IN DEFERENCE TO REFERENCE
by
Gab~iel
Ma,rk Aurel Segal
B.A., London lJnlversity, 1981
B.Phil., ()xfor'd University, 1984
SUBMI1.,rrED
AND
'"f() rrfi~ DEPARrr~~ENrr OF'
J.~lNGlJlS'['lCS
PHIL~OSOPf-IY
IN PAR TIAL~ FULFlLLMENT ()P
REQUIREMENrrS FOR l'HE DEGREE Ob'
rI't·l!~
in
Prl ILOSOPH y
at the
June 1987
~ Gab~iel Mark Au rel Sega.l 1987
rfhe author hereoy grants to M.I.rr. permission to t eproduce and to
distribute publicly caples of this thesis document in whole or in part .
1
Sign~ture
of Author
----~----,-,..-----::-
-----___=__~~-__c_'-
Certified by_
Ned Blockrrhesis Supervisor
Certified by_
Noam Chomsky
rrhesi::»
Accepted by _ _
--T_-~
"O-O'--------__
Supel~visor
------~--?---~-_:___-
George Boolos
Chair, l)epArtrnenlal Graduate Oornrnittee
'.'~~'f.!r.
'Dr~f/f~lE
JAJIM 2 6 1987
- 1 ..
Ufflffiffliiti
Arcnives
Abstract
This dissertation consists in three separate papers.
The topic of Iln Deference to Refernnts' is the semantic structure of
belief reports of the form 'A believes that pl. I argue that no existing
theory of these sentences satisfactorily accounts for anaphoric relations
among expressions inside and outside of the embedded complement
sentences. I propose a new account of belief reports, which assigns to
embedded expressions their normal semantic values, but which also
exploits Frege's idea of using senses to explain the apparent failures of
extensionality in the reports.
In 'Who's Afraid of Narrow Minds?' I defend the thesis that the
propositional attitudes ascribed by folk psychology possess a level of
content ('narrow content') that is independent of anything outside the
subject. I examine various objections to that conclusion, and show that
they fail.
of the
What is not Therel I examine the nature
In ISeeing
by the
states
perceptual
to
contents attributed
representational
computational theory of vision. I discuss Tyler Burge's claim that such
contents are not narrow, but depend essentially on aspects of the
external environment. I show that the claim is false, and that the
explanatory power of the theory of vision depends upon its employment
of a notion of narrow content.
- 2-
Acknowledgments
A large number of people have helped a geat deal in the preparation
of this dissertation. It is difficult to express gratitude in amounts that
are duly vast, without sounding smarmy. If I fail in my attempt to do
so, too bad.
I would like to thank:
- The members of my thesis committee,
Ned Block and Noam
the
extreme
with their time and
Chomsky. Both have been generous in
Without their guidance the dissertation would have been
energy.
considerably more feeble than it is, had it been at all.
- George Boolos, who has helped me with every aspect of my
philosophical thought, and in numerous other ways. His ability to make
accessible to me conceptually
and technically difficuit matters is .
unsurpassed, as is his kindness.
- Jim Higginbotham, in large measure, who
ready to offer explanations, advice and help.
seemed
always
to
be
- Jerry Fodor for much comment, criticism and discussion, during the
infancy of this work, and for being a source of inspiration.
- Richard Larson for discussions frequent, varied and often useful,
and for substantive help with matters of semantics and syntax.
- Peggy Speas for being ever available as a sounding board, for
explanations of linguistics and for the main ideas in the section of 'In
Deference
to Referents'
that is drawn
from a paper
that
we
co-authored.
- Rob Cummins who read draft after draft of 'Seeing What is not
There', and gave clear, lengthy, detailed and invaluable comments on
each one. Without his generosity a substantial part of this thesis would
scarcely be.
- James Blevins, Ned Block, George Boolos, Sylvain Bromberger,
Alex George, Jim Harris, Jim
Noam Chomsky,
Richard Cartwright,
Higginbotham, Paul Horwich, Richard Larson, Dan Leary, Jay Lebed,
Sarah Patterson, Thomas Ricketts, Jorge Rodriguez, Marc Ryser, Peggy
Speas, Carol Tenny, Judith Thomson, Tom Uebel and Crispin Wright for
helpful comments and/or valuable discussions
searching criticisms,
relating to 'In Deference to Referents'
-3-
- Ned Block, George Boolos, Sylvain Bromberger, John Carriero,
Noam Chomsky, Adrian Cussins, Jim Higginbotham, Jay Lebed, John
McDowell, James Page, Sarah Patterson, Mike Smith and Thomas Uebel
for contributions of the same sort to 'Who's Afraid of Narrow Minds?'
- Mike Antony, Ned Block, George Boolos, Noam Chomsky, David
Sarah Patterson and Paul Pietroski for such
Jay Lebed,
Kirsh,
contributions to 'Seeing What is not There'.
'. Vision Laboratory
- Ellen Hildreth, of the M. I.
informing me about the computational theory of vision.
for
patiently
David
- My teachers at Oxford: John McDowell, Chris Peacocke,
Pears and Peter Strawson. What philosophical wherewithal I now possess
is largely owed to them.
- Colin McGinn,
my first teacher,
who taught me how to think.
- Gottlob Frege for sense and reference.
- Tyler Burge,
to whose work I seem to turn at every turn.
for
Silva and Debra Stephenson
Marilyn
- Concepta Siembab,
generous and friendly help with various of the practical details involved
in getting a dissertation done.
- Dan Leary for keeping word-processors,
printers and disks in
working order, and making them write the dissertation for me, as well
as for philosophical discussion and comment.
- My parents for having me and keeping me.
I echo Michael Dummett in pointing out that although only I can take
responsibility for the contents of the thesis, I may not in fact hold sole
responsibility for any errors that appear in it.
-4-
TABLE OF CONTENTS
PAPER ONE: IN DEFERENCE TO REFERENTS
8
1 Introduction
9
1.1 Methodological Remarks
2 Frege's Problem
10
13
2.1 Constraints or a Solution
14
3 The Problem of Anaphora
18
4 Dummett's Solution
19
4.1 Criticism of Dummett's Solution
21
26
5 The Kaplanesque Proposal
5.1 Criticism of the Kaplanesque Proposal
28
32
6 Davidson's Account
6.1 Criticism of Davidson's Account
7 Higginbotham's Account
33
37
7.1 Criticism of Higginbotham's Account
8 Frege Vindicated?
41
45
-5
-
PAPER TWO: WHO'S AFRAID OF NARROW MINDS?
49
1 Introduction
50
1.1 background
51
1.2 Conceptions of Content
52
2 First Argument
59
2.1 Reply to First Argument
59
S62
3 Second Argument
3.1 Reply to Second Argument
4 Third Argument
64
65
4.1 Reply to Third Argument
67
4.2 A Modified Burgian Treatment of Proper Names
71
76
5 Fourth Argument
5.1 First Case: Perception
76
5.2 Reply to First Case
78
88
6 Second Case: Reference
6.1 Reply to Second Case
89
6.2 Points of View
91
- 6 -
PAPER THREE: SEEING WHAT IS NOT THERE
96
1 Introduction
97
1.1 Burge's Claim
97
1.2 Preview
99
2 Marr's Theory
99
2.1 First Point
103
2.2 Second Point
104
2.3 Third Point
1.06
3 Burge's Argument
109
4 Against Burge's Strong Claim
114
4.1 The First Case
114
4.2 The Straightforward Interpretation
114
4.3 The Devious Interpretation
117
4.4 Recapitulation and Remark
121
5 The Weaker Claim
123
5.1 Second Case
123
5.2 Undermining of the Devious Interpretation
125
5.3 Defence of the Straightforward Interpretation
128
6 A Specific Example
131
7 Exegesis
136
-7-
PAPER ONE
IN DEFERENCE TO REFERENTS
The wheel's still in spin
And there's no tellin' who that it's namin'
- Bob Dylan
In Deference to Referents
1 Introduction
sentences,
derive
for a language,
theory
The chief purpose of a semantic
how the semantic features of whole sentences
is to explain
from those of their parts.
Such explanations
rules that allow for the calculation of the
the semantics of propositional attitude attributions,
of
these
and (2)
sentences,
and
the
a set of
or logical form,
and
The topic of this paper is
the values of its semantically simple parts.
like (1)
require
value of a complex
;emantic
expression on the basis of its semantic structure,
that' sentences,
or a set of
particularly 'believes
I want to know the logical form
below.
conditions
that
the
deLermine
semantic
values of their constant parts, 'believes' and perhaps 'that'.
Two enterprises might lead a philosopher to seek for a semantics for
One
'believes that' sentences or indeed for any other sort of sentence.
such enterprise,
associated
with Frege,
and these days with Quine,
is
the development of a new language more suitable for scientific purposes
than the natural language which is the initial object of study.
with
such
sentences
a
will
concern
be
the
guided
by
choice
the
of
a
semantics
suitability
-9-
of
the
for
For one
'believes
resulting
that'
canonical
idiom - that is, the idiom that perspicuously displays logical form - for
use in a psychological theory.
with in this paper,
The second
words:
'the
however, is the other one.
enterprise
task
The enterprise that I shall be concerned
one Davidson
is the
of a theory of meaning
change, improve or reform a language,
it'.
1
distinguishes
as I conceive it,
with
is
the
not to
but to describe and understand
In seeking for an account of the semantics of attitude attributions
I seek to understand the workings of the actual English sentences.
The main purpose of this paper is to examine a number of alternative
accounts of the semantics of English 'believes that' sentences.
My aim is
partly to promote the attitude expressed in the title of this paper,
deference to referents,
with that attitude.
and tentatively
a
to offer a proposal in keeping
But my aim is also to enhance our grasp of what a
good explanation in the theory of meaning is,
by looking at the kinds
of evidence and arguments that may be relevant to it.
1.1 Methodological Remarks
Before launching into an examination of concrete cases,
I want briefly
to discuss in the abstract the question of what kinds of evidence
will
be relevant to the evaluation of specific theories of the logical form of
natural language sentences.
theory,
What is essential to a semantic
I.
Davidson,
D.,
'Truth and Meaning',
-
10 -
in the first place,
Synthese xvii,
1967.
is that it
succeed
in
correctly
the
predicting
observable
semantic
features
The form that such predictions take will depend
whole sentences.
what kind of theory one is working
with.
So
for example,
of
upon
a theorist
working within a Davidsonian theory of meaning - and I take myself to
be such a theorist - will try to prove the familiar T-sentences:
true iff p',
where S is a structure-revealing
description
of an object
and p is a statement, in the theorist's metalanguage,
language sentence,
of exactly the conditions under which
that object-language
sentence is
When we are
true. What is crucial here is to get the T-sentences right.
in the fortunate
'S is
position of constructing a theory of
own home language,
the evidence is not hard
hard to tell when a T-sentence is correct.
meaning
to collect,
for our
for it is not
The trick is to construct a
theory that proves it.
What
further
answerable
to,
evidential
constraints
will depend
one
upon exactly
wants
the
theory
what is to count
to
as a
be
good
description and explanation of the linguistic phenomena under scrutiny.
A methodological
knowledge
conservative might seek only
possession
of
which
would
theories
employ different
each other.
to
be
that
prove
axioms and
exactly
mechanisms,
They both get the right
achieved,
but
they
do
so in
need to seek for further evidence
-
to
interpret
utterances
For the methodological conservative
made by speakers of the language.
two semantic
suffice
to articulate a body of
the
same
are likely to
results.
different
-
but
be as good as
Each achieves what needs
ways.
that would count
11
T-sentences,
There
is
then
no
for one theory and
against the other.
From the
hand,
point of view of the
there is in principle
methodological
liberal,
on the other
no end to the kind of evidence that might
be relevant to the evaluation of a semantic
theory.
sorts of general constraints on good explanation,
There are also all
that can be applied in
semantics just as they can in any other domain.
I belinkve that there is much to be seen from the liberal perspective,
that might otherwise remain hidden.
advance of a search,
theory,
of time.
may
Since I see no way of saying,
in
what might count for or against a given semantic
I see no reason to think that such a search would be a waste
On the contrary,
count
illumination,
for
or
I think
against
that a detailed examination of what
particular
proposals
both of the natural language itself,
can
only
bring
and of the exigencies
of understanding how that language works. 3
The liberal position will stand or fall by the quality and amount of
the results yielded by research that follows its guidance.
This paper as
a whole may therefore be taken as a partial defence of liberalism.
2. The archetypical methodological conservative is Davidson. See in
particular his 'Belief and the Basis of Meaning', Synthese 27, 1974. See
also Quine, W. V. O., 'Methodological Reflections on Current Linguistic
Theory', in Harman, G., and Davidson, D., eds. The Semantics of
Natural Language, Humanties Press, New York, 1972, for a different
version.
3. For a defence of a liberal position, see Chomsky, N., Reflections on
Language, Pantheon, New York, 1975, and my 'The Topiary Theory of
Language', in Fukui, N., Rapoport, T., and Sagey, E., eds., M. I. 'T.
Working Papers in Linguistics Vol. 8.
- 12 -
2 Frege's Problem
We turn now to semantics.
(1)
Ralph believes that Tanya is a terrorist
(2)
Ralph believes that Patty
is a terrorist
Patty and Tanya are the same person.
Let us suppose that Ralph is
ignorant of this, and that he has heard,
from what he considers to be
a reliable
firmly
source,
believes
that Tanya
that
Patty
is a
is
no
terrorist.
terrorist.
On the other hand,
In
some
he
circumstances
it
would be wrong for me to report Ralph's belief as the belief that Patty
is a terrorist.
Let us suppose that the wrongness of the report amounts
to its falsity,
rather than, say,
its being merely misleading,
or lacking
in explanatory value.
It should be noted
could
hold
that,
codenotational
4
that such a supposition
contrary
expressions
to first
appearances,
in
content
the
attitude reports do preserve truth value.
would
then
considerations.
be
explained
is not mandatory.
by
intersubstitutions
sentences
of
One
of
propositional
Appearances to the contrary
pragmatic,
rather
than
semantic,
It would be worth exploring the consequences of such a
4. I shall use 'denotes' and 'has as semantic value' interchangeably, to
describe the relationship that holds, in extensional contexts, on any
roughly Fregean theory, between an expression and its semantic value.
My use of 'denotes' is not standard, but is easily understood.
- 13 -
view.
5
I take it that the view is
But that is not my present purpose.
.hat
and it is those
than the alternatives
prima facie less plausible
I
wish to explore in this paper.
and (2,
So (1)
in spite of the fact
may have different truth values,
that they are composed of expressions with the same denotations in the
same
denotations,
things;
reports)
attitude
propositional
sentences
content
inside
embedded
this
to
response
Frege's
order.
was
to
claim
that
(that
is
the
'that'
not
do
That is,
but rather denote their senses.
inside content sentences
one; expressions
two;
and
normal denotations,
denote
there
clauses
of
normal
their
Frege claimed two
do not denote their.
inside content
expressions
expressions
sentences
do
denote their normal senses.
2.1 Constraints on the Solution
Let
us
follow
in
Frege
supposing
'Tanya' and 'Patty' have some feature,
normal denotation,
truth
values.
that
"sense",
like
expressions
embedded
that differs from their
and that explains how (1) and (2) can have different
The
first
embedded expressions
question
contribute
I
want
to
ask
is
this:
How
do
their senses to attitude attributions?
In particular I want to know if it is necessary that senses play the role
of denotations of the embedded expresssions.
5.
For some such exploration see Salmon,
Press,
6.
Cambridge,
N.,
Or is it possible that the
Frege's
1986.
1 owe the idea of this paragraph to Crispin Wright.
- 14 -
Puzzle,
M.
I.
T.
embedded expressions contribute their senses in some other way?
It might look as though
the answer is mandated
by the
roles that
sense and denotation are designed to play.
What role does an expression's denotation play in the semantics? From
the point of view of a semantic
contributes
to
the
which it appears,
7
determination
a.d
(2)
it just is what the expression
of the
truth
and So the truth value of
just of the denotations
trle of (1) and (2)
theory,
of its component
in particular.
value
of sentences
a sentence is a function
expressions.
Suppose this is
Then whatever the expressions in (1)
contribute to the determination of the sentences'
truth
will just be what the expressions denote in those sentences.
be true for,
inter alia,
in
the embedded expressions,
values
This will
like 'Tanya' in (1).
What is a sense? It is what an embedded expression contributes to the
determination of the truth value of the attitude attribution in which it
appears.
It
thus
looks
as
though
the
only
way
the
embedded
expressions could contribute their senses is by denoting them.
So it appears that there is only one possible answer to my question
'How do embedded
attitude
have
expressions
attributions?'.
another
look
at
contribute their senses to propositional
They
denote
the
argument,
something like it provides
the
them.
I think
because
underlying
it
motivation
it
is
seems
important
to
me
to
that
for many modern
7. C. f. Evans, G., The Varieties of Reference, Oxford University
Press,
1982,
second chapter,
and Davidson,
D.,
'Reality
Without
Reference', Dialectica 31, 1977. In my view Davidson makes somewhat
too much of the point.
- 15 -
Fregean treatments of attitude attributions: that is treatments that have
words inside
content
sentences denoting
than their normal denotations.
It
sentence
normal
senses,
seems simple enough:
of an attribution is a function of,
embedded expressions.
their
rather
the truth value
among other things,
the senses of
Since it is also true that the truth value of a
is a function only of the denotations of its components,
senses of the embedded expressions
must,
in those contexts,
the
serve as
their denotations.
But the argument is invalid.
(3)
and
(4)
attributions.
e
denotations,
are
schematic
and
e'
so (3)
representations
different
of
senses
but
opaque
the
attitude
same
may have a different truth value from (4).
infer from
this,
least
component
one
have
and (4).
'...e'...'
(4)
'...e...'
(3)
Consider (3)
by
the Fregean
expression
component expression of (4).
principles,
in
(3)
normal
We may
that the denotation of at
differs
from
that
of
one
But it is not entailed that the expressions
that have different denotations must be e and e'.
Perhaps,
for example,
something else in the complex denotes the sense of e or e'.
That would
account for the opacity effects without our having to deny that e and e'
there denote their normal denotations.
Or perhaps there is some other,
more devious and complicated manner
in which the denotation of some
part of the whole
attribution is
affected by the senses of the embedded
expressions.
lSo
perhaps
expression
in
the
a
right
content
answer
sentence
to
our
question
contributes
- 16 -
its
is
sense
just
to
this.
An
the truth
value of the attitude attribution just by having it,
and the denotation
of something else in the attribution is affected by it.
There is thus the possibility
of using
senses to explain the opacity
effects in a manner slightly different from Frege's.
But does this idea
get us anywhere? Is there anything interesting to be done with it? That
depends
proposal,
upon whether
that is,
there is in fact some way of implementing
some way of actually doing the semantics.
the
We would
have to find some way of allowing the sense of one expression to be, or
somehow
to
determine,
the
denotation
of
another,
a
that
way
successfully predicts the opacity effects.
One reason why one might feel inclined to search for an alternative
account along those lines is this.
appears that expressions
There are various ways in which
even as they are embeddea inside a content
sentence,
are denoting their normal denotations.
sense
an
of
embedded
it
expression
without
being
to
affect
attitude
report
context,
then we could leave the expression
the
If we could allow the
the
expression's
truth
value
denotation
of
in
the
that
free to denote its normal
denotation, without losing our account of the failures of substitutivity.
In
the
next
section
I
shall
present
some
linguistic
evidence
that
suggests that embedded expressions do denote their normal denotations.
Then I shall investigate
evidence,
how well a Fregean theory might assimilate the
and whether some alternatives
- 17 -
fare better.
3 The Problem of Anaphora
The
linguistic
denote
their
among
expressions
evidence
normal
that
indicates
denotations
comes
that
embedded
from cases
expressions
of cross-denotation
inside and outside content sentences.
The clearest
examples are provided by what appear to be straighforwardly anaphoric
pronouns, like that in (5)
(5)
Betty believes that Barbarai is bright,
and she. is bright 8
1
In
(5)
'Barbara'.
determined
it looks as
'she' is straightforwardly
to
be
whatever
is denoted
this with the subscripts
of 'Barbara'
in
(5)
is a
by
and
that a person cannot
'Barbara'
cannot
denote
sense.
be a sense,
the
same
its antecedent
'iV. On the
though the denotation of 'she' is not a sense,
Assuming
anaphoric
But
it
Fregean
theory
certainly
the
one
the
looks
as
Barbara.
we have a problem:
if
I
'Barbara'.
but the person,
thing,
with
of 'she' is
That is to say it looks as though the denotation
have marked
denotation
though
'she'
denotes
a
person, and the other denotes a sense.
8. 'She is bright' is meant to be outside the scope of 'believes'. In
future examples possible ambiguities are always to be resolved in that
direction, so that the conjunction takes one of its conjuncts outside the
scope of the doxastic operator.
- 18 -
4 Dummett's Solution
9
Dummett,
notices
the
among
coreference
apparent
of
problems
like that illustrated in
expressions inside and outside content sentences,
He offers the following solution on Frege's behalf.
(5).
Since the account of opacity effects depends upon assigning a sense
to 'Barbara',
And Dummett wants to preserve
that is what we must do.
the intuition that (5)
contains the semantically monadic predicate (6):
Betty believes that x is bright and x is bright
(6)
a predicate that can only be satisfied by a single thing.
then
expressions
the
holders
'x',
is that
therefore,
denote
must
the
sense
'Barbara'
have
the
same
'she' denotes
of
the
and
'she',
denotation.
the person
expression
in
standing
What
for
the
it
Instead
which
is
place
denies,
Dummett
Barbara.
'Barbara',
If that is so,
the
must
latter
expression's denotation in (5).
The little sentence 'she is bright' in (5)
property of being
tries
to
bright to a sense.
now seems be attributing the
That cannot
rectify this by adjusting the semantic
outside the content sentence,
be right.
values
Dummett
of other things
so that overall it ends up with the right
truth conditions.
9.
Dummett,
1973,
M.,
Frege:
Philosophy of Language,
ch.9.
- 19 -
Duckworth,
London,
Dummett's first suggestion is that we assign to the predicate,
bright' as it
a special
sentence,
semantic
This special semantic value is a function that maps onto the True
value.
just
content
the
occurs outside
'x is
those
that
senses
are
of
senses
the
expressions
denotations
cf
which are mapped to the True by 'x is bright' in normal contexts.
In
occurring
predicates
then,
general,
in
contexts
wherein
they
but are predicated of an expression
appear outside a content sentence,
the value of which is determined within a content sentence,
get special
semantic values. These values are functions mapping onto the True just
the denotations of which
senses of expressions
that are
those senses
are mapped onto the True by the predicate in normal contexts.
utterer of a sentence like
(5)
something
will be saying
if I might be so bold as so to put it,
just in case,
Now the
that is true,
what she thought
she was saying was true.
But Dummett
finds
The view
himself unable to rest with this view.
required the assignment of special semantic values to predicates outside
content
sentences,
predicates,
can
have
in
the
like
pronouns,
their
semantic
cases
awkard
we
and
bound variables,
values
are
determined
considering.
subject
from
But
expressions,
within
a
content
sentence.
Suppose
that 'John is
here' is true.
here and
Dummett writes (ibid.
I did not expect that John
p.
276)
would be
that from this
we can ... infer 'John is something that I did not expect he
which is to be analyzed as
here),
would be' (namely,
'(3F)(F(John) and I did not expect that F(John))'.
- 20 -
just as before.
And now the argument continues
content
denote
Those inside
have
will
sentences
th3ir
to
codenotational
be
outside
those
inside.
with
must denote their
so those outside
senses,
Predicates
senses as well.
We must conclude that predicates like 'x is a bright',
in both their occurrences,
of
assignment
the
abandonned.
denote,
But the sense of a predicate is
their senses.
The proposed treatment which depended
not its special semantic value.
upon
in (5)
So, as Dummett says (p.
then,
must,
values
semantic
special
be
of a context like the second
277)
conjunct of (5):
we have to see the entire context, apparently transparent
but occurring within a sentence in which an opaque context
also occurs, as in reality opaque, with all expressions
contained in it having their indirect reference [=denoting
their senses)
How
can
utterer
the
see
suggests
that
we
operator
such as 'It
thought,
and 'It
of
(5)
sentence
the whole
is true
say
now
that'.
'That
is true' works as a
anything
as
at
prefaced
an
by
she is bright' just
predicate
Dummett
all?
the semantic
implicit
names
a
value
of
I have
a
which is a function that maps true thoughts onto the True.
What
are
general
we
to make of
expression
of
this
emended
skepticism,
a
Fregean
technical
account?
objection,
and
a
methodological criticism to offer.
4. 1 Criticism of Dummett's Solution
Here is
the general expression of skepticism.
- 21 -
It is
very
hard to see
what
utterances
ones,
present
upon
for
allow
could
mechanism
the
of
kind
after pausing
then,
he
on Dummett's
one in just the way that spreads opacity,
'Fred is the smartest student',
my
telling
about
psychological
Fred's
at t,
and then
Fred,
this to
states
t,
compatible
with
two
theory,
uttered
I had
not gone on to say anything
had
at
t'.
then
Well,
of
course
my
We have one initial utterance,
at
developments
It
t'.
is
overwhelmingly
natural to think that the semantics of 'Fred is the smartest
completely settled at t.
it will
paused to consider the wisdom of
utterance would have been transparent.
at
that
the other hand,
on
Suppose,
as well.
the
Since the first utterance is linked to the second
second word onwards.
be opaque
the
sense.
from
opaque
being
utterance
second
to my
is committed
at t',
hold off
cannot
Dummett
conclusion that my initial utterance of 'Fred' denotes its normal
For
future
I go on to say,
to consider the wisdom of my telling this to Fred,
he knows that he is the smartest'.
of
proposing.
is
Dummett
at t, 'Fred is the smartest student',
Suppose I say,
'And
effects
retroactive
the
And that is incompatible
student' is
with the consequences
of Dummett's theory.
There are,
of course,
many ways in which I can utter an expression
then later say something to pin down a denotation
for
e,
ambiguously,
it.
This happens with such mundane sentences as 'Before he spoke the
King cleared
his throat'.
But it
is
really determinate at the moment of utterance.
is
that the epistemic
utter e
mind,
with
and
the
the denotation
plausible that
All that can happen later
state of the audience may be improved.
intention
disambiguate
so
of denoting
as
to
make
- 22 -
one
it
of e is
thing,
look
as
a,
then
though
Even if
1
change
my
I meant
to
denote
b,
another,
epistemic
state of
be
all this could
I really
audience.
the
for
accounted
in
a,
denoted
terms
but
of
the
brought
it
about that my audience thought I was denoting b.
moreover,
The difficulty,
becomes aggravated: consider the dialogne:
Me: 'Fred is the smartest student'
Other: 'And he knows that he is the smartest'
On Dummett's theory,
both tokens of 'he' must denote a sense.
Is there anything which
to 'Fred'.
the first 'he' looks to be arnaphoric
But
prevents our familiar argument from implying that my utterance of 'Fred'
For there is no visible feature of this
denotes a sense? It appears not.
case that distinguishes it semantically from the case where I say it all
myself.
It is the
serial
different speakers,
utterance of
which brings
two
parts of the
about the
sequence,
final predication
by
of Fred.
But I do not see a way of exploiting this so as to allow that 'Fred' and
'he' do not share a dtnotation.
For what Other says stands in exactly
the semantic relation to my utterance,
as it would have if I had said it.
But surely the denotation of my
upon what Other
goes
on to
say.
utterance of 'Fred'
Such
a consequence
cannot depend
must
make
a
reductio ad absurdum of any view from which it issues forth.
A
defender
sentences
and
thought,
then
are
there
opaque
or
of
Dummett
really
the
would
might
composed
respond
of an opaque
by
suggesting
component
implicit 'it is true
that' operator.
be
whether
no
transparent,
question
which
of
question
what happened later.
- 23
-
would
a
If
given
have
to
that
that
names
that were
utterance
be
all
settled
a
so,
were
by
But
reply
the
up
gives
We
much.
too
expressions never denote ordinary
can
hardly
external objects,
that
suppose
without losing our
already precarious grip on how we manage to say things the truth of
which depends upon how things are in the world outside.
has something very like this point in mind,
And Dummett
when he writes (op.
cit. p.
198)
The denotation of an expression is its extra-linguistic
correlate in the real world: it is precisely because the
expressions we use have such extra-linguistic correlates
that we succeed in talking about the real world.
Let us move on to my second objection to the Dummettian extension of
Frege. Let
(7)
be
Sandra believed she was going to be late
uttered
by
other
someone
thian
On
Sandra.
Dummett's
account
in virtue of its relation to an opaquely embedded 'she',
'Sandra',
denote a sense.
denotes
the
supposes,
But
thought
because
which
that
Sandra
each of
In
sense?
'she
(7)
believed.
was
Moreover
denotes
its components
must
going to be late'
it
does
a sense,
so,
one
and
the
senses denoted are identical to the senses that compose the thought to
which Sandra is related.
So 'she' should denote the sense that stands in
subject position in the thought that Sandra believed.
the sense of the word 'she',
mouth.
This will not be
but the sense of the word 'I'
For Sandra had a first person belief about herself.
in Sandra's
But 'Sandra'
in someone else's mouth cannot denote that sense.
Dummett
anaphora.
might
perhaps
respond
The apparent anaphoric
by
weakening
relation in
- 24 -
(7)
and
the
theory
the like would
of
be
reconstructed
very
so that the expressions in question would not denote the
same Eenses,
denotation.
In
speaker
think
to
but
(7)
would
'Sandra'
of
would
Sandra,
featured in Sandra's
denote
senses
denote
and
own thought
'she'
that
the
sense
would
about
present
A believes
--- x--.-' and its fellows
fide monadic predicates.
in
these
cases,
expedient
of
the
same
the
sense
that
But to concede
this
would be to undermine the whole Dummettian approach.
'...x...
that enables
denote
herself.
the
On this proposal
would no longer be bona
And once one is prepared to give up anaphora
they
can
assigning
be
accounted
indirect
for
without
denotations
to
content sentences (see the Kaplanesque proposal,
the
problematic
expressions
outside
explained below,
next
section).
Perhaps neither of my two objections
But
even
so,
compliCations
appearing)
the
theory
(wholesale
looks
shifts
poor:
of
seem
highly
ad
Contexts outside
clauses
governed
exhibit
the
features
Intersubstitutions
inside them,
we
hoc.
would
to Dummett are unanswerable.
it
is
complicated,
denotation,
They
by
expect
of codenotational
preserve truth (try it).
lack
and
suppressed
independent
the
operators
motivation.
opacity inducing
verbs
do not
of
were
opaque.
different
senses,
them
expressions
if they
with
The hypothesized wholesale shifts
of denotation thus have no detectible properties.
So let
us put
aside
Dummett's
version
look for something better.
- 25 -
of the
Fregean
proposal
and
5 The Kaplanesque Proposal
Let us backtrack to the beginning of Dummiett's
story, and see if the
Fregean has other possibilities. The problem was that in (5)
that while 'Barbara' and 'she' are codenotationial,
and
the
other
denotes
a
person.
Dummett
it appears
one denotes a sense
tried
to
deny
that 'she'
denotes a person. But perhaps it is better to deny instead that the two
expressions are
they
denote
really
things
codenotationial.
that
stand
in
Could it
some
not
special,
be
that,
rather,
and
specifiable
content
sentences
relationship to each other?
The
idea
would
be
to
have
expressions
inside
denoting senses, but use some relation that holds between those senses
and the expressions' normal denotations,
to account for cross-referring.
How would this go?
Let us try (8):
(8)
R('Tanya', Tanya)
understanding
it
to
mean
something
like
'the
represents Tanya'.
We can exploit (8)
(9)
in a Fregean version of (1):
R('Tanya', Tanya) and Ralph believesd
rTanya is a terrorist
- 26 -
sense
of
"Tanya"
The
corner
expression
quotes
takes
a
create
sense
as
an
oblique
context
its
semantic
value.
within
The
which
corner
each
quoted
'Tanya is a terrorist' now denotes the sense of the sentence 'Tanya is a
terrorist'.
'Believes
d'
expresses a two place relation that holds between
believers and the senses or "dicta" of their beliefs.
Anyone
familiar
with
David
Kaplan's
paper
'Quantifying
recognize that I have constructed (9)
under its influence.
kind
'The
of
treatment
illustrated
in
(9)
Kaplanesque
note that Kaplan himself made no such proposal,
in another way,
In'10
will
I will call the
proposal'.
But
but used the materials
for another purpose.
With all this apparatus in play it easy enough to cope with anaphora,
thus
(10)
Ralph believes that Tanya. is a terrorist,
1
but she. is too honest
becomes
(11)
R('Tanya', Tanya i ) & Ralph believesd
rTanya is a terrorist7
but she, is too honest)
1
with 'she i
'
codenotationial
with
within the content sentence,
denoting a sense,
'Tanya '.
One
occurrence
of 'Tanya'
and there serves the Fregean
and another occurrence
is outside,
purpose of
and allows for the
10. Kaplan, D., 'Quantifying In', in Davidson. D., and
eds., Words and Objections, Reidel, Dordrecht, 1969.
- 27 -
is
Uintikka,
J.,
straightforward treatment of anaphora.
logical
of
form
expression
surely
It
(10).
no
occurring
'Tanya'
candidate
is a serious
that (11)
however,
I do not think,
be
cannot
the
the
necessary
to
have
times
in
a
three
than
less
for
correct
rendition of a sentence in which it occurs only once.
The underlying difficulty with (11)
'R('Tanya',
Tanya)',
and therefore
expression
'Tanya'
represents
is that it includes the clause (8),
actually says that the sense of the
something that is literally said by the sentence
to
seem
not
does
This
Tanya.
be
Rather it is the
(10).
fact that the sense of 'Tanya' represents Tanya that allows a hearer to
because (8)
is true,
but it does not say that it is true.
The sensible course at this stage would be to pull (8)
it can easily
appears in the theory,
not
the
be exploited
We would simply take
denotation
of its antecedent
out of (11)
As long as (8)
and place it somewhere else in the theory of meaning.
anaphora in (10).
functions
So (10)
ascertain the correct denotation of the pronoun 'she'.
to account
for the
the denotation of 'she' to be,
'Tanya',
but
object
the
that
is
represented by that denotation.
5.1 Criticism of the Kaplanesque Proposal
But how, exactly,
is (8)
must either be an axiom,
representation
relation
method of deriving,
going to figure in the theory of meaning? It
or derivable fromn
R is
to be
usable
for every embedded
- 28 -
some axioms.
in
general,
expression
Indeed,
then
we
if the
need
that determines
a
an
extension,
by the sense of that
what is represented
expression.
That
we need a general method of obtaining clauses of the form (12):
is,
(12)
R(e, b)
where e is an embedded expression and b is the object that the sense
of e represents.
But it is not easy to arrive at these clauses.
Consider
for example
(13):
(13)
John said that he loves [his wife.
1] ,
but he is rude to her.
How
we
are
denotation
going
of 'her'?
get
a
R('his wife',
that
allows
The simple disquotational
will clearly not work here.
(14)
clause
us
axiom
recover
suggested
by
the
(8)
What would we do with (14)?
his wife)
'His wife' unlike 'Tanya',
does not have a fixed denotation,
have its denotation determined by the context.
but will
The theory of meaning
thus cannot simply state which object is represented
'his wife'.
to
by the sense of
Rather it will have to contain a rule that specifies,
for each
context, which object the sense of 'his wife' represents in that context.
And
new
complexities
are
presented
constructions like those illustrated
(l5)(i)
John said he has
in
(15).
[some sheepij].
Olga will vaccinate themi
.
(ii) John said he has [a new wifei].
You will meet heri
.
- 29 -
by
quantifier/pronoun
Which relation holds between the sense of 'some sheep' in (15)(i),
that 'thenm' denotes?
the sheep
And
sense of 'a new wife' in (15)(ii),
I have
Fregean
no doubt
could
extensions
what
with senses.
is
the
the
and the denotation of 'her'?
appropriate
hard work the dedicated
mechanisms
for
But I think that if we reflect
required
between
relation holds
that with ingenuity and
construct
information
whi-h
and
to
get
us
to
the
associating
upon precisely
right
extensions,
a
non-Fregean line will appear much more pronising.
In each of our examples the appropriate extension for the anaphoric
pronoun is recoverable from the semantic interpretation that the content
sentence
as
a
transparent,
(10)
whole
would
had,
had
it
been
and given its normal semantic interpretation.
denotes
denotationially
what
in
would
'Tanya'
transparent.
would have denoted had (13)
Similarly
have
(15)(i)
and
And
'her'
have
in
denoted
(13)
denotes
as
Thus 'she' in
had
(10)
been
what 'his
wife'
been transparent.
(ii)
the
appropriate
extensions
pronouns would be recoverable
from their antecedents,
entire content
normal
sentences
treated
their
transparent
for
the
if we gave the
interpretations.
the
right
extensions from transparent treatments of the content sentences,
as is
the
theory
of meaning
will
have
the
resources
to
recover
For
apparent from (16):11
11.
For an account of these resources see Evans,
G.,
Quantifiers and Relative Clauses (1)', Canadian Journal of
vii,
1977.
- 30 -
'Pronouns,
Philosophy,
(16)(i)
John has [some sheepi].
Olga will vaccinate them i
(ii) John has [a new wife i J. You will meet her..
Knowledge of what the interpretation of the content sentences would
have been,
had they been transparent,
is thus sufficient
to determine
the extension of the anaphoric pronouns.
It is clear also that in some cases,
the examples (15)
nothing less than that will do.
you have to know what the transparent interpretation
of the entire content sentences would have been,
out
the
denotation
of the
unembedded
that 'them' denotes some sheep.
a
transparent
before you can figure
pronouns.
figure out the extension of 'them' in (15)(i)
give
In
It is
impossible
to
without first ascertaining
And you will only figure that out if you
interpretation
to
the
expression
'some
sheep'.
Moreover you will only figure out that the precise sheep in question are
the ones that John owns if -iou interpret the two expressions
'he' and
'owns' as if they had occurred transparently.
Similar remarks apply to (15)(ii).
the
denotation
of 'her' if
you
You will only be able to ascertain
interpret
each
element
in
the
content
sentence as if it occurred transparently.
To account for the anaphora in the sentences (15)
do
is
thus
transparently.
precisely
equivalent
to
treating
what we need to
the
for
the
contexts
Nothing more and nothing less than this is required.
It seems then that rather than attempt to formalize
representation
whole
relation,
failures
of
R,
the Kaplanesque
we should look for some way of accounting
substitutivity
-
while
31 -
yet
allowing
the
embedded
expressions to denote their normal denotations.
6 Davidson's Account
The
point
which
started
off
whole
this
discussion
of
alternative
was that if we could allow the denotation of one expression
proposals,
to be sensitive to the sense of another,
senses to block the
undesirable
then we might be able to use
substitutions,
while leaving
expressions free to denote their usual denotations.
embedded
But how could we do
this?
Donald Davidson's account of the logical form of 'says that' offers a
very attractive answer to this question.
Suppose that the logical form
of (1) is really (17):
(17)
Ralph believes that. Tanya is a terrorist.
with
'that'
as
a
demonstrative.
Davidson
himself
suggests
understanding of this that has no use for the notion of sense.
that is
one
already
have
of
the
main motivations
for the
the notion of sense in play,
proposal.
But
an
In fact
since
we
we might as well just have
the 'that' referring to the sense of the embedded sentence 'Tanya is a
terrorist'.
Now the opacity effects are easily explained.
Since the sense
of 'Tanya is a terrorist' differs from that of 'Patty is a terrorist',
may believe that
Tanya is a terrorist
but not believe that:
Patty
is
- 32 -
a terrorist
Ralph
in its own
Since the content sentence is also a sentence
right,
may assign to its component expressions their normal denotations,
we
and
among expressions inside
thereby solve our problems with codenotation
and outside content sentences.
Davidson's
manner,
seerms
is,
even
proposal,
as I said,
the
to read
very nearly
as
the
unDavidsonian
The reason it is attractive is that it
attractive.
semantics
in
interpreted
off the
explicitly
surface
structure,
On Davidson's account it is easy to see exactly
so.
But the whole story could not be so simple and elegant.
too
much
to
hope
for.
And,
what
how each contributes to the whole.
each familiar English word is doing,
be
or
indeed,
Davidson's
idea
That would
does
have
prima
facie
various problems.
6.1 Criticism of Davidson's Account
Davidson's
plausibility
proposal
the
from
perhaps
fact
that
derives
a
the indirect
demonstrative 'that' are spelled similarly.
by noting that in many languages
complementizer
completely
languages
of
different
(Japanese,
demonstrative,
does
indirect
(e.
not
even
have
and
the
- 33 -
discourse
and
'that'
the Romance languages)
and
Korean)
of
the
But this is quickly dispelled
g.
discourse
orthographical
Hebrew,
modicum
the
the
demonstrative
phonetic
forms.
complementizer,
superficial
form
of
In
unlike
a
the
have
some
the
separate
word, but rather of a mere affix.
English
also,
in
demonstrative
'that'
has
In
'believes
that'
spite
of
the
a
cluster
Notice
not.
of
first
Ralph believes
Dat
of
pronounced
however,
or destressing.
seems
One cannot
look atbat', or:
*Ralph believeshat
which interprets (18)
Second,
are
The demonstrative 'that',
never to undergo this phonetic reduction,
(19)
they
'that'
Ortcutt is a spy
•3•'.
shortening the 'that' to
say 'Hey,
that
the
that
properties
the
coincidence,
orthographical
One usually says
differently.
(18)
does
12
. Ortcutt is a spy.
in Davidson's way.
the 'that' of 'believes
that' and 'says that' is deletable.
One
can say:
(20)
Lemmy said he needed money
But the demonstrative 'that' cannot be dropped.
with the 'that' dropped.
You cannot say (21)
(21)
*Yes ...
is what I said
with the 'that' deleted,
(22)
instead of (22)
Yes that is what I said
12. German and Finnish,
the relevant respect.
on the
other
- 34 -
hand,
are
more
like
English
in
Now
there
generally
is
not,
accepted
to
explanation
involved.
It
demonstrative
referring
will
looks
knowledge,
explanation
invoke
as though
And
well-understood
of what determines
the
semantic
the
reason
pronoun 'that' is precisely
function.
any
I believe
phonological characteristics.
that
role
of
why one
because
similar
and
the possibilities of
But it is very plausible that,
deletion and destressing.
the
my
in both cases,
the
expressions
cannot
it has
delete
an
remarks
the
important
apply
to the
Stress often goes with semantic role; and it
is hard to see what else could account for the stressing rules in this
case.
13
Although
conclusive,
these
they
objections,
are
very
taken
well
by
worth
example of how semantics can interact
The
'that' of 'says
demonstrative
true
that
be
pronoun cannot.
the
demonstrative,
that' can
'that'
as
of
Davidson
themselves,
noticing,
and
destressed,
This is simply
that'
holds,
they
empirical
functions
then
it
or not is an empirical question
various parts of linguistics.
from
provide
an
must
the
English
data.
If it
is
semantically
like
a
be
false
explains the phonological and syntactic facts is semantic
this is so
far
with other parts of linguistics.
deleted
'says
for
are
that
role.
at the intersection
what
Whether
of the
The examples provide a nice demonstration
of methodological liberalism at work,
evidence that may help us choose
and point towards a rich field of
between competing
semantic theories,
13.
For a discussion of these and other syntactic difficulties
Davidson's proposal, see Speas, M., and Segal, G., 'On Saying
in Mind and Language, 1986.
- 35 -
with
jt ',
should we have such aspirations.
Another,
the
and more serious difficulty with Davidson's proposal is that
English
sentences
do
not,
in
important
semantically as they would if they were paratactic.
(23)
behave
Thus
Everyone said that they loved their mothers
is open
to two
interpretations.
It could mean
something like 'They loved their mothers',
with
(or
bound
individuals.
(24)
(24)
respects,
by)
'Everyone',
but
that
each
person
said
where 'they' is not anaphoric
refers
to
some
other
group
of
But it can also be readily understood as
Everyonei said that they i loved their mothers
means
that
each
person
claimed
of
herself
mother (saying something like 'I love my mother').
that
she loved
her
But the Davidsonian
rendition of (23)
(25)
Everyone said that. They loved their mothers.
is only interpretable in the first way.
inscription
of 'They loved their
(Imagine someone pointing to an
mothers' on a blackboard,
'Everyone said that!'. How would you interpret this?)
The
point of the example is
not that
samsaying relation that would allow
that (23)
is not,
(25)
Davidson
and saying
14
could
not define
to mean the same as (24).
as a matter of empirical
fact,
understood
a
It is
by English
14. This kind of example is due to Higginbotham, J., 'Linguistic Theory
and Davidson's
Program in
Semantics',
in
LePore,
E.,
ed.,
The
Philosophy of Donald Davidson: Inquiries into Truth and Interpretation,
Oxford, Basil Blackwell, 1985.
-
36 -
speakers as we would expect if Davidson's analysis were correct.
The empirical evidence against Davidson's proposal is not conclusive.
But it is good enough evidence to warrant a continuation of our search
for a satisfactory account of our problematic pieces of language. 15
7 Higginbotham's Account
What we were looking for was some way of making the denotation or
extension
of
some
part
of the
attitude
senses of the embedded expressions,
denote
expressions
suggested
normal
their
attribution
depend
upon
the
without having to deny that those
We
denotations.
found
the
that
simple solution of just having the 'that' denoting the senses
was not entirely satisfactory.
But another way of doing it is suggested
by Higginbotham (op. cit.).
The proposal is that we assign values to expressions only relative to
the larger expressions of which they form a part. So, in general,
have
an expression
e inside a larger expression,
instead of assigning
to e a value all by itself,
relative
point
to
S.
simultaneously
The
appear
of
this
in several
is
that
different
if we
say a sentence,
S,
we assign to e a value
since
an
embedding
expression
expressions,
may
we
15.
There are also objections to Davidson that concern the logical
These sentences appear to have
properties of 'says that' sentences.
implications that they would not, if their logical form were paratactic.
For a recent example see Burge, T., 'On Davidson's "Saying That"', in
LePore ed. op. cit.
- 37 -
may assign to it a value relative to one that differs from the value,
if
any, that we assign to it relative to another.
This idea of assigning values only relative to embedding expressions
is useful not just for attitude attributions,
but can solve problems with
sentences of a completely different kind.16 This is important.
If it were
not so, then we would be introducing sweeping changes in the semantic
theory,
changes that affected what we said about every sentence,
to cope
with
happened,
the
special
problems
with
attitude
attributions.
If
just
that
one might suspect that the cure was more radical than the
disease required.
The success of the proposal may ultimately depend upon just this.
the
trick
expressions
of
making
proves
assignments
fruitful
vindicate its use here.
in
a
of
range
values
of
relative
different
If on the other hand it idles,
other areas needlessly cumbersome,
Let us see how the idea applies
to
If
embedding
cases,
that
will
and simply makes
that may cast doubt upon it.
to belief attributions.
Consider
(1)
'Ralph believes that Tanya is a terrorist', carved up as in (26):
(26)
[S1Ralph believes [S,that [soTanya is a
terrorist] ] ]
(26) is articulated into the three phrases,
S O, embedded in S', which
is in turn embedded in Sl
16. Specifically, certain sentence connectives, such as 'if' and 'unless',
express different truth functions in different linguistic environments.
See Hligginbotham op. cit.
- 38 -
In order to cope with cross-denotation
content
sentences
to have
their normal
parts of the content sentence
SO,
itself as if it were unembedded,
we
want expressions
denotations.
So we
'Tanya is a terrorist'
inside
treat
the
relative to S O
and appeared in isolation. This means
that the expressions inside the content sentence just get their normal
denotations as desired.
This is expressed in (27).
altered Higginbotham's notation,
(27)
(i)
<'Tanya',
(I have significantly
for ease of assimilation).
SO> denotes Tanya
(ii) (Va)(a satisfies <'x is a terrorist', S 0 > iff a is
a terrorist)
from which it is easy to derive
(iii) <S O , SO> is true iff Tanya is a terrorist
The angle brackets are to be taken literally,
so that we are technically
assigning values to ordered pairs of expressions.
just understand
<'Tanya',
S0>
to mean,
But intuitively we can
"'Tanya" considered relative to
SO '
Now we look at S',
relative
to
terrorist'.
the
S',
logical
forms.
sentence
considered
verb 'believes'.
propositional
large
'that Tanya is a terrorist' and consider its role
Since,
(following
to be or
for
17.
Harman,
G.,
'Deep
Davidson eds., op. cit.
'Ralph
relative to S i ,
Higginbotham
attitudes
S1,
believes
appears as
Structure
as
- 39 -
Tanya is
a
the object of the
Harmanl ) takes the objects of
to be objects
Higginbotham,
that
the
"similar to"
objects
Logical
form',
to
interpreted
which
in
semantic
Harman
and
get assigned
values
the 'that' clause
are themselves logical forms,
Similarity is used
taken to be true of things that are similar to itself.
two things will
in conscious imitation of Davidson's samesaying relation:
be
similar
features.
to
each
The
other
role
of
if they
the
have
content
the
same or
sentence
is
may
similar
be
semantic
expressed
by
(27)(iv):
(27)(iv) (Va)(a satisfies <S',
S > iff a is similar to S O )
(iv) together with (v):
(v)
<'Ralph',
S I > denotes Ralph
and
(vi) (Va)(Vb)( <a,
b> sats <'x believes y',
SI> iff a
believes b)
along with simple interpretation rules give us (vii)
(vii)
<S I
,
SI> is true iff (3a)(a is similar to
S O and Ralph believes a)
(vii) is a T-sentence that tells us,
'Ralph
believes
that Tanya
is a
more or less,
terrorist'
is
that S I , the sentence
true
iff
Ralph
believes
something similar to the sentence 'Tanya is a terrorist'.
treatment
Higginbotham's
normal denotations.
has
embedded
expressions
This provides a simple
denoting
solution to the
their
problem of
anaphora that we have been looking at.
Let us
terrorist,
return to our example
but she. is too honest',
1
(10),
'Ralph
believes
that Tanya i is
now constructed as (28):
- 40 -
a
(28)
[SI Ralph believes that (SO Tanyai is a terroristiI
but
S2 shei is too honest)
The challenge was to find some way of getting the unembedded 'she.'
in S 2 to denote Tanya,
Thus all we
denotation
need
so that S 2 can get its correct truth conditions.
to do
is to
of x considered
find
relative
some
to
y,
pair
<x,
y>
is Tanya;
such
that the
as expressed in
(29):
(29) Den<x, y> = Den<'shei', S2> = Tanya
But this is now no problem.
sentence,
(30)(i)
'Tanya' considered relative to the content
will just get its normal denotation.
Den<'Tanyai',
S >=Tanya
S0>
(ii) Den<'she ', S 2 >=Den<'Tanyai',
So Qiii) Den<'shei',
The aspect
S2>=Tanya
of the theory that
their normal denotations
aspect
of the
failures
of
So we have (30):
theory,
provides
embedded
seems to be satisfactory.
the
substitutivity,
one
is
that
less
is
so.
However
supposed
How
are
expressions
to
these
the
account
with
other
for the
failures
to be
explained?
7.1 Criticism of Higginbotham's Account
'Ralph
believes that Tanya is
that Patty is a terrorist'
easily
enough.
embedding
The
attitude
is
false.
content
report,
a terrorist' is
is
true,
Higginbotham's
but 'Ralph
account allows for this
sentence,
considered
treated
a
- 41 -
as
believes
predicate
relative
true
of
to
its
things
similar to itself. The content sentences 'Tanya is a terrorist' and 'Patty
is a terrorist' are not similar to each other and obviously might fail to
Since each is true of things similar to
be similar to the same things.
itself,
they
will then be true of different things.
while the other is not.
Ralph's belief content,
to, and so true of,
One may be similar
It seems to me that the explanation of the failures of substitutivity is
The two content sentences 'Tanya is a terrorist'
not quite satisfactory.
and
'Patty
a
is
as
report,
attitude
extensions:
that
predicates
theory,
are
terrorist'
is
They
what accounted
of
their
to
predicates
are
for the opacity
content
codenotational
sentences,
they
expressions
different
with
effects.
But
the
on Higginbotham's
For the counterpart
expressions inside
the two content sentences will always get the same values.
the
embedding
that have,
expressions
denotations.
relative
treated,
predicates.
composed
the same
are
are
continue
assigned
to
their
share
normal
their
Relative to
values,
so
And
denotation.
relative to the belief report as a whole they are not assigned values at
all: only the whole content sentence gets a value relative to the larger
sentence.
Thus the extension
of the content
sentences
considered
as
predicates does not depend upon the values of their parts.
Now,
the departure from compositionality
a failing.
real
taken by itself, need not be
There is no a priori reason to believe that semantics
natural
language
will
everywhere
18. C. f. Hintikka, J., 'A Note on Anaphoric
Linguistic Inquiry,
Processing by Humans',
reference there cited.
- 42 -
be
for a
compositional.18
Pronouns and Information
1987,
p.
116,
and the
new
a
understand
may
There
parts.
known
sentence
that
be
is
built
other
up,
in
ways
in
ways,
from
explain
our
known
to
which
is built up
of new complex predicates
knowledge of how the extension
to be offered
But some such account needs
from old simple parts.
is that we may
of how it
is one kind of explanation
Compositionality
as treated by Higginbotham,
the particular case of content sentences,
for
if
the violation of compositionality is not to matter.
At this juncture
it will be well to reflect
upon
why the
failure of
compositionality does not afflict Davidson.
samesaying
Davidson's
meaning
to explain
the
is not invoked
relation
of any
extension
the
within
For
expression.
theory
of
him every
element in the whole attitude attribution has its extension determined by
some antecedently available part of the theory of meaning.
The content
right.
It is not a
sentence
is treated
predicate
true
only as a
other
of
sentence in its own
that
things
samesay
Consequently
it.
with
compositionality is not violated.
Moreover
Davidson has an acceptable
work.
is supposed to
among utterances
Earth
moves,
19
Davidson
and inscriptions.
then
samesaying
uses
explanation
of how samesaying
samesaying
only
If I say
relates
my
as a
relation
that Galileo said that the
utterance
of
the
content
19. The best source is his 1976 paper, 'Reply to Foster', reprinted in
Inquiries into Truth and Interpretation, Oxford University Press, 1984.
- 43 -
sentence
to a
similar
sort
of
thing.
20
arrive
To
at
samesays with Galileo's utterance one proceeds as follows.
utterance.
will deliver
of Italian.
conditions
truth
Galileo's
which
is true,
utterance
differences
options
be
will
flavour
of
Galileo's
century
Anierica.
soine
are
that
utterance;
more
Those suitably
better
centu y
17th
flavoured
suited
ways
to
among
But among
which will be too subtle for the theory of meaning to note.
these
as
there will be various different ways of stating the
Now,
under
This theory
in English).
(for Italian,
that have the same
sentences
English
the Italian one.
conditions
Take Galileo's
Construe it as an articulated linguistic structure
Apply to it a theory of meaning
that
something
the
capturing
Italy,
of stating
less
20th
the
truth
conditions of Galileo's utterance samesay with it.21
Higginbotham
thus
departs
from
Davidson
in
two
writes samesaying into the theory of meaning itself,
gives
the
extension
of
certain
complex
predicates.
treats samesaying as a relation among abstracta,
exactly,
ways.
First,
he
as a relation that
And
second,
not Fregean
he
thoughts
but "interpreted logical forms".
The second departure creates two lacunae that need filling.
know what it is to utter an utterance,
First,
we
but we lack an account of what
20. It is thus no accident that Davidson has never offer d an account
of the logical form of 'believes' sentences. Samesaying needs relata, and
Davidson refuses to recognize any analogue of an utterance, in the case
of belief. For an explicit avowel of this refusal see his 'Knowing One's
Own Mind', Presidential Address delivered before the Sixtieth Annual
Pacific Division Meeting of the American Philosophical Association, 1986.
21.
I am indebted
samesaying.
to
Bruce
Vermazen
- 44 -
for
my
understanding
of
it is to believe a logical form. So when similarity is used to account for
the extension of a content sentence,
which
of determining
provided
where
is
utterances
objects
specific
by the content sentence
the
other one?
And,
but we
to samesay,
we lack,
even in principle,
it relates.
itself,
second,
One
logical
a way
form is
as the theory treats it,
we
know
what
but
it is for two
lack an account of what it is for two
logical forms to samesay.
The first departure
filled,
from Davidson means that until the lacunae are
the semantics is radically incomplete.
a technical expression
in
For 'samesaying' appears as
the theory of meaning.
that is crucial to the account of how the semantic
It is
an expression
features of complex
expressions depends upon their structure and the semantic
features of
their parts. The theorist of meaning thus cannot leave it unexplicated.
I do not claim that
found.
the required
fillers of
the lacunae
are not be
But I recommend a different course.
8 Frege Vindicated?
Let us collapse
the
number of logical
Instead of trying to relate,
Galileo believes,
moves',
as separate
to the logical
forms in
things,
play
clown
to one.
the logical form that
form of the content sentence 'The
let us suppose that Galileo believes that very logical form.
22. This was Harman's original idea,
as I understand it.
- 45 -
Earth
22
We
may thus take a second step back towards Frege (the first, taken by
Higginbotham,
was
mentioned
four
paragraphs
back)
and
treat
the
content sentence (still relative to the embedding report, of course) as a
singular term, rather than a predicate.
And, since we are in the mood for a retreat towards Frege, we might
as well try a third step.
Let us assign to each
semantically
simple
expression in the content sentence a sense as its denotation.
Embedded
complex expressions,
will then
including the whole content sentences,
denote logical forms (their own) in which inhere senses. The extension
of a content sentence,
considered as a singular term,
will thus be a
function of the denotations of its component expressions (senses),
and
its structure: compositionality regained.
We are still left with a problem similar to that of saying what it is to
believe an interpreted logical form.
We now have to say what it is to
believe a thought, a logical structure of senses.
But we need not hurry to answer the question,
plays an important role in the semantics.
since it no longer
There is just a relation that is
denoted by the expression 'believes', a primitive and unanalyzed part of
the content of the familiar idiom of belief reporting. Our problem was to
explain
upon
how
the
semantic
those of their
parts.
features
of
We never
complex
promised
expressions
to
depended
give philosophical
theories of every primitive semantic value that we came across.23
23. These remarks echo Davidson's defence of his use of the samesaying
relation. See 'On Saying That' p. 104, of Inquiries into Truth and
- 46 -
So my suggestion is
(27)(iv) we write
that instead of the unhelpful
(31):
(31)
(i)
<'Tanya',
Sl> denotes rTanya
(ii) <'is a terrorist',
S1> denotes
is a terrorist
I assume that S' just inherits the value of S O considered
Sl
.
relative to
This gives us (iii):
(iii>
<S',
Si> denotes rTanya is a terroristp
the corner quotes once again signalling that the expressions inside them
denote senses,
rather than their normal denotations.
together
(31)(iii)
with the original (27)(v) and (vi) suffice to yield (29)
(32)
<S i
,
SI> is true iff Ralph believes rTanya is a
terrorist7
That is to say
Ralph believes
effects
are
'Ralph
the
believes that Tanya
is a
thought that Tanya is a terrorist.
explained
in
Frege's
expression denoting a sense.
original
way,
(ii),
expressions
we
can
easily
denote their
by
iff
So the opacity
each
embedded
But since we also have those very same
expressions denoting their normal denotations,
and
terrorist' is true
account
for
the
normal denotations.
as expressed
appearance
They
that
appear
in
(27)(1)
embedded
to because
that is just what they do.
Of course it would be rash to put a great deal of faith in the sketch
Interpretation,
and the references cited thereat.
- 47 -
of the account I have just offered.
more
complicated
inferences,
mechanics
cases,
and so
how
on and
We need to see
well
or
so forth.
badly
And we
of the proof of the T-sentences:
into difficulties with Frege's
hierarchy.
it
how it extends to
captures
need
a task
desirable
to spell out
that quickly
the
leads
And perhaps most importantly,
we had better be prepared to investigate whether anything could
the theoretical role that we have assigned to senses,
and if so,
play
what
sort of thing this would be.
I
confess
to
a
moderate
degree
Optimism about all the problems,
of optimism
however,
- 48 -
about
each
problem.
would be supererogatory.
PAPER TWO
WHO'S AFRAID OF NARROW MINDS?
- 49 -
The mind is its own place
- John Milton
Who's Afraid of Narrow Minds?
1 Introduction
of
purpose
The
this
is to
paper
defend
the
view
that all of
the
propositional attitudes that we ascribe to one another in our everyday
are
folk psychology
content'.
The
of what has come to be called
possessed
defence
will
by
no
be
means
complete,
since
'narrow
shall
I
neither be attempting to meet all varieties of objection to the view,
offering positive arguments in its favour.
nor
Rather I shall be concerned
This opposition comes
to fend off one major source of opposition to it.
from Gareth Evans , and John McDowell.2
In
this
dispute.
I
section
In
the
shall
remaining
give
a
brief
sections
I
general
shall
introduction
deal
with
to
Evans'
the
and
McDowell's objections to my view.
1. Evans,
1982.
G.,
The
Varieties of
Reference,
Oxford
University
Press,
2. McDowell, J.,'The Sense and Reference of a Proper Name' ('SRPN') in
Platts, M., ed. Reference, Truth and Reality, Routledge and Kegan
1986; 'Criteria,
in Synthese,
Paul, 1980; 'De Re Senses' ('DRS'),
British
Proceedings of the
('CDK'),
and
Knowledge'
Defeasibility
Academy, 68, 1982. Evans and McDowell appear to hold the same view,
down to a detailed level. The leading idea originates with McDowell.
- 50 -
1.1 Background
In
the
fifteen
last
years
so a
or
certain
thesis
about
reference,
largely owed to Kripke3 and Putnam,4 has become popular.
The
view,
which
I call the
are
expressions
'direct
An
"extension-dependent".
if its meaning
'Sandy' and 'Mandy'
are extension-dependent,
different
type-identical)
meanings.
6
if
is that certain
expression
upon its extension.
depends
extension-dependent
(even
view,
reference'
is
if
So
and refer to numerically
individuals,
they
And if 'Sandy' had referred to Mandy,
have
different
it would have meant
something other than what it actually does mean.
A
similar
thesis
is
applied
to
natural
kind
terms.
But
here
the
meaning of the expression depends not upon which actual things happen
to fall under it,
but upon the type to which those things belong.
There are two attitudes that the proponent
view
3.
might
Kripke,
take
S.,
towards
empty
of the direct reference
extension-dependent
Naming and Necessity,
expressions.
Harvard University Press,
'The Meaning of "Meaning"'
4. See e.g, Putnam, H.,
Language and Reality, Cambridge University Press, 1975.
in
his
1979.
Mind
5. 1 think the term 'extension-dependent' is owed to Wiggins.
6. Different extensions are sufficient for different meanings, but may
not be necessary. It is open to the direct reference theorist to hold
that two extension-dependent expressions with the same extensions may
yet have different meanings. McDowell and Evans do hold this, see
below section 4).
- 51 -
'Mandy'
when
empty
certainly
does
not
have
'Mandy' when it refers to a genuine Mandy.
reference view,
the
at least that empty extension-dependent
an empty 'Mandy'
extension
This entails,
not have the same meaning as do non-empty ones.
properly
same
as
on the direct
expressions do
One might hold that
has a different meaning from the name of a Mandy,
so-called.
McDowell and
Evans,
however,
hold
the
stronger
thesis that empty extension-dependent expressions have no meaning.
1.2 Conceptions of Content
It
seems
natural
to
take
the
direct
reference
concepts as well as to expressions of language.
view
Doing so,
to
apply
to
we arrive at
the idea that which content a given concept has will depend upon what
it refers to.
The extension of the direct reference view from words to concepts is
encouraged by the tight connections
that we find between the contents
of propositional attitudes and the sentences that we use to express and
report them.
If 'Sandy' and 'Mandy'
have different meanings,
would expect the beliefs they may be used to express,
have
different
contents.
7. Terminological matters.
is
a
psychological
syntactically
(Indeed,
on
A concept is
correlate
structured object
of
views
or report,
an
interpreted
has a content
to
expression's
a thought component.
an
and it
some
then we
sentence:
A thought
it
is
a
(or some contents)
which may or may not be essential to it. Thoughts are the objects of
propositional attitudes; they may be believed, doubted, desired-true
etc. 'Content' has to be understood more or less intuitively. It covers
at least semantic coiitent and phenomenal content, but claims that there
are other sorts of content are to be regarded as at least intelligible.
- 52 -
meaning is to be identified with its role in expressing beliefs; in which
case
the
extension
of
the
direct
thought would be automatic.)
have a different content
'Sandy'
has
no
reference
view
from
to
So the belief that Sandy is sullen would
from the belief that Mandy
referent,
language
and
is sullen.
And if
'Sandy
sullen'
so no
meaning,
then
cannot be used to express, or report,
a belief.
And this suggests that
is
the belief that Sandy is sullen depends for its existence on its object,
Sandy.
We
might
hypothesize
characteristics.
determined,
One
that
of
the
these
at least in part,
mental
kinds
states
of
have
two
characteristic
by things outside the
head.
kinds
would
of
be
The second
kind of characteristic would be logically independent of anything outside
the
subject.
Properties
of this
second
variations of the subject's surrounding
not
matter
there were,
whether
there were
a
kind
would
environment:
world
outside
be
insensitive
to them it would
the
subject,
nor,
if
what it was like.
Characteristics
psychological
of
states
the
with
those of the second kind,
call McDowell's
and Evans'
first
what
kind,
has
come
if
there
to be
if there are any,
are
called
any,
'wide
endow
content';
provide narrow content.
view the 'Wide Content Only'
('WCO')
the belief that Mandy is morose is
narrow content.
The only
the existence of its object,
I
view.
On the WCO view some psychological states have only wide contents.
content,
to
If
such a state, then that belief has no
content it has is of the kind that
requires
Mandy.
no wide
No Mandy,
no content at all.
- 53 -
no wide content;
For McDowell and Evans it is thus no accident that our specifications
are with the objects
states are bound up as they
of psychological
of
When the best description of a psychological state makes
those states.
mention of the state's object,
this is because what it is for a subject to
be in the state essentially involves her being related to that object.
if one is to believe that Mandy is morose,
So
one has to stand in a certain
relation to Mandy herself.
The
view,
view
that
I shall
defend,
the
"Narrow
Content
Also"
content.
is that all propositional attitudes have a narrow
(NCA)
So the
belief that Mandy is morose has a content that is not Mandy-dependent.
The belief that Mandy is morose perhaps has a wide content as well as
a narrow one.
Removal of Mandy would eliminate that.
But one would be
left with a belief endowed with a definite content nevertheless.
The existence of narrow content would be the existence of something
that simultaneously performs two separate tasks.
its
narrowness.
content
of
his
If
Zippy
mental
state
thinks
must
Zonker
be
is
that
The first accounts
zany,
much
then
the
of it that
for
narrow
is within
Zippy.
8
Suppose that Zippy's concept,
Zonker,
be the upshot of a combination of factors.
Zippy:
for example,
refers to Zonker.
This would
Some of these will lie outside
certain relations crossing the gap in
space and time
8. My use of the underlining device is to be understood as an attempt
to do something that is as much like quoting concepts and thoughts as
is possible.
- 54 -
between thinker and object of thought,
that make it the case that when
his thought will be
Zippy thinks some thought of the form Zonker is F,
But these external relations cannot
is F.
in particular,
true iff Zonker,
to let Zippy
be all that is required
by themselves
think
of
Zonker.
made at Zippy's end
Presumably there are also important contributions
It is a task of the concept's narrow content to make such
of the chain.
contributions. The concept's narrow content is thus determined by those
of its properties that are entirely independent of any links of the chain
that are outside Zippy.
The
second
task
for
content
narrow
Narrow content must deserve its name,
is
more
nebulous
what gives a subject's
at least,
presented
by
the
world
is
Mandy
is trendy makes
content.
But for me narrow
or subjective
perspective,
The narrow content of a belief is the way in which the
point of view.
perspective;
nature.
and be recognizably content. It
is not obvious what performance of this task entails.
content is,
in
the world
Sandy's
its subject.
belief to
appear
a certain
belief
way from
that
Sandy's
which way that is depends only upon the belief's narrow
Narrow content alone
must suffice to provide
something
that
should be counted as a subjective point of view.
The
espousal
partly motivated,
of a
theory
makes
essentially
I believe,
is
inner.
things
are
with
for
narrow
content
is
A subject's point of view of
properly thought of as intrinsically independent
of the world of which it is a point of view.
how
room
by the intuition that the mind is,
at least in my case,
to put it suggestively,
the world,
that
the
subject,
It should depend only upon
considered
in
abstraction
'p
-
55
-
from
her
points
of
mention
make
that
view
of
then,
For the friend of narrow content,
environment.
specifications of
be
must
objects,
external
specifications exploiting merely accidental features of what they specify.
The
WCO
theorist
is
skeptical
content are always performed,
The
WCO
theorist
extension-dependent
terms
and independent of the mind.
On the WCO view,
the
do
not
of such
determines the reference
tasks
have
narrow
narrow
by a single thing.
are
that
concepts
for
set
by
ascribed
For
content.
what
concepts is mostly outside the head
9
what there
these concepts does not,
two
the
or even performable,
that
holds
that
is in the
head
that corresponds
to
considered in isolation from its surroundings,
have anything that should be called 'content'.
It cannot by itself, in the
absence of its relations to an external referent,
count as a component
of a point of view.
Although I think
view
is
the
idea
that
that
the main underlying
the
internal
motivation
component
of
for the WCO
a
putative
extension-dependent concept is too exiguous a thing to be endowed with
9. C. f. Burge, ('Belief De Re', The Journal of Philosophy LXXIV,
1977), who writes 'features of the mental entity itself ... are not always
sufficient to pick out the relevant object' in these cases '[f]actors which
determine the objects about which the believer holds his beliefs ... are
cases,
most
in
indeed,
nonconceptual
counted
appropriately
noncognitive.' Oxi his view what determines reference is independent of
the mind in the strong sense of not being part of its cognitive domain.
McDowell (DRS, see also CDK) takes issue with this. For him the mind's
cognitive reach extends to include those factors that are literally (but
not metaphorically) outside the head. For him, the independence of
these factors thus consists only in the fact that they would exist, even
if the mind within which they were located, did not.
- 56
content,
I shall not have a great deal to say about the idea explicitly.
The worry is rarely articulated
what,
exactly,
statement
would be required
of the
misconceived.
very fully,
concern
and so it is hard to know
to alleviate
in section
6,
it.
below,
I shall take up
and
argue
I shall enter some intuitive considerations
that
it
is
that indicate
that what is ensconced entirely within the head is really content.
more than that would require a theory of narrow content.
one
Any
Once such a
theory was provided then it might be possible more seriously to take up
the question of whether or not what is narrow is also content.
I think that rather than launching into the development and defence
of such a theory,
posteriori,
it will prove more fruitful to fend off a priori (and a
for which see my 'Seeing
What is not There')
objections
to
the plausible and natural view that there is something there to have a
theory of.
With that end in mind,
of the WCO view,
I shall be focussing on a patrticular expression
that McDowell gives (in
SRPN).
We are to suppose
that a researcher is interested in the ways of a jungle tribe.
He learns
that the locals use the word 'Mumbo-Jumbol as the name of one of their
Gods.
Not
suppose
being
of
similar
faith,
he
concludes,
'Mumbo-Jumbo
brings
thunder',
then,
not
use
attributions
of
to
help
explain
native's
thunder
that
we
shall
that the name has no beArer.
If a native says something that intuitively
should
rightly
are
the
utterings
the
of
the
has the English translation
McDowell
belief
actions
sentences
- 57 -
holds,
that
-
the
interpreter
Mumbo-Jumbo
including
containing
those
'Mumbo-Jumbo'.
brings
actions
For,
according
words'.
to
brings thunder.
p.
'no
153)
proper names:
his
that Mumbo-Jumbo
the same is view is taken of
for the most part,
And,
by
belief is expressed
is no such belief as the belief
There just
other empty
(ibid.
McDowell
be used in content sentences
they cannot
that express or ascribe beliefs.10 So the existence of thoughts ascribed
depends upon the
by content sentences in which proper names feature,
existence of a bearer for the name: no bearer,
no thought.
I shall be dealing with what I take to be four distinct arguments that
11
explicitly distinguished,
of pages of SRPN,
places
possible
in
his
that
arguments
Evans'
book.
they
and most interesting,
contributions
are
is
therefore
wish
to
be
to
not
We
them.
shall
is supplemented
from
gathered
Interpretation
would
I attribute
are not
but appear all run together in a dense couple
and the last,
in detail elsewhere.
arguments
These
WCO view.
are offered in support of the
difficult,
associated
consider
with
the
different
and
it
all of
arguments
is
the
in
ascending order of depth and plausibility.
10. There may be exceptions, but these have special features
them apart from typical proper names. See section 4) below.
that set
11. Two additional arguments are to be found in Evans' book. I have
little to add to
I have relatively
these because
not dealt with
Blackburn's response to them; for which see Blackburn, S., Spreading
the Word, chapter 9. I do not endorse Blackburn's account of narrow
content.
-
58
-
2 First Argument
The
native
says
something
that
intuitively
has
the
translation
'Mumbo-Jumbo brings thunder'.
Here is what appears to be a brief,
almost self-contained
argument
(McDowell ibid.):
No belief is expressed by his words: they purport to
express
a
belief
which
could
be described
in
the
transparent style, but since no appropriate belief could be
thus described, there is no such belief as the belief they
purport to express.
A belief describable in
some object,
the transparent
style is a belief concerning
that it satisfies some property.
only such beliefs
that are attributed
On McDowell's
by content sentences
view it is
with proper
names.
I
am
not
sure
that McDowell
express an argument.
really
intends
the
quoted
words
to
But, in case he does, I respond.
2.1 Reply to First Argument
The argument would seem to proceed as follows.
words
purport
transparent
described.
to
style.
express
Second
Therefore,
a
belief
premise:
that
be
no appropriate
subconclusion:
belief the words purport to express.
could
there
is
Therefore,
- 59 -
no
First premise:
described
belief could
such
belief
conclusion:
in
be
as
there is
the
the
so
the
no
belief that they express.
What does it
The first is to the first premise.
I have two objections.
mean to say that the words 'purport to express a belief which could be
in
described
the
I
style'?
transparent
the
think
ihe belief
reasonably be purportedly purported is that there is a belief,
is
the
by
expressed
belief
the
and
words,
is
can
that
most
in
describable
the
transparent style.
McDowell,
When
however,
one
...
name
proper
utters
assertively
and
sincerely
a
containing
thinks otherwise:
not
does
one
sentence
a
mean
be
to
expressing a belief whose availability to be expressed is
indifferent to the existence or non-existence of a bearer for
the name.
Were
purport
this
true,
then
to
express
a
transparent
that
belief
but that
style,
words
the
was
of
the
only
not
utterance
native's
was
so essentially.
would
in
describable
But McDowell
the
has no
right to claim this.
Surely enough,
committing oneself to the existence
in the normal case anyway,
one is,
by the
But that is explained
of its bearer.
that it
practice
one's
words
is
lay
beliefs,
or
that one takes not just the truth,
singular
used
theories
into
written
beliefs
to express
than we
are,
to
depend
them.
upon
Those
including
truth of one's
the
- 60 -
implicit
somehow
for the
in
lay
but the very existence,
of
of
for
existence
less concerned
perhaps
utterance
There is no warrant
being dependent on this ontological fact.
claim
a proper name,
sentence containing
if one utters a
McDowell's
referents
with
native,
philosophical
are
not
up
to their necks,
knowingly or otherwise, in the WCO view.
The second premise,
the transparent style,
The
that no appropriate belief could be described in
is true.
subconclusion,
that
there
words purport to express would
were
interpreted
in
the
Then,
perhaps,
we
could
principle:
express
either
no
suggested,
strong
If,
a
on
thus allowing
such
follow,
perhaps,
of
belief
as
the
native
other
belief
the
seems to intend.
as
committed
describable
hand,
to be true,
the
if the first premise
that McDowell
transparently
the
it
is no
form
think
I express
belief.
12
we
belief,
interpret
then what
to
it
the
or
I
as
I
follows is merely;
there is no belief that has the property that the purportedly expressed
belief is purported to have.
It does not follow that there is no belief
purportedly expressed by the words.
My second objection
conclusion.
What
is
McDowell
to
the move
really
actually expressed by the words.
he has so far argued.
must also
be
from
needs
to
the
subconclusion
show is
that
no
to the
belief
is
But this in no way follows from what
It is not in general true that what is expressed
purportedly
expressed.
Why
should
it
be
true
in
this
case?
12.
Since McDowell's claim is undefended
and prima facie implausible,
I
do not undertake a defence of my opposed opinion. Were such to be
offered, it would be built around the frequent and unabashed uses of
empty proper names in descriptions of others' beliefs, that we indulge
in. Normal people do not have the sense that it is literally impossible to
believe that Santa Claus has a white beard,
- 61 -
whether or not he exists.
Even if McDowell did intend the rather feeble argument we have been
discussing,
he certainly would not want his WCO view to rest on it.
So
let us move on.
3 Second Argument
Here is an official defence of the WCO view (ibid.):
13
can be defended on these
The non-Fregean [=WCO] view
lines. An interpreter's ascription of propositional attitudes
to his subject is in general constrained by the facts (as the
This
is
partly
because
interpreter
sees
them).
intelligibility, in ascriptions of belief at least, requires
conformity to reasonable principles about how beliefs can be
acquired under the impact of the environment; and partly
because the point of ascribing propositional attitudes is to
bring out the reasonableness, from a strategic standpoint
constituted by possession of the at'itudes, of the subject's
dealings with the environment. Now, whether a name has a
bearer or not (in an interpreter's view) makes a difference
to the way in which the interpreter can use beliefs he can
ascribe to the subject in making sense of the subject's
behaviour. A sincere assertive utterance containing a name
with a bearer can be understood as expressing a belief
correctly describable as belief, concerning the bearer that
it satisfies some specified condition. If the name has no
bearer (in the interpreter's view), he cannot describe any
suitably related belief in that transparent style.
What follows the above quotation is a comment about what,
an interpreter could
empty proper name,
glean
from
the
utterance
given that it has no sense.
of a sentence
perhaps,
with an
Then comes the
brief
13. When McDowell calls the view 'Non-Fregean' he has in mind Frege's
famous claim in 'On Sense and Reference' that the thought that
Odysseus was set ashore at Ithaca while sound asleep 'remains the same
whether "Odysseus" has reference or not'.
- 62 -
argument I considered in the last subsection.
Neither of these sequentia
adds to the argument begun by the words in the long quote.
then,
reconstruct,
from those words,
an argument
Can we,
for the non-Fregean
view? Perhaps the following.
The argument
begins with
a statement
of the
principle of charity.
The ascription of attitudes is constrained by what the external facts are
taken to be,
that
make
in general,
sense in
possible (possible,
the
light
perhaps,
a name has a bearer,
containing
so that we can ascribe beliefs and behaviours
the name,
of
those
facts,
to whatever
extent
relative to certain other constraints).
is
When
we can attribute beliefs expressed by utterances
and
actions in a special way.
use these attributions
to make
sense of the
The special way involves taking the the belief
to be describablel4 transoarently.
Perhaps
subject's
McDowell
truck
environment.
with
would
the
which
he interacts.
their existence,
But
then,
that
in
environment
We can interpret
involve special cognitive
with
say
these cases
we
as
guided
his actions
rapports
Such
directly
by ascribing
can
see
the
by
that
attitudes
that
with those parts of the environment
rapports
are
logically
dependent,
for
upon their objects.
the
conclusion
would
be,
these
necessarily describable in the transparent mode,
attitudes,
which
must be distinct
are
from
14. 'Describable', not 'described'. The point is that the belief is taken
to be about a particular object.
It may be so taken even when
described by a content sentence in which intersubstitutions are not
truth-preserving.
- 63 -
any attitudes not so describable.
Such attitudes are simply not available
in the absence of an object.
3.1 Reply to Second Argument
If
is
that
the
argument,
then
is
it
the
to
vulnerable
following
objection. There is no valid rii.ve from the premise that the existence of
for the availability of a certain
for the attitude is requ.ired
an object
sort of description of the attitude,
to the conclusion that the object is
required for the availability of the attitude itself.
It might be true that
the existence of an object allows us to describe an attitude by relating
it to an object with which the subject is intimate,
without it being true
in another way,
that the very same attitude could not be described
way that made no mention of the object.
To show otherwise
McDowell
was dependent
upon the
that the attitude itself
would have to argue
a
object: and that is exactly the point of dispute.
The existence of descriptions of a subject's mind that make mention of
particular objects
provides no evidence
class of object-involving
these
descriptions
The
actions.
description
attitudes.
be,
might
question
may
is
for
existence of a special
useful
This is true no matter how
the
whether
correctly
for the
interpretation
what
be
seen
makes
as
of
true
the
a
subject's
transparent
composed
of
an
extension-independent attitude of the subject's and an external object to
which
it
stands
in
some
particular
relation.
Nothing
argument suggests a negative answer to this question.
- 64 -
in
McDowell's
4 Third Argument
The third argument to be found in Evans and McDowell results from
their conception of the proper form for a theory of sense.
one which
Their idea,
may serve
meaning
machinery
to
a
theory
for
each
as
derive,
is
I applaud,
that a
Such
of sense.
sentence
of
a
structure-revealing
and 'p' by a statement,
The
conditions.
description
of
will
values to simple expressions
include
will
theory
object
a
'S' would be replaced
sentence,
of that sentence's
clauses
of
have
language,
an object-language
in the meta-language,
machinery
a
the
T-sentence of the familiar form 'S is true iff p':
by
theory
Davidsonian
that
of the object language,
assign
truth
semantic
and clauses
that
allow for the computation of the semantic values of complex expressions,
on the basis of those of their parts.
states, then,
is what expressions denote.
as a theory of sense,
is,
it does
What the theory actually explicitly
so in
a
But the theory may yet serve
if in saying what the denotation of an expression
way that
displays,
or
reveals,
the
expression's
sense.
McDowell and Evans hold that proper names should be dealt with by
clauses of the form:
'a' stands for A
(1)
where
(ibid.
p.
'A'
has
143,
the same
c.f.
also
sense as
Evans,
'a'.
op.
- 65 -
Such
cit.
p.
a
clause,
35)
McDowell
says
or
more
'gives
-
as good as gives -
in that context,
strictly,
the sense of the name'.
their
doing
of
bit
in
object-language speakers.
for
A.
That
the
of
sense
making
capable
T-sentences
That context is a theory of meaning that delivers
actions
linguistic
of
The clause explicitly says only that 'a' stands
is something
by
the
but a quite different sense.
But
might
that
be
equally
expressed
clause:
'a' stands for B
(2)
where 'B' has the same referent as 'a',
only
first clause correctly
the
theory
that
as
correctly,
shows,
On a
theory
a
of
In
this way a
may
if
understood .
first
clause
sense.
It
suffice,
what
it
the senses of proper names
in
senses
of
so
of
because
suffices
not what it says.
theory of meaning that
McDowell's
proper
the
includes
shows the sense of 'a'.
suggested
manner,
gives
is
there
names that lack a referent.
giving
no
of
the
For the honest interpreter
cannot
enter into her theory of meaning the clause:
'Mumbo-Jumbo' stands for Mumbo-Jumbo
(3)
nor any other clause of the suggested form,
for an empty proper name.
McDowell does not say why the interpreter cannot accept
reason is obvious.
the interpreter
If this is
(3)
entails: (3y)('Mumbo-Jumbo'
(3),
stands for y),
but the
which
must take to be false.
the right way to go about things,
then empty Froper names
can have no senses; the moral that McDowell draws.
- 66 -
Evans
(ibid.
pps.
25-7)
appears
to make much the same argument.
He points out that a clause of the form of (1)
identifies the semantic value of the name in a way which
shows, or displays, its sense. When a theory of semantic
value meets this condition quite generally we may say that
This conception of
it can serve as a theory of sense ...
the relation between Fregean theories of sense and of
semantic value is attractive ... but it is clear that if we
suppose that expressions without semantic value may still
have a sense, we cannot avail ourselves of this conception.
And,
if a proper name has no sense,
an utterance that contains it.
then nothing is expressed
by
Hence McDowell's conclusion that no belief
is expressed by the native's utterance 'Mumbo-Jumbo brings thunder'.
The
form of the
argument
appears
to be
this.
We begin
with
the
premise that the best way to go about framing a theory of sense, is to
construct a theory of denotation that may do duty for one.
Peformance
of this duty consists in employment of clauses that state what needs to
be stated for the theory of denotation, but that do so in a special way.
The clauses must state what they state in a way that shows the senses
of the expressions
they deal
this
to
advice
seems
require
with.
the
meaning of clauses of the form (1).
names cannot be honestly handled
that these names have no senses:
For McDowell and Evans
incorporation
into
the
following
theory
of
Then they note that empty proper
by such clauses.
The conclusion is
and this is, or leads directly to,
the
WCO view.
4.1 Reply to Third Argumnent
The argument takes a false step.
The step from the constraints upon
- 67 -
clauses
the
of
the
clauses of form (1),
of
theory
to
denotation,
It is true that such clauses
is nowhere justified.
provide an admirable way of implementing
special case of proper names with
But McDowell and
referents.
the
if we
argument,
So it will
find clauses
can
that will do for the theory of sense,
some form other than (1),
Evans
provide the only
way of following the general program in the particular case.
possible to short-circuit
for the
the general proposal,
nowhere argue (that I am aware of) that the clauses
be
of
recommendation
the
of
but that
can be applied to empty proper names.
Evidently some clause like the following is our best bet:
(4)
(x)(N stands for x iff ... x...)
But what is to fill in the blanks?
What fills the blanks must
The question appears to be embarrassing.
same time must display the name's sense.
suppose
that
things will be abroad.
forthcoming
where
there
name,
which is
admit exceptions
property,
then,
to their
descriptive
(Some empty proper
I said
earlier
to
will only be
condition,
by those who understand
and generally held,
why
reason
both of these
it seems,
specific
some
to fix its reference.
that property,
might
is
is no
that does
Appropriate descriptions,
associated with the name,
the
But there
a description
for every name,
and at the
if it has one,
suffice to specify the referent of the name,
that
general thesis).
names may
McDowell and
have
Evans
If a name lacks this
though we might easily enough find a description that
specifies the name's
referent,
gives the name's sense (c.f.
we will have no right to suppose that it
Evans,
op.
- 68 -
cit.
p.
48).
Let us
not panic.
If the
name
description can always be found.
has
a bearer,
then
an
appropriate
We just have to combine the standard
disquotation trick with an '=' sign. So we could employ a clause like:
(5)
(x)('Mumbo-Jumbo' stands for x iff x=Mumbo-Jumbo)
If our metalanguage is regimented with a standard logic, then (5) will
not do as it stands.
The second occurrence of 'Mumbo-Jumbo' is in a
quantifiable position, and will permit existential generalization.
But there are two manoeuvres available to remedy this, both of which
I find independently advisable.
Intuitively it seems as though (5) is true rather than false or lacking
in
truth
value.
If
that
is
right
then
the
metalanguage
captures intuition will be based on a free logic,
permit
existential
generalization
in
(5).
that
best
a logic that does not
An appropriate
and elegant
regimentation of a metalanguage suitable for formulating a version of (5)
in such a way that it comes out true, has been provided by Burge.15
McDowell,
of course,
would not countenance
should be counted true.
the intuition that (5)
But that is not the issue. His claim was that a
theory of meaning that employed axioms for proper names that showed
the names' senses as it stated conditions upon their reference could not
handle
empty
language,
15. Burge,
Platts, ed.,
proper
names.
But
(5),
will serve just as well as (I)
T., 'Truth
ibid.
and
Singular
- 69 -
ensconced
in
the appropriate
for non-empty
Terms',
somewhat
proper names,
ironically
in
and is not falsified by empty ones.
The second manoeuvre for getting out of ontological troubte would be
to treat names
as predicates.
first would
to
be
semantically
take
complex
a
There are two ways of doing this.
leaf
from
Quine's
'= Mumbo-Jumbo'
book
and
collapse
the
the
semantically
simple
We would then avail ourselves of the (relatively)
ontologically
neutral
predicate 'is Mumbo-Jumbo',
into
16
The
with 'is' copular.
clause:
(x)('Mumbo-Jumbo' stands for x iff x is Mumbo-Jumbo)
(6)
Here
the
tnquoted
occurrence
of 'Mumbo-Jumbo'
is
term, and its position is not (first-order) quantifiable.
Mumbo-Jumbo' would
be uniquely
not
a
singular
The predicate 'is
true of Mumbo-Jumbo,
if he existed,
and otherwise would have empty extension.17
The
import
of
(6)
would
'Mambo-Jumbo'
would
Mumbo-Jumbo,
or of nothing.
be
function
that
as
a
the
object-language
descriptive
So 'Mumbo-Jumbo
phrase
expression
true
of
just
brings thunder' would
be elliptical for 'The Mumbo-Jumbo brings thunder'.
Employing the Quinean predicates in an account
proper
16.
names
is
relatively
The book is Quine,
1960, the leaf is
plausible
W. V. O.,
for those
of natural language
names
Word and Object,
M.
which
I.
T.
have
a
press,
179.
17.
Dummett makes just this response to Evans and McDowell,
in
Dummett, M.,
The Interpretation of Frege's Philosophy,
Cambridge,
Mass., Harvard University Press, 1981, p. 136.
- 70 -
unique,
generally
accepted
purported
referent
'Mumbo-Jumbo' would probably be an example,
some others.
in
the
community.
along with 'Pegasus' and
It does not seem to work very well for common or garden
names like 'Ike', 'Mike' and 'Spike'.
If a theorist of meaning were to tell
us that
(7)
(x)('Mike' stands for x iff x is Mike)
we would not learn very much.
(6)
does
speLify
'Mumbo-Jumbo*.
for 'Mike':
the
If there is one main Mumbo-Jumbo,
conditions
that
If there are many Mikes,
deter'mine
then
which thing dces x have to be,
(7)
the
then
value
of
does no such thing
if it is to be the value of
the word?
4.2 A Modified Burgian Treatment of Proper Names
Most ordinary
proper names are best treated as predicates
capable of being true of many
different individuals,
true of just a unique individual or nothing.
that are
rather than being
This treatment of namus is
due to Burge.18
Burge notes that names may without impropriety
nouns.
then
One can say,
(6)
does
specify
"Mumbo-Jumbo"',
'If
for
'An
18.
"Mike"'
Burge,
and
T.,
Philosophy LXX,
for example,
the
'If
coi.ditions
"Mandy"
'Reference
and
as common
there is one main Mumbo-Jumbo,
that
there are many Mikes,
empty
function
has
then
a
Proper
1973.
- 71 -
determine
(7)
different
Names',
the
value
of
does no such thing
meaning
The
from
the
Journal
of
properly so-called'.
name of a Mandy,
The meaning of proper names is apparent in such sentences.
Mandy
is
to
be
called 'Mandy',
is
to be
theory
someone
(perhaps
even
something)
who
is
whose name (or one of whose names) is 'Mandy'.
properly
of meaning
called 'Mandy'
proper.
is
a
question
that lies
To be a
properly
What it
outside
The theorist of meaning does not have
the
to
give an account of the conditions under which one is the genuine owner
of a proper name.
She is committed ornly to there being such an account
to be given.
Since there genuinely is such a thing as being a Mike,
Mike is necessary and sufficient
for 'Mike'
and being a
to be true of one,
we may
enter (8) into our theory of meaning:
(8)
(8),
(x)(x satisfies 'Mike' iff x is a Mike)
unlike (7),
does specify the conditions that determine whether or
not something is a semantic value of 'Mike'.
which Mike was in question,
(8)
Where (7)
left us wondering
tells us that any old Mike will do, and
so leaves nothing undetermined.
Burge (ibid p.
432)
recommends that when a proper
in a sentence as a singular term,
name
functions
the sentence should be formalized as
an open sentence: 'Aristotle is human' becomes
19. For a creditable theory of how names get hooked up to their
bearers see Evans, G.,
'The Causal Theory of Names',
Aristotelian
Society Supplementary Volume vol.
XLVII,
1973,
and chapter 11 ot
Evans op. cit.
- 72 -
(9)
Is-Human( [xi] Aristotle(x i ))
The free variable may be quantified from without,
but,
more standarly,
will receive its value from a contextual act of demonstration.
'Aristotle
is human'
Burge's view,
one is
referring to a particular
it is as if one said 'That
In saying
Aristotle.
Aristotle is human',
So,
on
indicating
the Aristotle that one has in mind.
I would,
however,
prefer 'The contextually definite NN is F' to 'That
NN is F' for capturing the flavour of uses of proper names as singular
terms.
For
uses of names
are
very
resembling an act of demonstration
that a speaker will,
particul,
rarely
by
on the speaker's part.
in using a proper name,
individual
accompanied
for his audience,
anything
It is rather
succeed in identifying a
if in the context of utterance
there is a most salient bearer of that name.
When the context is thus
propitious the speaker has to do nothing more than utter the name in
order successfully
to refer.
No demonstration
context is not thus propitious,
the speaker is
is
required.
And if the
then the audience will not know to whom
trying to refer.
(C.
f.
Evans ibid.
p.
374,
for some
considerations in a similar but not identical vein).
Burge offers two arguments for his demonstrative-ridder,
formulation,
neither of which are satisfactory.
His first argument is that 'Jim is 6 feet tall', like 'that book is green'
are (ibid.
p.
432):
incompletely interpreted - they lack truth value. The user
of the sentences must pick out a particular (e.
g.
a
particular Jim or book) if the sentences are to be judged
- 73 -
on
reliance
true
or
false.
It is
this
conventional
extra-sentential action or context to pick out a particular
which signals the demonstrative element in both sentences.
But
the
behaviour
of
'The
contextually
definite
F' does
relevantly different from demonstratives in this respect.
definite
contextually
assessing
the
truth
dog
is
hungry'
value of my
you
claim
are
than
no
seem
If I write 'The
more
if I write
not
capable
'that
book
of
is
green' or 'Jim is 6 feet tall'. If one is attributing to the object language
a bivalent
logic,
then
all three
sentences
attributing to it truth value gaps,
come
out
false.
If one
is
then all three will be seen to lack a
truth value.
Burge's
second argument
is that proper names,
always seem to take the widest possible scope.
of 'the contextually
ensures
a
that
'contextually'
definite'
definite
(compare
description
like demonstratives,
But this also seems true
'the actual' as an operator that
be
read
singles out the actual context,
rigidly).
The
modifier
and so makes the definite
description act similarly to a demonstrative.
So
I
would
recommend
that
(8)
be
employed
in
the
proof
of
T-sentences that tell us this sort of thing: 'Mike is F' uttered by s at t
is true iff in s's context at t,
Mike,
there is exactly one contextually definite
and he is F.
Our problem was to find way to construct clauses
that allow
us to
show the sense of a name that has a bearer,
by specifying its referent,
but that also allow us,
ontological
without
undesirable
convey the senses of bearerless names.
- 74 -
commitment,
to
We have found three ways of meeting the requirements.
use clauses like (5)
We can either
with a free logic or use Quinean predicates,
as in
(6); either of those two options would be suitable for names that have a
primary purported
amendment)
referent.
Or we could follow
and employ clauses like (8),
Burge (with suggested
with appropriate supplementary
mechanics for proving the right T-sentences.
The possibility of exploiting these methotds does not suffice to show
that empty proper names do have senses,
where this is to mean
they
genuine
can
be
used
to
express
or
report
argument that we were considering is met.
of
giving
a
name's
referent
assumption was rash.
- 75 -
But
the
That argument depended
the assumption that only clauses of the form (1)
function
thoughts.
that
and
on
could serve the duplex
showing
its
sense.
That
5 Fourth Argument
5.1 First Case: Perception
For
che
McDowell
countenancing
of
narrow
content
as
a
general
feature of propositional attitudes carries a commitment to 'the idea that
thought
relates
to
objects
with
an
essential
various ways in which this indirectness
description
theory of reference
might come up,
293),
There
are
for example a
for names and demonstratives,
resembles a 'sense data' account of perception.
(p.
indirectness'.
or what
As he puts it in DRS
crediting demonstratives with 'senses that determine objects in
such a way that that the senses are expressible
exist or not' ...
whether the objects
generates a falsification of demonstrative thought
akin to the falsification of perceptual experience that is
induced by representative realism. Representative realism
postulates items that are "before the mind" in experience
whether objects are perceived or not, with the effect that
when an object is perceived, it is conceived as "present to
the mind" only by proxy.
And Evans (op.
cit.
p.
199)
suggests that resistance to a WCO (he
calls it 'Russellian') theory of demonstrative thoughts might arise from a
picture of mind that
is already rejected when one acknowledges that if in
perception anything is before the mind, it is the public
not some internal representation of
objects themselves,
them. The thesis we have been considering [=the WCO
thesis] is
really no more than a corollary of this realism.
McDowell and Evans are claiming that the supporter of narrow content
- 7o -
is committed,
willy nilly,
and perception.
theory
of
to a faulty picture of
reference,
knowledge,
This picture suffers the defects of the old 'sense data'
perception,
Berkeleyan idealism,
and
has,
as
its
least
or solipsistic skepticism,
How does the argument go?
Realism (not
endearing
feature,
as a consequence.
"representative
realism",
view that McDowell thinks cannot but collapse into skepticism)
upon holding that
when one
perceives
an object,
mind' (as Evans puts it) really is the object.
but that what was before the mind
the object.
an
a
depends
what is 'before
the
Suppose it were not so,
was some mental representative
of
Then we should have to deduce the existence and nature of
external
object,
from
deduction could be sound,
as it is,
a
the
representative
itself.
But
no
since the representative could exist,
in the absence of any such object.
such
exactly
And it becomes a mystery
how we can even form a conception of any occupant
of a reality that
lies thus for ever beyond our ken.
What goes for perception goes for cognition
is,
the consequences
of an NCA
generally.
view are dire.
So,
the fear
If a sense or concept
always comes between us and the external objects of our thoughts,
from
these
senses
or
concepts
there
existence of the external objects,
is
no
sound
deduction
and
of the
how can we ever know the object is
there, or even understand what it would be for an object to be there?
We will be,
inhabited
by
as Dummett puts it 'inescapably
thoughts
and
their
components
- 77 -
locked into a third
...
unable
to
realm
reach
the
outer world of actual physical things'.
20
5.2 Reply to First Case
However,
I am disturbed by the threatened conclusion.
I cannot find
that leads into the mire.
anything that is essential to the NCA view,
and argue
of the view,
I shall try to state the minimal requirements
So
against Evans and McDowell that the paths I need tread are hardly more
perilous than those open only to the WCO theorist.
Embracal of narrow content involves no more than the claim that to
think (have an
distinct thing,
experience)
y,
of x involves
that both refers to (is
content of a sort that is independent
common component
quite rightly,
of the
dangerous
turn their backs on.
having in one's mind some
of)
of x.
views
x,
and that has genuine
Such a claim is indeed
that Evans
and
a
McDowell,
But to say just that would provide
the materials for only the most pitiful of inductive arguments.
And that
is not what McDowell and Evans intend.
Certainly there are ways of interpreting the relations that hold among
the
various
trouble.
elements
included
the
NCA
picture,
that
will
invite
In particular if we thought of the narrow content of a concept
as the object of the concept,
could perceive,
content,
in
then we would have to explain
or even conceive of,
out there in the
20.
Dummett,
ibid.
As
McDowellU/Evans argument.
something lying beyond the narrow
real world.
I
how we
implied,
- 78 -
Evans and
Dummett
McDowell
does
not
frequently
accept
the
saddle
the
NCA
view
with
this error.
The
misconstrual of the
position is apparent in the quotations provided above,
in other
that
places.
begins
with
McDowell
the
for example
premise
(CDK)
whether his experience is deceptive or not,
conclusion
that experience
provides
'at
best
and quite explicit
considers
that a subject may
an
argument
not be able
to
tell
and proceeds towards the
a defeasible
ground'
We shall be looking at a similar argument shortly,
knowledge.
NCA
for
but now I
want to draw attention to how McDowell elaborates on it.
He reformulates the argument:
In a deceptive case, what is embraced within the scope of
experience is an appearance that such and such is the case
falling short of the fact; a mere appearance. So what is
experienced in a non-deceptive case is a mere appearance
too.
Then,
in
the
next
paragraph
hle
refers
back
to
a
part
of
the
argument with the words 'As before the object of the experience in the
deceptive case is a mere appearance.' And he goes on,
understandably,
to claim that in the non-deceptive case the object of experience is not a
mere appearance,
but some part of the real outside world.
Surely enough some famous
works have contained
an
argument
that
began with the premise that one sometimes cannot distinguish deceptive
experiences from non-deceptive ones, and ended up with the conclusion
that what is
experienced,
the object
two cases.
The next
natural stopping
scepticism.
But that
is
that
McDowell
appeared
of experience,
an importantly
to
start
out
considering .
- 79 -
place is
the same in the
some sort
different
with,
is
argument
and
ought
of idealism or
from the one
to
have
been
The
latter
conclusion
moves
argument
that the experiences,
the same in the two cases.
of experience
are
to
the
in and of themselves,
are
from the
simply
considered
premise
stated
Such a claim does not entail that the objects
the same.
And the
sensible
move
at
this stage
is
simply to deny that they are. In the sense in which what I experience,
the object of my experience,
thing in the world,
experience,
is,
in the non-deceptive
case,
there is nothing that is experienced,
in the deceptive
case.
Yet the experiences
be the same in their non-relational
characteristics;
some real
no object of
themselves may
and that is all that
is required by the NCA theorist.
Availing myself of this essential distinction between the non-relational
character of an experience,
and the object of which it is an experience,
clears me of the crude version of McDowell's and Evans' objection.
underlying
thought,
the
crude
form
is
a
more subtle
and interesting
But
line
of
that proceeds without the needless slide.
To keep my feet on firm ground,
I must show that cognitive relations
to the physical world can constitute kr.owledge,
even if they consist in
having in one's mind something that could exist, even if the rest of the
world
did not.
So the
following sort of story
know that there is a keyboard in front of me,
must
be
intelligible.
I
because I see it there.
My seeing it involves my having a visual presentation that could be just
as it is even in the absence of the keyboard (I could be hallucinating).
Yet,
nevertheless,
I see the keyboard and I know that it is there.
To McDowell stories like the little one told by my last three sentences
seem
highly
problematic.
Objecting
to
- 80 -
a
'criterial'
view,
he
writes
(ibid.) that it
envisage [ s
ascribing
knowledge
on
the
strength
of
something compatible with the falsity of what is supposedly
known.
And it is a serious question whether we can
understand how it can be knowledge that is properly so
ascribed.
The question is serious for the following reason.
If I admit that the
basis for my claim to knowledge that p is the experience,
claim that my experience is compatible with not-p,
then I appear to be
committed to a pair of claims that sit uncomfortably
My experience legitimates my claim to know that p.
being what it is,
and I also
on the samt
knee.
But, my experience
I must also concede that for all I know,
not-p.
The
discomfort may be heightened by the observation that if for all I know,
not-p, then it follows that none among my items of knowledge rules out
not-p.
And that entails,
by elementary logic,
that the knowledge that p
is not one of those items.
I require it to be
possible
for an
experience
ground for a claim to knowledge that p,
with
the
failure
of
p
to
obtain.
I
anr4
must
to be both legitimate
such that it is compatible
therefore
deny
that
the
compatibility of the experience with not-p forces the concession that for
all I know,
is
not-p.
tantamount
And this seems right independently.
to the
quite
harmless
assertion
For that denial
that one
may
be in
a
position to claim to know that p on the basis of evidence from which p
does not follow logically.
McDowell
between
things
(ibid.)
accepting
not being so,
does not see how it is possible
that
everything
that
on the one hand,
-
81 -
one
has
to 'drive a wedge
is
compatible
with
and admitting that one does not
know that things are
so,
on the other.'
In
particular,
he
holds,
no
appeal to conventions governing the use of the word 'know' can help us
out
here,
since
we
would
have
to show
that
the
conventions
were
'well-founded'.
But,
(among which,
on this particular matter if few others, the NCA view is
to be counted)
concedes 'that the sceptic's complaints are substantially
correct'.
And
that
McDowell
says,
concession,
tht
he
'criterial'
says,
view
cannot
and its fellows
leave
room
for
any
appeal to the well-foundedness of the conventions.
It is perhaps unsurprising that each side thinks that it is the other
who concedes
too much to the sceptic.
All that I am conceding
to the
sceptic is that the experiences that are available to justify my claim to
know that p may be expected to be compatible with not-p.
how
this
undermines
the
well-foundedness
governing the use of the word 'know'.
Or,
the
'know'
same vein:
given
what
the
word
of
the
I cannot see
"conventions"
to make a stronger claim in
means,
and hence,
what
knowledge is, the sceptic is right at least in saying that what grounds
our claims to knowledge does not always establish it beyond all possible
doubt.
It is just a fact about knowledge,
that one's posession of it may
be based upon something that does not logically entail the truth of what
is known.
The contrary supposition must stem
from a need
that what distinguishes a subject who knows that p,
not,
must
considered
lie
in
in
and
the
characteristics
of themselves,
of
rather
the
than
from one who does
subjects'
in,
say,
thing to which the experience may be suitably related.
that which the sceptic holds,
to believe
experiences,
some
external
But it is exactly
and that we have no need to concede.
-
82
-
Consider two subjects whose experiences are the same,
warranting
a
claim
to
knowledge.
Suppose
that
in
in both cases
one
experience reveals the truth, and in the other, it misleads.
the subject really does know what he claims to know,
claim
is
in
fact
adequately
false,
grounded,
but
still
justified,
in
case
the
In one case
in the other, the
the
sense
of
being
so that the subject was quite right to make it.
McDowell (ibid.) goes so far as ask.
How can a difference in respect of something considered
cognitively
inaccessible
to
both
subjects
...
make
it
as
the
case that one of them knows how things are in that
inaccessible region while the other does not -- rather than
leaving them uoth,
strictly speaking,
ignorant of the
matter?
It is a fear that
McDowell
to
considered
question
hold
this sceptical
that
the
by themselves,
is
one
fearsomeness
groundless.
that
question has
experiences
be
the
McDowell
of
same.
also
no answer,
the
two
subjects
I shall argue
must
face,
or otherwise counts neither way.
that
in
leads
cannot,
later that the
which
case
its
But I think the fear is
If we take a closer look at the question we can see that it
lacks potency.
The
'difference
in
respect
of something
cognitively
inaccessible
to
both subjects' is simply a difference that does not effect the intrinsic
nature of the subjects' experiences,
outside
world.
inaccessible'
For if we
course
formulation
regions quite unnecessarily
take
even
difference
McDowell's
considered in abstraction
in
cognition
to include
the
proponent
the
surroundings
of
in
terms
as
narrow
content
of
two
- 83 -
'cognitively
invites the sceptic's challenges.
knowledge,
the
of
from the
we
should,
can
hold
subjects
is
then
that
of
the
'cognitively
accessible'.
No skeptical conclusions will then follow.
If we adopt a restrictive use of 'cognition',
then the region in which
hides the distinguishing feature of the two cases,
is indeed cognitively
inaccessible.
answering
question.
we
between
not
no difficulty
in
the obtaining of what the subjects
obtaining.
entails that p,
have
in the 'cognitively inaccessible'
The difference
difference
and its
But then
The
very
boring
fact
that
'S
McDowell's
region is the
claim to know,
knows
that
may know that p,
thus ensiu:es that one subject
p'
while
the other does not.
The proper way to respond to the sceptic is not to search for some
feature of our experiences that guarantees their accuracy,
it plausible
that even if
things not being so,
are
so,
hard.
on the one hand,
on the other".
Let me offer
"everything that one has
I do not think
but to make
is compatible with
one may yet know that things
that
the task would
the briefest of caricatures
be
very
of how the tale would
unfold.
'Everything that one has' means everything that one may legitimately
draw upon in ascertaining how things are.
one has is compatible with,
say,
To say that everything that
its not raining is just to say that it is
logically possible that everything that one has could be just as it is,
and it not be raining.
the visual,
auditory,
Suppose thae everything that you have includes
olfactory and tactile sensations typically associated
with being out in the rain.
just as it
order,
should
that
be;
Everything else is,
there is
you are being
as fa
as you can tell,
no evidence that your senses are out of
tricked,
or that anything
- 84 -
is
disturbing
the
status
quo.
raining.
If
You
have,
in
short,
overwhelming
that
it
is
But that is all you have.
you
have
overwhelming
evidence is caused,
evidence
that
in the usual manner,
you know that it is raining.
it
is
What more could possibly
your
legitimate
the claim to knowledge do not logically
raining.
claim
to
raining,
knowledge?
and
this
by its actually raining,
legitimate
must
evidence
But
still,
the
then
be required to
conditions
guarantee
that
that it is
To suppose that what grounds a legitimate claim to knowledge
logically
guarantee
that
what
is
claimed
to
oe
known
actually
obtains, is simply to confuse knowledge with a kind of certainty that is
a
philosopher's
confusion,
The
certainty.
One
might
expect
a
sceptic
to
make
this
but it is hardly prudent for the rest of us.
pr sition
adequate
invention.
room
that McDowell
for
It seems
the
attempts
required
to me,
to adopt is
distinction
however,
that
supposed
between
to leave
knowledge
the manoeuvres
and
required
to
make this room, lead McDowell into exactly the admissions he strives to
avoid.
On his articulation
two
subjects
of the WCO position
whcse
epistemi,
outside of their heads,
For one subject,
facts':
rather,
perceptually
be
'mere
states
only
in
what
went
manifest' to him.
This
on
states.
would be held not to 'fall short of the
experience
appearance'.
differed
between our
must be written into their experiential
his experience
his
the difference
would
just
For tae other,
poses
a
risk
be
the
what is
, our
fact
itself
'made
experienced
would
fallibility.
If
experience of its raining were truly different when it was raining,
-
85 -
the
from
how could we fail to distinguish the two? And
when it was not raining,
how could we be fooled by the
if we could alwals distinguish the two,
deceptive experiences?
is obvious
answer
McDowell's
enough.
content
The
a
of
one is not
at least,
experience is not fully accessible to introspection:
subject's
One
always in a positien to tell which sort of experience one is having.
may
raining,
fact
is the
experience
whether one's
not be able tell
that it
is
within one's cognitive reach, or a mere appearance that this is
Thus he concedes that 'there may be deceptive cases experientially
so.
indistinguishable
from
indistinguishability
of
themselves,
cL urse
the
in
experiences
but only in the fact that a subject may not know which of
So it is not conceded
the two he is enjoying.
there exist two experiences,
which experiences
conceded
that the mere appearance
(not
the
be
reside
cannot
This
(ibid.).
cases'
non-deceptive
same)
as
the
that in the two
are identical.
of its raining
reality
of
its
cases
It is only
may seem the same
raining,
when
that
is
experienced.
But even this meager concession is fatal. McDowell concedes that the
subject
cannot
tell
two experiences,
has
written
he
has
an
accurate
experience
or
a
But if a subject cannot tell the difference between the
deceptive one.
ascertaining
whether
then that difference is not something he may exploit in
which of the
into
the
two
he has.
subjects'
So the difference
experiences
cannot
"elevant.
McDowell (ibid.) attempts to answer this criticism.
- 86 -
be
that McDowell
epistemically
One could hardly countenance the idea of having a fact
made manifest within the reach of one's experience, without
supposing that that wodld make knowledge of the fact
available to one.
The root idea [behind the criticism) is that one's epistemic
standing on some question cannot intelligibly be constituted,
even in part, by matters blankly external to how it is with
one subjectively. For how could such matters be other than
beyond one's ken?
When someone has a fact made manifest to him,
the
obtaining of the fact contributes to his epistemic standing
on the question. But the obtaining of the fact is precisely
not blankly external to his subjectivity
The
criticism
is
not
answered
by
a
restatement
of
the
contested
conclusion (the first and penultimate sentences in the quoted sections),
and a reformulation ot the objector's point in a vocabulary that obscures
it
(the
external
second
paragraph).
to
it
McDowell
how
carl
is
certainly
with
Like
one
respond
'cognitively
inaccessible',
subjectively'
is
a
to
by
saying
its
use
experience is a fact within one's cognitive
reach,
simply from the concession
between the two experiences.
slippery
phrase.
that
if
an
then the fact is not
blankly external to how it is with one subjectively.
proceeds
'blankly
that a subject
But the objection
cannot distinguish
It does not matter whether the difference
between the experiences is counted as being part of how it is with the
subject subjectively,
things,
or not.
If he cannot distinguish between
then that distinction
is
not
something
that
he has
the swo
to
go on
when it comes to deciding how it is in the world around him.
The place at which we end up,
then,
McDowel]'s view and its consequences,
theorist
is
already at rest.
after winding our way through
is exactly the one where the NCA
We are left with the task of explaining
how
it can be that when "everything that one has is compatible with things
- 87 -
not being so,
on the one band,
on the other".
how
the
one may yet know that things are so,
And that of course,
contents
is just
of a 'cognitively
facing
inaccessible'
the question about
can make
region
the
difference between knowledge and lack of it.
6 Second Case: Reference
The
argument
McDowell
(SRPN)
goes
a
says
that opposition
little
differently
in
the
to his
conception of how thought relates to reality',
case
of
reference.
view involves a 'suspect
a conception which 'would
have to seek its support in the idea that' reference is mediated 'by way
of
a
blueprint
or
specification,
expressed
in
purely
conception
as
one
knowledge
general
that,
we discussed,
unable, in the end,
which,
terms.'
like
the
on
the
formulated,
McDowell
faulty
and
theory
would
Evans
of
be
see
this
perception
and
must keep the mind locked up within itself,
to conceive of anything outside.
Perhaps that last claim is correct.
for
if
suspect
conception
How would reference be accounted
of
how
thought
relates
to
reality?
Concepts would have to contain fully individuative
descriptive elements
to take up the burden of reference.
The concept
would then refer to
uniquely satisfied the description.
The likely difficulty
whatever
would
be
elements
object
to
find
descriptive
themselves
object-involving
would
components,
elements
have
to
or the
relates to reality would be forfeit.
that could
be
free
do the
of
any
original conception
And then,
- 88 -
trick.
These
indexical
or
of how thought
surely enough,
the theory
would never allow the mind to break outside of itself.
The argument
would take some spelling out, but it looks fair.
6.1 Reply to Second Case
However,
none of this need concern the proponent of narrow content.
It is no part of the
concept
general
NCA
view that the narrow
is a 'blueprint or specification which,
expressed in
purely general terms.'21
content of a
if formulated,
The NCA
position
would
be
requires only
that the narrow content of a concept be that in the subject which allows
him to refer to the object,
not have to take up the
counts.
The
first
and that it be recognizably content. It does
full burden
constraint
of reference
requires
only
that
on either of these
narrow
suitable to take up part of the burden of reference.
requires
that
referring
concepts
content that is reference-independent,
content
be
description.
think,
semantic
content
in
form
do not accept
we find one of the deepest
The second
some sort of mental
but does not
expressible
Evans and McDowell
have
be
The rest may be
supplied by external relations between subject and object.
constraint
content
of
require
an
this last;
points of disagreement
that thi1
individuative
and here,
I
between the
NCA and the WCO theorists.
21.
This is not violated by any of the treatments of names that I
recommended for the theory of meaning. True enough, 'is Mumbo-Jumbo'
may be used in a specification of necessary and sufficient conditions for
being the referent of 'Mumbo-Jumbo', and it shows the name's sense.
But nothing is implied about what it is that determines what the name
refers to.
- 89 -
view,
that on it
their extensions.
solipsist
[=NCA
representational
NCA
narrow contents must by themselves fully individuat,
Thus Evans (ibid.
wants
theorist]
thinking
recognizably
as an essential part of the
and McDowell take it
Both Evans
-
p.
202)
wiites 'The methodological
[=narrow
M-thinking
requires
which
at
least
content)
to
be
it
be
a
that
state.' And 'It is of the essence of a
representational
state that it be capable of assessment as true or false'.
It does follow
from this thesis that the narrow content of a referring concept
have to be fully individuative of its referent.
would
The reason for this is as
follows.
A thought of the form a is F where a is a referring component,
be true iff the referent of a satisfies F.
will
Now suppose that the thought
is assessable as true or false purely in virtue of its narrow
content.
Then the narrow content of a must determine which thing would have to
satisfy F,
reference-independent,
But it is not,
NCA view
to be true.
for the thought
recognizably
narrow
content
that
The condition
corresponds
thinking,
was
of a thought
only
to
that
it
without
thought must
its
being
be that
in
the
case
that
any part of the
must
be capable
that I placed
Evans'
deserving of the second part of its name.
status
narrow content is
Evans and McDowell notwithstanding,
that the
and
since
it could only do this by description.
assessment as true or false.
content,
But,
be
condition
recognizably
- 90
-
upon narrow
that
it
mental,
be
and
Narrow content can have that
the
narrow
content
virtue of which it is assessable
false.
of
of
any
as true or
Evans moves too quickly from the requirement that narrow content be
taken by
recognizably thinking to the requirement that it be assessable,
itself,
the
for truth value.
subject
he
when
appropriately
to a
Why shouldn't thinking just be what goes on in
undergoes
surrounding
states
context,
which,
become
There is nothing in the notion of a though
notion,
when
hooked
evaluable
(the lay,
for
up
truth?
not the F'regean
of course) that dictates that its pi ssession of truth conditions,
or truth
value,
given a context.
is intrinsic
to it,
rather
than something
it
has only
22
6.2 Points of View
Thought considered
be content,
purely
and recognizably
as the possessor of narrow
so.
Consider,
example involving McDowell's native.
specified,
case
with
there
concept
a
Mumbo-Jumbo
Mumbo-Jumbo,
Mumbo-Jumbo
properties,
upon
is
no
the
properties
subject
reft rs
himself,
a deluded
godlike
him.
the existence
viewed
may
a Twin Earth
One case is as McDowell originally
and
to
if you will,
content
in
But
native.
nature,
the
In the
and
native's
the
native's
individualistic
and nature of which depend
in
abstraction
from
other
any
only
particular
surrounding context, are the same in the two cases.
Whether or not the native's thoughts differ in truth conditions across
22. 1 think that any suggestion that the issue is becoming merely
terminological, would be misguided.
Questions about what should be
called 'thinking' or 'content' are better seen as questions about what
thinkillg and content are. There is
a genuine issue here, and an
important one.
- 91 -
the two cases,
to transplant
we were
would
not
If
the way the world appears to him will be the same.
notice.
the native
two
The
subject's point of view.23
to
from one context
cases
are
differences
Whatever
the
therefore
same
are
there
he
the other,
from
the
between
the
two cases, are ones that do not affect the way the world appears to the
native.
If the subject's point of view fails to determine the extension of
his concepts and the truth conditions of his thoughts,
then this shows
that there is a level of content that is not truth-conditional,
thought that is not,
a kind of
taken by itself, assessable for truth value.
It does
not show that the subject's subjective point of viewy has no content.
When
view,
want to
we
understand,
what motivates someone,
from
a sympathetic
human
what guides his thoughts
point of
and actions,
precisely what we try to find out is how things are from that person's
point
of
view.
for
Suppose,
example,
that
you
are
interested
in
understanding some aspect of Freud's thought that involved his concept
of the
man,
by
Unconscious.
I am imagining
and his picture of the world,
studying
conception
Freud.
of what the
Suppose
Freud,
the
not trying to learn about the world
that
Unconscious
expression 'the Unconscious',
trying to understand
to
begin
with
is supposed
you
to be.
have
little
You learn
an
with which you associate a concept that is
little more than a place-holder in your thought.
To
understand
Freud's
concept
23. A point of contact here
defence of narrow content.
with
better
you
Blackburn's
- 92 -
must
try
to
(Blackburn
develop
op.
cit.)
a
Freudian picture oft the world.
You must try to see the world through
Freud's eyes.
You ask yourself 'How
was
in
manifest
unconscious
enterprise
this
thoughts
would
attention
to
anything
else
of actions?',
might become
be
spoiled
whether
in it
series
if
Analogously
if
conscious?'
think
and so on.
constructing
the
Freud
the
picture
existed,
objective
or
world,
that
The whole
you
to
paid
whether
rather
than
a
If Freud believed in something that you
If Freud did
not believe in something that
you do not include the thing in your picture.
you
native's point of view,
does,
did
you still include the thing in your conception of the
world according to Freud.
you do believe in,
'How
Unconscious
part of
figment of Freud's imagination.
do not believe in,
in
or not the
was
would Freud say the Unconscious
want
to
understand
how
things
why he thinks what he thinks,
are
24
from
the
and does what he
you must pay no attention to whether or not Mumbo-Jumbo exists.
Your sympathetic
will suffer
understanding
if differences
of the workings
between
pictures of the world according
the
two cases
to the native.
of the
are
native's
built
into
mind
your
If your models of how
things appear from the native's point of view make it seem as though
things appear differently to him across the two cases,
you have made a
mistake.
If that much of the contents of the native's mind as appears in your
24. Here there is some harmony with Dennett, D., 'Beyond Belief', in
Woodfield,
A.,
ed.,
Thought and Object, Oxford University press,
1982, though we have different ideas about which world best represents
the subject's point of view.
-
93 -
model is insufficient,
this shows that
taken by itself,
to be
recognizably
to be assessed
tor truth value,
requires less than Evans
thinking
For the understanding that is given to you by the picture is
supposed.
an understanding of the native's conception of the world.
the
of
world,
thus
viewed,
solipsistically
is
A conception
thinking.
recognizably
Indeed thought is more easily recognized in this way than, in any other;
the thoughts from the inside,
for one recognizes
from the perspective
of the thinker.
Understanding
everyday,
person.
sort of way,
point
of
view,
in
this
sympathetic,
involves adopting that point of view in the first
One does not have to believe,
to believe,
merely
another's
or even try to believe,
or feign
that the world is as it appears from that point of view.
has to imagine
how
the world
would
appear to someone
One
whose
point of view it was.
To be
appear
sure,
from
another's
extra-cranial
strives
what one envisages
things.
to
point
But
understand
of
this
wide
while
view,
does
trying to grasp how things
is
not
rather
a
world
render
than
of
extra-mental,
the content
narrow.
The
that
one
content
is
understood in terms of how the world appears when conceived through
it,
by of a world,
not upon what there is,
I,
my
in
world,
can
understand
residents of Putnam's Twin Earth,
there
is
no
substance,
thoughts
water,
XYZ.
of a
and
When
Twin
the
to be conceived of.
how
the
by thinking of water.
extension
I think of
Earthperson,
of their
water,
I do
- 94 -
to
appears
world
concept
the
On Twin Earth
is
some
other
in order to understand
the
to
that
not
need
to
attribute
person a concept that has the same extension as the one I am employing
myself. The other's concept need not have any extension at all.
note only of how my concept presents its actual extension,
is thereby presented.
to
grasp
extension.
is
I take
not of what
The content of the concept that I thereby come
therefore
narrow;
The actual extension
ladder to the narrow content
it
remains
of my
(one must,
the
concept
is,
same
whatever
one might say,
so to speak,
its
a
throw it away
after one has climbed up on it).
McDowell (SRPN)
says this:
In practice an interpreter might say things like 'This man is
that Mumbo-Jumbo
brings thunder',
and might
saying
explain an utterance which he described in this way as
expressing the belief that Mumbo-Jumbo brings thunder.
That is no real objection. Such an interpreter is simply
playing along with his deluded subject - putting things his
way.
Understanding
playing along
how things are from another's
with
her,
putting
things her way.
think that the kind of understanding
putting things her way,
point of view requires
of a subject
I see
no
reason
to
that we derive
by
is inferior or defective in any way.
Nor do I
see any reason to think that what is thereby made accessible to us is
something less than the contents of a person's minrd.
- 95 -
PAPER THREE
SEEING WHAT IS NOT THERE
- 96 -
The Eye altering alters all
- William Blake
Seeing What is not There
I Introduction
1.1 Burge's Claim
1
Tyler
Burge
sets
out
to
vision 2 is not individualistic.
show
that
the
computational
theory
of
He tells us that
According to individualism about the mind,
the mental
natures of all of a person's or animal's mental states (and
events) are such that there is no necessary or deep
individuative relation between the individual's being in
states of those kinds and the nature of the individual's
physical or social environments. (Burge, pps. 3-4).
When Burge claims that Marr's theory is not individualistic,
means
is that
there
is some
necessary
or deep individuative
what he
relation
between some of the representational contents3 attributed by the theory
1. Burge, T.,
Jan. 1986.
'Individualism and Psychology',
The Philosophical Review,
2. 1 shall follow Burge in referring to it as Marr's theory. As Burge
notes,
very
substantial
contributions
have
been
made
to
the
and
most complete
The
others.
by various
theory
computational
accessible account of the theory is to be found in Marr's book, Vision,
San Fransisco: W. H. Freeman and Co.,
1982. All references to Marr
are from the book, unless otherwise specified.
3. By 'representational content' I intend something that is individuated
syntactic
Two identical
or intentional properties.
by its semantic,
- 97 -
to the visual system, and aspects of the extra-cranial environment.
The point of the qualifier 'necessary or deep' is to make it clear that
the kind of individuation in question is grournded in essential properties
would, ipso facto, be debarred from belonging to the individuated
on Marr's
claim is that,
subject,
in
a
such
that
environments
representational
kind.
are
contents
representational
to certain features of the world outside the
by reference
individuated
theory,
it
properties,
the individuating
just that if any state or event lacked
Burge's
it is
The essentialism need not be Aristotelian;
of what is individuated.
way
that
lacked
a
contents
the
contents
features.
those
subject
capable
is
could
not
exist
in
the
Consequently
of
having
are
not
determined solely by what we might call her individualistic properties,
that depend
properties
boundaries,
entails
down
contents,
I
solipsistically regarded.
that
to
think
only upon how things
two
the
last
subjects
whose
quark,
could
are within
Marr's theory
According to Burge,
were
type
identical,
representations
with
different
physical
have
her physical
states
if they inhabited suitably different physical surroundings.
that
straightforward
Burge
and
is
mistaken.
reasonable
Marr's
theory,
interpretation,
at
is
least
on
a
perfectly
individualistic.
different
if
you
prefer)
(or
have,
instantiate
tokens
may
representational contents. A representation is a syntactic object, though
typically it will have semantic
properties.
-
98 -
1.2 Preview
In the next section I shall explain relevant aspects of Marr's theory,
as I understand it,
pausing to note points that will be important to the
subsequent
argument.
In
that Marr's
theory
not
what
is
at
is
issue.
The
section
3 1 shall
individualistic,
issue
will
recount
and
turn
explain
out
to
description we should expect from Marr's theory,
counterfactual
surprising,
enjoyed,
conditional.
which
Each
conditional
rep resentational
contents
according
vary
across
to
Marr's
theory,
environments,
individualistic.
and
In sections 4,
own account of the
proper
passages
cohere
upon
subject
as
is
theory
reaction
to the counterfactuals.
my
would
is
giving my
The final
own
I shall show that
understanding
of
theory.
2 Marr's Theory
Marr (p.
99) formulates his task thus:
Our overall goal is to understand vision completely, that is,
to understand how descriptions of the world may efficiently
and reliably be obtained from [retinal] images of it.
The
- 99 -
not
Burge cites some passages from
takes to support his position.
with
have
Burge argues
that
the
not
would
contents
therefore
the
of a certain type of
5 and 6 I shall dispute this,
perfectly
precisely
her representational
section will be concerned with exegesis.
Marr that he (Burge)
more
concerns,
a
argument
depend
had she graced a suitably different environment.
that
the
Burge's
Marr's
example of a machine that can
human system is a working
make
...
descriptions
such
of
one
our
aims
is
to
at all levels: What kind of
understand it thoroughly,
information does the human visual system represent, what
does it perform to obtain this
kind of computations
it
represent
this
information,
and
why?
How
does
information and how are the computations performed and
with
what
...
algorithms?
How
are
these
specific
representations implemented in neural machinery?
As Burge
p.
24-27).4
notes,
i)
the enterprise
computational
The
has three separate parts
(Marr
theory
p.
'in
24)
(see Marr
which
the
performance of the device is characterized as a mapping from one kind
to another,
of information
the abstract properties of this mapping
are
and its appropriateness and adequacy for the task at
defined precisely,
hand are demonstrated.' ii) The theory of representation and algorithm: .
the information
must be
represented
with specific primitives that make explicit
computations
must
performed
be
by
representational
system,
specific information,
and the
in some
some
particular
algorithms.
The
second part of the theory describes the representations and algorithms.
The
iii)
system
theory
of
is physically
hardware
realized.
implementation:
In
what
follows
this
we
describes
how
the
shall be concerned
with the representational aspects of the theory.
The computational theory of vision is a theory that treats the human
visual system as a machine that constructs descriptions of the world in
terms of three dimensional objects,
given
computational
light
intensity
descriptions,
values
4. For a more sustained
View' in Haugeland, J.,
in
a
from retinal images.
called
'gray
two-dimensional
[he retinas are
arrays',
coordinate
that
represent
system.
Ihe
treatment see 'Artificial Intelligence: a Personal
ed. Mind Design, Cambridge, M. 1. T. Press,
1981.
- 100 -
computational
problem
is:
given
gray
arrays
as input,
how does the
system produce the complex descriptions of 3D objects as output?
The process breaks up into three stages (see Marr
The
first
has
two
parts.
Initially
the
system
p.
37,
must
p.
construct
description of the patterns of light and shade on the retinas;
42)
'representations
330).
a
(Marr p.
are obtained of the changes and structures in the
image'. These representations make explicit certain features that will be
useful
for
the
computations
construction
are performed
outside the viewer,
primal
see section 7,
sketch,
surface,
depth,
the
explicitly
in
a
constructs
describes
below).
system
reflectance changes
constructs
2.5D
a
on it.
This is
of this view of the primal
2.5D
information
coordinate
sketch
shapes and their
certain
At the second stage, on the basis of the
representing
the
Next,
a first sketch of the surface
(for a qualification
viewer-centered
from
representations.
to construct
that represents
called the 'primal sketch'
sketch,
of higher
a
of
the
about contour,
frame.
31)
orientations
sketch
Finally
model
relative
viewed
and
the
rough
system
representation,
that
to an object-centered
coordinate frame.
The passage from each stage to the next is regarded as the making
of an inference: Marr writes (p.
68) 'the true heart of visual perception
is the inference
from the structure of an image about the structure of
the
outside.'
real
world
information
explicit in
the
beyond anything available
poverty
of
stimulus
At
each
stage
representations
what
is
inferred
that are operated
from that information alone.
typical
of
problems
-
101
-
in
from
upon goes
There is
cognitive
the
thus a
science.
T[he
needs
theorist
to discover what a priori assumptions
5
bear by the system, that allow it to make the inferences.
c.
are
brought to
(Marr p. 219,
f. Burge p. 33)
The question to ask is, What assumptions are reasonable to
make - that we unconsciously employ - when we interpret
silhouettes ... as three-dimensional shapes?
If the visual system failed to make such assumptions, it would not be
able to infer the nature of the distal causes of its surface irritations
from
those
irritations.
representations
world
is,
in
the
Since
visual
to
important
very
actual
the
under normal conditions,
practice,
system
the
accurate
provides
structure
discovery
of the
the
of
assumptions built into the system.
The guiding
princio;le is that when
the
representations
constructed
this is evidence that the
are correct under a given set of conditions,
inferences that yield them are based upon assumptions that are true in
such
conditions.
example,
(Marr
6
p.
The
44)
kind
'the
of
assumptions
visible
world
can be
composed of smooth surfaces having reflectance
structure may be elaborate',
(Marr p.
51)
involved
are
regarded
that,
as
for
being
functions whose spatial
'if direction of motion is ever
5. 'A priori' is Marr's expression. The a priori assumptions are
necessarily true. The point is that they are innate, not learned.
not
6. The assumptions are not expected to be explicitly represented in the
system. So it may be that the mere fact that a system succeeds under
of
vindicate
the
attribution
conditions
is
sufficient
to
certain
Difficult philosophical
assumptions that are true in those conditions.
questions arise here, into which I shall not enter, and for a discussion
Computation And Cognition: Toward A
of which see Pylyshyn, Z.,
Foundation For Cognitive Science, Cambridge, M. I. T.
Press, 1984,
chapter 5.
- 102 -
at more than one point -
discontinuous
within
hold,
the
from
made
arrays
gray
come
the
When
idealizations.
reasonable
-
When the assumed conditions hold,
then an object boundary is present'.
inferences
for example,
along a line,
out
-
correct
assumed
at least,
constraints
do not
illusions tend to result.
I now turn to three points that will be important to our argument.
2.1 First Point
The
of
structure
objects
in
the
space.
computational
problem.
That
system
visual
representations
of the
The
nature
the
exigencies
is to
say,
of
they
of
from
infers,
three
the
dimensional.
lower descriptions
the
solving
must
a)
be
is
overall
recoverable
and b) be used in the level above.
from the level below,
Marr (p.
by
primarily
the
arrays,
gray
objective
determined
how
describes
theory
36) explains:
we arrived at the idea of a sequence of representations,
starting with descriptions that could be obtained straight
from an image but that are carefully designed to facilitate
more objective,
recovery of gradually
the subsequent
an
object's
shape.
properties
about
physical
The
process
inferential,
Each
of
constructing
and the theory of
inference
requires
additional assumptions
an
vision is the theory
input
(innate),
representations
(from
the
level
Each of these must be described
be
the
that
algorithm
regarded
as
of the inferences.
below),
probably
and an algorithm to compute an output
representation.
shown
is
works,
that
- 103 -
the
precisely,
required
and it
must
representations
(Marr p.
are actually computable in the manner described:
to arnalyze
exactly
emphasise,
then,
what
will work,
and
it'.
prove
The
75) 'we have
point
first
I
is that if a representation is attributed to the system,
it must be shown exactly
sequence of inferences,
how the
representation
from the gray arrays.
is arrived at,
by a
Representations require
a botton-up account.
2.2 Second Point
Since representations require a bottom-up account,
cheap.
more
more
computational activity required to derive it from the gray arrays.
Given
constraint
representation.
representation
specific
8
the
this
The
one
Such
needs
motivation
will satisfy,
the
a
informational
they do not come
solid
is
content,
motivation
provided
by
for
some
positing
need
that
some purpose to which it must be put.
a
the
In
the case of lower representations the purpose will be provided by what
is required of the computations at the next level:
The critical question is, What spatial relations are important
to make explicit now, and why? The answer to this, of
course,
depends
on
the
purpose
for
which
the
representation is to be used. For us, the purpose is to
infer the geometry of the underlying surfaces (Marr p.
80).
7. A small qualification: the gray arrays themselves are representations
of light intensity values on the retinas, but we are not offered an
account of how these representations are constructed.
lUltimately,
I
suppose, that account should be forthcoming.
8. More specific in the sense of being compatible with a narrower range
of distal causes: the informatic
that there is a crack on the viewed
surface is more specific than the information that a dark line appears
on it that could be either a crack or something else.
- 104 -
The information
that is made explicit at each stage is carefully chosen
to suit subsequent computations.
The
story
is
much
the
Marr's theory derives,
same
for
the
highest
the 3D model representations.
Stated baldly, the strong constraints come
representation is to be used for. (P. 326).
The representations,
of shapes.
representations
according to Marr,
Here again:
from what the
are used for the recognition
Such a purpose requires that the representation:
(1) use an object-centered coordinate system, (2)
volumetric primitives of various sizes, and (3)
modular organization. A representation based on a
natural axes (for example, the axes identified by
figure) follows directly from these choices.
What is represented
is
exploited
at
representions
'announced'
used
to
the
next
resulting
in the
account
stage.
from
2.5D
for
the
sketch.
All
the
early
What
our ability to
required
for
object
information
processes
appears
discern
in
explicit
is
(Marr
the
2.5D
surface
in
p.
the
268)
sketch is
features,
and
the 3D model representation.
elements of the 3D model representation
is
include
have a
shape's
a stick
at a given stage is tightly constrained by what
input to the algorithm that computes
what
Lhat
as
The
itself are carefully motivated by
recognition.
In
general
representations
require a top-down motivation. 9
9. This must not be taken to mean that the theory ascribes top-down
visual processing. Marr (p. 100) insists that, on the whole, it does
not. Processing is bottom-up, motivations are top-down.
- 105 -
2.3 Third Point
The theoretical role of a representation is given by its place in the
sequences of inferences that begin with the gray arrays,
and culminate
But
eventually in hypotheses about the shapes that give rise to them.
it
must,
course,
of
also
be
shown
that
the
system
visual
in the particular
employs the particular representations attributed to it,
processes attributed to it.
available
clues
to
that
constrain
introspective:
perception.
A variety of sorts of empirical evidence are
the attributions
for their discovery.
allow
actually
of representations
and
provide
The main source of evidence
we know a great deal about what is
revealed to
is
us in
The experimental context in which this sort of evidence is
mostly tapped concerns illusions.
So,
Chapter 3,
for example,
pages long,
are
3 of Marr's
book,
pairs
of
which
contains no less than 15 random dot stereograms.
square
dot-patterns;
the
two
squares
stereogram are the same except for a region in their centers,
displaced slightly in one square relative to the other.
in
each
which is
If one views them
(exactly one square visible to each eye)
stereoscopically
is 48
information from
and explains how we compute stereoscopic
the flat gray arrays,
These
section
(Marr p.
102)
'one vividly and unmistakably perceives a square floating in space above
the background.'
By
varying the density of the dots,
and the amount
of displacement,
one uncovers a great deal of information about exactly
w'hat is required
for the computation of depth.
Evidence is
fairly abundant at lower levels,
- 106 -
but tends to peter out at
There are a couple of global
the level of the 3D model representation.
326):
considerations (Marr p.
First, in order to manipulate objects and avoid bumping into
them, organisms must be able to perceive and represent the
disposition of the objects' surfaces in space. This gives a
minimal requirement of something like the 2.5D sketch.
Second, in order to recognize an object by its shape,
allowing one then to evaluate its significance for action,
some kind of three-dimensional representation must be built
from the image and matched in some way to a stored
three-dimensional representation with which other knowledge
is already associated.
And three additional pieces of evidence are cited.
stick figures (p.
327)
'are usually
The first is that
recognized easily despite the limited
amount of shape information they portray.' This suggests that the type
present
of information
in stick
figures
is important
in human
vision.
The second comes from certain illusions that have to do with the way
Whether or not a given shape at a
we perceive elements in a pattern.
given orientation is seen as a square or as a diamond may depend upon
to a larger pattern in which it apoears.
axes that we assignr
comes
from
recognitional
deficits.
Certain
capacities
that
brain
seem
canonical coordinate system (i. e.
All these forms of empirical
lesions
to
require
result
in
recovery
The third
losses
of
a
shape's
its stick figure).
evidence that particular
representations
and algorithms are being used are collected from behavioural data.
broader considerations
actions
that
we
circumnavigation
and
more
subjects
are
visual
and
specific
asked
concern
general
is
whether
requirements
perceivers
manipulation
data
of
collected
not
or
-
of
are
objects,
from
they
107 -
upon the
actually
and
the
discrimination
perceive
a
sorts of
capable
like.
The
of;
Narrower
experiments:
stereogram
as
a
single
raised
(a
shape
a
three-dimensional
unusual
two-dimensional
identify
can
from
example)
for
bucket,
they
whether
square,
representations of it or whether they see something as a square or as a
is incidental.
this
tests typically involve verbal behaviour,
Although these
diamond.
A dog could be trained to salivate when it saw a raised
square or a bucket.
What
is
Intentional
really
being
contents
behaviour by itself does not,
are.
tell you
to be sure,
capacit;.
discriminative
behaviour.
from discriminative
inferred
are
a mere
for is
tested
Such
what the contents
But it is assumed that a subject cannot discriminate between two
distal stimuli unless they cause representations with different contents;
so the discrimination tests reveal at least when different contents arise
from different arrays.10 And much of the computational
by
without
stimulations
between
the
yield
theorist
different
two stimuli,
then
more
knowing
contents.
there
If
must
arrays and some innate assumptions,
than
a
which
subject
be some
differences
can
features
the
two
contents
computations.
So,
for
will
example,
then
the
be
fact
exploited
that
a
in
discriminate
of
that allow her to do so,
theorist must search for an account of what these are.
between
theory can get
the
gray
and the
The distinction
in
subsequent
highlight
causes
a
10. The claim that different contents,
rather than merely different
are necessary for a discriminatory capacity is, of
representations,
course, controversial. But I think the claim is mandatory for anyone
who thinks that contents have an explanatory role to play. What Burge
for
a
contents
are
sufficient
dispute
is
that different
would
discriminatory capacity. For the no-content view see Stich, S., From
Folk Psychology
To Cognitive
Science:
The Case
Against
Belief,
Press, 1983. If Stich is right then there is a
Cambridge, M. I. T.
much shorter route to individualim than the one I am taking.
- 108 -
from an object
different representation
of
explanation
how
larger
we discern
boundary
scaled
will be
features
used in the
of
the
viewed
shape.
My
third
point,
then,
that
is
of
attributions
representations
are
checked against behavioural evidence.
3 Burge's Argument
argument
We are now in a position to consider Burge's
theory is
The argument
not individualistic.
steps, and a conclusion,
that Marr's
six separate
is stated in
as follows:
1) The theory is intentional. 2) The intentional primitives
carry
are
they
and the
information
of the theory
individuated by reference to contingently existing physical
items or conditions by which they are normally caused and
3) So if these physical
to which they normally apply.
conditions,
and possibly attendant physical laws,
were
regularly different, the information conveyed to the subject
her
visual
and
the
intentional
content
of
his
or
representations would be different. 4) It is not incoherent
to conceive of relevantly different physical conditions and
perhaps relevantly different (say, optical) laws regularly
individualistically
same
non-intentionally,
causing
the
individuated physical regularities in the subject's eyes and
nervous
system....
individual's
information
though the
...
might
perceptual
their
05)
In
such
a
case,
(by
3))
the
visual representations would carry different
and have different representational content,
person's whole non-intentional physical history
that some
6)
Assuming
remain
the same.
states
informational
are
or
identified
in
intentional
the
theory
content,
it
in
terms
follows
of
that
individualism is not true for the the theory of vision.
I
confess
argument.
to
being
Individualism
somewhat
about
by
the
structure
entails
that
mental
perplexed
the
mind
- 109 -
of
the
states
are
not individuated
external
(in a necessary or deep manner),
physical
individuative
necessary
items
relation
by
which
they
are
mentioned in step
or
deep
variety,
individualism.
Step
2)
then
attributes
more
the
conclusion
be
that
that
a)
if
individuated,
they make
up
(which
the
then
caused.
is obvious),
are
and
the
needed
to
refute
primitives
what
reference
for Burge's desired
intentional
nor
If
is of the
individuation by
So all that would be required
would
individualistically
is
intentional
anti-individualism demands of mental states;
to external things.
normally
2) of the argument
little
to
by reference to the
the
that
primitives
are
intentional
contents
b)
some perceptual
states are identified by their contents (which is stated in step 6)).
if 2) adverts to deep or necessary
individuation,
not
steps 3),
4)
So,
and 5)
do not contribute to the argument.
On the other hand if the individuative relation of 2)
or deep,
then surely step 3) would not follow from it,
For only if the individuative
sense
explained
would
it be
in
true
section
that
if
1)
the
conditions
above)
in 2)
to
conditions
the
is not necessary
as Burge claims.
were essential
individuated
differed,
then
(in
the
primitives,
so
would
the
contents.11
I think it is
best to assume
that the individuative
meant to be necessary and deep.
Steps
as an elaboration
anti-individualistic
what
the
is
Step 2) can therefore be seen as the
crucial step of the argument.
of
relation of 2)
3),
4)
and
5)
can then be treated
claim of 2)
11.
Burge is normally such a careful and precise
wonders why the argument is so disorganized.
- 110 -
thinker
comes to.
that
one
So I shall treat 3),
4)
and 5) as an explication of anti-individualism,
rather than as part of the argument for it.
Steps
given
3)-5)
invite
earthly
us to consider
subject
would
have
what
had,
representational
had
she
contents
inhabited
an
a
alien
environment in which the very same gray arrays as are here caused by
one
sort
of distal
cause,
there
result
from
a
distal
cause
of
some
different sort.
Let us give our subject a name, 'Visua'. When we consider how things
would have been with Visua, had matters been otherwise, we shall refer
to her as
Visua*.
Visua
she
as
environment.
Visua* is
would
Visua
have
and
thus Visua's
counterfactual
been,
she
Visua*
means that their individualistic
the
same
sequence
considered,
undergo
of
the
gray
had
inhabited
are individualistically
histories are the
arrays,
their
same machinations.
counterpart, .
identical.
same.
brains,
some
They
other
This
undergo
individualistically
But things
outside their
three dimensional physical boundaries may be varied.
The features
that are
to be varied
are some of those
contingently
existing items or conditions that normally cause Visua's representational
contents,
and to which the contents normally apply.
Marr uses such items to individuate contents.
uses them as a guide to discovering
In
Burge claims that
my view,
what the contents are.
he merely
So,
to pick
a neutral expression: the items and conditions to be varied are those in
terms of which Marr identifies contents.
In the counterfactual environment these mundane items and conditions
- 111 -
are
systematically
replaced
by
different
ones.
But these
environment cause
items and conditions in the counterfactual
the very
So there are
as do their actual counterparts.
same gray arrays
replacement
gray
array types the typical distal causes of which will differ across twins.
From
now
when
on
qualification,
same
given:
I
I mean
refer
the
to
that
counterfactuals
bodies,
that
environments
without
counterfactuals,
fit
the
further
just
outline
broad
are systematically
in
different
the way described.
There are two claims that Burge might make about Marr's theory and
If we take him at what seems to be his word,
the counterfactuals.
view
is
that
it
counterfactual
is
a
consequence
environment
of
of
Marr's
the
theory
envisaged
that
in
sort,
weaker
claim
wishes,
and
would
that would
perhaps
have
suffice
is what
been
to satisfy
he
really
different.
Burge's
intends.
For
the contents of the
he says that if the relevant items were different,
representations
any
Visua*'s
representational contents would have differed from her actual ones.
subject's
his
But
there
is
a
anti-individualistic
This
would be
that
there is at least one counterfactual environment in which Visua* would
have
had
different,
different
reptesentational
contents.
If
the
items
had
been
the contents of the subject's representations might have been
different.
The
truth
attributes
depend
of
the
weaker
representational
solely
environments
on
would
a
claim
contents
in
such
show
a
capable
-
of
endowing
112 -
that
way
that
Marr's
they
properties.
individualistic
subject's
be
would
a
given
theory
do
not
Different
subject
with
different representational contents, even if her individualistic properties
were held constant.
I shall
follows.
argue
First
against
I
both
shall
claims.
make
a
The
general
argument
point
will
about
legitimately suppose to vary across environments.
proceed
what
we
as
may
Then I shall outline a
particular test case and consider what Marr's theory should say about
it.
I will
theory,
offer
what
I call
upon which it
representations
with
a
would
the
'straightforward'
predict
same
that
intentional
interp
Visua
and
contents,
Lation
Visua*
even
of
compute
though
standard distal causes of their gray arrays are quite different.
then explain Burge's
my straightforward
Marr's
and
theory,
refute
the
I shall
"devious" interpretation of Marr, which contradicts
one.
I shall argue that the
deviously interpreted
incorrect.
the
This
will
L/
undermine
Burge,
the
predictions
yielded by
would be unmotivated
devious
interpretation,
Burge's strong claim that in any of the counterfactual
Visua* will have different contents
from Visua's.
then be in place to consider the more plausible,
The
and
habitats
background
but still false,
will
weaker
claim.
The general point is this.
retinas are identical,
including
the
patterns,
the
same.
exploit
But
both
Since the patterns of light on the twins'
and the very earliest representations
zero-crossings,
content
the
the
of
those
computations
input
at
a
minimum)
desriptions
about the structure of the world.
derive
and
The
- 113 -
the
higher
additional,
be
light
the
representations
innate
representational
those
must
representations
earliest
that
describe
(up to and
assumptions
content derived
a
by
given
thus
computation
A
assumptions.
given
of
computation
be
may
described,
(syntactically)
depends,
in
a
the
upon
part,
representation,
given
different
innate
formally
semantic
So, even if we know what the earliest descriptions are,
intepretations.
and what formal processes operate upon them, we cannot yet infer the
the resulting representations
attribute to the system.
How we interpret
thereby constructed.
content of the representations
must depend upon which assumptions we
If we regard Visua*'s system as constructing
different representational contents from Visua's, we must also regard it
as making different assumptions.
4 Against Burge's Strong Claim
4.1 'The First Case
Suppose that Visua*'s environment is one in which
we had artfully
attached a stereoscope to her face at birth. We have done this in such
a way that she is not aware of it.
the
stereoscope's
remote
control.
two screens,
We
identical to Visua's.
arrange
Her eyes receive only images from
we are capable of
which
things
so
that
her
adjusting
by
will
be
stimulations
The rest of her distal environment,
however,
will
be different from Visua's.
4.2 The Straightforward
r)iscussing
Interpretation
the assumptions that allow
- 114 -
us to infer a
3D
shape from a
silhouette Marr writes (P.
219):
if a surface violates these implicit assumptions, then we
should see it wrongly. Our senses should deceive us in the
sense that the shape we assign to the contours will differ
from the shape that actually caused them.
Visua*'s environment is one in which the viewed surfaces violate the
implicit assumptions.
Specifically,
retinal
from
images
arise
a
the system will assume
single
surfaces. On a straightforward
surface,
we
be able
should
illusions are (Marr p.
than
two distinct
interpretation of Marr we should predict
- exactly as commonsense dictates - that
Moreover,
rather
that the two
Visua*'s senses deceive her.
to determine
what
the
contents
of
her
100):
The brain is capable of measuring disparity and using it to
create
the
sensation
of
depth.
For
purposes
of
demonstration, a stereoscope from a souvenir shop will do:
when individual views are seen with just one eye they look
flat. However if you have good stereo vision and look with
both eyes the situation is quite different. The view is no
longer flat: the landscape jumps sharply into relief, and
your perceptions are clearly and vividly three-dimensional.
Visua*'s
brain
is
regularly
regularly delivers vivid,
this
because
those
relevant arrays.
are
confronted
illusory,
the
with
and
stereograms,
representations of 3D scenes.
representations
it
constructs
so
It does
from
the
If the gray arrays that in Visua result from 3D objects
were caused in Visua* by our stereoscopic
derive just the representational
facsimiles,
the latter would
contents that the former does.
would misrepresent flat distal causes as 3D scenes.
Burge writes (p.
35)
If the properties and relations that normally caused visual
what
they
are,
the
were
different
from
impressions
individual would obtain different information and have visual
- 115 -
Visua*
If the
content.
with different intentional
experiences
regular, law-like relations between perception and the
environment were different, the visual system would be
solving different information-processing problems.
But how would it obtain the different information? How would it solve
the
different
contributions
are
representations,
veridical
Marr's
carefully
very
separated.
The
usually,
to the
environment contributes the gray arrays and conforms,
assumptions.
information
There is no way that the environment can get the correct
into
the
system
visual
other
by
than
peripheries and conforming to the assumptions.
bombarding
It must
what caused the bombardment.
infer, from the gray arrays,
problem
theory is designed to tell us how the information
for
is
this environment
solved.
is solved
It
by
processing
the
brain's
construction of a representation: the representation is constructed,
the aid of innate assumptions,
will be correct
if and only
obtain
flukes,
an
(barring
unsound
conditions
if the
the
When
obtain,
will typically
correctly
be
a representation
is
assumed
by the
system
the description is arrived at by
conditions
representation will typically be correct,
problem
with
from the gray arrays. The representation
in which case
inference).
its
is given
The organism
only the gray arrays and its innate assumptions to work with.
Marr's
the
theory
to the business of forming
and organism,
of environment
On
problems?
processing
information
obtain,
then
the
and the information processing
When
solved.
constructed
do
that
the
cornditions
do
not
fails correctly to depict its
distal cause.
The construction
of
representations
information about the environment
is
thus
only
the
extraction
of
hold.
If
when the assumed conditions
- 116 -
the conditions do not hold,
representations.
then the system will fail to deliver correct
It won't magically solve another
problem.
It will screw
up.
If the informational content of a representation is defined by what it
regularly
actually
with,
covaries
of
then,
representations will carry different information.
course,
twins'
the
The question is whether
that sort of informational content is what the visual system manipulates
according to Marr's theory. I am about it to argue that it is not.
4.3 The Devious Interpretation
The
straightforward
predictions
the
theory
interpretation
delivers
about
requires
the
supposing
representations
that
the
constructed
from specified gray arrays are not limited in applicability to a range of
contexts that excludes the stereoscope world.
Burge would argue that the predictions are thus limited.
out that on Marr's theory (p.
He points
33):
the information carried by the representations
- their
intentional content - is individuated in terms of the specific
distal causal antecedents in the physical world that the
information is about and that the representations normally
apply to.
And concludes:
Thus
the
individuation
of
intentional
content
representational
types
presupposes
the
veridicality
perception.
And (p.
35):
- 117 -
of
of
The methods of individuation and explanation are governed
by the assumption that the subject has adapted to his or
her environment sufficiently to obtain veridical information
from it under certain normal conditions.
37)
And (p.
(or
state
visual
one's
of
content
intentional
the
representation) is individuated against a background in
which the relevant state is normally veridical.
devious
So the
assume
that in
causes
under
normal
under those conditions.
So,
will
from
R,
arrays
are
normal
then the representation will
by a thing of a certain type T,
conditions,
of a certain
under
caused,
standardly
correct
be
if a creature's
(the interpretation proceeds)
that
to
standard
have
visual system is such that it constructs a representation
type,
are
We
this.
representations
the
conditions,
like
representations
where
any environment
runs
Marr
of
interpretation
those causes as things of
have an intentional content that represents
type r.
is
interpretation
devious
The
for
analogue
the
thus
visual
12
representation of a causal theory of reference of the kind that Putnam
and that Burge
has applied to natural kind terms of natural language,
himselfl3
extended
has
to
content
the
of
propositional
the
attitudes
attributed in folk psychology.
On
12.
the
devious
Putnam,
H.,
interpretation,
'The
Meaning
Visua*
See
e.
g.
'Other Bodies',
assumed
of 'Meaning",
Minnesota Studies in the Philosophy
of Minnesota Press, 1975.
13.
is
in
of Science,
Woodfield,
Object, Oxford University Press, 1982.
- 118 -
in
to
see
Gunderson,
Minneapolis,
A.,
ed.,
correctly
K.,
ed.,
University
Thought
And
under
she will not be representing
single 30D scene,
that
the two surfaces of the stereogram
as a
but as two flat surfaces.
But Burge's argument
true
certainly
then,
It follows,
their typical normal causes correctly.
represent
will
representations
Her
environment.
normal conditions in her
for the devious interpretation
when
that Marr,
in earthly
humans
studying
It is
is invalid.
context,
typically supposes that what a representation represents is what causes
But it does not begin to
it under normal conditions.
had been
that if some other type of distal cause
follow
from this
the typical one,
the
representations
are
representation would have represented that.
What
theorist
the
needs
to
out
find
One good way
under what circumstances.
constructed
find conditions under
what
is
then to infer that what is seen is what is there.
the representations
representations
causes.
Since
well.
works
In an
Under those conditions
will represent their distal causes.
is
that
contents that typically
where
environment
propitious for vision,
Marr
so
the
are
may
to
attribute
true of their
normal
the representational
here,
For us,
it
are the ones in which our
is roughly true that the normal ,ircumstances
sight
to see correctly and
appears
which the subject
to do this is to
normal distal
circumstances
contents would
our
were
not
not typically
be true of their normal causes; they would typically misrepresent them.
There is
where our
those
not the slightest
representations
causes
plausibility,
had
correctly.
and is
certainly
to believe that in any environment
reason
normal causes,
Such
a
supposition
not supported
- 119 -
we
would be able to see
lacks
independent
by the computationdl
theory
of vision.
I argued above (in section
of
are
representations
checked
against
attributions
theory,
evidence.
behavioural
So,
for
Marr uses the observation that we successfully circumnavigate
example,
and
on Marr's
2.3) that,
manipulate
to
objects,
justify
representations
attributing
in
the
Attributions of representations that would be in line
form of 3D models.
with the devious interpretation could not survive even this crudest sort
of
appropriate
For
test.
observational
for one who
Visua*
was correctly
screens just in front of her eyes.
not
evidence
of
the
kind
act
perceiving
She would not,
in order to see what was
remove the screens,
Behavioural
would
in
manner
on two flat
images
for example,
try to
going on behind
them.
Marr
would
demanded
actually
any
by
therefore count against the devious interpretation.
evidence
The
straightforward
would
be
would,
that
quite
consistent
That interpretation
intepretation.
illuded,
be
however,
would
she
be
the
predicts that Visua*
representing
the
That is just the prediction
images as a single 3D scene.
with
stereoscopic
that fits the
data.
The straightforward
interpretation is also in line with
know what the phenomenology
like to view a stereoscopic
flat.
she
intuition.
We
of 3D vision is like. We know what it is
picture,
and
what it like to see things as
We know what it would be like for Visua* to have the experiences
has,
phenomenologically
experiences
like
that
could
construed.
have
It is
very
representational
- 120 -
hard
to believe
contents
that
that
were
true of flat surfaces.
14
4.4 Recapitulation and Remark
Let me recapitulate
that
Marr
the argument so far.
attributes
circumstances
here.
representations
Burge
inferred,
that
the
seer.
independent
A
good
better
reading
reason
of
is
that
that Visua*
that
that
we
under
the
normal
underlying
since
see
we
correctly
have
under
the contents the theorist attributes are ones
that in fact represent their typical distal causes.
supposing
correct
perceived
under normal circumstances
Marr
for supposing
normal circumstances here,
are
incorrectly,
principle was that vision is always correct
for
Burge correctly
sees correctly
in
her
reason to suppose that her representations
We have no reason for
environment,
and
so
no
depict their typical normal
causes.
The
content
behaviour.
of
higher
representations
interacts
substantially
When behaviour is appropriate in the light of distal stimuli,
and this fact is legitimately explained in terms of perception,
infer
that
perception
is
yielding
behaviour is not thus appropriate,
awry.
with
correct
representations.
we may
When
we should predict that perception is
15
14. Perhaps a supporter of Burge would hold that the phenomenal
properties of the twins' perceptual experiences would vary. That would
be an interesting extension of anti-individualism, and one that would
stand in
radical need of an argument.
15. Inappropriate behaviour is not, to be sure, conclusive evidence of
misperception. Peculiar desires or false non-perceptual beliefs may be
what is responsible for the wacky behaviour. But all that is required to
- 121 -
behavioural
evidence,
reasonable in
would be quite
that
One hypothesis
representations
have
Since Visua* is Visua's twin,
all of
Visua*Is
that
be
would
exactly the contents that Visua's do.
the light of the
Visua*'s movements would be ones that someone would make,
seeing
what
perceptions.
Visua sees and walks
When
a similarly
Visua* will travel
herself
is
rhinoceros.
states,
responding
and
sees,
actually
Visua
to
subject
the
illusion
that
will
It
too,
she,
It would thus be quite safe to attribute
including perceptual
that Visua* does,
states,
be
her
rhinoceros,
sleeping
a
sees
Visua*
if
as
mental
to Visua*
Nothing
that are just like Visua's.
would count against the
described,
non-intentionally
to
rationally
round a sleeping
route.
shaped
if she were
hypothesis.
The right way to make sense of what Visua* was doing would seem to
be
to
ascertain
appropriate.
Visua's,
in
what
sort
her
environment
of
The obvious candidate
motions
for such an environment
would be
and so attributing to Visua* contents that would be true,
she confronted with the distal causes of Visua's representations,
be
sensible.
But
Visua's
environment
which Visua* could reasonably
might
be supposed
not
be
the only
to be seeing
be
would
were
would
one
in
successfully.
What if there were another one? That is a topic for the next section.
undermine
the
devious
behaviour sometimes be
than other attitudes.
intepretation
explicable in
-
122 --
is
that
terms of
systematically
misperception,
wacky
rather
5 The Weaker Claim
My argument in the First Case depended on the point that successful
old
brain
old
in any
Where
environment.
the
behaviour
subject's
we should
without attributing correct visual representations,
explained
thus
is
There
representations.
correct
attribute
automatically
is
or appropriate in a way that can be
inappropriate to her environment,
not
for just any
cannot be presupposed
sight under normal circumstances
nothing to count against the straightforward interpretation.
5.1 Second Case
that
however,
Suppose,
environment
and that
attributable
not
to
the
success
fluke
is
Visua*
or
of
artful
her
but
to
her
it is
through
meanderings
contrivance,
to
adapted
perfectly
visual
her
In such a case Burge's argument could proceed without the
perception.
is automatically
that veridicality
false premise
assumed to result
from
the existence of normal causes.
then,
We shall assume,
can
see
successfully
some of her gray
counterpart
whatever
protagonist
are,
could
in her environment.
arrays must still differ
arrays.
they
that Visua* wears no stereoscopic glasses and
But
would
correctly
shall
we
The typical distal
from
suppose
the
lie
perceive
the
distal
- 123 -
the causes of
that
give
not
to
causes of
differences,
these
the
view
causes
Visua's
of
that
each
her
own
irritations.
This
could
happen
if,
subtle,
environments were small and
the
say,
the two cases played no significant
differences
between
across
or the items that differed
role in the subjects' lives,
the
or the
items were out of range of senses other than sight.
Burge's argument could then proceed
discover that
normal
for her.
under
that,
Visua*
system,
The
those
are
correct
that
are
true
world.
So,
And,
conditions
attributed
their
here,
attribute assumptions
for her world.
under
to be
depict
conditions,
false in Visua*'s
the theorist should
correctly
representations
constraints
hypothesis,
see
could
The theorist would
as before.
and
by
Visua's
argument would
to Visua*'s
just as Visua's
By
causes.
assumed
Burge's
be ones
would
normal
were
that
go,
system that were
assumptions allow
her to
construct correct representations of her distal stimulations on Earth,
that
Visua*'s assumptions allow her to construct representations
would
correctly
depict
her
distal
stimulations
on
her
stimulations and constraints differ across worlds,
world.
so
Since
the
Visua and Visua* will
construct representations with different contents.
The
case
Earth (E)
under
dispute
may
we have a constraint,
described
be schematically
Ce,
thus.
On
that is replaced on Twin Earth (T)
by a different constraint C t . On E the normal cause of gray array type
G is some item O ; on T the normal cause of gray arrays of type G is
some
other
sort
of
thing,
visual system assumes
Oe,
and
so
form
a
Ot
.
The
constraints
are
such
that
if
the
Ce it will infer from arrays G the presence of an
representation
satisfied by, and only by,
of
intentional
objects of type Oe.
- 124 -
type
Re,
that
is
But if it assumes C t it
will
infer
from
arrays
G
the
presence
of
an
Ot ,
and
representation Rt that is satisfied by and only by Ots.
form
(See Key,
a
part
(a)).
KEY
(a)
E: Earth
T: Twin Earth
G: A type of gray array
O : the type of object that normally causes Gs on E
Ot: the type of object that normally causes Gs on T
C e : A constraint that is true on E,
but false on T
Ct: A constraint that is true on T,
but false on E
R e: A representation that is true of, and only of, O s
e
e
R t : A representation that is true of, and only of, Ots
A visual system assuming Ce will construct Res from Gs
A visual system assuming C t will construct Rts from Gs
(b)
C n: A constraint that satisfied if either Ct or C
are
R n: A representation that is true of both Ots and Oes
A visual system assuming C
n
will construct R s from Gs
n
5.2 Undermining of the Devious Interpretation
On
the devious interpretation,
sees correctly on T.
On T,
we should
argue
be assuming
Visua's system,
Ct
,
follows.
Gs are typically caused by Ots.
correctly requires the construction of Rts from Gs.
must
as
and constructing
on the other hand,
- 125 -
C
e
So seeing
So Visua*'s system
representations
assumes
Visua*
Rt
from
Gs.
to construct R s from
e
Gs.
from Rts,
extensions
R s have difterent
Since
e
and
Visua
Visua*
undergo representations with different contents.
It does not follow from the fact that Gs
But this argument is invalid.
are
caused
must
by
Ots that correct
be representations
representations
of Ots,
hence
constructed
of intentional
type
from
Rt .
Gs
Any
representation that includes Ots in its extension will be correct on T.
In particular a representation that is satisfied by both Oes and Ots will
be
correct
on
T.
If
the
representations
Visua* are of some neutral type R
Ots,
by
Visua
and
that is satisfied by both O0s and
n
e
while
correctly
see
may
they
then
constructed
same
the
having
yet
representations.
Ascribing to both
Visua
and
Visua*
representations
that
would
be
correct in either environment is exactly what would be required by the
straightforward
interpretation.
The straightforward
interpretation thus
requires that both protagonists construct representations of the neutral
intentional type Rn.
would
be
broad
Correlatively,
enough
to
be
the assumptions
satisfied
in
systems would neither assume C t nor Ce,
C
Burge's argument that the comp¶Itational
reference.
language
The
reading into it a particular
(See Key,
part (b)).
theory is not individualistic
form of a causal
theory of
The Putnamian theory might be true of some parts of natural
and
automatically
perceptual
environments.
but some neutral assumptions
that would be true in both environments.
depends upon
both
we would attribute
even
of
assume
module.
some
that
it
parts
of folk
applies
to
Scientific psychology
- 126 -
psychology,
a
but
computational
will perforce
one
cannot
theory
of
depart
from
a
folk
One such departure might well be a failure
psychology in various ways.
16
to adhere to a causal theory of reference of the folk variety.
must
Burge
attribute
to
particular way of choosing,
content
of the
of
vision
a
from among the various ways of describing
the typical causes of a correct representation,
the
theory
computational
the
representation.
In
the one that will give
our example
the
recommended
selection for Visua* would have to he R t rather than Rn. I find nothing
in Marr's theory to support this interpretation.
itself and the particular Putnamian
himself has, without provocation,
However,
theory,
Both the causal theory
version of it are things that Burge
added into the computational theory.
even if the theorist of vision espouses no Putnamian causal
she might still attribute
devious interpretation.
contents in the way predicted by the
If that were the case,
it would be because the
balance of various sorts of evidence and argument happened to motivate
attributing to Visua and Visua* the specific representations
respectively,
I
shall
Re and Rt,
rather than Rn
argue
that
the
balance
would
never
tip
that
way.
The
attributions predicted on the straightforward interpretation are more in
line
with
Marr's
actual
methods,
and
with
the
dictates
of
good
explanation.
16. My own view is that folk psychology comes unstuck on twin earth
cases, dictating both that the twins do have the same psychological
states, and that they do not. I reserve that argument for another
paper.
- 127 -
5.3 Defence of the Straightforward Interpretation
and
her
place
Visua
and
a
in
Visua* to Earth,
Let us bring
the counterfactual.
Let us elaborate
together.
laboratory
is
Visua*
still
assumed to be an evolutionary product of and to have grown up in one
of
the
three-dimensional.
17
her
habitats,
counterfactual
to
visit
is
Earth
merely
In
Let us subject them both to two experiments.
the first experiment we arrange an experiment in which conditions C
are
and
violated,
objects
of
type
Ot ,
to
and
Ct obtain,
conditions
cause
of
arrays
subject
G.
type
the
subjects
In
the
e
to
second
and Gs are
experiment we arrange matters so that conditions Ce obtain,
caused by O s.
e
The
devious
sees
correctly
Visua*
experiment
would
interpretation
illusion that she is seeing O s.
e
us
have
but
Ots,
In the
that
say
is
Visua
second
the
subject
first
to
the
experiment it is Visua
correctly.
the illusions and Visua* who is seeing
who suffers
in
On the
straightforward interpretation they would both be seeing all the distal
causes as 0
n
s.
Representations
Top-down motivations
require a top-down motivation.
come from finding some use or purpose to which the content is put.
So
than mere O s,
we
to motivate
the attribution
of Ots and Os rather
would have to find some purpose that the more specific contents could
17.
If
the
strain
of
imagining
that
Visua*
counterfactually described, is now unbearable,
the twins are merely physically type identical.
- 128 -
you
is
Visua
may
herself,
suppose
that
serve,
but that the vaguer contents could not.
We would need to find
some higher computational process into which the contents would
feed,
and it would have to make some difference to this process whether its
input was on the one hand R
this
be?
Nothing
that
recognitional capacity.
or R , or, on the other, R . What would
e
t
issued
n
in
any
discriminative
For the twins are twins,
ability,
or
and will be the same in
every testable respect.
The twins' performances in all the tests we could devise would be the
same.
They will,
stereograms,
silhouettes,
example
be
for example,
able
make
the
distinguish
to
register
identify
3D effects
exactly
the
same
distinctions
among
shadow
boundaries
from
changes from object boundaries),
from
all the
same
3D
shapes
from
markings
(for
same
surface
highlights
from
texture
make the same groupings in patterns
and so on.
But it is tests of exactly this kind that the theorist actually uses to
determine
how
Attributions
many
distinct
of representations
contents
are
she
checked
needs
against
to
attribute.
just this
sort
of
behavioural evidence.
And there is good reason for this. It would be a
violation
razor
of
Occam's
to
invoke
invocation would be doing no work.
invoking the two contents,
theoretical
purpose
Consequently
served
There
by
the straightforward
the
global
contents
would just be
where one would do.
distinguishing
where
the
no point to
For there would be no
between
the
contents.
intepretation has economy on its side.
The devious one scratches where there is
Moreover,
different
considerations
- 129 -
no itch.
that
help
to
justify
the
same
the
be
would
contents
of
attribution
would come at exactly
behaviourally construed,
successes and failures,
For
cases.
two
the
in
the same points. The twins would make the same circumnavigations and
Visua* would duck to avoid flying objects, duck to avoid
manipulations.
sew buttons
read scrawly handwriting,
images of such in a 3D movie,
on a shirt, make drawings of what she sees, and so on and on, just as
It
Visi;a does.
all this
odd to suppose that throughout
would seem
Visua* would be systematically misperceiving something that Visua was
representing
different contents,
attributing
for
motive
no
be
would
There
correctly.
no explanatory gain to be achieved by doing so.
reason that
we would simply
interpretation
On the straightforward
both twins could solve the information processing problems set by both
We would infer that their representations were
Earth and Twin Earth.
of type Rn,
and that they were assuming neither C t not Ce but the
more conservative
representations.
Cn -
all that
required
was
to account
for their
And that is exactly what the evidence warrants.
To attribute contents deviously rather than straightforwardly
would
not only be out of line with Marr's actual practice, it would run directly
counter to basic canons of good explanation.
message
The general
distal
in
difference
evolutionary
is
cause,
background
we
that
can
regular,
however
must
make
never
a
To show that the differences between
showed
up
in
representations
perception,
were
specific
we
would
enough
- 130 -
have
to
a mere
ubiquitous
however
difference
content.
that
suppose
in
representational
the twins' environments
to
to distinguish
show
the
that
their
two sorts
of
this could
abilities,
Given the identity of discriminative
distal cause.
The best theoretical description will always be one in
never be shown.
which the representations fail to specify their extensions at a level that
distinguishes
the two sorts of distal cause.
It will always be better to
suppose that the extension includes both sorts of thing.
Being tickled
something is not tantamount
by
18
that
to knowing what
thing is.
6 A Specific Example
Burge
what
is
(p.
42-3)
wrong
straightforward
himself provides
with
the
devious
one.
It will be
an
example that
interpretation,
worth
nicely
and
going over the
illustrates
right
with
argument
the
again
with his specific example.
He
asks
us
to
imagine
a
subject,
P,
among
whose
early
representational types is one regularly caused by thin shadows.
visual
In such
slhadows.
If the
same perceptual type were on occasion caused by a small crack,
Burge
a case,
Burge argues,
the perception would represent
says, then P would misperceive the crack as a shadow.
normal circumstances
P may be fooled by a crack.
So, even under
But suppose that,
18. Note that refusing to proliferate contents beyond what is needed to
not special to the computational
explain discriminatory capacities is
theory of vision. The reason why we do not att-ibute, for example, a
concept of bank-managers to dogs is that dogs cannot discriminate
bank-managers (or banks) from other things.
- 131 -
the optical laws were different.
counterfactually,
In the counterfactual
world there are no instances of the right sort of shadow.
very
that
arrays
gray
P would be visually representing cracks as cracks. 1
We
Do
choice.
a
face
representations of shadows,
do
we
follow
representations
environments,
9
one
in
of crackdows
to
attribute
and
route
(thin,
environment
Or
of cracks?
representations
in another,
straightforward
the
P
to
attribute
we
Burge says,
In such a case,
caused by small cracks.
counterfactually,
are,
shadows
the
by
caused
factually
are
Instead the
dark,
in
both
marks
that
P,
could be either shadows or cracks)?
What
would
required
be
for
positing
the
contents would be a top-down motivation.
different
We would
representational
need a reason for
supposing that higher processes exploited the information that it was a
(crack)
shadow
marred
that
surface,
visible
the
rather
a
If the representation of a crackdow would account for all the
crackdow.
higher processing just as well as would the representation
(crack),
than
then it is a crackdow
of a shadow
representation that we should attribute.
Anything more discerning would be unmotivated.
But,
cracks.
by
We
explanation
19.
hypothesis,
could
P
cannot
therefore
of which would
between
discriminate
never
locate
a
capacity
require the attribution of
Note the analogy with water and XYZ:
shadows
of
the more
P's
and
the
precise
if there were a small amount
of XYZ on Earth we would probably misidentify it as water. We would
call it 'water', but we would be wrong, as chemical investigation might
show. But on Twin
predominates there.
Earth
the
word
- 132 -
'water'
is
true
of
XYZ,
which
content. No top-down motivation would be available.
We do form representations,
of about the specificity
that have contents
during early vision,
and ones that
of crackdow,
specificity of shadow or of crack.
have about the
It will be instructive to consider how
these are arrived at, and what motivates their attribution.
The
reliably
that
representations
first
covary
with
phenomena are formed when the outputs of the zero-crossing
conjoined (c.
Burge p.
f.
external
filters are
Marr attributes to these representations
33).
the content eJge.
"edge" representations
It is not abundantly clear whether
to represent
or features
of
In three places Marr states that representations
at
features of the retinal images themselves,
the viewed surface.
this level are still of the retinal images (p.
others (p.
41,
are meant
quoted above,
p.
93)
71,
p.
91,
p.
366); and in
strongly implies this.
But he also
tends to talk as if they can represent elements of the surface
p.
68,
p.
possibility
outputs
91); and this is how Burge seems to interpret him.
would be to think of these
of
the
representations
zero-crossing
representations
filters.
In
any
(p.
52,
Another
as being of the
case,
however,
that really do refer to external things must arise here
or soon after.20
It is interesting that the theory survives unclarity on this point.
It
20. My own view, unlike Burge's, is that we should avoid attributing to
the system representations of external things for as long as possible,
probably until the 2.5D sketch itself. At lower levels we do not need to
attribute such contents to explain what the system is doing.
- 133 -
does so precisely because nothing is at stake except the inferential role
their place in the sequence of steps that mediate
of the representations,
the
It
representation.
no
makes
segment
representations
surface,
or treats
to
array
gray
from
passage
as already
to
previous
edge
the
treats
visible
the
to marks on
referring
them as referring
one
whether
difference
model
3D
the
and
sketch
2.5D
and
representations
sees the later processes as inferring from them the presence of such a
mark.
Let
us,
in
any
follow
event,
in
Burge
that
supposing
representations do represent elements of the viewed surface.
important to realize that 'edge' has a very broad meaning.
edge
It is very
As far as the
early representation goes, an edge could be the boundary of a shadow,
a
highlight
a
on
surface,
content
boundary.
of an
21
unspecific
functions
thus
'Edge'
texture,
a
surface
real
one caused by a patch of ink),
reflectance change (for example,
object
of
property
a
representation
like
'crackdow'
the extension
to
or an
give
of which
the
is not
matched by that of any simple English expression.
The higher processes
"edge" is really
physical edge,
present,
perform
to decide
what
whether it is a shadow boundary,
or whatever.
further information,
computations
These computations
sort of
or a real
require exploitation of
for example about depth or motion.
21. It would therefore be impossible to construct a twin case for the
content edge. Since any reflectance change on a surface counts as an
"edge", any twin environment will either provide "edges" or be a world
of many illusions, If the distal stimuli were not even marks on surfaces,
then seeing successfully would be out of the question for creatures with
brains like ours.
-
134 -
What motivates the attribution to the system of a representation that
is really of, say,
a shadow boundar/,
rather than of an "edge", is the
fact that we can tell, under normal circumstances,
we are seeing:
boundary,
content
or a highlight,
is
manifest
discriminatory
needed
we can tell,
to
whether we are seeing a shadow
or a real physical edge.
us
behaviour.
to explain
that is,
what sort of "edge"
subjectively,
Representations
the discriminative
and
cleary
detectable
in
different
contents
are
and
provides
the
with
capacity,
The representational
this
top-down motivation for positing them.
It is only because we can normally tell which sort of "edge" we are
seeing,
that Marr attributes the different representational contents,
and
seeks for an account of how they are inferred from the gray arrays.
Similarly
for small
visually discriminate
it
be
rather
correct
than
to
cracks and
shadows.
between them,
attribute
crackdow
crackdows as shadows,
to
her
Only
if the
subject
under normal circumstances,
crack
representations.
or as cracks,
or
The
shadow
system
would
representations,
will
represent
rather than merely as crackdows,
only if it extracts this information from the retinal array,
of additional cues,
could
assumptions and computations.
system is doing this depends upon it,
Whether
with the aid
or not the
not upon what normally causes
the gray arrays.
It is no accident
that Burge
had to try to make up his own example,
rather than take one from the theory itself.
every
content
that
the
theory
actually
motivation and a bottom-up account.
does
The
reason is
simply that
has
a top-down
attribute
The result i. that it is too tightly
- 135 -
constrained
by
individualistic
the
to construct
testable dispositions,
general
argument,
the
shadow
arrays
and
the
Burge claims that his
'is
independent of the theory of vision that we have been discussing [i.
e.
seems to me,
includes
a twin case.
gray
example
Marr's theory].
which
the
facts,
and
crack
It supports and is further supported by that theory.' It
rather, that his argument flies in the face of the theory.
7 Exegesis
Burge's
examples
quotes.
devious
intepretation
of specific
In
this
theoretical
final
section
of Marr
claims,
I
will
is defended
and
show
by
two
lengthy
by a couple of additional
that
the
straightforward
interpretation is compatible with all Burge's evidence.
Here are the essentials of one of the specific claims (Burge p.
33):
In building up informational or representational primitives in
the
primal sketch,
Marr
states
six general physical
(for
assumptions that constrain the choice of primitives ...
example: ] a) the physical world is composed of smooth
surfaces
having
reflectance
functions
whose
spatial
structure may be complex: b) markings generated on a
surface by a single physical process are often arranged in
continuous spatial structures - curves, lines etc.... These
assumptions are used to identify the physical significance of
-
the
objective
information
patterns in the image.
conditions under which
given
by
-
certain
types
The computational theory
these primitives form to
information
about
items
in
the
[Clonditions are laid down under
representing
as
taken
may
be
condition; as being edge, boundary,
of
states
carry
physical
world ...
which certain patterns
physical
an
objective
bar or blob detectors.
I accept this as a fair account of something that Marr does.
But how
does this support the claim that 'the intentional primitives of the theory
- 136 -
...
are individuated by reference to physical items and conditions
by
reference
to
they
which
but
external contingencies,
words
the
content.
form
the
this to mean
I take
..
that
used
to
in
own
Burge's
informational
identify'
theorist
the
is
even
has,
reference
'assumptions
of
there
Certainly
caused'?
normally
are
the assumptions
uses
about the world in order to find out what the representations deployed
by
the
the
constrain
are.
system
visual
choice
of
When
seeks,
Marr
primitives,
he
trying
is
Burge
as
limit
to
says,
to
hypotheses
He is not laying
about which primitives there are in the visual system.
down conceptual constraints on what it is for there to be a particular
primitive in a system.
In his second
example
Burge
and representation
environment
the
expresses
thus (ibid.):
relation
'Marr motivates a central
representational primitive by stating physical constraints'.
and
true,
again
this
fails
support
to
between Marr,
Burge's
Again this is
The
interpretation.
physical constraints motivate the representational primitive by providing
evidence
about
represented.
what is being
Nothing
is entailed
about
individuation.
The conflation of what is taken to justify the ascription of a given
content
with
what
makes
apparent at numerous
Marr (p.
it
the
points in
case
that
the
is
content
Burge's exposition.
He
(p.
there,
27)
43) as saying
The purpose of the representations is to provide useful
descriptions of aspects of the real world. The structure of
role in
therefore
plays an important
real world
the
determining both the nature of the representations that are
used and the nature of the processes that derive and
maintain them. An important part of the theoretical analysis
- 137 -
is
quotes
is to make explicit the physical constraints and assumptions
that have been used in the design of the representations
and processes.
...
of
properties
with
interacts
intentionally
one
about
point
chief
non-individualistic
our
The chief point is 'what entities in the objective
thought experiments'.
world
the
generates
that
rererence
or
representation
to
'tantamount
is
this
that
says
Burge
types
representational
affects
...
and
...
the
semantical
we
individuate
how
So Burge takes 'determining the nature' to mean something like
them'.
Marr's words here are compatible with
'partly constituting the essence'.
such a reading, but they may more naturally be understood as making a
much more innocent claim.
and Idesign' Marr intends evolutionary
It is likely that by 'purpose'
If that is right, then the kind of determination in
purpose and design.
question would be evolutionary determination.
'nature
been very
to have
seems
visual systems'.
Marr says
and exact
careful
in
p.
75)
evolving
our
(Marr
For this reason we should expect what goes on in the
visual system to be highly sensitive to features of the real world.
The
and
the
'carefully
and
assumptions exploited
exactly' selected
real
and
world
evolutionary
the
discover
constantly
implied,
by the
by evolution.
of
nature
look to the
the
of
outside
relation
The
latter.
the
These have been
system.
representational
cause
primitives
representational
the
in
up
shows
sensitivity
content
Marr's
is
the
former
is
that
if
and
for evidence.
we
processes
Nothing
about individuation in the strong Burgian sense.
- 138 -
structure
the
that
point
representations
world
between
is
of
is
an
are
to
we
must
said,
or
A quote
from Marr that might seem to lend to support to Burge is
Burge p.
this (Marr p. 44,
27)
It is of critical importance that the tokens [representational
analysis I
theoretical
[in the
one obtains
particulars i
correspond to real physical changes on the viewed surface:
the blobs, lines, edges, groups and so forth that we shall
use must not be artifacts of the imaging process, or else
o the
inferences made from their structure backwards
be
meaningless.
will
the
surface
of
structure
At a casual perusal this passage might seem to indicate adherence to
it must correspond
the view that if a representation is to have meaning,
Lack of such a correspondence
to something in the real world.
the
render
token
actually saying.
But this is nothing
meaningless.
would
like what Marr
is
The comment is made in the course of a description of
the task of defining (ibid.)
a representation of the image of reflectance changes on a
surface that is suitable for detecting changes in the image's
geometrical organization that are due to changes in the
reflectance of the surface itself or to changes in the
surfaces orientation or distance from the viewer.
[The representation] should include some type of "tokens"
that can be derived reliably and repeatedly from images and
like
of attributes
values
be assigned
to
which can
size
...
brightness,
orientation,
The point is that if one selects tokens that do not correspond to real
physical changes then when one tries to construct from them a complex
sketch of what is
They
will
complex
priori
simply
not
be
representations.
stipulation.
There is
going on outside,
What
to
suited
Marr
else
is
one will end up with
the
making
would
an
task
an
of
empirical
empirical
no support for Burge's anti-individualism
passages.
- 139 -
building
up coherent
claim,
scientist
gibberish.
not
be
an
a
doing?
to be found in these