338 CREATR^ SCIENCE 



lay in the discovery of nuclear fission consequent to the misidenti- 

 fication of fission products as transuranic elements. Even when the 

 discrepancy is large enough to be noticed, it is most likely to be at 

 once dismissed from thought as experimental error. Polanyi empha- 

 sizes that, since the vast majority of discrepancies do not signify new 

 phenomena, such dismissal is generally the correct policy. 



In my laboratory I find the laws of nature formally contradicted at 

 every hour, but I explain this away by the assumption of experimental 

 error. I know that this may cause me one day to explain away a funda- 

 mentally new phenomenon and to miss a great discovery. Such things 

 have often happened in the history of science. Yet I shall continue to 

 explain away my odd results, for if every anomaly observed in my lab- 

 oratory were taken at its face value, research would instantly degen- 

 erate into a wild-goose chase after imaginary fundamental novelties. 



Conant suggests that a recognizable discrepancy is most likely to 

 be accepted as a significant discrepancy only when it far exceeds the 

 investigator's own estimate of probable experimental error. Take the 

 case of Rayleigh, who noticed and reported a discrepancy of 0.1% in 

 his measurements of the density of nitrogen. By the standard of or- 

 dinary gas-density measurements this was a trivial discrepancy, well 

 within the zone of experimental uncertainty, and it was dismissed as 

 such by those who read Rayleigh's report. But to Rayleigh, who had 

 confidence in his work, this discrepancy far exceeded any reasonable 

 estimate of experimental error, and he pursued the matter to discover 

 a genuine phenomenon— the occurrence in the atmosphere of the 

 new element argon. 



In some cases the effect observed is prodigious both absolutely and 

 relatively. But, as is plain in the case of Rayleigh's work, all that mat- 

 ters is the magnitude of the anomalous effect relative to the setting 

 of the TI. This setting varies with the state of the science concerned. 

 It is apt to be very high in areas already exhaustively explored: in the 

 area of the physics of simple mechanical assemblies the TI is today 

 almost indefinitely high. In certain recently opened areas in biochem- 

 istry it is very low indeed. But, even in a given scientific area at a 

 given time, there will be enormous variation in the TI from one scien- 

 tist to another. All workers will recognize the need for discrimination 

 in setting the TI neither too high nor too low. But sound discrimina- 

 tion in this sense can be seen in retrospect as one of the most distinc- 

 ti\'e marks of individual genius. Polanyi remarks an apt example: 



