Abstract
Understanding spoken language, although often thought to be effortless, is the result of several multi-stage interactive processes. These processes range from the identification of words from a continuous acoustic signal, to extracting semantic representations from the spoken content to form a ‘situation model,’ which allows listeners to make inferences about upcoming information. In situations where the quality of the incoming acoustic signal is less than ideal, extra time and effort are required for accurate comprehension. This dissertation consists of six experimental chapters designed to elucidate the cognitive and perceptual factors that underlie comprehension of spoken language under acoustically challenging conditions, with a particular focus on the real-world implications for listeners with age-related hearing loss.Following an introductory chapter, chapter 2 sets the context for this work with results from a large-scale longitudinal study of middle aged and older adult monozygotic and dizygotic twins on frequency detection thresholds for pure tones across the range of speech frequencies, a commonly used diagnostic measure for hearing loss. Biometric modeling revealed a strong and stable genetic component to hearing acuity even as hearing acuity declined across the age-span tested.
Subsequent chapters focus on speech perception and cognitive effort in laboratory-based recall or comprehension tasks for sentences and multi-sentence discourse. Chapter 3 begins at the sentence level, demonstrating that both younger and older adults can use sentence prosody (pitch contour, word-stress, intra- and inter-word timing patterns) spontaneously produced by speakers to facilitate accurate sentence recall. Chapter 3 also introduces the use of pupillometry (the measurement of dilation of the pupil of the eye as an index of processing effort). Results showed that the availability of supportive prosody has the adaptive feature of reducing processing effort.
Chapter 4 moves to the level of discourse, elucidating the benefit of allowing the listener additional processing time in recall of discourse passages presented at a rapid speech rate. Although older adults benefitted from the availability of extra processing time, the benefit was reduced for listeners with poorer hearing acuity. Pupillometric results suggest that poorer hearing acuity was associated with increased effort during the initial formation of a “situation model”.
Chapters 5 and 6 follow up on Chapter 4 to examine a unique case of auditory perception; the case of adults whose hearing loss has progressed beyond benefit from conventional hearing aids who hear via a Cochlear Implant (CI). As will be described, CIs transmit sound through direct electrical stimulation of the auditory nerve through a surgically implanted electrode array along the basilar membrane that delivers a spectrally impoverished acoustic signal. Using vocoder simulations with listeners with typical hearing and with CI users, Chapters 5 and 6 use two methods to examine how available processing time can be leveraged for comprehension and recall of spoken discourse.
Chapter 7 returns to the sentence level to challenge the notion of a single optimal sentence processing algorithm. Using both orienting instructions and a manipulation of spectral clarity, results demonstrate the flexibility of sentence processing. Results were further informed by application of drift diffusion modeling.
Taken together, the results of this dissertation depict spoken language processing as a dynamic, highly flexible system that can, within limits, operate under a variety of constraints.