cavalry. It seems to me indispensable either to insert an et in the one place or an ut in the other. The condition of things, therefore, at this point, seems to be as follows. A hard infantry battle is going on upon the slope of the hill, the war-chariots attempting ineffectually to support the infantry of their countrymen; the cavalry of the Britons has disappeared, while a part of the Roman cavalry must be understood to have accompanied the infantry directly into the fight. Now follows a very corrupt and perplexing passage: miêque equestres ea enim pugnae facies erat, cum egra diu aut stante simul equorum corporibus impellerentur. These words, taken by themselves, are wholly unintelligible. The difficulties lie chiefly in equestres and diu, which have been emended respectively into aequa nostris and clivo. I can see no reason for the first change. The rest of this chapter is occupied with a description of the confusion caused by empty chariots and riderless horses, and the next chapter begins with an attempt of the Britons to take the victorious Romans. in the rear. Evidently the battle was not minime aequa nostris at this juncture, and although I cannot suggest any entirely satisfactory expression of this hopelessly corrupt passage, yet it seems to me that it was the most natural thing in the world for Tacitus to say, as Livy does of the battle of Cannae,* that the look of the battle was not that of a cavalry engagement. The war-chariots on one side, and the three thousand cavalry on the other, were in full action; but they were so intermixed and confused with the infantry, that the writer only goes on to describe this confusion-equorum corporibus impellerentur-vagi currus-exterriti sine rectoribus equi―transversos aut obvios incursabant. The other emendation, clivo for diu, seems every way good; only I should read. with Halm and Dräger, adstantes (for aut stante), rather than instantes with Kritz. Adstantes not merely resembles more nearly the manuscript reading, but is the proper word to use for the Britons (see adstiterant, a few lines above), while instantes would naturally apply to the Romans making their way up the hill. *Book xxii. ch. 47, minime equestris more pugnae. The thirty-seventh chapter begins with an attempt of certain British forces which had up to this time taken no part in the battle, to take the Romans in the rear, circumire terga vincentium coeperunt; and this expression, showing that the Romans still had the better of their antagonists, completely disproves Kritz' reading minime aequa, with his interpretation of the previous words, referring the check in the words haerebant to the Roman cavalry. Agricola met this aggressive movement by bringing up four alae of cavalry which had until now been held in reserve; auxiliary cavalry of course, inasmuch as the term ala is confined to divisions of auxiliary cavalry. The mention of this reserve of four alae implies that the rest of this cavalry was already engaged, as I have assumed. It also serves to support the opinion already expressed that the equitum turmae were British; the word alae would probably have been used had it been the Roman cavalry. Not merely were the advancing Britons repulsed, but the fresh cavalry were themselves carried round to the enemies' rear, aversam hostium aciem invasere. The description of the slaughter which followed is in the picturesque style characteristic of Tacitus, but calls for no special remark. The enemy appears to have been steadily pushed up the hill until they reached a piece of woods which has not been mentioned before, where they again made a stand, and gained some advantages by their familiarity with the ground. And again Agricola called up fresh troops, this time infantry, validas et expeditas cohortes, who searched the woods in conjunction with mounted and dismounted cavalry. These fresh troops must also have been auxiliaries. It is true the word cohortes is not decisive, although this is the term regularly used for the auxiliary infantry, in contrast with the Roman legions; for the legions too were divided into cohorts, just as the alae were divided into turmae. But if it had been detachments from the legions, the word manipulos would have been more likely to be used, and at any rate the statement that they were light-armed, expeditas, proves that they were auxiliaries; for the Roman legion contained at this period no light-armed soldiers. The same thing is proved by the fact that the officer killed in this engagement, Aulus Atticus, is called praefectus cohortis; the term praefectus is never used for a legionary officer. This movement finally brought the resistance to an end; the British lines broke and fled, and the conquest of Britain was complete. It appears, if this sketch is correct, that the battle was fought by Agricola wholly with his front line, of auxiliaries; the legionary soldiers not being brought into the engagement at all. It appears further that there was never any serious check-except, indeed, at the very beginning of the contestand never any actual repulse. The battle nevertheless was in three distinct stages, or rather presented three distinct crises. First, when the fighting eminus proved ineffectual, a charge of Batavians and Tungrians was ordered, followed no doubt by a portion of the cavalry; the fighting was then upon the hill-side, confused and disorderly, especially by reason of the presence of the war-chariots. The second crisis was the attempt of the Caledonians to take the Romans in the rear; this was frustrated by the prompt bringing up of divisions of cavalry, which themselves came round upon the rear of the enemy, and effected a general slaughter. The third was the fight in the woods, determined likewise by the commander bringing up fresh troops, with which the victory was secured. VIII.-On Inconsistency in Views of Language. By W. D. WHITNEY, PROFESSOR IN YALE COLLEGE. If the study of language is to reach and maintain the rank of a science, those who arrogate to themselves the position and authority of teachers of it ought above all things to see that the views they put forth are fairly consistent throughout. Attention to this would remove at least a good part of the lamentable discordance of view which now prevails among them respecting even matters of fundamental consequence. There is a body of well-ascertained and undeniable facts within the reach of every linguistic student sufficient, if logically combined, to establish an abiding outline of scientific doctrine -one which, however much it may be filled in and made more definite by the labors of coming generations of investigators, will be neither swept away nor essentially altered. As a noteworthy example of the errors arising from carelessness of logical consistency, we may take the opinions brought forward by certain contemporary authorities respecting the relations to one another of dialects and languages.* There is a connected series of objectionable doctrines on this subject, which, so far as has been observed, began with M. Renan, being laid down by him in his work on the "Origin of Language" (second edition, 1858; it has been since more. than once reissued). In the eighth chapter of that ingenious and eloquent work, the author calls attention to the "impossibility of a homogeneous language spoken over a considerable territory in a rude state of society;" and adds that “civilization alone can spread languages through great masses of population." No careful scholar will think of differing with M. Renan on these points; nor, indeed, will he dispute the * In the paper as read before the Association, and since printed, in substantially the same form, in the American Journal of Philology (vol. i., pp. 327-343), this specimen of the common offenses against consistency was much more briefly treated; it has seemed best to make the fuller discussion of it here a substitute for the already published original article. theoretical soundness of the latter's inference: namely, that "at the beginning there were as many dialects as families-I might almost say, as individuals." Only, he will ask, how many families, or individuals, are we to regard as having been in existence "at the beginning"? M. Renan's conception of a beginning seems to be a somewhat peculiar one; for he immediately goes on to say that "each group of men formed its language on a foundation established, indeed, by an anterior tradition, but following its own instincts," and so on: a "beginning," accordingly, with an "anterior tradition" behind it! Such a "beginning" can, it should seem, only be that of our historical knowledge of the communities in question; and this, as abundantly appears from other parts of his work, is the author's real meaning. He everywhere evinces the utmost unwillingness to allow that the pre-historic periods of language-growth are to be investigated by analogical inference from its historic periods; as we actually find the facts to be when they come within our field of view, such they must have been from the outset. Hence it is that he holds every language to have been produced "at a single stroke," its whole characteristic structure complete at the start, "like Minerva sprung from the brain of Jove "-a comparison which provokes the reply that it precisely fixes the status of the author's doctrine: as science, the one thing is just as acceptable as the other. All our knowledge of the history of language shows that changes, and quite especially changes of structure, are of very gradual, or even extremely slow, progress and accomplishment; and the bringing into being of a tongue like the Indo-European by a single effort of its speakers could only be paralleled with the production of a full-grown and completely armed goddess from that womb of masculine products the brain, after a slight headache, with the cæsarean aid of a blacksmith's hammer. The facts appealed to by M. Renan in support of his theory of original indefinite dialectic variety are, among others, the extreme diversity of the Polynesian languages which, accordingly, represents to him the beginning of things in that wide-spread linguistic family. What to others appears the unquestionable certainty with reference |