94 THE SOCIAL LIFE OF ANIMALS 



and otherwise see that nothing untoward happens to 

 make him or anyone else question the results. 



After some twenty, twenty-five or thirty days of 

 such care, in which Sundays are included, again each 

 fish is photographed to scale, as they were also photo- 

 graphed at the beginning of the experiment; the 

 photographs are measured and the relative growth 

 determined for the fish that have daily been placed 

 into perfectly clean synthetic pond water, as com- 

 pared with those which daily have been put into 

 conditioned water, that is, into the water in which 

 other goldfish have lived for a day. 



During the course of an analysis of this problem 

 we have performed this simple basic experiment 

 many times. The first forty-two such tests involving 

 886 fish gave on the average about two units more 

 growth for the fish in the conditioned, slightly con- 

 taminated water, than for those in clean water (Fig- 

 ure 14). These results have a statistical probability 

 (P) of about one chance in a hundred million of 

 being duplicated by random sampling. Hence we 

 have demonstrated that under the conditions of our 

 experiments the goldfish grow better in water in 

 which other similar goldfish have lived than they do 

 when they are daily transferred to perfectly clean 

 water. 



The problem that has been occupying us for some 



