Stephen van Vlack

Sookmyung Women`s University

Graduate School of TESOL

Introduction to Linguistics

Fall 2006

Week 12 - Answers for Discussion

A Brief Review

Before we jump into what the authors try to tackle this week in the book, let`s quickly review what we have done with syntax to date. We learned that words are merged to form units called constituents in much the same way that sounds and morphemes are put together to form words. These constituents are then used to form phrases which are headed by one word, the lexical category which will dominate the phrase. We, thus, have phrase types determined by the lexical category of the head. It is important to remember that in generative theory all the elements are separate. Thus when doing syntax we worry not so much about the meaning but rather of the syntactic behavior of the phrases. All observation and explanation must be made using only ideas central to the syntactic component. Thus, theories of syntax are created by observing the syntactic behavior of words alone.

In the beginning of the generative movement it was observed that phrases of the same type behave in the same or similar ways in relation to where they are placed in relation to other phrases. That is, certain phrases typically occupy similar positions in a structure. Remember, the key here for generativists is simplicity and systematicity. They are looking to design a model for language that is not merely descriptive. Generative models should indicate how utterances can be created, not just describe in an idiosyncratic way the ones that are there. Generativists want their models to be simple yet powerful because they should, in theory, describe how every human language goes about doing this. For this reason generativists decided to look deeply into language. They have to try to find what all languages have in common despite the seemingly huge surface differences. Chomsky in developing his system took inspiration from an older theory already present in phonology, D-structure and S-structure. Thus, it is also attractive and maybe natural to assume that some of the underlying ideas inherent the systems we learned in phonology and morphology are the same or similar in syntax.

Utterances (We will avoid using the word sentence here for reasons we mentioned last week.) are made up of words which have been merged into constituents, and then phrases and finally clauses. A clause necessarily contains two parts, a subject and a predicate. A simple utterance is one which contains only one clause, while a complex utterance contains more than one clause. Again, complex utterances are created by merging clauses.

Trees

We also looked at how we can draw trees to show us how this merger works. Trees are drawn based on what are called phrase structure rules. It is assumed that these phrase structure rules are housed in the brain and that there is a severely limited number of them. There are rules which pertain to different arguments a clause can contain. Look below at the three phrase structure rules.

In class we looked at phrase structure rules from a very descriptive point of view and I mentioned that such simple/descriptive trees are not very useful. Nor are they systematic. To solve this problem, Jackendoff (1977) came up with the theory of X-bar.

The purpose of tree diagrams was originally to simply create a visual representation of the structural aspects of utterances. No one initially suspected at all that they might actually be used to do more than merely describe the structure of utterances in a very simple way. All this changed with the advent of x-bar theory. X-bar theory was first formalized by Jackendoff in 1977 with the publication of X-bar Syntax: A Study of Phrase Structure. From this point on x-bar became a theory in itself. This was forced by the basic underlying principle of x-bar theory, which was that all phrases are essentially the same. This simple idea forced the issue and the trees became more abstract. Once more, x-bar has played an important role in constraining and helping to simplify Chomsky`s overall theory of language. The simple principles which make up x-bar theory and the structures that are so obvious in the trees that are made using this theory have had a profound effect on syntactic theory over the years. Most of the principles and parameters that have emerged make use of this framework. Just look at binding relations, theta-marking, case-marking, and government, not to mention movement. The current descriptions of these important principles would be impossible without the accompanying concepts of x-bar theory. It is, then, possible to posit that there might be tree-like representations in the brain as well as part of UG. Thus x-bar became part of UG and not just a handy device that teachers used to use to teach their students grammar.

The underlying principle of X-bar is that all phrases, whether lexical or functional, behave in the same structurally. We can thus label all phrases as X. Leveling is important in X-bar and there are three levels inherent in each phrase. So a lexical item is projected into a tree and immediately the three other levels (X, X`, X``) appear. X-bar theory is extremely simple.

In X-bar theory there are only three real phrase structure rules.

NP -----> (D) N`      (Specifier Rule) XP -----> (YP) X`

(N` ------> N` (PP))   (Adjunct Rule) X` -------> X` (YP)

N` ------> N (PP)        (Complement Rule) X` --------> X (YP)

{PS. The parenthesis in the rules indicates that something is optional.}

Taking a look at this we find there are two things that we need to go over. The first one is the notion of tree diagrams. In the blue book they have been drawing flat trees with only two levels, the phrase level (NP, PP, VP, etc.) and the word level (N, P, V, etc.). We now want to argue that there are levels (at least three). We will call these levels double bar (V̋ = VP), bar (V`), and just the word level (V). As you can see from the phrase structure rules above, adjuncts occur at a different level in the tree than complements.

Let`s take a look now at how adjuncts and complements differ and make sure that this is the right idea. Look at the examples below. These examples are taken from Radford (1988).

(1) a student of Physics (complement)

(2) a student with long hair (adjunct)

We can differentiate complements from adjuncts semantically by saying that in the former their meanings are more essential to the head while the latter just offer addition, often unrelated information. So, for example, it is important or even essential to know what a student studies, but the length of their hair is irrelevant.

On a syntactic level, what is interesting is that the two behave differently.

(3) a student of Physics with long hair

(4) *a student with long hair of Physics

Here the second phrase is not grammatical. This tells us that complements must occur closer to their heads than adjuncts. Now look at this.

(5) *the student of Physics of chemistry

(6) the student with long hair in the corner

Here we can see that the adjunct rule is recursive while the complement rule is not. If this is true then we should be able to put our adjuncts in any order.

(6a) the student in the corner with long hair

Likewise if the adjunct rule is recursive we should be able to stack adjunct ad infinitum by simply adding

more X` levels to the structure. This is shown below in (6b)

(6b) the student with long hair in the corner wearing jeans with a tattered notebook etc...

Now, how about this?

(7) the student with long hair his father cut

Here we have an NP (student) which has an adjunct (with long hair) that in turn has another complement (his father cut). We can see that the second complement can`t be moved to modify the other NP in the utterance.

(8a) the student with long hair his father cut on Wednesday

(8b) *the student with long hair on Wednesday his father cut

We can plainly see that complements are inherently attached closer to their heads than adjuncts and therefore must occur on a closer level.

Now that we have looked at these more layered trees, we can mover forward. Up until Section 19 of the text, we looked at phrases in an idiosyncratic way. We have also really only looked at and used lexical phrases systematically.

There were as many different types of phrases as there are word types. Remember, in syntax a large part of what we want to do is limit the options available to us when we go to create sentences. One of the best ways of doing that is to limiting the number of categories.

The four lexical phrases in x-bar syntax are Noun Phrase (NP), Verb Phrase (VP), Prepositional Phrase (PP), and Adjective/Adverb Phrase (AP). The structure of the phrase is preserved by the endocentricity constraint, which states that all lexical phrases must have the same structure. They must all contain a head as well of the other levels necessary in the description. This means that all lexical categories must follow the same basic principles. This not only retains the structural uniformity of the tree, but works to curtail possibilities in the tree. For example, because of endocentricity a VP must go to V`, V, and finally the lexical representation, which carries the same features that are present at the VP level. The lexical category cannot change, nor can the specific nature of the VP. All the levels must be realized for each entry.

Inherent in the x-bar theory is also the idea of levels, as explained above. The idea for levels was developed based on sentences like the one shown in (9) below.

(9) The elephant whipped the farmer with its tail.

In this utterance the VP is composed of a verb and an NP and PP. After a cursory glance it is possible to see that the status of the NP and the PP is different. First of all the PP is optional, while the NP is obligatory.

(10) The elephant whipped the farmer.

(11) *The elephant whipped.

(12) *The elephant whipped with its tail.

Part of the problem with so-called `baby trees` is that they cannot show the difference between phrases like the farmer and with its tail. For this reason, we need to posit the idea of levels in the tree. On closer inspection we can see some further differences between the two. We can stack one but not the other.

(13) *The elephant whipped the farmer his wife with its tail.

(14) The elephant whipped the farmer with its tail, on Wednesday, in the mud, while performing a jig.

We can also change the order of one, but not the other. That is, we can move one around, but the other is stuck with the verb.

(15) The elephant, with its tail, whipped the farmer.

(16) *The elephant, the farmer, whipped with its tail.

The phrase, the farmer, is a complement and must be located near the verb it modifies on the tree. The phrase, with its tail, is an adjunct and is not as closely related to the verb. We can show these distinctions in the tree by putting them in at different levels.

Complements are the daughters of X` and the sisters of X. This places them close to the entity they modify. In fact, they are sisters.

Adjuncts are the daughters of X` and the sisters of X`. This puts them at a higher level on the tree and further from the lexical entity. They are like cousins of the word they modify, a much more distant relation than being sisters. This distance also allows them to occur recursively and to move rather freely. To add more adjuncts all we have to do is add more levels via X` to the tree structure.

Thus, we have now shown how the three levels of the tree structures work by describing complements and adjuncts. There is still one more grammatical function that we need to quickly go over, namely that of specifiers. Specifiers in x-bar theory are the daughters of XP and the sisters of X`. Once more, in English they branch to the left of their head. Now to understand the role or the position of specifiers in Syntactic theory we need to understand the difference between subjects and objects (arguments) are they are thought to relate to the verb (predicate). The predicate (verb) is seen by syntacticians a being the most important element of a clause. For the moment we can think about the predicate as being the head of a clause, but remember this will change as we add in functional categories soon. It is the predicate which chooses its complements. The choice of complements is part of a predicates lexical entry, as we discussed so long ago. Remember that in syntactic theory this does not mean that a specific verb picks a specific complement as in (17) below.

(17) The crocodile loves chicken.

The claim is not that the verb love chooses a specific word chicken. It is claimed that a specific type of verb (transitive) specifies that is requires a complement (or two, or none) and if a complement is required then the category if the complement, but not the word itself is specified. This should all be review, so we will move on. Now, the important part is that the predicate does not stipulate anything in regard to its subject. Subjects are outside of the verbs sphere of influence. This is reflected in the spaces they are claimed to occupy in the tree structure; namely, specifier position. We can see that spec (specifier) is in a different place than complements. It is much higher and therefore more distant from the verb. Thus, placement in the tree structure is intended to be indicative of the structural relationship (as defined in part by constituency) between elements. Such arrangements are made based on observations of how such elements behave syntactically regarding movement possibilities and other less overt behaviors such as case-marking and theta-role marking.

Central to the specifier - VP relation is the idea of TENSE and AGREEMENT. It has been observed and posited that it is the subject which imposes TENSE and AGREEMENT features in the verb. For this to happen the subject in the specifier position must be higher in the tree than the verb, which it is. This, on the basis of observations regarding selection restrictions of predicates (predicates select complements, not subjects) and TENSE and AGREEMENT marking it has been concluded in x-bar that subjects occupy specifier position. We are now ready to tackle the more abstract idea of functional categories.

New Functional Categories

In addition to the lexical categories (NP, VP, PP, AP) that are so familiar to us, linguists have also long posited that there need to be functional categories as well. This is not new idea as grammarians, both traditional prescriptive and descriptive, have long claimed that there are such words as function words which distinguish themselves from content words. Part of the problem with the traditional description of function words is that there were simply too many different kinds. One of the triumphs of modern generative syntax has been to limit and constrain the number of functional categories down to three or four major types. This has been achieved, following the same tactics used in limiting and describing lexical categories. Basically, linguists have looked at how such functional units behave syntactically not what they mean or the semantic/pragmatic functions they have. Focusing on the latter at the expense of the former will make it virtually impossible to understand modern generative linguistic theory. Thus, it is with the functional categories more than anywhere else that one must learn to think about language differently.

An interesting theory related to this is Wexler and Manzini`s (1987) Lexical Parameterization Hypothesis, which claims that all languages in the world are identical in the behavior of their lexical categories. This includes the mental lexicon. Thus, if a verb is a transitive verb in one language then its translation equivalent will be a transitive verb in all languages. Basically, all verbs in all languages have the same selection restrictions and generate the same basic deep structures. The same holds for all other lexical categories. According to this hypothesis they all behave the same cross-linguistically.

Functional categories, according to this hypothesis, however, are where languages differ. It is in functional categories that parameters apply. Thus while all language can be claimed as having the same functional categories these functional categories behave different across language as a result of different parameter settings. Below then are the functional categories that have been proposed for languages and the way they have been proposed to work in English.

IP

One of the things that the authors of the blue book proposed was that there is a new category called INFL or I. This was proposed based on the fact that auxiliaries (is, have, modals) and infinitive markers (to) in English are there and behave similarly as shown in the two examples below.

(18) The bear will eat the porcupine.

(19) The bear wants to eat the porcupine.

Both auxiliaries and infinitive markers serve to indicate TENSE and AGREEMENT features or the absence of them. This tells us that syntactic structures are projections of lexical items (words). These lexical items all have certain features and that representations of these features are held in the brain. There are head, specifier, and complement features. As utterances are constructed all these features must be checked to see that they match. Based on this we want to posit that I is the head of the clause in that is controls most of the features we just mentioned. Our trees will now be headed by I and the VP will be a complement of I. In x-bar there is no category called S.

INFL Phrase or IP is a functional category. It has been claimed to hold AGREEMENT and TENSE information. It is different from lexical phrases in that is may or may not be filled with a word. In the sentence, `I like trees`, the IP will be empty of any overt words and rather is filled with an abstract grammatical component. In the sentence, `Porky porcupine would kill you`, the IP is filled with a lexical item. In the tree IP appears above the level of VP. VP becomes the complement of IP. The big issue here is how the contents of IP and VP are merged, if necessary. So, for example in the utterance, `George loves dogs`, does the verb move up to IP to merge with the tense and agreement features to make the word `likes` or do the features trickle from IP down to the verb. Let`s see if we can figure out which happens.

(20) Bill loves pork.

(21) Does Bill love pork?

(22) *Loves Bill pork?

(23) What does Bill do?

       - Love pork.

(24) What does Bill do?

       - *Loves pork

From the examples above we can conclude that it is the features of TENSE and AGREEMENT which trickle down onto the VP if they are not being claimed by an overt unit in the IP rather than having the verb move into IP to claim the features. The evidence provided in the above examples shows that verbs in English do not move. We know this because if we want to form questions in English we can only use a dummy auxiliary verb and it is this dummy verb which is marked for TENSE and AGREEMENT as we can see in (21). Verbs are not able to move forward to make questions. Interesting as well is the phenomenon shown in (23) and (24). We could argue that in these constructions, as in imperatives, there is no IP since there is neither a subject or any TENSE or AGREEMENT allowed. Since we can tell when IP is not there as in the special examples above, it must also be true that IP is in fact usually there, but it is sometimes empty.

Now, in class we also argued that the category IP is in fact better dealt with when it is split into two different categories. These new categories are AGRP (Agreement Phrase) and TP (Tense Phrase). Much of the reasoning for claiming that these two are in fact necessary has come from investigations into exceptional case marking and exceptional clauses of the type shown below. Involved in the discussion is the covert category PRO (big pro). Little pro (short for `pronominal`) is different and not so important in English.

(25) The alligator wants [to eat pork]

(26) The alligator wants [John to eat pork]

(27) The alligator wants [him to eat pork]

(28) The alligator wants [him_i to eat pork himself_i]

(29) The alligator_i wants [to eat pork itself_i]

(30) *The alligator_i wants [him_j to eat pork itself_i]

There are a couple of things that we need to pay attention to in the examples above. The first of these is the case marking in the exceptional clause which is shown overtly on the pronoun. The other thing we need to look at are the binding relations of the elements in the utterance. From (27) to (30 pronouns are used to show case marking. In the examples we find that if there is an overt subject in the examples this subject is case marked for the accusative case. How is the subject being marked for case? It can`t be marked for case by the verb because we said in class that, due to a case filter, it is argued that a predicate can only assign case once and in one direction. For verbs and prepositions in English this is to the right presumably following the head parameter value for English. Interesting is how the subject John is marked for case. It is obviously not being marked for case by the IP. If it were it would be in the nominative case, but this is clearly not acceptable.

(31) *The alligator wants [he to eat pork]

Thus, in such constructions we make two claims based on the evidence. The first of these is that there cannot be a case marking category in the exceptional clause. The second claim is that the DP him must be case marked form the verb in the preceding clause. This is due to the fact that it is in accusative case which is exactly the case we would expect the verb to assign as shown in the following regular clauses.

(32) John loves Mary

(33) He loves her.

(34) *Him loves she.

So, the verb in the preceding clause marks the subject for case which means that there really must not be a case assigning category in the exceptional clause. We also know that IP is a case assigner as it is the element that has been claimed to assign nominative case to subject DPs which move into its spec. position for the sole point of receiving the case marking it cannot get from the verb. To claim then that exceptional clauses are headed by IP is a problem. We must therefore break IP into two different categories, one which is a case assigner (AGRP) and one which is not a case assigner (TP). So, in the exceptional clauses shown in (25)-(30) above the exceptional clause is headed not by CP or AGRP but TP, which is not a case assigner. Again, as the argument goes if these clauses have TP then there must be TP and AGRP in all regular clauses. Thus from this moment on we will draw our trees as being headed by AGRP with TP being a complement of AGRP and VP being a complement of TP. Subjects as well will originate in the spec position of VP but will have to bounce from spec VP position to spec TP to spec AGRP in its search for case marking.

Now before moving on we will quickly have to look at the binding relations in the examples (25)-(30) above repeated here for ease of studying.

(25) The alligator wants [to eat pork]

(26) The alligator wants [John to eat pork]

(27) The alligator wants [him to eat pork]

(28) The alligator wants [him_i to eat pork himself_i]

(29) The alligator_i wants [to eat pork itself_i]

(30) *The alligator_i wants [him_j to eat pork itself_i]

An interesting phenomenon in (29) is the fact that we can co-index alligator and itself across a clausal boundary. This is in violation of the regular constraint on anaphors only being bound within their local domain as shown in (35) below)

(35) John_i loves himself_i to much.

Once more, we see that when there is an overt subject in the exceptional clause, this subject blocks alligator from being co-indexed with itself as in (30). Based on such examples it seems that there must be some element in the exceptional clause which is acting as a bridge carrying over the features of the subject in the preceding clause. Linguists have posited just such a category and they call it `big pro` or more simply PRO. PRO is an example of a covert category. Covert categories like English`s PRO and the `little pro`, pro, of Korean subjectless sentences are there but they never have a form which is actually spoken in the language. From this we can see that PRO has no overt phonological or orthographic form but it must be there so that it can carry the features from the preceding subject. This also means that PRO cannot be case marked. If PRO were case marked it would not be able to hold the features of the subject from the preceding clause because it would have its own features.

CP

CP or Complementizer Phrase is another type of functional category. Linguists claim that a wide variety of different types of words are able to fill the CP node. As shown in the examples below.

(36) If you are stupid, try it.

(37) That you are stupid will make you try it.

(38) Since you are stupid you will certainly try it.

(39) Whether you really are stupid will be determined shortly.

(36a) Try it if you are stupid.

(37a) *It will make you try it that you are stupid.

(38a) You will certainly try it since you are stupid.

(39a) It will be determined shortly whether you really are stupid.

From the above examples we can see that complementizers come in different shapes but they all behave in similar ways. We can also see that though people tend to think of CPs as heading complement clauses or dependent clauses they can also come in the first clause. (37) provides an interesting example where the CP heading the dependent clause simply doesn`t work. Based on such examples and the idea of structural symmetry linguists have posited that all clauses then begin with CP (with the exception of course of the exceptional clauses we discussed above which have neither CP nor AGRP, but rather TP).

The idea of structural symmetry is an important idea in syntax in that it claims that all structures are essentially identical in their underlying structure regardless of what may appear on the surface. Structures have symmetry or congruity in that if a certain category can be shown to be there sometimes it must always be there. This keeps the idea of structure simple in that structures are assumed to be the same or at least very similar. Thus, CP, for example, is always there in the structure but is often simply empty.

DP

DP or Determiner Phrase works in similar ways as the already identified functional categories of AGRP, TP and CP in that it is often empty, but this doesn`t mean it is not there. Thus, based on the idea of structural symmetry we can claim that there is a category called DP and that all NPs are complements of DP. This is a recently proposed alternative to the old NP analysis which claimed that determiners were specifiers of NP. The DP analysis is seen as being better because it not only brings NPs into the head parameter constraint of right branching for English but it is able to account for many different kinds of noun markers.

(40) John`s friend

(40a) dear friend

(40b) my friend

(41) some water

(42) standing chicken

(43) the armadillo

All of the examples above show DPs but with different kids of determiners heading the nouns. A further advantage of the DP approach is that finally a distinction between pronouns and nouns can be clearly shown. Pronouns are claimed to be DP with empty NP complements. Easy!

Intro to Ling, pp. 304-337

1. What are empty categories and why do we need them?

The bottom line is that is that in a clause most elements (subjects [sometimes] and predicates are the exception) are optional to a degree. This is a problem for a theory of syntax which is based on a constraining underlying system. From a theoretical point of view, we don`t want all utterances to be completely idiosyncratic. They need to be more similar than they are different. For this reason we need to come up with a theory which allows us to develop simple, all encompassing models. This is achieved first by claiming that there are two different levels of syntactic structure, the deep structure level being the same for all utterances that might be very different on the surface due to movement. The other main way or simplifying syntax is the impose an idea of structural symmetry by which it is claimed that structures need to be as similar as possible. This is made possible only through the idea of empty categories. Because there are empty categories all structures will look the same even if some of the categories are empty.

(1) The crocodile likes chicken.

(2) The crocodile did like chicken.

(3) If the crocodile liked chicken, it would be happier.

(4) What if the crocodile liked chicken?

(5) What if the crocodile did like chicken?

(6) The crocodile liked porridge and its mother chicken.

(7) The crocodile liked chicken, didn`t it?

All of the utterances above all come from the same deep structure and all of them contain some but different empty categories in their surface strcuture.

2. What are some of the main arguments for the IP analysis and is it possible to equate these with our TP and AP analysis?

Tag questions (Bob beat Bill, didn`t he?)

Clitics (She had eaten cabbage last night. She`d eaten cabbage last night.)

          (She had a doll as a child. *She`d a doll as a child.)

3. What is PRO an how does it work?

PRO (Big pro) is an empty category proposed to occupy the subject position in what are termed exceptional clauses.

(1) The elephant_i tried [PRO_i to whip the farmer with its_i tail].

(2) The elephant_i tried [PRO_i to kill itself_i].

(3) *The elephant_i wanted [John_j to kill himself_j].

(4) *The elephant_i tried [PRO_i to kill themselves_i].

We can see that we are able o co-index the elephant with the anaphor in the next clause. This means that something must be in the clause with is helping to carry the features of the elephant over on to the next clause. This has been the main argument that PRO is there. Not only do all clauses in English require subjects and PRO helps to satisfy this requirement but something also seems to be there which seems to act like a regular NP in terms of feature holding. PRO holds the features of the preceding NP thus allowing us to co-index the distant DP elephant with its reflexive anaphor itself.

4. How do DPs work and do we really need them?

DPs allow us to retain the idea of structural symmetry through empty categories. The basic idea is that all NPs are complements of DPs even if the DP is empty. The DP analysis also makes sense when we argue the status of pronouns as DPs with empty NP complements.

5.Why does movement occur? (What causes movement?)

Movement usually occurs as a result of elements needing to move to satisfy certain conditions. For example, wh-heads and dummy verbs move into CP to satisfy the interrogative condition. NPs move from Spec VP position to Spec IP position to be case marked.

6.What is the difference between head movement and operator movement?

Head movement involves moving something from one head position to another as in the case of verb raising.

Operator movement involves moving an operator expression (wh-word, negative) from one type of position to another as in the case of Wh-movement where wh-words are moved from VP comp position to spec CP position, Wh-movement is also called non-A-movement as the Wh-word is being moved form an A (argument) position to a non-A-position.

(1) Who did you meet?

     I met Louise.

     You met who?

7. What are some other types of movement?

Passivization

Adjunction

Let`s not worry too much about these now.

Section 20 - Empty categories

Getting back to the idea of INFL or I, it is easy to see that we have to change our ideas of categories. There are many utterances where there is no infinitive or auxiliary.

(1) The elephant tripped down the steps.

Because one of the main ideas in syntax is simplicity and regularity we want to argue that all clauses necessarily have the same structure. It would not be productive to assume that such sentences as the one above are headed by the VP since there is no overt I. For this reason we say that there is an I, but the I is simply empty. The reasons for arguing this is not just to preserve the simplicity of our model. There is syntactic support for this.

One of the arguments in favor of this comes from cliticization.

(2) The armadillo had run away. = The armadillo`d run away.

(3) The armadillo had no real friends. = *The armadillo`d no real friends.

The fact that we can make clitics out of helping verbs, but no of main verbs clearly shows that there must be an empty I. The argument is that the I even though it is empty, must be there because it blocks the cliticization.

More evidence for empty I comes from tag questions.

(4) You haven`t eaten armadillo, have you?

(5) You can squeeze a porcupine, can`t you?

In the examples above we have utterances composed of two clauses. In the second clause there is no main verb, there is only an I. This clearly shows us that I must always be there. Even if we analyze the latter clause with the Is jumping to the V slot, then the I must be empty. Here is another example using tags.

(6) Boris killed the porcupine, didn`t he?

(7) Wendy loves armadillos, doesn`t she?

These examples are even clearer in that there is no overt I in the first clause, but there is an overt I in the second clause. In fact there is no V in the second clause. We can only assume that the I is there in the first clause and that it is carried over into the second one.

Based on this evidence, we can now argue that I must sometimes be empty (sometimes overt, sometimes covert). Now since our goal in linguistics is to simplify and systematize language, we can assume that I is not the only empty category. It seems that NPs can also be empty. Look at the examples below.

(8) The porcupine wants [to trip the elephant].

(9) The porcupine wants [the armadillo to trip the elephant].

There are good reasons for believing that the two sentences above have the same structure. The first reason is that we claimed earlier that all clauses necessarily require only two parts; subject and predicate. Well, the first sentence above has no overt subject. This causes problems with our basic idea of clause structure. The other reason is that there is no reason to assume that the two utterances should have a different structure. In fact, it goes with the basic tenets of syntactic theory that they have to have the same structure. We will therefore argue that the second clause in the examples above all have an NP subject but in one the NP is overt and in the other convert.

(10) The porcupine wants [PRO to trip the elephant]

Likewise, there is proof that this must be the case. Look at reflexives. We read in Radford et al. that reflexives have a severely limited domain of operation. Both the referent and the reflexive must occur in the same clause.

(11) The lemming wants [John to kill himself].

(12) *The lemming wants [John to kill itself].

Here the first utterance is acceptable because John and himself are contained within the same clause and their features match. The second utterance is unacceptable because the lemming and the intended reflexive itself are separated by a clause barrier, but what about the utterance below.

(13) The lemming wanted [PRO to kill itself].

This is acceptable despite the fact that lemming and itself are in different clauses. We assume that the empty NP PRO is there in the subject position of the clause and that it contains the features of the NP lemming. If there were no convert subject then the utterance would have to be unacceptable. There is, of course, more proof of this but for the sake of brevity we will leave it at that.

Now taking this one step further was can posit that if I and subject NPs can be covert, then anything should, according to the principles of syntax, be able to be convert. As it turns out this is the case. Take a look at the exchanges below.

(14) A: Do you like armadillos?

      B: I do (like armadillos).

(15) A: Can you eat raw meat?

      B: I can (eat raw meat).

According to the exchanges above it is possible to render entire complements covert. The fact that we can answer with or without the complement, depending on the situation, clearly shows that complements, too, are either convert or overt.

Notice that this is different than what we said earlier about adjuncts being optional. If something is optional it is either there or not. If something is overt or covert it is always there in the structure but the branches simply aren`t filled all the time.

The last thing we want to look at in this section is the status of noun phrases. It seems obvious that determiners, as well, are either covert or overt. Look at the examples below.

(16) The porcupines loved the worms.

(17) Porcupines love worms.

Again, it seems logical to assume that the two utterances above have the same structure and that the determiners are either covert or overt, but look at this.

(18) We armadillos hate to hug you porcupines.

This utterance has the same structure as the two above. What we have here is personal pronouns functioning as determiners. Since one of the main goals of syntax is to try to limit the number of categories and personal pronouns and articles (not to mention many other types of word types) behave the same way syntactically, it seems to make sense to argue that they are of the same syntactic category. Once we have done this we can take our argument one step further.

Personal pronouns, and pronouns in general, are often able to function by themselves.

(19) Give it to me.

Based on this we have to reevaluate how we look at NPs. Before we imagined that nouns were the heads of phrases like `the man`. This was quite different then the structure of the other phrases we saw. If we can say that all elements are potentially covert then we can reanalyze what we thought of as NPs as Determiner Phrases. Now, the determiner is the head, as evidenced in the sentence directly above and the noun is functioning as a complement of the determiner.

Section 21- Movement

       Now that we have clearly shown that there are overt and covert phrases in utterances. Let`s turn our attention to movement. The idea that elements in utterances move is not a new idea, but it was never formulated as eloquently as by Chomsky. In the spirit of all linguistic operations being basically the same, we can say that syntax, like phonology and morphology operates on at least two levels, the surface and underlying levels. We move things from the underlying level to form surface representations.

Think about it this way. How do we form questions in English? Look at the examples below.

(1) The porcupine likes me a lot.

(2) Does the porcupine like you?

The I lexeme seems to jump in front of the determiner phrase. This in itself is not a real problem. The problem comes from the question, where does it go to? To find the answer to this we have to look at some other examples.

(3) The bear wondered [if alligator tasted good].

(4) The landlord demanded [that I pay the rent on time].

(5) The landlord demanded [E I pay the rent on time].

In the examples above, the second clauses all have something in front of the determiner phrase whether it is overt or convert. In traditional grammar these were called complementizers. We will keep the same name here. What these constructions show us is that there must be something in the clause above the I level, for we have to have somewhere to put these words. From this point on we will no longer assume that I is the head of the utterance. Instead, C (for Complemetizer) is the head and the highest point on the tree. Lexemes like `if` and `that` in the examples above are complementizers and are the heads of the clause.

Comp or CP is a functional category which stands for Complementizer Phrase. This category was arrived at based on the data from such sentences as, `That he was hungry was a miracle.` and he wondered whether he could eat pork anymore. This category appears above the IP and IP is a complement of CP. Once we start to look at movement rules we will have a better idea of the full range of roles that CP plays in language.

Going back to our question forms, we now have a place to put the I when it moves. This type of movement is called head movement because we are moving something from one head position to another. In question formation the lexeme moves from the head position of I to the head position of C. Here are a few more examples.

(6) Did the armadillo fart?

(7) Will the bear have had it`s breakfast?

(8) Has the elephant had a bath?

Now, we assume that when something moves it leaves what we call a trace. This trace is like a mark which shows that something was there. It also works to prevent something else from moving into the position.

(9) *Will the bear`ve had it`s breakfast?

In this example the trace acts as a block for the verb `have` so that it cannot form a clitic on the noun. Now, what about some other question types?

Trace (t) is the residue that is left behind when an element moves from one position to another. We need trace in our theory of syntax because it seems true that when things move they leave something behind. For example, in the sentence, `what did Betty eat?`, we assume tat the NP `what` originated in the VP Complement position. After `what` is moved a trace occupies that position and it is not possible to insert anything into the position occupied by the trace.

(10) *`What did Betty eat the armadillo?`

This sentence is unacceptable because the trace is blocking anything from being able to take its position.

(11) How good did the armadillo taste to the bear?

What do we do about so-called wh-questions? Well, the answer is easy. We decided earlier that nouns are part of determiner phrases. To form wh-questions in English we simply move the determiner phrase in questions to the specifier position of the C. Questions are, however, not the only utterance types that have this structure.

(12) When did you get here?

(13) *Nobody can I beat at golf.

The big question that remains is where do these phrases come from if they do no originate in the C. The best way of figuring this out is by using echo questions.

(14) The armadillo tasted how good to the bear?

(15) You got here when?

(16) You can beat nobody at golf?

It seems that the DPs originate in the operator position and are moves to the specifier position of C. This is called operator movement.

Another type of movement is found in passive constructions.

(17) The porcupine ate bondaegi.

(18) Bondaegi were eaten.

Here we see the DP bondeagi is moved to ??? Look at the examples below to figure it out.

(19) How were the bondeagi eaten?

(20) *How was the porcupine the bondaegi were eaten?

Based on these examples, we can clearly see that the DP bondeagi must be located in the specifier position of the I, as the C and its specifier position can be occupied.

We will not go to deep into adjunction as a movement operation except to say that central to it is the idea of recursivity. In adjunction, we are copying categories and recreating them.

Cog Ling. Unit 10: Causation and Agency, pp. 157-169.

8. What is causation and how does it work in English?

     The idea of causation is closely related to the idea of AGENT. In a prototypical sense it is the causer who/which is the agent of a certain action. Again, prototypically, an AGENT is usually an animate object and more specifically a human being.

(1) Sweet, little Bambi died. (cause unknown or deemed unimportant)

(2) The stray bullet killed Bambi.

(3) The bullet fired from Bob`s gun killed Bambi.

(4) ?Bob`s gun Killed Bambi.

(5) *Bob`s gun murdered Bambi.

(6) Bob killed Bambi with a gun.

(7) ?Bob killed Bambi with a bullet.

(8) Bob killed Bambi with a bullet from his gun.

(9) Bob murdered Bambi.

(10) ?Guns don`t murder people. ..... People murder people.

9. How does causation work differently in Japanese?

     Basically causation or the idea of AGENT in Japanese works the same way as in English. Where they differ is how they are able to move away from the prototypical scenario of human agents and their causes. How each language, Japanese as a example, is able to extend from the prototype is a reflection of its culture. Not only can we establish chains of causation which would be deemed quite indirect being marked and interpreted as being direct in Japanese (the case of mothers and their responsibility to their children), but actual outcomes can be weaker as well. The same phenomenon are also prevalent in Korean. I have heard lower level Korean speakers of English say things like, ̋I found the phone but I couldn`t find it.̋ While this perplexed me at the time, there does indeed seem to be a clear explanation for such phenomenon in the cognitive model.

10. How does the cognitive treatment of causation help to further advance the cognitive model of language?

     The rather flexible way in which the cognitive model is able to treat causatives by incorporating different systems and types of information is a great shot in the arm for the model itself. Once again, the simple fact that the cognitive model is able to relatively easily account for linguistic phenomena that the generative model has trouble with is a shot in the arm for the model. It does seem that syntactic phenomenon like causation can indeed only be interpreted and explained through the world as some mediating organizer.