Is this a controlled perception?

I have an object in front of me, say, a copy of B:CP. I pick it up and
move it around; as I do so, certain visual and tactile perceptions of
mine change. But there appears to be another perception which does not
change: the perception of "this book". Does this perception fit into
the HPCT hierarchy? It seems to be a controlled perception, remaining
constant despite the changing inputs, but the behaviour which controls
it happens entirely within the brain.

-- 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.

[Avery Andrews 970824]
  (Richard Kennaway 970823)

I have an object in front of me, say, a copy of B:CP. I pick it up and
move it around; as I do so, certain visual and tactile perceptions of
mine change. But there appears to be another perception which does not
change: the perception of "this book". Does this perception fit into
the HPCT hierarchy? It seems to be a controlled perception, remaining
constant despite the changing inputs, but the behaviour which controls
it happens entirely within the brain.

The way I see it, there isn't a *controlled* perception of `this book'
until there's some action that doesn't stop until `this book' is
perceived. As for example you want to look at `this book', and
search around your house and office until you find it. As for the
question of what constitutes a perception of `this book', there's
a huge philosophical literature on this and related issues, for example
if you replace every plank of a wooden ship one at a time, it the
same ship when you're done? There's a lot of work do be done on
the nature of the higher levels of perception. One feature of
standard theories of this (which I've taken for granted in my occasional
musings on language structure) is that they have concepts of `entity'
and `identity of entities' (we perceive there to be an issue as to
whether the person we see in the black coat in the distance right now is
or isn't the person we met at the cocktail party yesterday). But
maybe this is all wrong too. Philosophers in the middle ages believed
in concepts of `form' and `substance' which are virtually
incomprehensible today.

Avery.Andrews@anu.edu.au

[From Bill Powers (970824.0137 MDT)]

Richard Kennaway (970824) --

I have an object in front of me, say, a copy of B:CP. I pick it up and
move it around; as I do so, certain visual and tactile perceptions of
mine change. But there appears to be another perception which does not
change: the perception of "this book". Does this perception fit into
the HPCT hierarchy? It seems to be a controlled perception, remaining
constant despite the changing inputs, but the behaviour which controls
it happens entirely within the brain.

Behavior controls _variables_. So you must ask, what is variable about the
perception we call "this book?" I think it helps to recall that in HPCT
labels are generated at the category level, and to remember to separate the
perception of configuration from labelling the perception by using another
perception ("this book", written or spoken, which you certainly cannot read
with the same result as reading the actual book). See Lewis Carroll and his
discourse on a song: the name of the song, what the song is called, what
the song _is_, and so forth.

The category to which we attach the noises or marks, "this book," is
exemplified by many different configuration perceptions and other
lower-level perceptions. Among them are a book lying open on a table, a
book on a shelf with only its spine showing, a book seen from many
different angles, a title "War and Peace", printed or spoken, passages from
the book, relations with the book ("library book," "my book"), sensations
("heavy book", "red book") and so on. At the lower levels these are all
variable perceptions that can, in some cases, be affected by our behavior.

It's useful to recall, also, that in HPCT perceptual functions are in
general many-to-one functions: a single scalar perceptual signal is a
function of many lower-level signals. This means that there are ways in
which the input perceptual signals can change that that leave the output
signal unchanged, and other ways in which the input signals can change that
do change the output signal. The mathematical way of putting this is to say
that the output perceptual signal is invariant under some transformations
of the input signals, but not under other transformations. These
transformations define, if memory serves, "orthogonal trajectories" in the
space defined by the input function. Changes along one set of trajectories
leave the perceptual signal unchanged; changes along the orthogonal set
alter the perceptual signal.

The perception of "this book" is invariant under many transformations:
rotation, translation, illumination, ownership, and so on. However, it is
not invariant under others: changing the title, the color, the number of
pages, what is written on the pages, or the author, for example. Of course
this depends on how the category is formed: you spoke of "a copy of a
book," so by "this book" you may actually mean any copy that has identical
attributes, or you may mean any printed volume of any size, color, or shape
that contains the same words and is attributed to the same author.

If a book's attributes change so it does not quite meet the conditions of
the category perceptual function, less of that category will be perceived.
We will say that what we're perceiving might be "this book", or is rather a
poor example of "this book" (a book collector might be quite fussy about
this). If the changes in the lower-level perceptions are sufficiently
great, we will say that what we are perceiving is not "this book" at all --
it's an imposter. It's a different book, although we still accept it as "a
book."

···

---------------------------------
[Avery Andrews 970824] says

The way I see it, there isn't a *controlled* perception of `this book'
until there's some action that doesn't stop until `this book' is
perceived.

I would agree with this, although there are other ways one can correct the
perception of "this book" by actions other than looking for it. If someone
tears out a page, for example, you can obtain another one and glue it in.
You resist disturbances that might change the perception of "this book" if
allowed to continue.

As for the
question of what constitutes a perception of `this book', there's
a huge philosophical literature on this and related issues, for example
if you replace every plank of a wooden ship one at a time, it the
same ship when you're done?

In terms of HPCT this is a mis-stated question. If you ask "is it perceived
as the same ship when you're done," there is no problem: of course it is,
if the restoration is perfect. What the philosophers are doing is reifying
perceptions. The question they think they're asking is unanswerable, or
else the answer is merely a matter of convention -- the meaning of "same."
If you simply admit that it's ALL perception, the problem goes away, or
changes into a tractable one.

There's a lot of work do be done on the nature of the higher levels of
perception.

I agree!

One feature of
standard theories of this (which I've taken for granted in my occasional
musings on language structure) is that they have concepts of `entity'
and `identity of entities' (we perceive there to be an issue as to
whether the person we see in the black coat in the distance right now is
or isn't the person we met at the cocktail party yesterday).

That could be as simple a problem as asking whether the configuration we
are seeing is the same as the one we remember. Or it could be a problem at
the category level: is this that person, or his twin who is called by a
different name? Or it could be at the system concept level: is this person,
who told me at the party day before yesterday that he was going to receive
electroshock therapy yesterday going to be the same person when I talk to
him today? When you start thinking in terms of levels of perception, a lot
of these old problems get a different look.

Best,

Bill P.

Bill Powers (970824.0137 MDT):

Behavior controls _variables_. So you must ask, what is variable about the
perception we call "this book?"

The degree to which it is present (although it's usually experienced as
an all-or-nothing perception).

The perception of "this book" is invariant under many transformations:
rotation, translation, illumination, ownership, and so on.

In this (i.e. HPCT) view, constructing the perception "this (copy of
this) book" out of the raw sensory inputs at the bottom level arises
from a causal chain of computation from the latter to the former. When
this chain is part of a control loop, the control is performed by
emitting muscular outputs acting on the environment.

I'm wondering what is happening when a perception on that level changes
without any apparent muscular outputs, as in the case of the Necker cube
I brought up a few weeks ago. Martin Taylor also talked about the
phenomenon of hysteresis in the category-level perception of an input
which changes smoothly from one which gives rise to one category
perception to one which gives rise to another. Without wishing to
reopen the argument over the hypothesis which he came up with to explain
that, I'm still wondering just what is going on in such cases.

-- 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 Bill Powers (9760824.0331 MDT)]

Richard Kennaway (970824) --

I'm wondering what is happening when a perception on that level changes
without any apparent muscular outputs, as in the case of the Necker cube
I brought up a few weeks ago.

First of all you have to realize that the number of perceptions of this
class is limited; most ordinary perceptions contain enough detail to
eliminate ambiguities. The Necker Cube and other similar phenomena occur
mainly when depth information is eliminated, leaving the way open to supply
different depth information from imagination. Imagination signals are
controlled just like real-time perceptions, except that no motor action is
required.

Martin Taylor also talked about the
phenomenon of hysteresis in the category-level perception of an input
which changes smoothly from one which gives rise to one category
perception to one which gives rise to another.

Again, while one can find positive examples of hysteresis, there are many
more examples of categories which are not mutually exclusive, such as "red"
and "sweet". For those cases where one can "morph" continuously from one
category to another, Martin Taylor's idea of flip-flop cross connections is
certainly one kind of circuitry that could accomplish the observed result.
However, there are other possible explanations, such as saying the the
exclusive-or is imposed at the logic level, along with local positive
feedback to generate the hysteresis. Finding ways to implement this effect
is not the problem -- the problem is to find experimental ways of
distinguishing between the possible models.

And it's time I went back to bed like a sensible person.

Best,

Bill P.

···

At 10:09 AM 8/24/97 +0100, you wrote