Our Neural Auto-correct For Interpreting Ambiguous Sounds

Your brain has a built-in “auto-correct” feature that you engage when re-interpreting ambiguous sounds, according to new research.

The study sheds light on how the brain uses information gathered after the detection of an initial sound to aid speech comprehension. The findings point to new ways we use information and context to aid in speech comprehension.

“What a person thinks they hear does not always match the actual signals that reach the ear. This is because, our results suggest, the brain re-evaluates the interpretation of a speech sound at the moment that each subsequent speech sound is heard in order to update interpretations as necessary,”

explains lead author Laura Gwilliams, a doctoral candidate in the psychology department at New York University and researcher at the Neuroscience of Language Lab at NYU Abu Dhabi.

Altered Perceptions

According to Gwilliams, your hearing can be affected by context occurring up to one second later, without you ever being aware of this altered perception.

“For example, an ambiguous initial sound, such as ‘b’ and ‘p,’ is heard one way or another depending on if it occurs in the word ‘parakeet’ or ‘barricade. This happens without conscious awareness of the ambiguity, even though the disambiguating information doesn’t come until the middle of the third syllable,”

adds principal investigator Alec Marantz, a professor in the linguistics and psychology departments, and co-director of the Neuroscience of Language Lab, where the researchers conducted their work.

Volunteers come to the lab for the experiment.

Volunteers come to the lab for the experiment. The first step is to make a 3D digital model of their head.
Credit: Kate Lord/NYU

The researchers have created examples of these stimuli, which you can listen to here.

Context Is Everything

It’s well known that the perception of a speech sound is determined by its surrounding context — in the form of words, sentences, and other speech sounds. In many instances, this contextual information is heard later than the initial sensory input.

This plays out in every-day life — when we talk, the actual speech we produce is often ambiguous. For example, when a friend says she has a “dent” in her car, you may hear “tent.” Although this kind of ambiguity happens regularly, we, as listeners, are hardly aware of it.

“This is because the brain automatically resolves the ambiguity for us—it picks an interpretation and that’s what we perceive to hear,” explains Gwilliams. “The way the brain does this is by using the surrounding context to narrow down the possibilities of what the speaker may mean.”

In the study, the researchers sought to understand how the brain uses this subsequent information to modify our perception of what we initially heard.

Brain Replay

To do this, they conducted a series of experiments in which the subjects listened to isolated syllables and similarly sounding words (e.g., barricade, parakeet). In order to gauge the subjects’ brain activity, the scientists deployed magnetoencephalography (MEG), a technique that maps neural movement by recording magnetic fields generated by the electrical currents produced by our brain.

Electromagnets are attached to the participant’s head

Electromagnets are attached to the participant’s head and help determine where the head is located inside the machine.
Credit: Kate Lord/NYU

Their results yielded three primary findings:

  • The brain’s primary auditory cortex is sensitive to how ambiguous a speech sound is at just 50 milliseconds after the sound’s onset.
  • The brain “replays” previous speech sounds while interpreting subsequent ones, suggesting re-evaluation as the rest of the word unfolds.
  • The brain makes commitments to its “best guess” of how to interpret the signal after about half a second.

“What is interesting is the fact that this context can occur after the sounds being interpreted and still be used to alter how the sound is perceived,”

Gwilliams adds.

For example, the same sound will be perceived as “k” at the onset of “kiss” and “g” at the onset of “gift,” even though the difference between the words (“ss” vs. “ft”) come after the ambiguous sound.

Inside the helmet are 208 magneto-encephalography (MEG) sensors

Inside the helmet are 208 magneto-encephalography (MEG) sensors, which measure the magnetic field created when neurons fire.
Credit: Kate Lord/NYU

“Specifically, we found that the auditory system actively maintains the acoustic signal in auditory cortex, while concurrently making guesses about the identity of the words being says,” says Gwilliams. “Such a processing strategy allows the content of the message to be accessed quickly, while also permitting re-analysis of the acoustic signal to minimize hearing mistakes.”

The study was funded by the NYU Abu Dhabi Research Institute, the European Research Council, France’s National Research Agency, and the National Institutes of Health.

Laura Gwilliams, Tal Linzen, David Poeppel and Alec Marantz
In spoken word recognition the future predicts the past
Journal of Neuroscience 16 July 2018, 0065-18; DOI: https://doi.org/10.1523/JNEUROSCI.0065-18.2018

Image: Digitalarti/Flickr

Leave a Comment