the Survey; but their analyses, with the exception of a single group to be noted later, do not fall within the scope of this paper.

Quite naturally, on account of the activity of the petrographers, the dominant feature of the laboratory work has been the analysis of rocks. These have been studied in great numbers and in the most thorough way. The results have appeared in widely scattered publications, official reports, monographs, bulletins, American and foreign journals, and so on. The object of this bulletin is to bring together this valuable material, together with such bibliographic and petrographic data as seems to be necessary in order to identify the specimens and to facilitate chemical discussion. Analyses of minerals have been included only when related to petrographic studies, appearing then in connection with the rocks to which they belong. Meteorites, of which twenty-nine have been analyzed, are brought into the work on account of their petrographic relations; and the groups of clays and soils have been admitted because of the bearing of these substances upon the study of slates and shales. The actual number of analyses given in the bulletin is as follows:

It may be observed that the classification thus indicated has not been rigorously followed. In a few instances the study of a sedimentary rock has been so related to that of its igneous neighbors that the analyses are best tabulated together; but these exceptional cases are few, and all are properly noted. The heading "igneous and crystalline rocks” has been used in the broadest and most liberal way, and doubtless many of the analyses given under it might properly be otherwise classified. In such cases of uncertainty, convenience has furnished the rule to follow.

Within each division of the analyses the classification chosen has been geographic. The petrographic grouping of the rocks would doubtless be best were petrographers agreed upon it; but their differences are many, and the chemist will do well to avoid them. The geographic method, moreover, has some advantages of its own; it facilitates the study of areas, it simplifies the bibliographic references, and it brings together, in great measure, the work of each petrographer for whom analyses have been made. Thus, most of Diller's work has been in California, most of Cross's in Colorado, and most of Iddings's in the Yellowstone National Park, and in each case the

analyses are massed, and their discussion is practically uniform in character. As regards nomenclature, each rock has received the designation given it by its describer, and no liberties have been taken. This plan may cause some lack of uniformity; but no other procedure seemed to be practicable. Whenever it was possible, however, I have inserted in italics the new names proposed by Cross, Iddings, Pirsson, and Washington in their classification of the igneous rocks. In most cases, these names were taken from Washington's great compilation of analyses, recently published by the Survey." This addition, it is believed, will assist petrographers in their study of the material thus brought together.

It will be noticed by anyone who uses this bulletin that the analyses. vary as regards completeness. Among the sedimentary rocks, especially, partial analyses are common, but in the igneous group thoroughness is more general. In the early days of the chemical division many analyses were made along the older lines, just as they are still made in many laboratories to-day-that is, only the main constituents, those having direct petrographic significance, were determined. In such analyses the minor ingredients, like titanium, phosphorus, barium, strontium, chlorine, etc., were ignored; and, although the results are satisfactory in some respects, they leave much to be desired. Latterly, greater completeness has been sought for, the work done has been much fuller, and the data obtained can be discussed with much higher approaches to accuracy. The old form of "complete analysis" is to be discouraged; it leads too often to erroneous conclusions; and only the best modern methods of work and of statement should be tolerated. The fuller analyses, moreover, have brought some interesting points to light; titanium now appears to be one of the more abundant elements, and barium and strontium are found to be almost universally diffused in igneous rocks in quite perceptible quantities.

On general principles the analysis of a rock and its petrographic description should be two parts of the same investigation, matching each other completely. In practice, however, this rule does not always hold, and the departures from it are in two opposite directions. For example, an analysis of the older type says nothing of titanium and phosphorus, while the microscope reveals the presence of sphene and apatite. In this case the petrographer has been more thorough than the chemist. On the other hand, a full and perfect analysis may be given, accompanied by a petrographic description of the most general kind, in which only the main mineral constituents of the rock are noted. Here the analysis has been incompletely used, and the petrographic discussion is defective. It is hoped that the publication of this material may lead to a clearer recognition of the mutuality which should exist between the chemical and the microscopic re

a Professional Paper No. 14, 1903. Abbreviations in Professional Paper No. 28.

searches, and so bring, in the future, both lines of investigation more into harmony. Hitherto the chemist and the petrographer have worked too much apart, and each has too often misunderstood the purpose of the other. If the study of the thin section could always precede the analysis, the petrographic problems could be stated more clearly, and the chemical evidence might be rendered much more pertinent and satisfactory.

During the preparation of this bulletin much assistance was rendered by the petrographers and geologists connected with the Survey, especially with reference to analyses hitherto unpublished. In each case credit has been given for the data thus added. Twenty-eight analyses of rocks from Montana, executed by or under the direction. of Prof. L. V. Pirsson, of Yale University, and having been made in connection with regular Survey work, are included in the tabulations. With this exception all of the analyses given were made in the Survey laboratories. To those executed in the laboratory at Washington "record numbers" are attached, which serve to identify them on the record books of the Division of Chemistry. Of the abbreviations used for bibliographic reference only four need explanation, and they refer to the official publications of the Survey. "Ann." for Annual Report, "Mon." for Monograph, "Bull." for Bulletin, and “P. P.” for Professional Paper are the four in question. The others relate to well-known journals, and are familiar to all geologists. The letters P. R. C., following the description of a rock, refer to the Petrographic Reference Collection of the Survey, and are followed by the number assigned to the rock in that series.

THE AVERAGE COMPOSITION OF ROCKS.

In a paper published some years ago," on the relative abundance of the chemical elements, I computed the average composition of the primitive crust of the earth from 880 analyses of eruptive and crystalline rocks. Of these analyses only 207 were from the laboratories of the Survey, while 673 were collected from various other American and foreign sources. A large proportion of them were incomplete, regarded from a modern point of view, and yet the results obtained were fairly conclusive.

In Bulletin No. 148 a similar estimate was given, based upon 680 complete analyses found in the Survey records, plus some hundreds of determinations of silica, lime, and alkalies. Again, in Bulletin No. 168, a third estimate was presented, representing 830 complete analyses and some partial determinations, all made in the Survey laboratories. In 1899 Harker' published a computation covering 397 analyses of

a Bull. U. S. Geol. Survey No. 78, 1891, p. 34.

b Geol. Mag., 4th decade, vol. 6, p. 220.

British rocks, and recently Washington" has worked out the average of 1,811 analyses given in his compilation.

For the more important constituents of igneous rocks the five estimates mentioned above agree remarkably well, and yet they are not thoroughly comparable. The 397 analyses discussed by Harker were in most cases incomplete, at least when considered from a modern standpoint. In only 34 of them was titanium taken into account, and in only 55 is phosphorus mentioned. These omissions affect the percentages of other things and lessen the value of the computation very materially. Excluding manganese the five averages may be tabulated as follows. Minor constituents will be considered later:

An examination of the foregoing table will show several discrepancies, and one of them is in the variable treatment given to water. In two columns hygroscopic water does not appear; in two others total water is given; in Washington's estimate a discrimination has been made. If we reject the figures for water and recalculate the remaining constituents to 100 per cent, the comparison of estimates will assume the following form:

care.

Between Washington's computation and mine there is one fundamental difference-a difference of method. In my own calculations analyses were averaged together as if each one was complete-an assumption which is not justifiable. In some cases minor constituents were not reported; in other cases they were determined with great The average of analyses varying in this way will obviously give too low a result for the rarer substances. Washington's policy was to consider the rock-forming elements separately, averaging each according to the actual number of determinations made, and thus to secure a truer group of estimates. For example, his average represents 1,811 determinations of AlО, and SiO,, 1,625 of Fe̱,O, and FeO, 1,139 of TiO2, 955 of P,O,, and only 731 of MnO. The data thus utilized, however, coming from many laboratories and representing various methods of analysis, are evidently of very unequal value; and this consideration counts for something against the result finally obtained. I have, therefore, thought it desirable to repeat the computation, applying Washington's method to the determinations made in the laboratories of the Survey alone, and using all of the analyses, partial or complete, for my purpose. Every determination made upon an igneous rock has been thus utilized. When an element was definitely reported as absent, it has counted as zero in the averaging.