The semantics of definite expressions and the grammaticalization of THE
Nikolas B. Gisborne
University of Edinburgh
This chapter explores the claim that definite expressions involve universal and existential quantification from the point of view of Word Grammar, in order to establish whether the quantificational view of definiteness is compatible with a particular cognitive theory of language, and to see how it compares with the familiarity treatment of definiteness. It is argued that the quantificational approach is superior to the familiarity approach in the analysis of a number of linguistic phenomena, and a number of Word Grammar analyses are presented. The chapter concludes with an investigation into the grammaticalization of the English definite article, in order to compare the merits of the two approaches, and argues that the quantificational approach delivers a simpler and preferable account.
Introduction
We know that the English definite article, the, is the result of the grammaticalization of the demonstrative pronoun se. Diessel (1999) and Hawkins (2004) trace a number of grammaticalization paths that demonstrative pronouns have taken cross-linguistically, and the English history is just one of the possible grammaticalization paths that demonstrative pronouns might take. It is also clear that this change is an example of grammaticalization; although both definite articles and demonstrative pronouns are grammatical words, and the main difference between them appears to be semantic, there is also a phonological difference. The is clearly a clitic: it varies in form before consonants and vowels just as a(n) does. However, I will argue that there is no categorial change or other syntactic change and that
this change involves just the loss of some semantic content that demonstratives have but which definite articles do not have, plus the phonological reduction to clitic status.
However, there are some problems. In particular, what is the right semantics of the, or of definiteness more generally? There is a debate in the formal semantic literature about how, precisely, the English definite article ought to be treated: is it a quantifier, as Russell (1905) claimed, or is it an expression of familiarity, as has been claimed in the tradition perhaps beginning with Christopherson (1939)? Russell's (1905) analysis has prompted a considerable amount of debate, from Strawson's (1950) criticisms, to more recent literature including Neale (1990), who argued for a quantificational analysis of the definite article, and Elbourne (2010), who argues against it, and who aligns himself with Heim's (1983) familiarity approach.
Another problem is to do with how we conceptualise the semantics. Russell works with quantification into the world, but cognitive semantics does not. That is to say that (crudely speaking) in a formal semantics, the meaning of a sentence is the fragment of reality it corresponds to, whereas within a cognitive theory, the meaning is a conceptual representation, anchored to reality by perception and memory. Given the theme of this volume, I want to ask how well a cognitive theory of semantics captures the differences and the debate in the formal theory, and whether, by taking a cognitive stance, we can sharpen up the question about how we should treat the semantics of the definite article.
In general, a cognitive theory of semantics helps with handling the semantic processes associated with grammaticalization: there is, after all, no reference to the world; the whole of semantics is within a conceptual structure, regardless of whether the bit of meaning that is under discussion is to do with sense, or reference, and so in a cognitive theory it is straightforward to account for changes that involve loss of content and more "referential" meanings. A cognitive theory should also make it possible to explain procedural meanings. This set of concerns is related to the issue of grammaticalization: we conceive of grammaticalization as a cognitive phenomenon, where semantic change is often intimately bound up with categorial change (Hopper and Traugott 2003). Furthermore, the cognitive theorist's rejection of encapsulation makes it much more straightforward to track systemic changes within a cognitive account.
The final problem is to do with the heuristic value of the theory we adopt. How well does the theory help us find and understand new data? Generative theory gave us the tools to discover island effects, and the raising-control distinction. What similar heuristic value might a cognitive theory have? Within a cognitive theory, what kinds of semantic distinction come into focus because of the theory that has been adopted?
In this chapter, I take the following line: I assume that the treatment of the as a quantifier is right. There is evidence from a number of linguistic domains, but especially specificational sentences, that argues in its favour. However, there are some problems with modelling Russell's specific proposals in a cognitive theory, because cognitive theories do not have unrestricted quantification into the world, so I argue that in a cognitive theory we need to develop a theory which is perhaps closer to the restricted quantification of say Neale (1990), although a cognitive theory does not work with Neale's truth conditional semantics.
However, both Langacker (1991) and Davidse (2004) treat the definite article as a proportional quantifier (which Langacker calls a "relative" quantifier), and any theory needs a treatment of proportional quantifiers, so my tack is to argue that the is a proportional quantifier, and to present an account of proportional quantifiers in Word Grammar.[2] I specifically formalise the account in Word Grammar because neither Langacker nor Davidse offer a formalisation, and I want to show how the analysis of the as a proportional quantifier interacts with other parts of the grammar in as precise a way as I can, so as to make it possible for scholars from different traditions to engage with the analysis.
Word Grammar is a cognitive theory of language which works with network representations. It was first reported in Hudson's (1984) monograph. The basic assumption of the theory is that language (and the rest of cognition, which language is embedded in) is a symbolic network. The theory is presented in Hudson (2007, 2010) and also in Gisborne (2010). I present such notions as the theory needs in the course of the chapter when they become relevant.
So, to summarise, the questions I want to tackle are:
- What is the right semantics of the to be able to capture the grammaticalization from the Old English demonstratives?
- How do we account for the right semantics within a (broadly speaking) cognitive theory of linguistic semantics?
- Are both treatments familiarity and quantificational (or more specific formulations of them) equally competent at accounting for the synchronic data?
There are several related points that I argue for in this chapter which are listed here.
- I argue that a treatment of the definite article as a quantifier within a restricted theory of quantification is better than the familiarity account for handling the synchronic linguistic facts.
- I argue that restricted quantification can be theorized within a cognitive theory of language structure, and present some of Hudson's (2007) analyses.
- I extend these findings to the, and explore some consequences of making these assumptions.
- Finally, I show that given some basic assumptions about the syntax, the quantifier approach makes for a simpler and more plausible diachronic story.
The chapter has the following structure. The next section, 2, introduces the two main approaches to definiteness, drawing on Lyons (1999) and Abbott (2004); 3 discusses reference and the cognitive theory of reference; 4 explores the familiarity theory of definiteness in Word Grammar (the presentation being taken from Hudson 2007); 5 presents the restricted quantification version of Russell's theory of definiteness; 6 models the proportional quantifier approach in WG; 7 presents some linguistic reasons in favour of the quantifier approach; 8 discusses the two theories' merits with respect to grammaticalization; and 9 presents the results and conclusions.
- [1] I would like to thank Willem Hollmann, Ekkehard König, and the anonymous referees for a number of suggestions improving the chapter. I should also like to thank the audience at the Societas Linguistica Europaea Conference in Vilnius, Ronnie Cann, and Geoff Pullum for useful comments and advice, and Dick Hudson for reading the chapter through and making extensive comments.
- [2] Langacker (1991: 92, 98) does not actually say explicitly that he views the as a proportional quantifier, although his semantic glosses are compatible with such a view. Langacker (1991: 98) says, "the designated instance ti of T is unique and maximal in relation to the current discourse space." I should note that Epstein (2002) challenges both uniqueness and identifiability approaches to definite expressions, and argues for a cognitive linguistic approach based on accessibility within a theory of mental models.
