Beyond Determinism and Indignity:
A
Reinterpretation of Operant Conditioning
(published in Behavior and Philosophy, Spring 1994)
W. Teed Rockwell
2419A Tenth St
Berkeley CA 94710
510/548 8779 Fax 510/548 3326
In
his most well-known philosophical work, Beyond Freedom and Dignity, B.F.
Skinner implied that psychological
research had somehow proven that freedom of choice was an obsolete concept;
that a scientifically ordered world would
necessarily be a determined world. Skinner claimed that freedom only
appears to be possible because we have clung to concepts derived from primitive physics, which conceived of
inanimate objects and persons in terms similar to the ones we use only for
persons today.
. . . Aristotle argued that a falling body
accelerated because it grew more jubilant as it found itself nearer home. All
of this was eventually abandoned, but . .
. Almost everyone who is
concerned with human affairs. . . continues to talk about human behavior in
this prescientific way. ( Skinner 1971. pp. 8-9)
We
now know, thanks to Newton and his colleagues, that these kinds of
descriptions, which presuppose something very like free will and purpose, are not correct for stones. Skinner claimed that if we were
consistent we would no longer apply these assumptions to living organisms either. From this assumption he concluded that
those people who think of themselves as free are really controlled by unknown
and disorganized contingencies, and they would lose nothing genuine if their
lives were controlled for the good of society. He consequently claimed that to
let human beings remain "free" has condemned society to chaos, and to take away that freedom would be
to deny individuals nothing that actually existed . This conclusion made many
people desperate to prove Skinner wrong, and fearful that he might be right.
What
seemed important about the appearance of Beyond Freedom and Dignity was
the sense of a major change in our world view and values, which seemed to
parallel what had happened with Copernicus and Darwin. In each of these previous
cases, important scientists revealed proven facts which had philosophical
implications that profoundly shrunk the difference between us and the rest of
reality . Copernicus revealed that we were not at the center of the universe, but only scuttling along
the surface of a spec of dust in an obscure corner of an obscure galaxy. Darwin
revealed that we were really nothing but very smart monkeys. Skinner appears to
have taken this progression one step further and said that we (and monkeys) are
really only machines, and that the only way to understand us is with the same
scientific laws that govern machines.
These rhetorical descriptions are overstated, of course. But
they do capture the sense of fear which gave Beyond Freedom and Dignity an aura of historical urgency when it
was first published. For anyone who believes that science reveals the truth, it
meant that one needed the philosophical courage to accept yet another
degradation of the human condition. Skinner was certainly one of the most important
psychologists of his generation, and if he said there was no free will, it had
to be taken seriously. Philosophers say depressing things like this all the
time, but seeing as they never agree with each other, there seems to be no real
urgency in this fact. But when a
scientist says something of this sort, you look like a plaintiff in the Scopes
Monkey trial if you don't agree with him.
However, if Skinner is to be considered to be another
example of this "scientist revealing the truth of the human condition" archetype, his claims have to be
evaluated only on the implications of the facts he discovered. Skinner's philosophical arguments date back to Hobbes and Hume, and his presentation of them is
in fact, not up to professional standards philosophically (see "Skinner Skinned" in Dennett
1978 ) The real question is: what scientific
evidence does Skinner have to
offer that would cause us to doubt the existence of free will? The answer is:
nothing whatsoever.
The
force of Skinner's argument rests
on the claim that what he and his
colleagues have discovered in the laboratory proves that there are no
significant differences between the laws that govern stones and the laws that
govern living organisms. However, a closer study of what Skinner had to say about
his own laboratory procedures (especially in the book Contingencies of
Reinforcement, written just before Beyond
Freedom and Dignity) reveals
something quite surprising. Most
of what Skinner achieved as a laboratory scientist resulted from his discovery
of an essential difference between
the laws of physics and the behavior of organisms, and this difference is just
the sort that would exist if there were such a thing as free will.
Almost
all of the important discoveries of physics and chemistry are expressed in
mathematical formulae, which make
it possible to predict exactly how
one quantity will vary when another quantity is varied. For example the formula
P= E x I/R tells us that if we increase a given voltage E from 3 volts to 5
volts, then the power P will
increase by a factor of exactly
5/3. Formulae of this sort are the strongest weapons in the arsenal of
Newtonian science, for they establish precise quantitative relationships
between qualities as diverse as heat and pressure, magnetism and electricity,
wavelength and frequency.
Many
psychologists, both before and after Skinner, understandably tried to find
similar mathematical formula governing the behavior of living organisms, in
order to give themselves the same ability to predict and control that is
possessed by the physical sciences.
But more than a century of careful psychological study failed to find these laws, and Skinner
made his most important contributions to laboratory research by accepting this
fact and finding ways of working around it.
Skinner assumed that the reason
Newtonian style laws could not be found in the psychology laboratory was epistemological
i.e. that the behavior of organisms
was so complex and difficult to trace that these laws could not be
discovered. It never occurred to
him to consider the possibility that this difference between the animate and
inanimate might be ontological i.e.
that the reason psychologists have found no mathematical formulae governing the
behavior of organisms might be that there aren't any such formulae. The fact is
that the existence of deterministic laws
governing the behavior of organisms remains an assumption, for which
psychological research provides no inductive evidence . The indeterminacy that psychologists
encounter in the laboratory could be an intrinsic aspect of living organisms,
and not just a reflection of our ignorance.
If
we were to take Skinner's research
at face value, It can be seen to
offer a very high level of replicability for data supporting the claim that
human behavior, unlike the behavior of inorganic matter, is not causally
determined by Newtonian laws, (and an even higher level of replicability for
data supporting the claim that the behavior of rats and pigeons is not causally
determined by Newtonian Laws). It
thus appears that the evidence which Skinner offers for determinism could be
the strongest body of evidence we have against determinism. Of course, evidence against determinism is not
the same thing as evidence for free will. But if there is an intrinsic
indeterminacy in the behavior of organisms that is not present in the behavior
of stones, it would mean that Skinner's collapsing of this distinction between
them would be less accurate than the common sense view which sees human beings
as free and stones as determined.
* * *
The
person who conceived of the goal of psychology as a quantified science most
vividly was Clark Hull who, in
such works as Mathematico-Deductive Theory of Rote-Learning and Principles
of Behavior, formulated mathematical functions
which attempted to do for psychology what Newton did for physics. There is no question that his basic
vision inspired and influenced many psychologists. Unfortunately, none of his formulae were actually capable of
predicting what an organism would do in a laboratory with anywhere near the
accuracy of physics equations. Hilgard's description of Hull as "giving a
foretaste of what psychology will be like when it reaches systematic
quantitative precision" (Hilgard 1940) still expresses the psychological
community's evaluation of his work. Everyone admires and shares
his vision of a quantitative, scientific psychology, but almost no one now
believes that he succeeded in becoming "the Newton of Behavior".
There
were, however, other aspects of the scientific method which other psychologists applied to their
subject matter with great success. Nobel prize winner Ivan Pavlov made precise
quantitative measurements, but
contented himself with predictions
that were only qualitative. For example, he was able to show that a
conditioned reflex could be created between the ringing of a bell and the
salivating of a dog. But he did not have a formula which enable him to predict
exactly how many grams of saliva would be produced per bell ring, or per
decibel level of the bell ring or any other quantifiable factor whatsoever.
This is why his Lectures on Conditioned Reflexes contains
many pages of accurate measurements, and
descriptions of his innovative and careful laboratory techniques, but no
mathematical functions with predictive pretensions. Nevertheless, he was able
to repeatedly produce certain effects on demand, to predict with previously
unequalled accuracy how they would fade , and to show how organisms could
discriminate between various kinds of stimuli. To be able to make vebalizable
predictions was obviously a tremendous step forward, so it was easy to postpone
the sense of urgency for mathematical predictions.
It
seemed at first that categories similar to those of physics and chemistry might
eventually be discovered by Pavlov's school of psychology. It was natural to assume that if
psychology developed a more and more precise taxonomy of stimuli and responses,
it would eventually be possible to classify organisms and/or S-R connections
into a system of natural kinds which could make the leap to the quantifiable
predictions of natural science. After all, chemistry has classified inanimate
matter into natural kinds, called elements, and once one has discovered which
of those elements are present in a chunk of stuff, one can predict with
replicable accuracy exactly what that chunk of stuff will do when you do
something to it. Why shouldn't psychologists eventually be able to discover
similar natural kinds if they were careful with their laboratory observations?
Instead,
Pavlov became at least somewhat aware that each organism he worked with was an
individual, not just an exemplification of a type, and had to be treated as
such if its behavior was to be predicted and controlled. Skinner spoke admiringly of Pavlov's
awareness that he was studying individuals, not members of a species.
Animal
psychology at that time was primarily concerned with the behavior of the
average rat. The learning curves
which appeared in textbooks were generated by large groups of organisms. Pavlov
was talking about the behavior of one organism at a time. (Skinner 1972, p.
594)
F.A.
Volgyesi, a student of Pavlov's, also remarked that the dogs Pavlov was testing
could only be understood if they were considered as individuals, with
individual characteristics.
One
animal is best suited for experiments with inhibitions, sleep, or hypnosis;
another is better suited for experiments related to arousal...Individual
differences of the nervous types are so subtle...there were as many differences
as we experience...with humans. (Volgyesi p.56-57)
To
be a laboratory scientist in the Pavlovian tradition thus required sensitivity
and flexibility, because the rules
that governed their laboratory procedures were not quantifiable
formulae, but intuitive rules of thumb.
It was extraordinarily lucky that he [Pavlov] began
with the salivary reflex. There seems to be no other response quite so simple.
Other glandular secretions, for example tears or sweat, are by no means so easy
to control, and we have heard today something of the enormous complexity of
conditioned cardiac responses.(Skinner 1972, p.595)
It is thus not
surprising that when Skinner extended Pavlov's work in new directions, he began
to stress that the rules sought by psychologists should not be theoretical
structures built with mathematical formulae, but rather verbalizable
generalizations that strayed as little as possible from the particular
individuals observed in the laboratory. In his first major work, The Behavior of Organisms, there are several laws of behavior stated, but not
one of them is a mathematical law. For example, the law of the magnitude of the
response says " the magnitude of the response is a function of the
intensity of the stimulus" (Skinner 1938 p. 13). But this cannot mean a
function in the strictly mathematical sense of the term, because it does not
show us how to use a known stimulus intensity to compute an unknown response
magnitude, or vice versa, (unlike physical laws, with which it is possible, for
example, to set up exact ratios between voltage and current in an electrical
system, or heat and pressure in a gas). Occasionally he illustrates these laws
with letters concatenated by operator-like symbols. But these are meant to be only heuristic devices,
because Skinner never claims that these letters can be translated into
quantifiable units.
Because
he believed that psychology must rely completely on the data, rather than on
theory and formulae, Skinner was willing to be agnostic about some of the
ontological presuppositions of deterministic physics. He admitted that the
terms "cause" and "effect" were not strictly applicable to
his findings. For his purposes, a
cause was merely a change in an independent variable, and an effect was a
change in a dependent variable. In other words, the cause did not necessarily
boot the effect in the seat of the pants or otherwise compel it to happen. The
two were merely linked by the more metaphysically neutral term "functional
relation". (Skinner 1953 P. 23). Consequently, Skinner was able to stand
the assumption of cause and effect underlying Pavlovian conditioning on its
head with his greatest contribution to psychology: the concept of operant conditioning.
Skinner
realized that although the reflex described by Pavlovian conditioning explained a great deal of
human and animal behavior, (he wrote his doctoral thesis on the subject), it
could not explain all of it.
It is neither plausible nor expedient to conceive of
the organism as a complicated jack-in-the-box with a long list of tricks, each
of which may be evoked by pressing the proper button ( 1953 p.49)
The
jack-in-the-box analogy does effectively describe the salivation that Pavlov
produced by ringing a bell, because the bell comes first, and thus
"causes" the salivation. Sometimes, however a stimulus can occur after a bit of behavior, and cause a similar kind of
behavior to occur again. For example, if a pigeon pecks a lever, and receives
food after doing so, he is more
likely to peck the lever again. This kind of stimulus was called a
reinforcement, and the conditioning that used reinforcements was called operant
conditioning. The principle that behavior could be shaped by reinforcement made
it possible to control kinds of behavior that were far beyond the reach of
Pavlovian conditioning. Not only could one control glandular secretions and
muscle twitches, it was now also possible to control what Skinner referred to
as "what an organism is doing." (Skinner 1938 p. 6). Operant
conditioning was so called because it controlled the operations performed by an organism in its attempts to survive
in the world, and this class of behavior seemed to be capable of including
almost any kind of behavior an organism could perform. Skinner was able to
train rats and pigeons to perform extremely complex forms of behavior (for
example, the controlling of guidance systems for missiles), and many of the
techniques he developed in the laboratory worked quite well on people as well.
But
this new-found control was bought at considerable price. The relationship
between stimulus and response in
operant conditioning was much harder to predict, and psychologists begin to
admit, at least implicitly, that they were now studying only probabilities, not
certainties. Skinner acknowledged this difference by making a distinction between eliciting a response (in Pavlovian conditioning), and setting
the occasion for the response (in
operant conditioning). One cash-value difference between these two terms was
that occasion-setting merely increased the probability of the response, it did
not guarantee that the response would occur. Consequently, Skinner usually
referred to the fundamental datum that he was measuring as probability of
response (1953, p. 65 ; 1969 p. 75,
80, 91, 117 ; 1984, p. 518), and in his (1984) he devotes quite a bit of space
to insisting that any attempt to describe these probabilities in terms that
posit the existence of actual entities is unjustifiably speculative.
How
then is it possible to have so much control over one's experimental subjects
when all of one's knowledge is only probabilistic? Skinner gives many maxims
that explain this. Apparently it helps a great deal to work with a small number
of subjects and get to know them really well.
The
number of organisms is usually much smaller than in statistical designs, but
the length of time during which any one organism is observed is greater. (Skinner 1969 p.81)
Operant
methods make their own use of Grand Numbers; instead of studying a thousand
rats for one hour each, or a hundred rats for ten hours each, the investigator
is likely to study one rat for a thousand hours. (ibid p. 112)
Like
Pavlov, Skinner believed that the behavior of organisms cannot be predicted
simply by classifying the organisms into natural kinds. In Methods and Theories in the
Experimental Analysis of Behavior, Skinner admitted that most
“satisfactorily smooth” rates occur when one is measuring an
individual subject, and that when one tries to analyze the responses of
populations of different organisms, “such a statistical datum raises more
problems than it solves” (p. 518).
But
more importantly, probabilistic knowledge is not as great a handicap when one
is more interested in controlling
behavior than in predicting or explaining it.
Relations among dependent and independent variables are
seldom explored according to a prior "experimental design" . . . The
contingencies depend in part on the behavior which the organism brings to the experiment.
Provision is usually made for changing the apparatus as the organism
changes, but seldom according to a predetermined plan. (italics
mine) (Skinner 1969 p. 81)
In
other words, when you don't know exactly what an organism is going to do, you need
to observe it carefully and adapt your behavior to it. (This is
what Skinner means by the sentence in italics.) Once you have produced the
behavior you want in an individual, you
have not thereby created an algorithm that can be applied mechanically to
all future subjects according to a "predetermined plan". The next
subject has to be granted the "dignity" of being considered an
individual, whose behavior can be shaped only by reinforcements tailor-made for
it. Skinner usually does not stress the fact that a behavioral psychologist
needs this kind of flexibility, and that
a physicist does not. But his frequent criticism of "the
hypothetical-deductive method" in psychology {which he admits works fine
for physics (1969 p. ix)} shows that he has learned from his years in the
laboratory that no psychological theory is guaranteed to predict exactly what
the individual subject is going to do. Better then to observe as carefully as
possible, and use theories as rough guides rather than as precise predictors.
When
control of the organism is your goal, "It is most efficient to explore
relevant variables by manipulating them in an improvised and rapidly changing
design" (1969 p. 112). Operant conditioning experiments can discover many
of the variables that reinforce behavior, and the knowledge of these variables
is what gives operant techniques their power. But discovering what variables to
recombine in each case requires repeated trials (or an occasional stroke of
good luck), and can thus never be reduced to a precise quantifiable law. It is
tempting to speculate that, like the "improvised and rapidly changing
design" created by a good jazz musician, a great deal of the success of
operant conditioning depends upon the intuitive artistry of the individual
researcher.
This
is not to say that Skinner's cautious anti-theoretical methods are universally
accepted. The Hullian dream of a quantifiable psychology is now strived for by
means of statistics. There are a number of behavioral laws that have been
recently discovered by combining Skinner-inspired laboratory techniques with
statistics, (for example, see Williams in Atkinson 1988). These laws can be
expressed using mathematical equations, and they fit the data about as well as
the laws used in some areas of the physical sciences. But this fit is obtained
only by averaging out variations that occur in individual cases. No predictions
based on statistical analysis are binding on any single member of the population, even though they can be
highly accurate about the population as a whole. Skinner himself was aware of
this, which is why he frequently felt the need to de-emphasize statistics.
Statistical
techniques cannot eliminate. . . individuality; they can only obscure and
falsify it. An averaged curve seldom represents any of the cases contributing
to it. (1969 pp. 111)
Nevertheless,
it is frequently useful to predict the behavior of large populations, and so
the demand for statistics in psychology remains high. Statistical studies of
different kinds of populations will probably always be used, not only in the
laboratory, but also with some degree of effectiveness for a variety of
business reasons, such as insurance actuarial tables and market research.
However,
for the purposes of this paper, the differences between statistical and
Skinnerian laboratory psychology are not as important as their similarities
. Both styles of psychology agree
that the fundamental datum of their discipline is what Skinner calls probability
of response. The main difference is
that statistics make it easier to measure this probability, and operant conditioning makes it easier to control it. Despite the apparent controversy over the two
methods, most behavioral psychologists use both statistics and operant
conditioning in varying amounts. And this means that while physicists are usually studying what will actually
happen, psychologists are usually
studying what will probably happen. When you put a piece of sulfur in a
test tube and heat it to a certain temperature, there is a prior experimental
design written in chemistry textbooks that will tell you what that piece of
sulphur will do, and this design holds (to several decimal places) for every
other piece of sulphur of comparable size, weight and purity. There are some
variations that do occur when an
experiment is replicated, but these are small enough that they can plausibly be
assumed to be laboratory errors, and not variations in the laws themselves.
Hull's assumption that this would be equally true for organisms was highly plausible,
and Skinner's discovery that Hull was mistaken is an important and profound
one.
This
does not mean that nothing can be said with certainty about living organisms.
But it does seem to indicate that there is a kind of hierarchy, in which
certainty decreases as we travel upward to the "higher functions".[1]
All
matter, living and dead, is governed by the laws of physics at all times. If
you drop a rat, a pigeon or a person from a tenth story window, their falls
will be governed by the same mathematical laws that predict the fall of a rock
or log of comparable weight, wind resistance etc. But there are also laws that
are unique to living organisms, and these seem to become less predictable as we
work our way up to those behaviors which are distinctively psychological. The
conditioned reflexes elicited by Pavlovian conditioning are not as predictable
as the cause-and-effect connections discovered by physics. And the operant
conditioning reinforced by Skinnerian techniques is even less predictable, and
consequently it can be described only as a functional relation that measures or
controls probabilities.
Why
then is inanimate matter so much more predictable in the laboratory than living
organisms? Skinner offers what seems to be the most plausible explanation.
The complex system we call an organism has an elaborate
and largely unknown history which
endows it with a certain individuality. No two organisms embark upon an
experiment in precisely the same condition. (It is characteristic of most
contingencies that they are not precisely controlled, and in any case they are
effective only in combination with the behavior that the organism brings to the
environment.) (1969 pp. 111-112)
In
other words, Skinner believes that the unpredictability of the individual
organism is a reflection of our ignorance, and is not a function of any
intrinsic freedom possessed by that individual. According to this view, the
entire universe, organic and inorganic, is equally determined by the same sort
of laws. The apparent freedom
possessed by living organisms is assumed to be an illusion produced by the
complexity and untraceability of the laws that determine their behavior. This
is probably the explanation that most practicing laboratory psychologists would
give if they were asked.
Although
this position is the explanation that comes most readily to mind, this does not
mean that it is the explanation that is the least speculative or the closest to
the facts. Metaphysical speculations which are firmly buried in the warp and
woof of shared opinion can easily masquerade as self-evident, whereas newly
formulated theories can encounter resistance even when they are less
speculative than the entrenched theories they are designed to replace. It is
impossible to disprove the claim that the behavior of organisms is completely
determined by Newtonian style laws, none of which have been discovered by over
a century of psychological research. There may be no smoke only because the
fire is very carefully hidden. My
only point is that this claim is a speculation, and that Skinner has offered us
no new evidence that it is true. On the contrary, Skinner's research seems to
indicate that, so far at least, the data we have discovered underscores the
difference between the laws governing organisms and the laws governing matter.
These differences are not proof
that organisms are ungoverned by
Newtonian laws, but I think they can be said to qualify as evidence for that claim. When Sir Edmond Hilary spent a week in the Himalayas looking for a Yeti
and did not find any, this did not prove (as he and many others claimed) that
there are no Yetis. But if he and thousands of others continued to look to the
best of their ability and still found nothing, one could say that each
unsuccessful expedition could be seen as a gradually accumulating body of
evidence that weakened the probability of their being any Yetis. Believers in
Yetis could always cite the size and inaccessibility of the Himalayas and the
cleverness of the Yetis as explanations for why they have not been found, but
there is no denying that each unsuccessful expedition (assuming it was
competently and sincerely conducted) counts as partial evidence against the
existence of Yetis. The number of
researchers trying to find Newtonian style laws for human behavior vastly
outnumbers those searching for Yetis, and these unsuccessful searches do count
as partial evidence for the non-existence of such laws.
Is
the deterministic explanation accepted by Skinner preferable because of Occam's razor? After all, we already
have the laws of physics, which are deterministic. Wouldn't it be unjustifiably
speculative to posit another set of laws different in kind from the ones we
already have? This objection is unfounded, however, because many branches of
physics now rely heavily on statistics and probabilities, and accept the fact
that their data cannot in principle be understood any other way. Quantum
statistical thermodynamics, which relates heat to changes in the motion of the
atoms and molecules of matter is described as follows by the Encyclopedia
Britannica.
...at a given time, the most that can be said about results of measurements on a
system is the probability of finding particular values of properties, such as
coordinates and momenta, rather than the particular values that will be
observed....the principle of indeterminacy... requires a description of the
state of a system in terms of probabilities. (italics mine)(v 18 p.311)
The
behavior of an individual atom of decaying radioactive matter is equally
indeterminate and is thus predictable only in terms of probability.
We
have, let us say, half a million atoms of radium in this room...the new laws
merely tell us that one of these atoms is destined to disintegrate today,
another tomorrow and so on. No
amount of calculation will tell us which atoms will do this; we must rather
picture fate as picking out her atom, by methods undiscoverable by us. (Jeans
p.150).
These
facts, and others like them, show that the objectivity of statistical phenomena
is accepted in post-Newtonian physics, and that there is no reason to
automatically assume that probabilities are only a function of our ignorance
when they appear in scientific data.
Similarly, the behavior of the population of a city can be predicted
using statistical methods, but the behavior of each person in the city cannot,
and there is no reason to assume that the unpredictability of each of those
persons is an illusion. Modern physics accepts the fact that reality can be
intrinsically statistical and probabilistic.
Although
individual atoms of radium may possess this kind of indeterminacy, the rate of decay of the piece of
radium in front of us can be predicted with unassailable replicability.
Consequently, it is usually assumed that these various probabilities are
present primarily at the submicroscopic level, and average themselves out in
such a way as to make causal laws possible at the macroscopic level. When science is studying medium-sized perceptible
objects traveling at medium speeds, Newtonian physics is usually adequate to
the task. Its weaknesses only became obvious when sophisticated equipment made
it possible to measure the movements of sub-atomic particles, and objects that
could travel near the speed of light.
Because living organisms are medium sized objects, there was no reason
to assume that the sub-atomic indeterminacies of quantum mechanics would be
relevant to understanding them. Consequently, behaviorist psychologists have
usually thought in Newtonian terms, and considered the ambiguities of quantum mechanics to be outside of their
domain. Skinner did borrow the term "functional relation" from
post-Newtonian physics, but he did this only to capitalize on the principle
that a contingency of reinforcement could
functionally determine a response even though it comes after the
response. There is no indication that Skinner ever considered the possibility
that the connection between behavior and reinforcement might therefore be
indeterminate.
However, now that thousands
of behavioral experiments have
been run, the entire cumulative record appears to indicate that living
organisms, like subatomic particles, are individually indeterminate, and
predictable only statistically in large populations. If we take these data at face
value, rather than assuming that there is a Newtonian structure hidden in it
somewhere, we could conclude that Skinner might be wrong in saying that our
behavior is determined in the classical sense.
However, as I said earlier, evidence
against determinism is not the same as evidence for free will. Indeterminacy is necessary for the
existence of free will, but it is not sufficient by itself to prove its
existence. One of the points I am making is that the triumph of Pavlovian and
Skinnerian techniques, with their strong emphasis on observation of the
individual, and the eclipse of the Hullian dream of absolute quantifiable
predictability, is another piece of strong inductive evidence for the
indeterminacy of living systems. We cannot infer from this that free will, as
common sense and/or philosophy understand the term, has thereby been
vindicated. If modern psychology
does come up with a theory that explains the indeterminacy of the behavior of
organisms, there is no reason to assume that it would resemble the common sense
or philosophical definitions of "free will", any more than the
physicist’s definition of “energy” resembles the common sense
concept of that word.
I
do think, however, that the points I have raised require a revaluation of Skinner's confident
assertion that we must simply replace the folk-psychological idea of freedom
with the Newtonian idea of determinism. Skinner liked to stress the
similarities between psychology and physics, and to assume that the differences
between the two stem from the fact that psychology is a very young science. If
we assume that some of these differences are inherent in living organisms,
however, it implies that Skinner has ignored a significant distinction. We now
know that there are two different kinds of causes: those which always produce
the same result under identical
conditions (call those Newtonian causes), and those statisical post-Newtonian
laws which only "incline without necessitating" (to use Leibniz's
helpful expression—see Kane p. 57).If the behavior of living systems is,
like the behavior of a decaying radium atom, genuinely indeterminate, this would mean that operant
conditioning only inclines without necessitating, (i.e. it increases the probability
of reinforced behavior occurring, but does not guarantee that the behavior will
occur).
This is obviously different from Newtonian causation
in a significant way, but can we assume that this difference is equivalent to
our concept of free choice? The subatomic particles of quantum physics
certainly do not have free choice as the term is ordinarily used and I have been told in conversation
with physicists that there are some inanimate macroscopic phenomena which are
also indeterminate. (such as the motion of a BB balanced on a razorblade.)
Before we can prove that there is such a thing as free will, we need a
definition that will distinguish free choice from ordinary randomicity. However, I don't have such a
definition, so I don't intend to claim that Skinner or anyone else offers
evidence for or against free will in the technical sense of that term. I am
only claiming that if operant conditioning only inclines but does not
necessitate, it is not incompatible with the concept of choice. This concept, although an essential part of the
concept of free will, is distinct from it, because it does not require the
metaphysical baggage that free will has attached to it. I am not claiming that
I or anyone else can define or prove the existence of choice. But I am saying
that neither Skinner's philosophy arguments nor his scientific data can
disprove the existence of choice once we acknowledge the possibility of
indeterminacy in human behavior.
Many
of the characteristics traditionally attributed to "free will" are
fraught with mystery and/or confusion.
Plato saw the mind as being able to somehow rise above the dictates of
desire through self control,
without acknowledging that the love of the good which propels
self-control must also be a desire (see Hocutt 1990). Kant had to posit a
noumenal self, transcending space, time, and knowledge, in order to have free
will. In Beyond Freedom and Dignity, Skinner rightly points out that this
ideal of freedom as an attribute of Autonomous Man, who floats untouched over
the world of causes, is primarily a cop-out to protect humans from the
embarrassment of scientific study. His argument against free will, derived
largely from Hume's An Inquiry Concerning Human Understanding, is based on the definition of a free action as an
uncaused action, and the fact that it
makes no sense to claim that free agents do things without cause. If
this is the only possible definition of free will, than freedom is not only
indefensible, it is not even worth defending. As Hume pointed out, this
definition of free will only leaves open the possibility that humans could
occasionally go completely bananas, and do things that are totally chaotic and
irrational.2
However,
we could still have something describable as choice, even if we are not
"autonomous", when there are aspects of our life where no single
cause can compel us to obey it. What this kind of indeterminacy gives us is the
possibility that sometimes when we experience ourselves as choosing a course of
action, we could have done otherwise than what we did. This means that during
the act of deliberation, the outcome is genuinely in doubt until the decision
is made. Certain causes can compel
us: If you prick us, we bleed, if you tickle us, we laugh, and if you poison
us, we die. But there are other
causes which apparently do not force us into this kind of cul-de-sac, which is
why we can respond to them in more than one way. The fact that operant
conditioning only increases the probability of response makes it likely that it
is the second sort of cause, and therefore it does not produce what Kane refers
to as "Covert Non-Constraining Control.(CNC) (1985 p.35) Kane argues, I
think correctly, that it was Skinner's belief in the possibility of CNC that
made him deny the existence of freedom.
If indeterminacy is real factor in human behavior, there would no longer
be any conflict between the existence of choice and the effectiveness of
operant conditioning. Operant conditioning eliminates the possibility of choice
only if an operant stimulus has only one possible response. As noone has ever
found an operant stimulus that produces its response with certainty, the
concept of choice remains unthreatened by the discovery of operant conditioning
Skinner's
research shows us that an organism's
behavior is never aloof from causal explanations, because all behavior
is shaped by reinforcing contingencies
in the environment. This means that there always is a reason for everything it
does. This need not imply, however, that there is only one thing that the organism could have done in that
circumstance. It only implies that if the organism had done something else,
there would have been a reason for that
as well (or instead). For example, if I am dieting I have the choice of
whether to eat chocolate cake or an apple for dessert. If I eat the apple,
there is a reason for that which is explicable in purely behavioristic terms (I
have been positively reinforced by hearing praise of attractive thin people,
and negatively reinforced by hearing ridicule of unattractive fat people). If I
eat the chocolate cake there would also be an explanation for that in
behavioristic terms. (I was positively reinforced by past occasions when I ate
and enjoyed chocolate cake, and negatively reinforced by past occasions of
being unsatisfied by having apples for dessert.). There is consequently no need
to posit that miraculous entity which Skinner calls Autonomous Man.
Nevertheless, these stimuli do not force me to take a single course of action,
because my response to each of them is only probable, and not certain.
Skinner,
however, was not aware of any of
this. He clearly maintained that
the effectiveness of operant conditioning did prove that there was only one
possible future for each of us, and therefore our belief that internal
deliberations could change our behavior was an illusion. For example, he criticized many Freudians for being
closet free-will advocates because they “have no hesitation in assuring
their patients that they are free to choose among different courses of action
and are in the long run the architects of their own fate.”(p. 20-21).
However, if operant conditioning only inclines but does not necessitate, then
Skinner's research offers no evidence that contradicts this possibility. He
does not realize this , however, because he assumes that freedom always implies the existence
of Autonomous Man, who does things
for no reason whatsoever. Skinner quotes Joseph Wood Krutch as stating a
position very similar to the point I am making here.
We can predict with a considerable degree of accuracy
how many people will go to the seashore on a day when the temperature reaches a
certain point, even how many will jump off a bridge...Although I am not, nor are you, compelled to do either.
(Skinner 1971 p.21)
Skinner’s
reply makes it clear that he considered Hume’s definition of freedom as
irrational chaos, or inexplicable autonomy, to be the only possible definition.
...he [Krutch]
can scarcely mean that those who go to the seashore do not go for good
reason, or that circumstances in the life of a suicide do not have some bearing
on the fact that he jumps off a bridge (ibid.)
His
definition of freedom made him ignore the fact that actuarial tables only
describe probabilities, not certainties, and therefore the reasons people have
for going to the beach do not eliminate the possibility of free choice.
He
cited as another example of the illusion of Autonomous Man the fact that
“we read that hundreds of people will be killed in traffic accidents on a
holiday weekend and take to the road as if personally exempt.” (p.20).
But people who drive on holiday weekends are not just playing Russian roulette.
There are a variety of stimuli that make drivers respond in ways that increase
accidents on the road during holidays. There are more opportunities to drink,
there is a greater tendency to relax and be careless when one is on vacation,
there are lots of other people on the road who are drunk and careless. But the
effects of any of these stimuli can only be predicted with probability, and not
with certainty. A certain number of people will choose not to respond to those
stimuli; they will not drink an hour before driving, they will become more
vigilant, instead of more careless, and/or they will drive defensively to
protect themselves from those who remain drunk and careless. The people who do
any or all of these things will do them because of other stimuli they are
responding to; they might have received advice from their parents or seen
friends die in car accidents. But these stimuli would only produce a probable
response; there would always be other people who would not have responded that
way to those exact stimuli. No
matter how far back we followed this chain of causes, we would find no
necessary responses, so none of these stimuli would have eliminated the choices
of those who responded to them. If Skinner’s behavioral engineers wish to
use their technology to decrease holiday traffic accidents, they have no need
to deny the validity of this kind of freedom. The conditioning they imposed
would decrease the number of accidents statistically, while leaving the choices
of each individual driver completely intact.
The
compatibility of operant conditioning and choice is perhaps best illustrated by
an apocryphal story about two children of an alcoholic, one of whom became an
alcoholic and the other a teetotaller. The explanation that each gave for his
behavior was "with an alcoholic father, what else could I have done?"
It
thus appears that Skinner the
scientist offers us no new data that compel us to consigned free will to the
ontological scrapheap next to caloric and the alchemical essences. There are plenty of problems with the
idea of free will, and Skinner often confuses us by juxtaposing his paraphrases
of these ancient conundrums with references to his laboratory research. There
is nothing in the research itself, however, that compels us to jettison the
idea of free will and use Newtonian causality to explain both the behavior of
organisms and the motions of matter.
On the contrary, Skinner's data indicate that radically different
methods are needed for each, and that the concept of free will, as confused as
it is, contains many elements which are more applicable to Skinner's subject
matter than is the notion of Newtonian cause.
Bibliography
Atkinson,
R.C. (ed.) [1988] Stevens
handbook of experimental psychology 2nd Edition New York Wiley.
Dennet, Daniel [1978] Brainstorms MIT Press Cambridge
Dennet,
Daniel [1984] Elbow Room MIT
Press Cambridge.
Hilgard,
E.R. review of Hull et al Psychological Bulletin 37:
808-815.
Hocutt,
Max Free Will (a talk for the 1990 meeting of the
southern society of philosophy and psychology)
Hull,
Clark [1943] Principles of Behavior
New York
Appleton-Century-Crofts.
Hull,
Clark et al. (1940) Mathematico-deductive Theory of Rote Learning; a Study
in Scientific Methodology. New Haven
Yale University Press.
Hume,
David An Inquiry Concerning
Human Understanding.
Jean,
Sir James [1945] Physics and Philosophy
Cambridge University Press.
Keenan
J.H. et al. [1982] “Thermodynamics,
Principles of “ in Encyclopedia Britannica Chicago William Benton.
Kane,
Robert [1985] Free Will and
Values Albany SUNY press.
Leahey,
Thomas [1991] A History of Modern Psychology New
Jersey Prentice-Hall.
Pavlov,
Ivan [1928] Lectures on Conditioned Reflexes New York
International Publishers.
Skinner,
B.F. [1938] The Behavior of
Organisms D. Appleton Century Company.
Skinner,
B.F. [1953] Science and Human Behavior New York Macmillan
Skinner,
B.F. and Ferster, C.B. [1957] Schedules of Reinforcement New
York Appleton Century Crofts
Skinner,
B.F. [1969] Contingencies of Reinforcement New York Appleton
Century Crofts
Skinner,
B.F. [1971] Beyond Freedom and
Dignity New York Alfred Knopf.
Skinner,
B.F. [1972] “Some responses to the stimulus
‘Pavlov’” in Cumulative Record; a selection of Papers
New York Appleton-Century Crofts.
Skinner, B. F. [December 1984] "Methods and Theories in the
Experimental Analysis of Behavior" in Behavioral and Brain Sciences Vol. 7
#4.
Stapp,
Henry P. [1990] A Quantum
Theory of the Mind-Brain Interface Invited presentation to the conference:
Consciousness within Science, UCSF.
Vitielo,
G. , and E. Delguidice, S. Doglia, and M. Milani. [1984] "Boson Condensation in Biological
Systems." in Nonlinear Electrodynamics of Biological Systems W.
Ross Adey and Albert F. Lawrence, Editors. Plenum Press.
Volgyesi,
F.A. [1963] Hypnosis of Man and
Animals Baltimore Williams and Wilkins.
Young,
Arthur [1976] the Reflexive
Universe Delacorte Press.
[1]The first person to make this point was Arthur Young, in his fascinating book The Reflexive Universe. He also pointed out there is a similar decrease in certainty as we travel downward from physical matter to atoms and to subatomic particles. More on that later.
2 Hume's definition of Free Will is not only unsatisfying, it is not actually a very accurate description of lunacy either. Neurotics and psychotics are unpredictable only to the majority of people who do not take the time to study them carefully. Most therapists usually find that their patients are more rigid and machine-like than normal people, and thus frequently become stuck in repetitive and self-destructive patterns.