lecture 5 Flashcards
what are the different associations
Stimulus-Stimulus, Stimulus-Response, Response-Stimulus
What did rescorla’s (199) SR SO association experiments show
Demonstrated that animals are able to learn complicated action-outcome relationships
Training:
Noise
Lever = grain
Chain = grain
Light
Lever = grain
Chain = sugar
Tone
Lever = sugar
Chain = sugar
Could learn which pairings meant which outcomes
Similar treatment for stimuli with shared associations
Extinction of light followed by testing of noise and tone carries over
Is SR learning goal-directed behaviour
If instrumental learning only involves acquiring (S-R) associations, this does not support true goal directed action (Thorndike, 1933)
This means there is no representational role for the consequences of behaviours
reward/punishment act like non-specific stimulation of neurons rather than part of a remembered experience
Early experiments on goal directed performance contradict the idea that drives motivate behaviour
Relative experience of outcomes determines the behaviour rather than behaviour being determined by a combination of drive state and strength of behaviour
Elliot (1928)-Rat runway Outcome experiment
Rats must run down a maze to retrieve food, decreasing the amount of food decreases running behaviour
Tinklepaugh (1928)- Monkey Outcome expectancy experiments
Monkeys taught to uncover shells to find a banana show irritated responses when banana reward is replaced with lettuce - outcome>habit
Manipulations of outcome value provide a primary route for studying goals vs habit
How does Hershberger (1986) distinguish between pavlovian or goal directed behaviour
The Pavlovian Chicken continues to run towards food following conditioning despite the introduction of a conveyor belt meaning they should run backwards
This insensitivity to the consequences of its behaviour result in a failure to learn
Only learning is visual approach signal for food
However approach was not under the control of the consequences of its actions
Cannot change behaviour to match contingency
Interfering with its ability to adapt to a new environment
what are the crtiera for goal directedness
Instrumental criterion
A representation of the instrumental contingency between response and outcome
Dickinson & Balleine (1994) suggested that an action is goal directed if its performance is mediated by the interaction of two representations.
Goal criterion
A representation of the outcome as a goal for the agent
This requires the animal learning the incentive value of a stimulus
If an animal performs an action when there is independent evidence that the outcome is not represented as a goal then we should doubt the evidence
If an animal stops performing a response for similar reasons the there is support for goal directed action
Reward devaluation or specific satiety
Support for R-O contingency- Criterion 1- show
sensitivity to R-O
Reinforcer Devaluation (Colwill & Rescorla, 1985)
Rats trained with two responses leading to two positive outcomes
R1→ O1
R2 → O2
O1 is made aversive by pairing it with LiCl (induces nausea)
Performance of the instrumental response was substantially attenuated during extinction testing
Results suggest these devaluation effects are mediated by instrumental contingency and that a detailed representation of the reinforcer is ended in instrumental learning
what evidence is there for outcome incentive valuation in humans
Neural Substrates for Human Goal Directed Learning using Incentive learning with selective satiety (Valentin et al., 2007)
Participants are presented with symbols which correspond to different probabilities of receiving different juices of different desirabilities
One of these icons is then devalues
Orbitofrontal context shows modulatory activity during selection of devalued and non devalued outcomes
Simple behaviour choices show incentive shifts such that tomato is devalued but chocolate is still valued.
What does Balleine et al., 2009 show about incentive learning
Animals learn R–> O
Outcomes change motivational states
and motivational states change behaviour
manipulated the hunger states rats were in when pre-exposed to food, trained with food and tested with food.
re-exposure to food while hungry, training while satiated and tested when hungry produces a spike in lever pressing
pre-exposing animals who are satiated to food, teaches them that food has no effect on their hunger state so they eat less
Found that goal-directed R-O learning is meadiated by DMS while habitual performance is DLS
how does incentive value link to OCD (Gillian et al., 2013)
OCD is characterised by anxiety and enhanced avoidance habits
participants are trained to pull a lever to avoid a shock
however once the outcome is devalued patients with OCD continue to pull the lever
demonstrating poor sensitivity to outcome
humans have causal perception of their actions - do rats? (Blaisdell et al., 2006)
Two conditions: observing or intervening
intervening rat presses a lever which turns on a light which produces cheese paired with a sound
wants to investigate if rats know the lever pressing is a prerequisite for producing the cheese or whether they just associate the tone with the food
nose pokes used as proxy measure
those that intervened exhibited fewer nose pokes than observers
interpreted as having some understadning of the consequences of the actions which implies causal knowledge
