In the northern British Isles, the Celtic tribes known as the Picts coexisted for centuries alongside literate cultures such as the Romans, the Irish and the Anglo-Saxons.

"They were the odd society out, in that they didn't leave any written record," says Rob Lee of the University of Exeter in England, save for some mysterious-looking sets of symbols on stones and jewels. In a paper published March 31 online in Proceedings of the Royal Society A, Lee and his coworkers now claim that the symbols are written language. Perhaps the Picts were not illiterate after all.

Pictish inscriptions—of which fewer than 500 remain, each not more than a few symbols long—have puzzled archaeologists for a century, Lee says. Could they be religious imagery? Tribe names? Or perhaps coats of arms? "Only in the last decade people have asked the question, 'Are they a language?'" he adds.

Lee's team attacked the problem with math. Written languages are distinguishable from random sequences of symbols because they contain some statistical predictability. The typical example is that, in the English language, a "q" is nearly certain to be followed by a "u"; and a "w" is much more likely to be followed by an "h" than, say, by an "s" or a "t".

This predictability is measured by the notion of Shannon entropy, one of the keystone concepts of information theory. (Shannon entropy explains for example why one can compress computer files into Zip archives, and puts a hard limit on how much they can be compressed.)

Lo and behold, the Shannon entropy of Pictish inscriptions turned out to be what one would expect from a written language, and not from other symbolic representations such as heraldry. "The paper shows that the Pictish symbols are characters of a lexicographic written language," Lee says, "as opposed to the most general form of writing, which includes things like the [non-verbal] instructions on your Ikea flat packs."

A different team used a similar technique last year to suggest that the script of the Indus Valley civilization, which flourished between 2600 and 1900 B.C., is also a written language.

But how about actually translating the mysterious scripts? The statistical methods unfortunately are not too helpful yet. And there is no Rosetta stone for deciphering Pictish script. But, Lee says, there are about a dozen stones that contain some very short inscriptions together with some Latin script, which might help interpret the language, at least in part. He adds: "I am somewhat hopeful that we may be able to do it."


Images: A Pictish stone showing symbols that may have been part of a written language (top)

Pictish silver plaques showing a "double disk and Z rod" symbol and a dog/seal's head symbol (bottom)

Credit for both is Rob Lee