It’s the bane (or savior) of every holiday cocktail party: the constant background buzz that makes it hard to hear the guy from accounting droning on. Now, researchers have taken a step toward understanding how people’s hearing systems pick out one sound among many.
The key is repetition, according to experiments reported the week of January 3 in the Proceedings of the National Academy of Sciences. Hearing a sound again and again in different auditory mixtures allows a person to separate it out and recognize it, the work suggests.
“Repetition alone may be a cue for sound segregation,” says Christophe Micheyl, a researcher in auditory perception and cognition at the University of Minnesota in Minneapolis, who was not involved in the study.
Scientists have long wondered how the hearing system can recognize individual sounds when natural environments rarely present a single sound in isolation to be learned. Researchers have accordingly dubbed it the “cocktail party problem.”
“The problem is that we think you need knowledge of individual sound sources in order to take a mixture and separate it, but if all you have are mixtures to begin with it’s not clear how we do that,” says Josh McDermott, an auditory researcher at New York University and leader of the team that performed the new work.
Experiments on the cocktail party problem are tough because people are already familiar with most natural sounds. For instance, playing a recording of a bottle dropping and a dog barking at the same time allows a person to draw on prior knowledge of each of those sounds, says McDermott.
To get around that, McDermott and his team synthesized new sounds by combining acoustic properties from spoken words and animal vocalizations with white noise. The researchers played these synthetic sounds in various combinations to college students, then played one sound back in isolation and asked if it had been in the mix the students had just heard.
Hearing two sounds played together once didn’t help the subjects improve their identification over chance, even if the researchers played the target sound first so the students knew what they were supposed to be listening for. But then the scientists played a sequence in which the target sound was mixed repeatedly with different “distractor” sounds. Suddenly the students became much better at picking out the target sound, perhaps because they heard it repeating through the different mixtures.
“What’s new about the study is it suggests one way the auditory system could get off the ground — that is, using sounds that occur repetitively,” says McDermott.
The work could help explain why certain animals, like birds and monkeys, often repeat calls in rapid succession, he adds. The creatures might be doing more than just trying to make sure their message is heard — they might be trying to help others pick out their sound in a noisy environment like a rainforest.
The findings make sense in light of other work that suggests repeating meaningless sounds embedded in noise allows people to create memories of those sounds, says Shihab Shamma, an electrical engineer at the University of Maryland in College Park who studies speech and hearing.
The new work might also open up new avenues of research using synthesized sounds, Shamma adds. Most earlier work on the cocktail party problem used only simple sounds, he says, but “this paper offers a way forward with more complex, natural-like stimuli.”
Indeed, McDermott is already planning more experiments using his novel sounds to study how familiarity with a sound affects how it is heard in a mix.
TRY IT YOURSELF
Listen to this sequence of sounds, and see if the final one in the series (the one that comes after the pause) was part of what you heard earlier. As the number of different sounds in the mixture increases, it should get easier to hear that the final sound was indeed included in the earlier sequence.
Credit: Josh McDermott
Target sound mixed with one sound:
With two sounds:
With three sounds:
With five sounds:
