First newsletter: The GLOW Manifesto

To give a more detailed insight into the ideals (and controversies) behind the founding of GLOW, we reproduce in these pages some of the discussions from the first issue of the GLOW Newsletter (September, 1978).

What GLOW stands for

From the outset, GLOW has aimed for internal intellectual cohesion rather than wide coverage. In order to provide a platform for our identity[,] a text was circulated which attempted to make explicit some of the beliefs characterizing an “idealized” member of GLOW. As such, the text, which has come to be called “manifesto” by now, does not represent a dogma, does not define a catechism that members of GLOW must adhere to. Rather, it represents a nucleaus of beliefs around which, in our opinion, the personal intellectual positions of the members of GLOW gravitate. As is to be expected under such an interpretation, some members share very few of the tenets of the manifesto and yet remain actively interested in GLOW, while others find themselves very much in agreement but do not partake in the activities of GLOW.

Understandibly, the manifesto has aroused a certain amount of reaction. We therefore reprint the theoretical part of the original text and the most elaborate of the reactions: that of Jean-Claude Milner. However, we do not intend to continue the publication of such statements, since we feel that no goal whatsoever is served by the proliferation of personal statements of belief about which a serious debate is impossible from the outset.

GLOW Manifesto

Concerning the object of inquiry

Linguistics is that part of human psychology that is concerned with the cognitive structures1 employed in speaking and understanding. Such cognitive structures can be viewed as steady states of the corresponding “mental organs”. As natural scientists, linguists are primarily concerned with the basic, genetically determined structures of these organs and their interactions, a structure common to the species in the most interesting case.

The theory of language is that part of linguistics that is concerned with one specific mental organ, human language. The growth of the latter organ, language learning, falls strictly within biologically determined cognitive capacities. It appears quite likely that the system of mechanismes and priciples put to work in the acquisition of the knowledge of language will turn out to be a highly specific “language faculty”. Stimulated by appropriate and continuing experience, the language faculty creates a grammar that determines the formal and semantic properties of sentences. The grammar is put to use, interacting with other mechanisms of mind, in speaking and understanding language.

The faculty of language, which is common to the species, provides a sensory system for the preliminary analysis of linguistic data, and a schematism, “Universal Grammar”, that determines, quite narrowly, a certain class of grammars. Thus, Universal Grammar (UG) is the system of principles, conditons, and rules that are elements or properties of all human languages not merely by accident, but by necessity.

“Thus UG can be taken as expressing `the essence of human language’. UG will be invariant amoung humans. UG will specifiy what language learning must achieve, if it takes place successfully. …What is learned, the cognitive structure attained, must have the properties of UG, though it will have other properties as well, accidental properties.”2

In principle, we should be able to account for the language faculty, and hence for linguistic theory (the theory of UG), in terms of human biology. With the progress of science, we may come to know something of the physical representation of the grammar and the language faculty — correspondingly, the cognitive state of the attained in language learning and the initial state in which there is a representation of UG but of no specific grammar conforming to UG. For the present, we can only characterize the properties of grammars and of the language faculty in abstract terms. At this point some comments are in order.

The term “grammar” is used with a systematic ambiguity; i.e., it not only refers to the rules and principles, the theory contructed by the linguist, but also to the corresponding structure internalized by the speaker-hearer. Grammar in the first sense may be interpreted as a theory of grammar in the second sense. The internalized structure represents the tacit knowledge or competence of the speaker-hearer. This postulated correspondence between linguistic theory and competence is a major source of controversy and misunderstandings. Critics often fail to see the methodological decision that is involved. The point is that any grammar can be interpreted as a theory of human linguistic competence. It is only under such a “realistic” interpretation that certain questions of empirical adequacy can be raised at all. Suppose, for example, that we decide to interpret a given traditional grammar as a theory of competence. Only then does it make sense to say that it is wrong in many respects; for instance, in that it fails to account for the infinity of the set of possible sentences, as well as for many less obvious facts. The question: “how do you know that the linguist’s theories represent the real psychological structures” is easily translatable into similar questions in other domains of science. How do you know, for instance, that the principles of quantum mechanics model the real principles of nature? The only answer is — just as in the case of linguistics — that it is the best theory scientists can devise, given the available evidence. Of course our theories will change as soon as a broader class of relevant evidence becomes available. If a grammar is interpreted realistically, the only matter that counts is whether it is true or not. Only the validity of a specific grammar (conceived as a psychological theory) can be challenged by evidence, not the decision to interpret it realistically, which is a matter of goals and interests.

Another source of continuing debate is the competence-performance distinction. As such the distinction — between what a person knows and what he does — seems hardly obscure. What usually is at stake is the legitimacy and use of the distinction. In our opinion much criticism appears incoherent when it is recognized that human cognitive behavior involves the interaction of diverse cognitive structures. Actual linguistic behavior involves structures like grammar, various kinds of knowledge of the world, and so on. A direct route to performance, use, process, and the like, seems ill-conceived, because it would involve the result of interacting factors that are themselves unknown. It is therefore hardly surprising that research that takes the direct route to language processing (much AI research for instance) usually tacitly assumes competence theories of a rather arbitrary and primitive sort. In our opinion, the analysis of cognitive structures has to preceed the study of the enormously intricate synthesis which we call behavior. This priority is not simply a matter of taste or interest, but a point of logic: if behavior is a consequence of interaction, it can only be understood by finding out of what interacting factors it consists. The kind of cognitive psychology we advocate therefore rejects the holistic study of behavior as hopelessly premature. A primary task should be the detailed description of the any cognitive structures (competences) that ultimately interact in behavior (performance). Grammar is one such cognitive structure that can be mapped out with some success. From this fact — if it is a fact — nothing follows with respect to our interest in other cognitive sturctures or our interest in their interaction, not to mention the totally unmotivated suggestion that for us “performance is everything that falls outside grammar”.

Some methodological considerations

From the beginning, generative linguists have implicitly or explicitly assumed the thesis of autonomy of formal grammar — the thesis that the language faculty constructs an abstract formal skeleton invested with meaning by interpretive rules — and have taken for granted the necessity of limiting the scope of inquiry to aspects of formal grammar. While it is quite clear that such fields as botanics3, the study of the artificial languages of logicians, astrophysics, the theory of truth are not part of linguistics, there is no real logical connection between the practical exclusion of certain domaines of inquiry by the generative linguists and the definition of their object. However, the limited scope of early transformational grammar was widely interpreted as a matter of principle, and not a matter of research strategy. In Europe, as elsewhere, much critical discussion was concentrated on the alleged “neglect” of meaning and language use, and not on the new goals, such as narrowing of the class of possible human grammars. Questions of restrictiveness (ultimately questions of explanatory adequacy) can only be raised where there are sufficiently explicit theories that have at least some content. Scope of description is an entirely irrelvant issue as long as the considered theories are without content, i.e., when the rules are unconstrained and do not forbid anything whatsoever. The descriptive means of early transformational grammar were so rich in expressive power that it was all too easy to go beyond the original limits. The expansion of the scope of linguistics theory in the hands of generative semanticists was simply an expression of a lack of content of the proposed hypotheses. Far from being a revolutionary Kuhnian paradigm, Generative Semantics was an essentially regressive development, which met the call for wider scope by relaxing the explicitness and explanatory goals of Chomskyan [sic.] linguistics.

In our opinion there are no a priori grounds on which to decide what the domain of a theory should be. The fundamental problem of linguistics in its present state is not how to incorporate meaning, pragmatics, or whatever, but rather how to develop a theory of a certain scope that is explanatory at all. As Chomsky puts it:

Our glimmerings of understanding can only be expected to illuminate some narrow range. If we hope to proceed beyond taxonomy, it is necessary to select and discard, to concentrate on facts that seem to have some bearing on such explanatory principles as we can devise, ignoring much else in the hope that it will ultimately be explained by deeper theories or perhaps on quite different grounds.4

These remarks are at the heart of the problem. As long as any attempt at idealization, “to select and discard”, is interpreted as the “neglect” of any aspect of language, linguistics shows its lack of a tradition comparable to the natural sciences. For example, the fact that certain aspects of formal grammar give rise to some explicit hypotheses and explanatory principles makes it only natural that attention focuses on these areas. In a mature linguistics, the hierarchy of interests is determined by the explanatory goals and possibilities, by the course of the theory itself, and not by pre-theoretic judgements about intrinsic interest of the diverse aspects of language.

The Chomskyan revolution has been a revolution of goals and interests as much as a revolution in method or content (formal description of strings of morphemes or phonemes, subject to the empirical test of acceptability judgements). In our opinion, generative linguistics acquired a new momentum in Europe after Chomsky’s “Conditions on transformations” (1973). This epoch making paper shifted the interest of linguistics from rather arbitrary rules to simple well-constrained rules operating under general common ground in the research program that grew out of “Conditions [on transformations]”.

Jan Koster
Henk van Riemsdijk
Jean-Roger Vernaud

 [Read Milner’s response to the manifesto]


1 Chomsky (1975) Knowledge of language; Chomsky (1975) Reflections on language

2 Reflections on language, p.29

3 Sic.; i.e., botany -[Craig Thiersch]

4 Chomsky (1977) [ Essays on form and interpetation ] [p.] 21