Imagine you are watching a soccer game, carefully attending to the player on your favourite team to determine which direction he will go. In studies on visual perception, the player is considered an “object”, which has multiple features: direction of motion, the colour of his outfit, etc. Can you voluntarily restrict attention to just the player’s direction of motion, the feature that is relevant for your decision, or does your attention unintentionally spread to also enhance the particular shade of orange in his shirt? Major theories of visual attention assume that attending to one feature of a visual object automatically results in selection of the whole object, including task-irrelevant features, but are they correct in assuming that the complete object comprises the basic unit of selection? The overall aim of this proposal is to reveal the neural mechanisms of selective attention by addressing this issue. I will focus on three major research questions: what is the unit of selective attention, how is attentional selection implemented in the visual cortex, and how does selective attention impact perceptual learning? I will address these questions using a combination of behavioural, computational modelling, and neuroimaging techniques, together with novel “decoding” methods to analyze brain activity patterns. This multifaceted approach will allow me to rigorously investigate how cortical selectivity for visual features changes with top-down attention, and to address these unresolved questions regarding the neural mechanisms of human visual attention.