470 Information Theory and Biology /25 : 3 



receptor unit. (There are about 7 x 10 6 cones and 10 8 rods in the 

 human eye.) 



Each receptor unit is believed able to respond in a characteristic 

 fashion to about 100 just noticeable differences in intensity between the 

 visual threshold and the pain threshold. The average information per 

 receptor unit then is 



H x = log 2 10 2 = 7 bits 



The eye sees a new picture about 10 times a second. Accordingly, the 

 rate of receipt of useful information is 



H[ = 70 bits/sec/receptor unit 

 or for the entire eye 



H' = 1 x 10 8 bits/second 



This can be compared with a television channel which carries about 

 10 7 bits/second. 



The optic nerve which carries this information has about 10 6 fibers. 

 Each carries a maximum of 300 spikes per second. Hence, in each 

 0.003 second, each fiber carries 1 bit of information. The optic nerve, 

 then, has a capacity for transmitting information of 



C = 3 x 10 8 bits/sec 



a number identical, within experimental error, with the receipt of useful 

 information by the eye. 



In the central nervous system, one can make crude estimates of the 

 information received that is associated with acuity and with color 

 vision. These lead to 



^ibrain = 5 x 10 8 bits/second 



The optic nerve is extremely well coded. Its channel capacity is not 

 many orders of magnitude larger than the auditory nerve. However, 

 the rate of information entering the conscious part of the brain is perhaps 

 10 7 times as large. 



The coding of the optic nerve may be compared to a television 

 channel. One of the limitations of television broadcasting is poor 

 encoding of information. Many engineers have realized that a system 

 which indicated changes of intensity only would be far more efficient in 

 transmission of information. (In other words, much narrower channels 

 could be used.) The high efficiency of the optic nerve as indicated by 

 the foregoing estimates suggests that such a system is used. Evidence 

 from electrophysiology and histology supports this view (see Chapter 7). 

 The approach of information theory helps to understand the histology 

 and the electrophysiological data. 



