Cognitive Psych: Exam #3 Flashcards
1
Q
Semantic memory definition
A
a vast repository of knowledge, facts and general information
2
Q
Benefits of categorization
A
1) reduces the complexity of the environment
2) reduces the need for constant learning
3) can guide you on what your next actions should be
4) enables the ordering and relation of classes of projects and events (ex: superordinate / subordinate)
3
Q
Superordinate definition
A
broader category
- items at this level have few features in common
4
Q
Subordinate definition
A
more specific
- items at this level have lots of features in common
5
Q
Categorization task definition
A
Items appear on a screen and you have to sort them into categories - you are not told the rule but receive feedback on whether you are correct or wrong.
6
Q
Types of rules
A
1) disjunctive: use the logical “OR” to relate stimulus attributes (ex: a bird has feathers or can fly)
2) conjunctive: use the logical “AND” to relate stimulus attributes (ex: a bird has feathers and can fly)
7
Q
Rule learning definition
A
Give participants the relevant attributes and have them learn the rule
8
Q
Attribute learning
A
Give participants the relevant attributes and have them learn the rule
9
Q
Critiques of the Categorization Task paradigm
A
1) task if artificial: things do not fit into perfect boxes in the real world
2) many items within a category share some attributes but not all
3) categories vary along continuous dimensions
4) Doesn’t capture hierarchal information
5) doesn’t capture typicality
10
Q
Characteristics of natural categories
A
1) hierarchal
2) typicality
11
Q
Study on natural categories and hierarchies
A
participants listed features for items presented – wanted to see how many features you could come up with at the superordinate, basic and subordinate levels
results: very few shared attributes at the superordinate levels but it increases as you move down the hierarchy (get more specific) – it is hard to come up with features that encapsulate ALL elements of a category
12
Q
Basic level categories are…
A
The most differentiated: avoid the extreme of sharing too many or two few attributes
- we probably learn these first
- carries the most information: best cognitive economy and highest cue validity (it is what we think of first when we think of the category – we say “dog” when presented with a picture of a dog not the specific breed or the general category “animal”)
13
Q
Role of expertise in categorization
A
if you are more experienced with items presented you are faster in all category levels (basic is usually fastest but here is basic AND subordinate)
14
Q
Typicality definition
A
some items are more representative of a category than others
15
Q
Hypothesis for typicality of categories
A
typical exemplars share many features of the category
- Study: items rated most typical share the most features with other members
- good for discrete categories but not continuous
16
Q
Multidimensional scaling
A
rate how similar the following pairs are – map results onto a graph (more similar items are closer)
17
Q
benefits of multidimensional scaling
A
1) typically outperforms the standard similarly ratings
2) provides a visual depiction that may lead to new insights
18
Q
prototype definition`
A
category example that has the average attribute values (best representation of a category member)
19
Q
Prototypes and phenomes
A
phonemes: variation across how we say syllables
study: you have to say when a sound switches from I to Y
conditions:
- P condition: comparison contains the prototype
- NP: comparison does not contain the prototype
results: if you play the prototype (P condition) you are not able to detect the change
20
Q
Animals and prototypical representation
A
they do not do this - only humans
21
Q
Exemplar rule of protoypes
A
You store every item of the category you have seen and compare it across the stimulus
- we do not use this method
22
Q
Prototype rule
A
a strategy that selects the category whose prototype is most similar to the classified item
- only retrieving the prototype
- downside: lose breadth of the category (only retrieving the average)
- we use this method
23
Q
When is each rule of prototypes used?
A
prototype rule: used more in early learning – when introduced to new things, usually only exposed to more prototypical patterns
- once you get better, you switch to exemplars to see the depth of a category (might be only used when the number of exemplars is relatively small)
24
Q
Goal derived categories
A
a category that is formed to fulfill a specific goal
25
Theory -based categorization definition
The importance of features for category membership depends on the role those features play in the theory underlying membership
26
Theory -based categorization study
Given 2 features (remove X, Y, or Z) and see if the participants can still guess the category
Conditions:
-Given no background information
-Given a background story that connects X→ Y → Z
Results:
-When you are missing X (the first feature) in the background condition you are likely to say it is not a member of the correct category
- You have stored that the first feature is the critical and leads to the other ones (root principle)
we expect features to fit together in a meaningful way
27
How should we learn categories?
Study:
-Center condition: shown things close to a prototype (average of features)
-Coverage condition: Shown things from all around the spectrum → define little clusters along the extremes and pick from the center of those clusters
Results:
1) Your are best when given what you previously saw (if you saw something in coverage, you are better with a coverage rock and same with a center rock)
2) You do better overall when you had the coverage condition: you want to see all the items and how variable the category is
28
Categorization and age
5-8: usually pick the most extreme example
9-adults: pick the average followed by most diverse most often → you are picking the average / coverage condition
*As you get older, you try to sample the category as much as you can*
29
taxonomic relations
organization of concepts based on their similarity of features
30
Thematic relations
organization of concepts that appear in the same context
31
Hierarchical Network Model
Theory: impose a strict hierarchical structure → common information stored at the highest level (minimizes redundancies)
32
Pros of the Hierarchical Network Model
- efficient
- meaning is accessed by traveling along pathways
- good for accounting for many aspects of behavorial data
33
study with the Hierarchical Network Model
hypothesis: Speed of access should depend on the distance between two “nodes” in the network
test: Test: sentence verification task what varies the distance between the nodes (ex: a canary can fly (1 level), sing (0 levels) or has skin (2 levels))
findings: the more levels you have to "go up" the longer it takes for you to respond
Another test: verify two sentences in a row
1) a canary can fly vs. a canary is a bird
2) a canary is a bird vs. a canary can fly
Which combination is fastest?
Findings: You are faster in the first condition because you are already in the “bird” category due to thinking about flying
34
Criticisms of the Hierarchical Network Model
1) Can not explain typicality effects: you are faster to say typical items but this model doesn’t account for that
2) Some sentence-verification results are problematic for the model (you are quicker to say a dog is an animal rather than an animal, and you are faster when the response is a "no")
35
Feature Comparison model
theory: the more overlap of features you have, the longer it takes you to determine the category
- unstructured, meaning is derived from the set of features and you determine the category from the comparison of shared features
- ex: is a shark a bird - quick to say "no" as they share few overlapping features
- dual process: 1st stage: fast, automatic overlap comparison of characteristic and defining features 2nd stage: slower → attentional overlap comparison of only defining features
36
Advantages of feature comparison model
Accounts for the typicality effect (more typical items are recalled faster)
Explains the quick “no” response reaction time
37
Criticisms of the feature comparison model
Major predictions are based on typicality ratings: little direct support that you are using feature overlap to make similarity ratings
Are you really computing a similarity value (comparing overlapping features) every time? → seems like this would be very slow
Can people really identify “defining features” of a category? → are we consciously aware of this
38
Spreading activation model
Theory: Unstructured network of interconnected nodes which represent conceptual information - activation spread to related nodes in the network (length of path varies by strength)
- explains priming: nodes that are connected should co-activate one another
- unstructured, spreading activation to access meaning, concepts are linked to related concepts
39
Mediated priming effect
priming effect for two concepts that are not directly connected
- we see this effect: is explained by the spreading activation model
40
Spreading activation model study
Semantic priming lexical decision task: you see two words → read the first one silently and then make a lexical decision judgment on the second (if the priming word is related, it should activate the node for the related target word → you should be quicker in these conditions)
Results: On average, you are faster within the related condition → “priming effect”
41
Criticisms of the spreading activation model
1) Flexible: somewhat arbitrary → how do you know a path is longer?
2) Hard to make “new” predictions based on this model
3) Determination of whether items in the network are connected is based largely on similarity ratings (no direct evidence for this → subjective accounts)
42
PDP model: parallel distributed processes
Knowledge about the meanings of the words and objects emerges from the interactive activation of perceptual, motor and linguistic representations across different modalities
- meaning arises from coactivation of the visual, sensory, motor, small, shape texture etc. regions of the brain for a specific item
- similar things have similar partial activation
43
Where are patterns of activation encoded?
Anterior temporal cortex:
- hub that provides access from one feature to another
- similar things have similar pattern of activation here BUT are different in modality specific regions
- Ex: flamingo and robin: both birds → similar semantic representation in the hub even though their features are different (different perceptual level/neurons)
44
PDP model accounts for...
1) Category effects: More similar things are close to / have overlap with their firing neurons
2) Typicality effects: less similar things have less neural firing patterns in common
3) False RT effect: you are faster to say “no” because the neural patterns are so different
4) Priming effects: similar concepts have overlapping neural firing
- more biologically plausible
45
Semantic dementia
progressive loss of knowledge about the meaning of words and objects
study: given sentence verification task
-You are usually fastest at base level category (animal is better than canary)
-Findings: specific attributes (canary) become distorted with semantic dementia but more general attributes (animal) remain intact
--- Accounted for in this model (PDP) but not in the feature comparison model (and other hierarchical models as they say you need lower level features to get to higher
46
Perceptual symbols model
Proposed that semantics are grounded in the same perceptual systems that permitted their acquisition in the first place
Ex: the meaning of chair is what it is because you sit on it → its based on how you interact with it
47
Perceptual symbols model studies
Study: read you a sentence and then shown a picture - asked was that picture in the sentence?
Ex: the carpenter pounded the nail into the WALL vs. FLOOR → you are faster to give the correct response when the picture matched the correct orientation (going into a wall (sideways) or the floor (down))
Study: given pairs of words:
Ex: blender-loud (auditory) → primed with either leaves-rustling (also auditory) or cranberries-tart (taste)
You are faster when you are primed with the same modality (sense)
Study: read a list of words while in a brain scan:
Conditions: either read outloud kick, pick, lick, sick or imagine moving your body in the way of the word (kick=move your foot)
Results: the overlap from reading words is a similar area of the brain as when you are actually imagining the movement happening
48
Criticisms for the perceptual symbols model
How do you represent abstract concepts that don’t have sensory experiences associated with them (peace, justice, love etc.)
49
MINERVA model
Semantic memory doesn’t truly exist but instead if defined as the summation of past episodic experienced
Ex: the meaning of “dog” to you is all the interactions you have had with a dog over time
Study: categorization task
Given many pictures of birds: varies on how similar they are to the exemplar
- The closer the image is to the exemplar, the faster/more accurate you are → based on the probe image so it is dealing with episodic memory (your memory of the probe)
50
Schemas definition
a general knowledge structure that provides a framework for organizing clusters of knowledge
51
Script definition
knowledge about what happens during a routine activity
52
Study on Higher order sorting
Study: given steps organized by centrality/importance vs. sequential order
- findings: when you are given a central idea you have a faster reaction time → whether it is early or late in the routine doesn’t matter, it’s the important that matters
Study: generate all events in a routine
Asked to generate from most to least important or for first to last
Findings: you come up with more items faster when you are listing items in order (harder to list importance)
*Overall: centrality and temporal order are important
Centrality: allows us quick access to a routine (script) and what we need to do
Temporal order: important to achieve all the steps to the event
53
False memory
Given many related items but you are not presented with the critical world
The related words spread activation to the non presented critical lure → it is activated through association
Can also get mediated false memory: wet (direct) → slippery (mediated)
54
Alzheimer's and semantic memory organization
Category fluency: name as many animals as you can (searching semantic memory)
- AD are deficient in this area (can not produce as many items)
- Normally: you are quicker if primed with a related word (spreads activation)
- Dementia: you are still faster → show priming effects (the node is not completely gone but you can not access it directly)
Now given kidney → organ → piano (words with multiple meanings)
- You are not primed → you constrained the activation of organ based on the first prime (kidney)
- Findings: people with dementia can not control what is activated (attention argument)
*The ability to constrain semantic activation based on context declines in Alzhimers*
55
propositional theory of visual imagery
the idea that all knowledge including spatial, can be expressed in semantic-based propositions
“The ball is red” → you are expressing the image of the red ball in a series of semantic statements
56
Visual imagery theory
visual images are processed in the same way as actual perceptual information (visualizing is akin to seeing)
57
Ways to test the theories of visual imagery
1) visual scanning
2) serial vs. parallel processing
3) mental rotation
4) interference
5) neuroscience
6) picture memory
58
Visual scanning study #1
Study: study a map for a minute or two and then asked to picture the image in your head and imagine a black dot moving from one location to the other and press a button when you reach it
If you are actually scanning an image, this task will take you longer based on distance
If you are just remembering semantic encoded things it will take the same time
Findings: it takes you longer the farther the two spots are away from one another → supports visual storage of images
Criticisms: participants might catch on to this and predict how long it “should take” (your guess for how long it would take maps directly on to the results) → you could just be guessing not actually scanning
59
Visual scanning definition
A shift of attention across a visual display or image
60
Visual Scanning Study #2
Study: given different shapes → asked how long the shape is (stretch them out linearly, ex: spiral)
We have a hard time stretching things out and trying to guess how long they are (we can not predict our scanning time)
Assumed lines were the longest - spirals and mazes looked less long → not the results found; suggests visual images are stored visually
61
Serial vs. parallel processing
Verbal operations are necessarily sequential
Study: Show pictures of a face OR have a verbal description of a face - ask if a face matches what you studied
Question: does the number of relevant features (amount of features changed) influence response time → if in parallel, should be the same response time, if serial, more features should take you longer
Results:
If you studied a picture: the number of features did not increase reaction time → visual features are processed in parallel
If you studied a description: the more features you have to go through, the longer it takes you → verbal codes are sequential
62
Mental rotation
The more you rotate an image, the longer it takes you to make a response on if the target and presented image are the same
In your head, you are actually twisting the image
- stored visually
63
Interference can be reduced by...
shifting categories (release from PI if you change content)
64
Interference study in visual images
Verbal condition: play a sentence and have the participant monitor the sentence for nouns (say out loud “yes” or “no” for each word)
Visual condition: study a block diagram and then answer “is this corner at the top or bottom of the object” as you mentally trace the figure in your head
- have different outpout modalities: verbal, tapping, pointing etc.
Findings: Shows that the visual condition (“F”) is stored as visual as visual interference interferes the most
Study:
Study a list of 11 letters (verbal code)
Study a list of 6 letters (verbal code) and 5 spatial location (still 11 things total)
Findings: remembered more in the 6 letters and 5 spatial → suggests that spatial things are stored separately as there is less interference
65
Neuroscience of visual imagery
Tasks using visual imagery activate the same brain regions as visual perception
66
Picture memory
very effective memory code
Study: participants studied 2912 pictures for three seconds each → asked if they studied it on a test
Novel: new to you
Exemplar: similar to what you studied but not exactly what you studied
Results:
You are very good at saying what you did not see (96% correct)
Even with many similar pictures (exemplars) you are still very good at saying you didn't see this (64 similar pictures, 76% correct) -- Picture memory is very good
67
Dual coding theory of picture memory
Two forms of elaboration:
1) Verbal associations
2) Visual elaboration
Abstract words are harder to remember → harder to form images for them (stuck with only verbal associations)
68
Imagery potential definition
the ease with which an item can be visualized
- abstract words have low imagery potential
69
Association value definition
the number of item associations generated for a particular concept
70
Study in association value and imagery potential
Study:
- Varied association value and/or imagery potential (H=high imagery potential, L=low)
Juggler-dress: H-H
Goblet-duty: H-L
Miracle-Hotel: L-H
Spirit-interest: L-L
Results: high imagery items are better recalled than low imagery AND better than high association
- Visual learning is very effective
71
Relational processing
information specifying how concepts are related
72
Does visual benefit come from the imagery potential of an item or how they are related?
Study: 12 pairs of items: concrete and abstract
- Integration condition: rate how easily these two words form a combined image
- Imagery condition: rate how each word on its own evokes a mental image
Results: The difference between a verbal vs. visual code is due to association → visual codes are only better when you are relate them to one another
73
Individual differences in spatial navigation
Study: have to walk through a map that is marked with building landmarks
Given test on items you saw within the pathway (on the route you took) OR on the way that connected you to the pathway (inbetween test)
Three groups found:
1) Integrators: put all the images into one big mental map
2) Non-integrators: better on the within route test, but not as good on the between-route test
3) Imprecise navigators: bad in all conditions
- People differ in the ability to integrate and update information into a visual image → not an “automatic” process
74
Limitations in how we store visual images
Given a picture and then asked if this is the picture you saw + are these shapes integrated into the picture (the picture is no longer shown)
duck vs. rabbit: You have a hard time reversing this image in your head (you can not mentally reverse it once its been stored)
75
Are images sensitive to bias?
Cities: Reno is more west than San Diego but we use bias that California is more west than Nevada and we get it wrong
Rate relative distances between cities in germany
Varied whether the city was within or across from the former wall
Evaluation of whether individual had positive or negative view of reunification
Results:
People who had a negative view of reunification rated the distance between the sides as very far
