The time term method of seismic refraction data analysis allows an areal arrangement of shot points and stations and yields directly a three-dimensional representation of underlying structure. This method was investigated analytically and through numerical experiments to determine implicit sources of error and to provide guidelines for its proper use. Error terms for computed refractor depths were derived for two site arrays over two structures, a simply dipping refractor surface and a symmetric anticlinal refractor surface. These error terms proved to be qualitatively reliable guides to the performance of the time term method using models of multi site arrays over planar and biplanar refractor structures. Neither continuity of time terms nor use of a known refractor velocity in the computations offered a guarantee of a more accurate refractor reconstruction. In the two models tested a refractor velocity greater than the true velocity yielded the best reconstruction. Sensitivity to angle of dip indicated preferred orientation of shot-station lines along gentle apparent dips when the underlying structure is known and their orientation along several azimuths when the structure is unknown so as to mute the effect of high dip angles. If the structure has refractor surfaces dipping more than a few degrees, the time term method will give inaccurate results unless previous knowledge of the refractor configuration permits model studies to determine error terms and arrangement of shots and stations. Unfortunately such knowledge is usually the goal of a seismic experiment.