In the Mood?: Some Thoughts on Post-Classical Literary Greek

In the Mood?: Some Thoughts on Post-Classical Literary Greek Geoffrey Horrocks, Cambridge gch1000@cam.ac.uk I. Introduction Linguistic analysis of post-classical Greek has routinely focused on developments attested in sub-elite written varieties on the grounds that these are closer to ordinary contemporary speech and testify more directly to ongoing change in the language. By contrast, Atticising literary 1 Greek of the post-classical period, a revival of the supposedly ideal form of Greek represented in the work of Attic writers of the classical era (5th and 4th centuries BCE), is usually taken to be an essentially artificial variety of little interest to historical linguists. But this does not mean that there are no worthwhile questions to be asked, as I hope to show. The creative use of a dialect that no longer had native speakers clearly obliged its users to ground their efforts in the study of classical models and the use of grammatical descriptions and lexica derived from these. This had important consequences for the Greek they aspired to write and some of these will be explored in the following sections. Section II seeks to show that the Atticists’ creative use of a ‘historical’ dialect that had had no native speakers since the 4th century BCE inevitably led to a distinct variety in its own right. Though this form of Greek was firmly anchored in the past in line with the aspirations of proponents of the Atticist programme, it was in fact also subject to internally driven change and external interference from its living counterpart, the Koine. It may therefore be seen as a literary dialect similar in many respects to those familiar from earlier periods , i.e. a variety 2 without native speakers that was linguistically tied to the usage of its original model but also regularly ‘refreshed’ through active use by later writers in the relevant genres. In this particular case, however, the idealisation of classical Attic made replication of the model a matter of paramount importance, at least in theory, and deliberate linguistic innovation was largely avoided. In the hands of competent practitioners this conscious commitment to archaism successfully inhibited modification of concrete and readily accessible elements of the target language like orthography, morphology, lexicon and constructional realisation. But the abstract rules and principles that had once controlled the less easily characterised aspects of 1 2 For a recent contribution see, for example, the chapters in Rafiyenko and Seržant 2020. See Cassio 2008, Tribulato 2010. 1 Attic grammar, such as those determining syntactic well-formedness and semantic interpretation, were not directly accessible and could not be so easily replicated. Unsurprisingly, therefore, the changes that did occur can be shown for the most part to have involved a range of subconscious processes that chiefly affected these ‘invisible’ components of grammar together with the outputs they permitted. Section III then offers a detailed case study of the Atticising Greek of Dionysius of Halicarnassus (fl. ca. 20 BCE), the supposed founder and high priest of the Greek Atticist movement. This provides detailed empirical support for the argument advanced in Section II through a data-driven analysis of his use of the optative, a mood that was all but dead in the living Greek of his time but retained as a key feature of his Atticising language. II. The linguistic implications of Atticism As just noted, the Atticists’ knowledge of inherently finite sets of concrete linguistic items could in principle be sufficiently complete to produce a high level of correspondence with the practice of Attic originals. But a mastery of more abstract underlying phenomena was always going to be less straightforward. The remainder of this section will focus on why this should be so. In matters of syntax and semantics native speakers rely on a shared and intuitive understanding of well-formedness and possible meaning that is applicable to any actual or potential sentence of their language, spoken or written. While words are simply learned and reused, every non-quoted sentence is a novel creation rather than a repetition of a pre-existing entity, and a proper description of what constitutes a grammatical sentence therefore demands predictive rules of formation and interpretation that exclude ungrammatical or semantically anomalous examples. Native speakers of classical Attic, we may presume, had internalised such a rule system, which they relied on whenever they used their language in speech or writing, including the composition of literary texts. A grammar of Attic compiled for didactic purposes may be seen as an attempt to provide a comprehensive account of the dialect that replicates as closely as possible the predictive accuracy of the rules tacitly employed by (a chosen subset of) its native speakers. As we have seen, this is more or less achievable for finite domains of directly accessible items, but abstract syntactic-semantic rules are not accessible in the same way. Nonetheless, reasonable inferences about them can be drawn from data that are accessible. If a language has native speakers, grammarians can draw not only on the passive evidence of corpora but 2 also on the active judgements of speakers about the form and meaning of any possible sentence. The availability of authoritative collective decisions about whether a putative sentence really is grammatical and meaningful, or rather ungrammatical and perhaps semantically anomalous, is an invaluable additional guide in the framing of accurate predictive rules. Unfortunately, as anyone who has learned a modern foreign language in adulthood can testify, even the best of such grammars are never more than approximations to the grammatical knowledge shared by native speakers. In particular, they can never guarantee that adherence to their prescriptions will always lead to sentences that native speakers would find fully acceptable and meaningful. By contrast, the sentences of ‘corpus languages’ form a finite set that cannot be added 3 to by native speakers, and grammarians have no native-speaker judgements, positive or negative, to appeal to when developing and testing their syntactic and semantic descriptions. They can only make inductive generalisations about the surviving corpus (without even a parallel corpus of ‘bad’ sentences to help them), and hope that these will make the right predictions should anyone wish to use their grammars as the basis for renewed creative writing. But there is obviously no fully reliable authority to appeal to. If, for example, all the formal properties of a novel sentence are sanctioned by the grammatical rules and replicated in illustrative examples from the corpus, it may be presumed to be well-formed (and so potentially acceptable to former native speakers), though with no absolute guarantee. If on the other hand such a sentence has particular properties that are not supported in exactly the required form, the absence may be accidental (a different corpus might contain validating parallels) or systemic (the sentence is ungrammatical and could never be validated by any corpus), but with no sure way of knowing which is the case, or indeed whether a native speaker might have accepted it. Consequently, despite access to extensive textual evidence and attempts at summative rules based on it, the Atticists were ultimately forced to rely on their own ability to make realistic assumptions about the status of what they themselves wrote in their version of Attic. In other words, they had quietly made themselves the only possible source of the grammaticality judgements required for active composition, and their output was in practice an exemplification of a ‘neo-Attic’ dialect under their own collective control. Viewed in this light, the introduction and acceptance of some degree of innovation vis-à-vis classical Attic usage was an unavoidable consequence of the implementation of their programme. Atticising 3 The term includes languages like Hittite, that ceased to have native speakers at some point, and earlier stages of languages that continued to be spoken but in a form that had changed significantly over time, like Greek. 3 Greek could be made to look very like classical Attic in lexicon and morphology in so far as good approximations to the ‘authentic’ rules could be successfully abstracted from the corpus. But without a comparably reliable basis to constrain their approximations to the syntacticsemantic rules, the language was always going to be something other than a simple continuation of classical Attic sensu stricto. This raises the interesting question of exactly how (and how consciously) Atticist writers filled the gaps in their knowledge, and indeed of whether they were aware of the effect of such supplementation on the language they were writing. A promising way to investigate these issues is through examination of the use of forms of classical paradigms that had dropped out of contemporary speech but were still required in literary composition. Recent research 4 has suggested inter alia that, in order to provide the fully-fledged syntactic-semantic framework needed to support the creative use of such paradigms, Byzantine Atticists supplemented what they could infer from classical and earlier Atticising texts by specific reference to their grammatical intuitions as native speakers of contemporary Greek. That is to say, if the classical use and meaning of certain items had become too obscure to be recovered with confidence from the corpus, they simply recycled these as high-register realisations of current syntactic-semantic categories drawn from the same overarching domain. A good example is provided by expressions of futurity (‘what will be’) and modality (‘what may, could, must be’). The two notions were not distinguished formally or functionally in vernacular 5 medieval Greek, where context of use was the principal guide to identifying the intended degree of epistemic commitment. The fact that ancient futures, subjunctives and optatives appear in virtual free variation in Atticising historiography of the period, just like the 6 contemporary realisations of futurity and modality in vernacular texts, gives us good reason to believe that Byzantine Atticists were following the underlying medieval rules in even their most stylistically ambitious compositions. Atticising Greek in that era is therefore best regarded as a ‘semi-living’ variety of medieval Greek in which Attic words and forms were substituted as a matter of course, but the inaccessible classical rules controlling their use and interpretation were modified and even replaced, inter alia through interference from writers’ tacit control of the grammar of the contemporary vernacular. 4 See, for example, Hinterberger 2014, Horrocks 2014. Horrocks 2014, 2017a and b, 2021. 6 Free variation at least in terms of function: optatives seem to have been viewed as stylistically marked variants of futures and subjunctives, and different authors no doubt had differing preferences. 5 4 Mutatis mutandis, this approach corresponds to how people generally read texts written in earlier varieties of their native language, namely by tacitly assuming (perhaps wrongly) functional correspondence to the modern language where words and word forms look similiar, and assigning (again perhaps wrongly) a contextually plausible modern function to unfamiliar forms. In the case of later Atticists, however, this instinctive behaviour also had the long-term effect of normalising many innovations. Since they were reading not only the original Attic texts but also what other Atticists had written, and then using the language themselves in novel compositions, many such changes were reinforced over time and eventually became standard practice in the highest literary registers. The reshaping process itself was almost certainly subconscious as well as inevitable in so far as educated Byzantines appear simply to have assumed that their intuitive control of the syntax and semantics of contemporary Greek was good enough for them to be able to replicate the relevant Attic rules acceptably, i.e. that the underlying constructions, if not their surface realisations, had remained essentially the same and that all they needed to master was how to put contemporary realisations of construction X into correspondence with its realisations in classical Attic. The idea that construction X might not have been a part of classical Attic grammar in quite the same form, or indeed at all, and that they might be reinterpreting the ancient in terms of the contemporary, seems not to have been an active issue for them. For the earliest period of Greek Atticism, however, when the target language was still much closer to the educated vernacular of the time (the Koine), we might imagine that a similarly instinctive reliance on native-speaker command of contemporary Greek, even if unavoidable for the reasons just outlined, would be somewhat less obvious in its consequences. But if we bear in mind that the temporal distance between the first Atticists and their models is similar to that between us and Shakespeare, i.e. 400 years or so, we should not be surprised that significant differences are nonetheless detectable. We would not, for example, expect contemporary writers in English to be able to replicate in a consistently authentic way the more extensive Elizabethan use of the now moribund English subjunctive. A priori, then, we might reasonably expect to find evidence of departures from classical practice even in the earliest Atticist compositions. We shall in fact see below that interference from the Koine was indeed a significant factor in promoting such change, but not the only one. As with all literary languages that follow earlier models, change could also be generated on the basis of internally driven rule extensions. Both types of change will be exemplified in the analysis in section III, which is based on 5 Dionysius of Halicarnassus’ use of the then moribund optative, as attested in Book 1 of Roman Antiquities. III. Dionysius of Halicarnassus and the Optative A. Background Dionysius’s exact dates are uncertain but he was active in Rome during the reign of Augustus (27 BCE-14 CE) and, as noted earlier, is often taken to be the prime mover of the Greek Atticist movement. Whether or not this reputation is justified, the similarities between his crusade 7 against Asianism and its earlier Roman counterpart are striking, in that both treat issues of style and rhythm ahead of grammar and lexicon, the committed treatment of which belongs to the period of the Second Sophistic (and mainly to the 2nd century CE). Nonetheless, Dionysius’ articulation of a ‘back-to-basics’ programme of archaism/classicism highlights some key elements of the later Atticist agenda, and we might reasonably expect that the careful study and imitation of Athenian predecessors would inevitably lead to a more Attic-like use of language as well as to a more Attic style. In this regard Dionysius clearly stands at a turning point between the literarised Koine of writers such as Polybius (2nd century BCE) and Diodorus Siculus (mid-first century BCE) and the fully-fledged Atticised Greek of the 2nd century CE. His writing, for example, still exhibits a characteristically Hellenistic discursiveness and he is not averse to drawing on useful elements of postclassical vocabulary. But his language overall is also demonstrably more Attic than that of his predecessors in a range of characteristic features such as an increased use of the historic present and optative, the avoidance of grammatical words typical of the Koine (e.g. διό/διόπερ), and the regular replacement of innovative Koine morphology (e.g. the replacement of thematic ἱστάνω with athematic ἵστημι). At the formal level, therefore, the instantiation of Dionysius’s proto-Atticist programme seems to have involved the careful avoidance of what he took to be the most conspicuous ‘markers’ of normal Koine usage in the early Roman empire, a semi-systematic practice followed by other writers of the period such as Strabo. 8 B. 7 8 Normal uses of the optative, 1st century BCE - 1st century CE See de Jonge 2008 for a general introduction to Dionysius and his work. See Kim 2010 for a general survey of Dionysius’ contribution to the movement. 6 The combined evidence of routine documentary papyri and ordinary biblical Greek (LXX and NT) points strongly to the conclusion that the optative was all but dead in the everyday vernacular of the period, surviving in just a few residual or formulaic uses exemplified in the table in (1) under three sub-categories: 9 (1) The optative in vernacular Greek (1st century BCE - 1st century CE) (i) in main-clause wishes: e.g. μὴ γένοιτο (cf. God save the queen) (ii) in ‘potential’ sense (+ ἄν) in main clauses: e.g. καλῶς ἂν ἔχοι, καλῶς ἂν ποιήσαις (iii) in prospective/hypothetical conditional protases: e.g. εἰ τύχοι ‘if it were to happen’ (> ‘perhaps/for example’), εἰ δυνατὸν εἴη (cf. English if I were you) The genuinely ‘optative’ use in (i) was subject to active competition from subjunctives (± ἵνα) or futures, 3rd person imperatives and imperfect indicatives such as ἤθελον, ἐβουλόμην + infinitive, and the attested examples are mostly drawn from a set of fixed phrases. The ‘potential’ use in (ii) was also very restricted in its range, specifically to polite clichés, parenthetic comments and short constituent questions of the type τί/πῶς ἄν τις…; It is rare in complete conditional sentences with overt protases, and was in general being replaced by the imperfect indicative (± ἄν), or by modal verbs such as ἐδυνάμην + infinitive. And finally, the ‘prospective/hypothetical’ use in (iii) seems to have been little more than a restricted stylistic variant of ἐάν + subjunctive (‘if ever X happens…’) in that co-occurring apodoses normally contain future or present indicatives. In stylistically more ambitious texts, however, such as high-level official papyri, the more literary books of the New Testament (e.g. Luke, Acts, Pauline letters) and pre-Atticising literary compositions, there is a slightly more productive use of all three types. These same sources also attest to occasional use of the optative in subordinate clauses in the ‘iterative’ sense of indefinite frequency in the past (‘if ever in the past X did something…’), a classical function for which past indicatives (± ἄν) were already the vernacular norm. The classical ‘oblique’ optative, used optionally in a range of subordinate clauses in historic sequence, is also occasionally employed as a past-time substitute for ἐάν + subjunctive (‘X said that if Y happened…’); but it is very rare indeed in indirect statements or final, temporal and relative clauses. On this evidence, we might also suppose that the formal speech of the literate classes would also have contained a few more optatives than normal everyday conversation. 9 Cf. the discussions of the optative in standard reference works such as Blass and Debrunner 1961 and Moulton 1963 (for the NT); Horn 1926, Mandilaras 1973, Mayser 1926-1938 (for the papyri). 7 In the specific case of the literary Koine, where we have a clear diachronic perspective, there is a visibly declining arc of overall optative usage over the centuries after the classical period, as the numbers given in (2) show: (2) Approximate number of optatives per 100 pages of Loeb text (my count) for: (i) Xenophon (5-4th c. BCE: Memorabilia) 350 (ii) Polybius (2nd c. BCE) 35 (iii) Diodorus Siculus (1st c. BCE) 15 In this connection, and in line with what was said earlier, it is interesting to note that there are 89 optatives in the 154 pages of Book 1 of Dionysius’ Roman Antiquities (Loeb text), making a round figure of about 60 examples per 100 pages - a significant increase over his immediate predecessors, as might be expected of our proto-Atticist, and a figure close to that of his near contemporary Strabo, see (3): (3) Approximate number of optatives per 100 pages of Loeb text (my count) for: (i) (ii) Dionysius of Halicarnassus (ca. 60 BCE - early 1st c. CE) 60 Strabo (64/63 BCE - ca. 24 CE) 75 The key issue now is to establish which of the residual functions of the optative show the greatest increase in numbers in Dionysius’ writing and why this might be so. The distribution of the 89 optatives in Book 1 is presented in (4): (4) Optatives by type in Dionysius’ Roman Antiquities 1 (my count) (i) main clauses: wish potential (+ ἄν) (ii) subordinate clauses: prospective/hypothetical (+ εἰ) past iterative oblique 0 32 6 17 34 The absence of wishes in a historical narrative is perhaps unsurprising, even though there are a few short passages of direct speech, put into the mouths of protagonists, where a wish might in principle have occurred. The attested examples are considered in detail in subsections C, D and E under three headings according to function. We begin with main-clause potential use and the use of optatives in hypothetical conditional clauses (protases), since the two are standardly combined to form ‘remote’ conditional sentences in classical Attic. The analysis of 8 past iterative and oblique uses follows, but a preliminary consideration of the numbers in (4) suggests that these are significantly more frequent than in contemporary Koine usage. 10 C. Potential and hypothetical optatives in Dionysius’ Roman Antiquities 1 The potential use with ἄν was a weak survivor in the vernacular of the period, but somewhat better represented in written registers of the Koine. Close examination of the corpus suggests that Dionysius must have built primarily on this ‘living’ foundation, in that his examples conform to the restricted range of contemporary practice rather than to the full range of classical usage - i.e. the examples almost always lack associated protases, and are used overwhelmingly as a vehicle for brief authorial comment on the narrative (see (1)(ii) and the associated discussion). The selection of examples in (5) illustrates what seems to have been a favourite construction for Dionysius (32 examples in the corpus): (5) (i) καίτοι θαυμάσειεν ἄν τις εἰ Πλακιανοῖς μὲν … ὁμοίαν διάλεκτον εἶχον οἱ Κροτωνιᾶται… Yet one might well marvel that the Crotoniats … had a form of speech similar to that of the Placians… 1.29.3 (ii) εἴη δ᾿ ἂν Ἑλλάδι φωνῇ Θεσπιῳδὸς τῇ νύμφῃ τοὔνομα· The nymph’s name in Greek would be Thespiôdos (“prophetic singer”) 1.31.1 (iii) …χωρίον ἐξευρόντες ἐπιτήδειον, ὃ καλοῦσι Ῥωμαῖοι Λουπερκάλιον, ἡμεῖς δ᾿ ἂν εἴποιμεν Λύκαιον. …once they had found a suitable site, which the Romans call the Lupercal, but we would call the Lyceum. 1.32.4 (iv) οὐδ᾿ ἄν τις αὐτὴν φαίη πολύκαρπον μὲν εἶναι … Nor could anyone say that it is rich in crops… 1.37.1 (v) χωρὶς δὲ τῆς ἐνεργοῦ πολλὴν μὲν ἄν τις εὕροι τὴν εἰς ποίμνας ἀνειμένην … And besides the [land] that is cultivated, one would find much that is set aside for pasturage… 1.37.3 Nonetheless, there are three examples of apparently complete conditional sentences in the corpus, i.e. with εἰ + prospective/hypotheticical optative in the protasis linked to a potential optative in the main clause (apodosis). This might in principle be seen as evidence that Dionysius also had an eye on classical models in that such sentences are relatively rare in the Koine but routine in classical Attic. Two of the three cases in Roman Antiquities 1 are given in (6) (the other, in (8)(ii) below, will be discussed shortly): 10 See the relevant sections of any of the standard grammars and reference works for discussion of the uses of the optative in classical Greek: e.g. Goodwin 1889, Kühner/Gerth 1889-1904, Rijksbaron 2002, Schwyzer/ Debrunner 1959, Smyth 1956, van Emde Boas et al. 2019. 9 (6) (i) περὶ ὧν πολὺς ἂν εἴη λόγος, εἰ βουλοίμην τὴν ἀκρίβειαν γράφειν …whose story would be a long one - if ever I chose (or ?choose) to write a detailed account… 1.23.2 (ii) “… σώζεσθαί τε ὑμᾶς περὶ πολλοῦ ἂν ποιησαίμην, εἴ μοι δῆλοι γένοισθε οἰκήσεως δεόμενοι ἥκειν…” (Latinus speaks) …and I would be very solicitous for your safety - if ever it became (or ?becomes) clear to me that you have come here in search of habitation... 1.58.5 If this limited use of complete hypothetical conditional sentences was really intended as a ‘marker of intent’ for his readers, it is hard to avoid the conclusion that Dionysius might have made a little more of the option. In support of an alternative interpretation, note that the two protases here are in fact postposed and that it makes no difference if they are translated as variants of the regular prospective/generic construction of ἐάν + subjunctive (‘if X does something’): ‘whose story would be a long one, if I ever choose to write it’; ‘I would be solicitous, if ever it becomes clear that...’ Neither example in fact sets up a true hypothetical condition for the fulfilment of a potential consequence, and both serve as loose, qualifying appendages that are better construed as commentary: in (6)(i) on a cautious authorial assertion made en passant (‘whose story would be a long one - assuming I ever choose to write it’), and in (6)(ii)) on a fictitious speaker’s potential offer of kindness (‘I would be solicitous - assuming it ever becomes clear that...’) On this reading, both the clauses in these examples conform, when taken separately, to current Koine practice (see (1)(ii) and (iii) + discussion), and their combination in a way that might be taken as a deliberate Attic usage is in fact misleading. Dionysius certainly uses εἰ + optative in prospective/generic sense (i.e. just like ἑάν + subjunctive) independently of potential apodoses, as the clearly generic example in (7) confirms - a truly hypothetical reading of the optative would be anomalous here, as the gloss suggests: (7) …ἣν φυλάττουσιν ἱερὰν …, οὐδὲν ἐπὶ τὸ σεμνότερον ἐξάγοντες, εἰ δέ τι πονήσειεν … τὸ λεῖπον ἐξακούμενοι... … [a hut] which they keep holy, adding nothing to render it more stately, but repairing the loss if any part of it suffers/??were to suffer damage... 1.79.11 The same is probably also true of the two further examples in (8), where the protases are once again at least compatible with prospective/generic readings (i.e. as marked variants of ἑάν + subjunctive): 10 (8)(i) εἰ γάρ τις … διαγνῶναι βουληθείη, τίς αὐτῶν ἀρχήν τε μεγίστην ἐκτήσατο …, μακρῷ δή τινι τὴν Ῥωμαίων ἡγεμονίαν ἁπάσας ὑπερβεβλημένην ὄψεται τὰς πρὸ αὐτῆς μνημονευομένας… For if one … wants (or perhaps wanted) to determine which of them obtained the widest dominion…, s/he will see that the supremacy of the Romans has far surpassed all those that are recorded before… (ii) 1.2.1-2 καὶ εἴ τις … χώρας ἀρετὴν ἐξετάζειν ἐθελήσειεν, … οὐκ ἂν εὕροι ταύτης τινὰ ἐπιτηδειοτέραν. And if ever one wants (or perhaps wanted) to examine countries in respect of their merits, … s/he would not find any more benign than this. 1.36.1 In (8)(i) this is clearly the more likely reading on the evidence of future ὄψεται in the apodosis. And if we read (8)(ii) with its potential apodosis strictly in the classical way, it would be the sole example (of only six cases of εἰ + optative in the corpus) in which the sentence as a whole displays the regular classical pattern. On balance, then, it is probably more realistic to see (8)(ii) as similar to the examples in (6), in which the clauses individually conform to Koine practice and the apparent overall correpondence with Attic is accidental. To sum up so far, the numbers and functions of optatives attested in potential and conditional clauses suggest that Dionysius was more or less following current practice in the 11 higher varieties of the written Koine, despite a few potential counter-examples that turn out on closer examination to be almost certainly illusory. In other words, his Atticising Greek in these specific respects was subject to the relevant rules and conventions of contemporary Greek, despite the strictly Attic orthography and morphology deployed throughout: Koine grammar was evidently the basis for neo-Attic grammar as far as the distribution of potential and hypothetical optatives are concerned. This leaves the uses of the optative in subordinate clauses to denote indefinite frequency in past time (iterative), and optionally to mark subordination in past-time environments as a replacement for indicatives or subjunctives in the relevant constructions (oblique). As noted 12 above, these entirely normal classical uses were all but unknown in the contemporary vernacular and very limited even in the most polished written registers of the Koine, where we may presume they had an archaising, and perhaps rather pretentious, ‘feel’ in the absence of 11 At most, one might conclude that he was pushing the numerical boundaries in his favourite use of the potential optative for authorial comment. 12 Despite continuing efforts to explain and motivate this rather strange convention, grammarians and linguists have consistently struggled to provide entiurely convincing accounts that do not require special pleading when confronted with examples chosen at random rather than as exemplifications of their various theories. See Aikhenvald 2003 for a general discussion of evidentiality, which plays a key role in some accounts, and Cristofaro 1996, Faure 2009, Jacquinod 1999 (relevant papers), Mendez Dosuna 1999, Neuberger-Donath 1983, Rijksbaron 2002, Van Emde Boas et al. 2019, 509-10, for a range of views. 11 even vestigial living roots. The high numbers in the corpus (17 and 34 examples respectively, see table (4)) are therefore likely to be significant. Though this experiment in the active revival of moribund uses of the optative might have offended some contemporary speakers on the grounds of taste, it is a reasonable assumption that Dionysius chose to extend these functions rather than those in residual vernacular use (cf. subsection B) precisely because they were an unambiguous indication that his prose was intended to be read as an instantiation of his Atticist agenda. Nonetheless, Dionysius and like-minded writers might well have tacitly relied on their own contemporary grammatical knowledge in order to shape the interpretation of the relevant constructions and thereby the basis for the reintroduction of their Attic realisations. This issue will be pursued inter alia in subsections D and E. D. Iterative optatives in Dionysius’ Roman Antiquities 1 The imperfect indicative (past imperfective) in classical Attic was used to present actions as ongoing/progressive (‘was doing’) or habitual/iterative (‘used to do’) in the past. In the latter case, the particle ἄν could be added optionally, albeit rarely, to make the iterative sense explicit; the same particle could also be added to an aorist (past perfective) indicative with the same meaning. But the iterative use of past indicatives with ἄν is in practice very 13 limited, perhaps because the same combination was used routinely to indicate hypotheticality/ counterfactuality (‘would do/would have done’), and the notion of repeated action in the past is routinely expressed by the bare imperfect, at least in main clauses. In subordinate temporal, conditional and relative clauses, however, the imperfect appears to have been used almost exclusively in its progressive function, and the iterative meaning was expressed instead by the optative (again perfective or imperfective). The two 14 headless relative clauses in (9) illustrate the normal classical usage: (9)(i) ὅσα ἔλεγε (imperfect indicative)… ‘whatever s/he was saying…’ (at a specific time) ‘anything/everything s/he was saying…’ (ii) ὅσα λέγοι/εἴποι (optative)… ‘whatever s/he said…’ (at any past times) ‘anything s/he (might have) said…’ 13 Since iterative/habitual action involves an open-ended sequence of individually complete events, it may be expressed by either aspect so long as the perfective is overtly marked in some way as denoting repetition. 14 Note that (9)(ii) can also be translated as ‘anything/whatever s/he might have said (at any past times)…’. From this perspective the optative is well motivated if it is taken to denote not actual repetitions of an action at specific times in the past but rather assumed ‘example’ occurrences of a type of event that was known to have taken place repeatedly over a period but with unknown frequency and at unknown temporal locations. 12 By Dionysius’ time, however, imperfect indicatives had all but replaced optatives in the relevant subordinate clauses, making these ambiguous between progressive and iterative readings - though the Koine still allowed for a distinction to be drawn through the optional addition of ἄν to a past indicative to enforce the iterative meaning, cf. (10)(i) and (ii): (10)(i) ὅσα ἔλεγε … (ii) ὅσα ἂν ἔλεγε/εἶπε … ‘whatever s/he said…’ (progressive or iterative) ‘whatever s/he said…’ (iterative only) Since the imperfect in (9)(i) corresponded formally to the imperfect in (10)(i), it would have been natural for Dionysius to assume that they also corresponded functionally (just as they did everywhere else), and that the optative construction in (9)(ii) was simply the ‘explicit’ correspondent to (10)(ii) (i.e. the version with ἄν), a construction unavailable in Attic in this environment. On this basis, the ‘unexpected’ absence of iterative imperfects in these contexts in Attic might have been seen as following from the inherent ‘modality’ (or ‘conditionality’) of referring to iterated actions that are known to have taken place in the past but at unknown times and at unknown intervals, i.e. events that are merely assumed to have occurred at some unspecified past times as a basis for continuing the argument (see footnote 14, and cf. Goodwin 1889, 122-126, Wakker 1994). 15 If this set of assumptions is correct, we might expect to find occasional unclassical iterative imperfects where iterative optatives should have been used, and this is indeed the case. Though the optative is used in all other relevant examples of adverbial and relative clauses in the corpus, there is just one example of an iterative imperfect. Consider the pair of relative clauses in (11): (11)(i) ἐλάμβανέ τε καταθέων ἐκ τῶν πέριξ χωρίων ὁπόσα εἰς τὸν πολισμὸν αὐτῷ ἦν χρήσιμα. And running down from the places round about, he used to take whatever was useful to him for building. 1.57.1-2 (ii) ᾐτιῶντο δὲ ἀλλήλους ἑκάτεροι θαμινὰ ἢ τὴν μὴ προσήκουσαν ὀργάδα κατανέμειν … ἢ ὅ τι δήποτε τύχοι. They often used to accuse one another of grazing the meadowland that did not belong to them … or of whatever there happened to be (sc. as an excuse for an accusation). 1.79.12 15 The use of the habitual/iterative imperfect in corresponding main clauses is unproblematical, however, in that the time reference is now specific: ‘on any occasion(s) when X might have occurred, on each of those occasions Y used to take place’. 13 Both examples have main clauses with imperfects describing habitual actions, and both contain headless relative clauses denoting unspecified entities that function as arguments of the main verb. Yet the relative clause in (11)(i) has an imperfect indicative, contrary to regular classical usage, while that in (11)(ii) contains an optative as expected. Note in particular the inherent modality/conditionality of the indefinite construction in (11)(i): ‘on regular occasions he used to run down and, if there were anything useful for building on those occasions, he used to take it’. We may perhaps conclude, albeit tentatively until further examples are found in a larger corpus, that Dionysius intended to follow classical usage throughout, but that his understanding of the rules in the light of the Koine occasionally led to a contemporary-looking realisation of the underlying iterative construction based on the ambiguity of the imperfect in (10)(i) (i.e. the construction without ἄν, since classical Attic did not allow this to be combined with past indicatives in subordinate clauses). E. Oblique optatives in Dionysius’ Roman Antiquities 1 There were not only no surviving traces of the oblique optative in the vernacular but no functionally equivalent Koine construction either. Nonetheless, the principle that the optative could be used optionally in place of an indicative or subjunctive in a subordinate clause dependent on a past-referring main verb was successfully accessed and then applied with considerable freedom in Atticising compositions. But the failure of modern grammarians and linguists to come up with a fully convincing account of how this use relates to any other uses of the optative or of how the oblique optative might have differed functionally from the indicative/subjunctive in the relevant contexts (see footnote 12), rather discourages the view that Dionysius, having cracked the mysteries of 5th/4th-century native-speaker competence in this domain, was in full control of what he was doing. On the contrary, we might expect an overall impression of control to be marred by occasional evidence of adaptation of the rules. In the absence of any realistic basis for Koine interference in this case, this would have to be the result of some internal analogy or generalisation. Consider first the examples in (12), in which optatives are used in conjoined pairs of leading subordinate clauses in primary (i.e. non-past) sequence after ὡς: (12)(i) ἔστι δέ τις καὶ ἕτερος λόγος ὑπὸ τῶν ἐπιχωρίων μυθολογούμενος, ὡς … ὁ Κρόνος ἐν τῇ γῇ ταύτῃ δυναστεύσειε, καὶ ὁ λεγόμενος ἐπ᾿ ἐκείνου βίος ἅπασι δαψιλὴς … γένοιτο. There is another legendary story related by the inhabitants, that … Saturn was lord in this land and that the fabled lifestyle of his reign was abundant for all. 1.36.1-2 14 (ii) λέγεται … ὡς καὶ δέησίν τινα ποιήσαιτο τῶν ἐπιχωρίων, … ὅπως ἀθανάτους αὐτῷ διαφυλάττωσι τὰς τιμάς…· καὶ ὡς διδάξειεν αὐτὸς τὰς θυσίας, ἵνα διὰ παντὸς αὐτῷ κεχαρισμένα θύοιεν... It is … said that he made a request of the inhabitants … to perpetuate the honours they had paid him…; and that he himself taught the sacrificial rites …, in order that their offerings might always be acceptable to him. 1.40.3-4 Elsewhere in the corpus, however, examples with a non-past main verb follow the regular rules of primary sequence, as in (13): (13) ὁ μὲν οὖν μυθικὸς … λόγος ὧδ᾿ ἔχει· ὡς δὴ … Ἡρακλῆς … ἀφίκετο … εἰς τὸ προσεχὲς τῷ Παλλαντίῳ χωρίον. The legendary … story is as follows: that Hercules … arrived in the neighbourhood of Pallantium. 1.39.1-2 where the relevant story is presented in the here and now and ὡς is followed by an indicative in primary sequence, as expected. In (12), however, the stories in question are not strictly speaking reported in present time, and the meaning is rather a temporally generic one: ‘there is a story (whose content has often been repeated in the past as well as reported now) that…’ But the generic present too was standardly treated as a primary tense in Attic, and classical examples like those in (13) routinely have indicatives in the subordinate clause: (14)(i) λόγος τις ἅμα καὶ φήμη ὑπορρεῖ πως, ὡς ὁ θεὸς οὗτος…διεφορήθη τῆς ψυχῆς τὴν γνώμην. There is a story-cum-rumour that somehow surreptitiously makes the rounds that this god [Dionysus] … was stripped of his soul’s judgement. Plato, Laws 672.b.3-4 (ii) ἕστι τοίνυν τις εὐήθης λόγος … ὡς ἄρ’ οὔπω Φίλιππός ἐστιν οἷοί ποτ’ ἦσαν Λακεδαιμόνιοι. There is then a simple-minded story … that Philip is not yet the sort of man that the Spartans once were. Demosthenes, Philippic 3.47.1-3 As a first step to explaining why Dionysius treated such cases differently, we should note that there are a handful of possible classical parallels to (12) that might have influenced his usage. Consider (15): (15) ἆρ’ οὖν δὴ οὐ μετρίως ἀπολογησόμεθα, ὅτι πρὸς τὸ ὂν πεφυκὼς εἴη ἁμιλλᾶσθαι…, καὶ οὐκ ἐπιμένοι ἐπὶ τοῖς δοξαζομένοις εἶναι πολλοῖς ἑκάστοις, ἀλλ’ ἴοι καὶ οὐκ ἀμβλύναιτο οὐδ’ ἀπολήγοι τοῦ ἔρωτος… Shall it not then be a fair defence on our part, that (sc. as we have previously said)… he was born to struggle towards reality and did not linger among the many individual things that are believed to exist, but moved on and did not lose heart or cease from his passion …? Plato, Republic 490.a.8-b.2 Here too a non-past main verb is followed by a set of leading subordinate clauses containing optatives. The customary explanation is that the defence in question is tacitly understood to 15 have been made on a number of past occasions as well as this one – i.e. that this is a repetition of a past defence made in the present for the future. There is, then, perhaps enough common ground between the very rare examples like (15) (assuming Dionysius had spotted them) and those in (12) for our writer to have made a connection and extended the optative construction to a context where it would not ordinarily have been used in classical Greek. Note in particular that this key marker of an unusual reading is carried through consistently across the conjoined leading statements after ὡς/ὅτι in both cases. Whatever the truth of the matter, this usage was subsequently extended to other types of generic statement that did not directly include such non-specific reference to the past, and the practice ultimately became the norm for the later Atticists of the Second Sophistic, as in (16): (16) οὐ γὰρ τόν γε τοῦ παντὸς ἐσφαλμένον ὡς ἀνέλοι σοφώτατον ἀνθρώπων πιστεῦσαι θεμιτὸν περὶ τοῦ θεοῦ. For it is not right to believe of the god [sc. Apollo] that he identified one who had failed in everything as the wisest of men. Aelius Aristides, To Plato: In Defence of Oratory 78-9 where the sense is ‘it is not right, now or ever, to believe…’ IV. Conclusions We have considered four different functions of the optative in Book 1 of Dionysius’ Roman Antiquities, and in three cases have found likely evidence of interference from the contemporary Koine in shaping their distribution and use. In the fourth, there was probable evidence instead of a wider range of usage being based on the analogical extension of a limited and poorly understood classical convention. The picture that has emerged from late Byzantine historiography is of a literary language shaped not only by fixed classical precedent but also by innovative internal generalisations and external interference from the ever-changing vernacular, especially in abstract areas of grammar. As understanding of the constructions of classical Greek became more tenuous, the ability to question assumptions about them made in part on the basis of native-speaker understanding of the contemporary language diminished, and Attic syntax was subconsciously adapted at the constructional level, if not at the level of realisation. In the light of the current discussion, it seems that these same factors were already relevant to literary Greek in the era of the Atticist programme as formulated by Dionysius of Halicarnassus, albeit with the difference that the classical language was still close enough, and 16 presumably well enough understood, to constrain more tightly this kind of subconscious developmental process. We should not be surprised that this is so, given the time interval between Dionysius and the classical period and bearing in mind the difficulty of fully grasping the underlying syntactic and semantic rules of classical Attic in the absence of native speakers. Both internally motivated change (analogical) and externally motivated change (involving interference from contemporary Greek) were inevitable from the start, and it is clear that Atticising Greek, even in its earliest manifestations, could never be a true replica of its model whatever the intentions of its proponents. But if Atticising Greek is viewed instead as a steadily evolving (or semiliving) literary language, the contemporary vernacular may fruitfully be seen as the source of many of the detectable deviations from classical practice. We should never forget that, for the fully literate and even semi-literate populations, spoken and written varieties were in constant interaction, and that the history of the Greek language cannot properly be told on the basis of one to the exclusion of the other. Interference evidently took place in both directions, and sustained forms of interference could eventually be normalised in the relevant registers. Bibliography Aikhenvald, A.Y. (2003), “Evidentiality in typological perspective”, in: A.Y. Aikhenvald A., and R.M.W.Dixon (eds.), Studies in Evidentiality, Amsterdam & Philadelphia, 1-31. Blass, F., and A. Debrunner (translated and edited by R. W. Funk) (1961), A Greek Grammar of the New Testament and Other Early Christian Literature, Chicago. Cassio, A.C. (ed.) (2008), Storia delle Lingue Letterarie Greche, Florence. Cristofaro, S. (1996), Aspetti Sintattici e Semantici delle Frasi Completive in Greco Antico. Florence. Dickey, E. (2011), “The Greek and Latin languages in the papyri”, in: R. Bagnall (ed.), The Oxford Handbook of Papyrology, Oxford and New York, 149-169. de Jonge, C.C. (2008), Between Grammar and Rhetoric: Dionysius of Halicarnassus on Language, Linguistics and Literature. Mnemosyne Supplements 301, Leiden and Boston. Faure, R. (2009), “L’optatif oblique serait-il un temps?” CLELIA (Colloque de Langue et de Littérature Antiques), Aug 2009, La Baume-lès-Aix, France. hal-00574471, 281-294. Goodwin, W.W. (1889), Syntax of the Moods and Tenses of the Greek Verb, London. Hinterberger, M. (ed.) (2014), The Language of Byzantine Learned Literature, Turnhout. Horn, R.C. (1926), The Use of the Subjunctive and Optative Moods in the Non-Literary Papyri, Ph.D. Diss., Philadelphia. Horrocks, G.C. (2014) “High-register Byzantine Greek: diglossia and what lay behind it”, in: C. Carpinato and O. Tribulato (eds.), Storia e storie della lingua greca, Venice, 4972. Horrocks, G.C. (2017a), “Georgios Akropolitis: theory and practice in the language of later 17 Byzantine historiography”, in: A.M. Cuomo and E. Trapp (eds.), Towards a Historical Sociolinguistic Poetics of Medieval Greek, Turnhout, 109-118. Horrocks, G.C. (2017b), “ ‘High’ and ‘low’ in Medieval Greek”, in: K. Bentein, M. Janse and J. Soltic (eds.), Variation and Change in Ancient Greek: Tense, Aspect and Modality, Leiden and Boston, 210-241. Horrocks, G.C. (2021), “Byzantine literature in ‘classicised’ genres: some grammatical realities (V-XIV CE)”, in: K. Bentein and M. Janse (eds.), Varieties of Post-Classical and Byzantine Greek, Berlin and Boston, 163-178. Jacquinod, B. (ed.). (1999), Les complétives en grec ancien: Actes du colloque international de Saint-Etienne (3-5 septembre 1998). Saint-Etienne. Kim, L. (2010), “The literary heritage as language: Atticism and the Second Sophistic”, in: E.J. Bakker (ed.) A Companion to the Ancient Greek Language, Chichester and Malden Mass., 468-82. Kühner, R., and B. Gerth. (1889-1904) Ausführliche Grammatik der Griechischen Sprache, 2. Teil: Satzlehre, Hannover. Mandilaras, B. G. (1973), The Verb in the Greek Non-literary Papyri, Athens. Mayser, E. (1926–1938), Grammatik der Griechischen Papyri aus der Ptolemäerzeit, Berlin. Méndez Dosuna, J.V. (1999), “La value de l’optatif oblique grec: un regard fonctionneltypologique”, in: B. Jacquinod (ed.), Les complétives en grec ancien: Actes du colloque international de Saint-Etienne (3-5 septembre 1998), Saint-Etienne, 331352. Moulton, J.H, (W.F. Howard and Turner N.) (1963), A Grammar of New Testament Greek, Vol. III Syntax, Edinburgh, 118-133. Neuberger-Donath, R. (1983), “Die Funktion des Optativs in abhängigen Aussagesätzen”, in: H. Shiro & I. Kazuko (eds.), Proceedings of the XIIIth International Congress of Linguistics. Tokyo & The Hague, 715-718. Rafiyenko, D., and I.A. Seržant (eds.) (2020), Postclassical Greek: Contemporary Approaches to Philology and Linguistics (Trends in Linguistics. Studies and Monographs [TiLSM], 335), Berlin and Boston. Rijksbaron, A. (2002), The Syntax and Semantics of the Verb in Classical Greek: an Introduction, 3rd edn., Amsterdam, Chicago. Schwyzer, E. and A. Debrunner (1959), Griechische Grammatik, 2 vols., Munich. Smyth, H.W. (1956), Greek Grammar (Revised by G.M. Messing), Cambridge, MA. Tribulato, O. (2010) “Literary dialects”, in: E.J. Bakker (ed.), A Companion to the Ancient Greek Language. Chichester and Malden MA, 388-400. Van Emde Boas, E, A Rijksbaron, L Huitink, and M de Bakker (2019), The Cambridge Grammar of Classical Greek, Cambridge, 438-446, 495-501. Wakker, G.C. (1994), Conditions and Conditionals: an Investigation of Ancient Greek, Amsterdam. Index Rerum 18 Atticism: linguistic implications for creative writing Dionysius of Halicarnassus: role in development of the Atticist programme subconscious linguistic innovation use of language use of the optative Linguistic interference: Attic-Koine Literary dialects Mood and modality: indicative subjunctive optative: hypothetical/prospective iterative oblique potential Index Locorum Aristid., Or. 2 (Ad Platonem), 78-79 Dem., Or. 9 (Philippica 3).47.1-3 Dion. Hal., Ant. Rom.: 1.2.1-2, 1.23.2, 1.29.3, 1.31.1, 1.32.4, 1.36.1-2, 1.37.1, 1.37.3, 1.39.1-2, 1.40.3-4, 1.57.1-2, 1.58.5, 1.79.11, 1.79.12 Pl., Leg. 672.b.3-4, Resp. 490.a.8-b.2 19