Log in / Register
Home arrow Language & Literature arrow Theory and Data in Cognitive Linguistics

The theories and grammaticalization

In this section, I argue that a set-theoretic treatment of definiteness, and therefore of the, is compatible with the grammaticalization facts and makes for a simpler story than the alternative familiarity approach. In making this argument, I rely on a particular syntactic assumption, that in the "NP" it is, in fact, the determiner that is the head, and that definite articles are, like demonstratives, just transitive pronouns. The difference between a definite article and a demonstrative is that the definite article is obligatorily transitive, whereas the demonstrative is optionally so. This version of the determiner-as-head hypothesis is from Hudson (1990: 268-276); it differs from Abney's (1987) version, which assumes a different position on the nature of the superordinate category, and which assumes a different structure within the NP. That is to say that I am not adopting Abney's version of the determiner-as-head analysis, because Abney assumes that pronouns belong in the determiner category; I am assuming that determiners belong in the pronoun category. The Old English demonstrative se is the source for modern English the (Mitchell 1985, Traugott 1992), and this is often treated as a prototypical grammaticalization path (Traugott 1982, Lyons 1999).

The story I offer is a simple one: I take it that the semantics for the that I have argued for, and worked out a WG/Cognitive representation for, is the most schematic of the meanings that definite expressions might have. If we recall Abbott's list of definite expressions in 2, it cannot be assumed to be the case that definite expressions only mean universal and existential quantification over a restricted set. Take 's: this is definite, and so must have the set-theoretic semantics of the, but it also means possession. The identity of the possessor has to be given separately, of course, but nevertheless, possession is an additional meaning that is not present in the. The grammaticalization of the does not involve categorial reanalysis; nor does its requirement to have a complement noun. The apparent categorial reanalysis and the complementation pattern are entirely explainable in terms of a very simple semantic change, so that essentially the semantic content of the reduces to the quantifier analysis I have given.

We can start with Diessel's (1999: 129-129) presentation of the grammaticalization of the English definite article from the demonstratives. Diessel writes,

The use of anaphoric demonstratives is usually confined to non-topical antecedents that tend to be somewhat unexpected, contrastive or emphatic ... When anaphoric demonstratives develop into definite articles their use is gradually extended from non-topical antecedents to all kinds of referents in the preceding discourse. In the course of this development, demonstratives lose their deictic function and turn into formal markers of definiteness. An example of such a definite marker is the article the in English.

He adds that there are three factors involved in this process:

- "demonstratives lose their status as free nominals when they become reanalysed as definite markers"

- "demonstratives are significantly more often inflected than articles, which suggests that adnominal demonstratives often lose the ability to inflect when they grammaticalize as definite markers"

- there is a cline: demonstrative>definite marker>specific indefinite marker.

How do the two theories handle the grammaticalization facts?

Because demonstratives are deictic, we need to start with thinking about deixis. The WG analysis of deixis relies on the situated nature of the sentence or utterance: because WG assumes that utterances are intramental, there is no distinction between utterance and sentence, both are conceptual tokens. We can therefore make the following claims about deixis:

- Words in utterances have a time, place and speaker.

- When I utter me, there is a relationship between me as speaker, the place and time of the utterance, and the discourse context.

- This is all modelled in conceptual structure.

- Deictic expressions are therefore subjective.

How does this play out in the semantics of the demonstratives? Let us take that as a case study. There are, I think, two main points: first, the demonstratives have to inherit the definite semantics of the. This is not to say that that Isa the, but to say that the meaning of that must inherit whatever it is that makes the definite, because demonstratives are also definite. This observation suggests that the meaning of that is definite in the ways that the meaning of the is definite, plus it has an additional layer of deictic meaning. The claim is very similar to what I claimed above for 's, that it inherited the definite semantics, with an additional layer of possessive meaning.

Figure 12 presents a WG analysis of that, showing both its definiteness and that it is deictic.

A partial WG lexical entry for that

Figure 12. A partial WG lexical entry for that

The diagram in Figure 12 is slightly underspecified, in that I have not shown the difference between that when it is anaphoric and that when it has an accompanying noun. For now, I want to concentrate on two main points: the word that clearly has the same model of definiteness as the, and it has an additional domain of meaning because its lexical entry includes the fact that the referent of that is (relatively) far from the speaker. Each utterance word has a speaker, and the lexical entry can model that fact; indeed, in the case of deictic expressions such as this and that, it is necessary for the lexical entry to show how the use of the word relates to the speech event. Crudely speaking, the change in Old English se from demonstrative to determiner is the loss of the deictic element in its lexical entry. However, as we shall see, whatever semantics we adopt for definiteness, the story will be more complicated than just the loss of the deictic element.

Below, I discuss how the known facts can be modelled given the quantifier account of the and, subsequently, I show how they could be modelled given the familiarity theory. However, before we move on to how the different theories model the change, we need to agree some ground rules: first, we need to establish what the most plausible syntactic account of demonstratives and definite articles will look like; and second we need to be clear about the main Old English facts, so that the models I propose can be evaluated in terms of their plausibility.

The syntax is straightforward, if a little non-standard. I am assuming that the determiners of English are all pronouns, and that they take common nouns as their complement. In this analysis, I am following Hudson (1990: 268-276). Note that the analysis is not the same as the Abney's version of NP structure, which makes the reverse categorial assumptions, that pronouns are determiners. The difference matters, because the pronouns-are-determiners analysis requires there to be a novel functional category, determiner, which has different distributional properties from the category noun, and which projects a novel functional node in a tree. In Hudson's WG story, the category "pronoun" is just a subtype of noun, so whether the noun selects the determiner or the other way around the NP is just headed by a noun.

We can see how the account works by comparing a(n), the, this, that, and my/mine.

(26) a. We saw a/the cat.

b. * We saw a/the.

(27) a. We saw this/that/my cat.

b. We saw this/that/mine.

The argument is simple: treat a(n), the, this, that, and my/mine as transitive pronouns. In (26a) and (27a) they occur with their complement; in (27b), this, that, and my/mine occur without their complement. In (26b) we see that a(n) and the are obligatorily transitive. As long as we treat my and mine as just conditioned variants of the same lexeme, where the different variant depends on whether the item has a complement or not, all of these items can be treated as members of the class of pronouns, which take complements. According to this story, "articles" and "demonstratives" are both subtypes of transitive pronouns; the difference between them is simply that an article must always have a complement and a demonstrative can have a complement or be elliptical or anaphoric.[1]

In broad terms, the WG analysis of "determiners" as transitive pronouns makes the analysis of the emergence of the English definite article very simple: an earlier demonstrative loses its ability to occur without a complement (and to be anaphoric, or interpreted elliptically) and has to occur at all times with its complement. At the same time, the ability of (singular count) nouns to occur without a determiner is also lost, giving rise to the mandatory appearance of determined NPs. The latter phenomenon is a reflex of the loss of case, and I shall not explore it here if the job of a determiner is to fit up a noun to refer, then this is a job that nominal inflection also does. With the loss of case comes the loss of being set up for reference.

Before I come to evaluating the different semantic arguments and how they accommodate the history of the English definite article, I briefly summarise the history of the English definite article. There are two main recent studies that I am aware of: Denison (2006) and Sommerer (2010). Denison's paper is a handbook article which addresses theoretical questions of how to identify word classes, and what their boundaries are; Sommerer's thesis is an investigation into the emergence of the English definite article from the Old English demonstrative se. Crucially, both of these studies assume that there is such a word class as "determiner", and so sets out to establish criteria for it as a word class, and to identify when this class might first be identified in the history of English.

Sommerer (2010) runs through a range of uses of the Old English demonstrative, as well as showing that in OE it was possible for bare nouns to occur in referring positions. The examples in (28) (Sommerer 2010: 14, example 1) show that Gothic, Old High German, Old English, and Old Saxon could all have referring nouns which occurred without a demonstrative, or anything otherwise similar to an article.

(28) a. ip sa inngaggands pairh daur hairdeis ist lambe but who goes through [the] door is [a] shepherd for [the] sheep Got (J.X.2)

b. uuantra giboran ist man in mittilgart because (it) was born [a] man in [the] world OHG (Tatian.174.5)

c. stonc da after stane stearcheort onfand feondes fotlast jumped then behind [the] stone [the] stouthearted, found enemy's footstep OE (Beo.2288)

d. ef eo man mid sulicun dadun dodes gesculdien if sometimes [a] man with such actions [the] death deserves OS (Heliand. 5244)

In (28), it is clear that the different nouns occur in a range of argument positions, and none of them expresses any kind of nominal determination. In (29), taken from Sommerer (2010: 24) we have an example where se is in a construction with a noun, but it is ambiguous between a demonstrative and a definite interpretation.

(29) Men ne cunnon secgan to sode ...hwa pcem hiceste onfeng people cannot say for sure ....who the/that cargo received (Beowulf 50)

In the next example, taken from Sommerer (2010: 21, example 5), se is anaphoric.

(30) and par da burh getimbrede, _ pas ilcan geares pa at Bricge and there the burgh built, and in the same year that at Bridgeworth[2]

Example (31) is also anaphoric, but arguably to a clausal antecedent rather than a nominal one.

(31) dada he was gebroht to geleafan mid dare grapunge, When he was brought to faith, with the touch, pa weard seo twynung purh pct us atbroden then was the uncertainty by that from us taken[3]

And the final example I shall give shows that the Old English demonstrative also functioned in relative clauses (of the correlative type, rather than clauses with the structure of modern relative clauses).

(32) Abel, Adames sunu, rihtwis and Gode andfenge, Abel, Adam's son, righteous and to God loyal, pone ofsloh Cain his brodor whom/this-one slew Cain, his brother[4]

Both Hawkins (2004) and Diessel (1999) observe that one possible grammaticalization path for relative markers is from demonstratives, so the final use is not surprising. This use of the OE demonstrative does not, however, last into Middle English. I take it that the loss of case reduces the functional utility of this construction as it can no longer serve to relativise low items on the Keenan-Comrie Accessibility Hierarchy.

The Old English facts are consistent with the account I have sketched where we treat both the Old English demonstrative and the modern English definite article as transitive pronouns, where the transitivity is optional in the case of the demonstrative. Which of the semantic approaches to definiteness might work best with this syntactic treatment, and what then would the diachronic story be?

First, whatever semantic story we might adopt, there will be no category emergence. Determiners will not, in my story, be a new category of English. What we see in the morphosyntax is a restriction of the definite article to a transitive structure or in other terms the transitivity becomes obligatory. I will treat the narrowing of the structures the definite article can occur in so that, unlike the demonstrative, it cannot occur in relative clauses as a consequence of the more general loss of case.

I present the familiarity story first and then move to the set analysis. As demonstratives are deictic, if definite articles are familiarity expressions, then what must happen is that explicit reference to something in the context (whether the demonstrative has a complement or not) is replaced with reference to some entity which is understood by the speaker to be part of shared knowledge. So, if I say at home the dog needs feeding, my family will understand that Jumble, our pet dog, is the dog at hand, not the mangy dog from the farm down the road which wanders around our neighbourhood.

How would this work? Deixis is subjective it is anchored in the speaker and the speaker-centred situatedness of the utterance. In order to make such an analysis work, we would need to assume that familiarity was inter-subjective, because it is the expression of a meaning of which the hearer is an essential part. Why, if I did not expect there to be a dog familiar to my interlocutor, would I use

the expression the dog? The expression creates a presupposition of familiarity for both speaker and hearer. According to this story, there would be several ways of creating "definiteness": an expression would be definite if it was deictic, because the referent could be contextually identified; it would be definite if it belonged in a shared knowledge of the world, because the speaker would be presupposing the hearer's familiarity with the object; and so forth. Definiteness, then, would be a heterogeneous category, perhaps organised in a family resemblance structure.

On the other hand, this way of construing how definite expressions are definite requires two things: First, it requires an extension to Traugott's (2010) notion of intersubjectivity. Traugott limits intersubjectivity to "addressee self-image", but if definite expressions are intersubjective as I have claimed a familiarity account must be then the notion of intersubjectivity is extended to shared knowledge, or the presupposition of shared knowledge. Second, this account requires that there is a semantic change: As demonstratives become determiners, they lose their deictic anchoring and acquire a new meaning: the presupposition of shared knowledge. In the next two diagrams, I give representations of the demonstrative in the phrase that man and the definite article in the phrase the man, which show the relevant changes. These diagrams allow us to compare this model of the change with the alternative Russellian account of definite descriptions.

That man in WG, without sets

Figure 13. That man in WG, without sets

In this diagram, and in the following one, I am treating both that and the as the head of the phrase. I am not complicating the diagrams by adding categorial information, but I am assuming that they are both pronouns. I am assuming that the only difference between that and the, and by extension between the and its demonstrative historical antecedent, se, is in the semantics.

The man in WG, without sets

Figure 14 The man in WG, without sets

As we can see from the diagrams, the difference between that and the comes down to the loss of a particular meaning situated in the utterance event, and the acquisition of what Hudson calls the "directed identity" link (see the discussion of Figure 2 above). In both examples, the NP actually "refers" (in the intramental sense of reference) in that there is a referential concept that the words identify. According to this theory, definiteness is heterogeneous: there are several different semantics associated under this single rubric. And according to this theory, the semantic change is as I described it above: from deictic to shared knowledge.

In the next diagram, I present a lexical entry for se, which adopts the semantics in Figures 13 and 14. The diachronic story, according to this theory, would be (i) that the semantics associated with the third sublexeme emerged, and then (ii) that the sublexemes split, to be associated with that and the respectively.

A lexical entry for se, on the familiarity story

Figure 15. A lexical entry for se, on the familiarity story

The lexeme se is where all of the inflectional information is collected. I have left this paradigmatic information out of the diagram in order to avoid unnecessary complexity. The sublexeme sei is the use of se we find in example (29) an anaphoric use of the pronoun. The sublexeme se2 accounts for the demonstrative interpretation in example (30), and the final sublexeme, SE3 accounts for the definite, non-demonstrative, interpretation in that example. All of the representations are necessarily somewhat schematic and there is clearly more complexity: for example, I have shown that SE2 and SE3 have complements, but I have not shown that those complements are nouns, or that their referents are instances of their senses.

The diachronic story that Figure 15 shows is simple: the semantics of SE3 emerge, and then in time SE3 comes to be realised by the; and sei and SE2 come to be realised by thati and that2. The and that come to be identified as separate lexemes, and the inflectional information attached to the lexeme se is lost. Apart from changing realisations, the main innovation is the semantic change from SE2 to SE3.

On the set approach to the semantics, that man looks like Figure 16, and the man looks like Figure 17.

That man in WG, with set-theoretic definiteness

Figure 16. That man in WG, with set-theoretic definiteness

The main difference between the analysis of that man and the man is simply the difference between a deictic expression and an expression that is not deictic.

The man in WG, with set-theoretic definiteness

Figure 17. The man in WG, with set-theoretic definiteness

According to this theory, both the and that express their definiteness in the same way: the difference between them is simply that that is deictic and the is not. Now we are in a position to see what a lexical entry for se would look like.

The lexeme se, with a set-based semantics

Figure 18. The lexeme se, with a set-based semantics

As in the other analysis, se1 represents the sublexeme for the anaphoric pronoun in (29) and se2 the lexeme that must exist on the demonstrative interpretation of (28). SE3 represents the definite article interpretation of the example in (28). As in the discussion of Figure 15, there is not much to the diachronic story: The semantics of SE3 emerge, and then SE3 comes to be realised by the, and sei and SE2 come to be realised by that. The main difference is in the claim about the semantics of definiteness, and the semantic change. The diagram is as schematic as the previous one.

The diagram in Figure 18 makes the claim that the only semantic change from transitive demonstrative to definite article was the loss of deictic meaning nothing else happened. At the same time, the only change in form was the realisation of the definite article by the, which is (as I noted above) a clitic a change in form that has to be accommodated on either semantic theory. I show the subsequent stage in the relevant lexical entries for Modern English in Figure 19.

Lexical entries for Modern English that and the

Figure 19. Lexical entries for Modern English that and the

In this diagram, I have shown that the is a separate lexeme, and that sei and SE2 have come to be replaced by thati and that2. I have also shown the morphosyntactic realizational facts, although I have not shown the two pronunciations of the, which are contingent on whether the following word begins with a vowel or a consonant. I will not present a diagram for these words on the familiarity approach it is this diagram, but with the semantics of Figure 15.

The diagrams show that either theory of definiteness can handle the grammaticalization of the. I think, though, that we can do better than that, and consider how the theories might compare. Both stories at least given the syntactic analysis I have adopted here make the grammaticalization of the simple. The common elements are that there is a lexeme-split (se to that and the) and that the evolves into a clitic. According to the account here, there is no categorial change. So the only difference between the two theories is that the familiarity story of definiteness requires us to assume that a new relation comes into play, which allows the speaker/hearer to identify a referential token in the (discourse) environment, whereas the set-theoretic account of definiteness merely involves the loss of the few network links that make the meaning of se or that deictic.

The familiarity approach requires an extension of the notion of intersubjectivity, whereas the set-based approach requires no such extension of the analytical tools. And finally, the familiarity approach requires us to treat definiteness as a heterogeneous category, whereas the set-based approach allows us to treat it as a unitary phenomenon, with the differences between the different meanings being due to such additional network links as are necessary for an account of the different meanings of the various definite expressions listed in 2 (at least with the exception of proper names). Given the advantages, for linguistic analysis, of treating definite expressions as proportional quantifiers that I pointed out in 7, and given the far simpler story of grammaticalization that it gives us, I conclude that functional linguistics working on grammaticalization, and cognitive linguists working on the phenomena discussed in 7, will want to work with a treatment of definite expressions as proportional quantifiers.

  • [1] Hudson (2004), following a paper by Van Langendonck (1994), revisits his earlier analysis, and argues that determiners (or transitive pronouns) are categorically as he described them in (1990), but that there is also a mutual dependency relation between them and their noun complements. The argument does not affect the analysis of the category of demonstratives and articles, so there is no need to visit all of the claims in the paper. Van Langendonck's arguments are mainly that the distribution of NPs follows from the noun and not the determiner. This, then, makes the noun the head. Hudson (2004) effectively splits headedness into two kinds: distributional headedness, and structural headedness (see also Rosta 2005). He argues that the determiner is the structural head i.e. the head that determines the internal structure of the NP and the noun is (or can be) the distributional head. For my purposes here, this two-headed analysis works. Determiners are still classified as (transitive) pronouns, and the distributional facts that Van Langendonck identifies can still be accommodated.
  • [2] (ChronC 96.31 (912))
  • [3] Sommerer (2010: 21) example (6); iElfric's Homilies i.234.23.
  • [4] Sommerer (2010: 22) example (11); iElfric's Homilies ii.58.25.
Found a mistake? Please highlight the word and press Shift + Enter  
< Prev   CONTENTS   Next >
Business & Finance
Computer Science
Language & Literature
Political science