| Abstract |
We perceive the world as stable and upright despite ego motion. To
construct a world-centered percept of our visual environment our visual system
must compensate for the orientation of the eyes in the head and of the head in
space. Previous work has suggested that the construction of such a world-
centered representation is accomplished at later stages in the cortical
processing of visual information, building on a fully retinal representation
of the visual world in striate visual cortex (V1). Contrary to this widely
held belief, we demonstrate that subsets of V1 neurons use information about
eye torsion to keep receptive fields stable relative to the head despite tilt-
induced ocular counter roll. This conclusion is based on the analysis of 89
neurons recorded from V1 layers 2-4 of a behaving monkey whose receptive fields
were compared with the monkey´s head and body kept in different roll
orientations. We obtained high-resolution receptive field maps using the
reverse correlation technique while the monkey was actively fixating a target
located in the middle of a computer screen. We repeated the same procedure
with the monkey and the monitor presenting the stimuli roll tilted by 50deg
and, finally, repeated the same procedure when the monkey was rolled back to
the upright position. 2-D eye position was monitored using the search-coil
technique and eye torsion was measured by means of a camera system. Roll-
tilting the monkey by 50deg induced 4 to 5deg of ocular counter roll. 88% of
neurons showed a shift of their receptive field location on the monitor,
suggesting that they were not able to compensate for ocular counter roll.
However, to our surprise, the remaining 12% exhibited receptive fields that did
not shift on the monitor despite tilt-induced counter roll. In other words,
they were head-centered. Head centered neurons were mostly located in the
superficial layers 2 (3 of 39) and 3 (7 of 32) whereas 17 of 18 layer 4 neurons
showed significant receptive field shifts, i.e. they were not head-centered.
This layer-specific distribution of coordinate frame preferences suggests that
non-retinal information on eye torsion is used to convert retinocentric
signals in the major recipient of geniculate input, layer 4, into a head-
centered format in supragranular primary visual cortex. On the other hand, the
lack of any evidence for world-centered responses in V1 suggests that the
integration of information about head orientation, requiring vestibular input,
is reserved to extrastriate visual cortex. In other words, the construction of
a world-centered representation of the perceptual visual world seems to take
place in at least two steps associated with distinct parts of visual cortex. |
