A Neurocognitive View on Speech Production

Producing language is a complex neurocognitive process that requires a whole system of functions.

Our language system consists of four different levels that vary in their degree of subtlety. The syntactic level is the “coarsest” level of the language system and is located at the sentence level while the phonological level is the most fine-grained level, which contains the smallest units of language – the “phonemes”. Around 100 phonemes are enough to build all languages in the world. But how are they integrated in our brain to produce meaningful sentences?

A well-known model by Levelt (1989) describes how speech is produced from a rough, conceptual idea to the movement of the speech organs. At the beginning, the utterance is in a conceptual phase: in this phase it is planned which contents are to be uttered and in which order. For this purpose, the uttered content is retrieved from short- or long-term memory in the form of declarative or procedural knowledge units (for more information see one of our recent blog entries). The resulting “preverbal message” is formulated in the next step, where the conceptual structure is translated into a linguistic structure. For this purpose, the concepts are translated into a suitable linguistic and grammatical form. This forms the surface structure of the utterance, which is then stored in the syntactic buffer. Based on phonological and morphological information in the lexicon, an articulatory plan is created. The articulatory system converts this plan into commands for the articulatory muscles.

Especially the mental lexicon, which Levelt (1989) hypothesized, has been taken up again and again in many other models and contexts (see Damsio et al. 1996). According to Levelt (1989) the mental lexicon is a relevant component of the language system for assigning words to the preverbal concept. It specifies properties of a word, such as grammatical category, pronunciation, and meaning. The way retrieval from the lexicon works reveals something about its structure. There are two prominent lexical retrieval phenomena. First, the neighborhood effect: words that are phonologically or orthographically very similar influence each other. Second, the frequency effect: frequently occurring words are available faster than less frequent words which indicates a flexible system that can adapt to needs. These effects suggest an efficient structure of the lexicon.

A structure of the mental lexicon, could also be realized at the semantic level. Regarding the organization of semantic knowledge, Collins & Loftus (1975) recognized that a semantic lexicon, organized as a network, can explain many cognitive psychological phenomena such as priming effects. The authors hypothesize that based on shared properties, two things that have many properties in common are arranged closer together than two things that share fewer properties. Activation from one object will spread to its neighbors, with decaying activity with growing distance.

Evidence for this model can also be found in the semantic verbal fluency task. In this task, the test person has to list as many things of a semantic category (e.g., animals) as possible within a given time. It turns out that semantic clusters (e.g., dog, cat, mouse / giraffe, elephant, cougar) overlap with temporal clusters (Linz, Tröger, Alexandersson & König, 2017). These findings imply that it is faster to generate a subsequent word that is semantically similar than a dissimilar one, supporting the semantic lexicon structure proposed by Collins & Loftus (1975).

Some pathologies are specifically known for impaired  retrieval processes from the semantic lexicon. (If you want to find out more about how semantic clusters differ between healthy people and people with Alzheimer’s: Click here).