THE DIAGRAMS 625 



of numbers 15 to the series shown in figure 2. Several other districts 

 are described petrographically as showing similar series, but quantitative 

 data are not available. 16 See the bibliography below. 



The "Trend" of Differentiation 



Several remarks on the curves plotted have almost the nature of im- 

 mediate conclusions. 



1. Out of about 40 connected series all but two or three show a linear 

 directional trend in the diaram, with relatively little scattering. 17 This 

 seems to be too strong a proportion to be accidental or the result of the 

 incompleteness of data. 



2. The scattering in a series with linear trend seems to occur at the 

 ends of the series rather than in the center. That is, at the basic sub- 

 alkalic end diverse monomineralic rocks separate from a single magma; 

 at the acid alkalic end there seems to be another sort of divergence, with 

 a good deal of variety in the silica and alkali content of the products of 

 one magma. It can not be assumed that during differentiation the liquid 

 magma passed through each stage represented by a rock specimen; but 

 just as the monomineralic basic rocks are precipitates from gabbro 

 magma, so the various acid and alkalic types at the other end probably 

 diverge from some point closer to the average or original magma. 



3. Magmas do not have to be of any special or particular composition, 

 to begin with, in order to develop into a differentiated series of rocks. 

 Magmas of the most various sorts have been strongly differentiated. 

 Neither monzonites nor pulaskite nor any common family is to be classed 

 as anchi-eutectic. 



4. The 'trend is in very different directions in different masses, even 

 if the different magmas were of similar compositions. Iddings men- 



15 These districts include the Kola peninsula ; the Adirondacks of New York ; a series 

 of Hawaiian flows ; the absarokite-shoshonite-banakite series from Yellowstone ; the 

 Roman "co-magmatic" region ; a group of dikes in Virginia ; Port Coldwell, Ontario ; the 

 Highwood Mountains as a whole ; Kirunavaara ; no doubt other series will be found to 

 be omitted. The Christiania rocks shown in the diagram by Iddings, as cited, range 

 widely, but a series from grorudite to tinguaite can be selected to show a linear arrange- 

 ment closely analogous to the granite series from Wisconsin. 



16 The flows from many volcanoes show partial series between gabbro and rhyolite, and 

 a few show more alkalic types. The rocks of Tasmania range from pyroxenite to 

 nephelite syenite with no granite. Series of rocks have been described from Bozeman, 

 Montana, from the Black Forest, from East Griqualand, etcetera, but these can hardly 

 be plotted. 



17 By "linear" it is not the intention to suggest straight lines, but a lack of scattering. 

 Harker has used the term for straight lines ; see Journal of Geology, vol. 8, p. 391, 



