[University home]

School of Languages, Linguistics and Cultures (legacy Site)

Germanic possessive -s : an empirical, historical and theoretical study

Research context

The impetus for this study comes from two separate but interrelated research directions with long histories in the linguistic literature:

HISTORICAL: The genitive, or possessive, -s in modern varieties of the Germanic languages developed from a common source, confined after Verner's Law to certain masculine and neuter stems.  The original genitive case marker was part of a wider case system and showed noun phrase-internal agreement. Limiting the discussion to national languages, two types of development can be distinguished. Faroese, German and Icelandic have retained a case system where genitive contrasts with other cases and is still marked by agreement (though the status of Faroese genitive is unclear). In English, Dutch and Mainland Scandinavian languages, the general case system has been lost except for some retention in the pronominal system. The genitive -s is, however, retained to express possession, albeit with properties different from the original case marker. In Dutch, English and Swedish, it has developed into a marker which occurs once only within the noun phrase, normally at the right edge of that phrase.  The details of the historical development of genitive -s in individual Germanic languages have been quite well studied, e.g. Allen (2003) for English, de Wit (1997) for Dutch and Norde (1997) for Swedish. However, we are not aware of any detailed comparative study of the development. This will be done as part of this project.
    For the purposes of our large-scale empirical study we will take Swedish as a representative of the Mainland Scandinavian languages, though variation within the group will be addressed too.
    The historical studies just cited show two main changes in the genitive:  from being marked most prominently on the head to being marked on the right edge, and from being marked by agreement to being marked once only. The first of these changes would in standard terminology be described as a development from a functional to a less functional element, from an affix to a clitic, and is therefore assumed to be an example of degrammaticalisation. The historical literature on possessive -s also makes a direct link between once-only marking and clitic status as evidence of degrammaticalisation (e.g. Norde 1997). However, a survey of the literature reveals numerous examples of features which are marked once only within a phrase but which show all the hallmarks of affixhood, e.g. the Danish definiteness marker (Börjars 1998) or the Finnish possessive (Kanerva 1987). A detailed comparative study of the development of the Germanic possessive is of direct relevance to the degrammaticalisation debate.  However, it also impinges on another theoretical debate.

THEORETICAL: The standard term 'clitic' is usually assumed to refer to an element which is bound to another word - hence similar to an affix - but whose placement depends on reference to syntactic structure - and hence shares some properties with full words. In the more recent literature, it is recognised that the two-way distinction is oversimplified, and additional categories such as 'phrasal affix' have been introduced (for a good example of such work, see Halpern 1995). The essence of these proposals is to separate two dimensions within which the distinction between affix and clitic can be made:  placement and degree of attachment. A phrasal affix is an element which is positioned with respect to a syntactic constituent but which shows affix-like strength of attachment (cf. also van Bergen 2003 for similar issues in the history of English).
    Yet even in the recent literature, where the need for a more subtle distinction is recognised, the English or Swedish possessive is commonly described as a clitic (e.g. Carstairs-McCarthy 1992:37). The possessive -s of present day Dutch is not relevant to this debate, since its use is limited to proper nouns and kinship terms and hence normally attaches to one-word phrases. The one notable exception to a straightforward clitic analysis of the English possessive -s  is Zwicky (1987), who provides evidence that the possessive marker is sensitive to the internal feature structure of the possessor phrase in a way that is unexpected for a clitic. Similar evidence has been provided for Swedish in Börjars (2003).  What is needed is a coherent and fine-grained theoretical account consistent across the languages concerned.
    Issues also arise with respect to the placement of the possessive -s  in both languages. It is clear that it occurs on the right edge of its phrase.  If it were a true clitic it should not be sensitive to the placement of the head within the possessor phrase. However, initial small-scale studies of both English (Greham 2004) and Swedish (Börjars 2003) indicate that possessive -s is avoided when post-modification of the possessor noun phrase would separate it from -s. Instead an alternative construction involving a PP is frequently chosen. The influence of semantic factors on the choice between possessive -s and ''an of-construction'' in English has been well studied, for instance by Rosenbach (2002). However, we are not aware of any large-scale study of the influence of structural factors on the choice, apart from Altenberg's (1982) study of 17th-century English. As part of this project, we will provide a detailed analysis of how structural factors - such as the presence of post-modification in, and syntactic weight (including pre-modification) of, the possessor - interact with semantic and information structural factors - such as animacy and topicality - to influence the choice between the two types of possessive constructions. If, as our initial studies (Börjars 2003, Greham 2004) indicate, the possessive -s still shows a preference for occurring on the head, this throws into question both the current status of -s as a clitic-like element and its proposed historical development.  Thus the theoretical and historical issues converge.