ἀφείλατο, ἐξαφείλατο
(Phryn. Ecl. 154, Antiatt. ε 31)

D. General commentary

Phrynichus (A.1) and the Antiatticist (A.2) address the middle aorist of the suppletive verb αἱρέω ‘to take, seize; (mid.) to take for oneself, choose’ and its prefixed forms, which in the post-classical period developed an alphathematic form εἱλάμην alongside the inherited thematic εἱλόμην. While Phrynichus – like most other Atticist lexicographers (B.1, B.2, B.4, B.5) – firmly rejects this innovationInnovative forms, the Antiatticist defends it on the basis of an (alleged) Herodotean attestation.

The classical aorist corresponding to the present αἱρέω is, as is well known, the thematic εἷλον (from an IE root variously reconstructed as *s(u̯)el- or *selh₁-: see DELG, EDG s.v. ἑλεῖν; LIV, 529; on the suppletive paradigm, see Kölligan 2007, 48–66), amply attested from the Iliad onwards (C.1). The thematic aorist remained the norm for this verb and its many prefixed forms throughout the classical Attic period, across all literary genres (cf. e.g. C.4, C.5; for a fuller list of occurrences, see Veitch 1887, 28–9). From the Hellenistic age onwards, however, this and several other thematic aorists began a shift towards the alphathematic inflection (a development further facilitated by the existence of older pairs such as εἶπον/εἶπα ‘I said’ and ἤνεγκον/ἤνεγκα/ἤνεικα ‘I brought’; see Fiori 2022, 138–47; AGP vol. 1, 320–2). The analogical process began in the active aorist indicative, where both the thematic and sigmatic aorist shared a 3rd-person singular ending -ε(ν), and was further motivated by the advantage of eliminating the ambiguity of the thematic conjugation, with its identical ending -ον for the 1st-person singular and 3rd-person plural (Hatzidakis 1892, 185–6). The first thematic aorist stems to be affected were those ending in a resonant (ρ, ν, λ) or σ, i.e. those in which the /s/ of the sigmatic aorist suffix would appear to have been lost by regular sound change, leaving only the endings to differentiate the two formations (Wackernagel 1897, 48). According to a proportional analogy of the type ἔλυσεν : ἔλυσαν = ἔπεσεν : X (where X = ἔπεσαν), the 3rd plural ending -ον was replaced by -αν, which had the advantage of being distinct from the 1st singular ending (for further discussion of this development see Schwyzer 1939, 753–4; Debrunner 1926; Joseph 1980). The alphathematic endings then spread to the remainder of the active and middle paradigm, also affecting the imperfect of thematic verbs. This process ultimately culminated in the merger of thematic and alphathematic endings into a single set of past-tense endings (see E.). Incidentally, the spread of alphathematic endings to formerly thematic aorists is also attested in certain dialectsDialects, such as Cretan, where it is not necessarily due to koine influence. For instance, the Cretan aorist active infinitive ἀφέλαι (IC 2.1.2.26, Allaria, early 2nd century BCE) exhibits the same development as ἀφείλατο (see Schwyzer 1939, 755; Bile 1988, 221). Nevertheless, given the inscription’s date and the presence of koine forms (ἐάν, προσθεῖναι) alongside Doric(ising) ones (εὐχαριστῶμες), the influence of the koine is difficult to rule out.

As far as εἷλον and its prefixed forms are concerned, the spread of the innovative alphathematic forms was relatively late and gradual. The earliest occurrences in Attic inscriptions and in papyri date from the early Roman period (see Threatte 1996, 532–3; Mayser, Gramm. vol. 1,2, 144). As Gignac (1981, 344–5) observes, the phenomenon remains rare in the 1st century CE, but becomes frequent in the following century, especially in private documents, though occasionally also in official ones. In the New Testament, the prefixed forms of εἷλον display alphathematic endings 2x in the active and 4x in the middle, as against the thematic infinitive ἐξελέσθαι in Act.Ap. 7.34 (in a quotation from the Old Testament; see Blass, Debrunner 1976, 63–4). In particular, the alphathematic aorist of ἀφαιρέω ‘to take away’, mid. ‘to take away for oneself, to prevent; to deprive; to set someone free’ first occurs in the Hippocratic treatise On Ancient Medicine (C.6), dated to the late 5th century BCE (see Craik 2015, 285; a 4th-century-BCE dating was advocated by Blass 1887 vol. 1, 89), and then regularly in the Septuagint (15x, e.g. C.7) and in koine authors (D.S. 6x, including C.8; D.H. 11.22.7; Ph. Legatio ad Gaium 276; Babr. 1.105.4, etc.).

The late attestation of the alphathematic forms and their frequency in Christian Greek rendered them unacceptable to the majority of Atticists. Accordingly, the proscription in Phrynichus’ Eclogue (A.1) is echoed by Philemon (B.1, on which see Brown 2008, 99–101), the Philetaerus (B.2), Orus (B.4; on B.5 and B.6, see F.2), the pseudo-Herodianic Περὶ τῶν ζητουμένων (B.3), and Hesychius (B.7). The Suda (B.8), by contrast, in an entry that Adler traced back to the Synagoge tradition (cf. Σ α 1144), merely notes the existence of the alphathematic forms alongside the thematic ones. The Antiatticist (A.2), on the other hand, typically quotes a classical source – here Herodotus – as evidence for the admissibility of (ἐξ)αφείλατο: although the quotation itself is imprecise (see F.1), alphathematic forms are indeed attested as variae lectiones in two Herodotean passages (C.2, C.3).

If the MS tradition is to be trusted, Atticising writers by and large resisted the confusion between thematic and alphathematic aorists, apart from those cases (εἶπον/εἶπα and ἤνεγκον/ἤνεγκα) in which classical Attic itself already exhibited oscillation (Rutherford 1881, 215–21 discusses the extent of this phenomenon); a rare exception is εὕραντο in Ael. VH 3.17 (see Schmid, Atticismus vol. 3, 40; vol. 4, 602–3). As Alpers (1981, 10 n. 23) rightly observes, however, the innovative forms were often removed from the MSS of imperial-period authors already in the Middle Ages. Thus, for example, Arethas corrected the text of Clement of Alexandria in cod. Par. gr. 451 (913–914 CE) in accordance with Atticistic norms, changing two occurrences of (ἀφ)είλατο to -ετο (Clem.Al. Paed. 2.1.17.1 and 2.10bis.112.1). Remarkably, ἀφείλατο is also transmitted in P.Vat. gr. 11 (= TM 59953) [Marmarica, 215–250 CE], which preserves Favorinus’ treatise On Exile (C.9), while εἴλατο occurs later in the same work (96.14). Since this author is known to have been a frequent target of Phrynichus’ criticism throughout the Eclogue (see e.g. entries διδόασι, διδοῦσι; εἶμι, ἐλεύσομαι; νῆες, ναῦς, νῆας; υἱεύς, υἱέως, υἱέα), it cannot be excluded that Favorinus’ usage underlies the rejection of ἀφείλατο in A.1. Finally, Alpers (1981, 158) points to yet another varia lectio in Galen (C.10), where the MSS are divided between ἀφείλατο and ἀφείλετο.

E. Byzantine and Modern Greek commentary

During the medieval period, the endings of the imperfect (originally identical with those of the thematic aorist) merged with those of the sigmatic aorist and the perfect, producing by the end of this period a single set of active past-tense endings. These endings were added to the imperfective stem to form the imperfect, and to the perfective stem to form the aorist indicative (see Horrocks 2010, 143–4; 318–9; CGMEMG vol. 3, 1531–43; 1571–91; 1613–27; entry ἐλέγοσαν, ἐγράφοσαν, ἐσχάζοσαν). In this new set of endings, /a/-vocalism predominated, while only a few relics of the inherited thematic aorist survived. By contrast, owing to the disappearance of the middle voice as an independent category, innovative middle aorist forms such as εἰλάμην have no direct continuation in Medieval Greek.

Alongside the innovative tendencies discussed so far, a different strategy for replacing the thematic aorist consisted in creating sigmatic aorist forms regularly derived from the present stem: thus, from αἱρέω speakers formed the sigmatic aorist ᾕρησα, attested from the imperial period onwards and not uncommon in Medieval Greek (see Psaltis 1913, 220 on ἀφῄρησα and similar forms in the language of Byzantine chronicles).

At the same time, medieval authors continued to employ the classical thematic aorist of αἱρέω and its prefixed forms. While higher-register writers adhered more closely to Attic inflection, alphathematic forms also continued to be used during this period, chiefly in lower-register texts such as hagiographies or chronicles (e.g., 9x in John Malalas’ Chronographia), but occasionally also by a few classicising authors (e.g. Photius 3x, Michael Psellus 5x). The predominance of alphathematic forms in the Septuagint and the New Testament arguably lent them a degree of prestige, despite their almost unanimous proscription by the Atticists.

F. Commentary on individual texts and occurrences

(1)    Antiatt. ε 31 (A.2)

The reference to Herodotus’ first book is mistaken, since no relevant forms occur there (Lobeck 1820, 183; Bredow 1846, 354). As suggested by Alpers (1981, 158; see also Valente 2015, 54), an original reference to Book 7 may have been lost, assuming that Η was corrupted to Α. Indeed, in Hdt. 7.190.1 (C.3) the MSS of the Roman family (d) read ἀνείλατο for ἀνείλετο; the same varia lectio occurs in 2.134.4 (C.2), and the stem (-)ειλα- is also the reading of the Roman family elsewhere (8.113.3, 9.85; incidentally, this peculiarity of d, noted by Rosén 1962, 130, remains unexplained: it is possible that alphathematic forms were perceived as an Ionic trait, by generalisation from εἶπα, which is well attested in Ionic literature from Homer onwards). The lemmatisation of the doubly prefixed form ἐξαφείλατο, which does not occur in Herodotean passages just mentioned, nevertheless remains difficult to explain. Valente (2015, 157) draws a parallel with Antiatt. ε 29Antiatt. ε 29 (ἐξαναλαβῶ· Θουκυδίδης α′ ‘ἐξαναλαβῶ (‘that I take on board from’): Thucydides in the first [book]’, cf. Thuc. 3.79.1 τοὺς ἐκ τῆς νήσου ἀναλάβωσιν ‘that they take them on board from the island’), where the preposition ἐκ was merged with the verb to yield the entry’s lemma. However, no such preposition occurs in Hdt. 7.190.1, unless – as the anonymous referee suggests – the prefix was taken over from the immediately preceding word ἐκβρασσόμενα. Finally, despite the contrast between A.2 and A.1, there are insufficient grounds to posit that the Eclogue’s entry was the target of the Antiatticist’s polemic. As Valente (2015, 54; see already Sicking 1883, 96) observes, other Atticists share Phrynichus’ rejection of ἀφείλατο (see D.), and none of them discusses Herodotean usage. Rather, A.2 is best understood within the broader controversy over Herodotus’ status as a linguistic model.

(2)    Σ^b α 2504 (B.5)

This gloss and another one (B.6) from the expanded Synagoge were identified by de Borries (1911, XXXIV n. 1; 170) as fragments (or even a single fragment: ‘fortasse in unum contrahenda sunt’) of Phrynichus’ Praeparatio sophistica, on the grounds of their allegedly analogistic tendency. Alpers (1981, 65 n. 44), however, objected that the entry’s actual standpoint is anomalistic, since it observes that the verbs’ conjugation in classical Attic is not the same as the one expected on the basis of analogy (τὸ ἀνάλογον). He therefore assigned the entry to Orus, as fr. A 16b, comparing it with B.4, preserved in Ps.-Zonaras and expounding a similar doctrine (on Orus’ anomalistic inclinations see entry Orus, Ἀττικῶν λέξεων συναγωγή).