A cognitive analysis of John's hat

Richard Hudson

last changed 11 January 2010

Bibliographical information

Based on a conference presentation in Manchester in April 2009, the first draft of this paper was written in November 2009. It is intended for a volume containing the conference proceedings: Morphosyntactic categories and the expression of possession, edited by Kersti Borjars, David Denison and Alan Scott (John Benjamins).

Abstract

This chapter has suggested that a child’s experience of suffix possessives could, and probably does, justify two different syntactic analyses, both of which survive into adulthood and co-exist happily in the adult grammar. The first is a direct descendant of the Old English inflected genitive, a single word inflected by the suffix {z}. But unlike the Old English genitive, it doubles as a determiner, so it is the head of the whole noun phrase, and not just a dependent modifier of the common noun; for example, John’s hat has hat depending on John’s rather than (as in Old English) the other way round. This reversal of the syntax means that dependent-marking case has been replaced by head-marking of the noun-phrase’s internal structure.

The second syntactic pattern is the ‘group genitive’, in which the {z} counts as a clitic, a separate syntactic word POSS realized by a mere suffix as in John and Mary’s house and the man across the road’s house. This diachronic innovation may have been encouraged by the parallel historical development of reduced auxiliaries which are also realized by {z}, as in John’s arrived and John’s working. For the modern child, of course, the historical development is irrelevant, and the main point is that experience includes examples like these which couldn’t be interpreted as inflected genitives, so an alternative interpretation is needed. Once again the interpretation needs to explain the determiner-like characteristics of John’s, so in this case POSS is classified as a determiner, which in Word Grammar means that they are pronouns.

Each of these analyses has attractions for a language learner. The inflected genitive gives a transparent one-to-one mapping from syntax to morphology, but at the cost of a relatively complex analysis of the possessor: in syntax it has a double classification as both a common/proper noun and a pronoun, and in semantics, it has an extra referent. This may well be the preferred analysis for single-word possessors such as John’s, though there may be considerable variation from person to person. In contrast, the group genitive is the only possible analysis for examples where {z} is separated from the head of a possessor phrase. Its advantages in other cases are a simpler mapping from syntax to semantics and direct mapping from syntax to morphology, rather than mapping via an inflectional rule. These advantages are offset by a more complex syntactic structure and the exceptional morphology of a clitic.

If grammar is learned from usage, then we may assume that an adult grammar reflects not only childish interpretations of experience, but also the frequencies of the patterns experienced, which are expressed in the grammar as activation levels attached to different stored patterns. For example, the tendency discussed in section 7 for suffixed possessives to be short may have two complementary explanations, both cognitive. It may be directly explained as a way to reduce memory load, but this in itself produces a bias in the usage that a child experiences and which in turn will be stored as activation levels in the child’s own grammar, thereby influencing the child’s own production independently of the processing benefits.

The main point of this chapter has been to show the benefits of cognitive analysis when applied to familiar data such as the English suffix possessive. These benefits include a theoretical framework which rests firmly on elementary principles of cognitive psychology, but perhaps the most important attraction for a linguist is the possibility of facing uncertainty and complication in the data without feeling obliged to arrive at a single coherent and complete analysis. If a multiple, messy and incomplete analysis is good enough for a language learner, it should certainly be good enough for us.