Legalization of cannabis for recreational use in many states created a need to study cannabis effects on driving performance. A double-blind parallel-arms (PA) study was designed to compare driving performance of healthy adults after being administered one of three treatments: 0% (placebo), 5.9%, or 13.4% ?9-tetrahydrocannabinol (THC) smoked cannabis. The participants’ performance on a driving simulator is being measured prior to THC administration (baseline) and at four later time points. Power was estimated under PA design and a crossover design for testing a hypothesis that changes in driving from baseline to post-smoking (outcome) differ between doses. Subjects serve as their own control in both designs as each visit consists of a baseline measurement and follow-up measurements. A PA study with n participants per arm (total=3n) has the same power as a crossover study involving 3 visits with n participants. Thus, compared to the PA design, the crossover design has no clear advantages, and important practical disadvantages, including learning effects, participant drop-out, increased study duration for inclusion of wash-out periods.