Carcinogenesis is a multi-stage and prolonged process. At the present time, our knowledge of biological activities along the process is incomplete, therefore, a variety of experimental data are used to assess health risk from exposure to environmental chemicals. However, experimental approaches may not be adequate unless human data are available to support the assessment. In this brief review, benzene (CAS No. 71-43-2), a well-established human leukemogen, will be used as an example to illustrate the challenge in assessing toxicological mechanisms and cancer risk. Benzene has been shown to form DNA-adducts in experimental animals but the adducts have proved elusive of detection in human. Several toxic metabolites of benzene have been identified but the metabolite(s) responsible for the carcinogenic activities is unknown. Furthermore, the significant differences between rodents and human in response to benzene exposure are not understood. Therefore, the bone marrow specificity for the induction of leukemia in human by benzene remains to be elucidated. These complications illustrate the complexity of the assessment process and identify serious information gaps. These information gaps can be viewed as research opportunities to provide more precise data for assessment of toxicological effects and health risk.