1. Psychology 485 September 28, 2010

2.  Introduction & History  Three major questions: What is learned? Why learn through classical conditioning? How does learning happen?

3.  Often contrasted to more cognitive approaches Watson & Little Albert  A premise: Study of simple learning processes will “scale up” to complex cognition

4.  Pavlovian (Classical) conditioning  Physiologist  Digestion Dogs  Conditional redirection of reflexes  Conditional reflexes

5. TIME Conditional Stimulus on off Unconditional Stimulus on off Unconditional Response (after enough pairings) Conditional Response

7.  “The originally neutral stimulus, through repeated pairings with the unconditioned one, acquires the response originally given to the unconditioned stimulus” Intro Psych textbook from 1987  What is wrong with this definition?

8.  Taste aversion  Idea of contiguity Temporal similarity between presentation of CS and US i.e. CS and US are presented at the same time  Contiguity is neither sufficient nor necessary

9. GroupInitial Training Second Training TestOutcome Control Group NothingTone + Light  Food Light  ?Moderate response to Light Blocking Group Tone  FoodTone + Light  Food Light  ?Little response to Light  During second training, tone & food are contiguous  Contiguity not sufficient

10.  Which CS would condition more easily? Contiguity is the same  CS2: US is contingent (dependent) on CS  Contingency, not contiguity CS1 CS2

11.  No contiguity between CS and US  CS signals absence of US Conditioned inhibitor  Contiguity is not necessary for conditioning CS1

12.  “The originally neutral stimulus, through repeated pairings with the unconditioned one, acquires the response originally given to the unconditioned stimulus” Intro Psych textbook from 1987

13.  CERs (Conditioned Emotional Response) Pair tone with shock When rat is shocked, it jumps and increases activity What tone is presented, rat freezes  Drug tolerance CSs for drug use cause body to prepare for drug Body prepares in opposite direction of drug

14.  Context  Hierarchical structure Second-order conditioning Occasion-setting  Expectancies

15.  What type of association is formed? Stimulus-Stimulus Stimulus-Response US CS Response

16.  So, how do you get rid of a response that is hard wired to a stimulus?  How can you get rid of a reflex? Habituation

17.  Less suppression in Habituation group (In other words, more responding)  Therefore, the connection MUST be S – SGroup Phase 1 Phase 2 TestHabituation L  N (startle) Noise (habituate) Light Control L  N (startle) NothingLight

18. NoiseStartle Light

20.  Expectancies CS helps you predict occurrence of US Makes animal more able to react to US  Biological relevance Not all CSs are created equal e.g. ‘bright-noisy’ water vs novel-tasting water  Hard to condition visual/auditory stimuli to nausea

21.  Blue Gourami Territory is defended more aggressively when competitor is signaled Winners become winners Losers stay losers

22.  Japanese Quail Signalling opportunity for reproduction Increases effectiveness of copulation (quicker and more ejaculate) Increases likelihood of fertilization

23.  Ant Lions Signal food presentation for larvae Build better pits Extract food more effectively Moult more quickly (quicker to reproduce)

24.  Baldwin effect If there is a reliable predictor of some important event across generations:  Learning faster is better  Learning becomes instinct? e.g. New predator in environment  Some behaviour makes it difficult for predator to kill prey  Learning behaviour provides survival advantage  Selection: ability to learn improves  Eventually behaviour becomes instinct

26.  Computational model of conditioning Widely cited and used Most important paper in animal learning?  Learning as a violation of expectations

27. Error Calculation

28.  On every trial: 1. Look around and examine all your stimuli 2. Use them to predict what will happen (V ∑ ) 3. Get a reward/US. How good/big was it? ( λ ) 4. How wrong was your prediction? ( λ - V ∑ ) 5. Take a portion of that error ( α and β ) 6. Change your prediction for next time ( Δ V)

29.  And Voila! You have a learning algorithm. Δ V = αβ ( λ - V ∑ )  λ = the maximum conditioning possible  α = saliency of the CS (between 0 and 1)  β = saliency of the US (between 0 and 1)  V X = associative (predictive) strength of a given stimulus  Δ V X = change in the associative strength of a given stimulus  V ∑ = total associative strength of all stimuli

30.  Equation describes a change in expectancies  Change in expectancy is based on: Features of the CS and US Total possible learning, minus what you’ve already learned  Based on expectation of US

31.  Does not account for many Classical Conditioning findings: Spontaneous recovery Savings CS pre-exposure (latent inhibition) Higher-order conditioning  Trial-by-trial based account Does not account for timing

32.  CS processing theories suggest properties of CS affect learning Attentional theories: it’s adaptive to pay attention to CSs that may signal important events Also adaptive to not pay attention to CSs that are not likely to signal important events  Pearce-Hall model Attention to CS changes across trials α can change from trial to trial