OBJECTIVE In previous publications (12, 13) measurements are described of the dose outside the primary beam (the peripheral dose (PD)) for 60Co gamma radiation to 25 MV photons. Comparison with data published by other investigators for different treatment machines, showed good agreement. This can only be explained when the contribution to the PD of radiation leaking through and scattering from the collimator does not differ considerably between treatment machines from different manufactures, and it is the purpose of this article to investigate whether this assumption is valid. METHODS A request was sent out to all radiotherapy departments in The Netherlands and one in Belgium to measure the dose outside the primary beam for as many machines as possible. The following geometry was given: field sizes of 10 x 10 cm2 and 20 x 20 cm2 at distances of 30 and 50 cm, for collimator angles 0 degrees and 90 degrees at the standard source surface distance. This, therefore, resulted in a dataset of eight measurements per photon energy. RESULTS Data were collected for four cobalt machines and 37 linear accelerators, from seven different manufacturers. All together 56 datasets were collected for 12 different photon energies. Although the variation of the leakage radiation dose is small, there can be differences of about 50% in the collimator scatter dose between collimator angles of 0 degrees and 90 degrees, depending on the collimator design or on the design of the flattening filter. For dual energy machines with a large gap between the low and the high energy, the values for the high energy are higher by about 40%. Old cobalt machines show higher leakage radiation dose than modern ones. CONCLUSIONS Although there is no large variation in leakage radiation dose between different makes of accelerators, some show higher collimator scatter dose than others. The magnitude of the leakage radiation dose is well within regulatory limits. As the PD not only consists of a contribution from collimator-related radiation, but also of patient scatter, the differences are negligible when estimating the peripheral dose for an individual patient.