Meta-Language and Learning How to Learn

One of the dilemmas faced by linguistics, semanticists and philosophers in recent years who study languages and their use is that one must use language in order to discuss language. In discussing the inability of most languages to describe ongoing, organic processes, for instance, one must make use of a specific language which is itself limited, static and unyielding to an accurate and vivid description of these dynamic processes. This paradoxical condition concerning the use of language to talk about language was addressed by Bertrand Russell (Whitehead and Russell, 1910) in his Theory of Logical Types.  This noted philosopher (and social activist) observed that any system, words, or taxonomies that are being used to describe a particular collection of objects, experiences and so forth, cannot itself be a part of this collection. In other words, we must somehow move outside of a system when we are trying to describe it.

Alfred Korzybski (2024) has similarly noted that a map of a territory is not itself the territory.  A map of Seattle Washington, for instance, is not Seattle, but only a map. Similarly, the word “cat” cannot scratch you. The word “chair” is not actually a chair, but only a representation of this type of furniture. These examples are obvious, and even absurd. Yet, often we find ourselves in the difficult and puzzling situation of not being sure whether we are addressing the real problem or only a representation of the problem.

We encounter people (often ourselves) who confuse the concept (e.g. “superego”) with the reality that this concept is supposed to represent. Thus, we search for the location of the superego in the cerebral cortex, rather than accepting the concept as a useful metaphor to describe a complex set of human activities and experiences. We must somehow be able to distinguish between the map and territory, between words and things, between “first-order” language that describes things and “second-order” language that describes how we use language.

In an entirely different field, experimental psychology, a similar problem was confronted in the 1940s and 1950s by Edward Tolman (1948). Animals which were being run through a maze not only learned how to execute this particular maze more rapidly and with fewer errors over time, they also were able to run through a new maze more rapidly and with fewer errors. Apparently, these animals learned not only how to run a specific maze, but also learned how to run mazes in general. This same phenomenon has been observed in the learning of many other types of tasks and puzzles by human as well as nonhuman subjects. This phenomenon has been labeled “the establishment of a learning set” or, more simply, “learning how to learn.”