Lecture 4 new.pptx

Post R-W: SOP model/more recent attentional models Reminder: Latent inhibition is one failure of R-W Unreinforced pre-exposure of the CS impairs subsequent learning. preexposure conditioning test Robert Lubow Wagner’s ‘Sometimes Opponent Process’ model (1976) Allan Wagner Rescorla-Wagner: expectatio...

Post R-W: SOP model/more recent attentional models Reminder: Latent inhibition is one failure of R-W Unreinforced pre-exposure of the CS impairs subsequent learning. preexposure conditioning test Robert Lubow Wagner’s ‘Sometimes Opponent Process’ model (1976) Allan Wagner Rescorla-Wagner: expectation of the US reduces its effectiveness Same could apply for the any stimulus (including the CS) SOP model provides a framework for understanding this effect. SOP model Based on classical ‘Opponent Process’ model famously used in studies of addiction (one of your PS2110 essay topics) ‘Sometimes’ opponent because the B process may not always be opposite in Wagner’s SOP model (Solomon, 1980) Central role of context in SOP model Presentations of the light in a particular context turns the context into an CS in its own right The context develops predictive value (anticipation of the light) Wagner (1976; 1981) Stimuli can be activated in STM either directly or associatively Full processing 1 A1 Strong response 0 pre se nta tio n Decay from A1 to A2 Sti mu lus Activation Facility to associate Inactive A2 tion a v i t ac e v i t ocia s s A Context can activate a representation of a stimulus (light, shock) in A2 Limited processing Weak response Difficulty to associate Wagner (1976; 1981) A2 activation either directly or associatively “Self-generated priming” “Associative priming” A1 and A2 (opponent) processes resulting from morphine injection A2 A1 Conditioning (with context as CS) activates the A2 representation (increased locomotion) Conditioned response is increased activity. m-e = conditioning m-hc = morphine in home cage (control) s = saline injection (control) Unconditioned eyeblink response has an A1 and A2 component Airpuff fast A1 (reflex arc) Eyeblink slow A2 (fear) Tone CS activates the slower A2 ‘fear’ response. Circuits anatomically dissociable. (Further elaborations not covered in this lecture). Problems with the R-W model: Latent inhibition Pre-exposure to the to-be-CS impairs conditioning. Light Light Light 80 Control Preexposed % leg flexion 70 60 50 40 30 20 10 0 1 2 3 Blocks of 20 trials 4 something happens during the preexposure to the to be CS that impairs Wagner (1976): Associative priming of Stimulus by Full A1 activation of stimulus context 1 A1 Strong Response High Processing Associations Activation Stimulus decays into A2 state 0 A2 A2 A2 Marginal activation by context Weak Response Low Processing No Association Inactive Context-Stimulus association Inactive Subsequent co-presentation Wagner (1981) Context—stimulus association accounts for Latent Inhibition Pre-exposed Light Context Context Light Light Control Preexposed Control Lubow (1965) Attentional models 1. Mackintosh (1975) model Core idea: The animal that rapidly detects signals for relevant events is more likely to survive than one that ignores these stimuli. Nick Mackintosh Animals are likely to pay more attention to stimuli that predict important events. Le Pelley et al. (2005) basic procedure Human participants are presented with trials in which a virtual patient experiences different reactions to different combinations of foods: Strawberry & Grapes O1] Banana O2] & Grapes  Allergy  Diarrhoea [Ax  [Bx  A and B are good predictors of Outcomes 1 and 2 x (grapes), is a poor predictor of those outcomes Le Pelley, M. E., Oakeshott, S. M., Wills, A. J., & McLaren, I. P. L. (2005). The Outcome Specificity of Learned Predictiveness Effects: Parallels Between Human Causal Learning and Animal Conditioning. Journal of Experimental Psychology: Animal Behavior Processes, 31(2), 226–236. In support of Mackintosh’s theory: Le Pelley et al. (2005) humans pay attention to good predictors of outcomes (allergies) AvO1 BvO2 CxO2 DxO1 ABCD: good predictors of O1 and O2 v and x: irrelevant Test: Prediction of O3 AxO3 CvO3 All cues relevant Participants were more confident that AC (relevant in stage 1) produce O3 than vx, irrelevant in stage 1 This result is based upon the predictive history of the stimuli AC vx The Mackintosh model Factors contributing to learning ΔVcs = (processing of CS) of US) Remember that (processing in R-W model α is fixed Now, α adopts values between 0.05 and 1 A new CS will have relatively high associability, α≈0.8 Experience will: increase αCS if perceived as a good predictor of the US = α (λ-V) cs cs decrease αCSΔV if perceived as a poor predictor of the US Δαcs > 0 Δαcs < 0 if if (Vcs > Vother cues) (Vcs < Vother cues) Le Pelley et al. (2005): changes in the salience of stimuli during stage 1 AvO1 BvO2 AxO3 CvO3 CxO2 DxO1 Assuming αAC=0.8 and αVX=0.8 αAC αVX (VAC > VVX) (VAC > VVX) αAC≈1 αVX≈0.05 AC vx Further assessment of Mackintosh The Hall-Pearce effect (Hall & Pearce, 1979) theory Group Phase 1 Phase 2 Pre-exposed A shock A  SHOCK! Control B shock A  SHOCK! Phase 1 Assuming αA=0.8 and αB=0.8 Phase 2 Pre-exp: αA≈1 Control: αA=0.8 Both αA and αB will increase as they reliably predict the outcome (shock). ΔVcs = αcs (λ-V) After Phase 1, αA≈1 and αB≈1 Pre-exp: ΔVA = 1 (1-0) = Control: ΔVA = 0.8 (1-0)= We can expect faster 1 0.8 Further assessment of Mackintosh The Hall-Pearce effect (Hall & Pearce, 1979) theory Group Phase 1 Pre-exposed A shock A  SHOCK! Control B shock A  SHOCK! Low 0.5 Suppression ratio 0.45 0.4 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0 Strong Phase 2 Preexp Contro l 1 2 3 Trials 4 5 Group Pre-exposed showed impaired learning According to that, αA decreased rather than increased in Phase 1 This result challenges the main assumption of Mackintosh’s theory 2. Pearce-Hall model (1980) This kind of results led Pearce and Hall to propose an alternative attentional theory of learning John Pearce Geoffrey Hall The Pearce-Hall model Basic assumptions ΔVcs = (processing of CS) (processing of US) Stimuli that fully predict their consequences will be processed automatically without attentional waste. Stimuli with low predictive value (followed by surprising events) attract attention and are fully processed (controlled processing) The Pearce-Hall model Basic assumptions ΔVcs = (processing of CS) (processing of US) ΔVcs = σcs (λ-V) Processing of the CS depends on σCS σ adopts values between 1 and 0.5 σCS ≈ 1 if VCS That is, the better a CS predicts an outcome (the 0attract. US), the less attention ≈ it will This is the opposite to the attentional parameter, α, ≈ 0.5(1975). if put forwardσ by CSMackintosh The Pearce-Hall model Empirical evidence: The Kaye & Pearce (1984) experiment The orienting response (OR): measures the attention to a stimulus A novel light stimulus elicits an OR; σLight=0.8 Repeated presentations of the light result in habituation of the OR; σLight=0.5 The Pearce-Hall model Empirical evidence: The Kaye & Pearce (1984) experiment Groups Pre-Test Test ‘None’ L L L L L L L … ‘Continuous’ L L+ L+ L+ L+ L+ L+ … ‘Partial’(low predictive value) L L+ L L+ L+ L L … L = light; + = food. Initial level of σLight = 0.8 Groups None and Continuous L is a good predictor of outcome (either + or its absence) Attention will gradually decrease Group Partial L is a bad predictor of outcome (+ and no+ in alternated trials) Attention will be kept high σLight = 0.8 The Pearce-Hall model Empirical evidence: The Kaye & Pearce (1984) experiment Pre-Test Test None L L L L L L L … Continuous L L+ L+ L+ L+ L+ L+ … Partial (low predictive value) L L+ L L+ L+ L L … Percentage of Observations (OR) 80 Orienting Response 60 None Continuous Partial 40 Series4 Series5 Series6 20 0 Pre-Test 2 4 6 8 10 12 14 Sessions Attention to L decreased when it was a good predictor of its consequences; and was maintained high when its The study by Le Pelley et al (2005) supports the Mackintosh model. Results reported by Hall & Pearce (1979) and Kaye & Pearce (1984) support the Pearce-Hall model. How can we reconcile this situation? 3. A hybrid attentional Model LePelley (2004) Pearce & Mackintosh (2010) α: salience parameter characterised by Mackintosh (1975) σ: salience parameter characterised by Pearce & Hall (1980) VCS= αCS σCS (λ-V) To recap: Different theories explain different aspects of learning: Recorla-Wagner (Wagner): US processing Mackintosh and Pearce-Hall: CS processing Development of these theories has led to wide range of experimental findings Recent attempts at combining existing theories to produce more powerful models It is not over: new theories are proposed to account for phenomena that lie beyond the scope of traditional theories reviewed here (e.g., HeiDI model; Honey, Iliescu & Dwyer, Psychological Review, 2020)