Again, there is the question as to the doors in the proscenium. If it was the background, it ought to have had three doors, the usual number in a Greek back-scene, as Pollux and Vitruvius tell us. But in most of the proscenia discovered there is only one door. In two of the proscenia, those at Megalopolis and Thespiae, there is no door of any kind. Even the single door, when it is found, is very narrow for the central door of the back-scene. At Epidaurus it is only four feet wide, at Oropus only 3 feet 8 inches, at Delos only 3 feet 3 inches.[479] A door so narrow as this would be altogether unsuitable as the central door of the palace, and quite inconsistent with the use of the ekkyklema. When we come to the Graeco-Roman theatres, where the wall at the back of the stage has in many cases been preserved, there we find everything corresponding closely with the descriptions of the grammarians. There is always the requisite number of doors, and the central door is of considerable width. At Termessos it is about seven feet.[480] As regards the absence of the three doors in the proscenium Dörpfeld gives the following explanation. These Hellenistic proscenia, as we see from the remains, consisted of an entablature resting on columns. The spaces between the columns were filled in with wooden boards. Dörpfeld suggests that when doors were required they might be provided _ad libitum_ by removing the intervening boards.[481] But if three doors were regularly required in the dramatic performances, it is most improbable that they should not have been provided as a permanent fixture in the proscenium. It is most improbable that the Greeks should have put themselves to the trouble of opening out these temporary doors at each festival. In any case we can hardly doubt that, if the proscenium had been the back-scene, the Greeks would always have provided at least one permanent door, and would not, as at Megalopolis and Thespiae, have erected proscenia in which there was no door of any kind. The absence of a door in these two places seems to prove conclusively that communication between the orchestra and the space behind the proscenium was a matter of no importance. [Illustration: FIG. 13.]