This paper reports the cyclic behaviour of chalk, which has yet to be studied comprehensively. Multiple undrained high-resolution cyclic triaxial experiments on low-to-medium density intact chalk, along with index and monotonic reference tests, define the conditions under which either thousands of cycles could be applied without any deleterious effect, or failure can be provoked under specified numbers of cycles. Intact chalk's response is shown to differ from that of most saturated soils tested under comparable conditions. While chalk can be reduced to putty by severe two-way displacement-controlled cycling, its behaviour proved stable and nearly linear visco-elastic over much of the one-way, stress controlled, loading space examined, with stiffness improving over thousands of cycles, without loss of undrained shear strength. However, in cases where cyclic failure occurred, the specimens showed little sign of cyclic damage before cracking and movements on discontinuities lead to sharp pore pressure reductions, non-uniform displacements and the onset of brittle collapse. Chalk's behaviour resembles the fatigue response of metals, concretes and rocks, where micro-shearing or cracking initiates on imperfections that generate stress concentrations; the experiments identify the key features that must be captured in any representative cyclic loading model.