lecture 2 Flashcards
How does probability of US presentation affect learning
faster acquisition and stronger behaviour for 100% consistent pairing
BUT more resistant to extinction/ stronger learning for 50% consistent pairing
How does the rate of US presentation affect learning?
Strength of acquisition and resistance to extinction are both also relative to the temporal duration of the cues
An increase in CS duration leads to reductions in response rate
Controlling for reinforcement rate with equivalent cumulative reinforcement rates removes this
Eg 30 sec CS 100% reinforced and 10 sec CS 33% reinforced elicit no difference in response rate (Austen & Sanderson, 2019)
Thus reflects sensitivity to reinforcement rate across cumulative exposures
conditioned stimulus
Classic: Any stimulus that provides a perceptual response that is temporally bounded
BUT animals can learn form cues about times, places and spaces - so not discrete but extended.
conditioned response
Often mimics the unconditioned response
BUT can sometimes be in opposition to the UR e.g one that prepares the animal to cope with the US such as homeostatic mechanisms and drug taking
The same stimulus with the same outcome can produce different response e.g changing time can alter a freeze response to a flee response
conditioned morphine tolerance (Siegel, 1977)
Drug tolerance: ability to withstand higher dosages of drugs after repeated use
It has been suggested that drug tolerance reflects a learned response, such that what the individual remembers determines tolerance.
E.g the high drug use among soldiers in the Vietnam war did not carry over when troops returned to the USA
Rats administered morphine in increasing doses, then split them into two groups: same context and different context.
When the same tolerated dose was administered in the unconditioned context, the rats overdose.
CR to drug is an opposite change in body temp (and other effects of drug / UR), removing the anticipation of high doses caused the overdose
Drug addiction is highly psychological
Highlights issues for current model of rehab where extinction of drug taking behaviour occurs in an environment where they are not conditioned to take drugs
what is configural learning
Learning to respond to two or more stimuli on the basis of their combination but not when either are experienced in isolation
Perhaps generating a unique representation for the two stimuli combined.
Simple model of learning breaks down when we take into consideration the other relations, such as this, that can be learned.
what is the relational hypothesis of hippocampus function (Rudy & Sutherland, 1989)
Hippocampus can relate space, time and place together in unique ways to create a representation that can be used to solve problems:
Negative Patterning where the compound is different from elements (A+ B+ AB-)
Conditional Context Discrimination where context signals change in contingency (Context X: A+ B Context Y: A- B+)
BiConditional where elements signal change in contingency (AB+ CD+ BC- AD-)
(McDonald et al., 1997)- All these discrimination seem to be acquired by ‘all’ species and provide clues to structure of learning and the hippocampus is required for some aspects of relational learning
Associative Learning or Episodic Memory (Iordanova et al., 2011)
Studied whether episodic memories can be explained by associative learning and how the Hippocampus is involved.
Tested in what/where/when conditions, manipulating lesioned/ intact hippocampus in three dimensional discriminations.
What: tone/click
Where: spots or squares on walls
When: morning/ afternoon
Does not interfere with the ability to learn about compounds of two or individual elements rather the more complex compounds of 3+ components.
Hippocampus and MTL implicated in a conjunctive process that generates unique configured cues
how is discrimination learning gradual and why can it appear as insight
Prosolution reversal: reinforcement schedule is reversed, such that V+, H- become V-H+. Before the solution is reached there is apparent impairment (Krechevsky, 1938).
Speed of responding: while the animals are responding to R, the response latencies to VR+ are faster than to the HR-.
S+ advantage: When animals stop responding to spatial position (e.g., R) they will more likely respond to VR+ than VL- HR- or HL- on the next trial. Suggesting that they have been learning about V throughout training.
jumping stand test
Jumping stand test demonstrates how insight learning is a gradual process involving associative learning
Rats performance jumps from 50/50 to 100% vertical which appears to indicate ‘insight’
However this can be explained as a gradual process because animals have a bias to one side
Rat is being taught to jump to vertical lines but will tend to start with a R/L bias.
On each trial the vertical and horizontal stimulus is altered from each side.
Rats will gradually acquire the correct response but the behaviour changes spontaneously from a side preference to an orientation preference.
animals can learn about a stimulus dimension (orientation V) even while systematically responding to another (spatial position R).
computational properties of error correction
ΔH=α(λ-H) H = habit strength α = salience λ = maximum amount of habit In extinction trials λ will = 0
Pavlov (1927) overshadowing
If cues are of different salience (intensity) there is good evidence that learning is weaker for weaker stimuli.
This suggests that independent gradual learning of each stimulus isn’t going on.
This effect is called overshadowing
Rescorla Wagner model
Original theories could not account for selective association effects that come about when you combine CSs:
Overshadowing (Pavlov, 1927)
Blocking (Kamin, 1969)
Contingency effects (Rescorla, 1968)
RW postulate associative strength (habit) between CS and US representations rather than response probability.
Crucially, the sum of all cues present on a trial determine the change in associative strength.
ΔV=αβ(λ-ΣV)
V = strength of association
α = salience of CS
β = salience of US
λ = max association
Blocking Demonstration in Humans (Eippert et al., 2012)
Within subjects experiment with humans learning about visual cues for electric shock in two phases (Eippert et al., 2012)
Exp group A+ AX+ Test:X
Control group B- BY+ Test:Y
Connectivity between amygdala and prefrontal cortex mediated learning
Significant blocking effect in fear ratings and amygdala
relative value
Early theories believed that changes in behaviour were a
function of CS-US contiguity or contingency.
Kamin’s blocking and Rescorla’s contingency effects made it
clear that, with Pavlovian conditioning, relative rather than
absolute experience is what matters. This is key evidence
that these behavioural effects are truly psychological events,
because they speak to how previous experiences are
combined with new experiences, sometimes generating
novel configural representations (like the context) requiring
a computational description not based on contiguity.
