The title of this collection might tend to suggest that the introductory pages should provide a definition of Empirical Semantics. Rightly or wrongly, I do not propose to give one. Instead I shall try - for a start - to delimit the field indirectly by way of two framing positions, which mark out the area of our concern. Then, the frame itself will be characterized in terms of what it may constitute and what it may not, before, finally, some points are made which may serve to survey what in fact this frame is to comprise, namely the papers presented in these volumes.
1. Animals studied by Americans rush about frantically, with an incredible display of hustle and pep, and at last achieve the desired result by chance. Animals observed by Germans sit still and think, and at last evolve the solution out of their inner consciousness. To the plain man, such as the present writer, this situation is discouraging. I observe, however, that the type of problem which a man naturally sets to an animal depends upon his own philosophy, and that this probably accounts for the differences in the results [1].
This ironical AI-evaluation of empirical approaches in `animal intelligence' research given by Bertrand RUSSELL more than fifty years ago is quoted here not - or not primarily - because of the intriguing though permuting analogies that contemporary `artificial intelligence' research has produced by substituting algorithmic simulation of programmable automata for experiental observation of animal behaviour. The reason for the quotation lies more in its implicitly expressed relativism which on the surface might seem to be associated with the general notion of observation but could equally well be traced to an even deeper layer of experience.
The latent parallelism alluded to here becomes evident when attempts are considered which tried at about the same time to reflect on the conditions of possible experiences and their contents and, in so doing, identified these as determined by the semiotic system and the structure of natural languages.
Language is not merely a more or less systematic inventory of the various items of experience which seem relevant to the individual as is so often naively assumed, but is also a self-contained, creative symbolic organization, which not only refers to experience largely acquired without its help but actually defines experience for us by reason of its formal completeness and because of our unconscious projection of its implicit expectations into the field of experience. [... Meanings] are not so much discovered in experience as imposed upon it, because of the tyrannical hold that linguistic form has upon our orientation in the world. Inasmuch as languages differ very widely in their systematization of fundamental concepts, they tend to be only loosely equivalent to each other as symbolic devices [2].
It may be argued whether Edward SAPIR's universal hypothesis of `linguistic relativity' and the rigorous scepticism of RUSSELL's `behavioural observation' can still be maintained or not; but there is no doubt that the conjunct of both characterizes the semanticists' dilemma in general and that of advocates of their - in whatever sense - experiential approaches in particular. Although acceptance of the two extremal positions might question the very possibility and necessity of a strict observational or experimentally controlled study of meaning, concomitantly they set up the poles between which - if at all - empirical semantics will intervene.
2. Empirical semantics being more of a label, hardly yet a program, and certainly not an encompassing and consistent strategy of research, serves as a frame for some of the recent approaches advanced by scholars from different disciplines active in the field.
Semantics, the study of meaning has become focal within a number of hitherto specialized disciplines. Scholars engaged in these research activities converge on a common interest in natural and artificial language systems and/or the communicative processes related to their use. They differ, however, in what from their points-of-departure constitutes semantics. The variety of aspects raised ranges from the analysis and representation of natural language meanings as conveyed by signs, words, sentences, and texts via conceptual knowledge, memory structure, thought processes and logical inferencing, to the procedural modelling of cognition and comprehension.
The empirical strategies which are being followed in order to relate this study to observable, experimental, and/or simulative control are as manifold as the topics under investigation. They differ in specificity and vary according to the formal assumptions and practical claims that can be made for them on the grounds of a discipline's more or less developed theoretical status. This status determines the empirical bearing of an approach more profoundly than any particular collection and analysis of data, model construction or simulation alone would signify. Thus, empirical semantics is more a continuous process, which consequently will be put into effect only gradually in developing and relating theories and their practical use, than a status which a study of meaning may (or may not) be found to comply with according to whether a decidable set of defining requirements is met or not. Empiricity is a matter of degree.
3. We constantly compare language with a calculus proceeding according to exact rules. This is a very one-sided way of looking at language. In practice we very rarely use language as such a calculus. For not only do we not think of the rules of usage - of definitions, etc. - while using language, but when we are asked to give such rules, in most cases we aren't able to do so. We are unable clearly to circumscribe the concepts we use; not because we don't know their real definitions, but because there is no `real' definition to them. [...] To the question "How do you know that so-and-so is the case?" we sometimes answer by giving criteria and sometimes by giving symptoms. [...] In practice, if you were asked which phenomenon is the defining criterion and which is a symptom, you would in most cases be unable to answer this question except by making an arbitrary decision ad hoc. [...] When we talk of language as a symbolism used in an exact calculus, that which is in our mind can be found in the sciences and in mathematics. Our ordinary use of language conforms to this standard of exactness only in rare cases. Why then do we in philosophizing constantly compare our use of words with one following exact rules? The answer is that the puzzles which we try to remove always spring from just this attitude towards language [3].
These notes from WITTGENSTEIN's Blue Books might well be understood as an early foreshadowing - or rather, "fore-lightening" - of a number of new approaches to a bundle of old problems which are gaining increasing importance in the field of cognition and language comprehension.
In this realm, traditional i.e. philosophically oriented investigations of knowledge, thought, and language have had a long-standing tradition of posing practical questions and producing theoretical answers. Meanwhile, the inverse has become characteristic of new approaches in the semiotic domain of cognitive and information sciences. Relevant research which has been and is currently being undertaken in the different disciplines concerned reveals, in spite of the theoretical problems raised in the questions quoted above, a growing tendency to come up with solutions which are practical in the sense that they are applicable to or reconstructable within operational models of some sort.
It is this kind of distinction between `theory' and `model' which - after having been made and practiced in the sciences and in engineering for some time - now seems to be becoming productive for some approaches in language and cognitive psychology and, of course, artificial intelligence. In these disciplines, general theories, as usual, may still be presupposed or developed in an informal or heuristic way, but only so that certain components of them may be specified as to be studied in a small-scale model. Preferably implemented as a computer program, this allows for the algorithmic simulation of its properties in order to test, evaluate, or modify the ideas, or fragments of them, developed in a large-scale theory.
It is felt that this approach might provide new insights and some - if not the only available - experimental control of differing answers to the questions
Apparently, this development has not yet gained momentum within linguistics, these ideas being taken up only reluctantly, if at all. Although linguists' research interests have undergone and still are undergoing considerable changes, these seem more to be confined to shifts from one object or domain to another, than to be initiated by a revised theoretical and/or methodological position comparable to that attained by the seminal impact of formal syntax theory on grammar in the late fifties', or of formal logics and model theory on semantics in the early seventies. Provoked more often by findings from non-linguistic disciplines than by those brought up within linguistics proper, shifts like for example, those from sentence to text or from syntax via semantics to pragmatics, and the separation of linguistics proper from hyphenated disciplines like psycho-, socio-, and ethno-linguistics reflect this tendency. Recent developments in the semiotically oriented cognitive and information sciences, however, have created a possibility if not the need for a new paradigm in lingutstic semantics.
4. There will probably be no argument among semanticists from any discipline that - however the above questions are answered and whatever else has to be dealt with in detail - the study of natural language meaning, its analysis and description, its representation and modelling, presents three major problems. Firstly, what is known as the denotational aspect of how the signs, words, and sentences of a language are related to the entities (objects and/or processes) they refer to, constituting referential meaning as a system of extra-lingual relations, secondly, what is known as the connotational aspect of how signs, words, and sentences of a language are related to one another, constituting structural meaning as a system of subsystems of intra-lingual relations, and thirdly, what is referred to as the dynamic aspect of how - and within what communicative frame - signs, words, and sentences of a language are related to functions which instantiate varying restrictions on possible choices of (referential/structural/procedural) meaning representations, constituting procedural meaning as a system of (recursively defined) operations that work on and simultaneously change the conceptual data of memory.
To start with the denotational aspect of meaning, referential semantic theory has developed along the lines of FREGE, RUSSELL, the early WITTGENSTEIN, and CARNAP. Their relevance to linguistics has only been recognized during recent years. Linguists' increasing interest in formal logics as a representational notation for language semantics has produced quite a number of different approaches since. These share the fiction though, that natural language meaning is essentially declarative, as opposed to associative, and should therefore be analysable in propositional structures which are either `true' or `false', or have a third value such as `indeterminate'. Like truth-conditions for formal predicates or propositions, those for natural language sentences are to be modelled analogously and hence be introduced in terms of classical set theory. Accordingly, the meaning of a word basically appears to be identifiable with its compositorial function in the propositions it may constitute. These, in turn, are interpreted by their denotations defined either extensionally as a set of points of reference, or intensionally as a set of satisfied properties in the universe of possible worlds, allowing truth-values to be assigned to any (declarative) natural language sentence thus reconstructed. These truth-value models of sentence-semantics now tend to exhibit all the formalisms and idealizing abstractions that the logical rigour of binary formal systems calls for. They do so, however, at the price of a rather limited coverage of basic and very obvious characteristics of natural language meaning, like for example, indeterminacy, vagueness, variation, con- and co-textual dependency, etc. All these phenomena cannot be accounted for adequately by these approaches but were considered neglectable noise factors. As they did not seem to perturb the formal assignment of truth-values, they consequently did not attract special attention as a problem of sentence-semantics unless this was founded on pragmatics.
Unlike referential semantics, structural semantic theory has primarily been concerned with word-semantics. As such, the afore mentioned characteristics were always considered fundamental to the constitution of natural language meaning. Instead of being topicalized, however, they were dispensed with all the more decidedly as exactly those problems which could not - or not yet - be solved in precise terms anyway. Structuralists have therefore been concerned with the question of how the lexical meanings of words - rather than being reconstructable from propositions relating language terms to extra-lingual entities - might be understood as being intra-lingually related to one another, constituting a (syntagmatically and/or paradigmatically) structured system of overlapping sub-systems (lexical fields) which organize the world as a universe of potential discourse. According to structural theory, the meaning of each term holistically depends on the position it occupies in that system. It is argued that - although the terms' references may be indeterminate, varying with their psycho-, socio-, modal-, and time-dependent con- and co-texts - the position of each term relative to another in any of these sub-systems will nevertheless be defined with precision. The idea of `structural determinacy' as opposed to `referential vagueness' cannot only be traced in the works of linguists like DE SAUSSURE, HJELMSLEV, and WEISGERBER down to COSERIU, GREIMAS, HALLIDAY, and LYONS, but has also inspired scholars from non-linguistic disciplines where - as in the ethnosciences or in cognitive and experimental psychology - the notions of `schema' and `prototype' were developed to cope with a range of varying but similar instances. The linguistic models' methods, and metaphors - however fertile and influential for some time and/or discipline - were based mainly upon intuitive introspection and the questioning of test persons. These approaches, which on the whole are abstracted from the pragmatic frame that any real communicants' language usage constitutes, apparently did not achieve either the theoretical consistency or the methodological objectivity that an empirical theory of cognitive and communicative processes calls for. Thus, the influence of structuralistic ideas in modern linguistics and its semantic theories was on the decrease until recently.
Unlike referential and structural notions of meaning, procedural semantic theory is primarily an instrumental approach, not an analytical or descriptive one. The relational but basically static structures established by both, the truth-functional analysis of denotating sentences, or the evaluating description of connotating word fields, are superseded by an essentially dynamic approach which identifies meaning with the execution of goal oriented procedures. These are activated by language terms and operate on and simultaneously modify conceptual structures that allow for the mapping of denotational and connotational notions of meaning. Represented either as conceptual cores (prototypes) determined by sets of functional and/or perceptual descriptors (schemata), or as linguistically labeled concepts (nodes) that are connected by identifying relations (links) between them, these model constructions of both the `memory'-type as developed in cognitive psychology and the `network'-type as advanced in artificial intelligence research converge on the procedural character of what meaning, and hence, cognition and comprehension, constitutes. Complex formal deduction as well as contents-guided inferencing involved in these processes are reconstructed by combinations of some apparently fundamental operations which psychologists have employed and tested in concept processing experiments of sorting, matching, and attainment tasks, and which information scientists have implemented and simulated in dynamic knowledge representation systems as storage, identification, and retrieval procedures. Although features of the descriptor-type and the network-type models were seminally combined by QUILLIAN and COLLINS, the gulf subsisting between psychology and artificial intelligence is only about to be narrowed and eventually perhaps be bridged on the basis of works by scholars like JOHNSON-LAIRD, MILLER, and ROSCH on the one hand, and MINSKY, SCHANK, and WINOGRAD on the other. Recent developments show that learning and the acquisition of knowledge as well as analogical reasoning as opposed to logical deducting processes become focal areas of current research to which linguists might contribute to a greater extent than has hitherto been realized.
5. Two fundamental problems stand out: How do people map natural language strings into a representation of their meaning? How do people encode thoughts into natural language strings? Because of a purported interest in the purely formal properties of language, linguists have consciously avoided both of these naturalistic problems. The second question seems, on the surface, to be closer to a linguist's heart. But linguists treat generation as a problem of determining whether a string is grammatical, i.e., whether it can be generated by the grammar they have set up. A grammar that generates natural language strings would be interesting and useful of course, if, and this is a big `if', it started at the right place. Linguists tend to start their grammars at the node S (for sentence). People, on the other hand, start with an already well-formed idea (or the beginnings of an idea) that they want to express. Linguists thus wind up concerning themselves with considerations of semantics at the level of `Can I say this string? Will it mean something?' People already know what they want to say and that it is meaningful. [4].
Views like these as advanced by SCHANK and ABELSON here, and others elsewhere, are still to be assimilated in linguistic semantics. Having been absorbed from some time in responding to challenges from outside by either hasty integration or ill-considered segregation, linguists will hopefully regain productive autonomy through the interdisciplinary status recognized for their discipline as anticipated by FILLMORE:
I think that everyone in linguistics and language research sees a need for an integrated view of language structure, language behaviour, language comprehension, language change, and language acquisition. I suspect that what strikes me as the current Zeitgeist in language research offers material to meet this need, though some of it is still somewhat hidden; and I keep getting the feeling that sooner or later it is going to be possible for workers in linguistic semantics, anthropological semantics, cognitive psychology, and artificial intelligence - and may even be language philosophy - to talk to each other using more or less the same language, and thinking about more or less the same problems [5].
Although one may have one's doubts as far as a common language is concerned, there is every reason to be positive in respect of the problems. Scholars active in the field are tackling - and sometimes solving - `real'-world problems instead of theoretical ones employing databased instead of formal procedures to produce testable findings instead of affirmative results. Their approaches show the rapidly growing interest that different disciplines are developing towards the semiotics of language systems in general and a related kind of applied research in particular.
This omnibus is intended to reflect - and perhaps even to intensify - this tendency. Thus, contributions were invited primarily - but not exclusively - from researchers known to be working on semantically oriented studies in linguistics, artificial intelligence, cognitive psychology, and related disciplines that are concerned with the development of operational and/or empirical, i.e. descriptive, quantitative, statistical, experimental, or simulative methods and models in their respective subject areas, rather than with formal theory and abstract model construction alone.
From the number of contributions received, twenty original papers written by twenty-six authors from nine countries were selected for publication. Covering a wide range of topics which overlap in several respects, the papers apparently deal with a few central issues that the arrangement of this collection tries to reflect. Volume One comprises some examples of Empirical Semantics meeting Linguistic Theory (BALLMER), Artificial Intelligence (BERRY-ROGGHE/ZIFONUN; CERCONE/GOEBEL; KNAUS; MARINOV; STEELS), and Fuzzy Theory and Application (UMANO/MIZUMOTO/TANAKA; WAHLSTER; ZADEH; ZYSNO). Volume Two contains examples of Empirical Semantics meeting Linguistic Application (ALLEN; DORFMÜLLER-KARPUSA/PETÖFI; KITTREDGE), Semiotic Models (NADIN; NOWAKOWSKA), Cognitive Theory and Memory Structure (HUBER; PELLETIER; RIEGER; SHOW/GAINES), and Catastrophe Theory (WILDGEN).
I would like to thank all contributors for their collaboration, in particular those who - having submitted their papers before the first deadline nearly two years ago - did not lose patience since. The same applies to the chief editor of this series, as well as to the secretary and the typists involved.
| Aachen, August 1981, | Burghard B. Rieger |
[1] RUSSELL, B. (1927): An Outline of Philosophy, London: Allen
&
Unwin Publishers, p. 33
[2] SAPIR, E. (1931): "Conceptual Categories in Primitive
Languages",
Science 74 [quoted from Schaff, A.: Language and Cognition,
New York: McGraw-Hill 1973, p. 187], p. 578
[3] WITTGENSTEIN, L. (1958): The Blue and Brown Books, New York/London: Harper & Row Colophon 1965, pp. 24-26