Electronic Thesis and Dissertation Repository

Thesis Format



Master of Science




Johnsrude, Ingrid S.


Our brains are proficient in learning recurring structures in the environment, in order to optimize perceptual inferences based on relevant information in a stochastic input. Sensory information is multi-dimensional, and the relationship between sound dimensions may be, in itself, a source of information. Many sounds in our environment covary dynamically, and these covariances may be learned, and therefore shape our perception, through exposure to them in our natural environment. In the present study we investigate how natural (long term), and experimental (short term), learning of statistical regularities in sounds may shape our ability to categorize them (Experiment 1) and to perceptually segregate them more easily from target speech (Experiment 2). Our results indicate that sounds that obey naturalistic pitch-speed relationships are more easily categorized than those that violate these expectations. However, these benefits did not translate into greater segregability of these naturalistic patterns from speech, although my method may have not been sufficiently sensitive to such effects. These findings highlight the ways in which long-term life experience may influence our auditory perception.

Summary for Lay Audience

We live in a very complex sensory world. All around us, we are constantly exposed to a variety of sights and sounds that compete for our attention. And yet, even if we don’t notice it, our brains are excellent at identifying patterns to help us better make sense of our environment: Light usually comes from above, so dark patches in the ground are often shadows; and if you hear loud thunder, the storm is probably pretty close! Not only that, but many properties of a stimulus may vary together (or covary) over time. For instance, many sounds in our environment seem to show a positive covariance between pitch and speed: As machines ‘power up’ and their parts move faster, they sound higher pitched, and faster speech is usually higher pitched. The relationship between these sound properties is quite strong and can even result in illusory effects where sounds that are played at a higher pitch tend to also sound ‘faster’ even when that is not physically true. Learning these covariances may be a valuable tool for providing extra ‘redundancy’ in environmental information, allowing us to infer additional information about sounds even in noisy or ambiguous listening conditions.

My project aims to investigate how sound patterns that are learned over time influence our perception. More specifically, here we look at how this long-term familiarity with positive pitch-speed covariances may result in: (1) more accurate and faster categorization of sounds that obey these rules; and (2) more effective segregation of noise that matches these rules from target sounds, allowing us to better pay attention to more relevant information.

This project will ultimately contribute to a better understanding of the ways in which long-term life experiences may shape our perception of the world around us.

Creative Commons License

Creative Commons Attribution 4.0 License
This work is licensed under a Creative Commons Attribution 4.0 License.