1.2 Word Categories

1.2.2 Categories

1.2.3 Morphological criteria for determining category

1.3 A Typology of Word Categories

1.3.6 Functionally underspecified categories

2.1.6 Labels

There are also other phrases associated with the verb (VPs), with adjectives (APs) and indeed with every kind of word category that we have discussed (noun phrases – NPs, inflectional phrases – IPs, CPs and degree adverb phrases – DegPs).

2.3.1 Substitution

This shows us that the string of words, under his mitre forms a constituent of the sentence, but the category of this constituent is not so easy to determine from the category of its replacement. We might suppose that there is a pronoun and therefore it replaces DPs, but this constituent is made up of a preposition (under) followed by a DP (his mitre) which does not distribute like a DP:

2.3.3 Coordination

The coordination test, however, needs to be carefully applied. Recall that the way coordination works is to take two elements and form them into a single element that has the same function as the two elements would have individually. It therefore follows that two elements cannot be coordinated if they do not have the same function, even though they may be constituents of the same category. For example, if we tried to coordinate a PP that was a locative modifier of a DP with one which was an instrumental modifier of a VP the result would be ungrammatical:

2.3.4 Single-word phrases

In such trees the dual status of these elements as both word and phrasal categories is clearly represented.

3.1.1 Rewrite rules and some terminology

Recall from chapter 2, rewrite rules which tell us how structures of various kinds decompose into their constituent parts. The rules in (1) are like these, only far more general. The generality is achieved through the use of category variables, X and Y, which stand for any possible category (nouns, verbs, prepositions, determiners, etc.). Thus these rules tell us how phrases in general are structured, not how particular VPs, PPs or DPs are.

3.1.1 Rewrite rules and some terminology

There are several things to note about this structure. First there are two immediate constituents of the X' (pronounced “X bar”): X, which is called the head of the phrase and the complement YP. The complement, which, as its label suggests is a phrase of any possible category, follows the head. This is a fact about English and in other languages the complement may precede the head.

3.1.1 Rewrite rules and some terminology

Again there are several things to note about this structure. Once more, there are two immediate constituents of the phrase. The specifier, a phrase of any category, precedes the X', the constituent just discussed containing the head and the complement. Again the ordering of these two constituents is language dependent: specifiers precede X's in English, but this is not necessarily so in all languages. Specifiers are a little more difficult to exemplify than complements due to complications that we have yet to discuss. However, the following are fairly straightforward cases:

3.1.1 Rewrite rules and some terminology

Note that the X' and the phrase share the same categorial status (X) and so if X' is P' XP will be PP, etc. As X' is the same category as the head, it follows that the whole phrase will be of the same category as the head. In this way, the head of the phrase determines the phrase’s category.

3.1.1 Rewrite rules and some terminology

The property of sharing category between the head, the X' and the phrase is called projection. We say that the head projects its categorial status to the X' and ultimately to the XP. If we put the two parts of the structure together, we can more clearly see how projection works:

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:

3.1.2 Endocentricity

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:

3.1.2 Endocentricity

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.

3.1.3 Heads and Complements

But if the structural rules of the grammar are themselves so general as to not make reference to categories how do categorial features come to be in structures? One would have thought that if all the grammar is constructed from are rules that tell us how phrases are shaped in general, then there should only be one kind of phrase: an XP.

3.1.3 Heads and Complements

The structure in (21) is still incomplete however and we must now consider how to complete it. Let us concentrate firstly on the complement position. At the moment this is expressed by the general phrase symbol YP. This tells us that only a phrasal element can sit here, but it does not tell us what category that phrase must be. Yet, it is clear that we cannot insert a complement of just any category into this position:

3.1.3 Heads and Complements

This restriction clearly does not come from the X-bar rules as these state that complements can be of any category, which in general is absolutely true. But in specific cases, there must be specific complements. Again it is properties of heads which determine this. Recall that part of the lexical entry for a word concerns its subcategorisation. The subcategorisation frame of a lexical element tells us what kind of complement there can be. For fall, for instance, it is specified that the complement is prepositional:

3.1.3 Heads and Complements

Thus through the notion of subcategorisation, the head imposes restrictions on the complement position, allowing only elements of a certain category to occupy this position:

3.1.4 Specifiers

These arguments are nearly always DPs and so, unlike the complement they do not seem to be restricted differently by different heads in terms of their category. This is reflected in the lexical entries of the relevant heads in the fact that subcategorised elements are always complements and subjects are never subcategorised for. Of course, there are restrictions placed on these specifier arguments from the predicate, but of a more semantic nature. The verb assigns a Θ-role to these arguments and so the argument must be semantically compatible with the Θ-role it has to bear. For example:

3.1.4 Specifiers

The verbs ask and wonder both have questions as their complements, but only with ask can this question be expressed by a DP like the time. Thus there are extra restrictions imposed on complements which go beyond the requirement that they be compatible with the Θ-role that is assigned to them. In short, specifiers are more generally restricted than complements as they tend to be a uniform category for different heads and merely have to be compatible with the meaning of the head.

3.1.6 Summary

Before moving on to look at other aspects of syntactic processes, let us consolidate what we have said in this section. X-bar theory is a theory of basic structure comprising of just three rules. These rules are generally applicable to all structures and substructures, no matter what their category: they are category neutral. The categorial status of a specific structure depends on the lexical elements it contains, in particular one word acts as the head of each phrase and this determines the category of the phrase by projecting its own categorial properties, established in the lexicon, to the X' node above it and ultimately to the XP.

3.2.2.1 D-structure and Theta Theory

Let us consider the nature of D- and S-structure a little more closely. An obvious question is why it is that some elements start off in one position and then move to another. To answer this question we have to ask about why elements occupy the positions they do at any level of description. This is a matter of distribution: there are grammatical principles which determine the range of possible positions of categories of certain types. X-bar principles obviously have a large role to play in this, determining head, complement, specifier and adjunction positions. But as both D- and S-structures conform to X-bar principles, this clearly is not what differentiates the two. Obviously there must be other grammatical principles holding at D-structure which are not applicable at S-structure and vice versa.

3.2.3 Traces

We can get some understanding of this situation if we make the following assumption: when an element moves, it leaves behind a copy of itself in the extraction site. This copy is called a trace and is envisaged to be identical to the moved element in terms of its grammatical and semantic properties. Thus the category of the trace, its role in the thematic structure of the sentence and its referential properties are the same as the moved element. The main way in which the trace differs from the moved element is that the trace has no phonological content and hence is unpronounced. We have already considered the possibility of phonologically empty but grammatically active elements when we discussed imperatives at the beginning of this chapter. A trace is another such element.

4.1 Why the Noun is not the Head of the DP

A further problem with this analysis is that adjuncts adjoined to XP or X' are phrasal. Only adjuncts adjoined to a head are X⁰ categories. But the determiner looks suspiciously like a word and to analyse it as a phrase by itself begs the question of why determiners never have complements, specifiers or adjuncts of their own:

4.1 Why the Noun is not the Head of the DP

Of course, the assumption of an empty category must have other motivations than just their necessity to make the analysis work. With the empty subject of an imperative we pointed out that this could act as the antecedent of a reflexive pronoun and with the trace we demonstrated how this prevents other elements from moving into a position vacated by another moved element. Is there any independent justification for either of the empty heads in (27) or (28)? If we consider the empty noun in (28), the only justification this has is to provide a head for the NP. The entire semantic content and the grammatical features of the phrase are contributed by the pronoun itself: the NP is third person singular because the pronoun is third person singular and the NP has a reference which is determined by the pronoun. Thus there is no independent support for the existence of this empty noun.

4.3 Multiple Determiners

The only difference between pre-determiners and other determiners is that when they appear in front of a DP the pre-determiner has an optional of whereas this is obligatory with other types of determiner. The fact that we can have a post-determiner in this position is interesting. We have stated that these come in specifier of NP, but where is the NP in (67c) that the post-determiner is in the specifier of? If there is an NP in this structure the head of which cannot be seen, it seems that we are forced to assume that this head is an unpronounced empty category and therefore the structure must be something like the following:

4.3 Multiple Determiners

As this phenomenon is restricted to structures involving group nouns, it seems that this is strong evidence in favour of the assumption that pre-determiner constructions involve a group noun in the position we have proposed an empty category.

5.2.3.2 Light verbs and ergatives

The disadvantages of this analysis are: i) there is an empty light verb and ii) the wrong word order is predicted. The supposition of the empty verb is, of course, not a problem in itself. We have seen a number of instances of empty categories that are well justified and enable us to provide accounts for phenomena that would otherwise be mysterious. As long as we can independently justify the assumption of an empty element, given that language apparently makes use of such things, there is no problem in the assumption itself. There is both semantic and syntactic evidence of the existence of the empty light verb. We will return to the latter, but the semantic evidence is fairly obvious: the structure is interpreted as a causative and the presence of this meaning justifies the assumption of a light verb which provides it. Similarly, the presence of a ‘causer’ argument justifies the assumption of a predicate which assigns the relevant Θ-role. As there is no such visible predicate which can do such things in (48b) our conclusion is that this predicate is ‘invisible’.

5.4.1 Adverbs

Another observation that can be made from the data in (170) is that adverbs may appear behind all verbal elements. There are a number of possible ways to capture this fact. One is to assume that adjunction is free from ordering restrictions. Indeed it does seem that different adjuncts can come on different sides of whatever they modify: the PP modifier, as we shall see, typically follows the verbal complex. Thus, adjunction in general is not restricted to a particular side as are complements and specifiers. Adverbs therefore may simply take advantage of this freedom and be adjoined either to the left or the right of the VP. The alternative would be to have adverbs generated on one side of the VP and then achieve the other position via a movement. Jackendoff (1977), for example, argued for this position on the basis of the similarity between adverbs and adjectives. Recall that in chapter 1 we analysed adverbs and adjectives as belonging to the same general category, so one might expect grammatical principles to apply to both in a similar way. Jackendoff’s observation was that adjectives typically precede the nouns that they modify:

Chapter6 Inflectional Phrases

In chapter 1 we introduced the category of inflection, consisting of modal auxiliaries, the marker of the infinitival clause to and tense morphemes. It is with this category that we will be concerned in the present chapter. We will show that providing a standard X-bar treatment of this element solves a number of problems that we have noted in previous chapters.

6.1 The structure of IP

Inflectional elements are word level categories such as will, can, may, must, etc. as well as to, ‑ed (and its numerous irregular manifestations) and ‑s. In chapter 1 we argued that these all belong to one category, ‘Inflection’ (I) because of their complementary distribution:

6.2.3 Tense and Agreement

These observations show that this zero morpheme is in complementary distribution with tense and thus the straightforward conclusion is that it IS tense. But how can this be if tense is an inflectional element and the zero morpheme is not in complementary distribution with modal auxiliaries, which are also inflectional elements? What the data show is that it is not modals that tense is in complementary distribution with, but the zero tense morpheme that accompanies the modal and hence the conclusion is that if modals are of the category ‘inflection’, then tense is not of this category. Given that tense is situated in front of the VP, we can assume that it is a head that selects a verbal complement and given that it follows the inflectional elements (i.e. modals) it must project a verbal phrase. In other words, tense is yet another light verb:

7.2 The Clause as CP

The generalisation is now that all verbs which select for a finite declarative complement select for a CP. This is fairly easy to capture in terms of the notion of canonical structural realisation principles. The idea behind this is quite simple. Basically, certain arguments are canonically realised by certain categories. For example, themes are typically realised as DPs and locations as PPs. This is their ‘canonical realisation’. It may be that a certain degree of non-canonical realisation of arguments is possible, for example the nominal home can realise a goal argument usually realised by a PP:

7.5.3 Negative fronting

What is the nature of XP and how does it interact with the other movements we have reviewed? If we are to maintain our view that functional categories select for a very limited selection of complements, then as XP is the complement of a complementiser this argues that it is something like an IP. However, presumably X is a functional element itself and it takes an IP complement, which makes it more like a complementiser. Given that complementisers are categorised as [+F, –N, –V] and inflections are [+F, –N, +V] categories, a category which shares properties of them both would be [+F, –N], with an undefined V feature. Let us refer to this category a little ‘i’, reflecting the use of ‘v’ to represent a verbal element with an undefined F feature. This element heads an iP and so the structure can be represented as:

7.5.3 Negative fronting

The difference however, between the fronted focus and the fronted negative is that the latter induces inversion to ‘i’, while the former does not. Referring back to the difference between wh-movement that triggers inversion and wh-movement that does not, we proposed that inversion is triggered when there is no head to agree with. In embedded interrogatives the complementiser position could be filled and hence there will be no inversion. If we project these ideas on to the current situation, we conclude that with focus the i head is filled by some abstract element but with negative fronting the head position is unfilled. We know that the negative head is a verbal element of the category ‘v’ and so it cannot be generated directly in i. The only way for the negative head to get to i is for it to move and yet we know that the negative is not able to move to I to support the inflection in English, as it is in Finnish. Thus, there is no way for the negative head to get to i and hence when a negative element moves to the specifier of iP it will induce inversion to provide a head for it.

8.1.2 Clauses without IP

The fact that heads subcategorise for their complements in terms of the complement’s category would seem to suggest that these clauses do indeed differ in terms of their categorial statuses.

8.1.2 Clauses without IP

But this proposal claims that the head of the predicate is no longer the head of the whole clause and hence it no longer determines the categorial status of the clause. It would be difficult therefore to account for the observations of (30) and (31) where different verbs subcategorise for different small clauses in terms of the category of the predicate. A separate question concerning (34) is what the status of X is. Haegeman (1994) argues that this head is an agreement element, i.e. what we have been calling I. Thus small clauses, according to Haegeman, are IPs (AgrPs in her terminology):

8.3 The Gerund

In this case, the ing element behaves exactly like a noun syntactically. For example it heads a phrase which can be the complement of a determiner. As we know determiners take NP complements, so the ing element must have the category of a noun. A further fact about nouns is that their DP complements cannot be assigned case, which is why there is an of preposition inserted into the structure. With a verb, its light verb will assign Case and hence there is no need to insert of. This can be clearly seen in the different behaviours of verbs and their derived nominal counterparts: