156 



EASTER ISLAND. 



Table 33- 



When we compare this with Table 32 we find that, including the ele- 

 ment of quality, the general order of the speech components remains 

 unaltered. This computation is too general to have value save as set- 

 ting conveniently in broad terms a bench-mark of specific comparison. 



Individually treated after this method, reference being made to ascer- 

 tained net stocks, the several languages give the following showing: 



Table 34- 



It must be understood, it is frankly acknowledged, that these figures, 

 while based upon exactly ascertained percentages, are essentially arti- 

 ficial and, although decimal in notation, are in no sense the designation 

 of any real ratio. But, since the four groups have been developed in 

 every part by exactly the same method, the resultant figures, though 

 artificially obtained, are fairly comparable. Their purpose is to weight 

 the ascertainable figures of each of these languages with an element 

 which may express the degree of the extension, through the province 

 and more remotely, of the several classes of word occurrence which our 

 studies have afforded us. Accordingly we bring the sums for each 

 archipelago into a single table, where they may be more readily com- 



