Principles of Plant Genetics and Breeding

428 CHAPTER 23
unit: in the field, these are plots (of rows of plants); in
the greenhouse, a unit could be a pot. For example,
applying different levels of fertilizer, planting at different
spacing, or evaluating different genotypes (as in breeding),
are all treatments that elicit variable responses from
plants. However, in addition to the expected variation
from a treatment, there is always some variation that is
unintended and cannot be accounted for, and is known
in research as experimental error. By definition, this is
the variation among plots that are treated alike. Major
sources of error are those due to soil heterogeneity,
the operator’s inability to conduct the experiment uniformly
as intended, and interplant competition.
Soil (site) variability
Soils are naturally heterogeneous, some more so than
others. Natural variation may originate from differences
in soil minerals, soil moisture, organic matter, or topography.
The tops of slopes are drier than the bottom
parts; nutrients wash down and accumulate at the
bottom; depressions may drain poorly. In addition to
these natural sources of soil variation, humans create
additional variation through how they manage the soil
in production. Before selecting and using a site for
genotype evaluation, the breeder should know about
the previous use of the land (history of use). The parcel
of land of interest may have been differentially managed
(e.g., different plant species were grown, or different
tillage practices were imposed). The breeder should
use all available data (e.g., yield records, management
records) and visual observations to identify general patterns
of variation at the proposed site. Experimental
design techniques and other tactics can be used to minimize
the effects of soil heterogeneity in field trials.
As previously indicated, breeders usually have selected
sites at which they conduct their yield trials (e.g., experimental
stations of universities). Are locations then a
fixed variable in ANOVA? Some argue that locations
are random effects since the breeder has no control over
the meteorological factors that occur at locations. Most
breeders also consider years of testing as random effects.
It is important to have at least two replications of each
treatment (genotypes) in each trial for estimating error
variance.
Tactics for reducing experimental error
In order to correctly and effectively evaluate the desired
variation, the researcher should eliminate or, more real-
istically, minimize extraneous variation. Some errors
come from natural soil variability whereas others are
human in origin. The plant breeder may use certain field
plot techniques to minimize experimental errors.
Use of border rows Different genotypes differ in various
plant characteristics to varying extents (e.g., growth
rate, size, height, nutrient and moisture uptake). When
planted next to each other, interplot competition may
cause the performance of one genotype to be influenced
by another in the adjacent plot. In early generation or
preliminary trials, which usually include large numbers
of genotypes, breeders often use fewer rows (e.g., two
rows) in planting a plot in order to save resources. In
advanced yield trails, four-row plots are customarily
used. Data are collected on the middle rows only, because
they are protected from border effects. Some breeders
minimize border effects by increasing interplot spacing
or using a common genotype for planting the border of
all plots in a test in which row plots are few (one or two
rows). To reduce interplot competition, the materials
may be grouped according to competitive abilities.
Proper choice of plot size and shape Several factors
affect the optimum plot size to use in field evaluation
of genotypes (breeding objectives, stage of breeding,
resources available, equipment). Evaluation of an F 2
segregating population is often based on individual
plant performance (i.e., individual plants are essentially
plots). Consequently, the plants are adequately spaced
to allow the breeder enough room to examine each
plant.
Some breeders use what are called microplots, consisting
of planting hills or short rows of test plants. This
tactic is used in the early stages of genotype evaluation
as a quick and inexpensive way of eliminating inferior
genotypes.
Row plots are commonly used by breeders for genotype
evaluation. The size and shape depends on the plant
species, the land available, and the method of harvesting.
Generally, row plots are rectangular in shape.
Adequate number of replications Replication or repetition
of treatment in a test is critical to statistical
analysis, providing a means of estimating statistical
error. The number of replications used usually varies
between two and four; fewer replications may be used in
early evaluation of genotypes while advanced yield tests
usually have four replications. The number of replications
depends on the accuracy desired in the analysis, the