Günther Grewendorf begins the preface to his "Minimalist Syntax", which is an account of the current generative theory, with the sentence:
"Der kognitiven Linguistik geht es darum, jene abstrakten strukturellen Gemeinsamkeiten aller natürlichen Sprachen zu ermitteln, an denen sich die genetischen Grundlagen einer angeborenen Sprachfähigkeit erkennen lassen." (Grewendorf, 2002)
("The aim of cognitive linguistics is to determine those abstract structural traits common to all natural languages which allow the genetic bases of an innate language faculty to be recognised.")
One may ask why we should accept the opinion that abstract and not concrete common traits form the object of research in cognitive linguistics and why the genetic bases should not be stated but only be recognisable. The reason for these formulations is evident: If we ever take the statements of generative linguistics literally, or if we try to confront them with reality, they obviously go wrong. Critics have repeatedly pointed out that there is a problem - without much result.
Models of speech production
One would expect that models of speech production - like the classical one by Levelt (1989) and its "implementations" by Roelofs (for example 1997) - which traditionally take their bearings from psycholinguistic experiments, are closer to biological realities. However the problem is that they are only limited by their output in relation to experimental data. The internal structures of the models are not compared with biological facts. Thus it is possible that they use "nodes" (which are not neurons) in networks whose properties cannot be found and are not even searched for in real neural structures.
Connectionist models of linguistic processes appear explicitly with a biological outfit. The notion "artificial neural nets" leads - in spite of the attribute "artificial" - to the assumption that in this case real biological basics are observed. However every connectionist who has read his classics with care knows that there are profound problems in taking the step from a connectionist net to the biological reality. The "artificial neurons" exhibit only a superficial similarity with their biological prototypes, and it is easy to show that they have properties which are in conflict with biology. Even the structure of a connectionist model as a whole shows only a superficial resemblance with structures of the cortex.
It seems that theories which are known as wrong, are continued to be accepted and developed due to the lack of alternatives. In the long run this may not be a good strategy. It would be important to at least look for criteria which may define a framework for acceptable constructs in linguistic science.
It is obviously naive to first construct complicated "mentalist" or "cognitive" theories of language and only afterwards verify (or not verify at all) whether the brain - which is nevertheless accepted as the location of the knowledge of language and ongoing language processing - allows the assumptions used. The excuse that we do not know how the brain really functions is threadbare if we do not even try to access the relevant information.
Cortical linguistics works the opposite way. Metaphorically: not from the roof to the foundation but from the foundation to the roof. The foundation consists in the overwhelming abundance of existing and somewhat stable findings about cortical structures and processes. The use of this information for linguistic purposes is enabled by a scientifically founded simulation method, the results of which may be checked by any available empirical evidence.
It is to be expected that the expense needed to come to lasting and useful linguistic constructions is reduced in this way.
Related fields: Cognitive neuropsychology (of language), cognitive linguistics
Interdisciplinary connections: Neurophysiology, psychiatry, neurology, informatics (computational linguistics), psychology (developmental psychology, psychology of language), musicology, philosophy (theory of cognition, philosophy of science)
Some keywords (alphabetically): Acquisition of concepts, Alzheimer type dementia, amnesia, aphasia, artificial neural nets, connectionism, coherence, dissociative personality disorder, hierarchies of models, language acquisition, language understanding, memory, mental lexicon, model building, multiple sklerosis, reading, series of models, Parkinson's disease, phonetics, phonology, prototype semantics, prototype theory, schizophrenia, semantics, short-term memory, speech production, stuttering, syntax, text linguistics, text understanding, thinking
Literature cited within this text:
Grewendorf, G. (2002) Minimalistische Syntax. Tübingen, Basel: Francke
Levelt, W. J. M. (1989) Speaking. From intention to articulation. Cambridge/Mass.: MIT Press
Roelofs, A. (1997) Syllabification in speech production. Evaluation of WEAVER. Language and Cognitive Processes 12, 657-693
Last update 9/15/2003