No. 641] 



THE DURATION OF LIFE 



487 



the food has partly cooled sprinkle on top of food the smallest possible 

 amount of pulverized dry magic yeast (shaken from a can with one pin 

 hole in cover). Put in breeding bottle a folded square of filter 

 paper, and stopper with cotton batting. 



The purpose of the filter paper is to furnish a dry 

 place for the larvae to crawl up and pupate ou, and also 

 to absorb some of the excess moisture wliieli often forms 

 on top of the food from llic ui-owiiiu' ymsl. l^^ilter paper 

 has not been used in the sm.-ill (hii-;ii ion of lif(^ bottles, 

 since no young are inipatiiiK tliei-e, and since it furnishes 

 too many hiding places for the flies in the frequent trans- 

 fers which have to be made in the duration of life tests. 

 Excess of moisture on top of the food may become a 

 source of ei-ror in duration of life experiments, because 

 flies may drown in a small drop of water. Throughout 

 our work we Inn e been constantly on guard against this 

 source of error and have tried a number oi' ]")lans, with 

 varying degrees of success to eliminnte it cntii-clw Some 

 of these experiments will be I'cportrd on in (i«'t;iil l;iier 

 on in this paper. In general it ina\ lie >ni(l In ri' that 

 this source of error from flies drowning need hcn vr W a 

 significant one if due precautions are taken. Wr know 

 that it has not been in our work. 



Occasionally the yeast l)ecomes too actixc at the edge 

 of the food and canses the whole food ma>s to rise in the 

 bottle. In an nttempt to rliminatc this nrri(h'nt yeast in 

 dilute solutions w;is ;id(h'd to the itoih'd hannnas and agar 

 and was ^].ra>ed on top <.r thr lood. Tli.. n-ults were 

 not parth-nlarh lax oi-al)!e. Thr ] .ul\ cri/cd .Irx > raM 

 added In the moM minnte M"afilitx po..Ihh- ti;. n.o.t 

 satisfaetory Manchinl nn-thod > et foun.l. We expect to 

 ccmtinue attempting to get the food conditions more and 

 more nearly ideal and identical in every experiment, but 

 we feel reasonably certain that in all of the experiments 

 we shall rei)ort. even includin- the very first in ])oint of 



guai-d a-a'inM accidental death N\ere Millieient to in^ire 

 statistic.-d accuracy in the results. Whatever environ- 



