the Perception of Colour. 43 
Table I. gives the actual observations, with their dates. 
Table II. gives the result of summing together each group 
of six equations. 
Each equation in Table II. has the sums of its positive and 
negative coefficients each equal to 600. 
Having obtained a number of observations of each combina- 
tion of colours, we have next to test the consistency of these 
results, since theoretically two equations are sufficient to deter- 
mine all the relations among six colours. We must therefore, 
in the first place, determine the comparative accuracy of the 
different sets of observations. ‘Table III. gives the averages of 
the errors of each of the six groups of observations. It appears 
that the combination IV. is the least accurately observed, and 
that VI. is the best. 
Table IV. gives the averages of the errors in the observation 
of each colour in the whole series of experiments. This Table 
was computed in order to detect any tendency to colour-blind- 
ness in my own eyes, which might be less accurate in discrimi- 
nating red and green, than in detecting variations of other co- 
lours. It appears, however, that my observations of red and 
green were more accurate than those of blue or yellow. White 
is the most easily observed, from the brilliancy of the colour, 
and black is liable to the greatest mistakes. I would recommend 
this method of examining a series of experiments as a means of 
detecting partial colour-blindness, by the different accuracy in 
observing different colours. The next operation is to combine 
all the equations according to their values. ach was first mul- 
tiplied by a coefficient proportional to its accuracy, and to the 
coefficient of white in that equation. The result of adding all 
the equations so found is given in equation (W). 
Equation (Y) is the result of similar operations with reference 
to the yellow on each equation. 
We have now two equations from which to deduce six new 
equations, by eliminating each of the six colours in succession. 
We must first combine the equations, so as to get rid of one of 
the colours, and then we must divide by the sum of the positive 
or negative coefficients, so as to reduce the equations to the 
form of the observed equations. The results of these operations 
are given in Table V., along with the means of each group of six 
observations. It will be seen that the differences between the 
results of calculation from two equations and the six independent 
observed equations are very small. The errorsin red and green 
are here again somewhat less than in blue and yellow, so that 
there is certainly no tendency to mistake red and green more 
than other colours. The average difference between the ob- 
served mean value of a colour and the calculated value is ‘77 
of a degree. The average error of an observation in any group 
