This book presents a new theory of grammatical categories - the Universal Spine Hypothesis - and reinforces generative notions of Universal Grammar while accommodating insights from linguistic typology.
INTRODUCTION This edited collection is the first book-length work on Canonical Typology (henceforth CT), a methodology for typological research presented in Corbett (2005) and since applied by various typologists (Baerman, Brown & Corbett 2005; Comrie 2003; Evans 2003; Polinsky 2003; Spencer 2005; Corbett 2006 is also written entirely in the CT framework, though CT is not its focus). While half the contributions come from the Surrey Morphology Group where CT has its origins, the remainder come from worldwide, attesting to CT’s growing reputation. Also included in this collection is a posthumous chapter by Anna Siewierska in conjunction with Dik Bakker. This volume is xiv + 312 pages (261 without the back matter), with Oxford University Press’s dependably high-quality construction and formatting.
Canonical Typology is a novel methodology in that it does not seek to establish the essential or defining features of a linguistic phenomenon. Instead, CT examines clusters of properties (‘criteria’) that converge on a canonical point and strongly co-vary, and yet are logically independent of each other. The typologist then maps out the ‘theoretical space of possibilities’ (or ‘base’) by examining all the ways a phenomenon might deviate from its canonical core along these dimensions. This method is useful for presenting multidimensional typologies, or shifting the focus from an epiphenomenal category to the components that interact to give that category shape. The papers in this volume apply this methodology to the exploration of a variety of disparate grammatical phenomena.
SUMMARY Dunstan Brown and Marina Chumakina begin with ‘What there is and what there might be’, where they present CT as a methodology designed to address the Correspondence Problem, i.e. “the issue of correspondence between similarly named features in different languages” (1). CT handles this problem by shifting the emphasis from definitions to the criteria that constitute those definitions. Whenever a construction shares some, but not all, of the properties of the canonical instance, CT forces the typologist to describe precisely which features are present or absent in crosslinguistically viable terms. After outlining CT’s basic terms and concepts, Brown & Chumakina provide examples of CT in practice (e.g. the IPA vowel chart), and establish historical precedents for CT, such as Sasse’s (2001) notion of ‘cardinal points’ of reference in describing ‘category squish’.
In 'A base for canonical negation', Oliver Bond questions how typologists should delimit the theoretical space (which he calls the ‘base’) for a given linguistic phenomenon. What’s the scope of investigation, and how does one determine what should be included in the typology? When typologists begin by selecting or arguing for the ‘best’ definition for a construction of interest, what makes these definitions the ‘best’? Bond suggests that instead of endlessly debating the necessary and sufficient criteria for these definitions, the base should include both the canonical and all the non-canonical, borderline cases from the literature as well -- all the variation known to exist for the phenomenon. Bond then applies this procedure in outlining a base for negation. This chapter is interesting in that it lays the framework for the crosslinguistic study of negation, but does not undertake such a study itself. It instead establishes the canonical points of reference that future studies of negation can calibrate to.
Greville Corbett’s chapter, ‘Canonical morphosyntactic features’, describes the interaction of morphosyntactic features (e.g. number, case) and parts of speech (Corbett’s preferred term, rather than ‘lexical category’ or ‘word class’). The problem motivating this chapter is that perfectly canonical morphosyntactic features would create parts of speech that are indistinguishable from one another -- every feature would be represented in every part of speech. This problem cannot be resolved by appeal to semantics, since often the morphosyntactic encoding of a certain feature value will also encode non-canonical meanings (e.g. the use of the second person plural morpheme to encode politeness in French). Corbett notes that when one group of lexemes lacks a certain morphosyntactic feature, or feature value, that set of lexemes constitutes a distinct part of speech. This chapter neatly shows how certain combinations of deviations from the canonical give rise to commonly-known morphosyntactic phenomena or word classes, such as gender (where certain lexemes may take only one feature value, depending on their gender) or pronouns (a small set of lexemes that are the only ones to exhibit the person feature). The chapter could however use more examples to illustrate the abstract ideas involved. And it is not until later in the chapter that Corbett mentions that one of his criteria for morphosyntactic features MUST be met, and therefore is part of the definition / base, leaving one to wonder why it wasn’t described as part of the base in the first place. Overall, though, this chapter is an exciting addition to the literature on parts of speech, suggesting that parts of speech are epiphenomenal in nature.
In ‘Some problems in the typology of quotation: A canonical approach’, Nicholas Evans takes CT beyond its traditional domain of morphology into the realm of discourse, and in doing so illustrates the usefulness of positing canonical ideals that may be completely unrealized in actual languages. Most typologies divide the phenomenon of quotation into direct speech, indirect speech, and other, which overlooks the diversity of the in-between cases. Variations include shifting person but not tense to reflect the original speaker’s speech (Russian), or coding some deictic spatial terms to the original speaker and others to the current speaker (Nez Perce). Previously, the coding perspective was treated as a binary choice between the original speaker versus the reporting speaker. Evans asks whether it is possible for a language to simultaneously represent the perspectives of both. Evans’ inspiration for this appears to have been “trirelational kin terms”, which simultaneously calculate kin relationships from two perspectives at once (90). While no known case of quotation is truly ‘biperspectival’, logophoric pronouns do approximate this, where the third person subject in reported speech is coreferential with the current speaker. The traditional direct vs. indirect speech dichotomy fails to capture the unique nature of logophoricity, demonstrating the usefulness of positing biperspectival speech as an only partially-realized canonical point.
Irina Nikolaeva debunks a similar false dichotomy in ‘Unpacking finiteness’, showing that finiteness is not a scalar notion from greater to lesser syntactic independence, as is usually proposed. Finiteness instead requires a multidimensional, canonical approach to capture its range and dimensions of variation. If canonical finite clauses are independent clauses, then finiteness is necessarily a language-specific concept, because different criteria must be used to determine syntactic independence for each language. Taking finiteness to be a property of clauses rather than verbs, Nikolaeva examines the assorted criteria, separating them into morphology, syntax, and semantics. Morphologically, (canonical) non-finite clauses show a reduction in the realization of features or feature values; syntactically, they are dependent on the main clause, and may be limited in licensing subjects; and semantically, they do not show independent temporal anchoring or make assertions. Nikolaeva shows that canonical non-finite clauses are those which exhibit all or most of these features, and that non-canonical finite clauses can be non-canonical in a variety of ways, depending on which criteria are being violated. The result is a robust typology of the properties that make up finiteness.
Andrew Spencer and Ana Luís present a typology of ‘The canonical clitic’. Clitics are unusual in the history of typology in that they have long been recognized to possess a mixture of the properties of affixes and function words -- a clear intersection of criteria amenable to description in CT. Spencer & Luís therefore define both canonical affixes and canonical function words, and then show how clitics are an amalgamation of the non-canonical properties from each. No study has yet successfully defined clitics as a simple combination of properties of affixes and function words, i.e. a point between affix and word. Spencer & Luís suggest that this is because the two points do not lie on a single continuum, but are orthogonal to each other. Put differently, a non-canonical affix is not necessarily a canonical function word, and vice versa. Clitics may simply be non-canonical along both dimensions at once. This chapter, like the previous two, again illustrates the success of CT in breaking down false dichotomies in typology, in this case through the use of ‘negative’ definitions. A small criticism of this chapter, however, is that some of the criteria put forth seem arbitrary and unsupported by the literature (whereas prior empirical research is an important part of defining the canonical base, as pointed out by Bond). An example is the proposition that the canonical affix is placed to the right of a word. This may be true, but no literature is cited to support the claim, and no explanation is given as to why this realization of suffixation should be viewed as more canonical than others.
In 'Passive agents: prototypical vs. canonical passives’, Anna Siewierska and Dik Bakker ask how we determine what is canonical and non-canonical for a given criterion. They investigate whether the expression of overt agents should be considered canonical for passives. Frequency-based accounts suggest that overt agents are not part of the canon, while exemplar-based accounts claim that they are. The authors note that one principle for determining canonical properties is whether those properties help to distinguish one type of canonical phenomenon from another. In the case of passives, overt agents do just that; a canonical property of anticausatives is that they do not exhibit an overt agent. Therefore, the presence of an overt agent is a distinguishing feature of the canonical passive. The authors also acknowledge that frequency contributes to what we consider canonical, and so they conduct a survey of 279 languages to determine the frequency and realization of overt agents in passives. They conclude that the overt expression of agents does correlate with many of the other canonical properties of passives, and so overt agents should be seen as part of the canonical passive. The chapter provides excellent guidance for how to proceed when there isn’t yet enough data on a phenomenon to determine what is canonical and what is not, and illustrates the success of CT in pointing the way in such an investigation.
Martin Everaert provides a canonical typology of reflexivization in ‘The criteria for reflexivization’. Like other chapters, the focus is on establishing a framework for a descriptive analysis of reflexivization, rather than a detailed empirical study. A problem with reflexivization is that languages, contrary to what is often assumed, frequently have more than one strategy for encoding reflexives. How can we describe these other than to simply call them all ‘reflexives’? Which strategies are more canonical? Everaert begins his answer by situating reflexivization as a type of anaphoric dependency (including deictic pronouns, logophoric elements, etc.). The distinguishing feature separating reflexives from other anaphors is the existence of an identity relation between two co-arguments. With this as his base, Everaert describes three sets of criteria for reflexivization: properties of the ‘binder’, the morphosyntactic encoding of the identity relationship, and the domain of the binder-bindee relation. He illustrates how this multidimensional framework neatly captures the observed variation for reflexivization, addressing the problem of multiple coding strategies. The chapter is straightforward, clearly written, and insightful.
In ‘Possession and modification -- a perspective from Canonical Typology’, Irina Nikolaeva and Andrew Spencer show how possession and modification, typically thought to constitute two ends of a continuum, actually emerge from the intersection of several semantic and morphosyntactic criteria. Interestingly, the non-canonical cases of possession and modification turn out to be other well-recognized phenomena, such as alienable possession or modification-by-noun. While canonical possessees are semantically relational nouns (kin terms, meronyms), and therefore constitute a type of inalienable possession, non-relational possessees like ‘hat’ are less canonical, and constitute alienable possession. Similarly, while canonical modifiers are adjectives, it is also possible for nouns to serve as non-canonical modifiers, creating modification-by-noun. Spencer and Nikolaeva’s is the first to relate these four phenomena in a systematic way.
The book ends with 'An ontological approach to Canonical Typology: laying the foundations for e-linguistics’ by Scott Farrar, advocating the use of the General Ontology for Linguistic Description (GOLD) in describing linguistic data, and pointing out parallels between CT and the principles for building an ontology. Farrar provides suggestions for how CT might be implemented in an ontology as part of a crosslinguistic database. An ontology-driven approach allows entities to be members of a canonical class, without assuming an identity between members of that class. In other words, the ontology could address the problem of crosslinguistic classification and comparison. Membership in a class is defeasible, so that the properties of that class are violable. Canonicity could then be measured by determining how many properties are violated by the specific linguistic construction under consideration. This chapter lays the groundwork for implementing a crosslinguistic database couched in canonical descriptions.
EVALUATION Many of these papers expand and refine the canonical approach in useful ways. Bond’s chapter is the first to make explicit the method of establishing the canonical base, while Evans’ illustrates the usefulness in describing theoretically possible but empirically unattested phenomena as canonical reference points for examining borderline or intermediate cases. The book does not aim to provide a discussion of theoretical or methodological issues, however. In fact, many of the authors, and Corbett especially, are oddly silent on the issue of crosslinguistic categories versus comparative concepts (Haspelmath 2010). When Corbett speaks of the Correspondence Problem, he is primarily interested in terminological rather than linguistic correspondence. This ambivalence is a potential merit of CT, in that any typologist can apply the method regardless of their theoretical assumptions. At the same time, throughout the book there is an implicit recognition of the need for crosslinguistically viable comparative concepts, couched in language-general rather than language-specific terms. I take this to be another merit of CT: in shifting focus away from superordinate categories like Noun or Finiteness and towards the criteria that constitute them, it encourages typologists to see the criteria as ‘basic’, i.e. having real status as a manipulable piece of the grammar of a language, and the superordinate categories as emergent. Despite lacking any explicit acknowledgement of this stance, this is precisely the position the authors in this volume appear to adopt.
Some readers might criticize this book for containing few exemplifying data. But while criteria are based heavily on reviews of previous literature, the aim of the chapters is not to provide comprehensive typologies based on crosslinguistic surveys but to describe a theoretical space in which future typological studies may be situated. Future studies of finiteness, for example, may now describe the exact realization of finiteness in a given language using the thirteen criteria set down by Nikolaeva.
To conclude, this volume successfully illustrates how Canonical Typology helps resolve a number of debates in typological description. In many cases, the canonical typologies presented expand the scope of investigation for a given topic to include phenomena previously thought irrelevant (e.g. Nikolaeva & Spencer’s links between alienable and inalienable possession and modification, or Evans’ links between direct and indirect speech and logophor). In all cases, these typologies pave the way for a much more robust and detailed description of these phenomena, which future linguists should find invaluable. The multidimensional nature of the criteria in CT is also highly amenable to multivariate statistical approaches to typology (Bickel 2011), making this an area ripe for exploration. I recommend this book to anyone seeking an introduction to the Canonical approach in typology, as well as to practicing typologists who are interested in ways they might improve or add more detail to their typological descriptions.
REFERENCES Baerman, Matthew, Dunstan Brown & Greville G. Corbett. 2005. “The syntax-morphology interface: a study of syncretism”. Cambridge: Cambridge University Press.
Bickel, Balthasar. 2011. Statistical modeling of language universals. “Linguistic Typology” 15(2). 401-414. doi:10.1515/LITY.2011.027.
Brown, Dunstan, Marina Chumakina & Greville G. Corbett (eds.). 2013. “Canonical morphology and syntax”. Oxford: Oxford University Press.
Comrie, Bernard. 2003. When agreement gets trigger-happy. In Dunstan Brown, Greville G. Corbett & Carole Tiberius (eds.), “Agreement: a typological perspective” (Transactions of the Philological Society 101), 313–337. Oxford: Blackwell.
Corbett, Greville G. 2005. The canonical approach in typology. In Zygmunt Frajzyngier, David Rood & Adam Hodges (eds.), “Linguistic diversity and language theories”, 25-49. Amsterdam: John Benjamins.
Corbett, Greville G. 2006. “Agreement”. Cambridge: Cambridge University Press.
Evans, Nicholas. 2003. Typologies of agreement: some problems from Kayardild. In Dunstan Brown, Greville G. Corbett & Carole Tiberius (eds.), “Agreement: a typological perspective” (Transactions of the Philological Society 101), 203-234. Oxford: Blackwell.
Haspelmath, Martin. 2010. Comparative concepts and descriptive categories in crosslinguistic studies. “Language” 86(3). 663-687. doi:10.1353/lan.2010.0021.
Polinsky, Maria. 2003. Non-canonical agreement is canonical. In Dunstan Brown, Greville G. Corbett & Carole Tiberius (eds.), “Agreement: a typological perspective” (Transactions of the Philological Society 101), 279-312. Oxford: Blackwell.
Sasse, Hans-Jürgen. 2001. Scales between nouniness and verbiness. In Martin Haspelmath & Ekkehard König (eds.), “Language typology and language universals”, Vol. 1, 495-509. Berlin: De Gruyter.
Spencer, Andrew. 2005. Towards a typology of “mixed categories.” In C. Orhan Orgun & Peter Sells (eds.), “Morphology and the web of grammar”, 95-138. Stanford: CSLI.
ABOUT THE REVIEWER:
Danny Hieber is a graduate student in linguistics at the University of California, Santa Barbara. Prior to grad school, he worked in Rosetta Stone’s Endangered Language Program to create language-learning software for the Chitimacha, Iñupiaq, Navajo, and Inuttitut languages. His primary research interests are language typology, documentary and descriptive linguistics, and the economics and praxeology of language. He holds a B.A. in linguistics and philosophy from The College of William & Mary. Learn more about his work at danielhieber.com.