Language production is the process of translating thoughts into linguistic expressions — spoken, written, or signed. While comprehension has received more research attention, production involves its own set of complex challenges: selecting the right words, assembling them into grammatically correct structures, computing the sound form of each word, and coordinating the precise motor movements of articulation, all at a rate of 2-4 words per second in fluent speech.
Levels of Production
Willem Levelt's (1989) influential model proposes several processing stages. Conceptual preparation formulates the pre-linguistic message — what the speaker wants to say. Lexical selection (lemma access) retrieves the appropriate words based on their meanings and grammatical properties. Phonological encoding retrieves the sound forms of words and assembles them into a phonological plan. Phonetic encoding translates the plan into motor commands. Articulation executes the motor commands to produce speech.
Speech Errors
Speech errors — slips of the tongue — provide a window into the production process. Exchange errors (saying "a leading list" instead of "a reading list") reveal that words are selected independently from their positions in the sentence. Blend errors ("a grizzly + ghastly → grastly") reveal competition between alternative words. Phonological errors (saying "fleaky" instead of "freaky") reveal the assembly of phonological segments. Victoria Fromkin and others showed that errors are not random but systematic, obeying the phonological and grammatical constraints of the language.
The tip-of-the-tongue phenomenon provides evidence for the two-stage model of lexical access. In TOT states, speakers have successfully selected the lemma (they know the word they want and some of its properties) but fail to retrieve its phonological form. This dissociation between semantic/syntactic access and phonological access supports the independence of lemma selection and phonological encoding stages in production.