Sample preparation remains a long and critical step for organic contaminant analyses in the environment. Classical extraction methods, like Soxhlet extraction, are time and solvent consuming. Some new techniques have been developed in the last few years (supercritical fluid extraction, microwave-assisted extraction, pressurized liquid extraction). Among them, microwave-assisted extraction at atmospheric pressure has appeared an interesting alternative method to Soxhlet extraction. In this present study the extraction of polycyclic aromatic hydrocarbons (PAHs) from solid environmental matrices (soils and sediments) by a microwave-assisted method has been studied. In a first part the experimental conditions have been optimised and different parameters adjusted using the SRM 1941a as a test matrix: time and power of irradiation, nature of solvent, percentage of water. Among all the parameters, the quantity of water is of primary importance to maximize the recoveries. With the optimised conditions applied to the SRM 1941a, the results obtained by the microwave-assisted extraction with 30% of water are in a good agreement with those obtained by Soxhlet. When compared to the certified values, the global recovery obtained by microwave-assisted extraction with 30% of water is 90%. The relative standard deviations obtained for the microwave-assisted extraction with 30% of water are in the same range as those given by the Soxhlet extraction and in good agreement with the certified values (relative standard deviations below 15%). The optimal conditions (30% of water, 30 ml of dichloromethane, 30 W, 10 min) have then been applied to the extraction of PAHs from various matrices (certified and natural ones; soils, sediments, sewage sludge) and the concentrations of PAHs obtained by microwave-assisted extraction are compared to those obtained by Soxhlet extraction. For all the samples tested in this study the recoveries for the microwave-assisted extraction using 30% of water are very good (more than 85%) in comparison to the Soxhlet extractions. The reproducibility of such an optimised procedure is also very correct (under 15% for ΣPAHs for all the matrices). The method developed has appeared to be not matrix-dependent.