Although the prognosis of Hodgkin's disease is relatively good, around 20% of patients do not benefit from current therapies and succumb to their disease. A large-scale molecular characterization of disease might help improve HD management. Using cDNA arrays, we studied the mRNA expression levels of approximately 1000 selected genes in 34 benign and malignant lymphoid samples including 21 classical Hodgkin's disease (HD) tissue samples. Hierarchical clustering identified three main molecular groups of HD tumours relevant with respect to histology and clinical outcome (response to therapy and survival). Samples from all bad outcome HD (BOHD) patients clustered in one group whereas the two other groups contained most good outcome HD (GOHD) cases. The nodular sclerosis GOHD samples overexpressed genes involved in apoptotic induction and cell signalling, including cytokines, while the BOHD samples were characterized by the upregulation of genes involved in fibroblast activation, angiogenesis, extracellular matrix remodelling, cell proliferation, and the downregulation of tumour suppressor genes. Our results establish a molecular taxonomy of HD correlating with response to therapy and clinical outcome, thereby suggesting the possibility of improving the current prognostic classification.