What is regression to the mean? First identified by Sir Francis Galton in the nineteenth century, regression to the mean simply means that offspring tend to be more moderate in characteristics than the parents. This is because the genes of the offspring come down from a large group of ancestors, so that extremes are often moderated because there are so many genes from so many individuals that are passed down to the current generation. As Galton put it, “Offspring tend to be more mediocre than parents.” This principle is one explanation for the often-heard wail, “How could these two beautiful parents have produced this litter of mediocre puppies?” In Shelties, we are on a quest for greatness. Who wants mediocrity?
I have used regression to the mean for many years as a way to get “improved moderates” for the show ring. Originally, this came about circumstantially, and eventually I realized what I was doing and how it was working. What regression to the mean signifies in breeding dogs, is that you can use rather extreme individuals to “push” the offspring toward correctness without too much risk of packing your bloodline with extreme characteristics.
In the distant past, I was an active expeditionary biologist, and so I sold my best dogs from every single litter. I was gone months per year on expeditions to the wild places of the world, and there was enough time and money for me to do breedings and travel to see dogs -- but not to campaign my own dogs, so I decided to sell every “best one” to someone who promised to show it. This meant that the cream of each generation was taken out of my program right away. (I learned that people very seldom actually SHOW a show prospect, but that is another story!) I followed the ones I sold, and most of them went right into the brood box, or were “interim generation males” for others (meaning that they used my dog a time or two, sold him, then showed the resulting generations). After a couple of generations of always selling the best ones off to supposed show homes, it became apparent to me that IN GENERAL, the supposed second-raters I kept were producing better quality puppies than the picks I had sold. It wasn’t entirely that the buyers of the picks were choosing less suitable mates, either. Often, litter sister “extreme” that I kept, was producing puppies that were more correct than the puppies of litter sister “show,” when bred to the same stud dog.
I found that after having sold the show prospects in a litter, the ones I kept from the remainder actually had more to-the-Standard excellences than the pick prospects I had sold, more correct “pieces.” The show dogs tended to be better in “overall look,” had better balance, and were more moderate. The second-rank ones I kept tended not to be as pulled together overall, not in as good a balance, but offered a number of excellences: extremely good angulation, for example, or extremely flat skulls, extremely clean skulls, extremely long tails, etc., etc. There were others I discarded from the breeding program (in addition to the pet quality ones). These were all-right dogs in most respects. I call these the “moderate average dogs.”
The extreme second-rankers I kept had obvious excellences and the moderate average dogs were just second-rate all over. They were mediocre, one might say, with a nod to Dalton. After a while it came to me that this practice of selection was using the principle of regression to the mean as a breeding tool.
Most of these second-rank extremes I kept as breeders were bitches, since I figured there was little point in using second-rank males. Now I realize that there would have been little difference in the quality of the resulting offspring. However, in practice, having a nice Ch. as the sire is a much better marketing tool! Also, there is better predictability in using a noted sire. A noted sire has many offspring, and so one can see easily how he is producing, making it is easier to choose the right bitch to take to him for the best offspring.
So, on to breeding a rather extreme bitch with excellences: If you have one of these second-rank “excellences” bitches with extremely correct shoulder angulation, for example, and you breed her to Ch. Moderate Male, since her shoulders are very much better than average, with regression to the mean you are going to swing that percentage of good shoulder structure farther toward Standard-perfect than if you had bred a moderately angulated show bitch to Ch Moderate Male. Note the extreme sable bitch above. Note that her extremes were excellences, things like, extremely angulated front, extremely correct stop, extremely long skull, extremely low-carried tail -- not things like, “extremely short tail,” or extremely shallow underjaw.” You can see that this gal, if bred to a moderate male, could shift several desirable factors quite a distance toward Standard-perfect. This is Banchory High Glow, Mae, who is in countless pedigrees today with four champions and two influential ROM sons to her credit. Even doubling and redoubling Mae did not load the pedigrees with extremes. That's regression to the mean.
Of course, you DO want to breed your moderate show gals, and you can get very good offspring from them. However, using regression to the mean as a breeding tool may enable one to gain ground fast with creatures that have excellences to offer but not that overall show “look.”
DanaQ
