Learning

Question 1

critical periods

Answer 1

periods in our lifespan where we can learn specific skills because of varying plasticity

Question 2

example of missing the critical periods (bad plasticity)

Answer 2

• visual deprivation
• two eyes misaligned (e.g. cataract)
• stronger eye will take over much of brain region shared by both eyes
• to correct must force use of weaker eye

Question 3

how visual deprivation effects neural responses? evidence for critical period

Answer 3

• Hubel and Wiesel on cats with ocular dominance
• closed contralateral eye from 1 week to 2.5 months (critical period) = no cells activated anymore in this eye
• closed contralateral eye for longer amount of time from 12 to 38 months (but not critical period) = a much smaller effect

Question 4

• retinotopic cortical map
• somatotopic map

feature of both

Answer 4

-systematically representing features at different visual locations
-representing different parts
of our body

BOTH show plasticity when input changes

Question 5

evidence for plasticity of somatosensation

Answer 5

(Mogilner et al., 1993).

• looked at somatosensory cortex in the brain.
• 26 days after syndactyly surgery, drastic changes in the brain mapping.

(Rutkowski and
Weinberger, 2005)
-cortical area devoted to sounds of a specific freq increases when those sounds are assosciated with a reward

Question 6

Perceptual learning can be used as a tool to understand brain representations (and
understand how we become experts)

Answer 6

(Karni & Sagi, 1991)

-trained subjects to detect pop-out stimuli in briefly
presented stimulus arrays
-subjects showed large improvement

(1) specific to the spatial location that was trained
(2) sensitive to slight changes in stimulus properties (e.g. change in orientation features)
(3) monocular training doesn’t transfer to other eye

Question 7

karni & sagi link to V1

Answer 7

Might occur in V1 because

(1) small receptive fields
(2) high feature selectivity
(3) it is the first stage where info from both eyes combined

(Li et al., 2008) confirmed significant V1 responses with similar pop-out tasks

Question 8

theory that links learning affects to brain areas

Answer 8

(Ahissar and Hochstein 2004)

If in low level (V1), learning affects are tuned to specific features

If in higher cortical areas, learning affects can generalise across specific features

Question 9

why is there less strong feature selectivity in highercortical areas?

Answer 9

-there are larger receptive fields

Question 10

when do we use V1 learning and when do we use higher

Answer 10

depends on task difficulty

to become an expert -> need low level

Question 11

what is the reverse hierarchy

Answer 11

(Ahissar and Hochstein 2004)

learning effects first occur in high-level areas, and
progressively transfer to early visual cortex when we become an expert

Question 12

Evidence for reverse hierarchy

Answer 12

(Sigman et al., 2005)

• study on shape discrimination
• during TRAINING, activity in high level areas involved in shape representation decreases while activity in early visual cortex increases

Question 13

Pavlovian conditioning

Answer 13

• acquiring a new response to a previously neutral stimulus

- result of predictive relationship between stimulus and response-evoking stimulus

Question 14

Unconditioned Stimulus (US)

Answer 14

A stimulus that has natural relevance

Question 15

Conditioned Stimulus (CS)

Answer 15

A stimulus that gains its relevance through learning

Question 16

Pavlov example

Answer 16

CS = bell
US = food (which evokes salivation naturally)
CR = salivation to bell

Question 17

two types of pavlovian conditioning

Answer 17

appetitive - pleasant US (e.g. food)

aversive - unpleasant US (e.g. electric shock/eye puff)

Question 18

neural basis of pavlovian conditioning?

Answer 18

cells in the brain stem (SUBSTANTIA NIGRA and VENTRAL TEGMENTAL AREA (VTA) project to forebrain structures (STRIATUM), where they facilitate dopamine release

Question 19

proof for dopamine involvement in appetetive

Answer 19

O’Doherty et al (2002)

• appetitive response with glucose solution, and a separate aversive response to salt solution
• fMRI showed that the substantia nigra and VTA were more active during the appetitive CS that aversive CS

Question 20

being aware CS - US relationship study

i.e. cognitive expectation causes the response

Answer 20

Lovibond (1992)

• paired one pic of plants with shock (US-) and one without (US+)
• showed that people aware of the link showed SCR response, so showed more anxiety

Question 21

cognitive expectation study with extinction phase

Answer 21

Hugdahl and Ohman (1977)

-flower (CS-) when people were told there would be no more shocks, SCR decreased immediately (took much longer when not told)

Question 22

evidence that pavlovian conditioning results from implicit learning

Answer 22

-spider (CS-) when people were told there would be no more shocks, SCR decreased much more slowly (fear relevant stimuli)

Implicit learning appears independent of cognitive expectation

Question 23

Bechara et al (1995)

Answer 23

• AMG and AMG/HC did not have SCR response to aversive conditioning (implicit learning)
• HC patient showed normal implicit conditioning
• HC and AMG/HC could not report which CSs were followed by US (explicit learning measure)
• AMG patient could report explicit learning

Implicit learning mediated by AMG
Explicit learning mediated by HC

Question 24

how does temporal contiguity affect the response?

Answer 24

• the longer the interval between the events, the less is learned about their relationship
• but depends on type of learning

Question 25

example of where temporal contiguity isn't necessary

Answer 25

- food aversion | - andrykowski and otis (1990) studied patients doing chemotherapy, found no relation to time and food aversion

Question 26

blocking?

Answer 26

when there is already a CS-UR relationship, the addition of a second CS will not be associated with the UR, this relationship is "blocked" by the previous link

Question 27

blocking experiment

Answer 27

Tobler et al. (2006) measured activation changes in ventral striatum Stronger activation when reward was surprising than when reward was result of a pre-learned CS-US relationship.

Question 28

how is blocking explained in terms of surprise

Answer 28

Learning proceeds in a negatively accelerated curve: with each CS-US pairing, expectation of the US increases and surprise decreases. I.e. the rescorla-wagner rule: ∆V = αβ(λ– ∑ V) where V = associative strength of CS λ= perfect prediction of the US (λ – ∑V) = prediction error αβ = learning rate parameters

Question 29

rescorla-wagner rule explaining dif types of responses?

Answer 29

explains the CS of both excitatory (presence of response, e.g. allergic reaction) and inhibitory associative strength (absence of response, e.g. allergic reaction) when λ = 1, it is excitatory when λ = 0, it is inhibitory

Question 30

example of inhibitory association using rescorla-wagner

Answer 30

According to Rescorla-Wagner, this happens when a CS is presented with an excitatory CS+ that has been fully learnt about previously. Under these conditions, λ is 0 as there is an absence of a US, whereas ∑VA is positive. Therefore CSB will acquire inhibitory associative strength

Question 31

On conditioned inhibitors

Answer 31

(Lovibond et al., 2000) - two phases: acquisition and extinction - first make E- a learned inhibitor - CS(d) paired with shock, but never with E-, in extinction phase the association weakened - CS(c) paired with shock, then later shown with E-, in extinction phase C did not lose its excitatory strength (BOTH in SCR experiment and self report expectation experiment)

Question 32

what is superlearning?

Answer 32

Turner et al (2004) - trained a banana with an allergic response - paired with mushroom -> omitted allergic outcome - then mushroom + pear -> allergic response - pears as seen as extremely allergic (SUPERLEARNING) - Showed activity in the right PFC, which is sensitive in particular not to what is being learnt, but to the degree of prediction error (very surprised)

Question 33

latent inhibition

Answer 33

Nelson and Sanjuan (2005) Preexposing stimulus prior to pairing it with the US retarded learning the association

Question 34

how can we describe LI?

Answer 34

- can't be explained by Rescorla-Wagner rule, because there is no prediction error during pre-exposure (an outcome is neither predicted not occurs) - alternative theories: (1) Mackintosh (1975) (2) Pearce & Hall (1980)

Question 35

Mackintosh (1975) theory?

Answer 35

The more reliable a CS is for predicting reinforcement, the more attention is paid to it

Question 36

Pearce and Hall (1980) theory?

Answer 36

The less reliable a CS is for predicting reinforcement, the more attention is paid to it -argue that when the outcome is perfectly predicted, we stop paying attention because there is nothing more to learn -so that's why it's harder to learn when we've been pre-exposed

Question 37

Experiment that showed one LI theory is better?

Answer 37

Hogarth et al. (2008) - presented (AX+) = loud white noise - presented (CX-) = no noise - presented (BX+/-) = half noise half no noise (i.e. uncertain) Participants paid attention to stimulus B, implying that Pearce-Hall (1980) was correct

Question 38

what is instrumental conditioning/who showed it first

Answer 38

IC is when there is a causal relationship between a behaviour and an outcome Thorndike (1911) demonstrated that cats can learn to press a lever to escape a box (known as trial and error learning)

Question 39

what are the principles of reinforcement?

Answer 39

positive reinforcer / negative reinforcer = probability of response increases punisher/pleasant outcome emitted = probability of response decreases

Question 40

schedules of reinforcement?

Answer 40

fixed ratio - e.g. getting reward every 6th time variable ratio - e.g. getting reward randomly fixed interval - e.g. getting paid every month variable interval - e.g. getting paid randomly throughout month

Question 41

comparing variable ratio and variable interval schedules?

Answer 41

Matthew et al. (1977) yoking experiment - one participant rewarded randomly as he was pressing lever (VR) - other participant rewarded every time part1 was (VI) VR schedule maintained a much higher rate of responding than VI

Question 42

Thorndike (1911) law

Answer 42

"law of effect" - positive and negative reinforcers strengthen link between stimulus and response (S-R)

Question 43

evidence for brain activity of instrumental conditioning

Answer 43

O'Doherty et al. (2004) -participant 1 chose between a stimuli where they have a high prob of receiving fruit juice outcome, and a stimuli with a low prob " " -participant 2 was yoked, but still saw the stimuli (essentially just pavlovian) -both saw ventral striatum BOLD activity -more BOLD activity in the dorsal striatum during the instrumental task.

Question 44

is IC due to S-R habits vs goal directed?

Answer 44

Klossek et al. (2008) - children trained to touch one icon for one cartoon, another for a dif cartoon - one cartoon was devalued by exposure to induce boredom - when given the choice, older children (>27months) used goal directed behaviour - younger children (<27 months) showed an S-R habit as they chose both options equally

Question 45

first fMRI study of S-R habit vs goal directed behabiour

Answer 45

De Wit et al. (2009) PART 1 - control: stimulus and outcome are different (S-R/S-O-R) - congruent: stimulus and outcome are the same (S-R/S-O-R) - incongruent: stimulus of one is same as outcome of the other (only S-R is accurate here to begin with) - compare control/incongruent and congruent/incongruent - vmPFC = goal directed - dmPFC = S-R habit PART 2 - instructed devaluation phase, where one outcome no longer produces points (affects S-R) - found that the degree to which the vmPFC was active correlated to the degree to which they were using goal directed learning!

Question 46

fMRI study of goal directed that shows the affect of the VALUE of reinforcers?

Answer 46

Valentin et al. (2007) PART 1 - control: tasteless liquid - two stimuli with common outcome (orange juice) - two other stimuli both with orange juice outcome, but also had a second outcome (chocolate or tomato) - compare neutral vs rewarding outcomes - medial orbitofrontal vmPFC active in rewarding trials PART 2 - subjects satiated by one outcome - greater medial and central PFC activity in the valued condition, suggesting this region is sensitive to the value of the outcome (how . much we want it)

Question 47

what is the generalising rule

Answer 47

learning generalises to a degree proportional to the similarity with the trained stimulus

Question 48

what if S+ and S- are similar (humans)

Answer 48

- if S+ is orange, and S- is blue, the excitatory and inhibitory gradients will overlap - net excitatory associative strength between them - in humans, super orange will be more associated with rule, as will super blue - RELATIONAL RULE (linear)(cognitive solution)

Question 49

what if S+ and S- are similar (pigeons)

Answer 49

slightly to the left of learnt colour of orange = the peak of responding -ASSOCIATIVE RULE (peak shift effect)

Question 50

Is it the case that humans use a cognitive system for a solution? Or is it that the cognitive system will search for a solution and dominate over an associative system?

Answer 50

Wills & Mackintosh (1998) used artificial dimension to make cognitive system impossible (set of icons labelled B to I) Had an average number of instances of each icon, allowed excitatory and inhibitory strength to develop through training When tested on new stimuli, people demonstrate the peak shift effect, so we can use it when stimuli don't make sense to us.

Question 51

two types of categories

Answer 51

highly typical/ close to prototype (e.g. cats and guinea pigs) -> "LOW DISTORTION" not very typical/ range from prototype (e.g. bat and platypus) -> "HIGH DISTORTION"

Question 52

prototype

Answer 52

example of most diagnostic of a category

Question 53

prototype effect

Answer 53

Even if the participant has never seen the prototype before, it will be more accurately categorized than the other novel exemplars

Question 54

typicality effect

Answer 54

Low distortion exemplars will be more accurately classified than high distortion exemplars

Question 55

exemplar theory

Answer 55

suggests that these two effects result from an explicit comparison of a new exemplar to stored memories of trained exemplars

Question 56

contrast to exemplar theory

Answer 56

the comparison might be IMPLICIT Squire & Knowlton (1995) observed that an amnesic patient with bilateral hippocampal damage demonstrated the prototype effect without explicitly recognising stimuli

Question 57

what explains squire & knowlton (1995)

Answer 57

prototype theory states that during training on a set of exemplars, the prototype is ABSTRACTED and generalization to new exemplars will be based on similarity to the prototype

Question 58

what about forms of learning that can't be solved associatively? can we solve them with cognitive rules?

Answer 58

Shanks & Darby (1998) - positive patterning: two foods did not cause allergy by themselves, but they did together - negative patterning: foods caused an allergy by themselves, but not together - Showed that people have different abilities at associative learning. - People that use associative learning a lot are poor at learning this. People that don't use it, use a generalised rule rule, and in this example this is more accurate

Question 59

is it better to use cognitive rules or associative rules?

Answer 59

Perruchet (1985) used gambling fallacy (after a run of bad luck, you are 'owed' a good one, or after a winning streak, you are 'owed' a bad turn) -gave people a puff of air to the face on 50% of trials -4321 - No eye puffs occurred -1234 eye puffs occur - The eye-blink CR behaved associatively, and the expectancy ratings obeyed the gambler’s fallacy. Therefore, here, we are using cognitive rules and falling into the fallacy. If we used association we'd know its 50/50.