Saturday, July 10, 2021

Weapons Effect Theory

A little while ago, I made mention that I had noticed the weapons effect, which I had always considered to be a phenomenon, referred to as a theory. In a way I found it amusing. In another way, I think the argument in favor of a weapons effect theory does have some merit. A good theoretical model would at minimum offer an explanation regarding a phenomenon and generate testable hypotheses. In the case of the weapons effect, it would be a relatively narrow theory. Then again, so too was frustration-aggression theory (itself an outgrowth of what was merely a hypothesis). 

We know the origins of what we could call weapons effect theory. We look no further than Berkowitz and LePage (1967). As the details of that initial experiment are detailed elsewhere, I will simply state that Berkowitz and LePage (1967) appeared to demonstrate that under conditions of high provocation, individuals experiencing short-term exposure to weapons showed higher levels of aggression (measured in number of electric shocks given) than those who had not been exposed to weapons. It goes without saying that the claim was highly controversial at the time, and that there were critics who could not replicate the original finding. That story has been told many times (including by me - see Benjamin 2019 or Benjamin, 2021), and bears no repeating here. What probably matters most is that a meta-analysis by Carlson, Marcus-Newhall, and Miller (1990) was supposed to have settled the matter. Short-term exposure to weapons under conditions of high provocation or frustration seemed to lead to a noticeably higher level of aggression than any other condition. Armed with Fail-Safe N as a means of assessing publication bias, Carlson et al (1990) concluded that the case was effectively open and shut. The weapons effect was viable, and it was time to move on. After that, social psychologists and some fellow travelers looked toward underlying processes responsible for this purported effect. That's where theory comes in.

Although the Anderson, Benjamin, and Bartholow (1998) paper referenced Anderson's then General Affective Aggression Model (which would be later abbreviated to General Aggression Model or GAM), I think it is safe to say that what we we actually did was to articulate a distinctive weapons effect theoretical model. Among social cognition models, it is a potentially "warm" theory in the sense that anger and arousal are considered potential antecedents. However, anger (affect) and arousal have never been adequately tested. Rather, testing of the model primarily focuses on short-term exposure to weapons priming of aggressive cognitions - think of these as behavioral scripts and schemas that include all of our semantic and episodic memories and concepts of aggression and violence as well as procedural memories of how to behave aggressively or violently. These memories may be implicit or explicit. Once aggressive cognitions have been primed, primary and secondary threat appraisals are primed, increasing the likelihood that an individual will be biased to perceive stimuli as more threatening than they might have otherwise, along with appraisals of how to best respond. Depending at what happens at the level of appraisal, an aggressive behavioral response might be the end result. Although primarily focused on the situational antecedents, the model keeps the door open to individual differences that might serve as antecedents (including personality traits and life experience). See the figure below. Note that technically this figure is the property of Sage Publications (from Anderson et al,, 1998), and if I am asked to take it down, I will do so:

It's a simple theory, really. One sees a weapon, which facilitates an increase in accessibility of aggressive cognitions, setting up primary and secondary threat appraisals, culminating in increase of aggression. The potential for weapons to prime anger and increase physiological arousal exist as well. It is a model that explains a body of results on a phenomenon, and offers some potential hypothesis tests. So far, so good. So, how well does the weapons effect theory hold up? Depending on whom you read, the weapons effect theory is either sufficiently established that we what we really need to do is to further explore interactions of person variables and short-term exposure to weapons (an endeavor that has barely been undertaken, and then only in a very scattershot fashion), or the body of research suggests the theory is enough of a nothingburger as to be swept into the dustbin of history.

When I finally published my meta-analysis (Benjamin, Kepes, & Bushman, 2018), I think any astute reader would hone in on Table 2 and realize that depending on how how publication bias is assessed, that there is nothing to be concerned about (if one believes random-effects trim-and-fill analyses) or quite serious (e.g., PET-PEESE). Most concerning are studies examining aggressive behavioral outcomes. The effect sizes are arguably negligible. Even when we look at the intervening variables in the model that are the underlying processes responsible for the presumed relationship between short-term weapon exposure and aggression (accessibility to aggressive cognitions and hostile appraisals) we have to keep in mind that the effects for these outcome variables are often small. Establishing accessibility of aggressive cognition is difficult, and numerous methods of measuring accessibility of aggressive cognitions have been utilized with varying degrees of success. Although much of the earlier cognitive priming literature for the weapons effect relied on either reaction times to aggressive versus non-aggressive words in lexical decision tasks or pronunciation tasks, more contemporary studies appear to rely on variations of a word completion task developed by Anderson - an instrument whose validity has been recently questioned. I wonder how many unpublished studies slipped through the cracks. Research on mostly primary threat appraisal has been more of a success story. Much of that work seems to build on research comparing phylogenitic and ontogenetic threats, with weapons being an ontogenetic threat. When individuals are shown arrays of objects with guns or knives embedded, studies appear to find evidence that individuals respond more rapidly to those arrays that have nothing but neutral objects. Effect sizes are small-to-moderate. One must also consider the possibility that arrays including unexpected objects could be just as effective in decreasing reaction times. So, although the pattern of findings looks promising, it's probably far from settled. But ultimately, for a cognitively based theory of the weapons effect to work, there has to be some establishment that aggressive behavioral outcomes are consistently positive. So far, that has not been the case. However aggressive behavior has been operationally defined - number of electric shocks, shock/noise blast levels, amount of hot sauce doled out to a presumed victim, point subtraction, etc., the results have been inconsistent. Some experiments appear successful, but many others do not replicate - either directly or conceptually - the original finding. Furthermore, it is not entirely clear that we are measuring aggression with these operational definitions, nor can we necessarily include intention to harm from the body of research thus far. Furthermore, there has been a tendency for those who do find positive effects to oversell their findings, tying their analyses to not merely the mild forms of aggression that we might be measuring, but to tie that work to acts of violence such as shootings. The theoretical model is not one that was designed to address violence per se, which means that even if we could hone in on consistently reliable findings, we can only speak to a narrow range of possible aggressive behaviors in everyday life, and even then with a good deal of caution and humility. All that being said, we have  a social cognitive model that states that short term exposure to weapons can trigger an aggressive behavioral response, to the extent that aggressive cognitions and hostile appraisals are successfully primed. Without a solid body of evidence pointing to an increase in aggressive behavior in this body of research, the theory falls apart. Unless or until the behavioral outcome piece of the theory is settled, it is a weak theory at best. If research surfaces that debunks any priming effect of weapons on aggressive cognitions, then the theory goes from weak to effectively moot. At that point, why even discuss the matter further?

Left unanswered in the theoretical model is the role of arousal. My impression is that early on, arousal was looked at as a nuisance variable to be measured and ruled out. I know of one explicitly reported arousal study, and it was a pilot study used to select stimulus materials.  As a "warm" theory, I've been a bit taken aback at the lack of interest among those best positioned to examine arousal and affect to actually take the time to do so and report their findings. Nor has the moderating role of individual differences been adequately explored. Aside from some one-offs, very little is known about the role of personality or life events as a moderator of the relationship between short-term exposure to weapons and aggressive behavioral outcomes. 

Research from the last decade has been discouraging. In the last two or three years, we've seen published some ecological valid behavioral work that was either adequately sampled, but showed a small effect size, or what on the surface appeared to be a robust effect, but in which the sample small enough that the statistical test was underpowered. Let's just say that historically experiments testing hypotheses derived from this theory rarely have samples of 15 or more in each cell, and even a sample of 15 per cell is probably inadequate. More traditional experimental research (i.e, in the lab) in recent years appears to suggest the behavioral effect is minimal at best. For example Guo, Egan, and Zhang (2016) found no main effect of weapons on aggressive behavior, and instead used a subsample of individuals who scored as high in external locus of control in order to craft a narrative for their findings. That's just the published research. I am aware of unpublished behavioral experiments (based on personal communication) that have found either a null effect, or even a suppression effect. That should give any of us who either have researched the weapons effect theory or who utilize this theory as part of our pedagogy pause.

Bottom line? As a theoretical model, I am very uncertain that the weapons effect theory is on solid ground. If anything, I am likely to agree with those who would argue that it is not on solid ground at all, and that it is a theoretical model worth abandoning. There appear to be small to moderate effects when it comes to weapons priming aggressive cognitions and hostile threat appraisal. The effect on aggressive behavior appears potentially negligible. I say that as someone whose professional identity was in some significant sense tied to this particular theory. I also say this as someone who has, in the past taught history of psychology to undergraduates, and who has a keen interest in my area's history. The conditions that made a link between short-term exposure to weapons and the very mundane aggression we can observe in the lab are ones in which there was both an increase in violent media consumption (in which weapons were ubiquitous) and an increase in real life violence (something Berkowitz goes into in a 1968 paper) seemed plausible. Since that time, the concept of media violence and real life violence has been effectively debunked. Whether or not a weapons effect theory holds up in the sense that, say frustration-aggression appears to hold up is questionable at best. I think a registered replication report of the original Berkowitz and LePage (1967) would be wise, assuming it were ethically and logistically doable, if for no other reason than to settle the matter once and for all.

No comments:

Post a Comment