ὀσμή, ὀδμή, ὄζω, ὄδωδα
(Phryn. Ecl. 62, Phryn. PS 97.21–2, Antiatt. ο 13, Poll. 2.75, [Hdn.] Philet. 304)

D. General commentary

Almost all the principal Atticist lexica (A.1, A.2, A.3, A.4, A.5), with the notable exception of Moeris, discuss the variation σ/δ in several nominal derivativesDerivatives from the root of ὄζω ‘to smell (intransitive)’, in particular the abstract noun ὀσμή/ὀδμή ‘smell’ and the second compound member -οσμος/-οδμος ‘-smelling’. Some lexica (A.3, A.4, B.1) address, alongside this issue, the apparently unrelated question of the correct tense for expressing the notion of ‘smelling’: the present ὄζειν or the perfect (with present meaning) ὀδωδέναι. In what follows, we focus primarily on the first problem, since it raises issues of dialect and register, as well as the role of euphony in Atticist prescriptions. Nevertheless, as will be argued below, the two issues may in fact be connected.

The perfect ὄδωδα (with ‘Attic reduplication’) and the present ὄζω (< *od-i̯e/o-) derive from the PIE root *h₃ed- ‘to smell’ (LIV 296). The corresponding abstract noun shows the variants ὀδμή < *od-mā (cf. Gheg Albanian amë ‘unpleasant smell’, on which see Demiraj, Neri 2024) and ὀσμή, whose relationship to the former has long been debated. According to one view, ὀσμή is the outcome of ὀδμή, reflecting a tendency (especially strong in Attic: see Threatte 1980, 657–8) to replace δ with σ before μ. Despite several parallels (e.g. Attic ἴσμεν ‘we know’, cf. Ion. ἴδμεν < *ϝιδ-μεν), it remains unclear whether this was a regular sound change, in view of counterexamples with preserved δ (e.g. μεσόδμη ‘tie-beam’), or rather an analogical development (see Lejeune 1972, 76–7 and nn. 5–6; Höfler 2017, 6–10 argues for a regular change *dm > sm word-internally as well as word-initially, partly obscured by analogical restitution of δ). Alternatively, ὀσμή may be derived from a different preform, *od-smā (Lejeune 1972, 75; DELG s.v. ὄζω) or *od-s-mā (i.e. a derivative of the s-stem noun *h₃ed-os-, reflected in Latin odor ‘smell’< odōs and possibly in Armenian hot; see EDG s.v. ὄζω).

Regardless of the precise linguistic relationship between the two forms, their distribution in classical sources – allowing for textual alterations, whether intentional or unintentional, in the MSS and in modern editions – largely follows dialectal lines. Only ὀδμή is attested in the Homeric language (Il. 14.415 = C.1, Od. 5x; also 5x in the Homeric Hymns), and, consequently, it also prevails in later epic poetry (Apoll.Rh. 7x; Nic. 8x; Opp. 8x, vs. ὀσμή 1x; Nonn. 12x). Forms with -δ- are also the norm in lyric poetry, including Theognides (1.9), Simonides (fr. 133 PMG [= fr. 323 Poltera]), and Xenophanes (2x), while Pindar (fr. 129.8 Snell–Maehler) has ὀδμά with non-Ionic vocalism; in Hipponax (C.2) the tradition is divided between ὀδμήν and ὀσμήν (see F.2). Ionic prose also shows a marked preference for ὀδμή, which is the norm in Herodotus (9x) and the Hippocratic corpus (44x vs. 9x ὀσμή); cf. also Democr. Diels–Kranz 68 B 11.8. Overall, then, the distribution outlined above justifies the ‘Ionic’ label assigned to ὀδμή by several erudite sources (A.1, A.2, A.5, B.2, B.6; see Valente 2015, 224 for further lexicographical parallels). The classification of ὀδμή as an Aeolic form, found elsewhere in erudite sources (B.5), is possibly based on its Homeric occurrences (cf. [Plu.] Vit.Hom.2 98, who includes the use of ὀδμή and ἴδμεν among allegedly Aeolic features in Homer), despite the fact that ὀδμή is not attested in the Lesbian poets (see Sowa 2024, 108). Another factor in this classification may have been the presence of ὀδμά in Pindar, an author in whom ancient scholarship recognised a mixture of Aeolic and Doric traits. Indeed, a different tradition, reflected in Eustathius (B.3, B.4), explained ὀδμή as a Doric form.

Classical Attic authors, by contrast, predominantly employ ὀσμή, both in drama (Aesch. 2x, including C.3; Soph. 5x; Eur. 5x; Ar. 5x) and prose (Thuc. 7.87.2 = C.5; X. 17x; Pl. 15x). Nevertheless, the stem ὀδμ- is not entirely absent from Attic literature. The non-Attic form ὀδμά occurs in [Aesch.] PV 115 (C.4), a lyric passage where it is readily explained as a Doricism (cf. its use in Pindar). Comedy presents a more complex picture. Remarkably, in a single passage of Hermippus (fr. 77 = C.7), preserved by Athenaeus, both ὀδμή (l. 3) and ὀσμή (l. 9) are transmitted. Comentale (2017, 311–2), following Dobree (1833 vol. 2, 295), suggests that in l. 9 ὀσμὴ θεσπεσία should be corrected to Ionic ὀδμὴ θεσπεσίη, since the line echoes Od. 9.210–1. The text of other comic fragments preserved in the secondary tradition shows similar oscillations: cf. Anaxandrid. fr. 42.36 (ὀδμαί codd. AE of Athenaeus : ὀσμαί cod. C; see Millis 2015, 208); Antiphanes fr. 159.10 (ὀσμῆς codd. CE : ὀδμῆς cod. A); Alexis fr. 195.3 (ὀσμάς all codd. of Athenaeus : ὀδμάς Clem.Al. Paed. 2.2.68.2). However, these are best regarded as interferences from later Greek, since ὀσμ- is otherwise the norm in Middle Comedy (see Favi 2024, 154 n. 11).

Whereas the forms with -δ- where perhaps admitted to some extent in Attic drama, Pollux (A.4) reports that Antiphon (C.6) was the only Attic prose author to employ forms in ὀδμ-. The Atticist lexicographers were seemingly unaware on the ancient debate on whether the orator Antiphon of Rhamnous and Antiphon ‘the sophist’ were the same person (for the modern debate, see Gagarin 2002, 37–52 for the ‘unitarian’ position, and Pendrick 2002, 1–22 for the ‘separatist’ one; see further entry καταριστάω). It is therefore impossible to determine which author and work Pollux had in mind: while Thalheim included Pollux’s quotation among the orator’s fragments, Pendrick (2002, 255) tentatively suggests that the fragment may derive from a discussion of sensory perception in the sophist Antiphon’s On Truth. On the other, hand, despite Phrynichus’ claim in A.1 (repeated by Thomas Magister = B.6), ὀδμή is not transmitted in Xenophon’s text. As already noted, the MSS of Xenophon consistently have ὀσμή (17x; in Smp. 2.3 ὀδμή is transmitted by cod. F [Vindob. phil. gr. 109] alone; on Xenophon’s use of ὀδμή see also AGP vol. 1, 209–10 with n. 236). Several scholars, including Lobeck (1820, 89) and Rutherford (1881, 160–4), have thus argued that Atticising scribes systematically replaced ὀδμή with ὀσμή in Xenophon’s text. Other scholars have suspected a textual corruption: Aly (1929, 155 n. 157) suggested that Ξενοφῶν is a mistake for Ἀντιφῶν, whereas Bossi (1980–1982) argued that Phrynichus originally mentioned Xenophanes to exemplify the Ionic use of ὀδμή, and that the text underwent a two-stage corruption: first Ξενοφάνης was replaced by Ξενοφῶν, then the latter in turn replaced Ἀντιφῶν in the following sentence. However, such conjectures seem unnecessary: as Favi (2024, 153–8) has recently argued, it is entirely possible that Xenophon employed ὀδμή (albeit not exclusively), as a more ‘international’ form, much as he used both θάρσος and θάρρος ‘courage’, and that the unanimous reading ὀσμή in our MSS is the result of later levelling. Favi further suggests that Antiphon may likewise have employed the forms with -δ- in order to avoid an Attic localism.

What is certain is that ὀδμή gained ground in Attic and koine prose from the 4th century BCE onwards, beginning with the works of Theophrastus (11x), where, however, ὀσμή still largely predominates (over 200x). It should be emphasised that in the text of major koine authors ὀδμή is never the only variant and is in fact often in the minority: cf. D.H. ὀδμή 1x (13.11.1) vs. ὀσμή 3x; Ph. ὀδμή 1x (De sacrificiis Abelis et Caini 44.2) vs. ὀσμή 23x; Str. ὀδμή 6x vs. ὀσμή 3x; Ios. ὀδμή 2x vs. ὀσμή 3x; Plu. ὀδμή 5x vs. ὀσμή 41x. In the Septuagint and in the New Testament, ὀσμή is actually the only form attested. The survival of ὀσμή in Modern Greek dialects (see E.) likewise suggests that ὀδμή never fully replaced it.

The distribution of the compounds broadly follows that of the simplicia: thus, εὔοδμος ‘sweet-smelling’ has a markedly non-Attic distribution (Pi. 2x; Hp. 2x; Theoc. 17.29; Call. fr. 43.13; Euph. fr. 451 col. 1.10, etc.), whereas εὔοσμος and its abstract εὐοσμία occur in Attic drama (εὔοσμος in Eur. Ba. 235; Achae. fr. 17.3; Stratt. fr. 71.7; εὐοσμία in Soph. fr. 370.2). Again as in the case of the simplex, Theophrastus is the first Attic author to adopt the forms in -δ- (εὔοδμος 8x, εὐοδμία fr. 4.51), though much less frequently than those in -σ- (εὔοσμος 96x, εὐοσμία 36x). The compound κάκοσμος, mentioned in the Philetaerus (A.5), has a few occurrences in Attic drama (Soph. fr. 565.1; Ar. 2x), whereas κάκοδμος is frequent in Hippocratic writings (27x).

The distribution of the forms is sufficient to explain the almost unanimous rejection of ὀδμ- by the Atticists. Pollux’s passage (A.4) is particularly interesting in this context: the lexicographer aligns with most other Atticists in expressing a preference for ὀσμή and, as already noted, observes that ὀδμή is foreign to Attic prose, with the exception of Antiphon. Pollux’s remark that many people found the forms in ὀδμ- ‘beautiful’ may point not only to their association with poetic language, but more specifically to the widespread aversion to the sound of sigma in Greek euphonic theory (see Stanford 1967, 53–4; Clayman 1987). The Atticists, too, paid attention to matters of pronunciation and euphony, including the avoidance of /s/ (see Vessella 2018, 1–12; AGP vol. 2, Phonology, forthcoming). Indeed, the fact that Attic happened to have fewer instances of /s/ than other dialects – owing to the reflex of *(-)Ki̯- as (-τ)τ- rather than (-σ)σ- and to the assimilation -ρσ- > -ρρ- – may have contributed to its perception as an especially agreeable-sounding variety. Nevertheless, the stricter Atticist did not usually allow euphonic considerations override other, more important criteria, such as attestation in the canonical authors (see entry βασμός, βαθμός, ἀναβασμός, ἀναβαθμός for a comparable case). As to the target of Pollux’s criticism, it is likely that by οἱ πολλοίοἱ πολλοί he refers here to practitioners of post-classical prose (Matthaios 2013, 102–3). Despite his own prescription, elsewhere Pollux (1.239Poll. 1.239) includes εὔοδμος among terms of praise for a meadow.

Other lexicographers, by contrast, do not comment on matters of register or literary genre, but insist on the dialectal affiliation of ὀδμή. In the Eclogue (A.1), Phrynichus, as already noted, criticises Xenophon for having transgressed against his native dialect by using an Ionic word. The lexicographer’s use of the verb παρανομέωπαρανομέω ‘to act against the law’, paralleled in Ecl. 86Phryn. Ecl. 86 (ἀτταγήν· καὶ τοῦτο παρανενόμηται τόνῳ καὶ θέσει· χρὴ γὰρ ἀτταγᾶς λέγειν ὥσπερ ἀλλᾶς, ‘ἀτταγήν (‘francolin’): This form, too, violates the norm in its accent and form; for one ought to say ἀτταγᾶς, like ἀλλᾶς (‘sausage’)’), is a notable example of the Atticists’ recourse to legalistic terminology, based on a puristic ideology that equated linguistic correctness with civic virtue (see AGP vol. 1, 79–83). Phrynichus, who repeats the same prescription in the Praeparatio sophistica (A.2), consistently prescribes expressions containing ὀσμή elsewhere in his lexica: cf. PS 91.3Phryn. PS 91.3 on νῆστις ὀσμή (lit. ‘starving smell’, i.e. bad breath resulting from prolonged fasting: see entry νῆστις ὀσμή) and Ecl. 126Phryn. Ecl. 126 on ἄχαρις ὀσμή ‘unpleasant smell’ (recommended as a more correct alternative to βρῶμος).

While other lexica, such as the Philetaerus (A.5), Photius (B.2), and Thomas Magister (B.6), agree with Phrynichus’ prescription, the Antiatticist (A.3), as usual, takes a more inclusive stance, reporting that some authorities promoted ὀδμή over ὀσμή and ὄδωδα over ὄζω. To understand why the Antiatticist treats two apparently separate issues in the same entry, we should consider the distribution of the two competing verbal forms and their discussion in other erudite sources. In the classical period, the present ὄζω, albeit relatively rare in prose (see Veitch 1887, 475), was normally employed (e.g. C.7); the use of the perfect ὄδωδα with a present meaning has precedents in Homer (the pluperfect ὀδώδει occurs 2x with present meaning in the Odyssey), but in prose it is typical of the koine (e.g. Plu. 11x). It is likely that strict Atticists rejected ὄδωδα on these grounds, although no explicit proscriptions survive; apart from the Antiatticist (whose own position on the matter is not entirely clear: see F.1), the extant lexicographical sources (A.4, B.1; see further Valente 2015, 224) also adopt a descriptive and tolerant approach, simply recording that both tenses were in use. It is possible – albeit difficult to prove with certainty – that the rejection of ὄδωδα was also connected with that of ὀδμή, since both forms share the consonant δ. Indeed, Eustathius (probably relying on Herodian: see Van der Valk 1971–1987 vol. 1, 235, and cf. Hdn. Περὶ παθῶν GG 3,2.372.9–15) claims in B.3 and B.4 that ὀδμή was derived from a fictional preform ὀζμή (with the ζ of ὄζω) through an allegedly Doric change ζ > δ, whereas ὀσμή was derived from the same preform through a ‘later’ change ζ > σ. Elsewhere, Eustathius (in Il. 1.235.15; in Od. 2.18.24–20.3 Cullhed–Olson = 1.199.29–31 Stallbaum) invokes the same mechanism to derive ὄδωδα from a fictional ὄζωζα. One may therefore suggest that the consonant δ in this lexical family was perceived as a marker of Doric origin, providing stricter Atticists with an additional reason to reject such forms. Conversely, ὀδμή and ὄδωδα could be regarded by others as mutually reinforcing. On balance, although the usual criterion of reliance on canonical authors suffices to explain the Atticists’ prescriptions, it is quite possible that in this case other considerations – including differing phonetic preferences and the perceived association of ὀδμ- with non-Attic dialects – played a supporting role.

The Atticising authors, as a rule, respected the classical Attic usage (see Schmid, Atticismus vol. 1, 207, 340; vol. 3, 143; vol. 4, 209, 358): only ὀσμ- is attested in Dio Chrysostom (5x ὀσμή + 1x εὔοσμος), Aelian (27x ὀσμή + 1x εὔοσμος), and Philostratus (2x ὀσμή + 1x εὔοσμος). Lucian constitutes a partial exception, since in his genuine works ὀσμή and ὀδμή each occur three times, although the former has the varia lectio ὀδμ- in two cases (both from the True Histories), and one occurrence of the latter is in Syr.De. 30, where it is justified as an imitation of literary Ionic. As far as the choice of tense is concerned, the post-classical use of ὄδωδα is not uncommon (Ach.Tat. 2x; Max.Tyr. 2x; Ael. NA 13.10; Philostr. Her. 3.4; Alciphr. Ep. 3.23.4), but it is avoided by a rigorous Atticiser such as Aelius Aristides, who employs only the present (3x), while other authors (Dio Chrysostom, Lucian) do not attest any form of this verb.

E. Byzantine and Modern Greek commentary

Both ὀσμή and ὀδμή are abundantly attested in medieval sources; however, while the latter is attested in the hundreds, the former occurs in the thousands, including in lower-register texts (see Kriaras, LME s.v.). In Modern Greek, the learned οσμή ‘smell’ coexists with the popular synonym μυρωδιά < medieval μυρωδία, derived, via the late adjective μυρώδης ‘smelling like unguent’, from the classical μύρον ‘unguent, perfume’ (LKN s.vv.). Another derivative of the same noun, μυρίζω, classically ‘to anoint with unguent’, developed the meaning ‘to smell’ (both transitive and intransitive) in Medieval and Modern Greek. Several modern dialects, however, preserve reflexes of ὀσμή, both in the classical sense ‘(sense of) smell’ and in the innovative sense ‘scent track’ (Andriotis 1974 s.v.). This fact, together with the distribution observed in Byzantine sources, suggests that ὀδμή, despite its relative frequency in the koine, did not ultimately succeed in displacing ὀσμή from the spoken language.

F. Commentary on individual texts and occurrences

(1)    Antiatt. ο 13 (A.3)

The formulation of the entry is perplexing, since in the Antiatticist the formula ‘οὔ φασι δεῖν λέγειν…, ἀλλά…’ usually ‘refers to a reaction of the author against a prescriptive rule’, where ‘the subject of φασί is to be identified in other lexicographers and/or Atticists having a more rigorous idea of the literary language’ (Valente 2015, 45–6). Here, however, the doctrine reported is at odds with that embraced by Phrynichus and other Atticists (see D.), thus giving the impression that the Antiatticist is in fact aligned with the Atticist mainstream on this point. Cobet (1877)    therefore proposed to emend the text of the entry to ὄζε<ι>ν· [οὔ] φασὶ δεῖν λέγειν, ἀλλ’ <οὐκ> ὀδωδέναι, ‘they say that one should say ὄζειν, but not ὀδωδέναι’. Nevertheless, since several more or less Atticising authors did in fact employ ὄδωδα, it cannot be excluded that this form was recommended by some contemporary lexicographers, who were in turn criticised by the Antiatticist.

(2)    Hippon. fr. 92.10–1 West² (= fr. 95.10–1 Degani) (C.2)

Tzetzes (Ex. Α 97–609, no. 21) quotes l. 11 with ὀδμήν, the expected Ionic form, whereas PSI 9.1089.11 (= TM 60198) [Oxyrhynchus, 2nd c. CE] reads ὀσμήν. The latter form is not impossible in Hipponax, since, according to Herodian’s Περὶ καθολικῆς προσῳδίας (cod. Vindob. hist. gr. 10 = p. 5 Hunger [1967]), the Ephesian poet used both Καδμῖλος (fr. dub. 197 Degani = iamb. adesp. 58 West²) and Κασμῖλος (fr. 155b West² = fr. 164 Degani); however, the alternation -δμ-/-σμ- in this name may also be due to its likely non-Greek origin (see Beekes 2004, 465–8). In contrast to earlier editors, who preferred the papyrus reading, both Degani (1991, 106) and West (1998 vol. 1, 138) print ὀδμήν, in light of the occurrences in Ionic authors (see D.) and of Phrynichus’ prescriptions (A.1, A.2). On the other hand, the papyrus’ ὀσμήν has been explained as an Atticistic correction (Hawkins 2013, 65–6).