12 



THE AMERICAN NATURALIST [Vol. XL1V 



Thus far, we have treated these groups as of equal 

 value in our computations. From our table it will be 

 seen, however, that the groups differ greatly in respect 

 to the number of individuals comprised, and in respect 

 to the magnitude of the differences shown. I have com- 

 puted the probable errors of the averages for those 

 seven groups which are large enough to make this worth 

 while. Taking into account the three characters (tail, 

 foot and ear) for the seven groups, we have, accordingly, 

 twenty-one probable errors for each of the contrasted 

 sets ( warm and ''cold")- A little figuring sll0WS 

 that in twelve of these twenty-one cases the difference 

 between two contrasted averages is two or more times 

 as great as the probable error of that difference; in one 

 case the difference is over three times its probable 

 error, and in three cases it is over four times its prob- 

 able error. 10 The significance of these facts will be 

 appreciated by any one familiar with statistical methods. 



Diagrams (Fig. 1 and 2) have been constructed per- 

 mitting of a comparison between the two contrasted sets 

 of mice, with respect to the mean length of tail, foot and 

 ear, for each of the size groups. These curves explain 

 themselves, and further comment seems unnecessary. 



The question naturally arises: How do these differ- 

 ences between the warm-room and the cold-room descend- 

 ants compare in amount with the differences which 

 were shown by the parents as a direct result of the ex- 

 ternal conditions! Unfortunately, the data necessary 

 for a direct reply to this question are not at hand, since, 

 in the case of the parents of this particular lot, foot 

 and ear length were not determined at the age of six 

 weeks I have at hand, however, a set of measurements 

 for a considerable number of mice (80 "cold» + 



