I would like to know if the principles derived from the experimental analysis of animal behavior usually apply to normal, healthy adults in real-world settings. This essay is the second report in Reward and Consent of a controlled experiment that uses aversive stimulation to suppress the behavior of normal, healthy adults. See my previous post, The Electric Shock of Smoking, for a brief review of Powell and Azrin (1968). Much study is needed to be able to claim that the science of human behavior is directly related to animal behavior.
Beginning behaviorists can read this paper. For each bit of behavioral jargon found in this essay, there is a context in lay language. It is not necessary to understand all the details of the electrical equipment described below, but it's included for readers interested in the kind of tinkering that B.F. Skinner and his followers employ with the gadgets they were inventing and collecting.
In a type of experiment they called "behavioral engineering," Azrin et al. (1968) adhered an apparatus to the upper backs of employees at Anna State Hospital in Illinois, and under the normal experimental conditions, it sounded a noise whenever it detected them slouching.
The subjects could wear the invention anywhere during their normal waking routines, but in this experiment they wore them on the hospital grounds. They installed the contraption under their outer garments.
To begin with, they taped the ends of an elastic cord horizontally across their second thoracic vertebrae and attached a "snap action switch" to the cord. They also ran a wire out from the switch, under their arms, around to their chests, and into a tone-signaling device, a transistor circuit with a hearing aid speaker that was hanging from a cloth ribbon around the neck. Under the usual conditions, whenever they curved their backs into a sustained slouching position, a loud noise was heard. You can see a picture on page 101 of the free .pdf report.
They defined slouching as "an uninterrupted closure of the switch for at least three seconds." When they rounded their backs, they stretched the chord. This activated the switch and it made a clicking sound. During three seconds of continuous back-rounding, the switch remained closed. After three seconds the switch opened and the hearing aid speaker emitted the 55-decibel, aversive tone which indicated they were slouching.
While the tone was on and with each subsequent slouching event, they measured the cumulative duration of total slouching by adding a time meter to the device. Electric current passing through the meter caused the slow movement of an electrolyte gap between two columns of mercury. They only allowed the current through the meter after the rounding of the backs had closed the posture switch for at least three seconds. The location of the moving gap between the two columns presented a miniature image of the running total of slouching-time that had elapsed. They photographed the images. When they erected themselves, the band relaxed, the switch opened, the tone stopped, and the gap stood still until the next trial.
They called this study an example of behavioral engineering because they not only defined slouching in behavioral language, "a rounding of the back for at least three seconds," they also defined it with physical properties that were detectable by electromechanical equipment. Rounding was "an increased distance between two points on the back." A curve in the back meant a stretch in the skin and a subsequent stretch in the elastic cord attached to the skin. With each stretch, the terminals of the cord spread further apart causing the tension that closed the switch. Three seconds of continuous stretching generated the noise and an electro-mechanical-behavioral chain of engineered responses took place.
Twenty-four healthy adults participated. They also included a patient who was living inside the psychiatric hospital. They were fully informed about what would take place, but only six received written instructions. Azrin and his colleagues did not say whether or not they all participated voluntarily, but some of them asked if they could join the study after they saw the benefit of improving their postures. We do not know if their continued employment in the hospital depended upon their participation.
Here are the stages, conditions, and results of the experiment. In the first stage, twenty-three of them wore the apparatus under two different conditions. (A) During the baseline, slouching did not produce the tone and (B) during the experimental condition, slouching did produce the tone. There was also (A) a return to the baseline condition after the tone condition. For all twenty-three subjects, the amount of time slouching in condition (B) was less than the amount in the first baseline condition (A). In the second stage, a control group of the remaining two subjects wore the apparatus under opposite contingencies. (1) During baseline, good posture did not produce the tone, but (2) during the experimental condition, good posture did produce the tone. They both slouched more and straightened their backs less in condition (2) than in condition (1). We can conclude that the noxious tone had suppressed the behavior it was contingent upon, whether or not the behavior was slouching in the first stage or keeping a good posture in the second. Punishment was in effect, but to give you a better understanding of the full meaning behind this conclusion, it is necessary to elaborate on some of the principles of behavior analysis.
It is useful to distinguish between positive and negative reinforcement and positive and negative punishment. "Positive and negative" indicate an addition or subtraction of a stimulus. "Reinforcement and punishment" indicate an increase or decrease in behavior strength. A consequence changes the probability that future responses will reoccur.
There are four categories of contingencies: (1) positive reinforcement occurs when a reinforcing stimulus is added after a behavior and the probability of behavior increases, (2) negative reinforcement occurs when an aversive stimulus is removed and the probability of behavior increases, (3) positive punishment occurs when an aversive stimulus is added and behavior strength decreases, and (4) negative punishment occurs when a reinforcing stimulus is removed and strength decreases. Notice how negative reinforcement is different from positive punishment. Lay people often say negative reinforcement when they should be saying punishment.
The field of applied behavior analysis has some inconsistent terminology for these contingencies and stimuli, but it becomes clear once you get the knack of it. The context usually provides enough clues for all of us to be clear about what we are talking about. Regarding the contingencies, they sometimes call positive reinforcement "reinforcement," without qualification as positive or negative, but they always call negative reinforcement "negative reinforcement." They often call positive punishment "punishment," also without qualifying it as positive or negative, but they have been calling negative punishment "penalty," "response cost," or "negative punishment." We can reinforce, negatively reinforce, punish, or penalize a response. I haven't found a good verb for "negatively reinforce." Regarding the stimuli, there are generally two kinds that can affect behavior strength, a reinforcer or a punisher. The same reinforcer that is added to reinforce a behavior can be also be removed to penalize one. A punisher that is added to punish a behavior can also be removed to negatively reinforce one. We can use "punisher" and "negative reinforcer" interchangeably as two synonyms that describe the same consequence, but it's not necessary to say "negative reinforcer." If you read Skinner, he used it a lot. I often use "punisher" as the noun for "negative reinforcer." Skinner also called it an "aversive stimulus." (See the comments below this post. One reader, Hannah, suggested a correction of these previous three sentences. Her comments should also be included in an analysis of behavioral terminology.)
Praise, attention, a slice of pizza, or money can be reinforcers. Punishers might include shame, disapproval, electric shock, cold water sprayed in the face, or sour lemon on the tongue. What is punishing for one person might be a reinforcer for another. A statement of disapproval can reinforce a child who is starved for attention. The first slice of pizza might reinforce the behavior of driving to the pizzeria, but the fifth slice of pizza might be avoided.
Reinforcement and punishment are not opposites. One strengthens the rate of behavior in the long run and the other suppresses it for the time being. Skinner and his peers did not say that a punisher has a sister reinforcer that can add the same number of responses over time that the punisher subtracts.
"Strengthening" or "weakening" are good words for what takes place during reinforcement and extinction. During extinction after reinforcement, behavior strength declines to pre-reinforced, naturally-occurring, low-levels of spontaneous emission, but extinction can take a long time to complete. For more on extinction, see my earlier post, Rat Behavior Extinguished. Ferster and Skinner (1957) withheld reinforcement from pigeons after shaping them up with a very-infrequent, intermittent schedule of reinforcement. This caused behavior to take an extremely long time to extinguish. That is why, according to Skinner, infrequent payouts at the slot machine cause gamblers to pull the levers of the slot machine more often, and that is why gambling is such a big addiction.
"Suppression" is a good word for what takes place under punishment and recovery. During recovery after punishment, behavior strength rises to pre-punished levels, but this generally doesn't take as long as extinction. In Azrin et al. (1968, p. 104, Table 1) after the noise was lifted, slouching quickly returned to higher rates of response. Powell and Azrin (1968 p. 69) electrically shocked smokers when they opened a cigarette case. When they halted the electricity, they found "an immediate recovery of the smoking rate to its pre-punished level." Skinner (1938, p. 154) showed that slapping the paw of a rat initially reduced the amount of lever presses per hour, but the rate recovered when the slapping ceased, and total presses emitted over time did not diminish. (For more on punishment, see my blog post, The Problem with Punishment.)
In Azrin et al. (1968) the tone was possibly an "unconditioned" punisher, because by itself, without being associated with any other punishers, it instinctively suppressed the behavior that produced it, just as it would function for a new-born baby. The click was possibly a "neutral stimulus" that became a "conditioned" punisher after its association with the tone. By hearing the tone soon after the click, the subjects at Anna State Hospital associated the click with the arrival of the tone. Pairing a neutral stimulus in time with a reinforcer or a punisher creates a newly conditioned stimulus. Pavlov discovered this with hungry dogs who salivated to food and did not salivate to a bell before conditioning. By ringing a bell during food delivery and then removing the food, the bell caused them to salivate. This is called "classical conditioning' because it was an early twentieth century form of behaviorism. Skinner called his mid-twentieth century discovery of all the contingencies "operant conditioning" because the behavior "operates' on the environment to produce a consequence.
Punishers not only play a role in punishment and negative reinforcement; they also determine the results in "escape and avoidance" contingencies. Did Azrin and his colleagues discover that the noise caused an escape response to occur? Yes. When slouching produced the tone, erecting the posture was the escape behavior that eliminated it. It was negatively reinforced; good posture was increased by the withdrawal of the aversive stimulus. They escaped from the tone after it went off.
Did they discover that the click caused an avoidance response to occur? Possibly. The click was conditioned as a discriminative stimulus that warned the subjects that the tone would follow unless they avoided it by straightening the back within three seconds after the click. Here they may have avoided the tone before it went off, thanks to the warning stimulus. Azrin et al. (1969) believed the subjects of their experiment had avoided the tone, but only because they told them so. Unfortunately, they didn't invent a mechanism that could record responses during this three-second interval. Behavior analysts consider verbal reports of behavior weak scientific evidence. They depend more upon the direct observation of clearly defined, measurable, and recordable units of behavior. They cannot say with assurance that avoidance took place.
Rounding and erecting the back are two incompatible behaviors. By definition, when one occurs, the other one cannot occur. They are mutually exclusive. Punishment wasn't necessary. A different experiment could have reduced the rate of slouching by reinforcing good posture and withholding reinforcement from slouching. Music playing through a speaker on the apparatus could have been a good reinforcer. Behavior analysts call this contingency the "differential reinforcement of incompatible behavior." Since punishment can cause aggression or escape from the punishing person, this would have been a more ethical intervention.
The change in the tone caused the change in behavior. Translating this into the scientific jargon, we can say the change in the "independent variable," tone on, caused the change in the "dependent variable," a decrease in the probability of slouching as measured by its slouching rate. Algebraically, the slouching rate can be defined as x/(x+y) where x = slouching time and y = non-slouching time. The non-slouching rate is y/(x+y).
Unknown variables can also cause changes in the dependent variable and invalidate the results of the study. If absolutely nothing else in the environment changes between the baseline and the experimental conditions, then they can deduce that the change of the independent variable has caused the change in the dependent variable. Therefore, assuming "ceteris paribus," the Latin phrase for "all other things equal," if all other values were constant, it necessarily follows that only the tone produced the decline in the rate of slouching.
How did they try to control for potential confounding variables? Did feedback from the click tell the subjects they were slouching and cause them to stop slouching? They said no, because the clicking was present in both the tone and the non-tone conditions. What about other variables? Azrin et al. (1969) implied that they had controlled for differences among subjects. The tone reduced slouching in men and women and in normal, healthy adults as well as the psychiatric patient. They admitted, however, that the tone may have embarrassed the subjects as they walked about the hospital grounds and feedback from other people in the hospital may have influenced the results.
It is impossible to control for all variables that might be imposing themselves into the experiment without the knowledge of the scientists. Skinner raised his own underweight, hungry pigeons, and managed their lifelong histories of reinforcement. He put them in soundproof experimental chambers. Control is more of a problem in applications with humans in their everyday settings. In our slouching experiment, somebody at the hospital might have said to one of the subjects, "Gee, I love the sound of that noxious tone that's coming from under your shirt, and it really makes you attractive." Maybe that person caused a unknown blip in the slouching rate and ruined the validity of all that work by Azrin, Rubin, O'Brien, Ayllon, and Roll. Maybe they should have performed the experiment on a group of college students taking a class of applied behavior analysis. Maybe they should quit the insane asylum for more predictable results.
How prevalent is the evidence that the behavior patterns of normal, healthy adults can be conditioned by consequences? I have yet to see a case to the contrary, but a more complete review of a body of primary sources is needed. How can we possibly account for each variable that might find its way into the broad spectrum of human affairs?
Skinner asked, "Do we have free will or is our behavior determined by the environment?" A chemist I knew said we'll never know. There is an infinite number of electrons floating about the universe. We would need a computer bigger than the universe to be able to account for them all. I believe, as some of Skinner's followers have said, that the behaviorists have taught us how to influence behavior, if we can't control it completely. Therefore, it is useful to proceed as though we can modify human behavior, as long as we do it ethically.
Azrin, N., Rubin, H., O'Brien, F., Ayllon, T., & Roll, D. (1968). Behavioral engineering: Postural control by a portable operant apparatus. [Electronic version]. Journal of Applied Behavior Analysis, 1, 99-108. Retrieved July 17, 2011 from U.S. National Institutes of Health PubMed Central database Web site: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC1310985/pdf/jaba00084-0003.pdf
Ferster, C. B., and Skinner, B. F. (1957). Schedules of reinforcement. New York: Appleton-Century-Crofts.
Powell, J. & Azrin, N. (1968) The effects of shock as a punisher for cigarette smoking. [Electronic version]. Journal of Applied Behavior Analysis, 1, 63-71. Retrieved July 15, 2011 from U.S. National Institutes of Health PubMed Central database Web site: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC1310976/pdf/jaba00083-0063.pdf
Skinner, B. F. (1938). The behavior of organisms: An experimental analysis (1991 ed.). Cambridge, MA: B.F. Skinner Foundation.
Skoivuma, (2009). Bad posture.jpg. [Electronic version]. Retrieved July 17, 2011 from Wikimedia Commons Web site: http://commons.wikimedia.org/wiki/File:Bad_posture.jpg. Wikimedia Commons says, "This file is licensed under the Creative Commons Attribution 3.0 Unported license. You are free to ... transmit the work.... You must attribute the work in the manner specified by the author or licensor (but not in any way that suggests that they endorse you or your use of the work)." The author does not necessarily endorse me or my work. (Dave, Reward and Consent.)