Perceptual Learning and Visual Short-term Memory
Author: Nicholas Michael Van Horn
Publisher:
Published: 2014
Total Pages: 154
ISBN-13:
DOWNLOAD EBOOKVisual short-term memory (VSTM) is commonly perceived as a temporary buffer into which information is moved for retention across relatively short intervals. Guided by work on the "standard model" of working memory, these buffers are assumed to reside in modality-specific areas of the prefrontal cortex (PFC). Recent converging evidence has begun to call these conventional views into question, with a growing consensus that working memory is the emergent result of attention-guided activity across many areas of the brain. This "emergent-property" view suggests that working memory is sustained in part by the very same sensory areas involved in encoding external input. If visual short-term memory is represented and maintained on the same neural populations involved in common perceptual tasks as evidence suggests, we might expect possible interactions. The current thesis describes a series of experiments designed to build a detailed profile of these interactions. In Chapter 2 we establish the utility of using visual aftereffects as investigative tools in two behavioral experiments. In the first experiment we measure the magnitude of static and dynamic motion aftereffects before and after perceptual learning on a motion discrimination task using identical stimuli. Our results indicate that learning does not affect the duration of the aftereffects, strongly suggesting that improvements in discrimination are not mediated by changes in the underlying perceptual representations of the stimuli. Rather, our evidence supports the view that practice changes the relative contributions of perceptual outputs to decision-making areas. A second experiment rules out a competing explanation in which the locus of learning does not share enough overlap with the neural populations responsible for the observed aftereffects. Next, in Chapter 3 we demonstrate the efficacy of analyzing multivariate response data over traditional reliance on accuracy or response time alone. We apply the diffusion model (DM) to accuracy and response time distributions on data from two perceptual learning experiments. Results reveal that practice-induced improvements in motion direction discrimination arise from the combination of stimulus-specific improvements in the quality of sensory information entering decision areas, and non-specific gains in the timing of decision-making processes. Following this, in Chapter 4 we leverage the methodological gains of the previous results to test the hypothesis that the contents of VSTM can influence the perception of stimuli during the memory retention interval. We use a dual-task design to test the effect of VSTM on perception. Results suggest that orientation information in memory has a repulsive effect on the perception of subsequently presented orientations, not unlike tilt aftereffects. Further DM analysis confirms that our results are perceptual in nature and do not arise from artifacts in response time shifts. Finally, in Chapter 5 we describe an experiment designed to test if the interactions observed in Chapter 4 share the same neural mechanism as memory masking, and what, if any, effect practice has on these two forms of interference. An asymmetrical effect due to memory-stimulus similarity, as well as contrasting effects due to training on the two sources of interference strongly indicate the involvement of two separate processes. Taken together, this collection of studies highlights the existence of another form of VSTM-perception entanglement that is complementary to memory masking, but behaves much differently, particularly under the influence of practice. The results indicate that VSTM and perception interact in systematic ways that are critical to understand, most notably in laboratory settings where the effects can disrupt the subtle differences upon which observers must make decisions. A novel model is proposed that parsimoniously reconciles observed patterns of perceptual learning, as well as the apparently disparate forms of interactions measured throughout. Collectively, the present work supports the emergent-property hypothesis of working memory, as well as a revised version of the standard model.Less
