Psych 110 Final UCLA (Wikenheiser)

novel cue

A _______ _____ always has an associative strength of 0 in the R-W model

prediction error

In the RW model, Associative strength can only change if _____________ != 0

Vsum

Changing ___________ causes all present cues to be changed in tandem

US

R-W model is a _____ -processing model because lambda sets the asymptote of learning and lambda is a property of this.

conditioned inhibitor

Phase 1: intermixed A+ and AX- trials

Phase 2: Intermixed A- and X-

R-W model predicts that X, which is a _______________ ______________, will extinguish

negative

Phase 1: light -> US

Phase 2: light + noise -> nothing

The associative strength of a conditioned inhibitor in R-W model will be ___________

summation test

Conditioned Inhibitors have negative V , and cues in compound have their assoc. strength added together : CI + CS+ = -1 + 1 = 0 , so no net US expectation, and no CR

(R-W explanation of __________)

impaired acquisition test

Because CIs start with a negative V, pairing with a US first must restore V to zero, and then push it to positive (R-W explanation of ___________________).

Implication: despite appearances, learning is always occurring during this test

Rescorla's Contingency

What happens to the associative strength of the US if it's not assigned to the CS ?

Context competes with CS for associative strength (R-W's explanation of ___________)

Blocking

In the RW model, ____________ occurs because the US is fully predicted, and w/ PE=0, there's no way for learning to occur for the novel cue in phase 2

over-expectation

Phase 1: light-> 1 food, tone-> 1 food, noise-> 1 food

Phase 2: light + tone -> 1 food, noise-> 1 food

On trial 1 of phase 2, _______________ represents the difference in conditioned responding to light+tone and noise. This is followed by a decrease in V for both cues of the compound!

* Conditioned responding is changing despite no change in US!

conditioned inhibitor

Phase 1: light -> US; noise-> US

Phase 2: Light + noise + X -> US

At phase 2, novel cue X has assoc. strength of 0. Over-expectation will cause Vsum to decrease, thereby making X a ___________________ despite always being paired with the US!

super conditioning

Phase 1:

Exp: A+, AB-

Control: A+, AB-

Phase 2:

Exp: CB+

Control: CD+ (both novel cues)

The US is extra surprising in phase 2 for the exp. group...

Adding an inhibitor to a compound increases learning of a novel cue C !

spontaneous recovery

RW model predicts basic curves of Extinction, but explains this phenomena by un-learning.

According to the model, and extinguished cue is no different than a novel cue (both have V=0)

This conflicts with ____________________

RW model

Difficulties of the ______________ include:

- Extinction

- CS pre-exposure/ latent inhibition

- Higher-order conditioning

- Sensory pre-conditioning

- CS/US belongingness

latent inhibition

RW model cannot explain ________________ because CS pre-exposure has a PE=0 so no change in V. In reality, there is difficulty learning familiar cues.

sensory preconditioning

RW cannot explain this because RW maintains a distinction between CSs and USs. A standard CS does not have a lambda to drive learning

Tone A --> Tone B

Tone B--> US

Test: A

Garcia Effect

The ____________ suggests that CS and US parameters are not independent. Each particular CS-US combination may have its own associability determined by evolutionary or other factors not specified by R-W

prediction error

Increased dopamine activity corresponds to positive ___________ _________, and so learning occurs

lambda

When a reward is predicted by a CS, dopamine neurons fire a burst of action potentials after the CS, and NOT after the reward. This suggests the CS seems to develop a _________ of its own (contradicts RW)

unblocked

standard blocking procedure:

phase 1: A+

phase 2: AX+

Test: X

Artificially activating DA neurons ___________ learning to X !

bottom-up

The alpha parameter of RW allows us to account for things like a loud noise overshadowing a quiet one; RW model is ______________ attention

limited capacity

Attentional models concern CS-processing, and assume that subjects have ______________ for sensory processing

Mackintosh

Says that better predictors command greater attention (increase alpha) and bad US predictors cause decreased attention

impaired acquisition

Mackintosh found that a Truly Random Control actually behaves much like a cue thats undergone latent inhibition; the TRC shows _______________

Learned Irrelevance

The slower learning that takes place when a CS and US are paired if they have previously been presented randomly with respect to each other

In contrast, RW predicts that the TR cue should have V close to 0 which mean it should condition the same as a novel cue

- Also, latent inhibition is distinct from this because TRC conditioned slower than LI cue

outcome specific

learned irrelevance is ___________________ ! Uncorrelated tone&water does not interfere with subsequent tone-> food learning

- Suggests this is not just a CS related phenomenon (like alpha in RW)

- Suggested to mackintosh that alpha is specific for particular pairs of CS&US

decrease

Mackintosh on blocking:

During phase 2, rats learn light isn't a good predictor of anything new, causing a ____________ in alpha and reduced learning on later trials

On trial 1 of phase 2, light is completely novel, so alpha is high resulting in a tiny bit of suppression in blocked group!

negative transfer

Learning in one situation (phase 1) affects subsequent learning in a diff situation (phase 2)

A-> Shock ; A-> SHOCK!

B-> Shock ; A -> SHOCK!

Mackintosh says that Group 1 should learn faster because A is a better predictor (so high alpha, fast learning), but this is not the case

Pearce-Hall

Says that uncertain predictors command more attention (increased alpha), and reliable predictors have decreased alpha

sign-tracking

Pearce Hall best explains this, because you pay most attention to cues that are least reliable (cue attention was highest for partially reliable cue)

Latent Inhibition

Both Mackintosh and PH models explain __________

M: pre-exposed CS is a bad US predictor so decreased alpha

PH: pre-exposed CS is a reliable predictor (of nothing) so decreased alpha

Blocking

Both Mackintosh and PH models explain _______

M: after trial 1 of phase 2, redundant cue X is no better at predicting US than pre-trained cue A (dec alpha)

PH: in phase 2, X is a reliable predictor of US, like A (dec alpha)

one trial blocking

Attentional models say that the extent to which a CS is a reliable US predictor affects how much V changes on the next trial... meaning that blocking cannot occur on the first trial

comparator model

Subjects learn many different associations during conditioning, but those associations are compared with one another to determine how each will influence behavior

associative strength

Problem with RW that Comparator model addresses: Learning requires direct experience in the RW model, so cues can't change _______________ on a trial unless they are presented

retrospective revaluation

a change in CR to a target cue is caused by learning that occurs when a target is absent

1. Un- Overshadowing

2. Recovery from blocking

3. Backwards blocking

within-compund

A _________________ association is when events that occur simultaneously (presented in compound) become associated with one another and evoke representations of one another

subtractive

Comparison is _________:

Response = directly activated US expectation - indirectly activated US expectation

indirectly

a prediction by comparison model:

anything that weakens the ____________ -activated US expectation should strengthen responding evoked by the target cue

RETURN HERE

law of effect

actions followed by satisfying consequences are strengthened, (more likely to occur in future), and actions followed by annoying consequences are weakened

reinforcement

increases behavior, can be positive or negative

punishment

decreases behavior, can be positive or negative

positive

Add a pleasant stimulus (reinforcement)

or Add a negative stimulus (punishment)

negative

Remove a pleasant stimulus (punishment)

or Remove a negative stimulus (reinforcement)

shaping

rewarding successive approximations of a desired behavior

Limitations:

- requires an "instructor"

- relies on random behavioral variation

auto-shaping

Pavlovian learning engages built-in behavioral mechanism related to the US that can be sculpted by reinforcement.

Eg. pavlovian sign-tracking responses could direct attention and US-appropriate response to parts of environment passively associated with an outcome

primary reinforcer

an innately/inherently reinforcing stimulus (warmth, food, water)

secondary reinforcer

neutral object that becomes associated with a primary reinforcer

continuos reinforcement

reinforcing the desired response every time it occurs

fixed ratio

reinforces a response only after a specified number of responses

- pause and run pattern

- post-reinforcement pauses lengthen as ratio increases

variable ratio

reinforces a response after an unpredictable number of responses

- steadier, high response rate

- as variability decreases, schedule approaches a FR

fixed interval

reinforces a response only after a specified time has elapsed

- scallop - accelerated responding as end of fixed interval nears

variable interval

reinforces a response at unpredictable time intervals

- very stable, steady responding (less slope than variable ratio)

progressive ratio

amount of responses required progressively increases with each reinforcer than subjects earn

- linear

- geometric (exponential)

breakpoint

Progressive ratio can be used to measure strength of reinforcer, or subjects motivation to obtain reinforcer , because as days go one,_____________ increases

partial reinforcement extinction effect

the greater persistence of behavior under partial reinforcement than under continuous reinforcement (slower extinction for partially reinforced group)

sequential

capaldi's ____________ model to explain Partial Reinforcement Extinction effect.

- memory of the sequence of previous trials becomes the stimulus that elicits future responses (but only for the PR group!)

frustration

Amstel's __________ model to explain PREE.

- emotional response on a previous trial become part of the stimulus that controls responding (only in PR group)

- in CR group, feeling of frustration first occurs during extinction, and is never followed by reinforcement

Response Outcome

Responding based on goal-directed behavior,

in contrast to Stimulus-Response learning (habitual behavior)

Plus Maze

Tolman's ___________ task showed that 'place' strategy was easiest to learn , so goal directed learning was easier than learning a sequence of actions

habitual

After extensive training in plus maze task, behavior transitioned to ______________.

dorsal striatum

After training until responding was habitual, lesioning the _________________ causes place strategy to re-emerge!

hippocampus

Lesioning the _________ prior to any training causes animals to learn a habit strategy (S-R) exclusively

reinforcer devaluation

Via sensory- specific satiety or conditioned taste aversion.

If habitual, would see no change in responding

If goal-directed, should see reduction in responding

goal-directed

reinforcer devaluation provides strong evidence for _____________ behavior.

As with spatial tasks, as training trials increases, goal-directed behavior --> habitual and so devaluation does not result in decreased responding in later trials

contingency degradation

destroy relationship between responding and outcome; this is the instrumental version of the Truly Random Control

- do this by allowing a primary reinforcer to be available for 'free'

- caused reduced responding for corresponding outcome, consistent with goal-directed behavior

goal-directed

Addiction can be ________________, as shown in experiment where rats had to execute novel sequence to earn cocaine. there was no chance for extended training so habitual directed responses could not be the cause

omission contingencies

if operationally Pavlovian learning is actually driven by an instrumental system, then arranging for the occurrence of a conditioned response to prevent outcome delivery should reduce conditioned responding

instinctive drift

tendency for an animal's behavior to revert to genetically controlled patterns