156 CARL DOWNEY LA RUE 



dark central cells. The difference between the mean appendage length of 

 early and late spores is 0.130.32ju. From SHEPPARD'S tables (PEARSON 

 1914) it may be found that the chances are fifteen to one that this difference 

 is due to random sampling. Accordingly it may be assumed that the two 

 sets of measurements are parts of the same frequency distribution. Thus it 

 appears that one may safely measure either early or late spores of a culture 

 without fear of introducing serious errors. In practice the cultures were 

 not measured until a vast number of spores had been produced and the 

 sample was then taken by drawing a sterile platinum wire over the fruiting 

 surface of the culture from the bottom of the agar slant to its top. In this 

 way a random sample was doubtless secured. 



The sample was now mounted on a clean slide on which a drop of a 

 weak solution of magenta-red in 40 percent alcohol had been placed. The 

 spores were mixed in the solution by stirring with the platinum wire, 

 covered, and the excess of solution removed with blotting paper after 

 which the spores and appendages were clearly defined in the remaining 

 solution. The appendages and the hyaline cells at either end of the spores 

 were stained deeply by the magenta-red, so that both spore length and 

 length of appendage could be measured accurately. The solution did 

 not cause plasmolysis or shrinking of either the spores or the appendages. 



Duplicate measurements were avoided by measuring all spores encoun- 

 tered in a path across the slide just below the upper edge of the cover- 

 glass and in another such path just above the lower edge of the glass. 

 The chances that any spore was encountered twice were thus rendered 

 exceedingly small. 



When the means of the ten cultures of the plus group and those of the 

 minus group had been measured in the manner described above, the 

 cultures with the longest and shortest means were selected to continue the 

 plus and minus selections respectively. 



As contrasted with the selection methods of other investigators, it will 

 be seen that only the range of variation happening to occur in a random 

 sample of ten variates is made available by this method. It may sometimes 

 happen that ten variates chosen at random will all fall near the modal 

 condition, but ten is a large enough number so that in general a random 

 sample of this size is likely to spread over a fair proportion of the range of 

 variation. To have made more than ten cultures in each line of descent in 

 each generation would have required an impossible amount of labor in 

 measuring. A larger number, as 100, for instance, would have insured the 

 detection of a greater number of extreme variates, but would also have 

 vastly increased the likelihood of selecting sporadic mutations rather than 



