O. H. Latter 
373 
Length 
Breadth 
Key 
Latter 
Key 
Latter 
Obs. 
Calc. 
UDS. 
oaic. 
Obs. 
Calc. 
Obs. 
Calc. 
1 
3-32 
3 
1-49 
50 
43-14 
22 
23-27 
1 
1-01 
2 
0-71 
185 
203-66 
123 
133-24 
74 
73-43 
47 
53-44 
380 
344-74 
300 
278-73 
459 
460-36 
379 
369-01 
185 
210-39 
201 
213-77 
302 
300-75 
269 
272-68 
46 
46-06 
61 
59-16 
19 
19-35 
18 
21-02 
8 
3-57 
6 
7-11 
0 
-10 
2 
0-14 
0 
-10 
1 
0-23 
9 
AC = 
16-36 
= 8-09 
= •62 
28-59 
i'<-025 
P = 
= -325 
P>-99 
P< -00004 
The odds are thus more than 40 to 1 against Rey's lengths of eggs fitting a 
normal distribution. But only once in 100 random samples of 85.5 eggs shoidd we 
expect a better result tlian we have got fo)- the breadths, if the distribution be leally 
normal. Turning to the 717 measurements we find the length distribution reason- 
able on tlie basis of a normal curve; once in every three random samples of 717 
eggs we should get a worse result, but the breadths are quite impossible. This 
impossibility arises, however, entirely from the two giant breadths of 18-8 and 19'2. 
They are undoubtedly abnormalities and if they be excluded, the fit is a good 
one, i.e. about 55 samples in 100 would give a worse result. Witli regard to Rey's 
length distribution the sources of improbability are seen on analysis to be (a) the 
crowding up of seven eggs on 25-0 mm., (6) of 131 eggs on 22-0, and (c) of 63 on 
22-5 mm. These contribute more than half the value of ! In other words there 
is little reason to doubt that the variability in length of Cuckoo's eggs would follow 
a normal curve, were it not for Rey's tendency to heap up observations on the -5 
and "0 nrillimetre groups*. It does not therefore seem worth while pursuing the 
distribution of variability further on the present observations. We see that it is 
quite possible that the normal curve would really suffice to describe the frequency. 
On the other hand, if the gentes theory be considered as established, we should 
naturally expect the heterogeneity to show itself in some deviations from any 
smooth curve of frequency. 
In conclusion I must gratefully acknowledge the generous assistance of 
Prof. Karl Pearson in the prepai-ation of this paper and in the statistical exami- 
nation of the measurements obtained : indeed without his aid I should have been 
quite unable to perceive their significance, nor would this paper have been written. 
* The evil of this heaping up may be easily seen if we take the first group 19-0-5 — •20-0.5 instead 
of 18-75 — 19-75 still grouping by mm. The mean will now be found to be 2'2-10 and the standard 
deviation 1-0014 instead of 22-27 and -9450 respectively I 
