Deadly crashes involving pot are on the rise. But stopping stoned drivers won't be easy.

To put it simply: proving that someone is driving stoned is a thornier problem than determining that a driver has had too much to drink. The body metabolizes pot in a way that makes it nearly impossible for scientists to agree on an appropriate legal limit for motor-vehicle operation, let alone come up with a toxicological test–like a simple breath-alcohol test–to measure how much a driver has inhaled. While it would seem obvious that driving while stoned is a bad idea, there isn’t enough evidence to prove it. Partly because of the roadblocks that years of illegality have posed, there is a dearth of scientific research on exactly how pot impairs driving and precisely how risky it is. But it is no surprise that solving the problem is a priority for public officials, since there is evidence to suggest that driving high is a real danger. From 1999 to 2010, during a period of widespread decriminalization, the rate of drivers who died in crashes with marijuana in their system tripled, from 4% to 12%, according to a review of some 23,591 driver deaths in six states. The data does not show whether marijuana caused those crashes, but it does tell us that the number of drivers on the road with pot in their system has been rising fast and at the very least correlates with mortality. It seems, at least for people at the wheel, there may be such a thing as being too mellow.

The problem of impaired driving goes all the way back to the repeal of Prohibition in 1933. Americans saw a spike in automobile accidents due to drunken driving, prompting intoxication research that established 0.15% as the acceptable limit for alcohol in the blood while driving. Drunk-driving laws proved hard to enforce at first, with cops forced to rely on subjective signs like alcohol on the breath or a flushed face, but that changed in 1954 with the invention of the Breathalyzer.

The body’s relationship with alcohol is straightforward.

With marijuana, it is not so simple. When you smoke pot, the psychoactive ingredient, THC, spreads throughout your body but leaves your blood quickly as it is absorbed by your fatty tissues and brain, so that the amount of THC in the blood is not a good indicator of impairment. It is even possible to have less THC in your blood when the effects of pot are at their peak, usually about 10 to 30 minutes after your last puff. A recent study of 1,046 drivers in New Zealand who were killed in car accidents showed that, counterintuitively, drivers with lower levels of THC in their system were actually more likely to be responsible for a crash.

The body’s metabolism of marijuana also makes it harder to equip law-enforcement officers with a toxicological test that can give an accurate measure of impairment when the driver was on the road. As anyone who has smoked pot and taken a drug test for work knows, urine tests can detect marijuana for days–even weeks–after the last puff, especially if you are a frequent smoker.

These challenges have made it hard for state legislators to write laws that are fair or effective. It is illegal to drive under the influence of marijuana in every state. The question is: How do you prove the driver was under the influence? A handful of states–Pennsylvania, Montana, Washington, Nevada, Ohio and Colorado–have set a numeric limit for THC in the blood, ranging from 1 nanogram (a billionth of a gram) of THC per milliliter of blood to 5 nanograms. States are in a quandary: set the limit too high, like the 5-nanogram limit in Washington, and Dr. Robert DuPont, president of the Institute for Behavior and Health, will tell you it’s a “license to drive stoned.” Set it too low–like the zero-tolerance policy for THC adopted in Wisconsin and 10 other states–and marijuana lobbyists will accuse you of convicting someone who might have last smoked a month ago.

Adding to the screening controversy is a more basic debate. Scientists don’t even agree on the level of risk that marijuana poses to drivers who are high. Marijuana reduces motor coordination, slows reaction time and impairs decisionmaking, according to NIDA. Studies have also suggested that marijuana may impair peripheral vision and the ability to concentrate, two vital skills at the wheel. And yet the body of scientific work on marijuana’s effect on motor-vehicle operation is small, and even then the results are all over the map.

Other nations are groping for legal and medical standards too. New Zealand uses a system that combines subjective signs of impairment with a zero-tolerance policy. If–and only if–a Kiwi driver fails the field sobriety test, any amount of the drug in his blood is illegal. It’s an interesting idea that might help address cases in which a driver has used marijuana and alcohol in small amounts that would have little impact on their own but can be deeply intoxicating when combined. Educating the public about the dangers of driving while high will help. Many Americans don’t know that smoking pot can impair their driving or that it is illegal to drive while high.

Technology may come to the rescue. A retired Canadian police officer and a physician have teamed up to create a new Breathalyzer that will detect marijuana use in the past two hours, a decent measure of impairment since pot’s effects are usually felt for about two to three hours after use. Co-developer Dr. Raj Attariwala says the technology will get more precise as legalization spurs research. “In the 1970s, we didn’t know a lot about alcohol either, so that’s basically where we are,” he says. “We are on the cusp. We’ve got a lot to learn.”

More From TIME Magazine: