reinforcers? One possibility is suggested by the observation of D.W. Zimmerman (1959) that the persistence of very first responses in the behavior chain controlled by the secondary reinforcing stimulus seemed to the crucial factor. He had trained rats to run from a start box through a short alley to a goal box where food pellets had been placed. Just before that start box door was opened, a buzzer was sounded. After this training, intermittent reinforcement was introduced, with some of runs being made to an empty goal box. Next, a bar was inserted in the start box and a press on the bar turned on the buzzer and opened the start box door. From this time onward, no food was ever placed in the goal box, and yet the animals continued to press the bar and scamper out of the start box time after time. Over the course of 10 or more daily sessions, each rat emitted thousands of bar presses. Thus, the stimuli from the buzzer and from the door being opened not only had shown capable of reinforcing a new response (bar pressing), but also were able to maintain this response with remarkable persistence. During the tests, running in the alley gradually extinguished; eventually, many of rats ran no more than a few inches into the alley. When picked up and replaced in the start box, however, they repeated to the process of pressing the bar and darting through the opening door. Because the rats continued pressing the bar only while their response to the buzzer remained vigorous, Zimmerman concluded that the secondary reinforcing properties of stimuli to case the next response in the response chain.
One serious limitation to the theory that the secondary reinforce always elicits (or otherwise controls) a response which previously was instrumental in bringing about the primary reinforcing event is found in free operant studies of secondary reinforcement. In these situations, there rarely is any evidence that a response (for example, an incipient approach to the food tray) consistently is made following the presentation of a secondary reinforce. For example, in the “chained schedule of reinforcement” procedure, a certain stimulus (such as a pattern of illumination on a pigeon’s key) may always be present when pecks on the key finally are reinforced with food. Now a different pattern is projected on the key, and pecks in the presence of this second pattern eventually result in the reappearance of the first pattern. Observations of the response rates, and especially of the changes in the “local rates” immediately preceeding and following the presentations of the first pattern have effects on key pecking similar to those to be expected if food had been presented, the first pattern is said to have secondary reinforcing properties.
The same logic as that used above can be applied to other types of reinforcement schedules (Kelleher, 1966). In all such situations it is clear that overt, incipient food-tray responses are not essential for the occurrence of secondary reinforcing effects. Of course, it could always be argued that prior training has extinguished the more gross features of such responses, but until some positive evidence is found that remnants of a response to a food tray are being made, it seems reasonable to conclude that the secondary reinforcing properties of stimuli are not uniquely associated with elicitation of the response which had been next in original behavior chain.
Some problems of interpretation In almost every experiment in which secondary reinforcement is said to have been demonstrated, one or more alternative explanations for the results might have been offered. Overall, the reaction of most psychologists has been one of interest in these alternative hypotheses, but they persist in relying on the use of the secondary reinforcement concept. The main reasons for this reluctance to substitute other explanations probably lie in the facts that secondary reinforcement is an ubiquitous concept, whereas the alternatives tend to be relatively specific to a few sets of conditions, and that secondary reinforcement has an intuitive appeal, while some of the alternatives seem rather implausible from the point of view of common sense. A brief survey of some of the hypotheses will illustrate their major characteristics.
DELAY OF REINFORCEMENT
“Delay of reinforcement” is a team that refers to the interval of time between the occurrence of a response and the delivery of its reinforcing stimulus. In each situation some criterion must be set which will define the moment of occurrence of the response. Where simple locomotor responses in an alleyway are studied, the moment the subject enters the goal box (as measured by depression of a movable floor or by interruption of a photocell beam) is usually taken as “the” instant that the instrumental response occurs, even though the movements involved are much the same as others which had been emitted earlier. In other case, specification of the “correct” response may be quite simple: Movement of a bar or key sufficient to close a switch defines the response, and in discrimination situations it is only necessary to add the requirement that such switch closures occur in the presence of certain stimulus conditions.