Transcript Document

Instrumental/Operant
Conditioning
Instrumental Conditioning involves three key elements:
 a response
• usually an arbitrary motor response
• relevance or belongingness
 an outcome (the reinforcer)
• bigger reward = better conditioning
• contrast effects
 a relation between the response and outcome
• contiguity
• contingency
Contiguity
 delay of reinforcement
• bridge the delay with a conditioned reinforcer
• marking procedure
 studies of delay of reinforcement show that a perfect
causal relation between the response and outcome is not
sufficient to produce strong instrumental responding
• even with a perfect causal relation, conditioning does
not occur if reinforcement is delayed too long
• led to the conclusion that response-reinforcer
contiguity, rather than contingency, was the critical
factor
Skinner’s Superstition Experiment
A landmark study in the debate about the role of
contiguity versus contingency
Method
Food presented to pigeons every 15 s regardless of the
behavior of the bird.
Result
Birds showed stereotyped behavior patterns as time
for food delivery approached.
Skinner’s operant conditioning explanation:
 Adventitious (accidental) reinforcement of the
bird’s behavior
 Stresses the importance of contiguity between
R and the reinforcer
Logic of Skinner’s explanation:
 Animal could pick out which response was being
reinforced
 Animal was insensitive to contingency
 Extinction was much weaker than acquisition
The Staddon and Simmelhag Superstition
Experiment
A landmark study that challenged Skinner’s
interpretation
Method
Basically the same procedure as Skinner, except fixed
time interval of 12 s, and birds were observed and
their behavior recorded on all sessions.
Responding
at Asymptote
Fixed R1
Interval
R2
R1
R7
Terminal
Responses
R8
R1 = magazine wall
Fixed
Time
R2 = pecking key
R3 = pecking floor
R4 = ¼ circle
R8
R3
R4
R4
R3
R5 = flapping wings
R6 = window wall
R1
R7 = pecking wall
R8 = moving along mag wall
R2
R11 = head in magazine
R7
Interim
Responses
R8
R9 = preening
R10 = beak to ceiling
R1
R6
R3
R7
R5
Importantly, early in training, terminal
responses were not contiguous with food,
interim responses were;
According to Skinner’s adventitious reinforcement
hypothesis, interim responses should have become
the terminal responses, but they did not.
thus superstitious key pecking is not due to
adventitious reinforcement.
Classical Conditioning
Explanation of Superstition
CS
Time since
food
US
Food
CR
Pecking
UR
Pecking
Animal Behavior Terminology
• Appetitive Behavior – Occurs when
reinforcement not available
• Consumatory Behavior – Occurs when
reinforcement is about to appear
Contingency
 effects of controllability of reinforcers
 a strong contingency between the response and
a reinforcer means the response controls the reinforcer
 most of the research has focused on control over
aversive stimulation
 contemporary research originated with studies by
Seligman and colleagues
 they investigated the effects of uncontrollable shock
on subsequent escape-avoidance learning in dogs
 the major finding was that exposure to uncontrollable
shock disrupted subsequent learning
 this phenomenon called the learned-helplessness effect
The learned-helplessness effect
The design for LH experiments is outlined in Table 5.2, page 153
Experiment by Seligman and Maier (1967)
- demonstrated the basic LH effect
- 3 groups of dogs
Phase 1:
Group 1: Escape
- restrained and given unsignaled shock to hindfeet
- could terminate the shock by pressing either of 2 panels on
either side of snout
Group 2: Yoked
- placed in same restraint and given same # and pattern of shocks
- could not terminate the shocks by pressing the panels; they were
shocked whenever Escape animals were shocked
Group 3: Control - just put in restraint
The learned-helplessness effect
Phase 2:
- all dogs were treated alike
- put in a 2-compartment shuttlebox and taught a
normal escape/avoidance reaction
- dogs could avoid shock by responding during a 10-s
warning L or escape shock once it came on by jumping
to other side of compartment
- if subject did not respond in 60 s, the shock was
terminated
Thus, the experiment tested whether prior inescapable
shock affected escape/avoidance learning
The learned-helplessness effect
Results:
60
Mean
escape 40
latency
20
E
C
Y
The Escape group learned as easily as the Control group
The Yoked group showed an impairment
This deficit in learning is the learned-helplessness effect
The learned-helplessness effect
 the yoked group received the same number of shocks as
the escape group, so the failure to learn is not simply due
to having received shock in phase 1
 rather, the failure to learn was due to the inability to
control shock in phase 1
 no matter which response they performed, their
behavior was unrelated to shock offset in phase 1
 according to Seligman and Maier, the lack of control in
phase 1 led to the development of the general expectation
that behavior is irrelevant to the shock offset
 this expectation of lack of control transferred to the new
situation in phase 2, causing retardation of learning
Explanations of the LH effect
The learned-helplessness hypothesis
• based on the conclusion that animals can perceive the
contingency between their behavior and the reinforcer
• so, the original theory emphasized the lack of control
over outcomes
• according to this position, when the outcomes are
independent of the subject’s behavior, the subject
develops a state of learned helplessness which is
manifest in 2 ways:
- there is a motivations loss indicated by a decline in
performance and heightened level of passivity
- the subject has a generalized expectation that reinforcers
will continue to be independent of its behavior
- this persistent belief is the cause of the future learning deficit
The LH hypothesis has been challenged by studies
showing that it is not the lack of control that leads to the
LH outcome, but rather the inability to predict the
reinforcer
 receiving predictable, inescapable shock is less
damaging than receiving unsignaled shock
• if inescapable shock is signaled, then see less learning deficit
• if you present a cue that tells the animal the shock is coming,
then see less learning deficit
• animal still can’t escape the shock (i.e., still uncontrollable),
but they know its coming
 presentation of stimuli following offset of inescapable
eliminates the LH deficit
• this was demonstrated in an experiment by Jackson & Minor,
(1988)
Jackson & Minor (1988)
4 groups of rats
Phase 1:
Escape Group:
- received unsignaled shock that rats could terminate by turning
a small wheel
There were 2 Yoked groups
Feedback Group:
- house-light was turned off for a few seconds when shock ended
No-Feedback Group:
- no stimulus was given when shock was turned off
No-Shock Control Group:
Phase 2:
- all rats trained in a shuttlebox where they could run to other
side to turn off shock
Results:
30
25
Mean
20
Latency
15
Yoked
Escape
10
5
No
Shock
Yoked/feedback
Trials
Escape and No shock groups performed better than the Yoked
Group – this is the typical LH effect
Yoked/feedback group learned as well as the Escape and No shock
groups
The Jackson & Minor (1988) experiment demonstrated
that receiving a feedback stimulus following shock
offset eliminated the typical learning deficit
So, the learning deficit is not due to lack of control, as
suggested by the LH Hypothesis, but rather, it is due
to a lack of predictability
Explanations of the LH effect
1. The learned-helplessness hypothesis
2. Attentional deficits
- inescapable shock may cause animals to pay less attention
to their actions
- if an animal fails to pay attention to its behavior, it will
have difficulty associating its actions with reinforces in
escape-avoidance conditioning
- if the response is marked by an external stimulus, which
helps the animal pay attention to the appropriate response,
the LH deficit is reduced
- demonstrated in an experiment by Maier, Jackson & Tomie
(1987) – described on p 154-155 of your text.
Chapter 6
Schedules of reinforcement
Schedules of Reinforcement
A schedule of reinforcement is a program, or rule, that
determines how and when the occurrence of a response
will be followed by a reinforcer
The simplest schedule is a continuous reinforcement
schedule, abbreviated CRF
- every response = reinforcer
- rare in everyday life
- more common to have partial, or intermittent, reinforcement
Schedules of Intermittent Reinforcement
There are 4 basic schedules of intermittent reinforcement
Ratio schedules
– reward is based on the number of responses
Interval schedules
– a response that occurs after a certain period of
time has elapsed is reinforced
Each of these classes is further divided into Fixed or
Variable schedules
Schedules of Intermittent Reinforcement
Ratio
(# responses)
Interval
(time)
Fixed
Fixed
Ratio
Fixed
Interval
Variable
Variable
Ratio
Variable
Interval
Fixed Ratio (FR)
There must be a specified number of responses to
obtain each reinforcer.
This number is FIXED (i.e., the same number to
obtain each reinforcer) thus the reinforcer is
predictable.
Effect: Post-reinforcement pause (PRP; longer
with higher ratios) followed by a ratio run
(NB: ratio strain when very high ratios required)
Example: ‘piecework’ in factories
PRP
Ratio Run
Variable Ratio (VR)
There must a specified number of responses made
to obtain each reinforcer
The specified number of responses required varies
from one reinforcer to the next; Thus the reinfrocer
is unpredictable.
Effect: generally no PRP; steady very high rate
Example: slot machines
Fixed Interval (FI)
A specified interval of time must pass during which no
reinforcer is delivered no matter how many responses
are made
Once the interval is over, then the next response
produces the reinforcer
The time interval is fixed (i.e., the same time interval
must pass before each reinforcer), thus the reinforcer is
predictable.
Effect: Postreinforcement pause (PRP) followed by
gradually accelerating response rate, producing the
characteristic FI scallop
Example: Exams spaced apart at certain intervals – study
more closer to MT and Final
Variable Interval (VI)
A specified interval of time must pass during which no
reinforcer is delivered no matter how many responses
are made
Once the interval is over, then the next response
produces the reinforcer
The specified length of the time interval varies from one
reinforcer to the next, thus the reinforcer is not
predictable
Effect: generally no PRP; stable rate of responding
Example: time between customers in a store