Professional Documents
Culture Documents
CONTENTS
Introduction ...........................................................................................................................................................2
Featural Approach............................................................................................................................................... 15
Consciousness ......................................................................................................................... 28
DREAMS ............................................................................................................................................. 38
Creativity ............................................................................................................................................. 41
INTRODUCTION
Concept: knowledge about something specific, based on features or a prototype, that isn’t true or false, it just is
Thinking: (Solso) thinking is a process by which a new mental representation is formed through the transformation
of information by complex interactions of the interactions between judging, abstracting, reasoning, and problem
solving
(Mayer)
1) Thinking is cognitive (occurs internally and only inferred from behavior)
2) Thinking is process involving the manipulation of knowledge or information
3) Thinking is directed and results in behavior (even brain activity) that provides the means to an end.
Questions:
These are all too rigid! They are not sequential; a 2 year old can be in the commitment stage and a 57 year old in
another. Very independent, doesn’t fit everyone so well.
There are assumptions we have about knowledge, how we learn and determine weight on what we know
(certainty)
Kitcher and King developed Reflective Judgment Model – concerned with beliefs people hold on what they know –
also sense of stages
Schommer – beliefs about knowledge are composed of independent dimension (versus being able to place in a
box)
• beliefs are independent – students may fall on certain point dimension because done it before
• our beliefs are based on where we lie on these 5 dimensions
Baxter Magolda…
September 4, 2006
• World has infinite variety, everything is different. How do we react to the variation? We couldn't react to
everything... would be sensory overload. we wouldn't be using knowledge from past experience.
• There are certainties that we don't have to attend to - 99 % of the world... categorization and concept
formation allows us not to.
We are stimulus response beings
• Little changes in environment aren't important to me. Salience is an important factor, what is and what
isn't important in the environment
• Classifying/categorizing an object allows us to know a lot about it already, so we don't have to learn new
things. Categorization brings set of assumptions.
• Context is important.
• Negative Aspect: We have a propensity to impose a classification on new events, ppl, objects, and bring to
mind a set of expectations about that thing. (stereotype) Our expectation could be wrong.
• We are cognitive economists: don't want to think more than we have to. But we can be wrong.
• Is heart of learning - developing child learns classifications and labels used by society (for ex, many
different colors of blue) learns variations under one concept roof.
• Criteria adopted for classifications varies on cultural group
• We are all human, and all do things in similar if not exact same way (process has to be same)
concepts:
• concept provides taxonomy of things in the world
The task that we did, first was implicit, second was explicit.
• Hammer and Elby (2002): Epis, sensitive to context (domain, religion, politics)
believe that despite everything, proving otherwise
construct is a broad way of thinking/believing that is too large: we can't test it
salient - outstanding
cognitive economics - humans don't think more than is necessary
Functions: little role in actual concept formation... is after the fact, the concepts are formed.
These all build upon the concept, they are not part of the construction process.
To have concept (dog, for example, X) is to know something about the entities that belong to the concept X, can
use these properties to classify novel objects. This is reversible as well - we can use a concept to infer features
about an object.
• words give us access to our lexicon (mental dictionary), and once the stimulus is there, it will go all the
way through fruition (ballistic movement)
• we always use labels - this is how children learn.
• if you have some prior knowledge of seeing a concept, X, then you don't need much evidence to
determine if a new object is an entity in X.
3 Views, believe in categorization and inference. (the only two things they agree upon)
1) Classical - oldest idea. concepts are relatively stable, mental representations, Internally within individuals (will
be same 50 years from now) and across individuals (same concept, same properties in mind). Concepts are
bounded units of knowledge. Classical view says we are 100% certain, it is or it isn't if taken all info
necessary/sufficient. inferences are deductions. we go from facts and come up with a final result. properties that
are inferred are necessary.
Rigidity and stability increase as we go up towards classical view. The views differ in the certainty that a person
has in these processes. classical view has complete confidence in any categorization that takes in all the relevant
properties. never 100% certainty in other two views.
concepts have dual functioning of categorization and inference; differ with certainty of accomplishment of such
things.
Global component property vs. holistic property is the same thing. Global component gives more information than
component.
Features vs. Dimensions: A dimension is a quantitative component, a feature equates with a qualitative
component
Dimension: has an amount, less or more (weapons- potential harm). can handle quantitative variations,
features can't
Feature: quality has or doesn't have (is or isn't)
If two concepts differ in respect to feature, one has it, one doesn't
If two concepts differ in respect to dimension, one has more of quality than another
Classical View: psychological theory about how concepts are represented in humans and other species. Hull
(1920) given credit for starting.
• Classical view is concerned with features
• Categorization is a dualism- there is representation of category, and process by which it is formed
The classical view is a proposal about representations, doesn't talk about process.
1) representation of a concept is a summary description of an entire class rather than a set of descriptions of
various subsets or exemplars of that class. (to represent bird we don't have different descriptions by species, we
have a summary description for all birds) It is parsimonious, reduces memory space and energy.
Three Characteristics:
more often than not is the result of an abstraction process
need not correspond to a possible, specific instance
representation applies to all possible test instances (applied to anything I think is a pen)
2) The features that represent a concept are singly necessary and jointly sufficient to define the concept. Every
instance of a concept must have that feature for it to be singly necessary. For a set to be jointly sufficient, every
entity that has that set must be a member/instance of the concept.
3) Two instances of the same concept cannot be totally disjunctive under classical view. They can be partially
disjunctive, so a red breast distinguishes a robin.
4) Nesting of features in subset relations. If concept X is a subset of concept Y, the defining features of Y are
nested in those of X. Defining features of bird are nested in robin... the subset must include defining features of
their own or they wouldn't be unique. Supersets are most basic, non existing abstract ideas.
These assumptions say nothing about the possible relations between features. Features are treated as
independent from one another. This is adequate for paradigms (kinship concepts, gender and age). This does not
fit with taxonomies, (animal concepts) where the features overlap and are clearly related. (don't worry about this
too much :) )
1) Functional features - we've said nothing, and they exist! The classical view deals only with structural features
that define an entity in isolation. Many concepts (especially made by humans) are defined by their functional
features; therefore the classical view can't handle all concepts.
But there's nothing in the above rules that says a feature can't be a functional one! They're in.
Why would someone think that the view only looks at classical features?
2) Classical view disclosed (partially) disjunctive concepts. Many concepts are disjunctive; therefore classical view
cannot handle all concepts. (to be member of concept in classical view, must have set of sufficient and defining
features, if they are there then it is X, if not, then not) Baseball strike is a disjunctive concept - hit and miss, foul
ball, all called strike.
Counter: do totally disjunctive concepts really exist, or is this a contrivance of man? (blue things, red
things with dot, etc - we are forming concept so anything goes.)
• Rosch - evidence - said there may be disjunctiveness when we are at the superordinate level because
there are very few features. But classical view says we never see superordinate level - we see a tree, not
tree.
Experimental Criticisms:
1) These findings deal with how people use concepts, reflect cat. processes. We cannot go directly from the
findings
1) If we put someone in lab, show work, ask if good X or Y must infer from behavior process that happened.
Constricted by representation. Can only use necessary features for classification. Left with task of interpreting
findings, inferring process, and generating models. We run models and see if produces same result as human
beings. Model is based on rep, so rep must be correct.
Water is an example of something defined as same thing, but has different defining features (ice vs. vapor)
they are partially disjunctive - share same characteristics or else wouldn't be water.
Quick Quiz
1) Perry scheme doesn't fit with what is going on because it is rigid - move from one stage to next, and there are
different stages
2) Three stages: objective, subjective, and combination of two. Cune, Cheney, and Weinstock.
3) Classical, exemplar, and probabilistic views both 1) categorize and 2) draw inferences about the thing
4) true
5) conservative focusing strategy - form hypothesis, focus on single feature hypothesis, get feedback and change
based on feedback.
Group Discussion
vvs4@duke.edu
rhl@duke.edu
maxine.mitchell@duke.edu
htc3@duke.edu
wee5@duke.edu
Animals
blooded
consumer
animate
reproduces
Mortal
stimulus response
Wolf
animal
fur
teeth
tail
pack
howl
carnivore
Bird
animal
feathers
beak
wings
lays eggs
Furniture
inanimate
man made
functional
decorative
convenient
Chair
furniture
provides sitting
grounded
flat surface
Stool
Furniture
provides sitting
(chair)
backless
Plant
chlorophyll
photosynthesis
cellular
living
Complexity Model
• Smith and Medine developed model to see if could make model based on classical view assumption that
would account for typicality effects
has classical features, nesting, etc
Two steps: Access and comparison
We immediately start accessing defining features of target and probe. Is a chicken a bird? As soon as defining
features are available, we compare those of target concept and those of probe. Then we say yes it is a member
only when every defining feature of target has matched a defining feature of probe.
We can say no as soon as any feature of target mismatches feature of probe.
Target = category
Probe = instance
Why when we are given a concept and asked to produce its instance, why do we name a typical instance?
• Acts as cue, memory probe, and the thing that comes to mind first is the most typical thing.
• Prototype is an abstraction, but it doesn't represent an entire class. Summary representation is thing that
is that entire class. (includes atypical members because they have necessary and sufficient features)
HAMPTON
Shared features are a good measure of typicality.
Some features that were listed for the concept were non-necessary ones (flies for bird)
Non-necessary features were correlated with categorization performance
So non-necessary features were being used for categorization. They have impact! It is difficult to
reconcile with models just based on the classical view.
Weak point in experiment: He asked people to list features... how does he know they were listing necessary vs.
non-necessary? He assumed features they listed contained necessary and non-necessary features. People can't
write down necessary features! (we couldn't do it for dog) assumption listed features correspond to true defining
features of concept. could argue features listed for particular concept are epi-phenomenon, that is, secondary to
the real thing... (this is said by classical view person)
1) categorization is based on defining features that are relatively inaccessible or at least very difficult to report on
based on introspection
2) these defining features are correlated with other non-necessary features that are easy to report (are accessible)
therefore the correlation between categorization and non necessary features is being mediated by defining
features.
Epi-phenomenon: appears that using non-necessary features to generate categorization, but we are using defining
features, just not aware of it. We don't have access to defining features, just non-necessary features.
How could we test this? Multi-dimensional scaling: used to eliminate need for subjects to list features.
intrinsic category: has more intrinsic features... take it in isolation; is it still what it is?
extrinsic category: gain membership because of relationship with other things. Leads to fuzzy category
membership. Clothing is very extrinsic category - almost anything could be considered clothing. taken in isolation,
don't have to be eaten (have extrinsic features) to be in category.
Asking people to produce supposed defining features is problematic, we don't know if naming non-necessary
features
Multi dimensional scaling: Basic paradigm: participants given pair of concepts from particular domain (robin,
sparrow) each pair contains two subsets of the generic concept (robin, hawk) Also, they have pairs that include
concept and one of subsets (robin, bird) and see up to 1,000 of these pairs and rate similarity. Hit number
between 1 and 7. Each pair repeated at least 5 to 10 times. All of these ratings are inputted into a scaling
computer program, which determines a "Hamming distance" and a geometric space is outputted. The items
involved in the ratings are represented by points in this 2D space.
Closer things clump together based on similarity. We can interpret dimensions (ex, size, verocity)
Dimensions should be decided before experiment.
We can also do this test with features, figure out what features cluster around what concept. If chicken is far away
from clump, then probably isn't good representation of bird with defining features of bird.
We are talking about these "dimensions" but we are still dealing with features.
People are judging similarity on non-necessary (perceptual) features. This flies in the face of pure classical view
What does this imply with regard to similarity judgments? Bird should be more similar to Chicken because
they have two things in common. X should be more similar to Y than to Z. Subset should be more similar to
immediate superordinate over any distant set. This doesn't always happen.
To test: Give subset to people and ask to generate superordinate. Say rose (most people say flower, but some
may say plant)
Loftus and Scheffe demonstrated that a subset was no more likely to produce its immediate
superordinate than a distant superordinate. What's wrong with this argument according to a classical view
theorist?
We are asking person to produce.. there may be something mediating this response, the person
may skip over something.
Better Technique: do direct similarity ratings of a subset to both immediate and superordinate (in same way does
multidimensional scaling)
(Rose, Plant)
(Rose, Flower)
Results: Most of subsets are rated as more similar to their immediate than to their distant superordinate, which
supports classical view. The problem is there are exceptions:
Exceptions: chicken and duck are consistently ranked as more similar to animal than to bird.
Because they are food. We eat animals, we don't eat birds. Although classical view holds for many cases, it
doesn't hold for all cases.
Another problem from nesting assumption: a probe concept should be categorized faster when the target concept
is a distant superordinate than when it is an immediate one. Why? there are fewer features to compare. But it
doesn't happen. Robin is judged more quickly as a bird than as an animal.
SUMMARY
Early studies show that categorization was faster when the target was an immediate rather than a distant
superordinate. Later studies showed inconsistencies in those results. The exceptions turn out to be cases when an
instance is rated as more similar to its distant than to its immediate superodinate. Chicken is a bird, chicken is an
animal, etc.
Similarity judgments: classical view predicts an advantage of higher similarity ratings for immediate over distant
superordinates and that is true for many cases, but there are exceptions.
Categorization Times: straightforward models based on the classical view predict an advantage (faster reaction
times) for distant over immediate superordinates. This is not true for the majority of cases. but there are
exceptions! (chicken)
1) simple typicality effects (ratings, categorization times, error rates, ease of learning, order of production, and use
as cognitive reference points) Not hard for classical view to deal with, can be explained using complexity model.
So not a huge problem. TYPICALITY SHOULD NOT EXIST in classical view.
2) Determinants of Typicality: typicality and the distribution of features across concept members are highly
correlated. (Gradients in category membership) This is tougher for classical view... needs add ons (like
extrinsic/intrinsic features) "They are there but you don't know what they are" There is a gradation, and there
shouldn't be.
3) We are using non-necessary features. Difficult for CV to deal with. Classical view says "it looks like we are using
these non-necessary features, but subconsciously using defining"
4) Nested concept problem (most solid evidence)
Taken together, these things form solid argument against Classical view.
• Salvaging continued...
Access links between concepts: Ex. Animal -> bird -> chicken. Animal is a bird. Animal is a chicken. Links between
concepts that people get while acquiring concepts. This is an addition to classical view. Led to widespread and
used theories of memory - spreading activation. People can check interconnect links rather than checking features
of concept when doing categorization facts. Information comes from links. Links of paths will vary. Longer paths =
longer processing time.
How does this account for typicality effects? An atypical member has a longer path than a typical member, so
takes more time.
Paths allow for shortcuts. This model is lacking in constraints. It can come up with something new very fast, can
get complicated. This also cannot explain typicality relations and results of many experiments. Also, we can make
something typical or atypical just by changing features. Also, this has nothing to say about use of non-necessary
features, which is apparent in many experiments. Doesn't acknowledge disjuncted concepts.
Drops nesting assumption all together. Takes care of unclear cases and all of the experimental evidence about
subset relationships. allows for translation between one feature and another. You can translate flies/egg laying
into animate, you have to be animate to do any of these things. This feature becomes one thing - one feature
translates into another.
Animate implies flying
Flying implies animate, etc.
Processing time requirements: the more times you have to translate something, the more time it takes, more
chance make an error, Typical things require fewer translations... accounts for typicality effects.
This is called a rule plus exception approach. Rule is classical view: the exceptions are translated (the atypical
ones) So something typical we use classical view, atypical we use translation effects. This approach has no way
dealing with use non necessary features because it has to translate them. Still relying on use of necessary and
sufficient features for categorization, gets rid of nesting. If we add enough levels of translation, then there isn't
any use for a set of defining features. If egg laying = animate, why even have animate? So in actuality they are
getting rid of defining features, so it isn't classical view anymore.
We assume concepts contain core features (that are consciously inaccessible) that are doing the brunt of the work.
To do categorization we focus on this core rather than perceptual features.
THIS view says maybe we don't need to focus on the core. If people do USE ID features, why not focus on them?
The core will always be there. The non-necessary features (perceptual ones) ARE important.
There is no need for nesting because these are non-necessary features... they don’t need to be there! If using ID
features instead of core, is there evidence connecting pictures and words? Studies show we list same features as
looking at things versus words... both bring up perceptual features.
This fails in discussing defining features - but it is classical view! Has difficulty with unclear cases - if given enough
time should be able to figure something out. Unconstrained, but not as much as previous two.
Shifting theoretical work away from defining features (core) so why bother with them at all? This view salvages
the classical view by throwing it away.
PROBABILISTIC VIEW
Three approaches
• Featural Approach
• Dimensional Approach
• Holistic Approach
Things in Common:
• Representation of concept is a summary description of an entire class
• The representation of a concept cannot be restricted to a set of necessary and sufficient conditions.
• Instead, it is some sort of measure of central tendency of the instances' properties or patterns.
Featural Approach
• Has summary description, representation of a concept is assumed to be results of an abstraction process
which is not necessarily realizable as an instance. and is used whenever a decision must be made about
membership in the concept. Only diff. with classical is we don't need necessary features
• Features that go into this summary representation are salient (tangible, perceptual, noticeable) ones.
They have a substantial probability in occurring in instances of the concept.
• If F1 is a features of XJ, a concept, Fi will be a feature of XJ to the extent that
one Fi is salient, either perceptually or conceptually, and
two, the probability of Fi given Xj is high, that is Fi tends to be true of instances labeled
Xj.
So there is a lot more flexibility for exceptions.
Featural approach dealing with modal features: the ones most likely to be there. The idea of a modal feature is
NOT the same as a modal instance. Modal features of representation may not be in any instance. Because they
occur in a lot of them doesn't mean they occur in all of them. A representation of modal features will be closer to
some instances than others, different than others of same concepts.
PROBABILISTIC HANDOUT
• Flying is salient property for most things called birds... could be big insect, bat, etc.
• Each feature has with it associated a weight that combines its salience and conditional probability.
Featured and winged have high salience and conditional probabilities. Higher weight = more important.
Weight is affected by salience and probability.
Salience going up, probability going down. Long necks and birds.
Representations presumably predict more modal features of class. Central tendency: mean is average, average is
all instances divided by total, medium is the one in the center of the distribution. The mode is the thing that
happens most often. So we are talking about modal occurrences. Makes an assumption about how we learn a
concept.
After encounter with first instance concept, we have only its salient features. As a hypothesis, we can hypothesize
that members of that concept has those features. Second instance has three salient features. (one isn’t there) so
the ones that occurred more often go into summary representation.
Key is that same features didn't all occur in same instance, came from different ones.
Featural Approach
Discrete features: everything is independent, but we have continuous types of properties (like size)
How is this represented in this approach? They are represented discretely. Size is still a modal feature,
because everything has size. Size is weighted lowly because is so variant. Certain features can be associated
with certain sizes. We would only access sizes if we needed a more fine grained analysis. This is just nesting...
features of small are nested in larger.
Apple: 5 features, F1 F2 F3 F4 F5
Is this specific apple a fruit? Add up weights of features of apple that are contained in fruit... we match F1, F2
from fruit (because apple has them) then we add up 1.0 and 1.0. We have to reach 3.5 to decide if it’s a fruit.
We have an internal counter that adds 1 every time we find a feature (cycle through process), then we
determine if we have a match, an we have an accumulator of information, and from that we make a
comparison stage to the criteria we've set, and then we make a decision. This is a time limited process
because it requires resources.
Doesn't say that although there is a high probability of occurrence, it occurs every time..
What if we are comparing a subset (the concept of apple) with the concept of fruit? The subset is a concept
itself, so it as associated weights as well. The difference in the process is the sum of both the weights is
accumulated.
Partially disjunctive: two examples of a concept that don't share all features.
According to feature concept, category membership is dependent on weighted sum. Since the same weighted
sum can be reached by combinations of various features, it follows that various feature sets can be used to
determine category membership.
4.9 3.9
Concept Furniture - F1(1), F2(1), F3(.9), F4(.8), F5(.5), F6(.7), F7(.4), F8(.6)
So they are both considered furniture, must meet 3.5. So furniture is a disjunctive concept... don't share all
features, we get sum from different features.
We do not yet have a measure of typicality.. but a higher sum might be a good indication of that. Explains why
typical members judged more quickly than non-typical (3.9 < 4.9)
There appears to be degrees of disjunctiveness. It can range from a simple 1 feature difference to 10 or 20
Unclear Cases
• The classical few was inconsistent with unclear cases. In general featural model, unclear case can arise
from two possible instances.
o If it is right at 3.5 or slightly below it
o If 3.5 was criteria for vegetable and fruit, and we reach them at the same time. If we reach
threshold for one or more different concepts at the same time.
• Are the ones weighted one defining features? There are no specification of defining features because
modal features go into representation.
Implicit vs. Explicit memory: implicit memory we can't remember context we learned in. Explicit we remember
specific context/time etc we learned.
Which level of Rosch hierarchical category do people list fewest features for? -superordinate.
Reaction times go down as typicality goes up... error rates go down as typicality goes up
Global vs. Conservative focusing... which one gives more information? Global focusing, we have global
assumption and it’s validated or invalidated. With conservative we can go on and on and on!
Classical view nesting: classical view says that instances of categories are either members or not, no uncertainty, if
defining features are present then it is, if not, then no. Nesting says that features of superordinate are in all
subclasses
Animal = superordinate
Bird = basic
Robin = subordinate
Results: Most defining features were rated as most typical. More features shared by a concept were most defining
or typical.
Strong correlation between non-necessary features and reaction times. The use of non-necessary features (judged
as typical) decreased reaction time.
Transformational Knowledge: sequence of images from point A to point B. Path leads to conceptual coherence
rather than naive theory that goes together.
1) to judge typicality of member directly affects the number of features it shares with its parent concept
Typical members match more features than non typical, so reach sum faster, less iterations through cycle, less
error, so typical members categorized faster and more accurately.
When listed, typical members listed before atypical ones. Target acts as a cue, cues most related thing in member,
which will be a typical member.
We use typical members as cognitive reference points because they are more like concept than atypical ones.
This model deals with use of non-necessary features very well, built on non-necessary features
Nested concepts: Similarity ratings: something is judged more similar to its immediate superordinate. Classical
view couldn't handle exceptions, but this can because weighted sum matches/reaches animal sum more quickly
than bird. so we say it’s an animal.
Categorization Times Themselves: since a typical thing share more features, it reaches criteria level faster,
classified more quickly
PROBLEM: very flexible at cost of making assumptions - are chicken's found on farms?
General Featural Approach deals with everything classical view couldn't. Works entirely with discrete features,
some people would argue that concepts are represented by continuous dimensions.
Spreading Activation
Collins and Loftus given credit for this.
• Grew out of computer science idea
• What kinds of evidence does a person consider when deciding whether or not something belongs to a
concept or not?
• In this model, concepts are represented as summary descriptions that contain many non-necessary
features. Each feature is weighted by its importance in conferring concept membership. This is called
criteriality. How much information does this feature give me? Represents information in an
interconnected network. The links carry the criteriality, the weights. Each concept contains a particular
feature that is connected to it by a labeled link that has criteriality in link. You only have to store features
once, and all types of birds are connected to these features. There isn't redundancy.
What is the process? Employ a spreading activation procedure. This is a continuous rather than a discrete process.
Once it is started, it has to continue until it is done. But it is time limited. Starts to decay as it goes. When a test
item in a concept is presented, activation begins to spread through the network. Activation from one source is
divided between all different paths from source. If two sources intersect at some feature, then the two paths
become available for further processing. These paths are then evaluated to see if they have the same label. If
they do, then those two concepts are considered to share that feature.
Two concepts might intersect as same feature, but links have different labels.
Spreading Activation
Key is that accumulating matches and mismatches. Once evaluated, matches and mismatches are treated in
similar fashion of general featural model. For every match, some amount of positive evidence is accumulated, the
degree of positiveness. This increases with the criteriality of the feature. For each of the mismatches, a negative
amount of evidence is accumulated. This continues until one of two things takes place
How does model deal with non-necessary features? They are built in! We use them!
Typicality effects: when dealing with a weighted sum, typical ones reach it faster than atypical
Can also explain priming effects: made it appealing to memory theorists. Prime: setup to elicit particular response.
Something that makes access to something else easier, In spreading activation, activation spreads to things most
closely associated faster and quicker.
Experiment by Rosch: asked people to determine if two simultaneously presented words were physically
similar or not. On certain trials, randomly the presentation of pair was proceeding by name of relevant concept
(fruit -> apple) Key finding is that relative to the trials that didn't have the prime, the concept prime facilitated
decisions only toward word pairs facilitating typical instances. So when we see apple activation is still spreading.
When prime is presented, it activates its associated concepts,. Activation spreads along all feature paths
emanating from those concepts, which results in activation of each member to extent member shares feature of
that concept. (based on shared features) When I see the next two items, they're activated as well.
Primes are basically context. You set up a context in which to perceive information.
An important finding: When we are deciding whether or not an item is a member of a concept, it appears that
people not only consider features that the item shares with the target concept, but they also consider features
that it has in common with other concepts as well. It's easier to decide an object is a chair, the less it looks like a
couch or stool.
Family resemblance: The greater the featural overlap with a contrast concept, the longer its going to take to
categorize an item correctly and the more likely the chance of categorizing it incorrectly AND the thing is going to
be rated as less typical. (Rosch and Mervis said was need to look at features called cue validity:
Cue Validity: originated with Warren and Resley, this was first application to concept formation:
The cue validity of a particular feature Fi with respect to a particular target concept Xj is such that it
increases with the probability that Fi occurs with instances of Xj and it decreases with the probability that Fi occurs
in instances of a concept that contrasts with Xj.
xk = contrast
We can use cue validity to determine weights (SO info about cue validity is being accumulated until we reach
threshold> it is reasonable to assume that when learning a concept, you often don't focus on the features of the
concept you're trying to learn, but rather the features of the concepts it contrasts. (Instead of focusing on "that's a
dog, focus on what it isn't)
• There is a summary representation using weighted non necessary features but this does not necessarily
rely on feature sums for processing. So when a person is given a test item and a target concept,
categorization is based on a two stage process
o 1) The person ignores all weights. they simply determine number of feature matches between
test item and target concept. Could be accomplished by simultaneously comparing each feature
of target with each feature test item. If the number of matches exceeds some high criterion, the
item mostly likely belongs to the concept, and we say YES. Processing is done! we don't need
stage 2. If first stage has some set of matches below low criteria, it's most likely not a member
and stage 2 processing unnecessary.
o 2) When first stage yields a number of feature matches between the high and low criteria, then
we execute 2nd stage processing. This is when feature weights come into play. The person
selects only those features of the test item and the target concept with high weights and
determines whether each highly weighted feature of the target matches such a feature of the
test item. If all such features match, the item is a member of the concept, otherwise it is not.
o Probability if second stage needed determines whether categorization is going to be rapid and
error free (based on stage 1) or slow and more prone to error based on stage 2.
Three Criticisms
1) Just listing features to describe something does not seem to go far enough, doesn't give us enough
information. Doesn't specify kind of knowledge we have about concepts. People know about
relationships between features and the variability that is permissible within that concept. Needs a
way to incorporate that information!
2) What features could be plausible? What's allowed and what isn't? Someone’s feature may not be
someone else’s. We may have pseudo features. Features should have some degree of generality.
There are no constraints on what can be listed as features. We know stability exists
3) Has little if anything to say about context. (The prime might be concept, so spreading activation
hints at context) There's not enough information given by discrete features.
Assumptions
Assumption 1) representation of a concept as a summary description that applies to all instances
Assumption 2)
a) any dimension used to represent a concept must be a salient one with at least some of its values having a
high probability of occurring in instances of the concept.
b) the value of a dimension represented in a concept is the subjective (an interpretation) average of the values
of the concept's subsets or instances on that dimension
Non necessary dimensions can be included like non necessary features. Weights are also used, and each
weight indicates the importance of variations in the associated dimension of concept membership.
Ex) Robins have three dimensions: animacy (1), size (.7), verocity (.4) The higher weight for the size
dimension than verocity means that I am more likely to classify a specific item as a robin/not a robin if it has
an inappropriate size than an inappropriate verocity value. Higher weights help make classification decisions.
This approach represents each physically represented dimension as a psychologically continuous dimension.
Therefore the difference between two concepts or between a concept and a specific instance is a matter of
continuous degrees. So a concept learner can combine the values of various instances or subsets by taking
their mean on each dimension. So a concept representation depicts the average or mean dimension values
for the entire class. (contrasts with featural approach... where each concept depicts modal features) With
means we can have something that is a member of the concept that looks like nothing that we see every day.
An instance with novel values on these dimensions would be totally dissimilar to the concept that contains the
modal properties. If made up of average rather than model properties could be similar to an instance that has
completely novel properties if matches the average. So with average, it MIGHT exist, with MODAL, maybe
not.
Assumption 3) Concepts with same relevant dimensions can be represented as points in a multidimensional
metric space
Relationship between any two concepts or instances of concepts can be judged based solely on distance estimates:
distances between corresponding pairs of items.
Categorization models can be based on distance computations rather than some sort of assumption of probability.
(no need to add up weights like in probability computations)
1) Simple Distance Model: relation between any pair of concepts or its instances is given by the distance between
points and a multidimensional representation.
FIGURE
* Circles are key. r = radius. r is the threshold distance. Circles designate one more assumption. Say that any
entity/instance is going to be categorized as belonging to a concept or being a subset of a concept iff the metric
distance between that thing and the concept is less than some threshold difference.
Assumption 4) AKA anything that fall in the threshold difference is in the category.
So categorization process doesn't need to explicitly consider dimensional values. It is the location that is key. All
that needs explicit designation is what points represent what. there is no decomposition of concepts into features.
ASSUMPTION 5) The closer item x is to concept Y the faster and more accurately X will be categorized as a member
of Y. (takes care of typicality effects)
How does this simple model deal with disjunctive concepts. (everything has relevant dimensions, doesn’t matter if
is large or small) don’t need same dimensional values as long as we're in a relevant dimension
• How does this model explain unclear cases? They are on the Limit (border) of the distance threshold
• What about failure to specify defining features: We don't need to because we only care about distances
• Determinants of typicality: shorter distance = more typical. Items used to define dimensions seem to
represent features in case of artificial concepts.
The Comparative Distance Model: Simple distance model doesn't consider influence of contrasting concepts. We
need a model that compares distance with test item (target concept) and the distance between the test item and
any item that might contrast with the target.
Assumption 1: An entity, x, is categorized as an instance or a subset of concept Y iff the metric distance between x
and Y is less than that between X and any concept that contrasts with Y.
Assumption 2: The greater the difference in the distance between X and Y on the one hand, and x in any contrast
of Y on the other hand, the faster and more accurately x will be categorized as a member of Y.
Russell, on aggression?
2) Basis of abstracted representation: one uses dimensional value, other uses modal feature
3) Processing strategy is different: one is comparison of distance computation, other is sum of weights
3) Both approaches allow for degrees of disjunctiveness by permitting different combinations of properties to
yield the same threshold quantity. weighted feature sum in featural approach and distances in dimensional
approach
4) Dealing with unclear cases - construe unclear cases very similarly. These are items that do not quite reach the
threshold quantity or are equally close to thresholds of one or more concepts. could be weighted feature sum or
distances
5) Explain many simple typicality affects by assuming that the typicality of a member reflects how similar its
properties are or how close it is to its parent concept.
6) Data on nested triples: Assume that properties of most concepts are more similar to those of the immediate
than the distant superordinate (what classical view says) Dimensional and featural can account for exceptions
because of idea of weighted sum. Chicken can reach necessary threshold for animal before bird.
7) Both can explain the use of contrast concepts by considering the relation between a test item and the target
concept and the test item and its contrast concept. Done implicitly in featural approach:
Given these, featural and dimensional can explain a lot of the same stuff.
1) Neither approach seem to represent all of knowledge contained in a concept, particularly relations between
concepts
2) Very few true constraints on these models, who is to say what are correct features/dimensions
3) Neither deals with context effects at all, spreading activation hints at it, but none of them talk about it explicitly
4) If you make the metric assumption (space is measurable) then sometimes the aspects of true metric space
(three things) don't necessarily coincide with what appears to be subjective aspects of space. To have a metric
space is an objective measurement. Cant assume this holds in subjective world of brain.
• Ex: Taversky demonstrated that the probability of judging two identical objects as the same
rather than different isn't perfect. As complexity increases, harder to identify object as one that I saw.
• Symmetry: invariably north Korea will be rated more similar to something like red china if given
in that order, ratings will be different depending on order with which receive information. Same distance
between two points, but symmetry doesn't hold. These items bring something different to mind
depending on order given.
• What about triangular inequality? For ex: Jamaica is always rated very similar to Cuba. Cuba is rated
similar to North Korea. Jamaica is never rated as similar to North Korea. So sum of other two doesn't
equal sum of other two because of CONTEXT, way that subjective mind sets it up.
Exemplar View
• Rationale: Concepts are represented by their exemplars, at least in part. Rather than by an abstract
summary.
• If concepts are represented by their exemplars, is there room for abstraction at all? Often term exemplar
is used ambiguously
Some exemplar view models do include a summary representation but the key is the exemplar plays the dominate
role because they are readily accessible, more so than any summary information
1) Kahnamen and Tevrsky: People had to estimate relative frequencies of occurrences of particular classes of
events. How often get hit by car. When people are asked to do this, they consistently retrieve one or a small set
of exemplars of the relative class and base their estimate on that example. (Think of someone getting hit by car,
specific instances, and base judgment on that)
2) Ask people if more words with k as first letter or third letter... THIRD! We think of examples, the ease which it
comes to mind. This is an exemplar.
3) Holyoak and Glass: demonstrated that participants in experiments often decide a test item is not an instance of
the target category by retrieving a counter example. T/F All birds are eagles. Think of a robin!
CRITICAL ASSUMPTIONS
1) representation of a concept consists of separate descriptions of some of its exemplars either instances or
subsets.
2) Exemplars themselves can be represented in different ways. it partly depends on whether the exemplars are
subsets or specific instances. If is a subset, representation can consist of instances of specific bird, relative
properties, descriptions of features. If it is a specific instance, then it is going to be represented by property
description
3) If the exemplar is an instance, it must be represented by the property description. representation is explicitly
disjunctive, and the properties of a concept are the sum of the exemplars properties. in direct contrast with idea
of summary description:
contains instances
contains subsets whose properties overlap enough to permit amalgamation:
Exemplar based representations show a substantially lack of abstraction than other models
So having specific instances is not criterion. In some models, different test items access different exemplars. On
the other hand, in some models the test item causes you to elicit all exemplars. Varies across models
EXTREME CASES
Proximity Model: This model violates all three of the summary representation assumptions.
OCTOBER 16, 2006 THE PROXIMITY MODEL
Medin: each concept is represented by all of its instances, all of its instances that I have ever encountered. When a
test item is presented and I have to decide if its a member, the test item automatically causes the retrieval of an
item from memory. It will be categorized as an instance of the category if the thing that it brought to mind is a
member of the category.
• The concept representation is lacking in abstraction, what comes to mind is a specific exemplar that exists
in the world.
• Every exemplar in the representation is realizable as an instance
• the information that you retrieve when making a decision about a particular concept membership varies
with the test item presented
PROBLEMS
• Model leaves no room for abstraction
• Drawback: sheer amount of stuff that I have to store! having to index through it all the time would slow
us down, this doesn't seem logical, saying that we are cognitively lazy. Key: there must be some way of
restricting exemplars that come to mind.
- Context Model
o Exemplar based model
o Abstraction based on context
o Differences from the exemplar model
1) deals with learning of exemplar model
2) deals with computation of similarity in the categorization process
o Case 1 - Experiment in artificial concepts
Subjects learn to classify schematic faces
• Choose category A or B
• Match distribution of facial features abstractly
Relative features treated as dimensions
• Eye height, Eye separation, Nose length, Mouth height
• Binary values – matrices
Strategy dependent choices – based on context in which you are viewing the
information
IF you attend to everything you see equally – you see the array as is in its entirety
If you attend selectively – in some cases some instances are gotten rid of
Can go from all instances to a particular feature of an instance
o Gives this model a way to restrict the representation of the concept to a very limited set of
exemplars
o Learner primarily attends to frequently
o Detailed representation of typical exemplars – contains focused on property
o Deprived representation of atypical members
o Best ex model – assumption that best examples dominated category learning
o Similarity computation: computed differences using additive computation. Is a computation of
the similarity between test instances and exemplars. It involved multiplying along component
dimensions. There is a similarity parameter.
High values = high similarity
between test instances and exemplars involves multiplying differences along
component dimensions
Association w/ each dimensional value is a similarity parameter. High=high similarity,
LOW=low similarity
Determinants of value of similarity between properties
• 1) Psychophysical difference between two value dimensions- actual physical
difference vs. perceived difference
• 2) salience of dimension: is it important?
• Given a fixed set of properties, multiply parameters to determine similarity. If
values difference = 0, then similarity is 1. Similarity computations are
computed and derive some value. Based on Luce’s decision rule, in or out of
concept membership. Multiplicative idea is why some instances learned faster
than others.
• Multiply values together accumulate derive final value
Models allows for correlations between properties to be computed from stored values
=> saves time
Lack of restraints is a problem.
Context effects – does well. Summary info: “bird lays eggs” is so far best model, allows
to attend certain information
Successes
o Why people various objects as equivalent
o How people use some properties to infer others based on previous knowledge
Problems
o Existence of disjunctive concepts (has one or more set of instances that share no features)
o Question of what is defining – can be non-necessary – “just because we can’t access doesn’t
mean they don’t exist”
o Unclear cases nesting
Pattern based on nesting doesn’t hold
Occurrence of typicality effects – there shouldn’t be graded membership
Family resemblance
o Members sometimes graded as more similar to distant as opposed to more immediate
superordinate category
- Salvage
o Creating links between concept
o Adding in translation process – translate between defining features and other features
o Allowing identification process to be used more accessible as mediating cores, non-necessary
features used because more accessible.
Liam Jaynes: Consciousness does not appear to be chopped up into bits. Consciousness flows.
Let’s call it the stream of thought/life
Aka. Consciousness is continually flowing and changing
The difference between what others see of us and what we perceive of ourselves. We have our own inner
awareness, and it’s different from this awareness of others.
The idea of self didn’t come about until 2000 years ago, looked at writings, no mention of I/me
*self awareness is key, and how it’s different from outside awareness
All metaphors are not explanatory in any way, just trying to describe what people perceive consciousness to be.
Questions:
Senescent MEAT!
1) a simplification function: helps us choose what organism does at any given moment
2) a guiding and overseeing actions function: at any moment, the content of consciousness is what we
are prepared to act on next
3) setting priorities for action: reflects internal means, survival usually wins
4) detecting or resolving discrepancies (HR, breathing)
3) Consciousness is the result of learning. This view holds that consciousness began after life had
already evolved. Beginning of consciousness equated with appearance of associate learning. If
animal can modify its behavior based on an experience, it must be aware of that experience,
conscious of that experience. So to study evolution of consciousness, must study learning. Grew out
of associatism. Interprets higher mental processes as resulting from combinations of sensory and/or
mental elements. Associations were formed based on three principles:
a. Contiguity( Things that happen close enough in time together)
b. Contingency (one thing results because of something else happening)
c. Similarity (similar things associated together)
Particle Physics: solid matter being solved as mathematical relations between particles.
So simply being conscious of something does not correlate with learning something, and learning something does
not mean we are conscious of it.
So conscious memory is not the storing of sensory images, we have to make an effort to encode
Mr. Wallace: “Man’s conscious faculties could not possibly have been developed by means of the same laws to
determine the progressive development of the organic world in general”
Proposed that some metaphysical force directed evolution at three critical points in man’s evolution:
1) Beginning of life
2) Beginning of consciousness
3) Beginning of civilized culture
This view steps outside bounds of natural science. Scientists didn’t buy in.
Materialistic view: holds that matter is the only reality. All objects, events, including psychological processes, can
be explained by the laws of matter. (reductionist approach) If we can figure out what parts of brain light up, then
we can explain the phenomenon. We can reduce it down to the mere matter, can be explained by reducing it
down. Assumes that consciousness does nothing at all, and cannot do anything, it is simply along for the ride.
Animals evolved, nervous systems increased in complexity, when some level of complexity is reached,
consciousness is there. What we do, see, think, etc is completely determined by wiring diagram of brain, and that
diagram interacting with environment. Consciousness is nothing more than the heat given off by the wires.
Hodgson: “Conscious feelings are mere colors laid on the surface of a mosaic, the mosaic is held together by its
stones, not the colors. We can see the colors, but they have nothing to do with it.
Consciousness cannot modify the working mechanisms of the body or its behavior.
A whistle can’t determine where the train goes, only the track does!
PROBLEMS:
5) Emergent Evolution
Something that is unexpected when you combine other things, and is usually greater than the sum of the
parts. (H2O)
So they would argue this is what consciousness is! Just as the property of wetness cannot be derived from the
properties of hydrogen and oxygen alone, so consciousness emerged at some point in a very derivable form from
its constituent parts. Evolution took pieces and put them together, and consciousness emerged. So there is
always the possibility for new emergence to happen.
This is anti-reductionist, we don’t have to go down to the molecule/atom to figure it out. There’s a brain, and this
other thing called consciousness. Most people LOVE this,
• seen as a liberation from physics and chemistry
• placed consciousness as the governor of behavior, and promised the possibility of newer and greater
things.
Questions
1) when did it happen?
2) In what species did it happen?
3) If it does exist, can we build it?
TAKEOVER! REVOLT! WAR! REVOLUTION! GUERRA!
Scientifically rigorous, experimentally based campaign to take over psychology. You could not do psychology if you
didn’t do behaviorism… behaviorists took over. To solve the problem of consciousness and its place in nature is to
deny that it exists at all.
The subject matter of behaviorism is overt, observable, and measurable behavior. That’s all there is, that’s all we
need to pay attention to! THERE IS NO CONSCIOUSNESS! This behavior has meaning in and of itself. It isn’t just a
manifestation of some mental event, it has meaning itself. We don’t have to interpret it in the brain.
John Brodus Watson: wrote a book. Couldn’t do anything unless pulled behaviorist line. Felt that study of some
inner consciousness was wrong direction for psychologists to go. Not making progress… introspection sucks.
HE SAVED PSYCHOLOGY
• Considered the notion of some interior mental state as a pseudo problem of science… something that was
made up by scientists so they would have something to argue about.
• True behaviorism was born. First similar to hapless spectator, consciousness was there but you couldn’t
study it.
• The defense of this view came the change that consciousness was nothing at all.
• Psychology and philosophy were entwined, in 60’s psychology started to separate with cognitive psych.
For first time, psychology is a science
• Behaviorism was a science born out of a method, and not a theory.
November 6, 2006
The study of behavior of present day species corresponding to various stages of development of these physical
brain structures should be able to reveal to us what consciousness is.
As complexity increases, the level of consciousness increases
This is one of the evolutionarily oldest parts of the brain, so the complexity argument doesn’t work well…
consciousness should maybe be in the cortex. If it is in this part of the brain, every species should have it!
We know which parts process what and when, but we can’t equate that with phenomenon
Why are things interpreted differently by different people?
What Consciousness ISN’T
• How extensive is consciousness? Is it everywhere? What does it mean to lose consciousness? They
haven’t lost consciousness, just reactivity. We are constantly reacting to things without being aware that
we are reacting.
• We react to things that we can’t be conscious of. When we see something with our eyes, the image is
bouncing around, but we see a stable image, because our brain stabilized it.
• Consciousness is the sum total of the mental processes occurring now. It is constantly moving forward. It
is probably a much smaller part of our mental life than we are conscious of because we cannot be
conscious of what we are not conscious of. The flashlight would be conscious of being on only when it’s
on. It doesn’t know it’s off when it’s off.
• Consciousness knits itself over time gaps (like gap in vision)
• Consciousness is not necessary in certain circumstances (knowing how I’m sitting) so this is non-conscious
(the definition of an automatic process)
• Consciousness functions in a decision as to what and or how and when to say something but then the rest
of it is done for us
• Consciousness isn’t as pervasive as people think, consciousness is not a copy of experience. Many have
emphasized the recording function of consciousness. The chief function of consciousness is to store up
experience and copy it like a camera so it can be used and reflected on in the future.
• Distinction between recognition and recall.
o Recognition: recognize something. Recognition memory is good.
o Recall: have to dredge something up out of nothing
• Conscious retrospection like this is not the retrieval of images. It is retrieval of what we have been
conscious of before and the reworking of those elements into an actual, plausible pattern. (so what I
remember it to be may not be what it was)
• When is the last time I actually saw myself swimming? NEVER! I was making it up, it’s not a valid
memory. This is why eyewitness testimony is the worst evidence ever – memory is totally reconstructed.
• Consciousness is not necessary for learning or for concepts.
o We form concepts based on experiences.
1) Learning of signals
2) Learning of skills
3) Learning of solutions
SIGNAL LEARNING:
(classical or pavlovian condition) Apparatus shoots puff of air in eye. Combined with light.
Light signal = conditioned stimulus (immediately followed by puff of air)
Puff of air = unconditioned stimulus
Consciousness has gotten in the way! We disrupted the natural process, and the learning was blocked.
SKILL LEARNING:
consciousness takes on role of hapless spectator, has very little to do. It directs us at the outset, takes us to the
task, gives us the goal we want to achieve, but from that point on, it has nothing to do with the action task itself
• (we threw coins in air, we’re conscious of goal, throwing in air, but once in air we just react!)
o Negative Practice:
1) Had one group practicing typing the the. The group that practiced the made fewer the
mistakes than the group that practiced the
Solutions: (how can we figure out how to make something happen more?)
• We couldn’t solve problem without being aware of solutions
• In animal learning this is called operant
Postman: Had participant sit across from partner, and participant asked to say any word that came to mind, but
asked to pause two to three seconds between each word. Partner was allegedly writing these words down, but
was supposed to respond to a certain type of word (an adjective, for example). Smile! Within ten minutes the
frequency of the responded to word increased two to three fold without the subject knowing they were saying
more of the type of word. Within 20 minutes, the participant would catch on. Consciousness broke the spell!
Key: Subject is not conscious that he or she is learning
Faculty Psychologists: Wanted to prove that reasoning is proof of consciousness, if we reason then we have logical
thought. Making up something to explain something don’t know what it is.
• The faculty or ability to reason is situated in consciousness: logic was meant to be the structure of the
conscious reasoning.
• Logic is how we ought to think if objective truth is our goal
• Logic is a science of justification of conclusions that we have reached by natural reasoning
• For natural thought process to take place consciousness is not necessary. The only reason we need logic
is because most reasoning is not conscious, so we have to come in after the fact and justify our thinking
with logic. (come up with logical theorem to demonstrate how/why we came to that conclusion)
Reasoning From Particulars (an expectation based on a generalization): There is no need for the conscious
collected together of past experiences in any conscious way to make a decision.
Consciousness has no location except for the one that we make up.
Is consciousness necessary?
1) Consciousness is not reactivity. We can react to the world w/o being conscious
2) consciousness is not involved in a host of perceptual phenomenon
3) it is not involved with the performance of skills and often hinders their execution
4) it need not be involved in speaking writing reading or listening
5) It does not copy down experience
6) It is not involved in signal learning and doesn’t have to be involved in the learning of skills or solutions
7) It is not necessary for making judgments or in simple thinking
8) It is not the seat of reason
9) It is not necessary for creativity
10) It has only an imaginary location
It would be entirely possible to have a race of beings that did everything without being conscious at all.
NON-CONSCIOUSNESS:
We know more about it than conscious processing
• Unconscious mental activity is that activity which is totally inaccessible to phenomenal awareness under
any circumstance. We are aware of product produced by mental state or activity but we are not aware of
the mental state or activity itself. This is referred to as…
• Procedural Knowledge: an event, a sequence of events or actions, that occurs so automatically or
effortlessly as to operate outside of awareness. (language, visual pattern recognition)
• Louicki talked about our language and how there are certain things that sound good. “The big red barn”
vs. the “red big barn”
• Foder: proposed that mind has set of innate cognitive modules that control these types of procedural
activities. Procedural knowledge can result from practice (typing, language, athletes, etc) These
processes are referred to as automatic vs. effortful or controlled
• Automatic:
o Has to happen in response to a given stimulus regardless of intent
o If something is automatic it will use up little or no attentional resources
o Automatic non conscious activity can be detrimental because you are not paying attention
o Spelkey had participants read unfamiliar prose and took dictation at the same time. Wrote down
words while reading – terrible at both. After six weeks of practice, they got to 90% accuracy at
dictation and passed a comprehension exam with at least 80%. The participants could not
remember words they dictated or any structure in list. Dictation was proceduralized
o This phenomenon is apparent in social interactions. I may like one face better than another but
be unable to tell why (unfamiliarity).
o Preconsciousness works on declarative memory (deciding what to do next) Declarative
knowledge is accessible, we can think about it, can be learned in as little as one try.
Procedure: Shown several word lists, told had to remember, and then given free recall or recognition.
Results: Amnesiacs performed significantly worse than normal controls because using explicit memory, had to go
back and recollect seeing words. In some of the trials, used
word fragment completion task - R A I N B OW (give stimulus, and had to complete word) Difficult even
for normals!
word stem completion task: - gave participants first three letters of word and had to complete rest. C H A
__
Both of these are implicit tasks, indirect measures of memory. The participant doesn’t have to recollect learning
episode with this test. The amnesiacs performed just as well as normal on both tasks.
HM: Epileptic that had debilitating seizures, so they removed most of hippocampus and amygdale, removed 40%
of brain. Had anteriograde amnesia. Doctor would re-introduce himself one day, put buzzer in hand, and after a
time HM didn’t want to shake his hand. He couldn’t remember why he didn’t want to shake his hand, but the
information was in.
Give Tower of Hanoi/ puzzles to amnesic, will eventually figure it out. Bring it back to them the next day, and they
figure it out even quicker, even though they can’t remember doing it.
POINT: Information is getting in, implicitly! Normals also retain information without conscious awareness.
• People can remember and learn things when they are fully anesthetized.
• Primary purpose of anesthesia is to make patient amnesic
• Fat Lady Syndrome: there are a lot of comments made during surgery during crew. Sometimes they say
mean things. Most of the time the people don’t realize that the patient may remember/hear what is
going on.
o Bennett: lawsuit filed by woman during surgery because he repeatedly referred to her as a
beached whale. She developed post operative complications. 3 days after, she remembered,
then she started getting better. (not sure if explicit or not)
1) The auditory functions are the last to go when you’re anesthetized
2) If patients can hear during surgery, think about implications when something goes
wrong.
• Levenson followed up on this, fabricated fake crisis during surgery. Had ten
dental patients anesthetized for surgery
o During surgery, anesthesiologist said “Stop the operation, I don’t like
the patients color, needs oxygen”
o Month later, brought back and hypnotized, 4/10 recalled words
verbatim. 4 others showed elevated anxiety
• Beth Loftus, work on false memory: tested recognition memory for list of 100 unrelated words, 28, 53
hours after having abdominal surgery. Performed no better than chance. It was an explicit task.
(Recognition and recall are explicit)
• Then she did a word completion task, and remember significant amount
Therefore, the person is still conscious! Reactivity is gone, but consciousness persists. So awareness is not
responsible for many things we do. We navigate world in way that is tantamount to being oblivious to what is
going on – do it for preservation/efficiency
We can only take in so much information at any time
• only see certain WL light
• only hear certain WL sound
• limited STM
• limited iconic memory
Second more flexible portion of mental system, might be consciousness. Allows us to shut off a lot of the world.
Serves as a gate for percepts that remain. Not a reasoned instrument, doesn’t plan, lives in moment.
There are cultural and language assumptions (gravity) that allow us to have a stable world around us. So how is
anything chosen to enter consciousness, if we are oblivious to most?
We have organized world in a particular way so we don’t have to think about things (months, alphabet,
etc)
Consciousness is the front page of the mind, which is like a newspaper. What’s important that day is on the front
page. Countless ordinary things are buried back on the last pages. They never make the headlines. The
unexpected makes the front page, and things of highest priority that must be attended to.
Today’s Theories
• Dreams can be very important to some people, life can be based upon them
• Jung felt that dreams aren’t garbled at all. They made the meaning as clear as possible. There is a lot of
debate about who’s right and who’s wrong!
• Early 50’s two French Michelle’s were studying the electric activity of cat brains. They discovered that the
activity of a sleeping cat fluctuates from quiet to active (as active as an awake cat). They could not think
of any explanation for why this could happen –
o A few years later Nathanial Clighton and Eugene A. discovered that people move their eyes is
rapid spasms during sleep. (REM sleep born)
1) The brain is as active when you are asleep as when you are awake!
2) Posterior part, if active, then there is sensory stimulation taking place
3) If anterior part is active, then there is information being sent out to motor neurons.
• Locus Cerulus: just before brain wakes up at night, this disconnects the motor and the sensory parts of the
brain. So dreaming is a result of some interior performance. This thing prevents us from standing up and
walking around
o We see our dreams for the same reason we see when we are awake – is this consciousness?
o If you cut the connection, then the cat will get up and walk around
o REM sleep led to physical theory of dreams
Cycle of Dreams
• 1-2 dreams per REM cycle
• 3-6 REM cycles per night
• Most dreams last from a few seconds to not much longer than 15 minutes. In dreams we lose sense of
time, could be faster or slower
o Dreams we Remember: Physicalist says it’s a fluke that we remember anything. It seems that
dreams sometimes tell us things.
o Hobson done several studies about looking at if we dream in cognitive styles. Is there meaning,
or is it random? Even if its random, we should have different random things going on in our
heads depending on our lives.
1) Looked at people who were nervous, tedious, anal, etc. Demonstrated that meaning
people found in their dreams varied depending on cognitive style
o Compared dreams of science and artist students. Students in creative arts had more vivid,
bizarre, and aggressive dreams. They also had more philosophical and religious themes. The
more meaning we perceive in the waking world, the more meaning in the dreaming world?
o Dreams have little if any inherent meaning. Images, objects, characteristics, settings, is a product
of semi random neuronal firing (because of connections)
• Stickgold at Harvard says “The mind becomes clinically insane for about 2 hrs every night. We hallucinate
wildly, see things that aren’t there, we basically become delusional.”
1) Proposed a new model of dreaming. Dreaming is a bottom up process. It is driven by
the stimulus rather than by any conceptual thinking.
2) Pons (means bridge) contains a portion of the reticular formation, that works in sleep,
muscle tonus, etc. These are nuclei that are very important to sleep and arousal. Also
has nuclei that project to cerebellum. One of the major portions of the pons is
• FTG (gigantic cellular field of the tegmentum) becomes very active during REM
sleep. Just before REM cycle begins, FTG kicks in and sends bursts of electrical
activity throughout brain – dreaming begins!
o Emotional state is the first thing that is attached to dream. It is made
available because FTG is stimulating amygdala (first).
o If the emotion assigned to dreams by the amygdale truly affects
everyday occurrences, then we should see specific gender differences
in dreams emotions (because men and women deal with emotions
differently)
Steve Foote: During REM sleep there is a low level chaos that occurs in the cortex. Neurons that receive signal to
fire often don’t, and others fire for no reason at all. There is a mismatch of firing and emotions, and the brain tries
to ascribe meaning to it.
What is difference between transformations and plot line shifts? Started waking people up during REM sleep, and
tested them. Tested ability to make associations between words.
• When woken up, people made strong associations better than normal with awake patients.
• With weak associates they were terrible
o Speculate that transformations are very constrained and reflect high associates.
o Plot shifts are far less constrained, can be anything
1) There is some method to the madness in dreams
Neurons responsible for things in our dreams were probably more active during the day.
• They are primed
• Takes less activation for them to get involved during sleep
• Dreams reflect personal history (memories and experiences)
• The problem comes when we try to impose an elaborate or symbolic interpretation onto dream.
• Meaning is constructed by waking mind, not the dreaming mind
Stephen King: I think that a lot of times dreams are nothing more than a mental or spiritual flatulence.
Why do we dream?
Rating Creativity
4 Shakespeare
6 Bach
5 Picasso
3 Eric Clapton
1 Dave Matthews
7 Issac Newton
7 Darwin
1 Coco Chanel
2 Michael Jordan
-2 Eminen
5 The Vétales
5 Marx
1 Usher
Poets, authors, scientists, singers, creativity is everywhere, but it is in the eye of the beholder.
Improvising Jazz, fixing broken engine, using sports skills, all things that other people see as creative. Creativity
might be correlated with skill. A certain level of skill might be necessary, but you don’t have to be an expert.
Dictionary:
Human Creativity
• can be occasional, sustained, psychology should be able to explain something so obvious and familiar.
• Creativity is unpredictable. It happens when we least expect it.
• Main goal of science is to be able to predict things
Question:
How novel does a novel thing have to be to count as creativity?
There is novelty in randomness, so is randomness creative
Every individual can think things that are novel with respect to their own personal thoughts
There is novelty in madness, unpredictability.
What is the distinction between madness and creativity?
You can have ideas/thoughts that as far as you know, no one has ever had.
It isn’t straightforward to say something is creative and something isn’t
INSPIRATIONAL APPROACH
• Creativity is mysterious, superhuman, or divine.
• Plato “A poet is holy and never able to compose until he has become inspired…”
• If creativity is the result of some unasked for, divine intervention, then the scientific study will be very
difficult, and you cannot teach it. You either have it or you don’t
ROMANTIC VIEW
• Less extreme. Creativity is not divine, but it is the exception rather than the rule. Creative people are
thought of as gifted with a specific talent that others lack. (intuition, insight, that normal people don’t
have)
• Creativity is fundamentally unanalyzable. You are born with intuitive, innate talent. The gift can be
squandered, but you cannot acquire it or be taught it. The most you can do to encourage creativity is to
identify people with this special talent and put them in an environment that will allow them to express it.
• Kessler: says this view is useless. Developed idea of
o biosociation of matrices: the juxtaposition or mingling of formerly unrelated ideas
1) “The moment of truth, the sudden emergence of a new insight, is an active intuition.
Such intuitions give the appearance of flashes or short circuits of reasoning… Like a
chain… so we have the beginning and end and we don’t know what goes in the middle.
• Cognitive science accused of ignoring social aspects of things it studies
o Creativity is one area where social creations impinge upon concerns of cognitive science… so we
have to take social conditions into account!
o How creative something is depends on two things:
1) The mental processes of the individual who produces the solution
2) The contextual factors surrounding the problem, person, etc.
• Some of these factors have little to do with actual thought process
• It could be argued that thought processes underlying creative thought aren’t
really different from other kinds of thinking.
• Once a creative solution has been found to a particular problem, it can and often is used as a model for
solutions to other problems, some similar and some not. These solutions are not considered as creative
as the one they are modeled after.
• Creative solutions to problems are the ones that are difficult to find in a particular circumstance, not
because they require unusual thought processes.
o They may require synthesis of large amount of information (expertise)
o May depend on connecting bodies of knowledge that weren’t connected before
o Many believe creativity can be boiled down to four foci:
1) Creative People:
2) The creative process
3) The products of creativity
4) The creative places (where creation takes place) and persuasive power of the creative
person
Four P’s
PEOPLE
PROCESS
PRODUCT
PLACE/PERSUASIVENESS
Simonson: View based on observation that once creative ideas are accepted, they can have profound influence on
people. Become accepted because creator convinces other people of their importance.
Creative Solutions: topics suited to mathematics and science. This is a Western approach to the study of
creativity. The majority of what people consider art were created or produced to fulfill certain goals.
• Many works of art are commissioned, and their function is specified by person commissioning art
• The artist is presented with a problem for which he or she must produce a solution
o Satisfy goal of commission with constraints of art form
o The principle difference between arts and sciences is in what is produced and not process. The
medium in which the creator works is different
PSYCHOMETRIC APPROACH:
• attempts to predict creative achievement, so must concern self with products of creativity
• Question: What makes a person in a particular environment creative?
o There are close connections between ideas of creativity and expertise
o We generally assume two people have same level of expertise
• Creativity almost always depends on the sort of familiarity with a subject area that constitutes expertise
• Much of the thinking of experts goes beyond the powers of the novice is not creative in the sense we
usually think of the term.
• One psychometric study of creativity tried to answer systematically what are creative people like?
o Personal characteristics
o A psychometric study is anything that tries to get a profile by asking a series of questions.
1) In these studies, the products of creativity are only indirectly important.
2) They looked at products, things people thought were creative, found people that
creative them, and then tried to figure out what makes those individuals creative.
• Roe 1952: examined north American scientists who were deemed creative by their peers
o More often than not, first born sons of wasp parents
o Usually isolated at school
o If they got married they married late in life
o Worked long hours, 7 days a week with little time off
o Considered very intelligent, introverted
o Were very self sufficient
1) Minor differences between social, biological, scientists
• Social scientists slightly more extroverted, aggressive, and if married, more
prone to divorce.
Population is a reflection of the social times! Can’t deny impact of social context on creativity
A similar study done by Revdoll and Katell 12 years later: it showed the same profile emerge in a group of artists
and writers.
Artists had more inner tension
• Early 90’s: a wide variety of personality traits and motivational qualities common amongst creative
people. Almost everyone has something
o One of most important cognitive characteristics that distinguishes creative people from highly
intelligent people is the ability to find or select the appropriate problem – to recognize what you
can do well and DO IT!
o Another important trait is the ability to defer judgment on possible solutions under sufficient
evidence or information is gathered and to keep these possible alternatives available for as long
as you can. EASER SAID THAN DONE!
Personal Traits
• Desire for originality
• Failure to conform to social pressure (self-sufficiency)
• Ability to tolerate ambiguity (defer judgment)
• Distinguished by personal form of self government (prefer legislative or rule creating existence rather
than an executive (real following) or judicial (rule assessor)
• Directing intelligence is as important as the intelligence itself
o KNOW what you’re good at and DIRECT it to the right place
• High IQ is not in and of itself sufficient for creativity.
o If psychometric tests can be used to ID creative people, can a test be used to predict who will be
creative?
* Guilford: president of American Psychological Association: gave presidential address for starting work on
creativity. In context of his speech, talked about
general theory of intellect: Distinguished between two types of thinking:
Convergent and divergent productions.
• Convergent Thinking: Needed to answer most items on an IQ test. Thinker is expected to converge on
best answer to a problem.
• Divergent Thinking: lead to many possible solutions to same problem. Requires originality and flexibility.
Relation to creativity harder to establish.
o If this is important, our education system is whack, because its based on convergent thinking.
• Wallas: studied biographical information, and has suggested four stage account of creative thinking
(based on biographical info)
o 1) Preparation – absorb background information
o 2) Incubation – problem is set aside, and unconscious processing occurs. Wallas tested what
activities are best for incubation period? Period of mental rest with light physical exercise was
best. Reading is the WORST because we are focusing our attention down certain path and not
letting it bounce around.
o 3) Inspiration -
o 4) Verification
Torrance: developed tests that can correlate moderately well with latent measures of creativity
• Torrance test of creative thinking
December 1, 2006
Stages
• not rigid, can shift from one stage to another
• gives a limited capacity for prediction (can’t say one person will move from one stage to another with any
degree of reliability.
Patrick
• asked people to think of poem in response to a picture
• asked then to talk throughout the process
• supported this progressive view:
o first gathered info, then inspired, and often revisited their ideas (close to incubation) provided
little evidence for nonconscious process -> all verbalized
o little difference in strategy for novices and expert and subjects, => airy? In quality of product
nd
2 study
• subjects given very difficult task
o one group worked for 20 minutes straight, another group allowed 5 minutes incubation period
• high skill subjects did better in 20 minute block
• low performance subjects did better after 5 min incubation period
• for the HS subjects, all the interruption did was impair their processing –
disrupted their train of thought
Problems:
1) in most cases we have no idea how well this person would have done without the 5 minute period,
could be that these people were just better at problem solving.
2) Also, might not be completely out of consciousness
Latent Inhibition: what occurs in most people, an animal’s nonconscious ability to ignore stimuli that experience
has shown to be irrelevant to its needs (COGNITIVE PARSIMONY). Tends to decrease with age, why people can
lose their “train of thought” and are distracted.
• brains of creative people appear to be more open to incoming stimuli than noncreative people.
• Creative people can remain in contact with these stimuli/possibilities for longer periods of time -> an
extreme end of this skill is psychosis, hyperactivity/ADD
• -> but, can this ability to remain in contact with outside stimuli actually lead to higher IQ. Ability to
remain in touch with many ideas while not losing focus. (still an experimental hypothesis)
o Low latent inhibition may be good when combined with high IQ (ability to keep many items in
working memory)
o Also with this ability comes the ability to discriminate and evaluate the validity of ideas.
o Low levels of latent inhibition and low levels of controlled focused cognitive thought may lead to
psychosis
o Notion of gathering and combining information (regardless of whether it is conscious or not) is
related to the idea that novelty results from connection of dissimilar, not normally associated
ideas.
Mendrick
• any creative idea almost inevitably brings together two previously unrelated ideas
• this is not enough, any explanation of creativity also has to incorporate an evaluation of the products
validity or fullness of this combo.
o Personal creativity and social creativity (use fullness)
o Not only implausible that all possible combinations of ideas are tried, it may be impossible.
o Constraints arise because of the domain within which one is working, time with which one has,
environment, and physical condition of the individual (ie arousal, tiredness) have to have the
requisite knowledge to make these judgments on the produced ideas
Creative: something creative/high creativity leads to a product that is novel and valuable.
• can fail in creative process doing verification by
o a) det. Ideas work when they don’t
o b) def. ideas don’t work when they do
sometimes highly creative people don’t verify their results
Parking: points out that many of Mendrick’s accounts are biased by the people who wrote biased sample, with
distorted accounts.
Problem Finding:
• have to find the thing that matches your ability (remember people are limited in their area of creative
expertise)
• Example: Art students asked to select the rational they were to paint, people who tried, experimented
and handled more items were judged more creative by independent judges
• These students that had shown higher rating on the pics were shown to have better careers later.
? – How would you teach this differing judgment, this looking-through of possibilities before making a solution?
What about environment’s conducive to creativity?
• Anabile? : * some better than others: especially environment where individuals see themselves as
internally motivated rather than trying to fulfill some externality set goal (not always so. (example, Sistene
Chapel)
• Internal goals seem to motivate creativity more than external goals
• Corporate “think tanks”
Environmental factors are important in determining what is useful
• the status of a product as creative or not depends on environment within which it was creative
• CSIKSZENT-MIHALYI: argued that three main forces shape creativity
1) Creative individual
2) Social field determines which ideas are worth retaining
3) Stability of society over time to ensure idea continues to be considered creative.
CM: likened the work of the social field to that of natural selection (a Darwinian spin on creativity)
• Processes that underlie creative thought do not differ in any basic way from other types of thinking I do
during the day
• Rule and Simon came up with problem solving rule that people want to relate to creative thought
o Simon suggests that you can use idea of problem solving when it comes to machine creativity
o Pure problem solving approaches emphasize role of memory and domain specific knowledge
o 1987 Simon and colleagues tried to apply this to machines – came up with computer programs
1) BACON: Main BACON program derived quantitative laws (planetary motion) by
numerical induction of raw data.
• Feed it data, give it laws, and it would devise quantitative laws. Is that
creativity? NO!
2) DALTON: Modules thinking of chemist John Dalton. Attempted to capture qualitative
reasoning.
• Taught an early version of atomic theory of matter
• Then asked to work out structure for molecules based on information about
chemical reactions. Is that creativity? NO
Johnson Laird :
• agrees that social factors are important in defining the constraints on a particular genre of art.
• Argues that their effects must be mediated by mind of individual. If you make an identical computer,
constraints are interpreted in the same way.
• Possession of skill set is important, a necessity but it isn’t sufficient
• Computers aren’t creative, but forgery is?
• Given that it is possible to say something about mental processes that underlie creativity – talk about two
attempts to say what is going on in mind.
Johnson Laird
• 1) the results of a creative process must be new at least for the creative person while they are produced
from pre-existing elements
• 2) results must not be produced by
o recall from memory,
o rote computation, or
o any other simple deterministic process
(this rules out MACHINES!)
• 3) the results must satisfy a set of criteria (influence of context to society)
• Also says that process must be distinguished from induction.
o The act or process of drawing general principles from particular instances or facts
Induction just increases amount of information, doesn’t change way we look at information.
Creativity is non-deterministic?
• creative processes appeared non-deterministic because we don’t know what we’re looking at, we don’t
know how to recognize if something is creative or not, so sometimes we get lucky
• Johnson Laird attempted to come up with computational account of creativity, suggesting three possible
classes of algorithms that might be used:
1) Neo-Darwinian – combine old elements in a random way. Results are subjected to a
selection process for which only viable combinations are retained. Take many
iterations, and viable creative product emerges
2) Neo-mamarchian – initial combination stage is not random. It still makes use of old
elements, but there are constraints placed on what can be combined with what. Usually
produces several viable combinations. Then there’s a random selection process. If
constraints are tight enough, then we come up with one thing!
3) A mixture of los dos!
Bowdoin:
• an idea is created for a particular person if that person could not have that idea before. The only way is If
they don’t have the knowledge or requisite skill set.
• It is historically creative if NO ONE could have had the idea before
• Defines could not in terms of what person’s initial representation of the process allowed
• Mental restructuring can take various forms, but it must be defined in computational terms.
o A creative solution may require a complete restructuring or devising a new technique for the
search of the solution. (So come up with new space or new way to search space)
• There have to be constraints to distinguish creativity and eccentricity (where society comes in)
o These constraints assign degrees of creativity
o Most of these constraints arise from previous acts, solutions, etc that were previously called
creative
o Something cannot be judged creative except by people who know previous ways of doing things.
So creativity can only be judged in historical context.
o Truscott demonstrated that origins of symphony are last total reconstruction of music in a long
time. What is prized in music isn’t development of symphony, but what is produced by
symphony
o Creative works of art must also appeal to us in some sense… scientific theorem doesn’t have to
appeal though! In science it may be restructuring that is key:
Cognitive scientists focus on processes of the mind of individuals, but a total account of creativity has to take more
than thought process into account. You have to make reference to social factors. Many may not impinge on mind
of creative individual. Psychologists have taken many approaches to study of creativity:
• psychometric tests
• tests of divergent thinking (Guilford)
• Biographical and Autobiographical aspects or accounts of creativity
• Wallace IDd four stages creative process
o These stages have been reproduced in lab
• ideas drawn from Darwinian ideas of natural selection
• Laird uses evolutionary ideas too!
• Bowdoin suggested that creativity depends on having ideas you couldn’t have had before, give you have
machinery necessary for task and a historically educated audience