Study 7: Schedules of Reinforcement

8 March 1999

Abstract

Rats that have acquired the operant response of food-reinforced lever pressing will be placed on variable interval schedules of reinforcement. To shape variable interval responding, rats will first receive reinforcement according to a continuous schedule of reinforcement. The variable interval requirement will then be introduced and subsequently increased in a systematic manner.

Introduction

As Skinner noted in his survey of his work (Skinner 1959), the study of reinforcement often involves different contingencies, or schedules of reinforcement. Indeed, Skinner once published a book consisting entirely of the analysis of cumulative records from different schedules (Ferster and Skinner 1957).

Our project for this week is less grand. For those of you who successfully shaped your rat in Study 6, (and, I think that includes everyone) this project will involve the measurement of performance on a variable interval (VI) schedule.

VI schedules have been studied so often that it is difficult to find a new point of view. Nevertheless, we will be collecting data that are relevant to a still unresolved issue regarding the processes that control VI responding. Specifically, we will be looking at interresponse times (IRTs) during VI responding. (Platt 1979) has proposed that VI reinforcement schedules serve to strengthen particular IRT distributions and has suggested a model of behavior based on this idea. Indeed, Staddon and Ettinger (1989) have even suggested that IRT analysis might be relevant to applying the new mathematical techniques of chaos theory to operant behavior.

Methods

Subjects:

Our Sprague-Dawley rats will serve as subjects.

Apparatus:

We will be using the six custom-constructed chambers to test our animals. Each of these chambers will be fitted with a response lever that the animal can press. In addition, we will use an electric clock to measure time intervals during training. Reinforcers will consist of chocolate sprinkles delivered to the food cup in each box after a tap on the chamber wall.

Procedure:

Our lab will again be 90 minutes long. During that time we will reinforce the rats according to a VI schedule. We will gradually work up to a VI 100 sec schedule. During the VI 100 phase, the time it takes the rat to perform a series of 5 responses will be measured.

Start by reinforcing your rat according to a CRF schedule. Deliver 5 responses on this schedule. Then deliver 5 reinforcers according to a VI 10 sec schedule. Begin with the first time interval in Table 1 (these tables will be distributed in class). Wait this amount of time and then reinforce the next response that occurs after this time interval has elapsed. Deliver 5 reinforcers on this schedule. Then deliver 10 reinforcers according to a VI 20 seconds schedule; intervals for this schedule are given in Table 2. Finally, switch to the VI 100 sec schedule. These intervals are given in Table 3.

During the VI 100 sec schedule, record the time it takes the rat to perform five responses. That is, note down the time that every fifth response is made and subtract the previous time to get the "waiting time" to the fifth response. Note which of these contains the delivery of a reinforcer.

After the session, weigh and feed your rat.

Results

We will try two different kinds of analysis of these data. The traditional means of analysis of operant behavior is by use of the "cumulative recorder". We don't have enough cumulative recorders for all the chambers in our class, but we can construct a cumulative record from the waiting times. Each waiting time marks an additional 5-response increment in the cumulative number of responses. If the cumulative total of responses is plotted against the cumulative time, we have a rough version of the cumulative record. Try plotting out your rat's behavior in this way. What kinds of biases are induced in constructing a cumulative record in this manner?

The second type of analysis is an IRT distribution analysis. Strictly speaking, this analysis requires the waiting time to single responses. For reasons of practicality, we're using waiting times to the fifth response, but the same considerations can be shown. Begin by constructing a frequency histogram of all the waiting times, excluding those that contain a reinforcer delivery (the time eating a sprinkle will complicate the analysis). Plot the histogram for the entire session. What shape does the histogram seem to show? Now, take the waiting times in the first third of the VI 100 sec phase (e.g., if you have a total of 60 waiting times, take the first 20 of these) and construct a histogram. Compare this to histogram of the last third of the phase. Do you think the VI 100 sec training has affected the VI distribution?

Skinner (1959) has argued for the importance of cumulative records in behavior analysis. Basically, he would argue that cumulative records provide better information than that from something like our IRT analysis. Do you think the IRT analysis provides any useful additional information?

References

Required

Platt, J.R. (1979) Interresponse-time shaping by variable-interval-like interresponse-time reinforcement contingencies. Journal of the Experimental Analysis of Behavior, 31: 3-14.