The Choice Paradigm

The Two-Key Procedure

In the laboratory, choice and preference are investigated by arranging concurrent schedules of reinforcement. Figure 9.1 shows a concurrent operant setting. In the laboratory, two or more simple schedules (i.e., FR, VR, FI, or VI) are simultaneously available on different response keys (Ferster & Skinner, 1957). Each key is associated with a separate schedule

FIG. 9.1. A two-key operant chamber for birds. Schedules of food reinforcement are arranged simultaneously on each key. Author Photo.

of reinforcement, and the organism is free to distribute behavior between the alternative schedules. The distribution of time and behavior among alternatives is the behavioral measure of choice and preference. For example, a food-deprived bird may be exposed to a situation in which the left response key is programmed to deliver 20 presentations of the food hopper each hour, whereas the right key delivers 60 reinforcers an hour. To obtain reinforcement from either key, the pigeon must respond according to the schedule on that key. If the bird responds exclusively to the right key (and never to the left) and meets the schedule requirement, then 60 reinforcers will be delivered each hour. Because the bird could have responded to either side, we may say that it prefers to spend its time on the right alternative.

Concurrent schedules of reinforcement have received considerable research attention, because they may be used as an analytical tool for understanding choice and preference. This selection of an experimental paradigm or model is based on the reasonable assumption that contingencies of reinforcement contribute substantially to choice behavior. Simply stated, all other factors being equal, the more reinforcement provided by an alternative, the more time and energy spent on that alternative. For example, in choosing between spending an evening with two friends, the one who has in the past provided the most social reinforcement will probably be the one selected. Reinforcement may be social approval, affection, interesting conversation, or other aspects of the friend's behavior. The experience of deciding to spend the evening with one rather than the other may be something like, "I just feel like spending the evening with John." Of course, in everyday life, choosing is seldom as uncomplicated as this, and a more common decision might have been to spend the evening with both friends. However, to understand how reinforcement processes are working, it is necessary to control the other factors so that the independent effects of reinforcement on choice may be observed.

Concurrent Ratio Schedules

Figure 9.2 shows a two-key concurrent-operant setting for humans. Consider that you are asked to participate in an experiment in which you may earn up to $50 an hour. As an experimental participant, you are taken to a room that has two response keys separated by a distance of 8 ft. Halfway between the two keys is a small opening just big enough to place your hand in. The room is empty, except for the unusual-looking apparatus. You are told to do anything you want. What do you do? You probably walk about and inspect your surroundings and, feeling somewhat foolish, eventually press one of the response keys. Immediately following this action, $1 is dispensed by a coin machine and is held on a plate inside the small opening. The dollar remains available for about 5 s, and then the plate falls away and the dollar disappears. Assuming that you have retrieved the dollar, will you press one of the keys again? In reality, this depends on several factors: Perhaps you are wealthy and the dollar is irrelevant; perhaps you decide to "get the best of the experimenter" and show that you are not a rat; maybe you do not want to appear greedy, and so on. However, assume for the moment that you are a typical poor student and you press the key again. After some time pressing both keys and counting the number of key presses, you discover a rule. The left key pays a dollar for each 100 responses, whereas the right side pays a dollar for 250 responses. Does it make sense to spend your effort on the right key when you can make money faster on the other alternative? Of course it does not, and you decide to spend all of your work on the key that pays the most. This same result has been found with other organisms. When two ratio schedules are programmed as concurrent schedules, then the alternative that produces more rapid reinforcement is chosen exclusively (Herrnstein & Loveland, 1975).

FIG. 9.2. A two-key operant chamber for humans. Pressing the keys results in money from a coin dispenser (middle), depending on the schedules of reinforcement. Author Photo.

Because ratio schedules result in exclusive responding to the alternative with the highest rate of payoff, these schedules are seldom used to study choice. We have discovered something about choice: Ratio schedules produce exclusive preference (in contrast, see McDonald, 1988, on how to program concurrent ratio schedules to produce response distributions similar to those that occur on interval schedules). Although this result is interesting, it suggests that other schedules should be used to investigate choice and preference. This is because once exclusive responding occurs, it is not possible to study how responses are distributed between the alternatives.

Concurrent Interval Schedules

Consider, however, what you might do if interval schedules were programmed on the two keys. Remember that on an interval schedule a single response must occur after a defined amount of time. If you spend all of your time pressing the same key, you will miss reinforcement that is programmed on the other alternative. For example, if the left key is scheduled to pay a dollar every 2 min; and the right key, every 6 min, then a reasonable tactic is to spend most of your time responding on the left key but every once in a while to check out the other alternative. This behavior will result in obtaining most of the money set up by both schedules. In fact, when exposed to concurrent interval schedules, most animals distribute their time and behavior between the two alternatives in such a manner (de Villiers, 1977). Thus, the first prerequisite of the choice paradigm is that interval schedules must be used to study the distribution of behavior.

Interval schedules are said to be independent of one another when they are presented concurrently. This is because responding on one alternative does not affect the rate of reinforcement programmed for the other. For example, a fixed-interval 6-min schedule (FI 6 min) is programmed to deliver reinforcement every 6 min. Of course, a response must be made after the fixed interval has elapsed. Pretend that you are faced with a situation in which the left key pays a dollar every 2 min (FI 2 min). The right key delivers a dollar when you make a response after 6 min. You have 1 hr a day in the experiment. If you just respond to the FI 2-min schedule, you would earn approximately $30. On the other hand, you could increase the number of reinforcers an hour by occasionally pressing the FI 6-min key. This occurs because the left key pays a total of $30 each hour and the right key pays an additional $10. After many hours of choosing between the alternatives, you may develop a stable pattern of responding. This steady-state performance is predictable. You should respond for approximately 6 min on the FI 2-min alternative and obtain three reinforcers (i.e., $3). After the third reinforcer, you may feel like switching to the FI 6-min key, on which a reinforcer is immediately available. You obtain the money on this key and immediately return to the richer schedule (left key). This steady-state pattern of responding may be repeated over and over with little variation.

Concurrent Variable-Interval Schedules

Recall that there are two major types of interval schedules. On variable-interval schedules (VI), the time between each programmed reinforcer changes, and the average time to reinforcement defines the specific schedule (e.g., VI 60 s). Because the organism is unable to discriminate the time to reinforcement on VI schedules, the regular switching pattern that characterizes concurrent FI FI performance does not occur. This is an advantage for the analysis of choice because the organism must respond on both alternatives and because switching does not result always in reinforcement. Thus, operant behavior maintained by concurrent VI VI schedules is sensitive to the rate of reinforcement on each alternative. For this reason, VI schedules are typically used to study choice.

Alternation and the Changeover Response

At this point, the choice paradigm is almost complete. Again, however, consider what you would do in the following situation. The two keys are separated and you cannot press both at the same time. The left key now pays a dollar on a VI 2-min schedule, and responses to the right alternative are reinforced on VI 6 min. The left key pays $30 each hour, and the right one delivers $10 if you respond. Assuming you obtain all programmed reinforcers on both schedules, you may earn $40 for each experimental session. What can you do to earn the most per hour? If you stay on the VI 2-min side, you end up missing the 10 reinforcers on the other alternative. However, if you frequently change over from key to key, most of the reinforcers on both schedules will be obtained. This is in fact what most animals do when faced with these contingencies (de Villiers, 1977).

Simple alternation between response alternatives prevents an analysis of choice, because the distribution of behavior remains the same (approximately 50/50) no matter what the programmed rates of reinforcement are. Frequent switching between alternatives may occur because of the correlation between the rate of switching and the overall rate of reinforcement (dollars per session). In other words, as the rate of switching increases, so does the hourly payoff. Another way of looking at this alternation is that organisms are accidentally reinforced for the changeover response. This alternation is called concurrent superstition (Catania, 1966) and occurs because as time is spent on an alternative, the other schedule is timing out. As the organism spends more time on the left key, the probability of a rein-forcer being set up on the right key increases. This means that a changeover to the right alternative will be reinforced, even though the contingencies do not require the changeover response. Thus, switching to the other response key is an operant that is inadvertently strengthened.

The Changeover Delay

The control procedure used to stop rapid switching between alternatives is called a changeover delay, or COD (Shull & Pliskoff, 1967). The COD contingency stipulates that responses do not have an effect immediately following a change from one schedule to another. After switching to a new alternative, a brief time is required before a response is reinforced (e.g., 3-s delay). For example, if an organism has just changed to an alternative that is ready to deliver reinforcement, there is a 3-s delay before a response is effective. As soon as the 3-s delay has elapsed, a response is reinforced. Of course, if the schedule has not timed out, the COD is irrelevant because reinforcement is not yet available. The COD contingency operates in both directions whenever a change is made from one alternative to another. The COD prevents frequent switching between alternatives. To obtain reinforcement, an organism must spend a minimal amount of time on an alternative before switching to another schedule. For example, with a 3-s COD, changing over every 2 s will never result in reinforcement. The COD is therefore an important and necessary feature of the operant-choice procedure for the investigator.

Experimental Procedures To Study Choice

The basic paradigm for investigating choice and preference is now complete. In summary, a researcher interested in behavioral choice should

1. arrange two or more concurrently available schedules of reinforcement

2. program interval schedules on each alternative

3. use variable- rather than fixed-interval schedules

4. require a COD in order to stop frequent alternation between or among the schedules

The Findley Procedure

An interesting variation on the basic choice procedure was described by Findley (1958). The procedure involves a single response key that changes color. Each color is a stimulus that signals a particular schedule of reinforcement. The color and the associated schedule may be changed when a response is made to a second key. This key is called the changeover key. For example, a pigeon may respond on a VI 30-s schedule that is signaled by red illumination of the response key. When the bird pecks a second changeover key, the color on the response key changes from red to blue. In the presence of the blue light, the pigeon may respond on a VI 90-s schedule of reinforcement. Another response on the changeover key reinstates the red light and the VI 30-s schedule. The advantage of the Findley procedure is that the response of changing from one alternative to another is explicitly defined and measured. Figure 9.3 compares the two-key and Findley procedures, showing that the Findley method allows for the measurement and control of the changeover response.

Current evidence suggests that the same principles of choice account for behavior in both the two-key and the changeover procedures. For this reason, researchers have not made a theoretical distinction between them. However, such a distinction may be important for the analysis of human behavior. Sunahara and Pierce (1982) have suggested that the two-key procedure may provide a model for social interaction. For example, in a group discussion

FIG. 9.3. Comparison of two-key and Findley procedures. Notice that the Findley method highlights the changeover response.

a person may distribute talk and attention to several group members. These members may be viewed as alternative sources of social reinforcement for the person. On the other hand, the changeover-key procedure may model role taking, in which an individual responds differentially to another person. In this case, the individual may change over between the reinforcement schedules provided by the other person as either a friend or a boss. For example, while at work the changeover may be made by saying, "Could I discuss a personal problem with you?" In other words, a person who is both your friend and your supervisor at work may sometimes deliver social reinforcers as a friend and at other times as a boss. Your social role may change when differential reinforcement (from supervisor or friend) is provided by the other individual.

0 0

Post a comment