A Critique of the Early Models of Transformational Grammar
Generative transformational grammar strongly affected theoretical studies of language. Linguists holding the theory explored syntactic problems, attempting to clarify relationships between syntactic patterns as traditional grammars had clarified morphological relationships. Patterns like passivization, ‘raising’, and so on, were scrutinized; if grammars were to account for the total structure of language economically, they might hope to solve, for example, with one set of rules, the process involved in raising subjects equivalent to the subject of the matrix verb as in (1) below, and those differing from it, as in (2) and (3).
.jpeg)
1. I expect to come
2. I expect him to come
3. I expect that he will come
Intensive investigations of such problems led not only to various solutions but also to great concern with syntax. That concern is a major contribution of transformational grammar.
Moreover, the emphasis placed on syntax has brought this segment of language greater prominence. A new large-scale handbook on English places far more emphasis on syntax than on morphology or phonology. Yet, this grammar does not adhere closely to the theory proposed by Chomsky. The effect of transformational theory may, then, be indirect, in focusing emphasis on syntax without determining the procedures used in presenting syntactic patterns.
Transformational grammar has also exerted an influence on the teaching of language. An example is the use of ‘sentence combining’. In learning to write, students often set down one simple sentence after another, without using devices like cleft sentences to indicate emphasis, without subordinating matters that are less essential, and so on. ‘Sentence combining’ instructs such students in understanding such relationships.
As we have noted, transformational grammar treats cleft sentences as variants of simple sentences. Similarly, relative clauses and other subordinate clauses are embedded through transformational rules. By applying this view of such complex sentences, teachers of rhetoric have been successful in improving the writing skills of such students.
A major impact of generative transformational grammar has had to do with psychological investigations involving language. Earlier psychological study of language was often limited to external, in keeping with the attention linguists gave to surface forms. Investigations were carried out on possible recall of lists of words. These yielded little information on language or on its control by the brain. The conjectures proposed by transformational grammar, on the other hand, offered possibilities of studying patterns of far greater interest. Cognitive psychology thereupon undertook to determine the reality of such constructs as subjects and predicates of clauses. Investigators of language acquisition examined the sequence of mastering complex structures, such as passive as opposed to active sentences. Such investigations are continuing, not only on language acquisition but also on language loss, as in speakers with brain injuries, on aphasics, as with neurotic and psychotic speakers. Investigations of all these kinds are carried out to determine the common features of language.
Features and characteristics common to all languages have come to be known as universals. They receive support from such observations as the ability of all infants to acquire any language. That ability must rest on basic features underlying all languages. One readily observed universal is the existence of sentences in all languages. Another is the universal use of consonants. But investigators seek more specific universals. The interest in universals has been heightened by understanding of control of language by the brain, especially by greater knowledge of areas that control semantic, syntactic, and phonological functions. As such understanding has been increased, further problems have been disclosed, among them the modifications of sentences in longer sequences, known as discourse, or text. These problems will not be examined here.
Having mentioned some of the insights of transformational grammar, let us look at those central concepts of grammatical theory for which nativists have received the most criticism over the years. Clarifying them will at least remove whatever share of the resistance to the theory is based on misunderstanding.
Competence and Performance
Probably no notion within transformational-generative grammatical theory has aroused more controversy than the competence/performance distinction. Criticisms of the distinction range from the assertion that it is ‘almost incoherent’ (Labov, 1972:110) to the conclusion that it is coherent enough, yet ‘too confining’ (Clark & Haviland, 1974:92), since so many systematic aspects of language do not fall under the generativists’ conception of competence. It might be pertinent to give this quotation in full:
many linguists have come to find the study of linguistic competence … too confining. There are simply too many interesting linguistic phenomena that do not fit under this rather small umbrella, yet are amenable to linguistic investigation… problems such as the structure of conversations …, the relation of meaning to context …, the perceptual difficulty of surface structures … the production of speech errors and speech hesitations..., and other similar phenomena. (Clark and Haviland, 1974:92)
If we bear in mind that ‘linguistic competence’ is no more than the name for the non-reducible core of language – those aspects of language that form the autonomous purely linguistic system characterized by a formal grammar, it becomes clear that many things we ‘know’ about language do not fall under the generativist’s competence as defined by Chomsky. Our competence is our tacit knowledge of the structure of our language. For example, we know that saying I’m hungry can convey a request to be fed, and we know that we should devoice our consonants and vowels when speaking in the library or in the church. This sort of knowledge is not competence because it is not strictly linguistic. The generalizations underlying this knowledge undoubtedly fall within the domain of cooperative communication and proper social behavior.
‘Linguistic performance’ refers to ‘the actual use of language in concrete situations’. There hardly exists an aspect of performance to which competence does not contribute. One could not describe fully (much less explain) a particular instance of whispering in a library or a church without reference to the social convention of silence in libraries or churches. Just as all non-marginal linguistic phenomena demand a theory of performance for their explanation, it seems inconceivable that any aspect of language use could owe its explanation only to competence. Competence simply represents the speaker’s knowledge of linguistic structure. Explaining speech production, therefore, requires a model of language in which the competence model is supplemented with other explanatory devices. Even in interpreting metaphorical explanations, if our competence did not supply a ‘normal’ reading, we wouldn’t even recognize it as metaphorical.
Another common dissatisfaction with the competence/performance distinction is that, for any given linguistic phenomenon, no hard-and-fast criterion exists to decide which aspects of that phenomenon should fall under competence. Some linguists have explicitly avoided referring to the distinction for that reason (e.g., ‘communicative competence’ has become fashionable). But still, enough is known about the systems interacting in language that the role competence plays in a particular phenomenon is often clear. For example, it hardly seems likely that the fact that in most English dialects two modal auxiliaries do not follow each other pertains to anything but competence, or that the fact that one does not normally shout obscene expletives in church has any explanation but a performance one. Although there are borderline cases, clear cases where the distinction between competence and performance is undoubtful are many, and they are enough to rely on.
Linguistic Universals
Another area of criticism of transformational grammar concerns linguistic universals. Scholars have argued that there is nothing ‘universal’ about generative transformational grammar since grammatical properties differ from language to language, especially as gleaned from the study of child language acquisition. Some scholars even attempt to overthrow the notion of ‘linguistic universal’ by hypothesizing the common origin of all languages. They argue that description of individual languages is more rewarding than searching for universals. However, these critics miss the mark.
The term ‘universal grammar’ is used to refer to that which is true of language by biological necessity. Hence, universal grammar is “taken to be the set of properties, conditions, or whatever that constitutes the ‘initial state’ of the language learner, hence the basis on which knowledge of language develops” (Chomsky, 1980:69). Particular aspects of universal grammar have typically been motivated by reference to the lack of the stimulus available to the child language learner. Why is it, for example, that a child born of Nigerian parents in England acquires English language the same way as English children? Take, for example, the way children form correct questions from declarative sentences. How are children trained to form them? This principle is not learned but forms part of the conditions for language learning. Whether languages have one ancestor or not, the problem for the child is the same – to construct a grammar on the basis of limited input available. Universal grammar studies the conditions that must be satisfied by the grammar of all languages. In other words, it characterizes human language as a whole. It should be noted that not all features common to all languages of the world are properties of universal grammar. There may be universals that arise independently of innate principles: it may be a universal that every language has a word for arm or that no language has more than thirty vowels, but it would be wrong to attribute these universals to properties of Universal Grammar. Also, it does not follow that every putative universal must be instantiated in every language. For example, for each type of grammar rule (transformations, deletions, filters, etc.) there may well be universal properties governing the behavior of rules of that type. Yet, it may be that not all languages manifest all rule types. If a language had deletion rules, one would hardly wish to conclude that the facts of that language ‘falsify’ some claim about the universal properties of such rules.
Simplicity and Evaluation
Another criticism concerns the ‘simplicity’ measure as a way of evaluating grammars. The criticism is that Chomsky, in insisting on simplicity of grammars, has implied that exceptions to grammatical rules should be very ‘costly’, and therefore, rules should be constructed so as to yield the smallest number of exceptions possible. They argue that transformational grammarians have implied, by so saying, that actual language can be wholly generated by rules. Many schools of linguistics, as a result, have rejected Chomsky’s ‘cost-benefit’ analysis of language.
However, this criticism of the simplicity measure turns out to be a rejection of scientific linguistics itself. Science, by definition, is the search for order in nature. Scientists take it for granted that their goal is to formulate the most elegant (i.e. the most order-reflecting) hypothesis possible, consistent with data, about the particular area under investigation. Although philosophers extensively debate scientific methods and the ontological status of constructs, this general point has hardly been questioned.
Linguistics, then, is the search for order in language. The goal of linguistics is to formulate the most elegant hypotheses possible about how language works, consistent with data. No formula exists for how to do this, nor is there an automatic decision procedure for picking out one of several competing theories as more elegant than the others. Yet, certain points are uncontroversial. One is that given two theories that cover the same range of facts, the one in which the facts follow from a small number of general principles is better than the one that embodies myriad separate statements and auxiliary hypotheses. Another is that it is methodologically correct to reduce redundancy within a theory; to reduce the number of postulates while preserving the scope of predications. Scientists (including linguists) often talk about ‘simplifying’ a theory or about constructing a theory with parsimony. Generally, they mean a less redundant theory as described above. It becomes obvious then that the simplicity criticism is unscientific.
Psychological Reality
The last criticism we shall discuss concerns introspection or the use of intuitions in grammaticality judgments. Critics claim that a concept is labeled ‘psychologically real’ only if it plays a particular kind of role in a particular kind of experimental procedure. They base their arguments on processing studies. More recently, other sorts of experimental data have been cited as evidence that grammatical theory is at odds with processing facts and hence lacks psychological reality.
It seems to me that negative conclusions based on processing studies about the psychological reality of grammatical theory are not compelling. Any theory that is based on psychological data and has as its principal goal the explanation of those data is ipso facto a psychological theory. A theory that does this correctly is a theory with ‘psychological reality’; a theory that fails to accomplish this lacks psychological reality.
Now, where does grammatical theory fit into this picture? Largely, but not exclusively, its data base has been speaker introspections about the well-formedness and interpretation of sentences. Surely, introspections count as psychological data if anything does. If that were all there were to it, then the constructs of grammatical theory could be considered psychologically real to the extent that they correctly explained such data.
But that is not all there is to it. Introspections are only one of many potential sources of psychological data about language. Data from studies of speech perception and sentence processing also count as psychological data. So do data gleaned from observations of aphasic speech, speech errors, and the acquisition of language by native speakers. A psychologically real theory of grammar must therefore be judged on more than its ability to explain intuitions; it must be shown to be compatible with all these sources of data.
Conclusion
We have tried in this unit to present a critique of Chomsky’s transformational-generative grammar. In spite of the criticisms discussed above, transformational generative grammar remains the most popular grammatical model in use today in the analysis of language.
Summary
In this unit, you have learned about transformational generative grammar.
Leave a comment behind