Agriculture Reference
In-Depth Information
Table 4.2 shows ISTA tolerances for determination of other seeds by number for tests on different
submitted samples, the second submitted sample being made in the same or different laboratory. These
tolerances are based on a one-sided test at the 5% level of signiicance where only higher levels of other
seeds than that indicated on the label are of concern. These tolerances apply if a 95% level of conidence is
desired and where a 5% probability of rejecting properly labeled lots (Type-I error) for containing in excess
of the labeled rate of other seeds is acceptable.
Tolerances for zero noxious Weed Seeds (Incidental Contamination)
In the 1961 Yearbook of Agriculture, Justice and Houseman (1961) pointed out that it is not realistic to
enforce prohibitions against the sale of seed lots labeled as containing zero noxious weed seeds, and that
a tolerance must be applied when seed is labeled to show no noxious seeds. Elias et al., (2000) reported a
similar approach. However, this is somewhat controversial. Just because tests on one or more samples indi-
cate a seed lot to be free of noxious weed seeds does not rule out the possibility that a noxious weed seed
may be found in subsequent samples. Niffenegger and Cox (1972) have reported the probability of negative
results in tests on samples contaminated with different levels of noxious weed seeds. They reported that
even for a noxious weed sample of 25,000 seeds, 10 percent of samples drawn from lots containing as high
as 21 contaminating seeds per pound could be expected to be free of noxious weed seeds.
The use of a tolerance for lots labeled as containing zero noxious weeds (or incidental contamination)
is disallowed in some if not all European countries. In Hungary, for example, the examination of as much
as 10 kg of lucern (alfalfa) and clover seed lots is conducted with the aid of an electro-magnetic machine
to “guarantee” complete freedom from incidental contamination by dodder. Although such lots may not
be guaranteed as dodder-free, they are disallowed for sale even if one dodder seed is found by oficial seed
inspectors.
Effect of Sample Size
Variability in test results is usually measured by the standard deviation of test results from repeated sam-
ples. The likelihood of committing a Type-I error (rejecting a properly labeled lot) may be reduced by either
increasing the size of the sample examined or by increasing the tolerance. However, since variability in the
occurrence of other seeds follows the Poisson distribution, true random sampling variation is independent
of the number of noxious weed seeds present in the sample, assuming a homogeneous population and cor-
rect sampling procedures. The factor involved in determining the tolerance is the estimate of the number of
noxious weed seeds in the seed lot based on a second test.
ToLErAnCES for GErMInATIon TESTS
Germination tolerances follow the binomial distribution. According to Miles (1963), germination toler-
ances apply to any one of the following: (1) percent normal seedlings, (2) percent abnormal seedlings, (3)
percent dead seeds, (4) percent hard seeds, and (5) the sum of any combination of these attributes. Miles
further stated that “theoretically, germination tolerances should be computed allowing for random sampling
variation only.” Thus, he computed tolerances for tests made in the same laboratory, based on experience
that variation among replicates in one laboratory had been found to be within that due to random sampling
variation. Miles went on to say that “experience has shown that the variation among laboratories has been
greater than that due to random sampling variation alone. Therefore, to be realistic, the tolerances he cal-
culated for tests made in different laboratories accounted for the amount of variation due to the amount of
“interlaboratory bias” or variation that existed in the 1950s. Since these tolerances have not been changed,
the presumption must be made that the amount of “interlaboratory bias” which existed in the 1950s still
exists today.
