THE ART OF SCIENTIFIC INVESTIGATION 



compared to data obtained in previous years. Evidence obtained 

 in this way is never conclusive, though it may be usefully sugges- 

 tive. "If when the tide is falling you take out water with a 

 twopenny pail, you and the moon can do a great deal." In 

 biological investigations there may be many unsuspected factors 

 that influence populations separated by time or geographically. 

 When general considerations have been satisfied, statistical 

 methods are used to decide on the necessary size of the groups, 

 to select animals according to weight, age, etc. and, while taking 

 these particulars into account, to distribute the animals into groups 

 without sacrificing the principle of random selection. 



No two groups of animals or plants are ever exactly similar, 

 owing to the inherent variabihty of biological material. Even 

 though great pains are taken to ensure that all individuals in 

 both groups are nearly the same in regard to sex, age, weight, 

 breed, etc., there will always be variation that depends on factors 

 not yet understood. It is essential to realise the impossibility of 

 obtaining exactly similar groups. The difficulty must be met by 

 estimating the variability and taking it into account when assess- 

 ing the results. Within reasonable limits it is desirable to choose 

 the animals for an experiment showing little variability one with 

 another, but it is not essential to go to great lengths to achieve 

 this. Its purpose is to increase the sensitivity of the experiment, 

 but this can be done in other ways, such as by increasing the 

 numbers in the groups. There are mathematical techniques for 

 making corrections in certain cases for diflferences between 

 individuals or groups. 



Another method of meeting the difficulty of variability in 

 experimental animals is by " pairing " : the animals are arrayed 

 in pairs closely resembling each other ( perhaps pairs of twins or 

 litter mates). Each animal is compared only with its fellow and 

 thus a series of experimental results is obtained. By using identical 

 twins one can often effect great economy in numbers, which is 

 important in investigations on animals that are expensive to buy 

 and keep. Experiments carried out in New Zealand on butterfat 

 yield showed that as much information was obtained per pair 

 of identical twin cows as from two groups each of 55 cows. In 

 experiments with growth rates, identical twins were about 25 

 times more useful than ordinary calves.* 



20 



