LIKE most historians, those who specialise in past cultures must normally rely primarily on written and printed sources. Although works of art and music, archaeological remains and other forms of material evidence can be pressed into service, most of our documents normally consist of texts and often especially of poems, plays and other works of ‘literature’. Through this material we try to reconstruct beliefs and practices that, in most cases, depended far less on writing and print than performances, rituals, oral discussions and other transient events that we can no longer directly witness. Faced with this predicament it is natural to hope that if we only read carefully and ingeniously enough, we can somehow gain access to every aspect of a vanished culture through words on a page. Especially if the work under examination is a rich and compelling one, and the scholar studying it a trained and dexterous reader, the range of latent historical meanings can seem almost boundless. Social customs, ideologies and even deep economic structures can all be discovered embedded within the text, allowing a total history to be reconstructed on the basis of literary analysis.

There is nothing intrinsically illegitimate about this methodology, which resembles all forms of historical scholarship in its attempt to extrapolate over-arching interpretations from patchy and incomplete bodies of documentary evidence. Works of literature always refer beyond themselves to a larger cultural universe that originally surrounded them. As Stephen Greenblatt has rightly argued, authors never truly create texts out of some mysterious inner quality of genius. They appropriate and manipulate materials borrowed from their surroundings, including language itself as well as meanings latent within material objects, political institutions and social customs. By examining these appropriations we can approach a text as a repository of past cultural attitudes and point of entry into the world that its writer and his original audience once inhabited.

But although valid in principle, the enterprise of reconstructing history through textual analysis raises several vexing methodological issues. To begin with, the perspective offered by a literary source may be incomplete, biased or misleading in significant ways, and unless we have some independent basis for controlling such distortions we risk being seriously misled.