Chapter 5: Lexical Relations and Grammatical Relations

5.1. Introduction

According to the Grammar we have developed so far, a sentence is grammatical in English just in case it corresponds to a P-marker which is generatable by our PSRs and whose terminal nodes each dominate an item listed in the Lexicon under the appropriate category. The Grammar enables us to characterize a large number of grammatical sentences in English, but it also generates many ungrammatical ones, thus failing to fully reflect our intuitions about the language. For example, our Grammar correctly characterizes a sentence like The boy saw Bill as grammatical because it has the structure (1) that instantiates the rules of our Grammar and the Lexicon, but it also incorrectly treats sentences like *The boy died Bill, with the structure (2), as equally grammatical:

(1) (2)

Both P-markers are legitimate according to our Grammar, since they can be obtained easily by applying the PSR rules and selecting a word of the appropriate part of speech under each terminal node. Similarly, our Grammar correctly generates the grammatical sentence The driver died with an appropriate P-marker but it also incorrectly generates the ungrammatical *The driver saw with a nearly identical P-marker:

(3) (4)

Examples like (2) and (4) indicate that our Grammar over-generates. Below are a few other examples of over-generation, repeated from the previous chapter:

(5) a. *The men would put the book.
b. *John explained Bill the theory.
c. *The man elapsed.
d. *The man from Ohio met.

Clearly, then, we must revise our Grammar to prevent it from over-generating ungrammatical strings. In Chapter 4, we augmented our Grammar by adding rules that add to its power to account for the problem of under-generation. In this Chapter, we will consider a few ways to constrain the Grammar.

5.2. The Organization of the Lexicon

5.2.1. Predicate Argument Structure

One area to look to constrain our grammar is in the Lexicon. Up to now, we have assumed that our (mental) lexicon consists of lexical items each of which is specified for its grammatical category (part of speech). But that is clearly a gross simplification of the content of our mental lexicon. When we say that the word die is in our vocabulary, that we know this word, we know more than the fact that it is a verb. In addition to its categorial membership, we know how it's pronounced, and we know what it means. Hence instead of simply listing of lexical items under each category as we did in (103) in Chapter 3, we must assume that our mental lexicon is more sophisticated. In particular, for each word that we know, our Lexicon must contain 3 sorts of information: phonology, semantic, and syntactic. We shall assume that phonological information is encoded by way of a phonemic representation which represents its pronunciation, as in /daI/ for the word die, and /hÊv/ for have, etc. The semantic information may be encoded in the form of a paraphrase. And the syntactic information about a word includes a specification of the syntactic category of which it is a member. Thus, for the word die, our Lexicon must have an entry containing the following information minimally:

(6)

In fact, it must include more. We know not only that die is a verb, but also that it is a verb that occurs with only one noun phrase to make a sentence--it is a "one-place predicate". For a verb like see, we know that it is a "two-place predicate". And for a verb like give, we know that it is a "three-place predicate". This is clearly the kind of knowledge we have that enables us to judge (1) and (3) as grammatical and (2) and (4) as ungrammatical. In particular, (2) is bad because a one-place predicate occurs with two "arguments", and (3) is bad because a two-place predicate occurs with only one "argument". Similarly, the ungrammaticality of the following arises from the fact that hand requires three arguments but see requires exactly two:

(7) a. *I handed a toy.
b. *I handed the baby.
c. I handed a toy to the baby.
d. I handed the baby a toy.
(8) a. *I saw the baby a toy.
b. I saw a toy.
c. I saw the baby.

In other words, when we know a word, say a verb, we also know its Predicate-Argument Structure (PAS). This knowledge must be part of our mental Lexicon. We shall represent the PAS of a lexical item by means of the notation {1}, {1, 2}, {1, 2, 3} indicating that a given item is a one-place, two-place, or three-place predicate, respectively. Thus, a fuller lexical entry for die has the form of (9):

(9)

And the lexical entries for see and hand should look like the following:

(10) (11)

The information about the PAS of a given verb is then one way to distinguish sub-categories among the members of the category called a verb. We can think of the PAS properties of a verb as indicating the number of "roles" required by the plot of a given dramatic performance. The performance of a drama denoting an event of dying, for example, requires a one-person cast--it is a monologue. An event of seeing can only be performed by two role-players, one acting as the see-er and the other acting as the seen. Similarly, the verbs give and put are three-place predicates because an event of giving or of putting requires three role-players.

5.2.2. S-Selection: Thematic Relations

For each given predicate with its associated PAS, we also know what roles its arguments play. For example, words like cry, laugh, sneeze have the same PAS as arrive, fall, and remain, i.e., {1}. The single role associated with the former group, however, is different from that associated with the latter. In the former case, the single argument plays the role of an Agent--an initiator of an action or event. In the latter case, the single argument plays the role of a Theme, some person or thing to whom or which an event has occurred. Since an Agent is by definition an animate being, the argument that plays this role must denote an animate entity; but a Theme is not required to be animate. This difference among these verbs accounts for the following contrasts:

(12) a. The baby cried.
b. Sincerity cried.
c. *The table sneezed.
(13) a. The baby arrived.
b. The parcel arrived.
c. John fell.
d. The tower fell.

Hence, predicates may differ from each other not only with respect to their PAS's, but also with respect to their S-Selection (semantic-selection) properties: i.e., verbs with the same PAS may differ with respect to the Thematic Roles that they select for their argument(s) to play. Some more examples of the relevance of S-selection are given below.

The verbs surprise and buy may each be used as a two-place predicate, having the PAS {1, 2}:

(14) a. His proposal surprised me.
b. They bought an orange.

However, they differ in their S-selection properties: surprise S-selects a Theme for its first argument, and an Experiencer for its second argument; whereas buy S-selects an Agent for its first argument, and a Theme for its second argument. (An experiencer is an entity whose psychological state is affected. Hence it is animate by definition.) The following sentences are ill-formed because they each violate the S-selection properties of their verbs.

(15) a. *They surprised the orange.
b. *His proposal bought me.

Similarly, although the verbs give and put are both 3-place predicates each having the PAS {1, 2, 3}, they differ in that the former S-selects a Goal for its 3rd argument, whereas the latter S-selects a Location for its 3rd argument. A Goal argument may be realized by a PP with the preposition to, and a Location may be realized with a preposition like at, on, etc., but not vice versa. Hence we have the following contrasts:

(16) a. John gave a toy to the baby.
b. *John gave a toy on the baby.
(17) a. John put a toy on the table.
b. *John put a toy to the table.

To correctly reflect our lexical knowledge, we need to include the S-selection properties of a lexical item in its lexical entry. The entries for buy, surprise, and put would look like the following:

(18) (19)

(20)

And for verbs like declare and persuade, exemplified in (21), we can have lexical entries like (22)-(23):

(21) a. The senator declares that the First Lady isn't a bitch.
b. John persuaded me that I should try this dish.

(22) (23)

5.2.3. C-Selection: Subcategorization

Another kind of knowledge that speakers have about their Lexicons that enables them to distinguish among certain predicates is based on what the lexical items C-select (Category-select) for their objects or complements. For example, there is good reason to believe that the verbs reside and inhabit have the same meaning, belong to the same category (V), have the same PAS (a two-place predicate), and S-select the same Thematic Roles (theta-roles) ({Agent, Location}). However, where the location is expressed through a PP in the case of reside, it is expressed directly with an NP object in the case of inhabit.

(24) a. Anteaters resided in Southern California.
b. *Anteaters resided Southern California.

(25) a. Anteaters inhabited Southern California.
b. *Anteaters inhabited in Southern California.

Similarly, the verb arrive may optionally be used as a two place predicate, and in this case it is identical to the verb reach with respect to its category (V), its PAS ({1, 2}), and S-selection ({Agent, Location}). However, while the Location is expressed directly by an NP in the case of reach, it can only be expressed with a PP in the case of arrive:

(26) a. They arrived at home.
b. They reached home.

The two verbs ask and wonder have identical categorial status (V), PAS (each a two-place predicate), and both S-select {Agent, Interrogative Proposition). The interrogative proposition (i.e., question) may be expressed by an S':

(27) a. John asked what time it was.
b. John wondered what time it was.

The question may be "concealed" within an NP also (see Grimshaw 1979). In the case of ask, the concealed question can be directly expressed by the NP. However, with wonder it must be expressed by a PP:

(28) a. John asked the time.
b. *John wondered the time.
c. John wondered about the time.

A similar contrast can be found between the two verbs tell and inform. Both are 3-place predicates S-selecting {Agent, Goal, Proposition}:

(29) a. The girl told me that she liked Clinton's 200-dollar haircut.
b. The girl informed me that she liked Clinton's 200-dollar haircut.

In place of a Proposition, both can also S-select a Theme. However, the two verbs differ how the Theme be expressed categorially: for tell the Theme can directly take the form of an NP, whereas for inform it must take the form of a PP:

(30) a. The girl told me the story.
b. The girl informed me of the story.
c. *The girl informed me the story.

In other words, as speakers of English we have knowledge that whereas reside and arrive C-select a PP as a complement, inhabit and reach C-select an NP. And whereas ask and tell C-select NP-NP (or NP-S'), wonder and inform C-select NP-PP (or NP-S').

We can represent this aspect of speakers' knowledge as a feature of each given lexical item. If a verb C-selects an NP, we represent this property with the notation: + [___ NP] (has the property of being followed by an NP); if it C-selects an NP and an S' in that order, we express this with: + [____ NP S'], and if it occurs as the sole daughter of a VP, it has the C-selection property of + [____]. Thus, a fuller entry for the verbs put and declare should look like:

(31) (32)

And the lexical entries for ask and wonder would contain identical syntactic specifications except for their C-selection features:

(33) (34)

Thus, lexical items of the same part of speech may be further sub-classified according to their PAS, their S-selection properties, and/or their C-selection properties. The phenomenon of C-selection has also been known under the term strict subcategorization. Lexical items may be strictly subcategorized according to the kind of categories that may occur as their complements. Recall that our VP expansion rule, PSR3, states that a VP contains a V obligatorily as its head, and one or more optional peripheries:

(35)

This rule defines 7 possible sentence patterns in English, illustrated by the following examples:

(36) a. John died. V
b. John saw Bill. V NP
c. John resides in Boston. V PP
d. John knows that I like him. V S'
e. John gave Bill a present. V NP NP
f. John put a book on the table. V NP PP
g. John persuaded Bill that he should try the dish. V NP S'

In fact, these also illustrate 7 subcategories of English verbs, in terms of the seven different strict subcategorization features corresponding to these possibilities. These seven subcategories correspond to certain traditional labels of verbal classification.

(37) a. + [____] (intransitive verbs):

die, cry, sing, dance, come, go, arrive

b. + [____ NP] (transitive verbs):

see, kick, hit, like, kill, kiss, hug, cheat, etc.

c. + [____ PP] (incomplete intransitives):

reside, live, lie, (arrive), etc.

d. + [____ S'] (transitives)

say, think, know, claim, believe, consider, try, prefer, etc.

e. + [____ NP NP] (di-transitives)

give, buy, mail, send, tell, teach, etc.

f. + [____ NP PP] (incomplete transitives)

put, place, assign, (give), explain, etc.

g. + [____ NP S'] (di-transitives or incomplete transitives)

tell, persuade, convince, promise, order, command, etc.

5.3. Lexical Insertion

We have seen that our lexical knowledge entails our knowing not only the categorial feature but also the PAS, the S-Selection, and the C-Selection features of lexical items. This knowledge ensures that a given lexical item can occur only in certain environments but not in others. For example, our knowledge of the word put tells us that it belongs to the sub-category illustrated by (37f), because it subcategorizes for (C-selects) an NP-PP sequence. This piece of knowledge ensures that a sentence containing put will be judged unacceptable unless the verb occurs in a P-marker that instantiates the option VP ----> V NP PP of the VP Rule PSR3. In order to capture the idea that it is this sort of lexical knowledge that prevents our mental grammar (our English pheno-type) from over-generating ungrammatical sentences, we must add to our hypothesized English Grammar the following condition:

(38) Condition on Lexical Insertion

Lexical items may be inserted under terminal nodes within a P-marker only if the environments into which they are inserted match the categorial, PAS, S-Selection, and C-Selection requirements of the lexical items.

Given this condition and the appropriate requirements of the lexical entries, we can now solve the problems of over-generation we have been considering. For example, the ungrammaticality of (2):

(2) *The boy died Bill.

follows because it violates the PAS requirement of die, which allows the occurrence of only one NP argument, leaving the other unaccounted for. It also violates S-Selection because the verbs selects only one theta-role. And it also violates C-Selection because the verb C-selects nothing as its object. The ill-formedness of (4):

(4) *The driver saw.

also follows because the verb's PAS, S-selection and C-selection requirements are not met. The same reason applies to (5a):*The men would put the book. As for (12b, c), (15a, b), (16b) and (17b), these sentences are out because they each violate an S-selection requirement of their verbs. And sentences (24b), (25b), (28b), and (30c) are ill-formed because the C-selection requirements of their verbs are not satisfied.

In addition to these commonly observed restrictions, certain lexical items may impose further idiosyncratic restrictions. For example, the semantics of elapsed requires its subject to be a denotation of a duration of time. And the verb meet requires that, when it is used intransitively, its subject designates a group of two or more entities. These requirements account for the ungrammaticality of examples like (5c, d):

(5) c. *The man elapsed.
d. *The man from Ohio met.

Up to now, we have been looking at the lexical entries of verbs. Lexical items of other categories are also treated similarly. For example, we can also sub-classify adjectives on the basis of their PAS, S-Selection and/or C-Selection properties. Consider the following examples:

(39) a. John is aware.
b. John is aware of Bill's success.
c. John is aware that Bill won the prize.
d. *John is aware Bill's success.
e. *The building is aware.

The facts displayed in (39) indicate that our Lexicon must include an entry for aware with a specification of its PAS {1, (2)} (a one- or two-place predicate), its S-Selection frame {Perceiver, (Theme)}, and its C-Selection feature "+ [____ (PP/S')]" (note the optionality as expressed by the parentheses). On the other hand, an adjective like new will be listed strictly as a one-place predicate that S-selects a Theme as its subject argument and C-selects nothing: "+ [____]."1

Prepositions may also be sub-classified on similar grounds. For example, of, at, into always take an NP object; they have the C-Selection feature "+ [____ NP]." Furthermore, these items may be distinguished by their S-Selection features: of selects a Theme, at a Location, and into a Goal:

(40) a. of the man; *of.
b. at the restaurant; *at.
c. into the hole; *into.

The preposition from C-selects an NP or PP, and S-selects a Source:

(41) a. He appeared from behind my back.
b. He arrived from Bosnia.

Off, in, beyond may optionally C-select an NP, or nothing at all:

(42) a. He jumped off (the table).
b. The thief broke in.
c. You should look beyond.

And nouns, of course, can have their own entries. Some nouns may have S-Selection or C-Selection frames like verbs:

(43) a. their destruction of the city +[____ PP]
b. his arrival at home +[____ PP]
c. our expectation that he will win +[____ S']

Other properties of nouns that can be included in their entries include semantic features such as [+human], [+animate], [+singular], etc. These features are important because they partially make up the sense of a noun that limits its possible denotations (e.g., being an Agent, or Experiencer, or Location, etc.). And these are in turn crucial in determining whether they match the properties of other elements in a P-marker into which they are inserted.

Summarizing, up to now our investigation has led us to the postulation of a mental Grammar with the organization depicted in (44):

(44)

5.3. Adjuncts, Complements, Arguments

We have just seen that verbs may be subcategorized according to the kind of elements that must or may follow them. Thus, die is classified as an intransitive verb because it C-selects nothing, and see as a transitive because it C-selects exactly one NP, etc. However, notice that in the following sentences die can be followed by an NP, and see by two NPs:

(45) a. The man died last year.
b. John saw Bill last year.

In fact, all of the seven verb classes we have identified can take a time expression which could be in the form of an NP (e.g., last year), a PP (e.g., in the afternoon), an S' (e.g., when I saw him), or an adverbial phrase (e.g., recently)--in addition to the elements which they C-select (for which they strictly subcategorize). Compare (36) with the following:

(46) a. John died very recently. V AdvP
b. John saw Bill when he came in. V NP S'
c. John resides in Boston these days. V PP NP
d. John knows that I like him now. V S' NP
e. John gave Bill a present in the morning. V NP NP PP
f. John put a book on the table at 3 p.m.. V NP PP PP
g. John persuaded Bill that he should try it yesterday. V NP S' PP

All verb types also allow an expression of place, manner or means, or reason, or a combination of these, to follow them:

(47) a. John died in Paris because he had a pneumonia.
b. John saw Bill here before Mary arrived.
c. John resides in Boston happily these days.
d. John knows that I like him because it is obvious.
e. John gave Bill a present in the classroom.
f. John put a book on the table with a loud noise.
g. John told Bill the story yesterday because of my request.

Certain manner expressions may also appear pre-verbally in the form of an AdvP:

(48) a. John suddenly died.
b. John unexpectedly saw Bill.
c. John happily resides in Boston.
d. John certainly knows that I like him.
e. John secretly gave Bill a present.
f. John kindly put a book on the table.
g. John very quickly persuaded Bill that he should try the dish.

Does the appearance of these adverbial expressions of time, place, manner, reason, etc. invalidate our earlier generalization that verbs may be subcategorized on the basis of what they C-select? The answer is no. What we need to do is distinguish between two kinds of elements that follow a verb: complements and adjuncts, and remember that predicates are classified on the basis of the complements that they C-select, but not on the basis of the adjuncts that may occur with them. The kind of expressions illustrated in (36) are complements, in the sense that they help complete the meaning of a sentence as required by a verb. For example, John saw sounds "incomplete" and the addition of the NP Bill makes it complete, as in John saw Bill. The NP Bill is a complement. (A complement that has the form of an NP is called an object.) Similarly, in Boston is a complement in the sentence he resides in Boston, because without it the sentence is out for its incompleteness. An adjunct, in contrast to a complement, does not help to "complete" the meaning of a sentence. The sentence John saw Bill is not felt to be "missing something" regardless of whether or not an adverbial expression of any sort is added to it.

It is important to note that only complements serve to subcategorize verbs, but not adjuncts. This is because almost all adjuncts can (optionally) occur with all verb types in addition to their complements. Predicates C-select their complements, not their adjuncts. Only complement-selection enters into C-selection and needs to be specified for the lexical entry of each predicate.

In addition to this distinction between complements and adjuncts, a more general distinction between Arguments and Adjuncts also needs to be made. Arguments are those elements that enter into the PAS of a predicate and its S-selection properties. They include all the complements, and also the element which serves as the subject of a sentence. (A complement of the category NP is an object.) Arguments are the participants--the place holders--of the PAS of a predicate (classified in terms of being n-place, n a positive whole number). They are also the role players of theta-roles in the S-Selection frame of a predicate. However, adjuncts do not enter into considerations of the PAS or S-Selection of predicates, any more than they enter into C-Selection relations with a predicate. More generally, adjuncts do not figure in the organization of the Lexicon. The grammaticality of sentences like (45)-(48) does not pose a problem for our conception of the Lexicon assumed up to now. We do, however, have to slightly modify our PSR3 to allow for the generation of sentences containing adverbial adjuncts. The following is a reasonable modification:

(49)

This revision adds an optional Adverbial Phrase in pre-verbal position, and an optional XP at the end. The category XP is intended to represent several of the categories that an adverbial expression may take, and the asterisk is used to indicate that more than one instance of the category may appear, to account for the fact that a sentence may contain several adverbial expressions.

The distinction between adjuncts and complements or arguments for verbs can be made for other categories as well. Some peripheral elements occurring with the head noun within an NP may be complements or arguments of the noun, whereas others are adjuncts to it. Consider the following examples:

(50) a. his destruction of my plan
b. their discussion of the issue
c. a student of music
d. a teacher of English

Each of these NPs contains a Det, N, and a PP. In each case, the of-PP is a complement to the head N. For example, just as my plan is a complement of the verb destroy in He destroyed my plan (it completes the meaning of the transitive verb destroy), so can we say that the PP of my plan is a complement to the noun destruction because it specifies what was destroyed. Similarly, the PPs of the issue, of music, of English are complements to their head nouns in the same sense that the NPs the issue, music, English are complements to the corresponding verbs:

(51) a. He destroyed my plan.
b. They discussed the issue.
c. Barry studies music.
d. Josephine teaches English.

On the other hand, PPs like on Sunday, in the classroom are adjuncts of the NPs:

(52) a. his arrival on Sunday
b. their discussion in the classroom

just as the same PPs would be adjuncts to the corresponding verbs:

(53) a. He arrived on Sunday.
b. They discussed the issue in the classroom.

Similarly, the PPs from New York, in blue jeans and APs like young, dangerous, do not serve any complement function; they are adjuncts:

(54) a. a student from New York
b. the teacher in blue jeans

So, we see that PPs that occur with a head noun can serve as complements or adjuncts to the noun. This also applies to S's. In the following examples, the underlined S' is a complement to its head noun in (b) as much as it is a complement to a verb in (a):

(55) a. I believe that he will win.
b. The belief that he will win turned out to be wrong.

(56) a. Mary claims that she never saw Bill.
b. I don't believe Mary's claim that she never saw Bill.

(57) a. John asserted that he won the debate.
b. John's assertion that he won the debate is not unreasonable.

But the underlined S's below, called relative clauses, are adjuncts to the nouns they modify. They are not complements; such nouns as man and books do not take complements (unlike event-denoting nouns like destruction, discussion, claim, etc.).

(58) The man who you met at the party is my brother.

(59) I have read many books that he wrote.

In summary, all phrases obligatorily contain a head, and optionally from one to several peripheral elements. Some of the peripheral elements are complements, while others are adjuncts. Complements denote objects and other elements which help to complete the meaning of a head. Their occurrence is often obligatory in the environment of a verb. Adjuncts provide additional information to modify their heads, but are not required by their heads. This distinction is important to an understanding of the theory of C-selection and of the organization of the Lexicon. C-selection refers only to complements of given lexical items, not to their adjuncts. Lexical items are marked for, and classified on the basis of, their C-selection feature (which specifies the categories of complements that they select or subcategorize for). Adjuncts may occur with any lexical items and they do not enter into C-selection at all.

5.4. Grammatical Categories, Grammatical Relations, and Semantic Relations

Up to now, we have employed several ways to refer to elements within a sentence. For example, we refer to certain strings within sentences as NPs, VPs, AP, PP, or S', or Det, etc. These are the grammatical categories we recognize in syntax. Membership within a grammatical category is defined in terms of its form or its substitutability relationship with other members of the same category.

We have also used terms like subjects, objects, complements, or adjuncts to refer to certain phrases. These terms designate the grammatical relations or grammatical functions of phrases. They designate relations or functions, because when we say that a certain NP is a subject, we mean it is a subject of a sentence or of a verb. The term subject is meaningful only when it means subject of something. For this reason, these terms might as well as be given as subject-of, object-of, complement-of, or adjunct of. Terms like head and predicate also refer to grammatical relations. In generative grammar, it has been suggested that, at least in some languages, grammatical relations can be defined in terms of grammatical categories and the configurations (P-markers) in which the categories occur. For example, we can define subject as referring to the phrase (NP or S') that is immediately dominated by S, predicate as the VP node immediately dominated by S, object of a verb as the NP immediately dominated by VP, etc.:

(60) (61)

In addition to referring to phrases by their categories and the grammatical relations they bear with respect to other elements within a sentence, we may also refer to them in terms of the semantic or thematic relations they bear to other elements. These refer to terms like Agent, Theme, Experiencer, Goal, Proposition, etc. They designate the semantic or thematic roles that given arguments may take with respect to given events denoted by the head. Note again that only arguments enter into thematic relations with their heads. Adjuncts do not.

These three ways of classifying linguistic elements are independent of each other. For example, a phrase classified as an NP under the categorial criterion may serve as an object, subject, or object-of-preposition function, and it may play the role of an Agent, a Theme, an Experiencer, or a Goal, etc. A phrase may be PP by category, a complement by function (or grammatical relation), and a Location by its semantic or thematic relation. A relative clause is an adjunct by function, and an S' by category. And so forth. Thus,

(62) a. John handed a toy to the baby.
b. A toy was handed to the baby by John.
c. The baby was handed a toy by John.

John, a toy, and the baby are Agent, Theme, and Goal, respectively in each of these sentences. But while John is a subject in (a), it is an object-of-preposition in (b)-(c); a toy is an object in (a) and (c), but a subject in (b); and the baby is a P-object in (a) and (b), but a subject in (c). In the examples below, the NP the boy serves as the subject of a sentences throughout, but the thematic roles it plays vary from one sentence to another (Agent in (a), Theme in (b), Goal in (c), Experiencer in (d)):

(63) a. The boy kicked the ball.
b. The boy was scolded by his sister.
c. The boy was given a toy.
d. The boy enjoyed the game.

In other words, the notions grammatical categories (NP vs. VP, etc.), grammatical relations (subject, object, adjunct, head, adjunct, etc.), and semantic/thematic relations (Agent, Patient, etc.) are independent notions. All of them are useful in the description of our syntactic knowledge.

5.5. Summary

In this chapter, we examined a problem of over-generation of our Grammar, and solved it by postulating an enriched Lexicon which specifies, for each lexical item, the idiosyncratic features of its syntactic category, its predicate-argument structure (PAS), and its properties of S-Selection and of C-selection, and by requiring that when lexical items are inserted into P-markers, their idiosyncratic features be matched by the environment into which they are inserted.

As a clarification of the nature of C-Selection, we also introduced the distinction between adjuncts and arguments, the latter including complements. We noted that adjuncts are optional modifiers that can occur with any sub-category of a given category, and that only complements enter into the specification of C-Selection. We also slightly revised our PSR3 to account for the occurrence of adjuncts in VPs.

In addition, we have seen that linguistic elements may be categorized in different ways. Grammatical categories are those that are defined in terms of the paradigmatic relationship among members of each class. Grammatical relations and thematic relations designate certain syntactic relations and semantic-role relations among elements within a sentence. These three ways to view given linguistic elements are independent of each other, in principle.

In discussing the adjuncts of a sentence, we had occasion to revise our PSR3 as in (49). In problem 1, Exercise 4, furthermore, we also saw that we should revise the PSR further in order to allow for the generation of APs within VPs, like:

(64) a. That son of a gun is aggressive.
b. This is totally unbelievable.
c. John must be very happy.
d. He looks stupid.
e. Mary felt very sad.

We can do this by adding the option VP ----> V AP to the PSR:

(65)

And in problem 2, Exercise 4, we saw that the AP rule should be revised to account for the following examples:

(66) a. She is afraid.
b. She is very afraid.
c. She is aware of the problem.
d. She is certain that the problems will arise.
e. John is proud of his sister.
f. I am extremely fascinated with your idea.
g. *I am afraid of Bill that he will bring problems to us.
h. *I am afraid him.
i. *Linda is fully clear the situation.

What we need to do is revise our AP rule as:

(67)

The Grammar, now with the enriched Lexicon and these further revisions, is more adequate than before. Let us refer to this as English Grammar, v. 1.3.

Proposed English Grammar, v. 1.3.:

A. PSRs

B: the Lexicon (sample listing)

a. N -----> books, ideas, mother, man, student, girl, house, friend, cement, pilot, . . . .
b. V -----> kick, laugh, cry, buy, live, tell, give, put, say, . . . .
c. A -----> good, bad, colorless, green, long, redundant, . . . .
d. P -----> at, in, under, on, through, up, . . . .
e. Art ----> a, the, some, . . . .
f. Dem ----> this, that, these, those
g. deg ----> very, extremely, . . . .
h. Conj. ----> and, but, or, . . . .
i. Comp ----> that, if, whether, for, why, who, etc.
j. Inf. ----> to
k. Modal ----> can, may, must, will, shall, could, might, ....
l. Perf. ----> have
m. Prog. ----> be
n. Adv. ----> quickly, suddenly, carefully, etc.

Homework 5

1. Consider the two verbs wait, await. (a) Make a few sentences to illustrate their different behaviors, (b) give a lexical entry for await with specification of its category, PAS, and S- and C-selection features, and (c) do the same for wait.

2. Observe the sentences in each group below, and give an appropriate lexical entry for each verb. (You can ignore the phonological and semantic information; just concentrate on the syntactic features.)

a. i. *Mary handed the baby.
ii. *Mary handed a toy.
iii. Mary handed a toy to the baby.

b. i. *John convinced that he should take over the job.
ii. Mary convinced me that he should take over the job.
iii. *Mary convinced me the proposal.
iv. Mary convinced me of the proposal.

c. i. John believed that Bill will win.
ii. *John believed me that Bill will win.
iii. John believed my story.
iv. *John believed me my story.
v. *John believed me of my story.

d. i. Mary gave a toy to the baby.
ii. Mary gave the baby a toy.
iii. Mary explained the question to me.
iv. *Mary explained me the question.

e. i. Mary is afraid that it will rain.
ii. Mary is afraid.
iii. *Mary is afraid the rain.
iv. Mary is afraid of the rain.

f. i. Mary fears that it will rain.
ii. *Mary fears.
iii. Mary fears the bad weather.
iv. *Mary fears of the weather.

3. For each of the underlined phrases, please identify (a) the syntactic category it belongs to; (b) the grammatical relation it bears within the sentence, and (c) the thematic relation it bears (if any: if the phrase is an adjunct, then (c) is not applicable).

a. John talked to Bill in the garden.
b. John put the plants in the garden.
c. That he won the prize surprised me.
d. I don't believe the claim that he won the prize.
e. The man who you met yesterday has left.
f. Bill was praised by the President.
g. That she has not arrived worries me.

4. On the basis of our proposed Grammar, v. 1.3, assign an appropriate P-marker to the following sentences.

a. My brother suddenly appeared at the doorstep.
b. John saw Bill when he came in.
c. John resides in Boston these days.
d. He will retire in style in a year.
e. The girl and her father said that they have been discussing the possibilities.

5. The sentences below are ambiguous:

a. He promised that he would do the job yesterday.
b. He told me that he would talk to Bill at the bar.

(a) means either that he made the promise yesterday that he would do the job sometime, or the promise was that he would do the job yesterday. (b) means either that he told me at the bar about something, or that he would talk at the bar to Bill. The ambiguities arise, in each case, from the fact that the adjunct phrases yesterday and at the bar can be construed as adjuncts of the embedded VP, or as adjuncts of the main clause VP. In other words, we can assign two different P-markers to each of the sentences to represent the two meanings they have.

Can you draw the two P-markers for each sentence?

1 In problem #2, Exercise 4, you were asked to write a PSR to account for the internal structure of an AP. The most probable (and correct) answer may be the following, given our current assumptions:

A major difference between A and V is that A does not C-select an NP. Hence there are few adjectival classes than verbal classes. There is only one exception to this generalization in American English: the adjective worth, as in This is not worth your time.