Thesis (Ph. D.)--University of Rochester. Department of Brain and Cognitive Sciences, 2018.
One of the central challenges in speech perception is the lack of invariance:
talkers sound different from one another, due to physiological features, such
as vocal tract length, and sociolinguistic features, such as dialect. To achieve
robust speech comprehensions, listeners must overcome this lack of invariance
in the speech signal.
One way that listeners may overcome this problem is by adapting to individual
talkers - i.e., learning how individual talkers sound - and generalizing
this learning to novel talkers. In Chapter 2, we explore the conditions for
adaptation and generalization: we identify strong evidence for generalization
following exposure to multiple talkers of an accent, and weaker evidence for
generalization following exposure to a single talker of an accent. Our experiments,
which use a larger set of exposure and test talkers than previous work,
set up the possibility for future exploration of talker-dependent adaptation
and generalization effects.
However, talkers differ in how they pronounce words for a number of
reasons, ranging from more permanent, characteristic factors such as having a foreign accent, to more temporary, incidental factors, such as speaking
with a pen in the mouth. One challenge for listeners is that the true cause
underlying atypical pronunciations is never directly known, and instead must
be inferred from (often causally ambiguous) evidence. In Chapters 3 and
4, we explore the extent to which causal inferences may underlie speech
perception. We find that in certain cases, the speech perception system may
be capable of considering causes of atypical pronunciations and maintaining
information about these competing causes. We discuss possibilities for why
we do not find this to be the case across all our experiments, as well as future
directions for research.