ἄρκτος, ἀπαρκτίας
(Phryn. PS 31.18, [Hdn.] Philet. 314, Σ^b α 2127, Phot. α 2265)

D. General commentary

Atticist lexica discussed the noun ἄρκτος ‘bear; Ursa major, north’ and its derivative ἀπαρκτίας ‘north wind’, and their variants ἄρκος and ἀπαρκίαςἀπαρκίας, which show simplification of the consonantal cluster ‑κτ‑ to ‑κ‑. While Phrynichus (A.1) – on whose doctrine Photius (A.4, B.3) depends, probably through intermediate sources – and the Philetaerus (A.2) agree in rejecting the forms without τ, a fragment attributed to Aelius Dionysius’ lexicon (A.3) accepts ἀπαρκίας but not ἄρκος. These entries are notable for their engagement with forms that likely reflect an actual phonetic phenomenon in contemporary spoken Greek whose reflexes have survived in several varieties of Modern Greek (see E.).

Greek inherited the noun ὁ/ἡ ἄρκτος from the widely attested Indo-European lexeme *h₂ŕ̥tḱo- ‘bear’ (NIL 343–5; Blažek 2017), which also gave Latin ursus, Vedic ŕ̥kṣa-, and Hittite ḫartakka-: the latter indicates that the original consonantal sequence must have been *-tḱ- (i.e., one of the so-called ‘thorn clusters’, on which see Schindler 1977; Kloekhorst 2014; Jasanoff 2018), which Greek regularly metathesised to -κτ-. This picture, however, is complicated by attestations throughout the Greek-speaking world of forms in -κ- (a representative selection of which is offered in C.): -κος, -κ(ε)ιος, -κύλος. Although forms without τ are largely extraneous to the classical literary language, they are not as late as modern reference works often portray them to be. As Dobias-Lalou (2000, 61) pointed out, ἄρκος is common in Cyrenaean inscriptions from the 4th century BCE, and the same stem is found in personal names from other regions, the oldest of which date to the 6th century (see Masson 1990–2000 vol. 2, 617–20; vol. 3, 147; 320; Minon 2023, 160–1). The compoundsCompounds of the type Ἀρκολέων ‘[Having the strength of] bear and lion’, Ἀρκόλυκος ‘[Having the strength of] bear and wolf’ are notable. It is true, however, that literary attestations of ἄρκος start from the Septuagint (18x, e.g. C.8), with the exception of a Hippocratic passage (C.7) in which the word refers to a crustacean (called ἄρκτος elsewhere, cf. Arist. HA 5649b.23), and that they are restricted to less refined koine prose (Ios. AI 6.183, NT Apoc. 13.2, Dsc. 2x, Gal. 5x), sometimes appearing as variae lectiones (see C.11). In later Greek, ἄρκος occasionally appears in papyri (e.g., P.Mich. 1.86.5–6 [= TM 1985], Arsinoites, 263–229 BCE) and inscriptions (cf. e.g. the faulty spelling ΔΡΚΟϹ for ἄρκος in IG 14.1302, the famous Nile mosaic of Palestrina [2nd-1st century BCE], on which see Meyboom 1995, 26 and 279 n. 59). The stem variant ἀρκ- is also attested in several derivatives, including the diminutives ἄρκυλλοςἄρκυλλος (schol. Opp. H. 2.248) and ἄρκιλοςἄρκιλος (B.4), both ‘bear cub’, the adjective ἄρκ(ε)ιοςἄρκειος ‘of a bear’ (Dsc. 1.125, D.Chr. 7.43), ‘northern’ (C.3), and, as already mentioned, the wind-name ἀπαρκίας (attested since the 1st century CE, cf. Arr. Peripl.M.Eux. 4.2 = C.9, 21.1; Peripl.M.Rubri 2.83). The latter also occurs in inscriptions (I.Erythrai Klazomenai 304.4, 1st century BCE; IG 14.1308) and in late LatinLatin as aparcias (Gloss. 4.19) / apartias (Gloss. 5.440) (literary sources employ aparctias, cf. Plin. Nat. 2.119, Apul. Mun. 11, etc.; but cf. Veg. Mil. 4.38.12 aparcias), suggesting that the variant without τ was common in spoken Greek of the imperial age. It should be noted that most of the above forms have variants with -τ-: ἀρκτύλοςἀρκτύλος ‘bear cub’ (Poll. 5.15Poll. 5.15, where cod. A has ἀρκύλοι; schol. Ael. NA 1.16), ἄρκτειοςἄρκτειος ‘of a bear’ (Heras ap. Gal. De comp. med. sec. loc. 12.399.5 Kühn, [Gal.] De succed. 19.743.12 Kühn), ἄρκτιοςἄρκτιος ‘northern’ (Hp. ap. Gal. Voc. Hipp. Gloss. α 153 Perilli [19.85.15 Kühn], Nonn. D. 38.329), while the better- and earlier-attested name of the north wind is ἀπαρκτίας (first attested in Strattis [C.6] and later in Arist. Meteor. 363b.14, Thphr. Sign. 2.33, etc.; see Orth 2009, 279).

To evaluate the distribution of the forms in ἀρκ-, a brief excursus on their possible origin and the nature of the change -κτ- > -κ- in this word family is necessary. As a purely phonetic change, it lacks exact parallels in Greek, although, certainly, no other lexeme with the synchronically unanalysable sequence -ρκτ- (as opposed to cases such as, e.g., ἄ-δερκ-τος ‘not seeing’, when the transparent relation to δέρκομαι would have arguably acted to preserve the cluster) is known. It is possible, then, that in at least some registers the heavy cluster -ρκτ- tended to become -ρκ- by a late simplificationSimplification (as maintained, e.g., by Lejeune 1972, 39 n. 12; for the treatment -ρκτ- > -ρτ- in inarticulate speech, see B.4 and F.1; on simplifications of other three-consonantal clusters in Attic see Threatte 1980, 571–5). A folk-etymological influence from ἀρκέωἀρκέω ‘ward off, defend’ is sometimes suspected (see DELG s.v.; the etymology of ἄρκτος from ἀρκέω, via ἄρκος, is found in Byzantine scholarship, cf. Et.Gen. AB α 1194 = B.2, schol. Opp. H. 1.12, 2.248). Some scholars, by contrast, have posited that the differentiation between ἀρκτ- and ἀρκ- is considerably older, ultimately deriving from different Proto-Greek (or PIE?) treatments of the cluster *-tḱ- (see Dobias-Lalou 2000, 61; EDG s.v. ἄρκτος). Again, this has no clear parallels, at least in word-internal position; in initial position, meanwhile, clusters of the shape *TK- (i.e., any dental stop followed by any velar stop) were simplified to *K- before a syllabic resonant, as proved by καίνω ‘kill’, if from the same root *tken- as κτείνω ‘kill’ (see entry ἀπέκτονα, ἀπέκταγκα), and by the pair χθών ‘earth’ : χαμαί ‘on the ground’ (a different explanation is provided by Kloekhorst 2014, 48–9), but *ártko- did not provide the relevant phonological environment anyway. Schindler (1977, 33–4), working under the assumption that the metathesisMetathesis *TK > *KT only applied when the cluster was tautosyllabic (but see Jasanoff 2018, 134), postulated two different ablaut grades with different syllabifications in the PIE word for ‘bear’ to account for the different reflexes in other daughter languages (zero-grade *h₂ŕ̥t.ḱo- > Hitt. ḫartakka-, full-grade *h₂er.tḱo- > Old Irish art; Latin ursus possible requires an o-grade *h₂or.tḱo-). Accepting this reconstruction, we may hypothesise that while *h₂er.tḱo- developed into ἄρκτος with metathesis, *h₂ŕ̥t.ḱo- initially preserved the cluster *-tk-, which then underwent simplification to -k- after another consonant (provided that it escaped the metathesis seen in *tít.kō > τίκτω). However, Schindler’s scenario can now be modified with the assumption that the metathesis of *TK followed Anatolian's separation from the rest of Indo-European (see Jasanoff 2018, 135, 138; Höfler 2024, 63 n. 24; on the reconstruction and syllabification of the Indo-European word for ‘bear’ in connexion with the problem of thorn clusters, see in greatest detail, Lipp 2009 vol. 2, 133–88). Even if Greek inherited a single ablaut grade, as is highly likely, one may suppose that the syllabification of the rare and marked cluster *-rtk- could oscillate between *-r.tk- and *-rt.k-, which were possibly treated in different ways. Finally, one should not forget that ‘bear’ is a lexeme often subject to tabooTaboo alterations, which may explain an irregular phonetic treatment (cf. Armenian arǰ for expected *arc‘).

At any rate, even if the variant ἀρκ- was indeed old in some dialects, it certainly did not belong to classical Attic; the single attestation of ἄρκειος in Aeschylus (C.3) may be regarded as a non-Attic element. The diffusion of ἀρκ- in the lower koine may be attributed to the influence of non-Attic dialects or to an independent simplification in lower-register speech that paralleled an earlier phenomenon in dialects such as Cyrenaean, or possibly to both factors. Consequently, to a 2nd-century-CE purist, a form such as ἄρκος would have sounded at once dialectally foreign to Attic and associated with substandard pronunciation. An additional reason for rejecting ἄρκος was the homophonyHomophony (albeit only in the nom. sing.) with the rare s-stem noun τὸ ἄρκος ‘defence’ (Alc.+), deriving from ἀρκέω (see B.1). While the unanimous rejection of ἄρκος may come as no surprise, it is interesting that a strand of scholarship possibly going back to Aelius Dionysius (A.3) nonetheless approved of ἀπαρκίας, justifying the latter form by appealing to ‘euphony’; Eustathius (B.4) also defends several other derivatives in -ἀρκ. Although Atticists generally exercised caution as far as tragedy was concerned, the early attestation of ἄρκειος in Aeschylus was adduced in support of the by-then-current forms in ἀρκ-. Nevertheless, the simplex noun ἄρκτος, amply attested since Homer (C.1, C.2), was likely too well-entrenched in the Greek literary tradition to permit a defence of ἄρκος, even on the part of those scholars who were prepared to admit other forms in ἀρκ-. The lexicographers adduced passages from Old Comedy (C.4, C.5, on which see Olson 2016, 504−5) to prove that ἄρκτος was the correct Attic form. Atticising authors use ἀρκτ-, except for the adjective ἄρκεια (δέρματα) ‘bear (skins)’ in Dio Chrysostom (C.10), which Schmid (Atticismus vol. 1, 156) catalogues among expressions indebted to post-classical prose. The reading ἄρκοι of some codd. in a passage of Aelian (C.11) stands isolated among the world’s 22 total occurrences in the same author, where the manuscripts are unanimous in transmitting ἄρκτ-.

E. Byzantine and Modern Greek commentary

That the variant ἄρκος, whether of purely phonological or folk-etymological origin, was a linguistic reality is proven by its survival in Post-classical Greek. The standard Modern Greek word for ‘bear’ is η αρκούδα, or more colloquially, το αρκούδι (see LKN s.vv.), with a diminutive suffix already attested in medieval sources (see Kriaras, LME s.vv. ἄρκος, ἀρκούδα, ἀρκούδιον). Pontic Greek notably preserves the unsuffixed noun άρκος (see Andriotis 1974 and ILNE s.v.). Throughout this period and up to the present day, άρκτος has remained in use as a learned form and is also used in the names of the constellations Μεγάλη/Μικρή Άρκτος ‘Ursa major/minor’.

F. Commentary on individual texts and occurrences

(1)    Eust. in Od. 1.215.40–216.7 (B.4)

In his commentaries on the Homeric poems, Eustathius expounds a doctrine about ἄρκ(τ)ος that he ascribes to a ‘rhetorical lexicon’ and that closely agrees with an entry in the expanded Synagoge (A.3). Based on this parallel, Erbse (1950, 109) identified the Synagoge entry as a fragment of Aelius Dionysius, since he held that by the expression ῥητορικὸν λεξικόν Eustathius usually refers to either Aelius Dionysius or Pausanias (Erbse 1950, 17). However, Cantore (2021, 185–92) points out that ῥητορικὸν λεξικόν is rather a generic designation and that we should be more cautious in assigning to Aelius Dionysius (or Pausanias) unattributed Atticist glosses in Eustathius (cf. Theodoridis 1982, XXXV–LX on the similar case of the ῥητορικόν mentioned in the Etymologicum genuinum). Cantore further suggests that the ῥητορικὸν λεξικόν mentioned in some fragments, including this one, may point to a version of the expanded Synagoge that was available to Eustathius. In support of ἀπαρκίας, Eustathius agrees with Σ^b in quoting ἄρκιος, to which he adds ἄρκιλος ‘bear cub’ (a hapax) and several forms whose etymological relationship with ἄρκτος is uncertain or unlikely, such as the ethnic Ἀρκάς ‘Arcadian’ or ἄρκυς ‘hunter’s net’. Quite interestingly, Eustathius then compares a different kind of cluster simplification found in Aristophanes, who depicts a character with a speech defect causing him to say ἄρτοςἄρτος ‘bread’ instead of ἄρκτος, and – more puzzlingly – τροφαλίς ‘piece of cheese’ instead of Τυρώ and σῦκα ‘figs’ instead of ἄστυ ‘city’. Although the interpretation of this fragment is debated (see Pellegrino 2015, 487; Lorenzoni 2017, 441–3; Bagordo 2020), it seems clear enough that the use of ἄρτος for ἄρκτος is a plausible mistake on the part of an ‘inarticulate’ character (a child?), while the confusion between Τυρώ (cf. τυρός ‘cheese’) and τροφαλίς pertains to the semantic level, and that between ἄστυ and σῦκα is perhaps a joking allusion to Athenian sycophancy. Eustathius correctly observes that ἄρκτος > ἄρτος engenders a change in meaning and, consequently, a comic effect, while ἄρκτος > ἄρκος does not. One may observe that, although both simplifications address the need to ease the pronunciation of a marked and unusual consonantal cluster, precisely the homonymie fâcheuse with ἄρτος ‘bread’ would have prevented the lexicalisation of an occasional learner’s error such as ἄρτος ‘bear’, while ἄρκος encountered no such obstacles.