Many seismic problems involve interfaces. Seismic modeling provides a means of passing from the mathematical formulation to the solution of such problems when suitable analytical techniques are not available. Accurate interface formulations are necessary if modeling is to be as effective as possible in this role. Application of improved techniques has provided more accurate and detailed results than possible in the original work (Toksoz and Schwab, 1964) on model interface formulations. The present investigation was limited to butt-joined, two-dimensional models constructed with metallic sheets and epoxy bonding agents. Experimental results were first compared with those predicted by the simplest of 'welded-contact' theories--that which totally ignores the bonding layer. For a layer thickness of 0.007 cm, fair agreement was noted for frequencies below 125 kHz. New theoretical results were then predicted by taking the finite thickness and elastic properties of the bonding layer into consideration. The metal-epoxy interfaces were assumed 'welded-contacts.' Irrespective of whether the formulation specified a perfectly elastic or an anelastic bonding material, the correlation between predicted and experimental results was quite poor. The disparity was much greater, over the entire frequency range, than that obtained from the formulation which totally ignored the bonding layer. All the welded-contact interface formulations tested in this investigation were found unsatisfactory for highly accurate work. Based on the technique used to test the welded-contact assumption, a mathematically accurate description of the interface was developed.