[From Bruce Nevin (05.28.2004 22:50 EDT)]
Please read carefully, our prompts have changed.
Bill Powers (2004.05.26.1848 MST) –
Bruce Nevin (05.26.2004 13:36 EDT)–
This is an interesting argument that we’ve seen before,
Just as saying is not telling, seeing is not understanding.
- Martin’s sketch
of “convergence” of reference values in autonomous
control systems indicated how this process might be modelled in the
individual specimens involved.
Is there really convergence of reference values?
The pair test and its variants demonstrates that there is.
Imagine an experiment: I read the above sentence, “Martin’s sketch
of ‘convergence’ of reference values in autonomous control systems
indicated how this process might be modelled in the individual systems
involved”, recording it on a sound recorder. At a different time and
place, you record yourself reading it, Mary records herself reading it,
and Bill Williams records himself reading it. (I assume good faith, i.e.
that these are well recorded, clearly spoken, with no masking or
interfering sounds, etc.) We present these sound recordings, repeated in
random order, to a succession of people attending the 2004 CSG
conference. After the playing of the first recording, we ask each time if
it is the same sentence as the previous time. We allow the previous and
the current one to be replayed on request, for comparison. After a time,
we ask the listener to write the sentence. My prediction: All listeners
will judge all the recordings to be readings of the same sentence, the
same arrangement of words. And the written transcripts will be the same.
And this despite manifest differences in the sounds that are
recorded.
When I say that they will be the same, I do not mean that physical
measurements that you or someone else could conduct will come out with
the same readings. I mean (to repeat what I said earlier) that a given
person will perceive them as the same, and that you will perceive their
reports that they are the same. We are not talking about the continua of
physical phenomena, we are talking about discrete perceptions, where
these sounds or this sequence of marks are perceived as this word or that
word. Language and culture are entirely on the binary side of the fence,
however much their performance is among the continua of physical
phenomena.
Nor am I talking about the meaning or the understanding of that sentence
being the same for you and for some other person, or for any two people.
(The only evidence that meanings are the same or not is in consequent
words that people say and things that they do, which are not themselves
meanings, but rather have their own meanings.)
- What is
neglected in this approach is, as Bill Williams says in the
later message quoted below
An infant specimen arrives into a context in
which there are patterns of
coordinated reference levels.
Are there really patterns of coordinated reference levels in the
world?
This follows from the above. But there is a better way to come at this
than talk of coordinated reference levels.
- In particular,
these patterns of convergence among reference levels that
facilitate coordination and enable individuals more easily to avoid
conflict with one another are cumulative over time, as well as
changing
through time.
Th[is] says the patterns of reference levels are cumulative over
time
and change through time. Are they really cumulative, and do they really
change?
I’ll repeat an account I have summarized previously on the net.
Eventually its significance will be apparent. In the 19th century, one of
the distinguishing features of the dialect where I live, aside from
retention of postvocalic r, was the raised and centralized position of
the first part of the diphthong /au/ in about, loud. (You
hear this frequently in Canadian dialects.) This had been gradually
merging with the more open /a/ heard in mainland dialects. This broader
sound, along with loss of postvocalic r (pahk the cah in Hahvahd Yahd),
had spread from dialects south of London over to the cities of the
American colonies, and thence spread to more rural areas. The older
pronunciations persisted in only a few regions, and were fading even
there. However, after the whaling trade died and as tourism became more
of a factor around the turn of the century, the traditional pronunciation
began a resurgence. The really puzzling thing was that the /ai/ diphthong
came along with it, in words like knife, slide, where it
had never been recorded in any dialect of English, certainly not in New
England. In a study that Bill Labov conducted in the early 1960s, it
turned out that adolescents who were leaving the Island for college or
jobs on the mainland changed their dialect to conform to that of summer
visitors from New York and Boston, while those who were looking ahead to
making their livelihood here continued to speak with the local dialect,
only controlling its distinctive features (as distinct from that of the
visitors) with higher gain, “more conservative than even the
Menemsha lobster fishermen.”
Here, we see evidence of people perceiving and controlling for norms, and
we see evidence for social processes by which norms change.
The most obvious reason that they must be cumulative is evolutionary:
they cannot have sprung into being full-blown by some miracle when, say,
a precocious subset of the offspring of a band of hominids woke up one
morning. Humans did not always tie their shoelaces with bow knots, the
arts of knot tying having developed by an accumulation of learning, and
some many generations before that they did not put shoes on their feet.
Given this explanatory notion, perhaps no more than a just-so story, one
may be willing to look for evidence for small accretions in know-how and
in custom, in how we learn from one another and make what seems a good
way to do things (or a good thing to do, or merely the sort of thing that
is done by the kind of people we esteem and perhaps aspire to be) a part
of our own ways of doing things or one of the things that we ourselves
do. The adolescent baboons that join the troop described in the article
at
http://www.nytimes.com/2004/04/13/science/13BABO.html
for whatever reason, by whatever means, come to behave normally for that troop, which is different from the way adolescent baboons conduct themselves in every other troop that has been observed. (See Bruce Nevin 2004.04.13 14:30 EDT and sequel.)
The field of culture contact is especially informative, just because the ways of doing things and the things that are done are obviously different, at least at first. Then members of each group may begin to adopt some of the ways of the other, or may begin changing their ways to accentuate the contrast with the ways of the other, perhaps outright rejecting features of their own heritage which fortuitously too much resemble how those rejected others do things. Adolescents controlling an identity as an islander accentuate perceived characteristics of their elders’ speech, even overshooting the mark, while adolescents controlling an identity as an off-islander change their way of speaking to sound like the summer visitors.
But this is all controlled without attention, and though it is happening around us all the time, and even in our own behavior, only rarely coming to awareness, it may be that you have to spend a protracted period being a foreigner learning how to be a native for this to be more than hearsay.
Something is missing from each of these statements: an indication that the
statement is about a perception in the person making it.
And something yet more: that these are verifiable statements, which have been verified by numerous people, and which are capable of test and verification by yourself – if you are willing. But it cannot be done alone.
I don’t know if the above three statements are true or
false. What I do know is that I hesitate to acquiesce to them, because my
reaction is not “How true,” but “Wait a minute.” When I say "Wait a
minute," I mean, “Is that really true?”
Find out. The keys to investigation are ready to hand. If you haven’t the interest to investigate and find out, then “Is that really true” is a rhetorical question covering disinterest at least.
Take the statement that there are convergent patterns of reference levels
in the world. It may prove true, if anyone could figure out how to test the
idea, that we can discern trends, parallels, similarities in various
dimensions among the reference levels people prove to employ. But we should
never forget Tom Bourbon’s demonstration of “learning curves,”
No one ever claimed that these learning curves are a social reality, a cultural artifact. The rings and arcs and following in lines of the CROWD program are precisely not cultural artifacts, because they are universal, and indeed the model shows why they are universal. Cultural norms, “social realities” such as we are talking about, are in their nature not universal, they are conventional, with the conventions differing arbitrarily (or, rather, with historical contingency) from one another, subject to variation and change governed by social factors such as the “what kind of person I present myself to be to others, based on my generalizations about what kinds of people there are” control of self image that seems to be operative in the change in dialect described earlier.
Attend closely to the social reciprocity in that specification: “what kind of person I present myself to be to others, based on my generalizations about what kinds of people there are”. Bear in mind that each of those other persons is concurrently controlling presentation of self image to others in the same way, and controlling generalizations about the others based on their behavior as they control generalizations about the others. It is the loop closure of this reciprocity that is the basis for convergence. And it can only be modeled in individuals concurrently controlling in a public.
I mean the term “a public” in Dewey’s sense: an association of people that carries on some task that is larger or more complex than can effectively be accomplished by an isolated individual. (Thanks to Bill Williams for clarifying this to me last February, in another context.) We are talking about conventionalization and institutionalization in a public. Conventionalization is similar to categorization, in that variants constitute tokens of a type, except that it only occurs in the recognizedly conventional context (in context of speech, what was heard as a sound is perceived as a word) and involves more than one individual. The nudge by the gorilla baby signifies a request to nurse. Institutionalization establishes a convention in a public. Each conventional nudge or poke is unique to each gorilla mother-baby dyad, but the word “please” is institutionalized among speakers of English.
A reminder from Bruce Nevin (2003.03.18 10:58 EST):
Captive chimpanzees have been observed to use a number of manual signals that share more features of words than do most other animal signals. The pokes by which infant chimpanzees indicate that they want to nurse are a good example (Tomasello, M., Gust, D., & Frost, G. T. A longitudinal investigation of gestural communication in young chimpanzees. Primates, 30, 35-50 (1989). These pokes begin when the infant simply tries to push its mother’s arm aside in order to get at the nipple. As mother and infant become accustomed to each other’s behaviour, the push can be simplified and conventionalized until the infant finally needs only to poke the mother in a characteristic way, and she will adjust her position to allow her baby to nurse.
[However, the convention is different for each mother-infant pair. The reason for this is obvious: there is no cross-communication from one such pair to another.]
http://www.psycprints.ecs.soton.ac.uk/archive/00000667/
The baby chimp can repeat this gestural word. There is reason to suppose that the precursors of language are in gesture rather than in vocalization (nice summary at http://www.humanevolution.net/a/signlanguage.html).
Conventionalization is necessarily social. It arises and is established through interactive usage. It is conventionalization which is at the root of culture and language.
I would like to raise the possibility that the convergent patterns of
reference signals may be something like the convergent pattern in the
universal learning curve. If you study how large numbers of people use
language, or principles, or money, or table manners, you will be able to
distill patterns very reliably out of the observations. But why should we
treat this pattern as anything more than the average of how people do
things? If we were to look at how each person behaved in these respects,
we would see, I propose, a continual flux of behavior as each person
interacts with and adjusts to the infinitely varied behavior of those
around him or her. It would of course be possible to sum up, to generalize,
to find average ways of interacting, but could that average way be said to
have caused the variety of behaviors of which it is the average?
There is a difference between an average and a norm. Put the other way around, what you have said above could be said as well of any sequences of performances controlling perceptions in accord with given reference values. The variations in performance could be averaged, but that average would not be the reference value(s), nor would the variability prove that there were no reference values.
Behavior is determined by reference values and disturbances. Reference values may be set by systems that control norms. Systems that control norms do so with varying reference values – more or less of the given norm – and with varying gain. When you have a socially established norm, individuals can conform to it (adopt it as a reference value) or not, electively. When the individual does conform to a norm, others who know the norm recognize it and perceive the social significance of their conforming to it; when the individual does not, that, too, has a perceived social significance. There are frequently competing norms, each of which constitutes nonconformity to the other(s), each of which along with other norms is associated with the “kind of person” who acts that way, so that in addition to the above a person may conform to a norm which is perceived as conflicting with the first norm, and this may be perceived by others as a particular way of controlling the first norm (aversively). No averaging can possibly approach an understanding of this, and the comparison to the learning curve business is, as Perry Mason used to say, irrelevant and immaterial.
Whatever statements we make, they are statements about our own perceptions.
Of course. But we are not isolated individuals each building our own simulations. Intersubjective test, verification, and agreement is possible. And for science, it is essential.
What is needed is a model of many individual specimens that includes a
mechanism by which they can influence each other,with a demonstration that
their reference values do in fact (in the modeel) converge to common
patterns rather than just adding up to an average pattern.
I agree, but failure to produce such a simulation would not demonstrate that this is false. There is evidence of very limited cultural norms in mammals and birds, but not in simpler organisms. I think we are far from being capable of creating a simulation that is sufficiently complex.
- What is neglected in this approach is, as Bill Williams says in the
later message quoted below
An infant specimen arrives into a context in which there are patterns of
coordinated reference levels.
Needed here is a model in which a pre-existing set of control systems is
joined by a new one, so we can demonstrate how the new entity’s reference
signals change in relation to the existing ones. First, of course, we must
establish that the coordinated reference levels are not just averages over
sets of very different reference levels
Again, I agree, but again, failure to to produce such a simulation would not demonstrate that this is false.
- In particular, these patterns of convergence among reference levels that
facilitate coordination and enable individuals more easily to avoid
conflict with one another are cumulative over time, as well as changing
through time.
And the same comment applies: we need a model in which patterns of
convergence can be observed, so we can see if they are cumulative over time
as well as changing through time.
And the same caveat applies, about drawing any conclusions from failure. But we are not limited to construction of simulations. We can observe and test people and gather the data that are prerequisite to modelling and building a simulation – the data that any simulation would have to replicate.
My point is that none of the original three statements involved statements
of fact, physical or social. They involved proposals. We always have a
choice when proposals are made: we can say, :Yeah, that strikes me as
reasonable," and acquiesce. Or we can say, "That would be interesting if
true, but before we go on from there we had better find a model that will
work that way, and that fits what people actually do."
Or you can say “Yeah, that strikes me as reasonable, so now I’m willing to invest the time and energy to gather (or to look at) data about what people actually do, see if the proposal continues to be consistent with the data, and begin to understand how a simulation might be organized. Because until I look carefully and deeply at the data I haven’t a clue what I’m talking about. Now that I know what we’re talking about, I can at last recognize what constitutes relevant data and begin a serious investigation.”
The crucial thing that has been missing is that other individuals are in the environment of that individual that we have singled out for observation and modeling, and she or he is part of the environment of each of the others. Well, OK, that is present in a limited way in the CROWD program, and in the dyadic stuff. What is missing is the further process of conventionalization that establishes a type-token relationship as a basis for communication, and the process of institutionalization that standardizes conventions in a public.
The starting place for this in the individual is generalization. That people generalize about others I take as obvious. (A generalization is not a category, though it may employ them, and it is not an abstraction, nor is it an average). Perhaps you disagree, or perhaps you take that as a proposal that must be modeled successfully before you can entertain it as credible. I further propose, then, if you will, that people perceive the generalizations that they have made or learned about others as norms to which (they perceive that) a given other either conforms or disconforms. And they indicate to others their own conformity to what they perceive as an accepted norm by means which become conventionalized, and eventually institutionalized within a public in which they participate.
It seems to me that we do not often or usually conclude that our generalizations about people are irrelevant for a given other, and evidence of discomfort at being unable to “peg” a person suggests that fitting them into one’s generalizations about people is itself a control process. This is easily tested by what is colloquially thought of as violating someone’s expectations. In our youth many of us deliberately engaged in violating the expectations of our elders, épater les bourgeois as companions of Baudelaire put it, as a way of resisting … what? What is “peer pressure”? Does this have something to do with conflict between people? By conforming to the generalization that you imagine that the other person perceives you fitting, do you avoid conflict so that you can proceed with tasks that require cooperation? Can the conflict attendant on your violating such an expectation avert a worse conflict by persuading the other person to fit you into a different generalization instead?
Marc Abrams (2004.05.27.0017)]
Control systems are not about reducing ‘conflict’. Control systems are
about getting what they need and want to minimize error. Because of
this, ‘conflict’ is the ‘norm’ among many competing control systems. We
need to understand better ways of ‘managing’ conflicts because we will
never avoid them and we cannot ignore them. You might side-step one,
but eventually one is going to ‘get’ you. The ‘vision’ of a ‘happy’
world free of conflict is a fantasy.
It is easy to demonstrate that people perceive conflict and control that perception, sometimes to avoid it, sometimes to foster it, and that is what is relevant here.
I’ll doubtless regret having said so much, because people will respond only to those parts that can be made consistent with prior understandings or which seem easiest to dismiss, preserving those understandings. So it goes.
/Bruce Nevin
···
At 08:27 PM 5/26/2004 -0600, Bill Powers wrote:
When we find a better way to