an Mendelian Ratios. 429 
4 
) 
referred to above, it is evidently no longer possible to give the full 
‘comparison with the binomial series as the range in number of 
‘seeds per plant is too great and the plants too few, and a com- 
‘parison of the actual standard deviation of the percentages for 
individual plants with the standard deviation (pq/H)? is all that 
‘can be effected. Table C shews the results. The first four lines 
‘Taste C. (Data from Second Report of the Evolution Committee 
| of the Roy. Soc. (W. Bateson and Miss Killby).) Comparison 
of actual and calculated standard deviations of percentages of 
recessive seeds in F, on different plants in peas. Hapectation 
Bot). 
| Nahe Mean Standard weve 
Table in Characters : of percentage | deviation and | pq/H 
| original dominant first ents of recessive | probable error | (per cent.) 
De seeds (per cent.) 
/ I Yellow and green 84 25°4 6°55 + °34 731 
| aa i . 86 24-3 4:86 = +25 5°30 
II Round and wrinkled 64 23°8 6:05 + -36 6:48 
ii, LOL Py mn 86 24°3 5°59 + *29 5°30 
| I, WI | Yellow and green 170 24°8 5-79 4-21 6-39 
‘| It, 111 | Round and wrinkled | 150 24:1 5°80 + °23 0°84 
Omitting from Tables I and II plants with less than 50 seeds, and from III 
plants with less than 100. 
| I Yellow and green 52 24-9 4-40 + -29 4-40 
| Sag # x 43 24-3 4-23 4-31 3-71 
| 1 Round and wrinkled 42 25°1 4-944. °31 4°38 
| Til a He 43 23°8 3°57 +26 3°71 
of the table give the standard deviations of the percentages of 
yellows for the plants included in Tables I and III of the original, 
and of the percentages of rounds for the plants tabulated in 
Tables II and III. In three cases out of the four the observed 
‘standard deviation is less than that calculated, though the 
differences are within the possible limits of fluctuations of sampling 
(taking the probable error of a standard deviation as given ap- 
proximately by the formula for the case of normality 0:67450//2n). 
Grouping together the data respecting yellows and greens from 
‘Tables I and III, and similarly the data respecting rounds and 
wrinkleds from Tables II and III, the results shewn in the next 
two lines are obtained: the observed standard deviation is less 
