Pavlovian (classical) conditioning Flashcards Preview

PSY2304 Biological Basis of Behaviour > Pavlovian (classical) conditioning > Flashcards

Flashcards in Pavlovian (classical) conditioning Deck (23)
Loading flashcards...
1

Pavlov initial experiment

food (UCS) --> salivation (UCR)

tuning fork (NS) --> no salivation (nonCR)

tuning fork + food --> salivation (UCR)

tuning fork (CS) --> salivation (CR)

2

pavlovian conditioning works

In all vertebrate species that have been tested; at least some invertebrates

For any CS an animal can detect (even a bell - though Pavlov probably never used one)

With spinal reflexes (e.g. in brain-damaged subjects)

In humans, e.g. eye-blink, GSR, knee-jerk - not used as much nowadays

3

common Pavlovian procedures

Eye blink to an air puff in humans, GSR

Key pecking in pigeons (“autoshaping”)

Taste aversion in rats (flavour plus illness)

Conditioned suppression (of lever pressing) in rats again

4

basic effects

1. Stimulus Generalisation
2. Extinction
3. Overshadowing
4. Blocking

5

stimulus generalisation

the further away from the initial CS the weaker the CR

e.g. touch, pitch, colour... – brighter/darker light, lower/higher pitch tone

see notes

6

extinction

Start with a well trained CS, then repeatedly present on its own – CR reduces

see notes

7

conditioned suppression

In this procedure the rat (typically) is trained to press a lever for food pellets.

After responding is established, the lever is withdrawn and the pairing of CS (e.g. light, tone) and US (normally shock) of interest takes place.

Then testing is performed with the lever back in the chamber.

The rate of responding during the CS (=RCS) and just before the CS (=Rpre-CS) is recorded, and a suppression ratio is calculated = "RCS" /"(RCS+Rpre−CS)" .

This measure has the property that for good conditioning it will approach 0, and for weak conditioning it will approach 0.5.
- lower score = stronger conditioning

8

overshadowing

Stage 1: Train compound of light and noise, or train the light and noise separately. For different groups use different intensities of noise (n=weak noise, N=intense noise).

Stage 2: Test the light and noise.

9

Mackintosh (1976)

Learning with noise overshadows the training with the light

Light also overshadows weak noise

Intense noise has some unconditioned properties as well – less of an effect of overshadowing

Train 2 stim together – one overshadows other and gains all the learning – usually more intense

see notes

10

blocking (Kamin, 1969)

Stage 1: noise →shock noise→ CR (fear)
Stage 2: noise + light → shock noise + light→ CR
Test: light light→ little suppression (0.45)

Surprise missing?

Control condition (no stage 1) gives suppression of 0.05 to the light on test.

With rats

Learnt very little from the light

Pre-training in stage 1 blocked learning to light in stage 2

see notes

11

blocking/overshadowing exp

A, B, C etc. are foods (Avocado, Bacon…). "+" means an allergic reaction. "?" asks for a rating – high = yes, low = no.

see notes

Class Results
The Y axis shows ratings. Higher means more likely to cause an allergic reaction. B shows blocking relative to C/D, and C/D show overshadowing relative to E.

The objection often made to demonstrations of this type is that people know what you’re doing, and you are testing their memory rather than their learning. We’ve recently addressed that point using an incidental paradigm in McLaren, Forrest, McLaren, Jones, Aitken and Mackintosh (2014, Neurobiology of Learning and Memory, on ELE). You still get overshadowing.

12

timing and conditioning

see notes

13

trace conditioning as a function of interval

Higher latency = better conditioning

4-8s = optimal conditioning

see notes

14

CS-US pairing not enough

Rescorla - “truly random control” (involves both partial reinforcement and free-US) leads to no conditioning despite some CS-US pairing: contingency is necessary

Kamin “blocking” - no acquisition when CS2 is paired with a US to which a strong CR to CS1 has already been established, if CS1 and CS2 are always presented together: is surprise necessary? – need learning to be novel – cannot be redundant

These and other phenomena were addressed by the Rescorla-Wagner theory

15

Rescorla-Wagner (R-W) learning rule

LEARN FROM TUTORIAL CARDS

see notes

Bear in mind that associative strength as computed by Rescorla-Wagner does not necessarily translate directly into responding – but we usually assume that the relationship between the two is monotonic. - see notes

16

contingency

Defined as p(O|R)-p(O|R')

Wasserman et al. (1983)

In this experiment your task is to find out whether tapping a telegraph key (the response, R) has any effect on the occurrence of a white light (the outcome, O).

It is to your advantage to tap some of the time and not tap (this is R') some of the time...

After each problem, choose a number on the -100 to 100 rating scale that best characterises the degree to which your tapping affected the occurrence of the white light.

The problems will be separated by a minute or two to allow you to make your rating.

Black – prob didn’t tap = .5
Blue – prob didn’t tap = .875

Ratings – have learnt contingencies

see notes

The Rescorla (1968) graph again.

This is evidence that rats can also track contingencies. As p(US|noCS) increases, the contingency,
p(US|CS) – p(US|noCS) decreases, and conditioning becomes less effective (higher suppression ratios).

see notes

17

timeline diagrams

see notes

18

significance of Pavlovian conditioning - then

A form of learning that could be described entirely objectively.

Provided a scientific underpinning for the concept of an association.

Contributed to the dominance of behaviorism in early twentieth century psychology; adopted as an account of all learning

19

significance of Pavlovian conditioning - now

Reliable, widespread phenomenon with determinable laws.

Practically useful, e.g. behavior therapy.

May account for involuntary, dysfunctional forms of human learning e.g. phobias.

Continued investigation of theories of Pavlovian conditioning (Rescorla-Wagner, Pearce-Hall, Dickinson, etc) – including some that are cognitive in nature, i.e. require animals to have mental representations of CS and UCS

It is not an explanation of all learning, but may give insight into fundamental learning processes. And these processes may well apply to humans.

20

human Pavlovian conditioning - case study

expectancy rating

SCR

Shown stimuli – paired with shock/light

21

autonomic conditioning

Note how over trials two things happen:

1) The participants expectancy of shock rises for S+ and drops for S-.

2) The change in skin conductance for S+ becomes larger than that for S-.

On this basis Lovibond and others have claimed that this is a form of Pavlovian conditioning in humans.

see notes

By separating out those participants who are aware of the contingencies (getting shocked to S+ not S-) from those who are not, it is possible to demonstrate that the effect is driven by the participants who have an expectancy of shock.

There is little evidence of conditioning in those who do not.

Hence, the claim that Pavlovian conditioning in humans requires a conscious cognitive expectancy of the US, and may not be the same as in other animals (e.g. the rat).

see notes

22

Hughdahl and Ohman (1977)

Further evidence in support of this view comes from the observation that if you now tell participants that no more shocks will occur - then the response quickly disappears (below left), in contrast to those tested in extinction (top left) but not given this information

see notes

If we use “fear relevant” stimuli (e.g. images of spiders and snakes) in this electrodermal conditioning paradigm then the results change.

There is now evidence that the responses are not entirely governed by conscious expectancy - and instead that there may be an automatic component that is more properly regarded as being due to Pavlovian conditioning.

see notes

23

McAndrew et al. (2012)

Given shocks – brown cylinder/pink square – 50% chance of shock

Rate expectancy of whether getting shock

Way respond depends on what recently experienced

A “run length” of -2, for example, means that we are looking at a trial following two non-reinforced (no shock)trials.

A run length of +2 would involve a trial following two reinforced (shock) trials.

Whether the current trial is reinforced or not does not come into consideration.

+1 = shock trial before

+2 = 2 shock trials before

-1 = no shock but brown cylinder – extinction

see notes

As no. Shock trial increase, expectancy of another one decreases – gamblers fallacy – but link between cylinder and shock incrwased

Expectancy not governing emotional reaction

see notes