548-B 29 OCTOBER 2021 • VOL 374 ISSUE 6567 science.org SCIENCE
INSIGHTS | PRIZE ESSAY
FEEDBACK FILLS IN THE GAPS
What is the mechanistic basis for this fun-
damental perceptual phenomenon? Using
optical recordings in awake mice, we found
that neurons in the primary visual cortex—
but not in earlier stages of processing—re-
spond to context alone. Contextual drive
must come from neurons that respond to
a stimulus in a surrounding region. These
neurons could be in the primary visual cor-
tex, in higher visual areas, or even in an ear-
lier stage of processing. However, each of
these scenarios makes different predictions
related to the time required for a signal to
travel from the retina to the neurons in the
primary visual cortex.
If contextual drive is inherited from sur-
rounding neurons in an earlier stage of pro-
cessing, we would expect transmission to
take as much time as a signal coding for a
stimulus in the center (the additional cable
length connecting the surrounding neurons
is negligible). Otherwise, we would expect
there to be at least one additional neuron
in the processing chain, which would gen-
erate a measurable delay. We found a delay
between the signal from a centered stimu-
lus and one from the context, making it
unlikely that contextual drive directly origi-
nates in earlier stages of processing.
To test whether responses to the visual
context are driven by activity in higher vi-
sual areas, we developed a method that al-
lowed us to silence neurons in higher visual
areas while recording the visual responses
of neurons in the primary visual cortex.
By silencing higher visual areas, we virtu-
ally abolished the contextual drive. Using
optical recordings from axons originating
in higher visual areas and projecting to the
primary visual cortex, we found that these
feedback projections carry the information
necessary for contextual drive—this reverse
flow of information likely filling in miss-ing information ( 12 ). In other words, when
there is nothing to see, we see what we ex-
pect to see: Context dominates when Waldo
is hidden in plain sight. jREFERENCES AND NOTES- H. Adesnik, W. Bruns, H. Taniguchi, Z. J. Huang,
M. Scanziani, Nature 490 , 226 (2012). - A. J. Keller et al., Neuron 108 , 1181 (2020).
- C. K. Pfeffer, M. Xue, M. He, Z. J. Huang, M. Scanziani, Nat.
Neurosci. 16 , 1068 (2013). - X. Jiang et al., Science 350 , aac9462 (2015).
- H.-J. Pi et al., Nature 503 , 521 (2013).
- R. von der Heydt, E. Peterhans, G. Baumgartner, Science
224 , 1260 (1984). - Z.-M. Shen, W.-F. Xu, C.-Y. Li, J. Physiol. 583 , 581 (2007).
- M. Fiorani Júnior, M. G. Rosa, R. Gattass, C. E. Rocha-
Miranda, Proc. Natl. Acad. Sci. U.S.A. 89 , 8547 (1992). - A. F. Rossi, R. Desimone, L. G. Ungerleider, J. Neurosci.
21 , 1698 (2001). - H. E. Jones, K. L. Grieve, W. Wang, A. M. Sillito,
J. Neurophysiol. 86 , 2011 (2001). - U. H. Schnabel et al., Sci. Rep. 8 , 17800 (2018).
- A. J. Keller, M. M. Roth, M. Scanziani, Nature 582 , 545
(2020).
10.1126/science.abl7124