The depth dose of the primary dose component, on axis and off axis of six different x-ray beams, has been determined from transmission measurements in narrow beam geometry with and without flattening filter using a Perspex column of a cross section large enough to ensure electronic equilibrium. In order to derive the primary photon fluence, a correction for the scatter from the column has been applied according to the following method: A number of spectra taken from the literature have been used for computing a scatter coefficient Sc at different depths by convolution of dose spread arrays. Using the relationship between Sc and the single attenuation coefficient mu i to represent each entire spectrum, it has been possible to correct the experimental transmission curves iteratively, until the corresponding values of mu were stabilized and representative of the primary. The measured attenuation coefficients were found to have a linear increase as a function of the distance from the central axis for all the energies and types of linear accelerators. For the same nominal energy, this increase is different from one accelerator to another. The same phenomenon was observed for the attenuation coefficients obtained without the flattening filter in the same experimental conditions. The results are tentatively interpreted considering the angular variation of bremsstrahlung energy spectra with and without a flattening filter as calculated by a Monte Carlo method and they are consistent and useful to take accurately into account the softening of the beam as the off-axis distance increases.