This paper introduces a new method for imaging lateral variations in the seismic velocity structure of the earth. The discussion is centered around the geometry of a pilot experiment conducted in Salt Creek vally near Bloomington, Indiana, but the methodology is more general in scope. In the pilot experiment, 24 explosions were fired at equal intervals around a circular area 190 m in diameter and recorded by geophones positioned diametrically opposite the source. Travel time residuals for the fan shots are inverted to estimate lateral velocity variations in a two-dimensional, bowl-shaped image reconstruction region under the circular array. A simple damped least squares inversion worked poorly when delay times were included as additional free parameters in the solution. A parameter separation procedure was more successful. The value of these data in determining structure was analyzed using synthetic data and resolving power calculations. Structure could be determined to high accuracy with little distortion in the center of the circular region where rays crossed from all directions, but results were comparatively poor near the fringes of the region where angular coverage was more limited. Inversion of the Salt Creek data indicates the observed variations in the residuals can be almost completely accounted for by variations in the weathered layer. The refractor velocity is nearly constant to a precision of 0.005 s/km, but there is a suggestion of a slight velocity decrease in the refractor at higher elevations above the water table.