300 THE INDIAN JOURNAL OF VETERINARY SCIENCE AND ANIMAL HUSBANDRY [VI, III

of the advantages of selecting breeding-stock homozygous for high productivity
if less than 1 per cent of the population is of that genotype, and if one is forced by
reproductive rates to save 10 or 50 per cent of the whole population for breeding.

The second point concerning the intensity of selection is that such intensity
is weakened (much more than is generally realized) by the inclusion of more and
more items in the ideal; that is, by considering many different characteristics in
making the selections. In the idealized case, where n independent, equally
important characteristics are to be considered in the selections and x is the pro-
portion of the population which must be saved for replacements, the selection
intensity for each characteristic singly is the same as if the nth root of x were the
proportion which must be saved. Thus, if one needs only to save one-tenth
of the offspring, but pays equal attention to eight independent characteristics, the
intensity of selection for each such characteristic is no greater than if attention has
been paid to the characteristic alone, but it had been necessary to save three-
fourths of all offspring born. This, I think, is the only general basis of real anta-
gonism between breeding for production and breeding for ' fancy points ' namely,
that each additional point considered must necessarily weaken the selection which
might otherwise have been practised. Nevertheless, the practical breeder must
pay attention to several things, even though he understands perfectly this weaken-
ing effect.

Besides estimating how intense his selections can actually be, one needs (if he
is to estimate the outcome at all accurately) to know first what portion of the
observed variance is genetic in the narrow sense, which includes only those gene-
combination effects that can be expressed by some additive scheme ; second, what
portion of the variance is due to gene-combination effects that cannot be expressed
additively ; and third, what portion of the initial variance is purely environmental
in origin. Of course such a division is not absolutely correct to the last detail,
since in actual practice there will be interactions between the three kinds of
variance. For instance, some genotypes may be more plastic to environmental
influences than others. If that be the case, and if the breeding system increases
the frequency of the more easily influenced genotypes, then the environmental
portion of the variance will likewise increase, although the environment does not
change. If extreme refinement is important, and if the necessary data are avail-
able, a more detailed division may be made specifying how much variance is due
to such interactions. However, in most practical problems, such refinements will
usually produce but small gains compared with the considerable amount of
approximation still remaining. Usually, the first approximation given by dividing
the variance into the three parts named above will be sufficient for the purpose
of practical application.

Only the genetic variance that can be expressed additively is subject to simple
mass selection. The refinements of selection methods, such as the progeny or the
judicious use of pedigree information, serve mainly to reduce somewhat the errors