Yes, this is a key problem with the study. THC metabolites are cleared very slowly. Testing positive indicates that you used marijuana within a week or two. It doesn't mean that you're stoned.

For alcohol, a comparable test might look at long-term changes in liver biochemistry associated with alcohol consumption. In that case, there probably wouldn't be any correlation with accident rates.

Blood tests can directly measure THC. The article seems to be talking about such tests.

The problem is that there is not a well established mapping between THC levels and impairment. It isn't even clear if such a mapping can be done.

Maybe there's a key distinction between THC levels in the blood and bioavailable THC. THC is extremely lipophilic, so it might get trapped for degradation.

> If a breathalyzer test on Tuesday picked up the beer you had last Saturday, there'd be no correlation between that test and accidents, either.

Not sure that's true: if the alcohol was still in your blood that long, it would still be affecting you and impairing function.