Twenty-seven laboratory experiments were conducted in a simulated smoking room to quantify rates of environmental tobacco smoke (ETS) leakage to a nonsmoking area as a function of the physical and operational characteristics of the smoking room. Data are presented for the various types of leakage flows, the effect of these leaks on smoking room performance and nonsmoker exposure, and the relative importance of each leakage mechanism. The results indicate that the first priority for an effective smoking room is to maintain it depressurized with respect to adjoining nonsmoking areas. The amount of ETS pumped out by the smoking room door when it is opened and closed can be reduced significantly by substituting a sliding door for the standard swing-type door. An "open doorway" configuration used twice the ventilation flow of those with smoking room doors, but yielded less reduction in nonsmoker exposure. Measured results correlated well with results modeled with mass-balance equations (R(2) = 0.82-0.99). Most of these results are based on sulfur hexafluoride (SF(6)) tracer gas leakage. Because five measured ETS tracers showed good correlation with SF(6), these conclusions should apply to ETS leakage as well. Field tests of a designated smoking room in an office building qualitatively agreed with model predictions.