Professional Documents
Culture Documents
Beyond Behaviorism 1973
Beyond Behaviorism 1973
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp
.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.
American Association for the Advancement of Science is collaborating with JSTOR to digitize, preserve and
extend access to Science.
http://www.jstor.org
The more closely the rat's acts are keep the behavioral result constant. The the presence of an inner representation
examined, the more variability is seen. "cue" hypothesis comes after the fact of the quantity in the form of a signal
Yet in every case the variations in the and overlooks too many practical diffi- or set of signals. It must contain or be
acts have a common effect: they lead culties to be accepted with any com- given something equivalent to a ref-
toward the final result that repeats fort. Yet what is the alternative? It is erence signal (or multiple reference sig-
every time. In fact, if precisely those to conclude that acts vary in order to nals) which specifies the "desired" state
variations did not occur, the final re- create a constant behavioral result. That of the controlled quantity. The sensor
sult would not be the same every time. implies purpose: the purpose of acts is signal and the reference signal must be
Somehow the different effects apparently to produce the result that is in fact compared, and the resulting error sig-
caused by the stimulus light are ex- observed. This is the alternative which nal must actuate the system's output ef-
actly those required to compensate for I recommend accepting. fectors or outputs. And finally, the sys-
differences in initial conditions on each tem's outputs must be able to affect
trial. This situation was clearly recog- the controlled quantity in each dimen-
nized by the noted philosopher of be- Feedback Control sion that is to be controlled. There are
haviorism, Egon Brunswik (5). other arrangements equivalent to this,
The accepted explanation for this Behaviorists have rejected purposes but this one makes the action the clear-
phenomenon of compensation is that or goals in behavior because it has est.
the changed initial conditions provide seemed that goals are neither observable This physical arrangement of com-
"cues," changes in the general back- nor essential. I will show that they are ponents is further constrained by the
ground stimuli, which somehow modi- both. There can be no rational expla- requirement that the system always op-
fy the effect of the main stimulus in the nation of behavior that overlooks the pose disturbances tending to create a
right way. There are three main prob- overriding influence of an organism's nonzero error signal; this is tantamount
lems created by this explanation. First, present structure of goals (whatever its to saying that the system must be or-
these hypothetical "cues" must act with origins), and there can be no nontrivial ganized for negative (not positive) feed-
quantitative accuracy on the nervous description of responses to stimuli that back, and that it must be dynamically
system employing muscles which, be- leaves out purposes. When purposes stable-it must not itself create errors
cause they are subject to fatigue, give are properly understood in terms of that keep it "hunting" about the final
anything but a quantitative response to feedback phenomena, acts and results steady-state condition. There is no point
nerve impulses. Second, these "cues" are seen to be lawfully related in a sim- in concern with unstable systems, be-
are hypothetical. They are never ex- ple and direct way. We will see this re- cause the (normal) behavior we wish to
perimentally elucidated in toto, and lationship using a simple canonical explain does not show the symptoms of
there are many cases in which one can- model of a feedback control system. dynamic instability-and we do not
not see how any cue but the behavioral Engineers use negative feedback con- have to design the system.
result itself could be sensed. Third, the trol systems to hold some physical quan- This system is modeled after Wie-
compensation explanation cannot deal tity in a predetermined state, in an en- ner's original concept (6). In the system
with successful accomplishment of the vironment containing sources of disturb- I describe, however, there are certain
behavioral result in a novel situation, ance that tend to change the quantity changes in geometry, particularly the
where presumably there has been no when it is uncontrolled. Every control placement of the system boundary and
opportunity for new "cues" to attain system of this kind must have certain the identification of the sensor (not ref-
control of responses. major features. It must sense the con- erence) signal as the immediate con-
The central fact that needs explana- trolled quantity in each dimension in sequence of a stimulus input. This is a
tion is the mysterious fashion in which which the quantity is to be controlled continuous-variable (analog) model,
actions vary in just the way needed to (Sensor function in Fig. 1); this implies without provision for learning.
352 SCIENCE, VOL. 179