Jas’ Elsner writes, in his review of E.W. Leach, The Social Life of Painting in Ancient Rome and on the Bay of Naples (Cambridge, 2004):
The one sustained attempt to build a social picture out of pictorial subject matter has been in the study of Roman sex, fundamentally indebted to Kenneth Dover’s explorations of Greek homosexuality from the evidential base of what was depicted on vases. If this is the model, one might argue that Leach has been wise to keep away: the post-Dover study of ancient sexuality, quite apart from the fundamental methodological problems involved in inevitable presentism, has insisted on a remarkably literalist interpretation of images so as to argue that what they show is what people did. It may indeed be that what is shown today in what the British coyly call ‘top shelf magazines’ (top shelf being where Newsagents display them) does indeed reflect some people’s actual practices, but of what socially meaningful group? And how does what is depicted in such publications relate to the activities of the majority (or even to their desires)? Likewise, when Greek pots or the cubicles where people left their clothes in Roman baths show scenes of pretty well every imaginable sexual activity (or perhaps my imagination is excessively limited here), we need at least a careful argument to show that this has anything to do with reality. The assumption that such imagery betrays actual practice, or is a positive and celebratory affirmation of contemporary mores, smacks of wishful thinking.
Perhaps the sheltered life of an Oxford academic is not the best vantage-point to discuss the range of sexual practice even in our own society, let alone another; and one may surmise that the purchasers of top-shelf magazines are for the most part less well placed to realize the fantasies therein portrayed or thereby inspired than those who stripped off at mixed-sex baths or drank at symposia in the company of hetairai, let alone the wealthy owners of Pompeian villas. But if actual practice was not reflected in the imagery that we have, what kind of imagery might we have expected to find instead if it had been? Without some such control, the flight from literalism becomes mere aprioristic speculation. It is not enough to say that things cannot be proved to have been so; we need some evidence that they were not to defeat the default assumption that they were.