Information and Inactivation of Biological Material 281 



3. E. C. Pollard, W. R. Guild, F. Hutchinson, and R. B. Setlow: The direct action of 

 ionizing radiation on enzymes and antigens. In: Progress in Biophysics, 5, 72-108, 

 Pergamon Press, London and New York (1956). 



4. C. B. Anfinsen: The limited digestion of ribonuclease with pepsin. J. Biol. Cliem. 11\, 

 405^12 (1956). 



5. W. GoRDY, W. B. Ard, and H. Shields: Microwave spectroscopy of biological substances. 

 Proc. Nat. Acad. Sci., Wash. 41, 983-1003 (1955). 



6. R. B. Setlow: A relation between cystine content and ultraviolet sensitivity of proteins. 

 Biochim. Biophys. Acta 16, 444-445 (1955). 



7. B. E. Proctor, S. A. Goldblith, E. M. Oberle, and W. C. Miller, Jr.: Radiosensitivity 

 of 5ac/////5.sM6r///.y under different environmental conditions. Radiat. Res. 3, 295-303 (1955). 



8. G. Stent and C. A. Fuerst: Inactivation of bacteriophage by decay of incorporated 

 radioactive phosphorus. /. Gen. Physiol. 38, 444-457 (1955). 



DISCUSSION 



Quastler: Dr. Morowitz's analysis of the informational aspects of radiation effects, 

 and his concept of information density, are very important developments. As a matter of fact, 

 I believe them to be so important that even small differences in interpretation are worth 

 mentioning, and this is the reason for making the following comments. 



To rephrase the situation: consider a structure (message) consisting of a distinct sub- 

 structures (words) of b elements (letters) each. Let H' be the information content per letter 

 and T' the information measure of constraints between letters. Then: 



H" = information content per word = b{H' — T') 



H" = information content of message = a(H" — T") 



(where T" represents the informational aspects of constraints between words) and 



H^jab = information density in bits per letter. 



If the 'letters' are atoms in living matter, then I suspect the constraints T' and T" to be quite 

 considerable and to reduce H"'lab to rather less than three bits per atom. 



We introduce noise of such a character that a single noise event results in the functional 

 destruction of a single letter, and examine the functional value of the message after it has 

 suffered a known number of noise events. If a single event destroys the functional value, then 

 all letters are functionally essential and the functional information density is H"'lab. If the 

 number of noise events needed averages more than one, then the informational density must be 

 less than maximum, and this can occur in two entirely different situations. 



(I) A number a^ of the letters in each word (or a number b^ of the words) is either irrelevant 

 or can be reconstructed, provided every one of the a — a^ essential letters (or b — h^ essential 

 words) is intact. Then the functional information density is 



H" a -ao H" b - bo 



ab a ab b 



and a single event can cause loss of function but does so only with probability (1 — aja) or 

 (1 — b^jb), respectively. This is the situation where the target size is less than the total size of 

 the structure. 



(II) Up to «r letters in every word (or up to br words) can be destroyed without loss of 

 function — and these letters or words do not belong to a predetermined sub-set but can be any 

 letters (or words). This is the case with error-correcting codes; in this case no single event can 

 cause loss of function, but Or + 1 events will every time. The functional information density 

 is again less than maximum, being 



H" a- ar H" b - br 



ab a ab b 



