Two Definitions of Punishment Per Holth i Norwegian Center for Studies on Conduct Problems and Innovative Practice, Oslo Two different definitions of punishment figure in the behavior-analytic literature. In a widespread definition, advocated by Azrin and Holz (1966), punishment is defined as a procedure in which (1) certain responses have consequences, (2) those responses decrease in frequency, and (3) the decrease in frequency occurs because of the response consequence relation, and not for some other reason. Keywords: punishment; theory; Skinner In Science and Human Behavior, Skinner (1953) had espoused another definition of punishment. According to Skinner s definition, punishment is a procedure in which responses are followed by either (a) the removal of a positive reinforcer, or (b) the presentation of a negative reinforcer (or aversive stimulus). The definition requires that positive and negative reinforcers are identified through the respective procedures of positive and negative reinforcement. The Azrin/Holz definition follows the functional-definition pattern of reinforcement by presupposing a particular behavioral result. Thus, according to Catania (1998): Reinforcement and punishment are symmetrical: The former increases responding whereas the latter decreases it (p. 91). A similar functional definition of punishment appears in Baldwin and Baldwin (1981), Donahoe and Palmer (1994), Fantino and Logan (1979), Grant and Evans (1994), Johnston and Pennypacker (1993), Latham (1996), Leslie (1996), Martin and Pear (1996), Novak (1996), Pierce and Epling (1999), and in Rachlin (1976), just to mention a few. Although most current behavior-analytic texts seem to have adopted the Azrin/Holz (1966) definition of punishment without even mentioning the existence of a different one, an exception is the behavior modification textbook by Martin and Pear (1996), in which Skinner s (1953) definition is briefly mentioned. A more distinct exception is Sidman s (1989) book, Coercion and its Fallout: Reinforcement differs in an important way from punishment. We define reinforcers positive or negative by their special effect on conduct: they increase the future likelihood of actions that they follow. But we define punishment without appealing to any behavioral effect: punishment occurs whenever an action is followed either by a loss of positive or a gain of negative reinforcers. This definition says nothing about the effect of a punisher on the action that produces it. It does not say that punishment is the opposite of reinforcement. It does not say that punishment reduces the future likelihood of punished actions. (p. 39) Two Theories of Punishment Apparently, the two different definitions of punishment largely coincide with two different theories of punishment. According to the first theory, reinforcement and punishment are not symmetrical. In the most extreme version of this view, which can be traced back to Thorndike s work in the early 30 s, punishment has no direct effect at all upon the rate of responding. According to Thorndike (1932, p. 34), "the wrong tendencies are not reduced in strength one jot or tittle by the punishment." In addition to Thorndike s own experiments, early support for such a denial of the effectiveness of punishment in reducing response rates (i.e., that the effect of punishment is simply the opposite of that of reinforcement) came from Skinner s (1938) experiments and from Estes (1944). Skinner (1938) found that when rats lever presses were punished during a limited time period, responding was just suppressed and when the punishment procedure was terminated, response rates increased such that the total number of responses emitted was unaffected by the punishment. Estes (1944) found that response-independent electric shocks suppressed responses about as well as response-dependent shocks. Hence, the results suggested a two-factor competing response theory of punishment according to which punishment is effective only to the extent that it elicits responses that 43
are incompatible with the punished response, or by producing conditioned aversive stimuli from which the organism can escape only by emitting behavior which is incompatible with punished behavior. In accord with such a competing response theory of punishment, Skinner (1953) described three main effects through which punishment could, indirectly, result in a reduced probability of the punished behavior: (i) An aversive stimulus may elicit responses which are incompatible with the punished response (e.g., "When we stop a child from giggling in church by pinching it severely, the pinch elicits responses which are incompatible with laughing and powerful enough to suppress it" [p. 186]). (ii) Previously punished behavior becomes the source of conditioned stimuli which evoke incompatible behavior. (e.g., "When a child who has been pinched for giggling starts to giggle upon a later occasion, his own behavior may supply conditioned stimuli which, like the mother's threatening gesture, evoke opposed emotional responses" [pp. 186-187]). (iii) Behavior which reduces the conditioned aversive stimulation arising from the behavior itself or from concurrent circumstances, will be reinforced. (e.g., following punishment, the subject may not simply stop engaging in the punished behavior, but may also show a marked resistance if manually prompted). Fowler and Miller (1963) conducted an experiment with rats that demonstrated how the elicitation of competing (or facilitating) responses can affect running down an alley to food. Rats that, upon arriving at the goal, were shocked in their fore paws slowed down their running, whereas rats that were shocked in their back paws speeded up. According to the second theory of punishment, reinforcement and punishment are symmetrical processes (e.g., Fantino & Logan, 1979). As pointed out by Fantino and Logan, this symmetrical theory seems to be directly inspired by the symmetrical definitions of reinforcement and punishment. In addition to such a conceptual inspiration, the theory also arose from the fact that reductive effects of punishment are not typically accompanied by the description of any specific observable competing responses (Fantino & Logan, 1979; Rachlin, 1976). Even if the theory of direct reductive effects of punishment wins out, or even if the theory could be demonstrated unequivocally, it does not automatically follow that punishment should necessarily be defined functionally in terms of that effect. Conceptual issues, issues related to parsimony, and certain practical concerns need to be addressed as well. Empirical Versus Conceptual Concerns Conceptual issues are sometimes treated as if they were empirical ones. As Smedslund (e.g., 1991; 1994) has shown, such confusion is widespread in psychology. Unfortunately, it happens in behavior analysis as well, and conceptual analyses are necessary to determine the extent to which propositions are empirical. For instance, using the functional definition of punishment following Azrin and Holz (1966), the issue of effects of punishment is no longer a purely empirical matter because punishment is defined by its reductive effect upon the response rate. Research questions regarding side effects or additional effects of punishment, as well as the question of whether a particular procedure decreases the response rate are still appropriate but, following the direct functional definition, it is as inappropriate to ask whether punishment works as it is inappropriate to ask whether reinforcement works. The Issue of Parsimony 44
When first introducing the functional definition of punishment, Azrin and Holz (1966) mentioned some advantages of their new definition as compared with Skinner s earlier one. For one thing, Skinner s (1953) definition of punishment as the response-contingent presentation of a negative reinforcer or the removal of a positive reinforcer requires that negative and positive reinforcers are actually identified initially through negative and positive reinforcement procedures, respectively. As Azrin and Holz (1966) pointed out, this is virtually never done in studies that involve the punishment designation. Hence, considering punishment alone, a definition similar to the definition of reinforcement, differing only in the direction of the change in response frequency, does seem simpler or more parsimonious. On the other hand, if the issue of parsimony is considered a matter of the total number of terms required, Skinner s (1953) definition is indeed more parsimonious than the Azrin/Holz (1966) definition. As Skinner commented, "we have not used any new terms in posing these questions [what is the effect of withdrawing a positive reinforcer or presenting a negative?] and hence need not define any" (p. 185). The Azrin/Holz definition, on the other hand, requires a definition of at least one new term, the punisher. A distinction between positive and negative punishment, paralleling the distinction between positive and negative reinforcement, requires two new terms: positive punishers and negative punishers. Practical Significance According to Azrin and Holz (1966), a second advantage of their definition of punishment directly in terms of response rate reduction is that it avoids an encouragement of hypotheses in terms of escape responses in order to explain the reduction. If punishment, like reinforcement, is defined directly in terms of a particular effect on behavior, it certainly appears moot to try to hypothesize what happens in terms of more basic phenomena. As Fantino and Logan (1979) pointed out: Usually, competing responses are not observed. Instead, occurrence of a competing response is inferred from the absence of the punished response... Thus such theories along with other theories postulating mediating variables tend to be less accountable to the data... It appears that while competing-response theory may provide insight into isolated cases of response suppression, it has little utility as a general theory of punishment. (p. 256) However, Skinner s (1953) definition of punishment as the response-contingent presentation of a negative reinforcer or the removal of a positive reinforcer is not unequivocally committed to a theory of punishment exclusively in terms of the occurrence and strengthening of competing responses. Competing responses are not always observed during the application of punishment procedures. However, when such competing responses do occur, any direct effect of punishment is presumably what remains to be explained when the effect of competing responses on the reduced probability of responding is subtracted. Rather than being committed to any particular explanation of punishment, or even a particular effect, using Skinner s definition, one is free to ask: "What will happen if...?" The practical significance of all this is that it seems very likely that when punishment is not by definition focused directly on a reductive effect, researchers and practicians alike may be more alerted to a potential diversity of effects of punishment procedure (e.g., Sidman, 1989). Conclusion In conclusion, both definitions are possible and there appear to be some advantages to each of them. The Azrin/Holz (1966) version avoids having initially to identify the positive and negative reinforcers through reinforcement procedures in order to specify the corresponding punishment 45
procedures. Designing a punishment procedure strictly according to Skinner's (1953) definition is obviously more cumbersome. Further, the Azrin/Holz definition may effectively diminish unproductive hypothesizing regarding unobserved processes. On the other hand, Skinner s definition may be more apt to foster important questions regarding additional effects when punishment procedures are being applied, such as what kinds of behavior are then likely to emerge instead. Skinner's additional interpretation of effects of punishment may be helpful in this regard even if not always relevant. Answers to questions about the general use of punishment in human affairs may not depend solely on empirical information about their effectiveness but also on conceptual concerns, such as how punishment is defined in the first place. It seems important, then, that behavior analysts are familiar with both of the two main behavior-analytic definitions of punishment. References Azrin, N. H., & Holz, W. C. (1966). Punishment. In W.K. Honig (Ed.), Operant behavior: Areas of research and application (pp. 380-447). New York: Appleton-Century- Crofts. Baldwin, J. D. & Baldwin, J. I. (1981). Behavior principles in everyday life. Englewood Cliffs: Prentice-Hall. Catania, A. C. (1998). Learning. New York: Prentice Hall. Donahoe, J. W., & Palmer, D. C. (1994). Learning and complex behavior. Boston: Allyn & Bacon. Estes, W. K. (1944). An experimental study of punishment. Psychological Monographs, 57, 1-40. Fantino, E., & Logan, C. A. (1979). The experimental analysis of behavior: A biological perspective. San Francisco: Freeman. Fowler, H., & Miller, N. E. (1963). Facilitation and inhibition of runway performance by hind- and forepaw shock of various intensities. Journal of Comparative & Physiological Psychology, 56, 801-805. Grant, L., & Evans, A. (1994). Principles of behavior analysis. New York: HarperCollins. Johnston, J. M., & Pennypacker, H. S. (1993). Strategies and tactics of behavioral research. 2 nd. ed. Hillsdale: Lawrence Erlbaum Associates. Latham, G. I. (1996). The power of positive Parenting. Logan: P & T inc. Leslie, J. C. (1996). Principles of behavioral analysis. Amsterdam: Harwood Academic Publishers. Martin, G., & Pear, J. (1996). Behavior modification: What it is and how to do it. Upper Saddle River: Prentice Hall. Novak, G. (1996). Developmental psychology: Dynamical systems and behavior analysis. Reno: Context Press. Pierce, W. D., and Epling, W. F. (1999). Behavior analysis and learning. Upper Saddle River: Prentice Hall. Rachlin, H. (1976). Behavior and learning. San Francisco: Freeman. Sidman, M. (1989). Coercion and its fallout. Boston: Authors Cooperative. 46
Smedslund, J. (1991). The pseudoempirical in Psychology and the case for psychologic. Psychological Inquiry, 2, 325-338. Smedslund, J. (1994). What kind of propositions are set forth in developmental research? Five case studies. Human Development, 37, 280-292. Skinner, B. F. (1938). The behavior of organisms. New York: Appleton-Century-Crofts. Skinner, B. F. (1953). Science and human behavior. New York: MacMillan. Thorndike, E. L. (1932). Psychology and engineering: The science of mind and behavior, and its relation to the principles and practices of human management. Mechanical Engineering, 54, (1932), p. 30-34. Author s note: Please address correspondence to: Per Holth The Behavioral Center P.O. Box 1565 Vika, 0118 Oslo, Norway e-mail: per.holth@atferd.unirand.no Tel.: Office: + 47 24147907 Home: + 47 64945045 Advertising in The Behavior Analyst Today Advertising is available in BAT. All advertising must be paid for in advance. Make your check payable to Joseph Cautilli. The copy should be in our hands 3 weeks prior to publication. Copy should be in MS Word or Word Perfect format and advertiser should include graphics or logos with ad copy. The prices for advertising in one issue are as follows: 1/4 page: $50.00 1/2 page: $100.00 vertical or horizontal Full Page: $200.00 If you wish to run the same ad in all four issues for the year, you are eligible for the following discount: issue 1/4 Pg.: $40 - per issue 1/2 Pg.: $75 - per issue -vertical or horizontal Full Page: $150.00-per For more information, or place an ad, contact: Joe Cautilli via e-mail at jcautilli@cctckids.com or by phone at (215) 462-6737. 47