The theory of pulsar radio emission is reviewed critically, emphasizing reasons why there is no single, widely-accepted emission mechanism. The uncertainties in our understanding of how the magnetosphere is populated with plasma preclude predicting the properties of the emission from first principles. Some important observational features are incorporated into virtually all the proposed emission mechanisms, and other observational features are either controversial or fail to provide criteria that clearly favor one mechanism over others. It is suggested that the criterion that the emission mechanism apply to millisecond, fast young, and slow pulsars implies that it is insensitive to the magnetic field strength.

It is argued that coherent emission processes in all astrophysical and space plasmas consist of emission from many localized, transient subsources, that any theory requires both an emission mechanism and a statistical theory for the subsource, and, that this aspect of coherent emission has been largely ignored in treatments of pulsar radio emission. Several specific proposed emission mechanisms are discussed critically: coherent curvature emission by bunches, relativistic plasma emission, maser curvature emission, cyclotron instability and free electron maser emission. It is suggested that some form of relativistic plasma emission is the most plausible candidate although one form of maser curvature emission and free electron maser emission are not ruled out. Propagation effects are discussed, emphasizing the interpretation of jumps between orthogonal polarizations.