One was an imposing statue of Mars, the Roman god of war. The other was an ancient bronze tablet, 5 feet wide and 3.5 feet high. It bore numerals in Roman but the words were in Gaulish, the extinct version of Celtic spoken by the inhabitants of France before the Roman conquest in the first century B.C.
The tablet, now known as the Coligny calendar, turned out to record the Celtic system of measuring time, as well as being one of the most important sources of Gaulish words.
Two researchers, Dr. Peter Forster of the University of Cambridge in England and Dr. Alfred Toth of the University of Zurich, have now used the calendar and other Celtic inscriptions to reconstruct the history of Celtic and its position in the Indo-European family of languages.
They say that Celtic became a distinct language and entered the British Isles
much earlier than supposed.
Though the Gauls were strong enough to sack Rome in 390 B.C., eventually the empire struck back. The Romans defeated the Celts, both in France and in Britain, so decisively that Latin and its successor languages displaced Celtic over much of its former territory. In the British Isles, Celtic speakers survived in two main groups: the Goidelic branch of Celtic, which includes Irish and Scots Gaelic, and the Brythonic branch, formed of Welsh and Breton, a Celtic tongue carried to Brittany in France by emigrants from Cornwall.
Because languages change so fast, historical linguists distrust language trees that go back more than a few thousand years. Dr. Forster, a geneticist, has developed a new method for relating a group of languages, basing it on the tree-drawing techniques used to trace the evolutionary relationships among genes. His method works on just a handful of words, a fortunate circumstance since only some 30 Gaulish words have known counterparts in all the other languages under study.
Dr. Forster and his linguist colleague Dr. Toth have used the method to draw up a tree relating the various branches of Celtic to one another and to other Indo-European languages like English, French, Spanish, Latin and Greek. In an article in today's issue of The Proceedings of the National Academy of Sciences, they say that soon after the ancestral Indo-European language arrived in Europe it split into different branches leading to Celtic, Latin, Greek and English.
Within Celtic, their tree shows that Gaulish - the continental version of the language - separated from its Goidelic and Brythonic cousins, much as might be expected from the facts of geography.
The researchers' method even dates the fork points in their language tree, although the dates have a wide range of possibility. The initial splitting of Indo-European in Europe occurred around 8100 B.C., give or take 1,900 years, and the divergence between the continental and British versions of Gaelic took place in 3200 B.C., plus or minus 1,500 years, they calculate.
These dates are much earlier than previously estimated. "The traditional date of the Indo-European family has been 4000 BC for some time," Dr. Merritt Ruhlen of Stanford University said. Dr. Ruhlen said the new method "seems pretty reasonable" and should be useful in tracing back the earlier history of the Indo-European language.
Specialists have long debated which country was the homeland of the Indo-Europeans and whether their language was spread by conquest or because its speakers were the first farmers whose methods and tongue were adopted by other populations. The second theory, that of spread by agriculture, has been advocated by Dr. Colin Renfrew, a Cambridge archaeologist.
Dr. Forster, who works in Dr. Renfrew's institute, said in an interview that the suggested date 8100 B.C. for the arrival of Indo-European in Europe "does seem to vindicate Renfrew's archaeological idea that the Indo-European languages were spread by farmers."
Agriculture started to arrive in Europe from the Near East around 6000 B.C., much earlier than the traditional date proposed by linguists for the spread of Indo-European. This timing would fit with the lower end of Dr. Forster's range of dates.
Dr. Forster said that his estimated date of 3200 B.C. for the arrival of Celtic speakers in England and Ireland was also much earlier than the usual date, 600 B.C., posited on the basis of archaeological evidence.
Dr. Forster said his method of comparing groups of languages was unfamiliar to historical linguists, many of whom study how words in a single language have changed over time. Asked what linguists thought of his method he said: "To be honest, they don't understand it, most of them. They don't even know what I'm talking about."
The method has two parts. One is to draw a tree on the basis of carefully chosen words; the second is to date the splits in the tree by calibrating them with known historical events. This is similar to the way geneticists date their evolutionary trees by tying one or more branch points to known dates from the fossil record.
Dr. April McMahon, a linguist at the University of Sheffield in England, said that Dr. Forster's method "seems to me to be a good start" and that it was reasonable to base a language family tree on just a handful of well-chosen words. She had less confidence in the dating method, she said, because language changes in an irregular way based on social factors like the size of the speaker's group and its degree of contact with others.
Geneticists often assume that the rate of mutation will average out over time, so that if one or two branch points in a tree can be dated by fossil evidence, the timing of the other branch points can be inferred.
Dr. Forster says he assumes that the rate of language change can also be averaged over time. But Dr. McMahon says she thinks that historical time, being much shorter than evolutionary time, is less friendly to averaging and that linguists should not even try, at least yet, to put dates on language trees.