The process sketched above was by no means continuous. The population was in slow but practically constant movement, much the same as that now taking place in the Zuni country; it was a slow migration. Outlying settlements were established at points convenient to cultivable fields, and probably were intended to be occupied only during the summer. Sometimes these temporary sites might be found more convenient than that of the parent village, and it would gradually come about that some of the inhabitants would remain there all the year. Eventually the temporary settlement might outgrow the parent, and would in turn put out other temporary settlements. This process would be possible only during prolonged periods of peace, but it is known to have taken place in several regions. Necessarily hundreds of small settlements, ranging in size from one room to a great many, would be established, and as the population moved onward would be abandoned, without ever developing into regular villages occupied all the year. It is believed that many of the single house remains of Mr Bandelier's classification[12] belong to this type, as do also many cavate lodges, and in the present paper it will be shown that some at least of the cliff ruins belong to the same category. [Footnote 12: See a paper by the author on "Aboriginal remains in Verde valley, Arizona," in 13th Ann. Rept. Bureau of Ethnology, p. 179 et seq.] The cliff ruins are a striking feature, and the ordinary traveler is apt to overlook the more important ruins which sometimes, if not generally, are associated with them. The study of the ruins in Canyon de Chelly has led to the conclusion that the cliff ruins there are generally subordinate structures, connected with and inhabited at the same time as a number of larger home villages located on the canyon bottom, and occupying much the same relation to the latter that Moen-kapi does to Oraibi, or that Nutria, Pescado, and Ojo Caliente do to Zuni; and that they are the functional analogues of the "watch towers" of the San Juan and of Zuni, and the brush shelters or "kisis" of Tusayan: in other words, they were horticultural outlooks occupied only during the farming season. Mr G. Nordenskioeld, who examined a number of cliff and other ruins in the Mancos canyon and the Mesa Verde region, adopts[13] a very simple classification, as follows: I. Ruins in the valleys, on the plains, or on the plateaus. II. Ruins in caves in the walls of the canyons, subdivided as follows: (a) Cave dwellings, or caves inhabited without the erection of any buildings within them. (b) Cliff dwellings, or buildings erected in caves. [Footnote 13: The Cliff Dwellers of the Mesa Verde, pp. 9 and 114.]