INTRODUCTION
In the previous unit, we considered two models of grammar proposed by Chomsky: the finite state grammar and the phrase structure grammar. From our discussions, it is immediately obvious that it would be impossible to account for all the sentences of a language in terms of the rules of any of the grammars. For example, the rules of phoneme succession cannot account for sentences, and it was also shown in the previous unit that a left-to-right generation, unit by unit, in terms of morphemes or words also cannot account for all the sentences of a language.
.jpeg)
There are processes, such as grammatical conjunction, which are not satisfactorily described in terms of phrase structure (John and Peter is neither John/Peter nor John and/Peter). There are cases, like the one of the active-passive relation (which exists between John admires sincerity and sincerity is admired by John) which, in spite of being grammatical rather than semantic relations, cannot be expressed by a phrase structure grammar. Difficulties of this kind are eliminated if the phrase structure grammar is limited to the sentence constituting the kernel of the language: a set of simple, declarative, active sentences (in fact, probably a finite number of these), while all other transforms of the kernel sentences are derived from these through the application of transformational rules. In fact, transformations do not operate on actual sentences but on more abstract structures, or strings of symbols underlying the sentences.
We shall discuss the transformational grammars propounded in Syntactic Structures (1957) and Aspects of the Theory of Syntax.
‘Syntactic Structures’ Model of Transformational Grammar
The first point that must be made is terminological. Whereas a phrase structure grammar consists solely of phrase structure rules, a transformational grammar (as originally conceived by Chomsky) does not consist only of transformational rules. It includes a set of phrase structure rules as well. The transformational rules depend upon the previous application of the phrase structure rules and have the effect, not only of converting one string of elements into another, but, in principle, of changing the associated phrase marker. Furthermore, they are formally more heterogeneous and more complex than phrase structure rules. We shall go on to give some examples of transformational rules, but we first need to introduce an appropriate set of rules. We will use those given by Chomsky in Syntactic Structures (p. 111) – with one or two minor changes – as follows:
1. Sentence: NP + VP
2. VP: verb + NP
3. NP: NPsing NPpl
4. NPpl: D + N + s
5. T: the
6. N: {man, ball, door, dog, book,…}
7. Verb: Aux + V
8. V: {hit, take, bite, eat, walk, open, …}
9. Aux: Tense (+N) (+ have + en) (+ be + ing)
10. Tense: present
11. N: {will, can, may, shall, must}
It will be observed that this set of rules allows for a wider range of choices. Both singular and plural noun phrases are accounted for by rule (3), and a large number of tenses and moods are introduced (instead of just the simple past tense of, for example, The man hit the ball) by means of the element Aux and its subsequent development. Rule (10) implies that every string generated by it must contain the element Tense and may contain, in addition, one or more of the other strings elements in brackets. (Elements like s in rule (5), and en or ing in rule (10) are morphemes rather than words. In fact, have, be, the, and all the elements listed on the right hand side of rules (7), (9), and (12) may also be regarded as morphemes. But we need not dwell here upon the difference between a ‘word’ and a ‘morpheme’).
Assuming that the lists given in rules (7) and (9) are considerably extended, this system of phrase structure rules will generate a large (but finite) number of what we may call underlying strings. It should be emphasized that an underlying string is not a sentence. The transformational rules have yet to be applied. One of the strings generated by these rules is:
the + man + present + may + have + en + open + the + door
(which, given the transformational rules of Syntactic Structures, underlies both the active sentence: The man may have opened the door and the corresponding passive: The door may have been opened by the man). Try to verify that this string is indeed generated by the rules and construct the associated phrase marker.
Chomsky derived passive sentences from underlying strings in Syntactic Structures by means of an optional rule which we may give, rather informally, as follows:
13. NP1 + Aux + V + NP2 → NP2 + Aux + be + en + V + by + NP1
This rule differs in various represents from the phrase structure rules. Not just one element, but a string of four elements, appears to the left of the arrow; and the operation that is carried out by the rule is quite complex – involving the permutation of the two NP’s (this is indicated by the subscripts) and the insertion of the elements be, en, and by at particular points.
There is, however, an even more important difference between phrase structure rules (1)-(2) and the transformational rule (13); and this has to do with the way in which we interpret the symbols which occur in the rules. In a phrase structure rule, a single symbol designates one and only one element in the string to which the rule applies. But in a transformational rule, a single symbol may refer to a string of more than one element, provided that the string in question is dominated by (i.e., derived from) this symbol in the associated phrase marker. It is in this sense that transformational rules are said to operate upon phrase markers rather than simply upon strings of elements.
We will first of all illustrate what is meant by this statement with reference to a purely abstract example. Given that the string a + d + e + b + f + c + g + h has been generated by a set of phrase structure rules which assign to it the phrase marker illustrated in Fig. 3 below (the reader can easily reconstruct these rules for himself), this string will be converted by means of the transformational rule B + D + E → E + B into the string c + g + h + a + d + e with the associated phrase marker shown in Fig. 4 below:
E B
c h a d e
In other words, since the string of terminal symbols itself constitutes part of the phrase marker, we can say that the rule converts one phrase marker into another; and this is the defining property of transformational rules. The rule we have just given has the effect of deleting everything dominated by D (including D itself) and permuting B and E, keeping their internal structure intact. The phrase marker given in Fig. 3 and the phrase marker given in Fig. 4 may be described, respectively, as underlying and derived, with respect to the transformation in question.
If we now look at our illustrative underlying string (the + man + present + may + have + en + open + the + door) and at the associated phrase marker (which I have left the reader to reconstruct for himself), we shall see that the + man is wholly dominated by NP, Present + may + have + en by Aux, V by V (this is an instance of ‘self domination’), and the + door by NP. This means that the transformational rule (13) is applicable and, if applied (for it is an optional rule), will convert the underlying string into 13(a) with the appropriate derived phrase marker:
13. (a) the + door + present + may + have + en + be + en + open + by + the + man
But what is the appropriate derived phrase marker? This is a difficult question. Granted that NP2 becomes the subject of the passive sentence, that be + en becomes part of Aux in the same way that have + en or may is (this, as we shall see, is necessary for the operation of subsequent rules), and that by is attached to NP2 to form a phrase, there are still a number of points about the structure of the derived phrase marker that remain unclear. Two possible phrase markers are given in Figures 5 and 6. It will be observed that they differ in that one takes by + NP1 to be a part of the verb phrase whereas the other treats it as an immediate constituent of the sentence, equivalent in ‘status’, as it were, to NP2 and VP. (It will also be noticed that I have put a question mark where the label for the bracketed phrase by + NP1 should be).
We have touched here on an important theoretical problem. The derived string produced by one transformational rule may serve as the underlying string for the operation of a subsequent transformational rule and will therefore need to have associated with it the appropriate derived phrase marker. Chomsky and his followers have worked on this problem and have tried to establish a set of conventions according to which a particular kind of formal operation (e.g., deletion, permutation, or substitution) is defined to have a particular effect upon the typology of the phrase marker it transforms; and we have followed these conventions when we decided that the effect of rule B + D + E → E + B operating upon the underlying phrase marker shown in Fig. 3 was the derived phrase marker in Fig. 4. But this was a very simple example from the point of view of the operations involved and the shape of the phrase marker to which they applied; and furthermore, it was a purely abstract example unaffected by any empirical considerations.
For want of space, we will introduce and briefly discuss a transformational rule in Syntactic Structures. This is the obligatory ‘number transformation’.
[elsewhere]
This is a context-sensitive rule, which says that present is to be rewritten as s if and only if it is immediately preceded in the underlying string by a sequence of one or more elements dominated by NPsing in the associated phrase marker but is to be rewritten in all other contexts as ‘zero’ (i.e., as the absence of a suffix). It is this rule which accounts for the ‘agreement’ between subject and verb manifest in such sentences as:
The man goes vs. *The man go or The man is tall vs. *The man are tall.
If it applied to 13(a), it yields:
14. (a) the + door + s + may + have + en + be + en + open + by + the + man
It will be observed that what we might call the ‘abstract’ verbal suffix s is here introduced in front of the element to which it is subsequently attached (in the same way that en and ing are introduced by phrase structure rule (10) in front of the element to which they are attached). We have called these ‘abstract’ suffixes because, as we shall see, they assume a variety of forms, including ‘zero’, in various contexts.
The rule by which these ‘abstract’ suffixes are placed after the appropriate stems (the ‘auxiliary transformation’) may be given as follows:
[ingbe be ing]
This rule says that any pair of elements the first of which is tense, en and ing and the second of which is M, have, be or V are to be (obligatorily) permuted, the rest of the string to the left and the rest of the string to the right remaining unchanged. If the rule is applied to 14(a), it will permute s + may (i.e., Tense + M), en + be and en + open (en + V), successively from left to right, yielding:
15. (a) the + door + may + s + have + be + en + open + en + by + the + man
One more transformational rule has yet to apply, which puts a word-boundary symbol (we shall use a space) between every pair of elements, the first of which is not M, have, be, or V and the second of which is not Tense, en, or ing. Applied to 15(a), this yields:
16. (a) the door may + s have be + en open + en by the man
And this is the form that our illustrative string would have after all the relevant transformational rules have operated.
Finally, in a grammar of the kind outlined by Chomsky in Syntactic Structures, there is a set of ‘morphophonemic’ rules, which will convert the string of words and morphemes into a string of phonemes. These would rewrite may + s as the phonemic representation of what is spelled may, open + en as what is spelled opened (be + s as what is spelled is, run + en as what is spelled ran, and so on). We end up, therefore, as we should, with the phonemic representation of The door may have been opened by the man.
Most of you who are unfamiliar with Chomsky’s system of transformational grammar may have found it rather tedious following this step-by-step derivation of a single sentence. But you will now have acquired a sufficient understanding of the way the grammar is designed and operates for you to appreciate the significance of some of the more general points made in this unit and what will be discussed in later units. At this stage in our discussion of transformational grammar, it may be helpful to introduce a diagram showing how the grammar outlined in Syntactic Structures was organized (see Fig. 7). The input to the grammar is
Fig.7
the initial element which generates a set of underlying strings by means of the phrase structure rules in the first ‘box’ of the diagram. The second ‘box’ comprises the transformational rules, of which some are optional and others are obligatory. These rules take, as their ‘input’, single underlying strings or pairs of underlying strings, and by successively modifying these strings and their associated phrase markers, generate as their ‘output’ all and only the sentences of the language, represented as strings of words and morphemes, and assign to each sentence its derived constituent structure. The third ‘box’ of rules then converts each of these sentences from its syntactic representation (i) The man opens the door.
(ii) The man did not open the door.
(iii) Did the man open the door?
(iv) Didn't the man open the door?
(v) The door is opened by the man.
(vi) The door is not opened by the man.
(vii) Was the door opened by the man?
(viii) Wasn't the door opened by the man?
They differ in that: (i) has had no optional transformation applied to the underlying string; (ii) has had the Negative transformation applied; (iii) the Interrogative; (iv) the Negative and Interrogative; (v) the Passive; (vi) the Passive and Negative; (vii) the Passive and Interrogative; and (viii) the Passive, Negative, and Interrogative. Of these eight sentences, the first (a simple, active, declarative sentence) is defined by Chomsky in Syntactic Structures as a kernel sentence. It should be emphasized that non-kernel sentences, such as (ii)-(viii), are not derived from kernel sentences, such as (i), but from a common underlying string. That is to say, there are no sentences generated without the application of at least a small number of obligatory transformations, including rules comparable in effect with rules (14) and (15) above.
Compound sentences in which two clauses are coordinated (e.g., The man opened the door and switched on the light) are generated by means of conjoining and embedding transformations, respectively, which take as 'input' a pair of underlying strings (e.g., the + man + past + open + the + door and the + man + past + switch + on + the + light) and combine them in various ways. Conjoining and embedding transformations constitute the class of generalized transformations in Syntactic Structures; and it is the repeated application of these rules which accounts for the existence of such recursive structures as: This is the… that lived in the house that Jack built or… a big, black, three-foot long … wooden box (p.48). All the generalized transformations are, of course, optional.
So much, then, by way of a general summary of the earlier version of transformational grammar presented in Syntactic Structures. Chomsky claimed that one of the advantages of this system, the third and most powerful of his 'models for the description of language,' was that it could account more satisfactorily than phrase structure grammar for certain types of structural ambiguity. To take one of Chomsky's famous examples: a sentence like Flying planes can be dangerous is ambiguous (cf. To fly planes can be dangerous and Planes which are flying can be dangerous); and yet, under both interpretations, the immediate constituent analysis would generate both interpretations the same way. But it would be possible to generate a sentence like Flying planes can be dangerous within a phrase structure grammar and to assign to it two different phrase markers – differing with respect to the labels assigned to the node dominating flying. But this would not be an intuitively satisfying account of the ambiguity; and it would fail to relate the phrase flying planes, on the one hand, to planes which are flying, and, on the other, to someone flies planes. The transformational analysis accounts for the ambiguity by relating two different underlying strings to the same derived string. Many other examples could be given of structurally ambiguous sentences which can be accounted for rather nicely in terms of transformational grammar.
In a similar way, sentences with similar surface structures can sign different deep structures. Consider the following sentences:
(i) John is anxious to help
(ii) John is difficult to help
Here, we are dealing with sentences which have similar surface structure, but different deep structures. The first would have a structure like: John pres + be anxious for John to help
and the second a deep structure like:
For someone to help John pres + be difficult
For now, we shall not go into details as we shall go into the derivation of these sentences in subsequent units. Let us highlight the features of the transformational grammar propounded in Aspects (1965).
SELF-ASSESSMENT EXERCISE 1
1. Generate five sentences from the following rules:
(a) SNP + Aux + VP
(b) NP Det + N
(c) Aux Tense (N) + (have + en) (be + ing)
(d) VP V + NP
(e) N man, ball, etc
(f) Vbit, took, kick, etc.
2. Write out five sentences of your own and generate phrase structure rules for each.
3.2 'Aspects' Model of Transformational Grammar
In the last section, we presented a summary of Chomsky's transformational grammar as contained in Syntactic Structures (1957). In 1965, in Aspects of the Theory of Syntax, Chomsky put forward a more comprehensive theory of transformational grammar, which differed from the earlier theory in a number of important respects. For our purpose, it will be sufficient to mention only the most important differences between the Syntactic Structures grammar and what we may call an Aspect-type grammar. Once again, a diagram may be helpful (cf. Figure 8 below).
The most striking difference between the two grammars, as represented in Figures 7 (previous section) and 8, is the additional 'box' of rules in the Aspect-type grammar labeled 'Semantic Component'. In Syntactic Structures, it was argued that, although semantic considerations are not directly relevant to the syntactic description of sentences, there are 'striking correspondences between the structures and elements that are discovered in formal, grammatical analysis and specific semantic functions' (p.10) and that 'having determined the syntactic structure of the language, we can study the way in which this syntactic structure is put to use in the actual functioning of the language' (p.100). In the years that followed the publication of Syntactic Structures, Chomsky and his collaborators came to the conclusion that the meaning of sentences could, and should, be submitted to the same kind of precise, formal analysis as their syntactic structure, and that semantics should be included as an integral part of the grammatical analysis of languages. The discussion of sentences such as: Colourless green ideas sleep furiously pointed out a sense in which grammar is independent of meaning. The subsequent distinctions concerning degrees of grammaticalness point out the need for a formal specification of how these degrees are to be judged. There are sentences that we would call ungrammatical yet acceptable, and there are other sentences that are grammatical but unacceptable. It is where semantic and grammatical criteria overlap that this problem is raised. The grammar of a language is now seen by Chomsky as a system of rules relating the meaning (or meanings) of each sentence it generates to the physical manifestation of the sentence in the medium of sound.
In Aspects, as in Syntactic Structures, the syntax falls into two parts. But the two syntactic components operate somewhat differently. It is now the base of the grammar (which is roughly comparable with the phrase structure part of the earlier system) rather than the transformational component that accounts for the semantically relevant options, including the possibility of forming recursive constructions. The difference between a declarative and an interrogative sentence, or between an active and a passive sentence, is no longer described in terms of optional transformations, but in terms of a choice made in the base rules. For example , there might be a base rule of the following form:
1. (a) VP Verb + NP (+Agentive)
and the selection of the element Agentive would distinguish the strings underlying passive sentences from the strings underlying the corresponding active sentences. There would then be an obligatory transformational rule corresponding to rule (13) above, operating if and only if the 'input' string contained the element Agentive. This proposal has the advantage that, if we formulate the transformational rule correctly, it gives us a label for the node dominating by + NP2 in the derived phrase markers associated with passive sentences.
The base rules generate an indefinitely large set of underlying phrase markers (which represent the deep structures of all the sentences characterized by the system); and these are converted into revised markers (which represent the surface structures of the sentences) by the transformational rules, most of which (apart from 'stylistic' rules) are now obligatory. The meaning of each sentence is derived, mainly if not wholly, from its deep structure, by means of the rules of semantic interpretation; and the phonetic interpretation of each sentence – its physical description as an acoustic 'signal' – is derived from its surface structure by means of the phonological rules.
We need not go into the more technical details which distinguish an Aspects-type grammar from the conceptually simpler system of Syntactic Structures. All that remains to be added to this account of the general characteristics of the later version of transformational grammar is that various semantically relevant grammatical notions are now explicitly defined in terms of deep structure relations. (This was merely hinted at in Syntactic Structures). We may note, in particular, the distinction between the 'logical' (deep structure) and 'grammatical' (surface structure) subject of a sentence. The 'logical' subject is that NP which is immediately dominated by S (= sentence) in the deep structure; the 'grammatical' subject is the left-most NP which is immediately dominated by (the top-most) S in the surface structure. For example, in a sentence like:
Jane was persuaded by Ann to take up teaching
the grammatical subject is Jane (it is this notion of 'subject' which is relevant to the statement of the agreement holding between the subject and the verb in English: Jane was persuaded vs They were persuaded, etc). But the deep structure of this sentence as shown below, consists of one sentence (S2)
S1 NP VP
V S2
NP NP VP
V NP
Jane Jane take up teaching
embedded within another (S1); and each sentence has its own logical subject. Figure 9 shows a representation of the deep structure of the above sentence. It will be seen that the logical subject of S1 (the matrix sentence) is Ann, and that of S2 (the embedded sentence) is Jane. Furthermore, the deep structure subject of S2 is identical with the deep structure object of S1 (that NP which is immediately dominated by VP). As Chomsky points out, it is these deep structure relations that are essential for the correct semantic interpretation of the sentence.
Another feature that seems to distinguish Aspects from Syntactic Structures is the more explicit recognition that traditional grammatical concepts are, in the main, quite correct, as is the traditional concept of universal grammar. There were obvious gaps in the traditional account of syntax, as well as some inconsistencies in the distinction of substantial versus formal universals. Chomsky believes that the distinction between deep structure and surface structure enables us to appreciate the lasting contribution of traditional grammar and to see why the structural linguist rejects universal grammar. The original form of the transformational grammar was tripartite; it involved:
(i) a phrase-structure grammar that produced terminal strings;
(ii) a transformational component that, through obligatory and optional additions, deletions, and rearrangements, produced the final grammatical representation of all sentences in the language, and
(iii) a morphophonemic component that rewrote these sentences into the proper sequence of phonemes.
The distinction between the phrase structure component (the base) and the transformational component suggests the difference between deep structure (what the base produces) and surface structure (the grammatical representation effected by transformations, and the phonological representation produced by the morphophonemic component). Chomsky finds that this distinction, though variously formulated, is as old as syntactic theory itself (Aspects, p.199).
The foregoing represents the major innovations in Aspects. In subsequent units, we shall discuss other modifications since the advent of Aspects.
CONCLUSION
In this unit, we have endeavored to present the popular models of Chomsky's transformational generative grammar as articulated in "Syntactic Structures" and "Aspects of the Theory of Syntax." While these two grammars exhibit some differences, Chomsky's original proposals have undergone continuous modification since the publication of "Aspects."
5.0 SUMMARY
Throughout this unit, you have gained insights into:
• the Transformational Grammar as proposed in "Syntactic Structures";
• the Transformational Grammar as proposed in "Aspects"; and
• the distinctions between these two grammars.
6.0 TUTOR-MARKED ASSIGNMENT
1. Compare and contrast two differences between the Transformational Grammar of "Syntactic Structures" and that of "Aspects."
2. What were the primary motivations for integrating the Semantic component into the grammar?
3. Provide deep structure trees for the following sentences:
(a) The child was bitten by the ant.
(b) Flying planes can be dangerous.
(c) He asked her to visit Lagos.
Leave a comment behind