lecture 12

advertisement
Chapter 6
Schedules of reinforcement
Schedules of reinforcement
Continuous Reinforcement Schedule
Every response is followed by the delivery of a reinforcer
(can also be called FR-1; one reward for one response)
Partial (Intermittent) Reinforcement Schedule
Not every response is followed by the delivery of a
reinforcer; that is, Rs are reinforced "intermittently"
according to the rule specified by the schedule
Four Simple Partial
Reinforcement Schedules:
1. Fixed Interval
2. Variable Interval
3. Fixed Ratio
4. Variable Ratio
FR
VR
FI
VI
Comparison of ratio and interval schedules
 both fixed ratio and fixed interval schedules have a
postreinforcement pause
 both FR and FI schedules produce high rates of
responding just before delivery of the next reinforcer
 both VR and VI schedules maintain steady rates of
responding, without predictable pauses
 BUT, there are differences between ratio and interval
schedules
 Ratio schedules produce higher response rates than
interval schedules
FR
VR
FI
VI
Comparison of ratio and interval schedules
VR schedules produce higher response rates
(responses per min) than VI schedules.
One possibility: Response rate higher when
reinforcement rate (reinforcer per min) is higher.
Will VR still produce higher response rate if rate
of reinforcement is equated on both schedules?
Reynold’s (1975) Experiment
Compared responses on a VI schedule yoked to a
VR schedule
One pigeon reinforced on VR schedule
One pigeon on VI yoked to a pigeon on VR so that
when the pigeon on VR was one response short of
the VR requirement, the next response by both
birds produced food.
Reynold’s (1975) Experiment
The yoked pigeon was on a VI schedule because:
 food availability depended on the time it took the
VR bird to complete its response requirement.
 this time interval varied from one reinforcer to
the next (dependent on # of responses the VR
bird had to make and how long it took the VR
bird to make them).
Reynold’s (1975) Experiment
Both birds received food at approximately the same time,
and therefore the rate of reinforcement (i.e., reinforcers
per min) was the same for both birds
Results
Despite the effort to equate rate of reinforcement, the VR
bird pecked much more rapidly than the VI bird
Thus, differences in reinforcement rate do not account for
differences in response rate
Another possible reason for higher response rates on
VR than VI:
 on a VR schedule a certain number of responses
must be made to obtain each reward
 however, on a VI schedule only one response
must be made to obtain each reward
 if the number of responses emitted to obtain
each reinforcer were the same on the two schedules,
then perhaps the rate of responding would be the
same
Experiment by Catania et al. (1977)
This study replicated Reynold’s finding (by equating
reinforcement rate) and also tested when equating
number of responses for each reinforcer by:
yoking the VR schedule to the number of
responses made by the VI subject.
i.e., the number of responses the VR bird had to make
to obtain each reinforcer depended on the number of
responses the VI bird had made during the interval to
obtain its reinforcer.
Experiment by Catania et al. (1977)
Again, even when the birds made the same
number of responses per reinforcer, the VR
birds responded at a higher rate than the VI
birds.
Cumulative Responses
Bird 402 on VI 30 s
Replication
Reynold’s
(1975)
Time (min) Bird 414 on VR 25
Bird 410 on VR, yoked so food
Bird 406 on VI, yoked so food
comes after same # of responses
comes at the same time as for Bird
as for Bird 402
414.
So, higher rate of responding on ratio schedules than
on interval schedules is not due to:
 differences in the rate of reinforcement on the
two schedules
 differences in the number of responses on the
two schedules
Why do ratio schedules produce higher rates of
responding than interval schedules?
A better way to explain the difference in response
rates between ratio and interval schedules is based
on the Inter-response time (IRT) – the interval,
or pause, between responses
Consider the probability of receiving a reward following
a given response
 on interval schedules, the probability of reward
increases with longer IRTs
 that is, the slower the animal responds, the more
likely it is that the next response will be reinforced
 BECAUSE, the next response is always closer to the
end of the interval
 this is not true for ratio schedules
 a low response rate under ratio schedules does not
change the probability that the next response will
produce reward
 in fact, long IRTs postpone reinforcement because
reward delivery is determined exclusively by the ratio
requirement, not the passage of time
On a VR schedule, short interresponse
times (IRTs) are more likely to be
reinforced, thus rapid responding is
reinforced.
On a VI schedule, long IRTs are more likely
to be reinforced, thus pausing (less rapid
responding) is reinforced.
Ratio schedules produce higher rates of
responding than interval schedules but
neither schedule requires that animals
respond at a specific rate
Can have procedures that specifically
require that a subject respond at a
particular rate to get reinforced
Response-rate schedules
Differential Reinforcement of Low
Rates of Responding (DRL)
 response is rewarded only after a certain amount
of time has elapsed since the last response
 DRL 15
• responses that are 15 seconds apart will be
reinforced (IRT 15).
• responses that occur with a lower IRT
(<15 seconds) will restart the timer
• 4 responses/min
 different than interval schedules because the timer
is reset
Differential Reinforcement of High
Rates of Responding (DRH)
 response is rewarded only if it occurs really quickly
after the last response
 DRH 5
• response is reinforced only if it occurs within 5 s
of the last response
• 12 responses/min or more
• if response rate drops below that, no reinforcement
(i.e., respond 6 or 7 seconds after last response, then
no reward)
Choice Behavior:
Concurrent Schedules
Measures of Choice:
Using Concurrent Schedules of Reinforcement
Typically two levers or keys with a schedule of
reinforcement associated with each. Choice is then
assessed by comparing an animal's rate of responding
on one lever with its rate of responding on the other.
e.g.,
Lever A
VI 1'
Lever B
VI 3'
Concurrent Schedules of Reinforcement
 Usually, reward on each lever is programmed
independently
 This means that if an interval schedule is programmed
on lever A, while responding on lever B, the timer for
lever A is running and reward availability is becoming
more likely
 Thus, with interval schedules the more time spent
responding on the other lever, the more likely the next
response on the interval lever will be reinforced
Typically there is a limited time frame: e.g., The
session is 60 min; have to obtain as many
reinforcers as possible in that time.
Thus, wait too long to respond on a lever (next
reward sits there waiting), then may not get the
maximum number of reward allotted for that lever
in the time allowed.
A formulation which describes the way animals
distribute their responding on the two levers is:
The MATCHING LAW:
• Relative rate of responding on a particular
lever equals the relative rate of reinforcement
on that lever:
# Responses on A
# Responses on A + # Responses on B
=
# Rewards on A
# Rewards on A + # Rewards on B
N.B. Reinforcement is what the animal actually receives;
NOT what he could receive
Download