44 



actual practice a difference of opinion as to the percentage of broken 

 seeds, even to the extent of 10 per cent, in a sample, would not aileot 

 uniformity of result if expressed in terms of pure germinating seed only, 

 and if, in the report, broken ' seeds and broken seedlings are classified 

 together, as they ought to be. 



Small Seeds.- — The ratio of the weight of seeds of Red Clover sieved 

 through a mesh of 1 • 25 mm. to those retained over this mesh varies with 

 the sample from 50 to 70 : 100. 



Small seeds of Red Clover contain a larger proportion of hard seeds 

 than do large seeds, but the proportion is not such as, in actual practice, 

 to make any great difference to the result, whether taken by weight or 

 numerically. 



Shrivelled (dead) seeds and immature seeds may be conveniently 

 considered together, as in general it is quite impossible to separate these 

 with any accuracy into germinable ajid ungerminable without reference 

 to a germination test. In Red Clover, the weight of a definite number 

 of shrivelled seeds is, approximately, 20 per cent, less than the weight 

 of the same number of good seeds. 



Empty glum,es — deaf seeds — shelled caryapses. — In samples of Cocks- 

 foot, the weight of a definite number of deaf seeds may vary from ^ to ^ 

 of the same number of full seeds, and in samples of Lolium from J to \. 



For the purpose of obtaining a true estimate of pure germinating seed, 

 which shall be a sufficient approximation to a weight percentage of live 

 seed, it is imperative that the " light seed "of greisses should be excluded 

 from the germination test. 



This is the more so because it is quite possible that a sample of 

 Cocksfoot, for example, may contain both the shelled kernels and the 

 husks from which they have fallen, and it is absurd to test both of these 

 for germination. 



The division of spikelets may, however, be carried to an unnecessary 

 degree of exactness, and it is here suggested that the conventional method 

 of treating a species such as Cocksfoot shoiild, in order to reduce the 

 work per sample, be modified to the extent of leaving intact the ultimate 

 deaf seed where it occurs along with a full seed. 



The difference between the percentage of pure germinating seed calcu- 

 lated by weight, and the percentage of pure germinating seed calculated 

 by number (presuming that the first mentioned calcxilation could in effect 

 be accurately made) reaches 5 per cent, only when either 20 per cent, of 

 the sample is ungerminable and the average ungerminable seed is 30 per 

 cent, less in weight than the average germinable one, or when 30 per cent, 

 of the sample is ungerminable and the weight of the average ungerminable 

 seed is 20 per cent, less than that of the average germinable one. 



This difference only exceeds the usual latitudes for germination when 

 the germination is less than 70 per cent, and, at the same time, the 

 average weight of an ungerminable seed is less than 60 per cent, of a 

 germinable one. 



In actual practice, samples which would show a difference of 5 per cent, 

 between a determination by weight and a determination by number are 

 never met with except where the so-called Irish method is practised on 

 samples of species such as Dactylis, Lolium, Poa and Alopecurus. 



The following is an account of calculations applied to a fairly normal 

 sample of Red Clover which had an ascertained purity of 99-3 per cent, 

 (broken seeds being reckoned pure seeds). The sample contained : — 



By Weight. Numerically. 



Broken seeds - - - 3-3% 3-7% 



Small seeds ... . 4.70^ 8-6% 



Large sound seeds - - - 91-3% 87-0% 



99-3% 99-3% 



Impurities - - . 0-7% 



100 



