The most common and natural human behaviours are often the most computationally difficult to understand. This is especially true of spoken language comprehension considering the acoustic ambiguities inherent in a speech stream, and that these ambiguities are exacerbated by the noisy and distracting listening conditions of everyday life. Nonetheless, the human brain is capable of rapidly and reliably processing speech in these situations with deceptive ease – a feat that remains unrivaled by state-of-the-art speech recognition technologies. It has long been known that supportive context facilitates robust speech perception, but it remains unclear how the brain integrates contextual information with an acoustically degraded speech signal. The four studies in this dissertation utilize behavioural and functional magnetic resonance imaging (fMRI) methods to examine how the normally functioning human brain uses context to support the perception of degraded speech. First, I have observed that text presented simultaneously with distorted sentences results in an illusory experience of perceptually clearer speech, and that this illusion depends on the amount of distortion in the bottom-up signal, and on the relative timing between the visual and auditory stimuli. Second, fMRI data indicate that activity in the earliest region of primary auditory cortex is sensitive to the perceived clarity of speech, and that this modulation of activity likely comes from left frontal cortical regions that probably support higher-order linguistic processes. Third, conscious awareness of the visual stimulus appears to be necessary to increase the intelligibility of degraded speech, and thus attention might also be required for multisensory integration. Finally, I have demonstrated that attention greatly enhances the processing of degraded speech, and this enhancement is (again) supported by the recruitment of higher-order cortical areas. The results of these studies provide converging evidence that brain uses prior knowledge to actively predict the form of a degraded auditory signal, and that these predictions are projected through feedback connections from higher- to lower-order order areas. These findings are consistent with a predictive coding model of perception, which provides an elegant mechanism in which accurate interpretations of the environment are constructed from ambiguous inputs in way that is flexible and task dependent. / Thesis (Ph.D, Neuroscience Studies) -- Queen's University, 2012-09-25 10:48:50.73

Human information processing depends critically on continuous predictions about upcoming events, but the temporal convergence of expectancy-based top-down and input-driven bottom-up streams is poorly understood. We show that, during reading, event-related potentials differ between exposure to highly predictable and unpredictable words no later than 90 ms after visual input. This result suggests an extremely rapid comparison of expected and incoming visual information and gives an upper temporal bound for theories of top-down and bottom-up interactions in object recognition.

In this study we sought to determine whether infants, like adults, utilize previous experience to guide figure/ground processing. After familiarization to a shape, 5-month-olds preferentially attended to the side of an ambiguous figure/ground test stimulus corresponding to that shape, suggesting that they were viewing that portion as the figure. Infants’ failure to exhibit this preference in a control condition in which both sides of the test stimulus were displayed as figures indicated that the results in the experimental condition were not due to a preference between two figure shapes. These findings demonstrate for the first time that figure/ground processing in infancy is sensitive to top-down influence. Thus, a critical aspect of figure/ground processing is functional early in life.