A method to assess micronucleus (MN) induction in Vicia faba roots by direct contact exposure to a solid matrix was developed. The procedure comprised a 5-d germination period, as in the well-known method using aqueous extracts. However, the seeds were here sown directly into the test soil whereas a culture period is necessary before exposing seedlings to a liquid medium. One soil under forest and two contaminated soils from areas affected by industrial installations and a coke works were used. Three durations of direct exposure were tested: 2, 5 and 7 d. The optimal duration was evaluated at 2 d to observe maximal MN induction without observing toxicity symptoms. The methodology using aqueous extracts was applied to the same three soils: MN frequency was higher than in the direct contact assay but the ratios of MN frequencies from tested soils in comparison to the negative control were lower. However, for each soil, both the direct contact method and the aqueous extract exposure led to the same risk assessment diagnosis. The evaluation of a concentration range of a polycyclic aromatic hydrocarbons (PAH)-contaminated soil showed a dose-dependent MN frequency when the seeds were allowed to germinate before sowing in the soil: the soil genotoxicity was the highest at intermediate doses. The direct contact method was found to be rapid, sensitive and well suited to the evaluation of soil quality.