Event:

The VBC framework is motivated by brain’s information bottleneck, so that
only a tiny fraction of sensory information is recognized.
For primate vision, this framework has three components: "V" for The V1 Saliency Hypothesis (V1SH),
"B" for the Bottleneck of attention, and "C" for the Central-Peripheral Dichotomy theory (CPD).
They motivate each other to shape the framework for vision.
V1SH states that neural responses in primary visual cortex (V1) to visual inputs form
a bottom-up saliency map of the visual field to guide attention. It has received converging
experimental support: e.g., V1 activity to a visual location is correlated with faster saccades to
that location in monkeys (Yan, Zhaoping, Li 2018), and human gaze is strongly attracted to a
location with a unique eye-of-origin of input which V1 responses would single out, even though
it is not perceptually distinctive (Zhaoping 2008). Since the saliency map guides visual
attention to center the attentional spotlight on the fovea, V1SH motivates the idea that the
attentional bottleneck, which limits the amount of information for deeper processing,
starts already at V1’s output to downstream areas along the visual pathway.
Together, V1SH and the bottleneck motivate the central-peripheral dichotomy (CPD) theory,
which hypothesizes distinct roles for central and peripheral vision (Zhaoping 2019):
(1) peripheral vision is mainly for looking (guiding gaze/attentional shifts) whereas
central vision is mainly for seeing (recognition); (2) top-down feedback from downstream
to upstream regions along the visual pathway should mainly target central vision to aid seeing
by querying for more information from upstream areas (e.g., V1).
I will review neural and behavioral evidence for this framework, particularly
the recent psychophysical confirmations of two predictions of the VBC framework:
(1) the novel reversed depth illusion, that is only, or more,
visible in peripheral vision; and (2) this illusion becomes visible in central vision
when top-down feedback is compromised by backward masking.
I will show how the VBC is related to but distinct from some classical and modern ideas, and
extend them to multisensory domains and across different animal species, for which we identity
the central and peripheral senses, attentional guidance behavior, and neural substrates
that may be sub-cortical and cortical, and provide falsifiable predictions.