Millisecond windows of time may be key to how we hear, study finds

(Press-News.org) You have 20 minutes of spare time, but the new episode of your favorite podcast is a few minutes longer. No problem; you can increase the listening speed and fit in those extra minutes.

Phew.

What happens when you listen to speech at a different speed? Neuroscientists thought that your brain may turn up its processing speed as well. But it turns out that at least the auditory part of the brain keeps “listening” or clocking in at a fixed time. That is the key finding of new research out today in Nature Neuroscience. The research was led by Sam Norman-Haignere, PhD, assistant professor of Biostatistics and Computational Biology, Biomedical Engineering, and Neuroscience at the Del Monte Institute for Neuroscience at the University of Rochester, in collaboration with researchers at Columbia University, including Principal Investigator Nima Mesgarani, PhD, of the Zuckerman Institute, and Menoua Keshishian, who completed his PhD in Electrical Engineering in his lab.

“This was surprising. It turns out that when you slow down a word, the auditory cortex doesn't change the time window it is processing. It's like the auditory cortex is integrating across this fixed time scale,” said Norman-Haignere, the study's first author, who started the study as a postdoctoral researcher at Columbia. “One of the key goals of this kind of research is to build better computational models of how the brain processes information in speech, which will increase our set of scientific tools and ultimately help us understand what goes awry when someone has difficulty understanding speech and language processing.”

The Complexities of Understanding and Modeling Speech

The auditory cortex, which consists of several layers and regions, is the brain area responsible for processing and interpreting sounds. Researchers know there are multiple regions in the brain that process speech—the primary auditory cortex, the secondary auditory cortex, and language areas beyond the auditory cortex. A fundamental understanding of how each region works and the hierarchy between and within these different regions is not well understood.

Understanding the complexities of the brain has been assisted by the development of computational models. These computer models use mathematical formulas or algorithms to understand sound and predict neural responses and human behavior.

The authors of this study used computer models to test whether their research method would distinguish between their two hypotheses: does the auditory cortex integrate information across speech structures—for example, words—or time? It turned out that some of the computer models learned to integrate across speech structures, unlike the auditory cortex. This finding was helpful in part because it helped to validate the methods the authors were using to study structure and time.

Accessing the Human Brain

Neuroscientists are typically limited in the types of neural data they can record from the human brain.  Electroencephalograms or EEGs provide researchers with the brain’s electrical activity read from the scalp, which is far away from the actual cells that produce this activity. Functional MRIs measure blood flow in the brain, which is an indirect measure of brain activity. Both tools have transformed our understanding of human brain function and disease; however, neither method is able to record spatially and temporally precise neural activity.

The researchers worked with epilepsy patients at NYU Langone Medical Center, Columbia University Irving Medical Center, and University of Rochester Medical Center to measure precise neural activity from inside the human brain. They worked with patients who were admitted to the hospital for epilepsy monitoring. As part of their monitoring, electrodes were temporarily implanted inside their brains so medical doctors could better determine the area of the brain where their seizures came from. These electrodes measure electrical responses right next to where neurons are active, providing much higher precision than standard methods such as EEG and fMRI.

The recruited participants were tasked with listening to a passage from an audiobook at normal speed, and then they were played the same passage at a slower speed. The researchers thought they might see a change in the neural time window that varied with the speed of speech. However, the differences they observed were none to minimal, indicating the fundamental unit of processing is physical time—for example, 100 milliseconds—and not speech structures such as words.

"This finding challenges the intuitive idea that our brain's processing should be yoked to the speech structures we hear, like syllables or words,” said Mesgarani, a senior author of the study and an associate professor of Electrical Engineering at Columbia. “Instead, we've shown that the auditory cortex operates on a fixed, internal timescale, independent of the sound's structure. This provides a consistently timed stream of information that higher-order brain regions must then interpret to derive linguistic meaning.”

“The better we understand speech processing, the better we think we'll be able to understand what is causing deficits in speech processing,” said Norman-Haignere. “One thing that is exciting about this line of work is that there are many people who have been studying hearing, and many people who have been studying language, but your brain needs to somehow transform the sounds that reach your ear into words, phrases, and sentences. So, figuring out how the brain goes from something more sound-based to something more language-based, and how to model this transformation, is an exciting space that we're working in.”

Other researchers include Guy McKhann, and Catherine Schevon of Columbia University, and Orrin Devinsku, Werner Doyle, and Adeen Flinker, NYU Langone Medical Center. The research was supported by the National Institutes of Health and a Marie-Josee and Henry R. Kravis grant.

 

END