Pavlovian Conditioning Flashcards
PAVLOVIAN CONDITIONING
WORKS FOR…
- all tested vertebrate species/at least some invertebrates
- any CS an animal can detect (ie. bell)
- w/spinal reflexes (ie. in brain-damaged pps)
- w/humans (ie. blinking/GSR/knee-jerk)
PROCEDURES
CS = > AIR-PUFF
- eye-blink -> air-puff (humans)
CS = > SHOCK
- electrodermal conditioning (humans)
AUTOSHAPING
- key pecking (pigeons)
FLAVOUR + ILLNESS
- taste aversion (rats)
LEVER PUSHING
- conditioned suppression (rats)
BASIC PARAMETERS/EFFECTS
STIMULUS GENERALISATION
EXTINCTION
TIMING
RESPONDING
STIMUUS GENERALISATION
- condition CS then test during gradual change
- the more unlike og CS it is => less responding
- graph curve = generalisation gradient
- the further away from initial CS -> weaker CR
- ie. touch/pitch/colour
EXTINCTION
- starting w/well trained CS then repeatedly present alone (ie. w/o reinforcer used in conditioning)
- CR gets weaker over presentations aka. undoing conditioning BUT NOT elimination of association
- animal learns opposing inhibitory association between CS/US negating og excitatory association effect
STIMULUS SUBSTITUTION
BEFORE CONDITIONING
- US (food) => UR (salivation)
- NS (tone) => NO UR
AFTER CONDITIONING
- conditioning aka. US (food) + NS (tone)
- CS (tone) tested; tone associated w/food so tone alone => CR (salivation)
- CS acts as US BUT weaker
CONDITIONED SUPPRESSION
- rat trained to press lever for food pellets
- lever withdrawn post response; CS (light/tone) + US (shock) paired
- testing performed w/lever back
- response rate w/CS (=RCS)/before CS (=Rpre-CS) recorded; suppression ration calculated (RCS/RCS + Rpre-CS)
- good conditioning = 0; weak/no conditioning = 0.5
TIMING
- schedule used has powerful effect on how well it works
- delay conditioning works well w/stimulus + shock/stimulus + food
- trace conditioning = temporal gap between CS/US; less effective is gap = long
- simultaneous conditioning doesn’t work well; oft no results
- optimal gap between CS/US onset varies according to stimuli/reinforcers
TIMING: TRACE CONDITIONING
- tone => shock conditioning procedure
- 8s interval = optimal
- too much (ie. 150) = no conditioning
- too little (ie. 0) = not as affective
- negative intervals = backwards pairing egs; also not as effective
TIME: RESPONDING
- associative strength doesn’t necessarily translate directly to responding BUT usually assume relation between them is monotonic
- can lead to “all or nothing” learning
- aka. learning = smooth BUT behavioural transition = abrupt; associative strength must pass threshold before responding occurs naturally
PHENOMENA
OVERSHADOWING
BLOCKING
RESCORLA-WAGNER
CONTINGENCY
HUMAN PAVLOVIAN CONDITIONING
OVERSHADOWING
STAGE 1
- train compound of light/noise or separately light/noise
- for dif groups use dif noise intensities (n = weak noise; N = intense noise)
STAGE 2
- test light/noise
OVERSHADOWING EXAMPLE
MACKINTOSH (1976)
- low suppression ratio = good conditioning
- light trained alone = good; light + weak noise = good as no overshadowing
- BUT loud noise (LN) => reduced conditioning (light ONLY) aka. loud noise = overshadowing
- weak noise trained alone = good conditioning BUT w/light = weaker (Ln) aka. light overshadowed noise
- loud noise + LN training = complex; high suppression/weak conditioning > n as loud noise energises beh counteracting freezing
BLOCKING
KAMIN (1969)
Stage 1
- CS (case noise) conditioned w/shock so presentation followed by shock administration; continues until good learning aka. rat hears noise => freezes
Stage 2
- conditioning continues BUT light added w/noise; conditioned in compound
- test light + noise; noise = good conditioning BUT light = hardly any
- conditioning to light blocked by earlier noise training
- if you condition both light/noise together from outset -> light = good conditioning/freezing when tested alone
BLOCKING: HUMANS
CAMBRIGE EXP
- blocking/overshadowing in getting ill after eating food
- overshadowing group = blocking control too
- lots of other foods/outcomes mixed (filters) to make sure overall chance of outcome = 50/50
MCLAREN ET AL (2014)
- common objections = people know what you’re doing so testing memory > learning BUT recent incidental paradigm exp still shows overshadowing
RESCORLA-WAGNER THEORY
RESCORLA
- “truly random control” (partial reinforcement/free US) -> no conditioning despite some CS-US pairing; contingency = necessary
KAMIN
- blocking; no acquisition when CS2 paired w/US to which strong CR/CS1 already established if CS1/CS2 always presented together
RESCORLA-WAGNER: LEARNING RULE
V = AB (Y = EV)
- alpha = learning rate parameter determined by CS salience
- beta = learning rate parameter determined by outcome salience
- V = associative strength
- y = V asymptote; positive value on episodes when outcome = US w/magnitude determined by US effectiveness
- EV = sum of associative strengths of all CSs present during learning
CONTINGENCY
WASSERMAN ET AL. (1983)
- aka. p (O/R) - p (O/R’)
- Q = if tapping telegraph key (R) has effect on white light occurrence (O)
- advantageous to tap some time but not all (R’)
- choose number -100 - 100 best characterising degree to which tapping affected occurrence of white light
- problems separated by minute to allow rating
CONTIGENCY: HUMANS RESULTS
RED LINE POINT 1 (0 rating)
- p (light coming on if key NOT pressed) = .125 always; p (light coming on if key pressed) = .125
- aka. pressing/not pressing = no dif
- aka. contingency = .125 - .125 = 0
RED LINE POINT 2 (.5 rating)
- p (light coming on if key pressed) = .5
- so contingency = .5 - .125 = .375 aka. good positive contingency; rating reflects
- overall pps ratings match calculated contingencies
CONTIGENCY: RATS
RESCORLA (1968)
- rats can also track contingencies
- p (US/CS) - p (US/noCS) decreases -> less effective conditioning (higher suppression ratios)
- but WHY?
CONTINGENCY: EXPLAINED
- puzzle is always how presentations of outcome when cue/CS = not present could influence associations from cue -> outcome
- these degrade contingency between cue/outcome
- how does RW capture this?
- ANS = presentations of outcome in absence of cue -> context/outcome associations blocking cue/outcome associations to some extent
RW X CONTINGENCY
WASSERMAN
- keypress/light co-occur = strong association
- context there BUT no effect; associations form/extinguish
- keypress sometimes w/o light = weaker association; trials w/keypress/no light partially extinguishes association
CONTINGENCY!
- light occurring alone => context/light association
- can partially blocks learning between keypress/light in after-trials => weaker association than usual
- associative strength between CS/US = proportional to contingency in RWR
SIGNIFICANCE
- could be described entirely objectively
- provided scientific underpinning for association concept
- practically useful (ie. beh therapy)
- may account for involuntary/dysfunctional human learning (ie. phobias)
- continued theoretical investigation (RW/Pearce-Hall/Dickinson); some cognitive (ie. requiring mental reps of CS/UCS in animals)
- doesn’t explain all learning BUT may give insight into fundamental learning processes; all may apply to humans
AUTONOMIC CONDITIONING
LOVIBOND (1992)
- 2 things occur over trials:
1. pps shock expectancy rises (S+) and drops (S-)
2. skin conductance change (S+ > S-)
- claimed that AC = Pavlovian conditioning form in humans
AUTONOMIC CONDITIONING: EXPLAINED
LOVIBOND (1992)
- by separating pps aware of contingencies (getting shocked in S+ NOT S-) from unawares => possible to demonstrate that AC = driven by pps expecting shock; little conditioning evidence in those who don’t
- aka. claim that human Pavlovian conditioning requires conscious cognitive US expectancy may not be the same in other animals (ie. rats)
AUTOMATIC CONDITIONING: FURTHER EVIDENCE
HUGHDAHL & OHMAN (1977)
- if you now tell pps that no more shocks will occur -> response quickly disappears in contrast to those tested in extinction BUT not given info
- if “fear relevant” stimuli (ie. spider/snake pictures) used in electrodermal conditioning paradigm => results change
- evidence that responses = not entirely governed by conscious expectancy
- there may be automatic component more regarded as due to Pavlovian conditioning
SUMMARY
- evidence for Pavlovian conditioning in humans
- establishing this = complicated by conscious cognitive expectancy issues BUT criterion of control helps
- if stimulus response = both acquired via CS=>US pairings/not under conscious control -> may be classical conditioning instance in rat format
