A theory is presented for the coupling between an ocean bottom seismometer (OBS), the sediments upon which it rests, and the surrounding water. Assuming that rotational and tilt effects are negligible (or have been made negligible through instrument design), the response of the OBS to forced harmonic motion is considered in both horizontal and vertical directions. Under these conditions it is concluded that the measured ratio of velocities when the OBS is on the seabed and when it is freely suspended in water (for an identical force) completely characterizes the OBS/seabed interaction. This enables the velocity transfer functions to be directly calculated without recourse to a detailed model of sediment/structure interaction. An OBS was designed and constructed with good coupling to seabed motion and reduction of rocking effects as principal design criteria, including an onboard shaker to conduct in situ coupling experiments. The amplitude and phase of coupling data collected on a clay seabed provide transfer functions due to horizontal and vertical seabed motion and horizontal water motion. In addition, a simple mass-spring-dashpot model of OBS/seabed interaction permits the analysis of amplitude-only coupling data. Good vertical coupling can be achieved over a wide bandwidth by designing the sensor package to have a large hydrodynamic added mass in the vertical. Good coupling to horizontal seabed motion is more difficult to achieve but is possible within a limited bandwidth, even on very soft seabeds. Finally, an example of seismoacoustic noise is presented. Hydrophone signals are compared with horizontal geophone signals received from sources within the ocean and within the seabed, and the differences are explained in terms of the coupling transfer functions. ¿ 1998 American Geophysical Union