Heritability

Heritability plays a vital role in the formulation of breeding plans for animal and plant improvement. An
important aspect of these plans is selection, that is, the choice of parents to produce the next
generation, on which the improvement depends. For the selection to be effective it is necessary that the
members of the population on which the selection is practiced vary in their genetic make-up with regard
to the character in question. In other words, it is only the genetically determined variation which can be
utilized for a permanent improvement of the production characteristics in a population. If all (or most) of
the variation existing in the population is attributable to environment, selection of phenotypically
superior individuals will not result in any (or material) alteration in the genetic composition of the next
generation. Hence, a basic pre-requisite to the planning of a breeding programme is that of the total
variability existing in the population, how much of this is caused by differences in the genetic make-up of
the individuals.

Estimation of the Heritability

As we know heritability is the ratio of the additive genetic variance to the total variance, and the additive
genetic variance is the chief determinant of the resemblance between relatives, this suggests that the
heritability can be derived from estimates of either a regression coefficient or an intra-class
correlation among related individuals. This is the basic principle on which all the different methods of
estimation of heritability rest. In addition to this direct approach, heritability can also be estimated from
the results of selection experiments. The selection differential, S, which is the mean phenotypic
superiority of individuals selected as parents, and the response to selection, R, which is the deviation of
the mean of the progeny generation from the original mean, are related to heritability as h2 = R/S .If R
and S are known, then S2 can be estimated as the ratio R/S, which is known as realized heritability.

Commonly used estimates based on regressions and correlations among related individuals along with
details about their estimation from the observational data are given below.

(i) Regression Estimates

Let bOD

denote the regression of offspring on dam, bOS

of the offspring on sire and bOP

that of offspring on mean of parents, obtained from a given body of data. Then three different estimates
of heritability can be obtained as follows:

(i) 2 bOD ; (ii) 2 bOS ; (iii) bOP

with sampling variance as 4V ( bOD ), 4V ( bOS ) and V ( bOP ) respectively, where

b = xz

x2

and

V (b ) =

z 2 b xz

( N 2) x2

(2.3)

in which x denotes the performance record on parent (or mid-parent) and z on the offspring, both x
and z are measured from their means, and N is the number of paired observation of x and z.

In the absence of maternal effects the genetic composition of all the three estimates is same and is equal
to

1
2+12

+12

+ ... .

2 A AA

4 AAA

Thus, all these estimates are nearly unbiased estimates of heritability in the narrow sense, because these
include only a little epistatic variation and no dominance variation. When maternal effects cannot be
ignored it is safer to go for the estimate based on bOS , which remains unaffected by maternal effects,
if any. This estimate is less commonly used because fewer degrees of freedom available for it and its
non-applicability to sex-limited traits (e.g. milk yield).

It may be noted that these estimates are free of the environmental sources of covariance as the parents
and offspring are usually measured over different time periods. Also, these