Perception of Speech in Context by Listeners With Healthy and Impaired Hearing

Overview

Recognition of speech sounds is accomplished through the use of adjacent sounds in time, in what is termed acoustic context. The frequency and temporal properties of these contextual sounds play a large role in recognition of human speech. Historically, most research on both speech perception and sound perception in general examine sounds out-of-context, or presented individually. Further, these studies have been conducted independently of each other with little connection across labs, across sounds, etc. These approaches slow the progress in understanding how listeners with hearing difficulties use context to recognize speech and how their hearing aids and/or cochlear implants might be modified to improve their perception. This research has three main goals. First, the investigators predict that performance in speech sound recognition experiments will be related when testing the same speech frequencies or the same moments in time, but that performance will not be related in further comparisons across speech frequencies or at different moments in time. Second, the investigators predict that adding background noise will make this contextual speech perception more difficult, and that these difficulties will be more severe for listeners with hearing loss. Third, the investigators predict that cochlear implant users will also use surrounding sounds in their speech recognition, but with key differences than healthy-hearing listeners owing to the sound processing done by their implants. In tandem with these goals, the investigators will use computer models to simulate how neurons respond to speech sounds individually and when surrounded by other sounds.

Participants in this study listen to speech played at a comfortable volume and respond by indicating what they heard, either in open-ended form or by choosing among a set of options displayed on a computer. They are seated inside a sound booth and complete the task at their own pace, with little intervention needed from the experimenter. Upon arrival at the lab, participants are given a brief description of the topic of the research (how earlier sounds influence our perception of later speech sounds) and are presented with a detailed informed consent form. Their demographic information is collected and then the experiment is demonstrated. Breaks are offered between testing blocks, which last about 10-15 minutes each. The main differences in the protocol consist of the various stimulus manipulations, which are designed to specifically control aspect of the voice that the participant hears. For example, the sound can be manipulated to emphasize higher or lower frequencies, or be spoken relatively slowly or quickly, or manipulated to sound degraded, as if the participant has a hearing loss. In all occasions, the manipulations are signaled to the participant. The outcome measure typically consists of the pattern of word identification, and specifically how that pattern changes depending on acoustic properties of the sounds heard before the target word. Additional comparison measurements are taken of the participant's ability to hear and repeat words or sentences in background noise. Once an experiment is ready to launch, participants are randomly assigned to different testing conditions ("interventions"). But in most of the planned experiments, participants complete identical protocols, except that the specific ordering of many speech stimuli is randomized within the testing session. After the participant completes all of the testing blocks, they are debriefed about the full nature of the study, the hypotheses and the larger scope of the project in the context of speech communication.