Thermoporometry is widely used for measuring pore size distribution in porous materials by differential scanning calorimetry based on melting point depression and the Gibbs-Thomson effect shown by a liquid contained in the pores. However, measurements on water-swollen cellophane showed that the shapes of heat how vs. temperature plots (and, therefore, derived pore size distributions) changed with heating rate. Results obtained in heating were the product of two sequential and overlapping effects: an endothermic melting of the smallest ice crystals within pores in the film, followed by diffusion of the water and an exothernic re-freezing on larger crystals at the film surface. This caused the volume of larger pores to be underestimated or missed completely. In contrast, di diffusion out of the smallest pores preceded freezing for measurements on cooling, so only the larger pores were observable. The rate of diffusion was estimated as 1 x 10(-13) m(2) s(-1) from the quantities of ice in pores and at the film surface observed at different heating rates. The value obtained was much slower than expected fur water-swollen cellulose, but consistent with water diffusion through dry cellulose, suggesting that the film surface had been 'freeze-dried' during the measurements.