In constructing two-dimensional layered seismic models, the most commonly used technique is to butt-join the layers with some bonding agent. Whether or not this type of bonding satisfies the "welded-contact" hypothesis which is assumed in theoretical treatments is investigated experimentally. Two pieces of a steel plate are bonded together with several agents (water, castor oil, Vaseline, epoxy resin, Devcon plastic steel, and two different solders), and the reflections of the normally incident compressional and Rayleigh waves from the interface are measured. From these, it is found that the bonds do not satisfy the "welded-contact" condition. Furthermore, comparison with the theoretically computed values demonstrates that the observed reflection coefficients are, in general, too large to be solely due to the contrasts between the elastic coefficients of the plate and the bonding materials. Imperfect adherence of the bonding agent to the edges is offered as an explanation.Theoretical curves are presented to demonstrate the significance of even the least possible effect of the bonding layer on the phase velocities of dispersive Rayleigh waves.