Effective dose, an indicator of the stochastic effect of radiation, has been widely used in dose evaluation in the environment. Though conversion factors have been used to obtain E from the air kerma or air absorbed dose, the variation of the conversion factors due to the change of exposure conditions has not been sufficiently investigated. This report documents an investigation of the variation of the effective dose per air kerma for environmental gamma rays depending on the exposure conditions using anthropomorphic phantoms and Monte Carlo calculations, taking into account the precise angular and energy distributions of the environmental gamma rays incident on the human body. As causes of the variation, posture of human bodies, biases of environmental source distributions, and body size were considered. The variation of effective dose in a prone position compared with that in a standing position was found to be within 30%. The bias of environmental sources causes the effective dose per air kerma to vary by 20% at maximum, but in some cases for low-energy gamma rays the variation was found to be up to 40% due to the change in the energy spectrum. The effective dose for a new born infant was estimated to be higher than that for an adult by a maximum of 80-90% for low-energy gamma rays from anthropogenic sources because of a lower shielding effect of the smaller body. The variation of the effective dose equivalent shows a similar tendency to the effective dose. Consequently, this study made it possible to estimate the uncertainties of effective dose and effective dose equivalent evaluated from air kerma or absorbed dose in air using the standard available conversion factors.