Note: The following article was originally published in the Spring 2016 issue of BAQ, Vol. 2(2).
By Benjamin N. Witts, PhD, BCBA
St. Cloud State University
The layperson might be confused at terms like negative reinforcement and positive punishment. “How can reinforcement be bad?” they ask, along with “What exactly is ‘good’ punishment?” The problem, as we know, lies with the terms negative and positive and the misconstruing of bad and good, respectively.
Furthermore, the behavior analyst recognizes that we cannot put value labels on reinforcement and punishment. The racist’s behavior persists because of reinforcement and the child refrains from touching hot surfaces perhaps because of a history of punishment for doing so. Here reinforcement supported “bad” behavior while punishment supported “good” (or rather, safe) behavior, respectively.
So we come to our typical procedural definitions of positive and negative reinforcement and punishment. Positive means to add something, negative means to take it away. Reinforcement is seen subsequently when members of an operant class increase in frequency (or maintain) because of the consequence, and we see punishment when members of an operant class decrease because of the consequence.
But not all follow the same guide in deciding if a particular arrangement was positive or negative, reinforcement or punishment. B. F. Skinner formulated these four processes around just two of them: positive reinforcement and negative reinforcement. Put simply, Skinner did not see punishment as necessarily its own process, but as derivative of reinforcement. This seems odd, but here’s how he did it.
Formulating Reinforcement and Punishment
A concern stems from whether we are talking of procedures or processes. We tend to mix the two when we speak of reinforcement. In the definitions of positive and negative above, I spoke of the procedure. In defining reinforcement and punishment, I spoke of the process. We might do well do formulate our definitions in terms of only one level—process—as Skinner did.
For Skinner (e.g., 1953), any stimulus presentation that, when following some response, led to an increase in responding was considered to be a positive reinforcer. A stimulus that would serve as a positive reinforcer would, upon occasion of removal, result in negative punishment. Here Skinner spoke of negative punishment on the grounds of positive reinforcement. Anything that resulted in preventing the stimulus’ future removal would reinforce that behavior. Thus, for Skinner, punishment does not lead to the cessation of behavior, but rather the support of alternative behavior.
Similarly, any stimulus that when removed would result in an increase in behavior would be termed a negative reinforcer. The presentation of that same stimulus would, therefore, result in positive punishment.
Now, of course, reinforcement is relative. What’s reinforcing one moment or under one particular set of conditions is not going to be so in another. My interpretation of Skinner’s argument, though, is that defining punishment through reinforcement suggests that these conditions would serve as positive reinforcers or negative reinforcers at that moment under those conditions.
Seldom do we see Skinner’s formulation of reinforcement presented. The last I saw of this formulation was presented by Murray Sidman in 2006, though this is probably owed more to my inability to keep up with a identifying a rarely used definition throughout all of behavior analysis.
I must admit… I have no idea if I prefer Skinner’s formulation or the modern version that accounts for punishment as its own process, resulting in behavioral reductions. Based on my reasoning, and an argument for parsimony, Skinner’s formulation seems to win. Yet behavior analysts pursue punishment in its own right.
Perhaps we are not as clear on our definitions. Perhaps we confuse processes with procedures. Perhaps we take the trending absence of a behavior to be a behavior change itself. I have no answers in this column, other than: I would love to hear what you think. Where have I gone wrong in my thinking? Where have I gone right? Do we need to return to old ways, accept new formulations, or push for some alternative account not yet presented to the scientific community?
Reinforcement is so ubiquitous in behavior analysis, but maybe, just maybe, we still haven’t gotten this thing pinned down.