The current WHO classification of Hodgkin's lymphoma (HL) generally distinguishes the relatively rare variant (approximately 5% of all cases of HL) of nodular lymphocyte predominant type from a second group, which comprises classical HL and is separated into four subtypes: lymphocyte rich type, nodular sclerosis type, mixed cellularity type and lymphocyte depleted type. The classical lymphocyte rich subtype is a new entity and based on the typical morphology, can be recognized by definition only by the immunohistochemical characteristics of the Hodgkin and Reed/Sternberg cells (HRS) (CD30+, CD15+, CD20-). Molecular single cell studies are consistent with the dichotomy of HL in nodular lymphocyte predominant and classical types and stress the exceptional position of the former, which shows similarities with non-Hodgkin's lymphomas in several aspects. On the molecular biology level the tumor cells of all kinds of HL turn out to be clonal B cells derived from germinal center cells. However, tumor cells of nodular lymphocyte predominant HL differ from those of classical HL by the pattern of somatic mutations. Considering the inability of HRS cells of classical HL to express a B cell receptor, they should perish under normal conditions. However, they escape from apoptosis by mechanisms so far only partially understood, such as genomic mutations of the l-kappa B gene and the fas receptor gene or probably by down-regulation of B cell markers. In rare cases, HRS cells of HL can also be derived from T cells, as could be demonstrated by single cell analysis. Also, it could be shown by single cell PCR that HL and non-Hodgkin's lymphoma of both B and T cell types can arise from a common precursor. These results suggest that future classifications of HL will not only take into account the morphological and phenotypical profile, but also mechanisms of transformation yet to be discovered.