More than 50 years of research, dozens of papers and several books by
multiple researchers from different countries have resulted from the
original Russian fox "star gene" experiments.  The results were
duplicated in red fox, mink, artic fox, and in other carnivores and
several rodent species.
 
The hypothesis is that selecting for tameness alone triggers the process
that ultimately leads to domestication.  I could document the process,
but since your question was regarding the regression to the means aspect,
in order to save time and space, I'll hold off until some other time.
 
If I understand regression to the means correctly (I am not a math
major), each generation of offspring would move towards the mean
(tameness), BUT they would show LESS overall tameness than the maximum
shown in the initial generation.  For example, if only tall parents are
bred, it results in tall offspring, but with a mean shorter than the
original parents.  In domestication, the overall degree of tameness is
increased, rather than finding that some some animals become tamer, while
others become slightly less tame.  This shows the population is not being
compressed towards a mean, but the frequency of expression is being
skewed towards a single end of the bell curve.  The way I understand it,
if this were a true regression to the means, the offspring of really tame
ferrets would less tame (moving towards the mean), while at the same time
the offspring of really wild ferrets would become tamer (again, moving
towards the mean).  In the fox experiments, ALL animals move toward the
tameness end of the behavioral continuum; wild fox become tamer, but tame
fox become even MORE tame.  This is NOT moving towards a mean, but rather
towards a single end of the continuum.
 
Still, only breeding tame animals should signal to ANYONE of the danger
of regression to the means.  You have a sharp eye, and should be proud
of your observation.  If domestion was ONLY the skewing of a single
behavioral trait, I would side with your observation.  However, since
specific structural changes can be documented in the migration of
melanocytes, in brain structure, and in the production of specific brain
chemicals, I would reject regression to the means.  We are not just
skewing a behavior, but rather a suite of physical, biochemical, and
behavioral changes.
 
Because the efforts have been duplicated in other lineages, by other
researchers from different countries, and in other species, regression
to the means is probably not taking place.  Understanding exactly HOW
breeding for tameness impacts so many physiological systems is another
matter and another post.  I apologize to any math majors if I have
misinterpreted or poorly explained this effect, and if so, I hope they
clarify my blunders.
 
Bob C
[Posted in FML issue 4126]