The performance of activated carbon in soil washing and subsequent selective adsorption for surfactant recovery from the washed solution was investigated. Sandy loam soil contaminated with phenanthrene at 200 mg kg(-1) was washed with four different nonionic surfactants: Tween 40, Tween 80, Brij 30 and Brij 35. The efficiency of soil washing was highest when using Brij 30 with the highest solubilizing ability for phenanthrene and low adsorption onto soil. In the selective adsorption step, surfactant recovery was quite effective for all surfactants ranging from 85.0 to 89.0% at 1 g L-1 of activated carbon (Darco 20-40 mesh). Phenanthrene removal from the solution washed with Brij 30 was only 33.9%, even though it was 54.1-56.4% with other surfactants. The selectivity was larger than 7.02 except for Brij 30 (3.60). The overall performance considering both the washing and surfactant recovery step was effective when using Tween 80 and Brij 35. The results suggest that higher solubilizing ability of surfactants is a requirement for soil washing but causes negative effects on phenanthrene removal in the selective adsorption. Therefore, if a surfactant recovery process by selective adsorption is included in soil remediation by washing, the overall performance including the two steps should be considered for properly choosing the surfactant. (c) 2007 Elsevier B.V. All rights reserved.