Seeing as the visual guidance of action

[From Fred Nickols (000213.0515 EST)] --

I spotted this on another list and, for some reason, the last sentence in
the first paragraph immediately reminded me of behavior as the control of
perception. Anyone else see a fit?

···

"In a classic study, Held and Hein raised kittens in the dark and exposed
them to light only under controlled conditions. A first group of animals
was allowed to move around normally, but each of them was harnessed to a
simple carriage and basket that contained a member of the second group of
animals. The two groups therefore shared the same visual experience, but
the second group was entirely passive. When the animals were released after
a few weeks of this treatment, the first group of kittens behaved normally,
but those who had been carried around behaved as if they were blind: they
bumped into objects and fell over edges. This beautiful study supports the
enactive view that objects are not seen by the visual extraction of features
but rather by the visual guidance of action."

-- Francisco J. Varela, Evan Thompson, Eleanor Rosch, The Embodied Mind;
Cognitive science and Human Experience (Cambridge, Mass.: MIT Press, 1991).
Describing research reported in R. Held and A Hein, 1958 "Adaptation of
disarranged hand-eye coordination contingent upon re-afferent stimulation"
Perceptual-Motor Skills 8:87-90.

--

Fred Nickols
The Distance Consulting Company
"Assistance at A Distance"
http://home.att.net/~nickols/distance.htm
nickols@worldnet.att.net
(609) 490-0095

Martin Taylor (2000.02.13 13:56)

[From Fred Nickols (000213.0515 EST)] --

I spotted this on another list and, for some reason, the last sentence in
the first paragraph immediately reminded me of behavior as the control of
perception. Anyone else see a fit?

"In a classic study, Held and Hein raised kittens in the dark and exposed
them to light only under controlled conditions. A first group of animals
was allowed to move around normally, but each of them was harnessed to a
simple carriage and basket that contained a member of the second group of
animals. The two groups therefore shared the same visual experience, but
the second group was entirely passive. When the animals were released after
a few weeks of this treatment, the first group of kittens behaved normally,
but those who had been carried around behaved as if they were blind: they
bumped into objects and fell over edges. This beautiful study supports the
enactive view that objects are not seen by the visual extraction of features
but rather by the visual guidance of action."

I've always interpreted the Held and Hein result (and others of the
same kind) somewhat differently. My interpretation has been that the
kittens did not learn what to extract from the visual field what
would later be important for the control of perception, not that
objects are "not seen by the visual extraction of features" at the
time they are perceived. On the contrary, I think these studies go
some way toward demonstrating that objects _are_ normally seen by the
extraction of features that these kittens had not learned to see.

One can go back to the kittens in related studies, that were not
exposed to vertical, or not to horizontal, contours during a critical
period. Physiologically, their visual systems later simply did not
repond to the kind of contour they did not see. My understanding of
that whole area has been that one learns to see by means of learning
to control, and what one has not learned to see, one does not see.
That intepretation seems to fit the behaviour of these kittens better
than an interpretation based on what the kittens are doing at the
time they are bumping into things.

Bill Powers (2000.02.13.0216 mst)

Thus it is possible that in this central region there is a 1:1 mapping of
some of the optic nerve fibers to sensory endings. Does anyone know if that
is true?

I believe it is true, but it is nevertheless not true that one
receptor connects to one optic fibre. There is a substantial spread
of cross-connections even in the central fovea. What goes up the
optic nerve is not an element-by-element replica of the intensity
pattern on the retina. It is a highly encoded function of the
temporal and spatial pattern of intensities.

Certainly the signal from any one _receptor_ is not a function only
of the intensity of light falling on it at the time. The receptor's
ouput depends on the intensity of light that fell on it in the recent
past, and (I think) on the recent intensity in its neighbourhood. So
if we can ever perceive the intensity of light (which I seriously
doubt), we can do it _only_ by generating some kind of a function
that codes the patterns of receptor outputs over time and space.

Relating to Erling Jorgenson's message about coding stages, some time
in the 1970s, I think it was, Christian van der Malzberg published a
series (a pair?) of papers in Kybernetic that simulated the
Hebbian(?) development of receptive fields. My memory is vague, but
what I remember was that he found that the first level encoded a
Laplacian ("difference of gaussian" or "on-centre-off-surround" or
"off-centre-on-surround") function, the next level connected these
into line detectors at different orientations, and in general the
coding functions were very like those observed in the visual system.
Their actions served--as any efficient coding system should--to
decorrelate the inputs when coded into their outputs. It all seemed a
very elegant set of simulation studies, at the time.

Martin

[From Rick Marken (2000.02.13.1330 PST)]

Fred Nickols (000213.0515 EST) --

I spotted this on another list and, for some reason, the last
sentence in the first paragraph immediately reminded me of
behavior as the control of perception. Anyone else see a fit?

The _sentence_ doesn't, but the _results_ of the Held and Hein
study sure do. Held and Hein's results show that kittens can't
learn to control visual perceptions by just passively watching
those perceptions go by. The active kittens learn to control
visual objects (they avoid visual obstacles, move to a visual
destination, etc) by _trying_ to control those objects -- ie.
by actively producing outputs that do or don't have an effect on
the seen objects. The passive kittens can't learn to control
because they are prevented from producing outputs (via the
harnessing) that would have an effect on (and, thus, allow then
to learn to control) their visual perceptions.

The sentence you quote:

This beautiful study supports the enactive view that objects
are not seen by the visual extraction of features but rather
by the visual guidance of action.

doesn't seem to have anything to do with PCT. It says that
kittens _see objects_ because they visually guide these objects
with their actions. This strikes me as _highly_ unlikely. One
reason is that people who can no longer affect their visual
perceptions with their actions (paraplegics, for example) still
report seeing what they can no longer affect with their actions
(objects that are out of reach, for example).

PCT says that perception basically _is_ a "feature extraction"
process; the features that are "extracted" are aspects of the
"intensity" signals present at the sensory surface: features like
sensations, configuration, transitions, sequences, etc.

Best

Rick

···

--
Richard S. Marken Phone or Fax: 310 474-0313
Life Learning Associates e-mail: rmarken@earthlink.net
http://home.earthlink.net/~rmarken/

[From Bill Powers (2000.01.14.0333 MST)]

Martin Taylor (2000.02.13 13:56)
Me:

Thus it is possible that in this central region there is a 1:1 mapping of
some of the optic nerve fibers to sensory endings. Does anyone know if that
is true?

Martin:

I believe it is true, but it is nevertheless not true that one
receptor connects to one optic fibre. There is a substantial spread
of cross-connections even in the central fovea. What goes up the
optic nerve is not an element-by-element replica of the intensity
pattern on the retina. It is a highly encoded function of the
temporal and spatial pattern of intensities.

My understanding is that these (mutually-inhibitory) cross-connections have
the effect of emphasizing edges, which removes some effects of optical
blurring at the retina. The amplification in the cross-connections is not
high enough to produce a flip-flop action. We can still end up with one
fiber per retinal element. I don't know what you mean by "highly encoded."
I think the relationships are still pretty much 1-to-1, with some modest
degree of sharpening due to the cross connections.

Certainly the signal from any one _receptor_ is not a function only
of the intensity of light falling on it at the time. The receptor's
ouput depends on the intensity of light that fell on it in the recent
past, and (I think) on the recent intensity in its neighbourhood. So
if we can ever perceive the intensity of light (which I seriously
doubt), we can do it _only_ by generating some kind of a function
that codes the patterns of receptor outputs over time and space.

You're overcomplicating the situation, in my opinion. Yes, there is a time
constant (about 0.1 sec at daylight levels of illumination) in the response
of the retina to light. And yes, there are effects from neighborhood
illumination. But my concept of this is simply that the cross-connections
correct somewhat for the scattering of photons as they pass through the
retina, and some of the negative feedback effects reduce the time constant
from what it would otherwise be. The overall effect is to make the retina
behave more like a simple array of receptors connected 1-to-1 to fibers in
the optic nerve. A CCD camera has less crosstalk than the retina, and
perhaps a somewhat shorter time constant, but otherwise it works similarly
to the retina. Not quite the same -- CCD cameras don't suffer the Mach Band
defect.

Re your last paragraph, I believe the line detectors are central, not in
the retina.

Best,

Bill P.

[From Fred Nickols (000227.1240 EST)] --

···

At 04:01 AM 02/24/2000 -0600, you wrote:

Rick Marken (2000.02.13.1330 PST)]

Fred Nickols (000213.0515 EST) --

> I spotted this on another list and, for some reason, the last
> sentence in the first paragraph immediately reminded me of
> behavior as the control of perception. Anyone else see a fit?

The _sentence_ doesn't, but the _results_ of the Held and Hein
study sure do. Held and Hein's results show that kittens can't
learn to control visual perceptions by just passively watching
those perceptions go by. The active kittens learn to control
visual objects (they avoid visual obstacles, move to a visual
destination, etc) by _trying_ to control those objects -- ie.
by actively producing outputs that do or don't have an effect on
the seen objects. The passive kittens can't learn to control
because they are prevented from producing outputs (via the
harnessing) that would have an effect on (and, thus, allow then
to learn to control) their visual perceptions.

The sentence you quote:

> This beautiful study supports the enactive view that objects
> are not seen by the visual extraction of features but rather
> by the visual guidance of action.

doesn't seem to have anything to do with PCT. It says that
kittens _see objects_ because they visually guide these objects
with their actions. This strikes me as _highly_ unlikely. One
reason is that people who can no longer affect their visual
perceptions with their actions (paraplegics, for example) still
report seeing what they can no longer affect with their actions
(objects that are out of reach, for example).

Hmm. I read the sentence a little differently. I took it to mean that our
actions (moving around, especially avoiding objects) are guided by our
visual perceptions and that if we have no need to guide our actions in
accordance with our visual perceptions then we are apt to bump into
things. I probably read too much into it.
--

Fred Nickols
The Distance Consulting Company
"Assistance at A Distance"
http://home.att.net/~nickols/distance.htm
nickols@worldnet.att.net
(609) 490-0095

[From Rick Marken (2000.02.27.1200 PST)]

Fred Nickols (000213.0515 EST) --

I spotted this on another list and, for some reason, the last
sentence in the first paragraph immediately reminded me of
behavior as the control of perception. Anyone else see a fit?

Me:

The sentence you quote:

> This beautiful study supports the enactive view that objects
> are not seen by the visual extraction of features but rather
> by the visual guidance of action.

doesn't seem to have anything to do with PCT.

Fred Nickols (000227.1240 EST) --

Hmm. I read the sentence a little differently. I took it to
mean that our actions (moving around, especially avoiding objects)
are guided by our visual perceptions and that if we have no need
to guide our actions in accordance with our visual perceptions
then we are apt to bump into things.

I took it to mean that too.

I probably read too much into it.

Maybe you just read it wrong. What you apparently _saw_ (and typed)
was "our actions...are guided by our visual perceptions". This
statement expresses the basic S-R view of behavior; perception is
the control behavior. But what you must have _read_ (and understood)
was "our perceptions are guided by our actions". This certainly
would remind me of the basic PCT view of behavior: behavior is the
control of perception.

Best

Rick

···

--
Richard S. Marken Phone or Fax: 310 474-0313
Life Learning Associates e-mail: rmarken@earthlink.net
http://home.earthlink.net/~rmarken/