t^- I 



54 THE SOCIAL LIFE OF ANIMALS 



TABLE I 



Survival in minutes of goldfish in colloidal silver 



NUMBER NUMBER DIFFERENCE STATISTICAL 

 GROUPED ISOLATED PROBABILITY 



7X lo 70x1 

 182 min. 507 min. 325 min. P < o.ooi 



Any biological experiment has a large number of 

 so-called variables, that is, of factors that it is diffi- 

 cult or impossible to bring under such complete 

 control that we can be certain that the experiment 

 will be exactly repeatable next time. Hence it is 

 customary to make experiments if possible as paired 

 experiments, in which one set of conditions (those 

 of the group in this instance) will differ from an- 

 other lot (those of the isolated goldfish) only by the 

 one difference, in this case of grouping and isolation. 

 Such results with these fish can then be analyzed 

 by statistical methods to find the probability of get- 

 ting like results merely "by chance." These methods 

 are now so simple that even I can make the calcula- 

 tions. They are as accepted a technique as is the 

 paired experiment. 



With the goldfish there is less than one chance in 

 a thousand of getting as great an average difference 

 with the same number of trials. Technically we say 

 that probability, or P, for short, is less than 0.0001. 



