Many languages have more than one way of forming possessive noun phrases. English, for example, has a choice between the "Saxon genitive" (Mary's house, the director's office, the day's end) and the "Norman genitive" (the office of the director, the end of the month), conditioned in part by the part of speech and semantics of the possessor but with considerable overlap and stylistic variation.
In contrast, many languages have an opposition of two (or sometimes more) forms of possessive marking whose choice is conditioned not by semantics or style, as in English, but lexically; and conditioned not by properties of the possessor but by the possessed noun, i.e. by the head noun in the construction. (The possessed nouns in the English examples above are house, office, and end.) Consider the examples in (1) from Mesa Grande Diegueño (Yuman; California), where the noun 'mother' takes the simple prefix ʔ- 'my' while 'house' takes the compound prefix ʔə-nʸ-.
This contrast of two formal types of possession, determined by the possessed noun, is possessive classification. We use the terms possessive and possession, as is traditional, to refer to all kinds of adnominal constructions regardless of whether the semantics is literal possession. We use the term classification because the marking of possession in examples like (1) divides, or classifies, the nouns of Mesa Grande Diegueño into two sets: those that behave grammatically in possessive NPs like 'mother' and those that behave like 'house'.
Though in many languages the two kinds of possession use the same kind of morphology (as in Mesa Grande Diegueño, where both kinds use prefixes on the possessed noun), a number of languages use entirely different kinds of morphology in the different possessive constructions. An example comes from Warndarang (Maran; Northern Territory, Australia). Here kin terms and certain others take possessive prefixes as in (2a), while other nouns take a separate possessive pronoun in the genitive case as in (2b).
‘my father’, ‘our father’
'(the/a) crocodile's tail'
'(the/a) crocodile's tail'
In Tiwi, either construction is possible for most nouns, but kin terms take only juxtaposition (as in 3a).
The distinctions shown in all these examples are classification because they are lexically determined by the possessed noun and result in an overt difference in the marking of possession. We consider only classes that are lexically conditioned, and have excluded any that are phonologically conditioned (or appear to be amenable to a phonological analysis). We note that, while lexical classifications are most often equipollent (one class of nouns takes possessive marker A and another takes marker B), they can also be privative (as in the Tiwi example just above: one class generally takes A but can sometimes or occasionally take B, perhaps with a semantic opposition; the other class takes only B but never A). Even if there is a semantic opposition available to one class, the overall classification is still lexical.
Most languages with possessive classification have a binary opposition like those illustrated above. Sometimes the classification is more elaborate. Nez Perce (Sahaptian; northwestern United States), for example, has three possessive classes: two different possessive prefix paradigms for different kin terms, and a preposed possessor in the genitive case for other nouns:
(prefix paradigm 1)
'my sister' (man speaking)
(prefix paradigm 2)
Burushaski (isolate; Kashmir, India and Pakistan) has four classes of possessive prefixation defined by allomorphy of the possessive prefixes. In (5), classes (a), (c), and (d) have different prefix shapes and (b) causes a stress shift.
[(d) example extracted from text, constructed]
'head of bed'
A few languages have complex systems of possessive classification, with ten or more classes. Amele (Madang; Papua New Guinea), for instance, has 31 different classes based on the allomorphy of the possessive suffix, as well as an open class which uses a postposition, for a total of 32 classes.
The following examples illustrate a few of the Amele possessive suffixal classes. The examples chosen all have stem-final –e to show that the endings are not phonologically conditioned in any obvious way.
Chichimeca-Jonaz (Oto-Manguean; Mexico) has twelve (or possibly even more) possessive classes of nouns, where the classes are defined broadly as involving tone changes, prefix-like elements, various internal changes, and combinations of these as well as suppletion. (Each of these classes, or at least some of them, contains both bound and non-bound nouns as these are defined in Chapter 58. In addition, the language has four possessive classifier nouns.) The examples in (9) display this variety.
(M = lenis nasal)
In these four languages with complex systems (Amele, Anêm, Chichimeca-Jonaz, Cayuvava), possessive classification has become a system of declension classes rather like the declension classes of case inflection in classical Indo-European languages. In Amele the differences between allomorphs are rather straightforward, in Anêm less so, and in Chichimeco Jonaz and Cayuvava they are great.
Map 59A shows the number of possessive classes in sample languages.
|No possessive classification||125|
|Three to five classes||20|
|More than five classes||4|
Binary possessive classification is fairly common everywhere except Eurasia, where it is entirely absent except for tokens in the Pacific Rim and the mountain enclaves (Caucasus, Himalayas). Multiple classes are not particularly common anywhere, though clearly best attested around the Pacific Rim, and the complex systems of ten or more classes are found only around the Pacific Rim. The larger systems of possessive classification in Melanesia are found on or near the coast and chiefly in the north of both Australia and New Guinea. Various kinds of morphology can be involved in possessive classification, as illustrated in the examples above, but all four of the complex systems use head-marking morphology. The complexity and opacity of the Chichimec and Cayuvava systems would appear to testify to a very great age for the head-marking type in these languages.
Binary possessive classification is usually called "alienable/inalienable" possession in the literature, partly because it often involves examples like those in (1), in which the house but not the mother can be bought, sold, etc., and partly because it is fairly common, especially in the Americas, for one of a binary set of possessive classes to be bound, i.e. obligatorily possessed (see Chapter 58). In fact, though, possessive classification is not a semantic or grammatical category but a purely lexical classification of nouns. Semantics is involved in that there is usually a default or open class and a specified or determinate class with a semantic common denominator shared by most but not all members of the class and also sometimes found in non-members. Kin terms often figure in the specified class, and usually several to many of the language's kin terms and a few other nouns take the non-default possessive form while all others, including a few kin terms, take the default possessive form. In addition, body parts, meronyms (parts in part-whole relationships), topological nouns, and property nouns often figure in non-default possessive classes.
Supported by U.S. National Science Foundation Grant 96-16448 (Nichols, P.I.), Swiss National Science Foundation Grants 08210-053455 and 610-062717 (Bickel, P.I.), and the Institute for Slavic, Eurasian, and East European Studies, UC Berkeley. We thank Aimee Lahaussois Bartosik, David A. Peterson, and Suzanne Wilhite for help with data collection.