Control of Language

[From Rick Marken (2003.01.03.1100)]

Bruce Nevin (2003.03.01 12:51 EST)

Bill Powers (2003.01.03.0848 MST) --

>I am happy to let you have had the last word on "controlling for." The
>term sprang up at a student-sponsored seminar I gave at Northwestern in
>1971-2, so I guess it's true vernacular. I had nothing to do with it.

Seems like language is out of control. ;->

I know that you're joking but I think control of language is a very interesting
topic. Obviously, people control language incredibly well. If they didn't,
cooperative accomplishment of very complex goals (such as building computer chips
and formulating models of human behavior) would be virtually impossible. Also, if
linguistic behavior were not under pretty tight control, phrases like "controlling
for" and "I see you have chosen" would not produce such strong opposition.
Clearly, "controlling for" is a disturbance to some linguistic goal for which you
are controlling just as "I see you have chosen" is a disturbance to a linguistic
goal for which I am controlling.

I think we use language to evoke imagery in others that corresponds to the imagery
in our own minds. But it's often hard to tell how well we are doing at this
"evoking the same imagery" business since, of course, we can't perceive what
another person is imagining. Perhaps the best way to tell if there is a
correspondence between what we are imagining and what image our words have evoked
in others is to see whether the person we address can reply with new language that
evokes in ourselves the imagery we hoped to evoke in that person. Bruce Gregory
did this extremely well with "controlling" and "controlling for". His use of the
example of a baseball pitcher to illustrate the difference between "controlling"
and "controlling for" indicates that the mental images evoked in him by these two
phrases is exactly the same as those evoked in me. So in this case, Bruce G.
demonstrated excellent control of language inasmuch as he was able to communicate
his intended meaning, which precisely corresponded to the meaning I would have
been controlling for had I used the terms "controlling" and "controlling for".

Best regards

Rick

···

---
Richard S. Marken, Ph.D.
Senior Behavioral Scientist
The RAND Corporation
PO Box 2138
1700 Main Street
Santa Monica, CA 90407-2138
Tel: 310-393-0411 x7971
Fax: 310-451-7018
E-mail: rmarken@rand.org

[From Bruce Nevin (2003.03.01 19:11 EST)]

Rick Marken (2003.01.03.1100)

I know that you’re joking but I think control
of language is a very interesting topic. Obviously, people control
language incredibly well. If they didn’t, cooperative accomplishment of
very complex goals (such as building computer chips and formulating
models of human behavior) would be virtually impossible.

The serious side of my quip

Bruce Nevin (2003.03.01 12:51 EST)
Seems like language is out of control. ;->
… hinges on the distinction between language and the use of language.
This may be familiar as de Saussure’s distinction between langue
and parole, and is somewhat parallel to Chomsky’s distinction
between competence and performance.
Viewed naturalistically, language is an emergent phenomenon with a
mathematically specifiable structure. In detail it varies across
communities of use and changes through time. It existed before you were
born, you learned it, and it will exist after you have died. But it can
exist only in its use, or in records of its use. What this
self-organizing system “emerges” from is its use by thee and me
and all the rest of us concurrently. It varies and changes pretty much
regardless of my personal preferences, or yours, or Bill’s. We know in
principle that we exert an influence, but with rare and maybe illusory
exceptions not enough to see or hear and say “I did that, I caused
that characteristic of English”. There’s no way I could prevent
“funner” from becoming current usage or to forestall talk of
“emails”. I cannot by myself cause change to norms (of
usage, of pronunciation, etc.)

One way to put this is as an instance of the dictum that you can’t
reliably control the behavior of another person, compounded to the
behavior of many other people. Neither Bill nor I can control whether
people use the phrase “controlling for”, so whether or not it
becomes or remains a norm here is “out of control” - although
nevertheless it is emergent from the control of perceptions that each of
us does by talking and writing.

So much for what I was saying. Now, on to what you were saying, which I
expect is of considerably more interest to you.

I think we use language to evoke imagery in
others that corresponds to the imagery in our own minds. But it’s often
hard to tell how well we are doing at this “evoking the same
imagery” business since, of course, we can’t perceive what another
person is imagining. Perhaps the best way to tell if there is a
correspondence between what we are imagining and what image our words
have evoked in others is to see whether the person we address can reply
with new language that evokes in ourselves the imagery we hoped to evoke
in that person.

The best test I know of understanding is paraphrase. (But this isn’t new
“language” in the above sense. That would be translation.)

Bruce Gregory did this extremely well with
“controlling” and “controlling for”. His use of the
example of a baseball pitcher to illustrate the difference between
“controlling” and “controlling for” indicates that
the mental images evoked in him by these two phrases is exactly the same
as those evoked in me.

His words evoked in you some perception which was also evoked in you by
the words of Bill Powers (2003.01.01.1232 MST) to which he was
replying:

“Controlling” is used when you are
successfully holding

something in a reference state. Controlling for a reference state

explains what you’re doing by saying what the intended outcome is,
but

without claiming success, yet. A bowler can say "I’m controlling for
a

perfect game," a golfer "I’m controlling today just for getting
all my

drives in the fairway." So “I’m controlling for X” means
"My goal is to

achieve X, and that explains my actions even when I don’t
succeed."

In terms of imagery, a bowler, a golfer, and a baseball pitcher
look rather different. What they have in common is either an
abstraction or a generalization.
I make what seems to me to be the same distinction using the words
“controlling successfully” and “controlling with
error”. An archer can say “I’m controlling my arrow hitting the
center of the target, but not very well yet. At least most of them are
landing on the target today!” or “I’m controlling my arrow
hitting the center of the target, but they’re going consistently to the
left. I must be plucking the string.” This seems to me to evoke the
same sort of perception as the above. Perhaps this is not the case for
you?
This perception can be described in words as a sequence of skilled
release, ballistic path, and approximation to an intended location, where
skill is increased by iterative practice. (Well, the path of a bowling
ball isn’t ballistic, but it’s subject only to laws of physics and
independent of any further activities of the bowler.)
Control is exercised in the momentary actions of the release (bowling
form, golf swing, baseball throw, draw and release of the bow), after
which you wait for the results. But what is relevant here is control
exercised through iterative refinement of the fine detail of the launch
and release, building skill and consistency. Does this mean that
“controlling for” is limited to iterative honing of skill, or
to activities involving such skill? If not, then perhaps some different
examples are needed to communicate all of the intended
meaning.

So in this case, Bruce G. demonstrated
excellent control of language inasmuch as he was able to communicate his
intended meaning, which precisely corresponded to the meaning I would
have been controlling for had I used the terms “controlling”
and “controlling for”.

I agree that paraphrase is a good test, one of the best we have, but it
is not absolute proof that you are having the same perception (whatever
that means).

    /Bruce

Nevin

···

At 10:54 AM 1/3/2003, Richard Marken wrote:

[From Rick Marken (2003.01.04.0900)]

Bruce Nevin (2003.03.01 19:11 EST)--

Neither Bill nor I can control whether people use the phrase
"controlling for", so whether or not it becomes or remains a norm here
is "out of control" - although nevertheless it is emergent from the
control of perceptions that each of us does by talking and writing.

I disagree. Anyone can control for "controlling for" (or any other
linguistic expression) becoming a commonly used phrase. I agree that
controlling for what other people do is not often going to be
successful. But that doesn't mean that people don't sometimes control
for it.

I make what seems to me to be the same distinction using the words
"controlling successfully" and "controlling with error". An archer can
say "I'm controlling my arrow hitting the center of the target, but
not very well yet. At least most of them are landing on the target
today!" or "I'm controlling my arrow hitting the center of the target,
but they're going consistently to the left. I must be plucking the
string." This seems to me to evoke the same sort of perception as the
above. Perhaps this is not the case for you?

Actually, no. When I say that an archer is "controlling for" hitting the
target I am not saying anything about whether or not this controlling is
being done successfully. An archer who hits the target regularly could
be controlling (successfully) for hitting the target or she could be
controlling (unsuccessfully) for hitting the outer rim. "Controlling
for" is just a way of saying that controlling is aimed at producing a
particular (reference) perception (the reference perception being the
implicit object of the prepositional phrase "controlling for"). It says
nothing (to me, anyway) about whether the controlling being done is
successful or not.

I agree that paraphrase is a good test, one of the best we have, but
it is not absolute proof that you are having the same perception
(whatever that means).

I agree. That, I believe, is why the conversation continues.

Best regards

Rick

···

--
Richard S. Marken
MindReadings.com
marken@mindreadings.com
310 474-0313

[From Bruce Nevin (2003.05.01 16:20 EST)]

Rick Marken (2003.01.04.0900)--

Anyone can control for "controlling for" (or any other
linguistic expression) becoming a commonly used phrase. I agree that
controlling for what other people do is not often going to be
successful. But that doesn't mean that people don't sometimes control
for it.

I believe Bill would say that anyone can "control for" the way people talk. I base this on the following:

Bill Powers (2003.01.02.1927 MST)--

I think the addition of "for" to "controlling" is a way of recognizing the difference between what you hope to achieve and what you actually achieve by controlling something. I find the distinction useful, in that a lot of people seem to think that controlling is an either-or proposition; either you're controlling some CV or you're not, with the perhaps unconscious assumption that if you're controlling, the error must be zero. If I say I'm controlling for three meals a day, this doesn't mean I always get them. It just says that my actions can be understood if you know I'm trying my best to keep that particular variable (number of meals per day) at that specific reference level (3).

There is an exact parallel as follows: If I say I'm controlling for PCTers using the expression "controlling for" in the way that I do, this doesn't mean that they always do so. It just means that my actions can be understood if you know I'm trying my best to keep that particular variable (PCTers using the expression "controlling for" vs. "controlling") at that specific reference level (using it in the way that I do).

The key point seems to be that "for" refers to a specific value of the reference perception:

Rick Marken (2003.01.04.0900)--

"Controlling for" is just a way of saying that controlling is aimed at producing a particular (reference) perception (the reference perception being the implicit object of the prepositional phrase "controlling for"). It says nothing (to me, anyway) about whether the controlling being done is successful or not.

It sounds like when you say
         Control x
then x is a controlled variable, a perceptual input;
and when you say
         Control for x
then x is a reference perception at a specified value.

A statement about either a controlled perceptual input or about a reference perception must implicitly or explicitly include a specification of a reference perception.

- If you have determined that the variable is being controlled, then you have determined some value(s) of the reference perception.

- If you have not inferred the reference by observing resistance to disturbance then you have not determined that the variable is controlled.

- If you know that the variable is controlled then you know that the reference approximates the observed value of the CV (precisely or approximately, depending upon loop gain).

Etc.

So presence or absence of knowledge of the value(s) of the reference perception cannot distinguish "control x" from "control for x".

Rick Marken (2003.01.04.0900)--

When I say that an archer is "controlling for" hitting the
target I am not saying anything about whether or not this controlling is
being done successfully. An archer who hits the target regularly could
be controlling (successfully) for hitting the target or she could be
controlling (unsuccessfully) for hitting the outer rim.

Sorry, non sequitur. The archer has told you that she is controlling her arrows hitting the center of the target. Having a good understanding of PCT, she is does not "think that controlling is an either-or proposition; either you're controlling some CV or you're not, with the perhaps unconscious assumption that if you're controlling, the error must be zero" (Bill Powers 2003.01.02.1927 MST). She understands that her control of her perceptual input (of the CV) can be disturbed from the value of her reference input (for the CV) by a number of factors, and in fact is alert to and controlling as many of those factors as she is able to. So unless she is lying to you, or is deluded e.g. by hypnosis, she is not aiming at the rim of the target.

"Controlling
for" is just a way of saying that controlling is aimed at producing a
particular (reference) perception (the reference perception being the
implicit object of the prepositional phrase "controlling for"). It says
nothing (to me, anyway) about whether the controlling being done is
successful or not.

So "controlling for" is just a rhetorical emphasis on the reference perception, which is explicitly or implicitly included when you say simply "controlling". There is no substantive distinction made. The rhetorical emphasis is useful for talking to people who "think that controlling is an either-or proposition; either you're controlling some CV or you're not, with the perhaps unconscious assumption that if you're controlling, the error must be zero", if it is used in circumstances where control is or can be poor, but in the technical language of PCT it makes no useful distinction. It's a synonym for "control" with some sort of implicit emphasis on the fact that there is a reference perception which is of the same kind as but might be in a different state from the observed CV.

         /Bruce Nevin

···

At 11:59 AM 1/4/2003, Richard Marken wrote:

[From Bruce Gregory (2003.01.05.1705)]

Bruce Nevin (2003.05.01 16:20 EST)

So "controlling for" is just a rhetorical emphasis on the reference
perception, which is explicitly or implicitly included when you say simply
"controlling". There is no substantive distinction made. The rhetorical
emphasis is useful for talking to people who "think that controlling is an
either-or proposition; either you're controlling some CV or you're not,
with the perhaps unconscious assumption that if you're controlling, the
error must be zero", if it is used in circumstances where control is or can
be poor, but in the technical language of PCT it makes no useful
distinction. It's a synonym for "control" with some sort of implicit
emphasis on the fact that there is a reference perception which is of the
same kind as but might be in a different state from the observed CV.

As I believe I said.

Bruce Gregory (2003.01.02.0730)

[From Rick Marken (2003.01.05.2210)]

Bruce Nevin (2003.05.01 16:20 EST)

A statement about either a controlled perceptual input or about a reference
perception must implicitly or explicitly include a specification of a
reference perception.

Not true. When I say that a thermostat is controlling temperature (a
perceptual input variable) I am not including a specification of what
temperature (70, 65, etc) is the reference perception. I am just referring to
the dimension of variation that is under control.

So presence or absence of knowledge of the value(s) of the reference
perception cannot distinguish "control x" from "control for x".

Right. What distinguishes "control x" from "control for x" is in what one
means to say. I use "control x" when I want to say that someone is controlling
the variable x, such as the position of a cursor, and I use "control for r(x)"
when I want to sat that someone is acting to keep x at a particular value,
r(x), such as keeping the cursor 1 cm to the left of a target line.

Rick Marken (2003.01.04.0900)--
>When I say that an archer is "controlling for" hitting the
>target I am not saying anything about whether or not this controlling is
>being done successfully. An archer who hits the target regularly could
>be controlling (successfully) for hitting the target or she could be
>controlling (unsuccessfully) for hitting the outer rim.

Sorry, non sequitur. The archer has told you that she is controlling her
arrows hitting the center of the target.

She did? If so, she sure has a funny way of talking. I can't really tell what
she meant. She might have meant that she is controlling the destination of the
arrows. She might also have meant that she is controlling for that destination
being the bull's eye. She might even have meant that she was controlling only
those arrows hitting the center of the target. If she meant the latter, then
what it is about those arrows that she is controlling for I don't know.

So "controlling for" is just a rhetorical emphasis on the reference
perception, which is explicitly or implicitly included when you say simply
"controlling".

Well, I guess that's a way to describe it. I would say that "controlling for"
is just a nice way of talking about what particular state of a perceptual
variable people are controlling for when they are controlling a known
perceptual variable.

The rhetorical
emphasis is useful for talking to people who "think that controlling is an
either-or proposition; either you're controlling some CV or you're not,
with the perhaps unconscious assumption that if you're controlling, the
error must be zero"

I think it's also useful for talking to people who know that controlling is
_not_ either-or but have difficulty understanding that the reference state of
a perceptual variable is secularly adjustable by the controller. People, for
example, like conventional control theorists who seem to think that they
_know_ what a person is controlling for when they know the variable the person
is controlling. For example, a conventional control theorist assumes that a
person who is controlling the position of a cursor in a manual tracking task
is controlling for the cursor being on target. In other words, this control
theorist assumes that controlling x (the position of the cursor) is the same
as controlling for r(x) (a particular position of x). In fact the person
doing the the tracking determines the value of x for which he or she is
controlling. Recognition of this simple fact -- that people themselves
determine the states of the perceptual variables for which they control-- is
one of the main factors that distinguishes conventional control theory from
PCT. In order capture this fact I think we have to be able to talk about the
fact that people can control for different state of the perceptual variables
they control.

It's a synonym for "control" with some sort of implicit
emphasis on the fact that there is a reference perception which is of the
same kind as but might be in a different state from the observed CV.

If you want to consider "control for" as a synonym for "control" that's fine
with me. But I think "control" and "control for" is a useful semantic
distinction so I think I'll stick with not treating them as synonyms.

Best regards

Rick

···

--
Richard S. Marken
MindReadings.com
marken@mindreadings.com
310 474-0313

means to say. I use “control
x” when I want to say that someone is controlling

the variable x, such as the position of a cursor, and I use “control
for r(x)”

when I want to sat that someone is acting to keep x at a particular
value,

r(x), such as keeping the cursor 1 cm to the left of a target
line.
[From Bill Powers (2003.01.06.0618 MST)]
Rick Marken (2003.01.05.2210)–
What distinguishes “control x” from “control for
x” is in what one
As far as I’m concerned, the last word belongs to you rather
than Bruce N. I think this set of definitions covers the usages I have
heard (and used) exactly. When you say you’re controlling x, x is the
name of a variable that can have many values: temperature, sweetness,
position, rate of spin, pressure, income. Controlling for x
requires x to be a specific state of a variable. Mixing the two usages
sounds strange: “I’m controlling three degrees” and “I’m
controlling for temperature;” compare with “I’m controlling
temperature,” and “I’m controlling for three degrees.” You
don’t have to control for temperature: everything is already at some
temperature and can’t be otherwise. And you can’t control three degrees
because that is a fixed temperature incapable of being either
disturbed or controlled.

Why do I doubt that this will settle the matter?

Best,

Bill P.

Convince a man against his will

He’s of the same opinion still.

[From Bruce Nevin (2003.06.01 15:45 EST)]

Rick Marken (2003.01.05.2210)–

I think “control” and “control
for” is a useful semantic

distinction so I think I’ll stick with not treating them as
synonyms.

You might be right. Time will tell.
In the meantime, no efforts by any one person to control the outcome can
determine it - the outcome being “‘control’ vs. ‘control for’ is a
useful semantic distinction”. “Useful” here refers to
communication. Whether or not something is useful for communication is
socially determined - I’m sure you recall the distinction between saying
and telling. More generally, social norms, including norms of language
usage, are reference perceptions that, as far as I can tell, are
determined (established as memories) by observation of others. They are
not determined in their status as observable social norms by any
individual’s control efforts. Hence, from an individual’s point of view,
they are ‘out of control’ in the same sense that the sunrise is ‘out of
control’.
Oddly, it is the controlling according to norms (as reference
perceptions) that reaffirms and recreates them as social realities. If
another person controls according to a different norm one often responds
by attempting to control the other person. The most common alternative to
that is to control with higher gain with respect to one’s own norm
(asserting, in effect, that I’m not that kind of person). A third
possibility is for each to control according to their own norms (Dixie
accent, Yankee accent) comfortably while interacting; but in fact the
(typical? always?) result of such amicable cooperation is a mutual
drifting of norms closer together. Why should this matter? Because we can
communicate and cooperate only when we control according to the same (or
systematically correlated) values of key reference perceptions - shared
norms. And norms are atypical in another way. Because they are reference
perceptions determined by observation of the behavioral outputs of
others, they are reference perceptions for controlling aspects of one’s
own behavioral outputs. In the canonical descriptions of control,
behavioral outputs are the means of control, they are not themselves
controlled.

I am not saying that this is all well understood. To the contrary, I am
saying that we cannot ignore it.

    /Bruce

Nevin

···

At 01:12 AM 1/6/2003, Richard Marken wrote:

[From Bruce Nevin (2003.06.01 16:06 EST)]

Bill Powers (2003.01.06.0618 MST)–

When you say you’re controlling x, x is the
name of a variable that can have many values: temperature, sweetness,
position, rate of spin, pressure, income. Controlling for x
requires x to be a specific state of a variable. Mixing the two usages
sounds strange

There is a ready alternative to mixture: when you say you’re controlling
x at reference r.

Why do I doubt that this will settle the
matter?

Probably because such things are not settled by fiat. I say this having
conceded the specific point to the extent that it is possible to, when I
replied to Rick:

Bruce Nevin (2003.06.01 15:45 EST)–

You might be right. Time will tell.

    /Bruce

N

···

At 08:29 AM 1/6/2003, Bill Powers wrote: