THE ORGANIZATION OF
KN WLEDGE
IN THE M ND
1
ICE BREAKER
2
WHAT IS THE STEP BY
STEP IN COOKING
CHICKEN ADOBO?
3
WHAT IS THE CAPITAL
OF THE PHILIPPINES?
4
DECLARATIVE VS
PROCEDURAL KNOWLEDGE
DECLARATIVE PROCEDURAL
Ddeclarative knowledge that can be expressed in This is knowledge about how to follow procedural
words and other symbols (i.e., “knowing that”) steps for performing actions (i.e., “knowing how”)
Declarative knowledge is about knowing facts or Procedural knowledge is about knowing how to do
information, something
Piece of information that a person is aware of Being aware of the steps needed to complete a
knowing task or a job
Knowing what Knowing how
5
ORGANIZATION OF
DECLARATIVE KNOWLEDGE
The fundamental unit of symbolic knowledge (knowledge of
correspondence between symbols and their meaning, for
example, that the symbol “3” means three) is the concept—
an idea about something that provides a means of
understanding the world
(Bruner, Goodnow, & Austin, 1956; Kruschke, 2003; Love, 2003).
6
ORGANIZATION OF
DECLARATIVE KNOWLEDGE
A category is a group of items into which different objects or
concepts can be placed that belong together because they
share some common features, or because they are all
similar to a certain prototype. For example, the word apple
can act as a category, as in a collection of different kinds of
apples. But it also can act as a concept within the category
fruit.
7
CONCEPT AND CATEGORIES
Concepts and categories can be divided in various ways.
One commonly used distinction is between natural
categories and artifact categories
(Kalenine et al., 2009; Medin, Lynch, & Solomon, 2000).
8
NATURAL CATEGORIES
are groupings that occur naturally in the world, like birds or
trees
ARTIFACT CATEGORIES
are groupings that are designed or invented by humans to
serve particular purposes or functions.
9
CONCEPTS
Concepts, on the contrary, are not always stable but can
change. Some categories are created just for the moment or
for a specific purpose, for example, “things you can write
on.” These categories are called ad hoc categories
(Dunbar, 2003; Thagard, 2003. (Barsalou, 1983; Little, Lewandowsky, & Heit, 2006)
10
CONCEPTS
Concepts are also used in other areas like computer
science. Developers try to develop algorithms that define
“spam” so that email programs can filter out unwanted
messages and your mailbox is not flooded with them.
However, spammers change the nature of their messages on
a regular basis so that it is hard to create an algorithm that
can catch all spam messages and can do so on a permanent
basis.
(Fdez-Riverola, 2007). 11
CONCEPTS
Concepts appear to have a basic level (sometimes termed a
natural level) of specificity, a level within a hierarchy that is
preferred to other levels.
(Medin, Proffitt, & Schwartz, 2000; Rosch, 1978).
12
CONCEPTS
Suppose I show you a red, roundish edible object that has a
stem and that came from a tree. You might characterize it as
a fruit, an apple, a delicious apple, a Red Delicious apple,
and so on. Most people, however, would characterize the
object as an apple. The basic, preferred level is apple. In
general, the basic level is neither the most abstract nor the
most specific. Of course, this basic level can be manipulated
by context or expertise.
13
(Tanaka & Taylor, 1991).
CONCEPTS
When people are shown pictures of objects, they identify the
objects at a basic level more quickly than they identify objects at
higher or lower levels. Objects appear to be recognized first in
terms of their basic level. Only afterward are they classified in
terms of higher- or lower-level categories. Thus, the picture of the
roundish red, edible object from a tree probably first would be
identified as an apple. Only then, if necessary, would it be
identified as a fruit or a Red Delicious apple.
(Rosch et al., 1976)
14
CONCEPTS
Now, how do people decide what objects to put into a category?
There are several theories that try to explain this process. One
theory suggests that we put an object only in one category if it
has several defining features. Another approach proposes that
we compare an object with an averaged representation (a
prototype) to decide whether it fits into a category. Yet another
is that people can categorize objects based on their own
theories about those objects. We will explore these approaches
in the next sections. 15
FEATURE-BASED CATEGORIES
The feature-based view of categories, rooted in the work
of Katz (1972) and Katz & Fodor (1963), proposes that
concepts can be broken down into defining features that
are both necessary and sufficient. A category is defined
by a set of essential features—if any are missing, the
object cannot belong to that category. For example, a
"bachelor" must have the features *male*, *unmarried*,
and *adult*. 16
FEATURE-BASED CATEGORIES
Each is necessary on its own, and together they are sufficient to define
someone as a bachelor. A 12-year-old unmarried boy is not a bachelor
(not an adult), nor is a married man (not unmarried), or an unmarried
woman (not male). Similarly, "wife" is defined as married, female, and
adult; "husband" as married, male, and adult. This approach is favored
by linguists for its systematic nature (Clark & Clark, 1977; Finley &
Badecker, 2009). However, it does not apply well to all categories. For
instance, the category "game" resists definition because it lacks a
consistent set of features (Wittgenstein, 1953)—some games are fun,
others are not; some involve multiple players, others do not. 17
PROTOTYPES AND CHARACTERISTICS
FEATURE
Prototype theory suggests that categories are centered
around a prototype—an abstract average of all previously
encountered examples. Objects are categorized by
comparing them to this prototype, which is defined by
characteristic features. These features are common in
typical examples but are not necessary for category
membership. 18
PROTOTYPES AND CHARACTERISTICS
FEATURE
For instance, the prototype of a game might include being
enjoyable, having multiple players, and offering a
challenge. However, a game can still be a game without
any of these features. Similarly, while birds often fly, flight
is not a defining feature—ostriches can't fly but are still
birds.
19
PROTOTYPES AND CHARACTERISTICS
FEATURE
In contrast to defining features (shared by all category
members), characteristic features are only shared by
many or most. This explains why some examples, like
robins, seem more typical of a category (like birds) than
less typical ones like ostriches.
20
PROTOTYPES AND CHARACTERISTICS
FEATURE
Rosch & Mervis (1975) found that people tend to list characteristic,
not defining, features for categories such as fruit or furniture (e.g.,
sweetness, being made of wood). They also showed that you can
compute a typicality score based on how many characteristic features
an instance shares with the prototype.
This principle extends to social categories: stereotypes are based on
average or characteristic features of a group (Medin, 1989; Dolderer
21
et al., 2009), not strict definitions.
CLASSICAL AND FUZZY CONCEPT
Psychologists differentiate two kinds of categories:
Classical concepts and Fuzzy concepts. Classical
concepts are categories that can be readily defined
through defining features, such as bachelor. Fuzzy
concepts are categories that cannot be so easily defined,
such as game or death. Their borders are, as their name
implies, fuzzy. 22
CLASSICAL AND FUZZY CONCEPT
Psychologists differentiate two kinds of categories:
Classical concepts and Fuzzy concepts. Classical
concepts are categories that can be readily defined
through defining features, such as bachelor. Fuzzy
concepts are categories that cannot be so easily defined,
such as game or death. Their borders are, as their name
implies, fuzzy. 23
THEORY-BASED OF
CATEGORIZATION
A departure from feature-based, prototype-based, and
exemplar-based views of meaning is a theory-based view
of meaning, also sometimes called an explanation-based
24
HOW DO PEOOLE USE THEIR
THEORIES FOR CATEGORIZATION
A theory-based view of meaning holds that people
understand and categorize concepts in terms of implicit
theories, or general ideas they have regarding those
concepts (Markman, 2003, 2007). For example, what
makes someone a “good sport”?
25
HOW DO PEOOLE USE THEIR
THEORIES FOR CATEGORIZATION
• In the componential view, you would try to isolate features of a good sport.
• In the prototype view, you would try to find characteristic features of a good
sport.
• In the exemplar view, you might try to find some good examples you have
known
in your life.
• In the theory-based view, you would use your experience to construct an
explanation for what makes someone a good sport.
26
INTELLIGENCE AND CONCEPT IN
DIFFERENT CULTURES
Culture influences many cognitive processes, including intelligence
(Lehman, Chiu, & Schaller, 2004). As a result, individuals in different
cultures may construct concepts in quite different ways, rendering results
of concept-formation or identification studies in a single culture suspect
(Atran, 1999; Coley et al., 1999; Medin & Atran, 1999). Thus, groups may
think about what appears superficially to be the same phenomenon—
whether a concept or the taking of a test—differently.
27
SEMANTIC NETWORK MODELS
Semantic-network models suggest that knowledge
is represented in our minds in the form of concepts
that are connected with each other in a web-like
form. In the following, we consider a model
developed by Collins and Quillian (1969) as well as
another model that is based on a comparison of
semantic features 28
COLLINS AND QUILLIAN’S
NETWORK MODEL
An older model still in use today is that knowledge is represented
in terms of a hierarchical semantic (related to meaning as
expressed in language—i.e., in linguistic symbols) network. A
semantic network is a web of elements of meaning (nodes) that
are connected with each other through links (Collins & Quillian,
1969). Organized knowledge representation takes the form of a
hierarchical tree diagram. The elements are called nodes; they are
typically concepts. 29
COLLINS AND QUILLIAN’S
NETWORK MODEL
The connections between the nodes are labeled
relationships. They might indicate category membership
(e.g., an “is a” relationship connecting “pig” to “mammal”),
attributes (e.g., connecting “furry” to “mammal”), or some
other semantic relationship.
30
COLLINS AND QUILLIAN’S
NETWORK MODEL
The connections between the nodes are labeled
relationships. They might indicate category membership
(e.g., an “is a” relationship connecting “pig” to “mammal”),
attributes (e.g., connecting “furry” to “mammal”), or some
other semantic relationship.
31
COLLINS AND QUILLIAN’S
NETWORK MODEL
. Thus, a network provides a means for organizing concepts.
The exact form of a semantic network differs from one
theory to another, but most networks look something like the
highly simplified network shown in Figure 8.2(next page).
The labeled relationships form links that enable the
individual to connect the various nodes in a meaningful way
32
COLLINS AND QUILLIAN’S
NETWORK MODEL
33
STRUCTURE OF SEMANTIC
NETWORK
In a simple semantic network, nodes serve as junctures
representing concepts linked by labeled relationships: a basic
network structure showing that relationship R links the nodes a
and b.
34
STRUCTURE OF SEMANTIC NETWORK
35
SEMANTIC REPRESENTATIONS
Another way to organize the many concepts
we have in our minds is by means of
schemas. First we will discuss schemas in
general and then have a look at scripts, which
are a particular kind of schema.
36
SCHEMAS
One main approach to understanding how concepts are related in
the mind is through schemas. They are very similar to semantic
networks, except that schemas are often more task-oriented.
Recall that a schema is a mental framework for organizing
knowledge. It creates a meaningful structure of related concepts.
For example, we might have a schema for a kitchen that tells us
the kinds of things one might find in a kitchen and where we might
find them. Of course, both concepts and schemas may be viewed
at many levels of analysis.
37
SCHEMAS
Schemas have several characteristics that ensure wide flexibility in their use:
1. Schemas can include other schemas. For example, a schema for animals
includes a schema for cows, a schema for apes, and so on.
2. Schemas encompass typical, general facts that can vary slightly from one
specific instance to another. For example, although the schema for mammals
includes a general fact that mammals typically have fur, it allows for humans,
who are less hairy than most other mammals. It also allows for porcupines,
which seem more prickly than furry, and for marine mammals like whales that
have just a few bristly hairs. 38
SCHEMAS
Schemas have several characteristics that ensure wide flexibility in their use:
3. Schemas can vary in their degree of abstraction. For example, a schema for
justice is much more abstract than a schema for apple or even a schema for
fruit.
39
SCHEMAS
Schemas also can include information about relationships (Komatsu, 1992).
Some of this information includes relationships among the following:
• concepts (e.g., the link between trucks and cars);
• attributes within concepts (e.g., the height and the weight of an elephant);
• attributes in related concepts (e.g., the redness of a cherry and the redness
of an apple);
• concepts and particular contexts (e.g., fish and the ocean); and
• specific concepts and general background knowledge (e.g., concepts about
particular U.S. presidents and general knowledge about the U.S. government
and about U.S. history). 40
SCRIPTS
A script contains information about the particular order in which
things occur. In general, scripts are much less flexible than
schemas. However, scripts include default values for the actors,
the props, the setting, and the sequence of events expected to
occur. These values taken together compose an overview of an
event.
41
THE “PRODUCTION” OF PROCEDURAL
KNOWLEDGE
Procedural knowledge representation is acquired through
practicing the implementation of a procedure. It is not merely a
result of reading, hearing, or otherwise acquiring information
from explicit instructions. Once a mental representation of
nondeclarative knowledge is constructed (proceduralization is
complete), that knowledge is implicit.
42
THE “PRODUCTION” OF PROCEDURAL
KNOWLEDGE
Psychologists have developed a variety of models
for how procedural information is represented and
processed. Each of these models involves the serial
processing of information, in which information is
handled through a linear sequence of operations,
one operation at a time
43
NON DECLARATIVE KNOWLEDGE
As mentioned previously, knowledge traditionally has been
described as either declarative or procedural. One can expand
the traditional distinction between declarative and procedural
knowledge to suggest that nondeclarative knowledge may
encompass a broader range of mental representations than
just procedural knowledge (Squire, 1986; Squire et al., 1990).
Specifically, in addition to declarative knowledge, we mentally
represent the following forms of nondeclarative knowledge:
44
NON DECLARATIVE KNOWLEDGE
• perceptual, motor, and cognitive skills (procedural knowledge);
• simple associative knowledge (classical and operant conditioning);
• simple non-associative knowledge (habituation and sensitization);
and
• priming (fundamental links within a knowledge network, in which the
activation of information along a particular mental pathway facilitates
the subsequent retrieval of information along a related pathway or
even the same mental pathway; see Chapter 4).
45
SEMANTIC PRIMING
In semantic priming, we are primed by a meaningful context or by
meaningful information. Such information typically is a word or cue
that is meaningfully related to the target that is used. Examples are
fruits or green things, which may prime lime.
REPETITION PRIMING
In repetition priming, a prior exposure to a word or other stimulus
primes a subsequent retrieval of that information. For example,
hearing the word lime primes subsequent stimulation for the word
lime. 46
INTERACTIVE MODELS FOR REPRESENTING
DECLARATIVE AND NON DECLARATIVE
KNOWLEDGE
So far, we have considered models for the representation of either
declarative or procedural knowledge. Next, we explore some models
that attempt to explain both. The first model is the ACT-R model,
which is based on semantic networks and production systems. Then
we look at findings that are using the human brain, rather than
computers, as a model. One such theory we will consider in detail: the
connectionist model.
47
COMBINING REPRESENTATION:
ACT-R
An excellent example of a theory that combines
forms of mental representation is the ACT (adaptive
control of thought) model of knowledge
representation and information processing
48
DECLARATIVE KNOWLEDGE WITHIN
ACT-R
Anderson’s declarative network model, like many other network
models (e.g., Collins & Loftus, 1975), contains a mechanism by
which information can be retrieved and also a structure for storing
information. Recall that within a semantic network, concepts are
stored at various nodes within the network.
49
DECLARATIVE KNOWLEDGE WITHIN
ACT-R
50
DECLARATIVE KNOWLEDGE WITHIN
ACT-R
51
PROCEDURAL KNOWLEDGE WITHIN
ACT-R
How does Anderson explain the acquisition of
procedural knowledge? Such knowledge is represented
in production systems rather than in semantic networks.
Knowledge representation of procedural skills occurs in
three stages: cognitive, associative, and autonomous
(Anderson, 1980). See Table 8.2 for examples of each of
these three stages. 52
PROCEDURAL KNOWLEDGE WITHIN
ACT-R
53
PARALLEL PROCESSING: THE
CONNECTIONIST MODEL
Computer-inspired information-processing theories assume
that humans, like computers, process information serially. That
is, information is processed one step after another. Some
aspects of human cognition may indeed be explained in terms
of serial processing, but psychobiological findings and
othercognitive research seem to indicate other aspects of
human cognition.
54
HOW THE PDP MODEL WORKS
The mental structure within which parallel processing is
believed to occur is a network. In connectionist networks, all
forms of knowledge are represented within the network
structure. Recall that the fundamental element of the
network is the node. Each node is connected to many other
nodes. These interconnected patterns of nodes enable the
individual to organize meaningfully the knowledge contained
in the connections among the various nodes. In many
network models, each node represents a concept. 55
KNOWLEDGE REPRESENTED BY
PATTERNS OF CONNECTIONS
Each individual unit (dot) is relatively uninformative, but when the
units are connected into various patterns, each pattern may be
highly informative, as illustrated in the patterns at the top of this
figure. Similarly, individual letters are relatively uninformative,
but patterns of letters may be highly informative. Using just
three-letter combinations, we can generate many different
patterns, such as DAB, FED, and other patterns shown in the
bottom of this figure. 56
KNOWLEDGE REPRESENTED BY
PATTERNS OF CONNECTIONS
In the brain, at any one time, a given neuron may be inactive, excitatory, or
inhibitory
• Inactive neurons are not stimulated beyond their threshold of excitation. They do
not release any neurotransmitters into the synapse (the interneuronal gap).
• Excitatory neurons release neurotransmitters that stimulate receptive neurons at
the synapse. They increase the likelihood that the receiving neurons will reach their
threshold of excitation.
• Inhibitory neurons release neurotransmitters that inhibit receptive neurons. They
reduce the likelihood that the receiving neurons will reach their threshold of
57
excitation.
HOW THE PDP MODEL WORKS
58
CRITISM OF THE CONNECTIONIST
MODELS
One general criticism is that connectionist networks neglect
properties that neural systems have, or that they propose
properties that neural networks do not have. Furthermore,
critics ask why any model should be more credible than
another for explaining cognitive mechanisms just because it
resembles the structure of the brain
59
COMPARING CONNECTIONIST WITH
NETWROK REPRESENTATION
lds up a knowledge base about a robin over time as more and
more information is acquired about robins. Note that
information about robins is embedded in a general network
representation that goes beyond just robins. One’s
understanding of robins partly depends on the relationship of
the robin to other birds and even other kinds of living things
60
COMPARING CONNECTIONIST WITH
NETWROK REPRESENTATION
In the network representation, the nodes represent concepts.
An individual builds up a knowledge base about a robin over
time as more and more information is acquired about robins.
Note that information about robins is embedded in a general
network representation that goes beyond just robins. One’s
understanding of robins partly depends on the relationship of
the robin to other birds and even other kinds of living things 61
HOW DOMAIN GENERAL OR
DOMAIN SPECIFIC IS COGNITION?
Cognitive psychologists should study both domain-general
and domain-specific processes. Domain-general processes
(like attention or memory) apply across many tasks, while
domain-specific processes (like face or language
recognition) are specialized and supported by evidence from
brain studies and expertise research. Recent approaches
suggest an integrated model, where general and specific
62
processes work together to support cognition.
