2013.07.04

A. M. Devine, Laurence D. Stephens, Semantics for Latin: An Introduction. Oxford; New York: Oxford University Press, 2013. Pp. x, 453. ISBN 9780199969524. $85.00.

Reviewed by D.M. Goldstein, Universität Wien (david.goldstein@univie.ac.at)

Version at BMCR home site

This is an exciting book—ambitious, penetrating, and a watershed in the description of Latin. On the face of it, Semantics for Latin is an introduction to the world of formal semantics, an approach to linguistic meaning intimately tied to, and partly born out of, mathematics and philosophy. At a deeper level, however, it is an attempt to enrich the conceptual tools of our field. By equipping the reader with a stock of new conceptual vocabulary (as well as a concomitant metalanguage), this book offers access to new dimensions of meaning. Readers familiar with earlier magna opera of Devine and Stephens will recognize the style: a terse presentation of theory, data, and analysis whose density is rivaled only by TLL articles. In contrast to The Prosody of Greek Speech (Oxford 1994, BMCR 95.07.05 and BMCR 95.10.09) or Latin Word Order: Structured Meaning and Information (Oxford 2006, BMCR 2006.09.33), however, Semantics for Latin is not a book for classicists interested in linguistics (that crowd is well served by the numerous introductions to formal semantics already on the market), but rather a book for every classicist. Meaning (and more on this last word in the next paragraph) is something that we must all wrestle with, both in print and in the classroom. By laying bare how meaning is produced, Devine and Stephens enable us to move beyond mere intuitive understanding. In sum, Semantics for Latin is an extraordinary accomplishment, and has set a new standard in the description of Latin. What faults it has are presentational: for all its undeniable success, this is not an easy book to use.

Before turning to the content of the book itself, a word about "meaning." To the question "What is (linguistic) meaning?" there is no answer, only answers. As far as formal semantics is concerned, meaning is referential, truth-conditional, and compositional. Referential, in that the meaning of a term like dog is the entity in the world (or a possible world) that it refers to. Truth-conditional, in that to know the meaning of a sentence is to know the set of conditions under which that sentence is true (see further p. 10). To know the meaning of the proposition Stefan Zweig was born in Vienna is to know (ignoring issues of tense) whether Stefan Zweig is a member of the set of people born in Vienna. Lastly, meaning is compositional, which is to say that the meaning of a complex expression (be it a noun phrase, a sentence, what have you) is calculated stepwise from its parts by functional application: the whole is nothing more and nothing less than the sum of its parts (an idea also known as the Frege Principle, in honor of the German philosopher Gottlob Frege). Readers should be aware that this is only one approach to linguistic meaning. Cognitive linguistics, lexicography, Construction Grammar, and pragmatics, to say nothing of other linguistic frameworks, or even other fields such as semiotics, literary theory, or anthropology, all take different approaches to the question of (linguistic) meaning. One feature of formal semantics that may surprise some readers, is that cognitive representations of meaning play no role whatsoever (p. 11). It accordingly has nothing to tell us, for instance, about the concept of virtus in the late Republic. In fact, lexical semantics is not discussed at all in this book: for this topic, one may want to consult the recent Lexical Analysis by Patrick Hanks (Cambridge, MA 2013). The goal of Semantics for Latin is not to argue for formal semantics as an approach, nor does it lay out how other theories would differ. (In fact, it remains to be seen whether the insights from formal semantics can be fruitfully combined with other approaches to meaning.)

Turning to the book itself, the Introduction (pp. 3-14) does an impressive job of concisely introducing the core concepts of set theory, relations, truth-conditions, type-theory, and lambda calculus. The first three chapters focus on verbal semantics: tense, aspect, and argument structure. The next chapter is devoted to locational phrases, while subsequent chapters address quantification, plurality, modification, and negation. The final chapter then returns to verbal semantics with a discussion of modality. A list of symbols (which does not contain all the symbols used in the book, e.g. those for upward and downward entailment), a bibliography, and index close out the work.

So what then does this book have to offer? While most who read it will not come away with mastery of the formal apparatus (which is in any case not the goal of the book), they will come away with a rich conceptual vocabulary for talking about grammatical meaning. Take for instance the first chapter, which is devoted to tense. Devine and Stephens begin by introducing the A- and B-theories of time. The question that distinguishes these two theories is whether or not time consists solely in serial relations such as precedence, simultaneity, etc. (as the B-theory, also known as tenseless theory, holds), or whether past, present, and future are in addition necessary (so the A-theory, which is also known as the tensed theory). Devine and Stephens opt for the B-theory, and introduce the following ontology: event time (E), utterance time (U), and reference time (R) (there is a further dimension, adverbial time, which I leave aside for the moment). The event time is the timespan of an event, the utterance time is the time of speaking/writing, and the reference time is the time "for which the truth of the sentence is evaluated" (p. 21). We can use this three-part system to describe the semantics of tense with considerable precision, as in the following example (Caes. Bell. Gall. 7.19): hunc (collem)...palus difficilis...cingebat, ‛a difficult marsh surrounded this (hill).' The utterance time is the time at which Caesar is writing. The reference time is a contextually-defined interval in which the marsh surrounded the hill. The event time, the interval in which the marsh surrounded the hill is more extensive than the reference time, as the marsh surrounded the hill before the time that Caesar is referring to, and continued to surround it afterwards. Devine and Stephens thus characterize (p. 27) the semantics as follows: R < U ⋀ τ(e) ⊃ U. This means simply that the reference time (the relevant time at which the marsh surrounded the hill in Caesar's narrative) precedes the utterance time (time of writing) and that the event time (the state of marsh surrounding the hill) includes the utterance time.

This division of labor also helps us get a better grip on the semantics of temporal adverbials that modify an event description. Devine and Stephens identify two classes of adverbials, inner and outer. Inner adverbials modify the event time. We see this type clearly in nominal examples like nocturno impetu ‛by a night attack.' Outer adverbials by contrast modify the reference time, as with postero die in the following sentence (Bell. Gall. 7.27): postero die Caesar...arbitratus est, ‛On the following day, Caesar...decided.' While clear in some cases, this distinction can be much harder to make in some cases. Outer adverbials often correlate with extraposed setting expressions, however.

Elsewhere, Devine and Stephens deliver new insights into long-standing questions of Latin. In the chapter on negation, for instance, Devine and Stephens lay out clearly that Negative Polarity Items (NPIs) like quisquam ‛anyone,' ullus ‛any,' umquam ‛ever,' usquam ‛anywhere,' usquam, etc., are not restricted to negative contexts (such as I haven't seen anyone), but can also occur in conditional sentences (If you see anyone there...) and questions (Did you see anyone?), among other environments. What these environments all have in common is that they are downward entailing, which is an inferential relationship between propositions. A sentence like We ate quickly entails We ate. As the set of events that constitute eating quickly is a subset of those involving eating, the direction of entailment is thus from subset (eat quickly) to set (eat). In downward entailing contexts, such as negation, this pattern is reversed: No one ate entails No one ate quickly, and the direction of entailment is now from set to subset. This insight is a major descriptive advance in our understanding of this class of words. The book is packed with insights like these.

To my mind Semantics for Latin suffers from two sets of problems. First, in its design it falls somewhere between a handbook, a textbook, and a monograph. This in-between status is reflected in the title of the work, which does not demarcate the intellectual territory of the work precisely enough (and what is "semantics for Latin" supposed to mean in any case?). The subtitle an introduction is arguably a misnomer. True, Devine and Stephens introduce a host of foundational concepts in formal semantics. Unlike prototypical introductory works, however, theirs does not always explain new technical terms as they appear, e.g. that of de re and de dicto. As the authors explain (p. 5), Semantics for Latin "is designed to be used in conjunction with a Latin grammar and a dictionary of linguistic and philosophical terms." But the typical classicist is in all likelihood going to need more than a dictionary of terms to understand some of this material. Despite this caveat the recommendations for supplementary reading that Devine and Stephens offer is surprisingly sparse (at the end of this review, I offer some additional recommendations). A further complication of the hybrid identity of the book is that it may be difficult for some readers to determine what is solid communis opinio and what is a novel claim, or where an analysis is open to other possibilities.

The second set of problems lies in the area of presentation. The presentation of examples is idiosyncratic: why the authors do not simply follow the standard Leipzig glossing rules is a mystery. The decision to limit bibliography to some recommended supplements at the end of each chapter makes for smoother reading, but sometimes leads to frustration, as when passing comparison is made (p. 357) to neg-raising in Danish with neither examples nor bibliography. In the same vein, that Devine and Stephens do not alert the reader to their own previous work on Latin is odd. On p. 215, they refer out of the blue to the "focus position" of the clause, but leave it to the reader to deduce where that must lie. Discussion of this point (to say nothing of actual demonstration) would have been useful, or at least a reference to their claims in Latin Word Order: Structured Meaning and Information (Oxford 2006). The index covers only the topical matter of the book; it does not list textual citations or words. So when the authors refer in chapter one (p. 16) to their discussion of dum in chapter eight, there is no way to quickly locate it (on p. 382).

For those with no previous exposure to semantics, Emmon Bach's Informal Lectures on Formal Semantics (Albany, NY 1989) is something of a classic, and Paul Portner's What is Meaning? (Malden, MA 2005) is one of the gentler introductions to the subject (the latter is also mentioned by Devine and Stephens). David Lewis' 1972 paper "General Semantics" (Synthese 22: 18-67) is a general plea for model-theoretic semantics. For those who want a better handle on the formalism, the two-volume Logic, Language, and Meaning by L.T.F. Gamut (Chicago 1990) is a standard in the field. Also useful is James D. McCawley's Everything That Linguists Have Always Wanted to Know about Logic . . . But Were Ashamed to Ask (Chicago 1993, 2nd ed.). For more on the distinctions of tense discussed above, see e.g. Laura A. Michaelis, "Tense in English," The Handbook of English Linguistics, ed. Bas Aarts and April McMahon (Oxford 2006) 220-243, as well as Eystein Dahl, Time, Tense and Aspect in Early Vedic Grammar (Leiden 2010).