The theoretical efficiency for the direct conversion of electrical energy into luminescent emission, that is, electroluminescence (EL), is first evaluated from thermodynamics. High-field, collisional excitation EL is then introduced and compared with low-field, minority carrier injection EL and with cathodoluminescence. Some general characteristics of thin-film high-field EL are discussed. The basic mechanism of high-field EL is reviewed. It consists of three steps occuring in sequence: generation of charge carriers, acceleration of these carriers to optical energies, and their inelastic collisions to excite luminescent centers. Ballistic, streaming and Maxwellian hot carriers are distinguished, and their distribution functions discussed. The incompatibility of large cross sections for collisional excitation and stability of the excited luminescent centers in high fields is considered, taking account of the energies of the electronic states of dopants with respect to band structure. The present understanding of the structure and electronic states of transition metal and rare earth (RE) luminescent centers in II–VI and II–VII 2 semiconductors is evaluated. Particular attention is given to probable structures of molecular dopants such as REF 3 in ZnS. Recent advances in thin-film, high-field EL are summarized. The emphasis has been on ZnS:Mn and ZnSe:Mn films sandwiched between insulating films such as Y 2O 3 and operated with a.c. voltages. All three steps of the high-field EL mechanism occur in the II–VI layer, with some carrier generation at the II–VI/insulator junction. The spatial separation of the sequential steps in contiguous layers in composite EL cells is proposed, and evidence for separation of the acceleration and collision excitation in SiO/ZnF 2:Mn/SiO cells is presented. Some quantum optical effects in thin-film EL cells are discussed. Finally, some basic problems of high-field EL are reviewed. These include: the origin of charge carriers, the distribution function of the energetic carriers, structure and energy levels of complex dopants, and sources of inefficiencies and of saturation. Possible methods for achieving higher efficiencies than the current (1%) are proposed.