The Status of Information Theory in Biology: A Round Table Discussion 401 



several distinct levels of organization. Some levels are more sharply delined 

 than others, and the hierarchy of levels is not always unambiguous. Still, 

 they are pervading enough that the intelligibility and validity of any statement 

 in biology depends on proper agreement with organizational hierarchy. Now, 

 one of the outstanding features in biological organization is that quite obviously 

 only a small amount of the features obtaining at a given level has observable 

 effects on the next higher level. Hence one of the most urgent problems, on 

 any level, is that of detennining what details are involved in the communication 

 to the next higher level — but this is precisely one form of the problem of the 

 'whole and its parts'. Thus, in studying the whole rather than its parts we 

 seem to act as organelles, cells, organs do. 



So we have good reasons to beHeve in the importance of the systems approach. 

 Still, it remains no more than a belief — and there exists an equally strong belief 

 that only intense preservation of details will yield major biological breakthroughs 

 and that it would be a 'young miracle' if really important contributions would 

 come to biology without intensive examination of details. So we have extreme 

 misgivings either way — and those misgivings seem to be destined to be with 

 us forever. There exists no rigid calculus telhng which formahsm must be 

 used on what data to achieve a major discovery. 



The present conference was arranged to explore the applications of infor- 

 mation theory to the study of living things. This is a new field, and one cannot 

 say, at this time, which approach is going to be most successful. Accordingly, 

 the scope of the program was extremely wide. It was natural to question how 

 much the various papers had contributed to furthering the purpose of the 

 meeting. There was general agreement that some papers had contributed 

 very much, some a moderate amount, some little or nothing; there was however, 

 notable disagreement about which papers belong in which category. 



There exist a few cases where information theory was used in dealing 

 with problems which could have been solved in other ways; and there are 

 very many cases where problems have been solved by various methods which 

 could have been, possibly, solved more easily by using information theory. 

 The coin problem that Rapoport talked about falls in this category; so do 

 the cryptographic studies of Gamov/ and YcAS. Information theory is so 

 general that its domain of applicability is very broad; one cannot name 

 one situation in which by the use of information theory one cannot get 

 some understanding on what is happening on an abstract basis. But one 

 is always beset by the niggling doubt that the application may not be proper. 

 One can in many situations obtain results which seem to clarify understanding 

 or increase the sharpness of a description to an extent which was not possible 

 prior to the use of information theoretic methods. On the other hand, such 

 results seem often suspended in mid air, away from the results of conventional 

 disciplines. The important question then, at this nascent stage of affairs, is some- 

 thing which is repellent to the scientific mind, the assessment of the 'worthwhile- 

 ness' of the answers information theory seems to give. It seems plausible to 

 assume that information theory should be useful where communication is 

 critical, where messages are to be transmitted in the presence of noise, and where 

 one might assume that some optimization is approximated; biologists are 

 inclined to invoke the Darwinian mechanism of random trials with perpetuation 



