Bryn Mawr Classical Review 2008.06.08
Helma Dik, Word Order in Greek Tragic Dialogue. Oxford: Oxford University Press, 2007. Pp. xvi, 281. ISBN 978-0-19-927929-6. $99.00.
Reviewed by D.M. Goldstein, University of California, Berkeley (firstname.lastname@example.org)
Word count: 3991 words
Table of Contents
Word-order variation in Greek and Latin remain thorny problems for philologist and linguist alike. In spite of this, systematic investigations of the topic are disconcertingly few. For this reason alone, one heartily welcomes Helma Dik's (= D.) new work, Word Order in Greek Tragic Dialogue. This volume follows her 1995 companion opus, Word Order in Ancient Greek (BMCR 95.11.08), which centers on Herodotos. As with the earlier work, D. argues here that word order in Greek is pragmatically conditioned. While couched in the framework of Functional Grammar developed by Simon Dik (no relation to the author), the book is nevertheless accessible to a general Classics audience: graduate students and more advanced scholars will have no trouble with it, and advanced undergraduates could probably benefit from it as well.
All in all, the work is a measured success. On the one hand, I recommend it without hesitation: one is hard pressed to think of another work on Greek word order that is as copiously illustrated or theoretically sophisticated. It is a potent antidote to the irritatingly prevalent notion that Greek word order is "free," and it will undoubtedly become required reading on the topic of word order in tragedy, as her 1995 work has already become for prose. At the same time, I cannot with equal alacrity second all her analyses and conclusions. While many of them seem successful, the work is too often too short on explanation (whether in the interests of mere clarification or for justification). Often I found myself wondering how exactly D. came to certain judgments about the pragmatic status of various constituents, or how she would defend her analysis against competing stories. Below I present a brief chapter-by-chapter overview of the book, followed by more detailed comments on selected chapters.
The book begins with a short chapter that introduces some elements of pragmatics, lays out the extent of the study, and offers a first illustration of the method and the type of questions asked. This is followed by a more substantial chapter two on the theoretical machinery of pragmatics--the essential ideas of Topic and Focus, as well as Theme, Tail, Setting, and Remainder. From here, we move from theory to illustration, as Dik examines constituent-order at the clausal level. To narrow her focus, she restricts her corpus to clauses with the verbs θνήισκω and ὄλλυμι. True to Denniston's belief that the reader "should be able to bathe in examples,"1 D. provides a veritable flood (both in this chapter and throughout the book): this is a conspicuous strength of the volume. From the clause we move to the noun phrase (= NP), to look at the internal ordering of noun and adjective. D. argues that with noun-adjective sequences, we are either dealing with a neutral configuration or one in which the noun is more salient; in adjective-noun sequences, the adjective is more salient. She bases her study on the adjectives μέγας, πατρῶιος, and possessive adjectives. From the NP we move back to the clause, this time the interrogative. D. examines the pragmatic ordering of both yes-no interrogatives and wh-interrogatives; within the latter set, she also looks at so-called "postponed interrogatives," that is, clauses in which the wh-word is not clause-initial. The sixth chapter then moves in a different direction altogether, and considers the relationship between word order and meter. (Up until this point, D. discusses tragic dialogue as if it were ametrical speech.) The penultimate chapter is a Leseproben of four passages from Sophocles' Electra in which Dik brings together all her pragmatic tools to discuss word order. A short conclusion caps the work.
Constituent-order at the Clausal Level
Past research on Greek word order has tended to rely on the categories subject (= S), object (= O), and verb (= V), and led to a debate about whether Greek is an SOV or SVO language. D. takes a different approach, however: she relies on informational categories like Topic and Focus, rather than structural ones, like subject and object. The over-arching claim of D.'s work is that Greek word order at the clausal level is organized according to the following pragmatic template:
Roughly speaking, the Topic is the constituent that a clause is "about." More fully, in D.'s words (p. 31), "Topic function is assigned to an element which the speaker regards as an appropriate foundation for constructing a message which is relevant to the subject matter of the discourse." The Focus, by contrast, is the most "salient," i.e. "important," piece of information in a clause. In her words (p. 32): "Focus function is assigned to an element expressing the information that the speaker considers the most urgent part of the message s/he wants to convey to the listener." Cross-linguistically, Topic-Focus ordering is by no means uncommon: if ultimately true for Greek, it would place the language among a number of similarly configured languages throughout the world. For classicists, the pragmatic approach will perhaps come as more rewarding or engaging than a more formal or generative-syntactic approach: D.'s study is anything but a dry analysis of words in sentences, and all parameters of communication are relevant: what is being said, by whom, to whom, when, why, where, etc.
Occasionally D.'s presentation here lacks clarity. On p. 17 she writes: "I believe that the fundamental characteristic of Greek word order is to place pragmatically marked constituents early in the clause." At the same time, Topic is organized before Focus; and Topic is supposed to be the grounding for the Focus. Why then is Topic regarded as pragmatically marked? D. answers this question by observing (pp. 31-32) that many clauses have no overt Topic,2 such as those in which there is no change of Topic from one clause to the next, or a Topic is otherwise unnecessary or irrelevant. So when a Topic is expressed, it is most likely new or contrastive. But this still allows for Topics that are expressed though neither new nor contrastive. Following D.'s scheme above, a Topic of this kind would be positioned before Focus; but if so, this then violates her general position that "pragmatically marked" (whatever that is supposed to mean exactly) constituents come early in the clause. A problem of this type is illustrated below by a passage from Ajax.
One problem besetting a pragmatic approach like D.'s is that the categories of Topic and Focus are notoriously difficult to define explicitly. Accordingly, judgments about what constitutes a Topic and what a Focus can be distressingly subjective, a problem that D. herself is aware of (pp. 33-34). To avoid this problem, she takes care to provide a bounty of examples, and goes through them in some detail. What she generally does not do, however, is explain how alternative pragmatic interpretations of a sentence could be ruled out. Take for example El. 582-583: εἰ γὰρ κτενοῦμεν ἄλλον ἀντ' ἄλλου, σύ τοι πρώτη θάνοις ἄν, εἰ δίκης γε τυγχάνοις. D. analyzes the lines thus: "From a general 'mortals' and 'we', the subject σύ is singled out as the Topic. πρώτη is Focus of the clause: 'You would die first.'" I have a hard time understanding why πρώτη should be Focus and not σύ τοι. For it seems possible to translate the sentence felicitously as, "the first to die would be you." Questions like this and the general subjectivity of the endeavor could have been removed (or at least diminished) if a more precise and rigorous definition of Topic and Focus had been provided.
I had a similar difficulty with D.'s discussion of Aj. 970, which runs: θεοῖς τέθνηκεν οὗτος, οὐ κείνοισιν, οὔ. Here D. identifies θεοῖς as Focus ("it is the gods that killed him, not they, no!"), which seems correct. Of οὗτος (= Ajax), which I thought patently Topic, she writes (p. 48): "... [it] follows the verb. It falls in my category of 'Remainder,' that is, of those constituents that do not have Topic or Focus function." But if it is neither Topic nor Focus, then what sort of pragmatic status is Remainder supposed to have? D. characterizes these post-verbal Remainder subjects as non-contrastive and "predictable" information. And indeed οὗτος could fit this description. But if it is pragmatically so weak, why is it realized morphosyntactically at all? And still one wants to know: why is it not Topic? The Topic of the previous line (if it is to be preserved) is Ajax, and here again he seems to be what the sentence is "about." D. writes (ibid.) further: "Placing it [= οὗτος] earlier in the clause would presumably have had an unwanted contrastive effect." Why? Are Topics restricted to new or contrastive units of information? These questions do not necessarily condemn her analysis or framework, but they are prime examples of how the book is too short on explanation. Problems of this sort could probably have been worked out with a more refined typology of pragmatic statuses.3
To return to the template above in (1), two extraclausal adjuncts are possible, Themes and Tails. Theme constituents precede the clause proper, while Tail constituents follow it. Both form their own intonational phrases. In the following example (Thuc. 6.9.3, quoted on p. 35), the prepositional phrase is a Theme:
(2) καὶ πρὸς μὲν τοὺς τρόπους τοὺς ὑμετέρους ἀσθενὴς ἄν μου ὁ λόγος εἴη ...
According to D., the "clause proper" begins with ἀσθενὴς, which she deduces according to the position of ἄν, which needs to be in "second" position. The pragmatic status of Theme and Tail is not described in any detail; indeed, their most prominent feature seemed to be prosodic as opposed to pragmatic. D. says that Themes are "especially suitable when the speaker introduces a new (Discourse) topic" (p. 35), which (2) nicely illustrates. She also points out that the host of the postpositive may also receive "additional prominence" (p. 35); this point is not developed, and if (2) illustrates this it is not clear how. She then adds (p. 36)-- and here her reasoning seems to stray a bit -- that "Theme and Tail constituents should by definition be analysed as Focus (the most salient part of the intonation unit), but the clause itself will have its own Focus constituent within it." But if Theme and Tail are always Focus, why is Theme said to be so suitable for introducing a new discourse topic? D. adds a syntactic requirement (ibid.): "An element that cannot be left out, without the resulting clause becoming ungrammatical cannot be a Theme or Tail." While often true, counterexamples to this claim are not exactly difficult to find, as we see here at Hdt. 6.98.18:
(3) τούτους μὲν δὴ τοὺς βασιλέας ὧδε ἂν ὀρθῶς κατὰ γλῶσσαν τὴν σφετέρην Ἕλληνες καλέοιεν.
Here the verb is καλέω, which, when used in the sense of 'call by name, name' (LSJ, s.v. II) typically selects for two accusatives (i.e., 'call X Y' where both X and Y are assigned accusative case). Here it selects for τούτους μὲν δὴ τοὺς βασιλέας and the anaphoric adverb ὧδε, which refers to NPs in the previous sentence. τούτους μὲν δὴ τοὺς βασιλέας forms its own intonational phrase, which causes postpositive ἄν to be positioned after ὧδε. Prosodically it looks just like the sentence-initial prepositional phrase in (2), but is lexically governed by the verb. Lastly, what I wanted to know from D.'s discussion is what pragmatically triggers the creation of a Theme or Tail (which are elsewhere known as preposing and postposing constructions, respectively), but this question is given only scant consideration.4
Constituent Order and Metrics
In chapter six, D. turns to the question of the relationship between meter and word order. Up until this point in the book, she has discussed word order in tragedy as if she were dealing with prose texts. Here D. claims that the end of the trimeter line is not a position of emphasis (by contrast, line-beginning is). D. cites the following example from OT 122-123: ληιστὰς ἔφασκε συντυχόντας οὐ μιαῖ | ῥώμηι κτανεῖν νιν, ἀλλὰ σὺν πλήθει χερῶν. She claims that the position of ληιστὰς and μιαῖ can be accounted for solely in pragmatic terms.
I was not persuaded by this claim, and thought her discussion at this point needed to be more refined. For there are three levels of positioning at issue here. The first is that of constituent order within the phrase, e.g. the adjective preceding the noun in μιαῖ ῥώμηι. The second is that of the position of the phrase within the clause. The third is the alignment of the clause and the trimeter. Constituent-ordering of the first and second levels predicts nothing about the positioning at the third level, although of course metrical line and morphosyntactic clause often do correspond in a one-to-one relationship. To illustrate this point with μιαῖ ῥώμηι, D.'s analysis will get the ordering adjective-noun and its place within its clause. What it will not get us is an explanation for why μιαῖ | ῥώμηι is positioned where it is in the line, or why it is split by a line-break.
D.'s point is that, since the sequence adjective-noun marks adjective salience (as claimed in chapter four), the marked pragmatic status of μιαῖ is already accounted for before metrical alignment even comes into the picture. Thus, line-end position does not engender any marked pragmatic status. Fine, but this still does not explain why μιαῖ is positioned where it is in the line, nor does it rule out the possibility that there is some other pragmatic effect created by line-end position. For given that line-end is an acoustically marked position (presumably characterized by pause and perhaps lengthening), how could it not have any pragmatic effect? A problematic assumption underlying her analysis is that if a constituent in line-final position is not salient (according to her definition of salience), then such a position cannot be "emphatic." But there are other reasons besides "salience" as to why a constituent might be "emphasized"--e.g. atypicality of the proposition or argument structure, general discourse prominence, constituent affirmation, affective meaning, register-heightening, etc.
Chapter six also contains an interesting discussion of enjambment. D. follows earlier discussions of this phenomenon by distinguishing two types, "necessary" and "adding" enjambment. "Necessary enjambment" refers to "syntactically necessary" enjambed constituents,5 as with θάνηις in the response of Orestes at El. 1503-1504: μὴ μὲν οὖν καθ' ἡδονὴν | θάνηις. Adding enjambment refers to the enjambment of syntactically non-necessary constituents. So Αἴγισθον from El. 954-957: νῦν δ' ἡνίκ' οὐκετ' ἔστιν, ἐς σὲ δὴ βλέπω | ὅπως τὸν αὐτόχειρα πατρώιου φόνου | ξὺν τῆιδ' ἀδελφῆι μὴ κατοκνήσεις κτανεῖν | Αἴγισθον.
D. claims that with necessary enjambment, it is not the enjambed item that is "emphasized," but rather the one preceding it (in the line previous). Only with adding enjambment is the run-on item "emphasized." The problem that I have with her claims here is the same one I had earlier with items at line-end: an enjambed item is acoustically marked, that is, you have (typically) one lexical item surrounded by two pauses (one brought about by the preceding line-end, and one brought about the by the fact that the enjambed item is the last in its sentence). Typically within a clause individual lexical items are not fenced in by pauses like this, unless you are dealing with, e.g., appositional phrases, or a pre- or postposed constituent. But if we follow D., this marked acoustic prominence has no pragmatic effect. If she is right about this, then her argument needed to address this question more fully.
To return to El. 1503-1504, D. claims that the Focus of the clause is καθ' ἡδονὴν. Here we again face the problem mentioned above regarding identification of pragmatic constituents.6 For I would identify the Focus as καθ' ἡδονὴν θάνηις. D. acknowledges the acoustic prominence of θάνηις (given its line-initial position and pauses on either side) but claims that there is "no real need to understand it as pragmatically marked" (p. 180). In support of her decision, D. could claim that Aigisthos' imminent death is contextually available information, and therefore there is nothing pragmatically marked about θάνηις. On the other hand, Orestes' response is contrastive and counter-presuppositional. He orders Aigisthos to walk ahead of him (ln. 1502), to which the latter responds, "in order that I do not get away from you?" In Aigisthos' question, we have his assumption of Orestes' behavior. From Orestes' rejoinder, however, we see that he has quite a different idea in mind. It seems that the whole negated purpose clause is designed to be contrastive or "marked" information, and not just the prepositional phrase. It is thus questionable whether καθ' ἡδονὴν alone can really be considered Focus. Regarding the separation between this phrase and θάνηις, this may well have been done to "emphasize" the atypicality of the collocation and the proposition as a whole. For, after the prepositional phrase, we might well have expected a verb like 'go,' 'behave,' or 'do.' While it is contextually derivable that Aigisthos is going to die, it is certainly not known that Orestes is preoccupied with the manner of Aigisthos' death--and herein lie both the newness and the atypicality of his proposition. This atypicality is not really composed until the final verb is delivered, and the enveloping pauses highlight this. To hold that θάνηις is pragmatically unmarked does not seem to me a viable position.
With enjambed proper names (pp. 192-196), D. comes to the opposite conclusion, that line-initial position serves a highlighting function. Most of the examples she analyzes are what she calls Tail constituents, that is, syntactically adjoined constituents co-referential with some previous NP. She considers the effect of their position more akin to clause-initial rather than clause-final position. So the address to Helios at Aj. 845-846: σὺ δ', ὦ τὸν αἰπὺν οὐρανὸν διφρηλατῶν |Ἥλιε. D. claims that the proper name here is "strictly speaking predictable from the preceding line" (p. 193). This is not the case: while the referent of the proper name has already been established, this in itself does not predict the appearance of the proper name itself. As for a line-initial highlighting function, here I would incline to a stance opposite that of D. Vocatives in non-metrical environments are often utterance-initial and prototypically fenced in by pauses. Given this, vocatives like that at Aj. 845-846 are acoustically no different from what we would find in a non-metrical text. In other words, unlike the example discussed in the previous paragraph, there is nothing acoustically non-canonical about the vocative here, and thus one wonders how D. envisages the position serving a highlighting-function.
As far as the big picture is concerned, my most serious reservation is that D.'s account allows little to no room for other components of grammar besides pragmatics. To be sure, D.'s aim is to show how Greek word order is pragmatically organized. But at various points I had the sense that to her pragmatics was everything (even if she does not consciously subscribe to this opinion) and that there was little room for other factors, such as prosody, metrics, semantics, or syntax. D. must be right that a good deal of constituent-order variation in Greek is pragmatically conditioned. But I at least am not persuaded that it can account for all that she thinks it can, nor that some of these pragmatic patterns have not been grammaticalized. For there may well be an underlying syntactic configuration to Greek (whether SOV or SVO), which is simply more prone to pragmatic reorganization than, say, English. Even D.'s template is not a purely pragmatic one, as she makes reference to the category Verb.
A related problem is that D.'s approach to pragmatics is narrow. She relies exclusively on one theory, namely that of Functional Grammar, which is a fairly insular and marginal school. Other approaches to pragmatics are theoretically richer and more nuanced (for examples, see the works cited in footnotes two and four). I have some worry that classicists will think that D.'s approach represents the field of pragmatics more broadly.
One over-arching question not discussed is what role prosody and specifically intonation played in signaling pragmatic status in Greek. In English (not to mention a vast number of other languages), the informational status of a constituent can be indicated via prosody (e.g. pitch, loudness) as opposed to word order, and similar mechanisms were likely to have been available in Greek. Obviously this is a very difficult issue to wrestle with when it comes to a corpus language, but nevertheless the question must be asked, as it has implications not only for the role that D.'s pragmatic template plays in ordering constituents but also in our broader understanding of Greek grammar. In addition to intonation, one wonders how particles (in their capacity to signal the information status of constituents) interact with D.'s pragmatic template.
A few small gripes. Some of the terminology is unfortunate, especially 'Mobile.' This term goes back to Dover (1960)7 and should be abandoned. It essentially refers to prosodically independent words, i.e. items that are neither pre- nor postpositive. The belief behind the term is that such words are relatively free as to where they can be positioned. There are, however, a good number of words that are prosodically independent but nevertheless are not so "mobile." Dover himself recognized this, and referred to such words as "preferential Mobiles." There may well be more such words (or contexts) than have so far been identified. Nothing is gained by the use of the term, and "constituent" or even "word" (in the sense of prosodically independent word) works just as well.
D. sometimes makes peculiar comments about postpositives. On p. 197, discussing line-initial σέ at Aj. 1015, she claims that it shows "more of a resemblance with postpositives," and suggests that its host is the final noun of the previous line. I at least am aware of no examples of clitics hosted by an item in a previous line, and if D. really does believe in such a phenomenon, something more should have been said on this point. On p. 220, she claims without explanation that the phrase πρὸς θεῶν is "used postpositively" at OT 1410, but I see no parallel between this phrase and traditional second-position items. Earlier on in the work (p. 18), she writes, "The rules that govern the placement of pre- and postpositives are predominantly syntactic and are relatively well established." Our understanding of clitic behavior is not nearly as good as she seems to think. Moreover, whether clitics in Greek are syntactically or prosodically arranged is debatable;8 what she means by "predominantly syntactic" is not explained, and would seem to contradict what we read on p. 21: ". . .it was probably prosodic peaks more generally, rather than only first words of clauses, that attracted postpositives." Her examples of postpositive-placement on p. 20 are not difficult to account for under a Kolon-type analysis.
The title of the work is slightly misleading: the phrase "Greek tragic dialogue" sets up the reader to expect a more encompassing study: while all three tragedians are represented to an extent, greatest attention by far is paid to Sophocles.
Lastly, the target audience. I was surprised and disappointed to find that examples were not quoted in both Greek and Roman scripts. The decision not to transliterate the Greek has severely limited D.'s audience: effectively, only classicists and the occasional linguist with some training in Greek will be able to read it. This is a topic that desperately needs the attention of both classicists and linguists.
To conclude, this is a valuable book, but one not without its problems. It can be read with profit as there is a good deal that D. must be right about. More, however, in the way of explanation was needed. Word-order variation is no doubt one of the most difficult problems in Greek, and my criticisms here should be taken as questions that remain to be answered: the nature of the topic is such that it will require ongoing consideration from both philologists and linguists.
1. J.D. Denniston, The Greek Particles (Oxford University Press, Oxford: 1954): vi.
2. Her discussion of whether clauses have Topics even if there is no overt morphological Topic constituent needed to be more nuanced. For consideration of this and related theoretical questions, see e.g. N. Erteschik-Shir, Information structure: the syntax-discourse interface (Oxford: OUP, 2007).
3. Here I have in mind something like that found in T.E. Payne, Describing Morphosyntax (Cambridge, CUP: 1997): 261-294.
4. Surprisingly there is no reference to the work of Gregory Ward on this or any other topic in D.'s book. He is one of the leading authorities on information structure and constituent order. See e.g. B.J. Birner and G. Ward, Information Status and Noncanonical Word Order (Amsterdam/Philadelphia, John Benjamins: 1998), as well as Ward's website.
5. Her terminology here is unfortunate. The first problem is that "necessary enjambment" seems like it should refer to a phenomenon whereby a constituent must be enjambed. The second problem is the notion of "syntactical necessity." Given the phenomenon of ellipsis, D. at points has to temper the idea of "necessity," e.g. p. 182 n. 28, pp. 184-185. Had she simply used the term lexically-governed enjambment instead, she would have avoided both problems.
6. Other points at which I wonder about her pragmatic judgments: (6.6) on p. 183; (6.9) on p. 186; (6.10) on p. 187; (6.11) on p. 188.
7. K.J. Dover, Greek Word Order (Cambridge: CUP, 1960).
8. For discussion on prosody versus syntax in accounting for clitic positioning, see e.g. A.L. Halpern and A.M. Zwicky, edd. Approaching Second: Second Position Clitics and Related Phenomena (Stanford: CSLI, 1996).