We report on the outcomes of a lysimeter experiment aimed at the measurement of travel time distributions of water and certain nonreactive solutes under non-stationary conditions to examine the kinematics of age mixing. In order to simulate the release of a compound in a receiving water body, it is common in hydrology to attribute a travel time probability distribution to each particle, which reflects the response of a catchment unit to a solute input. Hence, the concentration measured at a control section becomes the convolution between the travel time distribution and the concentration of the inputs throughout the past. This study aims at experimentally demonstrating that the tracer travel time probability distribution is, in fact, strongly dependent on the antecedent conditions at the time of tracer injection and the subsequent states experienced in the system. It is therefore a function of numerous transient processes such as hydrologic filtering in soils, climatic forcing or evapotranspiration patterns. A 2-meter deep weighing lysimeter was equipped with a discharge measurement system coupled with a sample collector, an array of water content sensors and an array of porous cups for soil water sampling at three different depths. Controlled random rainfall following a Poisson process was generated, and evapotranspiration losses from two willow trees planted in the lysimeter created an important soil-water storage deficit. Five species of fluorobenzoic acids were used as tracers, and sequentially injected through rainfall at different times. The measurement system installed allowed a precise and accurate monitoring of every input and output flux and water storage, which is crucial to determine the conditions influencing the travel time distribution and to calculate the mass loads and recovery rates. Breakthrough curves for multiple tracers measured at several depths within the lysimeter and at the lysimeter outlet provide support for non-stationary tracer travel time distribution. Additional data are being collected, and will be compared with an exact solution of non-stationary distribution assuming random sampling of ages.