103_4_full

103_4_full - PSYC 103 Winter 2010 Lecture 4 Nature of...

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: PSYC 103 Winter 2010 Lecture 4 Nature of associative learning A memory model of associative learning -  Detection of the CS or the US in the sensory register excites a memory representation of these events - The US also activates a response to produce the UR - Learning can result in S-R associations ( e.g. Aplysia) or CS-US associations 2 What is learned? Two theories S-R learning: CS elicits CR directly S-S learning: CS activates representation of US S-S / S-R Differentiation Nature of associative learning A memory model of associative learning CS-US learning: Colwill & Motzkin (1994): Rat subjects Conditioning Devaluation Tone → Food Test Food → Illness Tone vs. Light Light → Sucrose Sucrose → nothing Weaker responding to Tone than Light -  Results are consistent with CS-US account of learning but are difficult to explain with an S-R account of learning -  Holland (1990) conducted a similar experiment and found, in the final test stage, that when allowed to drink water during the test stimuli, rats made ingestive responses (e.g paw licking) to the stimulus whose US had not been poisoned, and aversive responses (e.g. head shaking) to the other stimulus 5 US devaluation: Example Holloway and Domjan 1993 Sexual Pavlovian conditioning in the domesticated quail Light (CS) paired with access to female Copulation at the end of the trial --> Light elicits approach response What happens when drive is reduced? Types of stimulus-stimulus learning A. Serial conditioning -  A sequence of stimuli precedes the US, e.g.: Light → Tone → Food Holland & Ross (1981) showed that after such training, the response normally evoked by the light (e.g. rearing) was replaced by the response normally evoked by the tone (e.g. head jerking) - The presence of the light excites a memory of the tone and causes the rat to respond as if it were present B. Sensory preconditioning Rizley & Rescorla (1972): Stage 1 Light → Tone Stage 2 Test Tone → Shock Light Fear CR At test, the light excites a memory of the tone, which in turn excites a memory of shock 7 Types of stimulus-stimulus learning C. Second-order conditioning Stage 1 Stage 2 CS1→US CS2 →CS1 CS2 comes to evoke a CR -  CS2 enters into an association with CS1, which in turn excites a memory of the US as a consequence of stage 1 training. -  Rashotte, Griffin & Sisk (1977): Autoshaping procedure with pigeons. White and blue key-lights as CS1 and CS2. CS1 extinguished after stage 2. Results showed that CR to CS was weakened by this treatment. Supports the above analysis. -  Rizley & Rescorla (1972): Conditioned suppression procedure with rats. Tone and light as CS1 and CS2. They also extinguished CS1 after stage 2. Results showed that CR to CS was NOT weakened by this treatment. Contradicts the above analysis. -  Rescorla (1980): When CS1 and CS2 are similar = CS2-CS1 learning. When CS1 and CS2 are dissimilar = CS1-CR learning. 8 Nature of US representations Two characteristics/properties of US representations: (1)  Specific – Characteristics that make the US unique and recognizable (e.g. its duration, flavor) CS can become associated with the specific characteristics of the US (e.g. its flavor) (2) Affective – Characteristics that the US has in common with other stimuli and reflect its motivational properties Blocking is unaffected by a change in the US from water to food between stages 1 to 2. Shows the CS must have become associated with the general affective properties of the US in stage 1 Associations between CS. Both these properties can form during excitatory cond. 9 Nature of the conditioned response The US can influence the CR A. Consummatory CR - Evoked when a CS retrieves specific properties of the US -  Mimics the response of the US Example: key pecks of pigeons differ depending on whether the CS signals food or water B. Preparatory CR - Evoked when a CS retrieves affective properties of the US. -  Appetitive CR might be an increase in activity … or approach -  Aversive CR might a reduction in activity… or withdrawal C. Compensatory CR - A CR that opposes, or compensates for the UR -  Frequently examined in studies of drug tolerance 10 Conditioned response : Influence of the US on the CR C. Compensatory CRs Siegel (1977) •  Reduction of the analgesic effects of morphine in rats with training •  Injection/handling cues = CS, Morphine = US •  Reduction in the effectiveness of CS (i.e. tolerance) = Conditioning •  Saline (placebo) injections lead to extinction; without injections full tolerance remains. Tolerance (compensatory CR) develops Saline injections No injection 11 Reflexive nature of the conditioned response CR can be maladaptive… Hearst & Jenkins (1974) - Pigeons autoshaped in a long arena. Key light and hopper separated by 60 cm. Subjects often missed the US as a consequence of not getting to the hopper in time. -  CR was still maintained! Williams & Williams (1969) -  Omission schedule -  Pigeons given conventional autoshaping. US omitted if CR (pecking) was performed. Pecking still continued! 12 Blocking Stronger CR in Group C Than Group E Kamin (1969) Group Element conditioning Compound conditioning Test Group E Noise → Shock Noise & Light → Shock Light Group C - Noise & Light → Shock Light “…perhaps, for an increment in an associative connection to occur, it is necessary that the US instigate some ‘mental work’ on behalf of the animal. This mental work will occur only if the US is unpredicted – if it in some sense ‘surprises’ the animal.” Kamin, 1969, p. 59 Led to the development of the influential Rescorla–Wagner (1972) model 13 Rescorla–Wagner model Typical learning curve Conditioning with a single CS Typical characteristics of a learning curve: -  At first CR is low (because CS is novel), -  Then, a rapid increase in CR, -  Finally, very little change in CR Rescorla & Wagner: Assumptions Adapted from Kehoe et al. 1994. 1.  Strength of the CS–US association = strength of the CR 2.  Thus CS-US association must be zero at first 3.  Then CS-US association grows quickly at first, before slowing 14 Rescorla–Wagner model Conditioning with a single CS Associative strength: Strength of the connection between internal representations of the CS and US. The change in associative strength ( ΔV ) on each conditioning is determined by: ΔV = α(λ-V) α = Learning rate parameter determined by the properties of the CS & US V= Current strength of the CS→US association λ = Magnitude of the US, reflects the maximum CS →US association (learning) possible Kamin’s “surprise” is expressed as (λ-V) : “What you got” - “What you expected” Learning on each trial ‘ΔV’ is proportional to (λ - V) 15 Rescorla–Wagner model Conditioning with a single CS ΔV = α(λ-V) An application to conditioning with a single CS Associative Strength (V) Before conditioning (Trial 0) assume λ = 100, α = 0.2, and, as no learning has yet taken place, V = 0 60 50 40 30 20 10 0 Trial 0 Trial 1 Trial 2 Trial 3 Trial 4 Trial 16 Rescorla–Wagner model Conditioning with a single CS ΔV = α(λ-V) An application to conditioning with a single CS On Trial 1 therefore the change in associative strength will be: ΔV = 0.2(100-0) = 20 Associative Strength (V) 60 50 40 30 20 10 0 Trial 0 Trial 1 Trial 2 Trial 3 Trial 4 Trial 17 Rescorla–Wagner model Conditioning with a single CS ΔV = α(λ-V) An application to conditioning with a single CS On Trial 2, the change in associative strength will be less, for now V= 20 Associative Strength (V) ΔV = 0.2(100-20) = 16 60 50 40 30 20 10 0 Trial 0 Trial 1 Trial 2 Trial 3 Trial 4 Trial 18 Rescorla–Wagner model Conditioning with a single CS ΔV = α(λ-V) An application to conditioning with a single CS On Trial 3, the change in associative strength will be even less, for now V = 36 ΔV = 0.2(100-36) = 12.8 Associative Strength (V) 60 50 40 30 20 10 0 Trial 0 Trial 1 Trial 2 Trial 3 Trial 4 Trial 19 Rescorla–Wagner model Conditioning with a single CS ΔV = α(λ-V) An application to conditioning with a single CS On Trial 4, the change in associative strength will be even less, for now V= 49 ΔV = 0.2(100- 48.8) = 10.2 Associative Strength (V) 60 50 40 30 20 10 0 Trial 0 Trial 1 Trial 2 Trial 3 Trial 4 Trial And so on until V = λ (100), and therefore ΔV = 0 20 Rescorla–Wagner model Extinction Value of 0 is substituted for λ : Suppose that conditioning had previously progressed to asymptote, i.e. V = λ On extinction trial 1, the the change in associative strength will be: Extinction ΔV = 0.2(0-100) = -20 On extinction trial 2, the change in associative strength will be less, because now V= 80 ΔV = 0.2(0-80) = -16 And so on until V = 0, and therefore ΔV = 0 21 Rescorla–Wagner model US intensity - Variations in the intensity of the US are simulated by varying the value of λ Large US (λ) Intense US = high value of λ (e.g. 100) Weaker US = lower value of λ (e.g. 50) - Reducing the value of λ has two effects: (1) Rate of conditioning is slower Small US (λ) (2) Ultimate level of conditioning is less 22 Rescorla–Wagner model US intensity The acquisition of conditioned suppression to a noise conditioned stimulus (CS) by two groups of rats that received either a 0.49 milliamp (Group Weak) or a 0.85 milliamp (Group Strong) shock unconditioned stimulus (US) (adapted from Annau & Kamin, 1961). Rescorla–Wagner model CS intensity - Variations in the intensity of the CS are simulated by varying the value of α Intense CS = High value of α (e.g. 1) Weak CS = Low value of α (e.g. just > 0) - Varying the value of α only influences the rate of conditioning, not maximum learning - Kamin & Schaub (1963), using a conditioned suppression procedure with rats confirmed faster conditioning with a 81 dB noise than a 49 dB noise, but no differences in the ultimate level of conditioning. (adapted from Kamin & Schaub, 1963). 24 Rescorla–Wagner model Conditioning with a compound CS -  When two or more stimuli are presented in compound e.g. A and X, their associative strengths are summed (VALL) to determine CR: VALL = VA +VX -  VALL determines the changes in associative strength on each trial: ΔV = α(λ-VALL) -  Thus, change in associative strength determined by the discrepancy between the combined associative strengths of all stimuli present and λ 25 Rescorla–Wagner model Conditioning with a compound CS Blocking Group Element conditioning Compound conditioning Test Group E Noise → Shock Noise & Light → Shock Light Group C - Noise & Light → Shock Light - V Noise = λ after the element conditioning for Group E Therefore (λ-VALL) = 0 during compound conditioning -  No increments in associative strength will take place to the light Overshadowing -  After compound conditioning in Group C, V Noise + V Light = λ -  Therefore V Noise or V Light alone < λ - Conditioning to the light alone would result in V Light = λ 26 ...
View Full Document

This note was uploaded on 03/24/2012 for the course PSYC 103 taught by Professor Pearlberg during the Spring '07 term at UCSD.

Ask a homework question - tutors are online