Mutant turtles, all too well known in popular culture, have long been an enigma to science.

For decades, zoologists have argued over how turtles are related to other reptiles, with whom they share a common ancestry but few obvious similarities. Now researchers have proposed an unorthodox theory for the origin of the order.

Reporting in the journal Nature, Robert Reisz and Michel Laurin of the University of Toronto argue that turtles evolved from a primitive sister-group of early reptiles known as the procolophonids, previously thought to be extinct.

The authors base their findings on new fossil evidence from South Africa: the bones of a small "parareptile" called Owenetta that lived about 220 million years ago. By comparing details of its skull with those of modern turtles, they conclude that today's species are more closely related to these lizard-like creatures without shells than to "any other Paleozoic reptile."

If confirmed, this theory could resolve a perennial mystery of paleontology. All reptiles, birds and mammals belong to the amniotes -- animals that produce eggs surrounded by a sac-like membrane that permits an embryo to become highly developed before birth, in contrast to amphibians that lay their eggs in water.

Conventional thinking divides the amniotes into two broad groups: One produced mammals; one branched to evolve birds and modern reptiles. Despite their apparent dissimilarity, members of the latter group share certain characteristics, chiefly the number and placement of openings in their skulls that facilitate muscle attachment.

Turtles, however, never fit in; they seemed to have taken some evolutionary offramp from the main path of amniote development. They lack skull openings, and they have distinctive shells and a unique design feature -- a rib cage outside the shoulder girdle. This configuration has persisted, largely unchanged, for 200 million years.

Before that, there is a 100 million-year gap in the fossil record, making it impossible to picture the turtle's evolutionary predecessors. In fact, says Gene Gaffney, a curator at the American Museum of Natural History in New York, "we can't even imagine what a half-turtle would look like."

The Nature article is expected to intensify controversy in the turtle evolution community. Gaffney deemed the Toronto theory "an interesting idea," but probably wrong. He is the leading proponent of the widely accepted notion that turtles descended from a different group of early reptiles called captorhinids.

Nicholas Fraser of the Virginia Museum of Natural History said further study may reveal that turtles represent a separate line unrelated to reptiles.