Purpose: Commercial EPIDs are normally used in indirect detection mode (iEPID) where incident x-ray photons are converted to optical photons in a phosphor scintillator, which are then detected by a photodiode array. The EPIDs are constructed from a number of nonwater equivalent materials which affect the dose response of the detector. The so-called direct detection EPIDs (dEPIDs), operating without the phosphor layer, have been reported to display dose response close to in-water data. In this study, the effect that different layers of materials in the EPID have on the dose response was experimentally investigated and evaluated with respect to changes in field size response and beam profiles. Methods: An iEPID was disassembled and the different layers of materials were removed or replaced with other materials. Data were also obtained on and off the support arm and with a sheet of opaque paper blocking the optical photons from the gadolinium oxysulfide (Gd S O:Tb) phosphor layer. Field size response was measured for field sizes ranging from 2×2 to 25×25 cm2, and profiles for the 25×25 cm2 beams were extracted from the data. Results: The iEPID configuration was found to be very sensitive to backscatter. The increases in output with solid water backscatter compared to the no backscatter case were 14.7% and 6.6% at the largest field size investigated for the 6 and 18 MV beams, respectively. The Gd S O:Tb phosphor layer had a large influence on field size response as well as beam profiles for 6 MV photons, while no major effects were observed for the 18 MV beam. For 18 MV large differences in dose response were found when the standard 1 mm Cu buildup was changed for dmax equivalent Cu or solid water buildup, indicating that head scatter largely influences dose response for this energy. When the optical photons originating in the Gd S O:Tb layer were blocked from reaching the photodiodes, both field size output data and beam profiles corresponded well with data obtained in the dEPID configuration as well as reference ion chamber data for both energies. Conclusions: As expected, changing the layers of material in the EPID had a dramatic effect on dose response, which was often quite complex. For 6 MV, the complex dose response is mainly caused by the optical photons from the Gd S O:Tb layer, while insufficient filtering of scattered radiation largely affects the dose response for the 18 MV beam. The iEPID was also found to be very sensitive to backscatter for both energies. Blocking the optical photons created in the Gd S O:Tb layer essentially changed the iEPID configuration into the dEPID configuration, thus demonstrating great potential for a system that can be optimized for both imaging and dosimetry. © 2009 American Association of Physicists in Medicine.