The Woof at the Door
Dogs may have been man’s best friend for thousands of years longer than we realized
Germonpré and her colleagues thought that researchers might have overlooked early prehistoric dogs in the fossil record of the Upper Paleolithic, so they analyzed skulls of large canids (wolves or dogs) from various European sites. The Upper Paleolithic time period spanned 40,000 to 10,000 years ago and is divided into sections based on the artifacts from those times. By convention, each span is named for a culture of people who made the artifacts, and the people, in turn, are usually named for the geographical location where the artifacts were found. The Epigravettian culture existed from 14,000 to 10,000 years ago; before that, the Magdalenian culture thrived from 18,000 to 10,000 years ago; and skipping back a few sections, the Aurignacian culture occurred from 32,000 to 26,000 years ago.
In order to identify the fossil skulls accurately, Germonpré’s team first analyzed a large reference sample of 48 wild, modern wolves and 53 dogs belonging to 11 different breeds. They also examined five skulls (including the ones found in Eliseevichi) that were firmly established as prehistoric domesticated dogs.
The team used statistical analysis of cranial and dental measurements on the skulls to sort the reference sample into six natural clusters. One cluster contained modern wolves. Another consisted of recent dogs of archaic proportions (such as chow-chows and huskies); a single specimen of a Central Asian shepherd was closer to this group than any other but fell outside it. A third cluster included dogs, such as German shepherds and malinois, which have wolflike proportions. These three groups overlapped each other in their cranial proportions. A fourth group of modern dogs has short toothrows—the length of the jaw that contains teeth—and includes such breeds as great Danes, mastiffs and rottweilers. This group overlapped slightly with the archaic-proportioned dog group but not with the others.
The fifth and sixth clusters were completely separate from all others. One consisted of dogs with extremely long, slender snouts, such as Doberman pinschers. The final group, which had long toothrows and short, broad snouts, was made up of the prehistoric dogs. Statistically, the team’s ability to identify any individual specimen as belonging to the correct group was highly significant and accurate.
Using these clusters as reference categories, Germonpré and colleagues used a statistical technique (called
discriminant function analysis
) to assign 17 unknown fossil canid skulls to the established categories. Not all of the “unknowns” were truly unknown, however. Five were immature modern wolves that might have had different proportions because of their age, two were wolves that had been kept in captivity, and one was the Central Asian shepherd that didn’t cluster into any of the groups. Additional unknowns were 11 fossil skulls from sites in Belgium, the Ukraine and Russia, although two of these fossil skulls proved to be too incomplete to classify.
The technique correctly classified all of the immature wolves as wolves, but the two zoo wolves were classified as recent dogs with wolflike snouts. Five of the fossil skulls also fell easily into the modern wolf group; although two of these specimens fell into the region of measurements that overlapped with the group of recent dogs with wolflike snouts, they had a higher statistical probability of being wolves. One fossil skull fit directly into the group of recent dogs with wolflike snouts, even though this specimen was clearly ancient.
The remaining three fossil skulls—one from Goyet Cave in Belgium and one each from Mezin and Mezhirich in the Ukraine—resembled each other closely. All three were classified as prehistoric dogs with probabilities of 99 percent, 73 percent and 57 percent, respectively, as was the (modern) Central Asian shepherd, with a 64 percent probability. In addition, the Mezin skull was odd enough in appearance (for a wolf) that another researcher has suggested it might have been a captive wolf. Germonpré and her team were delighted with these results.
The group also successfully extracted mitochondrial DNA (mtDNA) from seven ancient canid bones from Goyet Cave and Trou des Nutons in Belgium. Rather than damage precious skulls, they sampled only bones in which wolves and dogs differ little, so they presumed all of those they sampled for mtDNA were wolves. From each sample, they sequenced a segment of the mtDNA that is highly variable in living wolves and dogs. Each fossil had a unique mtDNA sequence, or
, in this region, which could not be matched with any known sequences for modern wolves (of which there are about 160) or modern dogs (of which more than 1,000 exist) stored in GenBank, a database of all publicly available nucleotide sequences.
“I was not so surprised at the rich genetic diversity of the fossil wolves,” says Germonpré, because there have been other studies with similar findings. Foxes and wolves underwent a severe bottleneck in population size at the end of the last Ice Age, and many genetic lineages went extinct at this time.
surprised at the antiquity of the Goyet dog,” Germonpré adds. “We expected it would probably be Magdalenian,” perhaps 18,000 to 10,000 years old. This outcome would fit with their results for the Mezin and Mezhirich skulls, which were found with Epigravettian artifacts roughly 14,000 to 10,000 years old. When the age of this specimen from Goyet was directly dated using accelerated mass spectroscopy radiocarbon-dating techniques, the team found that it was not 18,000 years old, but almost twice as old as the next oldest dog, placing the Goyet dog in the Aurignacian period.