In fast neutron therapy, the relative biological effectiveness (RBE) of a given beam varies to a large extent with the neutron energy spectrum. This spectrum depends primarily on the energy of the incident particles and on the nuclear reaction used for neutron production. However, it also depends on other factors which are specific to the local facility, eg, target, collimation system, etc. Therefore direct radiobiological intercomparisons are justified. The present paper reports the results of an intercomparison performed at seven neutrontherapy centres: Orléans, France (p(34)+Be), Riyadh, Saudi Arabia (p(26)+Be), Ghent, Belgium (d(14.5)+Be), Faure, South Africa (p(66)+Be), Detroit, USA (d(48)+Be), Nice, France (p(65)+Be) and Louvain-la-Neuve, Belgium (p(65)+Be). The selected radiobiological system was intestinal crypt regeneration in mice after single fraction irradiation. The observed RBE values (ref cobalt-60 gamma-rays) were 1.79 +/- 0.10, 1.84 +/- 0.07, 2.24 +/- 0.11, 1.55 +/- 0.04, 1.51 +/- 0.03, 1.50 +/- 0.04 and 1.52 +/- 0.04, respectively. When machine availability permitted, additional factors were studied: two vs one fraction (Ghent, Louvain-la-Neuve), dose rate (Detroit), influence of depth in phantom (Faure, Detroit, Nice, Louvain-la-Neuve). In addition, at Orléans and Ghent, RBEs were also determined for LD50 at 6 days after selective abdominal irradiation and were found to be equal to the RBEs for crypt regeneration. The radiobiological intercomparisons were always combined with direct dosimetric intercomparisons and, when possible in some centres, with microdosimetric investigations.