January 3, 1902.] 



SCIENCE. 



the constituent elements of organisms. The 

 study of physiological and pathological 

 variations that elucidate correlations will, 

 therefore, be a most powerful factor in the 

 discussion of the problem of variability. 



This point of view coincides closely with 

 that of Rudolf Virchow who has always 

 emphasized that the clue for the problems 

 of variability must be looked for in the 

 study of cellular variation. 



Under certain simplified assumptions the 

 problem, as here defined, may be made 

 amenable to statistical treatment. If all 

 the contributory causes of variation are 

 given equal weight, and if it is assumed 

 that the index of correlation for all the 

 groups is the same, then it is only necessary 

 to determine four unknown quantities : the 

 total number of causes (or correlated 

 groups) ; the probable number of causes 

 (or correlated groups) ; the amount of in- 

 fluence of each cause (or correlated group) 

 upon the whole organism; the variability 

 of the effect of each cause (or correlated 

 group) upon the whole organism. These few 

 data can be calculated without any great 

 difficulty from the averages of the first four 

 powers of the individual variations. This 

 method may be serviceable as long as the 

 actual correlations are unknown. 



Still another conclusion may be drawn 

 from our considerations. We have seen 

 that the variability of an organism depends 

 upon the correlations of its elements, and 

 that the variability must be the greater the 

 closer these correlations and the less the num- 

 ber of correlated groups. At the same time, 

 the variations will be the more likely to 

 have skew distributions, the less the num- 

 ber of correlated groups. A disturbance in 

 one element of an organism thus consti- 

 tuted must, therefore, result in a consider- 

 able variation of the whole. That is to 

 say, that in the case of skew distributions of 

 variations we may expect sudden transfor- 

 mations of type due to small causes. In 



organisms in which the variability is sym- 

 metrical, we may generally expect the 

 whole form to be controlled by many in- 

 dependent causes, or by many independent 

 groups of correlated elements. In this case, 

 the changes of form due to small changes 

 of conditions wiR probably be less marked. 



The principal advantage of the method 

 of considering variation that has been here 

 suggested is that the occurrence of varia- 

 tions in fixed lines, that are so difficult to 

 understand, may be considered as a result 

 of chance variation of small elements and 

 of physiological correlation. Both of these 

 are much more readily understood than a 

 variation of form that does not show any 

 immediate relation to the causes producing 

 the variation. 



It has often been asserted, or assumed, 

 that skew distribution of variations is a 

 proof of the effect of selection, or of some 

 other kind of instability of type. Our con- 

 siderations have shown that this is not 

 necessarily the case. Skew distributions 

 may be found in stable forms. On the 

 whole, it does not seem possible to discover 

 by purely statistical methods the causes 

 of skewness. The numerical material ob- 

 tained by measurements can be made to fit 

 satisfactorily many theories that would ac- 

 count for the skewness. It is necessary to 

 base such theories on biological investiga- 

 tions and to subordinate our statistical 

 methods to the biological point of view. 

 Otherwise the result of statistical inquiry 

 will be of little use and may even become 

 quite misleading. 



Franz Boas. 



ON THE TRUE NATURE OF TAMIOSOMA.* 



In 1856 the late T. A. Conrad described a 



remarkable fossil from California, imder 



the name of Tamiosoma gr eg aria, composed 



of large tubes with a longitudinal cellular 



'Communicated by permission of "the Director 

 of tlie U. S. Geological Survey. 



