There are some reasons to assume that δέ is procedural. The first is that its meaning seems inaccessible to consciousness. Of course, we have no access to native speaker intuitions for Ancient Greek, but earlier approaches to δέ’s meaning seem to point to
trouble in getting at its exact meaning. In fact, most analyses describe its meaning in terms of its (discourse) function, and not in terms of its semantic meaning – in other words, there is no obvious answer to the question what δέ means. According to Bakker, δέ points to discontinuity, but he does not provide a ready-made translation; the same is true for Black’s ‘low- to mid-level discontinuity’, and Levinsohn’s and Larsen’s description of its function in terms of ‘change’. Denniston (and others) do provide a translation for δέ, but one of his options (‘but’) is procedural as well, and the other (‘and’) is problematic, as it applies more straightforwardly to καί.16 Another oft-used
option is to leave δέ untranslated (Black 2002: 142), which does not help us at all. The authoritative LSJ dictionary, finally, does not provide a translation for δέ either – again, it is described in terms of what it does, whether that is ‘to introduce a proof’, ‘to begin a story’, ‘expressing opposition’ or ‘to resume after an interruption’ (s.v. δέ). This is not to say that translating conceptual items is easy, but even conceptual terms which are notoriously hard to pin down (such as λόγος) are provided with translations in the LSJ – as opposed to δέ, which is described operationally. Interestingly, the fifth-century lexicographer Hesychius does not discuss δέ, γάρ or οὖν in his exhaustive dictionary – although he does include items which he designates as σύνδεσμοι, such as δή, τοί and τοιγάρ. The only mention of any of the three DMs under consideration in this dissertation comes under Hesychius’ definition of δή, which he regards as synonymous with οὖν.17 This is not in any way to be construed as conclusive evidence that Hesychius
had trouble translating δέ (or γάρ or οὖν), but it is an interesting lacuna.
Moving on to non-compositionality, it is clear that δέ can occur with other items which are commonly regarded as DMs. The question is whether they combine to form a larger representation.
(30) τί μηχανᾶται, καὶ τί ποιεῖ; Μιμεῖται Ῥαψάκην τὸν Ἀσσύριον· ἐστρατήγει δὲ ἄρα τῷ βασιλεῖ τῶν Ἀσσυρίων Σεναχηρείμ. (Greg. Iul. I.648.11-14)
[Gregory is recounting how Julian attempted to win over the Jews in Jerusalem:] “What does he plan, and what does he do? He follows the example of Rabshakeh the Assyrian – he was a general [de ara] of Sennacherib, king of the Assyrians.”
There are similar instances in Heliodorus – δέ is located in second position along with ἄρα in an utterance where a person who was just introduced is identified further (see Hel. Aeth. 4.16.3.9; 5.8.2.2). Although δέ and ἄρα occur together, there is no reason to
16 Additionally, some scholars have argued that and is at least partly procedural (cf. Moeschler 2016: 123; Mauri
& van der Auwera 2012).
assume that they combine to fulfill a more complex function which they could not on their own – in (30), δέ still plays its normal role of indicating ‘continuity-within- discontinuity’ by marking an utterance which provides information about the previously introduced Rabshakeh (continuity), while also deviating from the main narrative line (discontinuity), in which Julian was compared to this Rabshakeh. Put differently, while the previous context is still relevant for the interpretation of the δέ segment, the δέ segment is also clearly discontinuous in that it is no longer concerned with Julian directly (note also the syntactic break – μιμεῖται is a present indicative, while ἐστρατήγει is an imperfect indicative). There is no reason to assume that this function changes due to the presence of ἄρα, which is usually taken to signal some sort of inference (Des Places 1929: 318; Sicking 1997: 173) or a comment by the speaker on the noteworthiness of the information she is presenting (Denniston 1954²: 33; Wakker 1997: 227).18 In (30), as in the examples in Heliodorus, it seems to work in this latter
sense, identifying an interesting fact about the character under consideration. At any rate, there is nothing to suggest that δέ and ἄρα function differently in light of each other, and their presence can be explained by the meaning of the DMs on their own – this should be considered the preferred method of doing semantics if we follow Grice’s Modified Occam’s Razor.
(31) Ὁρῶ μὲν γὰρ ὑμᾶς δακρύοντας καὶ ἀνθρώπινόν τι πάθος ἀναδεδεγμένους καὶ ἐλεοῦντας μὲν τὴν ἀωρίαν τῆς κόρης ἐλεοῦντας δὲ καὶ τὴν ἐμοὶ μάτην προσδοκηθεῖσαν τοῦ γένους διαδοχήν· ὅμως δ’ οὖν ἀνάγκη, καὶ ὑμῶν ἴσως μὴ βουλομένων, τῷ πατρίῳ πείθεσθαι νόμῳ, τῶν ἰδίων λυσιτελῶν τὸ τῆς πατρίδος ἐπίπροσθεν ποιούμενον. (Hel. Aeth. 10.16.5.1-6)
[Hydaspes, having been skeptical that Chariclea is his long-lost daughter, is finally convinced of the truth. However, his plan to sacrifice her to the Ethiopian gods, which dated from before he recognized her as his daughter, is still in place, as he himself confirms:] “I see you weeping and yielding to an impulse of human emotion in your pity for the untimely fate of this girl, and also in your pity for the frustration of my hopes for the succession; yet it is still [de oun] necessary, though it be perhaps against your will, that I obey our country’s law and set the claims of our fatherland above any personal advantage.”
In this instance, δέ occurs together with another DM under consideration in this dissertation, οὖν. However, there is again no reason to assume that they do not function
18 On a related note, it is high time that we get a more up-to-date analysis of ἄρα – see George (2015; fc.) for an
separately – δέ is balanced by the previous utterance, which is marked at its beginning by μέν. We get a concession here, marked clearly by the addition of ὅμως (usually translated as ‘nevertheless’ or ‘still’) in the δέ segment – the assumption in μέν could lead to the conclusion that Chariclea should not be sacrificed if it (a) provokes such an emotional response in the king’s retinue, and (b) cuts off the royal bloodline, but this conclusion is precluded by the δέ segment. In (31), then, δέ marks the suspension of an inference which could follow from the assumption in μέν – however, δέ as such only marks the assumption as being discontinuous against the background of the assumption in μέν, and the suspended inference is brought out by ὅμως (see Iten (2005: 179-183) on although, which is similar to ὅμως). As we will see, οὖν encodes the information that the assumption it marks is somehow mutually manifest already, i.e., is already part of the cognitive environment shared by the speaker and her audience. In this case, Hydaspes makes clear his intention to obey the laws (πείθεσθαι νόμῳ) and place his fatherland above his personal feelings (τῶν ἰδίων λυσιτελῶν τὸ τῆς πατρίδος ἐπίπροσθεν ποιούμενον) – but in the utterance preceding (31), Hydaspes had already stated that he intended to go through with sacrificing Chariclea (θεοῖς ἱερουργεῖν ὑπὲρ ὑμῶν ἐπείγομαι, or ‘I hasten to perform the rites for the gods on your account’ – Hel. Aeth. 10.16.4.8). As such, the assumption communicated in the segment marked by οὖν is already mutually manifest, based on what Hydaspes said before. As in (30), the DMs occur together, but function separately.19
(32) Ἄρκτος πολλάκις βαθυτάταις κατατρωθεῖσα πληγαῖς, ἑαυτὴν ἰατρεύει, πάσαις μηχαναῖς τῷ φλόμῳ τούτῳ ξηρὰν τὴν φύσιν ἔχοντι τὰς ὠτειλὰς παραβύουσα. Ἴδοις δ’ ἂν καὶ ἀλώπεκα τῷ δακρύῳ τῆς πίτυος ἑαυτὴν ἰωμένην. (Bas. Hex. IX.3.30-34)
[Basil argues that good Christians should take care of their souls, just as wild animals take care of themselves:] “The bear, which often gets severely wounded by very heavy blows, cares for himself and cleverly fills the wounds with mullein, a plant whose nature is very astringent. You could see [de an] the fox too heal his wounds with droppings from the pine tree.”
The DM ἄν marks the proposition as being either potential or unreal, i.e., as being non-real (the exact nature of that non-reality is specified by the mood and tense of the main verb – in this case, we get a potentialis) (Zakowski 2014d). Again, there is no reason to assume that δέ and ἄν combine to form a complex which is more than the sum of its parts – δέ marks the utterance as being discontinuous (the narrator transitions from a bear to a fox) but also continuous (the fox, like the bear, is presented as an
example of an animal which is able to take care of itself); ἄν marks the proposition as non-real. (33) Καὶ ὑπεξελθοῦσα τὸν Ἀρίστιππον ἔνθα προείρητο καταλαμβάνει, καὶ δεσμεῖν ἐπιστάντα τὸν μοιχὸν ἤπειγεν· ὁ δὲ εἵπετο καὶ ἐπιστὰς εἰστρέχει τε εἰς τὸ δωμάτιον καὶ τὴν κλίνην πρὸς μικρὰν τῆς σεληναίας αὐγὴν χαλεπῶς ἀνευρὼν «ἔχω σε» εἶπεν «ὦ θεοῖς ἐχθρά.» (Hel. Aeth. 1.17.3.4-9)
[Thisbe is plotting against her mistress Demaenete. She has convinced Demaenete to lay in bed and wait for the son of her husband, Aristippus – Demaenete is madly in love with her stepson. In fact, Thisbe goes to fetch Aristippus, tells him of Demaenete’s plan to cheat on him with an adulterer (whom she does not name) and leads him to the bedroom, acting as if the ‘adulterer’ is in bed with Demaenete:] “Thisbe slipped out, picked up Aristippus at the appointed place and incited him to bind the adulterer, who he would lie in wait for. He [de] followed her and suddenly dashed into the room. Making his way with some difficulty to the bed by a slight glimmer of the moon he said: ‘I have caught you, accursed woman!’”
Examples like (33) may seem to form the basis of a compelling case against the non- compositionality of δέ. As Bakker (1993: 293) points out, δέ had, by Homeric times, formed a collocation with the “unmarked anaphoric pronoun” ὁ (which eventually developed into the definite article). Without the addition of δέ, ὁ cannot occur as an anaphoric pronoun (as it does in (33)) – hence, δέ seems to form a larger representation with ὁ. However, two points should be made here. The first is that δέ still performs its usual function, i.e. marking continuity-within-discontinuity – in (33), it marks a shift in perspective to Aristippus, but the δέ segment still has to be seen against the backdrop of the previous utterance (Aristippus reacts to and follows Thisbe’s incitement to ‘surprise and bind the adulterer’). Bakker (ibid.) argues that this function harmonizes perfectly with switch-reference (or, in this case, perspectival discontinuity) – in other words, with a topic switch. As such, a constituent which marks a topic switch (ὁ) induces the use of δέ, while δέ is, of course, ideally suited to contexts containing a topic switch. But harmonization is not the same as composition, and the development of a collocation does not necessarily point to compositionality – there is no reason to suggest that δέ’s function is different in (33) than it is in, for example, (32). If anything, the function of ὁ changes (it cannot be used as a pronoun except with δέ), but that of δέ does not. The second point, then, is that δέ is still a marker in (33) and not itself compositional – compare (33) to Bezuidenhout’s examples in (52) in the previous chapter, where the DMs but and so were themselves the subject of a compositional process containing
conceptual items (‘but, and this is a big but’ … ‘so, although perhaps not obviously so’). This does not happen with δέ, even in examples like (33) – δέ does not combine with other elements, procedural or conceptual, to form a larger representation.20
The same is true for δέ’s use in conjunction with μέν. As we will see, μέν encodes the information that the assumption it marks will be relevant for the interpretation of an upcoming segment (which is usually marked by δέ). But δέ’s function remains the same – it signals continuity-within-discontinuity, with the assumption marked by μέν being the backdrop against which the information in the δέ segment has to be interpreted. As such, the assumption marked by μέν amounts to the ‘continuity’ component of the continuity-within-discontinuity marked by δέ – but it is marked by μέν, and not δέ.
The conclusion here should be that δέ is non-compositional. It does occur with other DMs, and does induce a unique use of ὁ, but, in the first case, δέ and the other DM function independently; in the second case, δέ still fulfills its basic function of indicating continuity-within-discontinuity – it does not combine with ὁ to form a complex constituent, but modifies the interpretation of ὁ.
Rigidity and non-polysemy are also characteristic of procedural items. These properties, however, are more difficult to assess for a dead language, as I pointed out in the previous chapter. Rigidity means that there is no possibility for the meaning of δέ to be influenced by the context of the utterance. There are some examples where δέ seems to preclude an interpretation which seems more straightforward:
(34) οὔτε γὰρ πᾶσι τὰ αὐτὰ, οὔτε ἑνὶ τὰ πάντα, οὔτε τὸν αὐτὸν τρόπον, ὡς τοῖς παρ’ ὑμῖν ἱεροφάνταις δοκεῖ, καὶ τοῖς τῶν θυσιῶν τεχνολόγοις. Ποῦ δὲ, ὥσπερ Λινδίοις, εὐσεβὲς τὸ καταρᾶσθαι τῷ Βουθοίνᾳ, καὶ τοῦτο εἶναι θεοῦ τιμὴν, τὰς εἰς αὐτὸν λοιδορίας; ἢ Ταύροις τὸ ξενοκτονεῖν, ἢ Λάκωσι τὸ ἐπιβώμια ξαίνεσθαι, ἢ Φρυξὶ τὸ κατατέμνεσθαι ὑπ’ αὐλῶν κηλουμένους; (Greg. Iul. I.640.1-8)
[Gregory is trying to draw a distinction between language and religion; at this point of his argument, he is demonstrating how there are many different sacrificial customs.] “Since all nations have not the same doctrines, nor any single one the sole possession of them; nor yet the same ceremonial, as it is laid down by your own sacred interpreters and directors of sacrifice. Where [de], as with the Lindians, is it a religious action to curse the Bull-eater, and is this a way of doing honour to the god, namely, the reviling of him? Or where, as with the Tauri, is it pious to sacrifice strangers; or as with the Laconians, to be flogged upon the altar? Or as with the Phrygians, to castrate yourself when bewitched by the sound of the flutes?”
In this case, we get an exemplification – the customs alluded to in the utterance preceding δέ are now given form by dint of an enumeration of strange, exotic ways of honoring the gods. As we will see, exemplifications are usually marked by γάρ (§4.2.3.3). Here, however, we get δέ, which points to a different inferential path for deriving positive cognitive effects – the upcoming utterance will noteworthily diverge from the previous one (we are transitioning to concrete examples of sacrificial customs), but the existing cognitive environment is still relevant as a background for interpreting the δέ segment (the δέ utterance should be interpreted against the assumption that there is great variety in religious doctrines, established in the utterance preceding δέ). Note also that the utterance which δέ marks is a question, as opposed to the preceding utterance, which is a statement – another factor which signals the discontinuity marked by δέ. As such, the utterance marked by δέ should not be taken as an exemplification, but as a separate segment which produces positive cognitive effects on its own, and takes the preceding context as a background for deriving those effects. In that sense, δέ forces a certain way of interpreting an utterance on the audience, and it compels that audience to construct an interpretation along the lines encoded by δέ. The following instance is similar:
(35) Προσετίθει δὲ κἀκεῖνα, ὡς σφόδρα χρὴ προσδοκᾶν καὶ ἐπιβουλεύσειν τῇ Ἀρσάκῃ τὸν Ἀχαιμένην, ἄνδρα δοῦλον μὲν τὴν τύχην (ἀντίθετον δὲ ὡς ἐπίπαν τῷ κρατοῦντι τὸ κρατούμενον), ἀδικούμενον δὲ καὶ εἰς ὅρκους ἀθετούμενον […]. (Hel. Aeth. 7.26.10.1-5)
[Arsace had planned to marry Chariclea off to Achaemenes, the son of her closest servant. Theagenes has managed to convince Arsace to call off the marriage, and is informing Chariclea of the good news.] “Theagenes further pointed out these things as well – that it was certainly necessary to anticipate that Achaemenes would even plot against Arsace, being a man of servile condition (inimical [de] in general is the subordinate to the master) and suffering injury too as the victim of a broken oath, […].”
In (35), the underlined δέ marks an utterance which could be taken to be an explanation for why a ‘man of servile condition’ would plot against his master – γάρ is often used to mark utterances which explain the ‘general rule’ underlying a particular instance of that rule (e.g. Bas. Hex. VI.7.59; Greg. Iul. I.557.31; Hel. Aeth. 4.5.7.9). In this case, the ‘general rule’ is that slaves are inimical to their masters, and the ‘particular instance’ is that Achaemenes will probably plot against Arsace. In any case, γάρ could have conceivably been inserted here; instead, Heliodorus inserts a δέ, which leads the audience down a different inferential path. The general rule outlined in the δέ segment is discontinuous in the sense that it is not a reason for Achaemenes to plot against Arsace (instead, it outlines one of the reasons for his plotting), but it is still continuous in that it should be seen against the background of his plotting. In relevance-theoretic terms, it achieves positive cognitive effects on its own (i.e., not as a part of the previous
context, which deals with Achaemenes’ plotting), but the derivation of those effects depends on the existing cognitive environment.
Of course, the notion of ‘rigidity’, especially for a dead language, is a largely subjective matter – for other scholars, (34) and (35) may not be evidence of δέ’s rigidity because they have a different view of γάρ’s function. For them, γάρ might not be expected at all in these instances. The problem here is that an analyst will always interpret a DM in light of his or her proposed function; as such, he or she will assume that its meaning is indeed rigid – that is, that problematic cases like (34) or (35) can be explained away by adjusting the context to fit the DM’s proposed function. As such, rigidity is less an aspect of the DMs under consideration and more a methodological bias in doing research into ‘dead’ DMs. While I think rigidity is a useful tool for distinguishing procedural from conceptual meaning in living languages for which we have access to native speaker intuitions, it is much less straightforward to apply it to δέ, γάρ and οὖν (and dead languages in general). As a consequence, the question as to whether δέ is indeed rigid will probably have to remain open indefinitely. However, examples like (34) and (35) indicate that the boundaries between different DMs (or, at least, between δέ and γάρ) are not always clear-cut, and that an approach which only has recourse to coherence relations cannot distinguish between them adequately – δέ and γάρ can both mark explanations, as (35) demonstrates.
Non-polysemy, the final trait of procedural expressions, is not clear-cut either. Whether δέ is polysemous depends largely on the perspective you take – if, like me, you start from Modified Occam’s Razor, you will probably find that every instance of δέ can be subsumed under a single meaning. On the other hand, Bakker (1993) seems to propose several distinct (yet related) functions for δέ – he is not explicit about whether he believes δέ is truly polysemous, but he seems to imply that he thinks it is at least polyfunctional. Bakker’s approach seems to be contiguous to that espoused by e.g. Hansen (1998) – words have different senses which are related “as extensions from a prototype” (1998: 241). The point is that the framework taken has a large bearing on whether you think δέ is polysemous or not. However, I have already argued that the minimalistic approach which flows naturally from Modified Occam’s Razor seems to be the most theory-neutral starting point for a semantic investigation, and I remain committed to that notion. Unless there is unambiguous evidence to the contrary, then, I will assume that δέ has a single distinct meaning.
In sum, there is reason to suggest that δέ is a procedural element. Rigidity and non- polysemy are difficult to ascertain for a dead language, but introspective inaccessibility and non-compositionality seem to be applicable and suffice to argue that, if δέ is not procedural, it is, at the very least, difficult to sketch in conceptual terms – indeed, I am not aware of any account which does not analyze its meaning in terms of what it does, and I have not come across any examples which even hint at compositionality. So, even