Agriculture Reference
In-Depth Information
Tolerances for Specific Tests
Tolerance tables are used to determine whether the difference between two test results is signiicant (mean-
ing out of tolerance) or only due to random sampling variation (meaning within tolerance). Both the AOSA
rules and the ISTA rules include tolerance tables to compare results of various seed quality tests such as
germination, purity, and TZ. The AOSA has tolerance tables for germination between 2 and 4 replicates
and tolerance between two tests, TZ between replicates and between tests, purity (regular purity and special
purity), noxious weed, seed moisture content, endophyte, seed moisture and luorescence. The ISTA rules
include tolerance tables for purity (two half-working samples, and two whole working samples), determi-
nation of other seeds by number, germination (between 4 replicates and between tests), TZ, seed moisture
determination, accelerated aging test, conductivity test, and tolerances for testing by weight replicates. The
sources of most tolerance tables in both the AOSA and ISTA is the Handbook of Tolerances and Measures
of Precision for Seed Testing (Miles, 1963). Examples of tolerances tables from both AOSA and ISTA are
included below.
In general, to determine the appropriate tolerance value, average the results from the two tests to be
compared, such as a irst and a second test or a label claim and a second test. The tolerance value is on the
line which has the average of the two tests in a given tolerance table. If the difference between the two test
values does not exceed the tolerance values, then the difference is not considered signiicant, perhaps only
due to random sampling variation, and the two test values are said to be within tolerance. However, if the
differences between the two test results exceeds the tolerance, the difference is considered signiicant, and
therefore the two test results are not within tolerance.
ToLErAnCES for PurITy TESTS
In theory, if all seed lots were completely homogeneous and equally free-lowing, and if all purity sepa-
rations were of equal dificulty, the same tolerances could be used for all kinds of seeds. However, in
practice, different seed kinds and different lots of the same kinds differ greatly. Some are nonchaffy and
free-lowing, while others are chaffy and non free-lowing. Still others contain mixtures of different types of
seed with different physical characteristics that encourage variability. Because experience has shown that
purity tests for chaffy seeds vary more than nonchaffy types, they require larger tolerances, as relected in
the AOSA purity tolerances.
The rules for testing seeds designate which genera and species are considered chaffy. For example,
wheat, clover, and soybeans are nonchaffy free-lowing types, while bluegrass, red fescue, and meadow
foxtail are chaffy non free-lowing types. A mixture is considered chaffy when at least 33% of the seed lot
is composed of chaffy types.
The application of purity tolerances is appropriate only when comparing a “irst” analysis (or test) with
a “second” analysis to decide if there is a real deiciency (i.e., signiicant difference). Usually, the “irst”
analysis/test is that which appears on the seed label, while the “second” is that obtained by a prospective
buyer or a law enforcement agency. Tolerances are not appropriate for use in distinguishing which of two
analyses are signiicantly better or poorer than the other except in the context cited above.
The Association of Oficial Seed Analysts (AOSA) has “regular” tolerances and “special” tolerances
for purity to determine whether apparent deiciencies in any component of the purity test (pure seed, other
crop seed, weed seed or inert matter) exceeds the labeled rate. The International Seed Testing Association
(ISTA) does not have “special” tolerances for purity tests.
regular Tolerances
Regular tolerances are used for either chaffy or nonchaffy seed lots which contain only one kind of pure
seed or more than one kind or cultivar which have nearly the same weight per seed as calculated by approxi-
