Archaeology Magazine Archive

A publication of the Archaeological Institute of America

Special Introductory Offer!

The carved slabs of northeastern Scotland known as the Pictish stones are complex and haunting, testimony to a sophisticated culture on the edge of medieval Europe. Made by the Picts, a group of Celtic tribes that lived there from the sixth to ninth centuries A.D., the monumental stones are among British archaeology's most enduring enigmas. So there was understandable excitement in March when a team of British scholars claimed their elaborate decorations are really a form of writing.

Only about 250 Pictish stones survive, mostly in museums or private collections. Their carvings mix Christian iconography with distinctive looping designs, symbols such as crescents and disks, and pictures of various creatures, from wolves and eagles to mermaids and dragons. (The classic "Celtic cross" was inspired by Pictish designs.) There are more than three dozen symbols—scholars disagree on exactly how many—and they usually appear in pairs.


A Pictish stone in the St. Vigeans Museum in Angus, Scotland

To see if the stones might represent a written form of the now-extinct Pictish language, Rob Lee, an industrial biology professor at the University of Exeter in England who simply has an interest in the Picts, gathered the known combinations of symbols in a database and analyzed them mathematically. In a paper published in the Proceedings of the Royal Society, he argues that the Pictish symbol sets share traits with known writing systems.

Lee applied a number of statistical methods to compare the order of Pictish symbols with that of characters in writing systems such as Egyptian hieroglyphs, Chinese, Anglo-Saxon, and Old Welsh. "One of the ways you can show whether things are writing is through mathematics," he says. "Is it some sort of random system, or not?"

By looking at the number of times symbols are paired together or appear in a certain order, Lee claims it's possible to classify Pictish as a written language. He determined that it is less repetitive than non-linguistic symbol systems, such as heraldry, and is structured more like written words.

Lee's report was greeted with enthusiasm in the press, just like a study last year in which a group of mathematicians claimed to have discovered similar trends in the Indus script, a 4,000-year-old collection of symbols found on seals in South Asia ("The Indus Engima," March/April 2010).

But critics say Lee's analysis is based on a fuzzy grasp of what defines writing and some wishful statistics. According to Richard Sproat, a computational linguist at Oregon Health and Science University (and an outspoken critic of the theory that the Indus script represents a language), a writing system must correlate sounds or spoken words with specific symbols.

In other words, non-random order alone does not a language make. Mathematical equations, for example, have rules and structure, but don't represent syllables or words. And even somewhat random processes, such as throwing dice, can create statistical patterns similar to writing. University of Pennsylvania computational linguist Mark Liberman argues that proving Pictish is writing would require ruling out all possible alternatives. "I don't believe there will ever be a mathematical test that in itself reliably proves that a particular collection of symbols is a form of writing, especially if the collection is small and mostly made up of one- or two-symbol sequences," he says.

There's no doubt whoever carved the Pictish stones had a message in mind. But without a key to unlock the meaning of the carvings, what they were trying to express will never be known.