The objectives of our study were to assess the feasibility of using ground-penetrating radar (GPR) to study roots over a broad range of soil conditions in the southeastern United States. Study sites were located in the Southern Piedmont, Carolina Sandhills and Atlantic Coast Flatwoods. At each site, we tested for selection of the appropriate antenna (400 MHz versus 1.5 GHz), determined the ability of GPR to resolve roots and buried organic debris, assessed root size, estimated root biomass, and gauged the practicality of using GPR. Resolution of roots was best in sandy, excessively drained soils, whereas soils with high soil water and clay contents seriously degraded resolution and observation depth. In the Carolina Sandhills, 16 1 x 1-m plots were scanned with the 1.5 GHz antenna using overlapping grids. Plots were subsequently excavated, larger roots (> 0.5 cm diameter) sketched on graph paper before removal, and all roots oven-dried, classified by size and weighed. Roots as small as 0.5 cm in diameter were detected with GPR. We were able to size roots (0.5 to 6.5 cm in diameter) that were oriented perpendicular to the radar sweep (r(2) = 0.81, P = 0.0004). Use of image analysis software to relate the magnitude of radar parabolas to actual root biomass resulted in significant correlations (r(2) = 0.55, P = 0.0274). Orientation and geometry of the reflective surface seemed to have a greater influence on parabola dimensions than did root size. We conclude that the utility of current GPR technology for estimating root biomass is site-specific, and that GPR is ineffective in soils with high clay or water content and at sites with rough terrain (most forests). Under particular soil and site conditions, GPR appears to be useful for augmenting traditional biomass sampling.