Running Head: SHAPING A RAT TO BAR PRESS
Shaping a Rat to Bar Press Using Fixed-interval
Schedules of Reinforcement
Christa A. Dinolfo and Marc A. Fernandes
Saint Bonaventure University
Abstract
Research has shown that animals can learn to respond on a fixed-interval schedule. Response rate is at its lowest in the beginning of an interval and highest at the end, showing that animals are capable of temporal discrimination. The present research attempts to observe temporal discrimination in a female rat as well as develop a characteristic of fixed-interval schedules known as the “fixed-interval scallop.” Researchers hypothesized that response rate would be at its highest before the end of a time interval and also that it would develop a fixed-interval scallop, responding at a low rate after reinforcement and high right before. The data supported these hypotheses.
Shaping a Rat to Bar Press Using
Fixed-Interval
Schedules of Reinforcement
Interval schedules are used in behavioral training to maximize response rate while delaying the inevitable effects of satiation. They give reinforcement only after a specified amount of time has elapsed since the last reinforcement was given (Alloway, 48). A fixed-interval schedule of reinforcement (FI) achieves the desired effects of an interval schedule particularly well because the time between reinforcements is set (and therefore not variable).
Research by Belke and Dunbar (1998) showed that rats can be trained to respond on a FI schedule. Rats were reinforced with fixed-intervals of wheel running after bar pressing on a fixed-interval schedule. Subjects were found to respond at the highest rate as the end of the interval approached and diminished response rate directly after reinforcement.
The most distinguished characteristic of fixed-interval schedules is known as the fixed-interval scallop. The fixed-interval scallop is an effect unique to the fixed-interval schedule that is noted by slow response rate at the beginning of an interval (or after reinforcement) which subsequently increases, reaching maximum response rate when the subject expects the end of the interval (thus bringing reinforcement) (Alloway, 50-51). The result is a scallop shaped line on a cumulative recorder created by a slow initial response rate and followed by a rapid ending response rate. The decrease in response rate after reinforcement noted during a FI schedule is similar to that of a post reinforcement pause seen in a fixed-ratio schedule noted by Dinolfo and Fernandes (2003).
Researchers have been able to achieve the scallop effect in several different organisms. Carlson et al (1965) demonstrated not only that a fixed-interval scallop was achievable but also that it was developed faster if a rat had been pre-trained on a FI schedule. A study by Boren et al. showed monkeys to keep the characteristic scallop despite strong punishment in an avoidance task. Some indications even suggest that human behavior can be described in terms of FI scalloping. A study by Roger Poppen (1982) implies that basic human behaviors related to “time-correlated stimuli” are inherently examples of fixed-interval scalloping.
Another purpose of FI schedules is to show that a subject can perform temporal discrimination, meaning that it can estimate and predict when the end of an interval is near. Presumably, when a subject expects the end of an interval to be near, it will respond at higher rates. The current study sought to replicate the temporal discrimination behavior as well as to demonstrate the fixed-interval scallop. Researchers hypothesized that the subject would respond slowly at first but progressing and reaching a high rate of response near the end of the interval, thus showing an FI scallop. We also hypothesized that the rat would “understand” the FI schedule and temporal discrimination, responding at a high rate soon before the end of an interval.
Method
The subject used was an approximately 75-day-old albino Rattus Norvegicus, Sprague-Dawley female rat, weighing approximately 200 g. Before each of the 5 experimental sessions, the rat was deprived of water for 48 hours. Treatment of the animal followed the guidelines of the “Ethical Principles of Psychologists and Code of Conduct” (American Psychological Association, 2001).
In each experimental session, the rat
was placed in an operant chamber, or “Skinner box”. A bar located inside the
Skinner box is activated with the force of approximately 5g. Approximately 3.8cm to the left of the bar is
a cup where 0.1ml of water drops into when either the rat pressed the bar
itself, or a feeder switch is manually activated on the control unit for the
Skinner box (Lavin 2). A distinguishable
noise, called the magazine, sounded every time the bar or the feeder switch is
pressed. The control unit also has the
ability to disable water from being dropped into the cup even if the rat
presses the bar or the manual feeder switch is pressed.
Prior to the experiment,
the subject was trained to bar press on a fixed-ratio (FR) schedule of
reinforcement. In the current
experiment, the subject was trained to bar press on a fixed-interval (FI)
schedule. As in the FR experiment, the
subject was only reinforced after certain bar presses. In this experiment, we manipulated a fixed-interval
of 120 s. This schedule required that a
bar press only be reinforced if 120 s elapsed since the previous
reinforcement. However, reinforcement
was only given after a bar press ended the interval, because we did not want to
reinforce superstitiously or accidental behaviors. We recorded the number of bar presses during
each 15 s interval during the FI 120.
This was repeated across the 5 experimental sessions.
Results
We calculated the mean
number of bar presses for each interval throughout each experimental day. A
one-way analysis of variance (ANOVA) was computed on our computer statistical
program, SYSTAT, to determine significance between intervals for each of the
five days. A within-subjects t-test was
used to compare the total number of bar presses during the first 2 intervals
(0-30 s) to the last 2 intervals (90-120 s) for each of the days. Alpha was set at .05.
Day 1 Intervals
A one-way ANOVA between
the mean bar presses for each of the 8 intervals in Day 1 was statistically
significant, F(7, 12) = 9.875, p = 0.000.
Figure 1 shows a graph of the mean number of bar presses for each
interval throughout Day 1.
Day 1 (0-30) vs Day 1 (90-120)
A t-test comparing the
first 30 s to the last 30 s of Day 1 showed that our data was significant. With t(18) = -4.606, p = 0.000.
Day 2 Intervals
A one-way ANOVA between
the mean bar presses in each of the intervals of Day 2 was statistically
significant, F(7, 15) = 23.589, p = 0.000.
Figure 2 displays the mean number of bar presses across each interval in
Day 2.
Day 2 (0-30) vs Day 2 (90-120)
There was a significant
difference between the number of bar presses in the first 30 s of Day 2 and the
number of bar presses in the last 30 s of Day 2. With t(21) = -7.540, p = 0.000.
Day 3 Intervals
ANOVA showed that each of
the means of the intervals across Day 3 were statistically significant from
each other, F(7, 15) = 39.027, p = 0.000.
Figure 3 shows a graph of the mean number of bar presses across the 8
intervals of Day 3.
Day
3 (0-30 s) vs Day 3 (90-120 s)
The mean of the first 30 s was 1.909 bar presses, while the mean of the
last 30 s was 7.545 bar presses. These
results were significant. With alpha at
.05, t(21) = -10.272, p = 0.000.
The intervals across Day 4
were statistically significant, F(7, 15) = 23.508, p = 0.000. Figure 4 shows a graph of the mean number of
bar presses across each interval in Day 4.
Day
4 (0-30 s) vs Day 4 (90-120 s)
The first 30 s interval
had a mean of 2.636 bar presses. The
last 30 s interval had a mean of 8.227 bar presses. The results were significant. With t(21) = -11.600, p = 0.000.
With an alpha level of
.05, the intervals of Day 5 were statistically significant, F (7,14) = 24.576,
p = 0.000. Figure 5 displays the mean
number of bar presses across each interval in Day 5.
Day
5 (0-30 s) vs Day 5 (90-120 s)
The mean of the first 30 s
in Day 5 was 1.857 bar presses. The mean
of the last 30 s (90-120 s) was 7.714 bar presses. There was a statistically
significant difference between the amount of bar pressing during these times,
t(20) = -10.669, p= 0.000.
Since each comparison was
found to be statistically significant, we can reject our null hypotheses. The subject responded more in the last 30 s
interval (90-120 s) than in the first interval (0-30 s), and its responses on
Day 5 were greater than those in Day 1 (see Figures 1 and 5).
Discussion
The present data confirms the original hypotheses. The subject responded at the highest rate in the last 30 seconds directly before the end of the 120 second interval. This suggests that the rat was able to expect and predict the end of the interval. Knowing that reinforcement was near, it responded at a higher rate as it expected reinforcement to be given. These findings are congruent to those found by Belke and Dunbar (1998).
The data also shows the development of a FI scallop. Although the data does not prove to be a “true” scallop, a comparison of means on the first and last 30 seconds shows the last segment to be significantly higher than the first segment. This is similar to the results found by Carlson et al. (1965).
Although a true scallop was not achieved, the data suggests that if the subject received more trials, the scallop shape would develop. Further research is needed to observe a more refined scallop shape. It is possible that after five or ten more trials the rat would perform at the rate of a true scallop.
The findings from
this research have larger implications and applications, particularly with
human behavior. As noted by Poppen
(1982), humans respond very similarly to animals on FI schedules, especially
with time-correlated stimuli. Fixed-interval
behavior shows itself in the student world as “cramming” and
procrastination. Students study for a
class very little at first but study for long periods directly before a
test. Students do not usually work hard
on a project the day it is assigned but do the most amount of work on it the
night before. Another example would be
if you were to imagine waiting for a train.
If the train is scheduled to arrive at
Fixed-interval schedules have few real world applications. It would not be in a boss’s best interest if their workers were only working at their hardest before a lunch break or the end of their work shift. The beginning of a sporting event might be a little boring to watch if players only played their hardest at the end of the game. So, fixed-interval schedules may not be best way to achieve consistent and high rates of response. It is necessary to study any time-correlated event to ensure response rate and consistency match the desired outcome.
References
Alloway,
T., Wilson, G., Graham, J., Krames, Lester K.
(2000). Sniffy the Virtual
Rat: Pro Version.
Belke,
T. W.,
Boren,
J. J., Merck, S., & Dohme Research Lab,
Carlson,
J. G., Myers, William A., Trapold, M. A.
(1965). The
Effect of Noncontingent Fixed and Variable Interval Reinforcement Upon
Subsequent Acquisition of the Fixed-interval Scallop. Psychonomic
Society. 2(9), pp 261-262.
Dinolfo, C. & Fernandes,
M. (2003). Shaping
a Rat to Bar Press Using Fixed Ratio
Schedules
of Reinforcement. Unpublished Manuscript.
Lavin, M.(2003). Lab 3 Instructions. Unpublished manuscript.
Poppen, R. (1982). Time Estimation by
Pigeons on a Fixed-interval: The Effect
of Pre-feeding. Behavioral Processes,
52(1), pp. 43-48.
Publication
Manual of the American Psychological Association (5th ed.). (2001).
Figure Caption
Figure 1. Mean number of bar presses across each 15 s interval in Day 1.
Figure 2. Mean number of bar presses across each 15 s interval in Day 2.
Figure 3. Mean number of bar presses across each 15 s interval in Day 3.
Figure 4. Mean number of bar presses across each 15 s interval in Day 4.
Figure 5. Mean number of bar presses across each 15 s interval in Day 5.
Figure
1
Figure
2
Figure
3
Figure
4
Figure
5