The prosecution of cannabis-presence driving offences (in the absence of any behavioural evidence of impairment) is ultimately based on the assumption that there is a tight causal relationship between positive toxicology for cannabis and impairment. The main purpose of this review is to examine the evidence for that relationship. We show that most experimental studies have failed to elicit statistically-significant cannabis-induced impairments for many of their possible outcomes. And many studies failed to demonstrate any impairment at all in regular users of cannabis (because of the development of tolerance). We argue that selective reporting by researchers, editors and the media has created the false impression that the evidence for cannabis-induced impairment is strong and consistent. Human beings are ‘over-engineered’ for the psychomotor skills required to drive safely. A benchmark level of cannabis-induced impairment is therefore required to distinguish unproblematic from ‘real-world’ impairment. The conventional benchmarks of statistical significance, effect size and BAC-equivalence are shown to be inadequate. However, a benchmark in terms of 30 years of normal cognitive aging has good face validity. The recent use of cannabis is indicated toxicologically by the presence of delta-9-tetrahydrocannabinol (THC) in blood or oral fluid. Evidence is provided that most THC-positive drivers are not impaired, and certainly not meaningfully impaired. It follows that the justice of stand-alone cannabis-presence driving offences must be questioned.