Friendship is a fluid concept. Not only does it vary in nature and prestige from culture to culture, but it is not to be identified by any legally recognized act or verifiable fact of birth or status. Rather than grapple with this Proteus, Williams has decided to limit his quest to Roman friendship in particular, and to the language employed rather than the social realities, ‘reading amicitia as a system of labels and categories to be interpreted more than as a set of relationships and events to be reconstructed’ (60).

In the long Introduction W. justifies his restriction to Latin-speaking areas of the Roman Empire, by noting that Greek philia has been more frequently studied, and by claiming that, via Cicero's De amicitia, Roman friendship has been more central than Greek to the history of Western friendship. He defends his concentration on the rhetoric of friendship, avoiding awkward questions of definition, by construing amicus and amicitia as performative utterances: whoever is called an amicus, is one; whatever is labelled amicitia, counts as such. This project, described as a more modest but more attainable goal (28), certainly avoids a number of problems. It allows W. to disregard the idealizing tradition of Latin texts in which types of friendship, notably those that Aristotle characterized as utilitarian, and friendships between social unequals, are rebranded as ‘political friendships’ and ‘patronage’ and denied the title of ‘true friendships’. But there are still difficulties in discussing the Latin vocabulary of friendship. Should one make use of English terms, or confine oneself to Latin terminology? In an interesting discussion (30–5) W. points to salient differences: there is no Latin equivalent of ‘just friends’ or ‘best friend’ (optimus amicus means ‘the best kind of friend’, not ‘the best of my friends’). In the end he sometimes leaves the Latin terms untranslated and sometimes uses English paraphrases, for which inverted commas are always to be understood and linguistic self-awareness advised (35). Then there is the fundamental problem of recovering ancient usage. As we have no opportunity for live encounters, and no access to unmediated speech, W. has recourse to what Bakhtin called ‘secondary speech genres’, ranging from traditional literary genres to inscriptional texts (37), adding a reassurance that there was no significant change in the vocabulary of social relations or in the ideal associated with them, between Republic and Principate.

The Introduction ends with an outline of the book's structure (60–2). Chs 1 and 2 deal with overall themes and problems. Ch. 1 shows that, whereas friendship is often represented in literary texts as a masculine prerogative, inscriptions and the letters from Vindolanda correct that impression. Moreover inscriptions use amicus and amica symmetrically, whereas in many literary genres, when a woman is linked with a man and called his amica, she is a sexual partner outside marriage (96). Ch. 2 explores the relation of love and friendship, showing that amicitia can be a subset of amor, which also covers erotic love, but that boundaries are very fluid between the two types of love. Chs 3 and 4 give closer readings of a selection of texts, both literary and inscriptional. The literary texts explored in ch. 3 are Catullus, Virgil, Horace, Propertius, Petronius, and the letters of Cicero and Fronto: Pliny, Statius and Ovid are not revisited. Particularly interesting is the discussion of amare and amor in Cicero and in Fronto's letters, where the differences from the usage of elegiac poets are clear. Benefiting from Hutchinson's 1998 study of the former's correspondence, W. points to Att. 9.10.2 where Cicero, explicitly invoking the Greek term τὰ ἐρωτικά, compares, in a simile, his disillusionment with Pompey to disenchantment with the tactless behaviour of a lover. The lavishly affectionate language that Cicero sometimes uses to his friends would not have been misunderstood as erotic by his readers; Shackleton Bailey is right to translate amor and amare in terms of affection, fondness and admiration (220). Fronto's correspondence with Marcus Aurelius surpasses Cicero in the use of such language, but W. succeeds in showing that here too what is really involved is affection and devotion (238–58). Not only are erotic relations sometimes marked as comparisons, explicitly (M.Caes. 3.14) or implicitly (Addit. 8 and 7), but amor here lacks passionate exclusivity. Instead, there is an emphasis on networks of friendship, while other members of their families, such as Fronto's wife, the emperor and Marcus' baby, are named as objects of amor, and of kisses desired and given, along with the correspondents. Then in Ad Verum 1.7 it is clear that kissing can be an envied mark of privilege. W. remarks that we learn here ‘something important about kisses throughout the Latin textual tradition and, as far as we can tell, in Roman social practice as well’ (257).

Earlier too, W. had admitted that the realities and subtleties of interpersonal relationships among Romans sometimes come through the texts being scrutinized for language (60), and it is partly because this is particularly true of the funerary inscriptions, that ch. 4 is the crowning glory of the volume. As W. says, one would not be aware from the Latin literary tradition or from scholarship on it, that ‘Romans were often buried in groups, individuals identified as amici not infrequently were members of these groups, and friends provided a key role in Roman commemoration of the dead’ (260). ‘More often than has been acknowledged, Roman burials and the inscriptions marking them perpetuated the memory of and thereby enacted the ties not only of kinship, marriage, or slavery, but of nothing more nor less than amicitia’ (337). Of course, comparing inscriptions with high literature is difficult: many are undatable, and many fail to mention the legal status of the individuals named. But epigraphists have established that most Latin inscriptions were produced between the first century a.d. and the beginning of the third century a.d., and that freedmen and freedwomen are over-represented. In the absence of legal ancestors, parents or siblings, they commemorated relationships with their former owners (where the language of amicitia is avoided), with their own freed slaves, with conliberti of their former masters, and with amici and amicae, citizen and slave. Indeed, slaves figure among the commissioners of inscriptions as well, commemorating friends as well as spouses and children, whom they call, without legal warrant, coniuges and filii.

Ch. 4 ends with a typology of the uses of the language of amicitia on epitaphs (296–354). The group commemorations do indeed give us ‘an indirect glimpse at the varieties of household structures that could arise in conjunction with slavery and manumission’ (324); the joint burials of two friends in a single tomb show that Martial 1.93 is not describing something unusual (339).

Reading Roman Friendship is itself well worth reading for its insights into Latin literature and Roman social history. Let us hope that the large number of typographical errors will be corrected in the reissue that it certainly deserves.