In everyday life, the visual environment is often far more complicated than can be faithfully represented by the brain. Decades of behavioral research has demonstrated that humans can flexibly prioritize different features of the world depending on their current goals. How does the quality of neural representations reflect this flexible prioritization, and in turn limit cognitive performance? Such representations must exist across populations of neurons spanning entire brain regions, which requires large-scale measurements of brain activity. Sprague takes advantage of the coherent retinotopic organization of visual brain regions, which is commensurate with the resolution of state-of- the-art neuroimaging methods, by applying newly-developed multivariate model-based image reconstruction techniques. These techniques, broadly termed ‘inverted encoding models’, are built to quantify neural representations encoded across entire populations, providing a unique window into human cognition. 

By visualizing neural representations across multiple brain regions, Sprague has tested key predictions of priority map theory, which posits that the relative importance of different scene elements is marked by the activity landscape over topographic ‘priority maps’. Across several studies, Sprague shows that the profiles of priority maps across retinotopic regions of occipital, parietal, and frontal cortex are modulated by the attentional demands placed on observers, and that they can be used to maintain information in visual working memory. Moreover, by relating changes in measured neural priority maps across task conditions to changes in behavioral performance, Sprague demonstrates a critical role for these neural representations in supporting spatial cognition.