In the late 1970s, Johansson et al (1978 Int. Symp. National and International Standardization of Radiation Dosimetry (Atlanta 1977) vol 2 (Vienna: IAEA) pp 243–70) reported experimentally determined displacement correction factors (pdis) for cylindrical ionization chamber dosimetry in 60Co and high-energy photon beams. These pdis factors have been implemented and are currently in use in a number of dosimetry protocols. However, the accuracy of these factors has recently been questioned by Wang and Rogers (2009a Phys. Med. Biol. 54 1609–20), who performed Monte Carlo simulations of the experiments performed by Johansson et al. They reported that the inaccuracy of the pdis factors originated from the normalization procedure used by Johansson et al. In their experiments, Johansson et al normalized the measured depth–ionization curves at the depth of maximum ionization for each of the different ionization chambers. In this study, we experimentally investigated the effect of air cavity size of cylindrical ionization chambers in a PMMA phantom and 60Co γ-beam. Two different pairs of air-filled cylindrical ionization chambers were used. The chambers in each pair had identical construction and materials but different air cavity volume (diameter). A 20 MeV electron beam was utilized to determine the ratio of the mass of air in the cavity of the two chambers in each pair. This ratio of the mass of air in each pair was then used to compare the ratios of the ionizations obtained at different depths in the PMMA phantom and 60Co γ-beam using the two pairs of chambers. The diameter of the air cavity of cylindrical ionization chambers influences both the depth at which the maximum ionization is observed and the ionization per unit mass of air at this depth. The correction determined at depths of 50 mm and 100 mm is smaller than the correction currently used in many dosimetry protocols. The results presented here agree with the findings of Wang and Rogers’ Monte Carlo simulations and show that the normalization procedure employed by Johansson et al is not correct.