From sounds to language: UCSF scientists study how the brain represents speech

Edward Chang, MD, professor and chair of the Department of Neurological Surgery at UC San Francisco, is marking the 11th year of funding for his research into how the brain processes speech with the renewal of his National Institutes of Health (NIH) Research Project grant.

“We've learned a lot about how the elements that are important in speech are encoded by electrical activity in neurons,” Chang said. “But some of the biggest questions are still unanswered.”

His lab focuses on the superior temporal gyrus, a region in the brain important in speech perception. He and his colleagues previously reported that populations of neurons in this part of the brain respond to the acoustic-phonetic features – the acoustic and articulatory parameters of the physical sound properties – in vowels and constants.

Edward Chang, MD, sitting at a table with graduate student  Ilina Bhaya-Grossman and David Moses, PhD.
Neurosurgeon Edward Chang, MD, is chairman of the Department of Neurological Surgery at UCSF. Dr. Chang meets with graduate student Ilina Bhaya-Grossman and postdoctoral scholar David Moses, PhD. Photo by Barbara Ries​.

Linguists describe phonemes as the smallest unit of sound in any language that distinguishes one word from another. Combinations of these phenomes generate all the individual vowel and constant sounds that make up syllables.

But researchers still do not know how the brain makes sense of the different speech sounds coming together during continuous speech.

ECoG array
ECoG array. Photo by Noah Berger.

Chang says that since pauses do not often mark the boundaries between words, the brain must interpret them instead. “How consonants or vowels get bound into single words is really unclear,” he said.

Now, he and his team are trying to get a better understand of how the brain integrates the information about speech sounds over time. By using high-density electrocorticography (ECoG) arrays, his lab can generate high resolution electrical recordings from the superior temporal gyrus and other regions of the brain involved in processing speech.

Chang and his colleagues are also interested in how the brain encodes information about other aspects of speech, like rhythm and intonation.

For example, people often change the pitch of their voice while speaking to stress certain words and create alternate meanings for the same sentence. In a prior study, Chang’s research group found that a population of neurons in the superior temporal gyrus detect these pitch changes.

Some speech sounds are also intrinsically louder than others, and the contrast between constants and vowels influences the cadence of speech.

Chang’s lab has demonstrated that some neurons in the back of the superior temporal gyrus respond to the onset of speech. Other neurons in the middle superior temporal gyrus respond to changes in loudness at the level of individual syllables.

He also wants to figure out how specialized this part of the auditory cortex in the brain is for recognizing speech over music. These studies will help scientists identify whether the brain encodes auditory information from music and speech in similar ways.

“What we’re seeing so far is it’s not as simple as all music or all speech,” he said. “There's a subset of the auditory neurons that actually can process both, and others that are really specialized for speech or music.”

No single cell or group of cells selectively responds to the individual words in speech. The challenge for his group now, Chang says, is coming up with alternative models to describe how the brain represents words.