1890.] 



Lord Rayleiyh's Colour Box. 



145 



Table III continued. 



Column T. IT. IIT. 



-3 4 6-0 



-2 7 6-0 



-1 7 6-3 



5 4-7 



-1-1 2 3-0 



+ 2 2 2-7 



+ 3 4 2-3 



+ 4 I 17 



The results of this table, in which the few observers showing large 

 differences are not included, is plotted in fig. 1, in which the numbers 

 in column I of the above table are taken as ordinates, and the 

 numbers in column 111 as abscissae. 



Fiff. I. 



We see at once that the greatest number of observers read between 

 I and 4. The curve falls rapidly on either side, but there are 

 more observers within the limits of the table apparently showing 

 large negative than large positive differences. Thus, for instance, 

 taking 3 as the mean value of all observers, there was nobody 

 amongst sixty-seven observers differing by eight-tenths or more of a 

 division at the positive, while there were as many as ten differing by 

 the same amount on the negative side. If we were from the given 

 curve to calculate the probability of such large differences, as shown 

 by the five observers, 5, 17, 28, 50, and 72, we should get exceedingly 

 small numbers ; this confirms Lord Rayleigh's statement, that diffe- 

 rences from normal vision do not seem to follow the law of errors. 

 For differences less than one division of the scale the curve is not 

 unlike the curve of errors, but not for the larger differences ; thus 

 half the total number of observers read within 3'5 units of the 

 average. If the difference from normal sight was to follow the law of 



