692 JOURNAL OF FORESTRY 



equidistant on each side to allow the insertion of wicks leading to the 

 water in the tank. The wick is attached to the center of a circular mat, 

 eight centimeters in diameter, made from all-wool blanketing and rest- 

 ing on the glass plates. Over the woolen mat is placed an equal-sized 

 mat of coarse cotton yarn, loosely crocheted, and over this is the filter 

 .paper on which the seed is laid. To allow air circulation, a hole, which 

 connects with one ground in the glass plate, is cut through all three 

 mats. The mats, filter paper, and seed are covered by an inverted glass 

 funnel. 



In this germinator the moisture conditions are regulated by the height 

 of the water in the tank, the size of the wicks, and the amount of air 

 admitted to the funnel. Heat may be furnished by alcohol or gas 

 burners, steam or hot water pipes beneath the tank. (The latter method 

 was used in this experiment in order to bring the temperature closer to 

 that of the sand tests, which were heated in the same manner.) 



For better circulation of the water and more uniform distribution of 

 the heat in the germinator, a glass plate somewhat smaller than the in- 

 side of the tank was placed horizontally within the tank, about ij^ 

 inches above the bottom. 



This germinator was used by the writer during the winters of 191 5- 

 1916 and 1916-1917 and the results were contrasted with those ob- 

 tained from the sand tests. Seeds of nine species of conifers native to 

 Montana and northern Idaho were tested by both methods and the ger- 

 minator was found to be the most satisfactory. Before considering the 

 results, however, it is important to discuss the methods used and the 

 care exercised, since these in a large measure influence the reliability 

 of the conclusions. 



The seed to be tested were separated from a larger lot by the use of 

 a sampling machine. The smaller lot from which the final seed were 

 counted was spread out one layer thick on a glass plate in such a manner 

 that the seed touched each other. From this equal portions were re- 

 moved for counting. Three samples, each containing the same number 

 of seed, were weighed separately and the average weight determined. 

 In case the weight of any one sample was more than 5 per cent above 

 or below the average for the three, new samples were taken and the 

 process repeated until the desired limit of error was reached. Three 

 samples each of 200 seeds were used in the sand and three of 100 seed 

 each in the germinator. 



The sand was sifted into the trays through a one-twelfth inch mesh, 

 after a given amount of coarser sand had been put in the bottom to 



