A network of multilevel samplers was used to evaluate the spatial patterns in contaminant extraction during an in situ cosolvent flushing field test. The study was conducted in an isolation test cell installed in a fuel contaminated site at Hill Air Force Base, Utah. Partitioning tracer tests, conducted before and after the cosolvent flush, were used to estimate the spatial distribution of nonaqueous phase liquids (NAPL) and the effectiveness of cosolvent flushing for removing NAPL. Samples collected during the cosolvent flushing test were used to visualize the extraction process. The results of these two analyses showed similar spatial trends in mass removal and were in general agreement with observations based on soil core data. In general, the cosolvents were more effective in the upper portion of the flow domain and had slightly lower mass removal effectiveness in the lower portion of the flow domain. In this region, tracers indicated slower transport rates and higher NAPL saturations. The spatial analysis also indicated that cosolvent was trapped in the capillary fringe increasing the time required to displace the cosolvent from the aquifer. These results demonstrate the value of spatial information for performance assessment and improving in situ flushing design strategies. ¿ 1998 American Geophysical Union