prototype theory

W5 – Categorisation and
1
idealized cognitive models
Chapter 11
2 1 A new approach to categorization and
the development of cognitive linguistics

For a couple of millennia we used classical

theory
When prototype theory entered, this seriously
challenged the conventional wisdom
Prototype theory itself was superseded, but its
core observations remain
especially effects

bird: this prototype in your mind has lots of attributes

what attributes are there that help us to form a catergory
 2 The classical theory
must be precise, have precise
3 definition
you have to develop a definition first all member in one
catergory is equal

Aristotle claimed that linguistic and

conceptual categories have definitional
structure
In order to be a member, you need to meet
both sufficient and necessary conditions all attributes have
to be achieved
BACHELOR is the classic example – male,
married, adult marriage--culturally informed

Suggests that we now have semantic

all basic units--male
or female, marriedprimitives, good for formal theory
or unmarried...
4 2.1 The definitional problem

Theory is precise, but its hard to identify the

conditions
The famous example is GAME, which resists
finding exact conditions
Considering CAT – what happens if it loses a
tail?
Following strict definitions has its problems
5 2.2 The problem of conceptual
fuzziness
Precise means boundaries
What about FURNITURE? What’s a carpet?
What about BIRD? Penguin is a bird right?
numbers: abstract system
ODD and EVEN numbers seems to work are not equally
accesssible, they
The point is, sometimes we have oddness should have the
same
membership
criteria
first think about the one which shared =attrbutes with most of the birds
6 2.3 The problem of prototypicality

Why do some members look better than

others?
This effect is called a typicality effect –
representing a graded category
Shouldn’t happen if it is binary right?
7 2.4 Further problems
no evidence: the easily the word is, the faster it will be recognised

For a definitional approach, it has no

psychological reality
Is a whale a fish? Some people happily say so,
but it’s a mammal most are on the land
Thus, you can have a concept, but not know
its definitions
a contradiction, concept is based on a wrong
definition
8 3 Prototype theory
3.1 Principles of categorization
Two key principles that underlie the human
categorization system (according to
prototype theory)
Cognitive economy
Principle of perceived world structure
9 3.1.1 Principle of cognitive economy

Very straightforward – we want to minimize

cognitive effort and save resources, so we
group things together finite resources and make the best use of it
save resources/effort; constrained
 get it from our daily life experience

10 3.1.2 Principle of perceived world

structure when we observed the world, there are some patterns

According to our perceptions, certain features

correlate
Feathers tend to be associated with flying
Fur tends not to be associated with swimming
 11 3.2 The categorization system

EX1

how inclusive/distinct something is

how much detail it contained

just the effect not in concept

 12 3.3 The vertical dimension

The top part contains the least amount of

details – inclusive
The lower you go, the more detail that
appears right amount of
detail
perfect of details andThe optimal level that has most inclusiveness
differencse among
each catergory
and detail, is called the basic level
Above, superordinate, below, subordinate
too abstract very specific/small detail
cannot applied to other tables
13 3.3.1 Attributes

Give someone 90 seconds, and you’ll find that

the basic level has the most features listed
 14 3.3.2 Motor movements

At the basic level, more properties are shared

share more attributes (e.g. WALK, RUN), as compared to very
in general
precise movements like PINCH, HOOK
more procise
15 3.3.3 Similarity of shapes

too general about shape

Comparing VEHICLE at the subordinate level
shows a lot of similarity, but the basic level
easy to be identified
shows the most inclusivity
16 3.3.4 Identification based on
averaged shapes
Averaging a shape over the superordinate
level isn’t done well, and again they showed
that the basic level was the most inclusive
 17 3.3.5 Cognitive economy versus level
of detail
Basic level categories at the most important as
differienate others
they are the most inclusive and informative
give me a chair vs. ...
(balancing the two)
This keeps the most information, while keeping
categories apart
Maximizes common within-category attributes,
exclude
and keeps high differences between-categoriesother
categories
Relates to cue validity – what do you use it for?
fuctional
Where is it found? Basic level maximizes this elements
18 3.3.6 Perceptual salience

Your perception is extremely important for

categorization
The basic level is the most abstract representation
of something
While you can think of a type of chair, you can’t
actually think of FURNITURE without going lower
19 3.3.7 Language acquisition

Children produce basic-level categories first

20 3.3.8 Basic-level terms in language

Base level units tend to be simple,

monomorphemic like CHAIR, subordinates tend to
one morphone
have two or more lexemes
Corpus studies show basic level words occur the
most
21 3.3.9 Are basic-level categories
universal?
This might be the case, however, it depends on
our interactions with the environment
Depending on the culture, different things may
occupy different levels

categories is based on/come from experience

the template/power to organize things is universal
22 3.4 The horizontal dimension
we always find patterns based on our experience
The world provides restrictions on categorization
Features that correlate more form categories
Members with more features are more
prototypical, leading to typicality effects
Not all members are the same – family
resemblance
share more attibutes--highly prototypical
23 3.4.1 Goodness-of-example ratings

To understand which is the most prototypical, rank

items in a category
24 3.4.2 Family resemblance

The most prototypical member has the most

attributes measure prototype based on people's instuition

To understand the structure of the category,

collect the attributes
The more frequent an attribute occurs in the
category for items, the higher the score

attribute can be ranked too

25 3.5 Problems with prototype theory
they are the same attribute, prototype cannot explain it

Why is 122337 less prototypical than 3 for ODD?

How can you have a concept but not be fully
aware of the properties?
Why aren’t there any prototypes for USA
MONARCH? predicate there must be a prototype

Why don’t categories combine well?

What gives rise to these effects?

important: what cause these problems?

26 4 The theory of idealized cognitive
models
Lakoff worked with a new theory of ICMs which
helps explain the ideas of prototypes how typicality effect arise
ICMs are stable representations, that have been
abstracted
27 4.1 Sources of typicality effects
4.1.1 The simplest type of typicality effects
Typicality effects arise from ICM mismatches
BACHELOR has been idealized, and it might not
match well with other similar ICMs
POPE is a bachelor, but not typical – POPE is
generally understood through the CATHOLIC
CHURCH ICM
pope is a ICM
pope is also unmarried
bachelar is a ICM this feature in two categories coincidentally cross together
28 4.1.2 Typicality effect due to cluster
models
Several ICMs can converge together to cause
typicality effects
MOTHER has many different ICMs, and they can
conflict

different models are interacted with each other which create typicality effect

you can refer back to differenct model

29 4.1.3 Typicality effects due to
metonymy
‘Washington said today that…’ – Washington
stands for the American government
A member stands for the whole category
The member becomes the reference point for
measurement, others becoming less prototypical
use other words/expressions
you can say that, but not so good
30
based on soical
stereotype
become the reference point
Stereotypical example of HOUSEWIFE given
Overlaps with MOTHER, but WORKING MOTHER is
deemed as less prototypical
If you know a non-womanizer BACHELOR, then
this is also prototypical
A typical bird is seen a lot – it becomes the
reference point
31

Ideals might give rise to effects – which others are

mentioned (ideal teacher – stereotypical lazy
teacher)
Paragons are members that represent ideals
Generators create the category, numbers 0-9
generate all/other
Salient examples stand out, e.g. Oxford numbers
32 4.1.4 Radial ICMs as a further sources
of typicality effects
EX2

which one/surrounding point is more influencing based on your experience

33 4.2 How the theory of ICMs resolves
problems with prototype theory
Generators are core to REAL NUMBER, so they are
judged more prototypical
come from mismatch of ICMs
No ignorance/error issue – typicality effects aren’t
representative of conceptual structure but
mismatches from ICM and so on
No prototypes for USA MONARCH, because
prototypes aren’t conceptual structure – and this
category is abstract (no experience)
PET FISH is not a combination, it is its own category
one ICM cross over another ICM, cannot combine to create a new one
34 4.3 The structure of ICMs

ICMs are structures of mental space, we review

some briefly
35 4.3.1 Image schematic ICMs

Image schemas like SOURCE-PATH-GOAL form the

basis of ICMs
36 4.3.2 Propositional ICMs

Some ICMs relate from properties that hold

between systems
37 4.3.3 Metaphoric ICMs

Abstract ICMs are based on the physical, like LOVE

38 4.3.4 Metonymic ICMs

We already discussed this in detail

39 4.3.5 Symbolic ICMs

Relates to semantic frames – buy/sell are only

understood in a commercial frame
Thus, the words are related to the system of
knowledge, resulting in symbolic units (words)