Project 4: Chapter 13: Exercises 41, 42

Stimulus Discrimination and Stimulus Generalization

This exercise is based on the materials in the Sniffy Pro manual.  To get a feel for the software download the Sniffy Demo software available on this site, and print off this document to follow step by step.  Some users may find it helpful to run the Tutorial, and see Project 4 before they try it by themselves.

Introduction to S+/SÐ Discrimination Learning

Learning to behave differently in the presence of different stimuli is called stimulus discrimination learning. The capacity to learn stimulus discriminations is the reason why a cat can learn to defecate in its litter box and not in the middle of the kitchen floor and why people learn to behave differently in a law courts and pubs.  The capacity to respond in a similar way in the presence of similar, but not identical, stimuli is called stimulus generalization. People do not have to relearn appropriate pub behavior when they visit a new drinking establishment, and a catâs appropriate toilet habits will usually survive a switch from one kind of litter box to another.

The Sniffy program uses tones to simulate two forms of stimulus discrimination learning. In simple discrimination learning, a tone is either on or off. If the tone is a positive discriminative stimulus (S+, a stimulus in whose presence a response is reinforced), then Sniffy is reinforced when the tone is on and extinguished when it is off. If the tone is a negative discriminative stimulus (SÐ, stimulus in whose presence a response is extinguished), then Sniffy is extinguished when the tone is on and reinforced when it is off. We call the second kind of discrimination learning that the Sniffy program simulates S+/SÐ discrimination learning. In this case, the S+ and SÐ are tones of different frequencies.

Once Sniffy has learned a simple discrimination using a tone as either the S+ or SÐ, you can measure stimulus generalization. Generalization tests are carried out under conditions of extinction. You would test Sniffy by presenting tones of several different frequencies (for example, 1.0, 1.25, 1.5, 1.75, 2.0, 2.25, 2.5, 2.75, and 3.0 kHz). In addition, you would probably want to create a stimulus generalization gradient, a graph in which you plot on the vertical axis (as your dependent variable) the number of responses that Sniffy makes during test periods of standard duration and on the horizontal axis (as your independent variable) the frequency of the different test tones.

The fact that Sniffy is a simulation and not a real animal simplifies the study of stimulus generalization for you. You can save a single discrimination-learning file when the learning process is complete and then revert to the status the file was in at the end of training for each generalization test. In this way, you create a situation in which each test stimulus is the ãfirstä stimulus that Sniffy receives immediately after the end of discrimination learning. Thus, you donât have to worry about presenting the test stimuli in different random orders and then averaging out the effect of extinction.

Stage 1: Exercise 41: S+/SÐ Discrimination Learning

In this exercise, you will teach Sniffy an S+/SÐ discrimination using the 2.0 kHz tone as the S+ and the 2.25 kHz tone as the SÐ.  Here are the steps that you would go through to set up a discrimination-learning experiment.

• Open a file in which Sniffy has been fully trained to respond with continuous reinforcement (either the file from Exercise 23 that we suggested you call Ex23-ShapeBP or the file named ShapeBP in the Sample Files folder). Examine the associations in the Operant Associations mind window. The soundÐfood association, barÐsound association, and action strength should all be at their maximum levels before you begin a discrimination-learning experiment.
• Use the Save As command to save the file under an appropriate new name (e.g., Ex41 S+2 SÐ2.25) in the Sniffy Files folder on your hard drive. This step is important because it preserves your original trained Sniffy file for use in future experiments.
• Select the Design Operant Conditioning Experiment command from the Experiment menu.
• In the dialogue box, select the stimuli that you want to use as the S+ (the stimulus in whose presence Sniffyâs bar presses will be reinforced) and the SÐ (the stimulus in whose presence Sniffyâs bar presses will not be reinforced).
  • To set up S+/SÐ discrimination learning, select tones of different frequencies for the S+ and the SÐ. For the tone-on S+ condition, point the cursor at the pull-down menu under S+, click your (left) mouse button, and then drag down to select a tone of 2.0 kHz. For the tone-on S- condition, point the cursor at the pull-down menu under SÐ, click your (left) mouse button, and then drag down to select a tone of 2.25 kHz.
• After selecting the S+ and the SÐ, click the OK command button at the bottom of the dialogue box to begin Sniffyâs discrimination training.
• Under the FILE menu you can set Preferences to "Sound Proof Cage". You may want to do this for this exercise because the computer generated tones can get to be hard to listen to.
• If you want to speed up the experiment, select the Isolate Sniffy (Accelerate Time) command from the Experiment menu.
• When Sniffyâs discrimination learning is complete, save the file.

During discrimination training, the S+ and the SÐ will be presented during alternating 1-minute intervals of program time. During the S+, Sniffyâs bar presses are reinforced. During the SÐ, his bar presses are not reinforced.  

During discrimination training, you can observe Sniffyâs progress by examining the Cumulative Record window, the Operant Associations mind window, and the DS Response Strength mind window. The cumulative record marks the occurrence of S+ and SÐ periods and shows how often Sniffy presses the bar during each period. The following figure shows the appearance of the cumulative record early in the S+/SÐ discrimination-learning experiment in which the S+ is the 2.0 kHz tone and the SÐ is the 2.25 kHz tone. Note how occurrences of these stimuli are denoted.

Very early in discrimination learning, Sniffy will press the bar equally often under the two stimulus conditions. As discrimination learning gets under way, he will begin pressing more often during the S+ than during the SÐ. When discrimination learning is complete, he will press the bar frequently during the S+ and very rarely, if ever, during the SÐ.

At the beginning of discrimination learning, the Operant Associations mind window will show that the barÐsound association and action strength are at high levels. Initially, when Sniffyâs bar presses are not reinforced during the SÐ, the barÐsound association and action strength will weaken. As discrimination learning proceeds, the strength of the barÐsound association and action strength will stop weakening and begin to increase. When discrimination learning is complete, the barÐsound association and action strength will be at or close to their maxima.

The DS Response Strength mind window shows Sniffyâs tendency to press the bar in the presence of tones of different frequencies. This mind window thus predicts the results that you would obtain if you performed generalization tests. At the beginning of discrimination learning, this window shows an equal moderate tendency to respond in the presence of all the different tones. During discrimination training, the DS Response Strength window will show the development of excitatory (above zero) and/or inhibitory (below zero) response tendencies. The following illustration shows the DS Response Strength mind window as it appears at the end of the S+/SÐ discrimination experiment in which the S+ was the 2.0 kHz tone and the SÐ was the 2.25 kHz tone. In this case, both excitatory tendencies to respond and inhibitory tendencies not to respond are associated with most of the tones.

Discrimination Learning Criteria:

Your criteria for deciding that Sniffyâs discrimination learning is complete should be based jointly on observations of the current Cumulative Record window and the Operant Associations and DS Response Strength mind windows. When discrimination learning is complete:

• The current Cumulative Record window will show that Sniffy is consistently (over several consecutive S+/SÐ cycles) responding frequently during the S+ and very rarely or not at all during the SÐ.
• In the Operant Associations mind window, the barÐsound association and action strength will approach their maxima.
• The appearance of the DS Response Strength mind window will have stabilized. The height of the bars will no longer be changing.

When Sniffyâs training is complete, choose the Save command from the File menu to save your results for use in tests of stimulus generalization. Your Cumulative Record window should resemble the CR shown below.

Stage 2: Exercise 42: Stimulus Generalization After S+/SÐ Discrimination Learning

In this exercise, you will measure stimulus generalization following the S+/SÐ discrimination learning that Sniffy acquired in Exercise 41. To perform the exercise, follow these steps.

• Open the file from Exercise 41 that we suggested you call Ex41 S+2 SÐ2.25, in which Sniffy has been fully trained on a discrimination-learning task.
• Check the criteria given above to make sure that discrimination learning is complete. If it isnât complete, run the file until it meets the completeness criteria.
• For the first of nine tests, select the 1.0 kHz tone for a generalization test. Select the Design Operant Conditioning Experiment command from the Experiment menu.
• In the Design Operant Conditioning Experiment dialogue box:
  • Select Extinction.
  • Be sure that there is a check mark in the box next to Mute Pellet Dispenser.
  • Place a check mark in the box next to Generalization Test by pointing the cursor at the box and clicking your (left) mouse button.
  • Select the tone frequency that you want to test from the pull-down menu under Generalization Test.
  • Click the OK command button
• When the test is complete, look at the Lab Assistant window to determine how many times Sniffy pressed the bar during the test. Record the number of bar presses on a piece of paper.
• Choose the Revert command from the File menu to return the file to the state it was in at the end of discrimination training.
• Repeat these steps to test Sniffyâs generalization to the other eight tones.

The Sniffy Pro program will automatically perform a generalization test during which the selected tone is presented continuously for 2 minutes of program time. During the generalization test, the Lab Assistant will count the number of times Sniffy presses the bar. On a piece of paper, record the frequency of the test tone and the number of times that Sniffy pressed the bar during the test.

To perform your next generalization test, choose the Revert command from Sniffy Proâs File menu and repeat the steps above.

• Warning 1: It is critically important to execute the Revert command to return the file to the state it was in at the end of discrimination training before performing the next test.
• Warning 2: It is also critically important not to save the file at the end of a generalization test. Saving the file at the end of a generalization test will leave you with a partially extinguished Sniffy and make it impossible to perform further unbiased generalization tests. If you save a partially extinguished Sniffy, your only option is to repeat the discrimination training before performing further generalization tests.

Use a spreadsheet, statistical analysis, or drawing program to draw a generalization gradient in which you plot the number of times Sniffy pressed the bar on the vertical (y) axis and the tone frequencies on the horizontal (x) axis. Your finished generalization gradient should resemble that shown next.

Thank you for exploring Discrimination and Generalization. When you get a chance go on to Project 5, an advanced operant conditioning project in which you train Sniffy to perform one of three tricks:  beg, front roll, or Face Wiping.