topics: |  cognitive | psychology | language | category | prototype-theory

This is the book that defined cognitive linguistics, well before that term came to the fore. Much of it is refreshed in Philosophy in the Flesh but the examples are muted there, and this book remains the sharper read.

The title comes from the fact that in the dying aborigine language Dyirbal, Women, and fire, and dangerous things appear to constitute a linguistic class, one of four groupings of nouns where each class must be preceded by one of four particles. These particles are called classifiers, and are a bit similar to determiners - e.g. romance languages like French require that determiners (la, le) agree with grammatical "gender". Dyirbal has four such groupings, and the second one, which must be preceded by balan, includes women, bandicoots, dogs, platypus... scorpions, anything connected with water or fire, sun and stars, shields, some spears, some trees, etc.

3 Prototype Effects in Language

the most widely accepted views of language within both linguistics and the philosophy of language assume that language is a separate "modular" system independent of the rest of cognition. The independence of grammar from the rest of cognition is perhaps the most fundamental assumption on which Noam Chomsky's theory of language rests. ... the very idea that language is a "formal system" (in the technical mathematical sense used by Chomsky and many other linguistic theorists) requires the assumption that language is independent of the rest of cognition. That formal-system view also embodies the implicit assumption that categories are classical (and hence can be characterized by distinctive features). Such views are also the norm in the philosophy of language, (e.g. Richard Montague, Donald Davidson, David Lewis, Saul Kripke, and many others).

Thus, the question of what linguistic categories are like is important in
two ways.

First, it affects our understanding of what language is. Does language make
use of general cognitive mechanisms? Or is it something separate and
independent, using only mechanisms of its own? How this question is answered
will determine the course of the future study of language.  Entirely
different questions will be asked and theories proposed depending on the
answer.

Second, the answer will affect the study of cognition, since it will
determine whether linguistic evidence is admissible in the study of the mind
in general.

Markedness

MARKEDNESS:  some morphological categories have a "mark" and
others are "unmarked."  Take the category of number in English.  Plural
number has a "mark," the morpheme -s, as in boys, while singular number
lacks any overt "mark," as in boy.  The singular is thus the unmarked
member of the morphological category number in English.

The intuition that goes along with this is that singular is, somehow,
cognitively simpler than plural and that its cognitive simplicity is
reflected in its shorter form. The idea here is that simplicity in cognition
is reflected in simplicity of form.  Zero-marking for a morpheme is one kind
of simplicity.

In phonology, markedness is often understood in terms of some notion
of relative ease of articulation. For example, the consonants p, t, and k
are voiceless, that is, they do not involve the vibration of the vocal
chords, while the minimally contrasting voiced consonants b, d, and g do
involve vocal cord vibration. Thus, one can understand voicing as a
"mark" added to voiceless consonants to yield voiced consonants, except
between vowels where the vocal cords are vibrating to produce the
vowels.  [b,d,g are +voice, p,t,k are -voice]

some lgs have only voiceless, not voiced.

in English, after initial S-, there is no contrast
between voiced and voiceless consonants. Only voiceless consonants may
occur. English has words like spot, but no contrasting words like sbot.
[in German at the end of words] only the voiceless consonants can
occur. Thus, for example, /d/ is pronounced as [t]... 60

In general, where the contrast is neutralized (that is, only one member of
the pair can occur), the one which occurs is "unmarked" in that environment.

neutralization in semantics:
tall-short, happy-sad. These contrastive pairs are not completely
symmetric. For example, if one asks How tall is Harry? one is not suggesting
that Harry is tall, but if one asks How short is Harry? one is suggesting
that Harry is short.
--> tall is unmarked in the tall-short contrast set.

Prototype effects at different levels in language

--Phonology: phoneme vs phone:

A phone is a unit of speech sound, while a phoneme is a cognitive element
understood as occurring "at a higher level" and usually represented by a
phone. For example, English has a phoneme /k/ (sometimes spelled with the
letter c in English orthography) which occurs in the words cool, keel, key,
school, and flak.

If attention is payed to details of pronunciation, it turns out that /k/ is
pronounced differently in these words: aspirated velar [kh] in cool,
aspirated palatal [k'h] in keel, unaspirated velar [k] in school, and
unaspirated palatal [k'] in ski. English speakers perceive these, despite
their differences in pronunciation, as being instances of the same phoneme
/k/. However, there are other languages in which [kh] and [k] are instances
of different phonemes, and others still in which [k'] and [k] are instances
of different phonemes.

Jeri Jaeger (1980) has replicated Rosch's experiments in the domain of
phonology. She suggests, on the basis of experimental evidence, that phonemes
are prototype-based categories of phones. Thus, the phoneme [k] in English =
category {[k), [kh], [k'), [k'h]}, with [k] as the prototypical member.

Phonemic categories in general are understood in terms of their prototypical
members. The nonprototypical phones are related to the prototype by
phonological rules.

Jaeger's results, if correct, indicate that phonological categorization, like
other cognitive categorization, shows prototype effects. Her results
contradict most contemporary phonological theories, which take the classical
theory of categorization for granted. They point in the direction of a
unification of phonology and other aspects of cognition.

Morphology

Bybee and Moder (1983) have shown that English strong verbs like
string/strung form a morphological category that displays prototype
effects. They argue that verbs that form their past tense with Lambda
(spelled u in English orthography) form a prototype-based category. The verbs
include:

	spin, win, cling, fling, sling, sting, string, swing, wring, hang, stick,
	strike, slink, stick, sneak, dig
	+ bring, shake (in certain dialects)

On the basis of experimental results, they argue that the category has a
prototype with the following properties:

	It begins with s followed by one or two consonants: sC(C)-.
	It ends with the velar nasal: /ij/.
	It has a lax high front vowel: I.

Although the verbs in the category cannot be defined by common features,
they all bear family resemblances to this prototype. String, sling,
swing, and sting fit it exactly.

* "one" difference from the prototype: cling, fling, and bring
   have two initial consonants, but no s; spin and stick have the right initial
   consonant cluster and vowel, but differ in the final consonant - spin
   has a dental instead of a velar nasal and
   stick has a velar stop instead of a velar nasal.
* two minimal differences: Win has no initial s and a final dental nasal
   instead of a velar. Strike has a nonnasal final consonant and a different
   vowel.

This category can be categorized by a central member plus something else.  In
this case the "something else" is a characterization of "minimal"
phonological differences: the lack of an initial s, the lack of nasalization,
a different vowel, the difference between a velar and a dental consonant,
etc.  [But no real] theory of what counts as a minimal difference for
morphological categorization...

Syntactic category : Nouns

In a number of studies ranging widely over English syntax, John Robert
Ross (1972, 1973a,b, 1974, 1981) has shown that just about every syntactic
category in the language shows prototype effects.

Ross's basic insight is that normal nouns undergo a large range of
grammatical processes in English, while less nouny nouns do not undergo the
full range of processes that apply to nouns in general. Consider the nouns
toe, breath, way, and time, as they occur in the expressions:

	to stub one's toe
	to hold one's breath
	to lose one's way
	to take one's time

Ross demonstrates that within these expressions toe is nounier than breath,
which is nounier than way, which is nounier than time. Ross (1981) gives
three syntactic environments that demonstrate the hierarchy. Starred
sentences indicate ill-formedness.

   I. Modification by a passive participle
	A stubbed toe can be very painful.
	* Held breath is usually fetid when released.
	* A lost way has been the cause of many a missed appointment.
	* Taken time might tend to irritate your boss.

   II. Gapping
	1 stubbed my toe, and she hers.
	1 held my breath, and she hers.
	*1 lost my way, and she hers.
	*1 took my time, and she hers.

   III. Pluralization
	Betty and Sue stubbed their toes.
	*Betty and Sue stubbed their toe.

	Betty and Sue held their breaths.
	Betty and Sue held their breath.

	*Betty and Sue lost their ways.
	Betty and Sue lost their way.

	*Betty and Sue took their times.
	Betty and Sue took their time.

   IV. Pronominalization (Lakoff)
	I stubbed my toe, but didn't hurt IT.
	Sam held his breath for a few seconds and then released IT.
	Harry lost his way, but found IT again.
	*Harry took his time, but wasted IT.

In each of these cases, the nounier nouns follow the general rule (that is,
they behave the way one would expect nouns to behave), while the less
nouny nouns do not follow the rule. As the sentences indicate, there is a
hierarchy of nouniness among these examples...

More recently, Hopper and Thompson (1984) have proposed that the
prototypical members of the syntactic categories noun and verb can be
defined in terms of semantic and discourse functions. They provide an account
with examples from a wide range of languages that indicate that
nouns and verbs have prototypical functions in discourses.

Subject, Agent, and Topic

Bates and MacWhinney (1982) proposed on the basis of language acquisition
data that prototype theory can be used to characterize the grammatical
relation SUBJECT in the following way:
	- A prototypical SUBJECT is both AGENT and TOPIC.

Van Oosten (1984) has found a wide range of evidence in English substantiating
this hypothesis and expanding it to include the following:

   - AGENT and TOPIC are both natural categories centering around prototypes.
   - Membership in the category SUBJECT cannot be completely predicted
	from the properties of agents and topics.

Perhaps the most striking confirmation of the Bates-MacWhinney hypothesis
comes from Van Oosten's study of the uses of the passive in English.
Van Oosten picked out passive sentences as they occurred in transcribed
conversation and compiled a list of all the uses. The list seemed
random. She then compared her list of uses of the passive with her list of
the properties of prototypical agents and topics. What she noticed was a
remarkable correlation - passive usage occurred whenever no single noun
phrase had all the (prototypical) agent and topic properties.

6 RADIAL CATEGORIES

"mother": as category, is structured radially, with a central subcategory
  defined by birth, nurture, etc, + non-central extensions (adoptive mother,
  birth mother, foster mother, surrogate mother, etc.) - which are extended
  from the central concept not by rules but by convention.  But these
  extensions are not random - they are determined by the central model and
  the extensions; birth mother and foster mother are not understood purely on
  their own terms, but wrt the central model of _mother. 91

Borges' taxonomy:

   These ambiguities, redundances, and deficiences recall those attributed by
   Dr. Franz Kuhn to a certain Chinese encyclopedia entitled Celestial
   Emporium of Benevolent Knowledge. On those remote pages it is written that
   animals are divided into (a) those that belong to the Emperor, (b)
   embalmed ones, (c) those that are trained, (d) suckling pigs, (e)
   mermaids, (f) fabulous ones, (g) stray dogs, (h) those that are included
   in this classification, (i) those that tremble as if they were mad, (j)
   innumerable ones, (k) those drawn with a very fine camel's hair brush, (l)
   others, (m) those that have just broken a flower vase, (n) those that
   resemble flies from a distance.  -- p.92

Borges of course, deals with the fantastic. These not only are not natural
human cateogires -- they could not be natural human categories. But part of
what makes this passage art, rather than mere fantasy, is that it comes close
to the impression a Western reader gets when reading descriptions of
nonwestern languages and cultures. The fact is that people around the world
categorize things in ways that both boggle the Western mind and stump Western
linguists and antropologists.

Dyirbal classifiers

An excellent example is the classification of things in the world that
occurs in traditional Dyirbal, an aboriginal language of Australia.
every noun must be preceded by one of four classifiers:
bayi, balan, balam, bala

These words classify all objects in the Dyirbal universe, and to speak
Dyirbal correctly one must use the right classifier before each noun. Here is
a brief version of the Dyirbal classifcation of objects in the universe, as
described by R.M.W. Dixon (1982):

1. Bayi: men, kangaroos, possums, bats, most snakes, most fishes, some birds,
   most insects, the moon, storms, rainbows, boomerangs, some spears, etc.
2. Balan: women, bandicoots, dogs, platypus, echidna, some snakes+fishes,
most birds, fireflies, scorpions, crickets, the hairy mary grub, anything
connected with water or fire, sun and stars, shields, some spears, some
   trees, etc.
3. Balam: all edible fruit and the plants that bear them, tubers, ferns,
   honey, cigarettes, wine, cake.
4. Bala: parts of the body, meat, bees, wind, yamsticks, some spears, most
   trees, grass, mud, stones, noises and language, etc.

Dixon observed that speakers learn the categ's systematically.
Proposes this anlaysis:

1. Bayi: (human) males, animals
2. Balan: (human) females, water, fire, fighting
3. Balam: nonflesh food
4. Bala: everything not in the other classes.

[NOTE: See also baudhAyana, q. in AK Ramanujan Indian way of thought] :
    There is a difference between the [Brahmins of the] South and the
    North on five points.  We shall describe the practices of the
    South: to eat with a person not having received Brahmanical
    initiation: to eat with one's wife; to eat food prepared the
    previous day; to marry the daughter of the maternal uncle or
    paternal aunt.  And for the North: to sell wool; to drink spirits'
    to traffic in animals with two rows of teeth' to take up the
    profession of arms; to make sea voyages.
	- Collected essays, A. K. Ramanujan, p.427 ]

[ALSO, this poem by dharmakIrti, in subhAshitaratnakoSha #478, in
 John Brough's translation:
	The grammar books all say that "mind" is neuter,
	And so I thought it safe to let my mind
	    Salute her.
	 But now it lingers in embraces tender:
	 For Panini made a mistake, I find,
	    In gender. (Poems from the Sanskrit, 44)]

Principles of Classifiers

Dixon further elaborates a "domain-of-experience" principle.
     If object A is associated w a domain or experience, then other entities
     in that domain are also in the same categ as A. 93

e.g. since fish are categ 1, fishing spears, fishing line, etc are in 1;
similarly trees that bear edible fruit are also in 3, though when the same
word is used to refer to the wood of such a tree, as in firewood etc, then it
is class 4.  Fighting implements are in the same class as fire and dangerous
things in class 2.

But "myth and belief" may override any rules.  E.g. though birds are animate,
they are believed to be the spirits of dead human females, so they are 2.

Sometimes, some objects with an "important property" may be classified
separately - most often this important property is "harmfulness".  e.g. most
fish are 1, but stone fish and gar fish are harmful, so 2.  trees bushes
grasses with no edible parts in 4, but stinging trees and stinging vine in
2. Other way also - hawks shd be with birds in 2, but since they are harmful,
they are placed in 1. 94-5

Dixon's achievement is remarkable... has provided a superb example of how
human cognition works.  General principles:

* Centrality: basic members of the category are central
* Chaining: complex categs structured by chaining - central members to others
	  etc. e.g. women linked to the sun, to sunburn, to hairy mary grub
* Experiential domains: may be culture-specific
* Idealized models: myths and beliefs
* Specific knowledge overcomes general knowledge
* The other: conceptual systems can have an "everything else" categ
No common properties: categs on the whole not based on any common properties

Dyirbal: language death

Dixon's research was in 1963.  By 1983, Dyirbal culture was dying under the
onslaught of English (taught in schools etc).  Study of changes based
on several generations of speakers.  [Annette Schmidt, 1985]:

age 45+: speak traditional Dyirbal
~35: intermediate, w. some simplified forms  --> categs are breaking down
     but mythic links are kept; sun, stars, bird are in 2 with women, as is
     fire.  Fishing is lost as a domain relev to categorizn, fish are in 1,
     but fishing spears/lines etc have gone to 4. water is still with 2.
     but speakers show variations; one speaker has lost
     the danger link (stonefish/gar in 1, nettles in 4), one spkr still has
     it.  two speakers lost the dog and bandicoot as exceptional animals in
     2, they went to 1 with other animals.

much younger: very simple - categ 3 lost completely.  Now:
  1. Bayi - human males, animates
  2. Balan: human females
  3. Bala: everything else

on the death of Dyirbal
Young people’s Dyirbal... as spoken by the "Rock’n-Rollers"

Loss of prestige: "Talking Guwal [Dyirbal] to a waybala [white person], it’s like singing an’ you’re ashamed of your voice. "

Correction of young semi-speakers by elders:

"If I say `Oh, that's my gaya [mother's younger brother] there', she'll probably say `You can't say gaya to me... You gotta say mugu [mother's elder brother]’."

Daughter:   nganaji gotta cook-iman bala you know.
                   we                     -TRANS  it
                "We've got to cook that, you know"

Mother:  nyajun!
                cook
            "You mean `cook'!"

Daughter:   nyaju, cook-iman ... bala
                cook,        TRANS    it
               "Nyaju, cook it, whatever..."

Mother:  nginda mijiji-bin!
               you white-woman-INTRANS
            "You've become a white woman!"

"That Phyllis, she don’t talk Guwal right. She mixes up the English... she got it wrong."
"They won’t think [in Dyirbal]. They sorta can’t get round their own language."

NOTE: These losses support the radial analysis - center is retained

Evan Pritchard on Lakoff:
How concepts are created: Lakoff

   In Dyirbal, an Australian aboriginal language, there are four categories:
   bayi, balan, balam, and bala. Everything that exists can be put into these
   four categories. As you might expect, if a people have only four
   categories for things, then some things could put together that we might
   not perceive to belong together. Specifically, women, fire, and dangerous
   things are all in the same category (balan).

   Lakoff explains how categorization in Dyirbal, or any other language,
   could have happened using three principles.

   * Domain of experience principle, so things that are experienced in the
     same way are categorized together.
   * Myth and belief principle, things that are linked by myth or belief are
     categorized together.
   * Important property principle, things that have an important property are
     categorized together.

   In Dyirbal, bayi and balan are categories for human male and human female,
   balam is for non-human living things, and bala is for everything
   else. Women are placed in the balan category (domain of experience
   principle), and women are believed to be related to the sun (myth and
   belief principle), which is related to fire, so fire is also placed in the
   balan category, but an important property of fire is that it's dangerous
   (important property principle), so dangerous things are also placed in the
   balan category.

Japanese classifier: hon

The Japanese classifier hon is typically applied for
long thin objects: sticks, canes, pencils, rope, etc.

Radial category - the center of the category constitutes of long thin
objects.

Non-central extensions:
•  Trajectories: hits in baseball, serves in volleyball
•  Communications (because of long, thin media: wires, and, in early
   Japan, scrolls; also, communications are trajectories according to the
   CONDUIT metaphor): letters, telephone calls, radio & TV programs
•  Activities done with long, thin objects: martial arts contests (swords and
   staffs), medical injections (needles)
•  Activities like those done with long, thin objects: judo contests (no
   staffs or swords), Zen koan contests

All these usages "radiate" from the central usage of "long thin objects"

11 The objectivist paradigm --> experientialism

BASIC REALISM: common to both objectivism and experientialism
- real world exists, external to human beings,
      and includes the reality of human experience
- links human concepts and other aspects of reality
- truth - not based merely on internal coherence
- stable knowledge exists in the world
- rejects the view that diff conceptual systems are as good a the other

two aspects:
- METAPHYSICS: nature of world, indep of human understanding
- EPISTEMOLOGY: nature of human cognition, language, knowledge

OBJECTIVIST METAPHYSICS
reality consists of:
   - entities,
   - properties of entities
   - relations that hold among these entities

maps to set theoretical models, which consist of 
     - entities
     - sets of entities (defined by the common properties of the members)
     - sets of n-tuples (corresponding to relations among entities)

--> classical theory of categories --> set-theoretic implementn is possible

objectivist metaphysics goes beyond the metaphysics of
basic realism.

Basic realism merely assumes that there is a reality of some
sort. Objectivist metaphysics is much more specific. It additionally assumes
that reality is correctly and completely structured in a way that can
be modeled by set-theoretical models-that is, in terms of entities, properties,
and relations.

OBJECTIVE CATEGORIES:
[since properties have an objective existence]
entities in the world form objectively existing categories based on theis
shared objective properties.  161

Essentialism: distinguishes shared "essential" properties, as opp to shared
accidental properties":

NATURAL KINDS: [characterized by] some 'essential nature' which the thing
shares with other members of the natural kind.  What the essential nature is
is not a matter of language analysis but of scientific theory construction.
[Putnam 75, p.104]

Objectivist Legacy

	According to the objectivist paradigm, true knowledge of the external
	world can only be achieved if the system of symbols we use in thinking
	can accurately represent the external world. The objectivist
	conception of mind mu~t therefore rule out anything that can get in
	the way of that: perception, which can fool us; the body, which has
	its frailties; society, which has its pressures and special interests;
	memories, which can fade; mental images, which can differ from person
	to person; and imaginationespecially metaphor and metonymy-which
	cannot fit the objectively given external world.

It is our objectivist legacy that we view rationality as being purely mental,
unemotional, detached-independent of imagination, of social functioning, and
of the limitations of our bodies and our memories.

the development of computer science is bound up with the development of the
foundations of mathematics. As a result, the most common kind of data base now
in use happens to fit an objectivist metaphysics: It stores representations of
entities, their properties, and the relations holding among them. In your
bank's computer, you might be represented by your bank account number,
together with your bank balance on various dates, your credit rating on
various dates, etc. The data base of your bank's computer is an
objectivist universe. In it you are your account number and your proper184
ties are your bank balance and your credit rating. People have been
treated as numbers and collections of records for a long time, and they
will be treated much more so in the future.

Such treatment serves an important function in our society. There is a major
folk theory in our society according to which being objective is being fair,
and human judgment is subject to error or likely to be biased.  Consequently
decisions concerning people should be made on "objective" grounds as often as
possible. It is the major way that people who make decisions avoid blame. If
there are "cbjective" criteria on which to base a decision, then one cannot be
blamed for being biased, and consequently one cannot be criticized, demoted,
fired, or sued.

[In this context, it is tempting to contrast some of the views from eastern
mores - e.g. A.K. Ramanujan's Is there an Indian way of Thinking?:

    [The concept of universalization] - putting oneself in another's place -
    it is the golden rule for the new testament, Hobbes' "law of all men" -
    "do not do unto others what you do not want done unto you."  The main
    tradition of Judeo-Christian ethics is based on such a premise of
    universalization - Manu would not understand such a premise.  To be moral,
    for Manu, is to particularize - to ask who did what, to whom and when.  -
    p.39

[LATER, quotes Blake]
"one law for the lion and the ox is oppression."  p.49

	- AK Ramanujan, Is there an Indian way of thinking

--Formalist Mathematics-- 

Formalism arose as an approach to the study of the foundations of
mathematics. 219

The discovery of noneuclidean geometries showed that the
axiomatic method, thought to be at the heart of mathematics, was
itself not properly understood.

Euclid's fifth postulate let's us conclude such "self-evident" truths such as: 

	Through a point outside of a straight line, L, there can be drawn
	exactly one straight line parallel to L.

But if we consider the geometry of the surface of a sphere, then a
"straight line" may mean a great circle.  Then, given a straight
line, L, no straight lines parallel to L can be drawn through a point outside
of L. 

And if we take the subject matter to be about a saddle-shaped surface
and we take a "straight line" to be a geodesic on that surface, then
there may be more than one line through P which are "parallel" to L
(and never meet). 

Even if we try to generalize Euclid's postulates so that 
instead of planes we may call them two-dimensional surfaces,
and instead of lines we could say geodesics, etc.  But even with 
such more general concepts, there might be still other geometries with still
different concepts than geodesics and surfaces. There was no guarantee that
any fixed concepts would be general enoughto avoid such problems
in the future. 221

Hilbert and Formalism

David Hilbert (see Kleene, Stephen. 1967. Mathematical Logic; chap 4) came
up with a solution that was completely general; his program of
formalism. Hilbert viewed mathematical proofs as merely matters of form,
with questions of meaning put aside to be discussed outside mathematics
proper in "metamathematics." Mathematics, Hilbert suggested, is the study of
meaningless symbols, and mathematical proofs are sequences of strings of
uninterpreted symbols, with the lines of a proof related to one another by
regular rules. In a formal axiomatic system, as Hilbert defined it, axioms
are strings of uninterpreted symbols, and theorems are other strings of
uninterpreted symbols derived from the axioms by rules. - 221

Similarly, mathematical logic is technically no more than the study of
sequences of symbols strings (proof theory) and the way symbol strings can be
paired with structures continaing entities and sets (model theory). What
makes it the study of reason? The answer is: objectivist philosophy plus a
way of understanding the models. It is only by assuming the correctness of
objectivist philosophy and by imposing such an understanding that
mathematical logic can be viewed as the study of reason in general. Such an
understanding has been imposed by objectivist philosophers. There is nothing
inherent to mathematical logic that makes it the study of reason.

Hilbert was wrong about mathematics being nothing more than the study of
meaningless symbols and their relationship to meaningless structures. Two
things make formal mathematics mathematics: (a) the way those symbols and
structures are understood as being about familiar mathematical domains and
(b) the detailed justifications for adopting such an understanding. The
assumptions of objectivist philosophy have been assumed to be sufficient
justification. But that is no justification at all. What is needed is
empirical justification. - 223/4

16 A New Realism

What we have referred to as objectivism is a special case of what Putnam calls
metaphysical realism. ... Putnam has argued that metaphysical realism is
internally incoherent. Its incoherence lies in its epistemology - its view of
meaning, reference, knowledge, and understanding.  The source of the
incoherence is what Putnam calls its externalist perspective, that one can
stand outside reality and find a unique correct way to understand reality.

Such an understanding, on the view of metaphysical realism, would involve a
symbol system standing external to the rest of reality and a reference
relation pairing symbols and aspects of reality. The reference relation is
assumed to "give meaning" to the symbols. First, Putnam shows that this is
logically impossible, without violating what we mean by "meaning." Second,
Putnam points out that in order for such an understanding to be unique and
correct, the reference relation itself must be part of reality. He then
observes that this too is logically impossible.

Thus, Putnam concludes, there cannot be such a thing as "exactly one true and
complete description of 'the way the world is'" -- that is, there can be no
God's eye view of reality. The crucial words here are "description" and
"view." They presuppose an external perspective: a symbol system external to
reality, related to reality by a reference relation that gives meaning to the
symbols. Putnam is not saying that there is no reality. And he is not saying
that there is no "way the world is." He is not denying basic realism. He is
only denying a certain epistomology. He is not saying that we cannot have
correct knowledge. What ihe is saying is that we cannot have a priviledged
correct description from an externalist perspective. - 260

(from http://cogweb.ucla.edu/CogSci/Lakoff.html):

On the Relativity of Knowledge and Truth

     "Knowledge, like truth, is relative to understanding. Our folk view of
     knowledge as being absolute comes from the same source as our folk view
     that truth is absolute, which is the folk theory that there is only one
     way to understand a situation. When that folk theory fails, and we have
     multiple ways of understanding, or 'framing,' a situation, then
     knowledge, like truth, becomes relative to that understanding. Likewise,
     when our knowledge is stable and secure, knowledge based on that
     understanding is stable and secure.

     Is such knowledge 'real knowledge'? Well, it's as real as our knowledge
     ever gets--real enough for all but the most seasoned skeptics." (300)

On the different kinds of cognitive frames

Lakoff argues that experience is made possible and structured by
preconceptual structures-- "directly meaningful concepts" roughly the same
for all human beings that thus provide "certain fixed points in the objective
evaluation of situations". He divides them into basic-level structures and
image-schema structures, and acknowledges there may be other
kinds. Basic-level structures arise "as a result of our capacities for
gestalt perception, mental imagery, and motor movement" and manifest as
basic-level categories such as hunger and pain, water, wood, and stone,
people and cats, and (perhaps more surprisingly) tables and houses
(302). Image schemas are spatial mappings such as source-path-goal,
center-periphery, and container. It is out of these basic cognitive tools
that more complex cognitive models of reality are constructed.


--- blurb
Focusing on studies of how humans categorize objects and ideas, this book
examines the new understanding of human thought which proposes that human
reason is imaginative, metaphorical, and intrinsically linked with the human
body.