30 



SCIENCE 



[N. S. Vol. XXVIII. No. 705 



by rolling the tube on ice the gelatin with the 

 sand and bacteria was cooled on its inner sur- 

 face. The sand used in this filter was between 

 .5 mm. and 1 mm. in diameter. 



The other type of filter tested consisted of 

 two short tubes, 1.5 cm. in diameter, arranged 

 in tandem, each containing 2.5 cm. of fine 

 sand, between .1 mm. and .3 mm. in diameter. 

 The sand in each tube was supported by bolt- 

 ing cloth on a perforated rubber stopper and 

 the tubes were connected by rubber tubing. 

 The apparatus is shown in Fig. 2. After 

 drawing air through this filter, the sand from 

 each tube was shaken out into ten cubic centi- 

 meters of sterile water and, after thorough 

 agitation, aliquot portions of the water were 

 plated. This method is essentially the one 

 used by Soper and by most recent observers. 



Each of these filter methods is open to possi- 

 bilities of error. Bacteria may be drawn com- 

 pletely through the filtering layer in either 

 case ; and in the second method there is danger 

 that bacteria filtered out may not be separated 

 from the sand or bolting cloth. My object 

 was to find out the magnitude of these errors 

 by direct comparison with the culture-bottle 

 method. For this purpose a number of ex- 

 aminations were made, of normal air, and of 

 air artificially infected with bacteria by spray- 

 ing with emulsified cultures. With the filtra- 

 tion method samples of 750 c.c. to 1,500 c.c. 

 were slowly drawn through the sand, the filtra- 

 tion occupying from two to three minutes. 

 With the culture bottles, samples of 100 c.c. 

 were generally used and the air was drawn in 

 more rapidly. The general results obtained 

 may be shown best by quoting a few typical 

 experiments in detail. 



Experiment III. — Examinations of air of 

 a city street on a winter day. Four suc- 

 cessive samples taken at intervals of fifteen 

 minutes showed: (1) 3 bacteria per liter, by 

 culture-bottle method; (2) lY bacteria per 

 liter by filtration method (fine sand) ; (3) 23 



bacteria per liter by filtration method (fine 

 sand) ; (4) 94 bacteria per liter by filtration 

 method (fine sand). Apparently the number 

 of bacteria in the air was increasing during 

 this experiment; but the results by the two 

 methods are concordant. 



Experiment IV. — A suspension of a culture 

 of B. coli was sprayed into a box and five 

 samples taken at intervals of about ten min- 

 utes. The results were as follows : (1) 2,640 

 per liter by filtration method (fine sand) ; (2) 

 100 by filtration method (fine sand) ; (3) 740 

 by culture-bottle method; (4) 40 by culture- 

 bottle method; (5) by sand-filter method 

 (fine sand). Evidently the bacteria were set- 

 tling out rapidly. With the exception of the 

 low sand-filter count in No. 2 the results of 

 the two methods check fairly well. 



Experiment V. — B. coli was sprayed into a 

 box four times, at intervals of about ten min- 

 utes, a sample of the air being examined after 

 each spraying. The results were as follows : 

 (1) 175 bacteria per liter, by sand filtration 

 (coarse sand) ; (2) 4,300 per liter by sand 

 filtration (fine sand) ; (3) 4,000 per liter by 

 sand filtration (fine sand) ; (4) 10,750 per liter 

 by culture-bottle method. Very probably the 

 repeated spraying more than balanced the set- 

 tling out and the number of bacteria in the 

 air of the box actually increased. The first 

 result with the coarse sand seems low, however. 



Experiment YI. — B. prodigiosus was sprayed 

 int» a box three times. The first two samples 

 were examined after the first spraying, the 

 third and fourth samples after the second and 

 third sprayings, respectively. Results: (1) 

 15,000 bacteria per liter, by sand filtration 

 (fine sand) ; (2) 14,000 per liter by culture 

 bottle method; (3) 5,300 per liter by sand 

 filtration (coarse sand) ; (4) 14,000 per liter 

 by sand filtration (fine sand). Again the 

 filtration method checked with the culture 

 bottle method when fine sand was used, but 

 gave low results with the coarse sand. 



