1.3 A Typology of Word Categories

Having introduced some of the basic concepts, let us now turn to look at what categories we need to refer to in the description of a language like English. In generative linguistics it is often seen as a positive aim to keep basic theoretical equipment to a bare minimum and not to expand these unnecessarily. This can be seen in the standard approach to word categories in terms of the attempt to keep these to as small a number as possible. In the present book we will mainly be concerned with eight basic categories. These come in two general types: thematic categories and functional categories. In the thematic categories we have verbs (V), nouns (N), adjectives (A) and prepositions (P) and in the functional categories there are inflections (I), determiners (D), degree adverbs (Deg) and complementisers (C). Thus we have the following classification system:

1.3.1 Categorial features

Before we start to look at the properties of individual categories, we will make the typology of categories described in (23) a little more systematic. One might wonder why there are these categories and why their division is so regular: four thematic categories and four functional ones. Moreover, we may have the feeling that the categories given in (23) are not completely unrelated to each other. For example, it is often felt that nouns and verbs are somehow opposites of each other or that adjectives have some things in common with nouns and other things in common with verbs. Even across the thematic/functional divide, we may see similarities. For example, words like the, these and some are determiners and these seem more related to nouns, which they usually accompany, than to verbs. Modal auxiliary verbs, such as may, can and must, which as we will see are classified as belonging to the inflections, are obviously more closely related to verbs than nouns.

1.3.1 Categorial features

To develop the system a little further, consider the intuitions that adjectives seem to have something in common with nouns, as they are typically used to modify nouns, as in crazy kid or thoughtful suggestion, but they also seem to have something in common with verbs, as they have certain distributional properties in common:

1.3.1 Categorial features

In this example, rich is an adjective and running is a verb and obviously they can both appear in similar environments. But if nouns and verbs are diametrically opposed to each other, how can adjectives be similar to both? The answer is that adjectives share different features with both nouns and verbs. Thus, we may categorise both nouns and adjectives as [+N] and both verbs and adjectives as [+V] and in this way adjectives will share features with both nouns and verbs. Of course, they will also have features different from nouns and verbs, but as we do not want to categorise adjectives as the same as the other categories, this is a positive aspect of this proposal. Adjectives can therefore be categorised as:

1.3.1 Categorial features

Having demonstrated that we can capture similarities and differences between word categories using binary features, let us turn to the issue of what categories there are. We will start this discussion by considering the two binary features [±N] and [±V]. So far we have shown how combinations of these features can be used to define nouns, verbs and adjectives. The two binary features can be combined in four possible ways, however, and hence there is one possible combination that we have yet to associate with a category. This is demonstrated by the following tab

1.3.1 Categorial features

However, this cannot be put down to good fortune. After all, categorising elements in terms of these features has consequences concerning what other categories are related to or different from these elements. Note that the feature combination in (29) predicts that while prepositions differ from nouns in that they are [–N], they are similar to nouns in that they are [–V]. Similarly, prepositions differ from verbs in being [–V], but they share the [–N] feature with them. Thus prepositions are predicted to be similar to nouns and verbs, but in a different way to how adjectives are similar to these categories. Indeed, while prepositions do not have similar distribution patterns as verbs, as do adjectives, they share another property with verbs. Consider the following observations:

1.3.1 Categorial features

In (30), we see that both verbs (see) and prepositions (to) can be followed by a word such as him, which is a pronoun. Nouns (portrait) and adjectives (mindful) cannot. We might claim therefore that the ability to be followed by a pronoun is restricted to the [–N] categories. Now consider the following:

1.3.1 Categorial features

As shown in (31), a noun like Sally and a preposition such as underneath can sit in the position between the words was and that in this English construction, known as a cleft construction. However, an adjective (stupid) and a verb (fishing) cannot occupy this position. We might claim therefore that this position can only be occupied by [–V] categories.

1.3.1 Categorial features

We see from the discussion above the predictive power of the system that we have set up: the system predicted that there should be a fourth thematic category that has certain properties and these fit the category of prepositions very well. We can take this as evidence in favour of this system of features. What else does the system predict? It is clearly predicted that if we add a third binary feature to the two we have just been discussing, then a further four categories will be defined. This again matches perfectly with the description of categories we started this section with, as seen in (23). With the third feature, [±F], there should be four functional categories which match the four thematic categories in terms of their feature settings for [±N] and [±V]. We have already seen how determiners and modals can be analysed as functional nouns and functional verbs, respectively. The expectation is that degree adverbs, such as so and too, and complementisers, such as that and if, should be related to adjectives and prepositions in the same way. As degree adverbs modify adjectives in a very similar way to how determiners modify nouns, it is not difficult to conclude that degree adverbs are functional adjectives. This leaves complementisers to fill the final place as functional prepositions. There is evidence in favour of this assumption, but it rests on notions not yet introduced, so we will have to wait until later to demonstrate it.

1.3.1 Categorial features

We see from the discussion above the predictive power of the system that we have set up: the system predicted that there should be a fourth thematic category that has certain properties and these fit the category of prepositions very well. We can take this as evidence in favour of this system of features. What else does the system predict? It is clearly predicted that if we add a third binary feature to the two we have just been discussing, then a further four categories will be defined. This again matches perfectly with the description of categories we started this section with, as seen in (23). With the third feature, [±F], there should be four functional categories which match the four thematic categories in terms of their feature settings for [±N] and [±V]. We have already seen how determiners and modals can be analysed as functional nouns and functional verbs, respectively. The expectation is that degree adverbs, such as so and too, and complementisers, such as that and if, should be related to adjectives and prepositions in the same way. As degree adverbs modify adjectives in a very similar way to how determiners modify nouns, it is not difficult to conclude that degree adverbs are functional adjectives. This leaves complementisers to fill the final place as functional prepositions. There is evidence in favour of this assumption, but it rests on notions not yet introduced, so we will have to wait until later to demonstrate it.

1.3.2 Predicates and arguments

So far we have mostly spoken of predicates that happen to be verbs, but it is not the case that all predicates are verbs. We have seen one case where this was not so, in (34b). Here we said the predicate was is tall. However considering the meaning of Tom is tall, we can see that the main semantic relations exist between Tom and tall and the is part simply expresses that Tom’s being tall is true at the present time (compare this with Tom was tall). Thus, we might claim that tall, which is an adjective also has a Θ-role as part of its lexical entry:

1.3.2 Predicates and arguments

Just like verbs, some adjectives express a relationship between two arguments:

1.3.4.1 Verbs

Other verbs take adjectival or adverbial complements:

1.3.4.2 Nouns

This inability to take nominal complements is something nouns share with adjectives, as we shall see. Verbs pattern with prepositions in this respect. Thus we can claim that whatever property it is that allows verbs and prepositions to take nominal complements, it is connected to the [–N] feature that they both share. The [+N] categories (nouns and adjectives) obviously lack this property.

1.3.4.2 Nouns

A final difference between possessors and arguments is that the possessor relationship is restricted to nouns whereas thematic relationships seem to be available to all thematic categories: we can find themes, experiencers, etc. for verbs, nouns or adjectives.

1.3.4.3 Adjectives

1.3.4.4 Prepositions

The last thematic category we will consider is prepositions. The pattern with nouns in being [–V] and with verbs in being [–N] and therefore do not share any feature with adjectives, apart from [–F] as they are both thematic.

1.3.4.4 Prepositions

The one complement that prepositions fail to have is clausal. While verbs, nouns and adjectives can all take clausal complements beginning with the word that, it seems that this is not possible for prepositions:

1.3.5.2 Determiners

Traditionally this group of quantifiers are known as post-determiners as they always follow other determiners (which are sometimes called central determiners). This terminology gives the impression that post-determiners are a subclass of determiner, which is likely to be inaccurate. These elements often have many adjectival qualities, including being able to be modified by degree adverbs and having comparative and superlative forms:

1.3.5.2 Determiners

For these reasons we will consider these elements as adjectival and will put off their discussion until a later section.

1.3.5.3 Degree Adverbs

So far we have looked at auxiliary verbs, which accompany verbs, and determiners, which accompany nouns, classifying these as functional equivalents of the categories they accompany. The obvious choice for functional adjectives, therefore, are the degree adverbs that accompany them:

1.3.5.3 Degree Adverbs

This would suggest that not all of these words should be categorised as degree adverbs, that is, as words with [+F, +N, +V] categorial features. Given that normal adverbs can be used to modify adjectives, some of the cases in (167) can simply be taken as adverbs, especially those that are formed from adjectives by the ‑ly morpheme:

1.3.5.3 Degree Adverbs

Others however are more difficult to categorise. Words like very do not appear to be able to be used as typical adverbs, modifying verbs or sentences, but are restricted to modifying adjectives as are the degree adverbs:

1.3.5.3 Degree Adverbs

Besides distributional properties, degree adverbs also have other properties that unify them. For example, it is typical for a degree adverb to appear alongside a clausal element which follows the adjective being modified:

1.3.5.3 Degree Adverbs

This clause specifies the bounds to which the degree of the property expressed by the modified adjective is given. Note that plain adverb modifiers of adjectives do not appear with such limiting clauses:

1.3.5.3 Degree Adverbs

There is an interpretation in which these kind of constructions are not ungrammatical. However, this is where the following clauses are associated with the adjectives or even the whole clause rather than the degree modifiers:

1.3.5.3 Degree Adverbs

Clearly, these behave exactly like the periphrastic constructions, and hence would seem to involve a degree adverb. The obvious choice would be the comparative and superlative morphemes themselves, which would suggest an analysis similar to what was proposed for verbal inflections: the comparative and superlative are independent lexical elements which are inserted into an expression separately into the degree adverb position and then by a syntactic process become attached to the adjective:

1.3.6 Functionally underspecified categories

If there is a non-functional non-thematic verb, then it is predicted that there must be non-functional non-thematic nouns, adjectives and prepositions. To what extent is this prediction fulfilled? There are nouns which do not appear to behave like thematic nouns and yet are clearly not categorised as determiners either. Consider the following examples:

1.3.6 Functionally underspecified categories

Next, let us consider possible non-thematic non-functional adjectives. Recall that post-determiners are elements which seem to have adjectival properties in that they have comparative and superlative forms and may be modified by adverbs:

1.3.6 Functionally underspecified categories

It is clear that these elements are not thematic and hence should not be analysed as adjectives such as pink, certain or keen, for example. For one thing, they cannot be used predicatively as adjectives can, making them more like degree adverbs:

1.3.6 Functionally underspecified categories

However, degree adverbs do not have comparative and superlative forms and so it would be inaccurate to categorise the post-determiners as functional adjectives (i.e. as [+F, +N, +V]). Therefore we propose that these elements be categorised as [+N, +V] elements. It seems that post-determiners always select nominal complements, which is why they have been confused with determiners and hence they have the following lexical entries:

1.3.6 Functionally underspecified categories

As we pointed out, nouns and adjectives seem not to be able to take nominal complements and hence the preposition of is inserted so that the complement is prepositional instead. This preposition plays no role in the thematic interpretation of these constructions however, the thematic relations hold between the noun or the adjective and the following nominal element. Another such preposition is the by which is found in passive clauses:

2.1.7 Rules

Although the system of rules in (36) is capable of describing the structures of a good number of English sentences, it is clear that we would need many more rules to attempt to describe the structures of all English sentences. For example, not every DP is made up of a determiner followed by a noun. Some may contain just a determiner, such as this for example, or a determiner, an adjective and a noun, such as a rusty kettle. A DP may indeed contain, amongst other things another sentence, such as the diagnosis that she had flu. It is clear that we would need many rewrite rules to capture all the possibilities for English DPs.

2.2.2 The object

We have also seen that the object is a more limited complement in some ways. For example, Verbs and Prepositions have objects, but nouns and adjectives do not:

3.1.5.1 Adjunction to X-bar

It is clear that the noun is the head of this construction as it can act as the complement of a determiner and determiners take nominal complements, not adjectival ones:

3.1.5.1 Adjunction to X-bar

The bracketed elements in (32a) and (b) have the same distribution and hence we can conclude they have the same categorial status. As this phrase in (32b) contains only a noun, we conclude that it is an NP. In (32c) however, the phrase following the determiner contains only an adjective and is ungrammatical. This clearly has a different distribution to the other two phrases, indicating that the adjective in (32a) is not the head of this phrase.

3.1.5.1 Adjunction to X-bar

However, there can be more than one adjectival modifier of a noun:

3.1.5.1 Adjunction to X-bar

Thus, the adjectival modifier is an adjunct of the noun. We will argue in a later chapter that adjectival modifiers follow the specifier of the NP and hence adjectival phrases are attached in a position between the specifier and the head. As we see in the following, this puts them as adjuncts to the N':

3.1.5.3 Adjunction to head

Finally, we will consider the case of adjunction to a head, using compound nouns for an example. There are a number of complexities which we will not go into here, sticking to more straightforward cases. Compound nouns are formed by putting two otherwise independent elements, usually an adjective and a noun or two nouns, together and use the resulting unit as a single noun:

3.1.5.3 Adjunction to head

Sometimes the spelling indicates that the two parts of the noun are put together to form one word, but other times it does not. We will not delve into the mysteries of English spelling here. Note that when compounds are formed from an adjective and a noun, the noun is second. Moreover, if there is a main semantic element of the two parts of the compound, this is also the second element: an armchair is a kind of chair not a kind of arm. We might claim therefore that the second element is the head and the first is a modifier of the head. The structures we get are:

3.1.5.3 Adjunction to head

In both cases the last noun is the head, but the adjuncts are related in different ways. In (43a), the adjective hard modifies the head to form a compound hard disk. We then add the second adjunct which modifies this. Thus we have the structure:

4.1 Why the Noun is not the Head of the DP

The assumption that the determiner is not the head leads to further problems for the analysis of determiners and pronouns themselves. First consider the determiner. If this is not the head then it is presumably an adjunct or a specifier within the NP. We should therefore expect it to behave as such. Determiners do not appear to be adjoined within the NP as they do not behave like adjectival modifiers, which we have analysed as N' adjuncts in the previous chapter. Adjectives are recursive modifiers of nouns and can normally be arranged in any order, as we might expect of an adjunct:

4.3 Multiple Determiners

So what is the proper analysis of multiple determiners? The easiest case to deal with is the post-determiner. We argued in chapter 1 that these are adjectival elements which are undefined for the F feature and hence are neither functional nor thematic adjectives. The fact that they may be modified in the same way as thematic adjectives, however, indicates that they head APs:

4.3 Multiple Determiners

Note that the object of a preposition is an accusative position and hence that prepositions are Case assigners. Inserting of then allows an otherwise Caseless DP to be assigned Case. Of-insertion is however a very limited phenomenon. It happens with the DP complements of nouns and adjectives and nowhere else which might be argued to be a Caseless position:

4.3 Multiple Determiners

It seems therefore that the appearance of the meaningless of is a good indication of the presence of a noun or an adjective. In the structure we are considering concerning the pre-determiner, the appearance of the of can be taken as strong evidence in favour of the presence of a noun even though one is not visible.

5.2.7 Phrasal verbs

A final problem for (127) is that it tends to go against the general pattern of compounding in English. When a complex head is formed from two heads by adjoining one to the other, it is generally the case that the head of the compound is the leftmost element. This is true in compound nouns and adjectives, but also with verbs:

5.2.7 Phrasal verbs

In all these cases of compounding, the rightmost element provides the compound with its syntactic and semantic properties. So an armchair is a kind of chair not a kind of arm and a white lie is a noun not an adjective.

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:

5.4.1 Adverbs

If we assume therefore that the basic position of the adjective is before the noun that it modifies, we might take this to indicate that the basic position of the adverb is before the verb that it modifies and therefore that its post-verbal position is a derived one. We are not really in much of a position to be able to evaluate either of these positions and therefore we will leave the matter unresolved.

8.2.1 Raising

Certain adjectives can also appear in raising structures. As adjectives do not assign Case, if an adjective takes a non-finite complement, the subject of that complement will not get Case and will therefore have to move. Furthermore, if the adjective does not assign a Θ-role to its subject, the subject position will be underlyingly vacant and will therefore either need to be filled by a pleonastic element or by a DP moving into it:

8.2.1 Raising

In this example, the subject starts off in the lowest clause as the subject of the adjective skilled. It then moves out of three clauses to the subject position of the raising verb. In principle, then, it might appear that there is no limit to how far a subject may raise. However, it is interesting that in order for this to happen, each predicate between the original clause and the final landing site of the raised subject must be either a raising predicate or a passive verb and moreover each intervening clause must be non-finite and have a vacant subject position. If any of these conditions is not upheld, the sentence is ungrammatical:

8.3 The Gerund

Here we see a verb refuse in its ing form. This form is used in a number of contexts which should be separated. Obviously it is the form used to express continuous aspect, as in he is running. It may also be used to form an adjective from a verb, as in his smiling face. Whether or not these are connected or totally distinct morphemes is an issue we will not attempt to fathom and we will concentrate only on the use of this morpheme in the gerund. In the gerund, the ing could be taken to be a ‘nominalising’ morpheme, turning a verb into a noun. Indeed, there are cases where this is exactly what it is:

8.3 The Gerund

As the ing element in (84b) has an of before its DP complement, we can assume that it is a noun not a verb. Finally, these kind of ing elements are modified by adjectives rather than adverbs again indicating their nominal status:

8.3 The Gerund

If we now refer back to (83), we note that some of these features are missing. For a start, there is no preposition before the DP complement. This suggests that this ing element is not a noun, but a verb. This is confirmed by the fact that this element is modified by an adverb, not an adjective: