Transformational Rules PS-rules and lexicon together generate deep structure. But there are some problems in our phrase-structure rules: 1. S does not have its head. 2. There is no way to specify 'tense' aspect in the deep structure with given set of PS-rules. 3. What do we do with auxiliary verbs as in the sentence like 'John could have been running'? Considering the problems, we revised our PS-rules as follows: a. S NP Aux VP b. Aux T (M) (have -en) (be -ing) c. T {present, past} *There is no 'future' tense in syntax. We are talking about structure, not meaning. The sentence "He would go" can be talking about a possible future event, but syntactically, that is, structurally, the tense of the sentence is past. Based on these PS-rules, we will see 5 transformational rules which will produce, from the deep structures, yes/no question, negation, and wh-question sentences. Most transformational rules have two parts: Structural Description which describes what conditions should be met for the rule to apply, and Structural Change which describes how they change. Unless the criteria described in SD are not satisfied, the rule cannot apply. First rule we will see is called Affix Hopping rule. In b above, we see T, -en, -ing which cannot be realized unless they are attached to some other elements. We will treat them as affixes, as in morphology. Now that we have a new set of PS-rules, the deep structure of every sentence will be different from its surface structure because it has T, -en or -ing in it. For instance, the deep structure of the sentence "He would go" would be now "He past will go." The link between these deep structure and surface structure is a transformational rule called Affix Hopping. (We will see later that these are not invented arbitrarily.) 1. Affix Hopping (obligatory) S.D. X {T, -en, -ing} {M, have, be, V} Y S.C. 1 0 3+2 4 This rule could be roughly translated as follows: If you see any one of T, -en, or -ing, (2) and any one of M, have, be, or a verb (V) (3)right after it, then you will attach that (2) after the (3). In this rule, X and Y are variables which can be anything including nothing. This rule is the most basic one, since it has to be applied to every deep structure. Therefore it is an obligatory rule, which means we have to apply this rule to every structure with affix in it. Or at least we have to try to apply the rule. (Sometimes we will see that it will be impossible to apply this rule because the structural description is not met. But we will discuss it later.) This rule is almost the last rule to apply. (Why almost? We will talk about it later.) Let's look at some examples. We will start with a simple one: This becomes interesting. What is the deep structure of this sentence? Remember the deep structure is made by PS-rules and lexicon. So anything which is not in PS-rules cannot appear in the deep structure. (Any other structure which cannot be produced by PS-rules and lexicon should be generated by some 1
transformational rules.) D.S. this present become interesting Now can we apply Affix Hopping rule to this deep structure? To answer that, we have to check whether we can identify each element in the SD in the rule. D.S. this present become interesting It seems this deep structure satisfies the SD in the Affix Hopping rule. Therefore we can apply the rule and as a result we will get: this 0 become+present interesting This is the surface structure. Actually it is still different from what we are saying. Syntax stops here. Then our morphological and phonological knowledge (PF) kicks in, providing the right form becomes'for 'present form of the verb become when the subject is third person singular' and logical form (LF) will give the sentence its meaning. Let's try another, a little more complicated example. This is getting interesting. What is the deep structure of this sentence should be? What is the tense of this sentence? Since we have is, it should be present. In addition we have a progressive. That means we should include that present progressive in the deep structure as a part of Aux. And is is in this sentence a verb or a part of be -ing? Since we have a progressive, we know it is a part of progressive. And the real verb of the sentence is get. D.S. this present be -ing get interesting Now, let's see whether we can apply the rule here: D.S. this present be -ing get interesting this 0 be+present -ing get interesting This way we took care of one affix, i.e. T. But we still have another affix which hasn't been attached to anything else. Since Affix Hopping rule is obligatory, we have to try to apply it whenever we find an affix. Can we apply it again? this be+present -ing get interesting this be+present get+ing interesting This is our final surface structure. Then, again, right forms of each verb will be supplied and as a result, we have "this is getting interesting." In one declarative sentence, this Affix Hopping rule can be applied upto 3 times. The example should be: "This should have been getting interesting." Try this for yourself. 2. NP-Aux Inversion (optional) Now, let's look at the so-called yes/no questions. Tom can go. Can Tom go? How would you make a transformational rule which will change the declarative sentences into questions? 1 st version NP-Aux Inversion 2
S.D. NP Aux X But what about "She is coming"? If we apply the above rule to the following deep structure, we would get the wrong sentence. DS she present be-ing come present be-ing she come Since we have affixes, we have to apply Affix-Hopping. The result would be: be+present -ing she come Then, we are left with stranded -ing, because Affix Hopping cannot apply to the above structure. It seems this data suggest that we should treat -ing part separately from be. And also we know that the tense has to have something to be attached after moving its position. Thus we get the following rule: 2 nd version of NP-Aux Inversion S.D. NP T{M, have, be} X Still, we have a problem. This rule implies that a part of Aux which should be moved must have tense and any one of M, have,or be but nothing else. It does not allow us to make any yes/no question from the deep structure which does not have any one of them. Consider a sentence like "I like him." The DS would be "I present like him." Can we apply the above rule to make a question? No, because it does not have any M have,or be. But since it should be the same process, we should be able to apply the same rule to this kind of structure as well. Taking this into consideration, we have the final version of our NP-Aux Inversion rule. NP-Aux Inversion Rule (optional) S.D. NP T({M, have, be}) X In this final version, tense is the only absolutely necessary part in the Aux to be moved. This rule is optional because we can apply this rule only if we want to make a question. If we don't apply this rule, the outcome will be a declarative sentence. If we apply this rule to the sentence above, we will get the following result: present I like him This is not a grammatical sentence, either. What should we do? We will get back to this question in a moment. But first, let's finish the discussion on question-making transformations. We call the above rule, not yes-no question rule, but NP-Aux Inversion rule, because the exact same rule is also used to make questions other than yes-no questions. Let's consider the following wh-question: What is this? Whenever you see wh-word, assume that it is in the deep structure. Then the DS of the above would be: 3
D.S. this present be what NP-Aux present be this what A.H 2 3 4 be+present this what Then we will get a sentence like "is this what." Here we need Wh-Movement rule. 3. Wh-Movement (obligatory) S.D. X Wh Y Basically, this rule says move wh-word into the beginning of a sentence whenever you find one. This rule is obligatory: that is, if you have a wh-word in the structure, you have to apply it. But this rule should not precede NP-Aux Inversion. Wh-questions and yes-no questions share the basic deep structure and the same transformational rule, i.e. NP-Aux Inversion. The only difference is in wh-questions, we apply one additional rule, which is wh-movement. Now let's get back to the question I raised at the end of the discussion on NP-Aux Inversion rule. What do we do with a structure like 'present I like him' in which Tense is stranded without any possibility of attaching itself to anything? To save this kind of "stranded" affixes, we have the following tranformation. 4. Do Support (obligatory) An occurrence of T that has not been able to undergo Affix Hopping must have a "do" inserted to the left of it. What does this mean? The best way would be go through an example. Let's derive the sentence "did he eat an apple?" The DS would be: D.S. he past eat an apple If we want to make a simple declarative sentence, all we have to do is to apply Affix Hopping rule. But since we want to make a question out of this, we will apply NP-Aux Inversion rule. D.S. he past eat an apple past he eat an apple Now since we have an affix (T in this case), we have to apply Affix Hopping rule. But we can't. Why not? Because Tense is separated from any possible element to which it can be attached. Tense can never be realized unless it is attached to some other forms because it is, let's say, a kind of bound morpheme. But here it doesn't have anything. Therefore we have to supply a "dummy" verb "do" so that the tense, which is a very important element in a sentence, can be realized. Our Do Support rule is doing exactly this. You have to careful because the rule says you insert "do" to the left of Tense. It is important. Let's apply the rule and see why. Do Support do+past he eat an apple Since we inserted "do" to the left of Tense, it will be automatically realized as "does" without applying Affix Hopping rule again. We cannot apply Affix Hopping after Do Support. We know we need a Do Support rule only after we tried to apply Affix Hopping. Once we applied Affix Hopping we cannot apply it again. Of course we may apply Affix Hopping rule upto three times, in a row. But we cannot apply Affix Hopping rule after we apply other transformational rules as 4
the result of not being able to apply AH at the first place. There is an interesting and important relationship between Affix Hopping and Do Support. They are kind of complementary. If one can be applied, the other never can. If you can successfully apply AH, there is no chance for Do Support to be applied. Therefore, in terms of ordering, Do Support is the last rule we apply, but only when AH fails. That's why I said AH is almost the last rule to apply. We try AH the last, and if it succeeds, that is it. But if it fails, then we need to apply Do Support. But how do we know that syntacticians just come up with this kind of ad-hoc explanation? Let's look at the negation process. How do we make negations? It would be a good exercise for you to go through some negative sentences and try to come up with the appropriate transformational rule. Then you will find the following rule: 5. Not Insertion (optional) S.D. NP T({M, have, be}) X S.C. 1 2 not 3 It happens that this rule is very similar to NP-Aux Inversion rule. We assume that negative sentences are all generated from the same deep structures as their positive counterparts. Let's look at an example. He may not have gone. D.S. He present may have -en go Not ins. he present may not have -en go A.H(1) he may+pres not have -en go A.H(2) he may+pres not have go+-en So we have "he may not have gone" which is grammatical. But what about "He did not go"? D.S. he past go Not ins. he past not go A.H Not Applicable When we tried to apply A.H., it didn't work because the tense is separated from the verb "go," which is a possible candidate for it to be attached to, by "not." So what do we do when Tense is stranded? Right. We apply Do Support. Not ins. he past not go A.H Not Applicable Do-Sup he do+past not go S.S. he did not go. So far we have talked about 5 transformational rules. I was not trying to cover all the rules in English, by any means. The important point is that seeming different types of sentences are generated by applying several transformational rules to the same deep structure. 5
Review Given the following PS-rules, S NP Aux VP Aux T (M) (have -en) (be -ing) NP (Det) N (PP) VP V (NP) (PP) PP P NP Give the deep structure for the following sentences, and derive the surface structure by applying appropriate transformational rules. ex) Tom is going. S NP Aux VP N T V Deep Structure : Tom present be -ing go AH: Tom be+present -ing go AH: Tom be+present go+ing Surface Structure: Tom is going 1. Tom could sing. 2. Tom has sung. 3. Tom might have been singing. 4. Will John go? 5. Should the cousin of the king have been counting the money? 6. Did Jack arrive at the party? 7. What did he do? 8. We did not take advantage of him. 6