Stephen van Vlack

Sookmyung Women`s University

Graduate School of English

Introduction to Linguistics

Fall 2003

Answers for Cook and Newson, `Chapter 5a`, pp. 159-173

1. What are predicates and arguments and how are they related to theta theory?

1. A predicate is something which says something about the elements in the sentence, their relationships or qualities. Generally predicates are VPs. The arguments are the things that the predicates has an effect on. Arguments are generally NPs or PPs. These are important terms in the study of syntax because they have a direct effect on the arrangement of data in the mental lexicon. The number of arguments a predicate has as well as their arrangement (ordering) is encoded in the mental lexicon. The basic idea is that we cannot really know the full meaning of a predicate unless we know what arguments it takes. The basic arguments that a predicate takes it called its argument structure, but there is more to the predicate-argument relationship than just that. In addition to the structural relationship there is a semantic one. This semantic relationship is encoded in what is called theta-roles. Every argument has a theta-role which reflects its semantic role in the clause. These theta-roles are given by the predicate and the same predicate always gives the same theta-roles.

2. What is a theta grid and how does it work?

2. A theta grid is, according to linguists, a part of the lexical entry for a predicate. Basically, it is a statement of the theta roles which the predicate gives its arguments. Again, the theta grid is part of a word`s lexical entry. The basic idea is that when a person is going to say something, they fire the lexical entry of the verb in their brain. As they fire that lexical entry they are activating a wide variety of information including the theta grid for that verb. The information in the theta grid is very important for an analysis of the utterance at the level of LF. The theta grid gives a lot of information about the meaning of the verb and its arguments. Different verbs will have different theta grids.

3. What is the difference between S-selection and c-selection and do we really need both?

3. S-selection (Semantic-selection) is basically the theta grid. It outlines the possible -roles that a predicate gives and in what order. C-selection (category-selection), on the other hand, identifies the categories of the arguments that a predicate can take. Basically, c-selection outlines the subcategorization information of the verb. It gives information on the number of arguments, their categories, and their order. Looking at this we see that S-selection is semantic in nature, while c-selection is structural, but that in many cases the information they give is the same. The verb KILL, for example has <AGENT, PATIENT> as its S-selection and [NP __ NP] as its c-selection. Since we know that AGENTS and PATIENTS are usually NPs, then the information given in the c-selection is redundant. People have come to question whether we actually need both. In many cases, actually, we don`t need both. Since the goal of linguistics is to come up with theories that are as simple as possible, while remaining powerful, maybe we should drop c-selection. But let`s not be too hasty. There are situations when we need both. Some verbs have the same s-selection, but different c-selection. C-selection is important here because it can explain the difference between the two. See Cook and Newson, pp. 162-163 for a good example of this.

4. What does a lexical entry contain?

4. No one really knows exactly what a lexical entry contains. Linguists make educated observations on what must be contained there by looking carefully at how language seems to work. Based on data from languages linguists can be relatively sure of what is contained there. The first and foremost, there are connections between lexical items. These connections, while strictly speaking are not part of a lexical entry`s contents, form a necessary and far-reaching part of the meaning of the lexical item. Any good semantics book, such as Saeed. 1997. Semantics. London: Blackwell, will provide a general description of these semantic-based connections in the lexicon. Most of the connections have been studied and identified through psycholinguistic experiments and you will find more about such connections in any good psycholinguistics book.

Apart form these connections there is overt semantic (grammatical) information that is housed in the lexicon in the entry for each lexical item (lexeme). The most primary thing is the lexical category of the lexeme. Words that are able to used as several categories are thought to have different, but related (connected) lexical entries. Some representation of the meaning of the lexeme must be there as well, but semanticists have yet to determine how this is represented. One possibility would be that each lexeme is made up of a series of abstract and finite semantic properties. This theory, while it seems to be quite sound, has had tremendous difficulties in formalization. In addition to category and meaning a lexical entry must contain information about the environment in which the lexeme can be used. This includes S-selection and c-selection, for verbs and the possible theta-roles that a noun might be able to take. Since Language Acquisition, according to Chomsky, is primarily lexical in nature then there has to be information about what each lexeme can occur with. This is, again, basically the S-selection. The information can also be more specific, for example, we know that certain verbs take only certain prepositional modifiers. The fact that the verb walk can take the prepositions on, out, above, and in, but not at as modifiers must be encoded as well in the lexicon. So, it seems that varied and large amounts or even specific information is housed in the lexicon. It also seems that the lexical entries for verbs are much more extensive then for any of the other lexical categories.

5. How is a lexical item projected into a structure?

5. A lexical item projected into a structure in a simple way. When we want to go and say something we form the concept in our mind. According to the specific rules and lexical composition of our language the intended concept is transferred to lexical form and constructed. The important thing to remember is that the lexemes are built from the bottom up. I am referring here to the tree structures that we are so used to analyzing. Trees are not built in our minds the way that we build them on paper. The lexical items are projected in to the structure at the lexical, bottom level. The rest of the structure is built on top of that. So, we stick a verb into the structure. Based on the Projection Principle and Endocentricity Constraint, the other levels of the VP are constructed on top of the lexical verb. This is how entire utterances are put together, once again, moving from the bottom up; i.e., by building the structure from the lexical items (whether overt or covert).

6. What`s one problem with the theta Criterion?

6. The theta Criterion states that each argument has only one theta-role and that only one theta-role can be given to each argument. This seems to be quite problematic because it is too restrictive. There seems to be cases in which some arguments seems to have more than one theta-role. Look at the following sentences.

A. Bob broke his arm.

B. Bob broke his arm on purpose.

C. Bob killed himself.

In sentence A, Bob is clearly the PATIENT/EXPERIENCER. This presents no problem for us. In sentence B, however, Bob is both the PATIENT and AGENT. Bob is the experiencer of the broken arm and is just as clearly the entity which caused this to happen. Sentence B is a clear violation of the theta Criterion. Such phenomena have caused Chomsky to try to rethink the basic concept of theta-roles. One idea that he has toyed with is the more abstract concept FI, which simply states that ever element in an utterance has a specific purpose. Everything is there for a reason. It is a mere question of economy. Just how this idea plays out in language theory has yet to be seen.

7. How are arguments assigned their theta roles?

7. Theta marking is the process by which theta-roles are assigned by predicates to their arguments. This is done based on the theta grid in the lexical entry of the predicate as well as the c-selection and S-selection requirements of the predicate. Chomsky believes that there is a sisterhood requirement on theta marking. Since complements are sisters to V, then there is no problem in theta marking for them, but that about subjects? Subjects are more distant from the verb that complements. Subjects are sisters to V`. In fact, it seems like subjects are given theta-roles by both the verb and its complement. Look at the examples below.

A. Bob broke the window

B. Bob broke his leg.

In A Bob is clearly the AGENT, but in B Bob is the PATIENT. How can this be if the verb is the same in both. It must be that the verb and complement do actually work together in marking the theta-role of the subject. We, therefore, need to posit two different types of theta marking, internal and external. Theta marking on the complement is internal, while marking on the subject is external.