was one of the party that attacked the Pequot fort near Mystic, Connecticut, to which he has reference, his account is of value as it supplements that of Wood:

"They choose a piece of ground, dry and of the best advantage, forty or fifty foot square (but this was at least two acres of ground). Here they pitch, close together as they can young trees and half trees as thick as a man's thigh or the calf of his leg. Ten or twelve feet high they are above the ground and within [the ground] rammed three foot deep with undermining, the earth being cast up for their better shelter against the enemy's dischargements. Betwixt these palisadoes are divers loopholes, through which they let fly their winged messengers. The door is for the most part entered sideways which they stop with bows and bushes as need requireth. The space therein is full of wigwams, wherein their wives and children live with them."

The palisades were set close together, but open spaces between logs not perfectly straight were unavoidable. These open spaces were probably used as loopholes. Underhill,1 describing the same structure, says:

"This fort or palisado was well nigh an acre of ground which was surrounded with trees and half trees, set into the ground three feet deep, and fastened close one to another, as you may see more clearly described in the figure of it before the book."

The illustration referred to, which appears in Underhill's News From America (1638), was evidently made by a wood engraver from a rough ground plan. It is of little value except as showing the fort to have been circular, with two entrances, one upon either side, each formed by overlapping the ends of the stockade, leaving a passageway between them. This fort is said to have contained about 60 or 70 wigwams.

Gookin says that at Natick "there was a handsome large fort, of a round figure, palisaded with trees." The fort at Penobscot was 70 feet long and 50 feet broad and within it were 23 wigwams.3 Philip's fort, the site of which is at South Kingston, Rhode Island,

1 Underhill's narrative in Orr's History of the Pequot War, p. 78, note.

2 Gookin's Historical Collections, Mass. Hist. Coll., first series, vol. 1, p. 181. 'Drake's Indian Wars, p. 325.

had "besides high palisades, an immense hedge of fallen trees of nearly a rod in thickness, surrounding it, encompassing an area of about five acres." It is said to have contained about 500 wig

wams.

acres.2

Another estimate gives the size of the enclosure at 3 to 4

A few instances are recorded of the apparent use by Algonquians of a trench and embankment without palisades as a defensive work under circumstances which probably rendered the erection of a stockade unpracticable. Two traditions current in past years among the New England Indians are as follows: A party of Nipmuc entrenched themselves on the shore of Quinebaug River against the Narraganset, where they remained three days. Fifty years ago these earthworks were visible (De Forrest, Indians of Connecticut, p. 268). We are also told (Mass. Hist. Coll., 3d series, vol. VI, p. 197) that a company of Mohegan invaded Block Island and were driven to a bluff, where they "by some means dug a trench around them toward the land, to defend them against the arrows of their enemies." This earthwork was known locally as

the Mohegan fort.

In all there are about twenty Indian forts mentioned by the early explorers and colonists of New England between the years 1605 and 1676, nearly all of which were in Massachusetts (including the province of Maine) and Connecticut. It seems evident from a study of the above accounts that the old earthworks described in this paper are the remains of Indian fortifications of the same general kind as those seen by the colonists. The one at Millis is of greater extent and its form is composite; the embankments and trenches, however, appear to be identical with historic examples.

The levelling by cultivation of portions of certain other earthworks in eastern Massachusetts, renders it difficult if not impossible to determine their original form. Some of them were extensive, and may have formed enclosures as great as the one at Millis. With our present knowledge, there seems to be no good reason for attributing these remains to other than Algonquian origin.

1 S. G. Drake, Indians of North America, fifteenth edition, pp. 218-219. 2 Ibid., p. 218.

That this people occupied the greater portion of New England for a long period seems certain, for they were probably the originators of most of the shell-heaps of our coast. With the exception of the Champlain watershed in Vermont, and possibly certain other small sections of western New England, the Iroquoian tribes do not seem to have occupied these states. There are indications, however, of the occupancy of eastern, and perhaps central, New England by a non-pottery-making people, possibly the Beothuk, but there seems to be no evidence that the Beothuk constructed fortified enclosures of the types known to have been common among the Algonquians, although they did build extensive deer fences with "half-moon breast works" at intervals. There are, doubtless, many embankments of the types described above in various sections of New England that are known but locally, and it is hoped that this brief account may prove an incentive to further investigation as to the distribution and origin of this class of remains in these

states.

PEABODY MUSEUM OF HARVARD UNIVERSITY,

CAMBRIDGE, MASS.

1

1 Rev. George Patterson, The Beothiks or Red Indians of Newfoundland, Trans. Roy. Soc. Canada, Sec. II, 1891, p. 133.

D'

INCORPORATION AS A LINGUISTIC PROCESS

By A. L. KROEBER

R Sapir's recent paper on "The Problem of Noun Incorporation in American Languages" is such a masterly interpretation of the evidence connected with this subject, even though the essay is avowedly a refutation of the thesis advanced by the present writer a few years before,2 to the effect that such incorporation is a chimæra, that it remains a cause of gratification to have taken the stand which has been productive of so novel and valuable a contribution.

One point of primary importance that Dr Sapir brings out clearly is the fact that noun incorporation has no necessary or inherent connection with pronominal incorporation, as it has been called, or "rather inflection," as he aptly designates it. Dr Sapir has gone farther than the writer in pointing out that there is rather an exclusion between the two processes, in that a pronominally incorporating language should find noun incorporation unnecessary, and vice versa; and even though, as he says, the facts do not entirely bear out this a priori consideration, it is nevertheless a conception of the greatest importance in the present state of our understanding of linguistic phenomena. The custom heretofore has been to assume that noun incorporation was merely a form or phase of pronoun incorporation, or even the reverse; and, as long as this view prevailed, there was no hope of a correct analysis of such evidence as was accumulating. In fact this assumption has been the cause of a persistent misunderstanding of the subject. That the present writer's argument, which was based on the contention that the assumed connection did not exist, overshot the mark and ended by doubting well-authenticated but unexplained facts that had been called noun incorporations, must perhaps be admitted. But this

1 American Anthropologist, (N. S.), XIII, 250-282, 1911.

2 Verh. XVI. Intern. Amerikanisten-Kongr., Wien, 569-576, 1909.

is of little moment in comparison with Dr Sapir's agreement that such "noun incorporation" as he has been able to establish has nothing whatever to do with so called pronominal incorporation. Until this point of view is conceded, or proved erroneous, the evidence on the question will continue to be misunderstood.

Dr Sapir takes issue with the writer's definition of noun incorporation as "the combination into one word of the noun object and of the verb functioning as the predicate of a sentence," on the ground that a morphological and a syntactical requirement are joined: in other terms, that the definition exacts not only a certain type of word formation, as is justifiable, but also a logical relation between the elements, which is unreasonable. This criticism is correct, and it can only be said in palliation of the definition that, inasmuch as the phenomenon to which it relates was not believed to exist, less attention was given to theoretical exactness of statement than to an endeavor to express what had customarily been meant by the phrase "noun incorporation." In short, the basis of the definition was historical rather than logical. As a matter of fact, one of the arguments advanced against the existence of noun incorporation as thus defined was the circumstance that incorporation of the subject noun had not been alleged, but would have to be expected in at least some cases if object incorporation were at all common. Here again Dr Sapir maintains a most commendable conservative attitude, and, instead of using the apparent absence of one form of incorporation as an argument against the existence of the other, demonstrates the occurrence of both, together with still other phases, such as adverbial and predicative. This leads to a new conception: incorporation is no longer an essentially objective process, as had usually been assumed and as the writer accepted for purposes of refutation, but is non-syntactical in its nature. However the evidence on the question may in future be interpreted, this is a logical point that compels recognition.

Dr Sapir also gives the solution of the problem-which would have been puzzling if it had not been so generally ignored-why in alleged incorporating languages incorporation sometimes takes place and sometimes does not. As the writer put this point,' the

1 Op. cit., p. 574.