ayesha khumaira

Trying to establish a general definition of what is and what is not a theory would not be a fruitful exercise in this kind of publication, but for the reader who is interested in such issues, Chalmers (1982) provides an eminently readable introduction and further references. Similar general issues are discussed specifically from the perspective of linguistics in the articles in Katz (1985). The relevant questions for our purposes are rather ‘When does a linguistic description turn into something more abstract, which we can call a linguistic theory?’ and ‘What is the relationship between description and theory in linguistics?’

With respect to the first of these questions, it is worth pointing out that every description that is not just a list of actually occurring sounds or phrases involves some degree of abstraction, so that for instance as soon as we refer to a unit such as a ‘phoneme’ or a ‘verb phrase,’ we are abstracting away from the pure data. A theory should of course predict (or generate in the sense used above) the correct set of data that it aims to deal with. However, it is often assumed that a good theory should do more than this. Chomsky (1964) defined three properties which a theory should have: they are known as ‘levels of adequacy’ and have played a central role not only within the Chomskyan approach to linguistics. The notion of generating the correct set of data which we have already discussed is referred to as the ‘observational adequacy’ criterion. In addition, a theory must be ‘descriptively accurate’ in that it must abstract away from the actual phrases and describe the principles which allow a theory to make predictions about the grammaticality of strings. Finally, a theory must possess ‘explanatory adequacy’: it must provide an explanation for how human beings can acquire the principles captured under descriptive adequacy. All linguists can be expected to agree on the necessity of observational adequacy. Even though there is some disagreement as to what the exact principles are which are captured under descriptive adequacy, the idea of a theory being required to have such principles is relatively uncontroversial. The idea that a linguistic theory should also explain processing and more generally the cognitive underpinning of language is also fairly widely accepted. However, exactly when a theory can be said to have explanatory adequacy in this sense is a very controversial issue.

Within the Chomskyan tradition, there is great emphasis on the aim of linguistic theory being the potential for explaining the knowledge of a language that is in a native speaker’s head and how it came to be there:

To put the matter in somewhat different but essentially equivalent terms, we may suppose that there is a fixed, genetically determined initial state of the mind, common to the species with at most minor variation apart from pathology. The mind passes through a sequence of states under the boundary conditions set by experience, achieving finally a “steady state” at a relatively fixed age, a state that then changes only in marginal ways. . . So viewed, linguistics is the abstract study of certain mechanisms, their growth and maturation. (Chomsky 1980: 187–8).

This general view of the ultimate goal of linguistic theory is shared by many theoretical approaches which differ from the Chomskyan tradition in other ways, as we shall see in the next section. In an introduction to Head-driven Phrase Structure Grammar (HPSG), we find the following statement on the aim of linguistic theory:

Indeed, we take it to be the central goal of linguistic theory to characterize what it is that every linguistically mature human being knows by virtue of being a linguistic creature, namely, universal grammar.(Pollard and Sag 1994: 14).

However, such assumptions are by no means a necessary part of a theory. Generalized Phrase Structure Grammar, which to some extent can be said to be a pre-cursor to HPSG, very explicitly did not contain any such assumptions:

In view of the fact that the packaging and public relations of much recent linguistic theory involves constant reference to questions of psychology, particularly in association with language acquisition, it is appropriate for us to make a few remarks about the connections between the claims we make and issues in the psychology of language. We make no claims, naturally enough, that ourgrammar is eo ipso a psychological theory. Our grammar of English is not a theory of how speakers think up things to say and put them into words. Our general linguistic theory is not a theory of how a child abstracts from the surrounding hubbub of linguistic and nonlinguistic noises enough evidence to gain a mental grasp of the structure of natural language. Nor is it a biological theory of the structure of an as-yet-unidentified mental organ. It is irresponsible to claim otherwise for theories of this general sort.(Gazdar et al. 1985: 5).

This approach would then not have the property of explanatory adequacy and hence would not be an acceptable theory according to the Chomskyan tradition.

In this context it is, however, important to keep in mind that our empirical knowledge and understanding of how the human mind deals with language is incomplete. Many accounts that claim explanatory adequacy only do so based on the assumptions made about the language faculty within their particular theoretical framework. To someone who does not share those particular assumptions, the theory would not be considered explanatory. Explanatory adequacy is a contentious issue.

To place linguistics in a broader context, we can say that those systems which we refer to as linguistic theories are essentially models of systems, on a par with a model of a chemical compound or a traffic situation. Models in this sense provide an abstract description of a system, in our case a language or a subset of a language. They are, however, not assumed just to describe, but also to enhance the understanding of that which it models. This way of looking at linguistic theory leads us to consider the relation between the model and that which it models, which comes down to the issue of the relation between the data described and the theory.

In this section so far, I have used ‘theory’ to describe whole frameworks, such as HPSG or Chomskyan theory. In a sense this boils down to including both the actual theory and the machinery used to express the theory under the term. Even though this is the way the term tends to be used, it is not entirely accurate to include under ‘theory’ the metalanguage which is used to express the theory. The distinction is sometimes articulated in linguistic writing, for instance by Bresnan (2001: 43) with respect to Lexical-Functional Grammar (LFG):
Note, however, that the formal model of LFG is not a syntactic theory in the linguistic sense. Rather, it is an architecture for syntactic theory. Within this architecture, there is a wide range of possible syntactic theories and sub-theories, some of which closely resemble syntactic theories within alternative architectures, and others of which differ radically from familiar approaches.Bresnan (2001:32).

For the sake of simplicity, I will continue to use ‘theory’ in the more common, less precise meaning. Current syntactic theories share some of their metalanguage, but they also vary substantially with respect to some of their fundamental assumptions. There are different ways of modeling the same data set. At a more abstract level, different theories would all like to claim properties such as ontological parsimony, i.e. a principle known as Ockham’s razor should apply: as little theoretical apparatus as possible should be used to explain a phenomenon within the theory. This is often captured in terms of a principle of economy in theories, but as we shall see, the effect which this principle is assumed to have varies drastically. Theories will also claim to have decidability – formal procedures exist for determining the answer to questions provided by the theory, like whether or not a particular sentence will be generated by the grammar–and predictability – the theory makes predictions about what does or does not occur.