180. The rules for the accentuation of compound verbs have been already given in § 88. They proceed on the general principle that (except in the augmented forms) the accent falls if possible on the preposition; either on the last syllable (as ἀπό-δος), or, if that is elided, then on the first (as ὕπ-αγε).
In regard to the other uses, and in particular the use with cases, the general assumption made by the Greek grammarians is that all prepositions are oxytone. They do not recognize the modern distinction according to which ἐν, εἰς, and ἐξ are unaccented. This distinction rests entirely on the practice of the manuscripts (Chandler, p. 254), and apparently arises from the accident of the smooth breathing and accent falling on the same letter (Wackernagel, K. Z. xix. 137).
Disyllabic prepositions, however, are liable in certain cases to become barytone. The exact determination of these cases was a matter of much difficulty with the ancients, and unfortunately we cannot now determine how far their dicta rest upon observation of usage, and how far upon analogy and other theoretical considerations. The chief points of the accepted doctrine are
- The disyllabic prepositions, except ἀμφί, ἀντί, ἀνά, and διά (except also the dialectical forms καταί, ὑπαί, παραί, ἀπαί, ὑπείρ, προτί), are liable to anastrophe; that is to say, when placed immediately after the verb or the case form to which they belong, they throw back the accent.
λούσῃ ἄπο (= ἀπο-λούσῃ)
Ζεφύρου ὕπο, etc.
Some held that the insertion of δέ before the preposition did not prevent anastrophe, and accordingly wrote ὦσε δʼ ἄπο, etc.
- Also, according to some, if the preposition stands at the end of a verse, or before a full stop (Schol. Α on Il. 5.283).
- Also, when it is equivalent to a compound verb (§ 177); as ἔνι, ἔπι, πέρι, πάρα (for ἔν-εστι, etc.). So ἄνα (for ἀνάστηθι); although ἀνά according to most authorities was not liable to anastrophe. Some wrote
πάρα γὰρ θεοί εἰσι καὶ ἡμῖν (Il. 3.440)
on the ground that in πάρ-εισι the accent is on the syllable παρ-.
- Twο prepositions are barytone in the adverbial use.
ἄπο when it is = ἄποθεν at a distance
πέρι when it is = περισσῶς exceedingly
Tο which some added ὕπο (as τρομέει δʼ ὕπο γυῖα, etc.).
- Monosyllabic prepositions when placed after the governed word take the acute accent (as an equivalent for Anastrophe); but only when they come at the end of the line. Some however accented Od. 3. 137
καλεσσαμένω ἀγορὴν ἔς πάντας Ἀχαιούς
Most prepositions, as appears from the Sanskrit accent, are originally barytone, and the so-called anastrophe is really the retention of the accent in certain cases in which the preposition is emphatic, or has a comparatively independent place in the sentence. Just as there is an orthotone ἔστι and an enclitic ἐστι (§ 87.1), so there is an orthotone πέρι and a proclitic περι, written περὶ before a governed noun, but in reality unaccented.
This view will serve to explain one or two minor peculiarities of Greek usage. Thus (1) it is the rule that when the last syllable of a preposition is elided before a case form, the accent is not thrown back. This is intelligible on the ground that the preposition is in fact without accent; and the same account will apply to the same peculiarity in the case of ἀλλά and τινά. On the other hand, (2) in the case of elision before a verb (as ὕπ-αγε) the accent is retracted, because the preposition is then the accented wοrd.1 Again, (3) the general rule of the Aeοlic dialect, that all oxytones become barytone, does not extend to prepositions, because they are not real oxytones.
The wοrd ἔτι (Sanskrit áti) is a preposition which happens to have survived (with the original accent) in the adverbial use only: cp. πρός = besides.
One or two suggestions may be added in reference to the prepositions which are generally said to be incapable of anastrophe.
ἀνά was thought by some to be capable of anastrophe, and this view is supported by the adverbial use ἄνα up!
ἀμφί is probably a real oxytone, like the adverb ἀμφίς. The corresponding Sanskrit preposition abhi is oxytone, contrary to the general rule.
The assertion that ὑπαί, παραί, προτί, etc., are not liable to anastrophe is difficult of interpretation. It may mean only that these words are not Attic, and by consequence that later usage furnished the grammarians with no examples.
If this is the true account of anastrophe, it is probable that the prepositions retained their accent in all quasi-adverbial uses, including tmesis—nοt only when they followed the verb or governed noun. The doctrine of the grammarians is unintelligible unless it admits of this extension. For if we write πάρʼ ἐμοί γε καὶ ἄλλοι because πάρα = πάρεισι, we must also write πάρα γὰρ θεοί εἰσι, where πάρα its equally emphatic. In Sanskrit too the preposition when separated from its verb is accented.
It is not so clear how far the later rules for prepositions in composition are to be applied to Homer. In Sanskrit there is an important difference between principal and subordinate clauses. In a principal clause the verb loses its accent, unless it begins the sentence (§ 87); the preposition (which usually precedes the verb, but is not always immediately before it) is accented. Thus we should have, on Sanskrit rules, such forms as πέρι δείδια, πέρι πάντων οἰδε, etc. But in subordinate clauses the accent is on the verb, and the preposition commonly forms one word with it, as in περιδείδια. If the preposition is separated from the verb, both are accented. In classical Greek twο changes have taken place: (1) the preposition and verb are inseparable, and (2) the accent is placed almost uniformly according to the "law of three syllables" (§ 88): if it falls on the preposition, as in σύμ-φημι, κάτ-εχεν, or on the verb, as in συμ- φήσει, κατ-έχει, the reason is purely rhythmical. The first of these changes had not taken place in the time of Homer. As to the second, we are practically without evidence. We do not even know when the law of three syllables obtained in Greek. It may be observed however that
- When a word of three syllables could not be unaccented, the form πέρι δείδια became impossible; but it does not follοw the πέρι lost its accent at the same time. An intermediate πέρι δείδια is quite admissible as a hypothesis.
- In many places in Homer it is uncertain whether a preposition is part of a compound or retains its character as a separate word. Thus we find
Il. 4.538 πολλοὶ δὲ περὶ κτείνοντο καὶ ἄλλοι
(Wοlf, from Ven. Α.)
Il. 16.497 ἐμεῦ πέρι μάρναο χαλκῷ
(πέρὶ sic Ven. A.)
Il. 18.191 στεῦτο γὰρ Ἡφαίστοιο πάρʼ οἰσέμεν ἔντεα καλά
Il. 1.269 καὶ μὲν τοῖσιν ἐγὼ μέθʼ ὁμίλεον (Ar.)
with the variants περικτείνοντο, περιμάρναο, παροισέμεν, μεθομίλεον. And the existing texts contain a good many compounds which we might write divisim without loss to the sense.
Il. 18.7 νηυσὶν ἐπικλονέονται
Od. 8.14 πόντον ἐπιπλαγχθείς
Od. 16.466 ἄστυ καταβλώσκοντα
Il. 2.150, 384; 3.12, 4.230, 5.332, 763, 772, 6.100, etc.
In reference to such forms we may fairly argue that the tendency of grammarians and copyists, unfamiliar with the free adverbial use of the prepositions, would always tend towards forming compounds; hence that modern critics ought to lean rather to the side of writing the words separately, and giving the prepositions the accent which belonged to them as adverbs.
With regard to the accent of prepositions in the ordinary use with case forms it is still more difficult to decide. A Sanskrit preposition generally follows the noun which it governs: hence it does not furnish us with grounds for any conclusion about the Greek accent.
180*. Apocope. Most prepositions appear in Homer under several different forms, due to loss of the final vowel combined (in most cases) with assimilation to a following consonant.
παρά and πάρ
ἀνά, ἄν, ἂμ (βωμοῖσι, φόνον)
κατά, κὰδ (δέ), κάβ-(βαλε), κάτ-(θανε), κὰρ (ῥόον), καμ-(μονίη), κὰγ (γόνυ), κὰκ (κεφαλῆς), κάλ-(λιπε), κὰπ (πεδίον)
προτί, πρός (for προτ-)
cp. ποτί, πός
ὑπείρ (for ὑπέρι), ὑπέρ
ἐνί, εἰν (εἰνί), ἐν
This phenomenon appears to be connected with the loss of accent which the preposition suffers when closely connected with a verb or case form. That is to say, from the adverbial forms πάρα, πρότι, κάτα, ἔνι, ἄνα (or ἀνά), etc., were formed in the first instance the unaccented παρ, πρὸς, κατ or κα, ἐν, ἄν. Then the pairs πάρα and παρ, etc., were used promiscuously. Finally one form was adopted as normal.
- 1. See Wackernagel, K. Z. xxiii. 457 ff. On this view, however, the original accent would be ἄπο-δος, ἔνι-σπες, πάρα-σχεs, etc. It may perhaps be preserved in the indicative ἔvι-σπες and imperative ἔνι-σπε (see § 88, where a different explanation of these forms was suggested).