Published and previously unpublished measurements of closure rates of five boreholes in polar ice caps are reviewed. The data cover effective shear stresses between 0.15 and 1.0 MN m−2, temperatures between −16° and −28°C, and strains up to 2.2. Curves of strain at the borehole wall (logarithm of the ratio of hole diameter to its initial diameter) against time show a stage of constant closure rate corresponding to secondary (steady state) creep of the ice followed by accelerating closure rate attributed to recrystallization of the ice (tertiary creep). Curves for low stresses also show an initial transient stage of decreasing closure rate. The onset of tertiary creep is largely determined by the strain; critical values range from 0.03 to 0.10, and the lower the temperature, the higher the critical value. Secondary creep rates in the different boreholes are consistent with each other; the data yield a creep activation energy of 54 kJ/mol and a flow law index close to 3. The borehole data reduced to a common temperature of −22°C are compared with the results of two laboratory experiments at this temperature. For a given stress the strain rates measured by Steinemann (1958a,b) are 2–3 times those in the boreholes, and for the experiments of Barnes et al. (1971) the factor is about 8. Differences between laboratory and glacier ice, probably in grain size, may explain the differences between the borehole data and the results of Steinemann. Some evidence is presented that the creep rates measured by Barnes et al. at this temperature may contain a significant component of transient creep; this might account for the large difference between their results and those of Steinemann. The ratio of tertiary to secondary creep rate increases approximately linearly with the strain. No steady state tertiary creep rate is observed even at a strain of 1.5, at which point the ratio of tertiary to secondary creep rate is about 10. However, the ice is not strained uniformly during borehole closure. Even if recrystallization has been completed in the ice near the borehole wall, the ice further away, having been strained less, may still be recrystallizing. This may account for the failure to observe steady state tertiary creep. Near the bottom of one borehole, creep rates (tertiary) are about 4 times those in the ice immediately above. The boundary between the two deformation regions corresponds closely to the boundary between ice deposited during the Wisconsin glaciation and ice deposited since that time. The crystals in the Wisconsin ice are smaller, much less variable in size, and more nearly equidimensional than those elsewhere. Moreover, the Wisconsin ice has a much higher microparticle content and a much lower content of salts of marine origin. It is suggested that one or more of these differences make the Wisconsin ice ‘softer’ than the remainder of the ice. The decrease in grain size is considered to be the mo