142 



M40 are 57.8 and 64.6 n, respectively, far below the mode of the original, 

 which was 81.6 M- M36 shows less striking difference, but this is still 

 marked. Comparison of the means shows those of M36 and M40 to be 

 approximately 17 and 18 -divisions (1 division = 3. 4 /*), while the original 

 was 23 divisions. In other words, the conidia of M36 wese-only about 

 three fourths the length of the normal conidium of H. No. 1. Such dif- 

 ferences as they appeared in the microscope are shown in PL XXVI. The 

 difference in variability is also strikingly large. 



Striking variation in conidial breadth, both relative and absolute, was 

 observed. Graphs and data of the more pronounced cases are presented 

 in Fig. Q and others are given later. In connection with Fig. Y (Graphs 

 114-138) are given summary data on the conidial length of saltants in- 

 cluded in this study. It is to be noted (Graph 6A, Fig. B) that whereas 

 the mode of the ordinary conidium stood at 20.4 /i and no conidia exceeded 

 a thickness of 23.8 /z, the modal thickness of M8-7 (Graph 75, Fig. Q) is 

 23.8 ju, with many conidia 27.2 ^ in thickness, one even 30.6 /z. Such dif- 

 ferences between saltants and the parental form are presented to the eye 

 in PI. XXVI. 



The ratio of conidial length to conidial breadth is perhaps still more 

 striking than the mere variation in length. In such variants as M6 (PI. 

 XXVI, b) and M8, while increased greatly in thickness the conidia were 

 at the same time absolutely shorter, thus emphasizing to the eye both 

 differences. The ratio of length to breadth in H. No. 1 is as follows: 



mean length _ 22.62 =*= .05 _ # 



mean breadth 6.03 .04 



while in a sample of one of its saltants this ratio is 



mean length _ 20.67 .22 ^ 



mean breadth 7.82 =*= .11 



and in another sample of the same saltant it is 

 mean length 19.58 .30 



mean breadth 7.30 = t .06 



2.67 .05 = 



*Probable error was computed according to the above formula kindly furnished me by Dr. J. A. Det- 



r> 



lefsen, where a <= probable error of A; b = probable error of B; and E = probable error of 



A 



