| Well into the Middle Ages students in the Western tradition knew the study of natural language as grammar. Since that time grammar, which was at first a rather general field covering language in all its uses, has become segmented into several separate studies. The year 1818 marks the time when the study of language became formalized into the study of linguistics. Until around the end of the 19th century scholars had called this specialty philology. |
Diachronic linguistics.The central focus of philology was to explain how the forms of language came to be as they are now. Philologists could see that Latin, which had been the lingua franca of the scholarly world, must have been an earlier stage of French, Spanish, and Italian. They devised rules to model the changes, rules that would describe how Latin could have mutated in a regular way into its various descendent languages. Linguists call the analysis of changes through various stages in time a diachronic approach, which results in a historical linguistics. |
| Everything that relates to the static side of our science is synchronic; everything that has to do with evolution is diachronic. | ||
| — Ferdinand de Saussure, 1916, Course in General Linguistics (trans. W. Baskin), I, Ch. 3* |
Synchronic linguistics.In contrast to the work of philologists, the majority of linguists of the present century carry out their analysis on the present state of a language; they compare versions of the same language as they observe it originating from different speakers and writers. This analysis is what they call a synchronic approach, which results in a descriptive linguistics. In many respects this study has separated from and contrasts with historical linguistics. |
Dialectic or symbiotic approaches?These two approaches to grammar result in quite different models of language. It is one thing to describe the changes in a particular state of affairs through time and quite another to describe that state of affairs as it varies between different manifestations in different places. It would seem that the latter would have to be resolved to make sense of the former. Before an investigator can adequately describe the temporal changes brought about to an object in a formal manner at all, it seems one would have to be prepared to describe the object that those changes affect as different from the objects not affected. The two descriptions are interdependent. Presumably the investigator would find that one can hardly observe anything that didn’t vary between instances (or change in time). To some extent synchronic studies would be prerequisite to diachronic studies, yet the results of diachronic studies could also feed synchronic studies. |
| A Grammar book does not attempt to teach people how they ought to speak, but on the contrary, unless it is a very bad or a very old work, it merely states how, as a matter of fact, certain people do speak at the time at which it is written. | ||
| — H. C. Wyld, 1925, Elementary Lessons in English Grammar, p. 12* | ||
Descriptive linguistics.Investigators have reduced the domain of grammar to a number of specialties; the domain of philology has now become descriptive and historical linguistics. In the present century scholars generally understand that the grammar of a language is the concern of descriptive linguistics. A grammar of English is a descriptive study of the forms of the language whether they are the written forms or the spoken forms. Presumably a good grammar would include examples of all the phenomena of English. The grammarian might give rules that would be general enough to allow the student to take any observed utterance and find an analogous form enumerated. |
Comprehensive descriptions.Early in this century the grammarian’s attempts were to make the linguistic description as comprehensive as possible in the sense that it would include all the necessary rules and copious examples. Henry Sweet was perhaps the first noted grammarian of this ilk (Henry Sweet, 1891). In the introduction of his two volume work he claimed that in comparing his syntax to other grammars, his work was fairly complete (Henry Sweet, 1891, p. v). More recently a quartet of linguists, Randolph Quirk, Sidney Greenbaum, Geoffrey Leech and Jan Svartvik, analyzed and classified a huge corpus of conversational English (Svartvik & Quirk 1980). They too felt that they were able to publish what they believed aimed at comprehensiveness (Quirk, et al. 1985, p. v) |
Some weaknesses of descriptive linguistics.Descriptivists tend to concentrate on just a few aspects of the scientific approach — observation and hypothesis — as they relate to formal description. The investigator documents a certain kind of utterance, analyzes and classifies it. At this point some feel their description of the language is complete. Any hypotheses are usually implicit in the presentation of observations. Often empirical work is actually limited to introspection — observations of one’s own thinking. The empirical laws are on this level actually principles of formal description. These might emerge from the descriptions, but are hence intuitive and may not cohere with the rigor sufficient and predictive power needed to build a mathematical model (scientific theory). In recent years the number and variety of models for describing languages seem to be proliferating without bound. |
New sciences — new analyses.Nineteenth century linguists may have been trying to emulate the work of the natural scientists of their generation. Natural scientists collected specimens and set about to study the morphological classes that occur in nature. Grammarians typically collect utterances of a natural language and study the various morphological classes that are apparent. The natural scientist studies his phenomena either in terms of their diachronic evolutionary relationships or as extensions of certain complex synchronic ecosystems. Nowadays the linguist usually approaches his work as either historical or descriptive. It seems unavoidable for scientists to view and analyze the various observables from within the framework of the current theories of their age. |
Chomskyan revolution.Even linguists tend to study individual phenomena within some particular theoretical framework. In the United States the predominant framework is the so-called generative grammar. The move to this orientation came about through the “Chomskyan revolution,” which began in 1957, when Noam Chomsky published Syntactic Structures (Chomsky, A. Noam 1957). It was accepted immediately. Subsequently one linguist (Stuurman, Frits 1990) has observed that the reason this seminal work attracted the serious attention of linguists so quickly was probably the respect that Robert B. Lees accorded it in his review article (Lees, Robert B. 1957). From the beginning Chomsky has continued to enlarge upon and expand his model considerably (Chomsky, A. Noam 1965, Chomsky, A. Noam 1979, Chomsky, A. Noam 1981). Unfortunately there is a double meaning in the term “generate,” which has caused no end of confusion. Consider his original definition of syntax, shaped to fit his goal: |
| Syntax is the study of the principles and processes by which sentences are constructed in particular languages. Syntactic investigation of a given language has as its goal the construction of a grammar that can be viewed as a device of some sort for producing the sentences of the language under analysis. More generally linguists must be concerned with the problem of determining the fundamental underlying properties of successful grammars. | ||
| — Noam Chomsky, 1957, Syntactic Structures, chapter 1* | ||
| On the one hand, “generate” is a mathematical term for a particular descriptive system. On the other hand, Chomsky seems to have applied it to mean that human language is generated in the human psyche of conceptualization directed by the social systems of acceptability. These two interpretations are on two different levels of investigation; one is a metalanguage for the other. The sad result of this latter interpretation has been that the investigators are given the false hope that generative grammar may directly provide us the explanitory power of scientific theories. Such theories are as yet forth-coming from developmental psychology and sociology. Somehow the generative grammar will magically connect the real world of language to our descriptions. These have as their object the system of signs in natural language, not the behavior of human beings — not the meaning we use language to convey. We cannot hope nor do we intend to explain our language capacity, only the variety of its syntactic structure. It is the claim of the majority of Chomsky’s adherents that somehow finding out the best way to describe natural language syntax in the generative framework will lead us to the way the human brain produces it. |
| You must lie upon the daisies and discourse in novel phrases of your complicated state of mind, | ||
| The meaning doesn’t matter if it’s only idle chatter of a transcendental kind. | ||
| — W. S. Gilbert, 1881, Patience, 1* | ||
Novelty in language.One fundamental principle underlying and running through the work of most generative grammarians is the fact that natural language utterances are creative and novel. Within certain constraints the language user is amazingly free to put together words and grammatical structures in an infinite variety of ways. The implication is that the descriptive accounts of the language, the documentation of utterances, are only a beginning. Their form must be generalized so that the grammar also serves to describe expressions that are as yet only potential. This may seem to correspond to the predictive power of a scientific theory. It is in fact the power of a descriptive grammar and serves to expand the data that supports the grammar. Just as a scientist makes observations that go into building the laws of some theory, so the linguist engages in objective observation. With work these must contribute to the development of “empirical laws.” In a theory of grammar the empirical laws are established principles relating directly to syntax, not anthropology. The study of language enlightens anthropology by way of telling us about the conceptual sytems of a society. These are choices of meanings for signs. The syntax of these signs can tell us something about the relationships between the concepts. But societies change. Languages change. The scientific approach suggests that hypotheses about the natural system of signs will have little to do with the content of the conceptual systems of society. What grammar describes is fractionated and faulted on its interface with what the science of anthropology seeks to explain. The linguist must construct models, maintain them with evidence and only later attempt to understand this interface between the two realms. |
“God’s truth.”The first half of the present century was dominated by the descriptive tradition. This approach relies heavily on inductive reasoning. In this tradition investigators attempt to collect facts without reference to a theory, because a theory would impose an organization on the facts and predispose the investigator to observe only certain facts. Objective unbiased observation of this kind is in the spirit of Darwin on the Beagle. Observation includes a good bit of classification and analysis, if only from simple to complex. The linguist in the Popperian spirit of this period, as many even today, was discovering, in the terminology of C. F. Householder, “God’s truth.” Linguists generally understood that they would have to emphasize thorough observation if they were to make progress. |
“Hocus pocus.”Growing out of the data collected through observation is the stage where the scientist develops certain general principles, rules, or laws. The scientist then determines what sort of facts he may deduce from these general principles in order to validate the laws. He must try to demonstrate the predictive power of his laws. Householder called this kind of invention of laws that actually work “hocus pocus.” These investigators understand that the development of empirical laws is necessary to make any headway in the science of linguistics. The reader of the present work will find a liberal number of observations ordered into a schema that go to support paraphrase as a useful if not true principle for language description — hocus pocus. To the degree that scientists can actually extend the systematic description so as to affect studies of language development and change, we are able to see some real scientific progress! |
Heuristic approach.Observation, hypothesis and empirical laws are all involved in the development of a heuristic. A heuristic is a set of methods for making discoveries. Zelig Harris has maintained that linguists could arrive at the grammar of a language by the systematic application of a certain proved set of discovery procedures (Harris, Zellig S. 1951). It seems that the feeling was that each language would have its own set of laws to be discovered. However, many linguists today continue to concentrate on finding linguistic universals — laws that apply to every language. Their usual heuristics seem now to be motivated by the hope that principles of child language development will be found to have universal application, and vice versa. |
Tension in linguistics.There were then two kinds of descriptive models that came to oppose each other. The one model wanted to build at the level of observing linguistic entities, then hypothesizing their properties in some unifying way. This came to be called an “item and arrangement” approach or the “element model.” The other model wanted to hypothesize the relationships between the entities using their models of linguistic change. This came to be called the “item and process” or simply the “process” model. Figure 3 diagrams the tension that developed in descriptive linguistics. In 1954 Charles Hockett pointed out how the evidence of the language could support these two very different models (Hockett, Charles F. 1954). Stuurman points out that back in 1928 Henrik Poutsma in writing his grammar of English wanted intentionally to refrain from coming forward with any kind of hypothesis, claiming that deduction would be premature (Stuurman, Frits 1990). According to Poutsma the linguist should avoid “wresting” the available evidence so as to bring it into harmony with some “pre-conceived theory” (Poutsma 1928, p. viii). Chomsky’s attempt to posit laws contrasts strongly with Poutsma’s hesitation. Chomsky requires that the linguist construct the grammar in accordance with a linguistic theory (Chomsky, A. Noam 1965, p. 62). What guides a Chomskyan is pertinent facts, and what is pertinent depends on the requirements of the theory (Chomsky, A. Noam 1979). For me there is nothing wrong with proposing certain hypotheses as an unproven theory. The understanding should be that the theory may fail after continued investigation. What linguists need to produce is a sufficient number of empirical laws that come to be accepted and can cohere into a theory that may stand the test of time. Many theorists formulate their linguistic principles in terms of a particular model (a model-theoretic approach). This is much like a cart pulling the horse. If a particular generalization cannot be expressed with their model, then maybe that generalization shouldn’t be made. The other approach is to adapt and incorporate models that can express whatever generalization needs to be made. Such an extremely eclectic approach could posit a different model for each observation. |
Models versus theories.Disagreements arise over whether a linguist is entitled to formulate a particular theory on the basis of the particular linguistic evidence that he cites. (Should an empirical law be raised to the status of a theory?) There are debates over whether what the theory predicts is in harmony with observation. As an example consider the “subselection model” for describing past tense morphology in English (Gleason, H. A. 1961:215). With this model the linguist describes the morpheme for past tense, {-D1}, as a choice between:
| |||
| The verb form is first selected, say “leak” /liyk/, and then based on this selection the correct allomorph is subselected, in this case /-t/. Against this model, i.e. device or apparatus, stands an “adjustment model.” In this approach the morpheme /liyk/ combines with the past tense morpheme /-d/ after which the combination undergoes the effects of a morphophonemic rule (as an element, not a process). In both cases the final form /liykt/ results. The difference between these two models is empirical, one driving decisions made by software engineers on a regular basis. The first model represents a table-driven solution whereas the second, a data-driven solution. The former is useful when the data class is closed, say irregular verbs that can be listed. The latter is more natural for data classes that are open, say regular, i.e., rule-driven verbs. We will illustrate this choice of models further in the chapter. |
Theory describes social conventions.The self-evident truth remains that a scientist, however sincerely objective, cannot come up with laws for language without some sort of guiding principles to constrain the form that the laws should take. Scientific theories do not typically come into existence in their final form. Current theories are ultimately based on numerous elements taken from earlier more primitive formulations. When it comes to true comprehensiveness, the most reasonable position seems to be that the linguist should describe any utterance at all which is capable of being produced, provided it is or would be accepted in the linguistic community as grammatical. This places the determination of what is grammatical on society, not the grammar.
| |||||
| Language is always in flux but its description must in practice be confined to a specific time, place, and social context. |
