READING RATES 507 



DISCUSSION' AND SUPPLEMENTARY EXPERIMENTS 



Conclusions from Principal Experiments 



The conclusions which can be reached with reasonable assurance from 

 these experiments are rather narrow. They might be stated: 



1. Information is best transmitted through a human channel by 

 means of well-chosen acts (reading well chosen words in this case) in- 

 ^•olving many bits per act, that is, much choice per act. Cutting down 

 drastically the bits per act does not substantially increase the speed at 

 which the individual act is accomplished. 



2. The lower bound of information transmission through the human 

 channel of rapid readers seems to be about 43 bits/sec. This estimate is 

 a little higher than that found bj'- Licklider,- and may be close to a 

 limiting rate. 



3. This limiting rate can be achieved by the simple act of reading 

 either randomized lists from suitably selected words or scrambled prose. 



-i. Both familiarity and length of words are important in determining 

 reading speed. The relative effect of these two variables on reading speed 

 is rather complex. 



Be\'ond these narrow conclusions, there is much understanding yet to 

 be achieved in the general field of the speed of human mental and 

 physical responses and operations. Thus, it seems worth while to men- 

 tion other experiments which were done in the course of the present 

 investigation and experiments carried out by other workers, and to 

 speculate somewhat concerning the whole of this experimental work. 



Multiple Tasks 



The reading- while-tracking experiments touch on an important prob- 

 lem. We have all heard of wireless operators who can receive and sub- 

 sequently type out a message while carrying on a conversation or playing 

 chess. There is nothing in this feat to indicate an information rate greater 

 than that we have found. Actually, the rate of receiving prose by Inter- 

 national Morse Code by ear is around 0.58 word/sec;^ this is slow com- 

 pared with the rates we have considered. 



Our experiments with tracking followed experiments in which words 

 in the lists were randomly printed in red or black, and in which the 

 subject spoke red words in a louder tone of voice than black words, or 

 pressed one key for red words and another for black words. In these 

 cases, the added information, one bit per word, was so small as to make 

 no clearly discernible difference in information rate for the large vo- 



