KNOWLEDGE step in a categoriza2on experiment is familiariza*on,
in which infants are exposed to a number of different
Back propaga2on - process by which learning can
examples within one category. For example, infants
occur in a connec2onist network, in which an error
might see a number of different kinds of cats, as
signal is transmi7ed backward through the network.
shown in ● Figure 9.28a, in which a number of pairs of
This backward transmi7ed error signal provides the
cats are presented to the infant. (In another set of
informa2on needed to adjust the weights in the
experiments it is determined that the infants can tell
network to achieve the correct output signal for a
the diff erence between the cats, so during
s2mulus.
familiariza2on they are seeing s2muli that look
Basic level - in Rosch’s categoriza2on scheme, the
different to them but share the characteris2cs that
level below the global (superordinate) level (e.g.,
make them “cats.”)
“table” or “chair” for the superordinate category
The second part of the experiment is the preference
“furniture”). According to Rosch, the basic level is
test, in which an example the infants have never seen
psychologically special because it is the level above
before from the familiarized category (cats, in our
which much informa2on is lost and below which li7le
example) is presented along with an example from
is gained.
another category, such as dogs
Categoriza2on - the process by which objects are
(Figure 9.28b). If the infant looks longer at the dog, it
placed in categories
is inferred that the infant has grouped the novel cat
Category-specific knowledge impairment - a result of
with all of the other cats (so it belongs to the category
brain damage in which the pa2ent has trouble
“cats”), but has not included the dog in that category
recognizing objects in a specific category
(the dog is in another category).
Cogni2ve economy – a feature of some seman2c
Family resemblance - in considering the process of
network models in which proper2es of a category that
categoriza2on, the idea that things in a par2cular
are shared by many members of a category are stored
category resemble each other in a number of ways.
at a higher level node in the network. For example,
This approach can be contrasted with the defini2onal
the property “can fly” would be stored at the node for
approach, which states that an object belongs to a
“bird” rather than at the node for “canary.”
category only when it meets a definite set of criteria.
Category - groups of objects that belong together
Global (superordinate) level - the highest level in
because they belong to the same class of objects,
Rosch’s categoriza2on scheme (e.g., “furniture”or
such as “houses,” “ furniture,” or “schools.”
“vehicles”)
Concept - mental representa2on used for a variety of
Specific (subordinate) level - the level in Rosch’s
cogni2ve func2ons, including memory, reasoning, and
categoriza2on scheme that is a level below the basic
using and understanding language. An example of a
level (e.g., “kitchen table” for the basic category
concept would be the way a person mentally
“table”)
represents “cat” or “house.”
Connec2onist network The type of network proposed
by the connec2onist approach to the representa2on
of concepts. Connec2onist networks are based on
neural networks, but are not necessarily iden2cal to
them. One of the key proper2es of a connec2onist
network is that a specific category is represented by
ac2vity that is distributed over many units in the
network. This contrasts with seman2c networks, in
which specific categories are represented at individual
nodes
Error signal - during learning in a connec2onist
network, the difference between the output signal
generated by a par2cular s2mulus and the output that
actually represents that s2mulus
Exemplar - in categoriza2on, members of a category
that a person has experienced in the past
Familiariza2on/novelty preference procedure -
makes use of the fact that when given a choice
between a familiar object and a novel one, infants
generally will look longer at the novel object. The first
in which infants are exposed to a number of different
Back propaga2on - process by which learning can
examples within one category. For example, infants
occur in a connec2onist network, in which an error
might see a number of different kinds of cats, as
signal is transmi7ed backward through the network.
shown in ● Figure 9.28a, in which a number of pairs of
This backward transmi7ed error signal provides the
cats are presented to the infant. (In another set of
informa2on needed to adjust the weights in the
experiments it is determined that the infants can tell
network to achieve the correct output signal for a
the diff erence between the cats, so during
s2mulus.
familiariza2on they are seeing s2muli that look
Basic level - in Rosch’s categoriza2on scheme, the
different to them but share the characteris2cs that
level below the global (superordinate) level (e.g.,
make them “cats.”)
“table” or “chair” for the superordinate category
The second part of the experiment is the preference
“furniture”). According to Rosch, the basic level is
test, in which an example the infants have never seen
psychologically special because it is the level above
before from the familiarized category (cats, in our
which much informa2on is lost and below which li7le
example) is presented along with an example from
is gained.
another category, such as dogs
Categoriza2on - the process by which objects are
(Figure 9.28b). If the infant looks longer at the dog, it
placed in categories
is inferred that the infant has grouped the novel cat
Category-specific knowledge impairment - a result of
with all of the other cats (so it belongs to the category
brain damage in which the pa2ent has trouble
“cats”), but has not included the dog in that category
recognizing objects in a specific category
(the dog is in another category).
Cogni2ve economy – a feature of some seman2c
Family resemblance - in considering the process of
network models in which proper2es of a category that
categoriza2on, the idea that things in a par2cular
are shared by many members of a category are stored
category resemble each other in a number of ways.
at a higher level node in the network. For example,
This approach can be contrasted with the defini2onal
the property “can fly” would be stored at the node for
approach, which states that an object belongs to a
“bird” rather than at the node for “canary.”
category only when it meets a definite set of criteria.
Category - groups of objects that belong together
Global (superordinate) level - the highest level in
because they belong to the same class of objects,
Rosch’s categoriza2on scheme (e.g., “furniture”or
such as “houses,” “ furniture,” or “schools.”
“vehicles”)
Concept - mental representa2on used for a variety of
Specific (subordinate) level - the level in Rosch’s
cogni2ve func2ons, including memory, reasoning, and
categoriza2on scheme that is a level below the basic
using and understanding language. An example of a
level (e.g., “kitchen table” for the basic category
concept would be the way a person mentally
“table”)
represents “cat” or “house.”
Connec2onist network The type of network proposed
by the connec2onist approach to the representa2on
of concepts. Connec2onist networks are based on
neural networks, but are not necessarily iden2cal to
them. One of the key proper2es of a connec2onist
network is that a specific category is represented by
ac2vity that is distributed over many units in the
network. This contrasts with seman2c networks, in
which specific categories are represented at individual
nodes
Error signal - during learning in a connec2onist
network, the difference between the output signal
generated by a par2cular s2mulus and the output that
actually represents that s2mulus
Exemplar - in categoriza2on, members of a category
that a person has experienced in the past
Familiariza2on/novelty preference procedure -
makes use of the fact that when given a choice
between a familiar object and a novel one, infants
generally will look longer at the novel object. The first
