248 Profs. Percy Frankland and Marshall Ward. 



my observations on this particular point are only by the way, and 

 much too few for any valuable conclusions on the wider question. 



As regards the species or forms met with, I defer their discus- 

 sion for the present ; suffice it to say that we carefully isolated and 

 tabulated the different forms in order to experiment with them 

 afterwards (see p. 285), and that a considerable number of the indi- 

 viduals (not species) were rapidly liquefying forms, rendering the 

 keeping of the plates difficult. 



Series B. 



This series, part of a more extensive set of experiments to be 

 referred to later, is calculated to show the kind of changes as regards 

 the number of normal water bacteria, undergone by the river- water 

 on standing. It should be regarded, therefore, as supplementing the 

 results in the last table. 



In each case the same procedure was adopted, and of course the 

 same care in collecting the samples and making the plates, &c. 



The collected water was placed in properly sterilised and plugged 

 flasks, and allowed to stand, at the temperature given, undisturbed 

 for a number of days. 



Plates were made forthwith to determine the initial numbers of 

 bacteria per 1 c.c. of the water, and then samples taken every twenty- 

 four or forty-eight hours, for several days, to determine whether any, 

 or what, increase or decrease in the total numbers had occurred in 

 the interval. 



It was to be expected, from the publications of others and from 

 our own experience, that such increase would occur, and my pre- 

 liminary experiments showed that in this case the increase is very 

 great. Consequently we had to take precautions against having our 

 plates too over-crowded with colonies, and this we did by adding 

 pure distilled water to each sample taken for analysis in such quanti- 

 ties that the bacteria in 1 c.c. of the original water were distributed 

 through 20 c.c, and making the plates from the diluted sample. 



It is true this method involves the risk of killing some of the 

 bacteria ; but the results show that the numbers obtained are very- 

 large, nevertheless. 



It is also true that the dilution method introduces a further source 

 of error in compelling us to multiply the ascertained result — always 

 a dangerous process in statistics. However, we have been unable to 

 see any way out of this difficulty, and have relied rather on the 

 general results expressed in the tables than on the actual numbers, 

 which it is evident must be looked upon as approximations only. 



