turns southwest until the mouth of the Des Moines River is passed, and then turns southward again.

The rock disappears in the bed on passing Keokuk; the water then deepens, flowing on a sandy bed, and resumes its width of about half a mile between its ordinary banks. On the rapids there are no considerable or permanent islands, but as you go above or below, you find them as soon as the rockbed is left.

At Warsaw, three miles below the foot of the rapids, the Carboniferous rocks show in the bluffs, and so does the unmodified glacial drift, covered with the loess. (See Diagram F)

The Mississippi Valley at Warsaw is about eight miles wide; part of it a sand terrace, but most of it subject to overflow.

Proceeding up the Des Moines from the mouth, we leave the limestone strata at the point H (Diagram E), and do not meet with it again, either in the river bed or bluffs, till we reach the point C, where we find it in both places, and thence all along up this valley.

In this distance between H and C the bluffs are only on the left bank or north side, and the material appears similar to that at Madison or to the loess at Warsaw, shown in Diagram F. If we examine the valley of Sugar Creek, we find the bluffs cut down as low as on the Des Moines between H and C, and all of similar material, clay and sand, but no rock in place. Nor could we find on any of the branches between ABC and DEH any rock in place or learn of any. Within these limiting lines those who have dug wells for water find it without reaching rock.

At the place marked K on the sand terrace a well fifty feet deep encountered no rock, although this was as far down as the level of low water in the river.

The width of the main valley above Fort Madison is nearly the same as below Keokuk, and if we prolong the line of bluff between these two places, it will include a space between them where there is no known rock in situ, and which it appears reasonable represents the ancient channel, since filled up.

The loess formation. This name has been given in this country to a fine material deposited over all the other formations, including the glacial deposits near the river, but not upon the river sand-and-gravel terraces, which are therefore the more recent. It has a thickness of many feet in places. It is generally regarded as having been deposited in quiet water. In many places it is nearly uniform in thickness, conforming with the previous irregularities of the surface, as snow lies where it has not drifted. This would be brought about by a silt laden fresh-water current spreading out over one more dense and quiet, through which the fine silt dropped. Such conditions would exist in salt water, and the silt would prevent the exist

ence of many kinds of marine life, and would contain the animal remains brought from the fresh-water streams.

The margins of the loess deposits are not well defined, and do not appear to have been investigated as they should be. The deposit is well shown near Rock Island, but does not appear as high up as Dubuque (so far as I know). It extends up the Missouri as far as Sioux City, and the Missouri River made immense deposits into the body of water in which the loess was laid down. From the absence of marine fossils, this body of water has been regarded as fresh and without connection with the Gulf of Mexico except through an outlet. Judging by what I know of the deposits, I should think it did not connect with our present Great Lakes, and that a tongue of land or promontory separated the arms extending up the Mississippi and Missouri Rivers.

The streams which brought this fine loess material which has been so spread out must have brought down heavier material along the bottom that fell as soon as it reached the enlarged section, just as rivers make their deposits of heavy material now on reaching their natural receptacles. A most interesting case in point is in the deposits made in the Great Salt Lake in quaternary times, where its terraces show a level nearly a thousand feet above its present surface. The streams, like the Weber, built out the terrace at their point of discharge in the fan shape of a river-bar up to this terrace level near the ancient shore, and extending back in the river valley, which must then have appeared as a fiord. Although the ancient Weber River deposit diminished in amount as it extended into the ancient lake, yet it terminated quite abruptly. I think, then, there is good reason to regard the bluff-deposits at Fort Madison and vicinity, on the west side of the valley, as belonging to the loess period, and not to glacial times. These deposits are where streams would bring them from the northwest, while the glacial deposits seem to have come from the northeast. This filling up of the ancient valley of Keokuk and Madison, and also in the neighborhood of Rock Island, is where the mouths of considerable rivers probably were during the existence of the large body of water in which the deposition of the loess was made.

Formation of rapids since the loess.-When the body of water of the loess period disappeared and the emergence of the land again took place, the Mississippi did not regain its ancient channel, but cut the new one now occupied. These rapids cannot be accounted for by any special hardness of the rocks, but their resistance is increased because of the dip of the strata being nearly that of the river-slope. When the river was at a higher level and made the sweeping bend above Montrose so as to wash the loess bluffs, but little more erosion would have

been required there to have carried the river back again to its ancient channel during some extraordinary flood, and yet it might have been that the new channel would even after this remain the permanent one for ordinary stages.

Such an explanation as this may be applicable to the cases at Fountain Bluff and at the Grand Chain on the Mississippi just above the mouth of the Ohio, and to the almost incomprehensible changes of course in the Lower Ohio itself, shown on the general map.

Another interesting supposition may be made that the Mississippi in the last terrace period might have succeeded in washing down the bluffs, separating it near Burlington from the Crooked Creek flowing into the Illinois. (See Diagram 1, sheet 4.) The new channel would have double the descent to the mouth of the Illinois of the existing one, and we might have gained a new course for the river, leaving a larger ancient channel occupied by a smaller stream, and there would have been set at work a new cause to modify all the valley of the Illinois River and all the Mississippi above.

Summary of principal points presented.—I will summarize the principal facts that seem to be made out along the course of the Minnesota and Mississippi:

1. That the Minnesota Valley and the Mississippi Valley above the Ohio have been, as a rule, formed since the deposition of the glacial drift, for this exists in unmodified and modified forms in the banks of the river; and that the Winnipeg basin drained out southward along it.

2. That the loess deposits, extending up to the neighborhood of Savannah, are later than the last glacial drift.

3. That channels like those at the Des Moines Rapids and river terraces in that vicinity are more recent than the loess.

Explanatory hypotheses.-I have advanced this hypothesis of southern elevation and northern depression several times before and illustrated its effect on the Minnesota and on the Wisconsin Rivers, and in the first instance considered what the results would be in contiguous regions, and how far facts seemed to correspond, and it is unnecessary to repeat them here.

The hypothesis appears to be in accordance with a number of very important facts, and is consistent with observations as to southern elevation and northern depression now going on. It explains, by one widely exerted influence, many effects which, on the grounds of glacial action alone, requires many special glaciers, and it will answer as an explanation of the coming on and disappearance of the loess body of water, and for the change in the drainage of Lake Winnipeg.

I think this change of relative elevation south and depression north has been probably reversed at some periods, and repeated. This is important, for, if we can show that any movement of

the earth's crust is a recurrent phenomenon, it may help us to trace out its cause.

Approximate practical conclusions.-The only practical conclusion which can be drawn from the preceding discussion seems to be that the origin of the excavation of the valley is comparatively modern, and that it was from the operation of forces producing probably uniform results, and in a way that we have some approximate comprehension of it in general, from our knowledge of special localities.

Note.-Descriptions of the best known sections of the valley follow in the report.

ART. LI.-On some points in Lithology; by JAMES D. DANA. [Continued from page 343.]

4. Containing Quartz or not.-Since quartz is the most universal of the materials of rocks, its presence is least entitled to be made a basis for distinctions among them. In sedimentary deposits, the original of many of the crystalline kinds, it is a very common ingredient owing to their mode of origin, and its more or less abundance is a matter of no great geological importance. Sufficient reasons exist, therefore, for the course pursued by recent writers on lithology in making the presence or not of quartz even in crystalline rocks a basis only for a subdivision under a kind of rock. Thus there is under dioryte, quartz-dioryte; under trachyte, quartz-trachyte; under felsyte, quartz-felsyte; and so in other cases..

Syenyte is defined by such authors as consisting chiefly of orthoclase and hornblende. Now a rock made prominently of these minerals often contains also quartz; and the name for the quartz-bearing kind, which a system of lithology using the above-cited terms would seem to require, would be quartz-syenyte. To call it "hornblende-granite," as is often done, is at variance with the system which uses the word quartz as an affix in other cases.

This term "hornblende-granite" is at variance also with the fundamental idea and nature of granite. Granite is eminently a potash-bearing rock. The feldspar is a potash-bearing species; and the mica, whether muscovite or biotite, yields on analysis little less potash than the feldspar, the amount being eight to twelve per cent. These two micas are both present in most granite, gneiss and mica schist: and they are so near akin that they sometimes occur combined in a single crystal-the presence of a little iron in the original material having apparently determined the formation of the latter where it occurs. On the contrary the hornblende of such rocks contains usually

less than one per cent of alkalies, and rarely in any kinds over five per cent. Looking to chemical and mineralogical constitution the true criterion as to identity among rocks-the strongly drawn line is between the mica-bearing series and the hornblende-bearing series. Granite belongs to a mica and potash-feldspar series; and syenyte, whether quartzless or quartzbearing, to a hornblende and potash-feldspar series.

Moreover, the original syenyte, from Syene, Egypt (to which the name "syenites" was applied by Pliny and other ancient writers) is a quartz-bearing "syenites." The larger part of the syenyte of all Archæan regions is quartz-bearing. The quartzless kind is seldom met with in Eastern North America, or, as far as explored, in the Rocky Mountain region. There are hornblende granites; but these are granites which contain hornblenbe in addition to the mica and other ingredients.

Transitions are common between granite, hornblende-granite and quartz-bearing syenyte; but they are so also between these and quartzless syenyte, between syenyte-gneiss and ordinary gneiss, between hornblende schist and mica schist, and between these and other rocks. They are throughout lithology a source of difficulty in characterizing kinds of rocks, as already stated. But they do not set aside the fact that the division between the mica and potash-feldspar series and the hornblende and potash-feldspar series is the most reasonable on mineralogical and chemical grounds.

5. Containing "Plagioclase."-The fact that the composition. of the triclinic feldspars between the extreme species albite, a sodium-aluminum tersilicate, and anorthite, a calcium-aluminum bisilicate, may be explained by supposing them combinations of these species through isomorphous substitutions of the tersilicate and bisilicate (the amount of sodium present determining the amount of tersilicate in the combination, and the amount of calcium that of bisilicate) was immediately followed by the assumption that these two silicates combined indefinitely, and, therefore, that all the triclinic feldspars were essentially one species, and for this reputed species the name plagioclase has been used. Some ground for the assumption was found in the analyses of the feldspars; but how much was uncertain, because, in several cases, mechanical mixtures of one species with another had been ascertained to exist in crystals. Now that Des Cloizeaux has proved, by optical investigations, that several of the species of triclinic feldspars are really species, that is, that the combinations of the two silicates, the tersilicate and bisilicate, are based on definite ratios, as in combinations in other departments of chemistry, and that there are not indefinite blendings, the term "plagioclase" has become merely a synonym for "triclinic feldspar."