Assessing neuronal correlates of salience and their adaptability with naturalistic textures in macaque V1 and V2

Abstract

Salience is critical to vision. It allows stimuli that are different from their surroundings to pop out, drawing our attention. Perceptual salience is postulated to be encoded via a saliency map, based on differences in neuronal responsivity to simple image features at different spatial locations. Simple image features such as luminance, orientation and color are known to affect saliency and many of these features are encoded in primary visual cortex (V1), which several influential theories propose instantiate a saliency map. However, the degree to which more complex image features can determine salience, and whether there are neural correlates of salience which are computed outside of V1, remains unclear. Here we use displays of naturalistic textures to test for neural correlates of salience--termed pop-out responses--in V1 and area V2 of anesthetized macaque monkeys. Sensitivity to higher-order texture statistics arises in V2, so pop-out responses for these displays, if they exist, would be expected to be computed after V1. We presented displays in which a target texture, presented within the neuronal receptive field, was surrounded by distractors. Distractors could differ from the target texture in either higher-order texture statistics only, or in both lower- and higher-order statistics. We found little evidence for pop-out signals in either V1 or V2, for either display type. However, brief periods of adaptation could induce pop-out responses in V2. This suggests that adaptation might define which features of the environment are most salient, even if those features would otherwise not evoke pop-out responses. Significance statementWe tested for neuronal correlates of salience--termed pop-out responses--using displays with multiple patches of naturalistic textures, whose higher-order statistics are encoded by neurons outside primary visual cortex. We found little evidence for pop-out responses in V1 or V2. However, brief periods of adaptation could induce pop-out responses in V2. Our results indicate that the computations that define bottom-up attention (i.e., salience) are malleable and continuously updated by our stimulus history.