Ozone production efficiency (OPE) can be defined as the number of ozone (O3) molecules photochemically produced by a molecule of NOx (NO + NO2) before it is lost from the NOx - O3 cycle. Here we consider observational and modeling techniques to evaluate various operational definitions of OPEs using aircraft and surface measurements taken as part of the 1999 Southern Oxidant Study field campaign in Nashville, Tennessee. A key tool in our analysis is a Lagrangian box model, which is used to quantitatively describe the effects of emissions, dilution, dry deposition, and photochemistry in an urban air parcel as it was advected downwind. After evaluating the model using the observed downwind concentrations of several key species, we show that the modeled NOx oxidation and O3 production rates as well as the associated instantaneous and cumulative OPEs depend on the time of day and the photochemical age of the air parcel. The observation-based OPEs are found to be consistent with the modeled values with the expected biases. A model sensitivity study suggests that downwind O3 concentrations in the Nashville plume are more sensitive to NOx emissions than anthropogenic VOC emissions. Because the OPE exhibits a nonlinear dependence on emissions and meteorological effects, it would be difficult to rely only on observations to map out the nonlinear response of O3 to a wide span of NOx and VOC emission changes. Properly constrained and well-evaluated models using a variety of observations are therefore necessary to reliably predict O3-NOx-VOC sensitivity for designing effective O3 control strategies.