506 H. S. JEVONS, H. I. JENSEN, T. G. TAYLOR AND C. A. SUS8MILGH. 



The significance of the above analyses in relation to the 

 quantitative distribution of minerals throughout the mass 

 will be considered in a subsequent section, and a discussion 

 of the bearing of the analyses upon the nomenclature of 

 the rock may be postponed to the next section, which is 

 devoted to that subject. It will be sufficient here to make 

 a comparison in regard to the composition of the essexite of 

 Prospect and its varieties with rocks of other well known 

 occurrences. With this object there are exhibited in the 

 usual manner in Tables I and III analyses of the Prospect 

 rocks, and of a number of rocks similar in chemical com- 

 position. 



The great difficulty which the eye experiences in grasping 

 rapidly a sufficient number of figures renders comparisons 

 from analyses presented in this form tedious and uncertain, 

 hence the attempts of several authors to introduce some 

 style of graphic representation of the chemical composition 

 of rocks. We have endeavoured here to facilitate com- 

 parison without cumbrous diagrams by simply using 

 numerals consisting only of the essential digits. The 

 principle on which the method is based is that errors of less 

 than about five per cent, of the stated quantity of the con- 

 stituent are negligible, for variations greater than this 

 between specimens are commonly found from points in one 

 mass less than a metre apart. If this margin of error be 

 adopted, it is unnecessary to use fractions at all in stating 

 quantities making over 10 per cent, of the rock, for the 

 number 10 may be taken as representing all quantities from 

 9*50 to 10*49, and the maximum possible error 0*5, becomes 

 less than 5 per cent, as the quantity increases above 10. 

 Similarly, quantities between 5 and 10 are sufficiently repre- 

 sented by the use of a fraction no smaller than a half. The 

 eye is assisted if a sign less complex than either '5 or \ be 

 used to represent one-half, and we have found a simple 



