II. The Nature of Explanation
A. Conceptual Precursor: Description, Prediction, Cause, and Explanation
A “description” details aspects of an event, e.g., the pigeon pecked the key 25 times; Johnny ran across the room. "Prediction" is the correct description of something which has not yet occurred, e.g., the pigeon will peck 25 times; Johnny will run across the room. "Cause" correctly specifies the efficient cause or the necessary and/or sufficient difference in the environment which preceded the resulting change in behavior. Following exposure to the fixed ratio schedule the pigeon pecked the key 25 times; when the spider fell on his arm, Johnny ran across the room. "Explanation" is the specification of the ancillary factors necessary and/or sufficient for the cause-effect relationship and how that relationship will change with changes in the context. Explanations enable the productive use of knowledge. For example, an explanation would tell us how the reinforcement rate alters the rate of an operant, what information lead to that general rule, and how to determine when the rate will apply to specific instances. It is also the case that a complete description of past, current, and future causal relationships under a variety of conditions is an explanation; so all the terms are related. Staddon has suggested that explanations must include a mechanism of variation and a mechanism of selection.
B. What is an Explanation?
Explanations are a special case of the vicarious system illustrated in Chapter 3 section VI. B. An explanation is like a metaphor. Explanations are one of the products of science. The word explanation is easily used. "The explanation for this phenomenon is ...." "We need to find an explanation for ..." In an earlier section (I. C. 2. e.) we discussed the general requisites of an explanation. In this section we discuss the characteristics and types of explanations. A little thought reveals that there are actually a variety of ways by which something can be "explained."
A paradigmatic framework creates the context which provides for acceptable
explanations. A paradigm is, in fact, created when a group of investigators
is interested in the same sort of phenomena and accept the same type of
explanation. Adequate explanation within a paradigm, therefore, is to a
degree determined by the curiosity of the pioneers of that subsequently
popular paradigm. Curiosity led those individuals to particular questions
and the satisfaction of their curiosity determined their explanatory system.
(Unfortunately, in some cases “explanations” worked too well.
Their curiosity was terminated and progress was also terminated.)
In one sense explanation can be considered as simply the activity which successfully pacifies curiosity because it makes an event not surprising because it is "understood." If it is understood it can be predicted, controlled, and synthesized. That specification is integrated within a coherent paradigm so that events never before experienced can also be better understood. An explanation for an event brings to bear a wide variety of "multiple converging evidence" by pointing out how that particular cause-effect relationship is only one of a great many similar events which can be said to follow the same rules. Following a discussion of the characteristics of an explanation, the types of explanations will be detailed.
C. Characteristics of Explanation
Explanations can be seen as very broad "if A then B" statements where A is some well integrated general conditions or rules, and the detailed specification of the conditions under which the effect will occur, e.g., if you hold a match under a piece of paper, then the paper will catch fire provided that the atmosphere ... the composition of the paper is ... and the temperature of the match is ... It occurs because of the rapid ...
Specifically, there is a difference between two sets of observations to be explained, e.g., most pieces of paper just lay there, this piece catches fire. This difference in observations is to be understood by relating it to the differences in treatments received by the two groups of paper and the integrated framework within which that result of the cause was the obvious result of the factors involved. Explanations are generally presented as a result (B) and then its explanation (A) e.g., the paper caught fire because a match was held under it and those conditions work in such and such a way. The explanation for why the pigeon pecked is the specification of the context for the set of causes for that result such that those specifications "make sense" out of that result. It allows us to understand the event and not be surprised by it.
A simple diagram may help depict the various requisites of an acceptable explanation. The various aspects are detailed in the sections below. The characteristics which follow could, for the most part, be seen as logical consequences of the demand for truth and understanding. The issues are reiterated in this section in terminology more typically applied to explanations.
The sets A, B, A' and B' and even the paradigm itself each include elements and exclude others. The elements comprising those sets (i.e., the Xs and Os) and the set boundaries must meet the requirements of “truth” detailed in a previous section.
a. Analogy Versus Explanation
Analogic models are advanced with little actual support, or simply by "analogy." They are not necessarily true. The model has only rhetorical or presumed similarity to the phenomenon to be explained and is, therefore, an unacceptable explanation. Recall at this point that acceptable explanations must have empirical, reliable, multiple converging, consensually valid and operationally defined support. Analogies do not have this type of broadly based necessary connection to nature. For example, "pressure builds up in the id until it breaks through and causes this behavior." This is a hydraulic metaphor with substantial support as a model of gas station grease racks but no support for being applicable to the behavior of people. When people say they are Freudian, more than likely what they actually believe is that hydraulic systems work that way (e.g., if pressure builds up enough then something will burst, thus "relieving" the pressure). They are so caught up in the plausible validity of the physics of a hydraulic mechanism that they forget to ask if it applies to human behavior. The syntax is so compelling they ignore the semantics (Chapter 3 Section VI. B.). Similar analogies can be drawn from computer science or whatever discipline is currently experiencing "glamorous" findings. Freud developed his theoretical model of human behavior when hydraulics were the popular craze in engineering. It is likely that superconductivity or nonlinear dynamical systems (chaos) will be offered as an explanation of human behavior within a few years. It may very well be right - but it may also be wrong even though it is at the height of fashion.
b. Folklore Versus Explanation
Often, notions creep into "common knowledge" without passing any test to assure that they are true. A surprisingly common pathway is that offhand, unproven possibilities presented in discussion sections are repeatedly invoked as if they were the point proven in the procedure. Soon it becomes common knowledge that a particular paper proved something that it did not.
An explanation must be explicit. The boundaries of A, B, A', and B' must be clearly stated. Which elements are included and which are not must be unambiguously specified. In order to properly explain a phenomena the communication must explicitly and clearly specify the proposed mechanisms or rules. Statements must be unambiguously connected to the empirical world. Ambiguous statements give only the illusion of successful explanation. In retrospect, they seem to have been right. However they actually say nothing because they could be construed to have contended anything. The oracle at Delphi once predicted to a king "if you attack Persia, a great kingdom will fall." The visitor attacked on the basis of the "favorable" prediction - and lost. Not realizing that the oracle would have been right either way. His own great kingdom fell. In sum, there must be a clear and unambiguous way to determine which elements are contained and which are not contained in the sets.
Sets may be defined in a number of ways. The value of any particular method of defining a set is governed by its productiveness with respect to the larger paradigm.
A set may be defined by enumerating each of its elements.
A set may be defined by a rule which determines which elements are included.
(1) Functional Definition
A set of behaviors may be defined as all those behaviors which result in the same “end”; examples would be as diverse as "key peck," "go home," or "clean up." As with any other type of definition, only to the extent that these functionally defined sets enter into orderly relationships are they meaningful.
i. Qualitative Explanation
This is a simple qualitative specification of the expected change in the dependent measure with the occurrence of some independent variable. For example, the theory could predict that the response rate should increase with the occurrence of a reinforcer. No attempt to specify a systematic change as a function of the independent variable is made.
ii. Quantitative Explanation or "Model"
Models specify a change in the dependent variable as a function of a change in the independent variable and also suggest an algorithm which publicly and reliably produces results similar to an actual subject. If it makes correct predictions, then in a sense the model explains by virtue of working. Ideally, the proposed mechanism or process breaks the mathematical specifications of the functional relationship into easy to conceptualize factors which may have empirical reality; but the specification may not lend itself to easy conceptualization in three dimensions. Physics gave up easy conceptualizations years ago. If the model does not specify outputs in terms of inputs it is of little real use.
Various implementations of theories based on models may or may not suggest additional dimensions such as: that its proposed mechanism is inside the organism (reductionistic); that nature works toward that end (teleological); or it may simply provide the relationship between inputs and outputs without adding hypothetical or intervening elements (correlative). There are therefore two aspects to models: their specificity and power to predict empirical relationships (qualitative versus quantitative) and secondarily the presumed empirical reality of the proposed process. These two aspects are often confused.
Quantitative models of behavior are becoming the method of choice in the analysis of behavior. Any model which correctly predicts and correctly quantifies the relationship between the environment and behavior is potentially of great importance. Competing models are to be judged on the basis of their ability to produce all those elements presented under “truth” and “understanding” and further elaborated in Chapter 4. A model must have generality, must be accurate and must minimize free parameters. It must also have theoretical machinery which correctly links it to other levels of paradigmatic molarity and time scales of adaptation. If the model suggests that it is anything more than a way of describing relationships, then that's another matter. The phrase, rule-governed behavior, versus rule-describable behavior captures this important issue. The first implies that an inner entity is using rules to make decisions. The second implies that we can describe what the animal did with a simple mathematical specification.
An explanation must not be tautological. The boundaries of A, B, A', and B' must be different. The A or A' cannot be a restatement of the B or B'. It is not acceptable to simply describe an event and then turn around and use that description as the cause for that event. The paper is burning because something caused it to catch fire. If you do that, you are not dealing with a functional relationship (an input and output, or a cause and effect). Rather you are simply talking about an effect and then using a synonym for that effect as the cause. Explanations for behaviors must add to our knowledge or understanding, not simply rearrange the words. If you were to consult a dictionary or an expert with the following question: "What does the principle of 'scientific manipulation' mean", and they responded: "that's when you manipulate things scientifically", then you would not be happy and you would not have gained much. Tautological explanations are similar: they explain why a phenomena occurs by nothing more than a verbal smoke screen. When a functional relationship between two elements is proposed, both elements must be unambiguously and independently connected to separate empirical events.
An explanation must be testable. The boundaries of A, B, A' and B' must allow someone to clearly predict whether an element is within or not within the set before being told by the proponent of the explanation. The boundaries cannot be a secret!
At least one element must be predicted by the rules and obtained as the result of empirical confirmation.
This second "testability" criterion is known as falsifiability or making a risky prediction. A useful theory or therapy must tell you something that isn't obvious to everyone. It must contribute to your knowledge or ability to solve problems or it has no value. If I tell you that a coin toss will result in a “head or a tail” then I haven't told you anything that you didn't already know even though future events will prove my prediction "true." If a theory predicts that a person will either get better, get worse, or stay the same then it hasn't said anything that any fool doesn't already know even though future events will prove that prediction "true." If a theory says that a person will not get worse but will either stay the same or get better then it has narrowed down the possibilities and has told you something useful. In doing so however, it has put itself at risk in that the person could actually get worse and prove the theory wrong. It can be seen therefore, that in order to say anything, a theory must take a stand and put itself at risk. A simple way to tell whether or not a theory would be useful therefore is to determine if it makes any risky predictions. Can it be falsified? Not all results can lead to confirming the theory.
The information transmitted (or the degree of understanding conveyed) in a communication is the amount of uncertainty reduced. Note the following figure. If you had to guess the value of a number between 1 and 10, and I told you that it was the number 6 (see line B), then I would have reduced your uncertainty to zero
||<- - - - Range of Problem - - - ->
I would have also told you something, (albeit less) if I only told you that it was between 4 and 8 (see line C). I would have reduced your uncertainty by half. I would not have told you anything at all if I told you it was between one and ten (line D). Your uncertainty would not have been reduced at all. In this last example it is also the case that I would have been correct no matter what happened no matter which was the right answer. You could never say I told you the number would be wrong when it was actually right. This of course explains why these correct but uninformative utterances are so popular. Philosophically these uninformative prophesies are said to make no "risky predictions." They cannot be disproved (in the sense of having said something that was wrong when it was right) in that they predict everything. Science demands that a theory make a risky prediction or be falsifiable, because you cannot be right (i.e., communicate information) if you cannot be wrong (i.e., some conceivable outcome would prove you wrong). In other words you cannot claim to be right if you have not reduced uncertainty. Science is interested in the size of the x'd in area in the preceding figure. Theories which make predictions that produce big x'd in areas are good theories, those which produce less x'd in areas are less acceptable. Let's face it, advisors are not much good if they tell you that anything you think of is right. They gave you no more new information than a tape recording of your voice.
If a theory makes predictions which have not yet been tested, it is pre-science.
As is illustrated in the following figure, explanations like decisions (see Chapter 4 A. 1.) have four possible occurrences.
1. Elements which should be in the cause or rule are in the cause or rule (x's contained within set), 2. elements which should not be in the cause or rule are excluded from the cause or rule (o's excluded from set), and the two types of errors which are 3. elements which should be in cause or rule which are not (x's left out of set) and 4. elements which should not be contained which are (o's included within set). Similarly, for the results: elements which should be covered by the explanation are covered (5), elements which should not be covered are excluded (6), and the errors which are elements which should be covered which are not (7), and elements which should not be covered which are (8). Science is the process of developing better rules or boundary for the set of rules so that the errors are minimized.
Explanations must be parsimonious. The boundaries of A and A' must be defined in such a way as to exclude as many elements which do not belong (i.e., 4) as possible. There must be the minimal number of causal elements or predictor rules. The most productive explanations are the simplest. Wildly complex theories do not clarify problems, they make them worse. For example, if I look out my window and “see” a prowler and then look again and see a bush in his place, I can assume a unparsimonious explanation by contending that the prowler changed into a bush or I could accept the more parsimonious explanation that I was mistaken the first time I looked out the window. Rube Goldberg machines are mechanically unparsimonious. If a child misbehaves you may assume that the child: has a need to self-actualize; a need to return to the primitive inorganic state recalled by the cellular substance; has a mean streak; is possessed by the devil; or has obtained consequences for that misbehavior which are maintaining it. The principle of reinforcement is a single straight forward notion which accounts for a very wide range of behavior in a wide variety of species and therefore provides a very parsimonious explanation. An alternate example which illustrates the successive loss of parsimony is provided in this sequence of increasing complex descriptions for the same event: a key peck occurred, the pigeon pecked the key, the pigeon learned to peck the key, the pigeon learned to peck the key in order to get food, the pigeon learned to peck the key in order to solve its need for food etc. Science has relied on simple explanations of wide generality more than complex explanations of specific phenomena. It has resisted complication or unparsimonious explanations.
Explanations must be general. The boundary of B' must contain as many of the x's (5) as possible. As many events as possible must be explained. If you understand architecture and the rules of physics you will be able to build a variety of structures in a variety of situations without any collapsing. Einsteinian physics explains more than Newtonian physics. The principle of reinforcement is one of the most general principles of psychology. The more phenomena covered and the wider the range the more general.
As was presented there are two types of errors on the rules sides and two types of errors on the results sides. Clearly having no errors on either side is best. But given that there will be errors, science attempts to optimize its task. The idea is to explain the maximum number of phenomena (minimize 7) while minimizing the number of rules necessary to accomplish that end (minimize 4). A false alarm or the inclusion of an unnecessary theoretical element on the rule side (4) is considered worse than having a necessary one missing (3). Secondly, it is considered worse to have a restricted range of applicability (7) on the results side than explaining something incorrectly (8). Science could be seen as trying to explain more results with less theory. Each scientist that moves our knowledge in that direction succeeds.
An explanation to be useful must be understood by the community. It must be presented in such a way that as many people as possible understand it. Unfortunately, as a science matures fewer and fewer participants qualify to understand the most recondite knowledge of a field. In fact, progress can be measured by the distance from the lay.
6. Systematic or Principled
An explanation must be systematically and coherently integrated within a larger frame of reference or paradigm. It is the interlocking and cross validating nature of the processes, explanations, and observations that defines truth. It's easy to be eclectic and choose whatever explanation fits the situation for now, but it should be remembered that the freedom of eclecticism is the freedom of ignorance over knowledge. Productive wisdom must have both the explanation for the local effect as well as a systematic way to have chosen that explanation.
D. Types of Explanatory Mechanisms
1. Explanation via Confusion, Emotion, Empathy; or “Rhetorical”
Rhetorical explanations are not really explanations at all and are therefore unacceptable even though they can be made to appear correct by fiction writers or our own confusion. They are nonempirical verbal statements chosen to produce agreement not explanation. For example, “I hit Johnny because .....” (where reasons cause empathy for the view); “evolution is false because humans are debased by a view which suggests that they evolved” (an invalid emotional argument); “things fall because of gravity” (a tautology empty of any meaning). I saw a movie that made me believe it was true. These “explanations” are actually nothing more than verbal confusion.
This is not to say that all information or knowledge is simply a matter of opinion and that a person gets to pick and choose whatever they like. Some things are facts. It is a fact that you will burn your hand if you put it on a hot stove. It is a belief based in rhetoric however if you believe it simply because of the person telling you rather than because of its factual basis.
Belief or faith in what was done in the past (tradition) in lieu of understanding and explanation. Note that the position argued for on the basis of tradition is not necessarily wrong, rather the explanation adds nothing. "I (we) always did it this way, therefore it must be true." Belief in what "was," is labeled belief via tenacity. An equally erroneous view is that whatever is new or nontraditional is correct: "I (we) need to move on and do it the modern way."
b. Authority / Antiestablishment
Unquestioning belief or faith in authority in lieu of understanding and explanation. "I was told it by XXXX (mother, leader, teacher) therefore it must be true." Belief in what you are told simply on the basis of their credentials is labeled belief via authority. Again, an equally erroneous view is that the "establishment" position is always wrong.
Belief or faith in knowledge derived from feeling states in lieu of understanding and explanation. "I feel in my bones that this is true." Clinical or theoretical judgments are often justified with this excuse. It is essential to realize that you would not want to go to jail because El Excellente felt you were guilty. Belief in what you feel is right is labeled belief via subjectivism. Nurse Ratched was wrong because she believed her feelings were adequate justification for what she did to her patients. If a mob lynches someone they are guilty of murder not because the person being lynched is always innocent but rather the mob terminated a life for an unacceptable reason: their strong feelings that they are right.
Belief or faith in knowledge apprehended directly without experience or reason in lieu of understanding and explanation. "I know that this is true." Belief in what you know is right is labeled belief via intuition or via a priori knowledge. It differs from subjectivism in that there is no emotional basis for the belief, it is simply "known." It differs from rationalism in that it was not “reasoned out.”
e. Rationalism / A Priori
Belief or faith in knowledge acquired through reasoning process alone in lieu of understanding and explanation. "I reasoned this out so it must be true." Often people make erroneous but seemingly logical predictions because they misunderstand what is happening or are unaware of all the facts. Belief in what you have reasoned out is labeled belief via rationalism. William F. Buckley, Jr. is a rationalist. He could convince you that black is white. Good salesmen are good at plausible arguments which are not true. The problem with rationalism is easily understood when you recall the time you talked someone into believing a lie by using what the other person thought was inescapable logic. Unfortunately, some people who claim to be scientists are more into winning than finding truth.
Belief or faith in pseudo-knowledge "created" when a synonym for an event is used to explain the event in lieu of understanding and explanation. Gravity caused the object to fall; the rate of responding increased because it was reinforced; the person believes someone is out to get them because they are paranoid. A classic analog of this type of pseudo-explanation is finding a word defined in a dictionary with a word which is in turn defined by the word you were looking up. These are called circular explanations because the explanation explains itself by referring back to itself. If the explanation for a problem is seen as "anchoring" or "supporting," the problem then a tautology is a problem with its supports tied back to itself. A porch swing tied to itself for support. Grabbing the back of your neck and lifting yourself off the ground. A snake completely devouring itself.
i. Adding Information which is not Factual (lie of commission)
ii. Removing Information which is of Importance (lie of omission)
Television programs often emphasize these rhetorical forms of knowledge. The writer of course sets it up so that these types of reasoning always win. This is because a generally uneducated audience cannot follow any other type of logic. They often do not have a very large or well-integrated paradigm. It requires considerable effort to learn about a wide variety of phenomena and check the explanations for each so that all of them are coherent. It is much easier to make up isolated explanations for each event. This is what it is to be ignorant.
A second and often unrecognized but serious problem develops when normally intelligent people are impressed over and over again, year after year, by simple "common sense" being correct in the make-believe TV world ("simple exposure effect"). People start using those rationales in their everyday interactions without realizing that they have used the erroneous, make-believe TV fantasy as their role model.
Imagine yourself being subjected to a decision (hang by the neck until dead) based on any one of these rhetorical types of logic. The "Quincy" or "Hawkeye" characters on television, with their error prone rationales, would not be admirable if the consequences for their logic was more like what would happen in real life.
If you believe that TV does not have an effect on how we think, then how about making a number of children’s shows which encourage children to ignore hard-won wisdom (e.g., don't take rides with strangers) and at the end of the show the kid is right and the parents are wrong and all of society is happy that the kid did exactly what wisdom would have rejected. We would not like shows like this on TV because they would provide very bad role models. We would think that they would subliminally teach people who didn't yet know better, to do what is harmful. Quincy and Hawkeye get to do things that virtually never work in real life and at the end of the show the writer makes everybody glad they did. The next time you watch either show, "identify" with some character other than Quincy or Hawkeye, or think about what would happen if they were wrong and your life or your career were at stake.
2. Explanation via Specification of the Goal or End; or “Teleological”
Nothing can ever read the future. The future does not exist in the present. Things cannot reach out of the future to cause things to happen in the present. Only by the twisting of a metaphor or in a simplistic shorthand can organisms work for a future goal. A pigeon cannot be placed on an FR 100 schedule and be expected to behave appropriately the first time. A teleological explanation is a shortcut description of what is actually the result of a history of exposure to the ontogenetic and phylogenetic contingencies. In this sense a teleological explanation is "backwards talk." What appears to be a common trend in a history of functional relationships becomes a future state and subsequently becomes a future goal and eventually a future cause. Note that teleonomic (covered in Chapter 1, Section III. 4. b.) should not be confused with teleological.
These explanations assume an entity which is aware of the contingencies and acts so as to benefit from that contingency. These explanations are therefore unacceptable for two reasons. First, nothing can read the future, the future does not exist in the present. Secondly, it's silly to explain why a person or pigeon does something by postulating a little man inside that does it. Even if that were the case, we should then be studying the little man inside. Possibly try to find the littler man within the little man?
A bird could be said to build a nest in order to protect its young (agency internal to the organism that "knows" what would be good in the future). The problem with this explanation is that it assumes the organism has some divine power to see into the future. Past experience with the phylogenetic or ontogenetic contingency must be at the root of the behavior. It could also be said that mother nature shaped the shark so it could swim through the water better (outside agent that knows what would be good in the future). The problem here is the assumption of an outside agency which can read the future, and then chooses to optimize the shark, and then carries out that desire.
b. Without Intentionality
This implies the speaker has a knowledge of the state of "perfection." The speaker extrapolates the presumed goal from the existing data, then states that the activity is directed toward that goal.
Often, belief in the reality of what began as an extrapolated goal, leads to erroneous predictions. For example: 1. a "goose tries to get the egg into the nest to protect it" even when the egg has been removed. It always uses its bill but never its wing or foot (both of which would be more efficient); and 2. a hen sits on eggs depending on her own temperature not the temperature "necessary to protect the eggs." These are clear examples that our extrapolated cause or teleological cause for the behavior could be totally wrong, and would lead us to make erroneous predictions, and erroneous theories.
3. Explanation via Appeal to Inner Cause; or “Reductionistic”
These explanations are based on an internal, more fundamental or elemental
process, or invoke additional more primitive elements to accomplish their
explanatory power. These types of explanations have been very fruitful in
medicine. However, they have had notoriously bad track records in psychology.
The issue raised by behaviorists is not that there are no processes operating
at a more reduced level, nor that only correlative explanations are
acceptable. The issue is not whether you have what you point to when
you say “consciousness”,
but whether it is productive to say that the reason you went to the store
is that you “wanted to.” The point is that reductionistic explanations are unnecessary, and that correlative explanations are typically the most productive path to an applied solution. An often overlooked fact is that reductionistic entities cannot be useful if they have not yet been anchored to environmental inputs and outputs (i.e., a well-developed correlative explanation comes first). Reductionistic explanations are sometimes also labeled “analytical” (“of what is the thing composed”).
This is Physical Reductionism. The classic example is a physiological explanation for behavior. For example, the behavior occurred because of activity in the brain. It is important to keep in mind that a reference to the brain does not necessarily make something true. An empirical reductionistic explanation must show empirical evidence that that entity "causes" the behavior. Few investigators would suggest that the brain has no role whatsoever in the behavior of the organism. An empirical reductionistic explanation must add substance and understanding to what everyone already accepts.
This is Nonphysical Reductionism or Conceptual Reductionism. This category includes any reductionistic entity without actual empirical support (including nonempirical physiologizing). Ultimately this approach can be seen as tautological. (Cause and effect are both the same single observation).
i. With Basis in Folklore
Unsupported phenomenological notions such as the mind are typical of this category. Spence referred to this class of explanation as animistic. These explanations are unacceptable because they have only flimsy support. Their support rests mainly in folklore and provide no "truth" or "understanding." For example, "when his mind realized the solution, John made the correct response" (often shortened to "when John realized the solution he made the correct response"). We of course know he realized the solution because he made the correct response.
ii. With Substantial Converging Support
This category is presumed to have what Dewey referred to as warranted assertability. A presumed reductionistic process generated from theoretical perspectives however is not necessarily always on firmer ground than primitive notions of the mind. Nonempirical entities which do not allow us to resolve disagreements other than by authority or opinion are not in the realm of science but rather are dogma. The description "rule-governed behavior" (as opposed to rule-describable behavior) implies a reductionistic entity which is using those rules to decide what behavior to emit. This type of Cognitive Psychology is therefore reductionistic. The obvious question becomes how does the entity which is using the rules work? The issue is returned to in Section III. D. 4. c. under correlative explanations. The following figure contrasts reductionistic frameworks.
c. Flaw of Reductionism as Psychological Explanation
We would immediately understand the fundamental problem with the reductionistic stimulus-process-response approach if the example were perceptual. For example, if I show you a red card and you say red - how do I know that in your brain you actually see red? In actuality, no one cares. Concern for what takes place in the mind is a metaphysical question not a scientific one. The relevant facts are presenting the card and documenting the answer.
If someone has a behavioral problem - how do we fix it? More than likely we change the environment, not reconnect the neurons in the brain. The difference between behavioral and reductionistic views is fundamental: pass the skin and you are a biologist or a philosopher, not a psychologist. Psychologists deal with environments and behaviors. If a TV picture isn't exactly what you want, you usually turn one of the control knobs. You don't open up the back of the TV and start changing transistors and ripping out wires. TV sets have evolved. Adjustments which have to be made often and locally are under the control of knobs responsive to the demands of the viewer. Channels can be changed, the volume can be raised or lowered, and so on. Ultimately, of course, all those things are mediated internally by circuitry, but this is not typically what we mean when we ask: “How do I change channels.” From this perspective, what causes the channels to be changed is turning the knob, not changes in the inductance of a circuit. The same goes for life forms and their behavior. They, too, have evolved. The adjustments which need to be made often and locally are under the control of processes responsive to contingencies in the environment. Organisms learn as the result of non random relationships in the environment. Ultimately, of course, all those things are mediated internally by the biology of the organism, but this is not typically what we mean when we ask: “How do I change that organism's behavior”, or “what caused the behavior to be changed.” The answer we want is the changes in the environment which lead to the new equilibrium, not changes in neuronal activity. The conceptual difference is the difference between psychology and biology. The difference between the success of the medical model (reductionistic) in medicine and the success of the behavioral model (correlative) with psychological problems can be
seen as the difference between seeing behavior as "blown transistors" and "knobs set wrong." Medical people fix transistors, psychologists turn knobs.
An additional fundamental problem with reductionistic explanations is that they are arbitrary. If a mechanism operating at a lower more fundamental level is always better, then obviously a process more fundamental than the brain, the mind, or the mental processing center must be sought. This leads to an infinite regress.
Note that the maximally reduced entity is typically postulated to work solely in terms of inputs and outputs, which is exactly what a correlative view of the entire organism argued all along (note that the lowest level used in each figure is the same as the figure for the correlative or behavioral explanation). In this sense reductionistic explanations are simply "tag team" explanations. Ultimately, all theories must account for the relationship between inputs and outputs; otherwise they are simple metaphysics.
i. Unobserved Verbal Activity as a Cause of Behavior
(This issue is discussed more fully in Donahoe and Palmer (1994)). It could be believed that some cognitive verbal process is the root cause of our behavior. For example, we could claim that a person thinks something through, arrives at a decision, and then behaves.
Typically, when research has specifically examined peoples' verbal processes with respect to what they do, feel, or say, the findings have been consistent with that particular lab's theoretical assumptions but not consistent with other assumptions of other labs. Introspection failed one of the basic requirements of "truth."
Freud (if you believe his views) has argued that verbal behavior does not or at best rarely reflects true causes.
Brain structures mediating verbal behavior developed very much later than much of the brain. It is unlikely that no behavior occurred until verbal behavior, and it is unlikely that verbal behavior centers of the brain are in touch with older centers.
(c) Desensitize Phobia Without Verbal
It is relatively simple to remove a phobia without the therapist talking about it or the patient being able to articulate what happened.
Cannot report what is seen but can act appropriately.
A child with a split brain had his left hemisphere exposed to a picture of a chicken claw and the right hemisphere exposed to a snowy scene. When given a set of pictures to match, his right hand (i.e., left hemisphere) chose a chicken while his left hand (i.e., right hemisphere) chose a snow shovel. When asked why the chicken, he said "it goes with the chicken claw." When asked why the shovel, he responded "to clean up the chicken coop." This study revealed the erroneous and tautological foundations of mentalistic psychology. Thoughts cannot be presumed to cause all (and therefore any) behavior. A thinking process (i.e., "oh yes, the shovel will be used to clean up after the chickens") did not determine the behavior of pointing to the snow shovel. Rather, the mental process that the child asserted caused the behavior was created after the fact, even though the child truly believed that the thought caused the behavior.
4. Explanation via Specification of Functional Relationships; or “Correlative”
The proper explanation of a behavior is the specification of the variables which control that behavior and the specification and quantification of the contingencies which modulate the behavior in terms of functional relationships. Functional relationships are the specification of how a behavior changes with changes in the environment. For example, when the red light is on, the behavior occurs, or as the reinforcement rate increases, the response rate increases hyperbolically. This class of explanation documents how elements are interrelated by specifying the functional relations among them. For example, if the magnitude of the reinforcer is increased, then the latency of responding decreases. This relationship specifies that latency is a function of reinforcer magnitude (among other things). These general statements become more quantified and can come to precisely specify an outcome given an input. For example "T = kTn" specifies in what way the average wait time is a function of the required time. At this point we would have a correlative explanatory model. In this sense, any meaningful psychological theorizing must ultimately be correlative in nature.
Unfortunately, the simple specification of the functional relationship (i.e., mathematical or logical relationship) between independent and dependent variables often evolves to some physical or quasiphysical model. What is first a functional description becomes a hypothetical process (not presumed to be a description of an internal intervening process) and eventually an intervening process (an actual internal process presumed to intervene between input and output). (Note that this usage is atypical: It follows Hull rather than Cronbach and Meehl. According to Cronbach and Meehl, a purely hypothetical variable which does not actually intervene between the input and output is labeled an “intervening” variable. A variable which is not simply hypothetical but rather is thought to actually intervene between input and output is labeled a “hypothetical” variable.)
A coherent explanation of a wide variety of behavior is possible by the specification of the functional relationships involved. Within this class of explanation, different areas of psychology can be seen as simply different time scales of adaptation. The impact of a demand for correlative explanations for psychological phenomena cannot be overstated. This change has a revolutionary and fundamental significance by focusing on what is the same about behavior across a wide range of organisms - not only those that contain a "mind." The knowledge obtained is applicable to all organisms not simply humans or not simply autistic children. We come to understand normal and abnormal humans and animals, from rural or urban areas, from one culture or another, younger or older, richer or poorer, doctors or lawyers, etc.
These explanations are sometimes also labeled comparative ("what are the characteristic properties").
i. Historical (time scale of adaptation)
An explanation for a phenomenon by the specification of the functional relationships which were necessary and or sufficient to generate the behavior. For example, the reinforcement history needed to generate an operant or the evolutionary history of a species needed to generate an instinct.
Note that these four items are for "second reading" convenience (they are discussed in Chapter 2, Section III. A. 3. a.
This is the explanation for a behavioral phenomenon by the specification of the functional properties currently controlling the behavior. Actually, the cause is the past experience with the contingencies and the current continuation of those contingencies. For example, a fixed-interval schedule as the explanation for the obtained distribution of behavior, obtained with a fixed-interval schedule.
iii. Potential or Functional
This is the explanation for a behavioral phenomenon by the specification of the functional relationships which characterize how the behavior will change with a given change in a parameter. Again, the cause is in the past, but in this case the past is used to enable a prediction of what would happen under changed conditions. For example, the specification of the dose effect curve as an explanation for the nature of a drug.
b. End State, Functional, or Teleonomic Explanation
If a marble is randomly rolled around in a frying pan with a hole in it, we can say "the marble will roll around until it falls through the hole." We are specifying an aspect of a variable series as its temporal end point. In fact, we could predict that if we repeated the procedure ten times, we would always get the same result. The marble will fall through the hole. It is a common property of all the variable series. We could say that water leaked into a basement because "it finally found" a hole and thereby flowed to a lower level. In the case of the marble and the water, we can make predictions about nature and we can confirm those predictions. But, we accept that neither of these are asserting a sentient or reductionistic process directing the object toward a goal. We are not saying that the water had the intention in advance to move toward the hole and thereby get where it was going, nor even that it was "mindlessly" searching for a lower level. Rather, it means that predictably the water appeared at the hole for whatever reason, and that the water moved to a lower level. We understand the process controlling the water, gravity, and fluidity.
It could be said that a trait exists in the population because it had the result of increasing the reproductive success of its bearer. For example, the vertebrate eye and the mollusk eye (octopus) evolved independently, but they are very similar. This is because the forces which result in variation and selection of the ability to detect visual stimuli are the same in both species. That the two eyes are similar is saying nothing more than that two marbles fell through the same hole in the frying pan. Neither the two eyes nor the two marbles were seeking the same goal, but rather were responding to the same laws.
Explanations of this type are termed functional explanations. The word explanation is applied because to the degree that we correctly understand the laws describing the behavior of a thing, we can predict "where" it will be in the future (the marble will fall through the hole). In some cases we have relatively strong confidence that we can correctly specify the function which relates the behavior to environmental changes and that by using that function we can predict where the future states of the behavior will be. It is only in this sense that can we know the "goal" of the behavior. While we can be satisfied that we understand something about the process and that we have in some way explained something when we can predict its end point, we cannot suggest that the prediction is the prior cause. Additionally, we rarely have sufficient knowledge of the end point or the functional relationship describing past and future occurrences; and our predictions may not be accurate in changed situations.
It is important to remember that our task is to specify why things are the way they are. If a person does something because "he wants to attain a goal" we are then left with the question why he wants to attain the goal. We are interested in how the behavior came about. For example, how does providing shelter for eggs come about, what alters it, and how is it put to rest? Often evolutionary perspectives ask "what purpose does it serve" the object being to discover its impact on reproductive success? Again, however, the better question is "what purpose did it serve?" We know that it exists and therefore it's a good guess that it's the result of its impact on reproductive success but we don't know that for sure. We may use the possibility as a beginning of a research program, but it is not an explanation. It is a description of our estimate of the common elements in past observations of functional relationships and in that sense is a tautology.
These explanations are sometimes also labeled teleonomic ("what ends do they serve"). A teleonomic explanation is similar to a teleological explanation in that its point of reference is an end state or a future behavior. The distinction between teleonomic and teleological is that teleonomic explanations accept that all causes are in the past, whereas teleological explanations consider the cause in the future.
c. Assembly Level of Explanatory Unit
The explanation of some obtained behavioral relationship at some assembly level is explained by appeal to relationships at a different assembly level but which are in the same unit domain. For example, without appeal to different units of analysis such as cellular or participational adaptation as an explanation for behavioral adaptation, we could explain a behavior by appeal to some different assembly level.
We could appeal to functional relationships occurring in smaller groupings or shorter time units. For example, we could appeal to the events occurring immediately before reinforcement to explain the higher rate maintained by a VR schedule than a VI schedule. It could be argued that it is more likely that reinforcement follows a later response in a burst of responses than an initial response in that burst, or an isolated single response. As a result, if bursting occurs, then it is more likely to be reinforced in a VR schedule than in a VI schedule.
Alternatively, we could appeal to functional relationships occurring in larger grouping or longer temporal unit. For example, we could appeal to the difference in the overall rate of reinforcement in a VR schedule under higher and lower response rates. A VR schedule results in higher reinforcement rates with higher response rates, while a VI schedule provides relatively equal reinforcement rates for all response rates.
d. Conceptual Follow-up: Mentalistic Versus Correlative (Behavioral) Explanations
A frequently invoked reductionistic metaphor for how an organism comes to behave correctly is a telephone switchboard and an operator. For example, a stimulus is presented to the organism. It is said to travel to the processing center where the switchboard operator evaluates the stimulus, decides on an appropriate course of action and activates the appropriate effectors. This metaphor brings great comfort to many students of behavior. For example, in order to explain how an actual telephone operator at the phone company functions, we could say that the operator (outer) receives a stimulus (outer). This stimulus is sent to the operator (inner), who decides what to do and activates the proper response (inner) which causes the operator (outer) to behave (outer) correctly. To argue that an operator knows what to do because the operator knows what to do is patently ridiculous as an explanation. Similarly, to argue that a rat learns a maze because a human telephone operator in the rat's head looks at a cognitive map, or to argue that a child behaves similarly on several tasks because a telephone operator in the child's head looks up a rule in a rule book is silly. The “explanations” are tautologies. The power of a telephone operator explanation to account for baffling empirical findings is an illusion. It is like the cartoon mathematical derivation which jumps over a difficult step with the note “and then a miracle happens.” The question to ask is how does the telephone operator come to behave the way it does. The difference in a behavioral and a mentalistic explanatory strategy is the difference in the wisdom given to the telephone operator (none versus infinite).
Similarly, a cognitive explanation will always appear to make sense out of the behavior of the subject and therefore will always appear to be a better explanation than a behavioral explanation. A behaviorist would argue that explaining how the outer animal behaves with the use of an internal processing center is simply explaining on "credit." The real explanation is simply put off for another day and will eventually have to be paid in full with interest. A critical point of focus in evaluating any theory, therefore, is the degree to which it invokes some power within the organism to decide what to do (e.g., the child uses a rule to decide which alternative to select, the rat retrospectively evaluates the correlation, etc.). To the degree that the behavior of an inner rat decides the behavior of the outer rat, the explanation is empty, and is, in fact, silly. Spending explanatory capital that you do not have requires that you then focus your effort on paying back your debt. Just what does it mean to say that: an inner rat is consulting a map, an inner child is using rules, or an inner switchboard operator is deciding what the outer switchboard operator should do.
In the same way mathematical models “fudge” over unknowns with free parameters (they can take on any value necessary to make the prediction work), so too can theoretical internal processes be used as free parameters. If a person uses a strategy to encode and decode information, then we have at least four free parameters or places where we can come up with whatever excuse is necessary to explain the obtained results. The person may or may not have had the correct encode strategy, they may or may not have used the encode strategy, they may or may not have the correct decode strategy and they may or may not have used it. If we add "wanting" to use (i.e., the person had the knowledge, and had the strategy, they just didn't want to use it), and "inhibition" (i.e., the person had the knowledge, and they had the strategy, and they wanted to use the information; it was just that they had an overactive strategy(or knowledge or wanting) inhibition center)) we would have more than enough free parameters to "explain" anything. Mathematical models declare their number of free parameters and lose credibility as they increase in number. Theories should be equally obliged to declare their degrees of freedom and be willing to be evaluated in that light.
The entire issue is brought into sharp focus with the principle "smart animals prove the experimenter stupid, stupid animals prove the experimenter smart." What this principle means is that we as professionals must know what causes behavior, not simply come up with impressive names for it. For example, we could presume that the herring gull was smart because it knows that it must retrieve its eggs when they get bumped from the nest in order to keep them from dying and in order to preserve the species. We could give the bird any number of complex realizations, processing centers, or divine inspirations. We could try to impress our colleagues by showing how smart the birds were. However, relatively straightforward research which varied the color, speckle pattern and size of artificial eggs showed that eggs were retrieved in the order green > yellow > brown > blue; more speckles > less speckles; and large > medium > small. This showed that stimulus conditions governed egg retrieval and that some unnatural stimuli worked better than natural stimuli (which were, in fact, brown, moderately speckled and a medium size). Similarly, a snail could be said to be smart because it knows to climb to the top of a tree in order to get to the most tender leaves. However, research has shown that the snail moves so that its shell pulls "back" (i.e., negative geotropism).
A herring gull can be made to retrieve very large pieces of green highly specked wood more than its own eggs and a snail can be guided toward the worst leaves by pulling on its shell. Both of these behaviors are inappropriate and in fact very destructive to the individual and species. They are stupid. Simply put, animals do things because of environmental cause not because of optimization. If you know the environmental causes: 1) you can make the animal do something stupid by controlling those causes, and 2) you can prove that you understand the psychological process controlling the behavior.
The Renaissance provides an exceptionally clear example of the importance of understanding empirical correlative causes of behavior and the vacuousness of even the most impressive sounding internal causation. Human beings react correctly to distance in the environment. They can throw an object to correctly land in a box placed 5 or 50 feet away. They can say "that thing is far away, this thing is close." Before the Renaissance, the "knowledge" of distance was an internal intelligent wondrous skill humans had. Humans reacted correctly to distance because they were smart A little research showed that "perspective" or the convergence of parallel lines in the distance made humans say "that thing is far away, this thing is close" even though both things were equidistant. They were shown to be incorrect (i.e., humans were shown to be stupid). The discovery of the environmental determinants of distance of depth perception (apparent convergence of parallel lines) in Florence in the early 1400s proved Alberti brilliant and changed the world forever.
Send comments/criticisms/speculations to
Date Last Reviewed : November 17, 2002