146 EXPERIMENT STATION RECORD. [Vol.40 



and smaller seeds three times the regular quantity was taken unless decidedly 

 impure. This is not suggested as a new basis, but is mentioned as a case of a 

 sliding scale used to advantage. The number of seeds of certain noxious weeds, 

 such as wild oats, mustard, etc., was calculated from the larger sample, which 

 is deemed quite an important point that should be carried out for any sorts that 

 are considered of special importance. 



It is maintained that results of seed tests should be accompanied by an indi- 

 cation of their accuracy expressed by writing the probable error after the 

 result, as 95±1.5 per cent, although for ordinary reports it is deemed desirable 

 to have some form by which the results can be stated more completely. Wbile 

 the probable error represents a very definite quantity, it is pointed out that its 

 practical application is somewhat difficult, and that it represents only an even 

 chance that the true result lies within the limits indicated. The chances are 

 as great that it lies beyond this, while if the probable error is doubled there are 

 about four chances in five that it lies within those limits, or that the results 

 will still be beyond this in 20 per cent of the trials. This is regarded as the 

 smallest reasonable allowance that can be made, and for the smallest practical 

 scale of allowance for variation the figures in the table should be doubled. The 

 observations indicate that the second decimal place is not necessary for the 

 calculation of probable error in such tests. 



The amount of seed used for tests (and therefore the degree of accuracy ob- 

 tained) must be regulated by two factors, viz, the degree of accuracy necessary 

 for dependable results and the amount of work which it is possible to handle. 

 From the data presented it would seem that for germination icsts 200 seeds in 

 a single test would be advisable for ordinary work, the number being increased 

 when desired. It is very Important that the probable error be known so that 

 such adjustments may be made. 



Duplicate tests appeared to be of little value, as, so long as only the factor of 

 probability in selection was present, variation between duplicates was not sig- 

 nificant, while if other factors are present the chances are probably as great 

 that duplicates which vary but little are unreliable. The necessity of duplica- 

 tion must be governed chiefly by judgment, whether duplicates vary or not, and 

 a test of 200 seeds will often require less time and space than two of 100 each. 



The observations of Rodewald (E. S. R., 1G, p. 881) relative to "accidental* 1 

 and "systematic" errors in germination and purity tests are briefly reviewed, 

 and his results and conclusions are compared with those of the author. 



Seed Reporter (17. S. Dcpt. Ayr., Seed Rptr., 2 {1918), No. 5, pp. 8. figs. 6).— 

 The principal feature of this number is a statistical report on the acreage for 

 1918, the usual yield of seed, and the general date of harvesting vegetable 

 seeds, by States and crops, as determined by the seed survey of July 1. The 

 data are supplemented by outline maps of the United States showing the coun- 

 ties in which the different crops were grown commercially for seed in 1918. 



Market notes are presented for Kansas City, Chicago, and the Eastern 

 States, and information Is given relative to transportation of seed and to ex- 

 portation of vegetable seed. The production of sugar-beet seed in the United 

 States and onion-set conditions in the Chicago district are briefly discussed. 



The usual data relative to the Imports of forage-plant seed permitted entry 

 into the United States are included. 



The revised agricultural seed law, C. P. Smith {Maryland Stn. BuL 219 

 {1918), pp. 9). — This bulletin contains the text of the revised seed law, pre- 

 viously noted (E. S. R., 39, p. 399), together with the rules and regulations 

 for its enforcement. 



