The Atticist lexica as metalinguistic resource for morphosyntactic change in Post-Classical Greek

In: Journal of Greek Linguistics
Ezra la Roi Ghent University Belgium Ghent

Search for other papers by Ezra la Roi in
Current site
Google Scholar
Open Access


While ancient metalinguistic resources such as lexica and scholia are increasingly studied in the field of ancient scholarship (Montanari 2020), they are investigated less within the historical sociolinguistics of Ancient Greek. Analysing the Atticist lexica by Phrynichus, Moeris and Aelius Dionysius, this article illustrates the historically persistent connection between social perception of and diachronic change within Ancient Greek. Although the historical relevance of Atticist prescriptivism has been observed, the evidence that these social evaluations provide for Post-Classical Greek language change is rarely assessed systematically (except for objectionable ideological reasons). I demonstrate that the Atticist lexica display metalinguistic awareness of the major morphosyntactic changes characterizing Post-Classical Greek (pace Lee 2013:286): paradigmatic (e.g. analogical levelling in verbal system of endings, voice and augment), category changes, category renewal (e.g. dual, pronouns, periphrasis), syntactic change (category expansion of ἔμελλον and τυγχάνω) and case changes (e.g. from case to prepositions).


While ancient metalinguistic resources such as lexica and scholia are increasingly studied in the field of ancient scholarship (Montanari 2020), they are investigated less within the historical sociolinguistics of Ancient Greek. Analysing the Atticist lexica by Phrynichus, Moeris and Aelius Dionysius, this article illustrates the historically persistent connection between social perception of and diachronic change within Ancient Greek. Although the historical relevance of Atticist prescriptivism has been observed, the evidence that these social evaluations provide for Post-Classical Greek language change is rarely assessed systematically (except for objectionable ideological reasons). I demonstrate that the Atticist lexica display metalinguistic awareness of the major morphosyntactic changes characterizing Post-Classical Greek (pace Lee 2013:286): paradigmatic (e.g. analogical levelling in verbal system of endings, voice and augment), category changes, category renewal (e.g. dual, pronouns, periphrasis), syntactic change (category expansion of ἔμελλον and τυγχάνω) and case changes (e.g. from case to prepositions).

1 Metalinguistic resources and the changing ancient Greek language

During the last decades a scientific field has taken its rightful place on the international stage, arguing for a rigorous study of all forms of so-called ancient scholarship, not only because they are relevant “themselves in their own right” but also because “the products of ancient scholarship are important, indeed fundamental, for what they tell us about themselves” (Montanari 2020: 4). With the arrival of a useful guide and comprehensive companions,1 scholars are now able to explore the wealth of information ancient sources provide about ancient sources and therefore about the Ancient Greek language of those sources. In fact, it can be a humbling experience for modern day grammarians of Ancient Greek to be reminded how strongly our modern ideas on Ancient Greek grammar not only were anticipated by scholars such as Aristarchus2 or grammarians such as Apollonius Dyscolus3 or Dionysius Thrax,4 but in some cases also still shape our current understanding of Ancient Greek.5

Moreover, there are distinct areas where ancient metalinguistic comments about the Ancient Greek language are of scientific interest to us now. Two areas which have been particularly relevant are those metalinguistic comments which deal with the social dimension and with the historical dimension of the Ancient Greek language. Metalinguistic comments aimed at the social dimension of Ancient Greek imposed social categorizations of reality such as barbarian vs. Greek (Hall 1989, 1997, 2002, Gruen 2012),6 non-Attic vs Attic (Colvin 1999, Willi 2003: 200–225), male vs. female (Willi 2003: 157–197; up-to-date references in Bentein & Janse 2021).7 Metalinguistic comments which deal with the history of the Ancient Greek language have received less attention,8 but are known to us from various source types. Although such comments are typically not made from within a clearly delineated theory of language,9 they can provide welcome insight into how the ancients conceptualized their language and its dimensions. The historian Herodotus, for example, in addition to recording the now-famous linguistic experiment by Psammetichus, himself deduced that Attic speakers must have spoken a different language before.

(1) εἰ τούτοισι τεκμαιρόμενον δεῖ λέγειν, ἦσαν οἱ Πελασγοὶ βάρβαρον γλῶσσαν ἱέντες. εἰ τοίνυν ἦν καὶ πᾶν τοιοῦτο τὸ Πελασγικόν, τὸ Ἀττικὸν ἔθνος ἐὸν Πελασγικὸν ἅμα τῇ μεταβολῇ τῇ ἐς Ἕλληνας καὶ τὴν γλῶσσαν μετέμαθε (Hdt. 1.57.9–12)

‘if (I say) one may judge by these, the Pelasgians spoke a language which was not Greek. If then all the Pelasgian stock so spoke, then the Attic nation, being of Pelasgian blood, must have changed its language too at the time when it became part of the Hellenes.’

In other literary works, language features that were the result of language change were ascribed to the person’s lack of education, (countryside) heritage or due to impact from other languages (see Blomqvist 2014).10 Such comments are more commonly found in various types of metalinguistic resources (e.g. scholia, Atticist lexica, linguistic treatises), which deal with the language of ancient sources. For instance, it is now well known that Aristarchus (as well as later scholiasts) noticed various types of diachronic change between the Homeric language, more recent poets (νεώτεροι) and contemporary language use (Nünlist 2012). Grammarians such as Apollonius Dyscolus, on the other hand, were less concerned with such diachronic matters, although not completely unaware of them (Matthaios, Montanari & Rengakos 2011).11 Specialized treatises dealing with linguistic ‘changes’ undergone by words (phonologically and morphologically) into their contemporary usage also seem to have existed, for example by Trypho (Dickey 2007: 84) and Herodian (Pagani 2015: 827). For Medieval Greek we possess various textbooks revealing how the Atticizing Greek which was taught in the education of the elite had changed over time (Cuomo 2017),12 thus demonstrating the intimate relationship of the social and historical dimension of Ancient Greek through time.

In the Atticism of the Second Sophistic,13 I argue, these two dimensions even become so closely intertwined that, at least in the prescriptive lexica of Phrynichus, Moeris and Aelius Dionysius,14 the social and historical are often two sides of the same coin. While ideas of linguistic correctness (hellenismos) have a long and complex history in both ancient philosophical and stylistic thinking (Pagani 2015, Matthaios 2020),15 the specific Atticism of these prescriptive lexica has recently received renewed interest, especially for their historical linguistic value.16 After all, as is well known, the words which those lexica rejected because they conflicted with their norm are typically the innovative and markedly Post-Classical words which occasionally survive even now (Horrocks 2010: 139, Lee 2013). Thus, these lexica constitute a rich metalinguistic resource of Antiquity for language change in Post-Classical Greek. However, the evidence which these lexica provide for language change in Post-Classical Greek is hardly ever studied systematically (cf. Tribulato 2019 for a plea to do so, and Monaco 2021: 15–81 for some detailed studies).17

Therefore, this study assesses the evidence which the Atticist lexica (by Phrynichus, Moeris and Aelius Dionysius18) provide for ongoing morphosyntactic changes in Post-Classical Greek, combining a historical and a sociolinguistic approach (Hernández Campoy & Conde Silvestre 2012). Although I am certainly not the first to notice the linguistic utility of the Atticist lexica, I argue that (i) the amount of available evidence for morphosyntactic change that they offer has been neglected thus far (ii) or at least only taken seriously for contestable ideological reasons, and (iii) without the proper caution, because the prescriptive lexica twist historical linguistic facts exactly because they are inherently ideological in nature. Already with the arrival of early editions, scholars noted the diachronic contrast between the words accepted and rejected by these Atticist lexica (Lobeck et al. 1820: passim, Hatzidakis 1892: 285, Rutherford 1986: passim). Still, typically the linguistic utility of these lexica results only in occasional mentions in grammars/histories of Post-Classical Greek or as exemplifications of Atticistic Greek (Dieterich 1898: 210–211, 242). Others which do take their linguistic utility more seriously (Thumb 1974: 4–8, Browning 1983: 44–50, Lee 2013) focus on the relevance for the diachrony of lexis. Lee 2013: 286 even goes on to claim that “we find no remark in the Atticists on many of the changes between Classical and Koiné Greek. This may be partly because the remains of their works are not complete; but some features, especially broader trends, simply slipped under their radar”. As I demonstrate, the Atticist lexica do in fact notice most changes familiar to us from linguistic histories as characterizing the changing morphosyntax of Post-Classical Greek, even though the majority of their prescriptions are aimed at matters of lexis (Kim 2010; Bentein 2021).

Still, there are scholars who do provide more comprehensive overviews, but only do so for ideological reasons, thus adding a layer of ideology to inherently ideological material. Whereas Hatzidakis 1892: 196–197, 204 contended that the Atticists lacked the Sprachgefühl for their language when they condemned certain morphological forms, Caragounis takes matters even one step further and claims that it is only because of such Atticists that Neohellenic is still Hellenic today (Caragounis 2010: 173). Caragounis reinterprets the Atticist lexica as successful resistance that continues Ancient Greek identity from ancient Athens to the modern day through shared language (Caragounis 2010: 155;165). He also concluded that “unlike Latin, which today lives only through its daughter languages, Greek is still the same language, having sustained the changes imposed by time, culture, religion, science and world-view” (Caragounis 2004: 21). Naturally such a theory-laden interpretation of the ideology of the Atticist lexica is not accepted by current specialists (Whitmarsh 2013a, Kim 2017) but it reveals a continuity on a different, more theoretical level: language change is viewed ideologically from the past to the present.

In fact, (historical) sociolinguists have often pointed out the discrepancy between language change and its social perception as some sort of decline (cf. the book by Aitchison 1998, with the apt title Language Change: Progress or Decay?).19 To illustrate, innovative linguistic variants are discouraged due to their association with certain dialects (Trudgill 1999: 12–15), innovative ideologies can shape language change (Romaine 2001: 155–165) and innovative usages which from a linguistic viewpoint fill a gap are held back because of ideology, e.g. the awkward use of you guys to address groups even if they include women in order to disambiguate from singular you (Bybee 2015: 10–11). This social dimension explains the strong link between the normative language of the Atticist lexica and the language change of Post-Classical Greek. Also, it resonates with still-existing ideologies of 19th century classicism which suggest that everything written in Post-Classical Greek is by definition of lesser quality on a literary, stylistic, and linguistic level (Whitmarsh 2013b: 1–5, Whitmarsh 2013a) and high-register Greek is treated as a “zombie” language even though it too was subject to language change (Horrocks 2010: 4).20

Finally, just as such socially motivated perceptions simplify current linguistic realities, the prescriptivism of the Atticist lexica simplify the historical realities of Classical and Post-Classical Greek. By virtue of their categorizations into accepted Classical and unaccepted Post-Classical, one loses track of the disparate trajectories of the usages in question along many dimensions, e.g. diachronically (e.g. what about differences between early and middle Post-Classical or condemned usages found in Classical Greek versus those which are not?), dialectally (e.g. which role does dialectal colouring and heritage play? (Cassio 2012: 261–263)),21 socially (e.g. to which social groups are language features ascribed? (Matthaios 2013)), and from a register perspective (e.g. what role does the model classical author in which the usage is attested play, the register it belonged to in Post-Classical Greek, and differences in accepted models across the lexica (cf. the contrast with the models of the Antiatticist, Valente 2015)?) After all, Post-Classical authors could change their style according to their need, as for example evidenced by the interchange of Atticist and non-Atticist diction in Lucian (Deferrari 1969; Adrados 2005: 201–202).

With these caveats in mind, this study first assesses the different types of language change which the prescriptivist Atticist lexica attest to (section 2).22 Subsequently, I detail the various types of morphosyntactic change from Post-Classical Greek evidenced by the lexica, such as paradigmatic change, category change, category renewal, syntactic change and case change. The findings that I present in section 3 on the area of morphosyntactic change stem from a close reading of all the prescriptive remarks of the Atticists Phrynichus, Moeris and Aelius Dionysius in the most recent editions,23 combined with searches in the Thesaurus Linguae Graecae (TLG).24 While I have tried to incorporate all remarks which relate to morphosyntactic changes in Post-Classical Greek, there are of course remarks which concern items that underwent other types of language change (as in section 2). Furthermore, by virtue of my focus on the commonality of the Atticists in detecting morphosyntactic change, I do not provide a full-fledged introduction of these Atticists separately,25 but only discuss relevant interpretative aspects of their work along the way. Translations are my own unless mentioned otherwise.

2 Types of Post-Classical language change in the Atticist lexica

As has been suggested before, diachronic changes in vocabulary and usage which characterise the Koiné were picked up by the Atticists and labelled as unfavourable contemporary language use. Such remarks thus indirectly reveal the Atticists’ metalinguistic awareness26 of language change in Post-Classical Greek. While I am not claiming that their metalinguistic abilities were of the kind of the modern linguist categorising linguistic changes (as I attempt below), their synchronic prescriptivist statements acknowledge ongoing diachronic changes. In fact, with the diachronic changes of early and middle Post-Classical Greek (3rd century BCE–3rd century CE)27 in mind, it becomes clear why the Atticist lexica discouraged certain forms: in Post-Classical Greek these (i) underwent semantic shifts (example 1), (ii) developed semantic doublets (example 2), (iii) underwent sound changes affecting orthography (example 3) or changed gender (example 4). In (1), the prescriptivist comment makes us aware of the semantic narrowing that παιδάριον has undergone in Post-Classical Greek. In (2), the near-synonym νομίζω, which increased its usage at the expense of οἶμαι, is discouraged, even though both were available in Classical Greek (Monaco 2021: 37). I use parentheses in the translation to mark the form which is commented upon.

(1) παιδάριον καὶ τὸ θυγάτριον Ἀττικοί· παιδάριον μόνως τὸ ἄρρεν Ἕλληνες. (Moeris π 62)

(Little kid) also (the little daughter) Attic: (little kid) only male Hellenic.

(2) οἶμαι καὶ οἴομαι Ἀττικοί· νομίζω Ἕλληνες. (Moeris ο 28)

(I think) and (I think) Attic: (I think) Hellenic.

The next testimony discourages a contemporary form of pronunciation (and orthography) from Post-Classical Greek as the delta in this context had already been changed to a theta (see Horrocks 2010: 170). Phrynichus urges non-use of these contemporary Post-Classical forms.28

(3) Οὐθεὶς διὰ τοῦ θ, εἰ καὶ Χρύσιππος καὶ οἱ ἀμφαὐτὸν οὕτω λέγουσιν, σὺ δὲ ἀποτρέπου λέγειν· οἱ γὰρ ἀρχαῖοι διὰ τοῦ δ λέγουσιν οὐδείς. (Phryn. Ecl. 153)

(Nobody) with the theta, even though Chrysippos and his school say it that way, you must refrain from saying it, for the ancients said it with the delta, (nobody).

Also, there are discouraging remarks on the use of forms in their contemporary Post-Classical gender.29

(4) Οἱ χόλικες ἀμαθές· οἱ γὰρ δόκιμοι θηλυκῶς αἱ χόλικές φασιν (Phryn. Ecl. 282)30

(The bowels) is dumb; for the esteemed say (the bowels) the feminine way.

These examples, then, illustrate the Atticists’ more familiar metalinguistic awareness of semantic and formal changes which made them discourage certain Post-Classical forms. The Atticists did not stop there, because they also discouraged forms which had undergone morphosyntactic change in Post-Classical Greek (i.e. changes affecting the correlation between form and syntactic function in different levels of structure e.g. verb/nominal phrase or clause) and therefore did not qualify as proper Atticistic Greek anymore in their eyes.

3 Atticist lexica on morphosyntactic changes

3.1 Paradigmatic changes and analogical levelling

It is a well-known fact that Post-Classical Greek saw an increase of analogical pressures on the verbal paradigms. For example, the εἶπα aorist of λέγω, which was formed by analogy with the sigmatic aorist endings, was only infrequent in Classical Greek but increased in frequency in Post-Classical Greek (Horrocks 2010: 109–110). In research on language change, this paradigmatic change is known as morphological analogy, “the re-making of a word based on similarity to other existing words in the language” (Bybee 2015: 93). In Post-Classical Greek the following verb forms undergo such morphological analogy, albeit according to different analogical models (see in parentheses). As such they became marked features of Post-Classical Greek Koiné (Horrocks 2010: 73–75; 82;103;144;154).31

Table 1

Paradigmatic changes observed by Atticist lexica

Prescribed form

Discouraged form



εἶπα (sigmatic aorist)

Aelius Dionysius


ἤνεγκα (sigmatic aorist)

Ael. Dion./Phrynichus


ἐλέγοσαν (athematic middle)

Aelius Dionysius


οἶδας (first person form)

Ael. Dion./Moeris


οἴδασιν (first person form)



ἐγρήγορα (present indicative)



ἤμην (middle imperfect)



ἦς (root aorist endings)



ᾔδεις (thematic pluperfect)



ᾔδειν (thematic pluperfect)



ζευγνύω (thematic present)



ζευγνύουσιν (thematic present)



δεικνύουσιν (thematic present)



ὀλλύουσιν/ὀμνύουσιν (thematic present)



ῥηγνύουσιν (thematic present)



ἀπεδώκαμεν/ἀπεδώκατε/ἀπέδωκαν (1st to 3rd person)



ἀποκριθῆναι (-θῆ aorist)



καταλεχθείς (-θῆ aorist)



παρῳκίσθην (-θῆ aorist)



πραθήσομαι/πραθήσῃ/πραθήσεται (-θῆ future)



γυμνασθήσεται (-θῆ future)



τιμηθήσεται (-θῆ future)



λογίσεται (-σ middle future)



μανήσεται (-σ middle future)



διαβιβάσω (-σ middle future)



ἐλάσω (-σ middle future)



ᾄσει (active future)



βοήσει (active future)



θηράσει (active future)



ὀμόσω/ὀμόσει (active future)



πράξω/πράξει (active future)


When these analogical forms actually eliminate an older form which was perceived as irregular, a so-called process of analogical levelling has been completed (Bybee 2015: 94–97). These paradigmatic changes have been picked up by the Atticists who naturally condemned the use of these new forms,32 most of which were actually already in use in Classical Greek.33 Thus, these new forms are not homogenous diachronically, since some innovative forms came through already in Classical Greek whereas others were used only later. I suggest that by saying that they prefer the older forms over what they view as markedly Koiné forms, these Atticist testimonies indicate their awareness of the ongoing processes of change. Compare the following remarks by Aelius Dionysius who prescribes the older forms εἶπον (5) and ἤνεγκον (6).

(5) εἶπον καὶ εἶπα· ἀμφότερα παρὰ Ἀττικοῖς, μᾶλλον δὲ τὸ πρότερον. καὶ τὰ προστακτικὰ δὲ ἀμφοτέρως καὶ εἰπὲ καὶ εἶπον [ὀξυτόνως]. καὶ αἱ μετοχαὶ ὁ εἰπὼν καὶ ὁ εἴπας. (Ael. Dion. ε 22)

(I said) and (I said); both with the Attics, but preferably the former. and the imperatives [are] both say! and say! [oxytone]. and the participles he who said and he who said.

(6) ἤνεγκα καὶ ἤνεγκον ἄμφω λέγουσιν, μᾶλλον δὲ τὸ ἤνεγκον· τὸ μὲν ἀπὸ τοῦ ἐνέγκαι, τὸ δὲ ἀπὸ τοῦ ἐνεγκεῖν. (Ael.Dion. η 10)

They say both (I brought) and (I brought), but preferably (I brought); the former from to have brought, the latter from to have brought.

Quite surprisingly, example 7 shows us that Phrynichus, who is generally believed to have been the strictest Atticist, approves both ἤνεγκον and ἤνεγκα.34

(7) ἤνεγκον: ἀπὸ τῆς ἐνεγκών μετοχῆς, ὡς ἀπὸ τῆς δραμών ἔδραμον. τὸ δὲ ἤνεγκα ἀπὸ τῆς ἐνέγκας. ἄμφω μὲν οὖν δόκιμα. (Phryn. PS 73.1–3)

(I brought): from the participle to have brought, as with I ran from to have run. (I brought) is from to have brought. both are approved.

In a similar vein, Aelius Dionysius’ comment (example 8) is not that strongly opposed to the new analogical form οἶδας, which contrasts with the comment by Moeris who marks the feature as belonging to the Post-Classical Greek Koiné, although perhaps to an intermediate level of Koiné, which some argue that Ἕλληνες refers to.35 These examples thus highlight the role of different standards between Atticist lexicographers.

(8) οἶσθα· ἀντὶ τοῦ οἶδας. λέγεται [καὶ] χωρὶς τοῦ σ· μετὰ δὲ τοῦ σ ποτὲ ἢ διὰ μέτρον ἢ διὰ τὸ μὴ συγκροῦσαι φωνήεντα (Ael.Dion. o 11)

(you know) instead of (you know). it is said [also] without the s; and with the s sometimes either due to meter or to not collide vowels

(9) οἶσθα χωρὶς τοῦ σ Ἀττικοί· οἶδας Ἕλληνες. (Moer. o 24)

(you know) without the s [is] Attic; (you know) Hellenic.

Ironically, οἶδας is actually an old Attic form (Rutherford 1881: 227), but is given another status by the Atticists.

Toward the new analogical forms for ἦν and ἦσθα, the Atticist lexicographers have stronger feelings, since all strongly prefer the older form. Phrynichus even calls the new form ἦς barbaric (example 12), something which he does not for the use of ἔφης instead of ἔφησθα (Phryn. Ecl. 206). If Ἕλληνες did indeed refer to an intermediate Koiné, then Moeris does not discourage it as strongly.

(10) Ἤμην· εἰ καὶ εὑρίσκεται παρὰ τοῖς ἀρχαίοις, οὐκ ἐρεῖς, ἀλλἦν ἐγώ. (Phryn. Ecl. 123)

(I was): even though it is also found with the ancients, do not say this, but (I was)

(11) ἦν Ἀττικοί· ἤμην Ἕλληνες. (Moer. η 2)36

(I was) Attic: (I was) Hellenic.

(12) Ἦς ἐν ἀγορᾷ σόλοικον, λέγε οὖν ἦσθα. ὀρθότερον δὲ χρῷτο ἂν ὁ λέγωνἐὰν ᾖς ἐν ἀγορᾷ“. (Phryn. Ecl. 118)

(You were) in the marketplace [is] speaking incorrectly, so say (you were). It [i.e. ἦς] would be used better when someone would say ‘if you will be in the marketplace’

Let us now turn to the remainder of the analogical changes in Post-Classical Greek that are vindicated. These changes stem from four major reorganisations of the Post-Classical verbal system: (1) increase of thematic endings at the cost of athematic and irregular endings; (2) increase of the -θη aorist over older middle aorists; (3) reorganization of the voice system of the synthetic future; (4) analogical increase of the augment. The first change is well known from our linguistic histories of Ancient Greek. The increasing use of thematic endings, as shown by the remarks by the Atticists below, not only affected athematic verbs such as ζεύγνυμι but also irregular verb endings such as the pluperfect ᾔδεισθα (example 25):

(13) ᾔδεισθα Ἀττικοί· ᾔδεις Ἕλληνες. (Moeris η 1)

(you knew) Attic: (you knew) Hellenic.

(14) ᾔδη Ἀττικοί· ᾔδειν Ἕλληνες. (Moeris η 3)

(he knew) Attic: (he knew) Hellenic.

(15) ζεύγνυμι Ἀττικοί· ζευγνύω Ἕλληνες. (Moeris ζ 1)

(I join) Attic: (I join) Hellenic.

(16) ζευγνῦσιν Ἀττικοί πληθυντικῶς καὶ περισπωμένως· ζευγνύουσιν Ἕλληνες. τὸ δὲ ζευγνύασιν τῆς δευτέρας Ἀτθίδος. (Moeris ζ 8)

(they join) Attic in the plural and perispomenon: (they join) Hellenic. (They join), from the second Attic.

(17) δεικνῦσι προπερισπωμένως Ἀττικοί· δεικνύουσιν Ἕλληνες· δεικνύασι δὲ οἱ δεύτεροι Ἀττικοί. (Moeris δ 29)

(they show) properispomenon Attic: (they show) Hellenic. (They show) the second Attic.

(18) ὀλλύασιν ὀμνύασιν Ἀττικοί· ὀλλύουσιν ὀμνύουσιν Ἕλληνες. (Moeris ο 15)

(they destroy) (they swear) Attic: (they destroy) (they swear) Hellenic

(19) ῥηγνύασιν διὰ τοῦ α Ἀττικοί· ῥηγνύουσιν ἢ ῥήσσουσιν Ἕλληνες. (Moeris ρ 5)

(they break) with the a Attic: (they break) or (they break) Hellenic.

(20) ἀπέδομεν ἀπέδοτε ἀπέδοσαν Ἀττικοί· ἀπεδώκαμεν ἀπεδώκατε ἀπέδωκαν Ἕλληνες. (Moeris α 19)

(we gave back) (you gave back) (they gave back) Attic: (we gave back) (you gave back) (they gave back) Hellenic.

Note that Moeris also keenly observes that Classical Greek itself already displayed variation in certain instances (see example 17 “second attic”).37 Similarly, he signals the later morphological variant of ῥήσσουσιν from Post-Classical Greek (example 19).

Furthermore, the Atticists warn against using the -θη aorists which from early Post-Classical onwards starting ousting the middle aorists from the more complex aorist voice system known to us from Classical Greek (Browning 1983: 30; Horrocks 2010: 103; Tronci 2018). As the following testimonies indicate the -θη aorist should not be used by Atticistic writers instead of the older middle/passive aorists even though the -θη aorists replace them in Post-Classical Greek.

(21) Ἀποκριθῆναι· διττὸν ἁμάρτημα, ἔδει γὰρ λέγειν ἀποκρίνασθαι, καὶ εἰδέναι, ὅτι τὸ διαχωρισθῆναι σημαίνει, ὡσπεροῦν καὶ τὸ ἐναντίον αὐτοῦ, τὸ συγκριθῆναι, ⟨τὸεἰς ἓν καὶ ταὐτὸν ἐλθεῖν. εἰδὼς οὖν τοῦτο ἐπὶ μὲν τοῦ ἀποδοῦναι τὴν ἐρώτησιν ἀποκρίνασθαι λέγε, ἐπὶ δὲ τοῦ διαχωρισθῆναι ἀποκριθῆναι. (Phryn. Ecl. 78)38

(Answer): double mistake, for he ought to have said (answer), and known, that it signifies being separated, as in fact also its reverse, being combined, the coming to one and the same. So know this and with returning the question say (answer), but with being separated (to be separated).

(22) καταλεγείς Ἀττικοί· καταλεχθείς Ἕλληνες. (Moeris κ 7)

(recounted) Attic: (recounted) Hellenic.

(23) παρῳκισάμην οὐ παρῳκίσθην. (Moeris π 22)

(I dwelled) not (I dwelled).

Now, for the future system there are three groups of remarks which are witness to the reorganization of the future system in early and middle Post-Classical Greek: (i) increase of -θη futures at the expense of middle futures; (ii) increase of sigmatic futures at the expense of Attic futures; (iii) replacement of futura tanta by active futures (see esp. Tronci 2020: 118–133). In the examples 24 to 26 we see that both future perfects (24) and middle futures (25 and 26) which were used in Attic Greek are to be preferred over their Post-Classical replacements, -θη- futures.

(24) πεπράσομαι πεπράσῃ πεπράσεται Ἀττικοί· πραθήσομαι πραθήσῃ πραθήσεται Ἕλληνες. (Moeris π 7)

(I will do) (you will do) (he will do) Attic: (I will do) (you will do) (he will do) Hellenic.

(25) γυμνάσεται Ἀττικοί· γυμνασθήσεται Ἕλληνες. (Moeris γ 24)

(he will exercise) Attic: (he will exercise) Hellenic.

(26) τιμήσεται Ἀττικοί· τιμηθήσεται Ἕλληνες. (Moeris τ 16)

(he will honour) Attic: (he will honour) Hellenic

In addition, we see that a variety of so-called Attic futures was preferred by the Atticists over their sigmatic replacements in Post-Classical Greek.

(27) λογιεῖται Ἀττικοί· λογίσεται Ἕλληνες. (Moeris λ 24)

(he will reckon) Attic: (he will reckon) Hellenic

(28) μανεῖται Ἀττικοί· μανήσεται Ἕλληνες. (Moeris μ 30)

(he will be furious) Attic: (he will be furious) Hellenic.

(29) διαβιβῶ Ἀττικοί· διαβιβάσω Ἕλληνες. (Moeris δ 19)

(I will cross) Attic: (I will cross) Hellenic.

(30) ἐλῶ, ἐλᾶ Ἀττικοί· ἐλάσω Ἕλληνες. (Moeris ε 24)

(I will drive), (he will drive) Attic: (I will drive) Hellenic.

Finally, there is a set of testimonies which concern so-called future tanta, forms which are active in the present but have an older middle future form with the same active meaning as the present. As with the previously mentioned changes, the following testimonies reflect the preference among the Atticists (esp. Moeris) that such new future formations are to be avoided.

(31) ᾄσεται Ἀττικοί· ᾄσει Ἕλληνες. (Moeris α 83)

(he will sing) Attic: (he will sing) Hellenic.

(32) βοήσεται Ἀττικοί· βοήσει Ἕλληνες. (Moeris β 37)

(he will shout) Attic: (he will shout) Hellenic.

(33) θηράσεται Ἀττικοί· θηράσει Ἕλληνες. (Moeris θ 7)

(he will hunt) Attic: (he will hunt) Hellenic.

(34) ὀμοῦμαι ὀμεῖ ὀμεῖται Ἀττικοί· ὀμόσω ὀμόσει Ἕλληνες. (Moeris ο 8)

(I will swear), (you will swear), (he will swear) Attic: (I will swear) (he will swear) Hellenic.

(35) πράξομαι Ἀττικοί· πράξω Ἕλληνες. (Moeris π 2)

(I will do) Attic: (I will do) Hellenic.

(36) πράξεται Ἀττικοί· πράξει Ἕλληνες. (Moeris π 3)

(He will do) Attic: (he will do) Hellenic.

A last process of analogical levelling seems to almost have been completed in early Post-Classical Greek (IIIBCEIBCE): the analogical use of the η-augment with verbs which strictly speaking did not need it. In Classical Greek, the η-augment from the verb ἐθέλω formed the analogical model for the creation of ἤμελλον ‘was about to’, ἠβουλόμην ‘wanted’, and ἠδυνάμην ‘could’, verbs without a vocalic beginning (Schwyzer & Debrunner 1950: 654; Mandilaras 1973: 124). In the Ptolemaic period, such analogically created forms become more frequent up to the point that they seem to have become the rule in the papyri (Mandilaras 1973: 115;116;120). However, in the Post-Ptolemaic period things start to change, as, for example, ἔμελλον and ἐβουλόμην predominate in the Post-Ptolemaic papyri and ἐδυνάμην gradually replaces its analogical form according to Mandilaras. It seems that the following testimony by Moeris reflects this change from the Roman period onwards.39 He suggests that the forms with the ε-augment belong to the Post-Classical Greek Koiné, which according to the distributions can only be applied to the Roman period.

(37) ἤμελλον ἠβουλόμην ἠδυνάμην ηὐξάμην διὰ τοῦ η· διὰ δὲ τοῦ ε Ἕλληνες. (Moeris η 5)

(I was going to), (I wanted to), (I could), (I wished) through the η; through the e [is] Hellenic

To sum up, the Atticist lexicographers provide ample awareness of different processes of analogical levelling which are so characteristic of the PostClassical Greek verbal system.

3.2 Category change

In the history of Ancient Greek multiple imperatives (e.g. ἄγε, φέρε, ἴθι and ἀμέλει) have undergone category change, as is recognizable from their acquisition of non-imperative functions and transition from imperative to discourse particle (la Roi 2022).40 The following examples witness aspects of change from the changes which the imperatives are undergoing. In example 39 the imperative singular is no longer used as a singular but used as a directive towards multiple people.41 This use indicates the ossification of the imperative form into a particle like construction that is used to increase “the directness of the speech act”, in this case the question (Zakowski 2014: 190).42

(38) εἰπέ μοι, τί τοῦτἀπειλεῖ τοὔπος, ἄνδρες δημόται, τοῖς Ἀχαρνικοῖσιν ἡμῖν; (Ar. Ach. 328–329)

Tell me, fellow demesmen, what does he mean by this threat against us Acharnians? (transl. Henderson)

(39) φέρε λόγων ἁψώμεθ᾽ ἄλλων. (E. Ion. 544)

Come, let us take a different tack (transl. Kovacs)

(40) ἄγε, βάδιζἀναστᾶσ’ (Men. Sicyonis 267)

Come on, stand up and walk. (transl. Arnott)

In example 39 φέρε is not literally an imperative to bring or carry someone/something anymore but strengthens the directive force of the exhortation that it precedes. (see Zakowski 2018 on the category change of ἄγε, φέρε and ἴθι). In example 40 the particle ἄγε can be left out of the sentence without changing the semantics of the sentence.

In Post-Classical Greek, the imperative ἀμέλει ‘do not worry’ becomes an intersubjective particle, meaning ‘of course’ (as in the following example), thus losing its possibility of a complement referring to the object of worry but gaining another meaning (cf. Blomqvist 1969: 103–107). Aelius Dionysius, a prime figure in early Atticist lexicography from the early second century CE, calques ἀμέλει in ways that betray his awareness of the predominantly Post-Classical Greek evolution into a particle.43 In example 41 he makes clear that he sees ἀμέλει as a subjective (δηλαδή, πάντως) adverb (ἐπίρρημα) signifying approval. The category of adverb was most probably introduced in the first century BCE (Wouters 1979; Kärnä & Matthaios 2007; de Jonge 2008).

(41) ἀμέλει· δηλαδή, πάντως· ἐπίρρημα γάρ ἐστι συγκατάθεσιν δηλοῦν. (Ael.Dion. α 97)

(no worries); clearly, undoubtedly; for it is an adverb signifying approval.

In another entry by Aelius Dionysius ἀμέλει is given as a close synonym for the particles δήπου and δηλονότι, which are similar in content.

(42) δήπου· ὡςδή’, ἀντὶ τοῦ ἀμέλει, δηλονότι. (Ael.Dion. δ 14)

(doubtless); as ‘evidently’, instead of ‘of course’, obviously.

While the above characterizations are telling testimonies of the main Post-Classical Greek use of ἀμέλει, they carry a blind spot. After all, do such statements imply that ἀμέλει was only used as adverb? This is not the case in Post-Classical Greek, where it could still be used in its original meaning with a complement.44 If so, such statements also do not take into account the synchronic coexistence of historical layers in one grammatical item.45

3.3 Category renewal

There is also evidence that the lexicographers signalled changes involving category renewal. The following remark by Moeris shows that they were aware of the renewal of wish particles in Post-Classical Greek. The originally conditional εἰ γάρ had already by the time of Archaic Greek turned into a wish particle due to insubordination (la Roi 2021) and was continued to be used as such in Classical Greek. In early Post-Classical Greek (3rd century BCE to 1st BCE), the new combination εἴθε γὰρ is introduced which fuses insubordinate εἴθε and εἰ γάρ.

(43) εἰ γάρ Ἀττικοί· εἴθε γάρ κοινόν. (Moer. ε 60)

(if only) Attic: (if only) common

The fact that Moeris uses the label κοινόν is significant. Although some had supposed that this refers to the contemporary colloquial Koiné (Jannaris 1897: 96; Thumb 1974: 6 note 1), Monaco (2021: 38) has rightly indicated that this explanation does not work for all items labeled κοινόν. In fact, the distribution of εἴθε γάρ provides some relevant evidence to the issue at hand. It is only found in high-register contexts in early Post-Classical Greek such as Callimachus’ poetry or falsely attributed to the Lydian king Croesus (la Roi 2021). At the same time, it is found in middle Post-Classical Greek in contexts marked by a register upgrade, such as to represent older language in histories (e.g. Plut. Dem. 38.8.1 or Cass. Dio. H.R. 64.11.1) or once in a high register official papyrus (PSI. 10.1103. ll. 15). Thus, the comment by Moeris seems to reveal the noteworthiness of a new high register feature which effects a change initiated from above.

The well-known loss of the dual in Post-Classical Greek (Horrocks 2010: 73) is also flagged by the Atticist lexica, as the older dual is recommended whereas its replacements from Post-Classical Greek, the numeral or the personal pronoun, are vindicated.

(44) Δυσὶ μὴ λέγε, ἀλλὰ δυοῖν. (Phryn. Ecl. 180)

do not say (two), but (two both)

(45) νώ δυϊκῶς Ἀττικοί· ἡμεῖς Ἕλληνες. (Moeris ν 2)

(both of us) dually Attic: (we) Hellenic

Similarly, the older set of pronouns σφεῖς were filtered out of the system in Post-Classical Greek (Jannaris 1897: 152) and therefore their ‘replacements’ as αὐτ- were discouraged by the Atticists. However, here we notice the importance of the Atticistic lens through which ‘proper Attic’ is viewed, since these αὐτ- pronouns were in fact in use in Attic Greek, but because they were not in equally strong use as personal pronouns in Post-Classical Greek their use is discouraged.

(46) σφεῖς Ἀττικοί· αὐτοί Ἕλληνες. (Moeris σ 3)

(They) Attic: (they) Hellenic

(47) σφῶν Ἀττικοί· αὐτῶν Ἕλληνες. (Moeris σ 4)

(Their) Attic: (their) Hellenic

(48) σφᾶς Ἀττικοί· αὐτούς Ἕλληνες. (Moeris σ 5)

(them) Attic: (them) Hellenic

Also the renewal of the reduplicated pluperfect with periphrastic alternatives in Post-Classical Greek has been flagged by the Atticists (cf. also Jannaris 1897: 492; Mandilaras 1973: 132–134). Moeris suggests to avoid the newer periphrastic formation which seeks to replace the reduplicated pluperfect ἐτετάχατο. This reduplicated pluperfect is in Post-Classical Greek only found in Atticistic writers such as Lucian and Arrian.

(49) ἐτετάχατο Ἀττικοί· τεταγμένοι ἦσαν Ἕλληνες. (Moeris ε 47)

(they were arranged) Attic: (they were arranged) Hellenic

Finally, we should notice that Atticists also note category renewals which do not seem to have taken place. A statement by Moeris implies that the combination of ὅπερ with habitual φιλεῖ has been replaced by the other habitual verb εἴωθα (for the habitual auxiliaries see la Roi 2020b).

(50) φιλεῖὅπερ φιλεῖ γίνεσθαιἈττικοί·ὅπερ εἴωθε γίνεσθαιἝλληνες. (Moeris φ 3)

(Is wont) “what is wont to happen” Attic: “what usually occurs” Hellenic.

However, both verbs had been grammaticalized as habitual auxiliaries already in Classical Greek and both were also still in use in Post-Classical Greek. For example, the habitual expression ὅπερ φιλεῖ is also used in Polybius and Flavius Josephus.

3.4 Syntactic changes

This section delves deeper into the metalinguistic awareness of syntactic change as reflected in three prescriptivist testimonies on ἔμελλον and τυγχάνω. Although at first these testimonies may seem insignificant,46 viewed from the perspective of theories on language change, they provide indirect testimony to the ongoing category expansion (Bybee 2015: 127) of both μέλλω and τυγχάνω.

(51) Ἔμελλον ποιῆσαι, ἔμελλον θεῖναι· ἁμάρτημα τῶν ἐσχάτων εἴ τις οὕτω συντάττει· τετήρηται γὰρ ἢ τῷ ἐνεστῶτι συνταττόμενον ἢ τῷ μέλλοντι, οἷονἔμελλον ποιεῖν“, „ἔμελλον ποιήσειν· τὰδὲ συντελικὰ οὐδένα τρόπον ἁρμόσει τῷ ἔμελλον (Phryn. Ecl. 313)

(I was going to/about to) do, (I was going to/about to) put: a mistake of the highest [kind] if someone arranges that way: for it is observed as either arranged with the present or the future, as (I was about/going to) do, (I was about to/going to) do: these completed things in no way fit (I was going/about to).

(52) Ἔμελλον γράψαι· ἐσχάτως βάρβαρος ἡ σύνταξις αὕτη· ἀορίστῳ γὰρ χρόνῳ τὸ ἔμελλον οὐ συντάττουσιν οἱ Ἀθηναῖοι, ἀλλἤτοι ἐνεστῶτι, οἷονἔμελλον γράφειν“, ἢ μέλλοντι, οἷονἔμελλον γράψειν“. (Phryn. Ecl. 347)

(I was going to) write: this composition is extremely barbaric; for the Athenians do not arrange (I was about/going to) with the aorist tense, but either with the present, as (I was about/going) to write, or the future, as (I was going/about to) write.

(53) Τυγχάνω· καὶ τούτῳ προσεκτέον· οἱ γὰρ ἀμελεῖς οὕτω λέγουσιν·φίλος σοι τυγχάνω, ἐχθρός μοι τυγχάνεις“, δεῖ δὲ τῷ ῥήματι τὸ ὤν προστιθέναι, „φίλος μοι τυγχάνεις ὤν, ἐχθρός μοι τυγχάνεις ὤν“. οὕτω γὰρ οἱ ἀρχαῖοι ἐχρήσαντο. (Phryn. Ecl. 244)

(I happen to be/I am); to this [something] should be connected; for careless speakers speak this way: (I am) dear to you, (you are) an enemy to me, but it is necessary that being is added to this verb, (you are) dear to me, (you are) an enemy to me. For this way the ancients used it.

The verb μέλλω has undergone a long grammaticalization process already from Homeric Greek onwards. In Homeric Greek the verb mainly had an epistemic modal meaning ‘be evident/probable’ and a relative future meaning of predestination ‘be destined to’. In Classical Greek it developed three new meanings: intention ‘intend to/going to’, immediate future ‘to be about to’ and a new lexical meaning ‘delay’ (see Allan 2017). The prescriptive comments by Phrynichus distort the distributions of this grammaticalizing future auxiliary, since Phrynichus indicates that he sees ἔμελλον as inherently future referring and therefore allegedly unfit for combination with the aorist (e.g. τὰδὲ συντελικὰ οὐδένα τρόπον ἁρμόσει τῷ ἔμελλον ‘these completed things in no way fit (I was about/going to)’). As Markopoulos has suggested, the increase in past future use of ἔμελλον with the aorist infinitive was a genuine phenomenon in Post-Classical Greek, as he records that the combination make up a third of the occurrences in the papyri of the Hellenistic Roman period whereas he claimed it is only hardly attested in Classical Greek.47 What Phrynichus thus indirectly seems to testify to is the functional rearrangement in the complementation system of Post-Classical Greek. After all, it has been shown by among others Markopoulos and Bentein (Bentein 2017: 8–9; Bentein 2018) that the loss of future infinitives in Post-Classical Greek had as a result that aorist infinitives became polyfunctional and, among other things, came to refer to future state of affairs as well. In the next papyrus the aorist infinitive refers to future state of affairs with the immediate future meaning, an aspectually relevant choice as the aorist aspect on the infinitive of a stative verb of sleeping provides an ingressive reading ‘go to sleep’.

(54) ἡνίκα ἤμελλον κοιμηθῆναι, ἔγραψα ἐπιστόλια β, ἓν μὲν περὶ Ταύγχιος τῆς ἐκ Θερμούθιος, ἓν δὲ περὶ Τετειμούθιος τῆς Ταυῆτος (chr.wilck.50. ll. 9–15)

Just before I was about to go to sleep, I wrote two letters, one about Taunchis the daughter of Thermouthis, and one about Tetimouthis the daughter of Taues (transl. Bagnall & Derow 2004)

However, note again that Phrynichus’ observations are made through an Atticistic lense, because combinations with an aspectually relevant aorist infinitive were also used in Classical Greek texts (e.g. Hdt. 8.40, Th. 6.31 or Lys. 1.17) but due to their infrequent appearance deemed only a marginal phenomenon (Markopoulos 2009: 28–30) even though they were aspectually relevant.

With regards to τυγχάνω, matters are slightly different. Even though Phrynichus condemns the use of τυγχάνω without a participial complement, such occurrences are to be found in Classical Greek already (esp. in Plato).48 As suggested by Bentein, the frequent periphrastic use of τυγχάνω with a present participle to express a durative event has contributed to the semantic bleaching of τυγχάνω. Consequently, τυγχάνω came to be used in the meaning ‘in the given circumstances it is the case that I …’, becoming nearly synonymous with εἰμί. As a result, τυγχάνω in this meaning lost the necessity of having a participial complement, a process which was strengthened in Post-Classical Greek by the increasing loss of participles in the complementation system.49 In other words, Phrynichus’ prescriptive comment is testimony to the ongoing morphosyntactic change of τυγχάνω in Post-Classical Greek.

3.5 Case change

There are three testimonies by Moeris on preferring older Attic suffixed cases over their diachronic replacements by prepositions which steadily started replacing them in Post-Classical Greek (Jannaris 1897: 138).50

(55) ἈθήναζεἈττικοί· εἰς ἈθήναςἝλληνες⟩. (Moeris α 52)

(to Athens) Attic: (to Athens) Hellenic.

(56) ἈθήνηθενἈττικοί· ἐξ ἈθηνῶνἝλληνες⟩. (Moeris α 53)

(from Athens) Attic: (from Athens) Hellenic.

Similarly, with the steady decline of the dative in Post-Classical Greek, the dative plural was increasingly replaced by a prepositional form, as witnessed by the following testimony from Moeris.

(57) ἈθήνησινἈττικοί· ἐν ἈθήναιςἝλληνες⟩. (Moeris α 54)

(in Athens) Attic: (in Athens) Hellenic.

By contrast, it seems that Moeris notices a case change of a verb characteristic of Post-Classical Greek only once, viz. κληρονομέω ‘inherit’ + genitive in Attic versus the Post-Classical κληρονομέω ‘inherit’ + accusative. Previously, the inherited object was expressed in the genitive, whereas in cases that did not concern inheritance specifically, viz. κληρονομέω as ‘obtain’, the accusative would be used. Thus, this remark refers to a change of case common to Post-Classical Greek (cf. Josephus AJ. 8.355 for an example of inheritance with an accusative).

(58) ἐκληρονόμησε τῆς οὐσίας Ἀττικοί· ἐκληρονόμησε τὴν οὐσίαν Ἕλληνες. (Moeris ε 37)

(he inherited) the property Attic: (he inherited) the property Hellenic.

More often with such matters, however, Moeris’ remarks would imply a diachronic contrast that is not a diachronic change but simply a synchronic Attic variant with a difference in meaning. For example, the difference between using προσεύχομαι with a dative or an accusative is that the former means ‘offer prays/vows to X’ whereas with the accusative it means ‘address X in prayer’.

(59) πρόσευξαι τὸν θεόν Ἀττικοί· πρόσευξαι τῷ θεῷ Ἕλληνες. (Moeris π 43)

(address) the god (in prayer) Attic: (offer prayers) to the god Hellenic.

4 Conclusion

I hope to have demonstrated that there is more to find for historical sociolinguists in the prescriptive Atticist lexica than their remarks have thus far been credited with. By adopting a historical sociolinguistic point of view, we were able to assess the diachronic dimension behind the social perception of Atticist prescriptivist statements, that is, to use unchanged Attic morphosyntactic structures (instead of the in their eyes bad ‘Post-Classical’ replacements). The Atticists’ disapproval of the forms/usages which characterize the changing grammar of early and middle Post-Classical Greek underlines their awareness of morphosyntactic changes in use (pace Lee 2013: 286). The types of morphosyntactic changes can be grouped (from our modern historical perspective) into: paradigmatic changes and analogical levelling, category change, category renewal, syntactic change and case change. Nevertheless, one should be aware that diachrony and social evaluation are not the only dimensions concerned here, since the rejected words belong to different registers and have different connections to certain dialects.

In addition, as highlighted in the introduction, many other testimonies could also be explained by adopting a historical linguistic point of view, for example because disapproved forms underwent sound changes in PostClassical Greek and may have changed their orthography accordingly.51 Also, it could prove illuminating to investigate the metalinguistic awareness of language change in Post-Classical Greek in other metalinguistic resources (cf. e.g. Nünlist 2012) or how Atticist norms are (re)negotiated in various later sources (see esp. Probert 2011). Since language change seeps through most parts of synchronic grammar, those that documented the grammar of the Ancient Greek language (e.g. ancient grammarians, lexicographers, teachers) all needed to decide how to explain instances of linguistic variation that were caused by language change. While some would have attributed them to the speaker, i.e. to the intellect, status, geographical origin etc. of the speaker, it stands to reason that it was only a matter of time before they started to point to time as the culprit.


I thank prof. dr. Klaas Bentein, Emmanual Roumanis, and dr. Chiara Monaco for many rewarding and enjoyable discussions on this topic. Any remaining errors are my own.


See Dickey (2007) for a guide; useful companions include Matthaios, Montanari & Rengakos 2011, Montanari, Matthaios & Rengakos 2014, and Montanari 2020.


See Matthaios 1999 and Schironi 2018 for comprehensive studies of the extent of linguistic knowledge that we can carefully infer to have been derived from Aristarchus.


See Sluiter 1990.


E.g. the discipline of grammar covering explanation of usage on the prosodic, ‘etymological’ and various other linguistic levels (Law & Sluiter 1998; Law 2003: 54–58). Early grammars transmitted on papyri also use an embryonic version of the parts of speech system familiar to us today e.g. noun, participle, article, verb, adverb (Wouters 1979; Law 2003: 55–57).


E.g. that εἴθε is a wish adverb distinctively marking a wish (Denizot 2011: 76–80), a view which goes back to Apollonius Dyscolus (la Roi 2020a: 231), even though it is now considered a fossilized conditional subordinator that marks main clause structures via insubordination (la Roi 2021).


To give an illustrative example, Aeschines gravely insulted Demosthenes by saying that he was a Scythian who spoke Greek, Aesch. 3.172: ἀπὸ τῆς μητρὸς Σκύθης, βάρβαρος ἑλληνίζων τῇ φωνῇ: ὅθεν καὶ τὴν πονηρίαν οὐκ ἐπιχώριός ἐστι (‘by his mother’s blood he would be a Scythian, a Greek-tongued barbarian—so that his knavery, too, is no product of our soil’).


For an account of how in-group and out-group thinking affected the treatment of multilinguals as reported by literary sources, see Leiwo (1996).


By contrast, histories of Latin make more extensive use of such comments (Clackson & Horrocks 2007: 190, 242, Adams 2013: 4–9; 12–22). See Adams 2013: 12–22 for a justification of the evidence which such comments can provide.


See, however, Gera 2003 for an attempt to reconstruct the theories of language held by the Ancient Greeks at various stages.


Note that such thinking affected the evaluation of language change in the Egyptian papyri for a long time as well (e.g. due to low education/heritage or bad command of Greek), but this superficial image is no longer believed (see Dahlgren & Leiwo 2020).


An example that is not noted by Lallot but does reveal the acknowledgment of language change is the treatment of fossilized imperatives by early grammarians. Apollonius Dyscolus (IICE) treats the fossilized imperative ἄγε as an adverb (Adv. 123.10–12 & Adv.128.21–22) on a par with insubordinate εἴθε in being restricted to a mood just as certain temporal adverbs cannot be used with all tenses (Synt. 3.19). Tyrannion (IBCE), a pupil of Dionysius Thrax, whose works we only possess in fragments, also suggests that ἴθι might be an imperative adverb: δύναται δὲ καὶ ἴθι ἐπίρρημα παρακελεύσεωςεἶναι ἀντὶ τοῦ ἄγεὡςτὸἴθἐκκάλυψον (fragment 6, Haas 1977) ‘(go!) can also be the adverb of exhortation instead of (come on) as in (go uncover!)’. The same classification as adverb can be found in Dionysius Thrax (D.T. 83). For the dating of the works of Dionysius Thrax, see Lallot (1998: 19–36), and Dickey (2007: 77–80) for further references. An important issue is, for example, that ancient references to the Technè tend to portray it in a more Stoic light than our version gives rise to, making it debatable whether the work as we have it was really compiled by Dionysius Thrax, a pupil of Aristarchus. For a parallel observation by the Latin grammarian Servius on the fossilized Latin imperative age, see Fedriani (2019: 74).


Another dimension of ancient sources which can provide insights into language change is their states of transmission, since different versions of a text may be explained as older vs. younger versions of a text (Cuomo 2017), stylistic rewriting to higher register which rids the text of innovative low features (Luiselli 2010; Bentein 2021) or spelling variation to match existing pronunciation (Dahlgren & Leiwo 2020).


For the historical background to Atticism, see Swain 1996 and Anderson 1993. For more compact introductions to Atticism, see Strobel 2009 and Kim 2017.


For compact introductions to these lexica, see Alpers 1990, Dickey 2007: 94–99, Strobel 2009, and Kim 2017.


See the important papers in Montanari & Richardson 1994, especially by Tosi and Schenkeveld, for relevant differences between the grammarians and Atticistic purism.


For example, Tribulato is heading an ERC project into linguistic purism as evidenced by the Atticist lexica (see and several scholars are providing new insights into the Atticist lexica, as evidenced by recent papers at the Post-Classical Greek conference 2021 by Roumanis (on register) and Monaco (on dialect). See also Monaco 2021 for detailed studies.


A recent exception is Vessella 2018 who systematically assessed their evidence for Atticist pronunciation.


I chose to focus on these lexica for matters of scope, but naturally other lexica deserve such a re-evaluation, e.g. Pollux’s Onomasticon, the Antiatticist, Philemon. The historical (socio)linguistic focus of this article also does not allow me to go into many philological details which are typically dealt with in the editions or philological studies such as matters of transmission or influence by other sources. Rather, I seek to generalize over the selected lexica for the historical linguistic utility of their prescriptivism (but for caveats, see discussion below).


This idea has a long history, but more than a century ago already faced critique by Jespersen 1894, who argued the reverse, that language change in fact constitutes progress from a linguistic standpoint.


Horrocks 2010: 141 outlines a more useful way to approach ‘mistakes’ in reference to Classical norms, viz. guided by both literary norms and subject to language change. See for earlier formulations of such an idea Schmid 1887: 4.733 and Thumb 1974: 8.


For example, in section 3.1 I discuss the condemnation of οἶδας, which is actually an old Attic form (Rutherford 1881: 227), but is condemned by Phrynichus. See also Palmer 1945: 25 who shows that Moeris wrongly classifies a Doric form διᾱνεκής ‘continuous’ as Attic.


I thus use the ancient scholars as colleagues in order to track language change. For other examples of such an approach, see de Jonge 2007 on ancient and modern notions of word order and Benedetti 2020 on Theodosius’ understanding of diathesis in the perfect.


See Fischer 1974 for Phrynichus, Hansen 1998 for Moeris, and Erbse 1950 for Aelius Dionysius. Note that Aelius Dionysius’ text has been derived from the many citations by the learned Eustathius (who still had a codex with at least lexica from Aelius Dionysius and Pausanias) and a multitude of other sources; see Erbse (1950: 7–22).


For the ambiguities of the stance towards Atticistic prescriptivism by the so-called Antiatticist (Valente 2015), see Tribulato 2021.


See note 18 above.


Note that metalinguistic awareness is also a concept that is widely used in studies on language acquisition (both L1 and L2) and generally used to describe an individual’s capacity to reflect on the nature and functions of language. See Roehr-Brackin 2018 for a complete overview. Obviously, the term is used in a slightly different vein here, as it refers to ancient direct and indirect awareness of the (changing) functions of language features.


I use the periodisation suggested by Lee 2007: 113 and applied by Bentein 2016: early Post-Classical IIIIBCE, middle Post-Classical IIIICE.


For the intricate relationship between prescriptions, sound change and orthography, see Vessella 2018.


For an overview of the changes of the gender system in Ancient Greek, see Coker 2009.


Interestingly, a similar negative evaluation is found in PS 125.3 χόλικες· οἱ πολλοὶ ἀρρενικῶς, οἱ ἀρχαῖοι θηλυκῶς where οἱ πολλοί ‘(bowels): most [say it] in masculine, [but] the ancients in feminine’ refers to contemporary usage, and in Moeris χ 12 χολάδας οἱ πρῶτοι Ἀττικοί, χόλικας θηλυκῶς οἱ μέσοιχόλικας ἑφθάς’ (Ar. Pax 717)· τοὺς χόλικας ἀρσενικῶς Ἕλληνες. ‘(bowels) [is] first Attic, [but] those in the middle (bowels) in feminine ‘boiled bowels’; (bowels) in masculine [is] Hellenic’ See Monaco 2021 for additional examples.


For references on specific paradigmatic changes in the verbal system of Post-Classical Greek, see the references provided in the discussion below.


Cf. Tribulato (2013: 207–209) who discusses how the Antiatticist deals with language change in imperative forms in α and the pluperfect. See also Thackeray (1909: 112) who criticizes Phrynichus (92) for condemning Ἀγήοχεν and proscribing ἤχασι even though the former occurs in inscriptions and biblical Greek.


Some Classical Greek examples of the analogical forms which become more frequent in Post-Classical Greek are X. Mem. (εἶπα), E. El. 13 (ἤνεγκα), E. Alc. 780 (οἶδας) and Hyp. Ath. 12.7 (ἤμην as I was).


See also Tribulato (2013: 203) who mentions some other remarks by Phrynichus that are not in line with his reputation as the strictest Atticist. Also, Menander oddly was a source for Attic in the works by Aelius Dionysius and Pausanias (Tribulato 2013: 204).


Anlauf (1960: 48–49). Strobel (2009: 102) rather suggests: “ ‘Hellenic’ here must mean the language spoken, or maybe even written, by the majority of Moeris’ contemporaries.” For earlier studies, see Maidhof (1912) and Thumb (1974: 4).


Hansen’s edition also makes note of the fact that the poorly transmitted Atticist glossographer Philemon says to avoid the new form: ἦν· μή λέγἤμην (Reitzenstein 1897).


See Monaco (2021: 47–51) for more background on this label.


Cf. Tribulato (2013: 208–214) who also discusses this form but from a different perspective.


The lexicographer Philemon says to use ἐβουλόμην instead, not ἠβουλόμην (see Reitzenstein 1897).


In the literature on grammaticalization, such change of category is called decategorialization (Hopper & Traugott 2003: 106–115).


The example was taken from Zakowski (2014) who analyses the evolution of εἰπέ μοι.


Interestingly, Spanish dime ‘tell me’ has undergone a highly similar evolution up to the point that the imperative and indirect object have been fused into one word.


See note 11 for the views from the grammarians.


See NT 1 Ep.Ti. 4.14.1, J. Ap. 2.152.1 and


A different perspective is provided by the Antiatticist, as argued by Tribulato (2021).


These syntactic remarks in Phrynichus are only shortly mentioned by Bentein (2021: 400).


See Markopoulos (2009: 54–59). In fact, he uses the comments by Phrynichus as one piece of evidence to support the authenticity of the evolution.


See Bentein 2016: 235–236 and Lorimer 1926.


In the Septuagint it occurs only once with a participle: 2 Ma. 3.9 (Jannaris 1897: 493). A Post-Classical Greek example without a participial complement of being is p.sorb.1.34 (230 BCE) and see upz.1.8 (after 161BCE) where it is combined as εἰμί with γεγονὼς, ‘[I have] been’. For the independent use of τυγχάνω in later Post-Classical Greek, see Horrocks 2010: 156.


Note that Phryn. Ecl. 66 observes a similar variation but explains it as caused by a genre difference: Ἀρχῆθεν ποιηταὶ λέγουσιν, τῶν δὲ καταλογάδην δοκίμων οὐδείς, ἀλλἐξ ἀρχῆς. ‘the poets say (from the start), but none of the approved [use it] in prose, but [use] (from the start).


For a recent investigation of the intricate relation between ancient orthography in documentary papyri and the norms of ancient grammatical treatises, see Stolk 2020.


  • Adams, James N. 2013. Social variation and the Latin language. Cambridge: Cambridge University Press.

    • Search Google Scholar