Lesson: Distinguishing negative a positive

Lesson: Distinguishing negative a positive
In a previous issue of The Behavior Analyst, we discussed the ambiguities that surround the distinction between positive and negative reinforcement (Baron & Galizio, 2005; see also Michael, 1975). Seven distinguished behavior analysts commented on our article, and their comments and our response to them are already in print (Baron & Galizio, 2006; Chase, 2006; Iwata, 2006; Lattal & Lattal, 2006; Marr, 2006; Michael, 2006; Sidman, 2006). We believe that this dialogue represented a constructive step toward clarification of an important concept within behavior analysis. Most notably, the discussion showed how difficult it is to arrive at a set of rules that can identify a reinforcer as positive or negative.
Since then, the journal has received two more thoughtful contributions that appear in this issue (Nakajima, 2006; Staats, 2006). A rereading of our original article and our previous response to our critics will reveal that much of the ground has already been covered. However, both Nakajima and Staats rephrase the issues in ways that call for a further response. We thank them for further broadening the discussion.
Nakajima (2006) defends the traditional distinction between positive and negative reinforcement in terms of the introduction and removal of stimulus consequences. His position is that any ambiguities can be easily resolved by focusing on changes that can be directly observed and manipulated; alternative changes are secondary because they are based on “speculations.” Thus, when a rat's responses produce food pellets with the consequence that a period of food deprivation is terminated, the pellets are “countable,” whereas reductions in hunger are no more than “a guess of the rat's internal state” (p. 269). Similarly, “money is also material for people, but ‘poverty’ is secondarily defined” (p. 269). Along the same lines, any questions about the reinforcer when a rat's responding turns on a heat lamp in a cold environment are resolved in favor of heat onset (primary) rather than cold termination (secondary). “The heat is a physical entity within the rat's environment but ‘coldness’ is a hypothetical state of the rat”
Nakajima has phrased the issue incorrectly. We noted in our article and repeated in our previous response that it is customary and generally preferable to focus on consequences that are directly observable when we are talking about reinforcers. This is the standard practice in behavior-analytic research, and we are not arguing against it. We are concerned with a different issue: the merits of designating reinforcers as positive or negative simply on the basis of whether the procedure involves onset or offset of stimuli. The ambiguity inherent in the positive–negative distinction is not limited to cases involving reference to unobservable events. Regardless of the ease with which we can observe the consequences of responding, in all cases the addition of a stimulus to the environment of necessity requires its previous absence and, symmetrically, the removal of a stimulus requires its presence.
Moreover, reinforcers always involve a range of potential consequences, and we are often uncertain about the relative contribution of each to strengthening effects. As we illustrated in our article, the rigid application of an onset versus offset standard across a wide range of cases quickly encounters contradictions. One that we mentioned (but not discussed by Nakajima) concerns the proverbial person who takes an aspirin to relieve a headache. Such behavior is usually viewed as a form of escape, that is, the response of consuming aspirins is negatively reinforced by relief from the headache. By comparison, the standard espoused by Nakajima demands that we classify the behavior as positively reinforced by the aspirin (countable, manipulable) rather than negatively reinforced by headache reduction (change in an internal state). Michael's (2006) comments about the stimuli that signify conditioned reinforcers are relevant here. For a human subject working for money, it is of no great consequence if the stimulus signaling that money has been earned involves the offset of a light already on rather than the onset of that light. To be sure, one could define the former contingency as negative reinforcement and the latter as positive, but, as Michael notes, this would be “obvious foolishness” (p. 117) because the assignment of offset or onset is an arbitrary procedural detail. These and other cases discussed in our article lead to the conclusion that reinforcing events are best described in terms of a change in stimulation, regardless of the direction of the change.
We are uneasy about Nakajima's approach for a different reason. Using Skinner (1950) as his authority, he argues that the researcher's ease of observing or talking about a particular consequence should be the exclusive factor in deciding whether the reinforcer is positive or negative. However, such a standard is too reliant on the resources available to a particular researcher at a particular point in time. The advance of science tells us that progress has been largely dependent on the development of technologies that broaden the range of phenomena that can be observed. In the analysis of the reinforcement process, what Nakajima refers to as “speculative” consequences may be the key to the variables that control behavior. In the case of token or conditioned reinforcement, for example, control is not limited to the events that are immediately contingent on responding. A complete understanding of the reinforcing contingency requires examination of more remote consequences, in the absence of which the token will no longer function as a reinforcer (so-called backup reinforcers). Also instructive is Dinsmoor's (2001) analysis of avoidance. The conventional wisdom has been that avoidance is reinforced by the obvious changes in shock density that accompany responding. By comparison, Dinsmoor pointed out that more subtle consequences of responding are the response-produced cues that signify shock-free periods, and he proposed that it is these cues that reinforce behavior. The task of a proper analysis of behavior is to specify controlling variables regardless of whether they are obvious or superficially salient.