371. There are various words beginning with one of these letters (the liquids ρ, λ, μ, ν, the spirant σ, and the media δ), before which a short final vowel is often allowed to have the metrical value of a long syllable. Initial ρ appears always to have this power of lengthening a preceding vowel; but in the case of the other letters mentioned it is generally confined to certain words. Thus we have examples before
λ: in λίσσομαι, λήγω, λείβω, λιγύς, λιαρός, λιπαρός, λίς, λαπάρη, λόφος, and occasionally in a few others, but not (e. g.) in such frequently occurring words as Λύκιος, λέχος, λείπω.
μ: in μέγας, μέγαρον, μοῖρα, μαλακός, μέλος, μελίη, μάστιξ, μόθος, but not (e. g. μάχομαι, μένος, μέλας, μάκαρ, μῦθος.
ν: in νευρή, νέφος, νιφάς, νύμφη, νότος, νητός, νύσσα, once only before νηῦς (Il. 13.472); not before νέκυς, νόος, νέμεσις, etc.
σ: in σεύω, σάρξ , once before σύ (Il. 20.434), and once before συφεός (Od. 10.238).
δ: in δέος, δεινός, δεί-σας, etc., (stem δϝει-), δήν, δηρόν (§ 394).
This lengthening, it is to be observed, is almost wholly confined to the syllables which have the metrical ictus; the exceptions are, πολλὰ λισσομένη (Il. 5.358, so Il. 21.368, 22.91), πυκνὰ ῥωγαλέην (Od. 13.438, etc.), πολλὰ ῥυστάζεσκεν (Il. 24.755). Further, it is chiefly found where the sense requires the two words to be closely joined in pronunciation: in particular
- In the final vowel of prepositions followed by a case form
- ἐπὶ ῥηγμῖνι
- ποτὶ λόφον
- ὑπὸ λιπαροῖσι
- κατὰ μοῖραν
- ἐνὶ μεγάρῳ
- κατὰ μόθον
- διὰ νεφέων
- ἀπὸ νευρῆφιν
- κατὰ συφεοῖσιν
- κατὰ δεινούς
- ἐπὶ δηρόν
and similar combinations.
- ἐπὶ ῥηγμῖνι
- In fixed phrases
- ὥς τε λίς1
- κλαῖον δὲ λιγέως2
- ἀπήμονά τε λιαρόν τε3
- καλή τε μεγάλη τε
- εἶδός τε μέγεθός τε
- Tρῶες δὲ μεγάθυμοί
- τρίποδα μέγαν
- Πηλιάδα μελίην
- ὥς τε νιφάδες
- σὺν δὲ νεφέεσσι κάλυψε
- ὅτε σεύαιτο
- οὔ τι μάλα δήν
and the like.
- ὥς τε λίς1
These facts lead us to connect the lengthening now in question with the peculiar doubling of the initial consonant which we see in compounds
and after the augment (§ 67), as ἔ-ρριψα, ἔ-ρρηξα, ἔ-ρρεον, ἐ-λλίσσετο, ἔ-μμορε, ἔ-ννεον, ἔ-σσευα, ἔ-δδεισα (so the MSS., but Aristarchus wrote ἔδεισα). The words and stems in which this doubling occurs are in the main the same as those which lengthen a preceding final vowel, and the explanation, whatever it be, must be one that will apply to both groups of phenomena.
With most of these words the lengthening of a preceding vowel (or doubling of the consonant, as the case may be) is optional. But there is no clear instance in Homer of a short vowel remaining short before the root δϝει- (e. g. in the 2nd aorist δίον, the 1st aorist ἔδεισα, the nouns δέος, δεινός, δειλός, even the proper names Δεισήνωρ, etc.), or the adverb δήν. The same may be said of ῥάκος, ῥήγνυμι, ῥύομαι, ῥητός, ῥίπτω, ῥίον, also μαλακός, μελίη, νιφάς. Lengthening is also the rule, subject to few exceptions, with λίσσομαι, λόφος, νέφος, νευρή, ῥινός, ῥόος, ῥάβδος, ῥίζα, and some others (La Roche, H. U. pp. 47 ff.).
372. Origin of the Lengthening.4
The most probable account of the matter is that most of the roots or stems affected originally began with two consonants, one of which was lost by phonetic decay. Thus initial ρ may stand for ϝρ (as in ϝρήγ-νυμι), or σρ (as *σρέω, Sanskrit sravāmi); λίς is probably for λϝίς (with a weaker stem than the form seen in λέϝ-ων); νυός is for σνύος (Sanskrit snushā); νιφ-άς goes back to a root sneibh (Gothic snaivs, snow); μοῖρα is probably from a root smer; σέλμα is for σϝέλμα (Curt. s. ν.); and δει- in δει-νός, etc., is for δϝει- (cp. δεί-δοικα for δέ-δϝοικα). It is not indeed necessary to maintain that in these cases the lost consonant was pronounced at the time when the Homeric poems were composed. We have only to suppose that the particular combination in question had established itself in the usage of the language before the two consonants were reduced by phonetic decay to one. Thus we may either suppose (e. g.) that κατὰ ῥόον in the time of Homer was still pronounced κατὰ σρόον, or that certain combinations—κατα-σρέω, ἐΰ-σροος, κατὰ σρόον, etc.—passed into κατα-ρρέω, ἐΰ-ρροος, κατὰ ῤῥόον (or κατᾱ ῥόον). There are several instances in which a second form of a word appears in combinations of a fixed type. Thus we have the form πτόλις, in ποτὶ πτόλιος, Ἀχιλλῆα πτολίπορθον, etc.; πτόλεμος, in μέγα πτολέμοιο μεμηλώς, ἀνὰ πτολέμοιο γεφύρας. Similarly a primitive γδοῦπος survives in ἐρί-γδουπος (also ἐρί- δουπος), ἐ-γδούπησε, and γνόος in ἀγνοέω. Cp. also the pairs σμικρός and μικρός, σκίδναμαι and κίδναμαι, σῦς and ὗς, ξύν and σύν. It is at least conceivable that in the same way the poet of the Iliad said μοῖραν and also κατὰ σμοῖραν, μειδιάων but φιλο-σμειδής, δὴν ἦν at the beginning of a line, but μάλα δϝήν at the end, and so in other cases.
It is true that the proportion of the words now in question which can be proved to have originally had an initial double consonant is not very great. Of the liquids, the method is most successful with initial ρ, which can nearly always be traced back to νr or sr. And among the words with initial ν a fair proportion can be shown to have begun originally with σν (νευρή, νυός, νιφάς, νέω, νύμφη). The difficulty is partly met by the further supposition that the habit of lengthening before initial liquids was extended by analogy, from the stems in which it was originally due to a double consonant to others in which it had no such etymological ground. This supposition is certainly well founded in the case of ρ, before which lengthening became the rule.