A theory of associative learning Flashcards
how did we learn about causality using associative learning?
- All species live in a world in which they need to predict and control the world around them to survive
- Associative learning found across animal kingdom – all vertebrates and invertebrates, and even in monocellular organisms such as planaria – and maybe plants?!
what is delay conditioning?
an unconditioned stimulus (for example, an electric shock) is introduced in the final moments of a conditioned stimulus (for example, a tone), with both ending at the same time
what is trace conditioning?
a form of associative learning that can be induced by presenting a conditioned stimulus (CS) and an unconditioned stimulus (US) following each other, but separated by a temporal gap
what is simultaneous conditioning?
when the conditioned stimulus and the unconditioned stimulus are presented at the same time
what is backward conditioning?
occurs when a conditioned stimulus immediately follows an unconditioned stimulus
Evidence for backwards conditioning, Mahoney and Ayres (1876)
- study on rats
- 4 second tone was paired with 4 second shock in variety of ways: delay conditioning, trace conditioning, simultaneous conditioning and backward conditioning
- Rats didn’t learn much if tone after shock
- learned more if tone simultaneous with shock
- learned most if tone before shock
- CR is suppression of licking. Measure is lick latency, how long it takes them to start.
- high latency = high fear
Importance of correlation in associative learning, Rescorla (1968)
- three groups of rats given 5 tones and some shocks
- group 1: positive - 2 tone = shock pairings, no more shocks
- Group 2: zero - 2 tone = shock pairings, plus extra shock
- Group 3: negative - shock never paired with tone
- group 1 showed the tone was positively correlated with shock
- group 2 showed the tone was uncorrelated with shock
- group 3 showed the tone was negatively correlated with shock
since pPositive and zero groups got the same number of tone →shock pairings but they did not learn the same - shows pairings are not everything and correlation is important
Kamin (1969)
- In Group 1 light conditioned with a pretrained noise N
- In Group 2 light conditioned with a novel noise N
- In Group 1 shock not surprising – predicted by pretrained noise
- In Group 2 shock not predicted, and is surprising
- found Less learning in Group 1, where shock not surprising
- The pretrained noise blocked learning about the light in Group 1
- This crucial observation, that pairings only produce learning when the US is surprising, is captured by Rescorla & Wagner theory
what does the Rescorla and Wagner (1972) theory describe?
describes how much association strength increases on each trial (i.e. CS→US pairing)
Rescorla and Wagner (1972) theory equation
∆V = αß ( λ - ∑V )
V is associative strength between CS & US
∆V is change in V after each pairing
α refers to salience of CS - its intrinsic perceptual intensity
ß refers to salience of US - its intrinsic perceptual intensity
λ refers to size of US - another physical property
what is α
salience of CS
what is ß
saliency of US
what is λ
the size of the US
Conditioning two CSs at once: Overshadowing, Mackintosh, (1976)
- Equal numbers of pairings of light with shock, but in overshadowing groups a noise also present – either quiet n or loud N
- most learning in control and with quiet n less learning with loud N
- When two stimuli conditioned in a compound, both help predict the US.
- On trial 1 learn about light and noise
- On trial 2 light and noise both help predict US
- Light and noise are competing for strength – if light gets more, the noise has to get less, or vice versa
- If α light= α noise they compete equally, and each gets half total strength
- But if noise more salient than light
- α noise»_space; α light
- noise will get more than half strength acquired on each trial
- In control group light gets all
the associative strength - In overshadowing group strength
is divided between light and noise - In blocking group noise gets all
strength in stage 1, so none
left for light!
what is ∑V?
- how much of the US is predicted
- the sum S of the associative strengths Vs of everything else that is present on that trial
therefore ( λ - ∑V ) is how surprising the US is
- if US is very large (λ big) it’s more surprising - if US is well predicted (∑V big) it’s less surprising
