In the classical theory, you have two choices for characterizing set membership: you can predict the members (by precise necessary and sufficient conditions, or by rule) or you can arbitrarily list them, if there is a finite list. The only choices are predictability (using rules or necessary and sufficient conditions) and arbitrariness (giving a list). But in a theory of natural categorization, the concept of motivation is available. Cases that are fully motivated are predictable and those that are totally unmotivated are arbitrary. But most cases fall in between-they are partly motivated.
- Lakoff (1987), a pag.146

Something in language or thought is motivated when it is neither arbitrary nor predictable. Motivated phenomena include category extensions [...], polysemy [...], related grammatical constructions [...], most idioms [...] etc. In natural language, motivation seems to be more the norm than the exception. Motivation is not the kind of phenomenon that algorithms were designed to characterize.[...] Motivation is a central phenomenon in cognition. The reason is this: It is easier to learn something that is motivated than something that is arbitrary. It is also easier to remember and use motivated knowledge than arbitrary knowledge.
- Lakoff (1987), a pag.346