The paper reviews present understanding of H-mode physics by summarising relevant experimental observations and discussing possible interpretation. The most important features of the H-mode are a minimum threshold edge temperature (threshold input power) required to achieve the H-mode; the bifurcation nature of the H-transition with instantaneous changes at the plasma edge; and the formation of a transport barrier at the plasma edge leading to pedestals in the density and temperature profiles. Global energy confinement times, are typically 2x to 3x longer in H- than in L-mode plasmas, reaching, for example, almost 1s in 3MA JET X-point discharges. τE is found to increase linearly with plasma current. Results of the variation of τE with input power are somewhat contradictory: no power dependence is found in ASDEX and DIII-D, whereas a degradation with power is indicated in JET and in JFT-2M limiter H-modes. Correspondingly, predictions for full power (40MW) 6MA X-point discharges in JET range from 0.6s to > 1s, depending upon which scaling is adopted. Two main theoretical models have been proposed to explain the H-mode with its heat barrier at the plasma edge. Such a barrier is predicted by the stability properties of ballooning modes close to a magnetic separatrix, corresponding to the transition to a second stable region above a certain threshold power. It could also arise from a critical temperature gradient model based on self-consistent stochasticity.