Blurb:

Relational Network Theory (RNT) assumes that the connections supporting linguistic and otherwise cognitive processes operate via loosely-yoked spreading activation. As other theories, RNT focuses on describing normal language processing, including operations which lead to successful production of what the speaker intends to say. However, as shown in this chapter, it can also account for multiple types of speech errors which abound in (non-pathological) verbal production. A five-stratum model is used leading from high-order cognition to speech output, with each stratum corresponding to different kinds of basic elements (emes). Input to the emes of a particular stratum is non-linear, but typically the emes are linearized by the tactic pattern, producing fully-ordered output to sound. However, because the linguistic system is biological in nature, asynchronies may occur if some of its parts take random rest periods. Moreover, imbalances may result when different potential realizations compete with one another. In either case, a speech error may result. Drawn from a large corpus collected in Polish and English, examples of the four most common spontaneously produced error types (timing errors, tactic pattern errors, unintended blends, substitutions) are presented and explained in RNT terms. The cumulative effect is to reify the five-stratum RNT model supported in other analyses with a basic distinction between realizational/semiotic relations and tactic patterns on each stratum.