We compare the merits of potential observatory sites on the Antarctic Plateau, in regard to the boundary layer, cloud cover, free atmosphere seeing, aurorae, airglow, and precipitable water vapour. We find that (a) all Antarctic sites are likely compromised for optical work by airglow and aurorae; (b) Dome A is the best existing site in almost all respects; (c) there is an even better site (“Ridge A”) 150 kms SW of Dome A; (d) Dome F is a remarkably good site except for aurorae; (e) Dome C probably has the least cloud cover of any of the sites, and might be able to use a predicted `OH hole' in the Spring.