| Popis: |
Representations of our external environment and of our internal memories together guide our awareness and actions. While large numbers of features and objects from the boundless external environment compete for sensory processing, working memory capacity is highly limited. Though extracting and integrating features into relevant objects are major objectives of perception, feature integration and well separated object representations may be basic characteristics of working memory. As discussed in Chapter 1, attentional selection of both external and internal content is pivotal, but its workings may differ according to the distinct representational formats. This thesis investigates how attention prioritises feature information in working memory to guide future actions. To test my hypotheses, I design and conduct behavioural studies, analyse neurophysiological recordings, and apply computational modelling. My empirical findings describe two main ways through which feature representations in working memory can be modulated: through goal-related prioritisation and through interacting with lingering sensory history. Chapters 2 and 4 describe tasks in which internally maintained information can be prioritised based on task demands. They demonstrate prioritisation of items and feature dimensions during working-memory delays, as well as the ability to protect and update features during maintenance. Chapter 4 also outlines how stimulus history modulates sensory processing and interacts with goal-based influences. Sensory content from same-trial or previous-trial information exerts opposing systematic performance biases. Applying neural decoding methods that I develop in Chapter 3, I reveal in Chapter 4 that both same-trial and previous-trial sensory history generates a repulsive neural bias at encoding, modulated by task relevance of previous stimuli. The findings suggest that other biases at post-perceptual stages may also contribute to the final opposing biases in performance. Taken together, the studies in the thesis extend our understanding of how visual feature information in working memory can be modulated and prioritised to serve future behaviour adaptively. |
