Macédones, Macedónia; púĕris, puerórum, puerirúmque; aúdiunt, audiebam, audiebámus.1 Arabic and Ethiopic: kátălă, katálta, kataltínna; jáktóló, jáktólū, jaktólna, jaktìlúna; jakúlū, jakīlúna; vállădă, valládākā, valladátăkă ; Substantives: kátēlu, kātélātu, katelina; medinătă, înátu, inātáni; also with enclitics: kollāmā, ‘éndāmā, but 'ělámă, etc. At first sight the Hebrew intonation, which puts the

1 So in the Romance languages (N. B. in so far as the Latin quantity has been preserved, especially in Italian), as créděrě, credéva, credevámo; amába, ābúmo (in Spanish abămos, because the penultimate is here shortened); benéfico, but pudico, etc.; in Spanish, even in case of apocope of the final vowel if the quantity is not thus, as usual, lengthened, as ángěl, órděn, difícěl, ámăn (Ital. ámăno, from amant). Even in English the Latin accentuation, under the same conditions makes itself felt; e.g. the words in -átor, -ity, -íěty, (Latin l-itas, íètas), '-ent (Latin -ens), -Iměn; adjectives in -tc, -tfic, id, -it, etc., e.g. prolífic, histórie, mosářc, intrépid, illicit, and many others which cannot be traced to any definite termination. It is true that in many other words from the Latin (or French) the accent, conforming to the English rhythm, which springs from the accent of the German element, and is applied to the Romanic, where it, of course, can only produce confusion, is thrown back further than in Latin. This, however, has always been first brought about by apocope, together with a shortening of the last syllables, while the rule for the position of the accent itself remains in general unshaken. Only in many forms, in which many syllables of formation are heaped together at the close, the German tendency to shorten these and draw back the accent as near as possible upon the root conflicts with the rule, e.g. ámicăblý, ámrăbleness, where the accent is crowded back upon the fourth and fifth syllable. The same thing is effected in Italian and Spanish, especially by the appending of short, unaccented suffixes to the verb; e.g. mánda-visěně, búscă-mělo. But these are consequences which, in a certain sense, are forced upon language against its will, and which in pronunciation are either as far as possible accommodated to the normal measure, or are entirely avoided, as for the most part in Italian. Since in all these languages the Latin quantity and accentuation is so powerful as to determine the pronunciation, the latter can of course be understood and fixed only by reference to the Latin forms; and it is quite preposterous to seek to fix it merely by mechanical rules — according to the general form of the final syllables without regard to derivation — rules which have just as many or even more exceptions, and are of no use at all, and through which thick volumes, especially on English pronunciation, are made fruitless and aimless.

2 When this Essay was first published (1852) I assumed that also in Sanscrit — whose vocalization rivals the Arabic and Ethiopic in purity and trueness to its original character- the etymological principle of accentuation, as in Greek, stands under the rhythmical law, all the more inasmuch as the remarkable law running through all forms the law of equilibrium between root and formation or inflexion endings— testifies to the power of this law. This assumption, never

accent for the most part on the last syllable, and only by way of exception on the penultimate, never on the antepenulti

theless, has not been confirmed by the later investigations respecting the actual Sanscrit accentuation as regards the final syllables, - in Benfey's Grammatik (Leipsic, 1852) and Bopp's Vergleichende Accentuationssystem des Sanscrit und des Griechischen (Berlin, 1854). Bopp makes it even exactly opposed to the rhythmical law, whose prevalence in the Greek language he considers a mark of degeneracy, and designates, as the principle of the Sanscrit accent, the "pushing back" of the tone as far as possible towards the beginning of the word, which, he says, was esteemed the "most dignified and forcible" accentuation (p. 16). This would surely be no principle at all; and so in his preface (p. v) he lays it down, over against the logical principle of the German languages and the rhythmical principle of the others, as a third, free or grammatical, in Sanscrit confined to no limits (without law therefore). Not having grasped the significance of the rhythmical law, which, as I have shown, is a law of nature, and finding, in the consequent fact that the accent is limited to the last three syllables, only an arbitrary confinement of the accent, he makes the principle of the Sanscrit accent only to consist negatively in freedom from this constraint, without stating clearly what the positive law is. This is, however, in fact, of a rhythmical character; for the Sanscrit accent, to express it briefly, follows the centre of gravity, ie. that point in the word where the equilibrium between different elements or factors of the word rests. Only it is not merely mechanical, determined by the outward gravity (length) of the syllables, as in most languages, but chiefly organic, determined by the significance of the syllable in question in its relation to the whole word and the other factors, i.e. it rests upon the syllable which is for the form the most important (as the soul of the word), and the outward weight of the other factors operates only conditionally. Accordingly the accent rests: (1) on the root, on the kernel of the word (as in German), where the root appears strong (i.e. when the formation-endings are defective, strengthened from the simple root), either absolutely strong or conditionally so, according to the strength of the terminations that are added; (2) on the syllable of formation which determines the sense: either (a) derivatively, from the root to the particular form in appended vowels, a, u (nu), nâ, these likewise partly absolutely, partly conditionally strong; or (b) derivatively, and at the same time serving as inflection syllables in prefixes — augment and reduplication (long and short); the first absolute and changeless, the latter conditioned by the strength of the inflection-endings, and hence changing except in the desiderative. But when the accent is conditioned and changeable, there is seen an oscillation between four elements (factors) - stem, vowel of formation, mode-vowel, and personal ending. First, between the mode-vowel and personal ending of weak stems, viz. in the modus obliquus (potential, subjunctive, where the significance of the word lies in the mode-vowel and the personal endings are shortened, as in the preterite); here the accent rests either on the strong mode-vowel before the weak personal ending, or on the strong personal ending after a shortened mode-vowel; on the other hand, in case of reduplication, on this; in the Subj. 1st Pers., either on the strong stem or vowel of formation, in spite of the strong mode-vowel â, and in

matejust contrary to the Arabic and Ethiopic intonation (cf. katál, jiktól with kátala, jáktolo) seems to deviate entirely from this rule. Yet a closer view shows that the variation is only accidental, and that there is a substantial agreement. The reason why the Hebrew accent is limited to the last two syllables is the same as operates similarly in other languages, viz. because these syllables, like the syllables in general, according to the punctuation of the Hebrew language handed down to us, are always long or heavy, partly by position, partly by nature, or by artificial lengthening of the vowels. That it mainly rests on the ultimate, arises from the fact that this syllable, on account of the general apocope of the original final vowel (still retained in the Arabic and Ethiopic languages), has become in all radical words compound, i.e. the product of a contraction of two simple ones (-tal and -tàl from -tălă, -tèl from -tělă, etc.), and therefore, in a language which has such rigid rules and so delicate a sense for syllabication in regard to the consonant as the Semitic have, must necessarily draw the accent to itself. And this holds true, of course, also of the inflected final syllables of the same kind.1 The accent can fall on the

the Imp. (2d and 3d Pers. exc. 3d Sing.), when the stem is weak and without mode-vowel, on the personal ending. Next, the oscillation takes place between the vowel of formation and the personal ending: when the strong vowel a occurs, unconditionally upon it; when the weak vowels u, nu, nî occur, conditionally, according to the weakness of the personal ending. Finally, between the stem and the personal endings (when no vowel of formation comes between, and the personal endings are immediately appended), on the strong stem before weak personal endings, and vice versa.

1 The same phenomenon appears also in the Romanic languages, especially in the Spanish, where in consequence of apocope compound final syllables arise, which are now prolonged, and draw the accent to itself; e.g. solver from the Latin solvěrě. In French this lengthening is often carried still further, so as to form diphthongs, as sápěrě, sabèr, savoir. Also in Arabic and Ethiopic this rule must hold, although not recognized. That the Arabian grammarians say nothing about it, is quite explicable from the fact that the thing with them is only an exception; and, considering the defectiveness of their rules on this subject (in most languages the one least noticed; cf. the meagre, shallow, and disconnected statements in Silvestre de Sacy, Gram. Art. i. p. 145, ed. 1), this omission can prove nothing against analogy and the nature of the case. An indirect, although distorted, testimony for it is found, moreover, in the droll statement

penultimate only in two cases: either when a word originally monosyllabic and ending in two consonants receives, according to the rules for syllables, a second unaccented one, as kódesch from kidsch, tíhu instead of tohv, chóli for cholj; or when an inflection-ending, forming an open syllable, has been added. According to this simple law, its position changes in words to which new syllables are added, according to the character of the final syllables; e.g. verbs: katal, katelti, ktalt m, ktalt hu, ktaltihem; jakému, jkumùn, j'kumáni, jkumúníchem; substantives: kôtel, kotlih, kotlim; m'dinah, m'dinth, m'dinatháim; dabar, d'būrim, dbrêchim. Yet many inflection-endings with open final syllables — these being always long by nature-draw the accent to themselves; either unconditionally, by their greater length and importance, as, e.g. the feminine ending ah (-) in the noun (in contradistinction to the unaccented accusative ending ah) and the pronominal suffixes i and particularly ô and áh (-), which, as being contractions from áhû and áha, must have the accent; or only in the connection of the sentence, when the open penultimate syllable, which strictly should have the accent, is thrown out; whereas at the close of the sentence (in pause) it reappears on the latter. Thus, kat lû, katálû; kil dù, hbédu (from kābèd); vajj hì, vajj{hê (cf. warpós, TaTéρos). Sometimes also the accent is drawn back from the ultimate upon the open penultimate for particular reasons,

(in Silvestre de Sacy, ibid. No. 7), that every compound syllable, as often as it occurs in a word is to be accented, e.g. éstáchrágtu! (This Bopp also, ibid Rem. 159, finds inconceivable.) That the compound final syllable of the imperative has the accent (e.g. októl) is stated by Ewald, Gram. Ar. i. § 142. The same must be true of the Future apocop. and the pronominal suffixes tom, kom and hom, as well as of the paragogic Future and Imperative in an (for ánna); with the penultimate compound syllable before open inflection-endings it is of course the case. In Ethiopic the unaccented pronunciation of the compound final syllable in the noun, as mánfás, déngél (nominative) from the full mánfăsă of the Acc. and status constructus, remains; and so is to be explained doubtless also that of the Fut. (which is here always apocopated) ígábět, jégbăr, i.e. it arises from the prevailing accent of the full forms on the antepenultimate (cf. the above-mentioned extension of the opposite Hebrew rhythm beyond its limits). This is true perhaps also of the Arabic Fut. apocop. But I cannot believe it to be original and normal.

especially when an accented syllable immediately follows, e.g. kárá, ló (instead of kārà), látět, l'chà (from lūtèt).1

But all this accords entirely with the rhythmical law of the alternation and equilibrium of rise and fall, and indeed can be explained only by it. The reason why the accent can stand only on one of the last three syllables is that a simple rhythmical department, the elevation in which is marked by the accent, can, strictly speaking, include only two syllables, and can include three only when in the descending part two short syllables (or syllables shortened in pronunciation) are equal to one long one; so that the regulative scheme or here alternates with the variations, according to the nature of the syllables; in which connection it is to be observed that the emphasis of the rising slide gives to the short syllables so much force that it can balance even a protracted downward slide, and that the lack of intonation in the downward slide shortens long syllables. In polysyllables the fore part of the word forms a sphere in which preliminary strokes introduce the accented part, and, when it consists of several syllables, secondary accents come in of themselves, which in most languages are recognized and fixed, and in many, as the Hebrew and the English, are rhythmically regulated. The verbal accent however must, in such words, select not the fore, but the latter, part for its sphere (which part, on account of the inflection-endings which modify the sense, is also the most important part of the word), because only here can the rhythm be prominent, since it would otherwise be destroyed again by the unaccented syllables following. So also the rhythm of a

1 A real exception to the rule of accentuation in question is found however in the German language in its present form, in which, on account of a weakening of the former full, strong endings, and on the other hand a lengthening and contracting of short syllables in the middle, the accent has gradually come to be put always on the kernel of the word, on the root, and thus depends on the sense, not the length of the syllable. Yet even here the exception is not uniform; e.g. cf. lebendig with lebend, wahrhaftig with wahrhaft, allmächtig with állmacht. On this transition in general, vid. J. Grimm, Deutsche Grammatik, i. 12 sq. (2d ed.).