Heterarchy

[Erling Jorgensen (950628.1325CDT)]

(Bruce Nevin 950626 19:52 EST)

Erling: Your words here suggest that comparators for perceptions of a given
level of the hierarchy "live" in a universe composed of perceptions of the
next lower level.

This is a place to start more than an ending point for me. It helps
me keep in mind the matter of "point of view." Every perception is
just one of a multiplicity of perceptions and "vantage points." It's
actually quite a radical and "postmodern" reminder in that respect.

Not too many months ago I asked (Thu 950306 11:08:16 EST)
whether this so tidy and attractive "concentric perceptual universes"
conception was valid. I could not make it jibe with the heterarchical I/O
connections that a category level must have.

I remember the discussion and found it extremely useful. If Categories
are constructing determinations of "sameness," there's no real reason to
restrict their inputs to the (presumably) next lower level of Relationships.
The Hierarchical PCT model has always postulated copies of lower level
signals ascending to higher levels_ anyway, so the input connections
are there in the model.

Bill Powers responded (950307.1625 MST) with
some other indications of heterarchy with respect to relationship
perceptions

He won't like to hear this, but I confess that I often watch for Bill's
"imprimatur", to help shape my sense of "acceptable modifications" to
the theory. (It's not an authority thing so much as he's just so darned
persuasive!) I agree with you that limiting connections to immediately
adjacent levels has seemed excessively tidy. That's been more a
function of our need for analytical simplicity, while we try to figure
the thing out. But there is a genius, I believe, in trying to build
things out of the simplest components possible. There's a lot that
can be done with replicating a good design all over the place.

With regard to levels of a hierarchy / heterarchy, it is not at all clear
that the central nervous system "knows anything" about these levels.
We need not directly perceive the structures that give rise to our
emergent abilities to perceive. The exception might be the "reorgan-
ization system," which might perceive (local? global?) degree of
error as part of its intrinsic variables, and/or the state of certain
parameters; (this is all still speculative.)

You quote Bill:

What all this suggests is that any level of perception can be a function
of perceptions of _any lower level_.

It helps me to consider each level as a _transformation_ of the lower
level signals, i.e., as a different type of perception -- different in
kind not just degree. If multiple lower levels could contribute input
to a higher level, perhaps we should be specifying the levels in
question. For instance: Transitions [Level 4] among Sensations [Lev.2]
(e.g., watching colors fade in a sunset); Relationships [L6] between
Configurations [L3] (e.g., is the chord major or minor); Categories
[L7]of Events [L5] (e.g., "it's as easy as riding a bicycle"); Sequences
[L8] of Relationships [L6] and Categories [L7] (e.g., "i before e,
except after c").

Continuing the quote from Bill:

     ...a function
of perceptions of _any lower level_. Why not higher levels, too? Well,
I've tried that on, and all I can say is that I can't make sense of it.
Maybe someone else can.

Wouldn't perceptual input signals from a "higher" level lead to
oscillations and instability? I don't think it's just a product of
modeling via discrete computer systems, that higher level systems
must work on a _slower_ time scale. Bill's 1979 article ("A Cybernetic
Model for Research...) suggests that it's a requirement of an analogical
hierarchy as well. Even though everything is happening simultaneously,
each level has its own definition of "an instant," which is necessarily
longer for a higher level system.

George Simon ("Revisiting the notion of hierarchy." _Family Process_,
32:147-155, 1993) has a similar concept: "hierarchy is not conceived
as levels of status, power, or authority, but as levels of temporality."
(p.149) He quotes Elisabeth Fivaz-Depeursinge ("Documenting a
time-bound, circular view of hierarchies: A microanalysis of parent-
infant dyadic interaction." _Family Process_, 30:101-120, 1991),
who gives evidence for "the typical durations of a level's episodes
in relation to others. The longer a level's time constant, the higher
the level is placed in the hierarchy." (p.110)

Her study of play dialogue in parent-infant interactions may be a
way to start thinking about _contexts_ of meaning. Granted, she
is layering various "episodes," which may all be within the Sequence
level of HPCT. Still, I think she's on to something about how
situations are framed. I'll quote one of her paragraphs to give you
a sense of it.

"Episodes of reciprocal gaze (that is, both partners being visually
oriented toward each other) have been demonstrated to form a
frame in which briefer episodes (expressive displays) are nested.
These gaze episodes are themselves embedded within longer
episodes organized by the body configurations. At the earliest
stages of infancy, these longer episodes consist in holding, namely,
the posturing and supporting of the infant by the adult and the
proprioceptive and tonic participation of the infant in these
postures." (p.103)

In essence, time duration becomes an ordering principle for
embedding and contextualizing different episodes. That lines
up with a system requirement that we think we're finding in the
hierarchical control of perceptions, such that, higher levels have
a longer / slower time frame. This also interests me as a way
to conceptualize (and/or measure?) _shifts_ of communication
context, which could be useful in counseling research.

Fivaz-Depeursinge addresses this latter point as follows:

"By recontextualization we mean a _process of reordering
levels_ by which a subsystem within a hierarchy _switches_
levels; for instance, gaze interaction in relation to holding
interaction switches from content to context level. . . . At
some point in the dialogue sequence, a reversal of the time
constants of the holding and gaze episodes was observed;
the holding episodes switched from nesting the gaze episodes
to being nested within them." (p.107)

I don't think any of this refers to an actual switch within the
HPCT levels. But it does raise the difficult issue of what forms
the context for what, and can those contexts reverse. Maybe
the circular causality of the control loop(s) is a way out of the
dilemma, where a (higher) reference signal is the context for a
(lower) loop, and a (lower) perceptual input is the context
for a (higher) loop. More likely, it is switching / reversing the
means for accomplishing a yet higher goal.

In the article, Fivaz-Depeursinge gave the following example
of this switch in contexts: "When a daughter showed that
she no longer liked the sitting position, her father gave in
to her and held her standing in front of him, both looking
intently at each other. She alternately danced and paused,
in brief episodes nested within long episodes of reciprocal
gaze." (p.108)

I think there is some potential for PCT analysis of this kind
of study, even though it moves into a thorny area of interactive
(and/or coordinated) control by each person of their variables.

In any event, thanks for raising the opportunity for getting into
this. I think it's important to restate every so often the emerging
consensus about aspects of the theory, and I thought you
summarized well what is at stake with a notion of heterarchy.

All the best,

Erling

(Bruce Nevin 950706 11:15 EST)

Erling Jorgensen (950628.1325CDT) replying to Bruce Nevin (950626 19:52 EST):

I have enjoyed your contributions. (That is, the ones I have read. I'm
gaining, the digest for 6/15/95 is next. Thanks for the heads-up about your
response so I could reply before August!) We could have a lot to say to
each other about Pearce's coordinated management of meaning, _Family
Process_, and I am sure much more. In a presently unlikely-seeming future
maybe we will be able to.

A problem with accounts like that of Elisabeth Fivaz-Depeursinge (which you
quote) and Pearce too as i recall after, hummn, well, since 1979 or 1980,is
that it is based on a taxonomy of behaviors that are observed by the
investigator. What are the participants perceiving, and which perceptions
are they controlling? "Episodes of reciprocal gaze (that is, both partners
being visually
oriented toward each other)" are probably not perceptions that the partners
are controlling; "seeing that you are looking at me" might be. "Expressive
displays" and their apparent (to the investigator) coherence and structure
are probably byproducts of the participant's control of perceptions that
the investigator has not identified. Body configurations do not organize
the longer episodes within which gaze episodes appear to be embedded,
though changes in what perceptions are being controlled in one episode vs.
another may be reflected in incidental changes in body posture that the
investigator may conveniently use as indicators of the changes. And so on.

While this sort of parsing of observed behavioral outputs into episodes
helps to organize the investigator's data, the suggestion that it (the
parsing, the structure) organizes the behavior itself is surely false.

(But on the other hand if the investigator can use a shift in posture as an
indicator of a change in what the participant is controlling, so may a
co-participant use it; and beyond that a participant in a community and its
culture may learn to use a conventionalized change of posture in ways that
other participants use it as a social signal of personal intent.)

That some of these episodes are on a longer time scale than others which
may be embedded within them is probably a function of controlling sequence
perceptions (as you suggested, I think), complicated by the social
coordination of autonomous control systems if that is the nature of the
sequence, rather than being a function of the differences of time scale
between a higher and a lower level of perceptual control.

When you get at what perceptions are being controlled I think you may find
that the seeming conundrum, whether gaze episodes are context for holding
episodes or vice versa and how they can reverse roles, dissolves. As you
say, "More likely, it is switching / reversing the means for accomplishing
a yet higher goal." Getting at what the goals are--controlled perceptions
and their preferred values --is the challenge.

        When a daughter showed that she no longer liked the sitting position,
        her father gave in to her and held her standing in front of him, both
        looking intently at each other. She alternately danced and paused,
        in brief episodes nested within long episodes of reciprocal gaze.

This perhaps could be restated as a supposition that the girl was
controlling a perception of sitting and then controlling a perception of
doing something that the investigator calls dancing, which precludes
sitting; and a supposition that the father was controlling a perception of
colluding in her control of each of these perceptions in turn. Perhaps the
"gaze episodes" are means for fathoming the other's intentions and of
communicating one's intention to collude with the other.

It would not be easy to apply the Test by disturbing what might be a
controlled variable. In linguistics, the Test has been described (Z.
Harris, _Methods in Structural Linguistics_, U Chicago 1952) as the pair
test. There, it is admitted that the test is not practical for gathering
and analyzing data in an actual field situation. It is useful for resolving
difficult points, yes, but mainly as a methodological standard by which all
less formal methods must be assessed, and in which they are ultimately
grounded. In practice, linguists follow all kinds of hunches, rules of
thumb, heuristics, and guesswork, subject to cross-checking and
confirmation as additional results become clear. In principle, all these
informal methods are based upon the pair test: roughly, pronunciation x is
offered as an instance of word X and a native speaker either accepts it, or
rejects it as peculiar, or accepts it as an instance of a different word Y
rather than X.

I think we need to follow this example and develop and learn to use a body
of less formal means for getting at what the controlled perceptions are,
with an eye to the grounding of these methods in the Test, and especially
their capability of being verified if necessary by using the Test formally
in a laboratory setting.

        Bruce Nevin

Erling Jorgensen (950720.0915)

Bruce Nevin (950706 11:15 EST)

A problem with accounts like that of Elisabeth Fivaz-Depeursinge
(which you quote) and Pearce too...
it is based on a taxonomy of behaviors that are observed by the
investigator

This limitation is faced by PCT research, too, I believe. Controlled
perceptions are first inferred through observations of the person's
behavior. A difference may be that PCT realizes it is looking either
at _corrective_ actions (yielding "correctified perceptions"), or
at irrelevant side-effects of behavior. The former can be analyzed
further via "the Test", whereas the latter are confounding observations
that will likely get in the way.

What are the participants perceiving, and which perceptions
are they controlling?

You're right. This is always the key issue.

"Episodes of reciprocal gaze (that is, both partner being visually
oriented toward each other)" are probably not perceptions that the
partners are controlling; "seeing that you are looking at me"
might be.

In one way I agree with you. If an episode - as a Sequence perception -
is being controlled at all, it is a composite of lower level percep-
tions. [I was going to call them "more basic, or more important
perceptions," but it seems all _controlled_ perceptions are presumably
"important" to the person.] For instance, I seem to recall from
Developmental Psych. that contrasts of light and dark, and specifically
the contrasts of the human eye, are more attended to by infants. So it
seems that the infant is stabilizing its perception of edge-Sensations
and/or circular-Configurations and/or color-Sensations, and/or
dilation-Transitions, etc. You suggested, "seeing that you are
looking at me." That sounds as if the Event-of-looking and/or the
directional "at"-Relationship may also be controlled here.

If those various layers of perception are being simultaneously
controlled, why not the Sequential perception of an "episode"
as well? Granted, it's a little more tricky - if Pearce and
Cronen are correct - because it may require "coordination"
between two people to stabilize it into an episode.

"Expressive
displays" and their apparent (to the investigator) coherence
and structure are probably byproducts of the participant's
control of perceptions that the investigator has not identified.

I agree that this is first of all an indicator of the investigator's
skill in chunking the observations into coherent perceptions. So it
tells us that humans (i.e. investigators!) can perceive mini-episodes
called "expressive displays." Whether these are Event- or Sequence-
or even Category-perceptions, I'm not sure.

But there also seem to be ritualized games that parents and infants
go through. While being held semi-standing on the parent's lap,
facing the parent, an infant can be observed smiling-dancing-
stopping-gazing and repeating the cycle several times. These actions
(Events in Sequence?) are more or less matched with gazing-smiling-
facial gestures, etc., on the part of the parent. The result is an
emergent ritual where _both parties seem to know their parts_. So can
we say there is coordinative-control of some perceptions here? Is this
a Program for Events in Sequence, or some such thing?

While this sort of parsing of observed behavioral outputs into
episodes helps to organize the investigator's data, the suggestion
that it (the parsing, the structure) organizes the behavior itself
is surely false.

I appreciate your caution here. In some sense, this parsing every-
thing into episodes runs the same danger as the behaviorists, where
behavioral-Events and contingent-Relationships are the only things
considered important; (perhaps an overstatement, but you get the idea.)
It is also true that behavior is "organized" according to whatever
will achieve stable control of important variables, within the given
field of disturbances. In other words, it will vary as necessary.
The organization happens viz a viz perceptual inputs, not behavioral
outputs.

Having said that, the proposed PCT hierarchy/heterarchy has a place
for things like episodes, as items of perception. My interest in
Fivaz-Depeursinge, as well as Pearce and Cronen, is that they
recognize some hierarchical organization of what might be perceptual
variables, _at the levels_ where communication and/or counseling
may come into play.

To understand counseling from a PCT frame of reference, I have to
consider "the social coordination of autonomous control systems,"
as you put it. Yes, this may veer into _emergent_ patterns,
which from the standpoint of analyzing mechanisms of control are
epi-phenomena. And there is no guarantee that just because the
investigator can stabilize the observations into a particular
pattern, that the participants must be doing it the same way. (This
is the error of the objectivist agenda of modernism, I believe.)

Still, human participants can perceive as many layers and types of
perceptions as can investigators. And it is certainly true that
they are controlling on many levels simultaneously. So how do we
investigate the _relative interactions_ of those levels?

I think you recognize some of these issues in your next remarks:

(But on the other hand if the investigator can use a shift in posture
as an indicator of a change in what the participant is controlling,
so may a co-participant use it; and beyond that a participant in a
community and its culture may learn to use a conventionalized change
of posture in ways that other participants use it as a social signal
of personal intent.)

A good example, I think, is a person's entry into a conversation two
other people are having, (what might be called, "cocktail party
dynamics".) There it seems that a person's proximity-Relationship
to others becomes "a social signal of personal intent." It can even
be used in the Test. Try varying the distance where you stop when
you approach an existing conversation. Notice at what point the
others tend to a) notice you, b) acknowledge noticing you, c) shift
posture to allow your entry into "their space," and d) shift the
conversation to include you (if at all). It _looks as if_ the
body configurations of variable (c) help to frame the more sporadic
(i.e. shorter time frame) interactions of variable (d).

That, too, can be tested. Play the part of one of the original
conversants. See what happens when one or both of you _don't_
shift your posture. Does the conversational shift of variable
(d) happen - i.e., does the "intruding" person break in or do
they tend to wander off to some other dyad?

I've played with some of these dynamics, and I have the impression
that body configurations are part of what shapes "what episode we're
conducting" and/or "who's included." I'm not sure the exact variable,
either for the "behaver" or the one picking up the signal. It may be
direction of gaze, it may be relationship to a person's back, it may
even include a metaphorical association to "the cold shoulder."

I do believe, as you mention, that conventionalized signals arise
among cultures and communities, which assist people in controlling
their own perceptions, and coordinating that control with others.

All the best,
   Erling