COSMOLOGY AND TECHNOLOGY 111 



there can no more be a perfectly "pure" science than, as Picasso re- 

 marks, there can be a perfectly abstract art: 



There is no abstract art. You must always start with something. After- 

 wards you can remove all traces of reality. There's no danger then 

 anyway, because the idea of the object will have left an indelible 

 mark. It is what started the artist off, excited his ideas, and stirred up 

 his emotions. 



Constraint. Perhaps the least unreliable forecast of the scientific 

 importance of any gi\'en inquiry turns less on the nature of the inves- 

 tigafioM than on the talent of the investigafor; he who has already 

 discovered much is not unlikely to discover more. But forced to wear 

 blinders, even the gifted scientist sees less. Priestley's comment on 

 his own career applies also to many other distinguished scientific 

 careers : 



In looking for one thing I have generally found another, and some- 

 times a thing of much more value than that which I was in quest of. 



Other things being equal, more will be discovered when the inves- 

 tigator retains power to pursue promising tangential leads opening 

 up during his investigation. A Rayleigh, prosecuting a self-assigned 

 investigation, is free to follow his research wherever it leads him. 

 Turning aside from his initial study, he traces a curious anomaly 

 back to its root, and so discovers a new element, argon. But a Hille- 

 brand, working toward a specified goal in a government laboratory, 

 could not stop to examine a similar anomaly which pointed to a 

 similar discovery: the terrestrial occurrence of the element helium. 

 In a retrospective reflection, Hillebrand wrote that 



. . . the chemical investigation had consumed a vast amount of time, 

 and I felt strong scruples about taking more from regular routine 

 work. 



For this and other reasons he failed to look into 



the suggestion made by one of us in a doubtfully serious spirit, that a 

 new element might be in question. 



The successes won by Rayleighs and lost by Hillebrands suggest 

 to Conant that the presence or absence of constraint is a distinction 

 far more important than that now doubtfully drawn betv/een pure 

 and applied science. One might attempt to distinguish as applied 



