512 THE BELL SYSTEM TECHNICAL JOURNAL, MARCH 1957 



bits/second for telephone and 50 million bits/second for TV). This 

 would not be surprising if the limitation we observe had been one of the 

 speeds at which words can be uttered, but it appears rather to be a 

 mental one, one of recognizing what is before the eyes. To the authors, 

 it seems reasonable that this mental limitation may apply to a human 

 being's ability to absorb information, that is, to the information rate 

 needed to present a satisfactorj- sensory input to a human being. If it 

 does, then why do we need so much channel capacity to convey to him 

 an acceptable sound or picture? 



This can be explained in part bj^ the inefficienc}^ of our present com- 

 munication methods. Despite its present imperfections, the vocoder 

 makes it clear that clearly understandable speech can be transmitted 

 using far less channel capacity than that required in ordinary telephony.^ 



However, it is quite likely that even with the most efficient of en- 

 coding means we will have to use far more than 43 bits/second for a 

 picture transmission channel. While only a portion of the image of the 

 transmitted picture falls on the fovea at any instant, we can cast our 

 eyes on any portion of the received picture. If the pick-up camera de\dce 

 and the received picture followed eye movements, a much less detailed 

 picture would serve. Even with our eyes fixed, we can concentrate our 

 attention on a particular part of our field of vision, and this is something 

 that the pick-up camera cannot track. There may be similar effects in 

 our apprehension of sounds. 



In the light of present knowledge it is impossible to estimate the 

 minimum channel capacity required to transmit sound and pictures in a 

 satisfactory manner. It will take work far beyond the measurement of 

 reading rates to enable us to make such an estimate. 



ACKNOWLEDGMENTS 



The writers are indebted to D. L. Letham, w^ho carried out some work 

 preliminary to that reported here, to J. L. Kelly for helpful comments, 

 to D. Slepian for help in putting the appendices in a mathematically 

 more acceptable form, to Miss Renee Hipkins who carried out most of 

 the experiments, and to three indefatigable readers, ^Irs. Mary Lutz, 

 S. E. Michaels and A. P. Winnicky. 



Appendix I 



ON OBTAINING GOOD VOCABULARIES 



The experiments in the body of the paper indicate that the choice of a 

 good vocabulary is important in attaining a high information rate in 

 reading lists of words. 



