24 HENRY S. WASHINGTON 



They are calculated to loo, free from water and the very minor con- 

 stituents. The analyses of the comendite and of the Zichidi tra- 

 chyte are not included, as their place in the succession is uncertain. 



la and Ih are the averages respectively of the early pantelleritic 

 trachytes and the aegirite pantellerites, I being the average of these, 

 representing the composition of the first phase. 11a and lib are the 

 averages of the trachytes and the hyalopantellerites, II being the 

 average of these and representing the composition of the second 

 phase. Ill is the average of the final basalts, including that of 1891. 



The first two phases show a marked repetition in the magmatic 

 succession, a beginning of which is apparently repeated in the basal- 

 tic phase. Starting with the pantelleritic trachytes, there is first 

 a rise in silica and a fall in the other constituents (that in K2O 

 being slight) to the pantellerites which formed the last flows of this 

 phase. After the formation of the caldera the magma returns 

 toward or beyond its original composition, as shown in the fall in 

 silica (and potash) and the rise in the others. Then the change 

 shown in the first phase is repeated, silica and potash rising and 

 the others falling in the hyalopantellerites. After the cessation 

 of these flows the basalts show a change in the magma in the 

 same directions (except the alkalies) but to a greater extent as 

 between phases I and II. The iron oxides do not conform to the 

 courses of the other oxides, but there is a steady increase in the ratio 

 of ferrous to ferric oxide, and first a decrease and then an increase 

 in their total amount. The averages of the whole phases (I and II) 

 show the general trend of the magma to a more femic composition. 



In the absence of any accurate data as to the relative volumes of 

 the various types no satisfactory estimate can be made at present 

 of the average Pantellerian magma, but a general consideration of 

 the various flows and cones suggests that probably the average 

 la roughly represents this. If this be so, the order of succession 

 corresponds well with that enunciated by von Richtofen and Iddings, 

 namely: beginning with the mean and ending with an extreme 

 (generally the most femic) after few or more alternations. 



The interesting feature about the present case is that the most 

 abrupt changes in the magma seem to be correlated with maxima 



