Introduction to associative learning theories_ conditions and contents of learning.pptx

Conditions of Learning Based on material developed by Jose Prados School of Psychology, University of Leicester Associative learning theory Deals with ability of living organisms to perceive contingency relations between events in their environment. Accounts for complex phenomena on the basis of a f...

Conditions of Learning Based on material developed by Jose Prados School of Psychology, University of Leicester Associative learning theory Deals with ability of living organisms to perceive contingency relations between events in their environment. Accounts for complex phenomena on the basis of a few simple principles. What are these principles? The primary issues to be addressed in the study of learning What are the conditions that bring learning about? What is learned? Robert A. Rescorla How does learning affect behaviour? Important reading suggestion: Rescorla, R.A. (1988) Pavlovian conditioning: It's not what you think it is. American Psychologist, 43, 151-160. (Available on-line.) What does the animal learn during conditioning? Events in ‘Real’ world Associative learning theory Central assumption of associative learning theories: conceptual nervous system consisting of 'nodes' links between nodes can form as a result of the conditioning link allows activity in one node to modify activity occurring in another node Tone Food … associations form in conceptual nervous system Associative learning theories study the principles that govern the establishment of these links, or ‘associations’. Is this how the real brain works? The famous ‘Hebbian synapse’ Contiguity principle - contiguity between events enough to guarantee learning However associative learning is more flexible… Let’s consider three cases where contiguity between events is not sufficient to establish associations. Contingency Blocking Stimulus specificity Which of these arrangements will produce better conditioning? Contingency, not contiguity drives learning informative value - relative probability of US CS in the presence vs absence of the CS: Signalled US Non-signalled US p (US – CS)> p (US – noCS) p=1 p=0 Tone CS Shock US In this case we should observe very good conditioning of the tone-CS: the US only occurs in the presence of the CS. Contingency, not contiguity drives learning CS informative value depends on the relative probability of the occurrence of the US in the presence and the absence of the CS: Signalled US Non-signalled US p (US – CS)= p (US – noCS) p=1 p=1 Tone CS Shock US In this case we should observe poor learning because the US is equally likely in the presence and in the absence of the CS. Contingency, not contiguity drives learning Rescorla (1968) Group Training Test 0-correlation p (Shock – Tone) = 0.4 p (Shock – noTone) = 0.4 Tone 0.4-correlation p (Shock – Tone) = 0.4 p (Shock – noTone) = 0.0 0.4/0.4 0.5 Suppression ratio 0.4 0.4/0.2 p (Shock – Tone) p (Shock – noTone) 0.3 0.4/0.1 0.2 0.1 0 0.4 / 0 0 Null correlation: NO learning p (US – CS) = p (US – noCS) 0.2 0.3 0.4 Positive contingency: good Contingency learning Contingency, not contiguity drives learning CS informative value depends on the relative probability of the occurrence of the US in the presence and the absence of the CS: Signalled US Non-signalled US p (US – CS)< p (US – noCS) p=0 p=1 Tone CS Shock US In this case the animal would learn that the CS is a good signal for the absence of the US, a case of inhibitory conditioning Contingency, not contiguity drives learning 1 p (US / CS) Excitatory conditioning 0 se e pr io t a t n ns o S C f d n a om d n a R S U Inhibitory conditioning p (US / noCS) 1 Example of Positive contingency: p(US/CS)=0.6 / p(US/noCS)=0 Ton e CS1 noCS1 Ton e CS2 noCS2 Ton e CS3 noCS3 Ton e CS4 noCS4 Ton e CS5 noCS5 Example of Zero contingency: p(US/CS)=0.6 / p(US/noCS)=0.6 Ton e CS1 noCS1 Ton e CS2 noCS2 Ton e CS3 noCS3 Ton e CS4 noCS4 Ton e CS5 noCS5 Example of Negative contingency: p(US/CS)=0.2 / p(US/noCS)=0.6 Ton e CS1 noCS1 Ton e CS2 noCS2 Ton e CS3 noCS3 Ton e CS4 noCS4 Ton e CS5 noCS5 Strength of conditioning 0.8 0.6 Excitatory 0.4 0.2 0 -0.2 -0.4 Inhibitory -0.6 -0.8 Positive Zero Negative Inhibitory learning No overt behaviour response How can we be sure learning has occurred? – Summation test – Retardation test Impaired learning in spite of contiguity with Kamin Blocking CS1  US, followed by…CS1 + CS2  US CS2 test shows impaired learning for CS2. Contiguity between stimuli does not guarantee learning. The surprise of the US seems to play a role here. Biological significance affects associability of stimuli Certain types of causes are more likely to produce certain types of effects: the cue-toconsequence effect Garcia & Koelling (1966) – ‘bright-noisy-water’ experiment (irradiation condition – nausea US) taste CS light/noise CS nausea US (irradiation) light/sound CS does not disrupt drinking Rats associated a taste, but not a light or sound, with illness. Avoidance of taste CS John Garcia Biological significance affects associability of stimuli (electric shock condition – pain US) taste CS light/noise CS pain US No fear response to taste CS Fear response to light/noise CS … In contrast, pain could be associated only with a visual or auditory cue, not a ta Biological significance affects associability of stimuli Licking response 4 3 Sucrose Tone-Light 2 1 0 Illness Shock no learning Conditions for learning Rescorla (Contingency): A stimulus will acquire the properties of a CS only if it is informative about the occurrence of the US Garcia: Associations between a CS and a US will establish if they are similar or biologically relevant Kamin: A stimulus CS will associate only with surprising USs Any learning theory should aim to account for Contents of Learning The three questions to be asked about any learning phenomenon Rescorla (1988) What are the conditions that bring learning about? What is learned? How does learning affect behaviour? Contents of learning: associative structures and how to demonstrate them empirically What does the animal learn during operant conditioning? The law of effect (Thorndike, 1910) https://www.youtube.com/watch? v=fanm--WyQJo Nodes (“neurons”) that respond to stimuli from the environment and control behaviour Training Leve r Res p. Test Leve r O O S Conceptual nervous system The law of effect postulates an association between stimulus and response, which is ‘stamped in’ by the outcome. R What does the animal learn during operant conditioning? Law of effect: The animal cannot predict the consequences of its behaviour, because there is no link between the response (or the stimulus) and the outcome. Animals react automatically to changes in their environment, and behave without purpose. Dotted arrow indicates that O only plays a role in reinforcing the strength of the S-R association but does not enter into an association e Law of Effect applied to Pavlovian conditioni (S-R Theory vs. S-S Theory) Challenge to S-R theory: StimulusStimulus Associations Sensory preconditioning: Rizley & Rescorla (1972) Preconditioning Conditioning Test Experimental tonelight lightShock tone Control tonelight Shock tone Conclusion on S-R vs. S-S learning S-R learning appears to be an oversimplified account for how animals learn. However there may be conditions which bias animals towards ‘habitual’ behaviour, akin to S-R learning. Learning theories contrast this to ‘goal-direct’ behaviour which depends on a representation of the outcome.