Written for Inclusion In: Theoxford Handbook Ofdeterminers, Edited By

[Written for inclusion in: TheOxford Handbook ofDeterminers, edited by SolveigaArmoskaiteMartinaWiltschko, OUP 2018?]

Arguments against the universality of D and determiners

Richard Hudson

Almost every modern grammar of English recognises a class of words called ‘determiner’ or ‘determinative’, including words such as an, the, some, this, which and every, and it is widely agreed that this set of words constitutes a distinct word class. The following definition is typical:

Determiners form a closed class of functional words which have the general property of not themselves permitting modification. The class of determiners includes articles, personal determiners, demonstratives, interrogative determiners, exclamatory determiners, and quality determiners. (Payne 2006)

This analysis is thoroughly modern and is one of the features that consistently distinguishes modern grammars from earlier ones. A similarly typical example of the latter is a recommendation from the 1911 report of the Joint Committee on Grammatical Terminology:

That English 'this'and 'that’, if used with a Noun, be called Demonstrative Adjectives, but if used without a Noun be called Demonstrative Pronouns; and that the same terminology be applied to the corresponding words in the other languages. (Joint committee on Grammatical Terminology 1911:19)

In 1911, if this wasn’t a pronoun, it must be an adjective, whereas modern grammarians would classify it in the second case as a determiner.

Distinguishing English determiners from adjectives is one of the achievements of modern grammarians, and I certainly don’t want to question this. From the very first grammatical use of deteminer (by Bloomfield), it was clear why they had to be distinguised from adjectives:

Our limiting adjectives fall into two sub-classes of determiners and numeratives... The determiners are defined by the fact that certain types of noun expressions (such as house or big house) are always accompanied by a determiner (as, this house, a big house). (Bloomfield 1933:203)

But notice that although Bloomfield recognised that determiners are syntactically special, he still classified them as a sub-class of adjectives.

However, four years before Bloomfield’s book, Harold Palmer had suggested a different analsis, albeit tied to the synonym determinative.

It is difficult, if not impossible, to draw any rigid line of demarcation between pronouns and determinatives, for most English determinatives are used both as modifiers of nouns and as pronouns. Some grammarians consider such determinatives as pronouns which can be used adjectivally (i.e. as modifiers of nouns), while others consider them as adjectives (i.e. modifiers of nouns) which can be used pronominally (i.e. to replace nouns). The simplest and most rational plan would seem to be to consider the personal pronouns alone as pronouns proper, and to place all others in the general category of determinatives, specifying in each case whether they may be used as modifiers, as pronouns, or as both. (Palmer 1924:42)

Determinatives. Under this heading are included a number of words which cannot be placed among the qualificative adjectives, because

(a) they cannot be used both epithetically and predicatively;

(b) they are rarely or never susceptible of comparison;

Notice that here too we have a recognition of ‘determinatives’ as grammatically special, though the distinctive characteristics complement the one noted by Bloomfield; but Palmer, unlike Bloomfield, considers and rejects the idea that determinatives are adjectives. Instead, he stresses their similarity to pronouns.

Modern grammarians of English agree on the distinctive characteristics of our determiners (or determinatives):

Any singular countable noun needs a determiner (in a book but not: *in book).
A determiner can’t be used predicatively (this big book or this book which is big but not: *big book which is this).
A determiner can’t be inflected for comparison or modified (some bigger books or some very big books but not: *somer books or *very some books).

We also agree that these characteristics show that determiners are not adjectives – and (pace Bloomfield) not even a special subtype of adjective – and that there is a signficant overlap between determiners and pronouns. Where we disagree is over how to explain this overlap and how to generalise from English to other languages.

English determiners are pronouns

Standard treatments of determiners ignore their overlap with pronouns or at best dismiss it as an uninteresting fact without consequences:

Given that many items can belong to more than one category … it is not surprising that some words can be used both as determiners and as pronouns. (Radford 1997:49)

However, the overlap between these two categories is surely significant, since almost all the clear determiners of English can also be used as pronouns (i.e. without a following common noun), with just four exceptions: the, an, every and relative whose(Hudson 1990:269). Admittedly the exceptions include the two most common determiners, but they are outnumbered 5:1 by the regular cases. The converse is not true: there are far more pronouns that cannot be used as determiners, so (pace (Postal 1966)) there is no case for treating pronouns as a sub-class of determiners.

This analysis faces two objections, one concerning form and the other meaning. Regarding form, there are six pronouns whose form varies with the presence or absence of a common noun: my~mine, your~yours, her~hers, our~hers, their~theirs, no~none. This variation can easily be handled in the morphosyntax as syntactically conditioned allomorphy. For example, the pronoun MY has a complement which may be overt or covert, and is realised as my in one case and mine in the other. Notice too that this variation isn’t restricted to possessive pronouns, since NO shows exactly the same pattern. In short, there is a minor generalisation about determiners that they may have two forms conditional on the presence or absence of an overt common noun.

The other potential objection to the analysis is that some pronouns have different meanings again depending on whether or not they combine with a common noun. The variation concerns the animacy-sex distinction male, female and neuter (as in he, she, it), which is relevant only in the absence of a complement (and only in the singular). For instance, what is compatible with any kind of common noun – what linguist, what child, what person as well as what book, what idea; but without a common noun it can only be used for inanimates (or at least non-humans). The same is true of the demonstratives this and that; so I know that girl is fine, but can’t be reduced to I know that. And of course, the personal pronouns distinguish he, she, it but the equivalent with a common noun is the, which neutralises this distinction. Once again, the semantic variation is easy to describe as a minor generalisation about the effect of syntax on meaning.

In earlier work I have argued for an analysis in which there is no categorial difference between pronouns with an without a common noun . My argument was that it takes more than one distinctive property to justify a distinct category, so (for example) although words are distinguished by their first sound, there is nothing to say about words that start with (say) /b/ other than that they start with /b/. Similarly, if the only difference between this in (1) and (2) is that one is combined with a common noun but the other doesn’t, then there is no case for a categorial distinction between them. However I now think this conclusion is premature because of these two minor generalisations about the interaction between syntax and form or meaning. I now believe, therefore, that the category ‘determiner’ is needed in order to allow these generalisations to be stated.

But the facts just reviewed do not support the standard analysis in which ‘determiner’ is a top-level word class, alongside ‘noun’, ‘verb’ and so on. On the contrary, they show clearly that determiners are also pronouns, distinguished from other pronouns mainly by their ability to combine with a following common noun. Moreover, the distributional similarities between pronouns and common and proper nouns suggest strongly that they all belong to a single super-class of ‘noun’ (Hudson 1984:90; Huddleston & Pullum 2002:327), so determiners are not only pronouns but also nouns.

This question of classification really matters for syntax. Suppose that determiners are indeed pronouns and nouns. In that case, this would be a noun in both these examples:

(1)I like this.

(2)I like this book.

And similarly for almost all the other determiners. But if this in this book is a noun, the obvious question is which of them is the head of the phrase. According to the DP hypothesis, the pronoun is the head of the phrase and the common noun is its complement (Hudson 1984:90–92; Abney 1987). But unlike the DP hypothesis, the analysis I am suggesting gives the phrase a noun as its head, so the phrase is a noun phrase – just as had always been suggested until the DP was born.

Applying this analysis to our examples, this is simply a pronoun in (1) but a determiner in (2); but it is the same lexical item (THAT) in both. These two possibilities are typical of determiners, but there are some that are only determiners (e.g. THE), and of course there are also some pronouns which are not determiners (e.g. ME). The analysis is displayed inFigure 1.

Figure 1: Determiners as pronouns

One of the many attractions of this analysis is to obviate the need for abstract invisible determiners in order to show the distributional similarities between noun phrases with and without determiners. For instance, if the object of a verb must be a DP, an abstract D is needed where there is no overt one, as in I like black coffee. But if it must be a noun phrase, and a determiner is a pronoun and a noun, then either the pronoun or the common noun will qualify as the head of this phrase.

Another attraction is to remove the need for ‘determiner’ as a grammatical function within a noun phrase, or ‘fused heads’ when there is no common noun, as advocated in some grammars(Huddleston & Pullum 2002:24–56). If the head of the phrase is the determiner, and this is classified as both pronoun and noun, then the phrase automatically qualifies as a noun phrase and the only functions needed are ‘head’ and ‘complement’.

What are the English determiners?

The distinguishing characteristics of our determiners are all syntactic, not semantic. Just to summarise, a determiner has the following properties:

A. It allows a singular noun to be countable.
B. It is a pronoun, so it inherits the common properties of pronouns such as not allowing modification by adjectives (though almost and hardly are sometimes allowed: almost every book, but not: *almost each book; hardly any book).
C. It excludes any other determiner.

This list includes the properties mentioned by Bloomfield and Palmer, with the addition of fact C, which they don’t mention. This property explains why we can’t combine (say)the, thator some with my, even though the meanings are compatible (and are expressible through other constructions, as in that book of mine or some friends of mine). It follows automatically if determiners are pronouns which allow a common noun (not a pronoun) as complement.

Given these identifying criteria, which words qualify as determiners? The following is, so far as I know, a complete list (except for two additions to be made shortly). (The asterisk is explained below.)

definite:
the
this/these, that/those
*my, *your, *his, *her, *its, *our, *their
‘s (see below)
*whose (relative)
indefinite:
a/an, *one, some
each, every
which, what, whichever, whatever
any, either
no, neither
*whose (interrogative)

The words listed all have the properties A-C, and typically they can occur either with or without a common noun as complement, though as mentioned earlier there are four exceptions where the common noun is obligatory.

The words marked with an asterisk may in fact be a single morphological realisation for two distinct syntactic words (written below in capitals). For example, one could be analysed syntactically as A ONE, to explain why this sequence isn’t available to match the one (as in the one thing I regret). In this analysis, the list of determiners would include A but not ONE, so there would be no precedent for including numerals (which otherwise combine freely with determiners: my two books, any three books, etc.).

Another case where morphology may be out of step with syntax is in all the possessive determiners. I have included ’S among the determiners because there are good grounds for recognising a clitic at the end of possessive phrases such as the King of England’s(Hudson 1990:276–82; Hudson 2004; Hudson 2013). One attraction of this analysis is that it explains why such a phrase acts like a determiner in relation to common-nouns: it acts like a determiner because its head is the determiner ’S. Now, if this analysis is right, then it allows an analysis of all the possessives in which they also contain ’S: most convincingly, perhaps, whose is a single morphological realisation for WHO + ’S and its = IT + ’S, but once these are recognised as two-word combinations, the same analysis extends easily to my = ME + ’S, your = YOU + ’S, and so on. This move greatly reduces the number of English determiners because there is just one possessive determiner, ’S, which combines with a number of ordinary pronouns just as it combines freely with full noun phrases; so all the starred words no longer qualify as determiners, leaving just 16 determiners.

The argument so far, then, leads to the conclusion that English has a relatively small number of determiners, forming a number of semantically-based groups:

article: the, a/an, some
demonstrative: this/these, that/those
possessive: ’S
interrogative:which, what, whichever, whatever
distributive: each, every
negative-polarity: any, either
negative: no, neither.

All these words allow a singular common noun to be countable, but of course some of them also combine with plural common nouns, as in these books or what books. But there are also pronouns that allow plural common nouns: we/us and you, as in we linguists and you students or you guys (which is rapidly turning into the default plural second-person pronoun). This addition gives us 18determiners, with various restrictions for number and countability.

Many grammars give a much longer list, so it is worth reviewing the words which do not qualify as determiners. This list is based on the one in a standard grammar(Huddleston & Pullum 2002:356), and includes my suggested alternative analyses:

numerals: two, three etc. – commonnouns which take other common nouns as complement (Hudson 1990:302–8).
universal quantifiers: all, both – pronouns which take almost any kind of noun as complement (all you boys, both the boys); if the complement is a personal pronoun, the usual order is reversed to give they all instead of the expected *all they.
another – AN + OTHER, but with two meanings, one of which matches that of SOME MORE (another beer = some more beer; another book = a different book).
a few, a little - A + common noun FEW/LITTLE, but with special syntax and semantics (possibly classifying FEW/LITTLE as adjective as well as noun in order to explain a very few/little alongside a good few and quite a few(Huddleston & Pullum 2002:392))
several, certain, various– restricted common nouns or adjectives (ibid)
much, many, few, little – both adjectives and nouns, to explain I didn’t see very much of it.
enough – a pronoun (to explain why it’s incompatible with a determiner: *my enough money)
sufficient – both an adjective and a noun, to explain I’ve got almost sufficient

Clearly the alternative analyses need to be fleshed out and justified, and a particular challenge is the rather odd pattern of words which allow almost andnearly: almostevery/all/any, but *almost each; nearly all but *nearly every/any/each – not to mention the equally odd difference between almost and nearly, whereby very is allowed by one but not by the other (very nearly but *very almost).

Leaving analytical details aside, the conclusion is clear: there is no evidence that any of these forms, with the possible exception of enough, is a determiner – a pronoun that takes just a common noun as its complement.It’s true that the rejected words all tend to occur at the start of a noun phrase, before the ordinary adjectives, but so do prepositions, and nobody would suggest that prepositions are determiners. This conclusion sits uncomfortably with the widespread assumption that logical quantifiers are always determiners (Keenan 2006); for instance, the universal quantifier is often identified with either all or every, but we have seen that one of these is a determiner while the other is not.

This detailed discussion of English was necessary because English is the most analysed language, and the language on which supposedly universal categories are often based. The category ‘determiner’ is probably the most prominent of such categories because it was specially invented for English, so it is important to recognise, first, that even in English, ‘determiner’ is not, in fact, a distinct ‘top-level’ word class, as usually assumed. The second point to note is that this rather ill-defined category has acted as the dumping ground for a disparate range of words that don’t fit comfortably into other categories. We now turn to the semantics of the English determiners, as defined here.

English determiners are semantically similar

Why have these particular words developed a shared grammatical profile? This calls for a functional approach in which each language’s grammar is seen as one ‘engineering solution’ to a particular set of functional challenges (Evans & Levinson 2009). The functional challenge here is how to identify a referent. Common nouns do this by invoking a general category such as ‘book’, which may be made more specific by means of modifiers such as ‘big book about linguistics’. Pronouns, on the other hand, identify referents more directly, typically by invoking features of the ongoing situation; so she identifies a female who is currently prominent, and this identifies something nearby and prominent. The contribution that determiners make is to combine these two methods by allowing a pronoun meaning to combine with the meaning defined by a common noun.