3.1.2 Endocentricity

An obvious consequence of the notion of projection is that we will never get a phrase of one category with a head of another. While this might seem a slightly perverse situation to want to prevent in the first place (why would verb phrases be headed by anything other than a verb?), it is certainly a logical possibility that there could be phrases of category X which do not contain a word of category X. For example the traditional view that preposition phrases can function adverbially could be captured under the following assumption:

In other words, a preposition phrase which behaves as an adverbial phrase is an adverb phrase headed by the preposition. Clearly this is something that would not be allowed by the X-bar rules in (1). Evidence favours the X-bar perspective and there is no reason to believe that just because something functions adverbially it is categorially the same as an adverb. For example, even when PPs are used adverbially, they still have different distributions to AP:

As we see in (9), a PP modifier of a verb must follow it, while an AP modifier may precede or follow it, even if the two modifiers have virtually the same interpretation. Thus a phrase headed by a preposition has a different categorial status to one headed by an adverb, supporting the X-bar claim that phrases have heads of the appropriate kind.

Moreover, the X-bar rules in (1) rule out another possibility if we assume that these are the only rules determining structure. While it might not make much sense to have a phrase with a head of a different category, the idea of a phrase that simply lacks a head is not so absurd. There is a traditional distinction made between endocentric and exocentric language elements. An endocentric phrase gets its properties from an element that it contains and hence this element can function by itself as the whole phrase. For example:

An exocentric phrase on the other hand contains no element that can have the same function as the whole phrase and so appears to have properties that are independent from the elements it contains. A standard example is:

The issue is rather complex. The traditional view mixes category and function in a way that is perhaps not helpful. The point is, however, that the X-bar rules in (1) claim that, categorially, all phrases are endocentric: in other words, all phrases have heads which determine their categorial nature.

There is one grammatical construction that seems at first to stand outside the X-bar system precisely in that it lacks a head: the clause. Certainly from a functional perspective the clause contains no element that could replace the whole construction: neither the subject nor the VP can function as clauses by themselves:

The examples in (12) all have very different natures, even categorially.

It might be argued that sometimes VPs can act as clauses:

However, such expressions have a special status and there is more to them than appears at the surface. The sentence in (13) is an imperative construction in which there appears to be no subject. However it is fairly clear that there is a definite subject understood in this sentence: you! An imperative cannot be interpreted as a command given to some third person and must be interpreted as directed towards the addressee. The question is then, what is the status of the subject of such sentences: are they only ‘understood’, present at some semantic level or are they merely ‘unpronounced’ though present at the grammatical level?

There is reason to believe that language makes much use of unpronounced elements that are nonetheless present grammatically and we will see many examples of such things in the following pages. One argument to support the assumption of an unpronounced subject in (13) comes from observations concerning the behaviour of reflexive pronouns such as himself. Unlike other pronouns, reflexives must refer to something else in the same sentence:

In (14a) himself can only be interpreted as referring to Fred and cannot, for example, be taken to mean someone else not mentioned in the sentence. Compare this to the behaviour of her in (14b). In this case the pronoun may either be taken as referring to Sue or to some other woman. We can say therefore that reflexive pronouns must have grammatical antecedents: some element present in the sentence which provides the reflexive with its reference. With this in mind, consider the following observations

As we see from (15), a by phrase containing a reflexive is interpreted to mean ‘unaccompanied’. In (15a), the reflexive refers to Pete and so it means that he was unaccompanied in eating the pie. In contrast, in (15b) the reflexive refers to the pie and so it means that the pie was unaccompanied (by ice cream for example) when Pete ate it. (16a) is grammatical even though there is no apparent antecedent for the reflexive. It is not surprising that the reflexive should be yourself however, as, as we have said, the understood subject of an imperative is you. Yet we cannot simply say that the antecedent ‘being understood’ is enough to satisfy the requirements of the reflexive as (16c) is ungrammatical. In this case the object is absent, though it is clearly understood that something was eaten in (16b). But this understanding is not enough to license the use of the reflexive in this case. So we conclude that the missing subject in (16a) is different from the missing object in (16c) and in particular that the missing subject has a more definite presence than the missing object. This would be so if the missing subject were present as an unpronounced grammatical entity while the missing object is absent grammatically and present only at the semantic level. In conclusion then, while imperatives might look like VP clauses which lack subjects, they are in fact full clauses with unpronounced subjects.

If the VP cannot be argued to function as a clause, one might try to argue that clauses and subjects have certain things in common. For example, a clause can act as a subject:

But this does not show that subjects are functional equivalent to clauses but quite the opposite: clauses may be functionally equivalent to subjects under certain circumstances.

Therefore it would appear that clauses are exocentric constructions, having no heads, and as such stand outside of the X-bar system. Later in this book, we will challenge this traditional conclusion and claim that clauses do indeed have heads, though the head is neither the subject nor the VP. From this perspective, X-bar theory is a completely general theory applying to all constructions of the language and given that X-bar theory consists of just three rules it does indeed seem that I-language principles are a lot simpler than observation of E-language phenomena would tend to suggest.