Non-random chromosomal translocations are specifically involved in the pathogenesis of many non-Hodgkin's lymphomas and have clinical implications as diagnostic and/or prognostic markers. Their detection is often impaired by technical problems, including the distribution of the breakpoints over large genomic areas. This study reports a fluorescence in situ hybridization (FISH) method which allows the detection of specific chromosomal breakpoints in tissue sections from routinely fixed, paraffin-embedded samples. Hybridization was performed after demasking the DNA. Previously validated locus-specific probes (cosmids, PACs) flanking the BCL1, BCL2 regions and similar new probes for the MYC breakpoint region were used. The cases studied were five mantle cell lymphomas (MCL) and five follicular lymphomas (FL), selected on the basis of a previously proved t(11;14) and t(14;18) and five randomly chosen Burkitt's lymphomas (BL), as well as 21 negative control samples. In all samples, hybridization signals of sufficient intensity were obtained. Three different algorithms were used to score the hybridization signals in tissue sections, two of them taking into account the nuclei and their signal distribution indicative of chromosomal break, and one only considering the colocalization or segregation of the signals. In control tissues, these algorithms resulted in cut-off levels of 9.1%, 1.3%, or 10.0%. In the 15 lymphoma samples the percentages of abnormal cells/signals ranged from 28% to 80%, 13% to 49%, and 40% to 70%, respectively. The results indicate that small locus-specific probes can be used in FISH for regular detection of translocation breakpoints on routine paraffin tissue sections. Copyright © 2002 John Wiley & Sons, Ltd.