216 The Inheritance of Wing Colour in Lepidoptera 



features, the colour-values of each family or group of families have been 

 expressed as a percentage frequency distribution. Among other advan- 

 tages this method tends to remove differences caused by the size of the 

 group dealt with. The following method was adopted when converting 

 one curve into the other. 



The colour scale was divided into intervals of 0*4 units eaph (e.g. 

 from 0-0 to 0*4, and 0*4 to 0*8, etc.). The number of individuals occurring 

 in each colour interval was found, and the result divided by four, so 

 that the percentages might all refer to the frequency in J^ of a colour- 

 unit. These average numbers of insects of any given colour-value were 

 then plotted as ordinates, and the colour- values as abscissae. On account 

 of certain errors mentioned below, an interval of 0*4 units was chosen, 

 since this gave a reasonably smooth curve ; though occasionally small 

 oscillations appear, which probably have no significance. 



The orange values are shown by the heavy line. Though of less 

 importance the yellow values have been treated by the same method and 

 are represented by a fine broken line. Roughly speaking it is seen to 

 run parallel to the orange curve, as might be expected. As in the case 

 of the distribution curves the colours of both parents have been shown 

 in the margin below the arrow which represents their orange values. 



(c) Experimental and statistical errors. 



Whenever a frequency distribution rises to two maxima they corre- 

 spond to those portions of the figures showing the distributions of the 

 colour-values which are flattest, and which consequently contain the 

 greatest number of insects of one colour^ The points at which the fre- 

 quency distributions reach a minimum (about 2*0 — 2 5) correspond to 

 those portions of the curve showing the distribution of the colour- 

 values where the fall is most rapid, and therefore to those points at 

 which segregation appears to occur. 



The question that must then be considered is whether the fall is 

 real or whether it can be more easily attributed to the various sources 

 of error. Errors due to technique, such as labelling, are, I think, insig- 

 nificant. There remain those due to taking the colour measurements. 

 The accuracy with which the readings can be taken varies considerably. 

 A difference may be just discernible between two colours, one measuring 

 1-00 and the other I'Ol units, but no change will be appreciated between 

 two other colours, one measuring 1000 and the other lO'Ol units. To 

 be discernible the increment in the last case should be about 01 



