To investigate the reasons for and the utilities of tropospheric tracer correlations, we examine the interrelationship between ethane and propane on the basis of the observations at northern middle and high latitudes (TOPSE) and over the tropical Pacific (PEM-Tropics B) and the corresponding global three-dimensional chemical transport model (GEOS-CHEM) simulations. We chose to examine the correlation between propane and ethane/propane ratio because it is more sensitive to mixing and is less dependent on temperature than that between ethane and propane. We show that the correlation generally follows the one determined by chemical (loss) kinetics and that the deviation from the kinetics slope reflects the difference between ethane and propane in the effect of mixing relative to chemical loss. At northern middle and high latitudes the model is generally in agreement with the observations in February and March but simulates a wrong seasonal change of the correlation from March to May. The model appears to overestimate the transport from lower to middle latitudes and the horizontal transport and mixing at high latitudes in May. Over the tropical Pacific the model reproduces well the observed two-branch structure of the correlation. The discrepancy between observed and simulated correlation slope values appears to reflect an underestimate of continental convective transport at northern middle latitudes and an overestimate of latitudinal transport into the tropics. In addition, we show that the correlation can be used to define the subset of observations, in which the coupling between chemistry and transport is simulated reasonably well in the model. Using the subset of observed and simulated data for inverse modeling would reduce (systematic) biases introduced by systematic model transport errors. On the basis of this subset (March at middle latitudes and February--April at high latitudes), we find that the model underestimates the emissions of ethane and propane by 14 ¿ 5%.