The history of electrostatic accelerators has been punctuated by a series of projects in which innovative designs have failed to meet the expectations of their designers. From the early, air-insulated Van de Graaffs at Round Hill to certain of the large pressurized heavy ion accelerators of the 1970s and 1980s, increases in size or changes in design and materials have not always led to the maximum voltages expected or extrapolated. Since these failures have continued beyond childhood into a mature technology, it is reasonable to assume that the causes of voltage limitation are varied and complex. They have remained poorly understood for a number of reasons: resources for an extended program of research into breakdown and failure of electrostatic generators have always been meager, especially for large machines devoted to nuclear research; the inaccessibility of pressurized generators makes instrumentation difficult and testing slow; the calculation of transient and dynamic effects is laborious and the results difficult to verify; voltage test experiments on operating accelerators are inhibited by the significant risk of damage due to energy release on breakdown: and the total voltages (though not the local fields) achieved in many electrostatic accelerators exceed those produced in any other man-made environment. In this review, the behavior of several generators of different designs is examined in order to assess the importance of the various design features and operating conditions that control the maximum voltage achievable in a working machine. ©1999 American Institute of Physics.