A series of model tests was conducted in a geotechnical centrifuge to investigate the variation of suction pressure during caisson installation in dense sand. The tests were carried out for caissons of different size and surcharge (or effective weight). Finite element (FE) simulations were also conducted to study and explain the experimental results. Data from two field installations and other model tests were investigated, and compared with the results from the centrifuge model tests. The suction pressure was observed to increase relatively linearly with embedment depth, following a distinct slope that is related to establishing a critical hydraulic gradient along the inner caisson wall. The total driving force (weight plus pressure differential) required for suction installation is significantly reduced compared with the force required to install the caisson by jacking. When a larger surcharge was used, the required suction pressure decreased for a given penetration depth, but the total force increased. By contrast, the initial caisson penetration depth did not appear to have any effect on the required suction pressure. The overall behaviour and the pressure variation with depth were similar for caissons of different sizes, and wall thickness. Subsequent FE analyses appeared to explain this observation well, and a consistent pattern of behaviour was established for all the model and field data.