Sampling and Subsampling - Seed Testing - page 37

Several kinds of mechanical dividers are commonly used for subsampling. Any of these dividers will
yield a representative sample; however, each has certain advantages over the others. The Boerner divider
is probably the most common; however, some chaffy grasses and other non free-lowing seed will not low
through it. The Gamet Precision divider requires electrical power to operate and is more suitable for free-
lowing seed and for certain seeds that do not low well through the Boerner divider. For non free-lowing
chaffy grasses, such as gramma grass, the Hay-Bates (also called the chaffy grass divider) or similar divider
such as the rifle divider should be used. For seeds of fuzzy or non-delinted cotton, super chaffy species such
as bottle-brush squirreltail, or extremely small-seeded species such as Juncus or Begonia , hand-dividing
methods of subsampling may be necessary; however, great caution must be taken to obtain a representative
sample.
Sample Size
The size of the sample to be submitted needs to be large enough to provide adequate seed for all desired
quality tests. In most instances, it should be large enough to provide enough seed for purity, germination,
and noxious weed seed testing. Whatever the minimum size, it should be larger than the absolute minimum
needed to allow for random selection during the subsampling process after it arrives in the laboratory
where it is to be tested. For certain tests (e.g., pathological or genetic purity testing), larger samples may be
needed, depending on the species. Table 3.1 provides suggested sample sizes (weights) for representative
species of different seed sizes.
Since the original sample drawn will generally be too large for the submitted sample, it should be
divided further before submitting to the seed laboratory. Further subdivision of free-lowing types should
be done by one of the dividers shown in Figure 3.6.
Sampling and Variation in Test results
Because of random sampling variation some contaminants that are present in the seed lot may not be
present in the submitted sample, and some contaminants that are present in the submitted sample may not
be present in the tested (working) sample. This kind of variation can cause differences in the test results
between two subsamples (sister samples) taken from the same submitted sample even if the analysis is
made by the same analyst.
Random sampling variation occurs because of the random distribution of contaminants within a seed
lot or a submitted or working sample. This kind of variation cannot be avoided entirely; however, it can
be minimized by cleaning the seed lot well and by assuring the homogeneity of the seed lot by thorough
mixing and blending.
A statistical procedure involving tolerances has been developed to determine whether the difference
in results between two tests is signiicant due to errors in sampling procedure, testing method, human error,
or due to unavoidable random sampling variation. Tolerance calculations take into account (allow) random
sampling variation, but usually not human or equipment errors. There are tables of tolerances in the AOSA
and ISTA rules for purity analyses, germination, TZ, and other seed quality tests. See Chapter 13 for more
details.