Purchase this article with an account.
Sonia Poltoratski, Sam Ling, Frank Tong; Contrasting Bottom-up Saliency and Top-Down Attention in the Early Visual Pathway. Journal of Vision 2015;15(12):222. doi: 10.1167/15.12.222.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
The visual system employs a sophisticated balance of attentional mechanisms: individuals can willfully guide attention to serve their goals, but still notice salient information in the environment outside of their current locus of attention. At its simplest, this saliency can be defined as a measure of feature contrast across the visual field, such that a local region that is unlike its surrounding context along one or more feature dimensions is deemed salient. Here, we used high-resolution fMRI at 7T to investigate whether the effects of bottom-up saliency and top-down voluntary attention operate independently or interactively in the human visual system, and to determine the stages of visual processing at which these mechanisms first emerge. We measured BOLD responses to a near-peripheral Gabor grating, presented among a field of other gratings. Depending on its orientation relative to the contextual patches, this stimulus patch could be either salient – for example, a vertically oriented Gabor patch amongst horizontal patches – or not salient, sharing all of its features with the surrounding context. To concurrently manipulate top-down attention, observers performed a demanding task that directed covert spatial attention to either the salient or non-salient patch. This design allowed us to directly compare the magnitude of BOLD responses to salient and non-salient items, and to attended and non-attended items. We found evidence of independent, additive effects of top-down attention and bottom-up saliency in retinotopic visual areas V1, V2, V3, and hV4. In contrast, fMRI response amplitudes in the lateral geniculate nucleus (LGN) showed significant enhancement due to top-down attention, but showed no evidence of reliable modulation by orientation saliency. We conclude that saliency representation emerges in feature-selective populations of neurons at early cortical stages of visual processing, and can be distinguished from independent effects of top-down spatial attention.
Meeting abstract presented at VSS 2015
This PDF is available to Subscribers Only