Determination of the physical properties and the geometry of features within the subsurface is a common application of active source seismic data. We use the example of unusually bright reflections identified in the lower crust of the Coast Ranges of northern California during the Mendocino Triple Junction Seismic Experiment to investigate the robustness of such determinations. The sources of these reflections are significant because they lie in a region of transition between subduction and strike-slip tectonics, but they also have an overburden of heterogeneous Franciscan terrane rocks. We use finite difference synthetic seismograms including a stochastic overburden representative of Franciscan rocks and a realistic target geometry to examine how typical seismic diagonostics are affected under these conditions. The addition of a realistic overburden to the models replicates variations in the amplitude of the observed Pg arrival and leads to variations in measures of the amplitude of the reflected arrivals that are also similar to those in the data. Some of these variations are due to the measures of amplitude used, some represent real variations at the target, and others are due to distortion of the illumination field by the overburden. Overall reflection amplitudes are reduced due to transmission losses within the overburden, and the P wave seismic velocity within the reflectors may be as low as 2.5 km s-1 rather than the 3.5 km s-1 estimated when the overburden was not taken into account. Migrations of simulations, including combinations of overburden, target dip, and uneven spatial sampling, recover somewhat consistent but inaccurate approximations to the true target geometry: Discontinuous migrated images may result from a discontinuous wave field even where reflectors were continuous. The simulations also imply that neither reversed reflection polarity nor converted shear waves are a reliable aspect of the reflected wave field if the overburden or target is complex. Comparison of migrated simulations to the data suggests that the reflections originate from two relatively continuous layers at the top and base of the lower crust, 300 m and 100--150 m in thickness, respectively. We believe that the reflections are from lenses of basaltic melt recently emplaced in the crust and generated by decompression melting in a slab-free window. ¿ 2000 American Geophysical Union