Can natural language be reduced purely to syntax?

In general, symbolism is a way of creating a map between a relatively small collection of objects, called the signs, and a much larger collection, which are the objects found in human experience.

The fact that the collection of signs is enormously smaller than the objects in experience is what makes symbolism useful. Signs serve as pointers or handles, reducing the complexity of experience so that it may be conveniently manipulated.

Symbolism generally works so effectively that we tend to forget that experience is immensely richer than the collection of signs. This means in particular that language can never truly communicate experience, although we habitually fall into the lazy fallacy which implies that it can.

This fallacy conceives of language as a conduit through which packets of experience can be transported and systematically unpacked at the receiving end. In fact, however, language can point only to experience that speakers already share.

Language is biological, not mechanical. It's a kind of meta-life that has co-evolved with humans.

No matter how many verbal descriptions a deaf person memorizes about a piece of music, he or she cannot experience the music through those descriptions -- although if the descriptions were sufficiently detailed, no one could prove otherwise.

A professor may give a brilliant lecture, but the students still may not understand the subject. When a student asks for clarification, the professor might reply, "I explained that already in the lecture!" The professor's response reveals an implicit belief that language conveys mental experience, when in fact the students cannot understand speech that points to what they do not in some sense already know. Such pointers are invalid.

In the miraculous game that we play with language, we use imagination cooperatively to create meaning and new mental experience by weaving together strands of experience that we trust to be already held in common by all interlocutors involved in the interaction. Speakers play a guessing game that is so rapid that they are usually unaware of being engaged in making a continuous stream of judgments and predictions.

Language, as practiced by human beings, is a profoundly creative process. Meaning is synthesized from human experience, not from abstract, syntactical properties inherent in language.

Therefore, although computers can be designed to process the rules of language to any desired degree of precision, unless computers are capable of having human experience the question of whether computers understand meaning is meaningless.

Human languages, or natural languages, are not at all the same as computer languages. A computer language systematically breaks up complex tasks into a series of extremely simple, perfectly defined computational steps. Natural languages do the opposite. They sweep away the vortex of tangled, endlessly varying experience, and replace it with fragments. Natural language depends on the mindís power to create stories that tie those fragments together.

It would be misleading to look at the study of computer "language" as a useful intellectual framework for studying natural language, although many people -- including linguists -- fell into that trap during the last century. Likewise, it is misleading to model human cognition in mechanical or computational terms.

Michael Webb, 2000

home ]