A macabre 1,700-year-old mass grave of people and horses, discovered in Normandy, poses perplexing new questions about the Roman conquest of France. Was there a small part of ancient Gaul which refused, Asterix-like, to surrender for 300 years?

The grave site, from the 3rd century, which was discovered by French state archaeologists at Evreux, appears to contain ritual arrangements of human and horse remains. In one, a human skull is clasped between two horse's skulls, like the two halves of a giant shell.

In Gaullish times, 300 years earlier, graves containing both horses and people were common. No such grave has ever been found from the Roman period, and even in the previous era, the remains were kept carefully apart.

In the recently discovered grave, about 50 miles west of Paris, the bones appear to have been intentionally mixed. The skeletons of 40 people and 100 horses have been found so far.

Was this a local - or maybe more widespread - survival of the Gaullish cult of Epona, the goddess of horses and warriors? Sylvie Pluton is leader of the dig for the |Institut National de Recherches Arc�ologique Pr�ventives (Inrap). She is also an expert on the Gallo-Roman period.

"With the Romans, you usually know what to expect," she said. "They were very organised. Their graves were very orderly. Not here. The bodies point in all directions ... Above all, there is extraordinary mingling of humans and horses. We could be looking at a cultural survival, previously unknown, such as a worship of the goddess Epona."

Roman graves often contained offerings of food, but Romans did not eat horse flesh. Nor can this have been a warriors' grave. Many of the human skeletons are those of children or women or old men.

Some Gaullish practices and beliefs did survive deep into Roman times, but there have been no previous finds as striking. One of the visitors to the site was Professor Christian Goudineau of the Coll�ge de France, the foremost expert on the period. He said: "Personally, I am reluctant to believe in some kind of cultural survival, such as a cult of the goddess Epona. Why would it survive for so long? And here, on the edge of what we know was a large Roman town?

"Perhaps these were slaves and horses which died in an epidemic and were just thrown here in a hurry and became mixed up," he added.

The problem, as Professor Goudineau himself pointed out, is that some of the remains seem to have been carefully arranged. Further digging on the site in the next two months, before it is covered by a new bungalow, may help to unlock the mystery.