210 ANNUAL KEPORT SMITHSONIAN INSTITUTION, l'J12. 



results which can not be attributed to chance. We should notseparate 

 what has been joined together. We could not break their frames, 

 only try to stretch them apart. And that we can not always succeed 

 in doing. The theory of equipartition explains so much that it 

 must contain a part of the truth; on the other hand, it can not all be 

 true for it will not explain all. We can neither abandon it nor 

 keep it without modification. Those modifications which we must 

 make seem so strange that we hesitate to reconcile ourselves to them. 

 At present we can only enumerate them without solving them. 



