Collecting data about behavioral regularities

Unedited posts from archives of CSG-L (see INTROCSG.NET):

Date: Sun Apr 18, 1993 9:42 am PST

Subject: Collecting data about behavioral regularities

[From Bill Powers (930418.0900)] Ken Hacker (930418) --

> You recently described how social and behavioral scientists amassfindings about behavioral regularities and how PCT scientists develop firstprinciples about human behavior. I believe you are right, but I do not agreewith the implication that there is anything wrong with collecting data aboutbehavioral regularities.

Before I talk about collecting data, let's make the "first principles"claim clearer. It's easy to let that claim slide gradually into a boast thatPCT can explain everything that happens in the animal kingdom without evencollecting any data.

When I say that PCT applies to all behavior all of the time, with noexceptions, I'm trying to convey the KIND of theory it is, not claim that PCThas finally wrapped up all the problems of life and we can all go home. I'mtrying to say that a principle like control of perception isn't meant to applyonly to a certain population under certain conditions, and only to someunspecifiable set of people within that population. If it's true, it neverstops being true and there's no person for whom it's not true.

The only comparable theory in psychology is stimulus-responsetheory which, regardless of semantic quibbles, is still the primary theory ofbehavior in the life sciences. The basic principle of S-Rtheory is that the environment acts on the organism through direct contact andthrough the sensory organs, and that out of these effects come the motoractivities and their consequences that we know as behavior. This theory, too,applies all of the time to all individuals with no exceptions. It is that KINDof theory. It is meant to be a universal principle like the law of gravitationwhich never turns off.

There are very few psychologists today who would admit to being S-Rpsychologists. This is because "S-R"has come to mean a special narrow application of the principle in which onlyspecific simple physical stimuli are assigned the role of cause, and behavioraloutputs --motor outputs --are linked directly to the stimuli without any attempt to characterize theintervening organism --reflexology. I have had psychologists tell me that of course S-Rtheory is dead --everyone now knows that the correct theory is S-O-R,stimulus-organism-response.So if you talk about what happens to stimuli on the way through the organism,if you talk about traits and tendencies and propensities and cognitions, youare not an S-Rpsychologist. But of course the underlying theory is that what happens to anorganism determines what it does, just as before, although now the process ismore complex and takes longer.

Skinnerians do not consider themselves S-Rpsychologists. They talk about classes of stimuli grouped according to theireffects and classes of behaviors grouped according to their consequences; theytalk about contingencies and reinforcing effects which are not stimuli thataffect the senses, but conditioning processes which alter the shape of behaviorsimply by being in existence. But if you ask any of them what, in the finalanalysis, determines the way organisms behave, they are very firm about theonly proper answer: the environment. If you approach it the other way aroundand ask what it is that the environment determines, the answer is "behavior."Water deprivation causes drinking, not thirst. The only difference from S-Rtheory is that under Skinner, it is no longer a theory but a fundamentalscientific principle. If you don't believe that the environment determinesbehavior, you're not a scientist.

There is one certain way to see whether any given approach to behavior isbased on S-Rtheory: look at its experimental methodology. This cuts through all the verbalBS and shows you what the underlying model is. Throughout the behavioralsciences, the almost universal practice is to hold all variables constant asnearly as possible, and then vary just one thing --the schedule of reinforcement, the stimulus, the situation --,while recording the change in behavior. The more careful scientists also rundummy experiments --odd that they should call them "control" experiments --in which the critical manipulation is replaced by some neutral operation, sothe effects of merely doing the experiment can be factored out. The results ofsuch studies are reported, almost universally, as "The effect of treatment A onbehavior B."

While philosophers of science have provided all sorts of modifyingstatements and disclaimers of causal implications in such conclusions, the factis that those who do experiments under this methodology believe that if it werenot for the treatment A, the change in behavior B would not have happened. Thebehavior changed BECAUSE OF the treatment, and without the treatment it wouldnot have changed in the same way.

This is still S-Rtheory, no matter how many scientists who use this method howl that they arenot, not, not S-Rtheorists. Those who object are simply looking at the situation too narrowly;it never occurs to them that the shape of behavior might NOT depend AT ALL onwhat happens to an organism. It's just that behavior doesn't depend on theenvironment in the simple-mindedway they associate with the term S-R.For most scientists in the behavioral sciences, the real principles of S-Rtheory apply to all organisms all of the time with no exceptions. They havejust stopped calling those principles "S-R."

PCT is incompatible with S-Rtheory because it says that the behavior of organisms is part of a closed loopof relationships, at the center of which is the organism's preference for whateffects the environment is to have on it. If organisms are organized as PCTsays they are, then S-Rtheory is incorrect under all circumstances and with respect to every organism,all of the time. It is incorrect now and it has always been incorrect. It wasincorrect every time it was used to design an experiment, and every time it wasused to interpret the results. There is no possible compromise: you can't havetwo incompatible universal principles operating at the same time. Since eithermust operate all of the time, they are mutually exclusive.

Now let's talk about "collecting data about behavioralregularities."

What does the term "behavioral regularity" mean? Usually it doesn't meanthat an organism regularly, and for no reason at all, emits some pattern ofbehavior. Usually what it means is that the behaviors we observe have someregular connection to an observable condition in the environment of anorganism. It means that if there is some change in the environmental situation --in happenings, arrangements, information, processes --we can learn to expect that some typical change in behavior will follow. The"regularity" in question isn't just a regular behavior like the swinging of thependulum of a Grandfather clock. It's a regular _relationship_ between behaviorand something else. The pendulum behaves regularly, but not as a consequence ofwhat's happening in the environment around the clock. The regularities thebehavioral scientist is interested in are those that can be seen asconsequences or influences of some antecedent event or situation.

So with that understanding, how would we go about gathering data aboutbehavioral regularities?

Why, we would look for or create changes in the environment of theorganism, and record consequent changes in the behavior, trying to find someregular dependency of behavior on the change in the environment. If theregularity is hard to see, we could apply the sophisticated techniques ofstatistical analysis to bring it out. But basically what we would be lookingfor is some regular way in which behavior is affected by the environment. Wewould be applying the fundamental principle of S-Rpsychology, taking it for granted. The S-Rprinciple would be at the root of all our interpretations of what theseregularities mean, as well as the method by which we go about findingthem.

What would be done differently if the S-Rprinciple were replaced _in toto_ by PCT? As we are talking about universalprinciples here, there is no halfway measure or compromise possible. To usePCT, you must totally abandon the S-Rprinciple. This means that every experimental result, every fact obtained sofar under the S-Rprinciple, must be re-evaluatedand reinterpreted --not necessarily discarded, but seen in the light of a different concept of whatbehavior is.

The first item to be reinterpreted is behavior itself. Under PCT, behavioris not simply a motor output of an organism or a subsequent effect in theenvironment. What has been termed behavior, in fact, turns out to be somethingquite different: a controlled outcome. And controlled outcomes are not justoutcomes that happen to be stable against disturbances; they are outcomes thatare _perceived_. What constitutes an outcome can no longer be determinedarbitrarily by the external observer. The nature of an outcome is defined bythe way the organism perceives the environment, and the particular outcomewithin the range of possible forms it might take is determined by the referencesignals inside the organism. The perceptual apparatus of the organismdetermines the kind of thing that will be seen as an outcome, and the referencesignal determines the particular state of the perceived outcome that will besought.

So when the PCT researcher and the S-Rresearcher look at a particular behavior, they see different things. The S-Rresearcher sees an act like answering a question as a response, something thatcomes out of the organism as a result of a question that went into theorganism. The PCT researcher sees the question as part of a controlledrelationship between a perception of the person's own utterances and aperception of the utterances of another person. The answer is not caused by thequestion; it and the question, together, form a controlled pattern. Thispattern can be disturbed, resulting in corrective action by either party; onemight say "Will you please stop answering my questions with questions?" or theother might say "When I answer your questions, I want some sort ofacknowledgement that you heard me." The controlled pattern spans more than onequestion and one answer.

In the PCT view there is no single locus of behavior. Behavior is acontrolled pattern of perceptions, and includes not only what the behaver isdoing but what the environment is doing. To understand behavior under PCT, itis necessary to understand the entire action-environmentrelationship as a continuing pattern under control by the organism andcontinually maintained near a state preferred by the organism.

This means that behavior can no longer be thought of as "caused." It is_maintained_. If the environment changes in some way that tends to alter thecontrolled pattern of perceptions, the motor activities of the organism shiftin the way required to maintain the controlled pattern in the same form, theform intended by the organism. This relationship between changes in theenvironment and changes in motor activities explains the appearance that theenvironmental changes caused the changes in action, but PCT shows that this isan incorrect interpretation. The focus of this changing relationship betweenactions and the environment is the constant pattern of outcomes beingmaintained by the organism.

The meaning of "collecting data about behavioral regularities" is nowcompletely different. On the way to finding such data, it may be useful torecord apparent cause-effectrelationships between changes in the environment and changes in motoractivities. But that is not the data desired; what is desired is to find theconstant patterns of outcomes that are maintained as a result of these shiftsand countershifts. Those constant patterns of outcomes are the regularities ofinterest under PCT, the regularities that tell us what the organism is reallydoing.

Hierarchical PCT, or HPCT, introduces another dimension into datagathering. At one level of understanding, we see behavior as a controlledpattern of consequences in which both actions and environment play a part. Butat another level, we can see that these controlled patterns can shift from oneform to another, and in doing so help to maintain controlled patterns of a moregeneral kind. So the data-gatheringprocess expands; we now recognize regularities that are maintained by _changes_in lesser regularities. We see how some controlled patterns are maintained asthey are, and maintained near changing reference conditions, as a means ofestablishing and maintaining higher-levelpatterns, more general ones.

Somewhere in the higher levels of control, these patterns explicitlyinvolve other people. The organism, by altering the behavioral patterns atlower levels, presents other organisms with environmental changes that becomeincorporated into the controlled patterns of the other organisms, and in thatprocess all organisms become part of each other's controlled patterns. This ishow social phenomena arise from the fundamental properties of individualorganisms. The data we gather about such phenomena remains the same: we want tofind the outcomes that are under control by each organism. But now the outcomesare defined partly in terms of how each organism affects and perceives thebehavior of other organisms, and what intentions each organism has for thedesired states of those perceptions.

So under PCT, as one would expect from any new universal principle, thesame object of study takes on a completely new appearance, and both theobjectives and the methods of scientific investigation become somethingentirely new. We are looking at exactly the same phenomena that the S-Rtheorist sees. But in those phenomena we discern entirely newrelationships.

Best, Bill P.