Electronic Thesis and Dissertation Repository

Thesis Format



Master of Science




Ingrid Johnsrude


Effortful listening is experienced by listeners when speech is hard to understand because it is degraded or masked by environmental noise. Pupillometry (i.e., measure of pupil size) can detect effortful listening: pupil size increases when speech is degraded compared to when it is clear. However, the pupil responds to a range of cognitive demands, including linguistic challenges such as syntactic complexity. Here I investigate whether it responds to the need to disambiguate words with more than one meaning, such as ‘bark’ or ‘bank’. Semantic ambiguity is common in English, and previous work indicates that it imposes a processing load. We combine this with an acoustic challenge in a factorial design so the pupil response to these two types of challenge can be directly compared. I found main effects of noise and semantic ambiguity on the pupillary area, indicating that pupil dilation can reflect processes associated with semantic disambiguation as well as noise. Pupil size reflect demands imposed by ambiguity both in the acoustic form of words (i.e. due to degradation) and in word meaning.

Summary for Lay Audience

It is difficult to listen to speech when it is either spoken unclearly, or if there is background noise affecting our ability to hear it. This difficulty in hearing requires higher mental activity called cognitive load, or more work for our brain when it comes to understanding speech. The pupils of our eyes change its size depending on how hard our brain is working, and the size therefore tells us something about the level of cognitive load. The way we can measure pupil size is through a method called pupillometry. Pupillometry can measure a change in pupil size that relates to whether speech is heard clearly or is noisy and degraded. While this is a useful method, the pupil can change its size when listening to speech for reasons that are not related to the quality of the signal. Cognitive factors involved in comprehension also make our brains work hard, such as when we can hear a sentence containing words with more than one meaning, such as “The shell was fired towards the tank”. The correct meaning of the bolded words must be deduced based on the other words in the sentence. In this thesis, our goal was to compare pupillometry measures when a listener has to work to understand a sentence, when words are hard to hear, and when their meaning has to be deduced from context.