DALY. — COMPOSITIONS OF IGNEOUS-ROCK TYPES. 213 



analysis which he has published in his treatise. These averages would 

 be chemical "center-points " in his system of classification as actually 

 applied to the terranes of the world. 



So far as the writer is aware, the preparation of these averages has 

 not hitherto been attempted to such an extent as to cover the chief 

 families and species of igneous rocks. An approximation to the 

 desired results is offered in the following tables. 



The work of computing the averages has been lessened very greatly 

 by the publication of Osann's " Beitrage zur chemischen Petrographie " 

 (2nd part, Stuttgart, 1905). This remarkable book contains, in con- 

 venient arrangement, the statement of most of the eruptive-rock 

 analyses (over 2400 in number) published in the interval between 

 1883 and 1901. The period of seventeen years lies within that during 

 which systematic petrography has been dominated by Rosenbusch's 

 names and definitions. In general, the number of analyses for each 

 rock-species is so large that their average would be but slightly modi- 

 fied by the inclusion of the analyses made since 1900. In many cases, 

 therefore, the extended labor required to search out from the literature 

 the additional analyses, has not been considered necessary for the 

 preparation of useful averages. For other averages it was necessary 

 to include analyses published since 1900. The sources of such infor- 

 mation are indicated below. Fortunately for the purpose, nearly the 

 entire period since 1884 has seen the application of more or less re- 

 fined methods of analysis ; so that errors of observation for the leading 

 oxides are relatively small. 



Method of Calculation. 



The method of computation used is essentially like that employed 

 by Washington and Clarke in their respective calculations of the 

 " average composition " of all igneous rocks. In general, only the 

 twelve more important oxides (including MnO) are recognized in 

 the following tables. Distinctly " inferior " analyses were not consid- 

 ered. In each case the average was computed according to the actual 

 numbers of determinations made by the analysts. Table I. shows 

 these numbers for the respective rock-types, each column being headed 

 by a key-number which corresponds with the named types of Table II. 

 For some of the rocks BaO and SrO were computed. Their sum 

 appears in the averages for CaO, as indicated in the tables. Similarly 

 C0 2 and Cr 2 3 were sometimes averaged and entered with H 2 and 

 Fe 2 3 respectively. As expected from the method employed, the 

 average totals nearly always ran well over one hundred per cent. All 



