It is generally regarded that the bulk of cosmic rays originate in the galaxy and that those below the ‘knee’ (the rapid steepening in the energy spectrum) at a few PeV, come from galactic supernovae, the particles being accelerated by the shocks in the supernova remnants. At higher energies, there are problems in that conventional SNR—which surely constitute the bulk of the sources—have a natural limit at a few tens of PeV (for iron nuclei). The question of the origin of particles above this limit is thus an open one. Here we examine a number of possibilities: a variety of supernovae and hypernovae, pulsars, a giant galactic halo and an extragalactic origin. A relevant property of any model is the extent to which it can provide the lack of significant irregularity of the energy spectrum, its intensity and shape together with structures such as the ‘second knee’ at the sub-EeV energy, in addition to the well-known ‘knee’ and ‘ankle’. Although it is appreciated that spectral measurements are subject to systematic as well as random errors, we consider that contemporary data are good enough to allow at least some progress in this new field. These aspects are examined for particles of all energies and it is shown that they can constrain some parameters of the proposed models. In the search for origin above PeV energies, we conclude that shocks in the galactic halo, whatever their source (galactic wind, relativistic plasmoids—‘cannonballs’, multiple shocks from supernovae etc) are most likely, pulsars such as B0656+14 and hypernovae come a close second although such a suggestion is not without its difficulties. What is most important is that trapping of particles in the halo is needed to reduce irregularities of the energy spectra both below and above the ‘knee’ caused by the stochastic nature of supernova explosions and other potential (discrete) galactic sources. We argue that precise experimental studies of spectral ‘irregularities’ will provide considerable help in the search for cosmic ray origin.