The overall predictability of smoke exposure indicators and the importance of different influencing factors were assessed in a cross-sectional study (n = 144), using multiple linear regression and bivariate correlation analyses. Respiratory CO, and plasma nicotine and cotinine concentrations were measured before and after smoking, for lip or holder smoking, and natural or standardized (30 puffs) puffing. The prediction of smoke exposure measures varied considerably across sampling times, smoking conditions, and dependent variables. The variance of plasma cotinine and nicotine were predictable to a considerable extent (30%; 19-41%) by cigarette yield, consumption and self-reported inhalation, whereas respiratory CO was less predictable (15-27%). Generally, consumption was the most important predictor, surpassed by nicotine yield for post-smoking plasma nicotine. Smoke exposure from a single smoking period could be predicted to a variable degree (CO, 11-42%; nicotine, 33-54%) by a subset of smoker's sex, cigarette yield, self-reported inhalation and puffing characteristics. The highest prediction was found under standardized smoking conditions (30 puffs through a holder), the lowest under natural smoking conditions. The best subset of predictors, especially with respect to puffing parameters, was found to vary considerably across smoking conditions and dependent variables.