This collection of papers is a first-of-its-kind exploration of deponency. The term ‘deponency’ was originally applied to a peculiarity in Latin grammar, where a lexically arbitrary set of verbs appears with passive morphology in syntactically active contexts (such verbs lack active forms altogether, having defective paradigms). In the volume under review, the term is extended to a more general set of cases, roughly characterized by morphological marking in the ‘wrong’ (i.e. non-typical) syntactic contexts.
As with most pioneering work, an important goal of this book is to delineate the boundaries of the subject matter. In the introductory paper, ‘Morphological typology of deponency’, Matthew Baerman defines deponency, using as a starting point the three characteristics of the Latin case: (i) mismatch between form and function, (ii) occurrence in a lexically specified set, and (iii) a defective paradigm. This preliminary definition is then relaxed along various dimensions (by Baerman and others) to include a wider range of morphological mismatches. The resulting category, ‘extended deponency’, is quite broad and sometimes difficult to distinguish from other categories, in particular syncretism. The advantage of proceeding in this way is that it allows us to modify the definition as needed to match the empirical facts (assuming, of course, that there is some natural class of facts). The danger inherent in this approach is that heterogeneous phenomena are lumped together in the same category. A related and legitimate worry is whether there is any such natural class as deponency to begin with (cf. my discussion of Jonathan Bobaljik's paper below).
Because of the issues discussed above, I will focus in this review on the question of what it is that we are investigating when we consider cases of extended deponency. While the pessimist's stance will be to provide grounds to doubt that deponency captures a single grammatical phenomenon, the optimist will search for evidence in favor of a general theory of deponency. I have organized my discussion of the contributions in this volume in terms of this dialogue between optimists and pessimists, which allows me to touch on most of the papers at an appropriate level of generality.
We begin with a general skeptical challenge raised by Peter Matthews in his contribution, ‘How safe are our analyses?’. Matthews cautions against relying on grammatical descriptions in defining linguistic terms, since this inevitably introduces theoretical biases. He makes his point by reviewing how Latin deponency was analyzed in antiquity. Back then, deponent verbs were seen as neither active nor passive, and, therefore, arguably did not involve a mismatch between form and function. The more general point here is that there are often several weakly equivalent ways of analyzing the same facts, and no objective criteria for choosing among these alternatives. This makes the task of defining deponency (or, for that matter, many other morphological notions) precarious. I agree: if our definitions depend heavily on our particular formalism, this is dangerous and may indicate a failure to grasp the essence of the phenomena in question. However, Matthews' skepticism seems to run deeper than this. He suggests that maybe there are no morphological facts to be captured, insofar as it is not clear that morphology is restricted by Universal Grammar (UG), posing the question ‘What is it, once more, that a “theory” of morphology would be expected to explain?’ (310).
While it is probably true that there are few universals in morphology (at least in comparison to syntax), this does not warrant the conclusion that morphology is completely unrestricted. UG is often thought also to include ‘soft’ cognitive biases that make certain patterns not impossible but harder to learn and, consequently, more marked and less widespread (see, for example, Wilson Reference Wilson2006, Pertsova Reference Pertsova2007, Moreton Reference Moreton2008). Such biases have consequences for typology, the trajectory of language change, errors during language acquisition, and facts about morphological processing. It is these kinds of facts (in addition to, of course, categorical universals) that a morphological theory might be expected to explain.
This more optimistic view is compatible with the approach taken by Greville Corbett in ‘Deponency, syncretism and what lies between’. Corbett proposes that languages have a set of ideal or canonical ways of coding inflectional distinctions, and that we can characterize all other patterns as deviations from the canonical to various degrees. As a result, we get a continuous scale from ideal to non-ideal patterns. On such a scale, deponency is close to the notion of syncretism, with some cases taking up an intermediary position between deponency and syncretism. While I generally sympathize with this approach, my main concern is that Corbett gives no explicit criteria for constraints on canonicity itself. A statement to the effect that something is canonical needs evidential support. One standard way of supporting claims about canonicity (which can be understood as claims about UG) is to consider whether the purportedly more canonical patterns are also more frequent within and across languages. However, Corbett maintains that canonical patterns are not likely to be frequent.
This claim is taken even further in Andrew Spencer's contribution, ‘Extending deponency: Implications for morphological mismatches’, where he asserts that the notion of canonicity is unhelpful for understanding facts about the statistical prevalence of various types of patterns (68). Yet, Spencer's claim does not follow from Corbett's: the majority of patterns may be close to canonical for non-accidental reasons (thus, canonicity would be helpful for understanding this prevalence), without the fully canonical patterns themselves being frequent. Similar situations arise in biology: while the prototypical gestational period for horses is 336 days, relatively few pregnancies last for exactly 336 days – though there will be a prevalence of pregnancies lasting close to the prototypical period. It appears that Corbett implicitly accepts this way of understanding canonicity, judging from the fact that he routinely describes deviations from canonicity as deviations from the majority pattern (28, 29).
Returning once again to the issue of constraints on deponency, Jonathan David Bobaljik's paper, ‘The limits of deponency: A Chukotko-centric perspective’, presents a somewhat weaker form of skepticism than the one espoused by Matthews. Bobaljik questions whether deponency, in its extended definition, exists as a natural class. He argues that some instances of deponency automatically follow from certain syntactic principles, while other at first sight similar instances do not. More specifically, Bobaljik considers the so-called spurious antipassive (SAP) construction in Chukchi. This construction employs non-active morphology, yet the verb-external morpho-syntax remains active and transitive. In this sense, SAP bears a superficial resemblance to the Latin case, but it turns out that SAP appears only with certain subject–object combinations, namely those that violate the person animacy hierarchy (1st>2nd>3rd person). Incidentally, a similar situation occurs in Yarok, a language briefly discussed by Baerman (8). Bobalijik lays out a morpho-syntactic analysis that straightforwardly predicts occurrences of SAP and that is embedded within current Minimalist theory. He concludes that Chukchi SAP is fundamentally different from Latin deponents, which are supposedly completely unpredictable. For Bobaljik, deponency (in its extended sense) is at best a useful descriptive label; it does not constitute a grammatical phenomenon with a single analysis. However, one wonders whether Bobaljik's conclusion holds under a stricter definition of deponency. Chukchi SAP does after all not affect a lexically arbitrary class of words and does not involve a defective paradigm. Thus, the question remains whether there is some non-trivial definition of deponency that captures a natural class.
‘Deponency in Latin’, by Zheng Xu, Mark Aronoff & Frank Anshen, provides optimism that the well-behaved Chukchi case is not so dissimilar from Latin after all. Contrary to the consensus view, the authors find Latin deponents not to be completely arbitrary but conditioned by subtle semantic differences. After examining 1,215 verb senses from the Oxford Latin Dictionary, using Levin's (Reference Levin1993) classification of verbs into syntactico-semantic classes, the authors argue that Latin deponents tend not to take physically affected objects and are low on Hopper & Thompson's (Reference Hopper and Thompson1980) scale of transitivity. Intriguingly, Xu et al.'s finding potentially ties together Latin deponency and Chukchi SAP, since canonical transitivity is connected to the notion of ‘agent high on the person hierarchy acting on a patient low on the person hierarchy’ (Maxwell Reference Maxwell, Corbett, Brown and Hippisley2008).
The optimism inspired by Xu et al.'s contribution is challenged by evidence from Greek presented in Nikolas Lavidas & Dimitra Papangeli's paper on ‘Deponency in the diachrony of Greek’. They claim that Greek middle/passive deponents cannot be explained either by a syntactic or by a semantic approach. Their argument rests on the finding that Greek verbs of various types (transitives, unaccusatives, unergatives, psych-verbs, etc.) have been both entering and falling out of the deponent class at different diachronic stages. However, for this argument to be convincing, one would like to see a more nuanced and quantitative investigation of the Greek verb senses and their semantic changes across different stages. Lavidas & Papangeli conclude that deponency is a truly morphological phenomenon that is due to the random realization of lexical feature specifications and/or idiosyncratic properties of some roots. This last claim is difficult to evaluate (or even to fully understand) since no concrete theoretical analysis of the facts is provided.
Several other papers in this volume include a diachronic perspective, which might help us understand how deponency may have arisen but, on the other hand, leads to the pessimist's viewpoint that what we are investigating is merely a product of historical accident. For example, in ‘Slouching towards deponency: A family of mismatches in the Bantu verb stem’, Jeff Good discusses the Bantu language Ganda, in which some verbs possess phonological properties that accidentally resemble causative morphology. These verbs have been reanalyzed by the speakers as causatives, despite lacking causative semantics. In ‘Pseudo-argument affixes in Iwaidja and Ilgar: A case of deponent subject and object agreement’, Nicholas Evans discusses two Australian languages, where the weakening of a gender agreement system and subsequent lexical shifts have resulted in something that looks like deponency. More specifically, in some verbs, the agreement affixes preserve their archaic and once-productive gender distinctions, while in other verbs, the agreement affixes have acquired new derivational and inflectional functions. While diachronic explanations are useful in understanding the roots of deponency, they leave one wondering about the limits of deponency. For instance, the question arises whether deponency is simply a temporary unstable stage in languages' development. This does not seem to be the case, given Lavidas & Papangeli's data showing that Greek has been retaining deponent verbs for centuries, but the question is then why such a non-canonical phenomenon persists across languages.
One answer may be that speakers find some use for deponency (as suggested by Xu et. al). Another possibility is raised in Ricardo Bermúdez-Otero's contribution, ‘Spanish pseudoplurals: Phonological cues in the acquisition of a syntax–morphology mismatch’, which approaches this question from a learning perspective. In explaining Spanish pseudoplurals, Bermúdez-Otero claims that learners are forced to assign a plural morphological structure to some singular nouns in order to satisfy phonological constraints. More generally, morphology is subject to ‘simultaneous and often contradictory pressures from phonology and syntax’ (266) which, depending on how such pressures are prioritized, might lead learners to adopt a system containing morphological mismatches. This view seems reasonable as well as optimistic. It allows us to see deponency as a solution to other problems in the grammar rather than as a global imperfection.
The last paper to be discussed here, Gregory T. Stump's ‘A non-canonical pattern of deponency and its implications’, distinguishes between two types of deponency: ‘form deponency’ (a mismatch between phonology and morphology) and ‘property deponency’ (a mismatch between morphology and syntax). Stump convincingly argues that Sanskrit middles are an example of property mismatch, and raises the possibility that the same is true of Latin deponents. Interestingly, this seems to have been exactly the view of the ancient grammarians, as far as I can tell from the discussion in Matthews's chapter. The gist of Stump's analysis is that verbs that show a full morphological distinction between active and middle in Sanskrit (or between active and passive in Latin) also have a semantic contrast between the two categories. However, verbs that lack one of these morphological instantiations (i.e. that are either always active or always middle/passive) are semantically neutral with respect to voice. In sum, only verbs that have fully contrastive paradigms have contrastive semantics. This analysis has intriguing implications for acquisition and morphological theory alike. For example, it suggests that during acquisition, discovery of semantic contrasts is driven by phonological contrasts, and that the window for detecting such contrasts is a single paradigm (i.e. at first, generalizations are not made across paradigms that differ in shape). Moreover, this view of deponency implies the existence of higher-level inflectional classes. A membership in such a class can have syntactic but not necessarily semantic consequences.
Overall, I highly recommend this volume for its wealth of interesting data and the quality of the discussion. Despite the fact that the question of ‘what is deponency’ remains unresolved, the book presents a valuable first step towards coming to grips with this phenomenon. In this review, I hope to have outlined the general issues that arise and the different avenues explored by the contributors in approaching the question of what constitutes deponency. The pessimists among the authors have raised legitimate worries that need to be taken seriously, but this reader is nevertheless left with hope that future work will elucidate the nature of morphological mismatches.
