Attaching labels

CSGnet Archive CSG2004
21

[From Bruce Gregory (2004.1102.0927)]

Bill Powers (2004.11.02.0623 MSR)

No, you are simply using the word "know" in a fuzzy way. You mean, I
think,
that you could quickly come up with the name if there were some reason
to
do so, such as being asked for it. This says nothing about the form
taken
by this knowledge. If you take the "association" idea literally, then
everything you see for which you "know" a word will evoke the word
every
time you see it -- you would live in a world in which everything
incessantly and annoyingly kept telling you in words what it is. The
din
would be terrible. I see no evidence at all for that sort of
explanation of
how words are "attached" to other experiences. I don't think they're
attached at all -- that implies an automatic process and I don't think
it's
automatic. I think we attach them, as needed. I see nothing to indicate
otherwise.

You seem to saying that every perceptual signal must be accompanied by
conscious awareness. Since I know you do not believe this, I find your
analysis less than fully persuasive.

Bruce Gregory

22

[From Bill Powers (2004.11.02.0752 MST)]

Bruce Gregory (2004.1102.0927)--

You seem to saying that every perceptual signal must be accompanied by
conscious awareness. Since I know you do not believe this, I find your
analysis less than fully persuasive.

Try the shoe on the other foot. Since you claim that perceptions of the
names occur whenever any named experience occurs, it's up to you to present
the evidence that this is true. All the evidence I have says it is not
true. Even your own postulate is that such evoked name perceptions are
unobserved. So the only reason I can think of for supposing that such
signals exist is that your theory requires that they exist. Do you have
some other reason?

Best,

Bill P.

23

[From Bruce Gregory (2004.1102.1027)]

Bill Powers (2004.11.02.0752 MST)

Try the shoe on the other foot. Since you claim that perceptions of the
names occur whenever any named experience occurs, it's up to you to
present
the evidence that this is true. All the evidence I have says it is not
true. Even your own postulate is that such evoked name perceptions are
unobserved. So the only reason I can think of for supposing that such
signals exist is that your theory requires that they exist. Do you have
some other reason?

I maintain that we associate names and objects by Hebbian learning. Do
you have another model in mind?

Bruce

24

[Fromn Bill Powers (2004.11.02.0839 MST)]

Bruce Gregory (2004.1102.1027)--

I maintain that we associate names and objects by Hebbian learning. Do
you have another model in mind?

Hebbian learning is not a model, it's a very rough idea with nothing in it
to make it work. How does the name "long division" get attached to your
perception of the processes you carry out to accomplish it? How does
"strengthening a synapse" cause association to take place? Why should
learning result from strengthening a synaptic response? No theory of
perception that has actually been modeled (such as neural networks) works
only by making the synaptic weightings greater. And exactly what does a
synapse have to do with perception in Hebb's system? He never said.

Hebb's theory was not a viable model, in my opinion.

Best,

Bill P.

···

Bruce

25

[From Bruce Gregory (2004.1102.1057)]

Bill Powers (2004.11.02.0839 MST)

Hebb's theory was not a viable model, in my opinion.

And you proposed alternative is?

Bruce Gregory

26

[From Rick Marken (2004.11.02.1100)]

Bruce Gregory (2004.1102.1057)--

Bill Powers (2004.11.02.0839 MST)

Hebb's theory was not a viable model, in my opinion.

And you proposed alternative is?

More irony?

My guess is that Bill's alternative would be perceptual control theory.

RSM

···

--
Richard S. Marken
MindReadings.com
Home: 310 474 0313
Cell: 310 729 1400

--------------------

This email message is for the sole use of the intended recipient(s) and
may contain privileged information. Any unauthorized review, use,
disclosure or distribution is prohibited. If you are not the intended
recipient, please contact the sender by reply email and destroy all copies
of the original message.

27

[From Bruce Gregory (2004.1102.1440)]

Rick Marken (2004.11.02.1100)

Bruce Gregory (2004.1102.1057)--

Bill Powers (2004.11.02.0839 MST)

Hebb's theory was not a viable model, in my opinion.

And you proposed alternative is?

More irony?

No.

My guess is that Bill's alternative would be perceptual control theory.

Then all perceptions are controlled and all learning is reorganization?
When a child learns that the furry creature is called a dog, what
perception is the child controlling and what is the reference level?

Bruce Gregory

28

[From Bill Powers (2004.11.02.1330 MST)]

Bruce Gregory (2004.1102.1440)]
Rick said: My guess is that Bill's alternative would be perceptual control
theory.

You said:

Then all perceptions are controlled and all learning is reorganization?

PCT does not say that all perceptions are controlled. It does not say that
all learning is reorganization. Read the book. God, you make it hard to be
civil.

Bill P.

29

[From Bruce Gregory (2004.1102.1547)]

Bill Powers (2004.11.02.1330 MST)

PCT does not say that all perceptions are controlled. It does not say
that
all learning is reorganization. Read the book. God, you make it hard
to be
civil.

Sorry to be such a trial. How does a child learn that the furry
creature is a dog? You obviously have model, why not share with us who
are too dense to understand B:CP?

Bruce Gregory

p.s. If this question is too hard, I'll understand.

30

[From Bruce Gregory (2004.1102.1740)]

Bill,

I fear I have been putting you in a difficult position. That was not my
intention. Please consider my questions withdrawn. If I want answers,
I'll look for them in B:CP.

Bruce Gregory

31

[From Rick Marken (2004.11.04.1020)]

I am still looking for data relevant to the question of whether there is a
reaction time (RT) difference when presented an instance and asked for the
category versus being presented the category and asked for the instance. My
cognitive psychologist friend in Minnesota (Dr. R. Kim Guenther, who wrote a
well-received textbook on cognitive psychology) pointed me to a study that
seemed, at first glance, to be perfect. It was done by Elizabeth Loftus in
1973 (JEP, 97, 70-74). According to Kim, "she compared RTs to cases where
the category was presented first (tree-oak) and the instance was presented
first (oak-tree)". But in this study both the category and the instance
were categories (according to PCT) inasmuch as they were both words. So
there is really no difference in RTs expected in the two cases, and
apparently there was no clear difference in reaction times for the two
conditions: the difference in RT "depended on how frequently the category
evoked the instance" according to Kim. But from the PCT perspective the
"evoked" instance is another category. So the study was not looking at
hierarchical relationships between perceptual variables. I am still looking
(with Kim's help) for a study like the Loftus study but where the instances
-- like oak, in the Loftus study -- are pictures instead of words. Does
anyone on the list happen to know of such a study?

Best

Rick

···

--
Richard S. Marken
MindReadings.com
Home: 310 474 0313
Cell: 310 729 1400

--------------------

This email message is for the sole use of the intended recipient(s) and
may contain privileged information. Any unauthorized review, use,
disclosure or distribution is prohibited. If you are not the intended
recipient, please contact the sender by reply email and destroy all copies
of the original message.

32

[From Bill Powers (2004.11.04.12:18 MST)]

Rick Marken (2004.11.04.1020)--

Musings: It would be good to find studies of category perception that have
already been done, but it would also be a good idea to do your own. There
are always questions about existing studies unless an extraordinary amount
of detail was reported.

Think about how the experiment would be done. Presenting a picture can be
done on a computer screen, but how is indicating the name done? If there is
an array of names to choose from, the selection being done by pointing with
the mouse, the search time and movement time would figure into the
"reaction time." On the other hand, if the person simply utters the name,
how does the computer time that utterance? From the first onset of sound
(using a microphone)?

And how do you do it the other way around to keep the situation
symmetrical? A person can't "utter" a picture. About all you could do would
be to have the person press a key when an image comes to mind, then check
later to see what the image was. So maybe the person should press a key in
either case to indicate that either a word or an image came to mind, with a
follow-up check to see what the image or word was that was evoked.

Best,

Bill P.

33

[Martin Taylor 2004.11.04.15.39]

[From Rick Marken (2004.11.04.1020)]

I am still looking for data relevant to the question of whether there is a
reaction time (RT) difference when presented an instance and asked for the
category versus being presented the category and asked for the instance.

I haven't dropped out of this discussion, but I've been swamped, and
I will be so until I go out of town for much of next wee, followed by
two meetings for which I have to do quite a bit of work, on Friday
and Saturday. After that, I may be able to resume. That means some
time around Nov 16.

However, I will make the point that I agree with Bill and Rick that
finding a label is often a purposive (control) phenomenon. Conscious
experience suggests that it isn't always, though. And whether
attaching a label is purposive or not, there's still the question of
modelling how a particular "appropriate" label is attached to a
picture, and vice-vers.

Conscious experience is a very fallible guide, so experiments are
need for two aspects of the problem:

(1) If there is a model for the evocation of picture by label and
vice-versa that does, and a model that does not involve control, can
an experiment be devised to determine which model best corresponds to
what people do in what seems to be a naturalistic situation?

(2) If it is accepted that all labelling of images and imaging or
labels is the result of a control process, is there a model or models
to account for the attachment of a particular label to a picture and
symmetrically of a particular picture to a label, and does that
(those) model(s) agree with the data for what people do?

I haven't actually seen a model (other than my own extended
flip-flop) that would be suited for testing, and I'm not at all sure
that even my model is in a state to give more than qualitative
results.

As for the experiments Rick is asking about, I said that I would ask
Lochlan Magee about his thesis work on picture-word Stroop (not the
word "red" written in green ink!), but I haven't seen him since then,
and I won't until at least Nov 16.

Martin

34

[From Rick Marken (2004.11.04.1400)]

Bill Powers (2004.11.04.12:18 MST)--

Rick Marken (2004.11.04.1020)--

Musings: It would be good to find studies of category perception that have
already been done, but it would also be a good idea to do your own. There
are always questions about existing studies unless an extraordinary amount
of detail was reported.

I agree. And I will certainly take your concerns about experimental design
into account if (or when) I do the study. And it looks like I will have to
do the study because Kim didn't know of any studies that are similar to what
I need. So there may be a new java demo up soon.

Best

Rick

···

--
Richard S. Marken
MindReadings.com
Home: 310 474 0313
Cell: 310 729 1400

--------------------

This email message is for the sole use of the intended recipient(s) and
may contain privileged information. Any unauthorized review, use,
disclosure or distribution is prohibited. If you are not the intended
recipient, please contact the sender by reply email and destroy all copies
of the original message.

35

[From Bill Powers (2004.11.04.1450 MST)]

Martin Taylor 2004.11.04.15.39 –

Conscious experience is a very
fallible guide, so experiments are

need(ed)

The problem with that idea is that we can’t find something to observe the
experimental results for us, or the outcomes of automatic analysis, so we
can’t trust the results of experiments, either. It’s all
somebody’s conscious experience.

Suppose I say “airplane” and you hit the space bar, then tell
me that you had an image of some airplane. You could hardly be mistaken
about that – that is, think you experienced an image when you didn’t. I
don’t think we need to worry about lying, do we?

I haven’t actually seen a model
(other than my own extended

flip-flop) that would be suited for testing, and I’m not at all
sure

that even my model is in a state to give more than qualitative

results.

I think we have to establish a phenomenon, first.

Best,

Bill P.

36

Re: Attaching labels

[From Bill Powers (2004.11.04.1450
MST)]

Martin Taylor 2004.11.04.15.39 –

Conscious experience is a very fallible
guide, so experiments are

need(ed)

The problem with that idea is that we can’t find something to observe
the experimental results for us, or the outcomes of automatic
analysis, so we can’t trust the results of experiments, either. It’s
all somebody’s conscious experience.
Suppose I say “airplane” and
you hit the space bar, then tell me that you had an image of some
airplane. You could hardly be mistaken about that – that is, think
you experienced an image when you didn’t. I don’t think we need to
worry about lying, do we?

I haven’t actually seen a model (other
than my own extended

flip-flop) that would be suited for testing, and I’m not at all
sure

that even my model is in a state to give more than qualitative

results.

I think we have to establish a phenomenon, first.

That has been the thrust of my earlier messages in this thread, but
somehow the topic keeps getting diverted from that.

One can’t establish whether an experiment is suitable before
establishing what the phenomenon is that raises the question the
experiment is to answer.

The initial description was a bit vague to permit a useful
experiment. It was that sometimes when one reads words or hears
them, relevant pictures flash, apparently unbidden, into the mind.
Also, and in the same person though at different times, there are
occasions when one sees something (a picture, or maybe a person), and
a word that we have been calling a “label” comes to mind
unbidden.

Now, as you have so often pointed out, there’s no point in doing
an experiment to see whether that happens. We want to know what is
happening in the head or in a fully fleshed out model when that
phenomenon is experienced.

Rick, and to some extent you, have brought up other phenomena
that are associated (non-technical use of the word:-) with labelling
or pictorial recall. Rick has asked whether the perception of a
potential label is faster than the perception of a picture for which
it might be a label, given that each is independently presented on
different occasions to a subject. Rick and you have both enquired
about the phenomenon that one often purposefully seeks a label for an
observed picture or object, and may also purposefully try to bring to
mind a picture that would go with a word. That’s clearly related to
the phenomenon that initiated this thread, and is worth following up
– IF THERE IS AT LEAST ONE PROPOSED MODEL, as is also the case for
experimenting with the initial phenomenon.

Suppose I say “airplane” and
you hit the space bar, then tell me that you had an image of some
airplane.

That’s a proposal for an experimental technique. I believe I
suggested a similar one in one of my messages. But what would the
results tell us that would help us to discriminate among possible
mechanisms?

Martin

37

[From Bill Powers (2004.11.04.1835 MST)]

I think we have to establish a phenomenon, first.

That _has_ been the thrust of my earlier messages in this thread, but
somehow the topic keeps getting diverted from that.

One can't establish whether an experiment is suitable before establishing
what the phenomenon is that raises the question the experiment is to answer.

I don't think it's that complicated. If we can't show that there is any
difference in reaction time when the image or the word occurs first, any
theory about why such a difference exists becomes irrelevant.

If a difference is found, we can start thinking of ways to test
explanations. If it's not found, we can start thinking of alternative
explanations.

Best,

Bill P.

···

The initial description was a bit vague to permit a useful experiment. It
was that sometimes when one reads words or hears them, relevant pictures
flash, apparently unbidden, into the mind. Also, and in the same person
though at different times, there are occasions when one sees something (a
picture, or maybe a person), and a word that we have been calling a
"label" comes to mind unbidden.

Now, as you have so often pointed out, there's no point in doing an
experiment to see whether that happens. We want to know what is happening
in the head or in a fully fleshed out model when that phenomenon is
experienced.

Rick, and to some extent you, have brought up other phenomena that are
associated (non-technical use of the word:-) with labelling or pictorial
recall. Rick has asked whether the perception of a potential label is
faster than the perception of a picture for which it might be a label,
given that each is independently presented on different occasions to a
subject. Rick and you have both enquired about the phenomenon that one
often purposefully seeks a label for an observed picture or object, and
may also purposefully try to bring to mind a picture that would go with a
word. That's clearly related to the phenomenon that initiated this thread,
and is worth following up -- IF THERE IS AT LEAST ONE PROPOSED MODEL, as
is also the case for experimenting with the initial phenomenon.

Suppose I say "airplane" and you hit the space bar, then tell me that you
had an image of some airplane.

That's a proposal for an experimental technique. I believe I suggested a
similar one in one of my messages. But what would the results tell us that
would help us to discriminate among possible mechanisms?

Martin

38

[Martin Taylor 2004.11.05.07.11]

[From Bill Powers (2004.11.04.1835 MST)]

I think we have to establish a phenomenon, first.

That _has_ been the thrust of my earlier messages in this thread, but
somehow the topic keeps getting diverted from that.

One can't establish whether an experiment is suitable before establishing
what the phenomenon is that raises the question the experiment is to answer.

I don't think it's that complicated. If we can't show that there is any
difference in reaction time when the image or the word occurs first, any
theory about why such a difference exists becomes irrelevant.

If there is a difference, or if there isn't, I still haven't
understood what kind of model is being proposed to account for the
bidirectional linking of label and picture. That's te phenomenon to
be explained. If the proposed model suggests that there ought to be a
difference in the timings, then the existence or non-existence of an
experimental timing difference becomes of interest. If it doesn't
then some other kind of experiment would be needed for a test.

If a difference is found, we can start thinking of ways to test
explanations. If it's not found, we can start thinking of alternative
explanations.

I'd rather think of what models could account for the phenomenon, and
then ask what the implications would be that could test the
plausibility of the model.

I think that all Bruce G was asking for in the first place was a
description of a plausible model compatible with a strict HPCT
hierarchy. I didn't follow the outline you provided sufficiently
clearly to be able to tell whether his question had been answered.
You asserted that the bidirectionality was the result of a control
process that in some way invoked the imagination loop in the
appropriate lower-level control element, but it was never clear to me
either how the HPCT structure would act to have this effect.

I'm not saying it wouldn't, but at the moment I'm missing the kind of
understanding that would permit me to make a mental (or a computer
simulation) model. To date, the only kind of model that I can make
work in a mental simulation involves cross linkages among perceptual
functions within a level. Bruce's (and my) question asks whether it
is possible without those linkages. Given a model that seems to
perform the required task, we can then ask whether it implies timing
differences in some experiment.

Martin

39

[From Rick Marken (2004.11.05.0930)]

Martin Taylor 2004.11.05.07.11]

I'd rather think of what models could account for the phenomenon, and
then ask what the implications would be that could test the
plausibility of the model.

I think that this is a good way to go when you are tuning up a model that
seems to work pretty well (in terms of explaining existing observations), as
in the case of tracking and baseball catching. But you're interested in a
phenomenon for which no _precise_ model exists and which is rather poorly
defined at that: the phenomenon of the bidirectionality of association
between categories and instances. That is, the phenomenon is the subjective
awareness that words like "dog" can bring to mind an image of a particular
dog and that the image of a particular dog can bring to mind the word "dog".

I suggested a very simple experiment that could "operationalize" this
phenomenon: ask people to indicate, as quickly as possible, the appropriate
category (name) after being shown a picture of an instance of that category,
and vice versa. A "qualitative" model based on PCT leads to the prediction
that the time to identify the category given the instance will be less than
the time to identify the instance given the category.

I'm still looking to see whether this experiment has been done. I'm sure it
has but it's hard to find because it's probably nestled in some study that
has to do with "priming and processing" or some such.

Regards

Rick

···

--
Richard S. Marken
MindReadings.com
Home: 310 474 0313
Cell: 310 729 1400

--------------------

This email message is for the sole use of the intended recipient(s) and
may contain privileged information. Any unauthorized review, use,
disclosure or distribution is prohibited. If you are not the intended
recipient, please contact the sender by reply email and destroy all copies
of the original message.

40

[From Bjorn Simonsen (2004.11.09,20:45 EST)]
I know it is long time since

From Bruce Gregory (2004.1102.0627)

...but, ....

It seems to me that we know the names of objects in our
immediate surroundings without having a purpose for this knowledge. Of
course, I may be wrong in this regard.

I don't thin we need a purpose to know the names of things. But I am sure we
had a purpose when we learned the names we remember today.
Bjorn