Nuclear power plants are responsible for the spent fuel management. Closely spaced racks submerged in a pool are generally used to store and to cool the nuclear fuel. A free-standing design allows to isolate the rack base from the pool floor and therefore to reduce the impact of seismic loads. However, the seismic response of free-standing racks is difficult to predict accurately using theoretical models given the uncertainties associated with inertial forces, geometrical nonlinearities and fluid-structure interactions. An ad-hoc analysis methodology has been developed to overcome these difficulties in a cost-effective way, but some dispersion of results still remains. In order to validate the analysis methodology, experimental tests are carried out on a scaled 2-rack mock-up equipped with fake fuel assemblies. The two rack units are submerged in free-standing conditions inside a rigid pool tank and subjected to accelerations on a unidirectional shaking table. A hydraulic jack induces a given acceleration time-history while a set of sensors and gauges monitor the transient response of the system. Accelerometers track the acceleration of the pool and units. Load cells measure the impact forces on the rack supports as well as the fluid forces at the centre of the rack faces. Video cameras record the transient displacements and rotations. Results provide evidence of a water-coupling effect leading to an in-phase motion of the units.