18 - Learning interactions

Question 1

Evidence for two types of learning

Answer 1

• In any classical SS training situation the animals may also make responses – can’t stop incidental instrumental learning
• In any instrumental RS training situation other stimuli will be present – can’t stop incidental classical conditioning
• This means that in addition to your intended pairings other incidental pairings might happen
• So in your classical conditioning task where a stimulus predicts food the animal may learn they are responding to get food – instrumental (e.g. bird always pecks the keylight that predicts food)
• Or in your instrumental conditioning task where a response predicts food, the animal may learn that the stimuli present predict food (e,.g. the rat just learns that the lever predicts food, and the CR is to interact with it – which results in a lever press)

Question 2

Are there two types of learning?

Answer 2

• But if animals could use classical conditioning in an instrumental task, or instrumental conditioning in a classical task, how do we know there are two types of learning at all?
• Maybe the behaviour we see in classical and instrumental training is ALL classical conditioning, or ALL instrumental conditioning and scientists like parsimony!!
• So are there two types of learning? Use the known differences between them to test this
• Instrumental response is rational and goal directed – can adapt as long as goal is reached
• Form of classically conditioned CR is often rigid – tightly linked to UR

Question 3

Can animals ever do rational instrumental conditioning? Omission training

Answer 3

• Holland, 1979: Train classical light = food association
• Light predicts food but also magazine approach CR prevents food
• Light = food Light + magazine entry = no food
• Light paired with food – classical conditioning; measured magazine behaviour
• Omission – every time rat made magazine response in CS, food cancelled
  a) If magazine responding is classical conditioning, they should persist
  b) If magazine responding is operant conditioning, they should stop
• Rats in Groups Omission, Yoked and Unpaired run in triplets
• Group Omission 10s light paired with food unless rat responded. This is the master rat.
• Group Yoked 10s light paired with food every time its partner in Group Omission got food i.e. when it didn’t make CR
• Group Unpaired as many foods as its partner in Group Omission, but unpaired with light
• Group Control 10s light always paired with food.
• Yoked and Unpaired – same number of USs as Omission group regardless of their behaviour
• Results: Omission training eliminated magazine behaviour (this is purely instrumental)
• Experiment 2 does it all over again, but puts omission schedule on rearing behaviour
• Light predicts food but also rearing CR prevents food
• Light = food Light + rear = no food
• Light paired with food – classical conditioning; measured rearing
• Results: Omission training reduced but did not eliminate rearing behaviour
• Mostly classical (but partly instrumental)

Question 4

Another way to confirm instrumental learning

Answer 4

• Form of classically conditioned response is constrained by nature of US… biting the lever more likely than pressing it
• If CR is to respond one way, it can’t be to respond the other..
• Guinea-pigs were taught to turn their heads to one side in response to the sound of a buzzer, with a food reward.
• The habit could be reversed by rewarding the animal for turning its head, on the sound of the buzzer, to the other side.
• It is difficult to explain these results in terms of Pavlov’s theory of conditioned reflexes

Question 5

Other interactions

Answer 5

• Involuntary classical CRs may also interfere with operant response you want to train
• Pigeon CR to a box signalling food is moving about – making it harder to peck the key
• Long box autoshaping
• Some responses difficult to elicit in operant training e.g. training animal to avoid shocks
• They might freeze, interfering with any active response e.g. species-specific defence reactions, but making passive avoidance easy they might run, which makes active avoidance easy, but passive avoidance difficult.

Question 6

Training animals to avoid shocks

Answer 6

• Training rats to do different things to avoid shock; running is more natural response when shock’s about than standing!
• The stronger the hard-wired UR to e.g. shock, the harder it is to train animals to escape or avoid it
• To train an operant response you may need to work with their classical CRs

Question 7

Classically conditioned CSs affect instrumental performance

Answer 7

• Where does learned instrumental/operant response come from?
• Although we don’t have a clear theory of why the response is made in the first place, we do know what factors can increase and decrease rates of responding
• Avoidance responses rewarded by removing aversive USs (e.g. shock) before they’ve begun
• But how do you know shock is going to happen?
• Rat in chamber: Buzzer comes on, followed by shock
• BUT if rat responds during buzzer, shock is cancelled
• Buzzer is a classically conditioning CS – a warning signal that predicts shock
• One reason for avoidance responding is the presence of a signal for fear

Question 8

Pavlovian – instrumental transfer: two process theory

Answer 8

• Pavlovian CSs affect level of instrumental responding
• If responding is motivated by something nice, CSs predicting something nice  respond more
• If responding is motivated by something nice, CSs predicting something nasty  respond less
• If responding is motivated something nasty, a CS predicting something nasty  respond more
• If responding is motivated something nasty, a CS predicting something nice  respond less
• Boosting relevant motivational state boosts responding and because you can’t feel happy and sad at the same time!

Question 9

General Pavlovian: instrumental transfer

Answer 9

1. Estes, 1948: Rats trained that a tone signals food
   - Then trained to press a lever for food
   - Finally allowed to press the lever but stopped food deliveries and occasionally present the tone
   - “In every case introduction of the tone was followed by a temporary increase in the rate of responding”
2. Rescorla & LoLordo, 1965: Dogs trained to jump a barrier to avoid shock on Sidman avoidance schedule – no explicit signals for shock, only time.
   - Then train separately
   - One stimulus  shock (CS+) nasty
   - Another signalled absence of shock (CS-) nice
   - Rate of avoidance responding: increased by signal for shock, decreased by inhibitor for shock

Question 10

Specific Pavlovian: instrumental transfer

Answer 10

• A Pavlovian CS signalling a reward energises instrumental responding for that specific reward
• (Estes, 1943): Specific Pavlovian-Instrumental transfer (PIT) involves two rewards:
• A stimulus predicting one reward elevates performance of responding for the same reward (congruent responding) more than responding for the other (incongruent responding).
• Pavlovian phase Instrumental phase Test
• Tone  food left lever  food tone: left > right
• Click  suc right lever  sucrose click: right > left
• The food-paired tone selectively elevates performance of the left leverpress that also signals food

Question 11

Pavlovian instrumental transfer and addiction

Answer 11

• A reason for studying PIT is its relevance to addiction.
• The reward of drinking or smoking becomes associated with the sight of beer and cigarettes, and hence their packaging.
• So seeing a picture of a beer bottle can make you want to have a beer (but not a cigarette)!
• This why cigarette displays are restricted in the UK, and packets no longer have branding

Question 12

Specific Pavlovian: instrumental transfer 2

Answer 12

• Pavlovian conditioning: one child gets food, the other gets drinks
• CS1 food
• CS2 drinks
• Instrumental Conditioning: participant presses z key to see pictures of food, and the m key to see pictures of drinks
• Response zfood
• Response m drinks
• Test: “now press the z and m keys to get the food and drink pictures! you will not see the pictures you are getting”
• Food responses during the food CS
• Drink responses during the drink CS
• Food responses during the drink CS
• Drink responses during the food CS

Question 13

Contextual control and state dependence

Answer 13

• Memories, or associations, can be cued by context that was present when you formed it. Contexts can be places, smells, drug states…state dependence (or contextual control)
• Cases in which the environment/context of learning helps you retrieve a memory
• Recall better if test context matches training context

Question 14

Other examples

Answer 14

• The same stimulus may be associated with two different outcomes. Which association is retrieved may be conditional on the context in which the stimulus is presented.
• This allows associations to represent knowledge in a versatile way.
• The context appears to control access to the appropriate CSUS association to use in that context – the one that was operating in that context.

Question 15

Discriminative/ occasion-setting control of associations

Answer 15

• Stimulus or context that is present when an association
• is formed can also control access to that association in a special kind of state dependence
• When the association is instrumental RS it is called a discriminative stimulus (Sd)
• When the associations is classical SS the controlling stimulus is often called an occasion setter
• The effects of these controlling stimuli do not depend on Pavlovian conditioning – they are something else

Question 16

Discriminative control of classical associations

Answer 16

• Two groups of animals:
• Group FP: Light  tone  food, tone  no food
• The light is followed by the tone, which is then followed by food.
• Animals respond more to the tone when the light has just been on
• Is the light a special controller stimulus for the toneàfood association?or is it just classical conditioning to the light?
• Group PP: Lighttonefood, tone no food, light no food
• Group PP gets the same, but ALSO the light ispresented alone so its associative strength extinguishes
• See classical conditioning to the light is absent in Group PP
• Performance is the same
• Light is controlling conditioned responding to the tone even when it doesn’t predict food itself
• Is the light a special controller stimulus for the toneàfood association?

Question 17

Discriminative control of instrumental associations

Answer 17

• Animals learn to only press the lever for food when the Sd is present – it controls access to Rfood
• Question: in Group Sd is the Clicker a classically conditioned CS or something else?
• If it is just a CS, it should block acquisition of associative strength by the light, just as in Group CS
• If it doesn’t have associative strength it should not.
• Learn about light in the control group where clicker has no strength  no blocking
• Learn less about the light in the CS group where clicker does have strength  blocking
• In Sd group clicker behaves like the light in the control group – as though it has no strength
• Clicker is controlling responding despite not being a signal for food
• Is the clicker a special controller stimulus for the responseàfood association?

Question 18

Latent inhibition: hard to learn about familiar things

Answer 18

• Subjects: four sheep and four goats
• CSs: rotor & flashing light
• US: shock
• Took more trials to learn with preexposed stimulus