Models must effectively represent velocities and celerities if they are to address the old water paradox. Celerity information is recorded indirectly in hydrograph observations, whereas velocity information is more difficult to measure and simulate effectively, requiring additional assumptions and parameters. Velocity information can be obtained from tracer experiments, but we often lack information on the influence of soil properties on tracer mobility. This study features a combined experimental and modelling approach geared towards the evaluation of different structures in the multiple interacting pathways (MIPs) model and validates the representation of velocities with laboratory tracer experiments using an undisturbed soil column. Results indicate that the soil microstructure was modified during the experiment. Soil water velocities were represented using MIPs, testing how the (a) shape of the velocity distribution, (b) transition probability matrices (TPMs), (c) presence of immobile storage, and (d) nonstationary field capacity influence the model's performance. In MIPs, the TPM controls exhanges of water between pathways. In our experiment, MIPs were able to provide a good representation of the pattern of outflow. The results show that the connectedness of the faster pathways is important for controlling the percolation of water and tracer through the soil. The best model performance was obtained with the inclusion of immobile storage, but simulations were poor under the assumption of stationary parameters. The entire experiment was adequately simulated once a time-variable field capacity parameter was introduced, supporting the need for including the effects of soil microstructure changes observed during the experiment. © 2019 The Authors Hydrological Processes Published by John Wiley & Sons Ltd