USE OF
TOLERANCES IN SEED TESTING 1. Testing of the same seed lot in different laboratories. 2. Seed lot testing under seed certification programme. 3. Seed lot testing for checking of seed certificate under seed law enforcement. 4. Statutory testing of seed lot for law enforcement to be used as evidence in law courts.
For the above situations and to ensure that analysis conducted within the laboratory is accurate, it in necessary at the part of the analyst to make sure that the result being reported is valid/reproducible. The application of an appropriate statistical method to test the results of seed testing enables the analyst to determine the validity of results within a calculated range of limits, the amount of this range in called the 'Tolerance'. For seed quality determination, replicated test is conducted from same sample or different samples drawn from the same lot and replicated tests are conducted in one laboratory or different laboratories. It is generally observed that the test results may not be the same.
No two samples taken from the same seed bag or same seed lot are likely to be identical. The size of a lot varies depending upon the seed size. Experience world wide, has shown that variations do occur in sampling process, among replicates even when recommended sampling procedures are used. However experiences have shown that variations between the laboratories have been greater than that due to random sampling variation. There are five main sources of variation are recognized" (a) bag to bag variation
(b) In bag variation
(c) Working sample variation
(d) Between analyst variation
(e) in analyst variation.
B. Basic Assumptions for Using Tolerances: There are three basic assumptions to use tolerances (a) The seed lot from which the sample is drawn should be relatively homogeneous (b) The sample must be drawn randomly as per seed sampling recommended procedure from container or locations in the lot (c) Bias must be avoided in conducting test.
C. Where to Apply Tolerances: C.l. Within Seed Testing Laboratory: Before release of seed analyst’s reports, a seed analyst has to make sure about the accuracy and reproducibility of results. Several situations normally arise in the seed. testing laboratory when tolerance should be applied to decide the validity of the tests. Some of them are: (a) Germination test: To decide whether the replicate wise performance of a germination test is comparable or retest is needed. (b) Purity analysis: Comparing the replicate wise results of physical purity of different components and to decide whether the test conducted is valid or retest is necessary. (c) Referee Test: To ensure that seed testing laboratories are achieving reproducible results, referee testing is arranged and tolerances are applied to evaluate the accuracy of their results. C.2. For Certification: (a) The Central Seed Committee (CSC), Government of India has prescribed the minimum seed certification standards for various crops. As a practice these standards are also taken as such for label information, even if the seed analysis results may be higher than the prescribed standards, the actual prescribed standards are only mentioned on the table. (b) Competition in the seed trade based on difference in seed quality in future is very likely. Moreover, the farmers' awareness for seed quality measurements may necessitate labeling as per the actual seed analysis report. Thus label information can be well above the prescribed seed standards or at par with the standards. In either case seed is fit for certification. (c) The seed standards are so formulated and prescribed that they ensure a reasonable level of quality for the seed user and that they can be achieved by a majority of seed producers i.e. it is neither kept high or very low. Therefore, for initial certification tolerances are not applied. (d) According to the prevalent practice, certification is valid for a period of eight months. On expiry of this period the certification agency draws a fresh sample of the lot and sends it to the laboratory .On receipt of results decision on extending the validity period i.e. revalidation/ recertification is taken. At present, revalidation is done only if the seed analysis results are at atleast meeting the prescribed seed standards without the application of tolerances. But the same situation is being viewed differently for the purpose of seed law enforcement for assessing the accuracy of label information tolerance tables are used. (e) From the above point (d) following points become clear: (i) For seed law enforcement, tolerances are used and thus labels are retained even if the laboratory report is less than the label information but difference is within tolerances units; and (ii) For revalidation/recertification, tolerances are not used and thus lot is declared as unfit if the laboratory result is lower than the prescribed standards irrespective of whether the difference is within or out of tolerance. C.3. Outside seed testing laboratory: (a) Seed law enforcement Label prescription: Under seed law enforcement, the information given on the label affixed/attached on each container of the lot may be above the prescribed seed standards but under prevailing practice in India the information about seed quality attributes given on the label is the same as per prescribed seed standards for each seed quality attribute. In several other countries the seller normally sells seeds at higher price when label information is above the prescribed seed standards. In such situations the analysis results of seed samples drawn by seed inspectors are compared with the actual information given on the label and tolerances are applied to determine if label information is close enough to analysis results. (b) Seed law enforcementSeed standard prescription:
The central seed
committee has prescribed crop wise minimum seed certification standards
class wise for labeling seeds. At present the seed certification
standards are as such taken for labeling, therefore information on the
labels given as per prescribed standards even though the sample might be
recording higher germination and purity than the standards. In other
words for law enforcement, tolerances are applied whether or not the lot
is at least equal to the prescribed minimum limit given on the label.
Several tolerance tables (1 to12) have been developed and are available in the reference materials such as the International Seed Testing association's "Hand Book of Tolerances and Measures of Precision for Seed Testing" (1963) and international Seed Testing Rules (1999.) Separate tolerance tables are available for different situations. Depending upon the situation the tolerance table to be used is decided. D.I. Use of Tolerance Table within and between laboratories Test Results (a)How to Use For Purity Analysis Results: To compute the tolerances of the components of purity analysis it is necessary to have the information about the magnitude of variation associated with each source of variation, which affects the percent, estimated on a sample from the seed lot.

Table1:Tolerances for comparing duplicate working samples from the same submitted sample for any component of purity for either chaffy or non chaffy seeds, at 0.05% probability.
*..... Indicates the other values in the series. Table2:Tolerance for any component of purity analysis between two laboratories , at 1% probability
*.... Indicates the other values in the series.
*... Indicates the other values in the series. 
Examples of Use of Tolerance Table for Purity Analysis Results: (i) To ensure accuracy in a laboratory, it is often desired that the submitted sample is divided first into working samples. Two different seed analysts then analyze each working sample. Suppose purity percentages of the whole working samples in two tests (by the two analysts) were 98.5 and 96.2. The question thus arises whether their difference is acceptable. To determine, this add the two values (98.5+96.2=194.7) and calculate the mean (194.7/2=97.35). Now in column 1 of table 1 find the range that compares with the average figure; it is 97.2597.49. The tolerance value given in column of the table is 1.20. The difference between the two working sample is 98.596.2=2.3. Since this difference is more than the tolerance the results are not equal, comparable or acceptable; hence fresh test has to be conducted by drawing another working sample. (ii) Sometimes samples of the same seed lot may be tested by two different laboratories. For e.g. one laboratory finds 98.5% pure seed and the other finds 96.2% pure seed. To find out whether these values are within tolerance and acceptable, calculate the average, refer table 2 and identify the range in which the average falls. This is 97.25 97.49 and the tolerance permitted is 1.39 for nonchaffy seed and 1.63 for chaffy seeds. The difference between the two laboratory results is 2.3. Therefore it is out of tolerance for both chaffy and nonchaffy seeds and the analysis should be redone. (iii) The examples elaborated in (i) above relate to purity analysis results reported on weight basis. Foreign seeds are also reported by number per unit weight. Tolerance
values are determined
as given above but using table 3 this table can be used in
comparing the number of seeds of single species or the total of two
or more species.
Germination tolerances should be computed allowing for
random sampling variation. Among the many causes of significant
differences between or among germination tests are 
Table4:Maximum tolerated ranges in germination percent for deciding whether to retest; allowing for random sampling variation only
*.... Indicates the other values in the series.
*..... Indicates the other values in the series.
*...Indicates the other values in the series. 
Examples of Use of Tolerance Table for Germination Test Results: (i) The germination test in a seedtesting laboratory is always conducted as replicated tests of 100 seeds each. The numbers of replications are normally four; for official evaluations such as in seed law enforcement work, the replications may be three; and for service sample it may be two. For working out the mean it is necessary that results of the replicated analysis are reliable and within the acceptable tolerance limits. For example, if the four replications showed germination percentage (Normal seedlings) as 90, 92, 93 and 83; the average worked out to be the 90 percent. In table 4 column 1, locate the range in which the average percent germination value of 90 falls. It is (8990). The maximum tolerance permitted amongst replicates in column 3, is 12. The maximum difference between these replicates is 9385=8. Thus the difference between the replicates is within the tolerance limits and therefore the results of the test are acceptable for adding and working out the average. (ii) Tolerance table5 is used when two series of replicated tests are made on a sample; for example, for a seed lot the first test replicates averages 85% and the second test replicate averages 89%. The average of these two independent tests would be 87%, while the difference between the average of the two series of tests is 8985=4 %. In table 5, column 1, the average percent germination of 87% falls in the range 8590%. The tolerance permitted here is 5, the difference between the average of the two series of tests is only 4; the results are hence within tolerance and therefore, the average of two tests can be reported. If the difference would have been otherwise then one more test should be made (iii) Table6 is used when the germination test results are compared between two laboratories. The details of the use of the Table are the same as given in (ii) above. E. Use of Tolerances under Seed Law Enforcement Programme: The abovementioned six different tolerance tables are used for different situations. The basic principle in the use of these tables is that they are used only when the results reported by the laboratory show lower germination or pure seed or higher inert matter, other crop seed or weed seed content than what is given on the label. But, during certification, labeling is done only when the laboratory results are in conformity with the prescribed seed standards. Thus on the label either the actual results or the prescribed seed standards may be printed. Under seed law enforcement two different situations may thus arise; (i) Comparing laboratory's results with the details on labels; and (ii) Comparing laboratory's results with the prescribed minimum seed standards indicated on the table. Specific tolerance tables for individual situations are, therefore, to be referred. (a) Table no.7 is meant for comparing the laboratory's results of purity test with the details on label. To use this table both the laboratories results and the details on label should be in percentage by weight.
(b) Table no.8 is used to compare laboratory's results on other crop seed and weed seed by number per unit weight with the details on label. To use this table both the laboratory's results and details on label should be in number per unit weight.
(c) Table no 9 is for comparing germination results reported by the laboratory with the details on table.
(d) Table no.1O and 11 are meant for comparing laboratory's purity analysis results with the minimum standards prescribed.
(e) Table no.12 is to be used for comparing laboratory's germination results with
the minimum seed
standards prescribed. The use of table no7 to 12 is elobarated by taking the following example
f) From the above (a to e) two points emerge:

Comparing the laboratory results with the label information: (i) Table 7 is meant for any purity component is percentage by weight. The following steps are involved in verifying whether or not the difference between laboratory result and the details on the label is within tolerance. Work out the average of laboratory results and the percentage on label: Pure Seed: 99.0 + 97.0 = 196 Average=196/2 = 98.0 (i) The difference between the laboratory result and the percentage indicated on the label is (99.097.0)=2.0. This difference is higher than the tolerance level and therefore, labeling is wrong. (ii) If the level of weed seed or other crop seeds are given in number on the label, Table 8 can be used to determine the accuracy of labeling for seed law enforcement. If the information given on the label compares satisfactorily with the number found upon testing labeling is regarded as acceptable. (iii) Table9 is used for comparing the laboratory result with the information given on label for germination. The procedure is: (iv) Calculate the average of the two viz. laboratory germination and the information on label; i.e. 80+94/2 = 87 (v) Fit the average suitably in Table 9, Column A & B as the case may be and find the tolerance level against this. The tolerance given in column C against the average viz, 87 is 6. (vi) The difference between laboratory result and the percentage indicated on label is 9480=14. This is higher than the tolerance level and therefore labeling is wrong. (g) Comparing the laboratory result with the minimum standards prescribed: (i) Table10 is meant for any purity component in percentage by weight. The procedure is: Work out the difference between the laboratory result and the standard prescribed; for example, pure seed. Laboratory result: 97.0 Standard: 98.0 Difference: 1,0 Tolerance level in column C, Table 10 against 98% is : 0.61 .The difference between the laboratory result and the standard in more than the tolerance level; hence lot is not fit to be sold as standard seed. (ii) The same procedure is followed to compare the information on other crop seeds reported by the laboratory in number per unit weight. Table 11 is referred for this purpose. (iii) For germination Table 12 is used. The difference between the laboratory germination and the standard prescribed in worked out. For example, it is (8580)=5. Taking the number of replicates as four of 100 seeds each the total number of seeds in 400. Fit the standard prescribed viz, 85% suitably in Table12 column C. The tolerance is 6 while the actual difference is only 5. Thus the laboratory result is in conformity with seed standards and seed lot can be soled as standard seed. 
(e) The same seed lot was found to be wrongly labeled when the comparison was made between the information given on the label and the laboratory result.
Table7 Tolerances for comparing the results of a purity test with the label information (when both the laboratory results and label information are in percentage by weight)
Table8: Tolerances for comparing the foreign seed numbers found in laboratory test with label information (when both the laboratory result and label information are given in number per unit weight)
A=Average of laboratory's result and label information B=
Tolerance Level
Based on 5% probability
Only the 5% probability column is used
A=Standard B=Rejected
Only 5% Probability column is used 