¹ Lg. 38.311 (1962). In what follows all phonemes and phonemic sequences are to be understood as initial unless otherwise labeled.

² In classical Greek there are a number of cases of the allomorph /a/ before a vowel other than /o/, and before /o/ in the sequence /oi/; but all or most of these can be traced back to forms that once had initial /w/. Before /o /, however, the allomorph /a/ occurs in cases where there never had been a /w/. A partial list of these cases, and a discussion of them, can be found in A. C. Moorhouse, Studies in the Greek negatives 47–9 (Cardiff, 1959). Moorhouse believes that forms such as /áozdos/ ‘with no branches’ are analogical after such a form as /aoíkεitos/ ‘uninhabited’, which derives from */awoíkεitos/. This explanation is indeed possible in some cases, but seems insufficient to explain the comparatively large number of /ao-/ compounds and the comparatively small number of /ae- ai- aa-/compounds.

³ ‘Prosodical digamma’ is Hoenigswald's term and seems a good one. Digamma was unknown to Homer, but had existed in certain old metrical formulas that he inherited from the epic oral tradition. Hence, though the consonant itself no longer existed, its metrical effects did. Hiatus, normally avoided in Greek, was allowed by the poet before certain words, which we know from other sources once began with /w/. Hence ‘prosodical digamma’ represents a sandhi rule in Homeric epic, though one that was not without exception. Few words that once began with /wo/ involve this sandhi rule, and this has led to the assumption that /w/ disappeared initially before /o/ earlier than before other vowels. /oi/ < */woi/, however, behaves prosodically like /e/ < */we/. See Pierre Chantraine, Grammaire homérique²1.123ff. (Paris, 1948). The same phenomenon has been observed in other dialects as well. See Eduard Schwyzer, Griechische Grammatik 1.225 f. (Munich, 1939); C. D. Buck, The Greek dialects 48 (Chicago, 1955).

⁴ We know that the initial sequence /wo/ lasted in Greek through the Mycenaean period, for the Linear B tablets contain numerous instances of (wo). See Ebbe Vilborg, A tentative grammar of Mycenaean Greek 43 (Göteborg, 1960). Hence there would have been relatively little time for the two historical stages necessary to preserve this hypothesis.

⁵ There is in Greek another prefix /a/, this one having developed from *[sṃ], which had only the one morphophonemic shape //a//; it could not appear as /an/ under any circumstances. Hence when /wo/ merged with /o/, such forms as */awollέ:s/ ‘all together’ were replaced by /aollέ:s/. No one objects to this form, and yet the only difference between it and /áosmos/ ‘without smell’ is that the latter contains a prefix that also has the phonemic shape /an/.

⁶ See Moorhouse, Studies 41 ff., Schwyzer, Gram. 1.431. This view, however, supposes that [ṇ] was phonemic in PIE; otherwise [ṇ] plus vowel should have yielded [n] plus vowel. J. Puhvel, Lg. 29.14 ff. (1953), has proposed a laryngeal approach which would account for the occurrence of /an/ before a vowel.

⁷ This statement of the morphophonemics may be oversimplified, for the allomorph /a/may have already occurred before a vowel in /áupnos/ ‘sleepless’ and /áatos/ ‘insatiate’. See Moorhouse, CQ NS 11.11 fn. 2 (1961).

⁸ The rule, as stated, implies that /an/ and /a/ were in free variation, as I believe they were immediately after the merger. In classical times, however, a list would be required, for some forms show only /an/, while others have primarily /a/ with one or two instances of /an/.

⁹ Hence I regard Moorhouse's suggestion (supra fn. 2) of analogical creation as essentially correct, but would set it back considerably in time. Analogy, however, as indicated by the above, is not the correct term, /o/ was alone in allowing the allomorph /a/ before it, and therefore /a/ tended to spread to all forms in /o/.

¹⁰ Presumably those dialects would have been all those which show the merger of /wo/ and /o/. We have direct evidence only for Arcadian, however. Most of the Doric dialects quite clearly preserved the distinction between /wo:/ and /o:/. See Buck, Dial. 48; Schwyzer, Gram. 1.226.I make no apologies for seeming to violate the canon ‘once a phoneme always a phoneme’: it may be that [w] initially before other vowels was phonetically different from [w] before [o o:]. I insist only on the fact that there was no contrast between [wo:] and [o:], though there was a contrast between [oi] and [woi], even in Arcadian.

¹¹ More precisely the dialect of Mantinea, Buck Dial. 198 #17. Elsewhere in Arcadia forms of this verb show no trace of /w/. The only other case of /wo/ in Arcadia known to me is the divine name (worthasia) found in E. Schwyzer, Dialectorum Graecarum exempla epigraphica potior a 326 #673 (Leipzig, 1923), a name that is most common in Laconia, and presumably spread from there. See Émile Bourguet, Le dialecte laconien 123 (Paris, 1927). F. Bechtel, Die griechischen Dialekte 1.321 (Berlin, 1921), concluded from this name that Arcadian preserved the sequence /wo/. A. Scherer, however, in Handbuch der griechischen Dialekte 2.129 (by Thumb and Scherer; Heidelberg, 1959), feels that the inscription is not even written in the Arcadian dialect.

¹² References to older work in Schwyzer, Gram. 1.226 fn. 1. The Mycenaean evidence has been taken into account by Michel Lejeune, Mémoires de philologie mycénienne 1.73 with fn. 58 (Paris, 1958). He rightly considers the Arcadian form a problem, but not one that vitiates the evidence of Linear B.

¹³ On this same inscription there occurs (anodeasas). This has been interpreted either as /án ô:d eássas/, e.g. by Buck, Dial. 198, or as /áno:d eássas/, e.g. by Schwyzer Dial. 320. Needless to say I assume /áno:d/, for /án ô:d/ should appear as 〈anwod〉 if my explanation of (wophlekosi) is correct.

¹⁴ That foreign influence is involved in the engraving of this inscription is rendered likely by a number of graphic innovations. For a discussion of the letter forms and the possible Argolic (or Achaean) origin of certain features of the script, see L. H. Jeffery, The local scripts of archaic Greece 212 ff. (Oxford, 1961). Alternatively we might imagine that an Arcadian, struck by the fact that the Laconians sometimes spelled with 〈wo〉 what was to him simply /o:/, introduced 〈wo〉 here, considering it merely an initial allograph of 〈o〉.