Topic 1. What Is Categorization?

# 2709 Ewa Mioduszewska

Topic 1. What is categorization?

Categories: natural – cow, dog,...; formal – prime number, odd/even number,...; other – happiness, state, ...

1.  Nominalism

Categories ---- linguistic (language) convention: dog – the range of entities may be called ‘dog’ (or it could be called otherwise). The entities may have nothing in common but the name.

2.  Realism

Categories ---- natural classes of entities in the world: dog – real-world dogs. The entities exist prior and independently of their names.

3.  Conceptualism

Categories ---- mental entities (concepts): between words and real-life objects: ‘dog’ --- DOG --- dogs. Speaker’s knowledge of the concept DOG allows him to categorize different entities as dogs.

3a Nominalist conceptualism

Concepts reflect linguistic conventions. Speaker’s concept DOG emerges from observation how the word ‘dog’ is used, which conditions its future use.

3b Realist conceptualism

Concepts reflect existing properties of the world, constituting our understanding of what the world “really” is like. They are not arbitrary creations of language.

I. Classical categorization

Ancient Greece, Aristotle (384-322 BC) Metaphysics

-  Essence and accidents of entities: necessary and sufficient conditions on the entities membership in a category.

-  Essence: all properties immanent in entities which define them and indicate their individuality.

man: two-footed animal; rational biped

If x is a man, x is two-footed

If x is a man, x is an animal

If x is a man, x is rational

Meaning postulates, truth-conditions, entailments

x is a man (y) – assigning entity x to the category y = checking properties of x against the features that define the essence of the category y (man)

Characteristics of category

1.  Categories are defined in terms of a conjunction of necessary and sufficient features.
The law of contradiction: ~(p&~p) x cannot both be and not be a member and a non-member of a category.

The law of the excluded middle” p v ~p x either is or is not a member of a category

2.  Sufficient and necessary conditions on category membership = properties/features of category members are binary: man two-footed: [+] or [-]; animal: [+] or [-]

3.  Categories have clear boundaries: an established category divides all entities into category members and non-members.

4.  All members of a category have equal status. An entity with all the defining features of a category is a full member of that category. An entity which does not exhibit all the defining features is not a member. No degrees of category membership.

Classical categorization in semantics

A feature approach in phonology, syntax and semantics – binary, primitive, universal, abstract and innate features (Katz Fodor 1963, Katz Postal 1964; Bierwish 1967, 1970; Leech 1981)

Bachelor [+MALE] [+ADULT] [+NEVER MARRIED] --- necessary and sufficient conditions on category membership

Advantages of such an approach:

a)  Economical and insightful statements about language structure (meaning)

b)  Determinable proportional relations within the lexicon: bachelor/spinster, girl/woman//boy/man

c)  Determinable selection restrictions: *sincerity admires John, *infant bachelor

d)  Determinable types of sentence meaning

The man is a bachelor – synthetic sentence

This bachelor is a man – analytic sentence (hyponymy: feature inclusion)

This bachelor is my sister – contradiction (incompatibility of features)

John is a bachelor ---- John is a man (entailment)

John is a bachelor ---- John is a man who has never married (paraphrase)

Features: ultimate, atomic constituents of word meanings potentially universal (Leibniz 17th c: alphabet of human thought, Wierzbicka 1980 Lingua mentalis, Chomsky 1985)

Problems: - number of features (ca 45 phonemes in English, ca 20 universal phonological features, semantics ?)

-  Status of features [ADULT] ≠ the meaning of the word ‘adult’ ≠ any of the attributes of adulthood in the real world. If the features are abstract, they must be innate (genetically inherited): semantic primes, primitives

-  Semantics autonomy hypothesis: G. Frege, J. Lyons: sense and reference of linguistic expressions

Autonomous semantics

1. Literal talk (language meaning)
2. Language meaning = speaker’s meaning
3. Truth-conditional semantics doesn’t tell us all about meaning
4. Distinction: semantics - pragmatics
5. Distinction: linguistic competence - communicative competence
6. Modularity of the mind
7. Formalism allowed and favored: precision, double checking, regularity and pattern tracing, comparisons, universality
7a. Formalism: do we invent or discover it
7b. Formalism: inherent to our minds or a burden to them

II. Prototypical categorization (cognitive linguistics)
1. How do we establish truth-conditions of the sentences:
a) x is a woman
b) x is a cup
c) x is a mug
d) x is a game
Conclusion: In the sentences b-d it is impossible
2. How do we know what cup, mug, game mean?
a. Family resemblance: Ludwig Wittgenstein - Philosophical Investigations (1953)
game: board-games, card-games (solitaires), ball-games, Olympic games, war games, marital games
- no set of common attributes
- criss-crossing network of similarities
- some members have nothing in common with others.
Conclusion: members of the category game can be distinguished by family resemblance.
- we learn the meaning of game from examples.
Problem: which example should we start with?
3. Prototypical categorization
- Labov 1973: cups, mugs, bowls, vases - drawings to be named:
- relevant attributes could be relational, functional or interactional
- the attributes formed a continuum
- attributes point not to the essence of objects in themselves but to the role they play in a given culture
- there are no criterial attributes of distinguishing between categories
Conclusion: We categorize in prototypes: it` s not necessary to make sure whether an entity has an attribute but how closely it comes to a prototype.
Rosch, 1973, 1975: ten categories: furniture, weapon, vegetable, tool, bird, sport, toy, clothing, fruit, vehicle; 60 items per each; 7 point scale; 200 college students.
Conclusion: degree of membership in a category is psychologically real; prototypes are psychologically real (verification time, naming); our categorization is prototypical in nature.
4. Basic level categories
Levels of categorization ( == hyponymy relations)

artifact
tool furniture dwelling place
TABLE CHAIR BED - basic level categories
dining-room kitchen dentist` s
chair chair chair
Basic level categorization is cognitively and linguistically more salient. Such categories are linguistically simple (kitchen-chair - compound, furniture - deviant, artifact - vague), short and of high frequency.
Assumption: people conceptualize things as perceptual and functional gestalts.
Problem: Why do we have basic categories ? Because they are useful: maximize the number of attributes shared by members of the category and minimize the number of attributes shared with members of other categories.
5. Prototypical members of categories.
Problem: Where do we get prototypes from ?
* inherent property of human perception: circle, square, triangle, vertical, horizontal -- perceptually more salient
* frequency
* learnt first
* mark what`s typical
* importance for a culture
* efficiency of prototype categories -- flexibility to accommodate new data without loosing its central stability
Definition of prototype
a) the central member or a cluster of central members of a category
b) schematic representation (mental image) of the conceptual core of a category
Establishing membership in a category
An entity belongs to a category if it is similar to a prototype. Similarity - a subjective notion, which underlies all categorization processes.
Similarity: a number of attributes shared
a number of attributes not shared
attributes of different weight (perceptual salience, high diagnostic value)
However, in cognitive linguistics a whole is more than its parts. Category - holistic, gestalt configurations and not attributes bundles. Attributes - dimensions along which different entities are regarded as similar.
Problem: How do you know that an entity does not belong to a category, e.g. cat = dog. You would need some criterial attributes.
There are categories with clear cut boundaries:
natural kinds (taxonomic divisions) versus nominal kinds (our definitions)
Category boundaries depend on how the world happens to be and what we know of it. Categories with clear boundaries are not typical of the notion of category.
6. Prototypes versus schemas
Schema: abstract characterization compatible with all members.
tree - oaks, elm, maples
schema: trunk, branches, leaves
PROTOTYPE tree1
+ pine
schema: trunk, branches
PROTOTYPE tree2
+ palm
schema: trunk
PROTOTYPE tree3
7. Folk and expert categorization
Experiments: odd and even numbers
Problem: the numbers are felt to have clear-cut boundaries; the data suggest they are perceived prototypically.
Within the truth-conditional approach: there is a core definition and identification procedure. Prototypical effects arise from the latter.
Cognitive approach: Schematic representation - clear-cut boundary
Prototypical representation - prototypical effects
Example: adult (water, gold)
what people believe what the legal regulations say
folk category expert category
In the case of a folk category, the schema comes from extracting what is commonly believed, what is prototypical and interactional. In the case of an expert category, the schema comes from imposing what people should believe; the schema is created, classical and resting on necessary and sufficient conditions on membership.
Some words can have only expert definitions: e.g. phoneme, or only a folk definition e.g. cup.
Language behaves as expert definitions, e.g. cup = mug. Yet, it has a device to relax the boundaries (i.e. hedges)
Examples: Loosely speaking, a telephone is a piece of furniture
?Loosely speaking, a chair is a piece of furniture
Strictly speaking, a bat is not a bird.
?Strictly speaking, a TV set is not a bird

8. Polysemy and metaphor
I. Polysemy
- in terms of family resemblance categories
- distinct though related meanings
- no need of a semantic component unifying the different uses (though it may be there)
- polysemous words, morphemes, structures, suprasegmentals
Example: diminutives in Polish
a) Basic/central/prototypical sense: smallness in physical space
b) Extension of the prototype: smallness on some dimension
c) Divergence from the prototype on experiential basis
Ad. a): domek, stolik, krzesełko, króciutki
Ad. b): skromniutki, szybciutki, prędziutko, króciutki, miłostka, awanturka, kolacyjka
Ad. c): - short temporal duration: piosenka, kolacyjka, wykładzik, rozmówka
- reduced extent or intensity: skromniutki, szybciutki
- attitude of affection: mamusia, Jaś, Zosia, sukieneczka, domeczek (metonymic transfer: small ~ affection)
- attitude of depreciation: wypracowanko, rozprawka, miłostka (metonymic transfer: small ~ unworthy)
- ambiguous affection: (accommodation of incompatible meanings in a single category): kolacyjka, awanturka, drzewko, córeczka)
- dismissive diminutive: fakcik, podanko, opowiastka, historyjka
- approximate diminutive: (expression of quantity): minutka, chwilka, godzinka.
All the instances are somehow linked to the central sense of „smallness in physical space”. The linking does not „assume” a common core of meaning.
Interesting facts about diminutives:
- productivity diminishes with lowering centrality of sense;
- diminutives often bring about independent lexical items: rozprawka, mazurek, krówka, bilecik. This is a case of semantic specialization leading to the extension of lexicon.
II. Metaphor
Origin: the basis of our conceptual system:
- spatial experience
- physical experience
- emotional experience
- emergent concepts
- experimental gestalts
Operation:

emotional experience spatial experience
domain domain
HAPPY up
------
UNHAPPY down
metaphorical links
Division of conventional metaphors
- orientational (spatial) metaphors: up-down, in-out, front-back, on-off, deep-shallow, central-peripheral. Example: Happy is up; sad is down
I` m feeling up. My spirits rose
I’m feeling down. I’m depressed. My spirits sank.
- ontological metaphors (experience with physical objects and substances). Example: Inflation is an entity.
Inflation is lowering our standard of living.
We need to combat inflation.
Inflation is backing us into a corner.
Example: The mind is an entity - The mind is a machine
Example: personification: caught Life has cheated me. Cancer finally caught up with him
Example: metonymy: The ham sandwich is waiting for his check
I`ve got a new set of wheels
He bought a Ford
- structural metaphors (culturally based)
Example: Argument is war
Your claims are indefensible
I demolished his arguments
You disagree? Okay, shoot!
Example: Communication is sending (Conduit metaphor)
I gave you that idea
His words carry little meaning
Time is money
- figurative metaphors
a) conventional metaphor: Theories are buildings; These facts are foundations of my theory
b) extension of the used part: These facts are bricks and mortar of my theory
c) unusual part: He prefers massive Gothic theories covered with gargoyles
d) novel metaphor: His theories are patriarchs who father many children
9. Cognitive linguistics - summary
* non-literal talk: mainly tropes
* meaning = speaker’s meaning
* tells us all about meaning
* no distinction among levels of language description: continuum
* no linguistic competence
* holistic view of the mind
* no formalism: meaning is culture and individual relative

Truth in autonomous semantics and in cognitive linguistics
- in the former: truth relative to a model
truth may be doubted: true/false/neither
Example: model
A B C D E
stołek taboret krzesło krzesło ?
Krzesło: { C, D}
taboret: {B}
stołek: {A}
X jest krzesłem:
A jest krzesłem: false
B jest krzesłem: false
C jest krzesłem: true
D jest krzesłem: true
E jest krzesłem: true
E jest krzesłem: neither true nor false
Truth condition: P(a) is true if the entity denoted by „a” is in the extension of P. P(a) is false if the entity denoted by „a” is not in the extension of P but is found in an extension of some other predicate of the model. P(a) is neither true nor false otherwise.
- special case: the model is the whole of our world. Then we speak about truth in the world ---- objectivism
- truth in cognitive linguistics: relative to the conceptual system. It avoids full subjectivism because individuals interact in their community. Truth is what members of a community believe to be true
OBJECTIVISM, SUBJECTIVISM, EXPERIENTIALISM
OBJECTIVISM
- the world is made up of objects which have properties independent of people (e.g. rock)
- knowledge of the world consists in experiencing the objects and learning about their properties
- understanding the objects in terms of categories and concepts which correspond to the properties of objects
- objective reality exists and we may describe it truthfully or falsely. Science guides the truthful pronouncements
- words have fixed meanings: they express concepts and categories in terms of which we think
- people can be objective and can speak literally
SUBJECTIVISM
- we rely on our senses and intuitions
- most important things: feelings, aesthetics, morality, spiritual awareness - purely subjective
- art and poetry - more important than reality: subjective (imagination)
- language: imaginative, metaphorical
- objectivism is dangerous for it misses the essence
EXPERIENTIALISM
- no absolute truth
- no imaginative idiosyncratic truth
- metaphor: unites reason and imagination. Reason comprises categorization and inference, imagination allows us to see one thing in terms of another
- truth relative to understanding
- objectivity relative to the conceptual system of a culture
- transcultural values and concepts possible (but not universal)
- we understand the world through interacting with it