1895.] into the Measurable Characteristics of Plants, fyc. 367 



taken as the quartile deviation through 1 mm. of growth. The results 

 are shown in Table III. 



Table III. Quartile Deviation of Frontal Breadths (Q) for various 

 Magnitudes of Carapace-length (C). 



The values here given are probably not very reliable, but they 

 show that in the youngest individuals the quartile deviation is dis- 

 tinctly less than at maturity ; that it increases with increase of size, 

 until a time arrives when it is distinctly greater than in adult life ; 

 and that finally it diminishes again. 



The initial features of this result, the smallness of the quartile 

 error at a young age, indicating relative infrequency of deviations, 

 and the increase during growth, have been observed by Bowditch in 

 the case of human stature. The result obtained by Dr. Bowditch 

 and that here described are both simply confirmations of Darwin's 

 statement, that many variations appear at a late period of develop- 

 ment. 



The initial increase in the quartile error may be attributed to the 

 fact that average young produce upon the whole average adults, 

 while animals which exhibit a deviation of known amount in the 

 young state, exhibit on the whole a greater deviation with advancing 

 age. If this view be the true one (and it is hoped that next year it 

 may be possible to test it by observation of living crabs, which can 

 be measured at various periods of growth), then, in a Plymouth crab, 

 which is of unit deviation when its carapace is 7 mm. long, the most 

 probable deviation when it has grown to be 12'5 mm. in length will 

 be 1O79/9-42 = about 1*15 units. The probable error of this expec- 

 tation is the expression of irregularities in the rate of growth, which 

 cannot at present, for want of knowledge, be adequately discussed. 



From the age represented by a length of 12'5 mm., the quartile 



* Of the four very abnormal values shown in the table, three occurred in this 

 group. They have been omitted in the determination of the quartile deviation, 

 which would otherwise become 9'92. 



