At first, it seems that the title of this book is too broad: how can a discussion of second-position words in Herodotus encompass any significant amount of Greek syntax? But in fact that’s exactly what happens: in the course of exploring Wackernagel’s Law, Goldstein manages to make all sorts of useful observations about word order, clause structure, and sentence structure in 5th-century Greek prose. As Goldstein puts it, his goal is “the demonstration of the value of clitics as diagnostics for syntactic and prosodic structure” (p. 294). All of the examples are transliterated, translated, and furnished with morpheme-level glosses, so the book will be accessible not only to classicists but to linguists in other areas.
In the first semester of Greek, we might explain Wackernagel’s Law as the tendency for the little words to come second, leaving vague exactly what we mean by “little words.” “Come second,” though, is also more complicated than it seems, and this is one of Goldstein’s main themes: clitics may be second in the sentence, in a clause, or even in a phrase, and a single Greek sentence may have two, three, or even more “second” positions. Moreover, identifying those positions tells us something about the structure of the sentence, both syntactically and rhetorically. Clitics can mark topic or focus, and can determine the relationship of a participle or infinitive to the main verb of the sentence. Goldstein uses a QUD framework (“question under discussion,” as developed by Craige Roberts and Daniel Büring, among others) (p. 29) and describes Greek prose in terms of the information structure of its sentences, nowadays a widely used approach (p. 35, with references and relationship to prior work). He argues that Greek word order is not based on an arrangement of syntactic constituents (the rule is not SOV), nor of Topic and Focus (the rule is not Topic-Focus-verb), but of parts of the answer to the QUD: a pre- posed topic usually relates to a hierarchy of sub-questions (p. 11), and a pre-posed focus can “counter an assertion in the Common Ground of the discourse” (p. 12), that is, introduce material that is new or even surprising. In short, “Greek word order exhibits systematic correspondences between structure and meaning” (p. 290), and the placement of clitics helps us figure out what those correspondences are.
The book is organized into three parts: “Foundations,” “The Left Periphery,” and “Clause Combining.” In the first part, after a brief introduction and an overview of Greek as a “discourse-configurational” language, Goldstein supplies an overview of the prosody of clitics (chapter 3). He argues that Greek clitics “are canonically hosted by the first prosodic word of the clause” (p. 69), or of whatever smaller domain they belong to, but some sentence-level clitics, notably γάρ and δέ, can be hosted by the definite article, as if the article has been promoted from proclitic to full word. Moreover, the host of a clitic doesn’t have to be a complete syntactic unit; the pattern adjective-clitic-noun is not uncommon, though determiner-noun-clitic is also quite common. Even though there is not much in this discussion that is really new, the chapter is a strong overview of what’s known, drawing evidence not only from Herodotus but from tragedy and from inscriptions as well.
In the next chapter Goldstein considers the syntax of clitics. The essential contribution here is the classification of these words by their domain: some operate at sentence level, like γάρ, some on clauses (for example, forms of εἰμί), and some within a phrase, like γε. Which level a clitic belongs to generally depends on its semantics, and there are exceptions: the modal particle ἄν, for example, is a word-level clitic when it generalizes or widens the application of the word that hosts it (like the -ever of English whenever or whoever), but a clause-level clitic when it has its modal sense, for example marking an optative as potential (p. 89). Clitic pronouns operate at clause level, and they are “syntactically deficient” (p. 86): they can’t be the antecedent of a relative pronoun, they are rarely the object of a preposition, and they don’t have attributive modifiers.
This classification by domain or scope explains why “second position” doesn’t mean “the second word of the sentence.” Rather, a clitic will go to second position in its domain. Thus ἄν is second in the clause when its scope is over the clause, but second in a phrase (that is, after the word it affects) when it has phrasal scope. As a result, there can be several clitics separate from each other in the same sentence; Goldstein delightfully calls this “splaying” of the clitics (p. 88). Not only that, though: some kinds of adverbs are hierarchically above clitics, and come before the word that is host to the sentence clitics, with the result that the clitics appear to be in third position. Exactly what kinds of adverbial expressions have this property is a bit difficult to pin down, but the important point is that they are adjoined quite high in the syntax tree (to S/CP or above). This chapter is full of examples; the treatment of pronouns is also particularly good.
The second part, “The Left Periphery,” comprises a chapter on topicalization (chapter 5) and one on focus preposing (chapter 6): in other words, these are the two mechanisms by which a phrase can move out of its expected position to sit at the left edge of a sentence. How can we tell them apart? By the accompanying clitics, Goldstein claims. In particular, a preposed Topic will be followed by μέν or δέ, but a Focus phrase will not (p. 121, 216). Typically, μέν and δέ are used when the QUD is explicit, and successive utterances run through the full list of answers to it. For example, in Hdt. 1.42, Croesus asks Adrastus to go hunting with his (Croesus’s) son. Adrastus’s answer is: under other circumstances (first possibility, marked with μέν), no; but as things stand (second possibility, marked with δέ), I will (p. 131–2). Then the so-called “μέν solitarium,” Goldstein observes, is a variant of this structure, in which the first answer to the QUD is given, and the remaining ones are left implicit.
Another topicalization pattern marks the end of a QUD (p. 140). A sentence starting with ταῦτα μὲν δή or the like, with an anaphoric expression and μέν, sums up. Although Goldstein doesn’t say so, often the next sentence will have an answering δέ, setting up the next QUD.
The third part is the strongest and most original. Chapter 7 treats participles, chapter 8, infinitives. Goldstein divides participial phrases and infinitive phrases into phrases strictly so called and clauses, which are phrases that act syntactically like clauses with finite verbs, usually adverbial clauses. This refines the familiar distinction among attributive, circumstantial, and supplementary participles: the circumstantial participles are the clauses, more or less (p. 221). More precisely, among circumstantial participle phrases there are “participial clauses, VP-participial phrases, and chained participles” (p. 223). Not surprisingly, clitic placement shows us the difference. A participial clause is its own domain for clitics, so there may be clitics both there and in the main clause; other participial phrases are part of the same clitic domain as the rest of the sentence (tidily summarized in a chart, p. 259).
Chained participles represent action that goes along with the main verb; English would typically use two coordinated main verbs, as “they would reap crops and sail” rather than θερίσαντες δ' ἂν τὸν σῖτον ἔπλεον with a participle (Hdt. 4.42.4, cited p. 254). As for the other two types, “whereas participial clauses typically provide information about a proposition (the finite clause), VP-participial phrases modify the internal structure of the event described by the finite clause” (p. 238). Participial clauses may be temporal, causal, concessive, or conditional (examples p. 227–228), but VP-participles merely clarify or elaborate on the action of the main verb, and may be the focus of the sentence. Genitive absolute phrases are usually participial clauses, occasionally VP-participles (an example is 5.106.5, cited p. 250). Supplementary participles are VP- participles, though the participles of “indirect discourse” with a verb of perception may be either VP-participles or clauses, the latter typically “when the embedded participial is not perceived visually, but mentally” (p. 248).
The situation with infinitives is similar. The essential distinction is the subject of the infinitive: if it is an argument of the main verb, then the infinitive phrase is part of the same clitic domain, but if the subject of the infinitive is not related to the main verb, then we have an infinitive clause with its own clitics (p. 261). Syntactically, that infinitive clause is its own S. Goldstein observes that this distinction is not simply the difference between infinitives in indirect discourse and other infinitives, for “one and the same predicate can select both a VP-infinitive and an S-infinitive” (p. 281); he gives examples with φημί, ἐλπίζω, and εὑρίσκω. There seems to be a semantic distinction, though Goldstein says the details are “beyond the scope of this investigation” (p. 284).
The ninth and final chapter is a brief conclusion with pointers to next steps. In particular, this work studies Herodotus. What about other authors? Beyond Thucydides, Goldstein suggests that metrical texts may also be interesting and claims that the patterns of clitic distribution in Attic drama “are far more diverse than in any other genre in classical Greek” (p. 293). He also observes that Proto-Indo-European presumably had second-position clitics, but we need more detailed studies of the daughter languages to flesh out how Wackernagel’s Law worked in PIE (p. 293, and footnote 3 on p. 4).
Throughout the book, Goldstein is honest about what he doesn’t know, and generous with suggestions for further work. For example, how are clitics arranged when two or more come together? “So little is known about the structure and ordering of clitic chains in Greek” (p. 91). In the “QUD-terminating” topicalization pattern, why is it sometimes μὲν δή and sometimes μέν νυν? “I leave for future research the difference in discourse function between the two” (p. 143). How does a VP-participial phrase fit into the syntax? They are somewhere under S, certainly, but “where exactly VP-participial phrases occur in S is difficult to determine” (p. 225).
With this study, then, Goldstein has shown what Wackernagel’s Law really is in classical Greek (or at least in the prose of Herodotus), and how clitics contribute to our understanding of the structure of a Greek sentence. This is a fascinating and useful study, and a sound basis for further work.