THE politics of friendship underpins the workings of premodern societies. Circles and networks of friends are being identified increasingly by historians as central to medieval and early-modern political structures. Friendship pervades the sources, with frequent references in practically every genre, from letters, treatises and romances to histories, charters and liturgical texts. Declarations of friendship have been read as evidence for everything from close affectionate relationships to negotiated allegiances—at one extreme being somebody's friend could mean no more than having made a pact to cease open hostilities.1 The extension of apparently personal bonds into the public or political spheres and their formalisation as types of allegiance, as well as the use of warm, emotionally charged language in diplomatic or business correspondence, have become current topics in the study of medieval politics as historians have been concerned to establish the connections between friendship, social network formation and political action, raising the question of what it actually meant to be called a friend. This problem is especially evident in medieval letters, where writers commonly directed professions of love and friendship to bare acquaintances, entire communities and even complete strangers, often adopting a far more intimate tone than when addressing close companions and confidants. While this has often been commented on in individual cases, we have as yet no satisfactory explanation of the conventions and practices governing these and other patterns of address and language use that are common to letter collections and that potentially offer the key to understanding the formation and operation of important political networks. Indeed, practically all studies of medieval letters have been concerned with individual relationships or particular exchanges of letters, taken in isolation from the collections of which they are a part and usually read as evidence of strong emotional bonds.2 There are at present few network studies of entire letter collections and no systematic surveys of the incidence of the language of friendship across different collections. This article presents the results of a new analysis of the use of friendly address and language in Peter the Venerable's letter collection, one of the most important of the twelfth century. The results, considered with the evidence from other letter collections where comparable studies have been made, suggest a reconsideration of the nature of the friendship bond and of letters of friendship as historical evidence.