[Martin Taylor 2009.02.17.11.23]
This message is inspired by, but not a response to, Rick’s continuing
misreading and misquoting of my position on “conventional
psychophysical experiments.” It is a start toward indicating under what
conditions I think it legitimate to use the results obtained by
experimenters who may not recognize that “all behaviour is the control
of perception.”
This being the case, every
action is the result of the organism controlling some perception. An
external observer can observe only the inputs and the action
outputs of any living thing – I ignore for the moment
neurophysiological and related measures – and none of the values at
the terminals of the individual components of the control loop except
for the environmental feedback function. For a
living thing, there can never be a straight-through input-output
function that connects observable action to observable input, and
therefore there can never be an “actual
transfer function that exists between a subject’s sensory input and the
subject’s output action” [Rick Marken (2009.02.17.2130)]. Rick has
frequently asserted that I claim the opposite, including in the message
just referenced.
Let’s make another generic claim: “All human psychology is an attempt
to find out something about the way a mind works”, where “a mind”
represents everything that goes on between the sense organs and the
muscular or chemical output processes. Over the centuries, many
different methods of making such attempts have been used, some more
likely than others to get at what really goes on. If we believe the PCT
mantra “all behaviour is the control of perception,” then methods that
do not take this into account produce useful data only by accident. But
such accidents do happen, and that’s the thesis of this message.
The basic problem is that all behaviour is the control of
perception. An external observer can observe only the inputs and the
action
outputs of any living thing – I ignore for the moment
neurophysiological and related measures – and none of the values at
the terminals of the individual components of the control loop except
for the environmental feedback function. Nevertheless, appropriate
experimental arrangements can offer useful information about at least
some of the component pathways in a control loop, and some of those
“appropriate experimental arrangements” are used by psychologists who
are blissfully ignorant of the control nature of life – “conventional
psychologists” doing “conventional research”.
I want to start a catalogue of circumstances in which data obtained by
experimenters who ignore the PCT mantra can be used to advance PCT
science. In this message I offer just one entry to the catalogue. Such
a catalogue is open-ended, since no matter how many such
circumstances are described, more may yet be noticed. Rick has made a
start, with this section from his “Revolution” paper p2:
"Essential to the validity of this approach is that the causal path
from IV to DV be one-way or open-loop. Only if this is the case can any
observed relationship between the IV and DV be considered a reflection
of the nature of the organism transfer function."
Of course, as noted above and many times in the “PCT Research and
Statistics” thread, there can never be such a thing as “the organism
transfer function” for a living thing, but Rick’s point remains valid
when applied to component transfer functions. Components with transfer
functions can and must exist in any control loop, and may exist even
when not connected in any control loop. Among the properties of mind
studied by “conventional psychologists” are such component transfer
functions, and that’s where the catalogue will start.
Case 1: E disturbs one input of a two input perceptual function p(S1,
S2), and allows the subject to influence the other input when
controlling. E observes the output O that influences the controlled
perception P through its effect on S2. In what follows, we label the
experimenters presentation (the disturbance) D.
Let’s consider the equations involved with this arrangement, using the
usual symbols D for disturbance, O for output, E for error, P for
perceptions and in this case S1 and S2 for the inputs to the perceptual
function, at whatever hierarchic level that function may be. Without
loss
of generality, we can set S2 = f(O), since we assume that the
experimenter introduces no disturbances on this pathway. The question
is whether
we can find a functional relationship between O and S1 – a
relationship between the disturbance E introduces at S1 and the
correction the subject introduces by influencing S2 in the course of
controlling P. Call the value of
S1 at time 1 S1.1, and at time 2 S1.2.
Because this is a control loop for which we assume an
invariant reference value set by the experimenter, if control is
perfect, then
p(S1.0, S2.0) = p(S1.1, S2.1), or p(S1.0, S2.0) - p(S1.1, S2.1) = 0.
An equation of the form f(a,b) - f(c,d) = 0 is a parametric
representation of f in the space of
its two arguments. If the values of a, b, and c are prescribed, then
for most
continuous functions, d can take on only a finite number of possible
values (it’s infinite if for that value of c, f does not depend on d at
least over some range of d). Since we usually consider only monotonic
functions, if you specify any three of a, b,
c, and d, the fourth can take on only one value.
S2 is a function of O, which the experimenter can observe. so we can
write p = p(S1, f(O)), and the expression above becomes a parametric
representation of a function relating S1 to O. We may not know what
that function is, but if S2 is a monotonic function of O, then O is a
monotonic function of S1. A change in S1 that would increase the value
of P is countered by a
change in O in the direction that would alter S2 so as to reduce P.
Colloquially, S1 and S2 are at the two ends of a see-saw.
What we are interested in is not really the transfer function between
S1 and O, but the one between the
disturbance D introduced by the experimenter and S1. We want to know
whether we can deduce anything about this transfer function by
observing the relationship between variations in D and variations in O.
In conventional psychology, D is the independent variable (IV) and O
the dependant variable (DV).
So far, we have shown that O is a function of S1. If S1 is a function
of D, then O is also a function of D, which is part of what we wanted
to know.
Knowing that O is a function of D, what can we say about S1 as a
function of D? That depends on what we can determine about the control
loop. The form of the function relating S2 to S1 is controlled by the
form of p(S1, S2). For example, if p == (S1 * S2), the answer would be
quite different than it would if p == (S1-S2). Under some
circumstances, we may be able to assume the form of p, or
test it by modelling the control loop. If we can determine the form of
the function that relates O to S1, and the function that relates O to
D, then we can determine the function D -> S1, which is what we
wanted to know.
Let’s consider the case in which p = (S1 - S2). This is the case in
which the subject is controlling the match between two variables, such
as the changing location of a marker and the location of a cursor, or
the brightness of the area to the left of a dividing line by the
experimneter and the
brightness of the area to the right controlled by the subject – or
reported by the subject as “brighter” or “darker”. In this case, if
control is
perfect, then deltaS1 = -deltaS2, and any remaining questions relate to
the function S2 = f(O). It’s the pursuit tracking problem. We may never
know exactly what p is, or what function relates the disturbance to S1,
but we do learn how the subject manipulates the environment in order to
make S2 match S1, and if we can determine the form of p by other
experiments (such as by allowing the subject to influence both S1 and
S2, for example, and modelling the control loop), then we could
determine the function D -> S1.
At another perceptual level, consider the case in which S1 and S2 are
category perceptions, and the reference is to make S2 (the answer)
match S1 (the presentation). In the diagram, there would usually be an
“imagination” connection in addition to the external connection from O
to S2, but this doesn’t affect the case. This is the situation that has
been the focus of the “PCT Research and Statistics” thread. In this
situation, we are rather better off, since the answer is reported
categorically, by a button push, a verbal statement, or something like
that. Assuming the subject is controlling for a match, we can determine
what characteristics of the disturbance correspond to what categories
or S1 – in other words we can determine the transfer function D ->
S1 as far as determining what values of D lead to what perceived
category, but not necessarily within the category. This is the signal
detection or discrimination situation, in which S1 is the subject’s
perception that the signal exists or is of type X. The subject’s
proportion of “correct” and “wrong” responses then can be used to
determine the information capacity of the pathway between disturbance
(presentation) and interpretation (creation of perception S1). A
“correct” response occurs when O represents a category different from
the category the experimenter intended to present as D.
The experiment by Schouten that I posted in [Martin Taylor
2009.02.17.17.35] is a more subtle case of the same situation. In that
experiment, the subject is still reporting the answer S2 that matches
the perceived category of S1 (the left or the right light was lit). For
any given timing of the answer relative to the presentation, the
information conveyed through the presentation-interpretation pathway
(disturbance -> S1) can be determined from the proportion of correct
and wrong responses, and by finding that proportion for different
timings, the channel capacity or information rate of that pathway can
be determined.
Case 1 is really quite general in its application. It depends on
setting up a situation in which the subject controls a perception
created by a function with two inputs, one of which is influenced only
by the experimenter, the other of which is influenced only by the
subject. The subject can control the perception only by using the
variable he controls to counter the experimenter’s disturbance to the
two-input perceptual function. This case covers standard signal
detection and discerimination experiments, as well as pursuit tracking
experiments in which the interest is usually, but need not be, on the
properties of the control loop itself. It could even cover things such
as Rohrsach tests, even though in that case the experimenter has no
“right” and “wrong” assessment of the subject’s responses.
Martin