Re-evaluation of ISTA tolerance tables for germination tests in diff erent laboratories

The data used to estimate the additional variance between laboratories in seed germination test results is now more than 50 years old. It is the scope of this article to model this variance with recent data from ISTA proficiency tests and to perform several comparisons and statistical analyses. It was found that the variation of replicate results within the same laboratory is significantly underdispersed while the variation of germination test results between laboratories is significantly overdispersed. After outlier exclusion according to Miles, there was no significant difference in the empirical variation between laboratories which Miles found and the empirical variation between laboratories in the new data set. When the species were divided into two groups, one with relatively high variation in between laboratory results (comprising mainly legumes) and with relatively low variation (comprising mainly grasses), this extension improved Miles' model significantly but the validity of this grouping structure should be verified in further analyses due to noise in the data. A trend in the variation between laboratory results might exist; the overdispersion of between laboratory results seems to become less in recent years.