No one can write an adequate commentary on a Greek play, or even edit it adequately, without producing it in his mind; that is to say, he must see in his mind's eye the exact location, movement and gesture of every character at every moment, arid must hear in his mind's ear the exact tone of every word. Admittedly, we do not know so very much about the appearance of the theatre in the fifth century B.C., or about the actors' costumes, or about Greek gestures; even our knowledge of the pronunciation of Greek at the relevant period is imperfect; and our ignorance of delivery and acting styles is total. Therefore, when we produce a play in our imagination we can hardly fail to import much which would be recognized as grossly erroneous if our evidence were suddenly to be increased. Yet, as so often happens in the study of the ancient world, we are confronted with a choice of risks. We can take the risk of an imaginative reconstruction which may be factually wrong but is at least reconcilable with all the evidence we have and is kept reconcilable, by modification, with whatever fresh evidence comes to light from time to time; or, alternatively, we can take the risk of missing the point of what the dramatist is saying and therefore of importing the wrong point. Even if we are not temperamentally disposed (as I am) in favour of the former risk, we are virtually forced to take it when we have committed ourselves to editing a text. The punctuation of a dramatic text is a matter for producers, not grammarians; a comma gives one instruction to the actor, a colon gives a different instruction—and few editors are likely to practise ars nescienti so pertinaciously as to refrain from giving the actors any such instructions at all.