OBJECTIVE Determine the effective point of measurement (EPOM) of 12 thimble ion chambers, including miniature chambers and three models widely used for clinical reference dosimetry. The EPOM is the point at which the measured dose would arise in the measurement medium in the absence of the probe: For cylindrical chambers, it is shifted upstream relative to the central axis of the chamber. Although current dosimetry protocols prescribe a blanket upstream EPOM shift of 0.6r, with r as the chamber cavity radius, it has been shown in recent years that the EPOM does, in fact, depend on every detail of the chamber design and on the beam characteristics. In the wake of this finding, the authors undertake a comprehensive study of the EPOM for a series of chambers in water. METHODS This work relies on EGSnrc Monte Carlo calculations for the central axis depth dose in a water phantom and in ion chambers. They use a full Elekta Precise linac treatment head simulation to generate realistic photon beams with nominal energies of 6 and 25 MV and fields sizes of 10 x 10 and 40 x 40 cm2. The correct EPOM shift for the 12 ion chambers, modeled in realistic detail, is taken as the one minimizing the deviation of the ratio between the dose to water and the dose to the gas of the chamber cavity, according to a method proposed and validated in previous work. RESULTS The analysis reveals that the actual EPOM shift is significantly smaller than the recommended value in current dosimetry protocols, by up to 25% for reference-class chambers and 80% for miniature chambers. The location of the EPOM also depends on the characteristics of the incident beam and varies in a well-defined way with the cavity length, the central electrode radius, and the thimble wall thickness. CONCLUSIONS The authors confirm that an upstream EPOM shift of 0.6r is too large for thimble ion chambers in high energy photon beams. Proper values for the EPOM shift could be tabulated per beam and per chamber, but they envisage that a single shift for all practical beams may prove sufficient. Moreover, the systematic dependence on chamber characteristics provides evidence that a universal parametrization in terms of a few design parameters is conceivable and has implication for the calculation of chamber correction factors.