It used to be the stuff of stoner comedies and "Just Say No" campaigns. Today, marijuana is becoming mainstream as voters across the country approve ballot questions for legalization or medical use.
In response, state governments are testing ways to ensure that the integration of this once-illicit substance into everyday life doesn't create new public health risks. These efforts are sparking a difficult question: At what point is someone too high to get behind the wheel?
The answer is complicated. Brain scientists and pharmacologists don't know how to measure if and to what extent marijuana causes impairment.
The reason: Existing blood and urine tests can detect marijuana use, but, because traces of the drug stay in the human body for a long time, those tests can't specify whether the use occurred earlier that day or that month. They also don't indicate the level at which a driver would be considered "under the influence."
"It's a really hard problem," said Keith Humphreys, a psychiatry professor and drug policy expert at Stanford University in California, the first state to legalize medical marijuana and where recreational pot use among adults became legal in 2016. "We don't really have good evidence -- even if we know someone has been using -- [to gauge] what their level of impairment is."
Marijuana is now legal for recreational use in 10 states and the District of Columbia -- including Michigan, where a ballot initiative passed in November took effect Dec. 6. In New York, the governor said Dec. 17 that legalization would be a top priority for 2019. And nearly three dozen states have cleared the use of medical cannabis.
For alcohol, there is a clear, national standard. If your blood alcohol content (BAC) is 0.08 percent or higher, you're considered cognitively impaired at a level that is unsafe to drive. Extensive research supports this determination, and the clarity makes enforcement of drunken driving laws easier.
Setting a marijuana-related impairment level is a much murkier proposition. But states that have legalized pot have to figure it out, experts said.
"You can't legalize a substance and not have a coherent policy for controlling driving under the influence of that substance," said Steven Davenport, an assistant policy researcher at the nonprofit Rand Corp., who specializes in marijuana research.
Marijuana, after all, weakens a driver's ability to maintain focus, and it slows reflexes. But regulators are "playing catch-up," suggested Thomas Marcotte, a psychiatry professor at the University of California-San Diego and one of a number of academics around the country who is researching driving while high.
States have put forth a bevy of approaches. At least five have what's called a "per se" law, which outlaws driving if someone's blood level of tetrahydrocannabinol, or THC, exceeds a set amount. THC is marijuana's main intoxicant.
Colorado, where voters approved legalization of recreational marijuana in 2012, has this type of driving law on the books. It took three years to pass amid fiery debate and deems "intoxicated" any driver who tests higher than 5 nanograms of THC per milliliter of blood.
Rhode Island, Pennsylvania and Indiana are among states that forbid driving at any THC level. Still others say drivers should be penalized only if they are impaired by the chemical -- a standard that sounds reasonable but quickly gets difficult to measure or even define.
None of these approaches offers an ideal solution, experts said.
"We're still definitely evaluating which policies are the most effective," said Ann Kitch, who tracks the marijuana and driving issue for the National Conference of State Legislatures.
States that set a THC-level standard confront weak technology and limited science. THC testing is imprecise at best, since the chemical can stay in someone's bloodstream for weeks after it was ingested. Someone could legally smoke a joint and still have THC appear in blood or urine samples long after the high passes.
There's general agreement that driving while high is bad, but there's no linear relationship between THC levels and degree of impairment. States that have picked a number to reflect when THC in the bloodstream becomes a hazard have "made it up," argued Humphreys.
"The ones who wrote [a number] into legislation felt they had to say something," he said. But "we don't know what would be the analogy. Is the legal amount [of THC] equal to a beer? Is that how impaired you are? Is it a six-pack?"
Roadside testing for THC is also logistically difficult.
Blood, for instance, needs to be analyzed in a lab, and collecting urine gets ... complicated.
In Canada, which legalized recreational pot just this year, law enforcement will test drivers with a saliva test called the Dräger DrugTest 5000, but that isn't perfect, either.
Some private companies are trying to develop a sort of breathalyzer for marijuana. But Jonathan Caulkins, a drug policy researcher at Carnegie Mellon University, said, "There are fundamental issues with the chemistry and pharmacokinetics. It's really hard to have an objective, easy-to-administer roadside test."
Some states rely on law enforcement to assess whether someone's driving appears impaired, and ascertain after the fact if marijuana was involved.
In California, every highway patrol member learns to administer "field sobriety tests" -- undergoing an extra 16 hours of training to recognize the influence of different drugs, including marijuana. Because medical marijuana has been legal there since 1996, officers are "very used" to recognizing its influence, said Glenn Glazer, the state's coordinator for its drug recognition expert training program.
That kind of training is taking off in other states, too, Kitch said. Lobbying groups such as Mothers Against Drunk Driving are pushing to bump up law enforcement training and rely on officers to assess whether a driver is impaired.
These tests, though, risk their own kind of error.
"They are subjective," Rand Corp.'s Davenport warned.
For one thing, officer-administered tests can be influenced by racial bias. Someone who has previously had poor experiences with law enforcement may also perform worse, not because of greater impairment but nervousness.
Indeed, relying on more subjective testing is in some ways the direct opposite of conventional wisdom.
"A general pattern of the last ... 40 years is to try to take human judgment out of decision-making processes when possible. Because we fear exactly these issues," Caulkins said. "The idea that you could come up with a completely objective test of performance ... is ambitious."
Researchers like Marcotte are trying to devise some kind of test that can, in fact, gauge whether someone is showing signs of marijuana impairment. But that could take years.
In the meantime, the public health threat is real. States with legalized pot do appear to experience more car crashes, though the relationship is muddled. "This is going to be a headache of an issue for a decade," Caulkins said.
Kaiser Health News is a national health policy news service that is part of the nonpartisan Henry J. Kaiser Family Foundation.