We analyse seismic polarisation of direct and reflected PS arrivals on multi-component ocean bottom (OB) data, by mapping vector fidelity as a function of azimuth and angle of incidence at the seabed. We compare data from a cable system with data from nodes, as well as buried and unburied cable sensors. As expected, nodes show overall best vector fidelity. For OB cable data, on the other hand, buried sensors show significantly better vector fidelity than unburied sensors, both for down-going direct P waves and up-going PS reflected waves. While both buried and unburied cable sensors exhibit vector infidelity to some degree in the direction along the cable, unburied sensors show in addition poor fidelity in the crossline direction. Vector infidelity of the unburied sensors may be misinterpreted as azimuthal anisotropy with a symmetry axis parallel to the cable.