One of the most frustrating features of joint mechanics is that all the important processes take place precisely where they cannot be seen or measured directly — the interface between contacting bodies. In order to achieve some insight into the mechanisms that give rise to the nonlinearities of joints one naturally turns to analytic or numerical models of interface mechanics. With such models, one can explore the significance of different assumptions of kinematics or surface mechanics and compare those with laboratory experiments on the integrated joints. Among the limitations of such modeling strategies are the twin problems of 1) employing suitable models for friction and 2) solving the resulting equations. There is evidence that the commonly used Coulomb friction model is inconsistent with the experimentally observed behavior of lap joints; it is necessary to explore the use of more complex models. Additionally, even when computing contact and sliding with the relatively simple Coulomb friction model, capturing the evolution of traction fields from one load set to the next in a physically plausible manner has been a continuing challenge. Obtaining fidelity to such path dependence for more complex models would be consequently more difficult. This issue has motivated the research reported here on the source of the difficulty in modeling path-dependent contact and possible solutions. A two-parameter Coulomb friction model is used to test a conventional contact algorithm and a newer one devised specifically to capture path dependence correctly. The evolution of lateral traction during cyclic loading is used to illustrate how the shear traction distribution at each load step evolves from that of the previous.