocks (Tedeschi &Quigley, 1996). The main criticism for both has been that, due to the cover stories of evaluating and teaching/helping, behaviour might be driven by altruistic motives, and not aggression (Tedeschi &Quigley, 1996). Other criticisms will be dealt with in a later section. Both parties in the debate agree that the following two paradigms are better measures and are more widely used and accepted (Tedeschi & Quigley, 1996; 2000; Giancola & Chermak, 1998). 2.1.2 The Taylor Competitive Reaction Time Task (TCRTT) In the original TCRTT, or Taylor Aggression Paradigm (Taylor, 1967), shock electrodes are attached to the wrist and care is taken to make the subjects believe there actually is an opponent in a nearby room. Subjects are told they are engaging in a competitive reaction time game. For each participant, the threshold is determined. At the beginning of each trial they are instructed to select (by pressing 1 of 10 buttons) any of the 10 intensities of shock they wish their opponent to receive. They are furthermore told that whoever loses the game receives the shock at the preset intensity, at the end of each trial. The intensity set by the ‘opponent’ gradually increases as the game progresses, providing different levels of provocation. The TCRTT is credited for not using a cover story (Tedeschi &Quigley, 1996; Giancola & Chermak, 1998). Although they are not exactly told what is being measured, this is quite normal for laboratory paradigms and generally not considered to be an issue (Giancola & Chermak, 1998). At least subjects are told that they are in competition with one another, and not that they are helping someone while in reality, what supposedly is being measured is a rather negative way of interacting. 2.1.3 Point Subtraction Aggression Paradigm (PSAP) The PSAP uses point and money subtraction as aversive stimulus. This paradigm was developed by Cherek (1967) and has been revised more recently (Cherek et al. 1992). Typically, subjects are told they are paired with an opponent (situated in another room) and that the aim is to earn money. There are three response options. Pressing ‘button A’ a 100 times is coupled with monetary reinforcement (usually 10 cents). Pressing ‘button B’, the point subtraction (PS) button, subtracts 10 cents from the opponent when pressed 10 times. And finally, ‘button C’ is typically an escape button, initiating an interval in which the participant is protected from monetary subtraction. Physical aggression is operationalised as number of times the point subtraction button was pressed. In the Cherek et al. (1971) study, all subjects indicated subtraction of money was experienced as very aversive. 13 Provocation is initiated by periodic subtraction of 10 cents, which was attributed to the other person. In several studies, aggressive responding (B presses) after provocation are reinforced by initiation of a protection interval. Whereas the original PSAP lasts three hours, more recent versions take about 30 minutes to complete (cf. Golomb et al., 2007). The PSAP is said to possess several strengths. Unlike in the TCRTT, subjects can choose not too subtract points, whereas in other paradigms, subjects are required to use noxious stimuli (e.g. Giancola & Chermak, 1998; Cherek et al., 2000). Second, the frequency of aggressive responses can be varied and aggressive responding can be initiated at any time (Cherek et al., 2000). And third, use of subtraction as aversive stimulus allows repetitive testing as well as testing in situations where shocks cannot be used, such as when testing children (Cherek et al., 2000). 2.2 VALIDITY THREATS Briefly defined, construct validity refers to the degree to which a particular measure accounts for the construct that it is supposed to measure. The following limitations have been pointed out by Tedeschi and Quigley (1996; 2000) as threatening the construct validity of laboratory aggression measurement. 2.2.1. Theoretical concerns Parrott & Giancola (2007) correctly note how absence of consensus regarding definition affects laboratory aggression measurement due to great variability in operationalisations of the construct (Parrot & Giancola, 2007). Clearly, paramount is for the experimental designs and its measures to correspond to the definition of aggression. This matter has also been carefully discussed by Tedeschi and Quigley (1996; 2000) and they note that only a small set of paradigms is responsible for a bulk of data obtainded by laboratory measurement of aggression, stressing the importance of determining validity of these measures. In their view, none of the traditional paradigms adequately assess the construct of aggression. According to Giancola and Chermak (1998), in the TCRTT (and its modified version) and the PSAP, physical aggression is operationalised in line with the Baron and Richardson (1994) definition. In their view, when subjects deliver a shock to the opponent, this can only be interpreted as intenteds to harm. They furthermore assume that it is clear, given the competitive nature of the task, that the opponents are motivated to avoid such harm, since point subtraction is harmful and threatening to one’s ego as well as financially. Although this is likely, it is presumptuous to say that this is ‘clearly’ so without consistently questioning the subjects themselves. Certainly not all subjects will feel experience the situation as a threat to their selfesteem, but there is no way to tell upfront, or adhoc, who will and who will not. Therefore, no statements such as these can be made without 14 sufficient evidence. Stimulus noxiousness’ might also influence their “aggressive” responding to the opponent. That is, subjects might set low intensities with the intent of persuading the opponent to conform to their strategy. Then, there is no intent to harm, but only motivation to avoid. If the intention of an actor is the basis for defining the concept of aggression, then measures of this intention need to be incorporated in laboratory experiments. For example, if subjects are told they are engaging in a competitive game, how can we be sure it the intent of their responses is to do harm and not merely to compete? What’s more, apparently, no correlation exists between competition and TCRTT responses, a finding used by Giancola and Chermack (1998) to support validity. However, it proves that subjects’ interpretation of the situation cannot assumed to be what experts tell them it should be. 2.2.2. Non-aggressive response options Tedeschi and Quigley (1996; 2000) furthermore note the absence of non-aggressive response options, and the inability of the paradigms to measure choice behaviour. The instruction for playing the TCRTT involves that participants are required to administer shocks to the loser of the game. This is the only manner for participants to respond to provocation and/or the only manner in which they can interact with the opponent. Choosing the lowest intensity has been argued to be a valid non-aggressive option because this intensity allegedly is not harmful (Giancola & Chermack, 1998). This was supported by limited evidence of verbal reports, indicating that when wanting to perform a nonaggressive act, participants chose the lowest intensity. Still, this does not prove that chosing the lowest intensity is always meant to be non-aggressive. For instance, it might be out of expectations of equal retaliation. Moreover, when response options are limited, and administration of shocks is required, behaviour might become influenced by (unintentional) cues from the experimenter suggesting what is expected from the participant. If there is only one way to respond, subjects easily grasp what is expected from them, regardless of whether they have the same interpretation of the situation. Some have suggested the PSAP is a better measure than the TCRTT because the PSAP is said to offer a nonaggressive response option (Cherek et al., 2000; Giancola & Chermak, 1998). Non-aggressive responding on the PSAP is operationalised as collection of points for oneself. Yet, as Tedeschi and Quigley (1996; 2000) convincingly point out, behaving “aggressively” on the PSAP is less effortful than behaving non-aggressively. It requires 100 button presses to collect one point, but only ten to subtract one. In addition, the only way for the participants to interact is by using the point subtraction button (Tedeschi & Quigley, 2000), and therefore, motives for showing ‘aggression’ might be different than the intent to hurt someone. Data of experiments where subjects do have alternatives options is surprisingly scarce. Tedeschi and Quigley (1996) refer to one study in which participants could respond with either punishment or 15 reward. Women frequently chose to reward a confederate who had insulted them, instead of choosing to punish them, suggesting this certainly is important to consider when reviewing current research. 2.2.3. Social interaction An important component of aggressive behaviour and bodily harm is social interaction. In real life, aggression is usually expressed towards someone in close proximity. Many incidents are even expressed against persons who are not even involved in the instigation for the aggressive feelings, simply because these victims are within reach. In all discussed paradigms, subjects and confederates are in different rooms and are not even able to see each other. Tedeschi and Quigley (1996) point out that when the noxious stimulation is administered at a distance, the generalisability of the findings is rather questionable. In support, they refer to a study by Milgram (1963) which demonstrates that “when physical contact is required to administer the ‘harm’, participants are quite reluctant to do so.” (Tedeschi & Quigley, 1996, p. 168). Then again, it is not surprising to find that a sample of the general population is not eager to do harm to another individual, that is why laboratory indicators of behaviour should not be made too explicit. Individual differences will probably not emerge under such extreme circumstances as in the Milgram experiment, so the necessity of increasing proximity and social interaction remains unclear. And of course, ethical considerations also complicate the issue. 2.3 EMPIRICAL EVIDENCE OF CONSTRUCT VALIDITY In this section we will briefly discuss the strength of some of the evidence brought forward by Giancola and Chermack (1998) in response to the critical comments expressed by Tedeschi and Quigley (1996; 2000). Generally, construct validity is assessed in two ways. Convergent validity refers to the degree to which an operation is similar to other operations that it theoretically should also be similar to. For example, to show the convergent validity of an aggression paradigm, responses on the experimental test should be highly correlated with other measures of aggression, such as questionnaires or other experimental designs. Discriminant validity is the degree to which an operation is distinct from other operations that it theoretically should not be similar to (Campbell & Fiske, 1959; www.wikipedia.org, 2008). 16 2.3.1 Convergent validity Any measure that assesses aggressive behaviour is assumed to correlate with related constructs such as violence, hostility and to correlate modestly with anger. With respect to the TCRTT, three of the four studies cited in favour of the TCRTT used the BussDurkee Hostility Inventory (BDHI). However, in a recent article, Parrot & Giancola (2007) convincingly argue the BDHI and its subscales are invalid measures of aggression, and that their meaning or value are unclear. No evidence was introduced to examine the relationship between different laboratory aggression measurements, which should be expected to correlate at least moderately. 2.3.2 Discriminant validitty To demonstrate discriminant validity of the TCRTT Giancola and Chermack (1998) refer to a study in which TCRTT responses did not correlate with competition (Bernstein et al 1987), among others. Although this might be expected considering that the dependant measure, selected shock intensity, is supposed to be uncorrelated to performance on the reaction time task. Still, Tedeschi and Quigley (2000) suggest this could also imply subjects are not following instructions (Tedeschi & Quigley, 2000). Clearly, the exact relationship between competition and ‘aggressive’ responses on laboratory tasks needs to be further investigated. There was no mention of studies examining discriminant validity for the PSAP, and this absence was not discussed. 2.3.3 Indirect evidence If a paradigm possesses construct validity, it should be able to differentiate between high and low aggressive individuals (Giancola & Chermack, 1998). There is some preliminary evidence for the potential of classic paradigms to discriminate between groups (cf. Giancola & Chermack, 1998), more recent evidence will be discussed in the following chapter. Giancola and Chermack (1998) argue that additional validity is gained by gathering indirect evidence demonstrating the theorised influence of assumed facilitators or inhibitors of aggression. However, it seems premature to gather indirect evidence, and so its value remains questionable. if the convergent and discriminant validity are still debated, 17 2.4 CONCLUDING REMARKS In summary, Tedeschi and Quigley (1996; 2000) argue convincingly that the major paradigms lack construct validity and have serious weaknesses. Of the traditional paradigms, the Teacher/Learner and the Essay Evaluation paradigms were the first two frequently adopted paradigms. Nonetheless, they share several weaknesses, one of which is the use of cover stories. This is particularly problematic for aggression measurement because of the importance of actor intentions in defining the construct. The Taylor Competitive Reaction Time Task and Point Subtraction Aggression Paradigm are considered less limited for measurement of aggression and have become the most frequently used methods for studying aggression in a controlled environment. All paradigms have been criticised for lacking construct validity and adherence to the definition because motivation and intentions are not measured. On the contrary, proponents of validity of laboratory measures argue this is an overstatement because the stimuli are aversive, and subjects are motivated to avoid them. But, these assumptions are not empirically supported. A second major limitation that was debated, also relevant to all four paradigm,s concerns the limited response alternatives that are available. Another strong argument in favour of the arguments of Tedeschi and Quigley (1996; 2000) is that there appears to be a serious lack in empirical evidence in support of construct validity for each of the paradigms. Moreover, most of the evidence that is referred to is either indirect evidence or obtained using invalid measures. Opinions on construct validity may be divided, all parties agree that some limitations need to be accounted for if we wish to meaningfully understand the concept of aggression. This lack of correspondence between theory and practice can be dealt with from two different approaches. One is to adjust the theory in such a way that it is possible for laboratory operationalisations to better adhere to the conceptualisation of the construct. This has been done, or attempted, as a modified definition as well as a more comprehensive social cognition theory have been offered. Another is to adjust the measurement tool so that these operationalisations fit the theory appropriately. Therefore, we will now discuss more recent laboratory measurement and examine the extent to which the criticisms have been incorporated and adjustments have been made or new paradigms have been developed. 18 CHAPTER THREE CONTEMPORARY LABORATORY AGGRESSION RESEARCH The extent to which the major limitations have been taken into account in theorising and methodology of aggressive behaviour will be evaluated here. What adjustments have been made to traditional measures and what new ideas have been brought forward? Following Bushman and Anderson (2001), in this chapter intent and motive will be referred to as proximate (or immediate) and ultimate (or primary) goals (see Tedeschi and Felson, 1995, for similar terms). 3. 1. MODIFICATIONS TO EXISTING PARADIGMS Here we will consider modifications of the Point Subtraction Aggression Paradigm and the Taylor Competitive Reaction Time Task in recent investigations. 3. 1. 1 The Taylor Competitive Reaction Time Task Anderson & Murphy (2003) presented with a significant addition directly relating to the main point of critique of Tedeschi and Quigley (1996; 2000). To assess subject’s goals, these researchers developed a short six item questionnaire (see attachment), consisting of two items to index instrumental motives, and four to assess hostile motivation (the ultimate goals). Importantly, this procedure has been adopted by other researchers (e.g. Anderson et al., 2008; Arriaga et al., 2008). Additionally, several studies modified the TCRTT to include a non-aggressive response option. Zeichner et al. (2003) explicitly mention the validity discussion and specifically aimed to develop a better tool by offering additional response options in their Response Choice Aggression Paradigm (RCAP; Zeichner et al., 1999). Subjects could choose between administering a shock after winning, losing, or not at all, better approximating real-life situations. Similarly, Reidy et al. (2008) modified the TCRTT and delivery of shocks was made optional; intensities were not required to be set at the beginning of each trial, nor were aggressive responses coupled to losing or winning. Subjects could choose to simply continue playing, instead of interfering with the opponents attempts to win, at the expense of their own gains. Finally, in several studies in which aggressive responding was operationalised by a loud auditory stimulus, non-aggressive responding was made optional by having subjects choose between intensities (volumes) ranging from zero to ten (e.g. Anderson & Murphy, 2003; Anderson et al., 2004; 2008; Arriaga et al., 2008). This is a clear improvement to the traditional 19 CRTT where the lowest intensity (1) is considered “clearly” non-aggressive (Giancola & Chermack, 1998). Several minor adjustments have been made as well. Aiming to increase the social interactive component of aggressive behaviour, Lotze et al (2007) made it possible for participants to view their opponents reaction when receiving the noxious stimulation (aversive pressure on the finger by means of compressed air). In other cases, additional dependant measures have been added, for instance, Zeichner et al. (2003) added measures of proportion of trials in which highest intensity was chosen, and unprovoked aggression (intensity and duration of the first shock, and the number of trials before the first shock was administered). 3. 1. 2 Point Subtraction Aggression Paradigm With respect to the availability of alternive response options and the incorporation of subjective interpretations, no consistent, delibarate adjustments have been made, suggesting the criticism was not taken seriously (e. g. Cherek et al., 2000; Tcheremissine et al., 2005; Bailly & King, 2006; Nouvion et al., 2007; Carré & McCormick, 2008; Lieving et al., 2008). 3. 2 OTHER VIEWPOINTS 3. 2. 1 The Hot-Sauce paradigm (HSP) Lieberman et al. (1999) developed a paradigm in which administration of hot sauce was considered an effective measure for physical aggression. Provocation was induced by either one of two possibilities (cf. McGregor et al., 1998). In some cases, subjects were instructed to read an essay supposedly written by another participant. Depending on the condition to which they were assigned, the content of this essay would either strongly agree or strongly disagree with the participants own cultural worldview. In other cases, subjects had to consume a sample of either a neutral or noxious juice which they were told was prepared for them by another participant. After the provocation manipulation, subjects are given a chance to respond to the virtual participant by determining the amount of very spicy hot sauce the other has to consume. They are instructed to prepare (in the context of a food tasting experiment) a plate with three chips and adequate (spicy) sauce. This paradigm has several interesting advantages. Importantly, subjects were aware that the other did not like spicy food and would have to consume the whole sample, so they knew their actions could result in actual physical harm in another person, geeft minder afwijking van definitie; betere 20 operationalisatie. Also, the paradigm possesses ecological validity because incidents with administering noxious food also take place in real-life, for instance, in child-abuse cases (Lieberman et al., 1999). Finally, the issue of requiring interaction is overcome by characteristics of the aversive stimulus. It is a potential means of administering physical harm but for this it is not necessary for those involved to be in close proximity of one another. 3. 2. 2 The Competitive Spatial Tracking Task (CSTT) Recently, the so-called Competitive Spatial Tracking Time Task (CSTT) was developed by J. Savage (in press). In developing this paradigm, care is taken to incorporate the limitations characterising earlier attempts to capture laboratory evoked aggression. In this task, subjects are told they are engaging in a competitive reaction time game with another opponent, which is unknown to them and in fact non-existent. Subjects are instructed to track a small dot across the screen by using the mouse pointer. Whoever does this fastest is the winner of the game and will be rewarded with extra points. They are furthermore told there is an option to annoy the opponent by pressing on a button on the screen by which they can deliver pressure to them. They themselves can also expect to receive “pressure” of different intensities. The pressure is only delivered at the end of each trials, and only when a trial was lost.. They are told the opponent received similar instructions. The participant’s finger is then attached in a box. Pressure is delivered, or received, through a small metal plate pressing down on the subjects’ finger. Before the start of the experiment, the participant’s maximum threshold is set. They are also told they will be able to view their opponent’s responses to the pressure via a web cam connection. In reality, these are pre-recorded images, and the same for all participants. This paradigm possesses several major improvements. Perhaps most important is that subjects are inquired about their motives and intentions at the conclusion of the experiment. To assess hostile and instrumental motives, subjects are asked whether they used the option to annoy the opponent, and if so, what their motives were. A second major improvement is that participants have a choice in how to interact and can also choose to compete and continue playing the game, trying to outperform the opponent. Additionally, while in the PSAP it might be more convenient to respond aggressively as opposed to non-aggressively, in this task it is less convenient. Pressure can only be adminstered at the expense of the likelihood of winning, as its necessary to move away from the competitive task in order to set the intensity of the pressure to be administered. Moreover, because the pressure is delivered at the end of each trial, it does not interfere with the opponents’ game. The aversive stimulation clearly does not directly interfere with the opponent’s peformance and merely annoys or irritates, a fine operationalisation of aggressive behaviour. And finally, frequency and initiation of aggressive responding are free to vary, and can be initiated anytime, unlike in TCRTT-based paradigms. 21 3. 2. 3 Bungled Procedure Paradigm (BPP) One interesting attempt of developing a new paradigm was undertaken by Russell et al. (1996; 2002). Participants are given the opportunity to shoot at a human target with a paintball gun. The target is a woman, and the task is presented as a ‘‘novel form of male entertainment’’ (Russell et al., 1996). Aggression in this paradigm is operationalised as the power of the gun chosen (from an array of guns of varying power) multiplied by the number of pellets elected to be used to shoot at the target. In reality, participants never actually shoot at the target, as they are told that there has been a mistake (the ‘‘Bungle’’ of the title), that they are in fact in the control condition, and that they are not therefore required to shoot at the target. A very powerful strength of this paradigm potential to measure intent, as the actual aggressive behaviour never takes place. It furthermore possesses ecological validity, since many real world violent acts, such as homicide, involve firearms (Russel et al., 1996). Finally, participants are not provoked beforehand but also allows the possibility to study more reactive aggression in response to provocation. Although this paradigm certainly possesses face validity, there is no additional mention in the literature of studies using this measure, and therefore, it will not be discussed as extensively. 3. 2. 4 Concluding Remarks While new ideas have been brought forward, the Taylor Competitive Reaction Time Game and the Point Subtraction Aggression Paradigm have remained the two most frequently adopted paradigms. Whereas several studies have been conducted with adapted versions of the TCRTT, no such attempts have been made in studies with the PSAP. Also, studies adopting the more recently developed paradigms are scarce. 3. 3. VALIDITY THREATS The interesting question is to what extent research has improved in the past ten years since the debate. In this section, the main limitations as outlined by Tedeschi & Quigley (1996; 2000) will be applied to the current research summarised above. Have their criticisms been satisfactory incorporated? Moreover, as Tedeschi and Quigley (1996; 2000) propose, one reason for neglecting to take subjects’ conceptions of the situation into account is the theory on which these paradigms (originating from 1967-1971) were built on. According to the authors, in Berkowitz’ (1993) neo-cognitive theory of aggression, motives are considered unimportant, or secondary, and this neglect has been adhered to in measrurement tools as well. However, by now, new models have been presented. Promising is the 22 General Aggression Model developed by Anderson and Bushman (2002; also see Joiremena, et al., 2003). Together with their model they present with a more specified definition. In both the model and the descriptive conceptualisation of aggression, motives and intent play an important role. Have these recent theoretical developments aided adherence to the efinition of aggression? And have researchers taken hold of the issues attributed to traditional aradigms? 3. 3. 1. Theoretical concerns An important addition to traditional measures is the consideration of subjects’ intentions and motivations by some researchers. For instance, Anderson & Murphy (2003) developed a six-item questionnaire to assess hostile and instrumental motivation, and subject motivations are also assessed through self report in the CSTT. Still, many researchers to date continue to neglect to consider this issue. Even when a modified definition has been offered, which specifies the role of motivation and intent to a greater extent, clear statements about how certain designs and paradigms might fit in there have not been encountered. It seems little effort to hypothesise about subjects’ ‘primary goals’ for their behaviour, and relatively little effort to later on confirm them. Apparently, the importance needs to be stressed once more. Tedeschi &Quigley (1996; 2000) made several suggestions of what is being measured if not aggression. This was true for the traditional paradigms but unfortunately, still holds for most recent attempts. It is possible that the ultimate goal is reciprocity, and that subjects retaliate at a level consistent with how they have been treated by the confederate. In effect, this is even what is expected in paradigms using provocation. Then, individual variations in responses reflect variations in feelings of injustice, instead of variations in aggression. Alternatively, the ultimate goal may be conformity to the strategy adopted by the opponent. Conversaly, the primary goal may be one of social control. Either aiming to persuade the opponent to refrain from using high intensities or subjects might set low intensities for the opponent, to persuade them to conform to their strategy. And finally, the behaviour might be aimed at self-preservation. The goal might be to do well, hoping for a positive opinion from the experimenter with respect to one’s adequacy compared to the other participant. For example, in the Hot-Sauce Paradigm, since participants are told they are participating in a food-tasting experiment. Intent to contribute to the aims of the experimenter may influence how subjects respond, i.e. how much sauce they chose allocate to their opponent. Other changes to increase operational adherence are researchers’ attempts to enhance the aversive stimulus’ noxiousness (e.g. Lieberman et al., 1999; Giancola & Parrott, 2008; Kramer et al., 2008;). For instance, in the Kramer et al. (2008) TCRTT study, aversiveness is enhanced by subtracting money, to ensure the subject is motivated to avoid, so that the operationalisation better adheres to the definition. In the Lieberman et al. (1999) HSP study, it was ensured that participants understood the 23 hot sauce would be harmful to the other person. Presumably, this is done to decrease the amount of alternative primary goals subjects might have. However, this could confound results as well by interfering with subjects’ original goal for administering the aversion. That is, variatons in aversiveness may cause variations in motivations to avoid. Finally, compliance to the Parrott & Giancola (2008) taxonomy reveals that the TCRTT (and the CSTT), PSAP, and HSP all measure different forms of aggression. That is, direct and active physical aggression in TCRT and CSTT paradigms, indirect and active physical aggression in the HSP, and indirect and active theft in the PSAP. Furthermore, the design of the HSP does not allow iterative measurements, in contrast to the other paradigms. These differences are not necessarily confounding, as long as researchers take note of them. What form of aggression is measured should be explicitly stated and should correspond to the research question at hand. 3. 3. 2. Non-aggressive response options With respect to reponse options, several studies have incorporated truly non-aggressive response options, followig the recommendations suggested by Tedeschi and Quigley (1996; 2000). This makes measurement of aggressive responding possible under conditions where alternative means of dealing with the situation or frustration are present. However, all of these studies used some variation to the TCRTT. No recent consistent adjustments have been made in studies using the PSAP. The argument of Giancola and Chermack (1998) that non-aggressive responding is possible is questionable. That is, there are two kinds of ‘non-aggressive’ responses. One kind is non-aggressive as in the opposite of aggressive, or neutral compared to aggressive. The other kind is non-aggressive as in not adhering to the operationalisation of aggressive behaviour, and could therefore be almost anything. Of course, in laboratory settings, a true non-aggressive response option should be made available, that is either neutral to, or opposite from, the operationalisation of aggression. Similarly, selecting the lowest intensity in CRTT paradigms is not truly non-aggressive. The fact that in most versions of the PSAP response options are unbalanced, as was pointed out by Tedeschi and Quigley (1996; 2000), has largely remained unaltered (e.g. Cherek et al., 1993; 2000; Bailly & King, 2006). The effort it takes for each of the three response options is unmatched, and it is not clear as to why this is so. In an recent variation of PSAP, response options were matched (Nouvion et al. 2007). In an attempt to assess instrumental aggression, participants were not provoked by occasional point subtractions, and points subtracted from the opponent were added to their earnings. The authors argued that this operationalisation of aggression indexes instrumental behaviour because responses were goal-directed (i.e. earning money) and not provoked. In this version, the number of presses needed to collect one point for oneself was equal to the number of presses needed to subtract money from the opponent. This directly relates to the critique of Tedeschi and Quigley (1996; 2000) 24 of unequal response options in the PSAP. However, again, the reasoning behind this adjustment is unclear. In an earlier study (Cherek et al., 1992) the number of presses needed for an aggressive response was varied and could be either 20, 40, 80 or even 100, creating one condition in which response alternatives were matched. But here, there is no mentioning of this being a reflection of instrumental aggression, apparently because participants were provoked. Moreover, while introducing the study, the authors even argue that TCRTT and PSAP measures reflect instrumental rather than hostile, reactive responding (Cherek, et al., 2000)! And importantly, the results showed that aggessive responding declined as the number of button presses needed increased, demonstrating the significance of the differening demands. There will certainly be a justification behind this reasoning, but it should be made more explicit as evidenced by these results. However, this study suffered from one major limitation: only four subjects were tested, and therefore, this procedure needs to be replicated. Contrary to what one might expect, till date, this has not been done. The HSP also does not offer a true non-aggressive response. Although there is the possibility for participants to choose the least intense option (a small amount of sauce), the same criticism as for classic TCRTT-based paradigms is applicable. If participants wish to behave non-aggressively it is very likely they will do so by choosing the weakest response. However, this does not mean that when participants choose this response, their motive is a non-aggressive one. In the CSTT participants are free to respond as they wish. If they are mainly motivated to compete, they can choose to simply play the game and not annoy the opponent at all. They will still be interacting with their opponent, even if they choose not to annoy them. Moreover, when response options are limited, and administration of shocks is required, behaviour might become influenced by (unintentional) cues from the experimenter suggesting what is expected from the participant. If there is only one way to respond, subjects easily grasp what is expected from them, regardless of whether they have the same interpretation of the situation. Alternatively, if aggressive responses can be initiated at any time and at any frequency (as in the CSTT), the situation becomes too complex for subjects to figure out/assess what behavioural pattern is expected. In the HSP, perceived experimenter demands might play a significant role. Subjects are instructed to administer “adequate sauce”. To researchers, increasing amounts of hot sauce are indicative of the harm subjects intent to do to the other person. Certainly not all subjects will have the same interpretation. Keeping in mind subjects are tested under the deception of participating in a foodtasting experiment, their reasoning could very well be based on what subjects think the experimenter might expect of them. Helpful might be to inform subjects their performance will remain unknown to the experimenter until the investigation has fully been conducted, reducing their need to satisfy the demands of the experimenter. 25 3. 3. 3. Social interaction The issue of proximity, or social interaction, has also largely been neglected. One reason is presumably the ethical concerns involved with having particpants engage in actual physical harmful contact. Therefore, a powerful advantage of the HSP is the fact that the issue of the interactive component of aggressive acts and the need of proximity for administering physical harm is overcome by characteristics of the aversive stimulus. It is a potential means of administering physical harm but for this it is not necessary for those involved to be in close proximity of one another. In the CSTT, players are still situated in different rooms, but because they can view eachother via webcams, this is potentially less confounding. 3. 4 CONCLUDING REMARKS In summary, Tedeschi and Quigley (1996; 2000) argue that it is possible that ultimate goals or underlying motives interfere or mediate the particular responses observed in laboratory settings. The degree to which recent studies of aggression have improved the measurement techniques to increase construct validity have been evaluated. The Taylor Competitive Reaction Time Task has undergone several meaningful adaptations. The task has been coupled with measures of subjects’ motives in several studies, which seems promising. Response alternatives have been increased by adding an intensity of zero. And occassionally adjustments have been made with the purpose of decreasing the amount of alternative interpretations of subjects’ responses, thereby increasing face validity. However, no significant changes have been made to the more traditional PSAP. Indeed, serious claims against the value and usefulness of the paradigm have been ignored. The more recently developed paradigms are promising, one more than the other. While the HSP also fails to consider individual differences in primary goals, offers no valid substitute for aggressive behaviour, and may be sensitive to demand cues, the paradigm possesses two notable advantages. The HSP holds superior ecological validity because in real-life, aggression is also sometimes expressed through administration of noxious food. Additionally, due to the stimulus characteristics, there is no need for participants to be in close proximity of one another, while the action remains one of physical harm-doing. In the CSTT, all major limitations that traditional paradigms suffered from have been adjusted. Subjects are inquired about their goals when responding aggressively, aggressive responding is entirely optional and can be assessed under both provoking and unprovoking circumstances. Several other adjustments have been made, attempting to account for all the minor and major limitations inherent in traditional measurement techniques. 26 CHAPTER FOUR RECENT EMPIRICAL EVIDENCE FOR CONSTRUCT VALIDITY Evidently, validity cannot be assessed based purely on theoretical and methodological considerations. Evidence for the paradigms in addition to that brought forward by Giancola & Chermack (1998) and Tedeschi and Quigley (2000) is presented. 4. 1 CONVERGENT VALIDITY CRTT responses have been found to correlate with state and trait hostility and aggression as assessed with the Buss and Perry (1992) Aggression Questionnaire (BPAQ; Anderson et al., 2004; 2008; Arriaga et al., 2008). These studies used a loud noise as aversive stimulus and included a nonaggressive option (intensity or volume of zero). Giancola & Parrott (2008) demonstrate positive correlations between CRTT responses and self report anger and aggression as assessed on the BPAQ, trait anger, scores on the ABC anger inventory, and a measure of aggressive beliefs. However, they used a CRTT version with shocks and intensities ranging from 1-10, thus lacking a non-aggressive choice (Giancola & Parro,t 2008). Evidence suggesting convergent validity of the PSAP comes from laboratory studies suggesting a relationship between aggressive responding on the PSAP and the Buss Durkee Hostility Inventory (BDHI; Buss & Durkee, 1957) self report measure of aggression (Gerra et al., 2001; 2004; 2007; Golomb et al., 2007). However, as mentioned before, the validity of the BDHI itself as a measure of aggression is questionable (cf. Parrott & Giancola, 2007). Another study that is referred to in support of convergent validity is the Cherek et al. (2000) study. Here, subjects were assessed on three measures of aggression: the BHVQ (Brown et al., 1979), the BDHI, and the Retrospective Overt Aggression Scale (ROAS; Sorgi et al., 1991). Together, the questionnaires produce eight relevant total and sub-scores to assess aggressive personality, creating 40 possible correlations, of which only six were significant. Finally, two recent studies (Tcheremissine et al., 2005; Lieving et al., 2008) did not find any significant correlations between aggressive responses on the PSAP and subjective measures of aggression (BPAQ) and scores on the Lifetime History of Aggression Questionnaire (LHA; Coccaro et al., 1997). Noteworthy, these studies presented with different response options; either with or without the additional escape option. 27 4. 2 DISCRIMINANT VALIDITY No evidence was found for discriminant validity of recent versions of the CRTT. It seems odd that no attempts have been made to investigate the relationship with competition, considering it is important that the aggressive resonses do not correlate with competition. Also, to date, no study investigating discrimant validity of the PSAP has ever been conducted. 4. 3 INDIRECT EVIDENCE Since many investigations have been conducted with respect to laboratory aggressive responding, and most of these can be argued to contribute to indirect evidence in support of the validity of research measures, only a few will be mentioned here. In a CRTT study using noise blasts as aversive stimulation and offering a non-aggressive response option, provoked aggressive responding was increased by brief exposure to a violent video game. However, the significance of the relation between violent media and aggressive behaviour has recently been debated (cf. Ferguson, 2007). Giancola & Parrott (2008) demonstrated CRTT aggression was higher in intoxicated individuals However, as noted by T&Q, the mechanisms of the effects of alcohol itself are not yet fully understood yet, therefore, this is not convincing evidence. Recent indirect evidence concerning the PSAP comes from studies showing differences in PSAP aggressive responding between violent and non-violent female parolees (Cherek et al., 2000; a similar relationship was found for male parolees, see Cherek et al., 1997) in a paradigm providing three response options (monetary reinforcement, aggressive responding and escape responding). On the other hand, in a version with two response options (no escape option; Bailly & King, 2006), the paradigm appeared unable to discriminate between high and low aggressive individuals (as measured by the MCMI-II; Millon, 1987). So far, only indirect evidence has been gathered in support of the Lieberman et al. (1999) Hot Sauce Paradigm. As these researchers hypothesized, distinct domains of self-esteem predicted aggression differently and this effect was mediated by provocation (Webster et al., 2006; Kirkpatrick et al., 2006). Also, the amount of sauce allocated to the other person was affected by provocation (Fisher & Greitemeyer, 2006; DeWall et al., 2007). Finally, increased testosterone predicted increased aggressive responding (Klinesmith et al., 2006). 4. 4 RELATED IISSUES A concern is that the TCRTT and PSAP are used to measure aggression differently in various studies. For example, CRTT studies vary in choice of aversive stimuli: some studies use blasts of loud noise as 28 operationalisation of aggressive behaviour (e.g. Anderson & Murphy, 2003; Kramer et al., 2008; Arriaga et al., 2008; Anderson et al., 2004; 2008), others use the more classic electric shocks (e.g. Giancola et al., 2006; Reidy et al., 2007; Philips & Giancola, 2008; Giancola & Parott, 2008; Zeichner et al., 2003), still others use finger pressure by means of compressed air as noxious stimulation (Lotze et al., 2007) or point and money subtraction (Miller & Lynam, 2003). Response options also differ across studies: in some experiments, there is a non-aggressive option available (e.g. Reidy et al., 2007; Anderson & Murphy, 2003; Arriaga et al., 2008; Anderson et al., 2004; 2008; Zeichner et al., 2003), but not in others (e.g. Giancola et al., 2006; Lotze et al., 2007; Kramer et al., 2008; Philips & Giancola, 2008; Giancola & Parott 2008). The PSAP has been subject to several different variatons too. Although in most studies subjects are offered three response options, this does not appear to be a consistent measure as some studies report using only two options, monetary reinforcement (Button A) and aggressive responding, or, point subtraction (Button B, or PS; e.g. Tcheremissine et al., 2005; Bailly & King, 2006). Monetary reinforcements also vary across studies. For example, one point obtained either 30 cents (Tcheremissine et al., 2005), 15 cents (Lieving et al., 2008), but usually 10 cents (Cherek et al., 2000; Bailly & King, 2006; Nouvion et al., 2007; Carré et al., 2008). Furthermore, aggressive responding is sometimes reinforced by initiation of a protection interval (e.g. Cherek et al., 2000; Tcheremissine et al., 2005; Carré & McCormick, 2008) and sometimes not (Bailly & King, 2006; Lieving et al., 2008). Also, the length of the protection interval is subject to variation (e.g. 62.5 seconds (Tcheremissine et al., 2005); 125 seconds (e.g. Cherek et al., 2000; Lieving et al., 2008) and 250 seconds (e.g. Carré & McCormick, 2008). This indicates that there is no standardized way of interpreting these paradigms, which allows random use of the measure and capitalization on chance. Data from different paradigms should therefore not be combined as indicative of the same underlying concept. Another isssue is that responses on the TCRTT and the PSAP have not yet been correlated. This should be interesting for future attempts. Even if these would tap into slightly different dimensions of aggression, there should still be a modest relation at least. 4. 5 CONCLUDING REMARKS In the past years, only few attempts have been made to confirm the construct validity of the paradigms, and the evidence is not convincing. If anything, support was found for convergent validity for CRTT measures only, and no evidence of discriminant validity has been found for neither measure. Studies examining convergent validity were only found for studies using the CRTT or PSAP. The studies demonstrating evidence in favour of validity used different variations of the task. Evidence for the 29 PSAP was even more limited. The self-report measure most frequently referred to possesses questionable validity itself, and two recent studies could not demonstrate positive relationships with two measures of subjective aggression. Strikingly, discriminant validity has yet to be gathered for all of the paradigms. This is a serious limitation, as no conclusions can be drawn with respect to its ability to meaningfully distinguish between aggression and related, but not similar, concepts. In summary, there is still ample evidence supporting construct validity of aggression paradigms. Overall, in proving empirically established validity, the PSAP and HSP heavily rely on indirect methodologies. Although some authors argue that the PSAP possesses validity (e. g. Giancola & Chermack, 1998; Cherek et al., 2000; Tcheremissine et al., 2005), but the studies they cite to prove validity are either outdated or use very indirect methodology to suggest that correspondence of results in correlational and experimental studies offers an indication of validity for the PSAP. Researchers already engage in validating even recent models and theories by using data obtained from these paradimgs. But to aim most efforts at proving indirect connections at the expense of collecting evidence to support convergent and discriminant validity seems premature, considering the construct validity is still debated. The significance of indirect evidence is limited if convergent validity has not been shown convincingly and discriminant validity has not been assessed at all. Moreover, inconclusive results are often ignored during the subsequent discussion (for an example, see Giancola & Parrot, 2008), failing to acknowledge the inconclusive nature of their findings, and hence, failing to contribute to the gathering of knowledge on the mechanisms involved in aggressive behaviour. The author does mean to necessarily imply that the paradigms lack validity itself, but there sure is a lack of empirical evidence in favour of construct validity. Notably, in assessing validity of measures (or significance of results in general, for that matter) it is not sufficient to report the strength of correlations, nor their significance. To accurately asses the value of a given effect, treatment or manipulation, effect sizes need to be calculated. Westen and Rosenthal (2003) present a simple method providing effect size estimates that represent a quantitive measure of construct validity. Moreover, according to the authors, the two measures of this method are easily produced and compared across studies, constructs, and measures. 30 GENERAL CONCLUSION The aim of this review was to evaluate and discuss the status of present day aggression research with respect to validity of laboratory aggression paradigms. Instigation for this objective was given by a discussion of the construct validity of aggression paradigms between Tedeschi and Quigley (1996; 2000) and Giancola and Chermack (1998). Having reviewed recent attempts at measuring aggression in controlled settings, the general conclusion remains largely unaltered to conclusions initially drawn by Tedeschi and Quigley (1996; 2000): aggression paradigms lack construct validity, or at least, evidence thereof. In the past ten years, theoretical adjustments have been made. The definition has been specified to better capture the complexity of human behaviour. Concurrently, a more extensive mediational model of human aggressive behaviour has been formulated. Taken together, several notable attempts have been made to adjust the classic paradigms to better adhere to the definition of the construct but few attempts have been made to empirically confirm the construct validity of the paradigms in response to Tedeschi and Quigley’s (1996) criticisms. At present, two of the paradigms on which the validity debate was centered, the Taylor Competitive Reaction Time Task and the Point Subtraction Aggression Paradigm, are still the most widely used and most well-accepted measures of laboratory aggression. Consideration of the major limitations of ignorance to the perceptions, motivations, and interpretations of the subjects, and inability to offer a clear nonaggressive response alternative, have been unsatisfactory. Taken together, several adjustments have been made to the TCRTT paradigm. Importantly, attention towards subjective interpretations of the situation and/or motivations for responses has increased. Also, several studies have included response alternatives that truly are clearly non-aggressive. These results are promising. Hopefully, attempts like these will be increasingly incorporated by other researchers as well. But, these attempts are not always adopted by others. What’s more, no consistent changes seem to have been made to the PSAP design and its methodology. Yet, variations of the design appear to be abundant. It seems that in every attempt at measuring aggressive responding on the PSAP, some variable or factor has been (slightly) altered. How can a multifaceted construct such as aggression be reliably assessed when measurement techniques are consistently adjusted? More recent designs for assessing behavioural expressions of aggression have also been developed, but these too have not (yet?) been adopted by contemporary researchers. The Hot Sauce paradigm has several weaknesses but it possesses ecological validity, forming a more familiar task. Equally or maybe more importantly, it enables examination of physical harm without the need for those involved to be in actual bodily contact or close proximity. A paradigm that incorporated all major and minor criticisms is the Competitive Spatial Tracking Task. However, this paradigm is still in development 31 but will certainly prove capable of providing a meaningful contribution to understanding human aggression. Convincing empirical evidence in favour of construct validity remains scarce, ambiguous, or absent. Researchers suggesting otherwise largely rely on indirect methodologies. In addition, there is much variation in the way in which the paradigms are applied. Instead of clarifying the concept of aggression, research conducted in this way may simply add to the confusion. One of the basic principles of research in human subjects is that conditions must be kept constant to reduce error in measurement due to subject variables. Therefore, construct validity yet remains to be demonstrated. To better adhere to the frequently used definitions, these paradigms still need to be enriched so that we can examine the motives and tactics used by people who use threats and punishments as means to achieve interpersonal objectives. Fortunately, the first attempts to reach this goal have been made. Preferably, a good laboratory aggression paradigm would consider subject’s motives and intentions, and would potentially allow aggression to emerge spontaneously as well as following provocation. Ideally, the aggressive behaviour will be one of a variety of offered response alternatives, including not only simply non-aggressive behaviours, but prosocial and communicative responses as well as. Obviously, a perfect measure of human behaviour, let alone a construct describing human behaviour, is not possible. However, we should continue to strive idealism and critically evaluate attempts to measure controlled aggressive behaviour to ensure progression in aggression research, otherwise, scientific understanding of risk factors for aggression and, ultimately, theory development will suffer. \ 32 REFERENCES Anderson, C. A., Deuser, W. E., & DeNeve, K. (1995). Hot temperatures, hostile affect, hostile cognition, and arousal: Tests of a general model of affective aggression. Personality and Social Psychology Bulletin, 21, pp. 434–448. Anderson, C. A., & Bushman, B. J. (2002). Human aggression. Annual Review in Psychology, 53, pp. 27–51. Anderson, C. A., & Murphy, C. R. (2003). Violent video games and aggressive behavior in young women. Aggressive Behavior, 29, pp. 423-429. Anderson , C. A., Carnagey, N. L., Flanagan, M., Benjamin, A. J., Eubanks, J., & Valentine, J. C. (2004). Violent video games: Specific effects ofviolent content on aggressive thoughts and behavior. Advanced Experimental Social Psychology, 36, pp. 199–249. Anderson, C. A., Buckley, K. E. & Carnagey, N. L. (2008). Creating Your Own Hostile Environment: A Laboratory Examination of Trait Aggressiveness and the Violence Escalation Cycle. Personality and Social Psychology Bulletin; 34; 462 Arms, R. L., Russell, G. W., Dwyer, R. S.,& Josuttes, D. ( 1996). Female targets of male aggression in the bungled procedure paradigm. Unpublished manuscript. In: Russell et al., 1996. Aronson, E., Wilson, T. D., & Akert, R. M. (1997). Social psychology.Englewood Cliffs, NJ: Prentice Hall. Arriaga, P., Esteves, F., Carneiro, P., Monteiro, M. B.(2008). Are the effects of unreal violent video games pronounced when playing with a virtual reality system? Agressive Behaviour, 34(5), pp. 521-538. Bailly, M. D., & King, A. R. (2006). Trait modulation of alcohol-induced laboratory aggression. Psychiatry Research, 142(2-3), pp. 129-138. Baron, R. A., Richardson, D. R. (1994). Human Aggression. New York: Plenum. 2nd ed. Barratt, E. S., & Slaughter, L. (1998). Defining, measuring, and predicting impulsive aggression: a heuristic model. Behavioural Sciences and the Law, 16(3), pp. 285-302. Berkowitz, L. (1993).Pain and aggression: some findings and implications. Motivation and Emotion, 17, pp. 277–293. Berkowitz, L. (1993). Aggression: Its causes, consequences, and control. Philidelphia, PA: Temple University Press. Bernstein, S., Richardson, D., Hammock, G. (1987). Convergent and discriminant validity of the Taylor and Buss measures of physical aggression.Aggressive Behaviour, 13, pp. 15 –24. Brown, G. L., Goodwin, F. K., Ballenger, J. C., Goyer, P. F., & Major, L. F. (1979). Aggression in 33 humans correlates with cerebrospinal fluid amine metabolites. Psychiatry Research, 1, pp. 131–139. Bushman, B. J., & Anderson, C. A. (2001). Is it time to pull the plug on hostile versus instrumental aggression? Psychological Review, 108(1), pp. 273-279. Buss, A, & Durkee, A. (1957). An inventory for assessing different kinds of hostility. Journal of Conslting and Clinical Psychology, 21, pp. 343–349. Buss, A. H. (1961).The Psychology of Aggression, New York,NY: Wiley. Buss, A. H., & Perry, M. (1992). The Aggression Questionnaire. Journal of Personality and Social Psychology, 63, pp. 452-459. Campbell, D. T., & Fiske, D. W. (1959). Convergent and discriminant validation by the multitrait-multimethod matrix. Psychological Bulletin, 56(2), pp. 81-105. Carré, J. M. & McCormick, C. M. (2008). Aggressive behavior and change in salivary testosterone concentrations predict willingness to engage in a competitive task. Hormones and Behaviour, 54(3), pp. 403-409. Cherek, D. R., Spiga, R., Steinberg, J. L., & Kelly, T. H. (1990). Human aggessive responses maintained by avoidance or escape from point loss. Journal of the Experimental Analysis of Behaviour, 53(2), pp.293-303. Cherek, D. R., Spiga, R., & Egli, M. (1992). Effects of response requirement and alcohol on human aggressive responding. Journal of Experimental Analysis of Behaviour, 58(3), pp. 577-587. Cherek, D. R., Roache, J. D., Egli, M., Davis, C., Spiga, R., & Cowan, K. (1993). Acute effects of marijuana smoking on aggressive, escape and point-maintained responding of male drug users. Psychopharmacology, 111(2), pp. 163-168. Cherek, D. R., Moeller, F. G., Schnapp, W., & Dougherty, D. M. (1997). Studies of violent and nonviolent male parolees: I. Laboratory and psychometric measures of aggression. Biological Psychiatry, 41, pp. 514–522. Cherek, D. R., Lane, S. D., Dougherty, D. M., Moeller, F. G., & White, S. (2000). Laboratory and questionnaire measures of aggression among female parolees with violent or nonviolent histories. Aggressive Behaviour, 26, pp. 291–307. Cornell, D. G., Warren, J., Hawk, G., Stafford, E., Oram, G., & Pine, D. (1996). Psychopathy in instrumental and reactive violent offenders. Journal of Consulting and Clinical Psychology, 64(4), pp. 783-790. DeWall, C. N., Baumeister, R. F., Stillman, T. F., Gailliot, M. T. (2007).Violence restrained: effects of self-regulation and its depletion on aggression. Journal of Experimental Social Psychology, 43, pp. 62 –76. Ferguson, C. J. (2007). Evidence for publication bias in video game violence effects literature: A meta-analytic review. (2007). Aggression and Violent Behavior, 12, pp. 470-482. 34 Fischer, P., & Greitemeyer, T. (2006). Music and aggression: the impact of sexual-aggressive song lyrics on aggression-related thoughts, emotions, and behavior toward the same and the opposite sex. Personality and Social Psychology Bulletin, 32(9), pp. 1165-1176. Gerra, G., Zaimovic, A., Raggi, M., Giusti, F., Delsignore, R., Bertacca, S., & Brambilla, F. (2001). Aggressive responding of male heroin addicts under methadone treatment: psychometric and neuroendocrine correlates. Drug and Alcohol Dependance, 65, pp. 85–95. Gerra, G., Zaimovic, A., Moi, G., Bussandri, M., Bubici, C., Mossini, M., Raggi, M., & Brambilla, F. (2004). Aggressive responding in abstinent heroin addicts: neuroendocrine and personality correlates. Progress in Neuropsychopharmacology and Biological Psychiatry, 28, pp. 129–139. Gerra, G., Zaimovic, A., Raggi, M., Moi, G., Branchi, B., Moroni, M., & Brambilla, F. (2007). Experimentally induced aggressiveness in heroin-dependent patients treated with buprenorphine: comparison of patients receiving methadone and healthy subjects. Psychiatry Research, 149, pp. 201–213. Giancola, P. R., & Chermack, S. T. (1998). Construct validity of laboratory aggression paradigms: a response to Tedeschi and Quigley (1996). Aggression and Violent Behavior, 3, pp. 237–253. Giancola, P. R., Godlaski, A. J. & Parrott, D. J. (2006).Perceptions of one's attacker's intentions following an aggressive interaction involving alcohol. Journal of General Psychology, 133(4), pp. 389-400. Giancola, P. R. & Parrott, D. J. (2008). Further evidence for the validity of the Taylor Aggression Paradigm. Aggressive Behaviour, 34(2), pp. 214-229. Golomb, B. A., Cortez-Perez, M., Jaworski, B. A., Mednick, S., & Dimsdale, J. (2007). Point subtraction aggression paradigm: validity of a brief schedule of use. Violence Victims, 22(1), pp. 95-103. Joireman, J., Anderson, J., & Strathman, A. (2003). The aggression paradox: understanding links among aggression, sensation seeking, and the consideration of future consequences. Journal of Personality and Social Psychology, 84(6), pp. 1287-302. Kaplan, H. I., & Sadock, E. J. (2003). Synopsis of Psychiatry, 9th Ed.. Kirkpatrick, L. A., Waugh, C. E., Valencia, A., & Webster, G. D. (2002). The functional domain-specifity of self-esteem and the differential prediction of aggression. Journal of Personality and Social Psychology, 82, pp. 756 –767. Klinesmith, J., Kasser, T., & McAndrew, F. T. (2006). Guns, testosterone, and aggression: an experimental test of a mediational hypothesis. Psychological Science, 17(7), pp. 568-571. Kramer, U. M., Jansma, H., Tempelmann, C., & Munte, T. F. (2007). Tit-for-tat: the neural basis of reactive aggression. NeuroImage,38(1), pp. 203-211. Lieving, L. M., Cherek, D. R., Lane, S. D., Tcheremissine, O. V., & Nouvion, S. O. 35 (2008). Effects of acute tiagabine administration on aggressive responses of adult male parolees. Journal of Psychopharmacology, 22(2), pp. 144-152. Lieberman, J. D., Solomon, S., Greenberg, J., & McGregor, H. A. (1999). A hot new way to measure aggression: Hot sauce allocation. Aggressive Behaviour, 25, pp. 331 –348. Lotze, M., Veit, R., Anders, S., & Birbaumer, N. (2006). Evidence for a different role of the ventral and dorsal medial prefrontal cortex for social reactive aggression: An interactive fMRI study. NeuroImage,34(1), pp. 470-478. McEllistrem, J. E. (2004). Affective and predatory violence: A bimodal classification system of human aggression and violence. Aggression and Violent Behavior, 10, pp. 1–30. McGregor, H. A., Lieberman, J. D., Greenberg, J., Solomon, S., Arndt, J., Simon, L., & Pyszczynski, T. (1998). Terrormanagement and aggression: Evidence that mortality salience motivates aggression against world-view-threatening others. Journal of Personality and Social Psychology, 74, pp. 590 –605. Milgram, S. (1963). Behavioral study of obedience. Journal of Abnormal Social Psychology, 67. pp.376. Miller, J. D., & Lynam, D. R. (2003). Psychopathy and the Five-factor model of personality: a replication and extension. Journal of Personality Assessment, 81(2), pp. 168-178. Millon, T. (1987). Manual for the MCMI-II. National Computer Systems, Minneapolis. Nouvion, S. O., Cherek, D. R., Lane, S. D., Tcheremissine, O. V., & Lieving, L. M. (2007). Human proactive aggression: association with personality disorders and psychopathy. Aggressive Behaviour, 33(6), pp. 552-562. Parrott, D. J., Giancola, P. R. (2007). Addressing ‘‘The criterion problem ’’ in the assessment of aggressive behavior:Development of a new taxonomic system. Aggression and Violent Behaviour, 12, pp. 280 –299. Phillips, J. P., & Giancola, P. R. (2008). Experimentally induced anxiety attenuates alcohol-related aggression in men. Experimental Clinical Psychopharmacology, 16(1), pp. 4356. Polman, J. D. M. (2008). Hot-headed or cold-blooded? Towards a clear distinction between reactive and proactive aggression in youth. Dissertation, Utrecht University. Reichman, M. (1991). Rating aggression in the clinical setting: a retrospective adaptation of the overt aggression scale. Journal of Neuropsychiatry and Clinical Neuroscience, 3, pp. S52–S56. Reidy, D. E., Zeichner, A., & Martinez, M. A. (2008). Effects of psychopathy traits on unprovoked aggression. Aggressive Behavior, 34(3), pp. 319-328. Ritter, D. & Eslea, M. (2005). Hot sauce, toy guns, and graffiti: A critical account of current laboratory aggression paradigms. Aggressive Behavior, 31, pp. 407 –419. Russell, G. W., Arms, R. L., Loof, S. D., & Dwyer, R. S. (1996). Men ’s aggression 36 toward women in a bungled procedure paradigm. Journal of Social Behaviour and Personality, 11, pp. 729 –738. Russell, G. W., Arms, R. L., Dwyer, R. S., & Josuttes, D. (2002).Females as targets of male aggression in the bungled procedure paradigm. University of Lethbridge. Savage, J. (in press). The competitive spatial tracking task. Cardiff University, Cardiff. Sorgi, P., Ratey, J. J., Knoedler, D.W., Markert, R. J., Reichman, M. (1991). Rating aggression in the clinical setting: A retrospective adaptation of the Overt Aggression Scale: preliminary results. Journal of Neuropsychiatry and Clinical Neuroscience, 3(2), pp. S52-S56. Taylor, S. P. (1967). Aggressive behavior and physiological arousal as a function of provocation and the tendency to inhibit aggression. Journal of Personality, 35, pp. 297–310. Taylor, S. P., & Epstein, S. (1967). Aggression as a function of the interaction of the sex of the aggressor and the sex of the victim. Journal of Personality, 35, pp. 474–486. Tcheremissine, O. V., Lane, S. D., Lieving, L. M., Rhoades, H. M., Nouvion, S., & Cherek, D. R. (2005). Individual differences in aggressive responding to intravenous flumazenil administration in adult male parolees. Journal of Psychopharmacology, 19(6), pp. 640-646. Tedeschi, J. T., & Felson, R. B. (1994). Violence, aggression, and coercive actions. Washington, DC: American Psychological Association. Tedeschi, J. T., & Quigley, B. M. (1996). Limitations of laboratory paradigms for studying aggression. Aggressive and Violent Behavior, 1, pp. 163–177. Tedeschi, J.T., & Quigley, B.M. (2000). A further comment on the construct validity of laboratory aggression paradigms: a response to Giancola and Chermack. Aggression and Violent Behaviour, 5(2), pp. 127– 136. Waschbusch, D. A., Pelham, W. E. Jr., Jennings, J. R., Greiner, A. R., Tarter, R. E., & Moss, H. B. (2002). Reactive aggression in boys with disruptive behaviour disorders: behaviour, physiology, and affect. Journal of Abnormal Child Psychology, 30(6), pp. 641-656. Westen, D., & Rosenthal, R. Quantifying construct validity: two simple measures. Journal of Personality and Social Psychology, 84(3), pp. 608-618. 37 ATTACHMENT 1 Anderson & Murphy (2003) six-item motivation questionnaire The questionnaire consists of six items on which participants were to specify the extent to which this motive describesd their your motive for deciding at what intensity to set the noise levels.’ The items used a 5-point scale from 0 (completely disagree) to 5 (completely agree). The six items were (1) I wanted to impair my opponent’s performance in order to win more; (2) I wanted to control my opponent’s level of responses; (3) I wanted to make my opponent mad; (4) I wanted to hurt my opponent; (5) I wanted to pay back my opponent for the noise levels he/she set; (6) I wanted to blast him/her harder than he/she blasted me. The first two items represent instrumental reasons for aggressing, and were combined to form a scale labeled ‘‘Instrumental Aggressive Motivation.’’ The latter four items represent a clearly revengeful type of aggressive motive, were combined to form a scale labeled ‘‘Revenge Motivation.’’ From: Anderson, C. A., & Murphy, C. R. (2003). Violent video games and aggressive behavior in young women. Aggressive Behavior, 29, pp. 423-429. 38