There are several conclusions that can be drawn from the data. The first thing is that there is no difference between personality variables for comic readers and non-comic readers. The lack of difference between readers and non comic readers is not surprising. Why would there be? There is no single personality factor that decides what people like, or are interested in. Therefore it would be an unusual turn of events if there was a single variable that controlled whether or not someone read comics. The data also shows that long- term exposure to comic books has no effect in the development of a HAB. The results support the contention that short-term exposure to EVCB causes a HAB. Individuals exposed to EVCB show more intent, retaliation and emotion after short-term exposure. Although the data showed that those highest in trait hostility responded most aggressively the proposed interaction with Trait Hostility and Stimulus was not significant. EVCB bias SIP towards aggressive responding and these effects are greatest in EVCB reading females responding to relational scenarios. For the overt scenarios those reading EVCB respond most aggressively. The findings concerning a link between EVCB and SIP are as expected since they have been found in the earlier Kirsch and Olczak studies (2002). That Trait Hostility plays a role in aggressive responding is similarly uncontroversial.

Where the results differ from previous studies is in predictions of female aggression. Earlier data suggested that male participants reading EVCB respond more negatively on ambiguous provocation story questions than females and males reading NVCB and that Females’ responses are largely dependent on trait hostility. Kirsch & Olczak 2002 paper examining relational aggression found that females responded more aggressively to relational scenarios than men. They also found was that males responded more vigorously to scenarios involving overt aggression. This study confirmed the high scores for EVCB reading females’ responses to relational scenarios but did not confirm the effect for EVCB reading males’ responses to overt scenarios. Males in the study reacted as they were expected to. They provided high responses to overt scenarios and low responses to relational scenarios. Females also reacted as expected by providing more aggressive responses to relational scenarios. However they responded equally as violently to the overt scenarios. Since females are equally as violent for both overt and relational scenarios it seems as if the females in this study are more aggressive than the males.

This difference however is primarily due to the lack of male responses to the relational scenarios. A typical male reply to relational scenario was “I don’t care”. Females on the other hand, being much more social, respond to the social sabotage much more enthusiastically. In a similar study involving relational aggression females were found to be more aggressive than males (Coyne, Archer J & Elsea, 2003). Most of the extreme responses to the relational aggression scenarios are due to few high hostility females. This is also the case for the earlier 2002 study. The females examined for this study are more Extraverted, higher in Trait Hostility, less Open and more Agreeable than the males. It’s suggested that aggressive networks may function differently for males and females. The Script Theory of aggression states that children observe violence in mass media and learn aggressive scripts (Anderson & Bushman 2002). Males and females may have different scripts because they like different things. Females, who prefer indirect rather than direct aggression, like and have more relational scripts than males who like and have more overtly aggressive scripts (Crick & Groepter, 1995). In this study it seems as thought he females have an equal supply of both overt and relational scripts.

A major difference between this and previous studies was in perceptions of comic books. Kirsch & Olczak found that each gender perceives similar levels of violence but prefer different ones. This meant that males liked more and showed more interest in EVCB than females who liked and had more interest in NVCB (Kirsch & Olczak, 2001; 2002). This study found no such data. The subjects perceived the same amount of violence which tends to suggest that the EVCB and NVCB were doing their role as stimulus, but there was no sign of the classic effects predicted by the earlier research. For interest and like there were no main effects or interactions. For humour there were Gender and Stimulus effects but no interaction. The reason for the difference between this research and prior research is down to the subjects. The subjects in the sample do not show evidence of the same polarisation of perception. The female subjects are much more enthusiastic about the EVCB than female subjects in previous studies. Males show no particular preference for EVCB and females show no particular preference for NVCB. Since the majority of the males are comic readers when they come to make an aesthetic judgment on the comic they have read they have a more sophisticated appreciation of it. Rather than simply responding to a simple “me like violence” heuristic they apply their years of experience to rate this comic on its true value. Informal debriefing on web sites suggests that the vocal majority of males were not very enthusiastic about the EVCB. The same debriefings showed that females were very enthusiastic not just about the EVCB but about the entire experiment in general. This may be a possible confounding factor. The female subjects of this study were largely psychology students. Perhaps any appreciation they may have for the comics may just be a reflection of their appreciation of the nuances of the experiment.

The study has some potential problems. One of the major ones is that there is not enough variability in the scores for overt and relational aggression. Initially the study was going to be simpler and consider only a single total aggression variable. For this three questions were deemed as enough to provide a robust enough variable for analysis. On examining the data post-test it was decided that a more involved analysis would be much more revealing and could be used to answer a more involved hypothesis. The problem is that the relational aggression variable is based on responses to a single question. In the Kirsch & Olczak 2002a study there are five scenarios. In the later 2002b analysis there are three. Although the data shows similar results, a larger spread of data would be preferable before any concrete conclusions can be drawn. The Overt aggression score is not much more reliable. This variable is formed by responses to only 2 questions. Again a larger spread of data would be required before more definitive conclusions are made.  

There are also potential difficulties with the way the experiment is designed. Previous studies were run in the lab and were supervised by experimenters. This test was self administered in whatever environment the subjects chose to work in. Since the subjects read the comics under real world conditions it gives the results ethological external validity. However it may present the problem of a lack of reliability of the results. The web site design was intended to allow for a wider range of subjects. It was also thought that this ease of use would encourage subjects to participate. The problem is there is no control. There could be any number of potential factors which can be controlled in the laboratory but are possible confounds in the real world. Mood and interference are two obvious examples. In the lab subjects are unmolested as they work on the study. In the real world who knows what background conditions are interfering. Similarly subjects in the lab are unlikely to be in an extreme mood. They will be restrained by the invisible bonds of society. At home in front of the computer people can be in whatever state of arousal they wish before taking the test. The design trades off control for subjects. The lack of control has meant more subjects however more data does not guarantee sound data.

This amount of exposure to comics is less than in the original study. In earlier experiments 20 minutes exposure to comics is used. Perhaps this extra time reading is a factor. The weapon priming, the paradigm most appropriate to this study, has an almost immediate effect so perhaps the minimal exposure is not a relevant complaint. A more solid worry is whether or not the experiments are actually measuring aggression. How one says they will act is not really same as acting. A bark is not the same as a bite. Perhaps some kind of physiological measure would be better. Anger or arousal is associated with increased heart rate, blood pressure and skin conductance (Bushman & Huesmann, 2006). A measure of one of these physiological indices of arousal may be a better measure of Aggressive response. This however would mean an involved laboratory set up. Here it becomes a trade off between external validity of results, an accurate measure of the dependent variable and the cost and complexity of running such an experiment.

Deception always plays a role in psychological experimentation. The second subjects know the purpose of an experiment is the second that response bias becomes an issue. Subjects informally questioned after completing the paper version of the test were unaware of the true intent of the research. For subjects taking the e-version of the test there was no way to check this. A simple check box added to the final page of the test could have been used to test for whether or not the deception was successful. This simple measure would have allowed for the removal of potentially compromised data. The uproar caused by the deception of subjects on the DC website would suggest that the deception was not realised. This deception led to the problem of active sabotage of the experiment.
Subjects were recruited from the comic readers’ message boards by a short advert detailing the researcher and the research. The research was not what was advertised but the true purpose of the experiment was explained during debriefing. Subjects were also advised that they were free to withdraw and to have their data removed at any stage. The process was straightforward and on the whole effective. After an initial request the community that inhabit the message board asked a few pertinent questions, answered the questionnaire and then moved on to the next flavour of the month. This was not the case on the DC web site. A self-appointed guardian, outraged by the deception they’d been subject to, posted the debriefing page immediately following the first post. A lengthy debate as to the ethics, morality and robustness of psychological methods then followed. Some were supportive of the method defending deception as being a necessary evil. Others were hostile and boasted of actively sabotaging the experiment by inputting false data. Fortunately such false data was generally gross enough to be detected. Six subjects were removed for their juvenile responses. A more pressing problem was that all the subjects from this sample group now knew the purpose of the experiment. To test for sabotage and potential contamination a 2(Gender) X 2(Sample) ANOVA was conducted for Total Response Aggression. The results are non-significant for Sample, F (1, 117) = 1.32, p= 0.266, Gender, F (1,117) = 2.16, p= 0.144 and for the interaction between Sample X Gender, F (1,117) = 0.63, p= 0.640. This seems to indicate that there were no significant differences between the samples. If the subjects were compromised it did not show in their responses.

A factor raised as a concern by the intelligentsia of the DC web site was that of self selection. It was argued that by volunteering to participate subjects’ results were biased. The results of the experiment could be due to a single underlying personality trait that affects all the subjects’ responses. Short of press-ganging subjects to participate self selection will be a problem for all experiments. Analysis of the personality data from the NEO-PI questionnaire does not reveal any prevalence of one specific personality factor over another. The subjects show a varied even spread across all personality variables. Volunteering would seem to come under the personality facet of Agreeableness. Since this factor has been entered as a covariate in the major analyses its influence has been controlled for.   
Another concern may be the web site selection process. Almost all the top selling comics are produced by DC and Marvel comics (these companies account for around 80% of the entire comic output in the US- Diamond, 2007). It was reasoned that subjects who read the message boards for these companies are representative of the average comic reader. In the UK 2000AD is the most popular magazine of its type. Readers of its message boards are therefore assumed to be prototypical members of the British comic reader. Critics have suggested that readers of message boards are not representative since their expressed views do not match sales figures. However this divergence may be due to the difference between subjective remembering of past events, which is coloured by individuals’ specific memory retrieval biases, with objective sales figures. The fact that there are no significant difference between the personality of comic and non-comic readers would tend to suggest that there are unlikely to be differences between comic readers who post on forums and comic readers who don’t. If there are no differences between two distinct groups then there would seem to low probability of differences between individuals within the same subgroup of comic readers. It thus seems a fair assumption that subjects are representative of comic readers. Since few subjects expressed any interest in EVCB they’re however not representative of violent comic readers. This may affect the strength of any results that relate to an increase in HAB after years of exposure to EVCB.

The small sample of comic books used may limit the generalizability and external validity of any findings (Kirsh & Olczak, 2002b). Whilst using a single issue means that there is better comparison of data it does present potential problems. Perhaps readers are responding to a single feature within a single issue. A wider selection of material would be better.  As well as wider reading material a wider net of web sites would also be an idea. As mentioned above subjects who read EVCB were not included in this study. Nor were readers of independent comics. Because these are niche markets the subjects were not deemed as being necessary to make general predictions. Since the majority of non-comic readers are university students this means they have their own particular intellectual and socio-economic biases. All are participants from the same institution and so may have shared norms and values including a common understanding of aggression (Bond, Bauer & Wingrove, 2004).  Replication of this study with a wider selection of subjects, both comic readers and non-comic readers would be an improvement.

The results were coded by an experimenter who was not blind to the experimental condition. This means bias may creep in and responses may be marked according to what is expected. This method did however have its advantages. When coding the data this researcher became aware of a trend. It seems as if a lot of female subjects were using the word “he” to describe the provocateur in the ambiguous provocation scenarios. The scenarios are written to be gender free so there may be a confounding factor at work. By means of illustration an image of a little girl crying because she had her toy taken from her by a larger male was presented on the instruction page before the ambiguous provocation scenarios. It’s possible that this image primed subjects with mental imagery which they focused on when replying to the questions. This may have been the reason why females personified the provocateur as male. If this image was as successful in priming as suspected then a variation of the ambiguous provocation stories laid out in comic form would be an excellent method for future analysis. It resonates with the source material and would allow for the testing of a more involved analysis (such as whether or not the gender of the provocateur has any effect on response). The design would have been better if the picture was not part of it. It would be better scientifically but not aesthetically. Aesthetic considerations were included in the design. It was reasoned that an eye pleasing experiment would encourage subjects to complete the study and to be honest and accurate. It did not stop attrition, 68 subjects were lost though incomplete questionnaires.

The projective aspect of the test may itself present problems. Participants are told to “Respond as a 10 year old would respond.” These child-like responses are said to be projections of participants’ current cognitive/emotional state. This instruction is to maintain the pretence that the study is about personality development. This deception fools participants to the actual purpose of the study and thus eliminates participant bias. Research has found projective tests to be predictive of significant aspects of personality, in certain situations (Kirsh & Olczak, 2002a). Still the opinions expressed for this test are not necessarily the opinions of subjects. A number of subjects added riders onto their answers qualifying their aggression as being how they imagined they might respond. This means that there is a difference between the response they give and the response they genuinely feel. This doesn’t seem a very accurate measure of aggression. Perhaps instead of coding subjects’ responses a Likert scale could be constructed. The questions tend to force certain responses. The question “How much would you retaliate? None at all, a little or a lot?” is generally answered by one of the 3 options provided. This could easily be converted onto a Likert scale. This scale would encourage more automatic responding. In such studies automaticity is a key ingredient. The faster the response the more honest and accurate it is. It would also be a lot easier to code and not as much of an interpretive science.
There are potential difficulties with the reliability of the measures. The full version of the NEO-PI questionnaire uses 140 items. This study, for the sake of brevity, used only 25. This must surely have an impact on the validity of any personality data. In defence the subjects’ replies to the PI were consistent. Those who scored high on one item relating to a specific factor scored high on all items. However since Neuroticism seems to be a factor in this study a more involved personality inventory should be carried out before any Nobel prizes are awarded. The five factor solution does not say all there is to say about personality, “whereas almost any personality construct can be mapped onto the Big 5, you cannot derive every personality construct from the Big 5 (Funder, 2001). There are also potential difficulties with the Aggression Questionnaire. The AQ, which was constructed in 1992, does not have the same 35 years of construct validity that the BDHI has. However the AQ has been cited in 530 studies. An analysis of a randomly selected sample of these papers indicates that the test is being used as a working, accepted measure of aggression. Thus concerns at the set up stage about the inferiority of the AQ compared with the BDHI seem to have been misplaced.  

The aggression exhibited by subjects in this test seems to be short lasting. All the studies show that EVCB cause short-term effects. The priming effect, like the weapon priming effect, is only temporary (Huesmann, 1988; Bond, Bauer & Wingrove, 2004). A few subjects primed by EVCB spoke of raised aggression which quickly dissipated after the first scenario. Closer analysis of the responses seems to confirm this. Aggressive responses do seem to be higher for the first scenario. This data was not collated at the time so any conclusions are based on subjective appraisal of the individual results. A closer examination as to levels of aggression through the testing stage may yield interesting findings.
Another factor to consider is the role of Neuroticism. In the analyses of overt aggressive responses Neuroticism co-varies with gender. Since neuroticism co-varies with Trait Hostility and Humour it would seem that there is some kind of interaction going on. A 2(Gender) X 2(Trait Hostility) ANCOVA for Neuroticism with humour as a covariate indicates main effects for the covariate humour, F (1, 122) = 10.80, p=0.001, gender, F (1, 122) =5.82, p=0.017 and trait hostility, F (1, 122) =7.25, p=0.008. The Gender X Trait Hostility interaction was not significant, F (1, 122) =0.01, p=0.914. Previous studies have found correlations between NEO-PI factors and scales of widely used clinical instruments (Costa & McCrae, 1992). Neuroticism and Extraversion have been found to be linked to psychopathology. So perhaps the fact that a link has been found between Neuroticism and Trait Hostility is unsurprising. These factors are related but how they interact is beyond the scope of this experiment.