[From Rick Marken (2013.09.01.1550)]
Since it’s a Holiday Weekend I am going to take a holiday from making a study guide for Ch. 9 (unless David is good enough to come up with one on his own). So read and comment on Ch. 9 by responding to this post.
Happy Labor Day
Best
Rick
[From Rupert Young (2013.09.06 21.00 BST)]
Chapter 9 extends the concept from the previous chapter of a hierarchy of control systems, with systems in a level dependant systems in other levels within the hierarchy. The configuration perceptions are a function of lower-level sensations. I don't have a lot to say but a couple of points I found particularly interesting:
* that the frequency of oscillation of instability tremors would indicate the level within the hierarchy of the problem,
* that the perceptual "representation" of an object, say, is a highly distributed collection of (sensation) perceptions that is not organised in any way similar to the experience of a 3-D object (an approach taken in computer vision)
Again my knowledge (ignorance rather) doesn't enable me to comment on the validity of the neural discussion. I would really like to see some research in this area. With the advances made in recent years on neural mapping it would seem much more possible to identify feedback loops. Surely we could tempt some neuroscientist with the promise of power and glory to be the first to investigate this new frontier!
Leading questions:
1. Tension of muscles, tightness of skin, keeping eyes open, position of head, eyes and hand.
2. Yes. Well, the muscular impression is different, but not the visual.
3. No. No, I can perceive edges etc, without needing to perceive an object.
4. Yes, I did that for dinner when I cooked Halloumi kebabs.
Regards,
Rupert
On 01/09/2013 23:49, Richard Marken wrote:
[From Rick Marken (2013.09.01.1550)]
Since it's a Holiday Weekend I am going to take a holiday from making a study guide for Ch. 9 (unless David is good enough to come up with one on his own). So read and comment on Ch. 9 by responding to this post.
Happy Labor Day
Best
Rick
--
Richard S. Marken PhD
<mailto:rsmarken@gmail.com>> rsmarken@gmail.com
<http://www.mindreadings.com>> www.mindreadings.com
[From Rick Marken (2013.09.08.1200)]
Rupert Young (2013.09.06 21.00 BST)--
Thanks for this Rupert. These are excellent observations. My own
summary of Ch. 9 will be coming up soon.
Best
Rick
Chapter 9 extends the concept from the previous chapter of a hierarchy of
control systems, with systems in a level dependant systems in other levels
within the hierarchy. The configuration perceptions are a function of
lower-level sensations. I don't have a lot to say but a couple of points I
found particularly interesting:that the frequency of oscillation of instability tremors would indicate the
level within the hierarchy of the problem,
that the perceptual "representation" of an object, say, is a highly
distributed collection of (sensation) perceptions that is not organised in
any way similar to the experience of a 3-D object (an approach taken in
computer vision)Again my knowledge (ignorance rather) doesn't enable me to comment on the
validity of the neural discussion. I would really like to see some research
in this area. With the advances made in recent years on neural mapping it
would seem much more possible to identify feedback loops. Surely we could
tempt some neuroscientist with the promise of power and glory to be the
first to investigate this new frontier!Leading questions:
Tension of muscles, tightness of skin, keeping eyes open, position of head,
eyes and hand.
Yes. Well, the muscular impression is different, but not the visual.
No. No, I can perceive edges etc, without needing to perceive an object.
Yes, I did that for dinner when I cooked Halloumi kebabs.Regards,
RupertOn 01/09/2013 23:49, Richard Marken wrote:
[From Rick Marken (2013.09.01.1550)]
Since it's a Holiday Weekend I am going to take a holiday from making a
study guide for Ch. 9 (unless David is good enough to come up with one on
his own). So read and comment on Ch. 9 by responding to this post.Happy Labor Day
Best
Rick
--
Richard S. Marken PhD
rsmarken@gmail.com
www.mindreadings.com
--
Richard S. Marken PhD
rsmarken@gmail.com
www.mindreadings.com
[John Kirkland 20130918]
A few comments and observations
P 122, here’s the neat idea of forms as visual configurations arising from stimulating the visual cortex. Presumably the visual cortex can be stimulated by other pathways too (Bill’s ‘Many different structures can effect the same function’ p 118 para 4). There are anecdotal reports about visual forms (nameless shapes) arising during hallucinations and erupting from exertion (sexual orgasm, as film makers depict with exploding skyrockets) and migraines. This leads me to ponder upon the familiar expression, ‘Go figure’. Do we do this figure-making intentionally, or does it just happen?
P. 122, the definition of configuration builds nicely from sensation vectors. Now, how about considering reverse engineering? For instance let’s assume there are stable, consistent, shared invariants; the basic evidence is pretty robust, for instance we tend not to bump into things except in the dark or other situations when there is sensory deficit/impairment. Of course, as philosophers have opined, there are then family resemblances arising from putting like with like. For instance if people were offered the same set of photographs depicting different ‘things’ they’d be able to place them into family groups, even without having to describe how they did so. Indeed any words may get in the way. By applying sophisticated analyses to these sorting data it would be possible to discover implicit invariants.
Crikey, I’ve been doing this for about 30 years. This is a two-part story. Here’s the part one, the practical. Give me a set of items (where N > 12) and a bunch of people who can make ‘reasonable’ (as opposed to arbitrary) decisions as to what goes with what in that finite set (it’s a bit more complicated as we’ve a routine here as to how these families are constructed) and I’ll provide a rendered aggregated rendered 3D ‘map’ with orthogonal axes. Yes, OK, it took us almost a decade to fathom out and create the maths involved from data gathered from dozens of simulated and other studies, many of these using others’ data (special thanks to those who chipped in freely). BTW, we’ve plugged in many types of items (sounds, colours, morphed faces, shapes, beers, wines, personality adjectives, cigarettes, cardiac oscultation, personality descriptors, interests, and so on and so forth) and the methodological suite hasn’t come up short yet: stable, robust maps emerge every time. Here’s one odd finding amongst many. Drawing from a stable 3D map generated from a corpus of single word personality descriptors I carried out yet another sorting study using a subset of this adjectival set but omitted almost an entire quadrant. Yet the intermap stress between the partial and the full one was respectably low. This suggests even if polar exemplars are not provided participants behaved as if the poles were still there. In short, it appears invariant axes are present even if corresponding items are not. I’m sure you can fill in the gaps.
That’s the practical part. And we are quite good at this too; we have dozens of corresponding research examples and articles. We even call ourselves psychological cartographers. Recall Kurt Lewin’s cute quip: there’s nothing so practical as a good theory? Well, all these years during which we’ve been piling up these sorts of practical results I’ve been looking, nay searching, for a theory. And, here is is. Thanks for part two Bill. It’s an epoxy.
Perhaps a sabbatical with a bona-fide PCTer might be beneficial.
If anybody would like to chip in and at the same time see this approach in action I’ve an item set on line that’s ready for sorting. It consists of single words describing sounds that engineers and others have difficulty communicating about, especially those working in high-risk settings like aviation. I want to try and uncover the dimensions (aka invariants) for distinguishing a knock from a dong from a tap and so on. It’ll take about 40 minutes. If anybody’s got some willing engineers, all the better. As with almost all of the research I unertake, it could be fun.
All right Rick, you asked what we’ve got out of B:CP so far. For me, a revelation. What makes this much worse is the delay since the book’s been around for as long as we’ve been doing this type of research. Ouch. And there’s a way to go yet.
With kind regards
JohnK
On Mon, Sep 9, 2013 at 6:59 AM, Richard Marken rsmarken@gmail.com wrote:
[From Rick Marken (2013.09.08.1200)]
Rupert Young (2013.09.06 21.00 BST)–
Thanks for this Rupert. These are excellent observations. My own
summary of Ch. 9 will be coming up soon.
Best
Rick
Chapter 9 extends the concept from the previous chapter of a hierarchy of
control systems, with systems in a level dependant systems in other levels
within the hierarchy. The configuration perceptions are a function of
lower-level sensations. I don’t have a lot to say but a couple of points I
found particularly interesting:
that the frequency of oscillation of instability tremors would indicate the
level within the hierarchy of the problem,
that the perceptual “representation” of an object, say, is a highly
distributed collection of (sensation) perceptions that is not organised in
any way similar to the experience of a 3-D object (an approach taken in
computer vision)
Again my knowledge (ignorance rather) doesn’t enable me to comment on the
validity of the neural discussion. I would really like to see some research
in this area. With the advances made in recent years on neural mapping it
would seem much more possible to identify feedback loops. Surely we could
tempt some neuroscientist with the promise of power and glory to be the
first to investigate this new frontier!
Leading questions:
Tension of muscles, tightness of skin, keeping eyes open, position of head,
eyes and hand.
Yes. Well, the muscular impression is different, but not the visual.
No. No, I can perceive edges etc, without needing to perceive an object.
Yes, I did that for dinner when I cooked Halloumi kebabs.
Regards,
Rupert
On 01/09/2013 23:49, Richard Marken wrote:
[From Rick Marken (2013.09.01.1550)]
Since it’s a Holiday Weekend I am going to take a holiday from making a
study guide for Ch. 9 (unless David is good enough to come up with one on
his own). So read and comment on Ch. 9 by responding to this post.
Happy Labor Day
Best
Rick
–
Richard S. Marken PhD
–
Richard S. Marken PhD
[From Erling Jorgensen (2013.09.19 1200 EDT)
John Kirkland 20130918
Hello John,
The work you've been doing sounds very interesting.
P. 122, the definition of configuration builds nicely from sensation
vectors. Now, how about considering reverse engineering? For instance
let's assume there are stable, consistent, shared invariants;...
For instance if people were offered
the same set of photographs depicting different 'things' they'd be
able to place them into family groups, even without having to describe
how they did so. Indeed any words may get in the way. By applying
sophisticated analyses to these sorting data it would be possible to
discover implicit invariants.
The first question I have is about what level of perception is being
constructed here. Is it Configurations or Classifications (i.e.,
Categories)? On the face of it, it looks like classification. But
then you go on to discuss the orthogonal axes that emerge, so that
suggests an underlying perceptual orderliness to those categories.
Give me a set of items (where N > 12)
and a bunch of people who can make 'reasonable' (as opposed to
arbitrary) decisions as to what goes with what in that finite set
(it's a bit more complicated as we've a routine here as to how these
families are constructed) and I'll provide a rendered aggregated
rendered 3D 'map' with orthogonal axes. ...
BTW, we've plugged in many types of items (sounds,
colours, morphed faces, shapes, beers, wines, personality adjectives,
cigarettes, cardiac oscultation, personality descriptors, interests,
and so on and so forth) and the methodological suite hasn't come up
short yet: stable, robust maps emerge every time.
My second question is whether the "rendered 3D 'map' with orthogonal
axes" means that you regularly get three orthogonal dimensions when
people go through such tasks? Should this be considered three
underlying factors, or three sets of polar opposites among six
overall factors?
The following question presumes a 'yes' answer that the matter of
three axes (regardless of task) is a stable phenomenon...
Does this imply a cognitive limit as to how such perceptions (whatever
we call them) are constructed? I'm reminded of Miller's classic
article, "The Magical Number Seven, Plus or Minus Two," about the
limits of working memory.
I believe every level of perception involves the construction of
implicit invariants, in a sense creating a new world of perceptions
out of (or on top of) the lower level(s). I like your idea of reverse
engineering, creating maps of the implicit perceptions involved. I
suspect that such models will be the only way to get past the level
of Sensations with their weighted sums, since the actual
neurophysiological wiring of higher levels seems incredibly complex.
Can you say more about those orthogonal axes? Are you able to assign
approximate labels to them after-the-fact?
All the best,
Erling
[From Rick Marken (2013.09.19.1400)]
John Kirkland (20130918)--
A few comments and observations
P 122, here's the neat idea of forms as visual configurations arising from
stimulating the visual cortex. Presumably the visual cortex can be
stimulated by other pathways too (Bill's 'Many different structures can
effect the same function' p 118 para 4). There are anecdotal reports about
visual forms (nameless shapes) arising during hallucinations and erupting
from exertion (sexual orgasm, as film makers depict with exploding
skyrockets) and migraines. This leads me to ponder upon the familiar
expression, 'Go figure'. Do we do this figure-making intentionally, or does
it just happen?
RM: I think the "figure making" discussed in this chapter is something
that "just happens" when certain lower level perceptual signals are
input to a configuration perceptual function.
P. 122, the definition of configuration builds nicely from sensation
vectors. Now, how about considering reverse engineering?
RM: That's what we are doing already, isn't it? We have a product (a
living organism) that has been engineered by evolution. We know what
it has been built to do - control -- we're just trying to figure out
how it was built to do it. B:CP can be seen, I believe, as a first
step towards reverse engineering the controlling done by living
systems.
For instance let's
assume there are stable, consistent, shared invariants; the basic evidence
is pretty robust, for instance we tend not to bump into things except in the
dark or other situations when there is sensory deficit/impairment.
RM:I don't see how that is evidence for perceptual invariants?
Of course, as philosophers have opined, there are then family resemblances
arising from putting like with like. For instance if people were offered the
same set of photographs depicting different 'things' they'd be able to place
them into family groups, even without having to describe how they did so.
Indeed any words may get in the way. By applying sophisticated analyses to
these sorting data it would be possible to discover implicit invariants.
RM: I think we must mean different things by "invariants". I think of
an invariant as a perception that remains the same despite changes in
its' components. So a square configuration is an invariant if a
square is seen despite variations in, among other things, the
thickness, color and orientation of the lines that make it up.
Crikey, I've been doing this for about 30 years. This is a two-part story.
Here's the part one, the practical. Give me a set of items (where N > 12)
and a bunch of people who can make 'reasonable' (as opposed to arbitrary)
decisions as to what goes with what in that finite set (it's a bit more
complicated as we've a routine here as to how these families are
constructed) and I'll provide a rendered aggregated rendered 3D 'map' with
orthogonal axes. Yes, OK, it took us almost a decade to fathom out and
create the maths involved from data gathered from dozens of simulated and
other studies, many of these using others' data (special thanks to those who
chipped in freely). BTW, we've plugged in many types of items (sounds,
colours, morphed faces, shapes, beers, wines, personality adjectives,
cigarettes, cardiac oscultation, personality descriptors, interests, and so
on and so forth) and the methodological suite hasn't come up short yet:
stable, robust maps emerge every time. Here's one odd finding amongst many.
Drawing from a stable 3D map generated from a corpus of single word
personality descriptors I carried out yet another sorting study using a
subset of this adjectival set but omitted almost an entire quadrant. Yet the
intermap stress between the partial and the full one was respectably low.
This suggests even if polar exemplars are not provided participants behaved
as if the poles were still there. In short, it appears invariant axes are
present even if corresponding items are not. I'm sure you can fill in the
gaps.
RM: Multidimensional scaling can be lots of fun; I particular enjoy
Roger Shepard's work. I think it can tell us something about how
people perceive the worlds at the category level, which involves
seening similarities between various entities. But I don't think it
tells us about invariants so much as what might serve as convenient
dimensions for a spatial representation of the relationship between
perceptions.
That's the practical part. And we are quite good at this too; we have
dozens of corresponding research examples and articles. We even call
ourselves psychological cartographers. Recall Kurt Lewin's cute quip:
there's nothing so practical as a good theory? Well, all these years during
which we've been piling up these sorts of practical results I've been
looking, nay searching, for a theory. And, here is is. Thanks for part two
Bill. It's an epoxy.
Perhaps a sabbatical with a bona-fide PCTer might be beneficial.
RM: I'm not clear about the relationship between your mapping work and
PCT. Maybe the sabbatical is a good idea. Or, better, how about an all
expense paid vacation for my wife and I in New Zealand and I'll talk
with you for at least 1 hour each day we're there!
If anybody would like to chip in and at the same time see this approach in
action I've an item set on line that's ready for sorting. It consists of
single words describing sounds that engineers and others have difficulty
communicating about, especially those working in high-risk settings like
aviation. I want to try and uncover the dimensions (aka invariants) for
distinguishing a knock from a dong from a tap and so on. It'll take about 40
minutes. If anybody's got some willing engineers, all the better. As with
almost all of the research I unertake, it could be fun.All right Rick, you asked what we've got out of B:CP so far. For me, a
revelation. What makes this much worse is the delay since the book's been
around for as long as we've been doing this type of research. Ouch. And
there's a way to go yet.
RM: I'm not completely clear about what that revelation is but after a
couple of weeks in NZ I should have it nailed down (absolutely no
bungee jumping for me though)!
Best
Rick
With kind regards
JohnK
On Mon, Sep 9, 2013 at 6:59 AM, Richard Marken <rsmarken@gmail.com> wrote:
[From Rick Marken (2013.09.08.1200)]
>Rupert Young (2013.09.06 21.00 BST)--
Thanks for this Rupert. These are excellent observations. My own
summary of Ch. 9 will be coming up soon.Best
Rick
> Chapter 9 extends the concept from the previous chapter of a hierarchy
> of
> control systems, with systems in a level dependant systems in other
> levels
> within the hierarchy. The configuration perceptions are a function of
> lower-level sensations. I don't have a lot to say but a couple of points
> I
> found particularly interesting:
>
> that the frequency of oscillation of instability tremors would indicate
> the
> level within the hierarchy of the problem,
> that the perceptual "representation" of an object, say, is a highly
> distributed collection of (sensation) perceptions that is not organised
> in
> any way similar to the experience of a 3-D object (an approach taken in
> computer vision)
>
> Again my knowledge (ignorance rather) doesn't enable me to comment on
> the
> validity of the neural discussion. I would really like to see some
> research
> in this area. With the advances made in recent years on neural mapping
> it
> would seem much more possible to identify feedback loops. Surely we
> could
> tempt some neuroscientist with the promise of power and glory to be the
> first to investigate this new frontier!
>
> Leading questions:
>
> Tension of muscles, tightness of skin, keeping eyes open, position of
> head,
> eyes and hand.
> Yes. Well, the muscular impression is different, but not the visual.
> No. No, I can perceive edges etc, without needing to perceive an object.
> Yes, I did that for dinner when I cooked Halloumi kebabs.
>
>
> Regards,
> Rupert
>
> On 01/09/2013 23:49, Richard Marken wrote:
>
> [From Rick Marken (2013.09.01.1550)]
>
> Since it's a Holiday Weekend I am going to take a holiday from making a
> study guide for Ch. 9 (unless David is good enough to come up with one
> on
> his own). So read and comment on Ch. 9 by responding to this post.
>
> Happy Labor Day
>
> Best
>
> Rick
> --
> Richard S. Marken PhD
> rsmarken@gmail.com
> www.mindreadings.com
>
>--
Richard S. Marken PhD
rsmarken@gmail.com
www.mindreadings.com
--
Richard S. Marken PhD
rsmarken@gmail.com
www.mindreadings.com
[John Kirkland 20130921]
Equinox…
Thanks for your helpful comments Rick. Let me mull these over. Yes, if I win Lotto we’ll have a get-together here in kiwiland.
Here’s what I’m pondering over.
When we aggregate sorting data there is a consistent MDS map obtained, usually rendered into 3D. As an side, I can provide one to anybody who’s interested as we designed a software suite for view these; they can be rotated with dimming effects (like viewing 3D through mist or fog where only the closest ones are seen. It’s fun trying to reconstruct the full map from these partial sightings. Years ago we used to say we were like surveyors taking line-of-sight readings and putting them together), and even having the option to turn on/off stereo views as well as red/cyan off-sets (about 3 degrees, the radial distance between eyes) and using those old 3D glasses. As an aside, when a person viewed the screen through these red/cyan lenses then always smiled when ‘seeing’ it).
Here’s a factoid. These maps tend to be highly consistent. For instance I was once interested in how interpersonal conflict was assessed. I tracked down a thesis which had compared results from two different pencil and paper instruments. Adopting our usual approach for such studies I assembled all the items, edited them lightly so they had the same signature and then used these for a sorting study. In our approach there are four phases (grouping, opposites, partitioning, adding) where each person assembles an individual dendrogram. These data are thrown into the big pot and analysed using our variation of MDS. A stable map can be assembled from about 30 data sets. Yes, as few as that. Of course we can then check the inter-item structure against other procedures, like factor analysis. Here’s factoid #1: when we remove or add items the MDS solution remains the same. I’m sure you know what happens when this is attempted with FA. And factoid #2: 30-odd seasoned police here in NZ completed the GOPA sorting task, as did 30-odd undergraduate psych students in the US (a mate of mine knows the ropes for collecting these data). We could say there was little similarity between these groups. Wait for it; their maps were over 95% agreement. Factoid #3: when we have had item sets translated (German, Japanese, and one of those Indonesian languages) the inter-map similarities are well in excess of 80%.
I’m seeking a means of accounting for these parallel outcomes. It’s hard because much of our (western) education is nurtured in the belief we all have different ‘constructions’ of the world. Well, at least that was the case until the Russians arrived and unpacked their bags to reveal social constructionism. I suspect there’s a convergence here between traditional hard and soft approaches; hard being phenomena such as squares and chairs (Ames’s room) and soft as language’s grammar (Chomsky). As a rapprochement I wonder if the term ‘invariant’ can serve as a useful term for both, and possibly everything in between as well. Broadly speaking then, what we are presented with (from biology and society) are some invariant dimensions within which we each negotiate our existence. My ‘revelation’ is that PCT may provide the mechanism accounting for these negotiations; we are nudged and eased into map zones of least resistance (lowest error). So, woe and betide anybody who tries to relocate us within this dimensional/invariant frame.
Let’s take as an example a sub-lexicon, a semi-finite word set such as personality descriptions, or emotion words, or anything from a concourse. As an illustration we once took data from every available published study reporting on the CCQ (Jack Block’s California Child Q-set) and focussed these data through the lens of one of our maps. Unbelievably perhaps, but we were able to demonstrate not only how this map’s three axes did a better job of accounting for variance than Jack’s ones but we can also show as vectors each datum (individual or collective q-sorts). Thus the vectors (analogously, factors) are related to a higher order abstraction, the map’s dimensions.
If I had the brains I could probably resolve this an other anomalies that appear to fly in the face of received wisdom.
Am I getting any closer to a resolution with contributions from PCT? Yes, as of now I think so. But, clearly, I need others to help out here. Years ago I opened a box whose contents have always been at the limits of my comprehension and understanding. I can sense where the abyss is too, just over there in the void of unknown at a higher level just beyond where I am now.
With kind regards
JohnK
On Fri, Sep 20, 2013 at 9:22 AM, Richard Marken rsmarken@gmail.com wrote:
[From Rick Marken (2013.09.19.1400)]
John Kirkland (20130918)–
A few comments and observations
P 122, here’s the neat idea of forms as visual configurations arising from
stimulating the visual cortex. Presumably the visual cortex can be
stimulated by other pathways too (Bill’s 'Many different structures can
effect the same function’ p 118 para 4). There are anecdotal reports about
visual forms (nameless shapes) arising during hallucinations and erupting
from exertion (sexual orgasm, as film makers depict with exploding
skyrockets) and migraines. This leads me to ponder upon the familiar
expression, ‘Go figure’. Do we do this figure-making intentionally, or does
it just happen?
RM: I think the “figure making” discussed in this chapter is something
that “just happens” when certain lower level perceptual signals are
input to a configuration perceptual function.
P. 122, the definition of configuration builds nicely from sensation
vectors. Now, how about considering reverse engineering?
RM: That’s what we are doing already, isn’t it? We have a product (a
living organism) that has been engineered by evolution. We know what
it has been built to do - control – we’re just trying to figure out
how it was built to do it. B:CP can be seen, I believe, as a first
step towards reverse engineering the controlling done by living
systems.
For instance let’s
assume there are stable, consistent, shared invariants; the basic evidence
is pretty robust, for instance we tend not to bump into things except in the
dark or other situations when there is sensory deficit/impairment.
RM:I don’t see how that is evidence for perceptual invariants?
Of course, as philosophers have opined, there are then family resemblances
arising from putting like with like. For instance if people were offered the
same set of photographs depicting different ‘things’ they’d be able to place
them into family groups, even without having to describe how they did so.
Indeed any words may get in the way. By applying sophisticated analyses to
these sorting data it would be possible to discover implicit invariants.
RM: I think we must mean different things by “invariants”. I think of
an invariant as a perception that remains the same despite changes in
its’ components. So a square configuration is an invariant if a
square is seen despite variations in, among other things, the
thickness, color and orientation of the lines that make it up.
Crikey, I’ve been doing this for about 30 years. This is a two-part story.
Here’s the part one, the practical. Give me a set of items (where N > 12)
and a bunch of people who can make ‘reasonable’ (as opposed to arbitrary)
decisions as to what goes with what in that finite set (it’s a bit more
complicated as we’ve a routine here as to how these families are
constructed) and I’ll provide a rendered aggregated rendered 3D ‘map’ with
orthogonal axes. Yes, OK, it took us almost a decade to fathom out and
create the maths involved from data gathered from dozens of simulated and
other studies, many of these using others’ data (special thanks to those who
chipped in freely). BTW, we’ve plugged in many types of items (sounds,
colours, morphed faces, shapes, beers, wines, personality adjectives,
cigarettes, cardiac oscultation, personality descriptors, interests, and so
on and so forth) and the methodological suite hasn’t come up short yet:
stable, robust maps emerge every time. Here’s one odd finding amongst many.
Drawing from a stable 3D map generated from a corpus of single word
personality descriptors I carried out yet another sorting study using a
subset of this adjectival set but omitted almost an entire quadrant. Yet the
intermap stress between the partial and the full one was respectably low.
This suggests even if polar exemplars are not provided participants behaved
as if the poles were still there. In short, it appears invariant axes are
present even if corresponding items are not. I’m sure you can fill in the
gaps.
RM: Multidimensional scaling can be lots of fun; I particular enjoy
Roger Shepard’s work. I think it can tell us something about how
people perceive the worlds at the category level, which involves
seening similarities between various entities. But I don’t think it
tells us about invariants so much as what might serve as convenient
dimensions for a spatial representation of the relationship between
perceptions.
That’s the practical part. And we are quite good at this too; we have
dozens of corresponding research examples and articles. We even call
ourselves psychological cartographers. Recall Kurt Lewin’s cute quip:
there’s nothing so practical as a good theory? Well, all these years during
which we’ve been piling up these sorts of practical results I’ve been
looking, nay searching, for a theory. And, here is is. Thanks for part two
Bill. It’s an epoxy.
Perhaps a sabbatical with a bona-fide PCTer might be beneficial.
RM: I’m not clear about the relationship between your mapping work and
PCT. Maybe the sabbatical is a good idea. Or, better, how about an all
expense paid vacation for my wife and I in New Zealand and I’ll talk
with you for at least 1 hour each day we’re there!
If anybody would like to chip in and at the same time see this approach in
action I’ve an item set on line that’s ready for sorting. It consists of
single words describing sounds that engineers and others have difficulty
communicating about, especially those working in high-risk settings like
aviation. I want to try and uncover the dimensions (aka invariants) for
distinguishing a knock from a dong from a tap and so on. It’ll take about 40
minutes. If anybody’s got some willing engineers, all the better. As with
almost all of the research I unertake, it could be fun.
All right Rick, you asked what we’ve got out of B:CP so far. For me, a
revelation. What makes this much worse is the delay since the book’s been
around for as long as we’ve been doing this type of research. Ouch. And
there’s a way to go yet.
RM: I’m not completely clear about what that revelation is but after a
couple of weeks in NZ I should have it nailed down (absolutely no
bungee jumping for me though)!
Best
Rick
With kind regards
JohnK
On Mon, Sep 9, 2013 at 6:59 AM, Richard Marken rsmarken@gmail.com wrote:
[From Rick Marken (2013.09.08.1200)]
Rupert Young (2013.09.06 21.00 BST)–
Thanks for this Rupert. These are excellent observations. My own
summary of Ch. 9 will be coming up soon.
Best
Rick
Chapter 9 extends the concept from the previous chapter of a hierarchy
of
control systems, with systems in a level dependant systems in other
levels
within the hierarchy. The configuration perceptions are a function of
lower-level sensations. I don’t have a lot to say but a couple of points
I
found particularly interesting:
that the frequency of oscillation of instability tremors would indicate
the
level within the hierarchy of the problem,
that the perceptual “representation” of an object, say, is a highly
distributed collection of (sensation) perceptions that is not organised
in
any way similar to the experience of a 3-D object (an approach taken in
computer vision)
Again my knowledge (ignorance rather) doesn’t enable me to comment on
the
validity of the neural discussion. I would really like to see some
research
in this area. With the advances made in recent years on neural mapping
it
would seem much more possible to identify feedback loops. Surely we
could
tempt some neuroscientist with the promise of power and glory to be the
first to investigate this new frontier!
Leading questions:
Tension of muscles, tightness of skin, keeping eyes open, position of
head,
eyes and hand.
Yes. Well, the muscular impression is different, but not the visual.
No. No, I can perceive edges etc, without needing to perceive an object.
Yes, I did that for dinner when I cooked Halloumi kebabs.
Regards,
Rupert
On 01/09/2013 23:49, Richard Marken wrote:
[From Rick Marken (2013.09.01.1550)]
Since it’s a Holiday Weekend I am going to take a holiday from making a
study guide for Ch. 9 (unless David is good enough to come up with one
on
his own). So read and comment on Ch. 9 by responding to this post.
Happy Labor Day
Best
Rick
–
Richard S. Marken PhD
–
Richard S. Marken PhD
–
Richard S. Marken PhD
[John Kirkland 20130921]
My apologies Erling, I should have addressed your questions and queries sooner.
EJ:The first question I have is about what level of perception is being
constructed here. Is it Configurations or Classifications (i.e.,
Categories)? On the face of it, it looks like classification. But
then you go on to discuss the orthogonal axes that emerge, so that
suggests an underlying perceptual orderliness to those categories.
JK: I’m not at all sure, though my hunch is these are deconstructed classifications, perhaps nascent configurations. Others may be able to assist with this matter.
EJ: My second question is whether the "rendered 3D ‘map’ with orthogonal
axes" means that you regularly get three orthogonal dimensions when
people go through such tasks? Should this be considered three
underlying factors, or three sets of polar opposites among six
overall factors?
JK: There’s usually a sharp elbow after the third dimension. If we find there are four then there’s one of two options. One is to produce two 3D outputs. The other is to adjust the initial 3 with weights obtained from the 4th. With the hundreds of studies we’ve done using this approach there have been only about half a dozen needing to include the 4thD’s contribution.
We’ve tended to consider this as a dimensional approach, with bi-polar axes. When we use the map to interpret ranking like data (yes, even Likert scales though we prefer the Method of Successive Sorts MOSS, after Jack Block, since this allows participants to review all their item distributions and rearrange these as they go. We have adapted and refined what we call the trilemma approach whereby 3 uncorrelated items, uncorrelated with respect to the map, are presented as a mini-q sort and participants are invited to choose most or least applicable, or least and most applicable – for partially- or fully-ranked data) we summarise all responses as a vector, much like a knitting needle piercing the 3D sphere from the origin. These vectors may be interpreted using the three orthogonal axes. And, what’s more, if you want to shift a person’s perception then provide adjustments across the axes using neighbouring items as stepping stones. I’m of the view these adjustments could be interpreted pretty well as PCT; shake the system and see it resolve itself into a new niche. Whatever. We use the term habitat to describe all possible places of occupation within the 3D map. What a person negotiates, where there is minimum stress (aka low error) is their niche, what they have carved out. As Jack Block used to say, carve nature at its joints.
On a related matter, one can take FA outputs from conventional rank scaling and use these data to validate the map since the map-making operation is carried out independently of the self-report (ranked) data; different people are involved. But you cannot create a map from the FA outputs. For map-making one needs sorting data.
EJ:The following question presumes a ‘yes’ answer that the matter of
three axes (regardless of task) is a stable phenomenon…
JK: It was a ‘yes’. What’s presented is a hollow sphere, where items are spattered on the surface of a ball. Clearly, it makes no difference if the 3 orthogonal axes are rotated within this sphere, the inter-item relations do not change. We have designed some software to realign axes. One may do this to assimilate existing theory (by assimilate I mean to align with but not change the underlying structure.
It used to bother me no end that we kept getting the same 3D map. A recent masters student had English and Japanese GOPA the ‘same’ items and recovered almost the same map, at least sufficiently to pool these data. Way back when we were scared of these outcomes (how dare we get consistency?) we used to collect more GOPA data than necessary and then took random split halves (or quarters etc) to compare sub-maps. They were always well over 90% agreement. We kept thinking we were doing something wrong. Why? Because from my Missouri stats training such consistency is suspect. Now I simply shrug my shoulders. Whatever we are doing it’s consistent.
EJ: Does this imply a cognitive limit as to how such perceptions (whatever
we call them) are constructed? I’m reminded of Miller’s classic
article, “The Magical Number Seven, Plus or Minus Two,” about the
limits of working memory.
JK: It could be, though as I recall Miller was talking about bits.
Over the years I have found younger students (say 13) or those who are more challenged (you know what I mean) can sort single words or short phrases quite satisfactorily. By satisfactorily I mean consistent with the aggregate (yes, we have a statistic to indicate reliability). But the longer and more abstract the items the older and less challenged the participants need to be to achieve consistency/stability. Over 80 items people are struggling. With 100 they are bending at the knees. And time taken goes up exponentially too: 50 minutes for 70, 90mins for 90 and so on. But, as you’d imagine we can cope with up to 160 items beyond which the matrices are too large. We have a means for mapping larger item sets using a split-half procedure.
My sense of it is that we have a relatively limited number of axes and these get applied to the phenomena under scrutiny. Osgood did much better and we too often find his E P A (evaluation potency activity) as the axes.
EJ: I believe every level of perception involves the construction of
implicit invariants, in a sense creating a new world of perceptions
out of (or on top of) the lower level(s). I like your idea of reverse
engineering, creating maps of the implicit perceptions involved. I
suspect that such models will be the only way to get past the level
of Sensations with their weighted sums, since the actual
neurophysiological wiring of higher levels seems incredibly complex.
JK: We are on the same line. If this was choral work we’d be harmonising. Man, this is music to my ears.
I am of the view each of Bill’s levels offers plenty of scope for scientific exploration; some may wish to work at the neuropsychological, others may delve into the sensory arena, others probe at the higher end (eg MoL). There’s plenty of opportunity. It’s what’s held in common which provides a unifying theme: closed loop, comparators, error, negative feedback and so on. In short B:CP.
Me, I just like tinkering.
EJ: Can you say more about those orthogonal axes? Are you able to assign
approximate labels to them after-the-fact?
JK: Yes, I can say lots and lots. By after-the-fact do you mean after we’ve analysed GOPA data? Yes, very much so. Participants have no idea about the underlying architecture they deploy for grouping items. Some want to label these groups (which is very time-consuming) but this is unnecessary because these groupings are low-level.
Let’s go back to a 3D map and its axes. We have designed software to mark groups on the map. And these can be labelled. After all MDS software can generate a 3D map and a 2D dendrogram. As one might expect the 2D aggregated dendrogram’s clusters (depending on where the cut-off is) relate to map zones. Whilst the dendrogram’s output is associated with eye-ball clusters (from the map itself) the latter are more parsimonious (since they are in 3D). But, the dendrogram cannot be used to create a 3D map and its axes. The map and the dendrogam talk to one another via a translator.
Erling, thanks for the compliment of ‘getting it’. It’s been a somewhat lonely road and fellow travellers are always welcomed.
With kind regards
JohnK
On Sat, Sep 21, 2013 at 8:14 AM, John Kirkland johnkirkland@gmail.com wrote:
[John Kirkland 20130921]
Equinox…
Thanks for your helpful comments Rick. Let me mull these over. Yes, if I win Lotto we’ll have a get-together here in kiwiland.
Here’s what I’m pondering over.
When we aggregate sorting data there is a consistent MDS map obtained, usually rendered into 3D. As an side, I can provide one to anybody who’s interested as we designed a software suite for view these; they can be rotated with dimming effects (like viewing 3D through mist or fog where only the closest ones are seen. It’s fun trying to reconstruct the full map from these partial sightings. Years ago we used to say we were like surveyors taking line-of-sight readings and putting them together), and even having the option to turn on/off stereo views as well as red/cyan off-sets (about 3 degrees, the radial distance between eyes) and using those old 3D glasses. As an aside, when a person viewed the screen through these red/cyan lenses then always smiled when ‘seeing’ it).
Here’s a factoid. These maps tend to be highly consistent. For instance I was once interested in how interpersonal conflict was assessed. I tracked down a thesis which had compared results from two different pencil and paper instruments. Adopting our usual approach for such studies I assembled all the items, edited them lightly so they had the same signature and then used these for a sorting study. In our approach there are four phases (grouping, opposites, partitioning, adding) where each person assembles an individual dendrogram. These data are thrown into the big pot and analysed using our variation of MDS. A stable map can be assembled from about 30 data sets. Yes, as few as that. Of course we can then check the inter-item structure against other procedures, like factor analysis. Here’s factoid #1: when we remove or add items the MDS solution remains the same. I’m sure you know what happens when this is attempted with FA. And factoid #2: 30-odd seasoned police here in NZ completed the GOPA sorting task, as did 30-odd undergraduate psych students in the US (a mate of mine knows the ropes for collecting these data). We could say there was little similarity between these groups. Wait for it; their maps were over 95% agreement. Factoid #3: when we have had item sets translated (German, Japanese, and one of those Indonesian languages) the inter-map similarities are well in excess of 80%.
I’m seeking a means of accounting for these parallel outcomes. It’s hard because much of our (western) education is nurtured in the belief we all have different ‘constructions’ of the world. Well, at least that was the case until the Russians arrived and unpacked their bags to reveal social constructionism. I suspect there’s a convergence here between traditional hard and soft approaches; hard being phenomena such as squares and chairs (Ames’s room) and soft as language’s grammar (Chomsky). As a rapprochement I wonder if the term ‘invariant’ can serve as a useful term for both, and possibly everything in between as well. Broadly speaking then, what we are presented with (from biology and society) are some invariant dimensions within which we each negotiate our existence. My ‘revelation’ is that PCT may provide the mechanism accounting for these negotiations; we are nudged and eased into map zones of least resistance (lowest error). So, woe and betide anybody who tries to relocate us within this dimensional/invariant frame.
Let’s take as an example a sub-lexicon, a semi-finite word set such as personality descriptions, or emotion words, or anything from a concourse. As an illustration we once took data from every available published study reporting on the CCQ (Jack Block’s California Child Q-set) and focussed these data through the lens of one of our maps. Unbelievably perhaps, but we were able to demonstrate not only how this map’s three axes did a better job of accounting for variance than Jack’s ones but we can also show as vectors each datum (individual or collective q-sorts). Thus the vectors (analogously, factors) are related to a higher order abstraction, the map’s dimensions.
If I had the brains I could probably resolve this an other anomalies that appear to fly in the face of received wisdom.
Am I getting any closer to a resolution with contributions from PCT? Yes, as of now I think so. But, clearly, I need others to help out here. Years ago I opened a box whose contents have always been at the limits of my comprehension and understanding. I can sense where the abyss is too, just over there in the void of unknown at a higher level just beyond where I am now.
With kind regards
JohnK
On Fri, Sep 20, 2013 at 9:22 AM, Richard Marken rsmarken@gmail.com wrote:
[From Rick Marken (2013.09.19.1400)]
John Kirkland (20130918)–
A few comments and observations
P 122, here’s the neat idea of forms as visual configurations arising from
stimulating the visual cortex. Presumably the visual cortex can be
stimulated by other pathways too (Bill’s 'Many different structures can
effect the same function’ p 118 para 4). There are anecdotal reports about
visual forms (nameless shapes) arising during hallucinations and erupting
from exertion (sexual orgasm, as film makers depict with exploding
skyrockets) and migraines. This leads me to ponder upon the familiar
expression, ‘Go figure’. Do we do this figure-making intentionally, or does
it just happen?
RM: I think the “figure making” discussed in this chapter is something
that “just happens” when certain lower level perceptual signals are
input to a configuration perceptual function.
P. 122, the definition of configuration builds nicely from sensation
vectors. Now, how about considering reverse engineering?
RM: That’s what we are doing already, isn’t it? We have a product (a
living organism) that has been engineered by evolution. We know what
it has been built to do - control – we’re just trying to figure out
how it was built to do it. B:CP can be seen, I believe, as a first
step towards reverse engineering the controlling done by living
systems.
For instance let’s
assume there are stable, consistent, shared invariants; the basic evidence
is pretty robust, for instance we tend not to bump into things except in the
dark or other situations when there is sensory deficit/impairment.
RM:I don’t see how that is evidence for perceptual invariants?
Of course, as philosophers have opined, there are then family resemblances
arising from putting like with like. For instance if people were offered the
same set of photographs depicting different ‘things’ they’d be able to place
them into family groups, even without having to describe how they did so.
Indeed any words may get in the way. By applying sophisticated analyses to
these sorting data it would be possible to discover implicit invariants.
RM: I think we must mean different things by “invariants”. I think of
an invariant as a perception that remains the same despite changes in
its’ components. So a square configuration is an invariant if a
square is seen despite variations in, among other things, the
thickness, color and orientation of the lines that make it up.
Crikey, I’ve been doing this for about 30 years. This is a two-part story.
Here’s the part one, the practical. Give me a set of items (where N > 12)
and a bunch of people who can make ‘reasonable’ (as opposed to arbitrary)
decisions as to what goes with what in that finite set (it’s a bit more
complicated as we’ve a routine here as to how these families are
constructed) and I’ll provide a rendered aggregated rendered 3D ‘map’ with
orthogonal axes. Yes, OK, it took us almost a decade to fathom out and
create the maths involved from data gathered from dozens of simulated and
other studies, many of these using others’ data (special thanks to those who
chipped in freely). BTW, we’ve plugged in many types of items (sounds,
colours, morphed faces, shapes, beers, wines, personality adjectives,
cigarettes, cardiac oscultation, personality descriptors, interests, and so
on and so forth) and the methodological suite hasn’t come up short yet:
stable, robust maps emerge every time. Here’s one odd finding amongst many.
Drawing from a stable 3D map generated from a corpus of single word
personality descriptors I carried out yet another sorting study using a
subset of this adjectival set but omitted almost an entire quadrant. Yet the
intermap stress between the partial and the full one was respectably low.
This suggests even if polar exemplars are not provided participants behaved
as if the poles were still there. In short, it appears invariant axes are
present even if corresponding items are not. I’m sure you can fill in the
gaps.
RM: Multidimensional scaling can be lots of fun; I particular enjoy
Roger Shepard’s work. I think it can tell us something about how
people perceive the worlds at the category level, which involves
seening similarities between various entities. But I don’t think it
tells us about invariants so much as what might serve as convenient
dimensions for a spatial representation of the relationship between
perceptions.
That’s the practical part. And we are quite good at this too; we have
dozens of corresponding research examples and articles. We even call
ourselves psychological cartographers. Recall Kurt Lewin’s cute quip:
there’s nothing so practical as a good theory? Well, all these years during
which we’ve been piling up these sorts of practical results I’ve been
looking, nay searching, for a theory. And, here is is. Thanks for part two
Bill. It’s an epoxy.
Perhaps a sabbatical with a bona-fide PCTer might be beneficial.
RM: I’m not clear about the relationship between your mapping work and
PCT. Maybe the sabbatical is a good idea. Or, better, how about an all
expense paid vacation for my wife and I in New Zealand and I’ll talk
with you for at least 1 hour each day we’re there!
If anybody would like to chip in and at the same time see this approach in
action I’ve an item set on line that’s ready for sorting. It consists of
single words describing sounds that engineers and others have difficulty
communicating about, especially those working in high-risk settings like
aviation. I want to try and uncover the dimensions (aka invariants) for
distinguishing a knock from a dong from a tap and so on. It’ll take about 40
minutes. If anybody’s got some willing engineers, all the better. As with
almost all of the research I unertake, it could be fun.
All right Rick, you asked what we’ve got out of B:CP so far. For me, a
revelation. What makes this much worse is the delay since the book’s been
around for as long as we’ve been doing this type of research. Ouch. And
there’s a way to go yet.
RM: I’m not completely clear about what that revelation is but after a
couple of weeks in NZ I should have it nailed down (absolutely no
bungee jumping for me though)!
Best
Rick
With kind regards
JohnK
On Mon, Sep 9, 2013 at 6:59 AM, Richard Marken rsmarken@gmail.com wrote:
[From Rick Marken (2013.09.08.1200)]
Rupert Young (2013.09.06 21.00 BST)–
Thanks for this Rupert. These are excellent observations. My own
summary of Ch. 9 will be coming up soon.
Best
Rick
Chapter 9 extends the concept from the previous chapter of a hierarchy
of
control systems, with systems in a level dependant systems in other
levels
within the hierarchy. The configuration perceptions are a function of
lower-level sensations. I don’t have a lot to say but a couple of points
I
found particularly interesting:
that the frequency of oscillation of instability tremors would indicate
the
level within the hierarchy of the problem,
that the perceptual “representation” of an object, say, is a highly
distributed collection of (sensation) perceptions that is not organised
in
any way similar to the experience of a 3-D object (an approach taken in
computer vision)
Again my knowledge (ignorance rather) doesn’t enable me to comment on
the
validity of the neural discussion. I would really like to see some
research
in this area. With the advances made in recent years on neural mapping
it
would seem much more possible to identify feedback loops. Surely we
could
tempt some neuroscientist with the promise of power and glory to be the
first to investigate this new frontier!
Leading questions:
Tension of muscles, tightness of skin, keeping eyes open, position of
head,
eyes and hand.
Yes. Well, the muscular impression is different, but not the visual.
No. No, I can perceive edges etc, without needing to perceive an object.
Yes, I did that for dinner when I cooked Halloumi kebabs.
Regards,
Rupert
On 01/09/2013 23:49, Richard Marken wrote:
[From Rick Marken (2013.09.01.1550)]
Since it’s a Holiday Weekend I am going to take a holiday from making a
study guide for Ch. 9 (unless David is good enough to come up with one
on
his own). So read and comment on Ch. 9 by responding to this post.
Happy Labor Day
Best
Rick
–
Richard S. Marken PhD
–
Richard S. Marken PhD
–
Richard S. Marken PhD
[From Erling Jorgensen (2013.09.29 1020 EDT)]
John Kirkland 20130921 / 15:27:21 +1200
Hi John,
I apologize for the delayed stutter-step kind of conversation here. I
tend to control for knowing what I'm talking about & making comments
worthwhile to the listener, or at least asking intelligent questions.
So I've had to read up on some of the things you've raised.
I have read or skimmed a couple of your articles online -- e.g., MDS of
facial expression & emotion, the CCQ study that you mentioned -- & other
examples of Multi-Dimensional Scaling. I also read some things by or
about Charles Osgood & his EPA studies, as well as Semantic Differential
research in general.
Let me continue with some of the micro-scale comments & questions, to
get more clarification about your approach.
EJ:The first question I have is about what level of perception is being
constructed here. Is it Configurations or Classifications (i.e.,
Categories)?
JK: I'm not at all sure, though my hunch is these are deconstructed
classifications, perhaps nascent configurations.
I wonder if they appear as configurations because they are composed of
dimensions that remain constant (I like that word invariant) _despite_ the
variety of vectors that comprise them. That's my understanding of how
Bill Powers in B:CP discusses configurations. Nonetheless, they do seem
to arise out of a series of classification tasks -- as you say, Grouping,
Opposites, Partitioning, Adding (GOPA).
we summarise all
responses as a vector, much like a knitting needle piercing the 3D sphere
from the origin.
Does this mean there is a vecotr summarizing each participant? Or does
each item response end up somewhere in the trait space & thus as an
implicit vector of whatever axes define the space? I think you are saying
the latter, but I don't know if you are also saying the former?
These vectors may be interpreted using the three
orthogonal axes. And, what's more, if you want to shift a person's
perception then provide adjustments across the axes using neighbouring
items as stepping stones.
Say more about how you see this matter of shifting a person's perception.
It sounds as though the stepping stones are _other_ instances of the
person's own perceptions (derived through their own sorting data), & you
would be trying to draw their attention to other salient aspects than the
ones that are currently clustering together.
I'm of the view these adjustments could be
interpreted pretty well as PCT; shake the system and see it resolve itself
into a new niche.
Wouldn't PCT suggest that it would resolve itself, through disturbance
resistance, back into its old niche? I suppose if the shaking is forceful
enough to induce sustained error, then reorganization might kick in. But
the results of reorganization are rarely predictable. There's no guarantee
the old perceptions would stick around as stepping stones. Reorganization
seems to shift things to a new location in the perceptual space.
We use the term habitat to describe all
possible places of occupation within the 3D map. What a person negotiates,
where there is minimum stress (aka low error) is their niche, what they
have carved out. As Jack Block used to say, carve nature at its joints.
I like the metaphor of carving a niche within a perceptual habitat. And
Block's epigram is intriguing, although I'm not quite sure what it is
saying.
EJ:The following question presumes a 'yes' answer that the matter of
three axes (regardless of task) is a stable phenomenon...
JK: It was a 'yes'. What's presented is a hollow sphere, where items are
spattered on the surface of a ball.
Why hollow? How do the ends of the vectors arrive on the surface of the
sphere? Surely the vectors do not give equal weight to all three
orthogonal dimensions?? Shouldn't the end points of those vectors end up
somewhere within the volume of the sphere?
Clearly, it makes no difference if the
3 orthogonal axes are rotated within this sphere, the inter-item relations
do not change. We have designed some software to realign axes. One may
do this to assimilate existing theory (by assimilate I mean to align with
but not change the underlying structure.
This is an interesting way to think about different theories -- i.e., as
axis rotations whose proponents may not realize they share a common
underlying structure with certain other theories.
It used to bother me no end that we kept getting the same 3D map. ...
Way back when we were scared of these outcomes (how dare we get
consistency?) we used to collect more GOPA data than necessary and then
took random split halves (or quarters etc) to compare sub-maps. They were
always well over 90% agreement. We kept thinking we were doing something
wrong. Why? Because from my Missouri stats training such consistency is
suspect. Now I simply shrug my shoulders. Whatever we are doing it's
consistent.
I don't mean to insult you by asking whether any of the consistency might
arise as an artifact of the measurement method itself. As I understand it,
you ask people to make comparisons, of both similarities &
dissimilarities. It's iterative, with successive sorts, & seemingly a
shifting task focus as they go (i.e., your GOPA designations). Then your
software positions things spatially, with similar items closer & dissimilar
items further apart, representing all the inter-item correlations at the
same time.
Could there just be something about geometry, that reconfigures comparisons
into a spatial form, without necessarily imparting a new increment of
understanding (aside from a bird's eye view of the data)? Do
the 'underlying dimensions' have their own reality, or do they just fall
out of an X-Y-Z coordinate system?
I'm sure you have thought about such things, & I don't mean to sound
boorish here. I think I've probably learned from Bill P. as much as
anyone about asking impertinent questions. To say it most polemically,
expanding upon Korzybski's dictum: A map is not the territory, & the
conventions of mapping are not any territory.
My sense of it is that we have a relatively limited number of axes and
these get applied to the phenomena under scrutiny.
My immediately preceding paragraphs are kind of asking about the "we".
_Who_ has the limited axes that "get applied" to the data? You do say a
bit later in your post: "Participants have no idea about the underlying
architecture they deploy for grouping items."
Having raised that side of the question, it is only fair to also consider
the other side of the questions -- i.e., whether there is an underlying
architecture, held in common when people sort & make comparisons, that
gets "deployed."
Osgood did much better
and we too often find his E P A (evaluation potency activity) as the
axes.
I had to look up this reference, & read up on Charles Osgood's Semantic
Differential approach. To summarize, for others who may be listening in --
When expressing attitudes & ranking items in a set of data, people often
use a good-versus-bad Evaluation distinction (which is often the strongest
loading when the data are analyzed), then a strong-versus-weak Potency
distinction, & an active-versus-passive Activity distinction. Apparently,
those dimensions are exceedingly common & stable in the research
literature.
Let's go back to a 3D map and its axes. We have designed software to mark
groups on the map. And these can be labelled.
My initial interest is less about the groupings within the 3D map. I'm
more curious about the axes themselves. You've already established that
their emergence is a consistent phenomenon when you analyze the sorting
data. When you don't use Osgood's EPA labels, how do you characterize
those axial dimensions? Does it depend on how you rotate the axes?
When we aggregate sorting data there is a consistent MDS map obtained,
usually rendered into 3D. As an side, I can provide one to anybody who's
interested as we designed a software suite for view these; they can be
rotated with dimming effects
Yes, I'd like to see an example.
For instance I
was once interested in how interpersonal conflict was assessed. ...
I"d be interested in seeing more detail about this study (or meta-study).
Perhaps some representative items that were sorted, along with the
instructions that were given about how to sort. I work as a
psychotherapist & have had training in mediation, so data about
interpersonal conflict is intriguing to me.
EJ: I believe every level of perception involves the construction of
implicit invariants, in a sense creating a new world of perceptions
out of (or on top of) the lower level(s). ...
JK: We are on the same line. If this was choral work we'd be harmonising.
... I am of the view each of Bill's levels offers plenty of scope for
scientific exploration; some may wish to work at the neuropsychological,
others may delve into the sensory arena, others probe at the higher end
(eg MoL). There's plenty of opportunity.
Agreed!
It's what's held in common which
provides a unifying theme: closed loop, comparators, error, negative
feedback and so on. In short B:CP.
Exactly!
Thanks for some stimulating posts, John. I await hearing more about it.
All the best,
Erling