Cognitive Neuroscience 2 Flashcards

You may prefer our related Brainscape-certified flashcards:
1
Q

What evidence do we have that grammar (Syntax) is important and unique to language?

A

We can assess grammar even when meaning is lost (Jabberwocky)
Animals can’t learn it

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
2
Q

What is recursion?

A

Embedding linguistics units within each other to make a sentence

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
3
Q

What is a constituent?

A

Words/phrases that can operate as a single unit in a hierarchical structure

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
4
Q

How do we parse sentences?

A

Identify local phrase structure through word category information
Compute syntactic and semantic relations by assigning thematic roles

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
5
Q

What is the extra step taken in parsing ambiguous sentences?

A

The integration of context

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
6
Q

What is a garden path sentence?

A

A grammatically correct sentence that starts in a way that means a reader’s first interpretation will probably be wrong

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
7
Q

What is a serial parsing model?

A

A model that suggests that the parser maintains one structure and then, if it is wrong, starts parsing the sentence again, accommodating the new information

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
8
Q

What is a parallel parsing model?

A

Parser maintains all possible structures and discards as new information becomes available

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
9
Q

What is an encapsulated parsing model?

A

The parser only uses syntactic information (bottom-up), frequency and prosody

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
10
Q

What is an interactive parsing model?

A

Parser uses lots of information including syntax, context, discourse, and visual information (bottom up and top down)

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
11
Q

What can ambiguity tell us about how we parse sentences?

A

Tells us our preferences in how we interpret sentences – suggests not all sentences are considered in parallel
May tell us how we handle unambiguous input – reaction times and eye movement tech as well as EEGs for auditory parsing

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
12
Q

What is the garden-path model?

A

Immediate, serial, encapsulated
Each word added to current syntactic structure, only one structure maintained at a time
Processing difficulties occur at points of ambiguity when preferred analysis disproved
Initial analysis only guided by syntactic info

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
13
Q

What is the two-stage model?

A

input > lexical processor > syntactic parser > syntactic structure > thematic interpreter > sentence meaning
context info only influences at a later stage

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
14
Q

What are some of the heuristics that cause garden path sentences?

A

Late closure – attach new parts of a sentence to a phrase/clause that is currently being processed rather than starting a new one
Minimal attachment – attach info to the phrase structure with the fewest nodes

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
15
Q

What are the characteristics of constraint-based models?

A

Immediate, parallel, interactive
Each word processed immediately
Multiple structures are maintained and ranked according to how well they fit constraints
All information types used to aid analysis – context, visual context, frequency of syntactic structure, syntax, prosody, word meanings

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
16
Q

What have eye movement studies shown us?

A

How garden path sentences affect parsing
When you reach the ambiguous section of the sentence, you have to go back and start it again to reinterpret what you read before you can continue reading

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
17
Q

How does the frequency of syntactic forms affect parsing?

A

If a word is most commonly used in one syntactic form, you are most likely to parse it that way which can lead to garden path sentences in the long run

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
18
Q

What is a P600 and when does it appear?

A

positive spike in ERP 600s after a syntactic anomaly e.g. a garden path sentence
also by morphosyntactic errors e.g, wrong ending/SVO disagreement

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
19
Q

What is an ELAN?

A

Early Left Anterior Negativity
100-200ms after syntactic anomly occurs
specifically when it’s impossible to build a phrase structure/word-category errors
more temporal

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
20
Q

Where does a P600 occur?

A

not lateralised

more parietal/frontal

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
21
Q

What do brain scans tell us about ELANs?

A

They correspond to the point in parsing when a phrase structure is built

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
22
Q

How has sign language been useful for brain scans?

A

confirmed findings about N400, P600 and ELANs

also right anterior negativity though which is unusual

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
23
Q

Can we use ERPs to understand whether sentence parsing is serial or encapsulated?

A

The fact that the ELAN comes before the N400 could suggest that syntax comes first

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
24
Q

What happens if a sentence contains both a word-category violation and a semantic one?

A

Only an ELAN is observed, no N400
a word that doesn’t doesn’t make syntactic sense doesn’t quality to be evaluated on semantic grounds
isn’t lexically integrated

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
25
Q

What happens if a sentence contains both a word-category violation and a semantic one and ps are told to focus on semantics?

A

N400 is restored and P600 is much bigger

suggests that P600 can be an interaction between syntactic and semantic information

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
26
Q

What areas have neural correlates of sentence processing implicared?

A
posterior MTG (medial temporal gyrus)
anterior STG (superior temporal gyrus)
posterior STS (superior temporal sulcus)
middle frontal gyrus
pars orbitalis IFG (inferior frontal gyrus)
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
27
Q

What is the relationship between Broca’s area and agrammatism?

A

Patients with Broca’s aphasia have impaired syntactic processing aka receptive agrammatism
rarely use function words
impaired processing of agents in complex sentences
struggle with pronouns in different word orders

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
28
Q

Which type of sentence did Broca’s and conduction aphasics struggle most with?

A

Object cleft

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
29
Q

When college students were tested on sentence comphrehension with compressed, low pass sentences, what happened? What did this indicate?

A

They showed results similar to those of Broca’s/conduction aphasics (very low performance on object cleft sentences)
suggests that Broca’s/conduction aphasics may struggle with those types of sentences because of demands on cognitive resources

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
30
Q

What caused increased activation in the Left Anterior Temporal Lobe?

A

more specific concepts when syntax was the same

also at the onset of adjectives

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
31
Q

How did the vmPFC respond to adjectives?

A

activation 200ms after onset

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
32
Q

What are the symptoms of amnesia due to hippocampus damage? What is a case study of this?

A

profound memory loss
intact faculties and intelligence
Clive Wearing

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
33
Q

What areas did Clive Wearing suffer damage?

A

left and right temporal lobes
hippocampus - both sides
left frontal - causes excessive repetition, highly emotional behaviour

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
34
Q

What types of learning are spared in amnesia?

A

perceptual-motor skills (mirror reading)
classical conditioning (learn to avoid aversive stimuli)
priming (prior exposure affects performance subconsciously)

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
35
Q

What is aterograde nmnesia?

A

amnesia where you have lost your long term memories but still have short term memory up until a certain point

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
36
Q

What is retrograde amnesia?

A

Where you’ve lost your short term memory but can still remember things from the past (can’t remember what happened just before you got amnesia but can remember things from your past)

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
37
Q

What are the subsystems of long-term storage? What can these be broken down into?

A

explicit (declaratice) memory - episodic and semantic

implicit (non-declarative) memory - procedural, priming, conditioning

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
38
Q

What are the dissociations between episodic and semantic memory?

A

anterograde amnesia involves a loss of both episodic and semantic memories from the onset of the amnesia
retrograde amnesia invoves the loss of episodic memory but the retention of semantic memory (general knowledge)

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
39
Q

What are the effects of early hippocampal damage?

A

striking episodic memory damage - can’t remember big things that happened in their lives
semantic memory is spared - reading, vocab
STM intact for recent info - digit recall, pics, stories

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
40
Q

Explain the theory about semantic memories being retained better rather than there being a double dissociation

A

semantic memories are retained better because they are stronger - semantic info is reinforced a lot more than episodic (which by nature only happens once) so maybe that’s why they’re retained better and not because of separate encoding methods

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
41
Q

What was Tulving’s rebuttal to the theory that semantic memories are retained better because they are stronger?

A

He used the case study of KC who could remember the difference between stalagtites and stalagmites but couldn’t remember an accident at a coal manufacturing plant that had caused his whole town to evacuate
surely second memory should be stronger because of emotional salience?

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
42
Q

What are the symptoms of semantic dementia?

A

loss of semantic memory - struggle with naming and using objects, picture matching
episodic memory is intact and so is working memory and problem solving
autobiographical info is spared but non-specific but doesn’t remember them as objective historical events, only in the way they were personally affected

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
43
Q

What cortical areas show damage in semantic dementia?

A

atrophy to anterior temporal lobe

bilateral but worse in left hemisphere

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
44
Q

Where does episodic memory seem to be localised?

A

the medial temporal lobe

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
45
Q

Where is content localised in episodic memories?

A

verbal - left hippocampus

visual - right hippocampus

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
46
Q

Where is content localised in semantic memories?

A

living/non-living - temporal lobe
fruit veg vs anmals - occipitotemporal
tools/actions - premotor/parietal

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
47
Q

What evidence is there that the MTL is important for encoding episodic memories?

A

items that ps remember correctly are associated with greater activity in the MTL during encoding

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
48
Q

What evidence is there that the MTL is important for retrieving episodic meories?

A

researchers tried to use hippocampal activity to figure out waht stimulus the participants were thinking about
could decode it about chance (~45% to 33%)

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
49
Q

What evidence is there that anterior temporal lobe is important for semantic memories?

A

when ps had to access info about tools, anterior TL activated

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
50
Q

What do we know about the storage of semantic memories?

A

Semantic memories are clustered all over the cortex in different categories although anterior TL may be important for accessing (the hub)
Huth et al 2012

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
51
Q

Explain the theory that suggests that rather than an episodic/semantic dissociation, there is a distinction between specifics and generalities

A

Hippocampus separates info, neocortex integrates it
want memory system to remember specifics and generalities about the world - stable and plastic
hippocampus remembers specifics, neo puts those specifics together to extrapolate general rules about the world

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
52
Q

What is the contrasting pattern of memory loss in people with semantic dementia and people with amnesia and what does it suggest?

A

amnesia - not complete retrograde loss of EMs, just most recent ones
case study PZ
reverse in patients with semantic dementia - increased EMs for closer things and loss of memories from further back
newer episodic memories reliant on MTL, esp newer ones
suggests that info gradually transferred from hippo to neo
older memories ‘semanticised’ - almost like a 3rd person memory

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
53
Q

What types of information (apart from memory) is the hippocampus associated with?

A

spatial info - Maguire taxi drivers, place cells
amnesic patients have poor spatial orientation usually
novel info in WM - Ranganath and Blumenfield, novel associations/objects
hippo damage leads to problems learning new associations

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
54
Q

What is the Complementary Learning Systems theory (McClelland et al)?

A

Theory that suggests that hippo is in charge of fast learning
creates arbitrary links and remembes specific details
neocortex is in charge of slow learning
integrates memories and creates consistent links
extracts stable generalities
slightly different different to specifics generalities - rather than different types of content, different types of learning

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
55
Q

How do the hippocampus and neocortex work together?

A

new information is rapidly stored in hippocampus and then gradually moved to the neocortex for consolidation
allows you to learn info rapidly and select what is worth keeping

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
56
Q

What is hippocampal replay?

A

hippocampus replays recent experiences during offline periods e.g. rest and during these periods the hippo and neo interact
predicted by CLS

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
57
Q

What did Takashima et al find and what is its significance for the CLS?

A

When you ask people to recall information that they have recently remembered, the hippocampus is activated, but when you ask them for info encoded 24h before, different areas of the cortex are activated
suggests that CLS is right and info is transferred from hippo to neo

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
58
Q

Why does learning exceptions provide an issue for the CLS?

A

need to be able to update new info that doesn’t fit with pre-existing rules without getting rid of those rules
hippocampus allows you to quickly learn an exception but gradually integrate it

59
Q

How do we know that memory is a reconstruction?

A

Not remembered exactly as it was encoded (War of the Ghosts)
memories biased to be consistent with prior knowledge
remember things better when we understand them - interaction with semantic memory seems important for understanding

60
Q

What evidence is there that the neocortex interacts with the hippocampus during encoding?

A

Grady et al found that when there were higher levels of interaction between the hippocampus and relevant cortical areas, memory was better

61
Q

What is the benefit of semantic memory supporting episodic memory in encoding?

A

means we have less to remember
can use semantic memory to give us a basis and then use episodic to flesh it out more
this is why there is interaction between neo and hippo in encoding

62
Q

What evidence is there that the hippocampus and neocortex interact during retrieval?

A

tested which regions are functionally correlated in terms of activity when people remember life events (Maguire et al 2001)
for life events - interactions between hippo and neo, PHG and temporal lobe

63
Q

What evidence is there that recall is really reactivation?

A

recorded brain activity with fMRI at encoding and then in free recall
attempted to decode cortical responses and found that they could decode to a high level of specificty (not just a face, but could differentiate betweeen faces) and could use this to reliably test computers

64
Q

What is the theory of associative learning/pattern completion?

A

A stimulus activates a cortical representation of a network of information
Hippo forms an association between knowledge representations to learn quickly
Persisting memory of that association
When you see his face again, it acts as a retrieval cue which activates the stored memory in the hippo and it activates the network and allows you to retrieve the rest of the information (pattern completion)
repeated exposure makes a shorter link which allows this process to happen quicker

65
Q

What does the dentate gyrus do?

A

keeps different memories separate

66
Q

What does the CA3 region do?

A

associated different elements of the same memory with each other

67
Q

Why do we categorise objects and events?

A

allows us to communicate
gives us economy of storage/efficience of processing
allows us to generalise
otherwise we would have to store all new information we learnt separately and basically all over again

68
Q

What is economy of storage?

A

extracting out common properties of objects and store them based on these properties
means you can store common information once and for all and you only need to remember identifying unique features

69
Q

What is a taxonomic hierarchy?

A

A hierarchy that gets more specific as it goes down

70
Q

What is a taxonomic hierarchy?

A

A hierarchy that gets more specific as it goes down and therefore gives you more relevant detail

71
Q

What do sentence verification tasks tell us about hierarchichal storage?

A

Collins and Quillian
reaction time increased for questions as they went from the bottom to the top of the heirarchy i.e. can a canary sing, does it have feathers, does it have skin
singing is a first order property of a canary while having skin is a property associated not just with canaries but with all animals therefore it takes longer to verify than something lower down

72
Q

What do habituation studies tell us about hierarchical storage?

A

habituation tasks
infants first learn broad categories then differences between them
speed of habituation to new toy shows how different they perceive it to be to the old toy
9 months old seem unable to differentiate between animals

73
Q

what do predicability trees tell us about hierarchical storage?

A

looking at what properties children feel comfortable assigning to objects
as children get older, they get more specific about what predicates can be applied to what subjects
suggests that broad concepts are gradually differentiated and become more accurate and specific

74
Q

What do patterns of loss in semantic dementia tell us about hierarchical structure?

A

loss of concepts in the opposite order than infants pick them up
differentiation lost at the bottom of the taxonomic hierarchy first
Patterson et al 2007 - go from swan - animal or miscategorisation
supports the idea of a hierarchy

75
Q

What about multiple taxonomic hierarchies?

A

dogs are animals and pets
concepts are context-dependent and so people can be flexible in how they consider them
dog is similar to wolf in animal terms
dog is similar to cat in pet terms
children show sensitivity to this as well

76
Q

What is the prototypicality effect?

A

the idea that some objects are better representations of their category than others
e.g. an apple is an OG fruit
other objects cluster around these prototypes at varying degrees of representability
people have quite high consensus when rating objects in terms of how much they fit the category except for on margins

77
Q

What are prototype effects?n

A

faster categorisation/verification
more frequently brought to mind
learnt first, spared in semantic dementia
recognised well even if never seen
less likely to make mistakes in identification
preference for typical things

78
Q

Do prototype effects only work for fuzzy concepts?

A

no - also work for very well defined things e.g. triangles

also grandmothers, odd and even numbers

79
Q

What is the basic-level advantage?

A

the fact that there is a level of description that we are most likely to use when referring to an object
could be more vague (fruit) or more specific (pink lady) but generally we use apple
basic level also learnt first in children
also more likely to be represented cross-linguistically

80
Q

Why are basic level concepts psychologically privileged?

A

because they hit the right balance of being inclusive and informative
can convey information at the most useful level
superordinate level often doesn’t add much more information that would be useful and subordinate level is too vague

81
Q

What are the limitations of the hierarchical structure?

A

it doesn’t capture the idea of prototypicality
it doesn;t capture differences in importance of different levels (basic level concept)
doesn’t capture how structure is learnt (seems intuitive)

82
Q

What is the general structure of a connectionist model?

A

input > input units > hidden units > output units > output
activity of output layers dictates behaviour of network
nature of the connections determines how info flows and is outputted
units are neurons/populations of neurons
connections can change based on learning

83
Q

What do simple processing units do?

A

they integrate inputs and send proportional utputs

neurons in the brain do this a lot

84
Q

Why do connectionist models favour a layered organisation?

A

captures organisation in the brain
similar to perception - v1 to v4
at each layer information is transformed and this depends on the nature of the connectons
learning happens by changing them

85
Q

What are the principles of neural processing?

A
neurons integrate information
neural activity reflects level of input 
layered brain structure
influence via connections 
learning alterns connection strengths
86
Q

What does Parallel Distributed Processing mean?

A

Information is passing through and represented across multiple units and through different connections at the same time

87
Q

If activity is a linear function of net input, then why is it best represented by a sigmoid/logistic function?

A

activity reaches a plateau because at some point the increase in acitivity of the input cant increase the net input anymore
asymptote also at the lower end
captures property of real neurones - they have a minimum and a maximum

88
Q

What is the equation of activity?

A

ai = 1/1+e^-netinputi

89
Q

What is the equation for inputi?

A

aj x wij

activity of j (preceding unit) x weight of connection between i and j

90
Q

What is the error-driven learning model?

A

Delta rule (Backpropagation)
network learns to reduce error between the correct response and its behaviour
animals learn to the extent that they are surprised by outcomes and outcomes are surprising to the extent that they are unpredictable

91
Q

What is the delta rule?

A

changes in the weight/strength in each learning episode = [difference in the level of activity that a unit should have - the observed level of activity] aka the error term x activity of the unit
equation produces a neg number when actual is greate than desired and vv

92
Q

How does backpropagation work?

A

If at the ouput layer the unit is too active, you have to reduce its activity at the layer that inputs to the ouput and then scale it through the learning rate (which is a small number as we learn slowly)

93
Q

What did Rumelhart’s model show us?

A

model of simplified form of semantic learning
has to learn which properties belong to which objects starting with no info - all connections are equally weighted
increase connection strengths when it receives positive error terms and weakens when it receives negative ones
this creates differentiated predictions
continues until there is no error term and it reaches a level of certainty

94
Q

When given unstructured input, what did Rumelhart’s model ‘learn’?

A

It learned a taxonomic structure after many learning episodes and started producing the right relationships/associations

95
Q

What is the name of the first hidden layer in the PDP model?

A

the representation layer

96
Q

What was the result of looking at hidden unit activity in Rumelhart’s model?

A

degree of difference was represented by slight changes across multiple units - not one unit or another showing differentiatin
things that are more similar have a greater degree of similarity across units than things that are more different
have learnt structure through a very simple learning rule

97
Q

What did Rumelhart’s model learn?

A

learnt to predict properties of objects and link them to output features
learnt to produce the right set of connections from activation patterns from hidden layer to output layer

98
Q

How does RUmelhart’s model develop knowledge?

A

It first learns broad properties e.g. animals v plants and then learns specifics
learning shared properties tend to compound but idiosyncratic ones cancel each other out until you have broad distinctions
at the beginning, bits of learning cancel each other out and the items wiggle around and little improvement is made and then at some point the learning about each individual item and the learning about the categories compound on each other and they split apart abruptly

99
Q

In Rumelhart’s model, what happens after each new distinction has been learnt?

A

Error-driven learning stops for the previous distinction and errors for finer distinctions become more salient, allowing differences here to be learnt

100
Q

How was semantic dementia modelled in Rumelhart’s model?

A

created ‘noise’ in hidden layer
SD caused by atrophy in ATL hidden layer can replicate this
if we add small random values (noise) to activity patterns it disrupts spatial representation and creates slightly different patterns

101
Q

What were the results of stimulating semantic dementia in Rumelhart’s model?

A

noise let to greater damage to properties specific to a canary than those that were specific to all animals (just like in SD)
might be because small disruptions in networ push you away from normal activity pattern of that specific item, but not the general activity that the item fits into

102
Q

How are prototypicality effects explained in Rumelhart’s model?

A

model learns categories in terms of shared and distinct categories
also see that items have shared features of the group and distinct ones that differentiate them
other items fit into the group but have properties that contradict the norm of the group
these are less prototypical than the ones who fit the average characteristics of the group

103
Q

How is the basic level explained in Rumelhart’s model?

A

basic level used more frequently so learnt quicker and more robust, more resistant to noice/damage
network experiences basic level more so given a large space in network
more robust

104
Q

What are the strengths of Rumelhart’s model?

A

similar concept to multi-dimensional space models
structures knowledge through learning - gives us flexibility to represent differences along different dimensions
based on the effects of coherent covatiation
basic building blocks are simple but allows for a complex model
fault tolerance - no single unit is crucial
complex representation - noise isnt catastrophic
access info by content
rich, graded representations which allow for exceptions

105
Q

What is coherent covariation?

A

noticing and learnign regular co-occurence of sets of proerties of objects

106
Q

What are the criticisms of Rumelhart’s model?

A

the model is oversimplified - model is told the right answer and has built in items and features, there’s more to concepts than similarity
how is knowledge used - how do we draw inferences, analogies, combine concepts
doesn’t account for causal information
learning methods aren’t plausible - backpropagation is too slow and too fragile
not a cognitive model - seems statistical and not constrained neough

107
Q

What are the two types of theories of rationality?

A

Normative theory - what we should do

descriptive theory - what we actually do

108
Q

What is the difference between formal logic and human reasoning?

A

Formal logic deals with certainty whereas human reasoning deals with uncertainty

109
Q

What is bounded rationality?

A

people think and behave rationally within the limits imposed by their cognitive capacities, available time, and access to information

110
Q

How can rationality be defined?

A

How well normative theories of how people should reason match up with how people actually reason

111
Q

What do normative theories of uncertainity use?

A

Probability - judges the likelihood of outcomes and events

112
Q

How effective are humans at reasoning probabilistically? What evidence do we have of this?

A

Lmao kinda bad

Monty Hall problem - 96% of people get it wrong

113
Q

What is base rate neglect?

A

ignoring the general prevalence (base rate) of an event and analysing its probability based on information explicitly provided
insensitivity to prior probability

114
Q

How do we successfully get around bounded rationality?

A

make them more simple and solvable by relying on heuristics

115
Q

What is the problem with using Bayesian probability theory in everyday life?

A

The computations can get very complicated very quickly, especially when you have to take lots of things into account

116
Q

What is the downfall of using heuristics to do probabilistic calculations?

A

Lead to serious and systematic errors when heuristics are misapplied
‘cognitive illusions’

117
Q

What is another way to conceptualise the frequency of an event? How has this been shown in experiments?

A

It’s availability, how easy it is to bring examples of it to mind
Tversky and Kahneman - how many words are there ending -n vs -ing
-ing voted 3x more even though -ing is a subset of -n so the latter has to have more

118
Q

What is another way of conceptualising the likelihood that something belongs to a particular category? How has this been demonstrated in experiments?

A

How representative is it of that category
Tversky and Kahnemann - Linda study, give demographic info about Linda and then ask which is more likely, either she’s a bank teller or a bank teller and a feminist
Statistically, the former has to be more likely because 1 possbility is more likely than 2 together
but 85% of ps picked option 2 because the information about her made her seem representative of feminism

119
Q

What is another way of estimating values (that you are unsure of)? How has this been demonstrated in studies?

A

Taking an initial value and adjusting it
called anchoring
Tversky and Kahneman
What % of African countries are in the UN? <10>? 25 avg
What % of African countries are in the UN? <65>? 45 avg

120
Q

What is the issue with the anchoring paradigm in Tversky and Kahnemann’s study and how was it resolved?

A

The presence of those values in the question kind of indicates that they should be used as an estimate otherwise why are they there
kinda gricey
this was resolved by the experimenters picking the relevant percentage by using a spinning wheel (Random selection) and anchoring effects were still present
BUT ALSO PPL LEGIT DON’T KNOW THE ANSWER SO THEY MIGHT AS WELL JUST USE WHATEVER INFO THEY’VE BEEN GIVEN

121
Q

What is Kahenman’s theory of fast and slow thinking?

A
system 1 (fast thinking, intuition) - parallel, automatic, effortless, slow-learning, associative, emotional, stimulus bound 
system 2 (slow thinking, reasoning) - serial, controlled, effortful, rule-governed, flexible, more logical, access to past, present and future
122
Q

What did Gigerenzer suggest about the method of communication and how it affects the way we reason?

A

we aren’t actually bad at reasoning/always using dodgy heuristics
we are just not used to questions presented in a statistical format
more used to dealing with frequencies

123
Q

What evidence supports Gigerenzer’s suggestion?

A

found that when given problems in terms of frequency rather than stats, people performed better
went from 22% in the other format to 48% other research found up to 80% correct
same thing happened with representativeness, went from 75 wrong to 25 wrong
but some things aren’t phrased in either freq or probability and people still get them wrong

124
Q

Under which circumstances is problem solving necessary?

A

There is a clearly defined goal
The path to the goal is uncertain
When it involves conscious planning

125
Q

What is the path through a problem state?

A

Initial state > intermediate states > goal state

navigate using operators

126
Q

What is the algorithmic approach to problem solving?

A

Go through all of the possible moves until you find a solution (and potentially carry on until you find the most efficient one)
not how humans actually solve problems

127
Q

What are the strengths and weaknesses of the algorithmic approach to problem solving?

A

guaranteed to solve the problem
inefficient and slow - becomes very long very quickly
very cognitively demanding

128
Q

What are the shortcuts that people use to search the problem space?

A
  • heuristics
  • depth-first search
  • breadth-first search
  • hill climbing (only picking moves that make current state more similar to goal)
  • means end analysis
129
Q

What is a local maximum?

A

a point where you have to go back to go forward because no further step takes you closer to the goal
helps people if you encourage them when they reach these points

130
Q

What is means-end analysis?

A

iterative method
set goal as making current state the same as goal state
set subgoal as eliminating difference between the 2 states
if it not possible to do this directly, create a subgoal to make it possible to do whatever it is you can’t do
recursive - keep doing it whenever needed until problem solved

131
Q

What evidence exists for means-end analysis?

A
  • matches verbal protocols (when people are solving tasks they say things that suggest they are thinking in this way)
  • models predict performance (can predict how long it takes people to solve a problem based on how many subgoals they create
132
Q

How is the prefrontal cortex useful for navigating problem spaces?

A

planning and goal setting

133
Q

How is the medial temporal lobe useful for navigating problem spaces?

A

task switching and response inhibition
remembering the past and thinking about the future
found activation in MTL when people are asked to imagine the future

134
Q

What are the parallels between spatial maps and cognitive maps?

A

hippo has place cells that activate based on location
entorhinal cortex has grid cells which fire when an animal is in different regions
evidence for neuroimaging that hippo maps abstract spaces of fractal patterns

135
Q

What does the Chinese ring puzzle show?

A

Chinese ring puzzle - ps struggled to solve it in person but solved it v quickly when it was online
struggled to solve it because they didn’t understand the rings and how the set up worked
had to define the problem space before they could work it out
showed that sometimes solving the problem is hard because the problem space isn’t understood

136
Q

What is the relevance of gestalt psychology in problem solving?

A

a way of looking at problems
restructuring the brain to see them differently
string dilemma - strings in the room which are far apart
have to hold both of them at the same time
objects on the table which you can use
have to use objects to make string swing
most ps didn’t get it until they were given a hint

137
Q

How is the multi-step problem solving model limited?

A

not all problems are solving gradually - with some you have no idea until you have an idea and then you sort it out very quickly

138
Q

What is functional fixedness?

A

The idea that we struggle to think of objects having any other use apart from the one they’ve been created/designed for

139
Q

What are some of the ways that we can restructure our brain to help us solve complicated problems?

A

changing the representation of the problem
(can change how you think about it)
incubation i.e. taking a break (weakens the strength of your previous assumptions and lets you think in different ways)
gets rid of functional fixedness

140
Q

To what extent do people solve problems by likening them to other analogies?

A

we can but we don’t do it naturally
tumor problem - 5% solved it
with analogy - 20% solved it
with analogy and hint that it’s related - 92%
problems are much easier to solve when they’re more superficial1

141
Q

What evidence is there that using analogies to solve problems can be helpful?

A

3 biology labs that used analogies were more successful than the other one that didn’t

142
Q

What is the nuance between the analogies people produce and those they remember?

A

people produce deep analogies (based on the structure of the issue)
but they recall superficially similar ones

143
Q

What is the nuance in how experts group problems together vs novices? What does this mean about problem spaces?

A

novices group problems based on surface similarities whereas experts do it based on underlying ideas and principles
being an expert changes the landscape of the problem space