Dynamic Selection History Guides Attention via a Path-Like, Habit-Like Mechanism

Abstract

Selective attention resolves the conflict between many competing stimuli in a complex visual environment. The locations or objects we attend to––those selected for further processing––depend on three primary factors: stimulus salience, current goals, and selection history. Researchers posit that, like stimulus salience and current goals, a history of prioritizing certain locations leads to increased weights at those locations within an integrated priority map. The location with the highest integrated weight is attended first. However, certain selection history effects, like location probability learning, are acquired via repeated shifts of attention along a vector and lead to biases that hold unique characteristics. In a study of reward learning that does not induce repeated attentional shifts, I show that non-dynamic selection history biases depend on awareness, unlike location probability learning. In a virtual reality study, I show that, like motor habits, the dynamic location probability learning biases are encoded in an egocentric spatial reference frame. And in a neurological study, I show that Parkinson’s patients retain the ability to acquire location probability learning. This sets location probability learning apart from related spatial attentional biases that do not involve repeated shifts of attention and are impaired in Parkinson’s disease. It also provides insights into the brain regions underlying this ability. These findings suggest that selection history can guide attention not just by modulating weights in a priority map, but by inducing dynamic biases. I conclude that location probability learning represents a dynamic, path-like, habit-like attentional orienting mechanism that is mechanistically distinct from the integrated priority map.