The Dialect of Sappho and Alcaeus and the Dialect of Epigraphic Lesbian

From Vol. 1, No. 1, Winter 2016

By Keita Kashima, Magdalen College, University of Oxford

ἀλλὰ κὰμ μὲν γλῶσσα ἔαγε λέπτον

δ ̓ αὔτικα χρωῖ πῦρ ὐπαδεδρόμηκεν…


The literary dialect of Sappho and Alcaeus, and its relationship with the vernacular, epigraphic Lesbian dialect, have been much discussed in the past. At first sight, the poets appear to write naturally in their vernacular dialect; one may therefore think that the dialect in which their poems were written is the same as the non-literary dialect. However, the current consensus is that many of their poems use epic diction, and that a number of other problematic factors, including uncertainties in textual transmission, emendation and the absence of useful archaic inscriptions, complicate our examination of the literary dialect and its similarities/differences with the epigraphic dialect. This paper has three objectives: first, to summarize briefly the problems involved in our examination; secondly, to demonstrate through a detailed linguistic commentary on IG XII 2.1 that, with a few exceptions, the individual, dialectal features which we find in the poems are very similar to those we find in inscriptions (§1); and finally, to discuss the preposition ὐπό and its problematic, potentially hyper-Aeolic form ὐπά (§2), showing its interesting but frustrating argumentum ex silentio.


This paper is a revised extract from my secondary school project in 2014[1] that examined the position of the Lesbian dialect or, to be more precise, the written dialect of Mytilene[2] within the Aeolic dialectal group. It also attempted to examine to what extent the epigraphic (hence non-literary) Lesbian dialect is the same as the dialect in which Sappho and Alcaeus originally wrote their poems. In the project, I discussed the issues in trying to establish the exact relationship between literary Lesbian and epigraphic Mytilenean/Lesbian, and I concentrated on several forms in the fragments of the poets, which are probably hyper-Aeolic (e.g., the doubling of nasals, -αι- for naturally long -α-) or the products of orthographic convention (e.g., the internal -ζ- as -σδ-, the initial δι- as ζ[3]). The introduction of such forms possibly resulted from the standardization of lyric texts and the hypercorrection of their dialectal forms sometime in the history of their transmission, particularly during the Hellenistic Period. The presence of such forms makes our comparison of the literary dialect with the epigraphic dialect extremely difficult, and this problem is combined with many other factors, which I mention briefly below:

(1) The Lesbian dialect is poorly attested in the Archaic Period and even in the Classical Period until the mid-4th century, so most Lesbian inscriptions that we have to compare with the dialect of Sappho and Alcaeus are over three hundred years after the poets floruerunt.

(2) There are numerous uncertainties for the transmission and the modification of their texts, and it is possible that their dialect became corrupt or Atticized because of non-Lesbian scribes.[4]

(3) Possible inconsistency in the poets’ dialect.[5]

(4) Influence of Homer’s language and of other early Greek poetry on the poets.[6]

§1 The linguistic analysis of IG XII 2.1, the oldest inscription from Lesbos that is long enough for a detailed study, seems to suggest that the ancient editors edited the texts of Sappho and Alcaeus accurately without any apparent dialectal alternation, as the dialectal forms in the fragments, except in some cases, seem to correspond to the Lesbian dialect in the Classical Period.[7] This may partially explain why the dialect of the Lesbian poets, who were regarded highly throughout the Classical World, was rarely imitated by later poets: it was extremely local.[8] If the extreme locality of their dialect was a factor that prevented later poets from imitating Lesbian poetry, then Theocritus’ Idylls 28-30 and Balbilla’s Epigrammata are truly exceptional. [9]

           — — — —Ε[— — — — — — — — — ὄττι]

             [δέ κε αἰ] πόλις [ἀμ]φότ[εραι ․․․․․․․․․․]

             [․․․․․] γράφωισι εἰς τὰ[ν στάλλαν ἢ ἐκκ]-

             [ολάπ]τωισι, κύ[ρ]ιον ἔστω. [τὸν δὲ κέρναν]-

5           [τα τὸ] χρύσιον ὐπόδικον ἔ[μμεναι ἀμφο]-

             [τέρ]αισι ταῖς πολίεσσι· δι[κάσταις δὲ]

             [ἔμ]μεναι τῶι μὲν ἐμ Μυτιλήναι [κέρναν]-

             [τι] ταὶς ἄρχαις παίσαις ταὶς ἐμ Μ[υτιλ]-

             [ή]ναι πλέας τῶν αἰμίσεων, ἐμ Φώκαι δὲ τ-

10        αὶς ἄρχαις παίσαις ταὶς ἐμ Φώκαι πλ[έ]-

             ας τῶν αἰμίσεω[ν]· τὰν δὲ δίκαν ἔμμεναι

             ἐπεί κε ὠνίαυτος ἐξέλθηι ἐν ἐξ μήννε<σ>-

             σι· αἰ δέ κε καταγ[ρέ]θηι τὸ χρύσιον κέρ-

             ναν ὐδαρέστερον θέλων θανάτωι ζαμι-

15        ώσθω, αἰ δέ κε ἀπυφ[ύ]γηι μ[ὴ] θέλω<ν> ἀμβρ[ό]-

             την, τιμάτω τ[ὸ] δικαστήριον ὄττι χρὴ α-

             ὖτ<ο>ν πάθην ἢ κατθέ[μ]εναι, ἀ δὲ πόλις ἀναί-

             τιος καὶ ἀζάμιος [ἔσ]τω. ἔλαχον Μυτιλή-

             ναοι πρόσθε κόπτην. ἄρχει πρότανις ὀ

20        πεδὰ Κόλωνον, ἐ[μ Φ]ώκαι δὲ ὀ πεδὰ Ἀρίσ[τ]-


          “Anything that both cities write… on the

          (stele) or erase, let it be valid. Anyone

          who alloys the gold is to be subject to

          trial to both cities. For the person who

          alloys in Mytilene, the judges are to be

          all the magistrates, more than half in

          number: in Phocaea (the judges are to

          be) all the magistrates, more than half in

          number. The trial is to take place within

          six months of the end of the year. If one

          is convicted of willingly debasing the

          gold, let him be punished with death: but

          if he escapes the conviction since he did

          not willingly commit the crime, let the

          court decide what is necessary to do,

          what he should suffer or should pay. But

          let the city be without any responsibility

          and be unpunished: the Mytileneans

          drew the lot to issue the coins first: the

          magistrate after Colonus puts into effect

          the agreement, and in Phocaea the

          magistrate after Aristarchus.”

Line 2

  • πόλις: this is very likely to be nominative plural, because the following subjunctive is in the plural, although we cannot reject the possibility that it is the direct object of γράφωισι.[10] The noun is likely to have a long iota. It is possibly an extension from the accusative plural.[11] This form is unique to Lesbian and no comparable form can be found in poetry.

Line 3

  • γράφωισι: the 3rd person plural active present subjunctive (also line 4: …τωισι). This form is found only in Lesbian and in northern Chios, where there is strong Lesbian influence. The original 3rd person plural active ending –nti (preserved in West Greek) underwent assibilation in East Greek to –nsi (preserved in Arcadian), which underwent further alternation in other dialects. In Lesbian, the n in the cluster –ns() in an intervocalic position or at word end after a vowel was lost gradually, but it influenced the preceding vowel, forming an i-diphthong (e.g., Lesbian λύοισι; contrast with Attic λύουσι,[12] both from λύοντι). In the case of γράφωισι, either the ending –nti was attached to the thematic subjunctive stem γράφω- and an i-diphthong was formed even with the long vowel ω, or the ending –ισι simply extended from the thematic indicative and was attached to the omega.[13] See also line 8.

Line 5

  • ὐπόδικον: see §2.2.

Line 6

  • πολίεσσι: dative plural of i-stems in inscriptions is –εσσι (e.g., Sappho 31.11), but this is not uniform in poetry, as –σι is sometimes found (e.g., Sappho, 105.c2, Alcaeus 117.b35); this may indicate that the poets used an older form of the dative plural, although this cannot be confirmed.[14] The ending –εσσι is not exclusively Aeolic, as other dialects also have this ending.[15] If there was originally only one sigma for a s-stem (e.g., Sappho 2.10), it is likely that the i-stem dative plural ending later spread to other stems, including the s-stem.[16]

Line 7

  • Μυτιλήναι: the long alpha remains unaltered (cf. line 10: τὰν δὲ δίκαν; lines 13-14: ζαμι-ώσθω; line 17: ἀ; line 18: ἀζάμιος). The same phonological rule applies in Sappho and Alcaeus, even in their ‘Homeric-toned’ poems (e.g., Sappho 44.34).

Line 8

  • ταὶς ἄρχαις παίσαις (also in line 10): from *tans arkhans pan(t)sans. It is the accusative plural, from –ns. The consonant cluster becomes –ις, affecting the preceding vowel (see line 3). By coincidence the same change happens in Elean. See for example Alcaeus 140. 4, Sappho 1.24, etc. αι sometimes replaces long α even when it is not phonologically possible (e.g. Sappho 1.14 μειδιαίσαισ’), and this is likely to be a form of hyper-Aeolism introduced by Hellenistic editors.[17]

Line 9

  • πλέας: from *πλέοας (< *plēyoh-n̥s<*plēyos-n̥s), found in Homer (e.g., Il.2.129). The Proto-Greek form was *pleh1-yos, and was replaced by the n-stem (πλείων).[18] In Attic, the older s-stem forms remain in the nominative and the accusative masculine and feminine plural πλείους (< *plē-yoh-es < *pleh1-yos-es). The Attic accusative form is based on the nominative form. No comparable form is found in the fragments of Sappho and Alcaeus.
  • αἰμίσεων (also line 11): cf. Attic ἥμισυς. ἥμι- derives from the lengthened e-grade of *semi- and the shift [e:] to [ai] is triggered by /si/ and /mi/,[19] and its form can be compared with Alcaeus 42.13 αἰμιθέων and Sappho 44.14 αἰμιόνοις as well as Αἰσίοδος for Ἡσίοδος cited by the grammarian Herodian from the 2nd century A.D.   It is unclear whether this sound change was regular (cf. ἦσι in Sappho 109[20]).  Since the prefix ἠμι- is found in Assos and Aegae,[21] αἰμι- may be a local feature of Mytilene, as Miller suggests.[22]

Line 11

  • ἔμμεναι: the Lesbian athematic infinitive (also likely in line 7; –μεναι), demonstrating the influence from the East Greek athematic ending -ναι,[23] as in Alcaeus 140.9. See line 12 for the explanation of the double nasal.

Line 12

  • κε: this conditional particle (also lines 8) is frequently used by Sappho and Alcaeus.
  • μήννε<σ>σι: the double nasal from the simplification of the intervocalic group, which consists of a resonant + s, or s + a resonant, or a resonant+ a semivocalic y, is a feature in Thessalian and Lesbian that is preserved from the Proto-Greek stage.[24] This form can be compared with the nominative singular σελάννα (cf. line 3 στάλλαν; compare also Attic σελήνην), which derives from the Proto-Greek form *selas-nā (e.g. Sappho 96.8). This double nasal form is likely to be the source of hyper-Aeolic forms, for example Sappho 1.16 κάλημμι; the form probably is not justifiable phonologically and most of the earlier inscriptions do not support the possibility of such forms as being genuine Lesbian.[25]

Line 13

  • αἰ: this conditional conjunction is frequently used by Sappho and Alcaeus.
  • καταγ[ρέ]θηι: dialectal[26] passive subjunctive for αἱρέω (Sappho 31.14: ἄγρει). ἀγρέω is from *h2ger-, the same root as ἀγείρω.[27]  
  • κέρ-ναν: the masculine nominative singular present participle of κέρνᾱμι (mix), contracted from *κερνάων,[28] where the combination α+ω gives a long α. No comparable form is found in the fragments of Sappho and Alcaeus, although Alcaeus has the athematic present participle κέρναις (338.6). The co-existence of thematic and athematic forms is found in Homer as well, for example κιρνάω and κίρνημι.[29]

Line 15

  • ἀπυφ[ύ]γηι: the prefix and the preposition ἀπύ is used by Sappho and Alcaeus in all places (e.g., Sappho 98.11).[30] Hodot 1990 agrees that ἀπύ et ἀπό are two Indo-European variants, and suggests the possibility that the two forms always co-existed in Lesbian, although this is impossible to determine because of the paucity of epigraphic evidence between the 5th and the 4th century B.C.[31] Alternatively, ἀπύ in Lesbian may be a later development. The final, original omicron may have changed to an upsilon (which occurs especially in Arcado-Cypriot), but there is no valid explanation for this in Lesbian. It is true that, in Lesbian, the change from [o] to [u] sometimes took place initially before [m].[32] In poetry, ἀπύ can be apocopated in some conditions (e.g., Alcaeus 371, probably Sappho 27.10), but this is not the case for Lesbian inscriptions.
  • ἀμβρότην: Lesbian (as well as Elean and Laconian) thematic infinitives end in –ην, ultimately from *-e-sen (see also line 17: πάθην; line 19: κόπτην). The Attic equivalent is ἁμαρτεῖν. The origin of the aspirate in Attic and Homer is unknown.  Both forms are from the zero-grade of *h2mert-. The etymological root for this verb is also unknown but, considering  *n̥-h2mert-es- > νημερτής (Attic), it is likely that the root had an initial *h2and that the aspirate in Attic and Homer is analogical.[33] It was once suggested that ἁμαρτεῖν was formed by adding the alpha privative to the root *smer, but this raises some phonetic problems.[34] The analogy to the root *smer, however, would explain the origin of the aspirate. The vocalic liquid develops into ρο/ορ in Lesbian (ρα/αρ in Attic), followed by the epenthesis of the beta in *amrot, like the delta in *anros (see ἀνδρός). Cf. Sappho 5.5 ἄμβροτε (strong aorist 3rd person singular). Homer also has the Lesbian form, and it is not aspirated.

Line 16

  • ὄττι: assimilation of *yodkwid (cf. Sappho 1.15, the neuter genitive singular in Sappho 16.3-4). The neuter nominative and accusative spread to other forms analogically (e.g. Sappho 26.2, 31.2, Alcaeus 38.A12). However, this is not the case for inscriptions. The attested forms show little variation, as Hodot 1990 (p.139) writes: ‘les formes attestées sont peu variées’. The author shows, for example, that the form ὄστις is attested five times, but ὄττις is not recorded (contrast Sappho 31.2).

Line 17

  • κατθέ[μ]εναι: apocopation and assimilation of κατά as a prefix are not always observed in the lyric fragments (Sappho 105.c2), and this is true also for the language of inscriptions. Apocopation seems to have depended on the following condition: ‘les caractéristiques phonologiques du mot du contexte’.[35]

Line 18

  • Μυτιλή-ναοι: Attic Μυτιλήναιοι. See Sappho 98.b3 dative singular Μυτιληνάωι. Hodot 1990 includes a ‘monetary caption’ (légende monétaire [501]), which has ΜΥΤΙΛΗΝΑΟΝ. It is genitive plural and possibly predates IG XII 2.1, where the Ionic alphabet is used. In front of an internal vowel, diphthongs ending in -i were lost and only the first element of the diphthong was retained as a short vowel. This phonological change seems to be regular (Sappho 16.5, 24.a4 etc.). But see Sappho 31.14 ποίας; this is metri causa.

Line 20

  • πεδά (also line 12): = μετά, unrelated in origin. It is found in both poets (Sappho 55.4, Alcaeus 50.4) and is used also as a prefix (Sappho 55.2). 
  • Φώκαι: Attic Φωκαίαι. See line 18 for the loss of the iota in an i-diphthong before an internal vowel. The alpha in Φώκαι is long because of the combination α+α.


§2 In this section, I talk about the difficulty in judging whether ὐπά, which is found in some fragments, is hyper-Aeolic or not. Many scholars either seem to accept that ὐπά is genuine Lesbian without giving any justification or do not mention this problematic preposition.[36] Two publications discuss the issue, Bowie 1981 and Hooker 1977, but their conclusions do not agree with each other:

Hooker: “There is no good reason to suppose that the preposition ὐπά was known to Alcaeus and Sappho. It is a grammarians’ form, constructed from ὐπό by analogy with κατά, and helped, no doubt, by the existence of ὐπά in WG dialects and of Homeric ὐπαί. ὐπά is found in the following papyri: Sappho 1.9, Alcaeus 6.14, 38a.7, 117b.8. … (and in a) quotation of Sappho 31.10 in the Codex Parisinus. [Footnote 41] it is interesting to notice that in Theocritus’ Aeolic poem 29.23 both papyri and codices read ὐποδάμναται. …That it (ὐπά) is not the form used by Lesbian, at least in the Classical Period, is shown by the presence of ΥΠΟΔΙΚΟΝ in the monetary agreement between Mytilene and Phocaea.” [37]

Bowie: “Hooker (25f.) would keep ὐπό, arguing that ὐπά is a grammarian’s creation, on the analogy of κατά: cf. ὐπά in W.Greek and ὐπαί in Homer. However, I do not see why they should have wished to create such a form: κατά is not regular Lesbian and the W.Greek form is hardly to the point. Nonetheless, Alcaeus could have used both ὐπά and ὐπό, which is the inscriptional form.” [38]

In short, Hooker completely rejects the view that ὐπά is genuine Lesbian, whereas Bowie is perhaps more cautious than Hooker and suggests that ὐπά was used by Alcaeus (and presumably Sappho), with the possibility that archaic Lesbian had both ὐπά and ὐπό. My aim is to support Bowie to some extent and question Hooker’s firm belief that ‘it is a grammarians’ form’, although I am overall inclined to follow Hooker’s view. I divide this section in the following way:

1. Formation of ὐπό and ὐπά

2. Discussion on ὐπόδικον

3. Theocritus 29.23

4. Epigraphic evidence

5. Analysis based on our epigraphic evidence

6. Conclusion


ὐπό is from the PIE *(s)upo,[39] and in Lesbian the breathing was lost through psilosis.[40]

The development of ὐπά from *(s)upo cannot be explained with utter certainty. Buck’s explanation of the formation of ὐπά is the same as Hooker’s: he attributes it to the outcome of the analogy with κατά.[41] Buck’s statement presumably takes into account the traditional, and perhaps the most convincing, explanation by Bechtel that: ‘die Dichter von Homer an ὐπαί [a locative-case form] gebrauchen: das Verhältnis von διά zu διαί, von κατά zu καταί, von παρά zu παραί fordert ὐπά neben ὐπαί’.[42]

More generally speaking, however, there seems to be a tendency for disyllabic prepositions to end with the alpha, and this phenomenon was perhaps a result of some analogical spread in the early post-Mycenaean history of Greek. For example, Mycenaean has pa-ro, which is replaced, in later Greek dialects, by παρά (in non Attic-Ionic dialects with frequent apocopation). ‘Alpha-Harmonie’ has been proposed by to theorize the correlation; it is argued that the vowel of the first syllable affects the second.[43] But the theory can be criticized, as it may not have occurred in the (expected) case of ἀπό. The publication attempts to prove that *apo/*h2epo once became *ἀπά, but its explanation is not satisfying. More importantly for our question, *(s)upo does not have an initial alpha and, should the theory apply to other vowels, it would be (h)upu,[44] not (h)upa. To conclude, little is known about the Post-Mycenaean development of Greek prepositions, and many theories are indeed shaky. ὐπά may have in fact been influenced by both the presence of prepositions ending in alpha and by the analogical proportion described in Bechtel.[45]

Buck claims that ὐπά is a feature of Lesbian, as well as of Boeotian, Locrian and Elean, but ὐπά is found only in compounds in the latter two dialects. ὐπά is therefore very rare in the West Greek dialects, which makes it doubtful that ὐπό was turned into ὐπά by their influence. Moreover, although κατά is not always apocopated (see line 17 of the commentary), apocopation is much more common. It is therefore doubtful that ὐπά was invented by ancient editors through analogy with the more rare, so Hooker’s explanation for the presence of ὐπά in the texts of Sappho and Alcaeus is not entirely convincing.


The term ὐπόδικον is borrowed from the Athenian legal terminology, and it literally means ‘subject to a trial.’[46]  Hooker fails to mention that the word is borrowed from the Attic dialect. Did ὐπόδικον undergo psilosis?[47] It cannot be confirmed because of the absence of an elision that gives a non-aspirate as the last letter of the preceding word. For this reason there are, at this point, two proposals:

  1. As it is an Athenian legal term, Mytilene had simply borrowed the term without any dialectal alteration; therefore the rough breathing in ὑπόδικον was conserved (but the modern editor of the inscription presupposed psilosis, though breathings are unmarked on the actual inscription).
  1. Psilosis did take place for this Attic term.

Both proposals are possible. However, as it is well known that initial psilosis is observed in Lesbian inscriptions, there is every reason to find proposal 2 more plausible. But it is worth considering that loanwords from Koiné can retain their rough breathings, which may suggest that words alien to Lesbian were not always dialectalized; this may also explain the presence of ὑπό rather than ὐπά.[48]

It may also be the case that Atticization was already taking place in the time of IG XII 2.1.[49] If that is the case, then the reason why ὐπά was not the form used in Lesbian at least in the Classical Period can be attributed to the influence of Attic.


As already quoted above, Hooker mentions Theocritus’ Aeolic Poem 29.23, where both papyri and codices read ὐποδάμναται. But this hardly helps us in our argument, because it simply leads to wild speculations, as I show below, concerning specifically the codices:

a. If Theocritus imitated the literary Lesbian dialect fully in Poem 29, we have further evidence of rejecting ὐπά as genuine Lesbian.

b. If Theocritus did not attempt to imitate the literary Lesbian dialect fully in Poem 29, we cannot be certain whether ὐπά is hyper-Aeolic.

c. If Theocritus did attempt to imitate the literary Lesbian dialect fully but used texts which had been Atticized and had ὐπό, even though ὐπά should have been the correct reading (and therefore genuine Lesbian) instead, we must conclude that Theocritus 29 was not fully written in the literary Lesbian dialect.

Proposal ‘b’ does seem true, and proposal ‘a’ false; in all of Theocritus’ ‘Aeolic’ Idylls, Koine and Lesbian forms coexist.[50] But did the poet choose what to imitate, or was his understanding of Lesbian insufficient (for the reasons given in proposal ‘c’)? But none of these possibilities can be verified with utter certainty and, though it is interesting, as Hooker writes, Theocritus 29.23 is not a piece of evidence that can decisively confirm that ὐπά is hyper-Aeolic, because Theocritus’ poem is not fully in the Lesbian dialect. So the preverb ὐπό, which we find in 29.23, is hardly a reliable proof (in addition to the possibility of dialectal corruption in the course of transmission).[51]


So far, Bowie’s argument seems more plausible than Hooker’s, but our epigraphic evidence clearly supports the latter. ὐπά (both as a prefix and a preposition) is never found in inscriptions from Mytilene, but only once from the 2nd to 3rd century A.D.[52] According to Hodot, ὐπό in contrast is recorded 33 times (at Mytilene as well as in other Lesbian-speaking cities) as a preposition with the genitive.[53] His examples include Koine and dialectally mixed inscriptions, but also Lesbian inscriptions too, such as IG XII 2.14, which is dated approximately to 300-275 B.C.


To account for the presence of ὐπά in some texts of Sappho and Alcaeus and its extreme rarity in epigraphic documents, the following possibilities can be proposed:

  1. ὐπά, which had been used in the time of Sappho and Alcaeus, was already obsolete by the time of surviving epigraphic documents, therefore the engravers/commissioners of inscriptions used the more modern form ὐπό.
  2. Both forms co-existed: the dialect of Mytilene always retained ὐπό as well as ὐπά, the latter being a post-Mycenaean development.
  3. Sappho and Alcaeus consciously used ὐπά, a word alien to their native dialect. Possible reasons may have been to enhance the artificiality of their poetic language and the final alpha that was analogical to many other disyllabic prepositions.
  4. The dialect of their poetry was modified through numerous transmissions and/or became standardized at some point. The non-Lesbian form ὐπά was interpolated.

Proposal D may first seem the least credible of the four. As shown in §2.1, there is no obvious reason to alter ὐπό and introduce the alien form ὐπά, which is rare even in West Greek. Moreover, as I have already demonstrated in the linguistic commentary (§1), the Hellenistic editors seem to have kept the local Lesbian dialect of Sappho and Alcaeus almost intact. It is also important to remind ourselves that ὐπά is never found in Attic-Ionic literature and inscriptions. Though highly speculative, suppose that an Attic-Ionic or Koie speaker copies a text of Sappho, and finds ὐπό in the copying material. Why would he alter it to ὐπά, which is potentially a hyper-Aeolic form, in the process of transmission and without any justification (assuming his lack of knowledge of West Greek dialects)?

Proposal C is possible, but not entirely satisfactory, given that it is generally true that the individual, dialectal features of their poetry are distinctly local, and there is no reason to change such a common preposition and employ a form that is not found even in Homer.

The fact that ὐπό was always used instead in both pure Lesbian and dialectally mixed inscriptions between the 5th century B.C. and 2nd century A.D. may make proposal B on its own unlikely.

Yet proposal A, or the combination of proposals A and B, is still possible.[54] ὐπα may have been the more common form or was the only form. Its absence in inscriptions may be explained by historical and linguistic changes such as the increasing influence of Attic-Ionic. This may have penetrated so far in the case of ὐπά that the preposition was virtually lost (for proposal A), forgotten or was considered unfashionable (for the combination of proposals A and B), even during the revival of the city’s old dialect later in her history.[55]


To summarize, given the epigraphic evidence (§2.4), it is not unreasonable to conclude that ὐπό was Mytilene’s vernacular preposition and prefix in the time of IG XII 2.1 and onwards, which is in accordance with Hooker’s assumption. But his explanation does have some weaknesses (cf. §2.1 and 2, possibly 3b and 3c). ὐπά may well be a literary form or was somehow introduced through emendation by Hellenistic scholars, but because we have an unrecoverable gap in epigraphic documentation of at least 200 years, there is no firm evidence to reject that it was the conventional or the alternative preposition and prefix that had been lost by the Classical Period (§2.5).

The lack of epigraphic evidence in the Archaic Period makes it difficult to judge whether ὐπά is hyper-Aeolic. What is clear, however, is that epigraphic Lesbian in general goes into decline in the mid-to-late Hellenistic Period under the influence of Koine, followed by a sudden resurgence in the late first century B.C. and onwards, perhaps inspired by the dialect of Lesbian monody. It may seem that we can use these late inscriptions for our examination of the literary dialect, but we are once again faced with problems in addition to (1), (2), (3) and (4):

(5) There seems to have been no systematic approach to archaizing the dialect.

Cassio speculates that local grammarians may have influenced the ‘dialectal revival.’[56]  If so,

(6) Did these grammarians know the Lesbian dialect a few centuries before their time?

(7) Did their contemporary dialect, which was influenced increasingly by Attic and Koine, affect and alter the dialect of the poems?

(8) Was the Lesbian dialect in standardized Alexandrian texts the same as in locally transmitted texts?

Our speculation can go on endlessly. There is essentially no way to prove or disprove that forms that are potentially hyper-Aeolic were in fact genuine archaic Lesbian. Some can be categorized as hyper-Aeolic more confidently (see lines 8 and 12 in the commentary), but the case for ὐπά is not easy to judge. Despite not being totally persuasive, Hooker’s judgment is at the moment more plausible than Bowie’s, mainly because of the epigraphic evidence. Yet this view may change dramatically if an inscription, which is substantial in length and is from the Archaic Period, is discovered.[57]


F. Bechtel, 1921. Die griechischen Dialekte, Weidmann, Berlin.

W. Blümel, 1982. Die aiolischen Dialekte: Phonologie und Morphologie der inschriftlichen Texte aus generativer Sicht, Vandenhöck& Ruprecht, Göttingen,

A.M. Bowie, 1981, The Poetic Dialect of Sappho and Alcaeus, Arno Press, New York.

V. Bubenik, 1989. “Hellenistic and Roman Greece as a Sociolinguistic Area”, Amsterdam.

C.D. Buck, 1955. The Greek Dialects, The University of Chicago Press.

D.A. Campbell, 1990. Greek Lyric I, Loeb Classical Library, Harvard, second edition.

A.C. Cassio, 1986. Continuità e Riprese Arcaizzanti Nell’Uso Epigrafico Dei Dialetti Greci: Il Caso Dell’Eolico D’Asia, Annali dell’Istituto Universitario Orientale di Napoli, Sezione linguistica 8, pp.131-146.

P. Chantraine, 1968. Dictionnaire étymologique de la langue grecque, Paris.

S. Colvin, 2007. A Historical Greek Reader: Mycenaean to the Koine, Oxford.

G.E. Dunkel, 2014. Lexikon der indogermanischen Partikeln und Pronominalstämme, Heidelberg.

T. Figueira, 2011. The Power of Money: Coinage and Politics in the Athenian Empire, The University of Pennsylvania Press.

M. Finkelberg, 2005. Greeks and Pre-Greeks: Aegean Prehistory and Greek Heroic Tradition, Cambridge.

J.L. García-Ramòn, 1975. Les origines postmycéniennes du groupe dialectal éolien, Salamanca.

A.J. Heisserer, 1984. The Monetary Pact between Mytilene and Phokaia, Zeitschrift für Papyrologie und Ephigrafik 55, pp. 115-132.

R. Hodot, 1990. Le dialecte éolien d’Asie: La langue des inscriptions, VII e s. a.C.–IV e s. p.C., Recherche sur les civilisations, Paris.

J.T. Hooker, 1977. The Language and Text of the Lesbian Poets, Institut für Sprachwissenschaft der Universität Innsbruck, Innsbruck.

G.C. Horrocks, 2010. Greek: A History of the Language and its Speakers, Wiley Blackwell, second edition.

D.G. Miller, 2013. Ancient Greek Dialects and Early Authors, De Gruyter.

H. Parker, 2008. The Linguistic Case for the Aiolian Migration Reconsidered, The Journal of the American School of Classical Studies at Athens, Volume 77, pp. 431-464.

W.B.Stanford, 1958. Aristophanes: Frogs, Bristol Classical Press.

A.L. Sihler, 1995. New Comparative Grammar of Greek and Latin, Oxford.

O. Tribulato, 2008. La lirica monodica, in A.C. Cassio (ed.), Storia delle lingue letterarie greche, Le Monnier Università, pp.145-173.

[1] I wish to acknowledge my considerable debt to my schoolteachers, James Burbidge and Katy Waterfield, as well as to Stephen Colvin, John Penney, Don Ringe, Matt Scarborough, John Taylor, Olga Tribulato and Jo Willmott for providing me with reading lists and for their indispensable suggestions. Without their generous help I would not have been able to complete my Extended Project.

[2] It is impossible to reconstruct the ‘everyday’ speech of the ancient Mytileneans only through the written evidence. The ‘dialect of Mytilene’ in the project was the language of inscriptions from Mytilene in IG XII 2 and IG XII Supplementum, both of which are used in Hodot 1990 and are, where possible, dated (whether precisely or approximately). I concentrated on Mytilene because the city is where we believe both poets were active. Where it is appropriate and there is no comparable form in IG XII 2 and IG XII Supplementum, inscriptions from other Lesbian-speaking regions will be compared with the dialect of Sappho and Alcaeus. Where it is sensible, statistics from the whole of Lesbos and Asian Aeolis in Hodot 1990 will be used. All the numerations of the poets’ fragments are those of Lobel-Page from the second edition of Greek Lyric I, Loeb Classical Library. Apart from one exception (see line 18 of the linguistic commentary), I have decided not to discuss any other form of direct, written documentation such as the curse tablets from Mytilene discovered in 1998, mainly because of the need to discuss the ‘curser’s bi-dialectism’ (p.194) and ‘vulgar’ Greek in a depth that is beyond the scope of this article. For more on this topic, see A Cypriot Curser at Mytilene, Zeitschrift für Papyrologie und Epigraphik 179 (2011), pp.189-198, by A.Dale and A.Ellis-Evans.

[3] See Hooker 1977, p.18, Bowie, 1981, p.138, Colvin, 2007, p.218 and Blümel, 1982, §65 and §128.

[4] An example of this may be the metrical corruption and partial Atticization in Artistophanes’ Wasps 1234-1235, a parody of Alcaeus 141. Attic can also become corrupt because of non-Attic scribes; see for example Aristophanes Frogs 454-9, imitated in an inscription from Rhodes (1st century B.C.); ἡμῖν in line 454 is replaced by ἡμεῖν (cf. Stanford, 1958, pp.112-113). See also Hooker 1977 (p.31) to appreciate how Koine can influence the quoted lines of Sappho and Alcaeus in the manuscripts of grammarians.

[5] For an example, see Hooker 1977, p. 47.

[6] Cf. Hooker 1977, pp.39-43, p.48ff. and Bowie 1981, pp.60-69.

[7] There has been much debate as to when this monetary agreement took place. Newton (quoted in Heisserer 1984, p.119) believes that the inscription ‘is not later than the 96th Olympian’ (396-392 B.C). Heisserer 1984 that this public decree is from ‘the second half of the fifth century B.C., more specifically ca. 440 to 410 B.C.’(p.122). He comes to this conclusion through his observation of the letterforms that are similar to other 5th century Aeolic inscriptions (pp.119-120). Hodot 1990 takes Heisserer’s paper into account and dates it to 426B.C. Figueira 2011 (p.488), however, seems to prefer to date it to the early fourth century, citing the study of the inscription’s letter forms by other scholars who oppose Heisserer’s observation. What is certain from all the related publications is that IG XII 2.1 is ascribed to sometime in the mid-Classical period. In the linguistic commentary below, references to the poetry of Sappho and Alcaeus have been included where possible.  Before translating the inscription, I read the following: Buck 1955, Colvin 2007 and Miller 2013.

[8] Other reasons may include the nature of Lesbian monody that is often occasion-specific and personal.

[9] See also §2.3.

[10] I owe this point to my tutor.

[11] Buck 1955, §109.3, supported by Blümel 1981, §266.

[12] Note that ου in λύουσι is a spurious diphthong (cf. Buck 1955, §25).

[13] In Attic, since ō cannot be lengthened any further in -ōnsi, simply became –ωσι.

[14] Bowie 1981, pp.119-120 and p.122.

[15] Examples include Locrian, Delphian, Elean. The argument put forward by García-Ramòn 1975 (p.84), who argues that the Aeolic migration from the mainland took place when both forms of the dative plural were used, and the borrowing of –εσσι by the neighbouring non-Aeolic dialects (cf. Finkelberg 2005, p.129) on the mainland, may question the theory put forward by Parker 2008. Parker in turn argues that there was no Aeolic migration from mainland Greece (the author also rejects the existence of an Aeolic dialect group).

[16] However, as my tutor pointed out, -εσσι may have in fact originated from s-stem nouns through the analogical proportion model δοῦλοι : δούλοισι, πάντες : πάντεσσι (see also Sihler 1995, §276.6.a). To go into deeper analysis is beyond the scope of this paper, but there is no doubt that the ending –εσσι is very problematic.

[17] See also Hooker 1977, pp.30-31, Blümel, 1981, §86.

[18] Miller 2013, p.251.

[19] Cf. Miller 2013, p.251and Blümel 1981 §85. Hodot 1990 (p.71) suggests that it is ‘une graphie inverse’.

[20] Though ἦσι may be an epic borrowing.

[21] Blümel, 1981, §85.

[22] Miller, 2013, p.251.

[23] Horrocks, 2010, p.25.

[24] In other dialects, the end result of simplification is the lengthening of the preceding vowel. This is the observation that Parker 2008 develops in his attempt to reject Boeotian from Aeolic and to go on to question the coherence of ‘Thessalo-Lesbian’. But see Finkelberg 2005, pp.109-139, for a persuasive account of the dialect geography of pre-historic Greece and the dialect continuum formed by Thessalian, Boeotian and Lesbian. Additionally, as my tutor pointed out, there is difficulty in knowing whether spelling represents the phonological reality.

[25] But see Cassio 1986, p.138, where he quotes προαγρημμένω, found at Cyme in an inscription from the 3rd century B.C. As Colvin 2007, p.219, notes, ‘it is odd that it [the double nasal] occurs only after η’. A further speculation for the explanation of the potentially hyper-Aeolic double nasal could be that in archaic Lesbian the eta was pronounced longer than (say) in Attic, and that this was represented by the double nasal.

[26] Buck 1955, §162.2.

[27] Miller 2013, p.251.

[28] Colvin 2007, p.109. Before Heisserer 1984, it was considered to be an athematic present infinitive of κέρναμι (Buck 1995, §155.3).

[29] Colvin 2007, p.109.

[30] Hodot 1990, p.148: ‘la forme ἀπό ne figure pas dans les fragments des Lyriques’. See footnote 55.

[31] Cf. Hodot 1990, p.148: ‘on s’accorde généralement à reconnaître que ἀπύ et ἀπό precèdent de deux variantes d’âge indo-européen…les deux variants aient toujours coexisté en Lesbien…dans les conditions que les maigres données dont nous disposons pour les Ve-IVe s. ne laissent pas préciser’. 

[32] Some examples are listed in Buck 1955, §22, and in greater depth in Blümel 1981, §47. This sound change is not regular either in inscriptions or in poetry, as Blümel 1981 (§47) writes: ‘auf den Inschriften und in der literarischen Überlieferung des Lesbischen schwankt die Schreibung für [o] vor [m] am Wortanfang zwischen [ο] und [υ].’ In general, older Lesbian inscriptions have [υ], and Hellenistic inscriptions have [ο]; this can be explained by the increasing influence of Attic-Ionic and Koiné. See also §2.2.

[33] This statement is based on the fact that in Greek *h2 does not have an aspirating effect.

[34] Cf. Chantraine 1968.

[35] Hodot, 1990, p.144.

[36] For example, Horrocks 2010, p.52, Buck 1955, §135.3. In Miller 2013, p.257, it is simply described as ‘poetic’. Blümel 1981 and Tribulato in Cassio (ed.) 2008, which is the most up-to-date survey of the dialect of Sappho and Alcaeus in relation to the epigraphic dialect, make no mention of this preposition. Colvin 2007 (p.218) does not include any epigraphic evidence from Lesbos/Lesbian-speaking area.

[37] Hooker 1977, pp.25-26.

[38] Bowie 1981, p.86.

[39] Cf. Sihler 1995, §406.8. Greek and Italic evidence suggest *supo, but this is not supported by other languages.

[40] See (§2, 2).

[41] Buck 1955, §135.3.

[42] Bechtel 1921, p.119.

[43] Dunkel 2014, Vol. 1, p.99.

[44] (h)upu is recorded only once in IG 14.871 from Cumae, dated to the 5th century B.C.

[45] I owe this point to my tutor.

[46] Cf. Horrocks 2010, p.77, and Colvin 2007, p.108.

[47] Although it is nowadays agreed that Lesbian was psilotic, some have denied it or explained the absence of initial aspirates by orthographic reasons. Cf. Hooker 1977, pp.16-17.

[48] See Hodot 1990, p.139. 

[49] See also §2.4.

[50] Bubenik 1989, p.68.

[51] Hooker 1977 (p.11) raises the possibility that the papyri had ‘a purer doctrine about the dialect of Sappho and Alcaeus’. But all the papyri fragments of the poets are also much later than the Archaic Period, therefore they may be corrupt as well. At any rate, the probability of corruption in papyri fragments is possibly lower, as they have avoided further transmission in manuscripts.  

[52] Hodot 1990, p.149. The inscription is IG XII 2.70. This inscription is a ‘fragment à des cultes’. 

[53] Some are found with the accusative, and as a prefix only rarely.

[54] This combination would mean that the two forms co-existed in the time of the two poets, but not by the time of our epigraphic evidence.

[55] It appears as though prepositions are highly susceptible to the influence of Koine, as Hodot would certainly agree in the case of ἀπύ, cf. Hodot 1990, p.148: ‘il est indéniable que c’est l’influence de la koinè qui explique la totale disparition de ἀπύ aux IIIe-IIe s.’ Even so, ἀπύ begins to reappear in the 1st century A.D., unlike ὐπά. It is highly intriguing that, supposing that the dialect of Sappho and Alcaeus did have a profound influence on the dialect of later inscriptions (as Hodot supposes, still in p.148: ‘c’est l’influence des Lyriques qui réintroduisit ἀπύ sous l’Empire’), ὐπά was never reintroduced but once. However, Hodot is often thought to be exaggerating the influence of the literary dialect on the epigraphic dialect: see Cassio 1986 (pp.139-140) for a critique of Hodot’s assumption that the poets’ dialect affected the language of inscriptions before the Imperial Period. The extreme rarity of ὐπά in inscriptions possibly indicates that Sappho and Alcaeus were not the absolute dialectal models in the Imperial Period, and hence not every dialectal feature of their poetry was used.

[56] Cassio 1986, pp.141-3.

[57] I would like to thank my tutor Evert van Emde Boas and the editors of Persephone for their kind help on this article. I am fully responsible for any factual mistakes and others forms of error.