82 MISSOURI BOTANICAL GARDEN. 
A comparison of the frequencies in Table III reveals dis- 
similarities too great to be attributed solely to the probable 
errors of random sampling.’ Even the two samples of 1905 
seem to differ far more, in the frequencies of the first three 
classes at least, than can be explained by the errors of samp- 
ling from a homogeneous population. Series (a) seems to 
have a significantly higher mean than series (b). The dif- 
ference is over thrice its probable error and confirms the 
impressions from the frequency distributions. Probably the 
difference in the method of collection of these two lots of 
material introduced some unknown biological factor to which 
the differentiation is due. In the 1906 series, the empirical 
mode is on 3 instead of 1 or 2. The mean and standard 
deviations are also higher and the range is wider than in 
the 1905 collections. 
The coefficients of variation show how exceedingly great 
the variability in this material really is. The reader will 
appreciate this by a comparison of the constants for several 
other species of plants, as given in Table V. 
3 How great such differences may be for individual classes may 
be seen by a comparison of the frequencies in series (c) and (d). 
Naturally, the reader will remember that these are actual, not rela- 
tive, frequencies, and that in one case, N = 1245, while in the other, 
N= 1331. 
4The difficulties of collecting really typical samples of material 
(already discussed in an earlier section), while at the same time retain- 
ing data for all the characteristics which it was desired to consider, 
was one of the chief reasons for discontinuing the study of Crinum. 
he? Se Sos tytn 
