It is no easy thing to reconstruct the syntax of a language that is only known through its descendants. But in the case of Proto-Indo-European (PIE), linguists have been sorely tempted to do so, partly because there are some inviting starting points, one of the most promising of which is the relative clause. Indeed, it is almost too promising, since the Greek relative ὅς and its Sanskrit cognate yá- point to an ancestral form *Hyo-, but Latin quī and Hittite kuiš go back to a stem *kwi-. Some Indo-Europeanists have suggested that one or the other of these two pronouns was the original relative, ousted by the other in some branches, but others have instead sought a functional difference between the two, especially since early Greek and Sanskrit relative clauses more often occur after their antecedent and further specify it or add new information (as in the first relative clause of the Iliad: Μῆνιν … ἣ μυρί’ Ἀχαιοῖς ἄλγε’ ἔθηκε), while relative clauses in early Latin and Hittite texts are often placed before their antecedent, which is typically a correlative pronoun, e.g. quam earum in iis locis optimam dicent esse, eam maxime serito (Cato, Agr. 6.1). But perhaps the apparent difference between the two types is only a side effect of a difference in register: Homeric epic and the Vedic hymns might give more occasion for one type of relative clause, the legal language of most early Latin and Hittite texts might elicit another. What has been needed to advance this discussion is a careful study of the actual usage of early Greek relative clauses. It is this gap that P.'s book fills admirably.
General linguistic and Indo-European concerns predominate in the first four chapters, in which P. gives a brief introduction to her corpus and the questions she will and will not cover (Chapter 1), situates the concept of subordination in early Greek in the broader context of the linguistic development from PIE to Greek, with a salutary reminder that, although we are accustomed to consider Homer paratactic, there is still quite an amount of subordination (Chapter 2), introduces the debate on relatives in PIE (Chapter 3), and defines relative clauses (Chapter 4), as well as some important subtypes – restrictive, non-restrictive and clauses like This was what he wanted , in which the antecedent is only implicit, making the clause as a whole equivalent to a noun, rather than an adjective. P. calls the last type ‘inherently maximalizing relative clauses’, because the noun phrase to which they are equivalent ‘picks out a unique entity, everything in a set, or a complete lot of stuff’ (p. 72). With Chapter 5, the book pivots towards material of more direct interest to classicists. It begins with a theoretical discussion of (in)definiteness that repays careful reading, partly because of its attention to the conditions under which Greek uses ὃς ἄν with the subjunctive, rather than ὅς with the indicative, partly because of its analysis of ὅστις as a ‘domain-widening’ pronoun, indicating that the set denoted by the relative clause has to be as broadly conceived as possible. That all this is not just theory for theory's sake becomes clear in the final section of the chapter, a seven-page discussion of the ὄττις ἐνάντιός τοι clause in Sappho 31, which applies all this terminological apparatus to analyse the numerous interpretations proposed for these lines.
Chapter 6 goes still further in working out a taxonomy of relative clauses in early Greek, first accounting for the various relative pronouns, then establishing a twofold division into (i) postnominal relative clauses and (ii) assorted inherently maximalising relative clauses, including free relative clauses, with no antecedent (ὃν οἱ θεοὶ ϕιλοῦσιν ἀποθνῄσκει νέος), relative clauses with incorporated antecedents, ‘semi-free’ relative clauses (like free relatives, but with a demonstrative pronoun as a placeholder antecedent), and relative-correlative constructions (resembling the semi-free clause, but with the relative clause occurring first, and the demonstrative resuming it in the main clause). As a final introductory chapter, Chapter 7 discusses both attractio inversa (attraction of the antecedent into the case of the relative pronoun) and the later phenomenon whereby the relative pronoun is attracted into the case of the antecedent, confirming that supposed examples in early Greek can all be explained otherwise.
With Chapter 8, P. begins the detailed survey of her data, starting with those genres in her corpus that, with comparatively few relatives, can be treated quickly: alphabetic inscriptions to 550 b.c., early literary prose, and iambic, elegiac and lyric poetry, also all to 550 b.c. The first conclusion drawn is that inherently maximalising clauses and postnominal clauses serve very different purposes: the former are restrictive and equivalent to a definite noun, whereas the latter modify indefinite nouns or are non-restrictive. Second, within the inherently maximalising group, free relatives are preferred if the case of the relative pronoun matches that of the notional antecedent; when the cases do not match, the syntax is harder to process, and so expressions with an explicit antecedent are preferred (relative-correlatives in inscriptions, prose and iambic poetry, but semi-free relatives in lyric and elegy).
Chapters 9–12 all deal with Homer, with the first, and longest, of them setting out the differences between the inherently maximalising and postnominal types (much the same as in the genres covered in Chapter 8), and the latter three each addressing more specific concerns: what determines which of the inherently maximalising constructions is chosen; the difference between relative ὅς and the relative use of what would in later Greek become the article; and the reasons for the different frequency and distribution of relatives in direct speech and narrative. Out of this rich discussion it emerges that relative-correlative sentences in Homer typically have a topic-comment structure, often in future-related speech acts, like wishes, predictions and generalisations, with the relative clause giving the broad category as orientation to the subsequent main clause; that postnominal ὅς-clauses are preferred in generalising clauses with τε or when the clause as a whole acts as an appositive to the antecedent, whereas ὁ-clauses state a new fact about the antecedent; that inherently maximalising clauses occur much more often in direct speech than narrative because they so often refer to future situations, in which all possibilities need to be accounted for, whereas narrative calls for more postnominal ὁ-clauses.
In Chapter 13, P. covers the last component of her corpus, Cretan inscriptions to 400 b.c., which are important as a genre in which the inherently maximalising relatives predominate (by a ratio of 56 to 21; in the Iliad, the opposite is true, with postnominal clauses favoured by 1094 to 255). As P. shows, this should not surprise us: the contexts in which the two main types of relatives are used are remarkably consistent from genre to genre and, given that the Cretan inscriptions are predominantly legal texts trying to cover all future contingencies, it is only natural that the inherently maximalising type should be so common.
At this point, P. can return to the larger historical linguistic questions posed at the start of the work, and Chapter 14 duly examines the status of supposed diachronic relics in the light of the preceding analysis, arriving primarily at negative conclusions, e.g. that apodotic δέ in relative-correlative structures is not evidence that relative clauses were once construed paratactically. The exception is epic τε, which points back to a time when ὅς τε-clauses were free relative clauses coordinated with what was later reinterpreted as an antecedent. This in turn leads P. to suggest an earlier system, in which postnominal ὅς-clauses were an innovation, with non-restrictive relative clauses previously having required the use of a demonstrative pronoun like ὁ instead.
Whether or not one wishes to go that far, the overall picture presented by P. is very convincing: the relative clauses of Greek form a coherent synchronic system that obtains in all genres. Thus, that they appear to be more closely related to Indic than to the Latin or Hittite variety is simply a side effect of the nature of the texts: if Indo-Europeanists read more Cretan inscriptions and less Homer, they would never have thought otherwise. P.'s work is, in short, a valuable survey of this important type of clause and, in its meticulous presentation of the primary evidence and thorough investigation of the secondary literature, could well serve as a model for research in historical syntax generally.
