388. In seeking to arrive at general conclusions as to the rules and structure of the Homeric hexameter, it was necessary to leave out of sight all the words whose metrical form is uncertain on account of the possible or probable loss of an initial consonant. It is time to return to this disturbing element of the enquiry.
The scholars who first wrote on this subject had few materials for their investigations outside of the Homeric poems. To them, therefore, the digamma was little more than a symbol—the unknown cause of a series of metrical anomalies. In the present state of etymological knowledge the order of the enquiry has been to a great extent reversed. It is known in most cases which of the original sounds of the Indo-European languages have been lost in Greek, and where in each word the loss has taken place. Hence we now come to Homer with this knοwledge already in our possession. Instead of asking what sounds are wanting, we have only to ask whether certain sounds, of whose former existence we have no doubt, were still living at the time when the poems were composed, and how far they can be traced in their effect on the versification.
389. Nature of the Evidence from Meter. The questions which are suggested by the discovery in Homer of traces of a lost digamma cannot be answered without some reference to the very exceptional circumstances of the text.
Whatever may be the date at which writing was first used in Greece for literary purposes, there can be no doubt that the Homeric poems were chiefly known for some centuries through the medium of oral recitation, and that it was not until the time of the Alexandrian grammarians that adequate materials were brought together for the study and correction of the text. Accordingly when these scholars began to collect and compare the manuscripts of Homer, they found themselves engaged in a problem of great complexity. The various readings, to judge from the brief notices of them preserved in the scholia, were very numerous; and they are often of a kind which must be attributed to failure of memory, or the license of oral recitation, rather than to errors of transcription. And the amount of interpolation must have been considerable, if there was any ground for the suspicions so often expressed by the ancient critics.
It follows from these circumstances that an attempt to restore the lost ϝ throughout the text of Homer cannot be expected to succeed. Such an attempt necessarily proceeds on the assumption that the text which we have is sound as far as it goes, or that it is so nearly right that we can recover the original by conjecture. With an imperfect text the process can only be approximate. We may be satisfied if the proportion of failure is not greater than the history of the text would lead us to expect.
The loss of the ϝ-sound, moreover, must have been itself a cause of textual corruption. It led to irregularities of meter, especially to frequent hiatus, and there would be a constant tendency to cure these defects by some slight change. The insertion of the ν ἐφελκυστικόν was almost a matter of course (see however § 391). The numerous alternative forms used in the poetical language, and the abundance of short particles such as γε, τε, ῥα, etc., made it easy to disguise the loss of ϝ in many places. We cannot be surprised, therefore, if we have often to make the reverse changes.
A few instances will serve to show the existence in pre-Alexandrian times of corruption arising from the tendency to repair defects of meter.
In Il. 9.73 the MSS. have πολέεσσι δʼ ἀvάσσειs, Aristarchus read πολέσιν ηγρ ἀνάσσεις. Both are evidently derived from πολέσιν δὲ ἀvάσσειs (i. e. ϝανάσσεις), corrected in two different ways.
In Il. 13.107 the MSS. have νῦν δʼ ἕκαθεν, the reading of Aristarchus; but Zenodotus and Aristophanes had νῦν δὲ ἕκας (i. e. ϝέκας).
In Il. 9.88 the reading of Aristarchus was τίθεvτο δὲ δόρπα ἕκαστος; other ancient sources had δόρποv (the reading of most MSS.).
In Il. 14.235 πείθεν, ἐγὼ δέ κέ τοι εἰδέω χάριν ἥματα πάντα, the order χάριν εἰδέω was preferred by Aristarchus.
Two very similar instances are
Il. 5.787 κάκʼ ἐλέγχεα, εἶδος ἀγητοί (Ar. ἐλεγχέες)
Il. 9.128 γυναῖκας ἀμύμονα ἔργα ἰδυίας (Ar. ἀμύμοναs)
In Od. 5.34 ἤματί κʼ εἰκοστῷ . . . ἵκοιτο the "common" texts of Alexandrian times (αἱ κοινότεραι) omitted the κʼ, which is not necessary, and may have been inserted in imitation of ἤματί κε τριτάτῳ κτλ. (Il. 9.363).
In Od. 1.110 οἱ μὲν ἄρʼ οἶνον ἔμισγον some MSS. omit ἄρʼ. So in Od. 3.472 most MSS. have οἶνον οἰνοχοεῦντες (vulg. ἐνοινοχ.).
In Od. 2.331, 8.174, 13.125 the ε of αὖτε is elided before a wοrd with ϝ. But in each case there is MS. authority for reading αὖ.
In Od. 8.526 the MSS. are divided between ἀσπαίροντʼ ἐσιδοῦσα and ἀσπαίροντα ἰδοῦσα.
It should be observed that the argument from these instances is equally good, whether the readings ascribed to Zenodotus, Aristarchus, etc. are conjectures made by them, or were derived (as is more probable) from older sources. They equally serve to illustrate the process by which traces of an original ϝ were liable to be gradually effaced. And it is not likely that there was any deliberate attempt to emend Homer on metrical grounds. It is enough to suppose that the meter helped to determine the preference given (consciously or unconsciously) to one or other of the existing variants.