Public health researchers generally agree that on balance, marijuana is a less harmful drug than alcohol. Booze, for instance, is more addictive than pot, and from a chemical standpoint, it’s far more toxic – and hence lethal – than marijuana.

Alcohol isn’t just more harmful to individuals than marijuana is; it’s more harmful to society, too. Drunken driving is a big factor in this – studies generally show that alcohol impairs driving ability much more than just about any other drug does. Alcohol also makes people aggressive, and is a factor in roughly 40 percent of violent crimes committed today.

What about a link between marijuana use and aggression? Most pot smokers will tell you that marijuana helps them relax. The popular stereotype of a heavy marijuana user is the guy stoned out of his mind on the couch, eating Funyuns and watching cartoons.

But surprisingly, research on the link between marijuana and aggression has been mixed. Marijuana seems to make most people relaxed, but it can also cause anxiety and paranoia, conditions which can occasionally manifest themselves in violent ways. There are occasional reports out of Colorado of marijuana users causing harm to themselves or to others.

So a recent study from the Netherlands, published in the journal Psychopharmacology, attempts to put this question to bed using the gold standard of scientific research: a random controlled trial. They recruited a group of 20 heavy alcohol users (three-plus drinks a day for men, two-plus for women), 21 heavy marijuana users who smoked at least three times a week, and 20 controls who didn’t use either drug heavily at all.

They then got the alcohol users drunk until their BAC measured 0.08, the standard threshold for impairment. They got the marijuana users high, by dosing them with 300 micrograms of THC per kilogram of bodyweight delivered via a vaporizer. The control group didn’t get to do any of this fun stuff, because they were controls.

Then they made all three groups complete a number of tests designed to get people riled up. The first, known as the “single category implicit association test,” had people match positive and negative words to photos depicting aggressive and violent behavior – punching, kicking, etc. In the second test, respondents played a computer game in which they were told they could win money by pressing buttons. They were pitted against an adversary who could undermine the players by taking money from them. The players were unaware that the “adversary” was actually controlled by the computer.

The researchers measured aggression, before and after the respondents took the test, by asking them how aggressive they felt on a 100-point scale. For good measure, they had the marijuana and alcohol users go through the whole thing again one week later, this time without getting high or drunk, as a kind of separate control.

They found, first of all, that “alcohol intoxication increased subjective aggression in the alcohol group.” The alcohol users, in other words, acted more aggressive when they were drunk than they did when they were sober. By contrast, the smokers became less aggressive when they were high.

These findings held through both the self-assessments – alcohol users rated themselves as more aggressive when drunk – and through the responses to the tests: The drinkers tried harder to undermine their computer opponents when they were drunk. But the smokers actually acted less aggressive toward their computer opponents when they were high.

“The results in the present study support the hypothesis that acute alcohol intoxication increases feelings of aggression and that acute cannabis intoxication reduces feelings of aggression,” the researchers conclude.

This is in line with other research. A study in 2014, for instance, found that marijuana use among couples was linked to lower rates of domestic violence. In a fun study from the 1980s, researchers gave undergraduates varying doses of marijuana and then asked them to administer electric shocks to people in another room. The more stoned the undergrads were, the less interested they were in zapping other people.

It’s true that as with any drug, marijuana can have unpredictable effects in some people. But this study strongly suggests that those unpredictable consequences – pot-crazed men jumping off ledges or shooting themselves – are tragic outliers. Those stories grab headlines because they’re exceptional, not because they’re common.