in Mendelian Ratios. 431 
tables in which the expectation is 25 per cent. the agreement is 
"good, very good considering the small number of cobs available. 
When these three tables are pooled together (without respect to 
the fact that they deal with different characters) the actual s.d. is 
2:05 per cent. against an expected value 1-99. 
; 
Taste E. (Data from Lock, Annals of the Botanic Gardens, 
Peradeniya, tt. 1906.) Comparison of actual standard 
deviations and s.d.’s of sampling im experiments with mazze. 
1 
i 
i: ae _~ arcs laid BS aa En 
; No Mean per- Standard 
‘Table in original Characters of | centage of | deviation and N pq/H 
lants| recessive probable error | (per cent.) 
P "| grains (per cent.) 
| I Starchy and sugary) 18 24°8 2044-23. 2°20 
Ii ieecaa %9 22 23-4 1-96 + -20 2:23 
XXXIV | Yellow and white 30 25°5 1-62-14 1°63 
on i 
[1, W, Xxxrv ue 70.| 247 205 4-12 1-99 
IL, IV, V, VI, VII |Starchy and sugary} 74 | 50-9 3174-18 2-60 
IH, IX, X, XI, XII| Yellow and white | 49 50-2 3°16 + -22 Oe 
ae XXXII a a 95 50-0 2-484 -12 2-16 
y 
| 
In the case of the DA x RR crosses, however, the agreement is 
_by no means so good, the actual exceeding the expected standard 
deviation by two to three times the probable error in each case. 
The divergence in the second group is largely due to the five 
plants of Table IX which give percentages of whites ranging from 
43°3 to 56°6 and 56:9 on 644, 648 and 515 grains respectively : 
with standard errors of 2 to 2:2 per cent., such deviations are not 
probable as the results of random sampling. But, as pointed 
out by Mr Lock, the data as a whole shew a rather greater 
-yariation than one would expect, and the result cannot be 
attributed merely to one or two particular tables. The difficulty 
‘is to imagine any cause for such excessive variation (such 
a fertilisation by pollen from DD’s or Df’s instead of RR’s, in- 
| 
correct sorting of doubtful yellows or unsuspected heterogeneity 
of the whites) which would not also affect the means obtained, 
and the means agree very well with expectation. The contrast 
between the DR’s selfed and the DR x RR crosses is curious. 
The former certainly do not suggest any significant fluctuation : 
the latter do, on the whole, suggest some source of disturbance 
or possibly error. I wish more data, for other plants or for 
