You’re probably already familiar with the National Rifle Association ad that ignited a national backlash earlier this summer. It showed protesters clashing with police, rioters damaging property, a man wearing a Trump shirt lying in a hospital with bandages over his head. The images were accompanied by an ominous claim from the narrator, Dana Loesch, who tells the viewer, “They use their media to assassinate real news. They use their schools to teach children that their president is another Hitler … All to make them march … smash windows, burn cars, shut down interstates and airports, bully and terrorize the law abiding.”

Critics, including some NRA members, interpreted the ad as a thinly veiled call to violence. Loesch vehemently rejected that characterization, even as she and other NRA personalities doubled down on their attacks on “leftists,” “fake feminists,” and “the liberal propaganda machine.” The group continued the attack on the media over the weekend, recirculating a video of an angry Loesch attacking the New York Times, saying: “We’re coming for you.”

As a researcher who studies political violence, I took an interest in the controversy. I need to be very plain here: I have not performed a qualitative assessment of the NRA ad; nor am I aware of anyone who has. But stepping back from the video in question, it is worth noting that, irrespective of a group’s or individual’s motives, rhetoric and imagery have a power of their own. A body of scientific evidence indicates that certain language and images can influence some viewers’ attitudes, and even lead them to support, or be inclined to engage in, aggressive or violent political acts.

The NRA video contains two powerful messages. The first is that “real” Americans (or, what my social science colleagues would call the “in-group”) should feel aggrieved — that wrong and unfair things are happening to them. The second is that this is an “us-versus-them” situation, and the blame for the injustice against the in-group clearly falls on others (the “out-group”).

A still of a protest featured in the NRA’s video.

Decades of research show that grievances can trigger political action — both violent and nonviolent — with the support and propensity for violence especially prone to increase when grievances are severe and the out-group is perceived as being “evil” or having dissimilar values.

Studies have further shown that using in-group/out-group comparisons increases prejudice. In more extreme cases, such dynamics have been used to justify torture, conflict, and even genocide.

Of course, violence doesn’t happen every time there are grievances or when a movement uses out-group dynamics — far from it, in fact. This leaves us with an important question that my co-researchers and I recently sought to answer. In circumstances where one group has grievances against another group, what are the specific factors that make some people more likely to favor political action and even violence?

To test that, we conducted an experiment with nearly 3,000 American adults, who were given one of two accounts of an intergroup conflict in a hypothetical country of Buchara. Half of the participants read a scenario where the Estamese, a hypothetical minority ethnic group, held low-level grievance:

“You are a university student. You occasionally hear comments about your ethnicity from fellow students. But on the whole you have been accepted and have good relations with students of many ethnicities. Your professors treat you fairly. Incidents involving discrimination against Estamese students are rare. On one occasion, you witnessed a fight that started when two Estamese students were attacked by Buchari students. The police arrested the Buchari students, and left the two Estamese students alone.”

The rest of the participants read a scenario where the intergroup grievances were high:

“You are a university student. You are discriminated against because you are Estamese. Fellow students make humiliating comments about your ethnicity. Your work is as good as that of your peers, but you consistently receive lower grades. Even Estamese like you with a good education are hired last and fired first. The authorities treat Estamese citizens unfairly. For example, the police stop Estamese without any reason. Estamese are not allowed to vote or to express their political views. Buchari leaders regularly deride the Estamese as unpatriotic and ridicule Estamese culture and language.”

We then asked participants about their perceptions of the out-group and their willingness to engage in political action.

Consistent with previous research, we found that people who read about higher levels of grievance were more likely to not only justify non-violent action against the Buchari but to also say they would engage in protest and violence against them. We also provided participants with a few dozen adjectives (e.g. cruel, capable, careful) and asked the extent to which each described the Buchari.

Participants who viewed the out-group (the Buchari) as being “oppressive” also were more likely to engage in and justify protest; and those who viewed the out-group as “evil” were more likely to justify violence.

Our research clearly shows that messaging matters: When grievances are emphasized, people are more likely to engage in and support violence.

Our research was based on a controlled experiment, allowing us to tease out these relationships with considerable confidence. It is difficult to isolate the effect of any single message or event in the real world, and we do not know what impact the NRA video will have on its audience. Yet our research does suggest that rhetoric and imagery designed to stir up strong grievances and highlight an us-versus-them mentality can motivate some people to political action — including support for violence. It may sound trite, but it’s terribly true: Words matter.