Dynamic load induced in the vertical direction by a single person walking across a footbridge structure is usually modelled as a resonant harmonic force. A comparison between modal responses generated by this harmonic force and a walking force measured on a treadmill confirmed that using the harmonic force is justifiable in the case of walking across a structure that does not vibrate perceptibly. It is further shown that using this harmonic force for predicting the response of a perceptibly vibrating footbridge can significantly overestimate the real footbridge response and therefore its degree of liveliness. This is because of the inability of pedestrians to keep pacing steadily when perceiving strong vibrations. A methodology for systematic comparison of the measured and simulated responses is proposed with the aim of identifying vibration levels which disturb normal walking. On two footbridges investigated this level was 0·33 m/s2 and 0·37 m/s2, in both cases being lower than the levels allowed by the current British footbridge design code. Finally, the observed effect of losing steady walking step can be expressed via either increased damping or modified harmonic dynamic force over time periods when test subjects lose their step. The average increase in the damping over the analysed time periods of imperfect walking was up to 10 times higher than the damping of an empty footbridge.