Bonifazi, Anna, Annemieke Drummen, and Mark de Kreij. 2016. Particles in Ancient Greek Discourse: Exploring Particle Use across Genres. Hellenic Studies Series 79. Washington, DC: Center for Hellenic Studies. http://nrs.harvard.edu/urn-3:hul.ebook:CHS_BonifaziA_DrummenA_deKreijM.Particles_in_Ancient_Greek_Discourse.2016.
I.2 From σύνδεσμοι to particulae
2.2 Early study of grammar
2.3 The Téchnē attributed to Dionysius Thrax
2.4 Early Definitions of σύνδεσμοι
The passage is broadly regarded as corrupt and highly opaque,  but what we may roughly deduce is that Aristotle here presents the σύνδεσμος as a word that can combine other significant (signifying) sounds into a significant whole. As is clear from the context, he is not discussing parts of speech (μέρη τοῦ λόγου) here, but parts of the lexis (μέρη τῆς λέξεως) – “diction” in Swiggers and Wouters’ translation – one of the six components of (good) tragedy. As such his focus is not to offer “a systematic treatment of word-classes, but [to offer] us a list of definitions of elements constitutive of the λέξις, (oral) poetic expression.”  A definition of ἄρθρον follows this passage; like the definition of σύνδεσμoς it too shows that ἄρθρον is not yet conceived as denoting only the article. In view of the opaque nature of the passage, as well as Aristotle’s non-grammatical concerns here, it is unclear which words fall in the category σύνδεσμος.  Swiggers and Wouters conclude tentatively that the connective and disjunctive particles would be covered by the term σύνδεσμος, whereas expletive particles would fall under ἄρθρον.
Clearly, the category has been redefined, and a crucial step has been taken from Aristotle’s philosophical comments on what the lexical category is to Diogenes’ attempt to describe the morphology (ἄπτωτον) and function of its members. The description provided in the text, “indeclinable” (ἄπτωτον), is a characteristic of particles that seems natural to the modern reader, but barring the Stoic tradition the feature was, in fact, not a requisite part of the category’s definition in antiquity until Priscian (sixth century CE). 
2.5 The scholia
This scholion hints at valuable insights, but the scholiast is not very helpful. First of all, it is not clear if his comment explains only ἀλλά or the whole construction ἀλλ᾽ ἄγε. The paraphrase ἀλλὰ φέρε suggests the latter, but it seems completely superfluous: surely ἄγε would be as natural as φέρε to the audience in later antiquity, if not more so. More problematic is the following description of “exhortative adverb.” Its sense is clear enough, but it is not entirely clear how ἐπίρρημα can be used to describe the combination ἀλλ᾽ ἄγε. ἀλλά on its own might be regarded as an adverb or “adverbial” in the sense that it co-occurs with a verb form (the imperative ἄγε). By extension the comment “or it is an adverbial μόριον” would refer just to ἀλλά, making this instance the only one in the scholia where μόριον describes a particle.  The weight of the evidence in the scholia, then, suggests that in early linguistic discussions σύνδεσμος was the normal term to refer to the words we would call particles. Note however, that although most particles were called σύνδεσμοι, not all σύνδεσμοι were what we would call particles.
2.5.2 σύνδεσμοι in the scholia
δηλοῖ σύνδεσμον παραπληρωματικὸν τὸν ἤτοι, ἴσον τῷ
δή, οἷον „ἐπειῆ πολὺ φέρτερόν ἐστι“ (1.169), καὶ παραλλήλως „ἦ δὴ
λοίγια ἔργα“ (1.518). δηλοῖ δὲ καὶ ἀπορρηματικὸν σύνδεσμον· „ἦ οὐχ
5 Ἑλένης ἕνεκ’ ἠϋκόμοιο; / ἦ μοῦνοι φιλέουσ’ ἀλόχους μερόπων ἀν-
θρώπων / Ἀτρεῖδαι;“ (9.339–41). δηλοῖ δὲ καὶ τὸ ἔφη· „ἦ καὶ κυανέῃσιν“ (1.528). |
καὶ σεσημείωται Ἀρίσταρχος ὅτι ὁ μὲν Ὅμηρος ἀεὶ
ἐπὶ προειρημένοις λόγοις ἐπιφέρει τὸ δηλοῦν τὸ ἔφη, ὡς ἐπὶ τοῦ προ-
κειμένου, ὁ δὲ Πλάτων μετ’ αὐτὸ ἐπιφέρει τὸν λόγον. |
10 ψιλούμενον δὲ καὶ βαρυτονούμενον δηλοῖ σύνδεσμον διαζευκτικόν· „ἦ εὖ ἠὲ κακῶς“
(2.253). ἔστι δὲ ὅτε καὶ ἀντὶ συναπτικοῦ τοῦ εἴ τίθεται, οἷον „οὐδ’ ἀφα-
μαρτοεπής, ἢ καὶ γένει ὕστερος ἦεν“ (3.215). ποτὲ δὲ παρέλκει· „ἀλλὰ
τίη με ταῦτα παρεξερέεσθαι ἕκαστα;“ (10.432).  δασυνόμενον δὲ καὶ
ὀξυτονούμενον ἄρθρον προτακτικὸν δηλοῖ· „ἣ δ’ ἑτέρη θέρεϊ προρρέει
15 εἰκυῖα χαλάζῃ“ (22.151). δηλοῖ δὲ καὶ ἄρθρον ὑποτακτικόν, οἷον „ἣ
μυρί’ Ἀχαιοῖς“ (1.2). δηλοῖ δὲ καὶ ἀναφορικὴν ἀντωνυμίαν· „ὣς ἥ γ’
ἀμφιπόλοισι μετέπρεπεν“ (6.109).
δηλοῖ δὲ καὶ τὴν σύναρθρον
ἀντωνυμίαν τρίτου προσώπου, συζυγοῦσαν τῇ ἐμή, σή. ὑποδείγματα
20 δὲ ταύτης παρ’ Ὁμήρῳ οὐχ εὑρίσκεται, ἐκ δὲ τοῦ ἀναλόγου νοεῖται·
αἱ γὰρ πλάγιοι πᾶσαι δι’ αὐτῆς παρ’ Ὁμήρῳ σώζονται.
is the filling combiner ἤτοι, similar to
δή, as in “ἐπειῆ πολὺ φέρτερόν ἐστι” (1.169), and elsewhere “ἦ δὴ
λοίγια ἔργα” (1.518). It is also the interrogative combiner: “ἦ οὐχ
5 Ἑλένης ἕνεκ’ ἠϋκόμοιο; / ἦ μοῦνοι φιλέουσ’ ἀλόχους μερόπων ἀν-
θρώπων / Ἀτρεῖδαι;” (9.339–41). And it also means ἔφη: “ἦ καὶ κυανέῃσιν” (1.528). |
And Aristarchus noted that Homer always
uses it to mean ἔφη after the words have been spoken, as in the
current example, while Plato starts the speech after it [sc. ἦ]. |
10 Unaspirated and with the grave accent, it is the disjunctive combiner: “ἦ εὖ ἠὲ
κακῶς.” (2.253) And it happens that it is even placed instead of hypothetical εἰ, as in:
“οὐδ’ ἀφαμαρτοεπής, ἢ καὶ γένει ὕστερος ἦεν” (3.215). And sometimes it is redundant:
“ἀλλὰ τίη με ταῦτα παρεξερέεσθαι ἕκαστα;” (10.432). Aspirated 15 and
with the acute accent, it is the prepositive article: “ἣ δ’ ἑτέρη θέρεϊ προρρέει
15 εἰκυῖα χαλάζῃ” (22.151). It is also the postpositive article, as in: “ἣ
μυρί’ Ἀχαιοῖς” (1.2). And it is also the anaphoric pronoun: “ὣς ἥ γ’
ἀμφιπόλοισι μετέπρεπεν” (6.109).
It is also the possessive
pronoun of the third person, to be added to ἐμή, σή. Signs
20 of this are not found in Homer, but from analogy it may be reasoned,
since all the oblique cases maintain that paradigm in Homer.
This long discussion of η (ἤ, ἥ, ἦ) is found as a whole in the Venetus A manuscript, but that does not mean it was conceived in its entirety by one person. It is typical of the scholia in containing an explicit references to a specific scholar, in this case Aristarchus (second century BCE). The embedded reference to a named scholar indicates that this scholion is probably a composite, that is, authored by someone who includes information from Aristarchus while adding other information drawn either from his own experience or, more likely, from other sources. 
2.5.3 Aristarchus on σύνδεσμοι
2.5.6 ἄν and κε(ν)
The authors of these scholia are rightly confused about the forms in these two constructions, arguing in both cases that the verb form should be replaced by an optative. In the first case the argument is understandable: “either κε is redundant, or the verb form should be optative.” The second, however, is confused: “[νοστήσω] instead of νοστήσαιμι, and κε is redundant.” Clearly, the scholia do not regard κε as an equivalent of ἄν, but it remains unclear what kind of function they attributed to the Homeric particle κε. 
Either the construction is a variant for ἄν plus the optative, or it is a future, and κε is simply superfluous. The only thing that appears to trigger the scholiast’s belief that this construction should be a potential optative is the presence of κε, which he replaces with ἄν in his paraphrase. There is some awareness, then, that κε could fulfill – or used to fulfill – a function similar to that of ἄν. However, they seem to have believed that κε could be used simply as a filler too, without any bearing on the reading of the verb. The few notes on redundant ἄν (see note 27 below) fall in the same group, but the longer discussions suggest that the scholiasts had more trouble accepting a redundant instance of ἄν than of κε. 
2.5.7 Noteworthy readings of σύνδεσμοι
2.6 The Téchnē and other early scholarship
δηρὸν ἔῃ κρειῶν, κέλεται δέ ἑ θυμὸς ἀγήνωρ
μήλων πειρήσοντα καὶ ἐς πυκινὸν δόμον ἐλθεῖν·
meat for a long time, and whose proud spirit urges him
to go and attack καί the closely built sheep-fold.
About καί in line 301 the scholion adduces the analysis and paraphrase of Dionysius Thrax:
The scholiast proposes that καί is redundant here, meaning it should not be translated: “to go and attack the closely built sheep-fold.” Dionysius, conversely, proposes a paraphrase containing the construction οὕτως … ὥστε (so hungry … that) but without finding an adequate synonym for καί. We propose that this is an attempt by Dionysius to render what we now call the scalar function of καί with the addition of οὕτως … ὥστε. With this paraphrase, then, Dionysius gives us a first attempt to express the force of καί as a scalar particle. 
Here we find a definition of σύνδεσμοι παραπληρωματικοί that resonates strongly with the definition of σύνδεσμοι in the Téchnē, but with a focus on filling the gaps (παραπληρῶν, compare the variant reading πληροῦσα in the Téchnē). Trypho explains this with a metaphor of stuffing between amphoras: σύνδεσμοι παραπληρωματικοί serve to keep the other words from becoming “harsh.” Trypho further argues that σύνδεσμοι should be regarded as words (as opposed to syllables). Elsewhere, Trypho has posited: “[if] they are [words] they must mean something.”  If Trypho pursued this line of thought, his discussion is no longer extant. However, the combination of these two thoughts will form the basis of Apollonius Dyscolus’ discussion of σύνδεσμοι παραπληρωματικοί.
2.6.2 Apollonius the Sophist
2.6.3 σύνδεσμοι in the Téchnē
This definition assigns two functions to a σύνδεσμος: on the one hand a σύνδεσμος knits together units of thought (διάνοια) in an utterance by imposing order, and on the other it has an effect on gap(s) in the expression. 
Beyond such brief definitions, the Téchnē only gives a few examples for each subcategory; i.e. καί is συμπλεκτικός, ἤ is διαζευκτικός, ἵνα is αἰτιολογικός et cetera. This discussion of σύνδεσμοι reveals none of the insight or interest found in Thrax’ comment quoted in the scholion to Iliad 12.301, and may serve as another argument that it is unlikely that chapter 20 of the Téchnē can be attributed to Thrax.
2.6.4 Demetrius’ Style
οὕτω δὴ οἶκόνδε φίλην ἐς πατρίδα γαῖαν… (Odyssey 5.204)
just like that, homeward to your beloved fatherland…
This explicit discussion of the πάθος that a σύνδεσμος contributes is unparalleled in early sources, other than Apollonius Dyscolus’ discussion of γε. 
2.7 Apollonius Dyscolus
Several different elements in this cryptic passage require explication. Lallot interprets Apollonius to mean that σύνδεσμοι have their own force, but that context determines which specific σύνδεσμος should be used, as well as the particular force that the chosen σύνδεσμος acquires in the sentence. 
And further down:
These passages outline Apollonius’ approach to those σύνδεσμοι he calls redundant, on which he expands in Combiners, the first and only work from antiquity known to be dedicated solely to σύνδεσμοι. The text of the treatise is damaged, apparently corrupt in places, and incomplete. Even so, it gives us more material on the subject than any other work. As Apollonius’ definitions in the Syntax already suggest, his Combiners contains innovative insights about the workings of σύνδεσμοι in general and in specific instances.
The resonance of the terms φράσις, συνδεῖν and most of all πράγματα must be significant.  However the lacuna is resolved, the explanation in Apollonius is more extensive than the definition in the Téchnē, but its meaning is not as clear. Moreover, the term διιστᾶσιν in the Téchnē’s definition seems more advanced than διαζευγνύουσιν, as it explains the category (διαζευκτικοί) without resorting to the same root. The definition in the Téchnē appears to be further developed than that given by Apollonius; yet another argument to regard this part of the Téchnē as composed later than Apollonius Dyscolus.
Apollonius answers this question by arguing that the majority of instances must have priority over the minority: copulative (συμπλεκτικοί) combiners often connect even if they are sometimes redundant – so they are rightly called copulative combiners. Similarly, σύνδεσμοι παραπληρωματικοί are filling in the majority of instances, hence their name. 
2.7.2 Important topics raised by Apollonius
The combination καίπερ is described as getting its force from περ, as both combiners have the same function. The concessive force of περ is clear enough, and repeated in all later descriptions of the particle, but more notable is the comment on amplification (αὔξησις). When he talks of “amplification,” Apollonius appears to be thinking of some form of emphasis: compare a similar comment in the Syntax, where Apollonius is more explicit: [σημαίνει] ἐναντιότητα ὁ πέρ μετ’ αὐξήσεως ἐμφατικῆς (“περ [conveys] a concessive force, along with emphatic amplification”).  What Apollonius has in mind by “amplification” is not self-evident, but we envision some kind of prosodic emphasis resulting from the addition of the enclitic περ. 
ὣς ὁ μὲν ἔνθ’ ἠρᾶτο (Odyssey 7.1).
Thus, he prayed there (Odyssey 7.1).
Apollonius compares the constructions οἱ μὲν δή and ὣς X μέν, both occurring often in transitional passages in Homer. Both constructions signal to the reader that one episode is left behind, and another begins.  Apollonius’ analysis of δή in this context shows a crucial new step in the study of σύνδεσμοι. Rather than discussing the function of a σύνδεσμος only with respect to the surrounding words or the host clause, he shows how important it is to consider the place of a σύνδεσμος, or string of σύνδεσμοι, in the larger discourse. Here he has noted the common occurrence (καὶ τῶν παραπλησίων) of δή in such concluding verses just before a transition to a new episode.
2.8 After Apollonius Dyscolus
2.8.1 Early grammars
τὰ μέρη τῆς ἑρμεν<ε>ίας.
the parts of the expression.
The simplicity of the definition is striking, especially compared to all the other extant definitions.  Scholars have therefore concluded that it is a simple school grammar, designed with the practical aim of listing and describing lexical categories rather than providing a philosophical foundation of the workings of language. This conclusion may be extended to most of the papyri from this genre, dating from the second to the sixth century CE. 
All separate aspects of Priscian’s definition can be traced back to his predecessors, but this is the first time we find the elements combined. In book XVI of the Institutions of Grammar, Priscian discusses only a small selection of combiners, and often has recourse to Greek translations. The list of subcategories that Priscian identifies comes closest to that found in Herodian, but he has added a few categories not mentioned before. After listing all possible kinds of combiners,  Priscian devotes only a few lines to most kinds and none to some. Two apparent additions to the list, the coniunctiones ablativae and praesumptivae, remain undiscussed, which makes it hard to guess what he meant with the terms (hence the lack of translation) and which combiners he had in mind.
2.8.2 Late Antique Scholia to the Téchnē
Note the striking similarity with Priscian’s definition, which we discuss above: “a combiner is an indeclinable part of the expression, which connects the other parts of the expression, with which it co-signifies, signalling either a force [of its own] or the arrangement [of the parts of the utterance].”  It is generally assumed, and may be observed in comparison, that Priscian based much of his Institutions on Apollonius Dyscolus’ work, which supports Pecorella’s hypothesis that this definition is in fact to be attributed to Apollonius. The definition combines different ideas in a way that fits well with our sense of Apollonius’ approach to σύνδεσμοι. It is, in addition, more intricate than the definitions we find in the different Téchnai. Since a definition of σύνδεσμοι is extant in Apollonius’ Syntax,  we find it more likely that the source both for Priscian and for Heliodorus was the definition of σύνδεσμοι in a lost part of Apollonius’ Combiners.