Title: TWO MOTION SENSOR POPULATIONS, AS REVEALED BY TEST PATTERN TEMPORAL FREQUENCY
1TWO MOTION SENSOR POPULATIONS, AS REVEALED BY
TEST PATTERN TEMPORAL FREQUENCY
M.J. van der Smagt, F.A.J. Verstraten W.A. van
de Grind
Neuroethology Group Padualaan 8 NL-3583 CH
Utrecht Netherlands M.J.vanderSmagt_at_bio.uu.nl
2Introduction
- Motion aftereffects (MAEs) tested with stationary
test patterns, such as Static Visual Noise (SVN)
occur for adaptation speeds up to about 25 deg/s. - It was shown recently1 that MAEs tested with
Dynamic Visual Noise (DVN) occur for much higher
adaptation velocities (up to 80 deg/s). - Static MAEs are dominant for lower adaptation
speeds, whereas dynamic MAEs are dominant in
the high speed range. (figure 1) - Transparent motion containing one fast and one
slow velocity results in an MAE opposite to the
fast vector when tested with DVN, and opposite
the slow vector when SVN is used as test.
3- Transparent motion containing one fast and one
slow velocity results in a transparent MAE2 when
a new test pattern is used, which contains both
SVN and DVN characteristics (see figure 2). Again
the DVN-component of this combined test pattern
seems to move rapidly opposite to the fast
adaptation vector, and the SVN-component slowly
opposite to the slow adaptation vector. - SVN and DVN patterns differ from each other
primarily in temporal characteristics. SVN is
refreshed at 0 Hz, DVN at 45 Hz. (note these are
temporal cut-off frequencies since these kind of
patterns are temporally as well as spatially
broad-band) - At least two temporal channels (sustained and
transient) have been shown to exist in human
motion vision.3 - Here we examine the effect of test-pattern-refresh
-frequency on the MAE, in relation to adaptation
speed.
4Figure 1 MAE durations as function of type of
test pattern (data from ref. 1). Arrows show the
speed combinations that are used in the present
experiment.
Figure 2 The space-time plots give an example of
adaptation and test stimuli that lead to the
transparent MAE (ref 2.)
5Methods
- The adaptation stimulus consisted of two
superimposed Random-Pixel-Arrays, that moved
transparently in orthogonal directions (figure 3)
behind a circular window. Three speed
combinations were used (see figure 1) - Slow 1.3 deg/s and 4 deg/s
- Fast 12 deg/s and 36 deg/s
- Mixed 4 deg/s and 12 deg/s
- The test stimulus consisted of DVN of which the
refresh frequency was varied across trials from 0
Hz (SVN) to 90 Hz. - Three observers adapted 45 s to the transparent
motion while fixating on a dot in the center. The
test stimulus was shown for 3 s, after which the
screen turned grey and an arrow appeared, which
was to be aligned with the MAE direction.
6Figure 3 Example of the adaptation stimulus
(left) and space-time plots of the some of the
test-pattern-refresh-frequencies (right)
7Results
- Irrespective of the test-pattern-refresh-frequency
, the observers indicated a more or less constant
MAE direction for both the Slow and the Fast
condition (figure 4). Although they reported weak
MAEs and the task to be difficult for high
refresh frequencies in the Slow, and low
frequencies in the Fast condition. - In the Mixed condition the MAE is more opposite
the fast (12 deg/s) adaptation component for test
frequencies gt 20 Hz, while more opposite the slow
(4 deg/s) adaptation vector for frequencies lt 20
Hz. - Around 20 Hz observers were inconsistent in their
direction judgements, sometimes judging the MAE
to be more opposite the slow adaptation vector,
sometimes more opposite the fast. There is no
smooth transition of MAE direction as function of
test frequency (figure 5).
8Figure 4 MAE directions as a function of
test-pattern-refresh-frequency for a typical
observer. Green diamonds are for the Slow
condition, blue circles for the Fast and red
squares for the Mixed condition. Curves are
sigmoidal functions fitted through the data. (r2
mixed gt 0.98 r2 slow lt 0.75 r2 fast lt0.1)
Figure 5 Same as figure 4 for three observers
(individual data points) and only the Mixed
condition. Note that there is no gradual change
in the MAE direction with increasing
test-pattern-refresh-frequency.
9Conclusions
- Adaptation to higher speeds is revealed in the
MAE when DVN patterns with refresh frequencies
above 20 Hz are used. - Adaptation to lower speeds is revealed in the MAE
when SVN patterns or DVN patterns with refresh
frequencies below 20 Hz are used. - The almost stepwise transition in the Mixed
condition indicates the existence of two rather
independent motion sensor populations which can
be identified by their velocity (slow and fast)
and temporal frequency (low and high) preference
(see figure 6). This finding is compatible with
the distinction between sustained and transient
channels3 in motion vision. - However, temporal frequency alone cannot explain
all the differences between the slow and fast
channel (see below).
10Figure 6 Schematic representation of the main
findings. Temporal frequency sensitivity does not
appear to increase with speed sensitivity. Rather
there appear to be two separate populations One
tuned to higher speeds and higher TFs, the other
to lower speeds and lower TFs.
11Temporal Frequency cannot be the whole story
- DVN test patterns are temporally broad-band and
high refresh frequencies are thus temporal
cut-off frequencies. - A Random-Pixel-Array of which each bright pixel
becomes dark and each dark pixel bright every few
frames is temporally more narrow-band (although
still broadband spatially see top right). - When such a contrast reversing pattern, with a
high reversal frequency, is used as test pattern,
adaptation to slow motion (as well as high speed
adaptation) yields strong MAEs (see bottom
right). - Although there are no low TFs in this type of
test stimulus, a recurring pattern is clearly
apparent to the observers. - Closer scrutiny of the MAE showed that for higher
speeds the MAE appeared perceptually different
from that at low adaptation speeds, again
indicating that we are dealing with two separate
motion channels.
12(No Transcript)