Along this part of our march there was much of what at a hasty glance seemed to be volcanic rock; but Oliveira showed me that it was a kind of conglomerate, with bubbles or hollows in it, made of sand and iron-bearing earth.  He said it was a superficial quaternary deposit formed by erosion from the cretaceous rocks, and that there were here no tertiary deposits.  He described the geological structure of the lands through which we had passed as follows:  The pantanals were of Pleistocene age.  Along the upper Sepotuba, in the region of the rapids, there were sandstones, shales, and clays of Permian age.  The rolling country east of this contained eruptive rocks—­a porphyritic disbase, with zeolite, quartz, and agate of Triassic age.  With the chapadao of the Parecis plateau we came to a land of sand and clay, dotted with lumps of sandstone and pieces of petrified wood; this, according to Oliveira, is of Mesozoic age, possibly cretaceous and similar to the South African formation.  There are geologists who consider it as of Permian age.

At Vilhena we were on a watershed which drained into the Gy-Parana, which itself runs into the Madeira nearly midway between its sources and its mouth.  A little farther along and northward we again came to streams running ultimately into the Tapajos; and between them, and close to them, were streamlets which drained into the Duvida and Ananas, whose courses and outlets were unknown.  This point is part of the divide between the basins of the Madeira and Tapajos.  A singular topographical feature of the Plan Alto, the great interior sandy plateau of Brazil, is that at its westernmost end the southward flowing streams, instead of running into the Paraguay as they do farther east, form the headwaters of the Guapore, which may, perhaps, be called the upper main stream of the Madeira.  These westernmost streams from the southern edge of the plateau, therefore, begin by flowing south; then for a long stretch they flow southwest; then north, and finally northeast into the Amazon.  According to some exceptionally good geological observers, this is probably due to the fact that in a remote geologic past the ocean sent in an arm from the south, between the Plan Alto and what is now the Andean chain.  These rivers then emptied into the Andean Sea.  The gradual upheaval of the soil has resulted in substituting dry land for this arm of the ocean and in reversing the course of what is now the Madeira, just as, according to these geologists, in somewhat familiar fashion the Amazon has been reversed, it having once been, at least for the upper two thirds of its course, an affluent of the Andean Sea.