[From Rick Marken (2000.02.10.1100)]
Has all interest in discussion of perceptual control theory
really disappeared? Isn't anyone doing anything with PCT
that's worth mentioning?
Best
Rick
--
Richard S. Marken Phone or Fax: 310 474-0313
Life Learning Associates mailto: rmarken@earthlink.net
http://home.earthlink.net/~rmarken
Well, what I've been doing is reading PCT and ecological psychology
concurrently. I'd appreciate someone's general take on ecological
psychology from a PCT perspective. Would it be a fair assessment to say
that the first sentence below is inconsistent with PCT while the rest is
quite compatible?
"It is not necessary to believe that anything whatever is transmitted
along the optic nerve in the activity of perception... We can think of
vision as a perceptual system, the brain being simply part of the
system. The eye is also part of the system, since retinal inputs lead
to ocular adjustments and then to altered retinal inputs, and so on.
The process is circular, not a one-way transmission... The
eye-head-brain-body system registers the invariants in the structure of
ambient light" (Gibson, 1979, "The ecological approach to visual
perception," p. 61)
Also, are there any published comparisons of the two theories that
anyone can recommend?
Thanks kindly,
Dan Palmer
Richard Marken wrote:
[From Rick Marken (2000.02.10.1100)]
Has all interest in discussion of perceptual control theory
really disappeared? Isn't anyone doing anything with PCT
that's worth mentioning?Best
Rick
--
Richard S. Marken Phone or Fax: 310 474-0313
Life Learning Associates mailto: rmarken@earthlink.net
http://home.earthlink.net/~rmarken
[From Richard Kennaway (20000211.1720 GMT)]
Dan Palmer writes:
Well, what I've been doing is reading PCT and ecological psychology
concurrently. I'd appreciate someone's general take on ecological
psychology from a PCT perspective. Would it be a fair assessment to say
that the first sentence below is inconsistent with PCT while the rest is
quite compatible?"It is not necessary to believe that anything whatever is transmitted
along the optic nerve in the activity of perception... We can think of
vision as a perceptual system, the brain being simply part of the
system. The eye is also part of the system, since retinal inputs lead
to ocular adjustments and then to altered retinal inputs, and so on.
The process is circular, not a one-way transmission... The
eye-head-brain-body system registers the invariants in the structure of
ambient light" (Gibson, 1979, "The ecological approach to visual
perception," p. 61)
I am baffled by the first sentence. Surely it *is* necessary to believe
that something is transmitted along the optic nerve, because you can stick
electrodes in and observe signals flowing along it. It is also necessary
to believe that these signals are involved in the activity of (visual)
perception, because if they are blocked, so is perception.
The next bit about circularity is unobjectionable, but doesn't go very far,
and doesn't mention the fact (so I've heard) that a large proportion of the
signals in the optic nerve itself go outwards. So there's more than just
the optical input - muscular output - optical input loop.
As for the last sentence, I can't make much of it, except that it suggests
the usual idea of the brain as a bucket into which externally existing
perceptions flow from the outside world. Perhaps more context would help,
but the library copy here is out on loan at the moment.
-- Richard Kennaway, jrk@sys.uea.ac.uk, http://www.sys.uea.ac.uk/~jrk/
School of Information Systems, Univ. of East Anglia, Norwich, U.K.
[From Dan Palmer (2000.2.12.1658)]
Richard Kennaway wrote:
As for the last sentence, I can't make much of it, except that it suggests
the usual idea of the brain as a bucket into which externally existing
perceptions flow from the outside world. Perhaps more context would help,
but the library copy here is out on loan at the moment.
My apologies, a little more context: In this part of his book Gibson has
critiqued the idea that in visual perception, the retinal image gets sent to
the brain as a whole, and suggests that "the only alternative has seemed to be
that [the retinal image] is transmitted to the brain element by element, that
is, by signals in the fibres of the optic nerve" (p. 60). In the quotation
you read, he was suggesting that both views are flawed: "We need not believe
that _either_ an inverted picture or a set of messages is delivered to the
brain" and later "The eye is not a camera that forms and delivers an image,
nor is the retina simply a keyboard that can be struck by fingers of light"
(both from p. 61).
My understanding of PCT is that "the visual image" does get sent inward in the
form of first-order intensity signals which are combined into various
higher-level "representations" or signals and compared with similarly packaged
reference signals. Thus there does seem to be a direct conflict between the
two theories on this point?
Thanks to Rick for his comments, I hope to catch up with the archived
discussions.
Cheers,
Dan Palmer
[Martin Taylor 20000212 09:57]
[From Dan Palmer (2000.2.12.1658)]
My understanding of PCT is that "the visual image" does get sent inward in the
form of first-order intensity signals which are combined into various
higher-level "representations" or signals and compared with similarly packaged
reference signals. Thus there does seem to be a direct conflict between the
two theories on this point?
I don't know about a theoretical conflict. Physiologically, you have
to recognize that there is a _lot_ of information processing going on
in
the retina. There are on the order of 10^8 individual light receptors
(rods and cones) but only 10^6 fibres in the optic nerve. There's no
way that the optic nerve signals reflect what happens in each single
receptor. What gets transmitted up the optic nerve is a function of
the spatial and temporal patterns of light hitting the retina. That
function presumably favours those relationships of time and space that
usually help us control what we need to control in order to survive,
and inhibits temporal and spatial patterns that rarely occur or that
don't matter.
Think of high-quality coding schemes for computer-based movies. Even
with our limited abilities, we can reduce the number of bytes by at
least one order of magnitude by noting correlations across space and
time within the movie. The retina does something like that, but does
it rather better than we know how to do.
It certainly seems odd to suggest that there is no relation between
the light patterns on the retina and the signals in the optic nerve,
and I don't read the sentence from Gibson you quoted as saying there
isn't. What I read it as saying is that there is no _element by element_
correspondence between a photon hitting a retinal receptor and an
impulse in the optic nerve. And there isn't. And I don't think anything
in PCT suggests there might be.
Martin
[From Bill Powers (2000.02.12.1711 MST)]
Martin Taylor 20000212 09:57--
It certainly seems odd to suggest that there is no relation between
the light patterns on the retina and the signals in the optic nerve,
and I don't read the sentence from Gibson you quoted as saying there
isn't. What I read it as saying is that there is no _element by element_
correspondence between a photon hitting a retinal receptor and an
impulse in the optic nerve. And there isn't. And I don't think anything
in PCT suggests there might be.
If that is what Gibson meant, why didn't he say that? I can't see any
reason to assume that what he said is not what he meant to say. Gibson
appeared to believe in magic -- if he didn't, he was an incompetent writer.
Best,
Bill P.
[From Rick Marken (2000.02.12.1720)]
Martin Taylor (20000212 09:57) --
Think of high-quality coding schemes for computer-based movies. Even
with our limited abilities, we can reduce the number of bytes by at
least one order of magnitude by noting correlations across space and
time within the movie. The retina does something like that, but does
it rather better than we know how to do.
I think this is a poor analogy. The retina is not an encoding
mechanism, like "winzip", producing an X Meg "zipped" version of
a Y Meg retinal array of inputs (where Y>>X). The retina is a
perceptual mechanism, producing time varying scalar outputs as a
_function_ of retinal array inputs. These scalar outputs are not
a coded version of something else (the inputs to the function);
they _are_ the perception itself. At least, that's the PCT view.
Best
Rick
--
Richard S. Marken Phone or Fax: 310 474-0313
Life Learning Associates e-mail: rmarken@earthlink.net
http://home.earthlink.net/~rmarken/
REFERENCE:
VISUAL INPUT FUNCTIONS:
[From Erling Jorgensen (2000.02.13.0130 CST)]
Dan Palmer (2000.2.12.1658)
My understanding of PCT is that "the visual image" does get sent inward in the
form of first-order intensity signals which are combined into various
higher-level "representations" or signals and compared with similarly packaged
reference signals.
This reply is not about ecological psychology, and the specifics of these
comments may or may not interest you, Dan. But I wanted to point out a
reference which suggests both a working model and some of the transfer
functions that might be involved physiologically in the visual system.
Last summer I was reading a book by Vadim D. Glezer, a Russian
researcher, called _Vision and Mind: Modeling Mental Functions_.
(Mahwah, NJ: Lawrence Erlbaum Associates, 1995). It was heavy going,
and many times I could only follow the main contours of his research,
not all the details. But he seemed to be deriving approximations of
visual input functions, using them in working models, and comparing
the results to experimental data.
In Hierarchical PCT we have Bill's proposals about different classes
of perceptions. But except for the lowest levels, and occasional other
suggestions (such as Martin's flip-flop mechanism of categories, or Rick's
logical functions), we have very few specific proposals about how to
model perceptual input functions (pif's) corresponding to recognizable
perceptions. I know I have wondered, ever since running across a
schematic proposal by Warren McCullough years ago, about how
"configurations" could be simulated. More recently Bill has posted
(2000.01.24.1036 MST) about acceleration, velocity, and position,
and an image sharpening procedure that uses "optically analogous
quantities" to those integrated perceptions, (although I have a hard
time visualizing what those correspond to).
As for Glezer, I find his math beyond me, but it is intriguing that he
suggests actual weighting functions for receptive fields in the retina
(p.4), lateral geniculate nucleus (p.13), and striate cortex (p.28f.), and
an algorithm for texture segregation in the peristriate and prestriate
cortices (p.127f.). He also ties these areas to different types of
visual processing. I believe he even presents a comparison process
for discrimination, with equations providing a best fit to experimental
data (p.108), that sounds somewhat like a control loop. He also
sums up his findings in an extended series of equations, defining "the
algorithm of invariant image coding" (pp. 219-221), which from the
verbal descriptions sounds remarkably like what we might call a
configuration.
I certainly do not think Glezer's level of detail is necessary for doing PCT
modeling of visual processes. Weighted sums, for instance, can define a
generic "sensation," which is sufficient for examining some elementary
forms of control. In the same vein, there are a _variety_ of algorithms for
simulating color on a computer, for example; it certainly need not be the
precise way the human nervous system implements it. However, if
someone wanted to make a close copy of what the visual system is doing,
and fit it into an elaborate control model of a person, Glezer's work could
be a valuable resource.
Partially for the sake of the CSGNet Archives, and partially to guide and
wet the appetite of anyone who would like to pursue it, I have reproduced
below (with occasional insertions) quite a number of excerpts from Glezer's
book. The book is not exactly for the faint of heart, but I think it does
provide some mathematical and physiological bulk to the perceptions we
frequently talk about.
------------------------------------------------------------------
Excerpts [with comments and clarifications in brackets] from --
Glezer, V.D. (1995). _Vision and Mind: Modeling Mental Functions_.
Mahwah, NJ: Lawrence Erlbaum Associates.
[One of Glezer's chief goals is to present a model of how modules in the
visual cortex work, built up from the organization and structure at lower
levels.]
"The neurons of the retina and the LGN [lateral geniculate nucleus] measure
the integrated light energy in the central summation zone of their receptive
fields. The existence of an inhibitory periphery causes the neurons at this
level to carry out a series of preliminary processing operations: separating
the signal from surrounding noise, emphasizing contours and high spatial
frequencies, and producing spatial and temporal decorrelations of the image.
... Because of the appearance of separate on- and off-systems, a foundation
is created for the transition to local spectral analysis at the next level,
in the striate cortex." (pp. 1-2)
"Because of the algebraic summation of the spatially organized excitatory
and inhibitory process, the resulting profile has the form of _difference
of Gaussian_ (DOG). ... What operations are performed by such a retinal
construction? One of the operations the retina performs is retinal
adaptation. The purpose of adaptation is to keep the visual response the
same when the level of illumination changes; in other words, adaptation
allows invariant description, despite changes in illumination." (p. 4)
[This sounds like sensation being created through weighted sums, which
can be kept constant despite changes in intensities (i.e., illumination).
Glezer goes on to describe figure-ground distinctions (creation of
configurations?), which are derived from transformations of sensations
such as texture, depth, and color.]
"The RFs [receptive fields] at this level [the striate cortex] of the visual
system form a two-dimensional lattice of spatial-frequency filters. ...this
operation may be defined as a segregation of the figure from the background.
This operation is performed by nonlinear cells, which measure piecewise
power spectra and extract the areas filled in by homogenous texture; by
directional cells, which extract the areas composed of elements moving in
the same direction; by binocular cells, which extract the areas lying at
different depths; and by color cells." (pp. 15-16)
[For those who purchase the book, helpful schematic diagrams of
receptive fields at various levels occur on pages: 2, 12, 27, 48, 52,
76, 94, and 167.]
[Glezer next describes the columnar modules of the visual cortex.]
"Columnar organization is important at higher levels of the visual system.
It allows for the combining of two antagonistic types of description:
retinotopical (point to point) and spectral (distributed). Columnar
organization creates the foundation for local spectral analysis in the
cortex." (p. 61)
"the module is a device that performs a Fourier analysis of the image...
which I call the model of modules." (p. 64)
"The weighting functions of the cells of the module form a harmonic
[logarithmic] series: 1, 1.41, 2, 2.83, 4. The weighting functions of
each harmonic are out of phase 0, 90, 180, and 270 [degrees]. The
modules of different size form a lattice that subserves the piecewise
Fourier analysis of images of different size and position in the field of
vision." (p. 88)
"The modules represented in Fig. 3.14d as columns actually provide
identical spectral descriptions of an object, irrespective of its position
and size (i.e., of its localization in three-dimensional space)." (p. 86)
[This sounds like configurations which do not change despite changes
in their constituent perceptions. At a later point, Glezer describes the
process this way -- ]
"A special meaning is attributed to one part, the figure, which we refer
to as being extracted from the background. ... We describe these as
areas of identical brightness, color, or texture; I refer to these as
_subimages_. In this term, I also include areas composed of elements
moving in the same direction at identical speed, or, in the case of
binocular vision, areas composed of elements with identical binocular
disparity. The process of extracting subimages is referred to as _primary
segmentation_." (p. 118)
[Glezer suggests that the harmonics in the Fourier analysis are behind
such abilities as detection and recognition.]
"It is important to stress that narrowly tuned filters appear in the case
of recognition. ... The 1st harmonic with a broad bandwidth is enough
for detection. The narrowly tuned harmonics are needed for recognition.
... Ginsburg (1976) studied the effect of filtering high frequencies on
the recognition of individual faces, and showed that 3 to 4 harmonics
per image are enough for image recognition." (p. 102)
"I want to stress that recognition is performed by harmonics; that is,
by relative frequencies, not by absolute frequencies. We can observe
an image from a distance and still recognize the object if 4 to 5
harmonics are kept (assuming, of course, that these frequencies can
be seen at this distance)." (p. 104)
[A later summary characterizes it this way -- ]
"The foundation for the appearance of some of the simplest types of
invariance are created by the organization of the modules of the
striate and prestriate cortices. The modules are chosen according
to the size and position of the image on the retina, and produce
identical patterns of excitation in the cells of the module, regardless
of the size and distance of the object, because the harmonic composition
of the modules is identical. The spectrum of the image is invariant with
transformations of size, distance, and position, but the number of the
module gives information about these changes." (p. 146)
[Glezer also tries to treat how such information is used by higher visual
processes, although not with the same detail and quantification of the
earlier processes. It is interesting that describing relationships and
classifying (which also appear in the HPCT taxonomy) are two of
the processes he deals with. A sample of this treatment follows.]
"perceiving a complex figure as a whole and describing the details
and spatial relations between them are performed by different
mechanisms. ... According to clinical data, these mechanisms are
localized in different areas of the visual cortex. The classification
mechanism is located in ITC [inferotemporal cortex], and the
mechanism for the description of spatial relationships is located in
the PPC [posterior parietal cortex]. In man, a lesion in the ITC
causes visual object agnosia (i.e., nonrecognition of objects...),
and a lesion in the PPC causes spatial agnosia." (p. 145)
"Patients suffering from object agnosia caused by lesions of the
ITC cannot recognize an object presented to them. They do not
recognize a pen or a comb, but describe them as long narrow
objects...; that is, they perceive their spatial characteristics."
(p. 152-3)
[Finally, for those with the requisite mathematical background,
a precise and elaborate "algorithm of invariant image coding,"
with equations for each step from the retina to his model of a
module in the striate cortex, is presented by Glezer on pages
219-221. I believe his "image code, invariant to position and
size (position in depth)" (p.221) is closest to what we in CSG
might mean by the term visual configuration.]
All the best to anyone who might want to pursue this further!
Erling Jorgensen
[From Bill Powers (2000.02.13.0216 mst)]
Rick Marken (2000.02.12.1720)--
The retina is not an encoding
mechanism, like "winzip", producing an X Meg "zipped" version of
a Y Meg retinal array of inputs (where Y>>X). The retina is a
perceptual mechanism, producing time varying scalar outputs as a
_function_ of retinal array inputs. These scalar outputs are not
a coded version of something else (the inputs to the function);
they _are_ the perception itself. At least, that's the PCT view.
I completely agree, Rick. The idea of "encoding" is a confused model of
perception which only postpones the "decoding" stage, after which we must
still explain how it is that various kinds of visual perception exist and
how it is that some perceptions are functions of others. Coding is not the
answer to Gibson. Perhaps Gibson's strange concepts of perception were
intended to be an answer to the coding idea.
The greatest mystery about perception is how we know about it. In the
visual realm, how do we know about the intensity type of perception? The
most obvious concept of knowing is that it takes place in higher regions of
the brain. However, if intensity signals are those which sensory receptors
emit, they must exist in the retina, and if we are aware of intensity at
high resolution, we must be aware of signals in the retina, because as
Taylor points out, there wouldn't be enough optic-nerve fibers to carry all
the intensity signals that are generated if they are generated with the
same spatial density over the whole retina. That doesn't square with the
idea that awareness is a higher brain function.
One possible solution may lie in the known reduction of visual acuity away
from the center of vision. The maximum acuity is achieved only over the
central 2 degree cone of rays, which is about 0.004 of the area of the
retina if I've done my steradian calculations right. It would be possible
to preserve full resolution over the central 2 degrees while leaving a good
fraction of the optic nerve fibers for carrying higher-order perceptual
signals generated in the retina.
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? This is the only way I can see in which we could get an impression
of full resolution over the whole visual field. Of course it's not true
that we have such full resolution over the whole visual field at any one
instant. To get the maximum resolution of anything in the field of view, we
must bring it to the center of vision, which makes what had formerly been
sharp more fuzzy. If an impression of clarity of the former region exists,
it can only exist as continued activity in a more central map, where a
wider range of full resolution would, in principle, be possible. As we move
our eyes from one fixation point to another, we are updating the central
map with maximum resolution in small regions of it. This might not preserve
the actual full resolution, but it might make possible preserving
higher-order information (the kind of letter we are reading) that depends
on resolving the visual field in a given place at high resolution for long
enough to distinguish the configuration and register its presence.
We know that intensity information does reach midbrain structures: the iris
reflex proves it.
Best,
Bill P.
[From Bill Powers (2000.02.13.0315 MST)]
Erling Jorgensen (2000.02.13.0130 CST)--
... I wanted to point out a
reference which suggests both a working model and some of the transfer
functions that might be involved physiologically in the visual system.Last summer I was reading a book by Vadim D. Glezer, a Russian
researcher, called _Vision and Mind: Modeling Mental Functions_.
(Mahwah, NJ: Lawrence Erlbaum Associates, 1995).
I haven't read Glezer, but from your excerpts it seems he is tackling some
very difficult problems. I think he's perfectly right to be looking for
some operation other than weighted summation to explain the perception of
configurations. I've always felt that the perceptron approach is fine for
recognizing _features_ (what I would class as sensations), but isn't right
for
configuration perception which has to be invariant in more ways than
sensations are.
The concept of edge enhancement or contouring is, I think, a red herring.
In discussing visual perceptual functions, one mustn't forget to open the
eyes once in a while and look around. Perhaps I'm peculiar, but I don't see
any contours in the scene around me. At the edges of objects, one color or
brightness simply ceases and another appears. It's hard to look _exactly
at_ an edge, because in fact there's nothing there to see.
If contouring does occur, its function must be that of restoring, to the
extent possible, resolution lost through interactions among adjacent
neurons or spreading of light energy as it filters through cell bodies on
the way to the sensitive volumes of the retina. Just the right amount of
edge enhancement can overcome some loss of resolution, but if it's
overdone, one gets a totally unrealistic-looking visual scene, like a badly
drawn comic strip. And in fact, in papers on visual image processing, the
scenes shown after contouring bear very little resemblance to the scenes we
normally see. I've always wondered how anyone could fail to see the
difference.
Glezer's approach, in which distributed perceptions such as configurations
are derived through distributed functions like the Fourier transform, seems
much more realistic to me. We aren't trying to get a set of signals that
"look like" the original object -- if that were the case, we'd still have
to construct a recognizer that could see the object in those signals. All
we need is a _single signal_ that covaries with the presence of a given
kind of configuration. The Fourier transform may not be the right one, but
it's the right kind of transformation for boiling down a distributed
property of a field of signals to a single signal.
The job isn't made easier by statements about perception that clearly
aren't true. Glezer says that figure/ground distinctions begin with figures
that have "areas of identical brightness, color, or texture." That may be
true of silhouette drawings of faces and vases, but it's not true of most
objects. You can distinguish your own hand as a figure against a
background, but this figure is made of a large variety of sensations which
are anything but identical within the boundaries of the hand. I think this
is a case where imagination has replaced observation. It may be that
commonalities within the boundaries do create the sense of a figure, but if
so they are something other than "brightness, color, or texture". Consider
the "figures" with which Julesz worked: they are made of collections of
random light and dark squares. All that creates a figure is the common
depth signal.
I find Glezer's discussions of Fourier components confusing. He seems to
shade over from _spatial_ frequencies to _temporal_ frequencies without
noticing it. A spatial frequence can't be converted to a temporal frequency
without specifying some scanning procedure, and in two or more dimensions
this can get rather tricky. Perhaps in the book Glezer discusses this
problem. You can't really speak of "tuning" to a spatial frequency; the
process wouldn't be at all the same as tuning to a temporal frequency, not
as I understand it. The resonance that underlies temporal tuning arises
from time derivatives and integrals, but as far as I know there is no
counterpart in spatial frequencies unless the spatial frequencies are
converted to temporal ones by scanning of some sort. And then the
properties of the scanning process become part of the final result. I
could, of course, be demonstrating my mathematical shortcomings here.
Those same shortcomings make me hesitate to tackle the math in Glezer's
book. I don't have the grasp of things like Fourier transform theory to
know whether the math I'm reading is the straight poop or just a lot of
impressive arm-waving. Just writing down a lot of equations copied out of
handbooks isn't enough to form a theory of a physical system. One has to
understand the underlying meaning of the mathematical expressions in terms
of physical relationships. I should think that someone like Richard
Kennaway, who is a Real Mathematician, would be the one to tell us if there
is substance there. It would be really lovely if there were.
Best,
Bill P.
[Martin Taylor 20000212 13:38]
[From Rick Marken (2000.02.12.1720)]
Martin Taylor (20000212 09:57) --
> Think of high-quality coding schemes for computer-based movies. Even
> with our limited abilities, we can reduce the number of bytes by at
> least one order of magnitude by noting correlations across space and
> time within the movie. The retina does something like that, but does
> it rather better than we know how to do.I think this is a poor analogy. The retina is not an encoding
mechanism, like "winzip", producing an X Meg "zipped" version of
a Y Meg retinal array of inputs (where Y>>X). These scalar outputs are not
a coded version of something else (the inputs to the function);
they _are_ the perception itself. At least, that's the PCT view.
Sorry, but I can't see a distinction between what I said and what you
say, other than the wording. Here's what I said just before the bit
you quote:
What gets transmitted up the optic nerve is a function of
the spatial and temporal patterns of light hitting the retina. That
function presumably favours those relationships of time and space that
usually help us control what we need to control in order to survive,
and inhibits temporal and spatial patterns that rarely occur or that
don't matter.
How is this different from your:
The retina is a
perceptual mechanism, producing time varying scalar outputs as a
_function_ of retinal array inputs.
I fail to understand how the output of such a function is "not a coded
version of something else (the inputs to the function)." Why is this
function unlike _all_ other functions, for which the output is
inevitably a coded version of their inputs.
Indeed, the scalar outputs _are_ the perception itself. At least we
can agree on that. Perhaps you can elighten me as to where you think
we disagree?
Or perhaps you think that the output of "winzip" is not a function of
the its input? I'm a bit baffled.
Here's another possibility that comes to mind as to where you might
be seeing a disagreement, but it's a bit far-fetched. We know it is
quite probable that the actual coding function is affected by an
"outflow" signal in the sense that the reference signal to a control
unit is an outflow signal. Were you suggesting that I was ignoring
that possibility? I wasn't, but I thought it an unnecessary
complication to the main point that there's no one-to-one
correspondence between what hits the optic receptors and what goes up
the optic nerve--as EVERY perception researcher has known for at
least a century, if not longer, Gibson included.
Martin
[From Rick Marken (2000.02.13.1500)]
Me:
The retina is not an encoding mechanism
Martin Taylor (20000212 13:38)
Sorry, but I can't see a distinction between what I said
and what you say, other than the wording. Here's what I
said just before the bit you quote:> What gets transmitted up the optic nerve is a function of
> the spatial and temporal patterns of light hitting the retina.
> That function presumably favours those relationships of time
> and space that usually help us control what we need to control
> in order to survive, and inhibits temporal and spatial patterns
> that rarely occur or that don't matter.How is this different from your:
> The retina is a perceptual mechanism, producing time varying
> scalar outputs as a _function_ of retinal array inputs.
Does Bill's post [Bill Powers (2000.02.13.0216 mst)] help? If not,
I'll give it a shot. I guess my basic complaint is that I don't
see _encoding_ as being anything like _perceiving_; encoding
is data reduction (necessary for data transport when there are
bandwidth limitations); perceiving is data representation
(necessary for control of any particular representation of the
data).
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 Bruce Nevin (2000.02.13.2037)]
"Encoding" is a metaphor that implies discrete bits of information. Is this
metaphor appropriate for an analog process? Or is the claim that between
retina and optic nerve already we have moved up to category level? Or are
"feature detectors" at this level already mapping analog data to discrete
(binary) values, and if so how are these not categorial?
Bruce Nevin
[From Bill Powers (2000.02.14.0318 MST)]
Martin Taylor 20000212 13:38--
I thought Rick's point was quite valid. If the function of the retina were
merely to compress data, then we would assume that in the midbrain we would
have decompression to restore the original resolution. That's what your
remarks
seem to imply:
> Think of high-quality coding schemes for computer-based movies. Even
> with our limited abilities, we can reduce the number of bytes by at
> least one order of magnitude by noting correlations across space and
> time within the movie. The retina does something like that, but does
> it rather better than we know how to do.
These coding schemes for computer-based movies require many fewer than the
original number of bits to transmit pictures, but before they can be
viewed, the compressed pictures have to be expanded again into a
full-resolution display. So your example implies that the original
resolution is recreated in the midbrain, and I don't think there's any
evidence that this takes place.
If, on the other hand, the retinal computations create higher-level
perceptions, the signals representing those perceptions would be
transmitted up the optic nerve and would be used as they are. If they were
sensation signals, they would not be unpacked again into intensity signals
at the destination (even if that were possible, which it probably isn't).
In a post yesterday I suggested that some fraction of the optic nerve
fibers could carry high resolution mappings of central vision, leaving
enough fibers to carry higher-order signals as well (though I don't know
how they would be sorted out at the destination). That scheme doesn't
involve any compression/decompression methods.
Best,
Bill P.
[From Rick Marken (2003.07.09.1130)]
Marc Abrams (2003.07.09.1200) re:
CSGNET list (86 recipients).
Not an impressive trend over the last 2 years
and
I don’t think it’s due to my posts…
Did
the list lose
10 people overnight
I am in the process of vetting CSG list. I sent an e-mail to all
128 e-mail addresses on the CSGNet membership list asking that people respond
“yes” if they are still on CSGNet. 31 of these messages bounced so
I deleted them from the membership list. So the CSG membership list actually
lost 31 names overnight and it now has 97 names. Some people have two addresses
and some of the names may be people who are no longer reading the list.
So the actual readership of CSGNet is probably considerably less than 97.
(I don’t know why the message you says 86 rather than 97; the membership
list currently contains 97 names).
My guess about the actual size of CSGNet (which was based on little
more than intuition) was that there were about 30-40 people actively involved
But it looks like the actual number may be as high as 50. I base this on
the fact that I have so far received 36 “yes” replies (not including myself
and Bill Powers, so that’s a total of 38 so far). And I still have not
heard from several people (like yourself) who I know are on the net because
they post.
I was thrilled to get “yes” replies from people who I haven’t heard
from in some time. It’s nice to see that there is a wonderful group of
people out there reading this list. I’m delighted that we may have
as many as 50 people on the list. It looks like it really does matter what
we say about PCT on this list! As they say at the movies “The audience
is listening”. So it looks like I will continue to reply to misinformation
(and disinformation) about PCT as best as I can.
One thing my work on the membership list shows is that the number of
people to whom messages are distributed on CSGNet (as per the acknowledgment
sent when you post a message) is not a good measure at all of the number
of people actually getting messages from CSGNet or the number of people
who are reading them. I’m trying to vet the list so we can get a
better idea of the size of its current readership. But even if I can make
sure that the list contains only the addresses of active participants I
don’t think the number of names on the CSGNet membership list will remain
a valid measure of the size if the CSGNet readership for very long.
I think the number of addresses on CSGNet will always be an overestimate
of the number of people actually reading CSGNet (because people will occasionally
leave without unsubscribing, for example) and an underestimate of the number
of people in the world who are interested in PCT (because there are people
out there who are doing PCT but who are not on CSGNet – at least I know
of one).
Best regards
Rick
–
Richard S. Marken, Ph.D.
Senior Behavioral Scientist
The RAND Corporation
PO Box 2138
1700 Main Street
Santa Monica, CA 90407-2138
Tel: 310-393-0411 x7971
Fax: 310-451-7018
E-mail: rmarken@rand.org
From [ Marc Abrams (2003.07.09.1200) ]
Not an impressive trend over the last 2 years and I don’t think it’s due to my posts.
Your message dated Fri, 21 Dec 2001 11:39:30 -0500 with subject “Any body
out there?” has been successfully distributed to the CSGNET list (102
recipients).
Your message dated Tue, 8 Jul 2003 17:55:21 -0400 with subject “Re: Levels
of analysis (Addendum to Bill)” has been successfully distributed to the
CSGNET list (96 recipients).
Today:
Your message dated Wed, 9 Jul 2003 11:27:45 -0400 with subject “Re: Levels
of analysis” has been successfully distributed to the CSGNET list (86
recipients).
Did the list lose 10 people overnight?
Marc
[From Rick Marken (2003.07.09.1250)]
Bill Powers pointed out to me that my efforts to get a valid list of CSGNet
readers may have led to the elimination of some actual members. I was only
eliminating from the CSG membership list e-mail addresses that bounced. one of
those that bounced was David Goldstein's, which I found puzzling because I new he
was on the list. Bill said this might have been the result of David's spam filter.
So now I have to confront the possibility that all the bounced mail (30 names) was
dues to spam protection. So I reinstated all the names bounced from the list
except 7 which I know to be obsolete.
So I guess my "vetting the list" days are over since I can't remove names based on
mail bounce since any bounce could be the result of pam filter problem. And I
don't want to eliminate people just because they don't acknowledge receipt of a
certified e-mail from me in a certain time because I don't know what the
appropriate time would be. And I couldn't get to people with spam filters anyway.
It is nice to get those "yes" responses to my personal posts, though. Although it
doesn't give an accurate total for the number listening at least it shows that
there are quite a few of you out there listening, and that's nice to see. So
thanks for the replies and if you haven't replied yet (either because my "CSGNet
Check" post was filtered out ot because you just haven't yet) I would appreciate
it if you would take the time to do it.
Best regards
Rick
--
Richard S. Marken, Ph.D.
Senior Behavioral Scientist
The RAND Corporation
PO Box 2138
1700 Main Street
Santa Monica, CA 90407-2138
Tel: 310-393-0411 x7971
Fax: 310-451-7018
E-mail: rmarken@rand.org
From [ Marc Abrams (2003.07.09.1631)
I’m just about ready to sign off. I am responding to Bill’s reply to my ‘I’m convinced’ post to wrap up my stay on CSGnet.
First, Rick, I received no e-mail from you concerning this list. I can’t reply to what I didn’t get.
Bill, If I misrepresented your views I apologize. I don’t apologize for your insistence that your model is an accurate portrayal of the CNS. I don’t believe it is even an accurate metaphor. Yes, I do believe that behavior is the control of perception. HOW that actually happens is where you and I diverge. I too believe it’s a control process. Actually 2. One cognitive, one intrinsic. I believe the ‘intrinsic’ one, the one I have been calling ‘reflexive’ is somewhat equivalent to your ‘reorganization’. The ‘cognitive’ is similar to your ‘learned hierarchy’, except I believe there are different kinds of discriminatory hierarchy’s embedded in networks, and I don’t think your concept of sensory binding is accurate. I don’t think your take on emotions are accurate, and I don’t think your memory model is a good one.
I am not saying these things to demean your work. The one basic principle of your theory I believe is rock solid (i.e. we control our perceptions not our behavior). Given the technology and research methods of the time I think you did an admiral job. But to stand there and reject all other work out of hand without even looking at it is the worst kind hubris. I’m not saying I am right about anything I have said. I believe what I said to be accurate based on the data I have today. That could change tomorrow.
I won’t be responding to this post on CSGnet. If you want to contact me CC me at mabrams@nvbb.net
----- Original Message -----
From:
Richard Marken
To: CSGNET@listserv.uiuc.edu
Sent: Wednesday, July 09, 2003 11:32 AM
Subject: Re: What’s up?
[From Rick Marken (2003.07.09.1130)]
Marc Abrams (2003.07.09.1200) re: CSGNET list (86 recipients).
Not an impressive trend over the last 2 years and
I don’t think it’s due to my posts… Did the list lose
10 people overnight
I am in the process of vetting CSG list. I sent an e-mail to all 128 e-mail addresses on the CSGNet membership list asking that people respond “yes” if they are still on CSGNet. 31 of these messages bounced so I deleted them from the membership list. So the CSG membership list actually lost 31 names overnight and it now has 97 names. Some people have two addresses and some of the names may be people who are no longer reading the list. So the actual readership of CSGNet is probably considerably less than 97. (I don’t know why the message you says 86 rather than 97; the membership list currently contains 97 names).
My guess about the actual size of CSGNet (which was based on little more than intuition) was that there were about 30-40 people actively involved But it looks like the actual number may be as high as 50. I base this on the fact that I have so far received 36 “yes” replies (not including myself and Bill Powers, so that’s a total of 38 so far). And I still have not heard from several people (like yourself) who I know are on the net because they post.
I was thrilled to get “yes” replies from people who I haven’t heard from in some time. It’s nice to see that there is a wonderful group of people out there reading this list. I’m delighted that we may have as many as 50 people on the list. It looks like it really does matter what we say about PCT on this list! As they say at the movies “The audience is listening”. So it looks like I will continue to reply to misinformation (and disinformation) about PCT as best as I can.
One thing my work on the membership list shows is that the number of people to whom messages are distributed on CSGNet (as per the acknowledgment sent when you post a message) is not a good measure at all of the number of people actually getting messages from CSGNet or the number of people who are reading them. I’m trying to vet the list so we can get a better idea of the size of its current readership. But even if I can make sure that the list contains only the addresses of active participants I don’t think the number of names on the CSGNet membership list will remain a valid measure of the size if the CSGNet readership for very long.
I think the number of addresses on CSGNet will always be an overestimate of the number of people actually reading CSGNet (because people will occasionally leave without unsubscribing, for example) and an underestimate of the number of people in the world who are interested in PCT (because there are people out there who are doing PCT but who are not on CSGNet – at least I know of one).
Best regards
Rick
Richard S. Marken, Ph.D.
Senior Behavioral Scientist
The RAND Corporation
PO Box 2138
1700 Main Street
Santa Monica, CA 90407-2138
Tel: 310-393-0411 x7971
Fax: 310-451-7018
E-mail: rmarken@rand.org
[From Rick Marken (2003.10.21.1400)]
Any progress on the "adaptive illusion"? Economics test bed? Chinese PCT? Paperback
B:CP?
I'm still plugging away at the error modeling.
Best regards
Rick
--
Richard S. Marken, Ph.D.
Senior Behavioral Scientist
The RAND Corporation
PO Box 2138
1700 Main Street
Santa Monica, CA 90407-2138
Tel: 310-393-0411 x7971
Fax: 310-451-7018
E-mail: rmarken@rand.org
Hi, Rick --
[From Rick Marken (2003.10.21.1400)]
Any progress on the "adaptive illusion"? Economics test bed? Chinese PCT?
Paperback
B:CP?
BCP is supposed to come out this month, if Alice's health holds up.
The other stuff is still on the shelf, waiting for some inspiration to
strike. I did write to Julius Soros to see if he would have any interest in
the economics stimulation, but haven't heard back.What about the economists
at RAND? By the way, is Rand an acronym or a name?
I'm sort of waiting for some sort of communication from Flach. If he gets
back in touch, I might give that MOL approach a try. I've also been
thinking about your suggestion of doing the adaptive illusion paper for two
orders of system dynamics instead of all three.
Also, reading that long thing from McRuer that you sent has been very
informative. The stick had a rather strong centering spring (two pounds
force at a 4-inch radius for a 15-degree deflection -- the total range was
+/- 15 degrees). Deflection was left-right only, so a forearm roll rotation
was being used as output. The dynamics were added electronically after the
measure of joystick angle, The disturbance was added after the dynamics,
and the "error" was shown on an oscilloscope screen as a deflection of a
spot from screen center. The screen was inscribed with a few lines so the
center could be located, but there was no target spot on the screen. This
was compensatory tracking only.
The disturbance itself was a sum of 10 sine-waves. Three disturbances were
used with 7,8, and 9 sine waves of 1/2-inch amplitude and, for each case,
the remaining 3, 2, or 1 sine-waves having only 1/10 that amplitude. These
smaller sine waves were at frequencies in the crossover region, where
amplitude was falling off rapidly as frequency increased. they were used
mainly for checking the frequency response at three points between 1 and 17
radians per sec. A list of the sine-wave frequencies was given so we can
reproduce the disturbance if we wish.
Haven't heard from Zhang Huaxia in a while. I'm still sporadically
practicing my pronunciation of Chinese from written Pinyin (romanized),
using a CD-ROM language program. Slow progress.
I spent the last three+ weeks nursing a very sore back -- turns out that
one disc is missing, at about L3, so the vertebrae are touching at an angle
instead of being separated and parallel. No idea when that happened -- it
might have been years ago. So I have to watch the heavy lifting and do some
exercises. Unpleasant.
I seem to be taking a vacation from programming right now, so nothing
startling to report.
Oh, one thing: Tim Carey is starting to work on a European CSG meeting for
June 2004. I expect it to slip to late 2004 or to 2005 as he finds out the
realities of getting people to commit to meeting. Scotland, of course.
Best,
Bill
At 11:58 AM 10/21/03, you wrote: