Learning and Behaviour Essay Plans

Page 1

Comparisons of the major theories: Thorndike vs.

Guthrie vs. Hull vs. Tolman vs. Skinner

Thorndike

- Cats and puzzle boxes

- Learning as gradual, trial and error

- Instrumental learning- consequences connect response to stimulus

- Learning by association

- 3 behavioural laws: law of effect, law of readiness, law of exercise

- Other contributions:

- Multiple response (like Skinner's creativity)

- Associative shifting (like Skinner's fading)

- Response by analogy (like generalisation)

- Response availability (like discrimination learning)

Guthrie

- Importance of contiguity of responses- responses that occur in conjunction with stimuli will be repeated

- Importance of S-R connection in therapies e.g. flooding, threshold method and counter-conditioning, incompatibility method

- Strength of learning determined by complexity of behaviour and stimulus

- Rewards: not essential, focus link between S-R, facilitative drive feedback

- All or none learning- Guthrie and Horton experiment

- Thorndike vs. Guthrie: gradual trial and error vs. correct response always there but needs to be produced in right situation, reinforcement makes behaviour more likely to be reproduced vs. reinforcement not essential but focuses behaviour on stimulus

- Flaws: experiment too simple (opening gate on first trial inevitable), learning isn't all or none, responses not fixed to stimuli (rat in maze with/out water), general movements reinforced not specific behaviour,

Hull

- Role of intervening variables

- Contingency and contiguity- learning and performance separate

- Drive reduction- finding appropriate R to S that reduces drives

- S-s-R relationship, s=trace of knowledge internalised after S- dissimilar to Skinner (internal processes)

- Habit strength- strength of association- like Guthrie and Tolman

- Drive strength- how much animal wants to take advantage of connection

- Reaction potential- likelihood of animal producing response (E= HxD)

- Drive reduction leads to convergence of behaviour, variability of responses decreases until drive reducing response dominates

- Behavioural convergence

- Habit family hierarchy- first preferred response

- Behavioural divergence- same environment, many potential responses

- Gradient of reinforcement or goal gradient: final reward reinforced all subsequent behaviours- final reinforcement generalised, secondary reinforcers

- Fractionary goal response

- Hull's behavioural system

Tolman

- Purposeful behaviour of animals- motivated to produce behaviour to achieve goals

- Cognitive maps to navigate environment

- Role of place: T-maze experiments, rats learned place not response

- Latent learning: learning without reinforcement- like Guthrie. Learning and performance different. Rats pre-exposed to maze without rewards learnt it faster, although food in pre-exposure did improve rate of learning.

Tolman and Honzik: 3 groups of rats, 17 days, 14 choice maze. Fed in maze, never fed in maze, not fed until day 12 (G3). G3 caught up with G1 from day 12 (something learned from unreinforced trials). Same with rats exposed to dry maze then wet maze

- Route variation studies- preference for shortest passage in maze, response hierarchy

- Punishment- animals in maze shocked and learned better- dissimilar to Thorndike or Skinner

- Vicarious trial and error- vacillate at intersection, suggests purposive action- dissimilar to Thorndike or Guthrie

- Learning paradigm- perception, subsequent action, goal. S-R-E, includes contiguity (S-R) and contingency (R-E)

- Expectancy value theory- animals reinforced 75:25 on 2 arms of T-maze, influenced Atkinson's EVT (P=MxExI)

- Influence: importance of contiguity and contingency, influenced EV theory, influenced LOC theory, distinguished learning from performance

Skinner

- Contingency not contguity, not S-R theorist

- Learning through consequences- similar to Thorndike, but consequences instead of association

- Operant paradigm- similar to Hull, but no distinction between learning and performance- learning inferred

- 3 term contingency (discrimination learning, generalisation gradient, extinction and spontaneous recovery)

- Behaviour modification (shaping and successive approximation, response chaining, forward chaining, fading, assisted/forced learning)

- Schedules of reinforcement (FI, FR, VI, VR), diagrams

Page 2

Causal reasoning and Illusion of control

- Contingency definition- P(reinforcement/response)>P(reinforcement/no response). Responses influence amount of reinforcement

- Contingency studies- Rats bar-pressing for water, no response= no water, 1/20 responses= water, responding didn't make much differences so learning unlikely. Rats ceased responding if food presented despite no response for 1/20 trials. Contingency influences conditioning

- Human contingency learning (conditions for forming/ defining causal impressions, example from Wasserman et al.)

- False contingency

- Superstitious conditioning (Skinner's pigeons and pecking, shape experiment, human examples)

- Illusion of control (definition, P(O/R)subjective>P(O/R)objective, examples)

- Why does illusion of control occur (motivational, attributional biases, perceptual/cognitive, situational confusion, evidence from Langer)

- Types of illusory control (primary and secondary)

- Factors influencing occurrence of illusion of control (situational: confusion of chance and skill, task framing, need for outcome, mood and anxiety; individual differences: LOC orientation)

- Matute- two factors of illusion of control: over-response in situations, struggle to withhold response to see what happens

- Cognitive research- Thompson et al. control heuristic

Page 3

Theories of avoidance learning

- Shuttle box

- Two-factor theory- CC and OC, support from Rescorla and LoLordo, Miller, Kamin et al.

- Problems with two-factor theory- as animals avoid problem, fear reduces e.g. car crash example

- One-factor theory- OC, Sidman, Herrnstein and Hineline

- Rival cognitive theory- Seligman

- Critiques- Bolles: artificiality, species-specific defence reactions, easier to teach behaviours that align with natural behaviour systems. Mazur: purely semantic, valid in own right

- Extinguishing avoidance+ applications (response blocking, flooding, used for phobias)

Page 4

Learned Helplessness (methodology and theory)

- Learned helplessness definition- attributed to motivational and cognitive deficit

- Maier and Seligman- dogs in shock harnesses, no training/controllable pretraining/uncontrollable pretraining --> uncontrollable lay down and whimpered. Design issue: uncontrollable exposed to failure and punishment

- Yoked design- triadic, uncontrollable group subjected to same sequence of shocks as controllable. Design issue: controllable group might solve task quickly, so fewer shocks near the end could be attributed to learning by yoked group (last-trials bias)

- Human helplessness experiments- loud noise stopped by pressing buttons, solvable post-task. Uncontrollable group performed worse in post-task, but did still learn (slower, less accurate)

- Learned helplessness and depression (similar performance, consistent with symptoms)

- Confounding of punishment and non-contingency

- Last trials bias- less aversive stimulation later on

- Confusion of failure and non-contingency- people who experience non-contingent outcomes also experience failure- Winefield et al. gave groups performance feedback after pretraining- uncontrollable group with failure feedback performed worse

- Individual differences- external LOC, depression, sensitive to punishment, high IQ

- Generalisation- students completed non-contingent task in one room, and then different task in another- didn't affect performance, effect doesn't generalise well- discrimination, cognitive fatigue, interference

- Lack of inevitability of learned helplessness effect (some more prone to LH than others, not just motivation and cognitive deficits)

- Revised theory of LH- attributional (internal/external, global/specific, stable/unstable)- little support for theory, difficult to measure

- Learned optimism and positive psychology

Page 5

How to treat behavioural disorders

Page 6

Comparison of major learning theories

Hull-Spence vs. Rescorla-Wagner

What are they measuring?- how the association between two stimuli changes over time depending on how many times the stimuli were presented together or apart

Formulas-

∆VA(t) = k(λ - VA(t-1))

∆VA(t) = k(λ - ƩV(t-1))

Learning curves- negatively accelerated (learning slows down until asymptote)

Extinction- A presented without US

HS- maximum surprise first extinction trial, loss in associative strength until asymptote, extinction learning curve like a wave

Blocking

HS- Doesn't predict blocking, suggests B will gain associative strength

RW- does predict blocking, B will stay as flat line on learning curve, A-US= strong CR for B, AB-US= weak CR for B

Inhibition-

HS cannot predict inhibition (B would stay neutral)

CS signals absence of US e.g. A-US= response, AB-no US= no response, B= conditioned inhibitor, B alone won't garner response

- Summation

A, AB, C

CB= weak, C= strong, then B is inhibitor

- Retardation

A-US, AB-no US, B= slow learning, D= fast learning

B loses associative strength below 0

Dopamine activity and RW

A-US, AX-US--> A blocks X

B-no US, BY= US --> Y surprising

HS predicts equal learning about X and Y

Less dopamine activity after X because A has blocked X, and higher dopamine activity because of surprise

Catastrophic interference problem-

RW model suffers catastrophic interference- new learning erases old learning (not true)

Latent inhibition

CS presented without consequence, then presented with consequence, learning is slower than with novel CS

HS and RW can't explain latent inhibition- nothing should happen to light in exposure phase, so AS is 0

Attention-

Automatic attendance to stimuli assumed, RW predicts Group Pretrain will learn faster due to associative strength (false)

Simultaneous discrimination procedures consistent with RW-

A-no US, AB-US--> B is excitatory (positive)

A-US, AB- no US --> B is inhibitory (negative)

Page 7

Comparison of major attention theories

Mackintosh Selective Attention Theory

- Can explain latent inhibition -> light not reinforced in exposure phase, so less attention in second phase

- Two-stage process- pay attention to stimuli, form associations between stimuli

- Inverse law of selective attention- limited attentional resource

- Attention allocation- if attended stimulus is reinforced, it is likely to be attended to again

- Blocking- attention allocated to A in Phase 1, little attentional resource for B in Phase 2

- Attention proportional to associative strength

Pearce-Hall Attentional Model

Mackintosh predicts fast learning for Group Pretrain (A-->little US, A-->large US= fast learning) -A reinforced in first phase, so more attention in second phase too- wrong

Group Pretrain actually learned slower because of diminished surprise- animals knew what A predicted after phase 1, so stopped paying attention

Assumptions: learning proportional to attention (like M), attention proportional to surprise, unpredictable events enhance attention, once US is fully predicted by CS attention diminishes

Kaye and Pearce

Orienting responses to light CS- standing on hind legs

Orienting response diminishes with habituation- not followed by US

Continuous reinforcement: light always followed by US (100%)

Partial reinforcement: light sometimes followed by US (50%)

No reinforcement: light never followed by US (0%)

Mackintosh predicts Continuous will display most ORs- best associative strength

PH predicts Partial will display most ORs- most surprising

Extinction is faster for CR- not what PH predicts, more attention for PR should mean faster extinction. Consistent with Mackintosh (loses associative strength quickly)

Attention is proportional to both surprise and associative strength, accounted for in hybrid models

Page 8

Hippocampus and neocortex theories of memory

Provides solution for catastrophic interference problem- memories in multiple brain structures

Similar versions of memories stored in the hippocampus and neocortex

Hippocampus: initial storage of new memories, associations learnt quickly, temporary store due to rapid synaptic plasticity

Neocortex: simultaneous gradual storage of memories, associations learnt slowly, new knowledge incorporated with existing knowledge, permanent store due to slow but stable synaptic plasticity

Damage:

Hippocampus: can't form new episodic memories, but can recall procedural memories. Retrograde amnesia (newer memories forgotten quicker), indicates older memories are stored in neocortex, independent from hippocampus

Celebrity faces over decades- all faces recalled, fMRIs showed activity in hippocampus for newer faces, and activity in entorhinal cortex for older faces

Page 9

Reconsolidation theory of memory

Synaptic plasticity

- Synapses between neurons gradually become more efficient at transmitting signals (learning)

- To permanently remember a memory, it must be consolidated- neurons produce proteins that strengthen synaptic transmission

- Drugs that affect protein synthesis during consolidation interfere with synaptic plasticity and prevent memory from forming

- Even consolidated memories must be reconsolidated when recalled- drugs that inhibit protein synthesis during retrieval lead to weakening of memory

Nader et al.:

- Rats and fear

- Fearful memories stored in amygdala and rely on intact amygdala for recall. Nader et al. injected protein synthesis inhibitor when memory was reactivated- drug was only effective if the rats were recalling memory, no effect if they weren't

Applications

- Attenuation of old memories- PTSD

PTSD patients recorded their story, injected with protein synthesis inhibitor, then listened to story again- emotional response was reduced

Page 10

Occasion setting: Holland's theory

- Hierarchical associations which inhibit or activate other associations

- Rely on specific situations

- Four discrimination procedures:

Simultaneous feature positive discrimination

A-no US, AB-US --> B is excitatory, reinforcement occurs in presence of B

Serial feature positive discrimination

A-no US, B-A-US--> B precedes A, A is excitatory but only in presence of B, B is neutral alone but sets occasion for A to be reinforced

Simultaneous feature negative discrimination

A-US, AB- no US--> B is inhibitory, conditioned inhibition, B passes summation and retardation tests for conditioned inhibition

Serial feature negative discrimination

A-US, B-A-no US--> B is negative occasion setter, inhibits A-US association, doesn't pass summation or retardation tests

- Simultaneous is consistent with RW, serial is not

- Serial procedures generate occasion setters due to lack of contiguity- longer delay between OS and US, making it less likely to be directly associated to US

- Ross and Holland- B controls responding on simultaneous, A controls responding on serial:

Simultaneous: tone-no US, tone+light-US--> rats developed rearing behaviour associated with light (B)

Serial: tone- no US, light-tone-US--> rats developed head-jerking behaviour associated with auditory (A)

- Holland- occasion setters generate hierarchical and direct associations simultaneously

Experiment- negative discriminations

Simultaneous-> A-US, AB-no US, B is inhibitory, A is excitatory

Serial--> A-US, B-A-No US, B is negative occasion setter/neutral, A is excitatory

B passes inhibition tests for simultaneous but not serial

Retardation test for simultaneous- B no longer inhibitory, slowly becomes excitatory. When paired with A again, it increases responding

Retardation test for serial- B is excitatory but keeps negative occasion setter properties for A- still decreased responses when paired with A

- Contextual stimuli as occasion setters- perform certain behaviours in certain settings

- Contextual cues e.g. anxiety associations in phobias, or CS-> craving responses in addiction-- treatment trying to directly extinguish context might not work