Concepts and categorization: June 18, 2013
Flexibility across time and relativity if something is “new” or “old” example of car Or
whatever two concepts.
Concept: mental representation of some object
Category: class of similar things
*Functions of categorization: understand individual cases you have not seen before
and make inferences about them. It reduces the complexity of environment and
requires less learning/memorization. You know how to act to something by linking
the new thing to something you have already seen.
1. Psychological Essentialism: people act as if things have underlying natures that
make them what they are (ex. Gender) often these molecular makeup are not right
available, we use more superficial cues not the essential factors (ex. Molecular)
a. Nominal kind concepts: clear definitions of necessary and sufficient features
(triangle – has to have 3 sides, 3 points etc.)
b. Natural kind concepts: classes of entities that exist in nature (bird)
c. Artifact concepts: class of objects used for different tasks (toothbrush)
Classical view: category membership determined by a set of defining (necessary
and sufficient) properties. Need all those properties to fit into the category.
Concepts are not representations of specific examples but a list of characteristics.
Membership in the category is clear cut, either in or out. All or none.
Problems with this: What is necessary and sufficient? Ex. Pope vs. bachelor. How
many features must there be? No defining features for many categories (eg. Games-
is poker a game of a business?) Also there is a lot of feature overlap
Typicality (a graded membership view) people judge members of a category as
differing in goodness.
Prototype view: An idealized representation of a class of objects. Include features
that are typical rather than necessary or sufficient. Combine previous into a
Levels of abstraction:
1. Superordinate (animal) allows grouping of similar entities
2. Basic (dog) compromise better super and sub
3. Subordinate (German shepherd) allows distinction between entities
Perceiving vs. constructing categories: rosch would argue main task in
categorizing is perceiving and knowing regularities in environment not arbitrarily
grouping clusters of attributes.
Determinants of typicality: prototypes serve as a reference point. “Smith brothers”
example, If they have light hair, bushy mustache, large ears and glasses then they
are the prototype but if they don’t have all of them than they are not the “ideal”.
Overlapping features predicts typicality therefore the more features the brothers
share the more likely you’re going to consider that they are brothers. THE MORE
FEATURES THEY SHARE WITH THE PROTOYPE THE MORE LIKELY THEY ARE
CONSIDERED A GOOD EXAMPLE. Problems with prototypes: limits on categories, what would the reference
point be in vast categories for example sports? Typicality and context is not fixed.
The exemplar view: Have stored previous instances and compare directly to
the stimulus you’re deciding on.
Concepts are composed of previous instances, categorization occurs by comparing
current instance with precious instances stored in memory.
Study: in training present with: builder and digger. If the creature has 2 of 3 features
then it is a builder (long legs, angular body and spots). During the test since they see
the digger “similar” shape has long legs and spots they didn’t choose it ven if it had
2/3. They chose the one that looked closer to the builder and didn’t have the body
shape of a digger. In the test they showed them 2 things that would qualify as a
builder one had (spots and long legs) and one had (angular body and long legs) they
store both examples given in the training and the digger interferes and rules out the
one with the body shape of digger, even though it does have 2/3 features. If they had
never seen the known digger in the training they would realize both qualify because
there is no interference.
If according to classical view and just going through ALL features they would know
both matched! That’s the difference between these 2 views.
Pros: discard less information than prototypes. Keep all examples to reference later.
Objects that are like more of the exemplars stored in memory are classified faster.
Cons: does not specify which exemplars will be used for categorization, requires
that we store lots of exemplars.