A Computational Fluid Dynamics (CFD) model has been formulated to evaluate axial mixing in Taylor flow. The residence-time distribution (RTD) curves of a Taylor flow unit cell (comprised of a liquid slug and two half bubbles) is calculated numerically from a step-tracer simulation by solving the steady-state Navier-Stokes and the transient convection-diffusion equations for a variety of conditions (10 < Pe < 10(5), 10(-5) < Ca < 10(-3)). The concentration calculated at the outlet of a unit cell gives the cell-density function, from which its RTD curve can be found. Two different trends in the unit cell RTD curves were observed with Pe. At high Pe, the curves have characteristic peaks, while at low Pe only one peak is found that decays slowly. The size and separation of the peaks is affected by the Pe, film thickness (delta) over bar, and slug length, but not Re. The moments of the RTD curves are then used to assess literature unit cell models (CSTR-PFR, two-region, and the model of Thulasidas et al.) For Pe (delta) over bar > 10 the CSTR-PFR model showed the largest difference from the CFD simulations, while the model by Thulasidas et al. showed reasonable agreement. The two-region model fitted the simulations, but only for Sh values significantly different from those. found by literature correlations. For Pe (delta) over bar < 10, the CSTR-PFR model gave the best predictions compared to the numerical simulations. A method for calculating the residence-time distribution of a reactor, based on the residence-time distribution of a unit cell by means of a convolution method, was also introduced and gave results which compared very well with experimental data. The short length of the reactor used in the experiments, however, could not allow proper differentiation from other models investigated, which also showed good agreement with the experimental data. (c) 2007 American Institute of Chemical Engineers