---

COMMENTARY ON PIERCE AND CHENEY, 3rd ed

Paul K. Brandon

Department of Psychology
Minnesota State University, Mankato

November 2, 2005

CHAPTER	PAGES
Ch 8:		STIMULUS CONTROL
		S-ave typically precedes negative reinforcement as well as punishment, since few aversive stimuli are inescapable and unavoidable. Therefore, I'd rather use the term S-p to denote a stimulus which signals punishment (and is usually also a conditioned punisher) and S-ave to denote a conditioned aversive stimulus, which can have both negatively reinforcing, punishing and eliciting functions.
		I think that footnote1 is a potentially confusing oversimplification that might obscure the specifics of different stimulus functions. Let's just say that the term discriminative stimulus denotes a situation in which a given contingency is in effect. Also note that the terms S+ and S- are also commonly used in place of S-dee and S-delta (See Catania 1999).
	p 210	Note that multiple schedules often involve two or more different reinforcement schedules neither of which is extinction (see p214: behavioral contrast). In this case we'd have to refer to S-dee1 and S-dee2.
	P 212	The "Bird-Brained" Pigeon. We used a similar procedure in the last rat experiment (Stimulus Discrimination) in Intro Behavior Analysis. In our case the schedule was a chain rather than a multiple schedule, but the purpose was the same: to separate responding at the end of the S-delta period from reinforcement during S-dee.
	p 217	Discrimination and generalization may be described as opposite ends of one continuum: the strength of stimulus control as indexed by response frequency. When stimulus control is strong (a large difference in response frequency in the presence and absence of the stimulus) we say that a subject discriminates. When the difference in response frequency is small, we say that the subject generalizes.
	p 220	Terrace's recent work with Rhesus monkeys on counting -- the relative stimulus control by the number of objects is another good example. The monkeys were trained to indicate the greater or less number of objects numbering one to five, and then tested on six to nine to show that in fact the relations greater and lesser had been trained; not just absolute stimulus discriminations.
	p 222	The book is not as clear as it should be; particularly in distinguishing between the cognitive psychological language of information processing and storage and the behavioral language of acquiring and emitting behavior (I had some commentary on it for the last edition of the text, and will redo it). Basically, the behavioral approach to memory is to say that the behavior acquired is reporting events to oneself, and then repeating that report to oneself (literally 'recalling it') at some later time. This recasts the situation in terms of the probability of the reoccurrence of a behavior, rather than using reifications such as 'memory' and 'information'. This approach also eliminates the need for positing intervening covert behaviors to account for recall. This is no more necessary that it would be to account for why any behavior acquired at one time is emitted at some later time.
	p 226	"Remembering" can also be analyzed as the acquisition of a reporting response, and the emission of that response after a period of time has passed. When you actively "notice" something (report it to yourself), you are more likely to be able to describe (report) it later. This is why active studying that involves the same verbal behavior that you want to be able to reproduce later is more effective than simply reading (or highlighting) the material. See Palmer, David C. (1991). A Behavioral Interpretation of Memory. In LJ. Hayes and P.N.Chase (eds.) Dialogues on verbal behavior (pp. 261-279). Reno, NV: Context Press.
	p 228	The ultimate probe on the pigeon's concept of human: In a similar experiment, Malott and Malott (citation?) found that their pigeons appeared to be consistently misclassifying one picture. Following Skinner's dictum that "the subject is always right" they examined the offending slide (which they had classified as not having a human figure in it) more closely and found that the subjects were right; there was a small picture of a person that they had missed, but the pigeons hadn't!
	p 229	One can define a concept behaviorally as a range of controlling stimuli where there is complete generalization within the stimulus bounds of the concept and complete discrimination beyond those bounds. Thus, rather than the usual triangular generalization gradient À there would be a rectangular one . The philosopher Gilbert Ryle (1949) has talked about category errors. One common such error is in assigning a label to a category of events and then treating that label as if it somehow explained those events. Conditional probabilities are at the heart of the 3-term contingency. We say that response B produces consequence C, given that situation A is in effect. Thus, the probability of occurrence of the response is conditional upon the stimulus situation in effect at a given time. One might say that all natural human discriminations are conditional ones. There is never a pure and simple stimulus situation; there is always some context on which stimulus control is conditional.
TOP
Ch 9:		CHOICE AND PREFERENCE
	p 236	As a slightly more rigorous definition of choice: Doing one thing (emitting an operant) when other operants are also possible (are in the individual's repertoire, physically possible, etc). The material on concurrent chains will develop this further. The term choice generally does not refer to a behavior, but to a relationship between two or more operants. The exception is a situation like the Findley procedure where there is an explicit choosing response.
	p 240	One procedure for increasing child compliance is called the Hi P/Lo-P procedure (see publications by Houlihan, Daniel). This is based on presenting a series of prompts for responses which have a high probability of occurrence, followed by a prompt for a behavior that is less likely to occur. This has been shown to increase the likelihood of compliance with the low probability prompts. This is consistent with the predictions made by the matching law (the prompter is increasing the frequency of reinforcement in her presence) and establishing operations (the high probability of compliance prompts establish the prompter as a source of reinforcement, increasing the likelihood of all behaviors reinforced by that person.
	p 257	Note that the formulation of the matching law is based on experiments with uniform responses and reinforcers. This does not happen in the real world of human behavior, making a precise analysis somewhat more difficult. However, when we're dealing with economics (see behavioral economics at the end of the chapter), reinforcers do have standard units. This, of course, includes many human behaviors of great interest to most of us!
	p 257	One could add a whole chapter on behavioral economics. As indicated, Richard Herrnstein is credited with originating this area of analysis. Steven Hursh and Howard Rachlin currently contribute much to this field, as does Fran McSweeney with her analyses of within session changes in the patterns of behavior maintained by reinforcement schedules. One contribution of behavioral economics is the distinction between open and closed economies. When a subject gets most of its food (or other reinforcers) outside of the experimental situation (an open economy) its behavior is quite different from when it gets all of its food under experimental conditions (a closed economy). There have even been a couple of human studies in which subjects lived for a couple of weeks in an experimental setting, with the consequences of most behaviors under experimental control. This term might also be applied to Walden Two type social experiments such as Los Horcones <http://tm.wc.ask.com/r?t=an&s=m&uid= 0FEFAD729EBC0D4F3&sid=13CBC3A5D30D9DE0 4&qid=383F929862F80046811295DDECCE2097&io= 0&sv=za5cb0d8d&o=0&ask=%22Los+Horcones%22&u ip=861d0b6c&en=te&eo= 100&pt=Los+Horcones%3a+A+Walden+Two+communit y&ac=24&qs=0&pg=1&ep=1&te_par=102&te_id=&u=http:// www.loshorcones.org.mx/>. The open/closed distinction is an example of a concept from classical economics being incorporated into behavior analysis, as is the concept of elasticity, which can be applied to the analysis of addictive drugs.
	p 262	The Quantitative Law of Effect. I don't expect you to be able to reproduce this math on tests. Its importance is as an illustration that behavior analysis can make some specific, quantitative statements about behavior -- go beyond a simple verbal narrative. It is also a reminder that behavior must always be analyzed in context!
TOP
Ch 10:		CONDITIONED REINFORCEMENT
	p 273	Another issue in backward chaining is the question of whether the verbal prompts necessary to produce the behavior are in the subject's repertoire. If they are, then we can use forward chaining. If not -- either because the subject lacks the necessary verbal repertoire or because the language lacks the necessary verbal responses -- than backwards chaining is the best choice. This is also related to the distinction between stimulus control and direct reinforcement control which will be developed in the next chapter.
	p 276	The distinction between the new response and the existing response methods of assessing the effectiveness of conditioned reinforcers boils down to the choice of a behavior to be reinforced. Do we use a conditioned reinforcer to shape a new behavior, or do we use it to maintain a behavior already in the subject's repertoire? The latter is often easier to accomplish. To use the Extinction experiment in Psych 207 as an example, as we did it we extinguished an existing response (the lever press) by removing both the primary (water) and conditioned (click) reinforcers. To make it an existing response test of a conditioned reinforcer we would simply empty the water dish (thus eliminating the primary reinforcer) but continue to operated the dipper so that the conditioned reinforcer of the click remains. The first project in this course, on the other hand, is an example of the new response method, since you will add responses to the chain be using previously established S-dee's as conditioned reinforcers.
	pp 286ff	Generalized conditioned reinforcers and social behavior. The authors are less clear than they might be here.Let's start with the concept of generalized conditioned reinforcers. First analyze the term (to analyze something is to break it down into its component parts). Reinforcers are consequences of behavior which maintain or increase that behavior. Conditioned reinforcers are effective because of their relationship with other reinforcers. Generalized (conditioned) reinforcers are effective in many situations (can reinforce many behaviors) because they are not dependent upon any particular establishing operation. This is true because they are maintained by their relationship to many different backup reinforcers. The text gives the classic examples. To apply this to human social interactions we must first look at the characteristics of generalized reinforcement. It is very consistent; always effective; not dependent upon specific situations or conditions. Therefore, if social interactions are maintained by generalized reinforcement they will be more consistent. If you are reinforced for maintaining a social relationship with someone, you will do so in most situations; not just when you are hungry or thirsty. This is important since much of our behavior is social in nature (involves the behavior of other people). Societies would be far less stable if they wsere based on something less predictable than generalized reinforcement!
	p 287	The behavior concept of countercontrol parallels the concept of checks and balances in the U.S. Constitution. In fact, the Constitution can be looked at as an attempt to establish and institutionalize countercontrol so that no portion of society can exercise disproportion control over another.
	p 290	Another advantage in applied settings of money or other tokens over primary reinforcers is the generalized conditioned reinforcers are less disruptive to ongoing operant behavior than is primary reinforcement. You don't stop to eat tokens!
	p 292	Natural or generic reinforcement contingencies refers to consequences of behavior that are not mediated by the intervention of a specific person acting on some plane.
		In other words, we can predict that they will occur whenever the specified behavior occurs in normal situations. The subject's normal environment will maintain the behavior once it is in the repertoire.
TOP
Ch 11:		CORRESPONDENCE RELATIONS
	p 295 ff	Note: Correspondence Relations refers to a particular type of stimulus control: one where the discriminative stimulus is itself the behavior of an individual, which exerts stimulus control (often of a point-to-point nature) over the behavior of another individual. Since the changes in the second individual's behavior are not always immediate (particularly in the case of rules) and work in conjunction with other sources of stimulus control we must also include function altering events (see p 320) in this discussion. Note also the danger of confusion with correspondence theories of truth (not something that we've concerned ourselves with in this course ;-): A statement may be considered true or justified in some sense if it corresponds with either other observations or some absolute criteria of the state of events. This is more an issue for the philosophy of science. It does arise when we consider private events (such as thoughts) in the next chapter.
	p 298 ff	Robert Epstein's work is portrayed in the video Cognition, Creativity and Behavior shown in IntroBA. His picture in the text is not current.
	pp 304ff	Generalized imitation can be looked at as a higher order response class acquired through induction. That is, as one is reinforced for imitating many different things, imitation itself has become a class of responses established and maintained by a history of (usually intermittent) reinforcement. We call it a higher order class of responses since it is a class that is itself made up of smaller classes of responses (individual types of imitation). If imitation itself is a response class, then we don't need to have a reinforcement history of imitating any given act or model in order to do it. This also leads to the question of whether induction itself is a phylogenetically determined process. That is, have we evolved the capacity to acquire general categories of behavior as a result of acquiring many related individual behaviors?
	p 312	Expectancy is not a behavioral construct, since it puts the cause after the effect. Rather than talking about expectancies, talk about a history of reinforcement for related behaviors in related situations. It may not seem as straightforward, but it is more logically defensible. It's the statement of the future probability of an event based on history.
	p 312	Promised rewards may be interpreted as establishing operations.
	p 314 ff	Precurrent behavior is actually a term originally coined by J.R. Kantor a decade or two before Skinner (a colleague of his) used the term. It refers to a behavior that makes it possible for the individual to emit another behavior, and is often used in the context of problem solving. One might look at precurrent behaviors in a behavioral chain; the chain cannot continue (and the primary reinforcer be obtained) unless certain behaviors occur first and set the occasion for the later behaviors. We may postulated that private events (thoughts) sometimes have this function.
	p 315 ff	Rule and contingency governance are not a dichotomy but rather a continuum. This is where the concept of higher order response classes comes in handy. Following rules is itself maintained by reinforcement contingencies (see my previous discussion). Therefore, the question is what specific type of reinforcement contingency is maintaining behavior; not rules or contingencies. In fact, given an adequately detailed analysis (c.f Brandon and Smasal, 2004); Smasal, 1987) both contingencies can be shown to be simultaneously in effect. Another point is that rule governance usually concerns the acquisition of new behaviors. Reinforcement contingencies are still needed to maintain behavior. Rules are often characterized as verbal discriminative stimuli. This is too simple. For a rule to be a verbal SD it would have to produce an immediate change in behavior. This often is not the case. Rules are often better described as function altering stimuli; they change the way in which a stimulus controls behavior. The rule tells you what to do in a given situation; your behavior doesn't actually change until you are in that situation (which is the actual SD).
	p 316 ff	Pliance and tracking are the listener's equivalent of manding and tacting for the speaker (this will be developed in the next chapter).
TOP
Ch 12:		VERBAL BEHAVIOR
	p 324ff	The term language leads us to regard words as things or objects that exist independently of behavior, rather than as behaviors which have separate response functions for speakers and discriminative stimulus functions for listeners.
	p 325	Chomsky vs. Skinner They are addressing different problems not really arguing with each other.
		Chomsky is interested primarily in the structure of language (although he has shifted a a bit recently). He is interested in identifying inherited mechanisms which make it possible for people to acquire language behavior. Skinner. On the other hand, was interested in the function of language. His question is how do people acquire language?. What is the function of language; why do people say one thing rather than another?
	p 326	While the authors start out by listing the most common categories of verbal behavior, the definition is a functional one. Verbal behavior is any behavior that is reinforced by a change in a listener's behavior through the mediation of a common intraverbal repertoire (words that are events having the same stimulus and response functions for the speaker and the listener). This definition should be credited to C.B. Ferster (1974). Another way of stating this is to say that the speaker and listener should be able to exchange roles. Thus, a dog who could respond appropriately to many verbal commands but could not emit them herself would not be said to be behaving verbally.
	p 326	Rule governed behavior might be defined as "behavior under the stimulus control of verbal behaviors, reinforced generically as a functional member of a higher order response class." In other words, the immediate reinforcement for following rules is not the outcome of following a specific rule, but rather the (intermittent) social reinforcement which maintains the broader class of behaviors which we label following rules. If we have a history of reinforcement for following rules (being under the control of verbal discriminative stimuli) we will do so even when they conflict with immediate reinforcement contingencies. Bear this in mind in the second project, when analyzing cases where your subjects' behavior may not be that which is predicted by the reinforcement schedule in effect.
	p 327	Some recent research (the article was in Science) indicates that gestures preceded vocal language. The blind use gestures even when talking to another person that they know is blind. From an evolutionary point of view bipedalism (which we know preceded speech and large brains) may have allowed a richer repertoire of gestures and thus set the stage for language.
	p 331	Skinner himself never did any research on verbal behavior; when he wrote the book no one had done any systematic data collection in the context of antecedents and consequences. This is why Verbal Behavior is best viewed as a statement of testable hypotheses. Rule governance may be looked at as a higher order class of responses (a class who members are themselves classes of responses).
	p 333	Tacting might also be looked at as a higher order response class. A reinforcement history of verbal responding under the stimulus control of nonverbal events is another way of describing the generic reinforcement contingencies which maintain this class of behavior.
	p 33x	We sometimes talk about the intraverbal repertoire; that is the the individual's repertoire of verbal responses ("words") and their dual response and stimulus functions. Again, this may be looked at as a defining characteristic of verbal behavior (see page 296).
	p 335	See my taxonomy of verbal behavior on the Behavioral Basics sheet.
	p 339	See my earlier comments about Chomsky and Skinner.
		You can see Jack and Jill in the film Cognition, Creativity and Behavior (video). BF 311 .C6 1982 (shown in Psych 207). This demonstration illustrates some other interesting phenomena such as the same person being both speaker and listener, and memory and note-taking.
	p 345	Chimpanzees have been taught to tact visual illusions, such as telling whether a rotating spiral is expanding or contracting (how do you think that this was accomplished?). Some other questions: Why is the tacting of private events inherently less precise than the tacting of public events (hint: how is tacting taught)? What is the name of the science that is based on the tacting of private events?
	p 347	Echoic behavior is not simply the duplication of sounds; the voices of the speaker and listener may be quite different. It would be better to say that the relationship is between functionally equivalent verbal stimuli of the same modality. Similarly, echoic and textual refer to different forms of correspondence relationships between stimuli and responses. The difference is not a basic one.
	p 34x	Saying and pointing are not necessarily functionally different. The statement give me that toy on the TV is a functional equivalent to pointing. Steven Hayes has introduced the terms pliance and tracking as the listener's counterparts to manding and tacting, respectively. Again, tacting can be analyzed as a higher order response class.
	p 34x	The interdependence of mand and tact functions may be due to the membership of words in equivalence classes.
	p 351	Ronald Schusterman (JEAB 19??) has demonstrated equivalencing in sea lions. There is a relationship between the degree of retardation and whether equivalencing can be acquired.
	p 351	According to Murray Sidman, equivalencing is a basic process (like reinforcement) and cannot be derived from more basic processes. This would be the first qualitative difference acknowledged by behavior analysis between humans and other animals. As the text indicates, there is some controversy as to whether equivalencing has been shown in subjects other than humans. There is some disagreement about this.
	p 352	Equivalencing and the emergence of untrained relationships is the key to the power of human language. Think about it (talking to yourself about it -- see the section on the same person being both speaker and listener).
	p 352	Skinner added another class of verbal responses : autoclitics. These are verbal behaviors whose function is to modify the effect of other verbal behaviors. Much of grammar falls into this category.
	p 352	A review of Moerk's work can be found in Salzinger, K. (1994). The lad was a lady, or the mother of all language learning: A review of Moerk's First language: Taught and learned. Journal of the Experimental Analysis of Behavior, 62, 323-329.
TOP
Ch 13:		APPLIED BEHAVIOR ANALYSIS
	p 359	Applied Behavior Analysis and Experimental Behavior Analysis are a continuum, not a dichotomy. Epling, Pierce, and Cheney fall someplace in the middle of this continuum in terms of their work.
	p 360	In terms of behavior analysis, there is no 'abnormal behavior', only abnormal contingencies. As the authors point out, with the (uncommon) exception of physical abnormalities due to disease or developmental defect, behavior is a lawful result of contingencies.
	p 362	Again, "rules vs. Contingencies" is not a dichotomy, since rule following is itself a behavior maintained by contingencies. The real question becomes a specification of the particular contingencies controlling a particular behavior. Are they the immediate consequences of the behavior, or are they verbal discriminative stimuli maintained by more remote reinforcement? What this discussion does highlight is the need for natural contingencies to maintain behavior. As I've pointed out before, rule following is itself a higher order response class which requires both a history of intermittent reinforcement and current reinforcement contingencies to be maintained. The real point is therefore not whether rules can change behavior; it's the need for some kind of natural contingency to maintain the changes produced by rules.
		------
		That's "Dr. Don Baer".
	p 364	There are two definitive texts on behavioral-analytic research. One is Murray Sidman's classic Tactics of Scientific Research (1963). The other is J.M. Johnston and H.S. Pennypacker's Strategies And Tactics Of Behavioral Research (2nd ed, 1993).
	p 364ff	These are some examples of the many possible single case research designs. Another one that's common in behavioral pharmacology (among other areasa) is the generation of dose response curves. Here, we demonstrate that as we systematically change the dose of a drug, some defined behavioral effect changes in a functionally related manner.
	p 369	There's a good quote from a couple of centuries ago on the effectiveness of lectures: Lectures were once useful; but now when all can read and books are so numerous, lectures are unnecessary. If your attention fails and you miss part of a lecture, it is lost; you cannot go back as you do upon a book.. Samuel Johnson (1799), cited in Boswell, p 462
	p 370 372	Celeration charts are logarithmic; that is they are based on ratios. They are usually used to describe changes in fluency; a joint function of rate and accuracy.
	p 373	There are limits to "the learner knows best". Yes, her behavior is lawful, but see Morris earlier on context. The fact that what a learner says is a lawful function of identifiable conditions does not necessarily make her statement a correct answer!
	p 374	Cultural replacement: Where do the students in the halls of Trafton come from?
	p 375	A reminder: Choose does not refer to an action; it's a description of an allocation of behavior across situations.
		See page 314 for an account of precurrent behavior.
	p 376	There is some question as to whether there is really any such thing as "self-reinforcement" in the sense of directly delivering reinforcers to oneself. The concept of contingency requires some separate event for reinforcement to occur. At best, one might present verbal conditioned reinforcers to oneself (see the speaker as listener under verbal behavior). These would still have to be backed up by primary reinforcers to remain effective.
	p 377	The term Applied Behavior Analysis is becoming a generic label for a specific set of procedures which apply Behavior Analysis to the treatment of autism. Strictly speaking, it's the application of behavior analytic principles to develop a set of contingencies which will produce an expanded set of social and verbal repertoires in those individuals labeled 'autistic'.
TOP
Ch 14:		LEVELS OF SELECTION
	p 391	Note that selectionism is based on a very different type of explanatory account from that typical of physics and chemistry. We are not looking for intervening mechanisms to account for behavior (that is the complementary science of neurophysiology). Rather, we are looking for historical explanations; descriptions of the conditions which caused a particular behavior to occur in a particular situation. This is the kind of account used in the study of evolution. Darwin described the basic system of natural selection long before we understood the genetic mechanisms involved. Darwin was not even aware of Mendel's work. Also note: Darwin did not disover or originate the Theory of Evolution; he described one theory of how the evolutionary process (widely accepted in his time -- his grandfather Erasmus Darwin wrote on that topic) worked. That is the process of Natural Selection.
	p 391	One can also draw a parallel between the processes of artificial selection and coercion. In both cases, selection is for the benefit of the selector; not the species or behaving organism whose characteristic is selected.
	p 392	While Skinner introduced the parallel with evolution in Science and Human Behavior (1953), he more formally developed the analysis in Selection By Consequences. Science, 1981, 213, 501-504.
	p 392-3	One can distinguish between contingencies and metacontingencies (cultural practices) by looking at the level on which fitness and propagation operate. For behavioral contingencies, it is the individual behavior which is selected; made more likely to reoccur. A metacontingency, on the other hand, while still dealing with behaving individuals, does not produce its defining consequences on the level of the behaving individual. Rather, a metacontingency is propagated (continues to occur) because it makes the culture (group of individuals) who practice it more successful (see also pp 410ff).
	p 393	Gould's use of the term contingency in his book Wonderful Life (1989) is rather different from ours. Gould is saying that the course of evolution was contingent upon very specific and local conditions. If we restarted the process -- 'rewound the tape' in his phrase - we might get a very different result. Thus, he is saying that the course of evolution was not predetermined. The book is a very good read, by the way.
	p 394	Even monozygotic (identical) twins show some physical variations (as do clones). A number of factors are involved: mutations, gene expression, differences in the prenatal environment, etc.
	p 394	The common point in natural selection and reinforcement is propagation: some aspect of the organism is made more likely to reoccur.
	p 397	FAPs illustrate the difference between structure and function in the analysis of behavior. While they look like operants in terms of complexity, they function like respondents in that they are determined only by antecedents, not by consequences. Once a FAP is modified by its consequences, it has become and operant.</