456 ON INCREASE IN SIZE [pt. m 



actual fact by Boas & Wissler, by Boas, and by Bowditch, who 

 studied exhaustively the growth of Toronto school-children. The 

 variability coefficient in that case followed exactly the curve of 

 yearly increment and reached an exactly simultaneous maximum at 

 the age of 15 years. The correlation coefficient behaved in the same 

 way. Boas & Wissler explained their results by saying that correla- 

 tions between measurements in one individual ought naturally to be 

 greater during periods of rapid growth than at other times, because 

 the variations in responsible factors will affect them all to an equal 

 extent. Variability is what governs correlation, so that would also 

 be expected to rise and fall in the same manner. But there is a 

 proviso that must be made here, for Boas & Wissler's use of the 

 term "growth-rate" is not the same as that of Brody, for instance. 

 Boas & Wissler mean by the time of greatest growth-rate the time 

 at which the largest increments are being added on to the organism 

 in unit time, i.e. the half-way point in the curve of the autocatalytic 

 equation. Brody means by the time of greatest growth-rate the time 

 at which the organism is adding on to itself the largest relative incre- 

 ments, i.e. the earliest stages of embryonic life, when the slope of 

 the log. weight/age curve is extremely steep, and the embryo doubles 

 its weight in an exceedingly small lapse of time. The variability 

 coefficient shows, therefore, that it is at any rate true to say that, 

 during the phase when the largest absolute increments are being 

 made, the widest variations from the mean tend to occur. This seems 

 very reasonable, but there is also evidence which shows that, when 

 the period of most rapid growth in Brody's sense is occurring, the 

 variability coefficient is also large. 



Other examples are numerous. Buchem found a coefficient of 

 variability of 0-4 for the early stages of the embryo cow and o-i later. 

 Edwards found a variability coefficient of 0-1347 for unincubated 

 chick blastoderms, but of 0-1087 for those incubated 24 hours. 

 Jenkinson gives a graph exactly analogous to those of Boas & Wissler 

 computed from Roberts' measurements of English artisans. Jenkinson 

 also worked on the trout embryo (or rather the alevin, for his first 

 point was 2-3 weeks after hatching, by which time the yolk was not 

 completely absorbed). He found that there was a close general 

 agreement between weekly increment and variability coefficient for 

 the first 10 weeks after hatching, true not only for the growth and 

 variability of the body-length but also for some of the parts such as 



