3 - Learning Theory Flashcards
Non-Associative Leraning
- when an animals behaviour towards a stimulus changes without the presence of an associated stimulus (reward/punishment)
Associative Learning
- ideas and experiences reinforce each other and can be linked to one another
> an animals behaviour towards a stimulus changes due to experience with that stimulus
Habituation
(opposite of sensitisation)
- to become accustomed to something
- decreased responding over a series of repeated exposures to the same stimulus
- habituation is not permanent
- dishabituation can occur
- habituation is stimulus specific
Sensitisation
(opposite of habituation)
- increased responding over a series of repeated exposures to the same stimulus
- can occur with intense stimuli
- can occur in the early stages of habituation
Dishabituation
- being increasingly stimulated by a new stimulus, relative to a stimulus you have been habituated to
Kamin’s blocking experiment and what it shows about surprise in learning
A = stimulus B = stimulus \+ = outcome (US) - = lack of outcome ? = no outcome applied but a human expectation of an outcome
Kamin’s blocking experiment comprised of two (learning) phases and a test phase
Phase 1
- the animal was exposed to a noise, Stimulus A (A), followed by a shock (+)
- A+
Phase 2
- the animal was exposed to A, then a light, stimulus B (B), followed by a shock (+)
- AB+
Test Phase
- exposed to B and was measured on it’s expectance of a shock (by it’s behaviour, freezing)
- B?
(the control group had only Phase 2 (CD+) and the Test phase (D+))
Results showed that the experimental group did not expect the outcome (+) when exposed only to the light (B), as they had been conditioned to expect + from A
The control group however, did expect + after exposure to D
- these findings showed that surprise is essential for learning
- shown by the experimental group not being surprised by the outcome (+) in phase 2 (AB+) due to the presence of A, even though B was present also, because they had already gone through the surprise of + linked to A and begun to expect it. Thus failed to learn an association between the light and the shock
- in this case the BLOCKED stimulus is B
Alternate phase 2 (UNBLOCKING):
- the experimental group is exposed to AB++ (double shock)
- this resulted in the experimental group expecting a shock when exposed to only B in the test phase
- as they had gone through the surprise of a second shock due to B in phase 2, and learned to associate B with +
Rescorla-Wagner model of classical conditioning
The Rescorla-Wagner model of learning is based on the principle of Error Driven Learning
- stating that if an outcome occurs that you haven’t predicted, you’ve made an error in your prediction, and the magnitude of error affects how strongly the association forms
- the strength of a Classical Conditioned Response depends upon the strength of the connection between internal representations of the CS and the US
> the associative strength of the CS - The change in associative strength is directly related to the discrepancy between an asymptotic value set by the magnitude of the US (λ), and the sum of the associative strengths of all the stimuli present on the trial
- The change in associative strength of a stimulus on any trial is determined by the discrepancy between λ and the sum of associative strengths of all the stimuli present on the trial
When graphed on a scale of associative strength due to number of trials, it produces an Asymptotic curve
What is shown when the Rescorla-Wagner model is graphed on a scale of associative strength due to number of trials?
an Asymptotic curve
Rescorla-Wagner model and Summation
𝚫V = λ - ∑V
- because the equation contains a ‘summed error’ term (∑V), if two stimuli (cues) are trained in compound (together) they both compete for associative strength with the paired outcome
- if trained with AB+, the association between A and + (and B and +) will be half as strong as if trained with only C+
Example experiment of conditioned inhibition and summation
Summation Experiment: Training phases: - A+ - AB- - C+ Test phase: - C? - BC? - AC? Results: - BC showed a lower conditioned response than C - AC showed twice the response of C - Thus B is the conditioned inhibitor, and reduces expectation of outcome
What is Conditioned Inhibition?
- specifically predicts the absence of an outcome
- requires an outcome to be expected (due to training)
Define retardation of acquisition and give an experimental example
Retardation of acquisition is a reduced ability to learn an association between stimuli, after having already learned a conflicting association
Retardation of acquisition test: Training: - A+ - AB- Test (learning stage 2): - A+ - B+ Results: - A+ shows an asymptotic curve of learning when the CS (A) is paired with a UCS (+) - B+ shows retarded learning when an inhibitor (B) is paired with a UCS (+)
What is Long-term Potentiation (LTP)?
A biological model of learning
Donald Hebb proposed a learning theory stating that when individual neural cells are fired at the same time, they establish connecting synapses or strengthen existing ones (what fires together wires together)
(connectionist)
Bliss and Lomo Experiment on LTP
Bliss and Lomo Experiment:
- brief pulses of current are delivered to a neuron and the magnitude of response is measured in neurons that are know to receive projects from neuron A
- this measures the stable baseline response (EPSP) to the stimulation
- then stimulation is changed to a burst of high frequency pulses. This is called the training stimulus
- after a brief rest period, the original test pulse is presented again, and the magnitude of response (EPSP) is now greater than before
- thus the test has shown LTP
- this synaptic change can be adopted as a model of how simple associative learning might occur biologically
- under optimal lab conditions, this enhanced response can persist indefinitely
Evidence that dopamine supports learning by signalling prediction errors (Schultz)
Shultz found that most midbrain dopaminergic neurons exhibit burst activity (phasic activation) following primary food and liquid rewards.
Also that dopaminergic neurons exhibit phasic activation following conditioned reward-predicting stimuli
Shultz found that dopamine release is sensitive to the prediction of an outcome and the outcome itself
Thus dopamine neurons code for error prediction
Dopamine neurons code for reward
(problematic that Shultz uses a reward system, when Kalin used a punishment system)
