Sniffy Pro Discrimination and Generalization PDF

Preview

Summary

This is the final research paper from Learning and Motivation. It involves working with SniffyPro and conditioning.(DO NOT COPY- put through turnitin)...

Description

Sniffy Pro Discrimination and Generalization April 1, 2018 PSY 0405 Learning and Motivation: Lab 1035, Lauren Ross

SNIFFY PRO DISCRIMINATION AND GENERALIZATION

2

Sniffy Pro Discrimination and Generalization Learning has been studied for centuries, and with each new finding, psychologists have established foundations for understanding more about individual learning processes. Throughout this history, conditioning procedures have been especially useful for generating novel findings that aid in clarifying specific learning practices. Operant conditioning, widely used to study associative learning, allows researchers to manipulate learning associations in subjects using reinforcement and punishment. Associations are situation dependent, and ensuing behaviors and responses suggest that reinforcement is particularly beneficial in creating lasting associations. Behavioral responses rely on certain reinforcers in various situations, therefore learning is reliant on the ability to distinguish between specific stimuli. The ability to associate certain behaviors and responses while experiencing different stimuli is called discrimination learning. Subjects learn responses to particular stimuli in different situations, and therefore do not behave the same in every situation. For instance, a child in school knows they can play tag at recess, but it is unacceptable (usually) to play tag in class. Discrimination allows learners to distinguish associations in certain situations, creating behaviors that are acceptable only when a specific stimulus is present. However, not every stimulus is exactly the same. For instance, different playgrounds exist between elementary and middle school, but students still know they can play tag outside. Alternatively, student’s classrooms look different each year, or even between teachers, but they know to behave similarly in every class. This process is called stimulus generalization. Learners can associate similar stimuli with similar behaviors and responses. This allows associations to be maintained between similar situations, despite lack of identical stimuli. In sum, discrimination allows learners to ignore other stimuli and respond to a specific stimulus association. Generalization, causes a response to something

SNIFFY PRO DISCRIMINATION AND GENERALIZATION

3

other than the main stimulus because the stimuli are similar. Because of this, discrimination learning usually comes first, later instigating generalization with stimuli resembling the primary stimulus. In discrimination learning, stimuli associated with changes in reinforcement are called discriminative stimuli. Discriminative stimuli can either be positive or negative based on whether reinforcement is present at the time of stimulation. When a response is reinforced at the time of a stimulus, this is called a positive discriminative stimulus, denoted S+. Subjects under this condition learn to associate the positive stimulus with reinforcement and the response behavior is expected to increase. When a stimulus extinguishes a response due to lack of reinforcement, this is called a negative discriminative stimulus, denoted S-. The response behavior of subjects under this condition decreases because reinforcement is not present and does not instigate association learning. In sum, when a stimulus is S+, responses are reinforced. When the response is not reinforced, and therefore extinguished, the stimulus is negative, or S-. Combined, the presence of an S+ stimulus implies a corresponding S- stimulus. Importantly, S+ and S- stimuli do not reinforce behaviors themselves, instead they refer to when reinforcement occurs (S+) or does not (S-). S+ and S- stimuli are only present in discrimination learning, not generalization. In generalization reinforcement is either always present or never present, therefore there are no corresponding S+ and S- conditions. In our current study, we examined discrimination learning and generalization for a virtual rat subject using Sniffy Pro software. The software, colloquially referred to as “Sniffy” allowed us to manipulate S+ and S- conditions during discrimination learning, as well as use previous extinction data to record generalization tests. To clarify, Sniffy is also the name of the virtual rat subject and he will henceforth be referred to as such. The Sniffy program uses a tone to represent

SNIFFY PRO DISCRIMINATION AND GENERALIZATION

4

stimulus-based discrimination learning. Sniffy’s responses are recorded based on bar-pressing in a virtual, operant chamber. During the positive discriminative stimulus condition (S+), Sniffy’s bar-pressing behavior was reinforced with a food pellet when the tone was on. His behavior was extinguished by lack of food reinforcement when the tone was off (S-). This is called simple tone-on S+ discrimination learning in the Sniffy program. During the S- or negative discriminative stimulus segment, Sniffy’s behavior is extinguished when the tone is on (S-) and reinforced when the tone is off (S+). This is called simple tone-on S- discrimination learning in the Sniffy program. During the simple tone-on S+ condition, Sniffy is expected to associate the tone with food reinforcement, and therefore bar-press significantly more while the tone is administered than during the S- condition when the tone is off. Alternatively, Sniffy is expected to associate lack of tone in the simple tone-on S- condition with reinforcement and therefore barpress significantly more when the tone is off than when the tone is administered. During the generalization test portion of the experiment, generalization was measured using prior conditions of extinction. Using Sniffy’s training with 2kHz tone during simple toneon S+ and S- discrimination learning, the goal of the generalization tests was to examine how similar stimuli needed to be for generalization to occur. In both conditions, the pellet dispenser was muted, therefore reinforcement was no longer occurring. In order to test this, we measured Sniffy’s bar-press response rates at nine levels of tone frequency: 1.00kHz, 1.25kHz, 1.50kHz, 1.75kHz, 2kHz, 2.25kHz, 2.50kHz, 2.75kHz, 3.00kHz. We tested him in two conditions, S+ and S-. In the S+ condition, Sniffy’s bar-pressing behavior was expected to occur only when the tone was administered and increase closer to the 2kHz baseline. In the S- condition, Sniffy’s barpressing behavior was expected to be lowest during the 2kHz frequency because bar-pressing in the simple tone-on S- condition was reinforced when the tone was off. For both discrimination

SNIFFY PRO DISCRIMINATION AND GENERALIZATION

5

and generalization, we hypothesized that, based on the condition, Sniffy’s bar-pressing behavior would increase during reinforcement periods and decrease when not reinforced. During the S+ discrimination condition, we expected Sniffy to bar-press more frequently when the tone was on, and less frequently when the tone was off and he was not being reinforced with food. During the S- discrimination condition, we expected Sniffy to bar-press more frequently when the tone was off and he was being reinforced, than when the tone was on and he was not being reinforced with food. In the S+ generalization test conditions, we expected Sniffy to bar-press most frequently at the 2kHz condition, and least frequently at the most extreme conditions. In the S- generalization test condition, we expected Sniffy to bar-press the least during the 2kHz condition, and the most at the highest and lowest of the nine frequencies. Methods Subjects and Design The subject involved with these experiments is Sniffy, the virtual rat. Sniffy came to be as a way to combat cost and problems when using live rats. Sniffy is used as a tool to witness learning within animals. The computer program that allows students and researchers to work with Sniffy is called the Sniffy Pro program. When working with the program, there are many crucial components depending on the intended experiment being conducted. Virtual Operant Chamber, Cumulative Record, Operant Association “Mind Window”, DS Response Strength Window, and Lab Assistant Window are all important components residing in the scope of our experiments. In opening the Sniffy Pro program, the Virtual Operant Chamber will be open, in the top left corner, along with a few other windows. Some of the other windows mentioned may either need to be opened manually or will open when the experiment has begun. The Virtual Operant Chamber is the main section of the Sniffy Pro program. This is where the viewable

SNIFFY PRO DISCRIMINATION AND GENERALIZATION

6

training of sniffy occurs. The important mechanisms within the Chamber are along the back wall. Inside there is a speaker to release tones of determined frequencies. There is a lever that allows us to record bar-pressing. This is accompanied by a dispenser where food pellets drop to be accessible to Sniffy. A water tube, that is not manipulated for the purposes of this experiment, is also located on the back wall. This chamber is where Sniffy’s reactions and learning can be witnessed by the researcher. The Cumulative record is a window that shows the amount of barpresses in a given amount of time. Specifically, for the discrimination part of the experiment the record would show the amount of bar-presses S+ or S- time periods. A line can be seen in this window where the slope is determined by the speed of Sniffy’s pressing. In the Operant Association “Mind Window” Sniffy’s cognition is measured by the behaviors he is exhibiting. There are three bars inside the “Mind Window”, two of which involve the main focus. Leading up to this experiment, Sniffy had to be taught to press the lever. To do this, a sound was connected to getting food, and then pushing the bar was connected to the sound, in turn connecting the bar to the food. Looking in the “Mind Window”, two of the bars are labeled ‘Sound-Food’ and ‘Bar-Sound’. These bars allow Sniffy’s mental associations to be known, outside of watching his repeated learned behavior. For the discrimination sections of the experiment, this window will tell us if these associations drop between the different time periods. The DS Response Strength Window provided information predicting what may occurs within the second half of our experiment, generalization. By showing Sniffy’s likelihood to bar-press when a tone is shifted up or down, we can compare the actual generalization test results to the predictions. This window originally starts with all bars, representing the different frequencies, equaled out. Following the end of the discrimination test, the bars shift, giving the generalization prediction. Finally, there is the Lab Assistant Window. This window is only used within the

SNIFFY PRO DISCRIMINATION AND GENERALIZATION

7

generalization tests. Following the completion of a generalization test, it tells you how many times Sniffy pressed the bar. Making the recording of data simple. These various windows make it possible to gain more insight into what Sniffy has learned and are useful throughout the entire experiment. Procedure This experiment was made up of two sections, Discrimination and Generalization, each of which had two parts S+ and S-. S+ and S- Discrimination The first portion of the experiment focused on having the tone as a positive discriminative stimulus (S+). Pressing the bar was reinforced (food pellet) only during the portion of time when the tone is on. After having Sniffy successfully conditioned with a Bar-Food association, the first discrimination test began. To ensure that Sniffy is conditioned properly, the Original Bar-Press training file was opened, and both the Sound-Food and Bar-Sound bars in the Operant Associations “Mind Window” were checked, making sure they were maxed. The experiment began with choosing the Design Operant Conditioning Experiment option. This was found in the upper left Experiment tab. When the window opened, the headings the were focused on were Reinforcement Schedule, Discrimination/Generalization, and Reinforcement Action. The continuous reinforcement schedule was selected. For this experiment, the S+ option, under Discrimination/Generalization, was set at two kHz and the S- was set at No tone. ‘Bar press’ was selected as the reinforcement action and the experiment is applied. The Sniffy Pro program alternated between tone, S+, and no tone, S-, in one-minute intervals. This experiment did not

SNIFFY PRO DISCRIMINATION AND GENERALIZATION

8

have a set time to measure until completion, but there were three criteria to follow to know when the experiment is complete. The first was that the Cumulative record showed Sniffy was rarely pressing the button during the S- intervals and pressed the button frequently during the S+ intervals, showing he was differentiating that the tone is when his actions were reinforced. The second sign that discrimination learning was complete, was done while focusing on the Operant Associations “Mind Window”. When the Bar-Sound bar was full during S+ intervals and very low during S- intervals, learned discrimination was near complete. The third criterion required paying attention to the DS Response Strength Window as the bars would eventually stop fluctuating. Once those three criteria were met, learning discrimination of S+ was complete. This file was then saved as Tone-On S+ Discrim. The procedure for S- was near exact. The major difference was that this portion of the experiment focused on having the tone as a negative discriminative stimulus (S-). Pressing the bar was reinforced (food pellet) only during the portion of time when the tone is off. Opposite to the previous portion. After having Sniffy successfully conditioned with a Bar-Food association, the second discrimination test began. The same bars were checked to ensure conditioning was complete. The set-up process of the experiment was the same, with one change. The reinforcement scheduled was continuous and the recorded actions were Sniffy’s bar-presses. The only change was made under Discrimination/Generalization. S- was set two kHz and the S+ was set at No tone. The Sniffy Pro program, again, alternated between tone, S+, and no tone, S-, in one-minute intervals. This experiment also did not have a set time to measure until completion, and the same 3 criteria were followed. Once those three criteria were met, learning discrimination of S- is complete. This file was then saved as Tone-On S- Discrim.

SNIFFY PRO DISCRIMINATION AND GENERALIZATION

9

S+ and S- Stimulus Generalization This portion of the experiment focused on generalizing the newly discriminated behavior of the Tone-On S+ Discrim file. It was to see how Sniffy generalized to 8 other tones after his discrimination training. This procedure was repeated 9 times with the S+ file. This segment started off by opening the previously saved document, Tone-On S+ Discrim. The file was checked to make sure that Sniffy had met the 3 criteria mentioned above. Similar to the first part, the Design Operant Conditioning Experiment window needed to be opened. After the window opened, the same headings were available. Extinction was now selected for reinforcement schedule. With this came two new options to be checked off. The pellet dispenser was muted, as well as Generalization Test being selected. By selecting this, a new drop down occurred and allowed one of 9 tone options the be chosen, ranging from 1kHz to 3kHz in .25 intervals (1.0, 1.25, 1.5, etc.). 1kHz was chosen as the first test. The generalization test then began. In this time, the program is played the given tone that was selected in the Design Operant Conditioning Experiment window and recorded how many times sniffy pressed the bar in a two-minute span of time. After that time is up, the number of bar-presses was noted in the Lab Assistant Window and manually recorded. This file was then save as S+ Tone-On Gen 1. This procedure was then followed 8 more times with the other frequency options and always started new from the Tone-On S+ Discrim file. Each file was saved with the respective frequencies in the title without decimals (Gen 125, Gen 15, Gen 175, etc.). The last portion of the experiment focused on generalizing the discriminated behavior of the Tone-On S- Discrim file. This portion of the experiment was to see how Sniffy generalized to 8 other tones after his discrimination training. This procedure was repeated 9 times with the Sfile.

SNIFFY PRO DISCRIMINATION AND GENERALIZATION

10

This segment started off by opening the previously saved document, Tone-On S- Discrim. The file was then checked to make sure that Sniffy had met the criteria mentioned in the discrimination training procedure. The set-up was exactly the same as the generalization testing for S+. The Design Operant Conditioning Experiment window was opened, and extinction was selected for the reinforcement schedule. The dispenser was muted, and Generalization Test was selected. 1.0kHz was selected for the first test. After clicking ‘OK’, the generalization test began. The program played the selected tone and recorded how many times sniffy pressed the bar in a two-minute span of time. After that time was up, the number of bar-presses was noted in the Lab Assistant Window and manually recorded. This file was then saved as S- Tone-On Gen 1. This procedure was then followed 8 more times with the other frequency options and always started new from the Tone-On S- Discrim file. Each was then saved with the respective frequencies in the title (Gen 125, Gen 15, Gen 175, etc.). Results Discrimination Training In the first discrimination test, where the tone was a positive discriminative stimulus (S+), Sniffy pressed the lever more times during S+ segments compared to the S- segments (See Figure 1). The two segments began with similar amounts of bar-presses, but then noticeable changes could be seen following the first minute. The trend of the S+ condition line increased in responses over time, while the trend of the S- line decreased in responses over time and even received none within the last 2 recorded minutes of the condition. In the second discrimination test, where the tone was a negative discriminative stimulus (S-), Sniffy pressed the lever more times during the S- segments compared to the S+ segments

SNIFFY PRO DISCRIMINATION AND GENERALIZATION

11

(See Figure 2). The two segments began with similar amounts of bar presses, and major changes were not seen until around the ninth minute of each segment. The trend of the S+ line gradually decreased in responses with random spikes and ended in no responses for the last 3 recorded minutes. The trend of the S- line stayed relatively consistent in the level of responses through the recording but was always above the S+ line after minute 16. DS Response Strength Figure 3 shows Sniffy’s behaviors were predicted to follow a normal response distribution following the S+ discrimination test. This distribution was centered at 2kHz and had 2 tail ends, one at 1kHz and the other at 3kHz. This predicts that in a generalization test, following S+ Discrimination, Sniffy would have the most responses while the 2kHz tone was playing, and his responses would lessen the further the tone got from the original testing tone (2kHz). Figure 4 shows Sniffy’s behaviors were predicted to follow an almost bimodal response distribution following the S- discrimination test. This distribution was centered with the lowest point being at 2kHz and had 2 peaks, one at 1kHz and the other at 3kHz. This predicts that in a generalization test, following S- Discrimination, Sniffy would have the least responses while the 2kHz tone was playing, and his responses would grow the further the tone got from the original testing tone (2kHz). Generalization Tests The first set of Generalization tests, following S+ discrimination, matched the prediction given by the DS Response Strength prediction in Figure 3. Figure 5 shows Sniffy’s behavior followed a normal distribution. Over all, the prediction est...