ABSTRACT
Unfamiliar accented speech poses a significant barrier comprehension in older adults, and there is insufficient
evidence to attribute it simply to age-related declines in hearing acuity or cognitive functioning. As the
population of adults 65 and older in the US has grown dramatically in recent years, it is critical that
interventions be developed to improve accent adaptation, but no such thing yet exists. In this proposal, we
examine the hypothesis that generalization lays at the root of older adults’ issues with accent adaptation;
specifically, older adults may be able to adapt to an accented talker with sufficient exposure, but fail to properly
apply this information to future encounters with talkers who share the same accent. This perspective is
informed by the Ideal Adapter model, a computational framework for speech perception wherein it is proposed
that listeners independently track the way the speech signal is altered by different qualities of the talker, like
age, sex, or accent, and use that information when adapting to an unfamiliar talker. Over the course of several
behavioral experiments, we will compare younger and older adults’ generalization of accent information across
talkers. In Experiments 1A & 1B, younger (18-22 years) and older adults (65+ years) will engage in lexically-
guided perceptual learning (LGPL) task, a well-established, accent adaptation paradigm that we will modify to
make more suitable for assessing generalization, such as adding a carrier sentence to each stimulus to
provide the listener with more exposure to the accent. Performance between the two age groups will be
compared, with the expectation that older adults will fail to generalize phonetic information to a novel talker of
the same accent. In Experiment 2, older adults will listen to a story spoken with an artificial accent which better
reflects naturalistic listening conditions, which both adds cognitive demands but also allows for the use of more
top-down heuristics. The magnitude of generalization in older adults on the LGPL task in Experiment 1B and
the story-listening task in Experiment 2 will then be compared. In Experiment 3, we will use the Ideal Adapter
model to attempt to specify the mechanism that leads to impaired generalization in older adults. We will create
several perturbations to the model that serve as proxies for age-related sensory/cognitive impairments like
hearing loss or slower processing speed, as well as more conceptual impairments like the downweighing of
previous experience. Output from the model will then be compared to older adults’ performance to adjudicate
which perturbation best reflects human data. This proposal will yield novel insight into the mechanisms of
accent adaptation in older adults, which in turn can be used by clinicians to better tailor interventions in the
future. In addition, it will provide the applicant with training opportunities in (1) expanding my research into
healthy older adults, (2) integrating computational modeling into my research program, and (3) networking with
interdisciplinary researchers of aging, with the intention of utilizing these to establish a research program as
the applicant moves towards independence in career as a researcher.