Complex radiotherapy techniques call for three-dimensional dosimetric methods with high spatial resolution. Radiation-sensitive polymer gel systems (e.g. commercially available BANG(TM) gel), read using MRI T2 mapping, offer a promising solution. A series of calibration test tubes is traditionally used to calculate the dose delivered to a larger, differently shaped volume of gel. In this work, we investigated the implicit assumption that the sensitivity of the gel is independent of shape and size. Phantoms of different shapes and volumes, and 20 glass test-tubes, were filled with BANG3 gel. T2 mapping of gels was performed pre- and post-irradiation using a 32 echo Carr-Purcell-Meiboom-Gill sequence and single exponential fitting. Gel irradiation was performed with a 6 MV Varian 6EX linear accelerator. The T2 values of both non-irradiated and irradiated gels varied with container volume. For containers of the same shape receiving the same radiation dose, larger volumes exhibited a lower T2 value than did smaller volumes. Containers of the same volume but different shape also showed a smaller variation in response to radiation. The greatest difference in T2 values at the same dose was seen between test-tubes and larger volumes. This would imply that if test-tubes alone are used to calibrate larger volumes, then up to a 35% error could be introduced into radiotherapy plan verification. This can be reduced to <10% error if the gel volume is normalized with an external measurement device. Consequently, the traditional test-tube calibration method would be unacceptable for clinical plan verification.