Abstract.
Reverse-phi motion is the illusory reversal of perceived direction of
movement when the stimulus contrast is reversed in successive frames.
Here we proposed a double synaptic veto mechanism that could account
for experimental observed responses to reverse-phi motion in V1 cells.
We carried out detailed biophysical simulation in NEURON and verified
our results with experimental data.
Reverse-phi motion was first demonstrated by Anstis (Anstis, 1970).
Subjects perceived the reverse direction of motion when the contrast
of a moving object reverses in the second frame of a two-frame shift
experiment. Livingstone (2000) showed that direction-selective cells
in striate cortex of alert macaque monkeys showed reversed excitatory
and inhibitory region when two different contrast bars were flashed
sequentially during a two-bar interaction analyses. Space-time plots
of reverse-phi motion show energy in the reverse direction (Fig 1A).
Although the bar movement direction is to the right, the left motion
energy unit aligns better with the stimuli and extracts more motion
energy than the right motion energy unit. Therefore both the motion
energy model (Adelson and Bergen, 1985) as well as the equivalent elaborated
Reichardt model (Santen and Sperling, 1985) can account for the reverse-phi
motion. While motion energy models predict the reverse phi response,
it is unclear how neurons can accomplish this. We carried out detailed
biophysical simulation of direction selective cell implementing a synaptic
shunting inhibition scheme. Our results suggest that a simple synaptic
veto-mechanism that with a strong direction selectivity for normal motion
cannot account for the observed reverse phi-motion effect. Given the
nature of reverse phi-motion, a direct interaction between ON and OFF
pathway, missing in the original shunting-inhibition model, is essential
to account for the reversal of response. We proposed a double synaptic
veto mechanism in which ON excitatory synapses are gated by both delayed
ON inhibition at their null side and by delayed OFF inhibition at their
preferred side and the converse for OFF excitatory synapses (Fig 1B).
Mapping this scheme onto the dendrites of a direction-selective neuron
permits the model to respond best to normal motion in its preferred
direction and to reverse-phi motion in its null direction (Fig 2). The
space-time receptive field maps and two-bar interaction maps of the
model cell have the same characteristics as the V1 cell maps. Receptive
field mapping showed a progressively decreased response onset time and
increased response transiency when stimuli flashing from preferred side
to null side. Two-bar interaction maps showed reversed excitation and
inhibition regions when two different contrast bars are presented. The
results suggest our proposed mechanism can account for experimental
observed reverse-phi and normal motion direction selectivity observed
in V1 cells.
Publications.
Adelson, EH. and Bergen, JR. (1985) Spatiotemporal energy models for
the perception of motion. J. Opt. Soc. Am. A 2:284-299. Anstis, SM.
(1970). Phi movement as a subtractive process. Vision Research 10:1411-1430.
Livingstone, MS., Tsao, DY. and Conway, BR. (2000) What happens if it
changes contrast when it moves? Society for Neuroscience Abstracts 26:162.6
Santen, JPH. and Sperling, G. (1985) Elaborated Reichardt detectors.
J. Opt. Soc. Am. A 2:300-320.
Figure
1. Space-time plot of normal and reverse-phi motion and connectivity
diagram of the model that accounts for direction and reverse-phi selectivity.
(A) Space-time plot of a one dimension white bar moving from left to
right in normal motion (left panel) and in reverse-phi motion (right
panel). A right motion energy unit aligns well with the normal motion
plot, but triggers a much reduced response for the reverse-phi motion.
Instead, this strongly stimulates a left motion energy unit. Adopted
from Fig. 16 in (Adelson and Bergen, 1985). (B) Connectivity diagram
of normal and reverse-phi motion direction selective model. Input to
LGN neurons comes from a one-dimensional array of 179 pixels. The intensities
from those pixels were summed through difference of Gaussian spatial
filters on to LGN cells. There are one ON and one OFF center geniculate
cell at one of six spatially offset locations. Each of the middle four
ON center LGN cells provided excitatory input to one branch of the model
cell dendrites, while the ON center LGN cell immediately to the right
and the OFF center LGN cell immediately to the left provide delayed
on-the-path inhibition. The converse connection scheme for the OFF center
LGN cells' excitatory inputs are not shown.
Figure 2. A double synaptic veto mechanism can account for both
normal as well as reverse-phi motion direction selectivity. (A) Connection
scheme of the double synaptic veto mechanism. ON excitation is gated
by a delayed ON inhibition at its preferred side and by a delayed OFF
inhibition at its null side and the converse for off excitation. (B)
Mapping the double synaptic veto scheme onto the eight-armed cable model
with spiking at the cell body. Each excitation-inhibition-inhibition
triplet is mapped onto its own dendritic branch with the excitation
at the far side of cell body. (C) Model cell's response to a bright
bar moving at 10o/s across its receptive field. The model cell responds
best when a normal motion stimulus moves in its preferred direction
and when a reverse-phi motion stimulus moves in its null direction.
Stimuli reverse rate is 75Hz.
