Canonical Typology PDF

Preview

Summary

The Oxford Handbook of Morphological Theory Jenny Audring & Francesca Masini (eds.) CANONICAL TYPOLOGY OLIVER BOND 1. Introduction Canonical Typology is a methodological framework for conducting typological research in which descriptive categories and theoretical concepts are deconstructed into ...

Description

Accelerat ing t he world's research.

Canonical Typology Oliver Bond

Related papers

Download a PDF Pack of t he best relat ed papers 

MORPHOLOGICAL T HEORY AND T YPOLOGY Pet er Arkadiev

MORPHOLOGY IN T YPOLOGY: Hist orical ret rospect , st at e of t he art , and prospect s Pet er Arkadiev T he Oxford Handbook of Morphological T heory Jenny Audring, Francesca Masini

The Oxford Handbook of Morphological Theory Jenny Audring & Francesca Masini (eds.)



CANONICAL TYPOLOGY

OLIVER BOND 1. Introduction Canonical Typology is a methodological framework for conducting typological research in which descriptive categories and theoretical concepts are deconstructed into fine-grained parameters of typological variation. Like other multivariate approaches to cross-linguistic research (Haspelmath 2007, Hyman 2009, Bickel 2010, 2011), Canonical Typology utilizes observations on a large number of empirically motivated variables to gauge the similarities and differences between linguistic structures (within or across languages). The method is distinguished from other contemporary approaches to typology by its appeal to the notion of the canon, a logically motivated archetype from which attested and unattested patterns are calibrated. While not a theory of morphology, Canonical Typology was first developed as a means to systematically analyse morphosyntactic and morphological phenomena, such as agreement and inflection (Corbett 2003a,b, 2005, 2006, 2009, 2015). It has proven to be especially adept as a tool to evaluate linguistic constructs and to elucidate the variation encountered in different morphological systems. More recently, it has been evoked as a means to characterise morphology that deviates from default relationships theorised in Paradigm Function Morphology (Stump 2013, Stump & Finkel 2013) and as a method for analysing a range of complex syntactico-semantic phenomena (see papers in Brown, Chumakina & Corbett 2013). In theory, the canonical approach to typology is compatible with any model of grammar in which variation can be decomposed into fine-grained variables. In practice, however, (inflectional) morphology is assumed to be an autonomous component of grammar by most proponents of Canonical Typology. 1 Consequently, the morphological component of grammar is usually understood to be inferentialrealizational (according to Stump’s 2001 typology) and morphological processes are concerned with morphological exponents, not morphemes. The role of features is central to the conception of the framework and these are commonly distinguished with respect to whether they participate in morphosyntactic processes (e.g. agreement), whether they are 1 The autonomy of morphology from syntax is well supported in the work of the major proponents of Canonical Typology, since much of this work demonstrates that inflectional morphology is characterized by its own rules and principles. It is less obvious if the same scholars think derivational morphology should be considered to be autonomous from the lexicon.



1

The Oxford Handbook of Morphological Theory Jenny Audring & Francesca Masini (eds.)

morphosemantic features that can only indicate semantic content (e.g. tense) or whether they are strictly morphology-internal (e.g. inflectional class).2 While much has been written on Canonical Typology, explicit descriptions of its methodology or the fundamental stages of its application are often open to interpretation. Here I deconstruct the framework by providing a stepwise introduction to the principles that can be used to identify the canon and the possible types associated with a particular linguistic domain. The first aim of the chapter is to review what it means to be ‘canonical’ and to provide guidance on how ‘canonical values’ can be identified (§2). The second aim is to provide an overview of the insights the method has already provided for inflectional and derivational morphology (§3). Having looked at how Canonical Typology has been employed to analyse what it means to be a ‘possible word’, we will turn to where it might be headed, and how it might develop as more morphological phenomena are investigated using the framework (§4). 2. Establishing a canon The types of categories and concepts linguists use to describe and compare languages frequently share the same label, but seldom share an identical set of characteristics. This variation can result from differences in descriptive tradition (e.g. gender vs. noun classes), theoretical approach (e.g. agreement in Minimalism vs. HPSG vs. LFG) or genuine linguistic diversity. Canonical Typology attempts to alleviate issues of categorisation by taking a bird’s eye view of a given linguistic phenomenon in all its dimensions. In doing so it provides a mechanism for demonstrating (i) how linguistic structures can differ from one another along a range of parameters; and (ii) how different (ontological or theoretical) categorisations of a phenomenon may map to these parameters. A central concept into this approach to typology is the CANON, a reference point from which to compare linguistic objects and descriptions. In Canonical Typology, the canon associated with a particular linguistic phenomenon (such as agreement, suppletion or gender) is a logically motivated archetype whose properties are determined through the application of the CANONICAL METHOD. Once the properties of the canon have been established, it is used as a reference point for describing real and hypothesised linguistic structures associated with that phenomenon. For example, a particular instance of agreement can be described as being canonical (or noncanonical) along a series of parameters. This allows us to assess how closely it resembles CANONICAL AGREEMENT (Corbett 2003, 2006) and compare instances of agreement in a principled way. The method pulls apart tangled problems to allow differences to be seen more clearly. It ensures linguists are talking about the same thing when comparing languages or structures, and avoids problems associated with terminology by placing cross-linguistics variation at the centre of analysis and description. 2 See

Corbett & Baerman (2006), Corbett (2012, 2013) and Kibort & Corbett (2008) for discussion of different types of features.



2

The Oxford Handbook of Morphological Theory Jenny Audring & Francesca Masini (eds.)

So when we speak of CANONICAL AGREEMENT, CANONICAL GENDER or CANONICAL INFLECTION, we are referring to a theoretical construct, defined by a set of logically compatible characteristics, that belongs to a specific notional domain. Since the canon is only a construct, a real life exemplar of it may not exist. The process of identifying the canon for a given domain – and crucially which properties define the canon – can be characterised by four stages, each of which will be described in turn. 2.1 Identification of the domain Canonical Typology is used to map variation encountered within a specific notional domain, typically one that is already established in linguistic description. A range of different morphological or morphosyntactic domains have been investigated from a canonical typological perspective so far, most notably agreement (Comrie 2003, Corbett 2003, 2006, Cormier, Schembri & Woll 2013, Cysouw 2011, Evans 2003, Palancar 2015, Polinsky 2003, Suthar 2006), morphosyntactic features and their values (Corbett 2008, Corbett 2013, Corbett & Fedden 2015, Round & Corbett 2016, Van de Velde 2013), and a wide range of topics related to inflectional and derivational morphology (see §3 for references and discussion).3 While seldom explicitly referred to in applications of the methodology, the formal identification of the domain of investigation is an essential part of conducting Canonical Typology. A domain is always at least implicit in applications of the framework because it delimits what types of linguistic structures should be considered as possible examples of a phenomenon. Bond (2013: 21-25) argues that suitable domains for investigation can be explicitly defined by characterising a broad and minimal relation between two or more linguistic elements. This process provides a BASE-DEFINITION or BASE, which sets out the domain of typological investigation, and therefore limits what properties must be associated with a structure for it to be considered part of the study. By way of example, the following base-definition for the agreement domain sets out the fundamental properties that characterise all examples of agreement (Bond 2013: 24, see also Corbett 2006 and Steele 1978): 3 Since Canonical Typology is a specific framework for carrying out cross-linguistic research, it

has been employed to discuss a wide range of topics. Some of these are related to properties of the verb or predicate, including expressive affixes (Fortin 2011), finiteness (Nikolaeva 2013), negation (Bond 2013), and reality status (Michael 2014). It has also been used for topics related to reference tracking and argument structure such as quotation (Evans 2013), reflexives (Everaert 2013) and passives (Siewierska & Bakker 2013), as well as the status of categories and relations in (morpho)syntax and semantics such as mixed categories (Nikolaeva & Spencer 2008), possession and modification (Nikolaeva & Spencer 2013), the argument/adjunct distinction (Forker 2014) and gender and classifiers (Corbett & Fedden 2015). The relationship between Canonical Typology and ontologies of description is discussed in Farrar (2013).



3

The Oxford Handbook of Morphological Theory Jenny Audring & Francesca Masini (eds.)

AGREEMENT DOMAIN: For the elements X and Y to be in an agreement relation, there must be a systematic covariance between a semantic or formal property of one element and a formal property of another. The agreement domain contains any relationships that exhibit these properties. An explicit characterisation of the domain ensures that sets of linguistic structures like those in (1) can be included in the typology since varying the number value of the subject element affects the formal properties of a second element, the verb. It is only by looking at two (or more) structures in parallel that this relation can be revealed. (1) a. The dog barks every night. b. The dogs bark every night. Equally, the base-definition for the agreement domain ensures that structures like those in (2) are not included in the domain of investigation, and thus do not influence the application of the method. In this pair of examples, varying the semantic properties of the subject does not affect the form of any other element in the clause. (2) a. The dog barked every night. b. The dogs barked every night. The base is designed to capture only essential information about the domain investigated; yet this sometimes means making a principled decision about how to limit an investigation before a full understanding of the phenomenon is reached. Consider the following base-definition for the inflection domain: INFLECTION DOMAIN: For a form to be an inflected one, it must bear a paradigmatic relationship to another form with the same lexico-semantic content. The inflection domain contains any sets of forms that exhibit this type of relationship. The need for this formulation should be clear. It ensures that our typology is limited to a clearly defined domain of morphology, but at a slight cost. It guarantees that the relationship between singular dog and plural dogs belongs to the inflection domain (even though dog itself does not bear any inflectional exponent). It equally ensures that the relationship between dog and dogged (which are not forms of the same lexeme) does not. This is achieved by explicitly stating that inflectional morphology concerns intra-lexemic not interlexemic relationships, and thus relies on an independent process for



4

The Oxford Handbook of Morphological Theory Jenny Audring & Francesca Masini (eds.)

identifying lexemes. It also relies on an explicit notion of a paradigmatic relationship (see §2.2). Once a domain has been identified, and a base-definition mooted, it is possible to begin mapping out parameters of variation for that domain. 2.2 Identification of parameters of variation If two linguistic structures (such as word-forms or sets of word-forms) in a given domain (such as the inflection domain or the agreement domain), are similar to each other, they will be identical along some (set of) parameter(s), yet differ in one (or more) other ways. In Canonical Typology (and typology in general), these parameters of variation provide the empirical base for establishing which variables are important when mapping out possible differences between languages and structures. Before we identify what these parameters are for inflection (introduced in §2.1), first consider the paradigm – a construct central to the conception of inflectional morphology. In its most basic conception, a paradigm defines the relationships between a set of forms that are manifestations of the same lexeme, but differ according to their featural specification. Consequently, in the simplest possible paradigm, a series of two forms vary according to their specification for mutually exclusive values of a single grammatical feature. For instance, this is the case for the majority of nouns in English, which exhibit a distinction between singular and plural values of the number feature, as illustrated with English cat in (3). It is clear however, that this distinction does not extend to all nouns in English. Abstract nouns like health do not have plural forms (e.g. *healths) and such lexical items do not meet the base-definition of the inflection domain in §2.1. (3) A paradigm with a single feature ‘cat’ SG PL cat cats Paradigms become more interesting when they are constructed from two or more orthogonal features. For instance, the paradigm in (4) is constructed from intersecting values for two agreement features, namely number and gender, giving rise to four distinct forms for the Spanish adjective alto ‘tall’.



5

The Oxford Handbook of Morphological Theory Jenny Audring & Francesca Masini (eds.)

(4)

A paradigm with forms distinguished by two intersecting features Spanish alto ‘tall’ SG PL MASC alto altos FEM alta altas

Each cell of the paradigm is populated by a unique form, providing strong evidence for distinguishing four feature values for this particular class of adjectives in Spanish. However, not all adjectives show this pattern. Many adjectives agree in number, but not in gender, demonstrating that this distinction is only relevant for some lexical items, as shown in (5) for the Spanish adjective inteligente ‘intelligent’. (5) A paradigm with forms distinguished by a single feature Spanish inteligente ‘intelligent’ SG PL inteligente inteligentes In Spanish, values for gender and number intersect to determine unique forms for some adjectives, but not for others. In other languages, this sort of pattern extends across every member of a distributional class, such that the intersection of values is only relevant for determining a part of the paradigm. For instance, in Macedonian (Slavonic, Indo-European), typical adjectives can agree in number, gender and definiteness (Baerman, Brown & Corbett 2009), giving rise to the paradigm structure in (6). While all of the forms in the paradigm are unique, the gender distinction is only relevant in the singular, leading to SYNCRETISM within the plural. (6) A paradigm with three intersecting features and syncretic forms Macedonian ubav ‘beautiful’ INDEFINITE DEFINITE SG PL SG PL MASC ubav ubaviot FEM ubava ubavi ubavata ubavito NEUT ubavo ubavato While typical adjectives in Macedonian have forms for three gender values in the singular, not all behave in this way. The adjective siromav ‘poor’ lacks feminine and neuter singular forms, resulting in the paradigm in (7).



6

The Oxford Handbook of Morphological Theory Jenny Audring & Francesca Masini (eds.)

(7)

A paradigm with three intersecting features and defective cells

MASC FEM

Macedonian siromav ‘poor’ INDEFINITE DEFINITE SG PL SG PL siromav siromaviot siromasi siromasite

NEUT Here several cells of the paradigm are greyed out, because no form is possible. This is known as defectiveness: the adjective is restricted to masculine singular or plural contexts, such that it is possible to say siromaviot čovek ‘the poor person’ (masculine), siromasite lugje ‘the poor people’, but not *siromavata žena ‘the poor woman’ (feminine) or *siromavato dete ‘the poor child’ (neuter). The paradigms in (3) to (7) vary in their structure in terms of whether these featural distinctions apply to all lexical items of a distributional class or a subset only (e.g. cats vs. *healths), how many features are involved in the construction of the paradigm (e.g. two for Spanish adjectives vs. three for Macedonian adjectives), how many values each feature has (two gender values in Spanish vs. three in Macedonian), whether all the cells determined by the intersecting features have a unique form (e.g. gender syncretism in Macedonian plural adjectives vs. gender specific forms in the singular), and whether the features intersect to create forms at all (e.g. defective feminine and neuters cells in Macedonian singular adjectives vs. gender syncretism in the plural). With a base-definition established, and parameters of variation identified, it is possible to formulate this variation in terms of sets of values, such that every example under consideration (that is, every paradigm, form, syntactic configuration, or phonological representation that has the characteristics of the base-definition) must be characterised by one of those values. Some possible binary values characterising variation in paradigm structure are given in (8). (8) A subset of possible binary values characterising variation in paradigm structure VALUE 1 VALUE 2 Only one feature defines More than one feature FEATURE intersections defines intersections Unique forms at feature Shared forms across UNIQUENESS intersections feature intersections All intersections have Subset of intersections FORMS forms without forms



7

The Oxford Handbook of Morphological Theory Jenny Audring & Francesca Masini (eds.)

If we consider the adjectival forms discussed above in light of these different variables, we instantly see that there is considerable variation across the patterns attested even within a small data set, as demonstrated by the matrix in (9). (9) Matrix of three variables with binary values for a set of adjectives FEATURE UNIQUENESS FORMS PATTERN V1 V2 V1 V2 V1 V2 Spanish ✓ ✓ ✓ A ‘tall’ Spanish ✓ ✓ ✓ B ‘intelligent’ Macedonian ✓ ✓ ✓ C ‘beautiful’ Macedonian ✓ ✓ ✓ D ‘poor’ Four different patterns are attested in this small data set, demonstrating that a fine-grained approach to determining variables is fruitful for unveiling variation. 2.3 Identification of canonical values Once parameters of variation (i.e. variables) have been established, the values for each variable are assigned to an ordered scale that distinguishes between ‘canonical’ and ‘non-canonical’ behaviour. In existing applications of Canonical Typology, these scales, known as ‘criteria’ are typically binary, such that one value for each variable (either VALUE 1 or VALUE 2) is canonical in nature, while the other is non-canonical.4 For instance, we might assume that all of the VALUE 1 properties in (8) are canonical properties of inflection. But how do we decide? Why would a paradigm defined by only one feature intersection be more canonical that one defined by multiple features? The process of determining which values are canonical is central to the application of the methodology, yet also the most opaque. Corbett (2010: 141) describes instances with canonical values as ‘the best, the clearest, the indisputable ones’, making re...