Transcript LT2Ch8a

PSY402
Theories of Learning
Chapter 6 (Cont.)
Chapter 8 – Cognitive Theories
Criticisms of Contiguity Theory


Guthrie conducted few studies to
support his theory.
Accurate parts:



Punishment can intensify inappropriate
behavior when it elicits a response
compatible with the punished response.
Contiguity is essential to prevent
conditioning of competing associations.
Not all environmental cues are noticed.
Impact of Reward

Guthrie’s view of reward has been
disproved.


If what happens after a response is not
rewarding, an S-R association is not
formed, even if the stimulus changes.
Noble – reward size predicts
response better than recency or
frequency (contiguity measures).
Single-Trial Learning


All-or-nothing (single-trial) learning
has been difficult to demonstrate.
Voeks – found single-trial learning
of an eye-blink response in humans.


Other studies report gradual learning.
Spence proposed a threshold
explanation of single-trial learning
using incremental learning theory.
Skinner


Emphasized the importance of
environment (reinforcers &
contingencies).
Validation of hypothetical constructs
interferes with analysis of the
variables controlling behavior.

Anti-theory
Spence’s Acquired Motives



Spence was a colleague of Hull.
Spence elaborated the idea that
reward size matters (K in Hull’s
theory).
It isn’t enough to say that reward
size matters – how specifically does
it affect behavior?

Spence proposed a mechanism.
Goal Responses



Reward elicits an unconditioned goal
response RG.
This response produces an internal
stimulus state SG that motivates
consummatory behavior.
Reward value determines the size of
the goal response RG.
Anticipatory Goal Responses

Cues become associated with
reward through classical
conditioning.



These produce an anticipatory goal
response rG.
Cues lead to internal stimulus
changes sG that motivate behavior.
Thus Pavlovian conditioning
motivates approach behaviors.
Amsel’s Frustration Theory

Amsel applied Spence’s theory to
avoidance of aversive events:




Frustration motivates avoidance.
Frustration suppresses approach.
Nonreward produces unconditioned
frustration response RF.
The stimulus associated with it SF
motivates escape behavior.
Anticipatory Frustration Response



As with goal states, classical
conditioning results in anticipatory
frustration response rF.
The conditioned stimuli associated
with them sF motivate avoidance of
a frustrating situation.
Example: car that won’t start.

SF motivates leaving the car, sF
motivates selling it.
Mowrer’s Two-Factor Theory


Mowrer proposed a drive-based
two-factor theory to avoid
explaining avoidance using cognitive
(mentalistic) concepts.
Avoidance involves two stages:


Fear is classically conditioned to the
environmental conditions preceding an
aversive event.
Cues evoke fear -- an instrumental
response occurs to terminate the fear.
Mowrer’s View (Cont.)



We are not actually avoiding an
event but escaping from a feared
object (environmental cue).
Miller’s white/black chamber – rats
escaped the feared white chamber,
not avoided an anticipated shock.
Fear reduction rewards the escape
behavior.
Criticisms of Two-Factory Theory

Avoidance behavior is extremely
resistant to extinction.


Should extinguish with exposure to CS
without UCS, but does not.
Levis & Boyd found that animals do
not get sufficient exposure duration
because their behavior prevents it.

Avoidance persists if long latency cues
exist closer to the aversive event.
Is Fear Really Present?

When avoidance behavior is welllearned the animals don’t seem to
be afraid.



An avoidance CS does not suppress
operant responding (no fear).
However, this could mean that the
animal’s hunger is stronger than the
fear.
Strong fear (drive strength) is not
needed if habit strength is large.
Avoidance without a CS

Sidman avoidance task – an
avoidance response delays an
aversive event for a period of time.


There is no external cue to when the
aversive event will occur – just
duration. Temporal conditioning.
How do animals learn to avoid
shock without any external cues for
the classical conditioning of fear?
Kamin’s Findings


Avoidance of the UCS, not just
termination of the CS (and the fear)
matters in avoidance learning.
Four conditions:




Response ends CS and prevents UCS.
Reponse ends CS but doesn’t stop UCS.
Response prevents UCS but CS stays.
CS and UCS, response does nothing
(control condition).
D’Amato’s Acquired Motive View

D’Amato proposed that both pain
and relief motivate avoidance.



Anticipatory pain & relief responses.
Shock elicits unconditioned pain
response RP and stimulus SP
motivates escape.
Classically conditioned cues sP elicit
anticipatory pain response rP that
motivates escape from the CS.
Anticipatory Relief Response



Termination of the UCS produces an
unconditioned relief response RR
with stimulus consequences SR.
Conditioned cues elicit an
anticipatory relief response rR with
stimulus consequences sR.
Example: dog bite elicits pain
response, sight of dog elicits
anticipatory pain, house elicits relief
A Discriminative Cue is Needed



During trace conditioning no cue is
present when UCS occurs and no
avoidance learning occurs.
A second cue presented during
avoidance behavior slowly acquires
rR-sR conditioning.
Similarly, in a Sidman task, cues
predict relief -- associated with
avoidance behavior, not the UCS.
How is rG Measured?

Anticipatory goal responses were
initially measured as peripheral
nervous system (ANS) response.


No consistent relationship between
such measures and behavior could be
found.
Now, Rescorla & Solomon propose
that these anticipatory states are
due to CNS activity (brain states).
PSY402
Theories of Learning
Chapter 8 – Cognitive Theories
Purposive Behaviorism




Tolman – behavior is goal-oriented.
Through experience we gain
expectations about how to use
paths and tools to achieve goals.
We expect specific outcomes to
follow specific behaviors.
If unrewarded, we seek other ways
to accomplish our goals.
Tolman’s View (Cont.)


We do not have to be reinforced in
order to learn.
We must be motivated:



Motivation produces internal tension
creating a demand for the goal.
Motivation determines what features of
the environment will be noticed.
Behavior is not fixed, automatic or
stereotyped, but flexible.
Place-Learning Studies


Demonstrate existence of spatial
expectations.
T-Maze – rat starts at different
location but reward always in same
end of maze.


Rats must turn different directions.
Alternate-path maze – rats choose
the shortest path after learning.

When blocked, take next shortest path.
Latent-Learning Studies


Investigate whether reward is
necessary for learning to occur.
Three conditions:




R –always got reward at goal
NR –never got reward at goal
NR-R – rewarded only on last 10 days
NR-R rats show rapid decrease in
errors when rewarded -- motivation
is needed for performance.
Problems with Latent-Learning

Difficulty replicating results:


MacCorquodale & Meehl found 30 of 48
studies could reproduce the results.
Motivation restricts attention to
relevant cues. Irrelevant rewards
are ignored.

No latent learning occurs when strong
but irrelevant rewards are provided,
even if they are relevant later.
Drive Response

Consistent latent learning occurs
when rats are not deprived initially.



Spence’s anticipatory goal response,
rG-sG was created to explain this result.
The anticipatory goal response is
formed but not apparent until there is
deprivation to activate the goal.
Handling animals may have been a
reward for Tolman’s NR-group.
Expectancies


Expectancy –mental representation
of event contingencies.
Dickinson – an expectancy contains
two kinds of information:


Associative link between two events –
classically conditioned, mechanistic.
Behavior-reinforcer belief –
consequences of action, operant,
intentional.
Testing Associative Links

Two groups trained to bar press:





One group reinforced with sodium (Na)
Other group reinforced with potassium
(K)
Both tested when deprived of sodium.
Irrelevant incentive effect – sodium
deprivation activated associative
link for Na rats but not K rats.
Could be due to beliefs not links.
Testing for Beliefs

Reinforcer devaluation effect – what
happens if the reinforcer is
diminished in value after training?




One group got sucrose for bar-pressing
and food regardless of behavior.
Other group got food for bar-pressing
and sucrose non-contingently.
Sucrose devalued during testing.
Bar pressing was lower when the
sucrose was behavior-contingent.
Importance of Disgust

Devaluation is a two-stage process:



A disgust reaction is associated with
the reinforcer (devalued by illness).
The reinforcer must be reexperienced.
Devaluation of the reinforcer occurs
when reexperience activates the
associated disgust.

Studied using ondansetron – a strong
anti-emetic (reduces nausea).
Importance of Habits



Dickinson acknowledged that habits
do exist and can control behavior.
Expectancies (behavior-reinforcer
beliefs) control actions before habits
are established.
Behavioral autonomy – control of
responding by habit rather than
expectancy.

Habit responds to devalued reinforcer.