The primary aim of this study was to validate predictions of human knee-joint contact mechanics derived from finite-element models of the tibiofemoral and patellofemoral joints (specifically, contact pressure, contact area and contact force) against corresponding measurements obtained in vitro during simulated weight-bearing activity. A secondary aim was to perform sensitivity analyses of the model calculations to identify those parameters that most significantly affect model predictions of joint contact pressure, area and force. Joint pressures in the medial and lateral compartments of the tibiofemoral and patellofemoral joints were measured in vitro during two simulated weight-bearing activities: stair descent and squatting. Model-predicted joint contact pressure distribution maps were consistent with those obtained from experiment. Normalized root-mean-square errors between the measured and calculated contact variables were on the order of 15%. Pearson correlations between the time histories of model-predicted and measured contact variables were generally above 0.8. Mean errors in the calculated centre-of-pressure locations were 3.1 mm for the tibiofemoral joint and 2.1 mm for the patellofemoral joint. Model predictions of joint contact mechanics were most sensitive to changes in the material properties and geometry of the meniscus and cartilage, particularly estimates of peak contact pressure. The validated FE modelling framework offers a useful tool for non-invasive determination of knee-joint contact mechanics during dynamic activity under physiological loading conditions.