In electron beam dosimetry the perturbation effect in the medium by the ionization chamber cavity is accounted for by introducing a replacement correction factor, P(repl). Another perturbation correction factor, denoted as P(wall), is due to the materials of the walls of the parallel-plate chamber differing from the phantom material. Because of the difficulties in separating these two components, we measure the overall perturbation factor, p(q) = P(repl)P(wall). A distinct advantage of parallel-plate ionization chambers over cylindrical chambers is that p(q) has been shown to be close to unity at the standard calibration depth, d(max). However, for many dosimetry applications it is necessary to know the overall perturbation factor at depths greater than d(max). We measured the overall perturbation factor at depths greater than d(max) (approximating the 95%, 90% and 50% depth dose) for a Farmer-type cylindrical ionization chamber and three parallel-plate ionization chambers. We assumed that p(q) for the NACP chamber is unity at these measurement depths. The depth dependence for the other chambers was then measured relative to the NACP chamber. The mean energy at depth, E(d), and percentage depth dose gradient ranges studied were 1.9-18.5 MeV and 0 to 4.5%/mm, respectively. For the other two parallel-plate chambers, we find p(q) to be unity at depths where the percent depth dose is greater than 90%, but it deviates from unity at deeper depths, where the dose gradients exceed about 2.5%/mm. For the cylindrical chamber, p(q) values at depths greater than d(max) were found to be in good agreement with those in TG 21, where the energy at depth, E(d), is used to evaluate p(q).