The likes of Brian under

In a scene in Monty Python's Life of B r ian , the hero is caught scrawling on the wall by a Roman centurion outraged at the poor grammar of his rebel graffito Romanes eunt domus (by which he intends to mean "Romans go home").

The necessary grammatical corrections are bullied out of the unfortunate subversive as if he were a humiliated schoolboy who has yet to perfect his command of Kennedy's Latin Primer .

But, of course, imperfect or even minimal knowledge of a language that dominates the government machinery and the defining cultural practices of a prevailing regime is no laughing matter, as countless parallels from other periods confirm.

J. N. Adams' authoritative work will be the standard textbook for classical philologists and others working in the field of sociolinguistics but, for all its occasionally dense technicalities, it should also engage those interested in broader cultural questions of empires, ethnicity and identity. This is not a reference book. Rather, Adams carefully and systematically introduces his methodology before launching into a series of detailed regional and thematic case studies. Unfortunately, while other languages are translated, competence in Latin is assumed - a shame, given the invigorating subject matter.

Language cannot be equated absolutely with ethnic and cultural identity but the link between them tends to be strong. As has been argued cogently in recent years, it was possible to become Roman in classical antiquity in a way that it was not really possible to become Greek.

To oversimplify, the Romans seem to have been more willing than the Greeks to bridge the gap between culture and ethnicity. Latin was the language of the Roman empire (although because of a complicated cultural pathology, Greek was also used), so, as the military machine gobbled up more territory - culminating with Trajan's conquest of Dacia (Romania) in the 2nd century AD - use and knowledge of Latin spread too.

There is little record of the conquerors being concerned to learn the indigenous languages; with the arrogance common to most empires, the Romans represented their expansion as a mission to civilise the natives, and their respect for native languages was therefore minimal. For their part, while newly conquered peoples were not forced to learn Latin, ignorance of it would leave them disenfranchised in certain ways (legally, for example), so for thousands of people across the "civilised" (that is to say, conquered) world some knowledge of Latin became beneficial. As the decades and centuries passed, the results of this process could be stark: men such as Martial, Ausonius and St Augustine - from Spain, Gaul and North Africa respectively - were celebrated Latin writers, while, without the Romans necessarily being aggressive, languages such as Iberian, Gaulish and Libyan died out. Although Adams studiously avoids articulating value judgements about Roman culture, he certainly provokes some.

But his focus is neither the languages killed off by Roman imperialism nor the self-conscious bilingualism of some elite Roman literati (predominantly Latin-Greek). His core material is sub-literary: letters, legal and other documentary texts, monumental inscriptions such as epitaphs, and inscribed artefacts. This is an important aspect of his study - and one that will demand the attention of other disciplines. Most studies in bilingualism concern themselves with the spoken word; Adams' material was written to be read or, in some cases, simply seen. The element of display inherent in much of this is an important determining factor in its interpretation.

Adams displays a magisterial command of his disparate evidence, most of the time employing four sub-categories: bilingual, transliterated, mixed language and cases in which the text implies a bilingual situation.

Categorisation can be a preoccupying feature of linguistics but, at his most accessible, Adams successfully engages the non-specialist reader by teasing out fascinating sociological contexts.

One example is a short bilingual Latin/Palmyrene Aramaic epitaph, found in South Shields, set up by a man called Barates for his wife. The Latin is faulty, the Aramaic fine, which suggests that Aramaic was the "dominant" language. But there are no Palmyrene units recorded on service in Britain, so hardly anybody in South Shields would have been able to understand the Aramaic half of the inscription. Why did Barates bother? Because, Adams argues, he was determined to display his ethnic identity as he memorialised his grief, and his choice of language was a powerful means of achieving that ambition. Furthermore, Adams argues, against earlier editors, that the Latin is at fault because Barates was bilingual in Aramaic and Greek, and that his acquaintance with Greek inscriptional idiolect accounts for the dodgy Latin. Adams' exposition of the widower's biographical context and dogged ethnic confidence is typically learned and sensitive.

Traders, as well as the military, had a reason to be more or less bilingual (and Adams is good on degrees of bilingual competence). He suggests that recently published texts from the underside of plates baked in the kilns of La Graufenesque in France in the middle of the 1st century AD reveal a richly diverse range of language use by many potters. Latin and the native Gaulish tend to be differentiated, but while some texts frequently "code switch" between the two languages, others do not. The role and linguistic preference of each potter is acknowledged, arguing against any claim for an institutionalised mixed language at this, a key stage in the process of the Romanisation of Gaul.

Adams is ever alert to the range of influences and motivations informing linguistic choices. In cases such as these from Britain and France, we are reminded that societies are made up of individuals, some of whom will not fit the models anthropologists construct. When it comes to the words we use, we are all individuals, as Brian would say.

Roger Rees is lecturer in classics, University of Edinburgh.