The Ancient Greek sentence left periphery A study on Homer

The pattern (Setting – Topic –) Focus . NB: The Verb always follows , which was proposed by H. Dik in order to describe AG’s left periphery, raises some issues. In particular, it presents a number of exceptions, which scholars (Matić and others) have variously attempted to resolve. In the present contribution, based on case studies drawn from Homer, the following pattern for the Homeric left periphery is proposed: ( Setting – Topic – Focus ). NB: Unmarked elements follow . This is not dramatically different from Dik’s pattern; rather, it is an extension of it.

Topic and Focus are pragmatically marked constituents. The Topic is an element that is usually provided in the discourse and that "the speaker regards as an appropriate foundation for constructing a message" (Dik 2007: 31). The Focus is an element that is usually new and that presents "the most salient piece of information in a clause" (Dik 2007: 32), e.g.
(2)  (Her. 5.3.1;cf. Dik 1995: 237-238). Dik (1995: 238) comments thus: "In this first paragraph of the descriptive excursus on Thrace, the overall subject (Θρηίκων δὲ ἔθνος) gets pride of place. Thrace has been mentioned earlier; in the previous clauses we have learnt that Megabazus has orders to subdue it […] The prominent position of Thrace […] marks the Topic shift. The first point Herodotus makes about Thrace is that it certainly merits our attention: μέγιστόν ἐστι (Focus)".
The Setting constituents are "adverbial phrases at the opening of clauses" that "tend to be part of the spatial or temporal (or causal) organization of the text" (Dik 2007: 36),5 e.g.
In this analysis, three points should be noted. First, the AG sentence features a division into two major parts: a left part, in which the constituents are ordered according to their pragmatic connotations (henceforth "left periphery," as it is usually called by syntacticians);7 and a right part, in which the order of the constituents is free-with the exception of post-verbal placement.8 Second, the pattern for the AG left periphery, including the important specification regarding Verb placement (henceforth "Dik's pattern" or "model" for the AG left periphery),9 is the following: (4) (Setting -Topic -) Focus. NB: The Verb always follows.
Third, about this pattern, (4), two remarks should be made: a) If the AG Verb always immediately follows the marked elements of the sentence, then everything in AG that is to the left of the Verb must belong Journal of Greek Linguistics 18 (2018)  to the left periphery. In other words, the AG left periphery coincides with everything that is on the left side of the Verb. b) In AG sentences, while the Setting and Topic are optional, the Focus is mandatory. Consequently, AG sentences begin with marked elements in every case, and there cannot be sentences without pragmatic constituents. The rest of this study focuses on the AG left periphery. Further analysis of this part of the sentence is conducted, and a (slight) modification to pattern 4 and points a and b above is suggested.

1.2
Uncertainties and problems Dik herself was aware that the pattern she proposed explains the word order well overall, but it does not apply to all of the AG sentences. She provided an (unsystematic) account of the "uncertainties and … problems" of her theory in various passages of her two books, particularly the chapters in which some "running" passages of Herodotus10 and Sophocles11 are analyzed.12 The main exceptions to the proposed pattern can be classified in two major groups: unmarked preverbal constituents and marked postverbal constituents.
I focus here on the first group because the second group is not relevant to the left periphery of the sentence. This first group generally includes constituents that for particular reasons cannot be considered either the Focus (e.g. because they are given or not important) or the Topic (e.g. because they follow the Focus). In order to be consistent with her theory, Dik considers such preverbal constituents to be only apparently unmarked and explains their position by referring to syntactic devices, such as complex foci and extended Topic units.
The following examples of these unmarked elements are provided (Dik's pragmatic 10 and 7.27-29, 38-39;cf. Dik 1995: 237-257. 11 El. 516-527, 23-37, 660-680, 1171-1226cf. Dik 2007: 225-248. 12 See also Dik (2007): 60-65, from which some of these examples are drawn. 13 In general, Unmarked elements preceding the Topic can easily be interpreted to be Setting constituents. κατὰ in πάντα all:nom.pl.n 'All these Thracians are alike in their usages ' (Her. 5.3.5;cf. Dik 1995: 238-239) Topic=νόμοισι, Focus=παραπλησίοισι πάντες, Unmarked Element=οὗτοι (Subject). Dik (1995: 239) surmises "that παραπλησίοισι πάντες can be interpreted as having complex Focus, but that apparently leaves us with two Topics … οὗτοι could be described as a necessary addition, which, however, is not allowed to take first position in the clause." Topic=ἐμοὶ, Focus=τἀληθὲς, Unmarked Element=σύ (Subject). Dik (2007: 243) remarks on this passage thus: "It seems clear that ἐμοὶ is Topic and τἀληθὲς Focus, but what are we to make of the σύ in between? I am tempted to classify it as Focus, but in any case, the vocative and the line break that intervene between σύ and τἀληθὲς allow for prominence for both those elements".  ' (Her. 5.4.3;cf. Dik 1995: 240) Topic=τὸν μὲν γενόμενον (περιιζόμενοι?), Focus=ὀλοφύρονται, Unmarked Ele-ment=οἱ προσήκοντες (Subject). Notice that here the verb has a pragmatic function. As Dik (1995: 239) observes, "τὸν μὲν γενόμενον is the Topic … the unexpected ὀλοφύρονται is the Focus of the clause, which in a Topic -Focus -Remaining Elements clause pattern seems to me to tip the scale toward construing τὸν γενόμενον with περιιζόμενοι". Nothing is said about the intervening οἱ προσήκοντες.  (Soph, El. 289-290;cf. Dik 2007: 63) Topic=/, Focus=σοὶ μόνῃ, Verb=τέθνηκεν, Unmarked Element=πατὴρ (Subject). Dik (2007: 63) suggests to consider πατὴρ a Focus: "I assume that we here have one of the rare instances where not one but two constituents have Focus, both σοὶ μόνῃ and πατὴρ: are you the only one who's lost a father?" Whether or not the pragmatic interpretations of these cases are correct, Dik claims that they are not statistically relevant. At the close of her analysis of Herodotean "running discourse", she states that "Despite the uncertainties and remaining problems, on balance the outcome is certainly favourable for a pragmatic approach" (Dik 1995: 256-257).14

1.3
Matić's pattern After the publication of Dik (1995), some scholars attempted to improve her model in order to better address the issues that remained unresolved. Most notably, Matić (2003: 579) proposed "some corrections to the pattern" in order "to explain the syntactic and pragmatic nature of sentences that do not allow 14 Dik does not provide numbers or percentages. According to Matić, Dik' For example, in (5) above, "The ConTop expression "they", οὗτοι, is placed after the contrastive Topic expression νόμοισι … when a ConTop expression is placed preverbally, it is frequently not directly after the FSTop expression, but rather within this expression, which in this way becomes a discontinuous phrase (νόμοισι … παραπλησίοισι)" (2003: 598).16 With this refinement of the model, Matić (2003: 615) could "account for the overwhelming majority of the attested AG sentences" (more than 90 % of the cases in his corpus). Still, he acknowledged that some material (9 %) remained "which is not, or only with unsurmountable difficulties, to be housed either in the sentence pattern based exclusively on preverbal slots, or in its somewhat more flexible variant proposed in Section 2" (2003: 579): postverbal narrow Foci and preverbal narrow Foci separated from the verb (2003: 615-625).17 For example, in (8) above, ὁποίῳ can be interpreted as a preverbal narrow Focus separated from the verb ὤλετ' by an expression (κεῖνος) that resembles a continuous Topic (this example is not quoted by Matić, but is similar to others presented in Matić 2003: 622, e.g. Xen. An. 2.1.10).18 15 "Not all Topic expressions … get assigned to this [i.e. left-peripheral] position, but only some of them … The preverbal Topic position is a structural position for referential Topic expressions denoting frame-Setting Topics … More precisely, it is used for contrastive Topics, Topics active in the same spatio-temporal frame or appearing in interaction, reintroduced or newly introduced Topics, etc. … Non-clitic Topic expressions that do not set new referential frames, i.e. continuous Topic expressions, are, with certain well-defined exceptions, placed immediately after the verb" (Matić 2003: 588 ff. passim). 16 Similarly, the other two examples of preverbal unmarked elements between Topic and Focus provided in §1.2 (Matić does not explicitly take them into account) can be explained: the intervening subjects σύ in (6) and οἱ προσήκοντες in (7)  "Narrow Focus expressions can be separated from the verb by clitics, negation, non-verbal parts of idioms, verb support constructions, etc. All these cases are accountable for in terms of the model proposed in Section 2. When the Focus expression is separated from the verb by a Topical expression, FSTop or ConTop, an explanation in terms of this model seems implausible" (Matić 2003: 624). 18 The other example of preverbal unmarked elements between Focus and Verb included in §1.2 (9) can be explained in the same way: the intervening subjects πατὴρ can be interpreted to be the Topical element between Focus and Verb.
Journal of Greek Linguistics 18 (2018) 172-210 In conclusion, Matić was not able to provide an univocal pragmatic explanation that is valid for all Greek examples and was compelled to posit two concurrent word order systems, so that "one has to put up with some free variation, meaning that the choice of the word order system used in a sentence is not to be predicted from its pragmatic, semantic or syntactic properties" (2003: 624).19

1.4
A new proposal In this study, an even more general pattern for the AG left periphery is proposed to explain the exceptions pointed out in both §1.2 and § 1.3.
The starting point for the AG left periphery (pattern 4, § 1.1) is Dik's pattern, which is presented again here: (10) (Setting -Topic -) Focus. NB: The Verb always follows.
The presence of preverbal unmarked elements between the Focus and the Verb (the case [Topic -] Focus -Unmarked Element -Verb; see § 1.2, b) could be better explained by substituting the specification "The Verb always follows" with "Unmarked elements follow," so that among the Unmarked elements, the Verb does not necessarily occupy the initial position. This is the modified version of the pattern: (11) (Setting -Topic -) Focus. NB: Unmarked elements follows.
And this is an example of a sentence in which the Verb is not the first Unmarked element: See n. 20: not without a new or salient element, but without a new or salient element singled out for a special position in the linear order of the clause. 22 In cases like this, note that the pragmatic interpretation of the constituents following the first Unmarked element is different than before (cf. 6). More precisely, constituents following the first Unmarked element are also considered unmarked. In the specific case, τἀληθὲς, following the unmarked σύ, is considered unmarked (very likely a "focal element": on the distinction between focal element and Focus, cf. n. 25). 23 As is explained in more detail later (b), in this model, Pragmatics plays an important but not exclusive role: parentheses show that there can be sentences without pragmatic constituents. On the other hand, the discussion summarized in § 1.2- § 1.3 shows the limits of an exclusively pragmatic explanation of AG word order. 24 If there is one; as I stated in (b), there can be sentences without pragmatically marked constituents. 25 The general presupposition of my proposal is the distinction, introduced by Dik & Hengeveld (1997: 312-313), between "informational status" and "pragmatic function". The "infor-Journal of Greek Linguistics 18 (2018) 172-210 Pattern 13 presupposes the following sentence structure (which is slightly different from the one proposed by Dik in pattern 1): (15) (Setting -Topic -Focus -) Unmarked elements. 'When a child is born, the kinsfolk sit round and lament' (Her. 5.4.3; = 7) mational status" is mainly related to the concepts of Topicality and Focality: "Topicality concerns the status of those entities 'about' which information is to be provided or requested in the discourse … Focality attaches to those pieces of information which are the most important or salient". The "pragmatic function" is related to the concepts of Topic and Focus: "certain topical elements are singled out for special treatment in the expression of the clause", particularly "a special position in the linear order of the clause". "Such elements will receive the pragmatic function of Topic … Likewise, if such special treatment is given to focal elements, these will be assigned the pragmatic function of Focus". The informational status characterizes every constituent of the sentence, while the pragmatic function only characterizes some of them.
As for the AG sentence, the assignment of a pragmatic function to some of its constituents, while compulsory according to the studies previously discussed, is optional according to the proposal raised here. Therefore, in AG, there can be "sentences without pragmatic constituents", that is, sentences whose elements, despite having an informational status, are not assigned a specific pragmatic function. More generally, in this article, when the expressions "pragmatic" (or similar: "pragmatically") and "marked" appear, what is meant is "with a pragmatic function," and when the expressions "not pragmatic" (or similar) and "unmarked" appear, what is meant is "without pragmatic function". Focus= σοὶ μόνῃ, Unmarked Elements=πατὴρ τέθνηκεν

1.5
Some further remarks Aside from the differences pointed out in the preceding section, the left periphery model proposed in this study (13) must not be considered to be radically in contrast with Dik's model (4) but rather as an extension of it. In fact, most AG sentences26 have quite a simple structure, with usually no more than two constituents to the left of the Verb. The structure of (the first part of) these clauses is often Topic -Focus -Verb, which is consistent with both patterns 4 and 13; cf. e.g. §1.1, ex. 2. Since pattern 4 explains the vast majority of AG sentences (with no more than two constituents to the left of the Verb), it is understandable that Dik's model could have been identified as a general pattern, that is, one that is valid for all Greek sentences.
On the other hand, the analysis of AG sentences with a more complex structure and at least three constituents to the left of the Verb (a case that is less frequent but not impossible to find) shows that Topic -Focus -Verb -Remainder is not the most general possible pattern. If the three preverbal constituents are not Setting, Topic, and Focus, Dik's model does not apply to describe these types of sentences unless devices such as complex foci and extended Topic units are adopted (in other words, if it is assumed that there is more than one Topic or Focus).27 The remainder of this study reveals the outcomes of research that was conducted regarding sentences with at least three constituents to the left of the Verb, in order to investigate whether or not this type of sentence can follow pattern 13. 26 Or more appropriately "cola"; see §2.2.3. 27 For its complexity, this explanation seemed "inadequate" to some scholars (cf. e.g. Matić 2003: 578).

2.1
The corpus The research was conducted based on a specific corpus: the poems of Homer.28 Given ther peculiar history, these works are particularly suitable for linguistic and syntactic analysis, because they reflect "to a large extent modes of preliterary oral communication" (Viti 2008: 206), in which the communicative and the pragmatic purposes prevail over the literary and artistic aims.29 Moreover, it is now commonly acknowledged that the poetic nature of the Iliad and the Odyssey should not be regarded as a crucial drawback for word order reliability: a number of studies (e.g. Dik 2007) have recently been conducted on the basis of the assumption that "word order is part of the 'common core' of the language, that part of the grammar that is shared by all registers of the language"; consequently, that "the rules of word order" are not "rendered invalid when one stops 'speaking prose'".30 These considerations explain the increasing interest of scholars in analyzing these poems31 in order to gain an overall understanding not only of Homeric word order but also of AG word order.32 28 The reference editions for AG text are Allen (1931) and Allen (1917Allen ( -1919. The commentaries used are Kirk (1985Kirk ( -1993 for Iliad and Heubeck & Hoekstra (1989) and Heubeck, West & Hainsworth (1990) for Odyssey. The translations quoted in this article belong to Murray (Murray 1919;Murray 1924) and Butler (Butler 1898;Butler 1900). 29 As Edwards (2002: 11) notes, based on the important studies of Bakker, particularly Bakker (1990), in the grammatical/linguistic structure of Homeric poems, "not just […] the poet's following of tradition and routine, but […] the result of his desire to communicate orally with his audience in the most effective way possible" should be recognized. 30 Quotations are drawn from Dik (2007) (p. 3, text and n. 3). This is especially true in the case of the meter, about which Dik's (2007) approach (specifically related to the meter of tragedy) seems particularly acceptable: "The fact that dramatic dialogue is metrical has prompted scholars to approach questions of word order and 'emphasis' in terms of metrical structure … If a word is considered emphatic, this is taken to be a consequence of its position in the trimeter line, rather than of its position in the clause (which may or may not coincide with the line)" (2007: 3). According to Dik (2007: 3), this assumption is incorrect: "the clause and not the line should be the unit of analysis". Nevertheless, she admits that some metrical aspects may also affect AG word order, particularly the emphasis on the beginning of the verse: "In poetry, the line boundary adds an extra opportunity for marking constituents as salient, and an author … to dole out the privilege of initial position … can place one constituent … in line-initial position" (2007: 13; see also pp. 168-224 for a detailed analysis). 31 See particularly the previously quoted works of Edwards (2002)  Furthermore, a powerful tool is available for all scholars studying the poems of Homer: the Ancient Greek and Latin Dependency Treebank (AGLDT). AGLDT is a database of Classical texts promoted by the Perseus Project where the morphological, syntactic, and lexical information for each sentence has been explicitly encoded,33 that consequently allows scholars to perform complex linguistic searches. Once completed, it will include all classical texts, but at the present time34 it is limited to (among AG authors) the entirety of Homer's Iliad and Odyssey, all of the works of Hesiod and Aeschylus, five tragedies of Sophocles, and some other single works (or parts of them).

2.2
The choice of examples 2.2.1 The method of Dik (1995) As Dik stated in the methodological introduction to her book on Herodotus (Dik 1995: 3-17), the main factors that are considered responsible for the AG constituent order include euphony, semantics, syntax, and pragmatics. To "identify the contribution of pragmatics to constituent order," she asserted that it is necessary to "isolate it from other possible factors" (1995: 15) Regarding euphony, this can be accomplished by studying authors ad hoc, who neither "favored certain rhythms over others" nor avoided hiatus (1995: 4).35 As for semantics, selecting sentences based on their verb may turn this variable into a constant (1995: 15):36 "This lexical selection ensures reasonable semantic similarity and allows for better comparison between states of affairs described and evaluation of the informational status of participants in those states of affairs".37 The same selection may also be useful for syntax.38 33 More information on the project and further bibliography is available at https://perseusdl .github.io/treebank_data/. 34 August 2017. More specifically, the research was conducted using AGLDT 1 (with Annis interface) and its guidelines (Bamman & Crane, 2008). Currently, a more recent version of AGDT is available (AGDT2), with up-to-date guidelines (https://github.com/PerseusDL/ treebank_data/tree/master/AGDT2/guidelines). 35 This is the case of the earliest prose authors, particularly of Herodotus, whom she studied. 36 More precisely, Dik (1995) selected sentences based on their main-clause predicate for the reasons stated on p. 16. For a similar approach, cf. Panhuis (1982). 37 More specifically, she analyzed sentences with the following predicates: στρατεύομαι (Dik 1995: ch. 4), ἄρχω / βασιλεύω (ch. 5), and λέγω (and similar verbs, ch. 6). 38 Notably, the examples presented by Dik (in the chapters quoted in n. 37) are characterized by not only the same predicate but also the same arguments (for example, in the case of στρατεύομαι, "the subject and an ἐπί-phrase designating the object of the expedition undertaken", 1995: 53). Therefore, these examples are analogous not only from a semantic point of view (because the context is the same) but also from the syntactic point of view Syntax and pragmatics More specifically, after a choice of a euphonic and semantic nature (see § 2.2.2), the study involved two steps. First, cases with the features noted in § 1.5 ("sentences with at least three different constituents to the left of the Verb") were isolated from the corpus based on syntactic categories ("syntactic analysis," mainly with the help of the AGLTD. Next, the possible pragmatic connotations of these examples were analyzed ("pragmatic analysis").
For further information on syntactic analysis, see § 2.3.2; on pragmatic analysis, see §2.3.3. Here, some clarification is provided about the expression "sentences with at least three constituents to the left of the Verb".
First, about the word "sentences": AG sentences consist of basic units called "cola", which can be distinguished by the position of ἄν, enclitics following Wackernagel's Law (unemphatic personal pronouns, some conjunctions, etc.), etc.42 Occasionally, sentences and "cola" correspond, but sometimes a sentence can be constituted by more than one "colon", as in the following example: ἕποντο follow:impf.3pl 'Even so the Danaans for a time ever followed on in throngs' (Il. 15.277) In this verse, there is only one sentence, but two cola, which are distinguished by the postpositive word μὲν in the second position: ὣς Δαναοὶ ‖ εἷος μὲν, etc.
(because they show the same syntactic constituents, as in the example above [subject, object, and verb]). 39 With few differences, cf. n. 41. 40 Dik (1995) also based her investigation on three case studies (see n. 37). 41 Differently from Dik (1995Dik ( , 2007, the syntactic analysis was not limited to the arguments of the predicates considered (cf. n. 38); rather, it was more general. On the reasons for this choice, see §2.2.4. 42 More information in Fraenkel (1964a), Fraenkel (1964b), Fraenkel (1965), Stinton (1977), and very recently, Scheppers (2011). Following Fraenkel's proposal, Wackernagel's Law is usually applied to cola rather than to sentences (cf. In consideration of this distinction, pattern 13 (see § 1.4) applies to "cola" rather than "sentences"; and in the definition "sentences with at least three constituents to the left of the Verb," the word "sentence" should be interpreted as "colon". Second, about the word "constituents": the focus of the research was only on the constituents in which the position is actually under discussion. Consequently, the following constituents, which appear in the first or the second position, were not considered: a) "Postpositive words" (according to the definition of Dover 1960: 12 ff.).
Their collocation is described by Wackernagel's Law, which requires them to occur in the second position of the sentence. The main words are: -Enclitic personal pronouns (μου, σου, οὑ …).
-The oblique cases of αὐτός in the anaphoric sense.

2.2.4
More than pragmatics This research conformed to the method of Dik (1995) as stated in § 2.2.2, and did not exclusively rely on a pragmatic analysis of the constituents, as the works on AG word order that were published after Dik (1995) usually have done (e.g. Dik 2007 andMatić 2003). Rather, as seen above, other levels of linguistic description (euphonic, semantic, and syntactic) were also considered and used as starting points. This seemed more appropriate to the specific nature of the research for both theoretical and practical reasons: -Theoretically: The study was explicitly focused on doubtful cases from a pragmatic point of view (i.e., as stated in §1.5, sentences with at least three constituents before the Verb); therefore, the pragmatic description could not be the starting point; rather, it had to be the end.

43
Only partially corresponding to Dover's "prepositive words" (Dover 1960: 13 ff.). 44 Not all Complementizers occur in the first position of the sentence. Others, such as γάρ, occur in second position as stated by point a (among "Postpositive words").
Journal of Greek Linguistics 18 (2018) 172-210 -Practically (particularly related to the syntactic level): The AGLDT provides syntactic45 and not (to date) pragmatic information for the sentence constituents.46 On the other hand, the pragmatics constituted the endpoint and the aim of the research. Therefore, after being analyzed from a syntactic point of view,47 each sentence was analyzed from a pragmatic point of view as well.

2.3
The analysis of examples 2.3.1 Introduction As stated in §2.2.3, after a choice of a euphonic and semantic nature, the research work involved two steps. First, "sentences with at least three constituents to the left of the Verb" were isolated from the corpus based on syntactic categories ("syntactic analysis"). Second, the possible pragmatic connotations of these examples were analyzed ("pragmatic analysis"). Here, some more information about the syntactic analysis (cf. §2.3.2) and the pragmatic analysis (cf. §2.3.3) is provided.

2.3.2
Syntactic analysis The syntactic analysis was based on a few fundamental categories. A detailed description of these categories is provided in the AGLDT:48 a) SBJ (Subject) Subjects […] come in a variety of parts of speech and phrases, including: -Nominative nouns -Accusative nouns -Genitive nouns -Infinitive verbs -Relative Pronouns Bamman & Crane 2008: 5-8 45 For information on the AGLDT, cf. Bamman & Crane (2008). 46 Moreover, the syntactic analysis allowed me to formulate a hypothesis regarding the order of the pragmatically unmarked constituents (to be confirmed via further research; cf. the Appendix). 47 "Syntactic analysis": as stated above, carried out after a choice of a euphonic and semantic nature. 48 Cf. Bamman & Crane (2008 In addition to some of the previous categories, there is a Complementizer (εἰ) and two Predicatives (νέῳ and πομπῆες). In this research, neither Complementizers nor Predicatives were taken into account. Complementizers were not considered because their position is usually not under discussion (as stated in §2.2.3, they are fixed in the first or the second position). The complex issues related to Predicatives are deferred to subsequent research. 51 Two (or more) constituents belonging to the same category or type (that is, coordinated constituents) were considered as only one.
To this general rule, the following exceptions were made. In the practical work of analysis, the examples were grouped according to the syntactic order of their three preverbal constituents (sentences beginning with SBJ -ADV -OBJ, sentences beginning with SBJ -OBJ -ADV, etc.: six possibilities). In the case of sentences with more than three constituents preceding the Verb (typically four, as in SBJ -ADV1-OBJ-ADV2-notice that in this case, since there are only three syntactic categories that are taken into account in this contribution [SUBJ, OBJ, ADV], at least one category must appear twice), the two constituents belonging to the same category were analyzed separately (in the preceding example, SBJ -ADV1 -OBJ and SBJ -OBJ -ADV2) and then the two different analyses were compared. The same method of analysis was used in points 2, 3, and 4 below. 53 Two coordinate OBJs were taken into account only when they were non-contiguous; see n. 51.
Journal of Greek Linguistics 18 (2018) ' (Od. 14,[317][318] τοῦ γὰρ φίλος υἱὸς=SUBJ, ἐπελθὼν=ADV (Temporal), αἴθρῳ καὶ καμάτῳ δεδμημέ-νον=ADV ("overcome as I was with cold and weariness", Modal) 4) Sentences with at least two constituents that were SBJs, ADVs, or OBJs, of which one was discontinuous, or split up into two parts (i.e. "hyperbaton"; in fact, according to the most comprehensive analysis on this subject, Devine & Stephens 2000, the two parts of the discontinuous constituent are in two different syntactic positions);56 e.g. 54 The behavior of Modal ADVs actually seems to be different from the behavior of all other ADVs. See also Kühner & Gerth 1898-1904. (more information is available in the Appendix. 55 Two coordinate ADVs, or two ADVs that were both Modal or Non-modal, were taken into account only when they were non-contiguous, cf. n. 51. 56 Devine & Stephens (2000) (and cf. also Stephens 1994 andStephens 2006) distinguishes between two main types of hyperbaton (with adjectives): a) Y1 hyperbaton (adjective in the first position, e.g. πολλὰ κατέλιπε χρήματα) in which the adjective is in Focus (cf. 45ff.) and b) Y2 hyperbaton (noun in first position, e.g. χρήματα κατέλιπε πολλά) in which the noun is the Topic or Focus (cf. 88ff.). Nonhead X hyperbata (e.g. πολλὰ ὁ ἄνθρωπος χρήματα κατέλιπε but also χρήματα ὁ ἄνθρωπος πολλὰ κατέλιπε) can be explained by postulating a double Focus (in the example above on πολλά and ὁ ἄνθρωπος, and other explanations are possible; cf. 121 ff. A rather frequent kind of Topic is called "in List" (see Benincà & Poletto 2004: 67ff.): "let us examine the case of the contrast within a given set, namely the case in which two elements Journal of Greek Linguistics 18 (2018) 172-210 As far as the analysis itself is concerned, it aimed to study the possible pragmatic connotations of the constituents of the sentences that had been isolated in the syntactic analysis; e.g.

Introduction
In this section, the outcomes of the research are given, with a focus on three particularly significant case studies: sentences where the Verb is ἕπομαι, ἄγω, and δίδωμι (see § 2.2.2).
For the first case (ἕπομαι), a complete report is provided, as follows. First, the most typical situation is presented: a sentence with at least three constituents, where one is a SBJ, one is an ADV, and one is an OBJ (cf. § 2.3.2, 1). Next, particular cases are taken into account: examples with at least three constituents, where one is a SBJ or an ADV, and the other two are OBJs of different types (cf. §2.3.2, 2); examples with at least three constituents, where one is a SBJ or an OBJ, and the other two are ADVs of different types (cf. § 2.3.2, 3); examples with at least two constituents that are SBJs, ADVs, or OBJs, where belonging to the same list of already known items are contrasted. We call this particular interpretation 'List Interpretation' (LI). In order to be clear, we add a context to our examples: Context: a farm producing a set of goods that are known to the people involved in the conversation.
(47) a. La frutta the fruit la it regaliamo, give for free, la verdura the vegetables la it vendiamo sell 'We give fruit for free, while we sell the vegetables' Here the two elements la frutta and la verdura are singled out from a list and attributed different predicates". beschi Journal of Greek Linguistics 18 (2018) 172-210 one is discontinuous, or split up into two parts (i.e. hyperbaton) (cf. § 2.3.2, 4). Finally, the interpretation of the examples is summarized in a "synoptic table".
For the other two cases studies (ἄγω and δίδωμι) the synoptic tables are given only for the sake of brevity.

3.2
ἕπομαι 3.2.1 Introduction The first case study involves the verb ἕπομαι, which is always used in the meaning of "to be or come after, follow" (see Liddell & Scott 1940, s.v.). It accepts two Objects (in the broad sense specified in §2.3.2, b): a person or thing that is followed, usually expressed with (ἅμα +) dat.; and a place toward which movement is directed, usually expressed with local expressions (εἰς + acc., etc.).

3.2.2
Sentences with at least three constituents, where one is a SBJ, one is an ADV, and one is an OBJ 3.2.2.1 SBJ -ADV -OBJ order This is the most frequent case. All of the examples appear to be pragmatically very similar: the Subject is usually a given element, while the new information is expressed by the Verb and its Object. Adverbials are temporal or instrumental and not modal (for an exception and explanation, see ex. 29). For example: αἰὲν ever ἕποντο follow:impf.3pl 'Even so then did the Trojans, high of heart, and their allies, gathered from many lands, smite great Aias, son of Telamon, with spears full upon his shield, and ever press upon him ' (Il. 11,[563][564][565] ὣς τότ' ἔπειτ'=ADVs (Modal / Temporal / Temporal), Αἴαντα μέγαν Τελαμώνιον υἱὸν=OBJ, Τρῶες ὑπέρθυμοι πολυηγερέες τ' ἐπίκουροι=SBJ The adverbs at the beginning, ὣς, which connect this clause as a whole with the preceding sentence, and τότ' and ἔπειτ' , which are temporal, can easily be interpreted to be "Setting." The Object (we may consider that Αἴαντα μέγαν Τελαμώνιον υἱὸν is logically governed by ἕποντο, even if it is formally governed by νύσσοντες from a grammatical point of view) is clearly pragmatically marked: Topic (Aiax has explicitly been mentioned in v. 556, and the simile of vv. 558-562 refers to him) or perhaps better Focus (as the way in which Ajax is referred to seems to show: not with the simple anaphoric ὁ[τὸν in this case], but with an heavy name with epithets).

3.2.2.5
OBJ -SBJ -ADV order There are two instances in which the Object is pragmatically marked. The rest of the sentence appears to be unmarked. Adverbials are both Modals and Nonmodals: αἰὲν ever ἕποντο follow:impf.3pl 'Even so then did the Trojans, high of heart, and their allies, gathered from many lands, smite great Aias, son of Telamon, with spears full upon his shield, and ever press upon him ' (Il. 11,[563][564][565] ' (Il. 15,204) πρεσβυτέροισιν=OBJ, Ἐρινύες=SBJ, αἰὲν=ADV (Temporal) πρεσβυτέροισιν here refers to Zeus, the Topic of the discourse, already quoted in v. 202. For a complete analysis of the passage, see Kirk (1985Kirk ( -1993   view (SBJ, ADV …) unless it has a pragmatic function (Setting, Topic …). There is no distinction between elements (Adverbials, Topics …) of the same type (e.g. in the case of three temporal adverbs only ADV [Non-modal] is written). There is no distinction between different types of Objects. An asterisk marks exceptions.

ἄγω, δίδωμι
The other two case studies (ἄγω and δίδωμι) have been analyzed in the same way as ἕπομαι. As stated in §3.1, for the sake of brevity, only the synoptic tables are provided (Tab. 2, Tab. 3).   The structure of the left periphery which results from this model is analogous with the one described in pattern 13 (cf. §1.4), which I present again here: (45) (Setting -Topic -Focus). NB: Unmarked elements follow.
The two main characteristics (cf. §1.4) are also comparable, as shown here: a) After the final marked element, the Verb does not necessarily follow. b) Not only are the Setting constituents and the Topic optional, but the Focus is also optional. Consequently, the analysis of the examples confirms the suitability of the new model for the AG left periphery suggested in §1.4. It is worth repeating that this analysis, and particularly pattern 13 (=45), is not dramatically different from Dik's pattern, but, rather, can be considered an extension of it. As noted in § 1.5, most of AG sentences (or cola)62 have a quite simple structure, and there are usually no more than two constituents to the left of the Verb. Therefore, their structure will often be Topic -Focus -Verb, which is consistent with both Dik's model and the model proposed in this study.
Finally, a note about the range of conclusions given in this contribution must be added. The starting point for this study was a general (that is, valid in general for AG) theory, which I proposed to demonstrate based on examples drawn from just one author: Homer. In §2.1, I assumed the validity of this method, showing that Homeric poems are more significant for linguistic and syntactic analysis than many other AG works. On the other hand, it is also true that further research into a more extensive corpus is needed to confirm the hypothesis behind this assumption. Consequently, at the present time, it would be prudent to consider the conclusions drawn here (and particularly the material in the preceding section) to be relevant to Homeric Greek at least, if not to AG in general. beschi Journal of Greek Linguistics 18 (2018) 172-210 whom the Homeric unmarked word order is SOV/SVO; etc. Nevertheless, at the present time, the existence of such an unmarked word order is usually denied, mainly because of its many exceptions. On the other hand, exceptions should not be taken to imply that a rule does not exist; they can be explained based on the important role played by pragmatics in AG.

2
Pragmatics and syntax This suggestion, if confirmed, would offer the opportunity to reply to an objection that may be raised to the model proposed in this contribution (13=45). It might seem that this model, by increasing the number and range of unmarked elements and, consequently, restricting the role of pragmatics in AG sentences, has less explanatory power than Dik's model. On the other hand, what is observed in §1 of this Appendix opens the way to another possible explanation-even if only tentative at this stage. In other words, the connection between the explanatory power of a word-order theory and the number and range of marked or unmarked elements is justified only if word order is explained on an exclusively pragmatic basis. If one postulates an interaction between pragmatics and syntax, things change.

3
Open questions On the other hand, the suggestion proposed in § 1 of this Appendix also raises some questions, particularly about the relationships that occur between pragmatics and syntax and the way in which two such factors "work together". Here, I restrict myself to saying that the co-occurrence of pragmatics and syntax in determining word order has been theoretically very well explained by a linguistic theory that "makes possible a very transparent approach to the interface between syntax and semantics -pragmatics" like the cartographic approach to sentence structure.65 I will not add anything more on this point since this 65 For an introduction to the cartographic approach, see Cinque & Rizzi (2010) (from which the preceding quotation is drawn, p. 63) and related bibliography (particularly, Cinque 1999). For cross-linguistic studies about the left periphery of the sentence in a cartographic perspective, see in particular Rizzi (1997), Benincà (2001), Benincà & Poletto (2004), and Benincà & Munaro (2011). The pattern of the left periphery proposed by Benincà & Poletto (2004) [ more general explanatory aspect-like the description of the right part of the sentence-goes beyond the scope of this contribution. As already stated in § 1.4, this contribution aims to provide a "pattern"-that is, a description-"for the AG left periphery", and so has two limitations: the descriptive aspect (more than explanatory) and the focus on a single part of the sentence. What is new in this contribution is pattern 13 (=45); for a theoretical explanation or a wider description (i.e. of the whole sentence), further studies are required.66