Skip to content

How Does the Brain Transform Thoughts into Spoken Words?

Key findings

  • The posterior-superior temporal gyrus (P-STG) region of the brain carries out contrasting linguistic computations in speech versus comprehension
  • Computations for selecting appropriate word forms during speech production are localized in the P-STG region of the brain
  • Aberrant neural recordings from the P-STG region of the brain during cortical stimulation correspond with a deficit in appropriate word form selection

Evidence from previous studies suggests that the posterior-superior temporal gyrus (P-STG) of the brain may be a core language hub, but what else is happening in the brain as a person’s thoughts transform into spoken words? And is speech altered in patients who have undergone surgical resection of lesions from the P-STG regions in their brain?

Massachusetts General Hospital researchers Ziv Williams, MD, a neurosurgeon at Mass General, and lead author Daniel Lee, a forth-year medical student at Harvard Medical School, reported in Nature Communications their team’s findings from a study aimed at better understanding these linguistic computations. The team devised word-production tests based on functional morphemes to investigate the division of labor in this aspect of speech.

The researchers assessed several mental processes from word production to articulation. A word-production test consisted of presenting a preamble (auditorily) to the subject followed by visual presentation of a target word to be modified and articulated by the subject. The control condition tested basic sensory perception and articulation by presenting a compatible target word. The test asked subjects to manipulate incompatible target words to assess four computation categories:

  • Recognition and transformation of the target word (phonetic transformation)
  • Retrieval of information from memory (lexical retrieval)
  • Comprehension and response using appropriate semantics (contextual processing)
  • Manipulation of a target word to select an appropriate inflection (morpho-syntactic processing)

Speech begins with forming a thought, and it ends with articulation of the thought. What occurs in between is a series of computations to select words and place them in the appropriate linguistic order to convey the intended meaning. Selection of the appropriate linguistic form of a given word may involve adding the suffix “ed” to the word “talk” to convey past tense. In this case, the suffix “ed” is a “functional morpheme,” an inflection or unit of speech that modifies the meaning of a word but does not have meaning by itself. Such elements of speech are critical for precisely conveying intended meaning.

To determine which computations are carried out in the P-STG, the Mass General team compared each subject’s test performance with and without cortical stimulation, a method for reversibly inhibiting neural processes. The cortical stimulation was delivered to various sites in and around the P-STG and was synchronized to begin with visual presentation of the target word and to end before speech production. Neural recordings were simultaneously collected from implanted intracranial electrode arrays.

The team also assessed patients who previously underwent surgical resection of lesions in the P-STG. Although pre-surgical function in these patients had not been evaluated for comparison, the team compared the data to that collected from control patients who had undergone surgical resection of lesions in another region, the anterior temporal lobe (ATL).

The team’s findings indicate that morpho-syntactic processing (measured as functional morpheme production) is a discrete step in the computations involved in speech production. Cortical stimulation of the P-STG selectively inhibited morphological morpheme production. They also reported a deficit in functional morpheme production in 2 patients with lesions in the gray matter of the P-STG region. Neural recordings from 7 sites within the P-STG corresponded with a deficit in functional morpheme production. The team also narrowed the timing of planned inflection to 1-2 seconds before articulation of the word.

This study revealed contrasting roles of the P-STG in speech production versus comprehension, which requires a related but different set of computations. The findings indicate that inhibition of the P-STG during speech production affects only morpho-syntactic processing, while previous studies of processing during comprehension show that lexico-semantic as well as syntactic processes engage this same region. Together these findings provide a more detailed view of the language circuitry and division of labor in the brain.

Learn more about Otolaryngology at Mass General

Learn more about the Neurosurgery Department at Mass General

Related topics


Many women who experience intimate-partner violence sustain at least one traumatic brain injury, which can lead to problems with memory, learning and cognitive flexibility. Mass General researchers searched for neural mechanistic evidence of an association between these TBIs and cognitive dysfunction.


Two very different risk factors, social isolation and low levels of brain-derived neurotrophic factor, are associated with increased risk of stroke and dementia. Mass General researchers investigated whether social relationships reduce stroke and dementia risk through the BDNF pathway.