more numerous original variables since they explain most of the variation in the
original data.
In practice a random sample of F3 or F4 lines is taken from a population in
which numerous plant type recombinations occur. From field experiments metric
data are recorded on an array of what are judged a priori to be important
morphological and physiological contributors to yield. There will be patterns of
characteristics that tend to be more highly correlated inter se arid with yie'd. The
analysis will group the variables into sets or factors, and from inspection the
investigator will interpret the factors in a biological context.
Some recent applications may help to clarify the method:
(1). Morishima and co-workers (1967) identified two principal factors related
to yield in rice, one expressed as "a panicle-length-panicle number" axis, in which leaf
length and width were also prominently involved, and a second factor characterized as
"mean internode length-elongated internode number," involving also leaf angle. These
are the morphologic patterns which characterize the new high yielding rice varieties
of south-east Asia.
(2). Walton has just reported (1973) the results of a factor analysis performed
on 14 characteristics measured in a 5 x 5 diallel of spring wheat. The 14 variables
were reduced to four main factors. Flag leaf area and duration of its activity were the
principal variables in Factor 1. The second factor was a stage of development factor,
with a long filling period having a negative loading, and a long period from
emergence to anthesis having positive loading. Factors 3 and 4 included the number
of heads per plant, kernels per head, and kernel weight.
(3)., In a recently completed thesis, Denis (1971), working with bean varieties
representing both North and Central American sources, performed a factor analysis
on data collected on 22 morphological characteristics of 16 varieties grown at two
locations. The 22 traits were reducible to two main factors, and a third slightly less
important factor.
Variables with highest positive loadings on the first axis included seed weight,
pod fresh weight, pod dimensions, basal internode length, and diameters of hypocotyl,
basal and upper internodes. Factor 1 is therefore essentially a weight factor,
equating size and diameter measurements with weight.
The variables with strong negative loadings on this axis included mostly those
with strong positive loadings on the second axis, namely a group of interrelated
variables expressing a number concept number of nodes bearing pods, total
number of branches, number of racemes, number of pods, etc.
Factor 3 included significant positive loadings only from three variables,
total number of nodes per plant, number of long internodes and average length of
long internodes. This factor is a structural factor, meaning many nodes, many leaves,
and upper leaves in particular spaced far apart by long internodes.
The use of isogenic lines or populations
Varieties with contrasting plant type-components are crossed, and repeated
backcrosses are made to both parents, with selection for the plant type component-
of the non-recurrent parent in each generation. After several generations of
backcrossing and selection, lines should be obtained which are essentially idential